MOP 2007 Black Group Integer Polynomials Yufei Zhao

Integer Polynomials
June 29, 2007
Yufei Zhao
[email protected]

We will use Z[x] to denote the ring of polynomials with integer coefficients. We begin by summarizing
some of the common approaches used in dealing with integer polynomials.

• Looking at the coefficients

◦ Bound the size of the coefficients

◦ Modulos reduction. In particular, a − b | P (a) − P (b) whenever P (x) ∈ Z[x] and a, b are distinct
integers.
• Looking at the roots

◦ Bound their location on the complex plane.

◦ Examine the algebraic degree of the roots, and consider field extensions. Minimal polynomials.

Many problems deal with the irreducibility of polynomials. A polynomial is reducible if it can be written
as the product of two nonconstant polynomials, both with rational coefficients. Fortunately, if the origi-
nal polynomial has integer coefficients, then the concepts of (ir)reducibility over the integers and over the
rationals are equivalent. This is due to Gauss’ Lemma.
Theorem 1 (Gauss). If a polynomial with integer coefficients is reducible over Q, then it is reducible over
Z.
Thus, it is generally safe to talk about the reducibility of integer polynomials without being pedantic
about whether we are dealing with Q or Z.

Modulo Reduction
It is often a good idea to look at the coefficients of the polynomial from a number theoretical standpoint.
The general principle is that any polynomial equation can be reduced mod m to obtain another polynomial
equation whose coefficients are the residue classes mod m.
Many criterions exist for testing whether a polynomial is irreducible. Unfortunately, none are powerful
enough to be universal. One of the most well-known criteria is Eisenstein’s criterion.
Theorem 2 (Eisenstein). Let f (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial with integer
coefficients such that p | ai for 0 ≤ i ≤ n − 1, p - an and p2 - a0 . Then f (x) is irreducible.
Proof. Suppose that f = gh, where g and h are nonconstant integer polynomials. Consider the reduction
mod p (i.e., apply the ring homomorphism Z[x] → Fp [x]), and let f¯, ḡ, h̄ denote the residues of f, g, h (i.e.
the coefficients are residues mod p). We have f¯(x) = a0 xn . Since Fp [x] is a unique factorization domain, we
see that the only possibilities for ḡ and h̄ are cxk for some integers c and k ≥ 1. Then, the constant terms
of g and h are both divisible by p, so p2 | a0 . Contradiction.
The most typical example for the application of Eisenstein’s criterion is to show that the cyclotomic
polynomial Φp (x) is irreducible for prime p:
Problem 1. Let p be a prime number. Show that f (x) = xp−1 + xp−2 + · · · + x + 1 is irreducible.

1
MOP 2007 Black Group Integer Polynomials Yufei Zhao

Solution. The polynomial f (x) is irreducible if and only if f (x + 1) is irreducible. We have

(x + 1)p − 1

p−1 p p−2 p p−3 p
f (x + 1) = =x + x + x + ··· + x + p.
(x + 1) − 1 1 2 p−2

Note that f (x + 1) fails the Eisenstein criterion for the prime p. Therefore f (x) is irreducible.
Note that the proof of Eisenstein’s criterion extends to other rings with similar properties. For instance,
to show that x4 + 5x + 5 is irreducible over the Gaussian integers Z[i], we can simply apply Eisenstein with
the Gaussian prime 2 + i.
The proof of Eisenstein’s Criterion can be slightly generalized to the following. The proof is more or less
the same, and so it’s left as exercise.
Theorem 3 (Extended Eisenstein). Let f (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial with
integer coefficients such that p | ai for 0 ≤ i < k, p - ak and p2 - a0 . Then f (x) has an irreducible factor of
degree greater than k.
We give one more result that relates to looking at the modulo reduction of polynomials, known as
Hensel’s lemma.
Theorem 4 (Hensel). Let a0 , a1 , . . . , ak be integers, and let P (x) = an xk +· · · a1 x+a0 , and let P 0 (x) denote
the derivative of P (x). Suppose that x1 is an integer such that P (x1 ) ≡ 0 (mod p) and P 0 (x1 ) 6≡ 0 (mod p).
Then, for any positive integer k, there exists an unique residue x (mod pk ), such that P (xk ) ≡ 0 (mod pk )
and x ≡ x1 (mod p).
The proof of Hensel’s lemma closely mimics Newton’s method of finding roots. We work up the powers
of p, and find the a zero of P (x) mod pk for k = 2, 3, . . . . The details of the proof are omitted here.

Root Hunting
When working with integer polynomials, it is often not enough to stay in Z. We have to think outside
the box and move our scope to the complex numbers. A lot can be a said about a polynomial if we know
something about its complex zeros. Many irreducibility problems hinge on placing bounds on the zeros of
the polynomial in the complex plane. We begin with a familiar example.
Problem 2. Let f (x) = an xn + an−1 xn + · · · + a1 x + a0 be a polynomial with integer coefficients, such that
|a0 | is prime and
|a0 | > |a1 | + |a2 | + · · · + |an |.
Show that f (x) is irreducible.
Solution. Let α be any complex zero of f . Suppose that |α| ≤ 1, then

|a0 | = |a1 α + · · · + an αn | ≤ |a1 | + · · · + |an |,

a contradiction. Therefore, all the zeros of f satisfies |α| > 1.

Now, suppose that f (x) = g(x)h(x), where g and g are nonconstant integer polynomials. Then a0 =
f (0) = g(0)h(0). Since |a0 | is prime, one of |g(0)|, |h(0)| equals 1. Say |g(0)| = 1, and let b be the leading
coefficient of g. If α1 , . . . , αk are the roots of g, then |α1 α2 · · · αk | = 1/|b| ≤ 1. However, α1 , . . . , αk are also
zeros of f , and so each has an magnitude greater than 1. Contradiction. Therefore, f is irreducible.
Next, we present a Perron’s criterion, which has a similar statement but a much more difficult proof
compared with the previous result.
Theorem 5 (Perron). Let P (x) = xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial with a0 6= 0 and

|an−1 | > 1 + |an−2 | + · · · + |a1 | + |a0 |.

Then P (x) is irreducible.

2
MOP 2007 Black Group Integer Polynomials Yufei Zhao

Again, the idea is to put bounds on the modulus of the roots of f . The key lies in the following lemma.
Lemma 1. Let P (x) = xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial with

|an−1 | > 1 + |an−2 | + · · · + |a1 | + |a0 |.

Then exactly one zero of P satisfies |z| > 1, and the other n − 1 zeros of P satisfy |z| < 1.
Let us see how we can prove Perron’s criterion if we have this lemma. Suppose that P (x) = f (x)g(x),
where f and g are integer polynomials. Since P has only one zero with modulus not less than 1, one of
the polynomials f , g, has all its zeros strictly inside the unit circle. Suppose that z1 , . . . , zk are the zeros
of f , and |z1 |, . . . , |zk | < 1. Note that f (0) is a nonzero integer, and |f (0)| = |z1 · · · zk | < 1, contradiction.
Therefore, f is irreducible.
Now, let us prove Lemma 1. We offer two proofs. The first proof is an elementary proof that uses only the
triangle inequality. The second proof invokes theorems from complex analysis, but it is much more intuitive
and instructive.
First proof of the Lemma 1. (due to Laurentiu Panaitopol) Let us suppose wolog that a0 6= 0 since we can
remove any factors of the form xk . Let’s first prove that there is no root α of P (x) with |α| = 1. Suppose
otherwise, then we have that

−an−1 αn−1 = αn + an−2 αn−2 + · · · + a1 α + a0 ,

thus

|an−1 | = |an−1 αn−1 | = |αn + an−2 αn−2 + · · · + a1 α + a0 |

≤ |αn | + |an−2 αn−2 | + · · · + |a1 α| + |a0 |
= 1 + |an−2 | + · · · + |a1 | + |a0 |.

This contradicts the given inequality. Therefore, no zero of f (x) lies on the unit circle.
Let’s denote with α1 , α2 , . . ., αn be the zeros of P . Since |α1 α2 · · · αn | = a0 , it follows that at least one
of the roots is larger than 1 in absolute value. Suppose that |α1 | > 1 and let

Q(x) = xn−1 + bn−2 xn−2 + · · · + b1 x + b0

be the polynomial with roots α2 , α3 , . . ., αn . Then,

P (x) = (x − α1 )Q(x) = xn + (bn−2 − α1 )xn−1 + (bn−3 − bn−2 α1 )xn−2 + · · · + (b0 − b1 α1 )x − b0 α1

It follows that bn−1 = 1, a0 = −b0 α1 , and ak = bk−1 − bk α1 for all 1 ≤ k ≤ n − 1. Then, using the given
inequality, we have

|bn−2 − α1 | = |an−1 | > 1 + |an−2 | + · · · + |a1 | + |a0 |

= 1 + |bn−3 − bn−2 α1 | + · · · + |b0 α1 |
≥ 1 + |bn−2 ||α1 | − |bn−3 | + |bn−3 ||α1 | − |bn−4 | + · · · + |b1 ||x1 | − |b0 | + |b0 ||x1 |
= 1 + |bn−2 | + (|α1 | − 1) (|bn−2 | + |bn−3 | + · · · + |b1 | + |b0 |) .

On the other hand, |bn−2 − α1 | ≤ |bn−2 | + |α1 |, so

|bn−2 | + |α1 | > 1 + |bn−2 | + (|α1 | − 1) (|bn−2 | + |bn−3 | + · · · + |b1 | + |b0 |)

and therefore
|bn−2 | + |bn−3 | + · · · + |b1 | + |b0 | < 1.

3
MOP 2007 Black Group Integer Polynomials Yufei Zhao

Then, for any complex number α with |α| ≥ 1, we have

|Q(α)| = |αn−1 + bn−2 αn−2 + bn−3 αn−3 + · · · + b1 α + b0 |

≥ |αn−1 | − |bn−2 αn−2 | − |bn−3 αn−3 | − · · · − |b1 α| − |b0 |
≥ |α|n−1 − |α|n−1 (|bn−2 | + |bn−3 | + · · · + |b1 | + |b0 |)
= |α|n−1 (1 − |bn−2 | − |bn−3 | − · · · − |b1 | − |b0 |)
>0

And so α cannot be a root. It follows that all the zeros of Q lie strictly inside the unit circle. This completes
the proof of the lemma.
In the polynomial P , the second term xn−1 is “dominating,” in the sense that the absolute value of its
coefficient is greater than the sum of the absolute values of all the other coefficients. In the above proof, we
managed to construct a new polynomial Q, whose leading term is dominating. While exactly one zero of P
is outside the unit circle, none of the zeros of Q is outside the unit circle. This observation generalizes to
the following result.
Proposition 6. Let P (z) = an z n + an−1 z n−1 + · · · + a1 z + a0 be a polynomial with complex coefficients,
and such that
|ak | > |a0 | + |a1 | + · · · + |ak−1 | + |ak+1 | + · · · + |an |
for some 0 ≤ k ≤ n. Then exactly k zeros of P lie strictly inside the unit circle, and the other n − k zeros of
P lie strictly outside the unit circle.
This is indeed true. The easiest way to prove this result is to invoke a well-known theorem in complex
analysis, known as Rouché’s theorem.
Theorem 7 (Rouché). Let f and g be analytic functions on and inside a simple closed curve C. Suppose
that |f (z)| > |g(z)| for all points z on C. Then f and f + g have the same number of zeros (counting
multiplicities) interior to C.
The proof of Rouché’s theorem uses the argument principle. It can be found in any standard complex
analysis textbook.
In practice, for polynomials, Rouché’s theorem is generally applied to some circle, and is useful when one
term is very big compared to the other terms.
Proposition 6 becomes very easy to prove with the aid of Rouché’s theorem. Indeed, let us apply Rouché’s
theorem to the functions ak z k and P (z) − ak z k with the curve being the unit circle. The given inequality
implies that |ak z k | > |P (z) − ak z k | for all |z| = 1. It follows that P has the same number of zeros as ak z k
inside the unit circle. It follows that P has exactly k zeros inside the unit circle. Also, it is not hard to show
that P has no zeros on the unit circle (c.f. first proof of Lemma 1). Thus we have proved Proposition 6.
Second proof of Lemma 1. Apply Proposition 6 to k = n − 1.
While we’re at it, let’s look at couple of neat applications of Rouché’s theorem, just for fun. These are
not integer polynomial problems, but they contain useful ideas.
Problem 3. (Romania ??) Let f ∈ C[x] be a monic polynomial. Prove that we can find a z ∈ C such that
|z| = 1 and |f (z)| ≥ 1.

Solution. Let deg P = n. Suppose that |f (z)| < 1 for all z on the unit circle. Then |f (z)| < |z n | for all z on
the unit circle. So, by Rouché’s theorem, f (z) − z n has n roots inside the unit circle, which is impossible,
since f (z) − z n has degree n − 1.
The Fundamental Theorem of Algebra is also an easy consequence of Rouché’s theorem.
Theorem 8 (Fundamental Theorem of Algebra). Any polynomial P (x) ∈ C[x] of degree n has exactly n
complex zeros.

4
MOP 2007 Black Group Integer Polynomials Yufei Zhao

Proof. Let P (x) = an xn + · · · + a1 x + a0 . For a sufficiently large real number R, we have

|an |Rn > |an−1 |Rn−1 + · · · + |a1 |R + a0 .

Apply Rouché’s theorem to the functions an xn and P (x) − an xn on the circle |z| = R, we find that P (x) has
exactly n zeros inside the circle. Also, since we may choose R arbitrarily large, so there are no additional
zeros.
Note that the above proof also gives a bound (although rather weak) for the zeros of a polynomial. This
bound is attributed to Cauchy.
Finally, the following result is a slightly stronger version of Rouché’s theorem.
Theorem 9 (Extended Rouché). Let f and g be analytic functions on and inside a simple closed curve C.
Suppose that
|f (z) + g(z)| < |f (z)| + |g(z)|
for all points z on C. Then f and g have the same number of zeros (counting multiplicities) interior to C.
There are many ways of bounding polynomial zeros on the complex plane. The following result is worth
mentioning, as it has proven useful quite a few times.
Proposition 10. Let P (x) = a0 + a1 x + · · · + an xn , where 0 < a0 ≤ a1 ≤ · · · ≤ an are real numbers, then
any complex zero of the polynomial satisfies |z| ≤ 1.
Proof. If |z| > 1, then, since z is a zero of (1 − x)P (x), we get

a0 + (a1 − a0 )z + · · · + (an − an−1 )z n − an z n = 0.

Thus,

|an z n | = |a0 + (a1 − a0 )z + · · · + (an − an−1 )z n |

≤ a0 + (a1 − a0 )|z| + · · · + (an − an−1 )|z n |
< a0 |z|n + (a1 − a0 )|z|n + · · · + (an − an−1 )|z|n
= a0 |z|n − a0 |z|n + a1 |z|n − a1 |z|n + · · · + an |z|n
= |an z n |

contradiction. Therefore, |z| ≤ 1.

It follows as a simple corollary that for any polynomial with positive real coefficients, P (x) = a0 + a1 x +
· · · + an xn , all its zeros lie in the annulus
ak−1 ak−1
min ≤ |z| ≤ max
1≤k≤n ak 1≤k≤n ak

Finally, we present one more irreducibility criterion, known as Cohn’s criterion. Essentially, it says
that if f (x) has nonnegative integer coefficients, and f (n) is prime for some n greater than all the coefficients,
then f is irreducible.
Theorem 11 (Cohn’s Criterion). Let p be a prime number, and b ≥ 2 an integer. Suppose that pn pn−1 · · · p1 p0
is the base-b representation of p, with 0 ≤ pi < b for each i and pn 6= 0, then the polynomial

f (x) = pn xn + pn−1 xn + · · · + p1 x + p0

is irreducible.
The following proof is due to M. Ram Murty1 .
As before, we begin with a lemma bounding the complex zeros of the polynomial.
1 M. Ram Murty, Prime Numbers and Irreducible Polynomials, Amer. Math. Monthly. 109 (2002) 452–458

5
MOP 2007 Black Group Integer Polynomials Yufei Zhao

Lemma 2. Let f (x) = an xn + an−1 xn−2 + · · · + a1 x + a0 belong to Z[x]. Suppose that an ≥ 1, an−1 ≥ 0,
and |ai | ≤ H for i = 0, 1, . . . , n − 2, where H is some positive constant. Then any complex zero α of f (x)
either has nonpositive real part, or satisfies
√
1 + 1 + 4H
|α| <
2
Proof. If |z| > 1 and Re z > 0, we observe that

f (z) an−1 1 1
≥ a n + − H + · · ·
zn z |z|2 |z|n
an−1 H
> Re an + − 2
z |z| − |z|
H |z|2 − |z| − H
≥1− 2 = ≥0
|z| − |z| |z|2 − |z|

whenever √
1+ 1 + 4H
|z| ≥ .
2
√
1+ 1+4H
It follows that α cannot be a zero of f if |α| ≥ 2 and Re α > 0.
To prove Theorem 11 for the case b ≥ 3, we notice that Lemma 2 implies if α is a zero of f (x),
then |b − α| > 1. Suppose that f (x) = g(x)h(x), where g and h are nonconstant integer polynomials.
Since f (b) is prime, one of |g(b)|, |h(b)| is 1. Say |g(b)| = 1, and the zeros of g are α1 , . . . , αk . We have
|g(b)| = |b − α1 | · · · |b − αk | > 1, contradiction. Therefore, f is irreducible.
The b = 2 case is special, and requires more analysis.
Lemma 3. Let f (x) = xn + an−1 xn−2 + · · · + a1 x + a0 with ai ∈ {0, 1} for each i. Then all the zeros of f
lie in the half plane Re z < 23 .
Proof. The cases n = 1 and 2 can be verified by hand. Assume that n ≥ 3. Then, for z 6= 0, we have

f (z) am−1 am−2 1 1 am−1 am−2 1
≥ 1 + + − + · · · + > 1+ + − 2 .
zn z z2 |z|3 |z|m z z2 |z| (|z| − 1)

If z satisfies | arg z| ≤ π/4, then we have Re(1/z 2 ) ≥ 0, and we get

f (z) 1
>1− 2 .
zn |z| (|z| − 1)
2 3
If |z| ≥ 23 , then |z|2 (|z| − 1) ≥ 32 9

2 − 1 = 8 > 1, and so f (z) 6= 0. On the other hand, √
if z is a zero
1+ 5
of f with | arg z| > π/4, and suppose that Re z > 0, then from Lemma 2 we have |z| < 2 , and thus
√
Re z < 1+ √ 5 < 3 . It follows that all zeros of f lie in the half-plane Re z < 3 .
2 2 2 2

To finish off the proof, suppose that f (x) = g(x)h(x), where g and h are integer polynomials. Since f (2)
is prime, one of |g(2)|, |h(2)| is 1. Say |g(2)| = 1. By Lemma 3, all the zeros of f lie in the half plane
Re z < 23 , which means that they satisfy |z − 2| > |z − 1|. Thus, if α1 , . . . , αk are the zeros of g, we have
|g(2)| = |2 − α1 | · · · |2 − αk | > |1 − α1 | · · · |1 − αk | = |g(1)| ≥ 1. So |g(2)| > 1, contradiction.

6
MOP 2007 Black Group Integer Polynomials Yufei Zhao

Problems
1. If q is a rational number and cos qπ is also rational, show that cos qπ ∈ {0, ± 12 , ±1}.
2. Let P (x) be a monic polynomial with integer coefficients such that all its zeros lie on the unit circle.
Show that all the zeros of P (x) are roots of unity, i.e., P (x)|(xn − 1)k for some n, k ∈ N.
3. If P (x) is a polynomial that such that P (n) is an integer for every integer n, then show that

x x x
P (x) = cn + cn−1 + · · · + c0 ,
n n−1 0

for some integers cn , . . . , c0 . (Note that the coefficients of P are not necessarily integers.)
4. Let f be an irreducible polynomial in Z[x], show that f has no multiple roots.
5. Player A and B play the following game. Player A thinks of a polynomial, P (x), with non-negative
integer coefficients. Player B may pick a number a, and ask player A to return the value of P (a), and
then player B may choose another number b and ask player A to return the value of P (b). After the
two questions, player B must guess P (x). Does player B have a winning strategy?
6. Determine all pairs of polynomials f, g ∈ Z[x], such that f (g(x)) = x2007 + 2x + 1.
7. (a) (USAMO 1974) Let a, b, c be three distinct integers, and let P be a polynomial with integer
coefficients. Show that in this case the conditions P (a) = b, P (b) = c, P (c) = a cannot be
satisfied simultaneously.
(b) Let P (x) be a polynomial with integer coefficients, and let n be an odd positive integer. Suppose
that x1 , x2 , . . . , xn is a sequence of integers such that x2 = P (x1 ), x3 = P (x2 ), . . . , xn = P (xn−1 ),
and x1 = P (xn ). Prove that all the xi ’s are equal.2
(c) (Putnam 2000) Let f (x) be a polynomial with integer coefficients. Define a sequence a0 , a1 , . . .
of integers such that a0 = 0 and an+1 = f (an ) for all n ≥ 0. Prove that if there exists a positive
integer m for which am = 0 then either a1 = 0 or a2 = 0.
(d) (IMO 2006) 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 (x) . . . ))
| {z }
k P 0s

Prove that there are at most n integers t such that Q(t) = t.

8. (IMO Shortlist 1997) Find all positive integers k for which the following statement is true: if P (x) is a
polynomial with integer coefficients satisfying the condition 0 ≤ P (c) ≤ k for c = 0, 1, . . . , k + 1, then
F (0) = F (1) = · · · = F (k + 1).
9. Let f (x) = x4 + 6x2 + 1. Show that for any prime p, f (x) is reducible over Fp , but f (x) is irreducible
over Z.
10. Let m, n, and a be positive integers and p a prime number less than a − 1. Prove that the polynomial
f (x) = xm (x − a)n + p is irreducible.
11. Let p be prime. Show that f (x) = xp−1 + 2xp−2 + 3xp−3 + · · · + (p − 1)x + p is irreducible.
12. (IMO 1993) Let f (x) = xn + 5xn−1 + 3, where n > 1 is an integer. Prove that f (x) cannot be expressed
as the product of two nonconstant polynomials with integer coefficients.
13. (Romania TST 2003) Let f (x) ∈ Z[x] be an irreducible monic polynomial with integer coefficients.
Suppose that |f (0)| is not a perfect square. Show that f (x2 ) is also irreducible.
2 This problem appeared in Reid Barton’s handout in 2005. Compare with the IMO 2006 problem.

7
MOP 2007 Black Group Integer Polynomials Yufei Zhao

14. Let z1 , z2 , . . . , zn ∈ Z[i] be Gaussian integers (i.e., complex numbers of the form a + bi, where a, b ∈ Z)
such that |zi − z1 | > 2 for all i > 1. Prove that the polynomial (x − z1 )(x − z2 ) · · · (x − zn ) + 1 is
irreducible over Z[i].
15. (Brazil 2006) Let f (x) be an irreducible polynomial, and suppose that it has two roots whose product
is 1. Show that the degree of f is even.

16. (MathLinks Contest) Let a be a nonzero integer, and n ≥ 3 be another integer. Prove that the the
following polynomial is irreducible over the integers:

P (x) = xn + axn−1 + axn−2 + · · · + ax − 1.

17. Let a1 ≥ a2 ≥ · · · ≥ an > 0 be positive integers. Show that the following polynomial is irreducible:

P (x) = xn − a1 xn−1 − a2 xn−2 − · · · − an

18. (MOP 2007) Let p(x) be a polynomial with integer coefficients. Determine if there always exists a
positive integer k such that p(x) − k is irreducible.

19. (Iran TST 2007) Does there existPa sequence a0 , a1 , a2 , . . . in N, such that for each i 6= j, gcd(ai , aj ) = 1,
n
and for each n, the polynomial i=0 ai xi is irreducible in Z[x]?
20. (China TST Quizzes 2006) Let n be a positive integer, and let A1 , A2 , . . . , Ak be a partition of the
set of positive integers. Show that for some i ∈ {1, 2, . . . , k}, there are infinitely many irreducible
polynomials of degree n and whose coefficients are distinct elements from Ai .
21. Prove that xn − x − 1 is irreducible over the integers for all n ≥ 2.
22. (Iran 2003) Let f1 , f2 , . . . , fn be polynomials with integer coefficients. Show that there exists a reducible
polynomial g(x) ∈ Z[x] such that fi (x) + g(x) is irreducible for i = 1, 2, . . . , n.

23. (IMO Shortlist 1997) Let f be a polynomial with integer coefficients and let p be a prime such that
f (0) = 0, f (1) = 1, and f (k) ≡ 0 or 1 (mod p) for all positive integers k. Show that deg f ≥ p − 1.
24. (IMO Shortlist 2005) Find all monic integer polynomials p(x) of degree two for which there exists an
integer polynomial q(x) such that p(x)q(x) is a polynomial having all coefficients ±1.

25. (IMO Shortlist 2005) Let a, b, c, d, e and f be positive integers. Suppose that the sum S = a + b + c +
d + e + f divides both abc + def and ab + bc + ca − de − ef − f d. Prove that S is composite.
26. (IMO 2002) Find all pairs of integers m, n ≥ 3 such that there exist infinitely many positive integers
a for which
am + a − 1
an + a2 − 1
is an integer.
27. (IMO Shortlist 2002) Let P (x) be a cubic polynomial with integer coefficients. Suppose that xP (x) =
yP (y) for infinitely many pairs x, y of integers with x 6= y. Prove that the equation P (x) = 0 has an
integer root.

28. (IMO Shortlist 1996) For each positive integer n, show that there exists a positive integer k such that

k = f (x)(x + 1)2n + g(x)(x2n + 1)

for some polynomials f, g with integer coefficients, and find the smallest such k as a function of n.
n
29. (Romania TST 1998) show that for any n ∈ N, the polynomial P (x) = (x2 + x)2 + 1 is irreducible
over the integers.

8
Winter Camp 2008 Inequalities Yufei Zhao

Inequalities
Yufei Zhao
[email protected]

1 Classical Theorems
Theorem 1. (AM-GM) Let a1 , · · · , an be positive real numbers. Then, we have
a1 + · · · + an √
≥ n a1 · · · an .
n
Theorem 2. (Cauchy-Schwarz) Let a1 , · · · , an , b1 , · · · , bn be real numbers. Then,

(a1 2 + · · · + an 2 )(b1 2 + · · · + bn 2 ) ≥ (a1 b1 + · · · + an bn )2 .

Theorem 3. (Jensen) Let f : [a, b] → R be a convex function. Then for any x1 , x2 , . . . , xn ∈ [a, b] and any
nonnegative reals ω1 , ω2 , . . . , ωn with ω1 + ω2 + · · · + ωn = 1, we have

ω1 f (x1 ) + ω2 f (x2 ) + · · · + ωn f (xn ) ≥ f (ω1 x1 + ω2 x2 + · · · + ωn xn ).

If f is concave, then the inequality is flipped.

Theorem 4. (Weighted AM-GM) Let ω1 , · · · , ωn > 0 with ω1 + · · · + ωn = 1. For all x1 , · · · , xn > 0,

we have
ω1 x1 + ω2 x2 + · · · + ωn xn ≥ xω1 ω2 ωn
1 x2 · · · xn .

Theorem 5. (Schur) Let x, y, z be nonnegative real numbers. For any r > 0, we have
X
xr (x − y)(x − z) ≥ 0.
cyclic

Definition 1. (Majorization) Let x = (x1 , x2 , . . . xn ) and y = (y1 , y2 , . . . , yn ) be two sequences of real

numbers. Then x is said to majorize y (denoted x y) if the following conditions are satisfied
• x1 ≥ x2 ≥ x3 · · · ≥ xn and y1 ≥ y2 ≥ y3 · · · ≥ yn ; and
• x1 + x2 + · · · + xk ≥ y1 + y2 + · · · + yk , for k = 1, 2, . . . , n − 1; and

• x1 + x2 + · · · + xn = y1 + y2 + · · · + yn .
Theorem 6. (Muirhead)1 Suppose that (a1 , . . . , an ) (b1 , . . . , bn ), and x1 , . . . xn are positive real numbers,
then X X
xa1 1 xa2 2 · · · xann ≥ xb11 xb22 · · · xbnn .
sym sym

where the symmetric sum is taken over all n! permutations of x1 , x2 , . . . , xn .

Theorem 7. (Karamata’s Majorization inequality) Let f : [a, b] → R be a convex function. Suppose
that (x1 , · · · , xn ) (y1 , · · · , yn ), where x1 , · · · , xn , y1 , · · · , yn ∈ [a, b]. Then, we have

f (x1 ) + · · · + f (xn ) ≥ f (y1 ) + · · · + f (yn ).

1 Practical notes about Muirhead: (1) don’t try to apply Muirhead when there are more than 3 variables, since mostly likely

you won’t succeed (and never, ever try to use Muirhead when the inequality is only cyclic but not symmetric, since it is incorrect
to use Muirhead there) (2) when writing up your solution, it is probably safer to just deduce the inequality using weighted
AM-GM by finding the appropriate weights, as this can always be done. The reason is that it is not always clear that Muirhead
will be accepted as a quoted theorem.

1
Winter Camp 2008 Inequalities Yufei Zhao

Theorem 8. (Power Mean) Let x1 , · · · , xn > 0. The power mean of order r is defined by
r1
xr1 + · · · + xn r

√
M(x1 ,··· ,xn ) (0) = n
x1 · · · xn , M(x1 ,··· ,xn ) (r) = (r 6= 0).
n
Then, M(x1 ,··· ,xn ) : R −→ R is continuous and monotone increasing.
Theorem 9. (Bernoulli) For all r ≥ 1 and x ≥ −1, we have

(1 + x)r ≥ 1 + rx.

Definition 2. (Symmetric Means) For given arbitrary real numbers x1 , · · · , xn , the coefficient of tn−i in
the polynomial (t + x1 ) · · · (t + xn ) is called the i-th elementary symmetric function σi . This means that

(t + x1 ) · · · (t + xn ) = σ0 tn + σ1 tn−1 + · · · + σn−1 t + σn .

For i ∈ {0, 1, · · · , n}, the i-th elementary symmetric mean Si is defined by

σi
Si = n .

i

Theorem 10. Let x1 , . . . , xn > 0. For i ∈ {1, · · · , n}, we have

Si Si−1
(1) (Newton) Si+1 ≥ Si ,
1 1
(2) (Maclaurin) Si ≥ Si+1 i+1 .
i

Theorem 11. (Rearrangement) Let x1 ≥ · · · ≥ xn and y1 ≥ · · · ≥ yn be real numbers. For any

permutation σ of {1, . . . , n}, we have
n
X n
X n
X
xi yi ≥ xi yσ(i) ≥ xi yn+1−i .
i=1 i=1 i=1

Theorem 12. (Chebyshev) Let x1 ≥ · · · ≥ xn and y1 ≥ · · · ≥ yn be real numbers. We have

x1 y1 + · · · + xn yn x1 + · · · + xn y1 + · · · + yn
≥ .
n n n

Theorem 13. (Hölder)2 Let x1 , · · · , xn , y1 , · · · , yn be positive real numbers. Suppose that p > 1 and q > 1
satisfy p1 + 1q = 1. Then, we have

n n
! p1 n
! q1
X X X
p q
xi yi ≤ xi yi .
i=1 i=1 i=1

More generally, let xij (i = 1, · · · , m, j = 1, · · · n) be positive real numbers. Suppose that ω1 , · · · , ωn are
positive real numbers satisfying ω1 + · · · + ωn = 1. Then, we have
!ωj  
Yn m
X Xm Yn
xij ≥  xij ωj  .
j=1 i=1 i=1 j=1

Theorem 14. (Minkowski)3 If x1 , · · · , xn , y1 , · · · , yn > 0 and p > 1, then

n
! p1 n
! p1 n
! p1
X X X p
p p
xi + yi ≥ (xi + yi )
i=1 i=1 i=1
2 Think of this as generalized Cauchy, as you can use it for more than two sequences.
3 Think of this as generalized triangle inequality.

2
Winter Camp 2008 Inequalities Yufei Zhao

2 A motivating example
In this section we discuss several common techniques in inequalities through following famous problem from
IMO 2001 by Hojoo Lee:
Prove that
a b c
√ +√ +√ ≥1
a2
+ 8bc 2
b + 8ca 2
c + 8ab
for all positive real numbers a, b and c.
The official solution is short and mysterious:

Official solution: First we prove that

4
a a3
√ ≥ 4 4 4 ,
2
a + 8bc a3 + b3 + c3
or equivalently, that
4 4 4 2
(a 3 + b 3 + c 3 )2 ≥ a 3 (a2 + 8bc).
Indeed, this follows from applying AM-GM as follows:
4 4 4 4 4 4 4 4 4 4 4 2 2 2 1 1 8 2 2
(a 3 +b 3 +c 3 )2 = (a 3 )2 +(b 3 +c 3 )(a 3 +a 3 +b 3 +c 3 ) ≥ (a 3 )2 +2b 3 c 3 ·4a 3 b 3 c 3 = a 3 +8a 3 bc = a 3 (a2 +8bc).

We have similar inequalities for the other two terms, and so

4 4 4
a b c a3 b3 c3
√ +√ +√ ≥ 4 4 4 + 4 4 4 + 4 4 4 = 1.
2
a + 8bc 2
b + 8ca 2
c + 8ab a +b +c
3 3 3 a +b +c
3 3 3 a + b3 + c3
3

While the solution looks nice and short, it leaves us wonder how in the world could anyone come up with
it. In particular, where did the exponent 34 come from? Here we provide some motivation.

Isolated fudging. It is not unusual to compare individual terms of an inequality to expressions such as
ar
ar + br + cr
because if the comparison turns out to be successful, we can finish off the problem right away. (Techniques of
this form are sometimes called isolated fudging, meaning that the effort is focused on manipulating individual
terms, as opposed to the inequality as a whole.) So, suppose we guess that it is possible to have an inequality
of the form
a ar
√ ≥ r .
a2 + 8bc a + br + cr
Now how can we pick the candidates for r? Blind guess and check will probably get us nowhere. Luckily,
there is a method that will give you the unique candidate for r (though with no promise that this r will
work). Be prepared, this method will require some calculus.
Suppose that some r works. Let us consider the function
a ar
f (a, b, c) = √ − r .
a2 + 8bc a + br + cr
What do we know about f (a, b, c)? Because of the inequality that we want, we need f (a, b, c) ≥ 0 for all
a, b, c > 0. Also, by considering the point of equality, we see that (1, 1, 1) must be a local minimum of f . So
consider the partial derivative of f with respect to a (denoted ∂f /∂a), meaning that we differentiate f with
respect to a while treating the other variables as if they were constants. Since (1, 1, 1) is a local minimum,
this partial derivative when evaluated at (1, 1, 1) must give zero. So let’s do this computation. (If you know
some multivariable calculus, it may be instructive to think in terms of ∇f = 0 at (1, 1, 1).)

3
Winter Camp 2008 Inequalities Yufei Zhao

We have √ 2
a2 + 8bc − √ a
∂f a2 +8bc rar−1 (ar + br + cr ) − ar · rar−1
= − .
∂a a2 + 8bc (ar + br + cr )2
3− 1
Evaluating at (a, b, c) = (1, 1, 1) and setting the value to zero, we get 9 3 − 3r−r
9 = 0, which gives r = 43 .
Aha! Now we found the candidate for r, we can now go through the steps of the official solution (which
were easy, given we know what r is).
Note that while we used calculus to motivate our solution, we do not need to include any
calculus in the solution! In fact, calculus is best avoided in olympiad solutions as it is generally viewed
unfavorably. However, since we are not required to provide the motivation to our proof, we do not need to
worry about this issue. Amazing, isn’t it?

Turning the table around. Seeing that the inequality is homogeneous (meaning that the transformation
(a, b, c) 7→ (ka, kb, kc) does not change anything), it is natural to impose a constraint on it. So let us assume
without the loss of generality that abc = 18 , so that we need to prove

a b c
q +q +q ≥ 1.
2 1 1 1
a + a b2 + b c2 + c

This does not seem any easier. Now, let us turn the table around and switch the roles of the constraint and
the inequality.
Let
a b c
x= q , y=q , z=q .
a2 + a1 b2 + 1b c2 + 1c
a q 1
Note that √ 1
= , so we can write a, b, c in terms of x, y, z:
a2 + a 1+ a13

x2 y2 z2
a3 = , b3 = , c3 = .
1 − x2 1 − y2 1 − z2

So, the inequality that we wish to prove is x+y +z ≥ 1 assuming abc = 81 . By considering the contrapositive,
it suffices to prove that abc < 18 assuming x + y + z < 1. That is, we need to prove that

x2 y 2 z 2 1
< 3
(1 − x2 )(1 − y 2 )(1 − z 2 ) 8

given x + y + z < 1. The square roots are now gone! The new inequality turns out to be extremely easy, as
it is merely a straightforward application of AM-GM. Indeed, we have
1 3 3
1 − x2 > (x + y + z)2 − x2 = y 2 + z 2 + xy + xy + yz + yz + xz + xz ≥ 8x 2 y 4 z 4

and similarly for 1 − y 2 and 1 − z 2 . Setting this back into about inequality gives the desired result.

How to use Jensen. Since the inequality is homogeneous, we can assume that a + b + c = 1. Note that
the function x 7→ √1x is convex. So we can use Jensen’s inequality as follows:

1 1 1 1
a· √ +b· √ +c· √ ≥p
a2 + 8bc 2
b + 8ca 2
c + 8ab a(a + 8bc) + b(b + 8ca) + c(c2 + 8ab)
2 2

1
=√ .
a3 + b3 + c3 + 24abc
So it remains to prove that
1
√ ≥1
a3 + b3 + c3 + 24abc

4
Winter Camp 2008 Inequalities Yufei Zhao

or equivalently,
a3 + b3 + c3 + 24abc ≤ (a + b + c)3 .
This is again extremely easy, as AM-GM gives

(a + b + c)3 − a3 − b3 − c3 = 3(a2 b + b2 a + b2 c + c2 b + c2 a + a2 c) + 6abc ≥ 24abc.

3 Problems
The following problems are selected from a packet written by Thomas Mildorf, which can be found at
http://web.mit.edu/tmildorf/www/Inequalities.pdf. Solutions can also be found there.

1. Show that for positive reals a, b, c

a2 b + b2 c + c2 a ab2 + bc2 + ca2 ≥ 9a2 b2 c2 .

2. Let a, b, c be positive reals such that abc = 1. Prove that

a + b + c ≤ a2 + b2 + c2 .

3. Let P (x) be a polynomial with positive coefficients. Prove that if

1 1
P ≥
x P (x)
holds for x = 1, then it holds for all x > 0.
4. Show that for all positive reals a, b, c, d,
1 1 4 16 64
+ + + ≥ .
a b c d a+b+c+d

5. (USAMO 1980/5) Show that for all non-negative reals a, b, c ≤ 1,

a b c
+ + + (1 − a)(1 − b)(1 − c) ≤ 1.
b+c+1 c+a+1 a+b+1

6. (USAMO 1977/5) If a, b, c, d, e are positive reals bounded by p and q with 0 < p ≤ q, prove that
r r 2
1 1 1 1 1 p q
(a + b + c + d + e) + + + + ≤ 25 + 6 −
a b c d e q p
and determine when equality holds.
7. Let a, b, c, be non-negative reals such that a + b + c = 1. Prove that
1
a3 + b3 + c3 + 6abc ≥ .
4

8. (IMO 1995/2) a, b, c are positive reals with abc = 1. Prove that

1 1 1 3
+ + ≥ .
a3 (b + c) b3 (c + a) c3 (a + b) 2

9. Let a, b, c be positive reals such that abc = 1. Show that

2 2 2
+ + ≤ 1.
(a + 1)2+ b + 1 (b + 1) + c + 1 (c + 1) + a2 + 1
2 2 2 2

5
Winter Camp 2008 Inequalities Yufei Zhao

10. (USAMO 1998/3) Let a0 , . . . , an be real numbers in the interval (0, π2 ) 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 .

11. (Romanian TST) Let a, b, x, y, z be positive reals. Show that

x y z 3
+ + ≥ .
ay + bz az + bx ax + by a+b

12. The numbers x1 , x2 , . . . , xn obey −1 ≤ x1 , x2 , . . . , xn ≤ 1 and x13 + x23 + · · · + xn3 = 0. Prove that
n
x1 + x2 + · · · + xn ≤ .
3

13. (Turkey) Let n ≥ 2 be an integer, and x1 , x2 , . . . , xn positive reals such that x21 + x22 + · · · + x2n = 1.
Determine the smallest possible value of

x51 x52 x5n

+ + ··· + .
x2 + x3 + · · · + xn x3 + · · · + xn + x1 x1 + · · · + xn−1

14. (Poland 1995) Let n be a positive integer. Compute the minimum value of the sum

x22 x3 xn
x1 + + 3 + ··· + n,
2 3 n
where x1 , x2 , . . . , xn are positive reals such that
1 1 1
+ + ··· + = n.
x1 x2 xn

15. Prove that for all positive reals a, b, c, d,

a4 b + b4 c + c4 d + d4 a ≥ abcd(a + b + c + d).

16. (Vietnam 1998) Let x1 , . . . , xn be positive reals such that

1 1 1 1
+ + ··· + =
x1 + 1998 x2 + 1998 xn + 1998 1998
Prove that √
n
x1 x2 · · · xn
≥ 1998.
n−1
17. (MOP 2002) Let a, b, c be positive reals. Prove that
23 23 32
2a 2b 2c
+ + ≥3
b+c c+a a+b

18. (Iran 1996) Show that for all positive real numbers a, b, c,

1 1 1 9
(ab + bc + ca) + + ≥
(a + b)2 (b + c)2 (c + a)2 4

6
IMO Training 2008 Polynomials Yufei Zhao

Polynomials
July 2, 2008
Yufei Zhao
[email protected]

1 Roots of unity
1. (USAMO 1976) The polynomials A(x), B(x), C(x), D(x) satisfy the equation

A(x5 ) + xB(x5 ) + x2 C(x5 ) = (1 + x + x2 + x3 + x4 )D(x).

Show that A(1) = 0.

2. A sequence a1 , a2 , . . . , an is called k-balanced if a1 + ak+1 + · · · = a2 + · · · + ak+2 + · · · = · · · =

ak + a2k + · · · . Suppose the sequence a1 , a2 , . . . , a50 is k-balanced for k = 3, 5, 7, 11, 13, 17. Prove
that all the values ai are zero.

3. Let P (x) be a monic polynomial with integer coefficients such that all its zeros lie on the unit
circle. Show that all the zeros of P (x) are roots of unity, i.e., P (x)|(xn − 1)k for some n, k ∈ N.

2 Integer divisibility
The main lesson, as illustrated by the first set of problems here, is that if P (x) has integer coefficients,
then a − b | P (a) − P (b).

4. (a) (USAMO 1974) Let a, b, c be three distinct integers, and let P be a polynomial with integer
coefficients. Show that in this case the conditions P (a) = b, P (b) = c, P (c) = a cannot be
satisfied simultaneously.
(b) Let P (x) be a polynomial with integer coefficients, and let n be an odd positive integer. Sup-
pose that x1 , x2 , . . . , xn is a sequence of integers such that x2 = P (x1 ), x3 = P (x2 ), . . . , xn =
P (xn−1 ), and x1 = P (xn ). Prove that all the xi ’s are equal.1
(c) (Putnam 2000) Let f (x) be a polynomial with integer coefficients. Define a sequence
a0 , a1 , . . . of integers such that a0 = 0 and an+1 = f (an ) for all n ≥ 0. Prove that if
there exists a positive integer m for which am = 0 then either a1 = 0 or a2 = 0.
(d) (IMO 2006) 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 (x) . . . ))
| {z }
k P 0s

Prove that there are at most n integers t such that Q(t) = t.

a b c a c b
5. Let a, b, c be nonzero integers such that both b + c + a and c + b + a are integers. Prove that
|a| = |b| = |c|.
1
This problem appeared in Reid Barton’s MOP handout in 2005. Compare with the IMO 2006 problem.

1
IMO Training 2008 Polynomials Yufei Zhao

6. (IMO Shortlist 2005) Let a, b, c, d, e and f be positive integers. Suppose that the sum S =
a + b + c + d + e + f divides both abc + def and ab + bc + ca − de − ef − f d. Prove that S is
composite.

3 Crossing the x-axis

For any continuous function (e.g. polynomial) f , if f (a) and f (b) have different signs for some a < b,
then there must exist a t ∈ (a, b) such that f (t) = 0.

7. (China 1995) Alice and Bob play a game with a polynomial of degree at least 4:

x2n + x2n−1 + x2n−2 + · · · + x + 1.

They fill in real numbers to empty boxes in turn. If the resulting polynomial has no real root,
Alice wins; otherwise, Bob wins. If Alice goes first, who has a winning strategy?

8. (USAMO 2002) Prove that any monic polynomial of degree n with real coefficients is the average
of two monic polynomials of degree n with n real roots.

4 Lagrange and Chebyshev

Lagrange interpolation. If (x1 , y1 ), . . . , (xn , yn ) are points in the plane with distinct x-coordinates,
then there exists a unique polynomial P (x) of degree at most n − 1 passing through these points, and
it is given by the expression
X n Y x − xj
P (x) = yi .
xi − xj
i=1 j6=i

You may have seen many problems that can be solved directly using interpolation (e.g., here are the
values of P (1), P (2), . . . , P (n), what’s the value of P (n + 1)). The following problems require more
subtle uses of interpolation.

Chebyshev polynomials. These are polynomials satisfying

Tn (cos θ) = cos nθ.

One can show using induction Tn is indeed a polynomial, and has integer coefficients, with leading
coefficient 2n . Chebyshev polynomials (including its variants) are often useful because they are nicely
bounded in [−1, 1], so that they often serve as equality cases. Specifically, we have

|Tn (x)| ≤ 1, whenever x ∈ [−1, 1].

Outside of [−1, 1], the values of Tn (x) can be found through Tn 12 x + x1 = 12 xn + x1n (why?).

A common variant of Chebyshev polynomials is the class of polynomials satisfying Pn (2 cos θ) =

2 cos nθ. One can show that Pn (x) is a monic integer polynomial. It also satisfies Pn (x+x−1 ) = xn +x−n .

9. Show that if f (x) is a monic polynomial of degree n − 1, and a1 , a2 , . . . , an distinct real numbers,
then
n
X f (ai )
Q =1
i=1 j6=i (aj − ai )

2
IMO Training 2008 Polynomials Yufei Zhao

10. (IMO Shortlist 1997) Let f be a polynomial with integer coefficients and let p be a prime such
that f (0) = 0, f (1) = 1, and f (k) ≡ 0 or 1 (mod p) for all positive integers k. Show that
deg f ≥ p − 1.
11. Let P be a polynomial of degree n with real coefficients such that |f (x)| ≤ 1 for all x ∈ [0, 1].
Show that |f (− n1 )| ≤ 2n+1 − 1.
12. Let P (x) be a monic degree n polynomial with real coefficients. Prove that there is some t ∈ [−1, 1]
such that |P (t)| ≥ 21n .
13. (Walter Janous, Crux) Suppose that a0 , a1 , . . . , an are real numbers such that for all x ∈ [−1, 1],
|a0 + a1 x + · · · + an xn | ≤ 1. Show that for all x ∈ [−1, 1], |an + an−1 x + · · · + a0 xn | ≤ 2n−1
14. Let x1 , x2 , . . . , xn , n ≥ 2, be n distinct real numbers in the interval [−1, 1]. Prove that
1 1 1
+ + ··· + ≥ 2n−2 ,
t1 t2 tn
Q
where ti = j6=i |xj − xi |.

5 Irreducibility through mods

In abstract algebra language, if A is a UFD (unique factorization domain), then so is A[x]. In particular,
fields are automatically UFDs, so that K[x] is a UFD whenever K is a field. Useful examples of UFDs
include: Z[x], R[x], C[x], Z[x, y], Fp [x].
The last example is especially worth mentioning. Yes, unique factorization holds even when the coef-
ficients of the polynomial is considered in mod p (where p must be prime). This means that when we
are considering factorizations of integers polynomials f (x) = g(x)h(x), it may be helpful to reduce the
problem to f¯(x) = ḡ(x)h̄(x), where the coefficients are considered in mod p.

15. (a) (Eisenstein’s criterion) Let f (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial with
integer coefficients such that p | ai for 0 ≤ i ≤ n − 1, p - an and p2 - a0 . Then f (x) is
irreducible.
(b) Let f (x) = an xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial with integer coefficients such
that p | ai for 0 ≤ i ≤ n − k, p - ak and p2 - a0 . Then f (x) has an irreducible factor of degree
greater than k.
16. Let p be a prime number. Prove that xp−1 + xp−2 + · · · + 1 is irreducible. (This is the example
that follows every exposition of Eisenstein’s criterion.)
n
17. Let n be a positive integer. Prove that (x2 + x)2 + 1 is irreducible.

6 Irreducibility through roots

Interestingly enough, when trying to prove that a certain integer polynomial is irreducible, it can be
usual to examine its complex roots.

18. Let f (x) = an xn + an−1 xn + · · · + a1 x + a0 be a polynomial with integer coefficients, such that
|a0 | is prime and
|a0 | > |a1 | + |a2 | + · · · + |an |.
Show that f (x) is irreducible.

3
IMO Training 2008 Polynomials Yufei Zhao

19. Let P (x) = a0 + a1 x + · · · + an xn , where 0 < a0 ≤ a1 ≤ · · · ≤ an are real numbers. Prove that
any complex zero of the polynomial satisfies |z| ≤ 1.

20. Let p be a prime. Prove that xp−1 + 2xp−2 + 3xp−3 + · · · + (p − 1)x + p is irreducible.

21. (Cohn’s criterion) Suppose that pn pn−1 · · · p1 p0 is the base-10 representation of a prime number
p, with 0 ≤ pi < 10 for each i and pn 6= 0. Show that the polynomial

f (x) = pn xn + pn−1 xn + · · · + p1 x + p0

is irreducible.

22. (Romania TST 2003) Let f (x) ∈ Z[x] be an irreducible monic polynomial with integer coefficients.
Suppose that |f (0)| is not a perfect square. Show that f (x2 ) is also irreducible.

7 Rouché’s theorem (optional)

The following theorem from complex analysis can be useful in locating the zeros of a polynomial.

Theorem (Rouché). Let f and g be analytic functions (e.g. polynomials) on and inside a simple closed
curve C (e.g. a circle). Suppose that |f (z)| > |g(z)| for all points z on C. Then f and f − g have the
same number of zeros (counting multiplicities) interior to C.

23. Let P (z) = an z n + an−1 z n−1 + · · · + a1 z + a0 be a polynomial with complex coefficients, and such
that
|ak | > |a0 | + |a1 | + · · · + |ak−1 | + |ak+1 | + · · · + |an |
for some 0 ≤ k ≤ n. Show that exactly k zeros of P lie strictly inside the unit circle, and the
other n − k zeros of P lie strictly outside the unit circle.

24. (Perron’s criterion) Let P (x) = xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial with a0 6= 0 and

|an−1 | > 1 + |an−2 | + · · · + |a1 | + |a0 |.

Then P (x) is irreducible.

25. (IMO 1993) Let f (x) = xn + 5xn−1 + 3, where n > 1 is an integer. Prove that f (x) cannot be
expressed as the product of two nonconstant polynomials with integer coefficients.

26. (Romania ??) Let f ∈ C[x] be a monic polynomial. Prove that we can find a z ∈ C such that
|z| = 1 and |f (z)| ≥ 1.

4
UMA Putnam Talk Lecture Notes

Determinants: Evaluation and Manipulation

Yufei Zhao
September 22, 2009

Appetizer problem
(This problem doesn’t actually use determinants.)
Problem 1. Do there exist square matrices A and B such that AB − BA = I?
Solution. No. Take the trace of both sides and using tr(AB) = tr(BA), we get that tr(AB−BA) = 0
while tr(I) 6= 0.

1. Introduction
In this talk I’ll discuss some techniques on dealing with determinants that may be useful for the
Putnam exam. We will focus on the evaluation and manipulation of determinants. I won’t talk
about applications of determinants to, say, combinatorics (maybe another time).
We will assume familiarity with basic properties of determinants. Just a reminder, if A =
(aij )1≤i,j≤n is an n × n matrix, then
X
det A = sgn(σ)a1σ(1) a2σ(2) · · · anσ(n)
σ∈Sn

where the sum is taken over all permutations of {1, 2, . . . , n}.

Here’s an outline of techniques used to deal with determinants.

Evaluation Manipulation
• Row and column operations • Assume invertibility
• Expansion by minors • Block decomposition
• Setting variables / Vandermonde • Conjugation and positivity
• Eigenvalues / circulant matrices

2. Evaluation of determinants
I’ll talk about how to evaluate determinants when the entries are given.
The most basic (and often extremely useful) method is row/column operations and minor
expansions. Though I won’t discuss them here, since I want to talk more exciting techniques.
The first example is everyone’s favorite Vandermonde determinant.
Problem 2 (Vandermond determinant). Let
xn−1
 
1 x1 x21 · · · 1
1 x2 x2 · · ·
2 xn−1
2

V = . . ..  .
 
. . .. . .
. . . . . 
1 xn x2n · · · xn−1
n
Show that Y
det V = (xj − xi ).
1≤i<j≤n
1
2

Solution. Let
p(x1 , x2 , . . . , xn ) = det V,
viewed as a polynomial in n variables. Now, suppose we view p as a single-variable polynomial in
x1 with coefficients in Q(x2 , . . . , xn ). If we set x1 to xi , for any i 6= 1, then two rows of the matrix
are equal and hence the determinant vanishes, and therefore (x1 − xi ) must be a factor of p.
Similarly, every (xi −xj ) for i 6= j is a factor of p(x1 , . . . , xn ). But the degree of p is 21 n(n−1) (from
looking at the matrix), and we just showed that 1≤i<j≤n (xj − xi ) (which has degree 21 n(n − 1))
Q
divides p. Therefore, Y
p(x1 , x2 , . . . , xn ) = k (xj − xi ),
1≤i<j≤n
for some constant k. Comparing the coefficient of the term x2 x23 x34 · · · xnn−1 shows that k = 1.
Our next example is the circulant matrix.
Problem 3 (Circulant matrix). Let
 
a0 a1 a2 · · · an−1
an−1
 a0 a1 · · · an−2 

C = an−2 an−1
 a0 · · · an−3 

 .. .. .. . . .. 
 . . . . . 
a1 a2 a3 · · · a0
Then
n−1 n−1
!
Y X
jk
det C = ζ ak
j=0 k=0

where ζ = e2πi/n .
Solution. We know that the determinant equals to the product of the eigenvalues. The eigenvectors
of C are        
1 1 1 1
1  ζ   ζ2   ζ n−1 
4
   2     2(n−1) 
v0 = 1 v1 =  ζ  v2 =  ζ  · · · vn−1 = ζ .
       
 ..   ..   ..   .. 
.  .   .   . 
2
1 ζ n−1 ζ 2(n−2) ζ (n−1)
They are independent because of the Vandermonde determinant, so they form a complete set of
eigenvalues. The corresponding eigenvectors are
λ0 = a0 + a1 + a2 + · · · + an−1
λ1 = a0 + a1 ζ + a2 ζ 2 + · · · + an−1 ζ n
···
2
λn−1 = a0 + a1 ζ n−1 + a2 ζ 2(n−1) + · · · + an−1 ζ (n−1)
Thus det C = λ0 λ1 · · · λn−1 .
Now you have the tools to solves the following problem, which appeared as Putnam 1999/B5.
The highest score on his problem was 2 points by one contestant! By this measure, it is one of
the most difficult Putnam problems in history; but knowing the above technique is becomes not so
bad.
Problem 4 (Putnam 1999/B5). Let n ≥ 3. Let A be the n×n matrix with Ajk = cos(2π(j +k)/n).
Find det(I + A).
3

3. Manipulation of matrices
Now I’ll discuss some techniques on dealing with determinants of matrices without knowing their
entires. We will make repeated uses of the fact that det AB = det A det B for square matricies.
Problem 5. Let A and B be n × n matrices. Show that det(I + AB) = det(I + BA).
Solution. First, assume that A is invertible. Then
det(I + AB) = det(A(I + BA)A−1 ) = det A det(I + BA) det(A−1 ) = det(I + BA).
Now we give two ways of working around the assumption that A is invertible.
Method 1. For any t ∈ R, let At = A − tI. Then At is non-invertible precisely when t is
an eigenvalue of A. Thus, if t is not an eigenvalue, then det(I + At B) = det(I + BAt ). Now,
det(I + At B) − det(I + BAt ) is a polynomial in t which vanishes everywhere except for the finitely
many eigenvalues; hence det(I + At B) − det(I + BAt ) = 0 for all t. Setting t = 0 gives the result.
Method 2. View the entries of A and B as indeterminants, so that what we are proving is a
polynomial identity in {aij } ∪ {bij }. Work over the field Q(a11 , . . . , b11 , . . . ). Then in this field, A
is invertible, and the proof works.
Remark. The set of invertible matrices form a Zariski (dense) open subset, and hence to verify a
polynomial identity, it suffices to verify it on this dense subset.
Remark. The statement is also true when A and B are not square matrices. Specifically, suppose
that A is an n × m matrix, and B an m × n matrix, then det(In + AB) = det(Im + BA). To prove
this fact, extend A and B to square matrices by filling in zeros.
The technique of assuming invertibility is very powerful. Let us give another example.
Problem 6. Let A, B, C, D be n × n matrices such that AC = CA. Prove that

A B
det = det(AD − CB).
C D
Solution. First assume that A is invertible. Then

I O A B A B
= ,
−CA−1 I C D O D − CA−1 B
so that

A B A B
det = det = det A det(D − CA−1 B)
C D O D − CA−1 B
= det(AD − ACA−1 B) = det(AD − CB).
(We used the fact that A and C commute.)
Now we need to get rid of the invertibility assumption. Let At = A − tI. Since AC = CA, we
get At C = CAt for all t. It follows that

At B
det = det(At D − CB).
C D
whenever t is not an eigenvalue of A. But this is a polynomial equation in t, which holds for all
but finitely many t’s, and hence it must hold for all t. In particular, setting t = 0 gives the desired
result.
Finally, let us look at a few problems involving inequalities.
Problem 7. Let A be a square matrix with real entries. Show that det(A2 + I) ≥ 0.
4

One way to solve this problem is to look at the eigenvalues of A. If the eigenvalues of A are {λi }
i.e., counting multiplicities), then the eigenvalues of A2 + I are {λ2i + 1}, and hence
(as a multiset, Q
det(A2 + 1) = i (λ2i + 1). Finally use the fact that all non-real eigenvalues λi come in conjugate
pairs.
Here is a much slicker solution.
Proof. We have A2 + I = (A + iI)(A − iI). So
det(A2 + I) = det(A + iI) det(A − iI) = det(A + iI)det(A + iI) = |det(A + iI)|2 ≥ 0.
Problem 8. Let A, B, C be n × n real matrices that pairwise commute and ABC = O. Show that
det(A3 + B 3 + C 3 ) det(A + B + C) ≥ 0.
Solution. Recall the identity
A3 + B 3 + C 3 − 3ABC = (A + B + C)(A + ωB + ω 2 C)(A + ω 2 B + ωC)
where ω = e2π/3 is a third root of unity. We used the assumption that A, B, C pairwise commute.
Hence,
det(A3 + B 3 + C 3 ) det(A + B + C) = det(A3 + B 3 + C 3 − 3ABC) det(A + B + C)
= (det(A + B + C))2 det(A + ωB + ω 2 C) det(A + ω 2 B + ωC)
= (det(A + B + C))2 det(A + ωB + ω 2 C)det(A + ωB + ω 2 C)
≥ 0.
Problem 9. Let A and B be two n × n real matrices that commute. Suppose that det(A + B) ≥ 0.
Prove that det(Ak + B k ) ≥ 0 for all k ≥ 1
Problem 10. Let A be real skew-symmetric square matrix (i.e., At = −A). Prove that det(I +
tA2 ) ≥ 0 for all real t.
UMA Putnam Talk
LINEAR ALGEBRA TRICKS FOR THE PUTNAM

YUFEI ZHAO

In this talk, I want give some examples to show you some linear algebra tricks for the
Putnam. Many of you probably did math contests in high school, but you might not have
had much experience with linear algebra problems since they do not appear on high school
contests. My goal here today is not to teach linear algebra. Instead, I want to show you how
to use linear algebra. I will remind you some linear algebraic facts along the way as we use
them. If there’s anything that I quote that you have not seen before, then it’s a probably a
good idea for you to look them up when you go home.

1. Determinants
Let us start with the classic example of the Vandermonde determinant.
Problem 1 (Vandermonde determinant). Let
1 x1 x21 · · · x1n−1
 
1 x2 x22 · · · x2n−1 
V = ... ... .. . . ..  .
. . . 
1 xn x2n · · · xnn−1
Show that Y
det V = (xj − xi ).
1≤i<j≤n

The most standard way of evaluating the Vandermonde determinant comes from noting
that the determinant vanishes when two of the variables are set be equal, thereby allowing
us to extract a linear factor.
Solution. Let
p(x1 , x2 , . . . , xn ) = det V,
viewed as a polynomial in n variables. Now, suppose we view p as a single-variable polynomial
in x1 with coefficients in Q(x2 , . . . , xn ). If we set x1 to xi , for any i 6= 1, then two rows of
the matrix are equal and hence the determinant vanishes, and therefore (x1 − xi ) must be a
factor of p.
Similarly, every (xi − xj ) for i 6= j is a factor of p(x1 , . . . , xn ).Q But the degree of p is
1
2
n(n − 1) (from looking at the matrix), and we just showed that 1≤i<j≤n (xj − xi ) (which
has degree 12 n(n − 1)) divides p. Therefore,
Y
p(x1 , x2 , . . . , xn ) = c (xj − xi ),
1≤i<j≤n

for some constant c. Comparing the coefficient of the term x2 x23 x34 · · · xnn−1 shows that k =
1.
Date: November 22, 2011.
1
2 YUFEI ZHAO

The above approach for evaluating the Vandermonde determinant is very slick. However,
for other determinant evaluation problems such tricks might not be available. A more versa-
tile method for evaluating determinants is by elementary row/column operations. Sometimes
you’ll have to get your hands dirty and play with the algebra. Let me illustrate this concretely
with the 4 × 4 case of the Vandermonde determinant.
We know that subtracting a multiple of one column from another column does not change
the determinant of a matrix. So let us do this successively with the goal of getting some
zeros entries in the matrix.
     
1 a a 2 a3 1 a a2 0 1 a 0 0
1 b b2 b3  1 b b2 b2 (b − a)  1 b b(b − a) b2 (b − a) 
 2 3
  2 2
  
1 c c c  1 c c c (c − a)  1 c c(c − a) c2 (c − a) 
1 d d2 d3 1 d d2 d2 (d − a) 1 d d(d − a) d2 (d − a)
   
1 0 0 0 1 0 0 0
1 b − a b(b − a) b2 (b − a)  0 b − a b(b − a) b2 (b − a) 
   
1 c − a c(c − a) c2 (c − a)  0 c − a c(c − a) c2 (c − a) 
1 d − a d(d − a) d2 (d − a) 0 d − a d(d − a) d2 (d − a)
Here we first subtracted a times the 3rd column from the 4th column, thereby eliminating
the (4, 1) entry. Next we subtracted a times the 2nd column from the 3rd column. Next we
subtracted a times the 1st column from the 2nd column. Finally we subtracted the first row
from each of the other rows. When evaluating the determinant, note that we can take out
the common factors from each row. Thus
1 a a2 a3
b − a b(b − a) b2 (b − a) 1 b b2
1 b b2 b3
= c − a c(c − a) c (c − a) = (b − a)(c − a)(d − a) 1 c c2 .
2
1 c c2 c3
d − a d(d − a) d2 (d − a) 1 d d2
1 d d2 d3

Now we get a smaller Vandermonde determinant. We can then repeat the process (or use
induction) to finish the computation.
This method of row elimination comes very handy when we need to evaluate the deter-
minant of other large structured matrices. In particular, this is how I solved the following
problem on the 2008 Putnam.
Problem 2 (Putnam 2008/B6). Let n and k be positive integers. Say that a permutation
σ of {1, 2, . . . , n} is k-limited if |σ(i) − i| ≤ k for all i. Prove that the number of k-limited
permutations of {1, 2, . . . , n} is odd if and only if n ≡ 0 or 1 (mod 2k + 1).
At first the problem might not seem related to determinants. However, recall that deter-
minants in some sense encode permutations, especially if we work mod 2 so that we do not
distinguish between plus and minus signs. Indeed, if the matrix is A = (aij )ni,j=1 , then
X
det A = sgn(σ)a1σ(1) a2σ(2) · · · anσ(n) ,
σ

where the sum is taken over all permutations σ of {1, 2, . . . , n}. To count (mod 2) k-limited
permutations, consider the matrix A where we set aij = 1 if |i − j| ≤ k, and aij = 0
otherwise. Then the number of k-limited permutations has the same parity as det A. For
LINEAR ALGEBRA TRICKS FOR THE PUTNAM 3

instance, when n = 6, k = 2, the matrix is

1 1 1 0 0 0
 
1 1 1 1 0 0
1 1 1 1 1 0
 
0 1 1 1 1 1
 
0 0 1 1 1 1
0 0 0 1 1 1
It remains to compute the determinant of this matrix mod 2. This can be done by successive
row/column manipulation. It gets a little messy, but do-able. I refer to Kiran Kedlaya’s
write-up1 for details.

2. Inverses commute
Now let us look at some matrix algebra problems.
Problem 3. Let A and B be n × n matrices satisfying A + B = AB. Show that AB = BA.
Solution. The given equation can be rewritten as I = AB − A − B + I. We can factor
the right-hand side as (A − I)(B − I) = I. So A − I and B − I are inverses. Hence
(B − I)(A − I) = I. Expanding, we get A + B = BA. Hence AB = BA.
The key fact used in the solution is:
for square matrices X and Y , if XY = I, then Y X = I.
In other words, left inverse and right inverse are the same for matrices. This is a very
useful fact that allows us to get additional relations from matrices “for free.” It is actually
somewhat nontrivial (as in, if you try to prove it by writing down the equations relating the
matrix entries, you might quickly get stuck). Furthermore, the analogous statement is false
for infinite dimensional matrices2.
Problem 4 (Putnam 1986/B6). Suppose A, B, C, D are n × n matrices, satisfying the con-
ditions that AB t and CDt are symmetric and ADt − BC t = I. Prove that At D − C t B = I.
(M t is the transpose of M .)
Solution. The information given imply that
t
A B D −B t ADt − BC t −AB t + BAt I 0
= = .
C D −C t At CDt − DC t −CB t + DAt 0 I
Here the matrices are (2n) × (2n) written in block form. It follows that
t
D −B t A B I 0
= .
−C t At C D 0 I
Equating the lower-right block gives us −C t B + At D = I, as desired.
Here’s a cute problem just for fun.
Problem 5. Do there exist square matrices A and B such that AB − BA = I?
1http://amc.maa.org/a-activities/a7-problems/putnam/-pdf/2008s.pdf
2Take X to be the infinite matrix of with only 1’s in the positions immediate above the main diagonal, and
Y to be the transpose of X.
4 YUFEI ZHAO

Solution. No. The key fact here is that for any matrices A and B, tr(AB) = tr(BA)
(important: it is NOT true that tr(AB) = tr A tr B in general). So we have tr(AB −BA) = 0
while tr(I) 6= 0.

3. Linear independence
So far we’ve been just dealing with matrices. Linear algebra also comes with even when
we are not working with matrices. A very useful concept is that of linear independence.
Recall that vectors v1 , . . . , vk ∈ V said to be linearly independent if c1 v1 + · · · + ck vk = 0
implies that c1 = c2 = · · · = 0.
Problem 6 (Putnam 2003/B1). Do there exist polynomials a(x), b(x), c(y), d(y) such that
1 + xy + x2 y 2 = a(x)c(y) + b(x)d(y)
holds identically?
It is possible to do this problem by writing down some polynomial coefficients and playing
with them. Though it might be get somewhat messy. A better way to view R[x], the set of
polynomials in x, as a vector space, with basis {1, x, x2 , . . .}.
Solution. No they do not exist. If such polynomials existed, then setting y = −1, 0, 1
shows that 1 − x + x2 , 1, 1 + x + x2 are in the linear span of a(x) and b(x). However,
{1 − x + x2 , 1, 1 + x + x2 } is a linearly independent set, so it cannot be contained in a two-
dimensional span, thus a contradiction.
Remark. We only needed a(x) and b(x) to be polynomials, while c(y) and d(y) could be ar-
bitrary functions. Similarly, if c(y) and d(y) are polynomials and a(x) and b(x) are arbitrary
functions, then the same conclusion holds, by switching the roles of x and y in the proof.
Problem 7. Let A = (aij ) be a real n × n matrix satisfying
X
|aii | > |aij |
j6=i

for all 1 ≤ i ≤ n. Prove that A is invertible.

Remark. Such matrices are called (strictly) diagonally dominant.
Solution. We know that A is invertible if and only if its columns are linearly independent.
Suppose not, there would exist real numbers c1 , . . . , cn , not all zero, such that
Xn
cj aij = 0
j=1

for each i = 1, . . . , n. In particular, let us pick i so that ci has the largest absolute value
among c1 , . . . , cn . Then
X cj
aii = − aij .
j6=i
c i

Since |ci | ≥ |cj | for each j, we have

X cj X |cj | X
|aii | = aij ≤ |aij | ≤ |aij | < |aii | ,
j6=i
ci j6=i
|ci | j6=i
LINEAR ALGEBRA TRICKS FOR THE PUTNAM 5

a contradiction. Therefore the columns of A are linearly independent, and hence A is invert-
ible.

Problem 8 (Oddtown). In a town with n people, m clubs have been formed. Every club
have an odd number of members, and every two clubs have an even number of members in
common. Prove that m ≤ n.

Previously we worked with linear independence over the real numbers. It turns out that
we can actually do linear algebra over any field. In particular, in this problem, we want to
work in mod 2, so it is natural to use the field F2 with two elements. We can talk about
vectors and linear independence over F2 just as in the real case.

Solution. For each i = 1, 2, . . . , m, let vi ∈ Fn2 be the vector encoding the membership data
for the i-th club, so that the j-th component of vi is 1 if person j is in club i, and 0 otherwise.
Then vi · vi is the number of members in the i-th club, which we know is odd, so vi · vi = 1
in F2 . Similarly, vi · vj = 0 whenever i 6= j since this count the number of people who are in
both club i and club j.
We claim that the vectors v1 , . . . , vm are linearly independent. Indeed, if c1 , . . . , cm ∈ F2
satisfy

c1 v1 + · · · + cm vm = 0.

For each i = 1, . . . , m, then the dot product with vi shows that ci = 0 for each i.
So v1 , . . . , vm are m linearly independent vectors in a n-dimensional vector space Fn2 , we
have m ≤ n.

Problem 9 (Fisher’s inequality). Let k be a positive integer. In a town with n people, m

clubs have been formed. Every two clubs share exactly k members. Prove that m ≤ n.

The idea is similar to the previous problem. We use incidence vectors of the clubs, and
note that dot products correspond to intersections. However, now we work over the reals.

Solution. Let v1 , . . . , vm ∈ Rn , where the vi is a {0, 1}-vector encoding the i-th club: the
j-th component of vi is 1 if person j is in club i, and 0 otherwise. Then vi · vj = k whenever
i 6= j.
If v1 , . . . , vm are linearly independent, then m ≤ n. Otherwise, then there exists c1 , . . . , cm ∈
R, not all zero, such that

c1 v1 + · · · + cm vm = 0.
6 YUFEI ZHAO

Taking the dot product of the left-hand side sum with itself, and then expanding, we find
that
0 = (c1 v1 + · · · + cm vm ) · (c1 v1 + · · · + cm vm )
Xm X
= c2i vi · vi + 2 ci cj vi · vj
i=1 i<j
Xm X
= c2i |vi |2 + 2k ci cj
i=1 i<j
Xm n X
X n
= c2i (|vi |2 − k) + ci cj
i=1 i=1 j=1
m n
!2
X X
= c2i (|vi |2 − k) + ci .
i=1 i=1
Since every two clubs share exactly k members, every club has at least k members. So
|vi |2 ≥ k for each i. So all the terms in the final sum are nonnegative, and hence must be
zero. Since some ci is nonzero, it follows that |vi |2 − k = 0 for some i, so that some club C
has exactly i members. It follows that all other clubs must all contain members of C but
are otherwise disjoint. It follows immediately that there can be at most n clubs.
4. Some general advice
Putnam is right around the corner. Here’s some general advice that I have for preparing
for the competition:
• Work through some past papers3. Right now the most useful thing to do for prepara-
tion is to think about time management. Three hours can fly by really quickly when
you’re writing the Putnam. It’s a good idea to do some practice tests in a timed
setting, so that you have some sense of what you can accomplish in three hours.
• Every year, you’ll hear some people complain that they did not receive points for
problems they thought they solved. Often this is due to poor write-up. Since you
have limited time during the exam, you might end up rushing your write-ups. Make
sure to always check your write-ups to make sure that they’re understandable, and
that no important points were left out.
• Get some good rest before the exam. Don’t cram the night before the exam, instead,
go to bed early. The Putnam, being six-hours long, can drain a lot of energy. Also,
set your alarm clock; you don’t want to be the guy who shows up after the exam has
started.
• Most importantly, have fun and enjoy the Putnam!

3A good source is Kiran Kedlaya’s Putnam directory http://amc.maa.org/a-activities/a7-problems/

putnamindex.shtml. There are also some books archiving with older exams and solutions
Bijections
Yufei Zhao

In this lecture, we will look at using bijections to solve combinatorics problems.

Given two sets A and B, a bijection (also called bijective correspondence) is a map
f : A → B that is both injective and surjective, meaning that no two elements of A
get mapped onto the same element in B, and every element of B is the image of some
element of A . This gives us a way of pairing up every element from A with some
element of B. In particular, when the sets are finite, the existence of a bijection implies
that |A | = |B|.
This explanation of a bijection may seem a little abstract. Let us take a look at some
examples of how bijections can be used.

A B

Figure 1: Illustrating a bijection from A to B.

Problem 1. Determine the number of walks from (0, 0) to (m, n) allowing only unit
steps up or to the right.

The idea of using bijection is that we want to tranform the problem into something
that we know better how to count. In this case, we can encode the path as a sequence
letters U and R, correponding to whether the step taken was a unit up step (U) or a
unit right step (R). The path displayed above, for instance, is encoded as

RRURRU URRURRU U URRU.

Note that the resulting sequence of letters always has m copies of R and n copies of U,
since it takes in total of m right unit steps and n up unit steps to move from (0, 0) to

1
(m, n). Also, you can check that for any sequence consisting of m copies of R and n
copies of U, we can construct a corresponding path encoded by this sequence. Thus we
have constructed a bijection between the following two sets:

the set of walks from (0, 0) to (m, n) using only unit up or right steps
l
the set of sequences consisting of m copies of R and n copies of U.
m+n
We know how to count the latter set. It has exactly m elements, which can be
thought of as choosing m spots to place the R’s in a sequence of length m+n. Therefore,
m+n
the number of lattice walks from (0, 0) to (m, n) must be m as well.

The above example illustrates the use of the bijection technique. Counting lattice
paths is a fairly important problem by itself that we will come back to it later in the
lecture.
Usually, a proof involving a bijection between two sets A and B should explain the
following:

1. How to obtain an element of B from any element of A .

2. How to recover the element of A from any element of B.

3. Why the above two constructions are inverses of each other.

Can you spot these elements in the solution presented above? Sometimes, like in the so-
lution above, once the map A → B (i.e., the first step) has been constructed, it is easy
to construct the inverse map. However, there are times when the inverse construction
is not at all obvious.
Next, let us look at a problem involving partitions. A partition of a postive integer
n is a way of writing n as a sum of positive integers, where the order of the summands
is irrelevant, though we usually write the summands in non-increasing order. For in-
stance, the number 4 has exactly five partitions:

4=1+1+1+1
4=2+1+1
4=2+2
4=3+1
4=4

We will also write a partition as a non-decreasing tuple of parts, so the partitions of 4

are denoted by (1, 1, 1, 1), (2, 1, 1), (2, 2), (3, 1), (4).

2
Problem 2. Let n and k be positive integers. Show that the number of partitions of n
with exactly k parts equals the number of partitions of n whose largest part is exactly
k.

To solve this problem, we consider a graphical representation of a partition. For

each partition n = a1 + a2 + · · · + a r , where a1 ≥ a2 ≥ · · · ≥ a r > 0, we consider
a diagram with ai dots on the i-th row, all left aligned. This is known as the Ferrar
diagram of the partition. For instance, the partition 15 = 5+3+3+2+1+1 corresponds
to the following diagram:

The key insight here is that there exists a natural bijection between partitions of n and
the set of Ferrar diagrams with n dots. The number of parts of the partition corresponds
to the number of rows in the Ferrar diagram, and the size of the largest part corresponds
to the number of columns of the Ferrar diagram.
For each partition, let us consider its conjugate, whose Ferrar diagram is formed by
reflecting the original diagram across the main diagonal. The following diagram shows
a pair of conjugate partitions.

(5, 3, 3, 2, 1) (6, 4, 3, 1, 1)

Observe that the conjugation operation switches the number of rows and columns
in a Ferrar diagram. In particular, we get a bijection between the number of Ferrar
diagrams with k rows and the number of Ferrar diagrams with k columns, thereby
giving us a bijection between the set of partitions of n with exactly k parts and the set
of partitions of n whose largest part is k.
By the way, the conjugation operation is known as an involution, since applying it
twice gives us back the original element. Every involution is a bijection of a set with
itself (why?).

Here is a very classic result about partitions orginally discovered by Euler.

Problem 3. Prove that the number of partitions of n into distinct parts is equal to the
number of partitions of n into odd parts.

3
We will construct a bijection between the set of partitions of n into distinct parts
with the set of partitions of n into odd parts.
Starting from partition of n into distinct parts, let us write each part of n as a · 2 b ,
where a is odd, and then split the part into 2 b parts all equal to a. This gives a partition
of n into odd parts. For instance, starting from the partition (12, 7, 6, 4, 1) of 30, we
have

30 = 12 + 7 + 6 + 4 + 1
= 3 · 22 + 7 · 20 + 3 · 21 + 1 · 22 + 1 · 20
= (3 + 3 + 3 + 3) + 7 + (3 + 3) + (1 + 1 + 1 + 1) + 1
= 7 + 3 + 3 + 3 + 3 + 3 + 3 + 1 + 1 + 1 + 1 + 1.

So we obtain the partition (7, 3, 3, 3, 3, 3, 3, 1, 1, 1, 1, 1) into distinct parts. For the re-
verse direction, we start with a partition of n into distinct parts. Support that there
are k parts equal to a, where a is odd, let k = 2k1 + 2k2 + · · · + 2kr for distinct positive
integers k1 , k2 , . . . , k r (this is equivalent to the binary representation of k). Then create
the parts a · 2ki for each i. Note that the results parts are all distinct, since every integer
can be uniquely written as a·2 b where a is odd. For instance, starting from the partition
(7, 3, 3, 3, 3, 3, 3, 1, 1, 1, 1, 1) of 30, we have

30 = 7 + 3 + 3 + 3 + 3 + 3 + 3 + 1 + 1 + 1 + 1 + 1
=7·1+3·6+1·5
= 7 · 20 + 3 · (22 + 21 ) + 1 · (22 + 20 )
= 7 · 20 + 3 · 22 + 3 · 21 + 1 · 22 + 1 · 20
= 7 + 12 + 6 + 4 + 1
= 12 + 7 + 6 + 4 + 1.

So we get back the partition (12, 7, 6, 4, 1). You should convince yourself that the two
procedures just described are inverses of each other. Thus we have shown that the
number of partitions of n into distinct parts is equal to the number of partitions of n
into odd parts.

Problem 4. A triangular grid is obtained by tiling an equilateral triangle of side length

n by n2 equilateral triangles of side length 1. Determine the number of parallelograms
bounded by line segments of the grid.

4
One approach is to carefully count all the parallelograms directly. But one has to
be careful to not make mistakes, and we end up with a summation that has to be dealt
with. We present a different, much more elegant solution using bijections.
Observe that the parallelograms come in three different orientations:

By symmetry, there is an equal number of parallograms of each orientation, so let us

just count the paralellograms with the middle type of orientation (i.e., no horizontal
sides).
Extend the triangular grid by one extra row at the bottom. The key (and clever)
observation is that starting from any such parallelogram in the original grid, we can
extend its sides to meet the lines to meet the bottom edge of the new row in the large
triangular grid, and there would be four distinct intersection points, as shown below.

Conversely, starting from any four distinct grid points in new bottom edge, we
can extend 60◦ lines from the first two points and 120◦ lines from last two points
to obtain a parallologram in the original grid. This gives us a bijection between the set
of parallelograms in the original grid with no horizontal sides with set of four distinct
n+2
points in the new bottom edge, and hence there must be 4 of them. Accounting for
all three orientations, we find that the total number of parallolograms in the original
n+2
grid is 3 4 .

5
Next, let us explore the Catalan numbers, which is a sequence is that comes up in
many counting problems, which provide a rich source for bijections.

Problem 5. Let n be a positive integer. Determine the number of lattice paths from
(0, 0) to (n, n) using only unit up and right steps, such that the path stays in the region
x ≥ y.

We saw previously that the total number of lattice paths from (0, 0) to (n, n) without
2n
the x ≥ y restriction is equal is n . Let us count the number of paths that goes into
the x < y region. Call these paths bad paths.
Suppose that P is a bad path. Since P goes into the region x < y, it must hit the
line y = x + 1 at some point. Let X be the first point on the path P that lies on the line
y = x + 1. Now, reflect the portion of path P up to X about the line y = x + 1, keeping
the latter portion of P the same. This gives us a new path P 0 .

P P0

We claim that this gives us a bijection between the set of bad paths to the set of
lattice paths from (−1, 1) to (n, n) using only up and right unit steps.
Here is the inverse construction. For any lattice path Q from (−1, 1) to (n, n), let X
be the first point on the path lying on the line y = x + 1, and let Q0 be constructed from
Q by reflecting the first portion of Q up to X through the line y = x + 1 and keeping
the rest the same. Then the inverse of the bijection given above sends Q to Q0 .
To complete the proof of this claim, we need to check a number of details, which
we outline below. The reader should think about why claim is true.

• The inverse construction is well defined. That is, we can always find such a point
X , and also, the resulting Q0 is a always a bad path.

• The two constructions are inverses of each other.

The number of bad paths is equal to the number of lattice paths from (−1, 1) to
2n
(n, n) using only unit up and right steps, and there are n+1 such paths (using again

6
Problem 1). Therefore, the total number of “good” paths, i.e., those that do not go into
the region x < y, is equals to
2n 2n 2n n 2n 1 2n

− = − = .
n n+1 n n+1 n n+1 n
This is our first example of something that is counted by the Catalan numbers. There
are many more to come.

The n-th Catalan number is

1 2n

Cn = .
n+1 n

Problem 6. Show that the n-th Catalan number counts the number of expressions
containing n pairs of parentheses which are correctly matched. E.g., for n = 3,

((())) (()()) (())() ()(()) ()()()

We could solve this problem by counting it using techniques similar to the one
used to count paths above. A much quicker solution is to find a bijection between
these parentheses expressions and the lattice paths counted in the previous problem.
Indeed, note that by interpretating each ( as a unit right step and each ) as a unit up
step, we obtain the desired bijection. The condition that the parentheses expression is
correctedly matched corresponds exactly to the condition that the lattice path do not go
into the x < y region (why?). This bijection shows that the number of expressions of n
pairs of parentheses which are correctly matched is also equal to n-th Catalan number,
as desired.
The previous bijection was rather simple. Let us look at a more involved Catalan
number bijection.
A plane tree is an object with the following structure. We start with a root vertex
(drawn at the top), and then with each node we attach a number of new vertices
(possibly none), where the order of the attached vertices matters. For instance, there
are exactly 5 plane trees with 4 vertices:

Problem 7. Show that the n-th Catalan number counts the number of plane trees with
n + 1 vertices.

We will produce a bijection between plane trees and the parentheses expressions
considered in the previous problem. We first describe an algorithm to turn a plane tree
into a parentheses expression.

7
Given a plane tree, starting from the top vertex, let us perform a depth-first search
walk, meaning that we go as further down as possible until we hit a dead-end, and
then backtrack to a branch point, where we then explore a new branch. We will always
explore the branches of a vertex in order from left to right. For instance, starting with
the plane

we obtain the following walk, where the steps are labeled in order.
1 20

12 13

2 11 14 19
5 1617
6 15 18

3 4 7 8 9 10

Now we record the sequence of steps we took, writing down a ( each time we stepped
downward along an edge, and a ) every time we stepped upward along an edge. For
example, the above walk corresponds to

((())(()()))(()()()).

A plane tree with n + 1 vertices always produces a correctly matched expression of n

pairs of parentheses correctly matched (why is it correctly matched?). Conversely, given
an expression of n pairs of correctly matched parentheses, it is possible to reverse this
construction to produce a plane tree that corresponds to it. You should first convince
yourself that this is the case why writing down a few parentheses expressions and then
figure out what the corresponding tree is. Then, you should write down a description
of this bijection.

Practice problems:

1. Let n be a positive integer. In how many ways can one write a sum of at least two
positive integers that add up to n? Consider the same set of integers written in a
different order as being different. (For example, there are 3 ways to express 3 as
3 = 1 + 1 + 1 = 2 + 1 = 1 + 2.)

2. Let m, n be positive integers. Determine the number of m-tuples of positive inte-

gers (x 1 , x 2 , . . . , x m ) satisfying x 1 + x 2 + · · · + x m = n.

8
3. Let m, n be positive integers. Determine the number of m-tuples of nonnegative
integers (x 1 , x 2 , . . . , x m ) satisfying x 1 + x 2 + · · · + x m = n. (Hint: can you find a
bijection that reduces this problem to the previous problem?)

4. Let n be a positive integer. Prove that the number of partitions of n equals the
number of partitions of 2n with n parts.

5. Show that the number of ways of stacking coins in the plane so that the bottom
row consists of n consecutive coins is Cn . E.g., for n = 3,

6. Show that the number of triangulations of a convex (n + 2)-gon into n triangles

by n − 1 diagonals that do not intersect their interiors is the n-th Catalan number,
Cn . E.g., for n = 3,

7. Show that the number of complete binary trees with n internal vertices is the n-th
Catalan numberm Cn . E.g., for n = 3,

8. Show that the number of ways to tile a stair-step shape of height n with n rectan-
gles is the n-th Catalan number, Cn . E.g., for n = 3,

9. Show that the number of ways that n + 1 factors can be completely parenthesized
is the n-th Catalan number, Cn . E.g., for n = 3,

(((a b)c)d) ((a(bc))d) ((a b)(cd)) (a((bc)d)) (a(b(cd)))

10. (Putnam 2003) A Dyck n-path is a lattice path of n upsteps (1, 1) and n down-
steps (1, −1) that starts at the origin and never dips below the x-axis. A return is
a maximal sequence of contiguous downsteps that terminates on the x-axis. For
example, the Dyck 5-path illustrated has two returns, of length 3 and 1 respec-
tively.

9
Show that there is a one-to-one correspondence between the Dyck n-paths with
no return of even length and the Dyck (n − 1)-paths.

11. (USAMO 1996) An n-term sequence (x 1 , x 2 , . . . , x n ) 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.

12. Form a 2000×2002 screen with unit screens. Initially, there are more than 1999×
2001 unit screens which are on. In any 2 × 2 screen, as soon as there are 3 unit
screens which are off, the 4th screen turns off automatically. Prove that the whole
screen can never be totally off.

13. (IMO 2002) 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 colored 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.

10
MOP 2007 Black Group Algebraic Techniques in Combinatorics Yufei Zhao

Algebraic Techniques in Combinatorics

June 26, 2007

Yufei Zhao

Linear algebra
Useful facts in linear algebra

• Any set of n + 1 vectors in an n-dimensional vector space is linearly dependent. That is, we can
find scalars a1 , . . . , an+1 , not all zero, such that a1 v1 + · · · + an+1 vn+1 = 0.

• Almost all linear algebra results (especially the ones related to rank) are true over any field. For
instance, the field F2 is often useful for working with parity or incidence. However, note that we
might not be able to use eigenvalues.

• Suppose that A is an m × n matrix, then the subspace spanned by its columns has the same
dimension as the subspace spanned by its rows. This common dimension number is called the
rank of A. In particular, we have rank A ≤ min(m, n).

• Rank-nullity theorem: rank A + nullity A = n.

• rank(AB) ≤ min(rank A, rank B)

• rank(A + B) ≤ rank A + rank B

• A is invertible if and only if m = n = rank A. If A is a square matrix, then it is invertible if and

only if there does not exist a nonzero vector v such that Av = 0. Or equivalently, none of the
eigenvalues of A is zero.

• The determinant of a square matrix A can be evaluated by taking a sum over all permutations
of {1, 2, . . . , n} X
det A = (−1)sgn σ a1σ(1) a2σ(2) · · · anσ(n) .
σ∈Sn

A matrix is invertible if an only if its determinant is nonzero. In F2 , we no longer have to worry

about the sign in front of the product.

• If W is a subspace of V , denote W ⊥ = {v ∈ V | vt w = 0 ∀w ∈ W }, then

dim W + dim W ⊥ = dim V.

However, it’s not necessarily true that W ⊕ W ⊥ = V ! E.g., take W = {(0, 0), (1, 1)} in F22 , then
W⊥ = W.

• Suppose that we have a set of maps fi : V → R, and points vi ∈ V , 1 ≤ i ≤ n. If fi (vi ) =

6 0 and
6 j, then f1 , f2 , . . . , fn are linearly independent.
fi (vj ) = 0 whenever i =

1
MOP 2007 Black Group Algebraic Techniques in Combinatorics Yufei Zhao

Problem: (St. Petersburg) Students in a school go for ice cream in groups of at least two. After
k > 1 groups have gone, every two students have gone together exactly once. Prove that the number
of students in the school is at most k.

Solution. There is a combinatorial solution which is somewhat long and non-intuitive. However, a
much quicker (and more intuitive) solution is available using the tools of linear algebra.
Let there be n students. Note if some student went for ice cream only once, then everyone else has
to have gone with that student, due to the constraint that every pair of student have gone together
exactly once. Furthermore, since each group consists of two students, no other groups can be formed.
However, k > 1, so this situation cannot occur. Therefore, every student went for ice cream at least
twice.
Let us construct incidence vectors vi ∈ Rk , 1 ≤ 1 ≤ n, representing the students. That is, the
j-th component of vi is 1 if student i went with the j-th group, and 0 otherwise.
The condition that every two students have gone together exactly once translates into the dot
product vi · vj = 1 for i 6= j. The condition that every student went for ice cream at least twice
translates into |vi |2 ≥ 2.
We want to prove that n ≤ k. Suppose otherwise. Then {v1 , . . . , vn } consists of at least k + 1
vectors in Rk , so they must be linearly dependent. It follows that there are real numbers α1 , . . . , αn ,
not all zero, such that
α1 v1 + · · · + αn vn = 0.
Let us square the above expression (i.e., taking the dot product with itself. We get

n n n n
!2
X X X X X X
αi2 |vi |2 + 2 αi2 |vi |2 + 2 αi2 |vi |2 − 1 +

0= α i α j vi · vj = αi αj = αi .
i=1 i<j i=1 i<j i=1 i=1

However, the RHS expression is positive, since |vi |2 ≥ 2 and some αi is nonzero. Contradiction.
Alternatively, we can use an incidence matrix to obtain a somewhat faster (albeit more tech-
nically involved) solution. The incidence matrix M is a k × n matrix, where Mij is 1 if the i-th group
includes student j. Using the notation in the previous solution, we can write this as
 
| | ··· |
M =  v1 v 2 · · · v n 
| | ··· |

Consider the product M t M . Since vi ·vj = 1 for i 6= j and |vi |2 ≥ 2 for all i, we see that M t M = J +A,
where J is the n×n matrix with all entries 1, and A is a diagonal matrix with positive diagonal entries.
Note that J is positive semidefinite (meaning that xt Jx ≥ 0 for all vectors x) and M is positive definite
(meaning that xt Jx > 0 for all x 6= 0). Therefore, their sum J + A = M t M is positive definite. In
particular, this means that M t M is invertible, so its rank is n. Since M is a k × n matrix, we conclude
that k ≥ n.

2
MOP 2007 Black Group Algebraic Techniques in Combinatorics Yufei Zhao

Partially ordered sets (posets)

Definition. A partially ordered set (or poset for short) P is a set, also denoted P , together with a
binary relation denoted ≤ satisfying the following axioms:

• (reflexivity) x ≤ x for all x ∈ P

• (antisymmetry) If x ≤ y and y ≤ x, then x = y.

• (transitivity) If x ≤ y and y ≤ z, then x ≤ z.

An example of a poset is the set of all subsets of {1, 2, . . . , n} under the relation ⊂. This poset is
sometimes called the Boolean algebra of rank n, and denoted Bn .
The Hasse diagram is a simple way of representing (small) posets. We say that x covers y if
x > y (i.e., x ≥ y and x 6= y) and there is no z ∈ P such that x > z > y. If x covers y, then we
draw x above y and connect them using a line segment. Note that in general, x > y if and only if x
is above y and we trace a downward path from x to y in the Hasse diagram. The Hasse diagram for
B3 is depicted below.
{1, 2, 3}

{1, 2} {1, 3} {2, 3}

{1} {2} {3}

There is one result about posets that has proven useful for olympiad problems. Before we state
this result, let us go over some more terminology.
Two elements x, y of a poset are called comparable if x ≥ y or x ≤ y, otherwise they are called
incomparable. A chain is a sequence of elements a1 < a2 < · · · < ak , and an antichain is a set
of pairwise incomparable elements. The length or a chain or antichain is the number of elements
contained in it.
Now we are ready to state Dilworth’s Theorem.
Theorem 1. (Dilworth) Let P be a finite poset. Then the smallest set of chains whose union is P
has the same cardinality as the longest antichain.

There is also a dual version of this theorem that’s much easier to prove.
Theorem 2. Let P be a finite poset. Then the smallest set of antichains whose union is P has the
same cardinality as the longest chain.

The length of the longest chain and the length of the longest antichain are often referred to as the
height and width of a poset, respectively. Another way of stating the two theorems is that if a poset
P has height h and width w, then P can be covered with h antichains and w chains. One simple
consequence is that |P | ≤ h · w.

3
MOP 2007 Black Group Algebraic Techniques in Combinatorics Yufei Zhao

Instead of proving the the above two theorems, which you can do yourself1 , let’s see how we can
apply them.

Problem: (Romania TST 2005) Let n be a positive integer and S a set of n2 + 1 positive integers
with the property that every (n+1)-element subset of S contains two numbers one of which is divisible
by the other. Show that S contains n + 1 different numbers a1 , a2 , . . . , an+1 such that ai | ai+1 for
each i = 1, 2, . . . , n.

Solution. Use the divisibility relation to obtain a poset on S (that is, x“ ≤ ”y iff x | y. Check that
this makes a poset). The condition that there does not exist an n + 1 element subset of S that no
element divides another translates into the condition that there does not exist an antichain of length
n + 1 in S. So the longest antichain in S at length at most n, and thus by Dilworth’s theorem, S can
be written as the union of at most n chains. Since S has n2 + 1 elements, this implies that one of
these chains has a length of at least n + 1. This implies the result.

A very similar result is the Erdös–Szekeres Theorem, which states that within any sequence
of ab + 1 real numbers, there is either a nondecreasing subsequence of a + 1 terms, or a nonincreasing
subsequence of b + 1 terms. This result is also a simple consequence of Dilworth’s Theorem.
More examples are given in the problems section.

Problems related to linear algebra

1. (Nonuniform Fisher inequality) Let A1 , . . . , Am be distinct subsets of {1, 2, . . . , n}. Suppose that
there is an integer 1 ≤ λ < n such that |Ai ∩ Aj | = λ for all i 6= j. Prove that m ≤ n.

2. (a) (China West 2002) Let A1 , A2 , . . . , An+1 be non-empty subsets of {1, 2, . . . , n}. Prove that
there exists nonempty disjoint subsets I, J ⊂ {1, 2, . . . , n + 1} such that
[ [
Ak = Ak .
k∈I k∈J

(b) (Lindstrom) Let A1 , A2 , . . . , An+2 be non-empty subsets of {1, 2, . . . , n}. Prove that there
exists nonempty disjoint subsets I, J ⊂ {1, 2, . . . , n + 2} such that
[ [ \ \
Ak = Ak , and Ak = Ak .
k∈I k∈J k∈I k∈J

3. (Russia 2001) A contest with n question was taken by m contestants. Each question was worth a
certain (positive) number of points, and no partial credits were given. After all the papers have
been graded, it was noticed that by reassigning the scores of the questions, any desired ranking
of the contestants could be achieved. What is the largest possible value of m?

4. Oddtown and Eventown. In a certain town with n citizens, a number of clubs are set up. No
two clubs have exactly the set of members. Determine the maximum number of clubs that can
be formed under each of the following constraints:

(a) The size of every club is odd, and every pair of clubs share an even number of members.
1
if you are really stuck, then you may consult the proof at, for example, http://ocw.mit.edu/NR/rdonlyres/
Mathematics/18-997Spring2004/FC143A49-2C1F-4653-85C4-C08A74990438/0/co_lec6.pdf

4
MOP 2007 Black Group Algebraic Techniques in Combinatorics Yufei Zhao

(b) The size of every club is even, and every pair of clubs share an odd number of members.
(c) The size of every club is even, and every pair of clubs share an even number of members.
(d) The size of every club is odd and every pair of clubs share an odd number of members.

5. (Moldova TST 2005) Does there exist a configuration of 22 distinct circles and 22 distinct points
on the plane, such that every circle contains at least 7 points and every point belongs at least
to 7 circles?

6. (Iran TST 1996, Germany TST 2004) Let G be a finite simple graph, and there is a light bulb
at each vertex of G. Initially, all the lights are off. Each step we are allowed to chose a vertex
and toggle the light at that vertex as well as all its neighbors’. Show that we can get all the
lights to be on at the same time.

7. Let a1 , a2 , . . . , an be integers. Show that

Y ai − aj
i−j
1≤i<j≤n

is an integer. (Hint: use the Vandermonde determinant.)

8. (a) Let G be a graph with v vertices. Let f (n) denote the number of closed walks in G of
length n. Show that there exists complex numbers λ1 , . . . , λv such that

f (n) = λn1 + λn2 + · · · + λnv

for all positive integers n.

(b) Let g(n, m) denote the number of sequences (x1 , x2 , · · · , xn ), with terms from {1, 2, · · · , m},
such that x1 = 1, xn 6= 1, and xi 6= xi+1 for any i. Show that
1
g(n, m) = ((m − 1)n + (m − 1)(−1)n ) .
m

9. (Crux 3037) There are 2007 senators in a senate. Each senator has enemies within the senate.
Prove that there is a non-empty subset K of senators such that for every senator in the senate,
the number of enemies of that senator in the set K is an even number.

10. (Classical) Let a1 , a2 , . . . , a2n+1 be real numbers, such that for any 1 ≤ i ≤ 2n+1, we can remove
ai and separate the remaining 2n numbers into two groups of n numbers with equal sums. Show
that a1 = a2 = · · · = a2n+1 .

11. (Russia 1998) Each square of a (2n − 1) × (2n − 1) board contains either +1 or −1. Such an
arrangement is called successful if each number is the product of its neighbors (squares sharing
a common side with the given square). Find the number of successful arrangements.

12. (Graham–Pollak) Show that the complete graph with n vertices, Kn , cannot be covered by fewer
than n − 1 complete bipartite graphs so that each edge of Kn is covered exactly once.

13. (Iran 2006) Let B be a set of n-tuples of integers such that for every two distinct members
(a1 , . . . , an ) and (b1 , . . . , bn ) of B, there exist 1 ≤ i ≤ n such that ai ≡ bi + 1 (mod 3). Prove
that |B| ≤ 2n .

5
MOP 2007 Black Group Algebraic Techniques in Combinatorics Yufei Zhao

14. (a) (Frankl–Wilson) Let A1 , A2 , . . . , Am be distinct subset of {1, 2, . . . , n}. Let L be the set of
numbers that occur as |Ai ∩ Aj | for some i 6= j, and suppose that |L| = s. Show that

n n n
m≤ + + ··· +
s s−1 0

(b) (Ray-Chaudhuri–Wilson) Let 0 < k ≤ n be positive integers, and let A1 , A2 , . . . , Am be

distinct k-element subsets of {1, 2, . . . , m}. Let L be the set of numbers that occur as
|Ai ∩ Aj | for some i 6= j, and suppose that |L| = s. Show that m ≤ ns .

Problems related to posets

1. (Sperner) Let A1 , . . . , Ak be subsets of {1, 2, . . . , n} so that no Ai contains another Aj . Show
that
k
X 1
n
≤ 1.
i=1 |Ai |
n

Conclude that k ≤ bn/2c .

2. (Romanian TST 2006) Let m and n be positive integers and S be a subset of {1, 2, 3, . . . , 2m n}
with (2m − 1)n + 1 elements. Prove that S contains m + 1 distinct numbers a0 , a1 , . . . , am such
that ak−1 | ak for all k = 1, 2, . . . , m.

3. (Iran 2006) Let k be a positive integer, and let S be a finite collection of intervals on the real line.
Suppose that among any k + 1 of these intervals, there are two with a non-empty intersection.
Prove that there exists a set of k points on the real line that intersects with every interval in S.

4. (Slovak competition 2004) Given 1001 rectangles with lengths and widths chosen from the set
{1, 2, . . . , 1000}, prove that we can chose three of these rectangles, A, B, C, such that A fits into
B and B fits into C.

5. Let G be a simple graph, and let χ(G) be its chromatic number, i.e., the smallest number of
colors needed to color its vertices so that no edge connects two vertices of the same color, and
suppose that G is colored using χ(G) colors as such. Show that there is a path in G of length
χ(G) such that all χ(G) vertices are of different colors.

6. Suppose that A and B are two distinct lattice points in Rn with non-negative integer coordinates.
We say that A dominates B (denote by A > B) if all the components of A − B are nonnegative,
and A 6= B.

(a) Suppose that S is an infinite sequence of lattice points in Rn with non-negative coordinates.
Show that there exists an infinite subsequence satisfying Si1 < Si2 < Si3 < · · · .
(b) Suppose that T is a set of lattice points in the box [0, t1 ] × [0, t2 ] × · · · [0, tn ] (where
t1 , t2 , . . . , tn are fixed nonnegative integers). It is known that no element in T dominates
another element. What is the maximum value of |T |?

6
MOP 2007 Blue Group Tiling Yufei Zhao

Tiling: Coloring and Weights

June 15, 2007

Yufei Zhao

[email protected]

Main discussion: Packing boxes with bricks

Probably the most basic problem about tiling is to show that an 8 × 8 chessboard with two opposite
corners removed cannot be tiled with dominoes. Many olympiad problems play on variations of this
idea. (A related question worth considering: if one black square and one white square are removed
from the chessboard, then can we always tile the rest with dominoes?) For example, consider the
following generalization:

Problem 1: Let k be an integer. Which m × n boards can be tiled with 1 × k tiles (rotations
allowed)?

You could assign colors (labelled 1, 2, . . . ), or you could assign roots of unity. For k = 3 the
coloring schemes are shown below (ω is a third root of unity).

1 2 3 1 2 3 ··· 1 ω ω2 1 ω ω2 ···
2 3 1 2 3 1 ··· ω ω2 1 ω ω2 1 ···
3 1 2 3 1 2 ··· ω2 1 ω ω2 1 ω ···
1 2 3 1 2 3 ··· 1 ω ω2 1 ω ω2 ···
2 3 1 2 3 1 ··· ω ω2 1 ω ω2 1 ···
3 1 2 3 1 2 ··· ω2 1 ω ω2 1 ω ···
.. .. .. .. .. .. .. .. .. .. .. .. .. ..
. . . . . . . . . . . . . .

The idea is that each tile covers all three colors exactly once, so all the colors must appear the same
number of times. In the roots of unity method, each tile covers a sum of zero since 1+ω+· · ·+ω k−1 =
0, and the sum of the whole board must be zero. It’s easy to check that this happens if and only
if k divides m or n. (check this!)
In this problem, these two approaches amount to the same thing. Each method has its own
advantages. In this discussion, we show how the roots of unity approach can be extended to other
problems. Many problems given for practice at the end use the coloring approach extensively.
Consider the following continuous analogue of Problem 1. It is also a generalization, as it implies
the Problem 1 (why?).

Problem 2: Show that if a rectangle can be tiled by smaller rectangles each of which has at least
one integer side, then the tiled rectangle has at least one integer side.

There was a paper that contains 14 proofs of this result!1 How many can you find?
1
Stan Wagon, Fourteen Proofs of a Result About Tiling a Rectangle, Amer. Math. Monthly, 94 (1987) 601–617

1
MOP 2007 Blue Group Tiling Yufei Zhao

Let’s see how we can use the idea in Problem 1. In Problem 1, we are essentially assigning
2πi
the square with coordinates (i, j) with the value ω i+j = e(i+j) k . How can we extend this to the
continuous case? One way to do this is to construct the function f : R2 → C.

f (x, y) = e2πi(x+y) .

This function assigns “weights” to the points of the plane, just as we assigned weights to the squares
of the board in Problem 1. In Problem 1, the sum of the roots of unity covered by a single tile is
0. Here, note that the integral of f over any horizontal or vertical line segment with integer length
is zero. This is the key insight.
So, let us place the large rectangle on the 2-D Cartesian coordinate, with its bottom left corner
at the origin (assume that all rectangles are placed with its sides parallel to the axes). We can
check that the integral of f over any rectangle with one integer dimension is zero. Thus, if a tiling
is possible, then the integral of f over the entire large rectangle must be zero as well. You can
check that this is possible only when one of the sides of the large rectangle has integer length as
well. The technical details of this argument can be summarized in the following integration:
Z bZ d Z b Z d 2πib
− e2πia
2πid
− e2πic
ZZ
2πi(x+y) 2πix 2πiy e e
f dA = e dydx = e dx e dy = .
a c a c 2πi 2πi
[a,b]×[c,d]

This is a perfectly valid solution. However, it uses calculus (gasp!). Moreover, it uses calculus
with complex numbers! Can we get an elementary solution out of this? Well, let’s see if we can at
least reduce the solution to just calculus over the real numbers. Consider the following function:

f1 (x, y) = sin (2πx) sin (2πy)

The two weight functions f1 and f share many common properties, and it turns that the solution
still works if we had used f1 instead of f . (However, the weight function cos(2πx) cos(2πy) does
not work. Why?)
We can simplify a bit more. Notice that we never really used anything about the exact curvature
of sin, as we more or less only need the property that the integral of the function over [x, x + 1] is
0. So, why don’t we “straighten” out our weight function, and use the following:
(
1 {x} ≤ 21 ,
f2 (x, y) = g(x)g(y), where g(x) =
−1 {x} > 12 .

You can check that f2 also does the trick! Moreover, f3 allows us to come up with a combinatorial
formulation of the solution—just consider the weight function as a black and white coloring of the
board. You can work out the details yourself.
Excellent! We just found three of the fourteen solutions.
Now, let’s go back to the discrete case, but let’s move up a dimension. Can you tile a 6 × 6 × 6
with 1 × 2 × 4 boxes? What can we say in general?

Problem 3: (de Bruijn) If the box A1 × · · · × An can be tiled with bricks a1 × · · · × an , then show
that for each i, ai divides some Aj . (Note that this does not necessarily mean that the box is a
multiple of the brick, e.g. the box 1 × 5 × 6 and brick 1 × 2 × 3.)

We say that a d-dimensional box A is a multiple of a d-dimensional box B if box A can be tiled
by B in such a way that all the copies of B’s are placed in the same orientation. Equivalently, A

2
MOP 2007 Blue Group Tiling Yufei Zhao

is a multiple of B if there is some permutation σ on {1, 2, . . . , n} such that the i-th component of
A is an integer multiple of the σ(i)-th component of B for each i.
Problem 3 isn’t too difficult. In fact, it’s pretty much the same as Problem 1. Let’s focus on one
particular ai , and use the same idea as in problem 1 with k = ai to show that one of the dimensions
of the box is divisible by ai .
Now, let’s restrict ourselves to a certain special case:

Problem 4: (de Bruijn) Suppose that the brick a1 × · · · × an satisfies the divisibility relations
a1 | a2 | · · · | an . Then the box A1 × · · · × An can be tiled with the brick if and only if it’s a multiple
of the brick.

Let’s work our way down through the ai ’s one by one, starting from the largest. By Problem 3,
there is some Ai that is divisible by an , say an | An . Now, dropping the last coordinate and take
a cross-section of dimension A1 × · · · × An−1 . If any tiles of this box has a side of length an , then
the divisibility condition allows us to cut up the tiles into a1 × · · · × an−1 tiles. So now we end up
with a tiling of the box A1 × · · · × An−1 with the bricks a1 × · · · × an−1 . Repeat the argument.
However, if the brick does not satisfy the chain of divisibilities, then there is always some box
that can be tiled by the brick without being a multiple of the brick (exercise: prove this!).
Now, what if we are allowed to use two bricks, but with restricted orientations? Then, there is
a definitive criterion for when we can tile a box with the bricks. Note that we don’t even require
the dimensions of the bricks to the be integers!

Problem 5: (Bower and Michael) Prove that the d-dimensional box R can be tiled by translates
of two given d-dimensional bricks B1 and B2 if and only if R can be partitioned by a hyperplane
into two sub-boxes R1 and R2 such that Ri is a multiple of Bi for i = 1, 2.

Note that the not every tiling has to be bipartite, but the existence of a non-bipartite tiling
implies the existence of a bipartite tiling.
The most difficult part of the problem is the 2-D case. However, we have done most of the work
for that already! (Where?) The rest is left as exercise.

3
MOP 2007 Blue Group Tiling Yufei Zhao

Problems
Key ideas: color the board in some way that gives some constraints or invariants. Often, this
involves marking a certain subset of the board. Looking at parities and other modulos is a good
idea. Sometimes you may have to use more than one coloring schemes simultaneously2 to solve a
problem. Finally, don’t forget that some problem have two (related) parts—proving a constraint
and constructing an example.

1. Is there a closed knight’s tour on a 5 × 5 chessboard?

2. For which n is there a closed knight’s tour on a 4 × n chessboard?

3. (Tournament of Towns 2004) Given two rectangles A and B, such that one can tile a rectangle
similar to B using copies of A, show that one can tile a rectangle similar to A using copies of
B.

4. (Canada 2007) What is the maximum number of dominoes which can be placed on an 8 × 9
board if six are already placed as shown below?

• •
• •
• •
• •
• •
• •

5. What is the smallest number of squares on an 8×8 chessboard which would have to be painted
so that no 3 × 1 rectangle could be placed on the board without covering a painted square.

6. Which single square can be removed from a 7 × 7 board so that the rest can be tiled with
1 × 3 trominos.

7. Prove that a 4 × 11 rectangle cannot be tiled by L-shaped tetrominoes.

8. (Russia 1996) Can a 5 × 7 board be covered by L-trominos, not crossing its boundary, in
several layers, so that each square of the board is covered by the same number of trominos?

2
You may have seen certain “cool” coloring proofs that use the Klein four-group. Those might not actually be
as fancy as you thought. The Klein four-group is the group with four elements {e, a, b, c} satisfying the relations
a2 = b2 = c2 = abc = e, but in fact, it is isomorphic to the group of two-dimensional coordinates in mod 2 under
addition, i.e., {(0, 0), (1, 0), (0, 1), (1, 1)}. Thus, any coloring proof that uses the Klein four-group can be replicated
by multiple applications of a black-white coloring scheme.

4
MOP 2007 Blue Group Tiling Yufei Zhao

9. (Russia 2002) A rectangle is partitioned into 100 L-trominos and some 1 × 3 tiles. Someone
chosen chosen 96 of these L-trominoes and merge them in pairs into 48 2×3 rectangles. Prove
that one can translate the remaining 4 L-trominoes and merge them into two 2 × 3 rectangles.

10. A 6 × 6 board is tiled by dominoes. Show that there is a line that cuts the board into two
parts without cutting any domino.

11. (Iurie Boreico) Let n be a positive integer such that gcd(n, 6) = 1, and let k and l be positive
integers. The entries of a k × l table are all +1 or −1. One can simultaneously change the
signs of any n consecutive entries horizontally, vertically, or diagonally. Prove that one can
eventually make all the entires negative numbers if and only if n divides k or n divides l.

12. (Tournament of Towns 1993) On a 10×10 square board we are trying we place ten rectangles:
one 1 × 4, two 1 ×3, three 1× 2 and four 1×1. Prove that if we arbitrarily place the rectangles
on the board but in the aforementioned order, then at each step, it is always possible to fit
the rectangle into the board so that it does not share a point (even on the boundaries) with
an existing rectangle.

13. There is a 5 × 5 array of lights, such that at each step, we may toggle all the lights in any
2 × 2, 3 × 3, 4 × 4 or 5 × 5 sub-square. Initially all the lights are switched off. After a certain
number of toggles, exactly one light is switched on. Find all the possible positions of the
light.

14. (APMO 2007) A regular (5 × 5)-array of lights is defective, so that toggling the switch for one
light causes each adjacent light in the same row and in the same column as well as the light
itself to change state, from on to off, or from off to on. Initially all the lights are switched
off. After a certain number of toggles, exactly one light is switched on. Find all the possible
positions of the light.

15. (USAMO 1998) A computer screen shows a 98 × 98 chessboard, colored in the usual way. One
can select with a mouse any rectangle with sides on the lines of the chessboard and click the
mouse button: as a result, the colors in the selected rectangle switch (black becomes white,
white becomes black). Determine the minimum number of mouse clicks needed to make the
chessboard all one color.
√
16. (Balkan 2000) Determine the maximum number of 1 × 10 2 rectangles that can be placed
on a 50 × 90 rectangle without overlap and so that the small rectangles have its sides parallel
to the large rectangle.

17. (IMO Shortlist 2002) For n an odd positive integer, the unit squares of an n × n chessboard
are colored alternately black and white, with the four corners colored black. For which values
of is it possible to cover all the black squares with non-overlapping L-trominos? When it is
possible, what is the minimum number of L-trominos needed?

18. (IMO Shortlist 2000) A staircase-brick with 3 steps of width 2 is made of 12 unit cubes.
Determine all integers for which it is possible to build a cube of side n using such bricks.

(The staircase-brick is made up of two such layers.)

5
MOP 2007 Blue Group Tiling Yufei Zhao

19. (IMO 2004) Define a “hook” to be a figure made up of six unit squares as shown below in the
picture, or any of the figures obtained by applying rotations and reflections to this figure.

Determine all m × n rectangles that can be tiled with hooks? (i.e. no gaps, no overlaps, and
no part of a hook lies outside the rectangle).3

Fun facts about tiling

• (Fish and Temperly 1961; Kasterleyn 1961) The number of tilings of a 2m × 2n rectangle
with 2mn dominoes is equal to
m Y
n
mn
Y jπ kπ
4 cos2 + cos2
2m + 1 2n + 1
j=1 k=1

This is a remarkable formula. It’s not even clear that the product is an integer!

• (Laczkovich and Szekeres 1995) For which x can a square be tiled with finitely many rectangles
similar to a 1 × x rectangle (in any orientation)?
The answer is that this is possible if and only if x is the root of a polynomial with integer
coefficients, and all the roots of the minimal polynomial of x has positive real part.
√
For example, we cannot tile a square √ with rectangles similar to 1 × 2, but we can tile a
2 + 17

square with rectangles similar to 1 × 12 .

3
For an analysis of a large number of polyomino tiles in terms of which rectangles each can tile, see http:
//www.math.ucf.edu/~reid/Polyomino/rectifiable_data.html. This database contains the analysis of the “hook”
polyomino way before the 2004 IMO. Consequently, the problem received some complaints on MathLinks after the
contest because it was “well-known.” Nevertheless, only 11 students solved the problem at the IMO, and no one got
6 points (only one contestant got a 5 . . . and there’s an interesting story behind that . . . ).

6
MOP 2007 Black Group Counting in Two Ways Yufei Zhao

Counting in Two Ways

Incidence Matrices
June 26, 2007

Yufei Zhao

[email protected]

Problems to be discussed in lecture

Problem 1. In a certain committee, each member belongs to exactly three subcommittees, and each sub-
committee has exactly three members. Prove that the number of members equals to the number of subcom-
mittees.

Problem 2. (IMC 2002) Two hundred students participated in a mathematical contest. They had six
problems to solve. It is known that each problem was correctly solved by at least 120 participants. Prove
that there must be two participants such that every problem was solved by at least one of these two students.

Problem 3. (IMO 1998) In a competition, there are a contestants and b judges, where b ≥ 3 is an odd
integer. Each judge rates each contestant as either “pass” or “fail”. Suppose k is a number such that, for
any two judges, their ratings coincide for at most k contestants. Prove that k/a ≥ (b − 1)/2b.

Problem 4. (Iran 1999) Suppose that C1 , . . . , Cn (n ≥ 2) are circles of radius one in the plane such that
no two of them are tangent, and the subset of the plane formed by the union of these circles is connected.
Let S be the set of points that belong to at least two circles. Show that |S| ≥ n.

Problem 5. Let S1 , S2 , . . . , Sm be distinct subsets of {1, 2, . . . , n} such that |Si ∩ Sj | = 1 for all i 6= j. Prove
that m ≤ n.

1
MOP 2007 Black Group Counting in Two Ways Yufei Zhao

Counting in Two Ways

Incidence Matrices
June 26, 2007

Yufei Zhao

[email protected]

A large number of combinatorics problems involve looking at a quantity in at least two different ways.
This technique is often called “double counting.” In this note, we will mainly focus on how to use incidence
matrices to help us set up the counting. While the incidence matrix is a powerful tool, there are many
problems which cannot be tackled using incidence matrices—you will encounter some of this in the problems
section.

Incidence Matrices
Let’s start with a really simple (and silly) example.
Problem 1. In a certain committee, each member belongs to exactly three subcommittees, and each sub-
committee has exactly three members. Prove that the number of members equals to the number of subcom-
mittees.

Here’s how we usually set up the incidence matrix. In our incidence matrix, each row represents an
individual, and each column represents an organization. An entry is 1 if the individual corresponding to its
row belongs to the organization corresponding to its column; otherwise, the entry is 0. Of course, the roles
of rows and columns may be interchanged. Two examples of the configuration described in Problem 1 are
shown below.
 
  1 1 1 0 0 0 0 0 0
1 1 1 0 0 0 0 
 0 0 0 1 1 1 0 0 0 


 1 0 0 1 1 0 0 


 0 0 0 0 0 0 1 1 1 


 1 0 0 0 0 1 1 


 1 0 0 1 0 0 1 0 0 


 0 1 0 1 0 1 0 


 0 1 0 0 1 0 0 1 0 


 0 1 0 0 1 0 1 


 0 0 1 0 0 1 0 0 1 

 0 0 1 1 0 0 1  
 1 0 0 0 1 0 0 0 1 

0 0 1 0 1 1 0  0 1 0 0 0 1 1 0 0 
0 0 1 1 0 0 0 1 0

Counting the number of 1’s. To solve Problem 1, we could ask ourselves, how many 1’s are there in the
incidence matrix? Suppose that there are n subcommittees and m members. Then the incidence matrix is
a m × n matrix. The given conditions tell us that each row contains 3 ones, so there are 3m ones in total.
On the other hand, each column contains 3 ones, so there are 3m ones in total. Equating the two counts,
we see that 3m = 3n, so m = n, which is what we wanted to prove.
The following result follows from the same idea.

Proposition 1. If A = (ai,j ) is a r × c matrix with row sums Ri , i = 1, 2, . . . , r, and column sums Cj ,

j = 1, 2, . . . , c, then
X r Xc
Ri = Cj .
i=1 j=1

2
MOP 2007 Black Group Counting in Two Ways Yufei Zhao

Counting pairs of 1’s.

Here’s an approach that appears frequently in combinatorial problems. Oftentimes, we are given some
restriction that applies to every pair of organizations (or individuals). For example, it may be that every two
organizations share exactly one common member. In this case, counting the number of 1’s as we did above
does not incorporate all the given information, and thus would likely be unsuccessful. Fortunately, such
problems can usually be approached by counting pairs of 1’s. Specifically, we are interested in the number
of pairs of 1’s that lie on the same column (or row).
Proposition 2. Let A = (ai,j ) be a r × c (0, 1)-matrix with column sums Cj . Suppose that for every two
rows, there exist exactly t columns that contain 1’s from both rows, then
X c
r Cj
t = .
2 j=1
2

Proof. Let T denote the set of all unordered pairs of 1’s that lie in the same column. Let us count the
elements of T in two different ways.
Counting by rows: For any two rows, there are t pairs of 1’s among these rows that belong to T , so
|T | = t 2r .

Counting by columns: In the j th column, there are Cj 1’s, and thus C2j pairs. Counting over all the

Pc
columns gives |T | = i=1 C2j .

The result follows by equating the above two expressions.

Inequalities. In most problems of the type, we are not given enough information to directly produce a
combinatorial identity. Instead, we have to work with inequalities and bounds.
Many incidence matrix problems strive for the existence of a certain subconfiguration. These problems
can usually be approached through contradiction. Under the assumption that the opposite result holds, we
can count a particular set (e.g., the set of all pairs of 1’s that belong in the same column) in two different
ways, once by rows and once by columns. Furthermore, we would want to establish an upper bound in one
of our counts, and a lower bound in the other count. If the upper bound turns out to be less than the lower
bound, then a contradiction is reached.
Problem 2. (IMC 2002) Two hundred students participated in a mathematical contest. They had six
problems to solve. It is known that each problem was correctly solved by at least 120 participants. Prove
that there must be two participants such that every problem was solved by at least one of these two students.

Solution. Let us assume that the contrary is true. That is, for every two students, there is some problem
that neither of them solved. This prompts us to count the pairs of students with their unsolved problem.
Let us consider the incidence matrix of this configuration. We have six rows, each representing a problem,
and 200 columns, each representing a student. In light of the above remark, we make an entry of the matrix
1 if the student corresponding to the column did not solve the problem corresponding to the row, and make
the entry 0 otherwise. The setup is illustrated below.
 
Problem 1 0 1 0 ··· 0
Problem 2 
 1 1 0 ··· 0 

Problem 3 
 0 0 0 ··· 1 

Problem 4 
 0 1 1 ··· 1 

Problem 5  1 0 1 ··· 1 
Problem 6 0 1 0 ··· 0

Let T denote the set of pairs of 1’s that belong in the same row. Let us consider the cardinality of T
from two different perspectives.

3
MOP 2007 Black Group Counting in Two Ways Yufei Zhao

Counting by columns: We assumed that for every two students, there was a problem that neither of them
solved. Thus, for every two columns, there is at least one pair of 1’s among these two columns that
belong
in the same row. So we can find an element of T in every pair of columns. Since there are 200 2 pairs of
columns, we have |T | ≥ 200

2 = 19, 900.
Counting by rows: We are told that each problem was solved by at least
120 students. This means that
there are at most 80 ones in each row. So each row contains at most 80
2 pairs of 1’s. Since there are six
80

rows, we have |T | ≤ 6 2 = 18, 960.
Combining the above two inequalities, we get 19, 900 ≤ |T | ≤ 18, 960, which is clearly absurd. Therefore,
our initial assumption must be false. So there must be two students such that every problem was solved by
at least one of these two students.

Convexity. As we are often interested in counting pairs of 1’s, the function f (n) = n2 appears quite

frequently. Let us extend this function to the real numbers in the obvious way: f (x) = 21 x(x − 1). Note that
f is a convex function. Using Jensen’s theorem, we can get the following inequality.

a1 a2 a3 an s(s − n)
+ + + ··· + ≥ ,
2 2 2 2 2n

where a1 , a2 , · · · , an are positive integers and s = a1 + a2 + · · · + an .

However, this bound is not always the best possible, since the equality is attained at ai = s/n for all i,
which may not be achieved as ai needs to be an integer. Using the fact that the ai ’s must be integers, we
can get the following tight bound, either using Karamata’s majorization inequality, or more simply through
discrete smoothing.

a1 a2 a3 an k+1 k
+ + + ··· + ≥r + (n − r)
2 2 2 2 2 2

where a1 , a2 , · · · , an are positive integers, and s = a1 + a2 + · · · + an = nk + r, where k and r are integers

such that 0 ≤ r < n.
Problem 3. (IMO 1998) In a competition, there are a contestants and b judges, where b ≥ 3 is an odd
integer. Each judge rates each contestant as either “pass” or “fail”. Suppose k is a number such that, for
any two judges, their ratings coincide for at most k contestants. Prove that k/a ≥ (b − 1)/2b.

Solution. Let us form an incidence matrix as usual. Let there be b rows, each representing a judge, and a
columns, each representing a contestant. Make the entries 1 or 0, representing “pass” and “fail”, respectively.
Let T denote denote the set of pairs of entries in the same column that are either both 0 or 1. Again,
we will count T in two different ways.
Counting by rows: Since the ratings of any two judges coincide for at most k contestants, for every two
rows, at most k pairs belong in T . Since there are 2b ways to choose two rows, we have |T | ≤ k 2b = kb(b−1)
2 .

p
Counting by columns: For a particular column, suppose there are p ones and q zeros, then there are
+ 2q pairs in T . Note that p + q = b is odd, using smoothing, we get

2
b+1 b−1
p q 2 2 (b − 1)2
+ ≥ + = .
2 2 2 2 4
a(b−1)2
Since there are a rows, we must have |T | ≥ 4 .
a(b−1)2 kb(b−1) a(b−1)2 kb(b−1)
Combining the two inequalities, we get 4 ≤ |T | ≤ 2 . It follows that 4 ≤ 2 and
thus ka ≥ b−1
2b .

4
MOP 2007 Black Group Counting in Two Ways Yufei Zhao

Counting with weights

Problem 4. (Iran 1999) Suppose that C1 , . . . , Cn (n ≥ 2) are circles of radius one in the plane such that
no two of them are tangent, and the subset of the plane formed by the union of these circles is connected.
Let S be the set of points that belong to at least two circles. Show that |S| ≥ n.

Let us set up a matrix with n columns, each representing an unit circle, and |S| rows, each representing an
intersection point. An entry is 1 if the corresponding point lies on the corresponding circle and 0 otherwise.
Since no circle is disjoint from the rest, every column contains at least two 1’s as no two circles are tangent.
As well, by definition, each row must contain at least two 1’s.
Let one focus one 1 in the incidence matrix, say aij = 1. Each one on row i distinct from ai,j corresponds
to a circle that goes through the point represented by row i. Any such circle meets the circle Cj at exactly
two points as no tangency is allowed. So we will associate each one in row i distinct from ai,j with a one
from column j different from ai,j that represents the second intersection. Note that no one in column j is
associated with two different 1’s on row i, as this would mean that three different unit circles are passing
through the same two points, which is impossible. Hence, there is an injection from the 1’s in row i to the
1’s in column j.

..
 
 . 
 1 → 1 
..
 
 

 ↑ . 

 ··· 1 ··· ai,j = 1 · · · 1 ··· 
 
 .. 

 . ↓ 


 1 ← 1 

..
.

How can we use this information? Well, this is where the weights come in.
Let us revisit the idea of counting 1’s. However, this time, we will assign a “weight” to each 1. For
example, if an incidence matrix has three 1’s on each row, and we assign a weight of 13 to each 1, then the
sum of all the weights is r, the number of rows. We will see momentarily why this might be useful.
Using the same idea, if we associate each one with a weight, in such a way the weights of all the 1’s
in each row sum to 1, then the sum of the weights of all the 1’s in the matrix equals to r. The following
proposition comes from this idea.
Proposition 3. Let A = (ai,j ) be an r × c matrix with row sums Ri , and column sums Cj . If Ri > 0 for
1 ≤ i ≤ r, then X ai,j
= r.
i,j
Ri

Similarly, if Cj > 0 for 1 ≤ j ≤ c, then

X ai,j
= c.
i,j
Cj

Proof. We have  
X ai,j r c r X r
X 1 X X 1
=  ai,j =
 Ri = 1 = r.
i,j
Ri i=1
Ri j=1 i=1
Ri i=1

The proof of the second part is done in a similar manner.

The following proposition leads to an application of this idea.

5
MOP 2007 Black Group Counting in Two Ways Yufei Zhao

Proposition 4. Let A = (ai,j ) be a r × c (0, 1)-matrix with row sums Ri and column sums Cj such that
Ri > 0 and Cj > 0 for 1 ≤ i ≤ r and 1 ≤ j ≤ c. If Cj ≥ Ri whenever ai,j = 1, then r ≥ c.
ai,j ai,j
Proof. When ai,j = 1, Ri ≤ Cj implies that R1i ≥ C1j . It follows that Ri ≥ Cj for all 1 ≤ i ≤ r and
1 ≤ j ≤ c. From Proposition 3, we have
X ai,j X ai,j
r= ≥ = c.
i,j
Ri i,j
Cj

Back to the problem. Since there is an injective mapping from the 1’s in row i to the 1’s in column j, we
see that Cj ≥ Ri whenever ai,j = 1. Therefore, therefore r ≥ c, and the result follows. .
We will play one more variation on this technique. Sometimes we may not be able to compare Ri and
Cj when ai,j = 1, but we may be able to make the comparison when ai,j = 0. The next proposition is an
analogue of Proposition 4.
Proposition 5. Let A = (ai,j ) be an r × c (0, 1)-matrix with row sums Ri , and column sums Cj . If
0 < Ri < c and 0 < Cj < r for 1 ≤ i ≤ r and 1 ≤ j ≤ c, and Cj ≥ Ri whenever ai,j = 0, then r ≥ c.

1 1
Proof. Suppose on the contrary that r < c. Then 0 < r−Cj < c−Ri whenever ai,j = 0. Hence, c−Ri < r−Cj ,
which implies
Ri Cj
< .
c − Ri r − Cj
Let M denote the number of 1’s in A, we have
r r r c
!
X X Ri X X Ri X (1 − ai,j )Ri
M= Ri = (c − Ri ) = (1 − ai,j ) =
i=1 i=1
c − Ri i=1 i=1
c − Ri i,j
c − Ri
c r
! c
X (1 − ai,j )Ci X X Cj X Cj
< = (1 − ai,j ) = (r − Cj ) =M
i,j
r − Cj j=1 i=1
r − Cj j=1
r − Cj

This is clearly impossible. Therefore, r ≥ c.

Problem 5. Let S1 , S2 , . . . , Sm be distinct subsets of {1, 2, . . . , n} such that |Si ∩ Sj | = 1 for all i 6= j. Prove
that m ≤ n.

This problem is a special case of Fisher’s inequality. It has a very simple and element proof using linear
algebra. However, in this note, we give the combinatorial solution to the problem following the models that
we have developed so far.

Proof. The result holds trivially if the collection is empty (m = 0) or m = 1. So we may assume that m ≥ 2.
It is easy to see that none of the sets Si are empty. So assume m ≥ 2 and all of the sets are non-empty.
As usual, we consider the incidence matrix A for the collection of sets. The m rows of A correspond to
sets and the n columns correspond to the elements, where ai,j is 1 if element j belongs to set Si , and is 0
otherwise.
Now let us show that the hypotheses of Proposition 5 are satisfied. If any row has all 1’s, say the first
row, then the constraint |S1 ∩ Si | = 1 for all i 6= 1 forces |Si | = 1, which, along with |Si ∩ Sj | = 1, implies
that m = 2, and n ≥ 2 because the sets are distinct. If any column has all zeros, then that element belongs
to none of the sets and we may simply remove that column. We may do this until every column satisfies
Cj ≥ 1 because if the result holds for this reduced matrix, it certainly holds for the original A. Finally, if
any column has all 1’s, say the first column, then |Si ∩ Sj | = 1 implies that no other column may contain
two 1’s. As well, at most one row may contain a single one (on the first column), and each of the other r − 1
rows must have the second one on distinct columns. So the number of columns must be greater than or
equal to the number of rows, giving m ≤ n in this case as well. We are now ready to employ Proposition 5.

6
MOP 2007 Black Group Counting in Two Ways Yufei Zhao

..
 
 . 
 1 → 1 
..
 
 

 . ↓ 

 ··· 1 ··· ai,j = 0 · · · 1 ··· 
 
 .. 

 ↑ . 


 1 ← 1 

..
.

Let us consider any ai,j = 0. By the given condition, for every one on column j, its corresponding
subset must intersect with Ai . So we may correspond each one on Cj with an one on row i such that the
element represented by the one on Ri also belongs to the subset represented by the one on Cj . Note that
this correspondence is injective, since having two 1’s on Cj both corresponding to the same one in Ri implies
that some two subsets intersect in at least two elements. The injective mapping implies that there must be
at least as many 1’s on ith row as there are on the j th column.
Thus Ri ≥ Cj for any ai,j = 0. It follows from Proposition 5 (with the roles of rows and columns
interchanged) that m ≤ n.

Final remarks Incidence matrices can be very useful for visualizing the combinatorial configuration.
However, when writing up a solution, it’s usually easier to avoid the incidence matrix and simply stick with
set theory notation instead.

Problems
1. (China 1993) A group of 10 people went to a bookstore. It is known that
(1) Everyone bought exactly 3 books;
(2) For every two persons, there is at least one book that both of them bought.
What is the least number of people that could have bought the book purchased by the greatest number
of people?
2. (IMO 2004 Shortlist) There are 10001 students at a 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. Suppose that the
following conditions hold:
(i) Each pair of students is in exactly one club.
(ii) For each student and each society, the student is in exactly one club of the society.
(iii) 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.
Find all possible values of k.
3. Let X be a finite set with |X| = n, and let A1 , A2 , . . . , Am be three-element substs √
of Xsuch that
|Ai ∩ Aj | ≤ 1 for all i 6= j. Show that there exists a subset A of X with at least 2n elements
containing none of the Ai ’s.
4. Let A1 , A2 , . . . A7 be subsets of M = {1, 2, . . . , 7}, such that each pair of elements of M belongs to
exactly one of the subsets, and |Ai | ≥ 3 for each i. Show that |Ai ∩ Aj | = 1 for all i 6= j.
5. An organization has n members, and it has n + 1 three-member committees, no two of which have
identical membership. Prove that there are two committees that share exactly one member.

7
MOP 2007 Black Group Counting in Two Ways Yufei Zhao

6. (China TST 1992) Sixteen students took part in a math competition where every problem was a
multiple choice question with four choices. After the contest, it is found that any two students had at
most one answer in common. Determine the maximum number of questions.
7. (China TST 1995) Twenty-one people took a test with 15 true and false questions. It is known that
for every two people, there is at least one question that both have answered correctly. Determine the
minimum possible number of people that could have correctly answered the question that most number
of people are correct on.
8. (China 1996) Eight singers participate in an art festival where m songs are performed. Each song is
performed by 4 singers, and each pair of singers performs together in the same number of songs. Find
the smallest m for which this is possible.

9. (Canada 2006) In a rectangular array of nonnegative real numbers with m rows and n columns, each
row and each column contains at least one positive element. Moreover, if a row and a column intersect
in a positive element, then the sums of their elements are the same. Prove that m = n.
10. (Iberoamerican Olympiad 2001) Let X be a set with n elements. Given k > 2 subsets of X, each with
nk
at least r elements, show that we can always find two of them whose intersection has at least r − 4(k−1)
elements.
11. (IMO 1989) Let n and k be positive integers and let S be a set of n points in the plane such that

(i) no three points of S are collinear, and

(ii) for any point P of S there are at least k points of S equidistant from P .
Prove that:
1 √
k< + 2n
2
12. Let A1 , A2 , . . . , Ak be subsets of S = {1, 2, . . . , 10} such that

(1) |Ai | = 5, i = 1, 2, . . . , k;
(2) |Ai ∩ Aj | ≤ 2, 1 ≤ i < j ≤ k.
Determine the maximum possible value of k.

13. (Burnside’s Lemma) Let G be a finite group that acts on the set A. For g ∈ G, let Fix(g) denote the
number
P of elements in A that are fixed by g. Show that the number of orbits of G on A is equal to
1
|G| Fix(g).
g∈G

14. Let A be a set with |A| = n, and let A1 , A2 , . . . , An be subsets of A with |Ai | ≥ 2, 1 ≤ i ≤ n. Suppose
that for each two-element subsets A0 of A there is a unique i such that A0 ⊆ Ai . Prove that Ai ∩ Aj 6= ∅
for any i 6= j.

15. (IMO 2001) Twenty-one girls and twenty-one boys took part in a mathematical competition. It turned
out that
(a) each contestant solved at most six problems, and
(b) for each pair of a girl and a boy, there was at least one problem that was solved by both the girl
and the boy.
Prove that there is a problem that was solved by at least three girls and at least three boys.
16. (IMO 2005) In a mathematical competition 6 problems were posed to the contestants. Each pair of
problems was solved by more than 25 of the contestants. Nobody solved all 6 problems. Show that
there were at least 2 contestants who each solved exactly 5 problems.

8
Winter Camp 2008 Combinatorics Yufei Zhao

Combinatorics
Yufei Zhao
[email protected]

1 Bijections
Basic examples
1. (a) Let n be a positive integer. In how many ways can one write a sum of at least two positive integers
that add up to n? Consider the same set of integers written in a different order as being different.
(For example, there are 3 ways to express 3 as 3 = 1 + 1 + 1 = 2 + 1 = 1 + 2.)
(b) Let m, n be positive integers. Determine the number of m-tuples of positive integers (x1 , x2 , . . . , xm )
satisfying x1 + x2 + · · · + xm = n.
(c) Let m, n be positive integers. Determine the number of m-tuples of nonnegative integers (x1 , x2 , . . . , xm )
satisfying x1 + x2 + · · · + xm = n.
2. Determine the number of paths from (0, 0) to (m, n) following the gridlines and moving in the up or
right directions.

1.1 Catalan numbers

Let us define the nth Catalan number Cn by

1 2n
Cn = .
n+1 n
So
(C0 , C1 , . . . ) = (1, 1, 2, 5, 14, 42, 132, 429, . . . ).
There are a huge number of combinatorial interpretations of these numbers1 , and we’ll discuss some of them
in lecture, and leave a few more as exercise in the next section.
There are two main way of handling these problems: bijection and recurrence. The bijections often
tend to be very elegant, while the recurrence method tend to be more routine. We focus on the bijection
perspective but we also briefly discuss the recurrence method.

1. Show that the number of lattice paths from (0, 0) to (n, n) using only up moves and right moves, and
never stepping above the x = y line, is Cn . E.g., for n = 4,

1 See http://www-math.mit.edu/~rstan/ec/ for a list of (currently) 161 interpretations.

1
Winter Camp 2008 Combinatorics Yufei Zhao

2. Show that the number of expressions containing n pairs of parentheses which are correctly matched is
Cn . E.g., for n = 3,
((())) (()()) (())() ()(()) ()()()
3. Show that the number of plane trees with n + 1 vertices is Cn . E.g., for n = 3,

4. Show that the number of complete binary trees with n internal vertices is Cn . E.g., for n = 3,

5. Show that the number of ways that n + 1 factors can be completely parenthesized is Cn . E.g., for
n = 3,
(((ab)c)d) ((a(bc))d) ((ab)(cd)) (a((bc)d)) (a(b(cd)))

A word about recursion. Try to show that every interpretation above gives the recurrence relation
Cn = Cn−1 C0 + Cn−2 C1 + · · · + C1 Cn−2 + C0 Cn−1 .
This is not too hard. It is mostly about how to break up a “Catalan problem” into two smaller “Catalan
subproblems.”
Now, if we turn the table around and ask: given
recurrence relation (and the initial conditions of course),
1 2n
how can we arrive at the formula Cn = 2n+1 n ? This is also rather difficult unless you have seen it before.
The most standard way is through generating functions (if you know generating functions, you should try
to work out this computation yourself).

1.2 More Catalan exercises

1. Show that the number of triangulations of a convex (n + 2)-gon is Cn . E.g., for n = 4,

2. Show that the number of ways to tile a stairstep shape of height n with n rectangles is Cn . E.g., for
n = 4,

2
Winter Camp 2008 Combinatorics Yufei Zhao

3. Show that the number of ways of stacking coins in the plane so that the bottom row consists of n
consecutive coins is Cn . E.g., for n = 3,

4. Show that the number of ways of drawing n nonintersecting chords joining 2n given points on a circle
is Cn . E.g., for n = 3,

5. Show that Cn equals to the number of (unordered) pairs of lattice paths with n + 1 steps each, starting
at (0, 0), using steps (1, 0) or (0, 1), ending at the same point, and only intersecting at the beginning
and end. E.g., for n = 3, The 64th William Lowell Putnam Mathematical Competition

r r
Saturday, December 6, 2003

r r r r r r r r r r
r r r r r r r r r r r r r r
r r rA1 Let
r be a fixed rpositive
r integer.
r
Howr many
r ways
r are r B1 rDo there
r exist
r polynomials %_4 & # %`4 & ? %`a & b %_a & such
,with
there to write as a sum of positive integers,



an arbitrary positive inte-
that
c 4da 4 a J %_4 & ? %`a & # %`4 & b %`a &
1.3 More bijections there are four ways: 4, 2+2, 1+1+2,
ger and ? For example, with

A2 Let "! and # # $ # ! be nonnegative

1+1+1+1.
holds identically?
an equilateral triangle of side length n by n2 equilateral triangles
1. A triangular grid is obtained by tiling
!& (' ! % # # # !& (bounded
% ofparallelograms
real numbers. Show that
'! B2 Let X be a positive integer. Starting with the sequence
e f e@he@ fiMe3 ggX ofge@3g the
,%jXform L L L en-
of side length 1. Determine the number
*) % # +& % # & % ! # !&-,
' !
by line segments
tries
grid.

a new
& e X %
sequence
& of
by taking the

L X theentries,
averages of two consecutive entries in the first se-
quence. Repeat the averaging of neighbors on sec-
.0/1324 56 /4 87 9 2:4 56$7 4 /; 5 4 <5 / 5 4=.
A3 Find the minimum value of ond sequence to obtain a third sequence of
4 final sequence4 produced X e . consists
4 of a single number ! . Show that !lk
and continue until the

A4 Suppose that > # ?@ AB CD E are real numbers, H F G B3 Show that for each! positive integer n,
for real numbers .

and AIJ F G , such that

2. Form a 2000 × 2002 screen with unit screens. . 4 # 4 Initially, ? . . A 4 there C 4 KE are . more than 1999 ×nm"2001 pgo q r 5unit slt" X screens u vwxeyjzM{M
as soon as4 there are 3 unit screens which are off, the 4th screen
E . be totally off.(Here r 5s denotes the least common multiple.)
which are on. In any 2 × 2 screen, for all real numbers . Show that
turns off automatically. Prove that the whole . # :L Mscreen ? . . C LcanMA never
% 1)& andand n downsteps B4 Let |
3. (Putnam 2003) A Dyck n-pathA5is AaDyck lattice %
L is a& lattice
path of path n upsteps of upsteps(1,
N %_} & J(1, }M~ −1) %S} # }M %_that ? } starts
}<downsteps b }L <€ %_} L
4
-path
downsteps
at the origin and never dips belowdips the x-axis. that
A starts
return
below the -axis. A return is a maximal sequence
at the is origin
a maximal and
4 -axis. has two returns,P of length
never sequence of
J

contiguous
L
& L S
% }
P P  3& P ~ &
&
that terminates on the x-axis. For example, the Dyck 5-path illustrated
where > # ?@bh€ are integers, H F G . Show that if
of contiguous downsteps that terminates on the

Pthen P P P isisaarational number and P ‚ P ƒF P  P ~ ,

and 1 respectively. For example, the Dyck 5-path illustrated has two re-
turns, of length 3 and 1 respectively.
rational number.
B
A
C , and E be equidistant points on the circumfer-
ence of a circle of unit radius centered at N , and let „
B5 Let

be any point in the circle’s interior. Let > # ? be the dis-

tance from „ to A C† E , respectively. Show that there
is a triangle with side lengths h # ? and that the area of
% L |with depends even from „ to N .
Show that there is a one-to-one correspondenc between
Show that there is a one-to-one the correspondence
Dyck
& -paths.
-paths with no return between of even the lengthDyck and the n-paths this triangle
_
% 4
no return only on the of distance
Dyck
A6 For a set O of nonnegative integers,
length and the Dyck (n − 1)-paths.
let P@Q
% & denother &
S
% R @
R & R U
T on the interval ) G , . Show that
O
B6 Let be a continuous real-valued function defined

R VT O , R triangle
the
4. (Canada 2005) Consider an equilateral number
W F RX , and ofR Y side
of ordered pairs R Z[length such that ,
n, which is divided
‡ ‡ into | unit | triangles, ‡ . | %`4 . 4
C in such % integers
the % triangle
P]^ path
. Is it possible to
A .
ˆ ˆ & & b triangle `
% 4 _
% a . 4 †
a
b ˆ &b ‰
a way that P]\ in& our & for allshare
as shown. Let f (n) be the number partition of the paths nonnegative from into two setsin and the top row to the middle
in the bottom row, such that adjacent triangles ? a common edge and the path never
travels up (from a lower row to a higher row) or revisits a triangle. An example of one such path is
illustrated below for n = 5. Determine the value of f (2005).

3
f (n) be the number of paths from the triangle in the top row to the middle triangle in the bottom
row, such that adjacent triangles in our path share a common edge and the path never travels up
(from a lower row to a higher row) or revisits
Winter Camp 2008
a triangle. An example of one such path
Combinatorics
is illustrated
Yufei Zhao
below for n = 5. Determine the value of f (2005).

2. Let (a, b, c) be a2005)

5. (Putnam Let S = {(a,
Pythagorean b) | ai.e.,
triple, = 1,a2,triplet
. . . , n, b of
= positive integers with
1, 2, 3}. Determine a2 + b2of=paths
the number c2 . in S
starting at (1, 1) and ending at (n, 1), with unit steps and passing through all the points in S exactly
that (c/a + c/b)2 > 8.
once each.
a) Prove
b) 6.Prove
(IMOthat there
Shortlist doesLet
2002) not
n exist any integer
be a positive n for
integer. Eachwhich
point we
(x, can
y) infind a Pythagorean
the plane, where x andtriple
y are(a, b, c)
non-negative integers with
2
satisfying (c/a + c/b) = n. x + y < n, is colored red or blue, subject to the following condition: if a
0 0 0 0
point (x, y) is red, then so are all points (x , y ) with x ≤ x and y ≤ y. Let A be the number of ways
to choose
3. Let S be n blue
a set of n ≥points with in
3 points distinct x-coordinates,
the interior and let B be the number of ways to choose n blue
of a circle.
points with distinct y-coordinates. Prove that A = B.
a) 7.Show
(USAMOthat1996)
thereAnaren-term
threesequence
distinct(xpoints a, b, c ∈ S and three distinct points A, B, C on the
1 , x2 , . . . , xn ) in which each term is either 0 or 1 is called a
circle
binarysuch thatofa length
sequence is (strictly)
n. Let acloser
n be theto A than
number any other
of binary point
sequences of in S, bn is
length closer tonoBthree
containing than any
consecutive
other point terms equalc to
in S and is 0,closer
1, 0 in
tothat
C thanorder.any Let other
bn be the number
point in S.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
b) Show that
bn+1 = 2anfor
for no value ofintegers
all positive n cann.four such points in S (and corresponding points on the circle)
be guaranteed.
8. Along a one-way street there are n parking lots, and n cars numbered 1 to n enter the street in that
4. Let ABCorder.beEach driver ofwith
a triangle the ith car heads to his
circumradius R, favorite parking
perimeter lot aiarea
P and ∈ {1,K.
2, . . .Determine
, n}, and, if it
theis free,
maximum
he occupies3 it. Otherwise, he continues to the next free lot and occupies it. But if all succeeding lots
value ofareKP/R .
occupied, he leaves for good. How many sequences (a1 , a2 , . . . , an ) are there such that every driver
can park?
5. Let’s say that an ordered triple of positive integers (a, b, c) is n-powerful if a ≤ b ≤ c, gcd(a, b, c) = 1,
an(IMO
and 9. + bn Shortlist
+ cn is divisible
2002) Let by
n be
a+a positive integer.
b + c. For A sequence
example, (1, 2, of
2) nispositive integers (not necessarily
5-powerful.
distinct) is called full if it satisfied the following condition: for each positive integer k ≥ 2, if the
number
a) Determine k appears in thetriples
all ordered sequence (ifthen
any)sowhich
does the
arenumber k − 1, and
n-powerful moreover
for all n ≥ 1.the first occurrence
of k − 1 comes before the last occurrence of k. For each n, show that there are n! full sequences.
b) Determine all ordered triples (if any) which are 2004-powerful and 2005-powerful, but not 2007-
(IMO Shortlist 2005) In an m × n rectangular board of mn unit squares, adjacent squares are ones with
10.powerful.
a common edge, and a path is a sequence of squares in which any two consecutive squares are adjacent.
Each square of the board can be colored black or white. Let N denote the number of colorings of the
[Note that gcd(a, b, c) is the greatest common divisor of a, b and c.]
board such that there exists at least one black path from the left edge of the board to its right edge,
and let M denote the number of colorings in which there exist at least two non-intersecting black paths
from the left edge to the right edge. Prove that N 2 ≥ M · 2mn .

2 Counting in two ways

1. (IMC 2002) Two hundred students participated in a mathematical contest. They had six problems to
solve. It is known that each problem was correctly solved by at least 120 participants. Prove that there
must be two participants such that every problem was solved by at least one of these two students.

2. (IMO 1998) In a competition, there are a contestants and b judges, where b ≥ 3 is an odd integer.
Each judge rates each contestant as either “pass” or “fail”. Suppose k is a number such that, for any
two judges, their ratings coincide for at most k contestants. Prove that
k b−1
≥ .
a 2b

4
Winter Camp 2008 Combinatorics Yufei Zhao

3. Show that if the edges of K6 , the complete graph with 6 vertices, are colored in 2 colors, then graph
contains two monochromatic triangles. (Hint: count the number of monochromatic “angles”)
4. (Russia 1990) There are 30 senators in a senate. Each pair of senators, the two senators are either
friends of each other or enemies of each other. Every senator has exactly six enemies. Every three
senators form a committee. Find the total number of committees whose members are either all friends
or all enemies of each other.
5. (China 1993) Ten students ordered books. Each student ordered 3 different books. Each pair of
students had ordered at least one same book. The book Mathematics Olympiads was the one which
most (a tie being allowed) students ordered. What was the minimum number of students who ordered
Mathematics Olympiads?

6. (USA TST 2005) 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
(a) each element of T is an m-element subset of Sm ;
(b) each pair of elements of T shares at most one common element; and
(c) each element of Sm is contained in exactly two elements of T .
Determine the maximum possible value of m in terms of n.
7. (China TST 1992) Sixteen students took part in a math competition where every problem was a
multiple choice question with four choices. After the contest, it is found that any two students had at
most one answer in common. Determine the maximum number of questions.
8. (IMO Shortist 2000) Let n ≥ 4 be a fixed positive integer. Let S = {P1 , P2 , . . . } be a set of n points
in the plane such that no three are collinear and no four are concyclic. Let at , 1 ≤ t ≤ n, denote the
number of circles Pi Pj Pk that contain Pt in their interiors, and let m(S) = a1 + a2 + · · · + an . Prove
that there exists a positive integer f (n), depending only on n such that the points of S are the vertices
of a convex polygon if and only if m(S) = f (n).

9. Let X be a finite set with |X| = n, and let A1 , A2 , . . . , Am be three-element substs √

of Xsuch that
|Ai ∩ Aj | ≤ 1 for all i 6= j. Show that there exists a subset A of X with at least 2n elements
containing none of the Ai ’s.
10. (Canada 2006) Consider a round-robin tournament with 2n + 1 teams, where each team plays each
other team exactly once. We say that three teams X, Y and Z, form a cycle triplet if X beats Y , Y
beats Z, and Z beats X. There are no ties.
(a) Determine the minimum number of cycle triplets possible.
(b) Determine the maximum number of cycle triplets possible.

3 Binomial sums exercises

Prove the following identities through combinatorial interpretations. (You can assume that the variables are
nonnegative integers and that all the expressions make sense.)

n−1

n
1. Prove that k =n .
k k−1
n−k

n m n
2. Prove that = .
m k k m−k

n n n
3. Prove that + + ··· + = 2n .
0 1 n

5
Winter Camp 2008 Combinatorics Yufei Zhao

n n n
4. Prove that 1 · +2· + ··· + n · = n · 2n−1 .
1 2 n
n−1 n−2

n n+1
5. Prove that + + + ··· = .
k k k k+1
n 2
2n − 1

X n
6. Prove that k =n .
k n−1
k=0

k
n+m X n m
7. (Vandermonde’s identity) Prove that = .
k i=0
i k−i

k
n−1

X n
8. Prove that (−1)i = (−1)k .
i=0
i k
n
n−k

X n n m
9. Prove that =2 .
k m−k m
k=0

n
n−k

X n 2n + 1
10. (China 1994) Prove that 2k = .
k b(n − k)/2c n
k=0

6
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Lecture 1 : Pigeonhole Principle

Yufei Zhao
July 17, 2007

1. Let a and m be positive integers. Show that the sequence a, a2 , a3 , a4 , . . . is eventually periodic
mod m.

2. Let Fn be the Fibonacci numbers, defined by F1 = F2 = 1 and Fn+2 = Fn+1 + Fn . Show that
for some n ≥ 1, Fn ends with 2007 zeros.

3. Let S be a subset of {1, 2, 3, . . . , 2n} with n + 1 elements.

(a) Show that there are two elements in S which are relatively prime.
(b) Show that there are two elements in S, one divisible by the other.

4. The edges of K6 , the complete graph with 6 vertices, are each colored in red or blue. Prove
that there is a monochromatic triangle.

5. Let a1 , a2 , . . . , a20 be distinct positive integers not exceeding 70. Show that there is some k
so that ai − aj = k for four different pairs (i, j).

6. Let a1 , a2 , . . . , an be positive integers. Prove that we can choose some of these numbers to
obtain a sum divisible by n.

7. Suppose that a and b are relatively prime integers. Show that there exist integers x and y
such that ax + by = 1.

8. Let p and q be positive integers. Show that within any sequence of pq+1 distinct real numbers,
there exists either a increasing subsequence of p + 1 elements, or a decreasing subsequence of
q + 1 elements.

1
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Problem Solving Session

July 17, 2007

1. Five lattice points are chosen in the plane lattice. Prove that you can choose two of these
points such that the segment joining these points passes through another lattice point. (The
plane lattice consists of all points of the plane with integral coordinates.)

2. Given 7 lines on the plane, prove that two of them form an angle less than 26◦ .

3. A closed unit disc contains 7 points such that any two of them are at least unit distance
apart. Show that the center of the disc is one of the 7 points.

4. At a party, certain pairs of individuals have shaken hands. Prove that there exist two persons
who have shaken the same number of hands.

5. Consider the set M = {1, 2, 3, . . . , 2007}. Prove that in any way we choose the subset X with
15 elements of M there exist two disjoint subsets A and B in X such that the sum of the
members of A is equal to the sum of the members of B.

6. A chessmaster has 77 days to prepare for a tournament. He wants to play at least one game
per day, but not more than 132 games. Prove that there is a sequence of successive days on
which he plays exactly 21 games.

inside a 3 × 4 rectangle. Show that two of the points in S have a

7. Let S be a set of 6 points √
distance not greater than 5.

8. (Canada 2004) Let T be the set of all positive integer divisors of 2004100 . What is the largest
possible number of elements that a subset S of T can have if no element of S is an integer
multiple of any other element of S?

2
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Lecture 2 : More Pigeonholes and Some Coloring

Yufei Zhao
July 18, 2007

1. Show that there is some n for which 111 · · · 111 (with n ones) is divisible by 2007.
√ 1
2. Show that there exists some integer n such that the number n π differs by at most 1000000
from its nearest integer.

3. Let α be some irrational number. Show that for any a, b satisfying 0 < a < b < 1, there is
some positive integer n such that a < {nα} < b.

4. Prove that there exists a positive integer n such that the four leftmost digits of the decimal
representation of 2n is 2007.
a−b √1 .
5. Prove that among any seven real numbers there are two, say a and b, such that 0 ≤ 1+ab ≤ 3

6. (a) Can you tile a 6 × 6 board with 1 × 4 tiles?

(b) Let k be a positive integer. For which positive integers m, n can a m × n rectangle can
be tiled with 1 × k tiles?
(c) Let a, b, c be positive integers such that a|b and b|c. Suppose that a rectangular box can
be tiled with a × b × c bricks. Show that one can tile the box with all the a × b × c bricks
arranged in the same orientation.

7. What is the smallest number of squares on an 8×8 chessboard which would have to be painted
so that no 3 × 1 rectangle could be placed on the board without covering a painted square?

8. For which n is there a closed knight’s tour on a 4 × n chessboard?

9. (USAMO 1998) A computer screen shows a 98 × 98 chessboard, colored in the usual way. One
can select with a mouse any rectangle with sides on the lines of the chessboard and click the
mouse button: as a result, the colors in the selected rectangle switch (black becomes white,
white becomes black). Determine the minimum number of mouse clicks needed to make the
chessboard all one color.

3
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Problem Solving Session

July 18, 2007

1. Show that there is a positive integer n so that | sin n| < 10−10 .

2. Let A1 A2 · · · A2n be a convex polygon. Let P be a point in the interior of the polygon, such
that P does not lie on any of its diagonals. Prove that there exists a side ` of the polygon
such that none of the lines P A1 , P A2 , . . . , P A2n intersects the interior of `.

3. Let S be the set of 25 points arranged in a 5 × 5 unit square array. Show that among any 6
points in S, we can always find three of them so that the area of the triangle they form is at
most 2.

4. Several chords of a unit circle are chosen such that no diameter intersects with more than
four of them. Prove that the sum of the lengths of the chords is at most 13.

5. (Canada 2007) What is the maximum number of dominoes which can be placed on an 8 × 9
board if six are already placed as shown below?

• •
• •
• •
• •
• •
• •

6. Prove that a 4 × 11 rectangle cannot be tiled by L-shaped tetrominoes.

7. Show that if a rectangle can be tiled by smaller rectangles each of which has at least one
integer side, then the tiled rectangle has at least one integer side.

8. (IMO Shortlist 2003) Let D1 , . . . , Dn be closed discs in the plane. 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 other discs Di .

4
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Lecture 3 : Binomial Coefficients

Yufei Zhao
July 19, 2007

n
X n k n−k
n
1. Show that (x + y) = x y .
k
k=0

n+1 n n
2. Show that = + .
k+1 k+1 k
m
X n+k n+m+1
3. Show that = .
k m
k=0

n n−1

n
4. Show that =
k k k−1
n
n
= n · 2n−1 .
X
5. Show that k
k
k=0
n
X
2n
6. Evaluate k .
k
k=0

2007 2007 2007 2007
7. Evaluate + + + ··· + .
1 4 7 2005
8. Stirling numbers

n
(a) Stirling numbers of the first kind: Let denote the number of permutations on
k

n+1 n n
n elements with exactly k cycles. Show that = +n .
k k−1 k

n
(b) Stirling numbers of the second kind: Let denote the number of ways to
k

n+1 n n
partition {1, 2, . . . , n} into k non-empty sets. Show that = +k .
k k−1 k
n
X n k
(c) Show that x = x(x + 1)(x + 2) · · · (x + n − 1).
k
k=0
n
n
x(x − 1)(x − 2) · · · (x − k + 1) = xn .
X
(d) Show that
k
k=0
(
X
n−k m k 1 if m = n,
(e) Show that (−1) =
k≥0
k n 0 if m 6= n.

5
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Problem Solving Session

July 19, 2007

n−m

n k n
1. Show that = .
k m m k−m
n
n
= 2n .
X
2. Show that
k
k=0

X n X n
3. Show that = .
k k
odd k even k
n
X 1 n
4. Evaluate .
k+1 k
k=0
n
n−k

X n m n
5. Show that =2 .
k m−k m
k=0
m
m n−1

X
kn
6. Prove that (−1) = (−1) .
k m
k=0

p
7. Let p be a prime number, and let k be an integer such that 0 < k < p. Show that is
k
divisible by p.
X n
a+b a b
8. (Vandermonde) Show that = .
n k n−k
k=0

2p
9. Show that − 2 is divisible by p2 for all primes p > 2.
p
10. Let n be a nonnegative integer. Show that
n 2
2n − 1

X n
k =n .
k n−1
k=0

11. A license plate consists of 8 digits. It is called even if it contains an even number of 0s. Find
the number of even license plates.

12. Let Fn be the Fibonacci numbers defined by F1 = F2 = 1 and Fn+2 = Fn+1 + Fn . Prove that
n
n−k+1
X
= Fn+2 .
k
k=0

6
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Lecture 4 : Bijections
Yufei Zhao
July 20, 2007

1. (a) Let n be a positive integer. In how many ways can one write a sum of at least two positive
integers that add up to n? Consider the same set of integers written in a different order
as being different. (For example, there are 3 ways to express 3 as 3 = 1 + 1 + 1 = 2 + 1 =
1 + 2.)
(b) Let m, n be positive integers. Determine the number of m-tuples of positive integers
(x1 , x2 , . . . , xm ) satisfying x1 + x2 + · · · + xm = n.
(c) Let m, n be positive integers. Determine the number of m-tuples of nonnegative integers
(x1 , x2 , . . . , xm ) satisfying x1 + x2 + · · · + xm = n.
2. Determine the number of paths from (0, 0) to (m, n) following the gridlines and moving in
the up or right directions.

3. A triangular grid is obtained by tiling an equilateral triangle of side length n by n2 equilateral

triangles of side length 1. Determine the number of parallelograms bounded by line segments
of the grid.

1 2n
4. Catalan numbers. Let n be a positive integer and let Cn = .
n+1 n
(a) Show that the number of lattice paths from (0, 0) to (n, n) using only up moves and
right moves, and never stepping above the x = y line, is Cn . E.g., for n = 4,

(b) Show that the number of expressions containing n pairs of parentheses which are correctly
matched is Cn . E.g., for n = 3,
((())) (()()) (())() ()(()) ()()()

7
AwesomeMath 2007 Track 1 — Combinatorics Week 1

(c) Show that the number of plane trees with n + 1 vertices is Cn . E.g., for n = 3,

(d) Show that the number of plane binary trees with n + 1 leaves is Cn . E.g., for n = 3,

(e) Show that the number of ways that n + 1 factors can be completely parenthesized is Cn .
E.g., for n = 3,

(((ab)c)d) ((a(bc))d) ((ab)(cd)) (a((bc)d)) (a(b(cd)))

(f) Show that the number of triangulations of a convex (n + 2)-gon is Cn . E.g., for n = 4,

(g) Show that the number of ways to tile a stairstep shape of height n with n rectangles is
Cn . E.g., for n = 4,

(h) Show that Cn satisfies the recurrence relation

Cn = Cn−1 C0 + Cn−2 C1 + · · · + C1 Cn−2 + C0 Cn−1 .

8
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Problem Solving Session

July 20, 2007

1. (AHSME 1992) Ten points are selected on the positive x-axis, X+ , and five points are selected
on the positive y-axis, Y+ . The fifty segments connecting the ten points on X+ to the five
points on Y+ are drawn. What is the maximum possible number of points of intersection of
these fifty segments in the interior of the first quadrant?

2. (HMMT 2007) On the Cartesian grid, Johnny wants to travel from (0, 0) to (5, 1), and he
wants to pass through all twelve points in the set S = {(i, j) | 0 ≤ i ≤ 1, 0 ≤ j ≤ 5, i, j ∈ Z}.
Each step, Johnny may go from one point in S to another point in S by a line segment
connecting the two points. How many ways are there for Johnny to start at (0, 0) and end at
37th
(5, 1) so that he neverCanadian Mathematical
crosses his own path? Olympiad
Wednesday, March 30, 2005

3. A triangular grid is obtained by tiling an equilateral triangle of side length n with n2 equi-
lateral triangles of side length 1. Determine the number of rhombi of side length 1 bounded
by line segments of the grid.
1. Consider an equilateral triangle of side length n, which is divided into unit triangles, as shown. Let
4. (Canada 2005) Consider an equilateral triangle of side length n, which is divided into unit
f (n) betriangles,
the number of paths from the triangle in the top row to the middle triangle in the bottom
as shown. Let f (n) be the number of paths from the triangle in the top row to the
row, such that adjacent triangles
middle triangle in the bottominrow,oursuch
paththat
share a common
adjacent edge
triangles andpath
in our the share
path anever
commontravels up
(from aedge
lower
androwthetopath
a higher
never row)
travelsorup
revisits
(from aa lower
triangle.
row An
to aexample of one
higher row) such path
or revisits is illustrated
a triangle.
below for
An nexample
= 5. Determine
of one suchthe value
path of f (2005).
is illustrated below for n = 5. Determine the value of f (2005).

2. Let (a, b, c) be a Pythagorean

5. Determine the number of triple,
ways ofi.e., a triplet
stacking coinsofinpositive integers
the plane so thatwith a2 + b2row
the bottom 2
= cconsists
.
of n consecutive coins.
a) Prove that (c/a + c/b)2 > 8.
b) Prove that there does not exist any integer n for which we can find a Pythagorean triple (a, b, c)
satisfying (c/a + c/b)2 = n.

3. Let S be a set of n ≥ 3 points in the interior of a circle.

6. Determine the number of ways of drawing n nonintersecting chords joining 2n given points
a) Show
on a that
circle.there are three distinct points a, b, c ∈ S and three distinct points A, B, C on the
circle such that a is (strictly) closer to A than any other point in S, b is closer to B than any
other point in S and c is closer to C than any other point in S.
b) Show that for no value of n can four such points in S (and corresponding points on the circle)
be guaranteed.

4. Let ABC be a triangle with circumradius R, 9perimeter P and area K. Determine the maximum
value of KP/R3 .

5. Let’s say that an ordered triple of positive integers (a, b, c) is n-powerful if a ≤ b ≤ c, gcd(a, b, c) = 1,
AwesomeMath 2007 Track 1 — Combinatorics Week 1

7. (HMMT 2007) A subset S of the nonnegative integers is called supported if it contains 0, and
k + 8, k + 9 ∈ S for all k ∈ S. How many supported sets are there?

8. (IMO Shortlist 2002) 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 colored red or blue, subject to the following
condition: if a point (x, y) is red, then so are all points (x0 , y 0 ) with x0 ≤ 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.

10
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Lecture 5 : More Bijections

Yufei Zhao
The
July64th
21, William
2007 Lowell Putnam Mathematical Competition
Saturday, December 6, 2003

1. (a) Suppose that n lines are drawn on the plane, no two parallel and no three concurrent.
Determine the number of regions created by these lines.
(b) Suppose that n pointsA1are Let chosen on a integer.
be a fixed positive circleHow andmanyline
segments
ways are B1 Do are drawn
there exist %_4 & # %`4 & ? %`a & b
connecting
polynomials
%_a & such
every pair of them, suchthere that
to write

no three
,with


as a sum of positive integers,

line

segments
an arbitrary
that
are concurrent.
positive inte-
c
4da Determine
4 a ofJ answers
%_4 & ? %`a & the
# %`4 & b %`a &
number of regions of the ger there are four ways: 4, 2+2, 1+1+2, 1+1+1+1.
disk and formed by these? For lineexample, with
segments. (The sequence
begins as 1, 2, 4, 8, 16.) A2 Let "! and # # $ # ! be nonnegative holds identically?

!& (' ! % Initially,

% screens. # # !& (' ! there
real numbers. Show that
e X f e@he@ fiMe1999
are morethan 3 ggX ge@3g ,%jXform
× L 2001 L & byof taking en-the L
# X sequence
% unit
B2 Let be a positive integer. Starting with the sequence
2. Form a 2000 × 2002 screen with unit
*
) % +
& % & % !
! -
& ,
' ! &
a new
e
screens which are on. In any 2 × 2 screen, # as # soon as # there are 3 unitaverages tries
screens of two which are off,in the first se-
L X the sec-
consecutive entries
the 4th screen turns off automatically. Prove that the whole screen can
quence. never Repeat be the averaging totally off.
of neighbors on
.0/1324 56 /4 87 9 2:4 56$7 4 /; 5 4 <5 / 5 4=.
A3 Find the minimum value of ond sequence to obtain a third sequence of
4 ! 4 l
! k X
e
entries,

4 the number of partitions ofofa single

and continue until the final sequence produced consists
3. Let m, n be positive integers. Let p(n, m) n into numberm . parts, Show that and .

A4 Suppose that > # ?@ AB CD E are real numbers, H F G B3 Show that for each! positiveq(n,
for real numbers .
q(n, m) the number of partitions of n whose largest part is m. Prove that p(n, m) = integerm).
I
A J
F G
n,

Let n be a positive integer. Let f (n). 4 denote

and
# 4 ? . the. A 4 number
, such that
C 4 KE . of ways to partitionnm" npgq into o X u vwxeyjzM{M
5slt"distinct
4.
4 thatn into odd parts. Show that f (n) = g(n).
r
. # :L M . ? Show
. . C L MA E . (Here 5s denotes the least common multiple.)
parts, and g(n) the number offorwaysall real of partition
r
numbers

5. (Putnam 2003) A Dyck n-path is a lattice path of n upsteps %

& Let | n downsteps (1, −1)
(1, 1)B4and
A5 A Dyck -path
that starts at the origin O and never dips % L 4is a& below
lattice path of upsteps
the x-axis. N and A
and
return is a maximal %_} & J }M~ L # sequence }M <L ? } b }L <of€ L
downsteps
contiguous downsteps that terminates on the x-axis. For example,
dips below the
that starts at
-axis. A return
the origin
is a maximal 4
sequence
never
the Dyck 5-pathP & illustrated J
S
% }
%_} P & %S} P  & %_} P ~ &
where > # ?@bh€ are integers, H F G . Show that if
of contiguous downsteps that terminates on the -axis.
has two returns, of length 3 and 1 respectively.
Pthen P P P isisaarational number and P ‚ P ƒF P  P ~ ,
For example, the Dyck 5-path illustrated has two re-
turns, of length 3 and 1 respectively.
rational number.
B
A
C , and E be equidistant points on the circumfer-
ence of a circle of unit radius centered at N , and let „
B5 Let

be any point in the circle’s interior. Let > # ? be the dis-

tance from „ to A C† E , respectively. Show that there
is a triangle with side lengths h # ? and that the area of
% L | with of from „ to N .
Show that there is a one-to-one correspondenc between
Show that there is a one-to-one the Dyck
correspondence & -paths.
-paths with no return of even length and the
between the Dyck n-paths this triangle
_
% 4
dependsno onlyreturn
on the distance
Dyck
a set O of nonnegative integers, let P@Q
% & denother &
S
% R R R on the interval ) G , . Show that
even length and the Dyck (nA6−For1)-paths.
@
& U
T O
B6 Let be a continuous real-valued function defined

R VT O , R W F RX , and R Y R Z[ . Is it possible to

the number of ordered pairs such that
A
,
‡ ‡ . | %`4 | %_a . 4 a†‰ ‡ . | %`4 . 4
partition the nonnegative% integers% into two sets and
C in such a way that P]\ & P]^ & for all ? ˆ ˆ & & b b ˆ & b

11
AwesomeMath 2007 Track 1 — Combinatorics Week 1

Evaluation Test 1
July 21, 2007

Instructions: Read the problems carefully. Write your solutions neatly and concisely, but make
sure to justify all your steps. Start each new solution on a new page and write your name and the
problem number on every page.
Each problem is worth 10 points.

1. Twenty five boys and twenty five girls sit around a table. Prove that, no matter what the
arrangement, it is always possible to find someone sitting in between two girls.

2. Is there a closed knight’s tour on a 5 × 5 chessboard?

3. Evaluate
n 2 n 2 n n+1 2 n
−2 +3 − · · · + (−1) n .
1 2 3 n

4. Let n be an integer greater than one, and let Tn be the number of nonempty subsets S of
{1, 2, 3, . . . n} with the property that the average of the elements of S is an integer. Prove
that Tn − n is always even.

12
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Lecture 10 : A Contest of Contests

Yufei Zhao
July 31, 2007

1. (IMC 2002) Two hundred students participated in a mathematical contest. They had six
problems to solve. It is known that each problem was correctly solved by at least 120 partic-
ipants. Prove that there must be two participants such that every problem was solved by at
least one of these two students.

2. (IMO 1998) In a competition, there are a contestants and b judges, where b ≥ 3 is an odd
integer. Each judge rates each contestant as either “pass” or “fail”. Suppose k is a number
such that, for any two judges, their ratings coincide for at most k contestants. Prove that
k b−1
≥ .
a 2b

3. (China TST 1992) Sixteen students took part in a math competition where every problem
was a multiple choice question with four choices. After the contest, it is found that any two
students had at most one answer in common. Determine the maximum number of questions.

4. (IMO 2005) In a mathematical competition 6 problems were posed to the contestants. Each
pair of problems was solved by more than 25 of the contestants. Nobody solved all 6 problems.
Show that there are at least 2 contestants who each solved exactly 5 problems each.

1
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Problem Solving Session

July 31, 2007

1. Show that if the edges of K6 , the complete graph with 6 vertices, are colored in 2 colors, then
graph contains two monochromatic triangles. (Hint: count the number of monochromatic
“angles”)

2. (China 1993) Ten students ordered books. Each student ordered 3 different books. Each pair
of students had ordered at least one same book. The book Mathematics Olympiads was the
one which most (a tie being allowed) students ordered. What was the minimum number of
students who ordered Mathematics Olympiads?

3. Twenty-five people form some committees with each committee has 5 members and each
pair of committees have at most one common member. Determine, with justification, the
maximum number of committees.

4. (USA TST 2005) 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

(a) each element of T is an m-element subset of Sm ;

(b) each pair of elements of T shares at most one common element; and
(c) each element of Sm is contained in exactly two elements of T .

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

5. (APMO 2006) In a circus, there are n clowns who dress and paint themselves up using a
selection of 12 distinct colors. Each clown is required to use at least five different colors. One
day, the ringmaster of the circus orders that no two clowns have exactly the same set of colors
and no more than 20 clowns may use any one particular color. Find the largest number n of
clowns so as to make the ringmasters order possible.

6. (IMO Shortlist 2004) There are 10001 students at a 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. Suppose that the following conditions hold:

(a) Each pair of students is in exactly one club.

(b) For each student and each society, the student is in exactly one club of the society.
(c) 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.

Find all possible values of k.

2
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Lecture 11 : Generating Functions I

Yufei Zhao
August 1, 2007

Suppose that (an )∞ 1

n=0 is a sequence of (complex) numbers. The generating function of the sequence
(an ) is the following (formal) power series
X
F (x) = an xn = a0 + a1 x + a2 x2 + a3 x3 + · · · .
n≥0

1. Vandermonde’s identity. By comparing the coefficient of xn in (x+1)a+b = (x+1)a (x+1)b ,

Show that X n
a+b a b
= .
n k n−k
k=0

2. Find the closed form expressions for the following generating functions:
X
(a) xn
n≥0
X
(b) nxn
n≥0
X m + n
(c) xn , where m is some positive integer.
m
n≥0

3. Fibonacci numbers. Determine the generating function for the Fibonacci sequence, defined
by F1 = F2 = 1 and Fn+2 = Fn+1 + Fn .
1
2n n −2
4. (a) Show that = (−4) .
n n
X 2n

−1/2
(b) Show that (1 − 4x) = xn .
n
n≥0
n
X 2k 2(n − k)
(c) Show that = 4n .
k n−k
k=0

5. Catalan numbers. The Catalan numbers Cn , satisfies C0 = 1 and

Cn = Cn−1 C0 + Cn−2 C1 + Cn−3 C2 + · · · + C0 Cn−1 , n ≥ 1.
Find the generating function for Cn , and use it to obtain an explicit formula for Cn .
6. Root of unity filter. Let n be a positive integer, and let ζ = e2kπi/n for some 0 < k < n.
For any polynomial F (x) = f0 + f1 x + f2 x2 + · · · (with finitely many terms), show that the
sum f0 + fn + f2n + · · · is given by
1
F (1) + F (ζ) + F (ζ 2 ) + · · · + F (ζ n−1 ) .

f0 + fn + f2n + · · · =
n
1
In this course, we will only work with ordinary generating function, since this is the type that occurs in olympiads
most often. Other useful generating functions include exponential generating functions, and Dirichlet series (used in
number theory).

3
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Problem Solving Session

August 1, 2007

1. Find the generating function for the sequence (an )∞

n=0 defined by a0 = 2, a1 = 0, a2 = −2 and
an+3 = 6an+2 − 11an+1 + 6an , n > 0.

2. Express
2 2 2 2
n n n n n
− + − · · · + (−1)
0 1 2 n
in closed form.
∞
X
3. Express in closed form: n2 xn .
n=1

4. (HMMT 2007) Let S denote the set of all triples (i, j, k) of positive integers where i+j+k = 17.
Compute X
ijk.
(i,j,k)∈S

∞
X Fn
5. Let Fn be the n-th Fibonacci number. Evaluate the infinite series .
4n
n=0

6. (Romania 2003) How many n-digit numbers, whose digits are in the set {2, 3, 7, 9}, are divis-
ible by 3?

7. Adrian tosses 2003 fair coins, Andrea and Claudia each toss 2004 fair coins, and Zachary tosses
2005 fair coins. Show that the two events are equally likely, and determine the probability.

(i) Claudia gets exactly one more head than Andrea does,
(ii) Adrian and Zachary get exactly the same number of heads.

8. (Britain 1994) An increasing sequence of integers is said to be alternating if it starts with an

odd term, the second term is even, the third term is odd, the fourth is even, and so on. The
empty sequence is counted as alternating.
Let An denote the number of alternating sequences involving only the integers 1, 2, . . . , n.
Find A(20).

4
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Lecture 12 : Generating Functions II—Partitions

Yufei Zhao
August 2, 2007

A partition of an integer n is a nonincreasing sequence of positive integers a1 ≥ a2 ≥ · · · ≥ ak such

that n = a1 + a2 + · · · + ak . We say that the ai ’s are the parts of the partition.
For example, p(5) = 7, as the 7 partitions of n = 5 are 5, 4 + 1, 3 + 2, 3 + 1 + 1, 2 + 2 + 1, 2 + 1 +
1 + 1, 1 + 1 + 1 + 1 + 1. By convention, p(0) is defined to be 1.

1. Let p(n) denote the number of partitions of n. Show that the generating function for p(n) is

X
n 1 1 1
p(n)x = ··· .
1−x 1 − x2 1 − x3
n≥0

2. Let n be a positive integer. Let f (n) denote the number of partitions of n with distinct parts,
and let g(n) denote the number of partitions of n with all parts being odd numbers. Prove
that f (n) = g(n).

3. (a) Show that every positive integer can be uniquely written as a sum of distinct powers of
2.
(b) Show that every integer can be uniquely written as a sum of the form ∞ k
P
k=0 ak 3 , where
ak ∈ {−1, 0, 1}.

4. (Putnam 1957) Let α(n) be the number of representations of a positive integer n as sum of
1’s and 2’s, taking order into account. For example, since

4 = 1 + 1 + 2 = 1 + 2 + 1 = 2 + 1 + 1 = 2 + 2 = 1 + 1 + 1 + 1,

we have α(4) = 5. Let β(n) be the number of representations of n that are sums of integers
greater than 1, again taking order into account. For example, since

6 = 4 + 2 = 2 + 4 = 3 + 3 = 2 + 2 + 2,

we have β(6) = 5. Show that α(n) = β(n + 2).

5. (IMO Shortlist 1998) Let a0 , a1 , a2 , . . . be an increasing sequence of nonnegative integers such

that every nonnegative integer can be expressed uniquely in the form ai + 2aj + 4ak , where
i, j, k are not necessarily distinct. Determine a1998 .

6. Euler’s Pentagonal Numbers Theorem. Show that

∞
Y ∞
X
n
(1 − x ) = 1 + (−1)k xk(3k−1)/2 + xk(3k+1)/2
n=1 k=1
= 1 − x − x2 + x5 + x7 − x12 − x15 + x22 + x26 − · · · .

5
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Problem Solving Session

August 2, 2007

1. Let n be a positive integer. In how many ways can we fill a bag with n fruits subject to the
following constraints?

• The number of apples must be even.

• The number of bananas must be a multiple of 5.
• There can be at most four oranges.
• There can be at most one pear.

2. Show that every integer

P has a unique base (−4) representation. That is, every integer can be
written uniquely as ∞ a
k=0 k (−4)k , where a ∈ {0, 1, 2, 3}.
k

3. Let n be a positive integer. Show that the number of partitions of n into parts which have at
most one of each distinct even part (e.g. 1 + 1 + 1 + 2 + 3 + 4) equals the number of partitions
of n in which each part can appear at most three times (e.g. 1 + 1 + 1 + 2 + 2 + 4 + 4 + 4).

4. Does there exist a subset S of the positive integers satisfying: for each positive integer n, the
number of partitions of n, where each part occurs at most twice, equals to the number of
partitions of n into parts that are elements of S?

5. Let n be a positive integer. Show that the number of partitions of n, where each part appears
at least twice, is equal to the number of partitions of n into parts all of which are divisible
by 2 or 3.

6. How many polynomials P with coefficients 0, 1, 2 or 3 have P (2) = n, where n is a given

positive integer?

6
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Lecture 13 : Generating Functions III

Yufei Zhao
August 3, 2007

1. Suppose the positive integers have been expressed as a disjoint union of arithmetic progres-
sions {ai + ndi }∞
n=0 , i = 1, 2, . . . , k.

1 1 1
(a) Show that + + ··· + = 1.
d1 d2 dk
a1 a2 ak k+1
(b) Show that + + ··· + = .
d1 d2 dk 2
(c) Show that di = dj for some i 6= j.

2. (Putnam 2000) Let S0 be a finite set of positive integers. We define finite sets S1 , S2 , . . . of
positive integers as follows: the integer a is in Sn+1 if and only if exactly one of a − 1 or a is in
Sn . Show that there exist infinitely many integers N for which SN = S0 ∪ {N + a : a ∈ S0 }.

3. (Putnam 2003) For a set S of nonnegative integers, let rS (n) denote the number of ordered
pairs (s1 , s2 ) such that s1 ∈ S, s2 ∈ S, s1 6= s2 and s1 + s2 = n. Is it possible to partition the
nonnegative integers into two sets A and B in such a way that rA (n) = rB (n) for all n?

4. Let (a1 , a2 , . . . , an ) and (b1 , b2 , . . . , bn ) be two different unordered n-tuples of integers such
that the sequences

a1 + a2 , a1 + a3 , . . . , an−1 + an (all pairwise sums ai + aj , 1 ≤ i < j ≤ n)

and
b1 + b2 , b1 + b3 , . . . , bn−1 + bn (all pairwise sums bi + bj , 1 ≤ i < j ≤ n)
coincide up to a permutation. Prove that n is a power of two.

5. (IMO 1995) Let p be an odd prime number. How many p-element subsets A of {1, 2, . . . , 2p}
are there such that the sums of its elements are divisible by p?

7
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Problem Solving Session

August 3, 2007

1. (Putnam 2001) Adrian has the coins C1 , C2 , . . . , Cn . For each k, Ck is biased so that, when
tossed, it has probability 1/(2k + 1) of showing heads. If the n coins are tossed, what is the
probability that the number of heads is odd? Express the answer as a rational function of n.

2. Each vertex of a regular polygon is colored with one of a finite number of colors so that the
points of the same color are the vertices of some new regular polygon. Prove that at least
two of the polygons obtained are congruent.

3. A standard die is labeled 1, 2, 3, 4, 5, 6 (one integer per face). When you roll two standard
dice, it is easy to compute the probability of the various sums. For example, the probability
of rolling two dice and getting a sum of 2 is just 1/36, while the probability of getting a 7 is
1/6.
Is it possible to construct a pair of nonstandard six-sided dice (possibly different from one
another) with positive integer labels that nevertheless are indistinguishable from a pair of
standard dice, if the sum of the dice is all that matters? For example, one of these nonstandard
dice may have the label 8 on one of its faces, and two 3’s. But the probability of rolling the
two and getting a sum of 2 is still 1/36, and the probability of getting a sum of 7 is still 1/6.

4. Let p be a prime. Compute the number of subsets T of {1, 2, . . . , p} such that p divides the
sum of the elements in T .

5. (Putnam 1997) Let am,n denote the coefficient of xn in the expansion of (1 + x + x2 )m . Prove
that for all k ≥ 0,
b2k/3c
X
0≤ (−1)i ak−i,i ≤ 1.
i=0

6. A finite sequence a1 , a2 , . . . , an of real numbers is called k-balanced if

am + am+k + am+2k + · · ·

is the same for any k = 1, 2, . . . , p. Suppose that the sequence a0 , a1 , . . . , a49 is k-balanced
for k = 3, 5, 7, 11, 13, and 17. Prove that a0 = a1 = · · · = a49 = 0.

8
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Lecture 14 : When Worlds Collide—Algebraic Combinatorics

Yufei Zhao
August 4, 2007

1. (St. Petersburg) Students in a school go for ice cream in groups of at least two. After k > 1
groups have gone, every two students have gone together exactly once. Prove that the number
of students in the school is at most k.

2. Oddtown and Eventown. In a certain town with n citizens, a number of clubs are set up.
No two clubs have exactly the set of members. Determine the maximum number of clubs
that can be formed under each of the following constraints:

(a) The size of every club is odd, and every pair of clubs share an even number of members.
(b) The size of every club is even, and every pair of clubs share an even number of members.

3. (a) (China West 2002) Let A1 , A2 , . . . , An+1 be non-empty subsets of {1, 2, . . . , n}. Prove
that there exists nonempty disjoint subsets I, J ⊂ {1, 2, . . . , n + 1} such that
[ [
Ak = Ak .
k∈I k∈J

(b) (Lindstrom) Let A1 , A2 , . . . , An+2 be non-empty subsets of {1, 2, . . . , n}. Prove that
there exists nonempty disjoint subsets I, J ⊂ {1, 2, . . . , n + 2} such that
[ [ \ \
Ak = Ak , and Ak = Ak .
k∈I k∈J k∈I k∈J

4. (Russia 2001) A contest with n question was taken by m contestants. Each question was
worth a certain (positive) number of points, and no partial credits were given. After all the
papers have been graded, it was noticed that by reassigning the scores of the questions, any
desired ranking of the contestants could be achieved. What is the largest possible value of
m?

5. (Crux 3037) There are 2007 senators in a senate. Each senator has enemies within the senate.
Prove that there is a non-empty subset K of senators such that for every senator in the senate,
the number of enemies of that senator in the set K is an even number.

6. Let a1 , a2 , . . . , a2n+1 be real numbers, such that for any 1 ≤ i ≤ 2n + 1, we can remove
ai and separate the remaining 2n numbers into two groups with equal sums. Show that
a1 = a2 = · · · = a2n+1 .

7. Odd vertex cover. (Iran TST 1996, Germany TST 2004) Let G be a finite simple graph,
and there is a light bulb at each vertex of G. Initially, all the lights are off. Each step we
are allowed to chose a vertex and toggle the light at that vertex as well as all its neighbors’.
Show that we can get all the lights to be on at the same time.

8. (Graham–Pollak) Show that the complete graph with n vertices, Kn , cannot be covered by
fewer than n − 1 complete bipartite graphs so that each edge of Kn is covered exactly once.

9
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Evaluation Test 3
August 4, 2007

Instructions: Read the problems carefully. Write your solutions neatly and concisely, but make
sure to justify all your steps. Start each new solution on a new page and write your name and the
problem number on every page.

1. Short answer questions: correct answers are worth full points in this problem.

(a) [5] Let α, β be complex numbers. Determine the generating function for the sequence
(an )∞
n=0 defined by a0 = α, a1 = β, and an = an−1 + an−2 , n ≥ 2.
(b) [5] Determine the generating function for the sequence (bn )∞
n=0 defined by b0 = 0, and
bn = 2bn−1 + n for n ≥ 1.

2. [10] Let n be a positive integer. Show that the number of partitions of n into odd parts
greater than 1 is equal to the number of partitions of n into unequal parts none of which is
a power of two.

3. [10] Seven students take a mathematical exam. Every problem was solved by at most 3
students. For every pair of students, there is at least one problem that they both solved.
Determine, with proof, the minimum number of problems on this exam.

4. [10] Find the number of subsets of {1, 2, . . . , 2007}, the sum of whose elements is divisible by
17.

10
AwesomeMath 2007 Track 1 — Combinatorics Week 3

Evaluation Test 3
Solutions
August 4, 2007

1. Short answer questions: correct answers are worth full points in this problem.

(a) [5] Let α, β be complex numbers. Determine the generating function for the sequence
(an )∞
n=0 defined by a0 = α, a1 = β, and an = an−1 + an−2 , n ≥ 2.

Solution: Let A(x) = n≥0 an xn be the generating function. Then

P

X X
A(x) = an xn = α + βx + an xn
n≥0 n≥2
X X
= α + βx + an−1 xn + an−2 xn
n≥2 n≥2
2
= α + βx + x(A(x) − α) + x A(x).

α + (β − α)x
Solving for A(x) gives A(x) = .
1 − x − x2
(b) [5] Determine the generating function for the sequence (bn )∞
n=0 defined by b0 = 0, and
bn = 2bn−1 + n for n ≥ 1.
Solution: Let B(x) = n≥0 bn xn be the generating function. Then
P

X X X X x
B(x) = bn xn = bn xn = 2bn−1 xn + nxn = 2xB(x) + .
(1 − x)2
n≥0 n≥1 n≥1 n≥1

x
Solving for B(x) gives B(x) = .
(1 − x)2 (1 − 2x)
2. [10] Let n be a positive integer. Show that the number of partitions of n into odd parts
greater than 1 is equal to the number of partitions of n into unequal parts none of which is
a power of two.

Solution: Let an denote the number of partitions of n into odd parts greater than 1. The
generating function for (an ) is
1
A(x) = .
(1 − x3 )(1 − x5 )(1 − x7 ) · · ·

Also, let bn denote the number of partitions of n into unequal parts none of which is a power
of two. The generating function for (bn ) is

(1 + x)(1 + x2 )(1 + x3 )(1 + x4 ) · · ·

B(x) = .
(1 + x)(1 + x2 )(1 + x4 )(1 + x8 ) · · ·

Note that we have

(1 − x2 )(1 − x4 )(1 − x8 ) · · · 1
(1 + x)(1 + x2 )(1 + x4 )(1 + x8 ) · · · = 2 4
= ,
(1 − x)(1 − x )(1 − x ) · · · 1−x

11
AwesomeMath 2007 Track 1 — Combinatorics Week 3

and
(1 − x2 )(1 − x4 )(1 − x6 ) · · ·
(1 + x)(1 + x2 )(1 + x3 )(1 + x4 ) · · · =
(1 − x)(1 − x2 )(1 − x3 ) · · ·
1
= .
(1 − x)(1 − x3 )(1 − x5 )(1 − x7 ) · · ·

Combining the two, we find that A(x) = B(x). Therefore, an = bn for all n, as desired.

3. [10] Seven students take a mathematical exam. Every problem was solved by at most 3
students. For every pair of students, there is at least one problem that they both solved.
Determine, with proof, the minimum number of problems on this exam.

Solution: Assume that there are n problems. Let T denote the number of pairs (P, S),
where P is a problem, and S is a pair of students who both solved P . On one hand, since
every problem was solved at at most 3 students, we have T ≤ 32 n = 3n. On the other hand,

since there is such a pair (P, S) for every pair of students, we have T ≥ 72 = 21. Thus,

3n ≥ 21, so n ≥ 7.
Let us show that n = 7 can be attained. Label the problems 1, 2, 3, 4, 5, 6, 7, and the let
each of the following set denote the set of problems that an individual a student has solved:

{1, 2, 3}, {1, 4, 7}, {1, 5, 6}, {2, 5, 7}, {2, 4, 6}, {3, 4, 5}, {3, 6, 7}.

Therefore, the minimum number of problems on this exam is 7.

Remarks: Although the construction can be found by trial and error, it can be motivated by
the Fano plane, which is depicted below. Each vertex represents a problem, and each line
(including the center circle) represents a student who solved the problems that the line passes
through.

6 2
7

5 4 3

4. [10] Find the number of subsets of {1, 2, . . . , 2007} (including the empty set) the sum of whose
elements is divisible by 17.

Solution: Consider the polynomial

f (x) = (1 + x)(1 + x2 )(1 + x3 ) · · · (1 + x2007 ).

Every subset {a1 , a2 , . . . , ak } ⊂ {1, 2, . . . , 2007} corresponds to a term in the expansion of

f (x), namely xa1 +a2 +···+ak . Therefore, for each m, the coefficient of xm is f (x) equals to the
number of subsets of {1, 2, . . . , 2007} the sum of whose elements is m. Therefore, we need the

12
AwesomeMath 2007 Track 1 — Combinatorics Week 3

find the sum of the coefficients of x0 , x17 , x34 , . . . , x2006 . This can be done using the roots of
unity. Let ζ = e2πi/17 . Then, the answer equals to
1
f (1) + f (ζ) + f (ζ 2 ) + · · · + f (ζ 16 ) .

17
We must evaluate this expression.
First, notice that since the zeros of x17 − 1 are 1, ζ, ζ 2 , . . . , ζ 16 , so

x17 − 1 = (x − 1)(x − ζ)(x − ζ 2 ) · · · (x − ζ 16 ).

Setting x = −1 yields the identity

(1 + 1)(1 + ζ)(1 + ζ 2 ) · · · (1 + ζ 16 ) = 2.

Note that for 1 ≤ k ≤ 16, {0, k, 2k, 3k, . . . , 16k} is a permutation of {0, 1, 2, 3, . . . , 16} as
residues in mod 17. It follows that

(1 + 1)(1 + ζ k )(1 + ζ 2k ) · · · (1 + ζ 16k ) = 2.

Consequently, since 2007 = 17 · 118 + 1, we have

f (ζ k ) = (1 + ζ k )(1 + ζ 2k ) · · · (1 + ζ 2007k )
h i118
= (1 + 1)(1 + ζ k )(1 + ζ 2k ) · · · (1 + ζ 16k ) (1 + ζ k )
= 2118 (1 + ζ k ).

Also, f (1) = 22007 . Therefore,

1
f (1) + f (ζ) + f (ζ 2 ) + · · · + f (ζ 16 )

17
1 2007
+ 2118 (16 + ζ + ζ 2 + ζ 3 + · · · + ζ 16 )

= 2
17
1
= 2118 21889 + 15 .

17
2118 (21889 +15)
Thus, the answer is 17 .

13
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

Lemmas in Euclidean Geometry1

Yufei Zhao
[email protected]

1. Construction of the symmedian.

Let ABC be a triangle and Γ its circumcircle. Let the tangent to Γ at B and C meet at D. Then
AD coincides with a symmedian of △ABC. (The symmedian is the reflection of the median
across the angle bisector, all through the same vertex.)
A
A A

B C
B M F C E
M' C B
P

D
D Q D

We give three proofs. The first proof is a straightforward computation using Sine Law. The
second proof uses similar triangles. The third proof uses projective geometry.

First proof. Let the reflection of AD across the angle bisector of ∠BAC meet BC at M ′ . Then
∠BAM ′
BM ′ AM ′ sin
sin ∠ABC sin ∠BAM ′ sin ∠ABD sin ∠CAD sin ∠ABD CD AD
= sin ∠CAM ′ = = = =1
MC
′
AM ′ sin ∠ACB sin ∠ACD sin ∠CAM ′ sin ∠ACD sin ∠BAD AD BD

Therefore, AM ′ is the median, and thus AD is the symmedian.

Second proof. Let O be the circumcenter of ABC and let ω be the circle centered at D with radius
DB. Let lines AB and AC meet ω at P and Q, respectively. Since ∠P BQ = ∠BQC + ∠BAC =
1 ◦
2 (∠BDC + ∠DOC) = 90 , we see that P Q is a diameter of ω and hence passes through D.
Since ∠ABC = ∠AQP and ∠ACB = ∠AP Q, we see that triangles ABC and AQP are similar.
If M is the midpoint of BC, noting that D is the midpoint of QP , the similarity implies that
∠BAM = ∠QAD, from which the result follows.

Third proof. Let the tangent of Γ at A meet line BC at E. Then E is the pole of AD (since the
polar of A is AE and the pole of D is BC). Let BC meet AD at F . Then point B, C, E, F are
harmonic. This means that line AB, AC, AE, AF are harmonic. Consider the reflections of the
four line across the angle bisector of ∠BAC. Their images must be harmonic too. It’s easy to
check that AE maps onto a line parallel to BC. Since BC must meet these four lines at harmonic
points, it follows that the reflection of AF must pass through the midpoint of BC. Therefore,
AF is a symmedian.
1
Updated July 20, 2007

1
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

Related problems:

(i) (Poland 2000) Let ABC be a triangle with AC = BC, and P a point inside the triangle such
that ∠P AB = ∠P BC. If M is the midpoint of AB, then show that ∠AP M +∠BP C = 180◦ .
(ii) (IMO Shortlist 2003) Three distinct points A, B, C are fixed on a line in this order. Let Γ
be a circle passing through A and C whose center does not lie on the line AC. Denote by
P the intersection of the tangents to Γ at A and C. Suppose Γ meets the segment P B at
Q. Prove that the intersection of the bisector of ∠AQC and the line AC does not depend
on the choice of Γ.
(iii) (Vietnam TST 2001) In the plane, two circles intersect at A and B, and a common tangent
intersects the circles at P and Q. Let the tangents at P and Q to the circumcircle of triangle
AP Q intersect at S, and let H be the reflection of B across the line P Q. Prove that the
points A, S, and H are collinear.
(iv) (USA TST 2007) Triangle ABC is inscribed in circle ω. The tangent lines to ω at B and C
meet at T . Point S lies on ray BC such that AS ⊥ AT . Points B1 and C1 lies on ray ST
(with C1 in between B1 and S) such that B1 T = BT = C1 T . Prove that triangles ABC
and AB1 C1 are similar to each other.
(v) (USA 2008) 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.

2. Diameter of the incircle.

A

B D F C

Let the incircle of triangle ABC touch side BC at D, and let DE be a diameter of the circle. If
line AE meets BC at F , then BD = CF .

Proof. Consider the dilation with center A that carries the incircle to an excircle. The diameter
DE of the incircle must be mapped to the diameter of the excircle that is perpendicular to BC.
It follows that E must get mapped to the point of tangency between the excircle and BC. Since
the image of E must lie on the line AE, it must be F . That is, the excircle is tangent to BC at
F . Then, it follows easily that BD = CF .

Related problems:

(i) (IMO Shortlist 2005) In a triangle ABC satisfying AB +BC = 3AC the incircle has centre I
and touches the sides AB and BC at D and E, respectively. Let K and L be the symmetric
points of D and E with respect to I. Prove that the quadrilateral ACKL is cyclic.

2
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

(ii) (IMO 1992) In the plane let C be a circle, ℓ a line tangent to the circle C, and M a point on
ℓ. Find the locus of all points P with the following property: there exists two points Q, R
on ℓ such that M is the midpoint of QR and C is the inscribed circle of triangle P QR.
(iii) (USAMO 1999) 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.
(iv) (USAMO 2001) Let ABC be a triangle and let ω be its incircle. Denote by D1 and E1 the
points where ω is tangent to sides BC and AC, respectively. Denote by D2 and E2 the points
on sides BC and AC, respectively, such that CD2 = BD1 and CE2 = AE1 , and denote by
P the point of intersection of segments AD2 and BE2 . Circle ω intersects segment AD2 at
two points, the closer of which to the vertex A is denoted by Q. Prove that AQ = D2 P .
(v) (Tournament of Towns 2003 Fall) Triangle ABC has orthocenter H, incenter I and circum-
center O. Let K be the point where the incircle touches BC. If IO is parallel to BC, then
prove that AO is parallel to HK.
3. Dude, where’s my spiral center?
Let AB and CD be two segments, and let lines AC and BD meet at X. Let the circumcircles of
ABX and CDX meet again at O. Then O is the center of the spiral similarity that carries AB
to CD.
C
B
X

Proof. Since ABOX and CDXO are cyclic, we have ∠OBD = ∠OAC and ∠OCA = ∠ODB. It
follows that triangles AOC and BOD are similar. The result is immediate.

Remember that spiral similarities always come in pairs: if there is a spiral similarity that carries
AB to CD, then there is one that carries AC to BD.
Related problems:
(i) (IMO Shortlist 2006) Let ABCDE be a convex pentagon such that
∠BAC = ∠CAD = ∠DAE and ∠CBA = ∠DCA = ∠EDA.
Diagonals BD and CE meet at P . Prove that line AP bisects side CD.
(ii) (China 1992) Convex quadrilateral ABCD is inscribed in circle ω with center O. Diagonals
AC and BD meet at P . The circumcircles of triangles ABP and CDP meet at P and Q.
Assume that points O, P , and Q are distinct. Prove that ∠OQP = 90◦ .
(iii) Let ABCD be a quadrilateral. Let diagonals AC and BD meet at P . Let O1 and O2 be
the circumcenters of AP D and BP C. Let M , N and O be the midpoints of AC, BD and
O1 O2 . Show that O is the circumcenter of M P N .

3
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

(iv) (USAMO 2006) Let ABCD be a quadrilateral, and let E and F be points on sides AD and
BC, respectively, such that AE/ED = BF/F C. Ray F E meets rays BA and CD at S and
T , respectively. Prove that the circumcircles of triangles SAE, SBF , T CF , and T DE pass
through a common point.
(v) (IMO 2005) Let ABCD be a given convex quadrilateral with sides BC and AD equal in
length and not parallel. Let E and F be interior points of the sides BC and AD respectively
such that BE = DF . The lines AC and BD meet at P , the lines BD and EF meet at Q,
the lines EF and AC meet at R. Consider all the triangles P QR as E and F vary. Show
that the circumcircles of these triangles have a common point other than P .
(vi) (IMO Shortlist 2002) 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.
(vii) (USA TST 2006) In acute triangle ABC, segments AD, BE, and CF are its altitudes, and
H is its orthocenter. Circle ω, centered at O, passes through A and H and intersects sides
AB and AC again at Q and P (other than A), respectively. The circumcircle of triangle
OP Q is tangent to segment BC at R. Prove that CR/BR = ED/F D.
(viii) (IMO Shortlist 2006) Points A1 , B1 and C1 are chosen on sides BC, CA, and AB of a triangle
ABC, respectively. The circumcircles of triangles AB1 C1 , BC1 A1 , and CA1 B1 intersect the
circumcircle of triangle ABC again at points A2 , B2 , and C2 , respectively (A2 6= A, B2 6= B,
and C2 6= C). Points A3 , B3 , and C3 are symmetric to A1 , B1 , C1 with respect to the
midpoints of sides BC, CA, and AB, respectively. Prove that triangles A2 B2 C2 and A3 B3 C3
are similar.
4. Arc midpoints are equidistant to vertices and in/excenters
Let ABC be a triangle, I its incenter, and IA , IB , IC its excenters. On the circumcircle of ABC,
let M be the midpoint of the arc BC not containing A and let N be the midpoint of the arc BC
containing A. Then M B = M C = M I = M IA and N B = N C = N IB = N IC .

IC
N
A

IB

B C

IA

Proof. Straightforward angle-chasing (do it yourself!). Another perspective is to consider the

circumcircle of ABC as the nine-point-circle of IA IB IC .

4
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

Related problems:

(i) (APMO 2007) Let ABC be an acute angled triangle with ∠BAC = 60◦ and AB > AC. Let I
be the incenter, and H the orthocenter of the triangle ABC. Prove that 2∠AHI = 3∠ABC.
(ii) (IMO 2006) Let ABC be a triangle with incentre I. A point P in the interior of the triangle
satisfies ∠P BA + ∠P CA = ∠P BC + ∠P CB. Show that AP ≥ AI, and that equality holds
if and only if P = I.
(iii) (Romanian TST 1996) Let ABCD be a cyclic quadrilateral and let M be the set of incenters
and excenters of the triangles BCD, CDA, DAB, ABC (16 points in total). Prove that there
are two sets K and L of four parallel lines each, such that every line in K ∪ L contains exactly
four points of M.

5. I is the midpoint of the touch-chord of the mixtilinear incircles

Let ABC be a triangle and I its incenter. Let Γ be the circle tangent to sides AB, AC, as well
as the circumcircle of ABC. Let Γ touch AB and AC at X and Y , respectively. Then I is the
midpoint of XY .
A A

P
Y Y
I
I

X C X C

T
B B

Proof. Let the point of tangency between the two circles be T . Extend T X and T Y to meet the
circumcircle of ABC again at P and Q respectively. Note that P and Q are the midpoint of the
arcs AB and AC. Apply Pascal’s theorem to BACP T Q and we see that X, I, Y are collinear.
Since I lies on the angle bisector of ∠XAY and AX = AY , I must be the midpoint of XY .

Related problems:

(i) (IMO 1978) In triangle ABC, AB = AC. A circle is tangent internally to the circumcircle
of triangle ABC and also to sides AB, AC at P, Q, respectively. Prove that the midpoint of
segment P Q is the center of the incircle of triangle ABC.
(ii) Let ABC be a triangle. Circle ω is tangent to AB and AC, and internally tangent to the
circumcircle of triangle ABC. The circumcircle and ω are tangent at P . Let I be the
incircle of triangle ABC. Line P I meets the circumcircle of ABC at P and Q. Prove that
BQ = CQ.

6. More curvilinear incircles.

(A generalization of the previous lemma) Let ABC be a triangle, I its incenter and D a point on
BC. Consider the circle that is tangent to the circumcircle of ABC but is also tangent to DC,
DA at E, F respectively. Then E, F and I are collinear.

5
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

A A

F F/F' I
I

B E C E C
D B D

Proof. There is a “computational” proof using Casey’s theorem2 and transversal theorem3 . You
can try to work that out yourself. Here, we show a clever but difficult synthetic proof (commu-
nicated to me via Oleg Golberg).
Denote Ω the circumcircle of ABC and Γ the circle tangent tangent to the circumcircle of ABC
and lines DC, DA. Let Ω and Γ touch at K. Let M be the midpoint of arc BC d on Ω not
containing K. Then K, E, M are collinear (think: dilation with center K carrying Γ to Ω). Also,
A, I, M are collinear, and M I = M C.
Let line EI meet Γ again at F ′ . It suffices to show that AF ′ is tangent to Γ.
Note that ∠KF ′ E is subtended by KE d in Γ and ∠KAM is subtended by KM d in Ω. Since KE
d
d
and KM are homothetic with center K, we have ∠KF E = ∠KAM , implying that A, K, I , F ′
′ ′

are concyclic.
We have ∠BCM = ∠CBM = ∠CKM . So △M CE ∼ △M KC. Hence M C 2 = M E · M K.
Since M C = M I, we have M I 2 = M E · M K, implying that △M IE ∼ △M KI. Therefore,
∠KEI = ∠AIK = ∠AF ′ K (since A, K, I ′ , F ′ are concyclic). Therefore, AF ′ is tangent to Ω and
the proof is complete.

Related problems:

(i) (Bulgaria 2005) Consider two circles k1 , k2 touching externally at point T . A line touches
k2 at point X and intersects k1 at points A and B. Let S be the second intersection point
of k1 with the line XT . On the arc TcS not containing A and B is chosen a point C . Let
CY be the tangent line to k2 with Y ∈ k2 , such that the segment CY does not intersect
the segment ST . If I = XY ∩ SC . Prove that:
(a) the points C, T, Y, I are concyclic.
2
Casey’s theorem, also known as Generalized Ptolemy Theorem, states that if there are four circles Γ1 , Γ2 , Γ3 , Γ4
(could be degenerated into a point) all touching a circle Γ such that their tangency points follow that order around the
circle, then
t12 t34 + t23 t14 = t13 t24 ,
where t12 is the length of the common tangent between Γi and Γj (if Γi and Γj on the same side of Γ, then take their
common external tangent, else take their common internal tangent.) I think the converse is also true—if both equations
hold, then there is some circle tangent to all four circles.
3
The transversal theorem is a criterion for collinearity. It states that if A, B, C are three collinear points, and P is
a point not on the line ABC, and A′ , B ′ , C ′ are arbitrary points on lines P A, P B, P C respectively, then A′ , B ′ , C ′ are
collinear if and only if
AP BP CP
BC · ′ + CA · ′ + AB · ′ = 0,
AP BP CP
where the lengths are directed. In my opinion, it’s much easier to remember the proof than to memorize this huge formula.
The simplest derivation is based on relationships between the areas of [P AB], [P A′ B ′ ], etc.

6
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

(b) I is the excenter of triangle ABC with respect to the side BC.
(ii) (Sawayama-Thébault4 ) Let ABC be a triangle with incenter I. Let D a point on side BC.
Let P be the center of the circle that touches segments AD, DC, and the circumcircle
of ABC, and let Q be the center of the circle that touches segments AD, BD, and the
circumcircle of ABC. Show that P, Q, I are collinear.
(iii) Let P be a quadrilateral inscribed in a circle Ω, and let Q be the quadrilateral formed by
the centers of the fourcircles internally touching O and each of the two diagonals of P . Show
that the incenters of the four triangles having for sides the sides and diagonals of P form a
rectangle R inscribed in Q.
(iv) (Romania 1997) Let ABC be a triangle with circumcircle Ω, and D a point on the side BC.
Show that the circle tangent to Ω, AD and BD, and the circle tangent to Ω, AD and DC,
are tangent to each other if and only if ∠BAD = ∠CAD.
(v) (Romania TST 2006) Let ABC be an acute triangle with AB 6= AC. Let D be the foot
of the altitude from A and ω the circumcircle of the triangle. Let ω1 be the circle tangent
to AD, BD and ω. Let ω2 be the circle tangent to AD, CD and ω. Let ℓ be the interior
common tangent to both ω1 and ω2 , different from CD. Prove that ℓ passes through the
midpoint of BC if and only if 2BC = AB + AC.
(vi) (AMM 10368) For each point O on diameter AB of a circle, perform the following construc-
tion. Let the perpendicular to AB at O meet the circle at point P . Inscribe circles in the
figures bounded by the circle and the lines AB and OP . Let R and S be the points at which
the two incircles to the curvilinear triangles AOP and BOP are tangent to the diameter
AB. Show that ∠RP S is independent of the position of O.

7. Concurrent lines from the incircle.

Let the incircle of ABC touch sides BC, CA, AB at D, E, F respectively. Let I be the incenter
of ABC and M be the midpoint of BC. Then the lines EF, DI and AM are concurrent.
A A

E N E
P Q
F
I F
I

B C
B D M C D M

Proof. Let lines DI and EF meet at N . Construct a line through N parallel to BC, and let it
meet sides AB and AC at P and Q, respectively. We need to show that A, N, M are collinear,
so it suffices to show that N is the midpoint of P Q. We present two ways to finish this off, one
using Simson’s line, and the other using spiral similarities.
Simson line method: Consider the triangle AP Q. The projections of the point I onto the three
sides of AP Q are D, N, F , which are collinear, I must lie on the circumcircle of AP Q by Simson’s
theorem. But since AI is an angle bisector, P I = QI, thus P N = QN .
4
A bit of history: this problem was posed by French geometer Victor Thébault (1882–1960) in the American
Mathematical Monthly in 1938 (Problem 2887, 45 (1938) 482–483) and it remained unsolved until 1973. How-
ever, in 2003, Jean-Louis Ayme discovered that this problem was independently proposed and solved by instruc-
tor Y. Sawayama of the Central Military School of Tokyo in 1905! For more discussion, see Ayme’s paper at
http://forumgeom.fau.edu/FG2003volume3/FG200325.pdf

7
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

Spiral similarity method: Note that P, N, I, F are concyclic, so ∠EF I = ∠QP I. Similarly,
∠P QI = ∠F EI. So triangles P IQ and F IE are similar. Since F I = EI, we have P I = QI, and
thus P N = QN . (c.f. Lemma 3)

Related problems:
(i) (China 1999) In triangle ABC, AB 6= AC. Let D be the midpoint of side BC, and let E be
a point on median AD. Let F be the foot of perpendicular from E to side BC, and let P
be a point on segment EF . Let M and N be the feet of perpendiculars from P to sides AB
and AC, respectively. Prove that M, E, and N are collinear if and only if ∠BAP = ∠P AC.
(ii) (IMO Shortlist 2005) The median AM of a triangle ABC intersects its incircle ω at K and
L. The lines through K and L parallel to BC intersect ω again at X and Y . The lines AX
and AY intersect BC at P and Q. Prove that BP = CQ.
8. More circles around the incircle.
Let I be the incenter of triangle ABC, and let its incircle touch sides BC, AC, AB at D, E and
F , respectively. Let line CI meet EF at T . Then T, I, D, B, F are concyclic. Consequent results
include: ∠BT C = 90◦ , and T lies on the line connecting the midpoints of AB and BC.
An easier way to remember the third part of the lemma is: for a triangle ABC, draw a midline,
an angle bisector, and a touch-chord, each generated from different vertex, then the three lines
are concurrent.
A

E
T
F I

B D M C

Proof. Showing that I, T, E, B are concyclic is simply angle chasing (e.g. show that ∠BIC =
∠AEF ). The second part follows from ∠BT C = ∠BT I = ∠BF I = 90◦ . For the third part, note
that if M is the midpoint of BC, then M is the midpoint of an hypotenuse of the right triangle
BT C. So M T = M C. Then ∠M T C = ∠M CT = ∠ACT , so M T is parallel to AC, and so M T
is a midline of the triangle.

Related problems:
(i) Let ABC be an acute triangle whose incircle touches sides AC and AB at E and F , respec-
tively. Let the angle bisectors of ∠ABC and ∠ACB meet EF at X and Y , respectively,
and let the midpoint of BC be Z. Show that XY Z is equilateral if and only if ∠A = 60◦ .
(ii) (IMO Shortlist 2004) For a given triangle ABC, let X be a variable point on the line BC
such that C lies between B and X and the incircles of the triangles ABX and ACX intersect
at two distinct points P and Q. Prove that the line P Q passes through a point independent
of X.
(iii) Let points A and B lie on the circle Γ, and let C be a point inside the circle. Suppose that
ω is a circle tangent to segments AC, BC and Γ. Let ω touch AC and Γ at P and Q. Show
that the circumcircle of AP Q passes through the incenter of ABC.

8
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

9. Reflections of the orthocenter lie on the circumcircle.

Let H be the orthocenter of triangle ABC. Let the reflection of H across the BC be X and the
reflection of H across the midpoint of BC be Y . Then X and Y both lie on the circumcircle of
ABC. Moreover, AY is a diameter of the circumcircle.
A

B C

X Y

Proof. Trivial. Angle chasing.

Related problems:

(i) Prove the existence of the nine-point circle. (Given a triangle, the nine-point circle is the
circle that passes through the three midpoints of sides, the three feet of altitudes, and the
three midpoints between the orthocenter and the vertices).
(ii) Let ABC be a triangle, and P a point on its circumcircle. Show that the reflections of P
across the three sides of ABC lie on a lie that passes through the orthocenter of ABC.
(iii) (IMO Shortlist 2005) Let ABC be an acute-angled triangle with AB 6= AC, let H be its
orthocentre and M the midpoint of BC. Points D on AB and E on AC are such that
AE = AD and D, H, E are collinear. Prove that HM is orthogonal to the common chord
of the circumcircles of triangles ABC and ADE.
(iv) (USA TST 2005) Let A1 A2 A3 be an acute triangle, and let O and H be its circumcenter and
orthocenter, respectively. For 1 ≤ i ≤ 3, points Pi and Qi lie on lines OAi and Ai+1 Ai+2
(where Ai+3 = Ai ), respectively, such that OPi HQi is a parallelogram. Prove that

OQ1 OQ2 OQ3

+ + ≥ 3.
OP1 OP2 OP3

(v) (China TST quizzes 2006) Let ω be the circumcircle of triangle ABC, and let P be a point
inside the triangle. Rays AP, BP, CP meet ω at A1 , B1 , C1 , respectively. Let A2 , B2 , C2 be
the images of A1 , B1 , C1 under reflection about the midpoints of BC, CA, AB, respectively.
Show that the orthocenter of ABC lies on the circumcircle of A2 B2 C2 .

10. O and H are isogonal conjugates.

Let ABC be a triangle, with circumcenter O, orthocenter H, and incenter I. Then AI is the
angle bisector of ∠HAO.

Proof. Trivial.

Related problems:

9
IMO Training 2007 Lemmas in Euclidean Geometry Yufei Zhao

(i) (Crux) Points O and H are the circumcenter and orthocenter of acute triangle ABC, re-
spectively. The perpendicular bisector of segment AH meets sides AB and AC at D and
E, respectively. Prove that ∠DOA = ∠EOA.
(ii) Show that IH = IO if and only if one of ∠A, ∠B, ∠C is 60◦ .

10
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

Cyclic Quadrilaterals — The Big Picture

Yufei Zhao
[email protected]

An important skill of an olympiad geometer is being able to recognize known configurations.

Indeed, many geometry problems are built on a few common themes. In this lecture, we will explore
one such configuration.

1 What Do These Problems Have in Common?

1. (IMO 1985) A circle with center O passes through the vertices A and C of triangle ABC
and intersects segments AB and BC again at distinct points K and N , respectively. The
circumcircles of triangles ABC and KBN intersects at exactly two distinct points B and M .
Prove that ∠OM B = 90◦ .

B
M

N
K

O
A C

2. (Russia 1995; Romanian TST 1996; Iran 1997) Consider a circle with diameter AB and center
O, and let C and D be two points on this circle. The line CD meets the line AB at a point
M satisfying M B < M A and M D < M C. Let K be the point of intersection (different from
O) of the circumcircles of triangles AOC and DOB. Show that ∠M KO = 90◦ .

C
D
K

M
A O B

3. (USA TST 2007) Triangle ABC is inscribed in circle ω. The tangent lines to ω at B and C
meet at T . Point S lies on ray BC such that AS ⊥ AT . Points B1 and C1 lies on ray ST
(with C1 in between B1 and S) such that B1 T = BT = C1 T . Prove that triangles ABC and
AB1 C1 are similar to each other.

1
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

B S
C

C1
B1 T

Although these geometric configurations may seem very different at first sight, they are actually
very related. In fact, they are all just bits and pieces of one big diagram!

2 One Big Diagram

O
B
P
D
A

R
Q
M

Figure 1: The big picture.

In this lecture, we will try to understand the features of Figure 1. There are a lot of things
going on in this diagram, and it can be frightening to look at. Don’t worry, we will go through it

2
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

bits and pieces at time. In the process, we will discuss some geometric techniques that are useful
in other places as well.
(Can you tell where to find each of the problems in Section 1 in Figure 1? You probably can’t
at this point, but hopefully you will be able to by the end of this lecture.)

3 Miquel’s Theorem and Miquel Point

Fact 1 (Miquel’s Theorem). Let ABC be a triangle, and let X, Y, Z be points on lines BC, CA, AB,
respectively. Assume that the six points A, B, C, X, Y, Z are all distinct. Then the circumcircles of
AY Z, BZX, CXY pass through a common point.

Y X

A
Z B

Figure 2: Diagram for Fact 1 (Miquel’s Theorem).

Exercise 1. Prove Fact 1. (This is very easy. Just chase1 a few angles.)

Fact 2 (Miquel point). Let `1 , `2 , `3 , `4 be four lines in the plane, no two parallel. Let Cijk denote
the circumcircle of the triangle formed by the lines `i , `j , `k (these circles are called Miquel circles).
Then C123 , C124 , C134 , C234 pass through a common point (called the Miquel point).

Exercise 2. Prove Fact 2. (Hint: apply Theorem 1)

We want to specialize to the case of a cyclic quadrilateral.

C
B
D
A

Q R

Figure 3: Miquel point for a cyclic quadrilateral ABCD.

1
If you are bothered by configuration and orientation issues (and you should be!), use directed angles.

3
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

Fact 3. Let ABCD be a quadrilateral. Let lines AB and CD meet at Q, and lines DA and CB
meet at R. Then the Miquel point of ABCD (i.e., the second intersection point of the circumcircles
of ADQ and ABR) lies on the line QR if and only if ABCD is cyclic.

Exercise 3. Prove Fact 3. (This is again just easy angle chasing.)

4 An Important Result about Spiral Similarities

A spiral similarity 2 about a point O (known as the center of the spiral similarity) is a composition
of a rotation and a dilation, both centered at O.

Figure 4: An example of a spiral similarity.

For instance, in the complex plane, if O = 0, then spiral similarities are described by multi-
plication by a nonzero complex number. That is, spiral similarities have the form z 7→ αz, where
α ∈ C \ {0}. Here |α| is the dilation factor, and arg α is the angle of rotation. It is easy to
deduce from here that if the center of the spiral similarity is some other point, say z0 , then the
transformation is given by z 7→ z0 + α(z − z0 ) (why?).

Fact 4. Let A, B, C, D be four distinct point in the plane, such that ABCD is not a parallelogram.
Then there exists a unique spiral similarity that sends A to B, and C to D.

Proof. Let a, b, c, d be the corresponding complex numbers for the points A, B, C, D. We know that
a spiral similarity has the form T(z) = z0 + α(z − z0 ), where z0 is the center of the spiral similarity,
and α is data on the rotation and dilation. So we would like to find α and z0 such that T(a) = c
and T(b) = d. This amount to solving the system

z0 + α(a − z0 ) = c, z0 + α(b − z0 ) = d.

Solving it, we see that the unique solution is

c−d ad − bc
α= , z0 = .
a−b a−b−c+d
Since ABCD is not a parallelogram, we see that a − b − c + d 6= 0, so that this is the unique solution
to the system. Hence there exists a unique spiral similarity that carries A to B and C to D.
2
If you want to impress your friends with your mathematical vocabulary, a spiral similarity is sometimes called a
similitude, and a dilation is sometimes called a homothety. (Actually, they are not quite exactly the same thing, but
shhh!)

4
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

Exercise 4. How can you quickly determine the value of α in the above proof without even needing
to set up the system of equations?

Exercise 5. Give a geometric argument why the spiral similarity, if it exists, must be unique.
(Hint: suppose that T1 and T2 are two such spiral similarities, then what can you say about
T1 ◦ T−1
2 ?)

Now we come to the key result of this section. It gives a very simple and useful description
of the center of a spiral similarity. It can be very useful in locating very subtle spiral similarities
hidden in a geometry problem. Remember this fact!

(Very Useful) Fact 5. Let A, B, C, D be four distinct point in the plane, such that AC is not
parallel to BD. Let lines AC and BD meet at X. Let the circumcircles of ABX and CDX meet
again at O. Then O is the center of the unique spiral similarity that carries A to C and B to D.

C
B
X

D
A

Figure 5: Diagram for Fact 5.

Proof. We give the proof only for the configuration shown above. Since ABXO and CDOX are
cyclic, we have ∠OBD = ∠OAC and ∠OCA = ∠ODB. It follows that triangles AOC and BOD
are similar. Therefore, the spiral similarity centered at O that carries A to C must also carry B to
D.

Exercise 6. Rewrite the above proof using directed angles mod π so that it works for all configu-
rations.

Finally, it is is worth mentioning that spiral similarities often comes in pairs. If we can send
AB to CD, then we can just as easily send AC to BD.

Fact 6. If O is the center of the spiral similarity that sends A to C and B to D, then O is also the
center of the spiral similarity that sends A to B and C to D.

Proof. Since spiral similarity preserves angles at O, we have ∠AOB = ∠COD. Also, the dilation
ratio of the first spiral similarity is OC/OA = OD/OB. So the rotation about with angle ∠AOB =
∠COD and dilation with ratio OB/OA = OD/OC sends A to B, and C to D, as desired.

Exercise 7. Deduce Fact 6 from Facts 2 and 5.

Now, let us apply these results to our configuration in Section 2.

Fact 7. Let M be the Miquel point of quadrilateral ABCD. Then M is the center of spiral
similarity that sends AB to DC, as well as the center of the spiral similarity that sends AD to BC.

5
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

Exercise 8. Prove Fact 7.

Let us specialize to a cyclic quadrilateral, and continue the configuration in Fact 3

Fact 8. Let ABCD be a cyclic quadrilateral with circumcenter O. Let lines AB and CD meet at
Q, and lines DA and CB meet at R. Let M be the Miquel point of ABCD (which lies on line QR,
due to Fact 3). Then OM is perpendicular to QR.

M2 O
B
M1
D A

R
Q
M

Figure 6: Diagram for the proof of Fact 8.

Proof. Let T denote the spiral similarity centered at M which sends A to D and B to C (Fact 7).
Let M1 and M2 be the midpoint of AB and DC, respectively. Then T must send M1 to M2 . So
M is the center of unique spiral similarity that sends A to M1 and D to M2 (Fact 6), and thus it
follows that M, M1 , M2 , Q are concyclic (Fact 5).
Since M1 and M2 are the midpoints of the chords AB and CD, we have ∠OM2 Q = ∠OM1 Q,
and so O, M1 , M2 , Q are concyclic, and OQ is the diameter of the common circle. It follows that
O, M, M1 , M2 , Q all lie on the circle with diameter OQ. In particular, ∠OM Q = 90◦ , as desired.

5 A Criterion for Orthogonality

In this section, we give another proof of Fact 8 and introduce a very useful computational criterion
for orthogonality.

(Very Useful) Fact 9. Let A, B, C, D be points in the plane. Assume that A 6= B and C 6= D.
Then lines AB and CD are perpendicular if and only if AC 2 + BD2 = AD2 + BC 2 .

Proof. The result follows immediately from the following identity.

~ − C)
(A ~ · (A
~ − C)
~ + (B
~ − D)
~ · (B
~ − D)
~ − (A
~ − D)
~ · (A
~ − D)
~ − (B
~ − C)
~ · (B
~ − C)
~ = 2(B
~ − A)
~ · (C
~ − D).
~

Note that the LHS is zero iff AC 2 + BD2 = AD2 + BC 2 and the RHS is zero iff AB ⊥ CD.

Another proof of Fact 8. Let r be the circumradius of ABCD. Using Power of a Point on the
circumcircles of ABCD and ABRM , we get

QO2 − r2 = QA · QB = QM · QR = QM · M R + QM 2

6
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

(the strategy here is to transfer all the data onto the line QR). Similarly, we have

RO2 − r2 = RA · RD = RM · RQ = QM · M R + RM 2 .

Subtracting the two relations, we get

QO2 − RO2 = QM 2 − RM 2 ,

and it thus follows from Fact 9 that OM is perpendicular to QR.

6 Radical Axis
Given two circles in the plane, their radical axis is the locus of points of equal power to the two
circles. It turns out that this is always a line. If the two circles intersect, then the radical axis is
the line passing through the two intersection points (i.e., the common chord). If the two circles are
tangent, then the radical axis is the common internal tangent.

Figure 7: An example of a radical axis.

Exercise 9. Use Fact 9 to deduce that the radical axis is always a line.

It is well known (and easy to prove) that, given three distinct circles, their pairwise radical axes
are either concurrent or all parallel. If the three radical axes meet at a common point, we say that
the common intersection point is the radical center of the three circles.
For instance, using the setup from Fact 8, we see that BC is the radical axis of circles ABCD
and BCQM , AD is the radical axis of circles ABCD and ADQM , and QM is the radical axes
of circles ADQM and BCQM . So the lines AD, BC, QM meet at common point, R, the radical
center of the three circles: ABCD, ADQM, BCQM .

Fact 10. Use the setup from Fact 8. Points A, C, M, O are concyclic, and points B, D, M, O are
concyclic.

Exercise 10. Prove Fact 10. (This is pretty easy angle chasing.)

Fact 11. Use the setup from Fact 8. The lines AC, BD, OM are concurrent.

Proof. Consider the three circles: ABCD, AOCM, BODM . Lines AC, BD, OM are the three
radical axes, and thus they must concur.

Exercise 11. Show that M O bisects ∠CM A as well as ∠BM D.

7
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

O
B
P
D
A

R
Q
M

Figure 8: Diagram for the proof of Fact 11.

7 Inversion and Polarity

In this section, we assume some prior knowledge of inversion, as well as poles and polars. Here is
a quick review:
Let C be a circle, with center O and radius r. The inversion with respect to C is a transformation
(in fact, an involution) that sends a point P (6= O) to a point P 0 on ray OP such that OP ·OP 0 = r2 .
Inversions “switches lines and circles.” Specifically, a line that pass through O gets sent to itself; a
line not passing through O gets sent to a circle through O; a circle that pass through O gets sent to
a line not passing through O; and a circle not passing through O gets sent to a (possible different)
circle not passing through O.
Suppose that P (6= O) is a point, and ` is a line passing through the inverse of P and also
perpendicular to OP , then we say that ` is the polar of P , and that P is the pole of `. Polar maps
satisfy the principle of duality. For instance, the P lies on the polar of Q iff Q lies on the polar of
P ; `1 passes through the pole of `2 iff `2 passes through the pole of `2 ; three poles are collinear iff
the three corresponding polars are concurrent.
Let us return to the configuration.

Fact 12. Let ABCD be a cyclic quadrilateral with circumcenter O. Let AC and BD meet at P ,
lines AB and CD meet at Q, and lines DA and CB meet at R. Let M be the Miquel point of
ABCD. Then P is the inverse of M with respect to the circumcircle of ABCD.

Proof. Since P is the intersection of AC and BD, under the inversion, it must be mapped to the
intersection (other than O) of the circles OAC and OBD, which is M (Fact 10).

Note that this gives another proof of Fact 11, which says that O, P, M are collinear.

Fact 13. The line QR is the polar of the point P .

Proof. This follows from Fact 8 and Fact 13.

Given a circle, C, we say that a triangle is self-polar if each side is the polar of the opposite
vertex.
Now we are able to prove an extremely useful result in projective geometry.3

(Very Useful) Fact 14. The triangle P QR is self-polar with respect to the circumcircle of ABCD.

8
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

O
B
P
D
A

R
Q

Figure 9: P QR is self-polar.

Proof. There is nothing in the proofs that required us to have A, B, C, D in that order on the circle.
By permuting the relabels of A, B, C, D, we can deduce from Fact 13 that P R is the polar of Q,
and P Q is the polar of R. This gives the desired result.

Fact 15. O is the orthocenter of P QR.

Proof. This follows immediately from Fact 14, since OX ⊥ ` for any pole-polar pair (X, `).

8 Summary
This concludes our analysis of the diagram in Section 2. Here is a summary of the key results that
came out of it. (Refer to Figure 1.)

Theorem. Let ABCD be a cyclic quadrilateral with circumcenter O. Let AC and BD meet at
P , lines AB and CD meet at Q, and lines DA and CB meet at R. Let line OP meet QR at M .
Then

(a) The circumcircles of the following triangles all pass through M : QAD, QBC, RAB, RDC,
AOC, BOD. (In particular, M is the Miquel point of the quadrilateral ABCD.)

(b) M is the center of the spiral similarity that carries A to B and D to C, and also the center
of the spiral similarity that carries A to D and B to C.

(c) OM ⊥ QR. In fact, M is the inverse of P with respect to the circumcircle of ABCD.

(d) The triangle P QR is self-polar with respect to the circumcircle of ABCD.

Remember this configuration! Many olympiad geometry problems are basically just a portion
of this one big diagram.
3
For what it’s worth, here’s a very quick sketch of a proof of Fact 14 using projective geometry: Let line RP
intersect AB and BC at E and F , respectively. By applying perspectivities from P and R, we find that (A, B; E, Q) =
(C, D; F, Q) = (B, A; F, Q), from which it follows that (A, B; E, Q) and (C, D; F, Q) are both harmonic. It follows
that EF is the polar of Q, and hence P R is the polar of Q. Similarly we can show that QR is the polar of P , and
P Q is the polar of Q. So P QR is self-polar.

9
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

9 Problems
0. Work through all the exercises.

1. (IMO 1985) A circle with center O passes through the vertices A and C of triangle ABC
and intersects segments AB and BC again at distinct points K and N , respectively. The
circumcircles of triangles ABC and KBN intersects at exactly two distinct points B and M .
Prove that ∠OM B = 90◦ .

2. (China 1992) Convex quadrilateral ABCD is inscribed in circle ω with center O. Diagonals
AC and BD meet at P . The circumcircles of triangles ABP and CDP meet at P and Q.
Assume that points O, P , and Q are distinct. Prove that ∠OQP = 90◦ .

3. (Russia 1999) A circle through vertices A and B of a triangle ABC meets side BC again at
D. A circle through B and C meets side AB at E and the first circle again at F . Prove that
if points A, E, D, C lie on a circle with center O, then ∠BF O = 90◦ .

4. Circles ω1 and ω2 meet at points O and M . Circle ω, centered at O, meet circles ω1 and ω2
in four distinct points A, B, C and D, such that ABCD is a convex quadrilateral. Lines AB
and CD meet at N1 . Lines AD and BC meet at N2 . Prove that N1 N2 ⊥ M O.

5. (Russia 1995; Romanian TST 1996; Iran 1997) Consider a circle with diameter AB and center
O, and let C and D be two points on this circle. The line CD meets the line AB at a point
M satisfying M B < M A and M D < M C. Let K be the point of intersection (different from
O) of the circumcircles of triangles AOC and DOB. Show that ∠M KO = 90◦ .

6. (a) Let A, B, C, D be four points in the plane. Let lines AC and BD meet at P , lines AB
and CD meet at Q, and lines BC and DA meet at R. Let the line through P parallel
to QR meet lines AB and CD at X and Z. Show that P is the midpoint of XZ.
(b) Use part (a) and Fact 8 to prove the Butterfly Theorem: Let C be a circle and let
EF be a chord. Let P be the midpoint of EF , and let AC, BD be two other chords
passing through P . Suppose that AB and CD meet EF at X and Z, respectively, then
P X = P Z.

7. Let ABCD be a cyclic quadrilateral with circumcenter O. Let lines AB and CD meet at R.
Let ` denote the line through R perpendicular to OR. Prove that lines BD and AC meet on
` at points equidistant from R.

8. (USA TST 2007) Triangle ABC is inscribed in circle ω. The tangent lines to ω at B and C
meet at T . Point S lies on ray BC such that AS ⊥ AT . Points B1 and C1 lies on ray ST
(with C1 in between B1 and S) such that B1 T = BT = C1 T . Prove that triangles ABC and
AB1 C1 are similar to each other.

9. Let ABC be a triangle with incenter I. Points M and N are the midpoints of side AB
and AC, respectively. Points D and E lie on lines AB and AC, respectively, such that
BD = CE = BC. Line `1 pass through D and is perpendicular to line IM . Line `2 passes
through E and is perpendicular to line IN . Let P be the intersection of lines `1 and `2 . Prove
that AP ⊥ BC.

10. (IMO 2005) Let ABCD be a given convex quadrilateral with sides BC and AD equal in
length and not parallel. Let E and F be interior points of the sides BC and AD respectively

10
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

such that BE = DF . The lines AC and BD meet at P , the lines BD and EF meet at Q,
the lines EF and AC meet at R. Consider all the triangles P QR as E and F vary. Show
that the circumcircles of these triangles have a common point other than P .

11. A circle is inscribed in quadrilateral ABCD so that it touches sides AB, BC, CD, DA at
E, F, G, H respectively.

(a) Show that lines AC, EF, GH are concurrent. In fact, they concur at the pole of BD.
(b) Show that lines AC, BD, EG, F H are concurrent.

12. (China 1997) Let quadrilateral ABCD be inscribed in a circle. Suppose lines AB and DC
intersect at P and lines AD and BC intersect at Q. From Q, construct the two tangents QE
and QF to the circle where E and F are the points of tangency. Prove that the three points
P , E, F are collinear.

13. Let ABCD be a cyclic quadrilateral with circumcenter O. Let lines AB and CD meet at E,
AD and BC meet at F , and AC and BD meet at P . Furthermore, let EP and AD meet at
K, and let M be the projection of O onto AD be M . Prove that BCM K is cyclic.

14. (IMO Shortlist 2006) Points A1 , B1 and C1 are chosen on sides BC, CA, and AB of a triangle
ABC, respectively. The circumcircles of triangles AB1 C1 , BC1 A1 , and CA1 B1 intersect the
circumcircle of triangle ABC again at points A2 , B2 , and C2 , respectively (A2 6= A, B2 6= B,
and C2 6= C). Points A3 , B3 , and C3 are symmetric to A1 , B1 , C1 with respect to the
midpoints of sides BC, CA, and AB, respectively. Prove that triangles A2 B2 C2 and A3 B3 C3
are similar.

15. Euler point of a cyclic quadrilateral

(a) Let ABCD be a cyclic quadrilateral. Let HA , HB , HC , HD be the orthocenters of BCD,

ACD, ABD, ABC, respectively. Show that HA HB HC HD is the image of ABCD under
a reflection about some point E (i.e. a 180◦ rotation about E).
Point E is called the Euler point of ABCD. (Aside: why is it called the Euler point?4 )
(b) Show that E lies on the nine-point-circles of triangles ABC, ABD, ACD, BCD.
(c) Show that E lies on the Simson line of triangle ABC and point D.
(d) Show that E is also the Euler point of HA HB HC HD .
(e) Let MXY denote the midpoint of XY . Show that the perpendiculars from MAB to CD,
from MBC to DA, from MCD to AB, and from MDA to BC, concur at E.

4
Hint: Recall that the Euler point of a triangle is another name for the center of the nine-point-circle.

11
Winter Camp 2009 Cyclic Quadrilaterals Yufei Zhao

10 Hints
0. Really, they are not hard.

1. Find the configuration in the big diagram. Fact 8 is the key.

2. This is the same as the previous problem! (Why?)

3. We’ve done this too many times already!

4. Use Facts 8 and 10.

5. See previous problem. (Do we need AB to be a diameter?)

6. (a) There is a one-line solution using projective geometry (try a perspectivity at Q). (b) Use
OP ⊥ QR.

7. Butterfly, metamorphized.

8. To see how this fits into the big diagram, try using BCC1 B1 as the starting cyclic quadrilat-
eral.

9. Repeatedly apply Fact 9.

10. Do you see a spiral similarity? Where is its center?

11. Use the self-polar diagonal triangle of EF GH.

12. Use the self-polar diagonal triangle of ABCD.

13. Through power of a point, it suffices to show that F B · F C = F M · F K.

14. Use Fact 5, and see that 4C2 BA ∼ 4C1 A1 B1 ∼ 4CA3 B3 , and similarly with the other
three vertices. Deduce that ∠B2 A2 C2 = ∠B3 A3 C3 .

15. Complex numbers may be helpful. For (c), recall the following fact: the Simson line of ABC
and D bisects DHD . For (e), observe that a dilation of ratio 2 centered at B sends MAB to
A and E to HB .

12
Winter Camp 2010 Three Lemmas in Geometry Yufei Zhao

Three Lemmas in Geometry

Solutions
Yufei Zhao
Massachusetts Institute of Technology

1 Diameter of incircle
1. (IMO 1992) In the plane let C be a circle, ` a line tangent to the circle C, and M a point on `. Find
the locus of all points P with the following property: there exists two points Q, R on ` such that
M is the midpoint of QR and C is the inscribed circle of triangle P QR.
Solution: Let C touch ` at D, and DE be a diameter of C. For any such P, Q, R described in the
problem, the line P E must intersect ` at a point F such that M D = M F , by the lemma. The
point F depends only on M , `, and C. It follows that P must lie on the ray F E beyond E.
Conversely, given a point P lying on the ray F E beyond E, let the tangents from P to C meet ` at
Q and R. By the lemma we must have QF = RD, from which it follows that M is the midpoint of
QR. Therefore, the locus is the ray F E beyond E.

2. (USAMO 1999) 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.
Solution: Observe that F is the center of the excircle of ADC opposite to A (since the center
satisfies the two defining properties of F ). Let line AC touch this excircle at X. Then, using fact
that GACF is cyclic, we have

∠GAF = ∠GCF = ∠XCF = ∠AGF,

and therefore AF G is isosceles.

3. (IMO Shortlist 2005) In a triangle ABC satisfying AB + BC = 3AC the incircle has centre I and
touches the sides AB and BC at D and E, respectively. Let K and L be the symmetric points of
D and E with respect to I. Prove that the quadrilateral ACKL is cyclic.
Solution: Let x = AD, y = BD = BE, z = CE. Then AB = x + y, BC = y + z, AC = x + z,
so the condition that AB + BC = 3AC is equivalent to x + y + y + z = 3x + 3z, or equivalently
y = x + z.
Let line CK meet AB at M and line AL meet BC at N . Then by the lemma, BM = AD = x,
so M D = BD − BM = y − x = z. Similarly N E = x. By comparing the lengths of the legs, we
see that the right triangles M DK and CEL are congruent, and so are ADK and N EL. Therefore
∠M KD = ∠CLE and ∠AKD = ∠N LE. Adding gives ∠M KA = ∠CLN , so ∠AKC = ∠ALC,
and hence ACKL is cyclic.

4. (Nagel line) Let ABC be a triangle. Let the excircle of ABC opposite to A touch side BC at D.
Similarly define E on AC and F on AB. Then AD, BE, CF concur (why?) at a point N known
as the Nagel point.
Let G be the centroid of ABC and I the incenter of ABC. Show that I, G, N lie in that order on
a line (known as the Nagel line, and GN = 2IG.

1
Winter Camp 2010 Three Lemmas in Geometry Yufei Zhao

Solution: Let the incircle of ABC touch BC at X, and let XY be a diameter of the incircle. By
the lemma, A, Y, D are collinear. Let M be the midpoint of BC. Then M I is a midline of triangle
XY D, so IM and Y D are parallel. The dilation centered at G with ratio −2 takes M to A, and
thus it takes line IM to the line through A parallel to IM , namely the line AD. Hence the image of
I under the dilation lies on the line AD. Analogously, it must also lie on BE and CF , and therefore
the image of I is precise N . This proves that I, G, N are collinear in that order with GN = 2IG.

5. (USAMO 2001) Let ABC be a triangle and let ω be its incircle. Denote by D1 and E1 the points
where ω is tangent to sides BC and AC, respectively. Denote by D2 and E2 the points on sides
BC and AC, respectively, such that CD2 = BD1 and CE2 = AE1 , and denote by P the point of
intersection of segments AD2 and BE2 . Circle ω intersects segment AD2 at two points, the closer
of which to the vertex A is denoted by Q. Prove that AQ = D2 P .
Solution: From the lemma we know that D1 Q is a diameter of the incircle. Let the incenter of
ABC be I, its centroid be G, and the midpoint of BC be M . Note that P is the Nagel point of
ABC. From the previous problem, we know that the dilation centered at G with ratio −2 sends
M to A and I to P , and hence sends segment IM to P A, thus P A = 2IM . On the other hand,
a dilation centered at D1 with ratio 2 sends IM to QD2 , so QD2 = 2IM = P A. Therefore,
AQ = P A − QP = QD2 − QP = D2 P .

6. (Tournament of Towns 2003 Fall) Triangle ABC has orthocenter H, incenter I and circumcenter
O. Let K be the point where the incircle touches BC. If IO is parallel to BC, then prove that AO
is parallel to HK.
Solution: Let KE be a diameter of the incircle, and let line AE meet BC at D. Let M be the
midpoint of BC. By the lemma, M is also the midpoint of KD. Since IO is parallel to BC, KM OI
is a rectangle. Since I is the midpoint of KE and M is the midpoint of KD, we see that O must
be the midpoint of ED. Thus lines AE and AO coincide.
Let G be the centroid of ABC. A dilation centered at G with ratio −2 takes M to A and O to H
(by Euler line). So it takes segment M O to AH, and hence AH = 2M O = EK. Since AH and
EK are both perpendicular to BC, it follows that AHKE is a parallelogram, and hence HK is
parallel to AE, which coincides with line AO.

7. (IMO 2008) Let ABCD be a convex quadrilateral with |BA| = 6 |BC|. Denote the incircles of
triangles ABC and ADC by ω1 and ω2 respectively. Suppose that there exists a circle ω tangent
to the ray BA beyond A and to the ray BC beyond C, which is also tangent to the lines AD and
CD. Prove that the common external tangents of ω1 and ω2 intersect on ω.
(Hint: show that AB + AD = CB + CD. What does this say about the lengths along AC?)
Solution: Chasing some lengths using equal tangents yields AB + AD = CB + CD (details
omitted). Let ω1 and ω2 touch AC at P and Q respectively. Then AP = 21 (AB + AC − BC) =
1
2 (CD + AC − AD) = CQ.
Let P P 0 be a diameter of ω1 , and let QQ0 be a diameter of ω2 . By the lemma, B, P 0 , Q are collinear,
and P, Q0 , D are collinear.
Choose point T on ω so that the tangent to ω at T is parallel to AC and furthermore puts ω and
B on different sides. Then the dilation centered at B that sends ω1 to ω must send P 0 to T , so
B, P 0 , Q, T are collinear. Analogously, the dilation centered at D (with negative ratio) that sends
ω2 to ω must take Q0 to T , so P, Q0 , D, T are collinear.
Now, P P 0 and Q0 Q are parallel diameters of ω1 and ω2 , and lines P 0 Q and P Q0 meet at T . It
follows that there is a dilation with positive ratio centered at T that takes ω1 to ω2 , and hence T
is the intersection of the common external tangents of ω1 and ω2 . Since T lies on ω, we are done.

2
Winter Camp 2010 Three Lemmas in Geometry Yufei Zhao

2 Center of spiral similarity

1. (IMO Shortlist 2006) Let ABCDE be a convex pentagon such that

∠BAC = ∠CAD = ∠DAE and ∠CBA = ∠DCA = ∠EDA.

Diagonals BD and CE meet at P . Prove that line AP bisects side CD.

Solution: Since A is the center of the spiral similarity sending BC to DE, by the lemma we know
that ABCP and AP DE are both cyclic. Furthermore, since ∠ACD = ∠ABC, the circumcircle of
ABCP is tangent to CD. Since ∠ADC = ∠AEC, the circumcircle of AP DE is also tangent to
CD. Let AP meet CD at M . Then by power of a point, M C 2 = M P ·M A = M D2 , so M C = M D,
as desired.

2. (USAMO 2006) Let ABCD be a quadrilateral, and let E and F be points on sides AD and
BC, respectively, such that AE/ED = BF/F C. Ray F E meets rays BA and CD at S and T ,
respectively. Prove that the circumcircles of triangles SAE, SBF , T CF , and T DE pass through a
common point.
Solution: Note that ABCD cannot be a parallelogram, for otherwise F E and BA would be
parallel. Let P be the center of spiral similarity that carries AD to BC. It must also carry E
to F since AE/ED = BF/F C. Since the spiral similarity carries AE to BF , it follows by the
lemma that P AES and P BF S are concyclic. Similarly, since the spiral similarity carries DE to
CF , P EDT and P F CT are concyclic. Therefore the circumcircles of SAE, SBF, T CF, T DE all
pass through P .

3. (China 1992) Convex quadrilateral ABCD is inscribed in circle ω with center O. Diagonals AC
and BD meet at P . The circumcircles of triangles ABP and CDP meet at P and Q. Assume that
points O, P , and Q are distinct. Prove that ∠OQP = 90◦ .
Solution: Let M and N be the midpoints of AC and BD, respectively. Let T be the spiral
similarity that carries A to B and C to D. By the lemma (and the fact after it), Q is the center of
the spiral similarity. Since T carries AC to BD, it preserves midpoints, so T brings M to N . Using
the lemma again, we see that M, N, P, Q are concyclic. Since ∠OM P = ∠ON P = 90◦ , points
O, P, M, N are concyclic with diameter OP . Therefore, M, N, P, Q, O are concyclic with diameter
OP , and therefore ∠OQP = 90◦ .

4. Let ABCD be a quadrilateral. Let diagonals AC and BD meet at P . Let O1 and O2 be the
circumcenters of AP D and BP C. Let M , N and O be the midpoints of AC, BD and O1 O2 . Show
that O is the circumcenter of M P N .
Solution: Let the circumcircles of AP D and BP C meet at P and Q. Let T denote the spiral
similarity that sends AD to CB. Then T is centered at Q by the lemma. Let id denote the
identity transformation, and consider the transformation R = 21 (id + T). This is another spiral
similarity centered at Q (if you’re not convinced, think about it in terms of multiplication by
complex numbers). Then R(A) = M , R(D) = N , R(O1 ) = O. Since O1 is the circumcenter
of QAD, the transformation yields that O is the circumcenter of QM N , whose circumcircle must
again pass through P by the lemma. This proves the desired fact.

5. (Miquel point of a quadrilateral) Let `1 , `2 , `3 , `4 be four lines in the plane, no two parallel. Let Cijk
denote the circumcircle of the triangle formed by the lines `i , `j , `k (these circles are called Miquel
circles). Then C123 , C124 , C134 , C234 pass through a common point (called the Miquel point).
(It’s not too hard to prove this result using angle chasing, but can you see why it’s almost an
immediate consequence of the lemma?)

3
Winter Camp 2010 Three Lemmas in Geometry Yufei Zhao

Solution: Let Pij denote the intersection of `i and `j . Let C134 and C234 meet at P . Then by the
lemma, P is the center of the spiral similarity that sends P13 to P23 and P14 to P24 . It follows that
P is also the center of the spiral similarity that sends P13 to P14 and P23 to P24 . Applying the
lemma again, we find that C123 and C124 also pass through P , as desired.

6. (IMO 2005) Let ABCD be a given convex quadrilateral with sides BC and AD equal in length
and not parallel. Let E and F be interior points of the sides BC and AD respectively such that
BE = DF . The lines AC and BD meet at P , the lines BD and EF meet at Q, the lines EF and
AC meet at R. Consider all the triangles P QR as E and F vary. Show that the circumcircles of
these triangles have a common point other than P .
Solution: Let S be the center of the spiral similarity that carries AD to CB, then it must also
carry F to E. Using the lemma, we see that SP AD, SRAF , SQF D are all cyclic. Note that they
are the Miquel circles of the quadrilateral formed by the lines AD, AP, P D, QF , and thus S is the
Miquel point of these circles. The remaining Miquel circle passes through P, Q, R, S, and hence S
lies on the circumcircle of P QR. Note that S is the desired point, as it does not depend on the
choice of E and F .

7. (IMO Shortlist 2006) Points A1 , B1 and C1 are chosen on sides BC, CA, and AB of a triangle ABC,
respectively. The circumcircles of triangles AB1 C1 , BC1 A1 , and CA1 B1 intersect the circumcircle
of triangle ABC again at points A2 , B2 , and C2 , respectively (A2 6= A, B2 6= B, and C2 6= C).
Points A3 , B3 , and C3 are symmetric to A1 , B1 , C1 with respect to the midpoints of sides BC, CA,
and AB, respectively. Prove that triangles A2 B2 C2 and A3 B3 C3 are similar.
Solution: By the lemma, C2 is the center of the spiral similarity that takes A1 to B and B1 to A.
So triangles C2 A1 B1 and C2 BA are similar. But C2 is also the center of the spiral similarity that
takes A1 to B1 and B to A. Then because BA1 = CA3 and AB1 = CB3 ,
C 2 A1 BA1 CA3
= = .
C2 B 1 AB1 CB3
Since ∠A1 C2 B1 = ∠A3 CB3 , triangles CA3 B3 , C2 A1 B1 , and C2 BA are similar. So ∠CA3 B3 =
∠C2 BA. Similarly, ∠BA3 C3 = ∠B2 CA. Then

∠B2 A2 C2 = ∠B2 AC2 = 180◦ − ∠B2 C2 A − ∠C2 B2 A = 180◦ − ∠B2 CA − ∠C2 BA

= 180◦ − ∠BA3 C3 − ∠CA3 B3 = ∠B3 A3 C3 .

Similarly ∠A2 B2 C2 = ∠A3 B3 C3 , hence triangles A2 B2 C2 and A3 B3 C3 are similar.

3 Symmedian
1. (Poland 2000) Let ABC be a triangle with AC = BC, and P a point inside the triangle such that
∠P AB = ∠P BC. If M is the midpoint of AB, then show that ∠AP M + ∠BP C = 180◦ .
Solution: Since ∠CAB = ∠CBA, ∠P AB = ∠P BC implies that ∠P AC = ∠P BA, and thus
the circumcircle of ABP is tangent to CA and CB. It follows by the lemma that line CP is a
symmedian of AP B, and therefore ∠AP M = 180◦ − ∠BP C.

2. (IMO Shortlist 2003) Three distinct points A, B, C are fixed on a line in this order. Let Γ be a circle
passing through A and C whose center does not lie on the line AC. Denote by P the intersection of
the tangents to Γ at A and C. Suppose Γ meets the segment P B at Q. Prove that the intersection
of the bisector of ∠AQC and the line AC does not depend on the choice of Γ.
Solution: We follow the method of the first proof of the lemma. Let the bisector of ∠AQC meet
A at R, then AR/RC = AQ/QC, so it suffices to show that AQ/QC does not depend on Γ.

4
Winter Camp 2010 Three Lemmas in Geometry Yufei Zhao

Applying Sine law repeatedly, we find that

PQ AB
AQ sin ∠ACQ sin ∠P AQ sin ∠AQP sin ∠AQB AQ sin ∠ABQ AB QC
= = = PAP
Q
= = CB
= .
QC sin ∠CAQ sin ∠P CQ sin ∠CQB sin ∠CBQ CB AQ
CP sin ∠CQP CQ
p
Thus AQ/QC = AB/CB, which is independent of Γ, as desired.
3. (Vietnam TST 2001) In the plane, two circles intersect at A and B, and a common tangent intersects
the circles at P and Q. Let the tangents at P and Q to the circumcircle of triangle AP Q intersect
at S, and let H be the reflection of B across the line P Q. Prove that the points A, S, and H are
collinear.
Solution: We will only do the configuration where B is closer to line P Q than A. You should
think about what happens in the other configuration, which is analogous.
Since AS coincides with the symmedian of AP Q, it suffices to show that H lies on this symmedian.
Note that AB coincides with a median of AP Q. Indeed, let line AB meet P Q at M , then by Power
of a Point, M P 2 = M B · M A = M Q2 , so M P = M Q.
Since ∠P HQ = ∠P BQ = 180◦ − ∠BP Q − ∠BQP = 180◦ − ∠BAP − ∠BAQ = 180◦ − ∠P AQ, we
see that AP HQ is cyclic. Then ∠HAQ = ∠HP Q = ∠BP Q = ∠BAP . Since AB coincides with
a median of AP Q, it follows that AH coincides with a symmedian of AP Q, and hence A, H, S are
collinear.
4. (USA TST 2007) Triangle ABC is inscribed in circle ω. The tangent lines to ω at B and C meet
at T . Point S lies on ray BC such that AS ⊥ AT . Points B1 and C1 lies on ray ST (with C1 in
between B1 and S) such that B1 T = BT = C1 T . Prove that triangles ABC and AB1 C1 are similar
to each other.
Solution: Let M be the midpoint of BC. Since BT is tangent to ω, we have ∠T BA = 180◦ −∠BCA.
By the lemma, we have ∠BAT = ∠CAM . Applying Sine law to triangles BAT and CAM , we get
BT sin ∠BAT sin ∠CAM MC
= = = .
AT sin ∠T BA sin ∠BCA AM
Since T B = T C1 , we have T C1 /T A = M C/M A. Note that ∠T M S = ∠T AS = 90◦ , so T M AS is
cyclic, and hence ∠AM S = ∠AT S. Therefore, triangles AM C and AT C1 are similar. Analogously,
triangles AM B and AT B1 are similar. Combine the two results, and we see that ABC and AB1 C1
are similar.
5. Let ABC be a triangle. Let X be the center of spiral similarity that takes B to A and A to C.
Show that AX coincides with a symmedian of ABC.
Solution: Let the tangents to the circumcircle of ABC at B and C meet at D. Let AD meet the
circumcircle of BCD again at X 0 . Then
∠ABX 0 = ∠BX 0 D − ∠BAX 0 = ∠BCD − ∠BAD = ∠BAC − ∠BAD = ∠X 0 AC,
and analogously we have ∠ACX 0 = ∠X 0 AB. Therefore X = X 0 , and it lies on the symmedian AD.
Second solution, not using the lemma: Let Y be a point on the circumcircle of ABC so that AY
coincides with a symmedian of ABC. Let X 0 be the midpoint of AY . Let M be the midpoint of
BC and N the midpoint of AC. Since AY is a symmedian, ∠BAY = ∠M AC. Additionally we
have ∠BY A = ∠M CA, so triangles ABY and AM C are similar. Since X 0 is the midpoint of Y
and N is the midpoint of AC, we see that ABX 0 and AM N are similar. Hence
∠ABX 0 = ∠AM N = ∠M AB = ∠CAX 0 .
Analogously ∠ACX 0 = ∠BAX 0 . Therefore X = X 0 , and it lies on a symmedian.

5
Winter Camp 2010 Three Lemmas in Geometry Yufei Zhao

6. (USA TST 2008) Let ABC be a triangle with G as its centroid. Let P be a variable point on
segment BC. Points Q and R lie on sides AC and AB respectively, such that P Q k AB and
P R k AC. Prove that, as P varies along segment BC, the circumcircle of triangle AQR passes
through a fixed point X such that ∠BAG = ∠CAX.
Solution: Let X be the center of spiral similarity T that carries B to A and A to C, as in the
previous problem. Triangles BRP and P QC are similar as the corresponding sides are parallel.
Since AR/RB = QP/RB = QC/RP = CQ/QA, we see that T must carry R to Q. Thus triangles
ARX and XQC are similar, so ∠ARX = ∠XQC, and hence ARXQ is cyclic. Note that ∠BAG =
∠CAX since X lies on the symmedian. Therefore X has the required properties.

7. (USA 2008) 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.
Solution: Let X be as in Problem 5. Then ∠ABX = ∠XAC = ∠BAM = ∠ABF , and analogously
∠ACX = ∠ACF , so X = F . Then this problem follows as a special case of the previous problem.

8. Let A be one of the intersection points of circles ω1 , ω2 with centers O1 , O2 . The line ` is tangent
to ω1 , ω2 at B, C respectively. Let O3 be the circumcenter of triangle ABC. Let D be a point such
that A is the midpoint of O3 D. Let M be the midpoint of O1 O2 . Prove that ∠O1 DM = ∠O2 DA.
(Hint: use Problem 5.)
Solution: We have ∠AO3 C = 2∠ABC = ∠AO1 B, and triangles AO3 C and AO1 B are both
isosceles, hence they are similar. So the spiral similarity at A that carries O1 to B also carries O3
to C, and it follows that AO1 O3 and ABC are similar. Analogously, triangles ABC and AO3 O2
are similar. So triangles AO1 O3 and AO3 O2 are similar.
Compare triangles AO1 D and ADO2 . We have ∠O1 AD = 180◦ − ∠O1 AO3 = 180◦ − ∠O3 AO2 =
∠DAO2 . Also AO1 /AD = AO1 /AO3 = AO3 /AO2 = AD/AO2 . It follows that triangls AO1 D
and ADO2 are similar. It follows by Problem 5 that DA is a symmedian of DO1 O2 , and thus
∠O1 DM = ∠O2 DA.

6
Trinity Training 2011 Power of a Point Yufei Zhao

Power of a Point
Solutions
Yufei Zhao
Trinity College, Cambridge

April 2011

Practice problems:

1. Let Γ1 and Γ2 be two intersecting circles. Let a common tangent to Γ1 and Γ2 touch Γ1
at A and Γ2 at B. Show that the common chord of Γ1 and Γ2 , when extended, bisects
segment AB.

B
A

Solution. Let the common chord extended meet AB at M . Since M lies on the radical
axis of Γ1 and Γ2 , it has equal powers with respect to the two circles, so M A2 = M B 2 .
Hence M A = M B.

2. Let C be a point on a semicircle of diameter AB and let D be the midpoint of arc AC.
Let E be the projection of D onto the line BC and F the intersection of line AE with
the semicircle. Prove that BF bisects the line segment DE.

Solution.

E
D
F C

A B

Let Γ denote the circle with diameter AB, and Γ1 denote the circle with diameter BE.
Since ∠AF B = 90◦ , Γ1 passes through F . Also since ∠DEB = 90◦ , Γ1 is tangent to
DE. From Problem 1, we deduce that the common chord BF of Γ and Γ1 bisects their
common tangent DE.

3. Let A, B, C be three points on a circle Γ with AB = BC. Let the tangents at A and B
meet at D. Let DC meet Γ again at E. Prove that the line AE bisects segment BD.

Solution.

1
Trinity Training 2011 Power of a Point Yufei Zhao

A C

D B

Let Γ1 denote the circumcircle of ADE. By Problem 1 it suffices to show that Γ1 is

tangent to DB. Indeed, we have

∠ADB = 180◦ − 2∠ABD = ∠ABC = ∠AEC,

which implies that Γ1 is tangent to D.

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

Solution.

Γ2
N

Γ1

P M Q
C D
A B ℓ
E

Extend N M to meet AB at X. Then by Problem 1, X is the midpoint of AB. Since P Q

is parallel to AB, it follows that M is the midpoint of P Q. Since ∠M AB = ∠M CE =
∠BAE and ∠M BA = ∠M DE = ∠ABE, we see that E is the reflection of M across AB.
So EM the perpendicular bisector of P Q, and hence EP = EQ.

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

Solution.

2
Trinity Training 2011 Power of a Point Yufei Zhao

Q
A

H
R
P

B D C

Let D be the foot of the perpendicular from A to BC, and let H be the orthocenter of
ABC. Since ∠ADB = 90◦ , the circle with diameter AB passes through D, so HS · HR =
HA · HD by power of a point. Similarly the circle with diameter AC passes through D
as well, so HP · HQ = HA · HD as well. Hence HP · HQ = HR · HS, and therefore by
the converse of power of a point, P, Q, R, S are concyclic.
6. Let ABC be an acute triangle with orthocenter H. The points M and N are taken on
the sides AB and AC, respectively. The circles with diameters BN and CM intersect at
points P and Q. Prove that P , Q, and H are collinear.

Solution.
A

E
P
F
H N
M

B C

We want to show that H lies on the radical axis of the two circles, so it suffices to show
that H has equal powers with respect to the two circles.
Let BE and CF be two altitudes of ABC. Since ∠BEN = 90◦ , E lies the circle with
diameter BN . Hence the power of H with respect to the circle with diameter BN is
HB · HE. Similarly, the power of H with respect to the the circle with diameter CM is
HC · HF .
Since ∠BEC = ∠BF C = 90◦ , B, C, E, F are concyclic, hence HB · HE = HC · HF by
power of a point. It follows that H has equal powers with respect to the two circles with
diameter AB and BC.
7. (Euler’s relation) In a triangle with circumcenter O, incenter I, circumradius R, and
inradius r, prove that
OI 2 = R(R − 2r).

3
Trinity Training 2011 Power of a Point Yufei Zhao

Solution.

O
I

B C

Let AI extended meet the circumcircle again at D. The power of I with respect to the
circumcircle is equal to
−IA · ID = IO2 − R2 .
Let us compute the lengths of IA and ID. By consider the right triangle with one vertex A
and the opposite side the radius of the incircle perpendicular to AB, we find IA = r sin A2 .
We have

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

2R
Thus ID = BD = sin A
, where the last equality follows from the law of sines on triangle
2
ABD. Hence
A 2R
R2 − IO2 = IA · ID = r sin · = 2Rr.
2 sin A2
The result follows.

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

Solution.

A D B M C

4
Trinity Training 2011 Power of a Point Yufei Zhao

Using power of point, we have AE · AF = AB 2 = AD · AC. Therefore, D, C, F, E are

concyclic. The intersection M of the perpendicular bisectors of DE and CF must meet
at the center of the circumcircle of DCF E. Since M is on DC, it follows that DC is the
1 DC 1 3 3
diameter of this circle. Hence M is the midpoint of DC. So M C
AC = 2 AC = 2 · 4 = 8 .
AM 3
Thus M C = 5.

9. Let ABC be a triangle and let D and E be points on the sides AB and AC, respectively,
such that DE is parallel to BC. Let P be any point interior to triangle ADE, and let F
and G be the intersections of DE with the lines BP and CP , respectively. Let Q be the
second intersection point of the circumcircles of triangles P DG and P F E. Prove that the
points A, P , and Q are collinear.

Solution.

N
P
D E
F G
M

B C
Q

Let the circumcircle of DP G meet line AB again at M , and let the circumcircle of EP F
meet line AC again at N . Assume the configuration where M and N lie on sides AB
and AC respectively (the arguments for the other cases are similar). We have ∠ABC =
∠ADG = 180◦ − ∠BDG = 180◦ − ∠M P C, so BM P C is cyclic. Similarly, BP N C is
cyclic as well. So BCN P M is cyclic. Hence ∠AN M = ∠ABC = ∠ADE, so M, N, D, E
are concyclic. By power of a point, AD · AM = AE · AD. Therefore, A has equal power
with respect to the circumcircles of DP G and the EP F , and thus A lies on line P Q, the
radical axis.

10. (IMO 1995) Let A, B, C, and D be four distinct points on a line, in that order. The
circles with diameters AC and BD intersect at X and Y . The line XY meets BC at Z.
Let P be a point on the line XY other than Z. The line CP intersects the circle with
diameter AC at C and M , and the line BP intersects the circle with diameter BD at B
and N . Prove that the lines AM , DN , and XY are concurrent.

Solution.

5
Trinity Training 2011 Power of a Point Yufei Zhao

M
X N

P
A B C D

By power of a point, we have P M · P C = P X · P Y = P N · P B, so B, C, M, N are

concyclic. Note that ∠AM C = ∠BN D = 90◦ since they are subtended by diameters AC
ad BD, respectively. Hence ∠M N D = 90◦ + ∠M N B = 90◦ + ∠M CA = 180◦ − ∠M AD.
Therefore A, D, N, M are concyclic. Since AM , DN , XY are the three radical axes for
the circumcircles of AM XC, BXN D, and AM N D, they concur at the radical center.

6
IMO Training 2008 Circles Yufei Zhao

Circles
Yufei Zhao
[email protected]

1 Warm up problems
1. Let AB and CD be two segments, and let lines AC and BD meet at X. Let the circumcircles of
ABX and CDX meet again at O. Prove that triangles OAB and OCD are similar.

2. (Miquel’s theorem) Let ABC be a triangle. Points X, Y , and Z lie on sides BC, CA, and AB,
respectively. Prove that the circumcircles of triangles AY Z, BXZ, CXY meet at a common point.

3. (Also Miquel’s theorem) Let a, b, c, d be four lines in space, no two parallel, no three concurrent.
Let ωa denote the circumcircle of the triangle formed by lines b, c, d. Similarly define ωb , ωc , ωd .

(a) Prove that ωa , ωb , ωc , ωd pass through a common point. This is called the Miquel point.
(b) Continuing the above notation, prove that the centers of ωa , ωb , ωc , ωd , and the Miquel point
all lie on a common circle.

4. (Simson line) Let ABC be a triangle, and let P be another point on its circumcircle. Let X, Y, Z
be the feet of perpendiculars from P to lines BC, CA, AB respectively. Prove that X, Y, Z are
collinear.

5. Let ∠AOB be a right angle, M and N points on rays OA and OB, respectively. Let M N P Q be
a square such that M N separates the points O and P . Find the locus of the center of the square
when M and N vary.

6. Let ABCD be a convex quadrilateral such that the diagonals AC and BD are perpendicular,
and let P be their intersection. Prove that the reflections of P with respect to AB, BC, CD, DA
are concyclic.

7. An interior point P is chosen in the rectangle ABCD such that ∠AP D + ∠BP C = 180◦ . Find
∠DAP + ∠BCP .

8. Let AB be a chord in a circle and P a point on the circle. Let Q be the projection of P on
AB and R and S the projections of P onto the tangents to the circle at A and B. Prove that
P Q2 = P R · P S.

9. Let ABC be an acute triangle. The points M and N are taken on the sides AB and AC
respectively. The circles with diameters BN and CM intersect at points P and Q. Prove that
P , Q, and the orthocenter H are collinear.

10. 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 a
common point, or no two of them have any common point. (The line through the midpoints of
AC, BD, EF is called the Newton-Gauss line.)

1
IMO Training 2008 Circles Yufei Zhao

2 Tangent circles
In this section, we explore the following two configurations.1

1. Let chords AC and BD of a circle ω intersect at P . A smaller circle ω1 is tangent to ω at T and

to segments AP and DP at E and F respectively.

B C

F
E

(a) Prove that ray T E bisects arc ABC of ω.

(b) Let I be the incenter of triangle ACD and M be the midpoint of arc ABC of ω. Prove that
M A = M I = M C.
(c) Let F 0 be the common point of ω1 and line EI other than E. Prove that I, F 0 , D, T are
concyclic.
C
M

F'
E
I

(d) Prove that DF 0 is tangent to ω1 . This means that F = F 0 , so that E, F, I are collinear.
(Remember this fact.)
(e) Let J be the incenter of triangle AP D. Prove that T, D, F, J are concyclic.
1
Thanks to Oleg Golberg for providing some of these problems.

2
IMO Training 2008 Circles Yufei Zhao

B C

F
E J

(f) Prove that ray T J bisects angle AT D.

2. Now we consider a special case of the part (d) in the previous problem. Try to find a short proof
of the following result. (Hint: use Pascal’s theorem)
Let ABC be a triangle and I its incenter. Let Γ be the circle tangent to sides AB, AC, as well
as the circumcircle of ABC. Let Γ touch AB and AC at X and Y , respectively. Show that I is
the midpoint of XY .

X
Y
I
B
C

3. Let Ω be the circumcircle of ABC. A circle ω is tangent to sides AB, AC and circle Ω at points
X, Y and Z, respectively. Let M be the midpoint of the arc BC of Ω which does not contain A.
Prove that lines XY , BC and ZM have a common point.
(Can you prove the result when Ω only passes through B and C and contains A in the interior?)

4. A circle ω is tangent to sides AB, AC of triangle ABC and to its circumcircle at points X, Y
and Z. Segments AZ and XY meet at T . Prove that ∠BT X = ∠CT Y .

5. (Sawayama-Thébault) Let ABC be a triangle with incenter I. Let D a point on side BC. Let
P be the center of the circle that touches segments AD, DC, and the circumcircle of ABC, and
let Q be the center of the circle that touches segments AD, BD, and the circumcircle of ABC.
Show that P, Q, I are collinear.

3 Projective quickies
Warning: The introductory text in this section is very rough. It is not meant for you to learn from.
Rather, I assume that you already know the basics of projective geometry, and this should serve as a
quick review of the important concepts.
You should be familiar with the notion of pole and polar. A quick definition is as follows:

3
IMO Training 2008 Circles Yufei Zhao

Definition. Suppose that ω is a circle with center O, and P and P 0 are inverses respect to ω (i.e., P 0
lies on ray OP such that OP · OP 0 = r2 , where r is the radius of ω). Let ` be the line through P 0
perpendicular to OP . Then we say that ` is the polar of P and P is the pole of `.

Here are a few diagrams that may help you to remember the common setups of poles and polars.

polar

pole
pole
pole pole

polar

polar polar

Make sure that you understand the duality behind poles and polars. If you need to prove that something
is true, it suffices to prove its polar dual. The polar map transforms points to lines, lines to points. It
transforms the intersection of two lines to the line joining the two points, and vice-versa. For instance:

• To show that line ` passes through point P , it suffices to show that the pole of ` lies on the polar
of P .
• To show that three points are collinear, it suffices to show that their poles are concurrent.

The cross ratio of four collinear points A, B, C, D is defined as

AC · BD
(A, B; C, D) =
AD · BC
where the lengths are directed. Cross ratios are preserved under projective transformations. (Besides
coincidence, this is pretty much the only thing that’s projective-invariant.)
The most significant case is when (A, B; C, D) = −1, in which case we say that (A, B; C, D) is harmonic.
This notion of cross ratios and harmonic quadruples is not limited to collinear points, but also applies
to pencils of lines, and concyclic points. Harmonic quadruples arise frequently, so learn to recognize
them!
Possibly the most useful fact from polar geometry is the self-polarity of the diagonal triangle of a
cyclic quadrangle. (Part of the reason why it’s powerful if because it is not easy to prove simply using
Eucliean geometry.)
Theorem. Let ABCD be a cyclic quadrilateral with circumcircle ω, and let AD∩BC = P , AC ∩BC =
Q AB ∩ CD = R. Then P QR is self-dual with respect to ω. That is, P is the pole of QR, Q is the
pole of P R, and R is the pole of P Q.

In particular, we see that O, P, Q, R forms an orthocentric quadruple, where O is the center of ω. (This
means that any of the four points is the orthocenter of the triangle formed by the other three).
P

A
B
E
Q F
R
C
D

4
IMO Training 2008 Circles Yufei Zhao

Here is a quick sketch of the proof. Let line QR meet AD and BC at E and F respectively. Then from
the configuration of lines, we see that (A, D; P, E) and (C, B; P, F ) are both harmonic. It follows the
polar of P is EF , which coincides with QR.
You should develop the skill of recognizing when to use projective geometry. Here is a gallery of
diagrams showing the patterns that tend to hint the use of projective geometry, in particular the use of
poles and polars. Try to figure out the projective significance of each diagram. Ask Yufei if you need
help.

1. Let ABC be a triangle and I be its incenter. Let the incenter of ABC touch sides BC, CA, AB
at D, E, F respectively. Let S denote the intersection of lines EF and BC. Prove that SI ⊥ AD.
2. Let U V be a diameter of a semicircle, and let P, Q are two points on the semicircle. The tangents
to the semicircle at P and Q meet at R, and lines U P and V Q meet at S. Prove that RS ⊥ U V .
3. A circle is inscribed in quadrilateral ABCD so that it touches sides AB, BC, CD, DA at E, F, G, H
respectively.
(a) Show that lines AC, EF, GH are concurrent. In fact, they concur at the pole of BD.
(b) Show that lines AC, BD, EG, F H are concurrent.
4. (China 1996) Let H be the orthocentre of triangle ABC. From A construct tangents AP and
AQ to the circle with diameter BC, where P , Q are the points of tangency. Prove that P, H, Q
are collinear.
5. (China 1997) Let quadrilateral ABCD be inscribed in a circle. Suppose lines AB and DC
intersect at P and lines AD and BC intersect at Q. From Q, construct the two tangents QE
and QF to the circle where E and F are the points of tangency. Prove that the three points P ,
E, F are collinear.
6. (Butterfly theorem) Let ω be a circle and let XY be a chord. Let M be the midpoint of XY ,
and let AB and CD be two chords of ω, both passing through M . Let XY meet chords AD and
BC are P and Q respectively. So that M P = M Q.
7. Points C, M, D and A lie on line ` in that order with CM = M D. Circle ω is tangent to line `
at A. Let B be the point on ω that is diametrically opposite to A. Lines BC and BD meet ω at
P and Q. Prove that the lines tangent to ω at P and Q and line BM are concurrent.
(Hint: what is the significance of midpoints in terms of harmonic conjugates?)

5
IMO Training 2008 Circles Yufei Zhao

8. A very important fact. Let ABC be a triangle and Γ its circumcircle. Let the tangent to Γ
at B and C meet at D. Then AD coincides with a symmedian of 4ABC. (The symmedian is
the reflection of the median across the angle bisector, all through the same vertex.)
(Hint: consider a reflection about the angle bisector of ∠A.)

9. Let ABCD be a quadrilateral (not necessarily cyclic), and let X be the intersection of diagonals
AC and BD. Suppose that BX is a symmedian of triangle ABC and DX is a symmedian of
triangle ADC. Prove that AX is a symmedian of triangle ABD.

10. Let ABC be a triangle and tangent at point C to the circumcircle of ABC meets AB in M .
The line perpendicular to OM in M intersects BC and AC in P and Q respectively. Prove that
M P = M Q.

11. (Iberoamerican 1998) The incircle of triangle ABC is tangent to sides BC, CA, AB at D, E, F ,
respectively. Let segment AD meet the incircle again at Q. Show that the line EQ passes through
the midpoint of segment AF if and only if AC = BC.

12. Let ABCD be a convex quadrilateral (not necessarily cyclic). Let lines AB and CD meet at E,
AD and BC meet at F , and AC and BD meet at P . Let M be the projection of P onto EF .

(a) Show that ∠AM P = ∠CM P .

(b) Show that ∠BM C = ∠AM D.

(Hint: what do internal and external angle bisectors have to do with harmonic divisions?)

13. Let ABCD be a circumscribed about a circle quadrilateral with the incenter I. Let M be the
projection of I onto AC. Prove that ∠AM B = ∠AM D.

14. Let ABCD be a convex quadrilateral (not necessarily cyclic). Let lines AB and CD meet at E,
AD and BC meet at F . Suppose X is a point inside the quadrilateral such that ∠AXE = ∠CXF .
Prove that ∠AXB + ∠CXD = 180◦ .
(Hint: apply a polar transformation centered at X and see what happens. And yes, you don’t
start with any circle. It’s kind of weird I know . . . )

15. Let ABCD be a cyclic quadrilateral with circumcenter O. Let lines AD and BC meet at E and
lines AB and CD meet at F . Let M be the projection of O onto line EF . Prove that M O bisects
∠BM D.

16. (IMO 1985) A circle with center O passes through the vertices A and C of triangle ABC, and
intersects the segments AB and BC again at distinct points K and N , respectively. The circum-
scribed circles of the triangle ABC and KBN intersect at exactly two distinct points B and M .
Prove that angle ∠OM B = 90◦ .

17. Let ABCD be a cyclic quadrilateral with circumcenter O. Let lines AB and CD meet at E, AD
and BC meet at F , and AC and BD meet at P . Furthermore, let EP and AD meet at K, and
let M be the projection of O onto AD be M . Prove that BCM K is cyclic.

18. Let ABC be a triangle with incenter I. Fix a line ` tangent to the incircle of ABC (not BC, CA
or AB). Let A0 , B0 , C0 be points on ` such that

∠AIA0 = ∠BIB0 = ∠CIC0 = 90◦

Show that AA0 , BB0 , CC0 are concurrent.

6
IMO Training 2008 Circles Yufei Zhao

19. Let ABC be a triangle with incenter I. Fix a line ` tangent to the incircle of ABC (not BC,
CA or AB). Let ` intersect the sides of the triangle at M, N, P . At I, erect perpendiculars to
IM, IN and IP and let them intersect the corresponding sides of the triangle at M0 , N0 , and P0
respectively. Show that M0 , N0 , P0 lie on a line tangent to the incircle.

20. Let ABC be a triangle, and ω its incircle. Let ω touch the sides BC, CA, AB at D, E, F respec-
tively. Let X be a point on AD and inside ω. Let segments BX and CX meet ω at Q and R
respectively. Show that lines EF, QR, BC are concurrent.

Food for thought. Let ABC be a triangle with inscribed circle ω. There exists a projective trans-
formation that preserves ω but sends ABC to any other desired triangle inscribed in ω! Can you prove
this?
If a geometry problem is completely projective-invariant, then perhaps sending ABC to an equilateral
triangle or an isosceles right triangle might be a good idea. Note that when the circumcircle is not
present, we can simply apply an affine transformation to get what we want.

7
IMO Training 2007 Similarity Yufei Zhao

Similarity
Yufei Zhao
July 12, 2007
[email protected]

1. Let ABCD be a convex quadrilateral. Show that

AC 2 · BD2 = AB 2 · CD2 + AD2 · BC 2 − 2AB · BC · CD · DA cos(A + C).

2. (IMO Shortlist 1998) Let ABCDEF be a convex hexagon such that ∠B + ∠D + ∠F = 360◦ and

AB CD EF
· · = 1.
BC DE F A
Prove that
BC AE F D
· · = 1.
CA EF DB
3. A circle ω is inscribed in a quadrilateral ABCD. Let I be the center of ω. Show that
AI · BI · CI
BI 2 + = AB · BC.
DI

4. (Turkey 1998) Let ABC be a triangle. Suppose that the circle through C tangent to AB at A
and the circle through B tangent to AC at A have different radii, and let D be their second
intersection. Let E be the point on the ray AB such that AB = BE. Let F be the second
intersection of the rat CA with the circle through A, D, E. Prove that AF = AC.

5. A circle with center O passes through the vertices A and C of triangle ABC and intersects
segments AB and BC again at distinct points K and N , respectively. The circumcircles of
triangles ABC and KBN intersects at exactly two distinct points B and M . Prove that ∠OM B =
90◦ .

6. Circles ω1 and ω2 meet at points O and M . Circle ω, centered at O, meet circles ω1 and ω2 in
four distinct points A, B, C and D, such that ABCD is a convex quadrilateral. Lines AB and
CD meet at N1 . Lines AD and BC meet at N2 . Prove that N1 N2 ⊥ M O.

7. (Crux) Points O and H are the circumcenter and orthocenter of acute triangle ABC, respectively.
The perpendicular bisector of segment AH meets sides AB and AC at D and E, respectively.
Prove that ∠DOA = ∠EOA.

8. (IMO Shortlist 2000) Let ABCD be a convex quadrilateral with AB not parallel to CD, and let
X be a point inside ABCD such that ∠ADX = ∠BCX < 90◦ and ∠DAX = ∠CBX < 90. If the
perpendicular bisectors of segments AB and CD intersect at Y , prove that ∠AY B = 2∠ADX.

1
Trinity Training 2011 an ± 1 Yufei Zhao

an ± 1
Solutions
Yufei Zhao
Trinity College, Cambridge

April 2011

Practice problems:

1. A primitive root mod n is a number g such that the smallest positive integer k for which
g k ≡ 1 mod n is φ(n).

(a) Show that 2 is a primitive root mod 3n for any n ≥ 1.

(b) Show that if g is an odd primitive root mod p such that p2 - g p−1 − 1, then g is also
a primitive root mod pn and 2pn for any n ≥ 1.
n−1
Solution. (a) Since φ(3n ) = 2 · 3n−1 , the problem amounts to showing that 3n - 23 −1
n−2
and 3n - 22·3 − 1 (when n ≥ 2). The first claim follows from reduction mod 3, and the
n−2
second claim follows from the exponent lifting trick, as 3 k 22 −1, so that 3n−1 k 22·3 −1.
n−2
(b) Since φ(pn ) = φ(2pn ) = (p − 1)pn−1 , it suffices to show pn - g (p−1)p − 1 and
n−1
pn - g dp − 1 for any divisor d of p − 1 with d < p − 1. The first claim follows from
n−2
pn−1 k g (p−1)p − 1 by the exponent lifting trick as p k g p−1 − 1 by assumption, and
the second claim follows from the fact that p | g m − 1 if and only if (p − 1) | m as g is a
primitive root mod p.

2. (Cyclotomic polynomials) For a positive integer n, define the polynomial Φn (x) by

Y 2πik
Φn (x) = (x − e n ).
1≤k≤n
gcd(n,k)=1

= xn − 1, where the product is taken over

Q
(a) Prove the polynomial identity d|n Φd (x)
all divisors d of n.
(b) Prove that Φn (x) is an integer polynomial.
(c) Let m and n be positive integers, and let p be a prime divisor of Φn (m). Prove that
either p | n or n | p − 1.
(d) (Special case of Dirichlet’s theorem) Prove that for every positive integer n, there
are infinitely many primes p with p ≡ 1 (mod n).

2πik
Solution. (a) The right-hand side polynomial xn − 1 can be factored as nk=1 (x − e n ).
Q
2πik
For1 ≤ k ≤ n, each factor x − e n appears exactly once in the left hand side (in Φd (x)
n
for d = gcd(n,k) ) and all factors in the left hand side are of this form.
(b) Use induction on d. We have Φ1 (x) = x − 1. Suppose Φd (x) is an integer polynomial
for all d < n. Then by (a) Φn (x) is the quotient of two monic integer polynomials, and
hence it must also be an integer polynomial.
(c) Suppose p - n and n - p − 1. We have p | Φn (m) | mn − 1 by (a). So p - m, and
hence p | mp−1 − 1 by Fermat’s little theorem. Thus p | mgcd(p−1,n) − 1. Since n - p − 1,

1
Trinity Training 2011 an ± 1 Yufei Zhao

n k gcd(p−1,n) − 1.
gcd(p − 1, n) < n. Since p - n, we know p - gcd(p−1,n) . Suppose p k m By
the exponent lifting trick, we have pk
k mn − 1. However, from (a) we know that the
xn −1
polynomial Φn (x) is a factor of xgcd(p−1,n) −1 . Setting x = m gives us a contradiction, since
mn −1
on one hand we have p | Φn (m) but on the other hand p does not divide mgcd(p−1,n) −1
as
it is the quotient of two numbers both exactly divisible by p . k

(d) It suffices to show that for every positive integer n ≥ 2, there is at least one prime p
with p ≡ 1 (mod n), since then we can find infinitely many such p by finding primes pk
such that pk ≡ 1 (mod kn) for each k ≥ 1.
For n ≥ 2, |Φn (n)| > 1. Let p be a prime divisor of Φn (n). Since Φn (n) | nn − 1 by (a),
p - n, so n | p − 1 by part (b). This gives us the desired prime.

3. (IMO 2003) Let p be a prime number. Prove that there exists a prime number q such
that for every integer n, np − p is not divisible by q.
−1 p
Solution. Let q be a prime divisor of Φp (p) = pp−1 = pp−1 + pp−2 + · · · + p + 1 with
p2 - q − 1 (this must exist since Φp (p) 6≡ 1 (mod p2 )). By problem 2, p | q − 1. If np ≡ p
q−1 q−1
gcd( q−1 ,p)
(mod q), then p p ≡ nq−1 ≡ 1 (mod q). We have q | gcd(p p −1, pp −1) = p p −1,
2
which equals to p − 1 since p - q − 1. However, we cannot simultaneously have q | p − 1
and p | q − 1. Thus np − p is not divisible by q.

4. (a) Prove that Φm (x) and Φn (x) are always relatively prime as polynomials for m 6= n.
(b) Show that if for some integer x, Φm (x) and Φn (x) are not relative prime, then m/n
is an integer power of a prime.

Solution. (a) The zeros of Φn (x) and Φm (x) are distinct, since the zeros of Φn (x) are
precisely the primitive n-th roots of unity. Thus the polynomials are relatively prime.
(b) Suppose some prime p divides both Φm (x) and Φn (x). By replacing x by x + p if
necessary, we may assume that x > 1. Let us deal with the p = 2 case separately. We
claim that if Φm (x) is even then m must be a power of 2. Indeed, otherwise let q be an
odd prime divisor of m, and let m = qs, then by the previous problem, Φm (x) divides
xm −1 (q−1)s + x(q−2)s + · · · + xs + 1, which is always odd. The p = 2 case follows.
xs −1 = x
Now assume that p > 2. By the previous problem, p divides xm − 1 and xn − 1, and hence
m n
p | xgcd(m,n) − 1. Let pk k xgcd(m,n) − 1. One of gcd(m,n) and gcd(m,n) is not divisible by
p, and assume that it is the latter. Then by the exponent lifting trick, pk k xn − 1. If
xn −1
gcd(m, n) < n, then Φn (x) divides xgcd(m,n) −1
, which is not divisible by p by the above
analysis. This contradicts p | Φn (x). Hence gcd(m, n) = n, i.e., n | m.
We claim that m m
n is a power of p. If not, then pick some prime q dividing n . We have
p | Φn (x) | xn − 1 | xm/q − 1. By the exponent lifting trick, the same power of p divides
m
both xm − 1 and xm/q − 1. But Φm (x) divides xxm/q−1−1
, which contradicts p | Φm (x). Thus
m
n is a power of p.

5. Let p1 , p2 , . . . , pk be distinct primes greater than 3. Let N = 2p1 p2 ···pk + 1.

(a) (IMO Shortlist 2002) Show that N has at least 4n divisors.

k−1
(b) Show that N has at least 22 divisors. (Hint: use cyclotomic polynomials)

Solution. (a) Observe that if a and b are coprime odd numbers, then gcd(2a + 1, 2b + 1) =
3, since their gcd must divide gcd(22a − 1, 22b − 1) = 2gcd(2a,2b) − 1 = 22 − 1 = 3. Since
2ab + 1 is divisible by both 2a + 1 and 2b + 1, it must also be divisible by 31 (2a + 1)(2b + 1).

2
Trinity Training 2011 an ± 1 Yufei Zhao

We use induction on k. When k = 1, 2p1 + 1 is divisible by 3 and greater than 9, so it must

have at least 4 divisors. Let a = p1 · · · pk−1 and b = pk . Suppose that 2a + 1 has at least
4k−1 divisors. Since 2a + 1 is coprime to 31 (2b + 1), the number M = 13 (2a + 1)(2b + 1) must
have at least 2·4k−1 divisors (for each divisor d of 2a +1, we get two divisors d and 13 (2b +1)d
of M ). Also M | N and N = 2ab + 1 > M 2 (since 2ab + 1 > 2ab > 22(a+b+1) > M 2 ). So
N has at least 4k divisors (for each divisor d of M , we have divisors d and N/d). This
completes the induction.
(b) It suffices to show that N is divisible by at least 2k−1 distinct prime. We have
Q
p1 ···pk 22p1 ···pk − 1 d|2p ···p Φd (x)
Y
N =2 + 1 = p1 ···p = Q 1 k = Φ2d (2).
2 k − 1
d|p1 ···pk Φd (x)d|p1 ···pk

Consider the set of divisors d of p1 · · · pk with an odd number of prime factors. There are
2k−1 such divisors d, and they provide mutually coprime Φd (2) by Problem 4. Take one
prime divisor from each such Φd (2) and we get what we want.
2n + 1
6. (IMO 1990) Determine all positive integers n such that is an integer.
n2
Solution. We claim that the only solutions are n = 1, 3. Suppose n ∈ / {1, 3}. Let p be
the smallest prime divisor of n. Then p | 2n + 1, so p | 22n − 1. By Fermat’s little theorem,
we also have p | 2p−1 − 1. Thus p | 2gcd(p−1,2n) − 1. Since p is the smallest prime divisor
of n, we must have gcd(p − 1, 2n) = 2. So p | 22 − 1 and hence p = 3.
Suppose 3k k n. We have 3 k 22 − 1. So by the exponent lifting trick, 3k+1 k 22n − 1. If
n2 | 2n + 1, then 32k | 22n − 1. Thus 2k ≤ k + 1, hence k = 1. Thus 3 k n.
Let n = 3m. Suppose m 6= 1. Let q denote the smallest prime divisor of m. By the
same argument as above, we have q | 2gcd(q−1,6m) − 1, and gcd(q − 1, 6m) ∈ {2, 6}, so q
divides either 22 − 1 = 3 or 26 − 1 = 63 = 7 · 32 . Since 3 k n, q 6= 3, so q = 7. However,
2n + 1 = (23 )m + 1 ≡ 2 (mod 7), so 7 cannot divide 2n + 1, contradiction. This shows
that 1 and 3 are the only solutions.

7. (IMO 2000) Does there exist a positive integer N which is divisible by exactly 2000
different prime numbers and such that 2N + 1 is divisible by N ?

Solution. Yes. We will show by induction that for any m ≥ 1, there exists a positive
integer N divisibly by exactly m different prime numbers such that N | 2N + 1.
When m = 1, choose N = 3.
We will use the following variant of the exponent lifiting trick: if p is an odd prime, a ≥ 2,
k, m ≥ 1, ` ≥ 0, n odd, pk k a + 1, and p` k n, then pk+` k an + 1. This in fact follows
from our usual exponent lifting trick, as neither a − 1 nor an − 1 are divisible by p (since
a ≡ −1 (mod p) and n is odd), so the claim follows as pk k a2 − 1 implies pk+` k a2n − 1.
Now suppose N = pa11 · · · pamm satisfies N | 2N + 1, where p1 , . . . , pm are distinct prime
and ai ≥ 1. Suppose pbi i k 2N + 1 for each i. Write this as pb11 · · · pbmm k 2N + 1. Then
`
by above variant of the exponent lifting trick, we have pb11 +` pb22 · · · pbmm k 2N p1 + 1. For
` `
` sufficiently large, we also have pb11 +` pb22 · · · pbmm < 2N p1 + 1, so that 2N p1 + 1 has some
`
prime divisor pk+1 distinct from p1 , . . . pk . Then N p`1 pk+1 | 2N p1 pk+1 + 1, and hence we
can choose N 0 = N p`1 pk+1 to complete the induction.

8. Let N be a positive integer ending in digits 25, and m a positive integer. Prove that
for some positive integer n, the rightmost m digits of 5n and N agree in parity (i.e., for

3
Trinity Training 2011 an ± 1 Yufei Zhao

1 ≤ k ≤ m, the k-th digit from the right in n is odd if and only if the k-th digit from the
right in N is odd).

Solution. We will prove by induction on m that there exists infinitely many n that works.
This is trivial when m = 1, 2.
For the inductive step, it suffices to prove the following claim: if n ≥ m ≥ 2, then the
m−2
rightmost m digits of 5n and 5n+2 agree in parity, but the (m + 1)-th digit from the
right differ in parity.
m−2
By the exponent lifting trick, we have 2m k 52 − 1 as 22 k 5 − 1. It follows that
2m−2 +n n m m
5 − 5 is divisible by 10 but not 2 · 10 . The claim follows.

9. (Hensel’s lemma) Let

f (x) = cn xn + cn−1 xn−1 + · · · + c2 x2 + c1 x + c0

be a polynomial with integer coefficients. Its derivative f 0 is a polynomial defined by

f 0 (x) = ncn xn−1 + (n − 1)cn−1 xn−2 + · · · + 2c2 x + c1 .

Suppose that a ∈ Z satisfies p | f (a) and p - f 0 (a). Prove that for any integer k, there
exists an integer b satisfying pk | f (b) and p | b − a.

Solution. We use induction on k to find, for each k ≥ 1, an integer bk , satisfying b1 = a

and
bk+1 ≡ bk (mod pk )
and
f (bk ) ≡ 0 (mod pk ).
Note that this implies bk ≡ b1 = a (mod p).
When k = 1, we can just take b1 = a. Now assume that k > 1 and bk−1 has already been
chosen. Set
bk+1 = bk + pk r
for some integer r to be decided later. We have
n
X
f (bk+1 ) = f (bk + pk r) = cj (bk + pk r)j
j=0
k
cj (bjk + jpk rbj−1
X
k 0
≡ k ) = f (bk ) + p rf (bk ) (mod pk+1 ),
j=0

where the modulo equivalence comes from binomial expansion. (This is related to the
taylor expansion in calculus: f (x + ) ≈ f (x) + f 0 (x).) From the induction hypothesis,
we know pk | f (bk ). Also bk ≡ a (mod p), so p - f 0 (bk ), and hence f 0 (bk ) has an inverse
mod p, say t ∈ Z, satisfying f 0 (bk )t ≡ 1 (mod p). Then setting r = − f (bpkk )t , we have

f (bk+1 ) ≡ f (bk ) − f (bk )rf 0 (bk ) = f (bk )(1 − rf 0 (bk )) ≡ 0 (mod pk+1 ).

since pk | f (bk ) and p | 1 − rf 0 (bk ). This completes the induction step.

4
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Lecture 6 : Divisibility and the Euclidean Algorithm

Yufei Zhao
July 24, 2007

1. If a and b are relatively prime integers, show that ab and a + b are also relatively prime.

2. (a) If 2n + 1 is prime for some integer n, show that n is a power of 2.

(b) If 2n − 1 is prime for some integer n, show that n is a prime.
12n + 1
3. Show that the fraction is irreducible for all positive integers n.
30n + 2
4. Let x, a, b be positive integers, show that gcd(xa − 1, xb − 1) = xgcd(a,b) − 1.

5. (a) Let p be a prime number. Determine the greatest power of p that divides n!, where n is
a positive integer.
(m + n)!
(b) Let m and n be positive integers. Show that is an integer (without referring
m!n!
to binomial coefficients).

6. (USAMO 1972) Show that

gcd(a, b, c)2 lcm(a, b, c)2

= .
gcd(a, b) gcd(b, c) gcd(c, a) lcm(a, b) lcm(b, c) lcm(c, a)

7. (a) Show that if a and b are relatively prime integers, then gcd(a + b, a2 − ab + b2 ) = 1 or 3.
(b) Show that if a and b are relatively prime integers, and p is an odd prime, then

ap + bp

gcd a + b, = 1 or p.
a+b

8. Let n be a positive integer.

(a) Find n consecutive composite numbers.

(b) Find n consecutive positive integers, none of which is a power of a prime.

9. Let n > 1 be a positive integer. Show that

1 1 1
1+ + + ··· +
2 3 n
is not an integer. (Try not to use any powerful results about the distribution of prime
numbers.)

1
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Problem Solving Session

July 24, 2007

1. Let a, b be positive integers. Show that gcd(a, b) lcm(a, b) = ab.

a
2. Let a, b, c be positive integers. Show that a divides bc if and only if gcd(a,b) divides c.

21n + 4
3. Show that the fraction is irreducible for all positive integers n.
14n + 3
4. Let n be a positive integer. Find gcd(n! + 1, (n + 1)!).

5. Find all positive integers d such that d divides both n2 + 1 and (n + 1)2 + 1 for some integer
n.

6. Let a and b be positive integers such that a | b2 , b2 | a3 , a3 | b4 , b4 | a5 , . . . . Prove that a = b.

7. Let n ≥ 2 and k be positive integers. Prove that (n − 1)2 | (nk − 1) if and only if (n − 1) | k.

8. (AIME 1986) What is the largest positive integer n for which n3 + 100 is divisible by n + 10?
(2m)!(2n)!
9. Let m and n be positive integers. Show that is an integer.
(m + n)!m!n!
10. Prove that n2 + 3n + 5 can never be a multiple of 121 if n is a positive integer.

11. Let a and b > 2 be positive integers. Show that 2a + 1 is not divisible by 2b − 1.

12. Let a, b, n > 1 be positive integers. Show that an + bn is not divisible by an − bn .

n m
13. Prove that if m > n then a2 + 1 is a divisor of a2 − 1. Show that if a, m, n are positive
with m 6= n, then (
m n 1 if a is even,
gcd(a2 + 1, a2 + 1) =
2 if a is odd.

14. Let n > 1 be a positive integer. Show that

1 1 1
1+ + + ···
3 5 2n − 1
is not an integer.

2
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Lecture 7 : Fermat, Euler, and Wilson

Yufei Zhao
July 25, 2007

Notation: Let Zn = {0, 1, 2, . . . , n − 1} denote the complete residue system mod n, and let
Z∗n = {d | d ∈ Zn , gcd(d, n) = 1} denote the reduced residue system mod n.

1. Fermat’s little theorem. Let p be a prime number.

p
(a) Show that if k is an integer with 0 < k < p, then is divisible by p.
k
(b) Show that if a ∈ Z, then (a + 1)p ≡ ap + 1 (mod p).
(c) Show that if a ∈ Z, then ap ≡ a (mod p).
(d) Show that if a is an integer not divisible by p, then ap−1 ≡ 1 (mod p).

2. Let a and m be relatively prime positive integers. Show that the following two sets are
identical in mod m:

{a, 2a, 3a, . . . , (m − 1)a} and {1, 2, 3, . . . , m − 1}.

3. Let a and m be relatively prime positive integers. Show that there exists an integer x such
that ax ≡ 1 (mod m).
We say that x is the multiplicative inverse (or just inverse) of a in mod m, denoted by a−1
when the context is clear.

4. Another look at Fermat’s little theorem. Let p be a prime number, and a an integer
not divisible by p.

(a) Show that {a, 2a, 3a, . . . , (p − 1)a} ≡ {1, 2, 3, . . . , p − 1} (mod p).
(b) Show that a · 2a · 3a · · · · · (p − 1)a ≡ 1 · 2 · 3 · · · · · (p − 1) (mod p).
(c) Conclude that ap−1 ≡ 1 (mod p).

5. Euler’s totient function. We use φ(n) to denote the number of elements in {1, 2, . . . , n}
that are relatively prime to n. That is, φ(n) = |Z∗n |.

(a) Compute φ(7) and φ(24).

(b) Compute φ(pn ), where p is a prime and n is a positive integer.
(c) Show that if m and n are relatively prime integers, then φ(mn) = φ(m)φ(n).
(d) Find a formula for computing φ(n) in terms of the prime factorization of n.

6. Euler’s Theorem Let a and m be relatively prime integers.

(a) Let Z∗n = {r1 , r2 , . . . , rφ(n) } be the set of positive integers less than m and relatively
prime to m. Show that

{r1 , r2 , . . . , rφ(n) } ≡ {ar1 , ar2 , . . . , arφ(n) } (mod m).

3
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

(b) Show that aφ(m) ≡ 1 (mod m).

7. Wilson’s Theorem Let p be a prime number.

(a) Show that the set of residues {2, 3, . . . , p−2} can be paired up into multiplicative inverses.
(b) Show that (p − 1)! ≡ −1 (mod p).

8. Let p > 2 be a prime number.

(a) Suppose that p ≡ 1 (mod 4), show that x2 ≡ −1 (mod p) has a solution. (Hint: use
Wilson’s theorem)
(b) Suppose that x2 ≡ −1 (mod p) has a solution, show that p ≡ 1 (mod 4). (Hint: use
Fermat’s little theorem.)

9. Let n be a positive integer. Show that all divisors of 4n2 + 1 have the form 4k + 1 for some
integer k.

4
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Problem Solving Session

July 25, 2007

1. Let p be a prime number. If x is an integer, then show that x2 ≡ 1 (mod p) if and only if
x ≡ ±1 (mod p).

2. Let n be a positive integer. Show that if (n − 1)! ≡ −1 (mod n), then n is prime.

3. Let p, q be distinct prime numbers. Show that every integer a satisfies the congruence
apq−p−q+2 ≡ a (mod pq).

4. RSA public-key cryptography. Alice and Bob are sending cryptic messages to each other.
Let p and q be distinct primes and n = pq and t = (p − 1)(q − 1). Let e, d be positive integers
such that ed ≡ 1 (mod t). Alice takes a message, M (an integer relatively prime to n, and
sends C = M e to Bob. Bob receives C and computes M 0 = C d (mod n). Prove that M ≡ M 0
(mod n).

5. Let m be an even positive integer. Assume that

{a1 , a2 , . . . , am } and {b1 , b2 , . . . , bm }

are two complete sets of residue classes modulo m. Prove that

{a1 + b1 , a2 + b2 , . . . , am + bm }

is not a set of complete residue classes.

6. Let p ≥ 3 be a prime, and let

{a1 , a2 , . . . , ap } and {b1 , b2 , . . . , bp }

be two sets of complete residue classes modulo p. Prove that

{a1 b1 , a2 b2 , . . . , ap bp }

is not a complete set of residue classes modulo p.

7. Find all non-negative integer solutions to 4ab − a − b = c2 .

Q of integers x, 1 ≤ x ≤ n, such that both x

8. For an odd positive integer n > 1, let S be the set
and x + 1 are relatively prime to n. Show that x∈S x ≡ 1 (mod n).

5
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Lecture 8 : Residue Classes

Yufei Zhao
July 26, 2007

1. Wilson’s Theorem. Let p be a prime number.

(a) Show that the set of residues {2, 3, . . . , p−2} can be paired up into multiplicative inverses.
(b) Show that (p − 1)! ≡ −1 (mod p).

2. Let p > 2 be a prime number.

(a) Suppose that p ≡ 1 (mod 4), show that x2 ≡ −1 (mod p) has a solution.
(Hint: use Wilson’s theorem)
(b) Suppose that x2 ≡ −1 (mod p) has a solution, show that p ≡ 1 (mod 4).
(Hint: use Fermat’s little theorem.)

3. Let n be a positive integer. Show that all divisors of 4n2 + 1 have the form 4k + 1 for some
integer k.

4. Chinese remainder theorem

(a) If m and n are relatively prime integers greater than one, and a and b are arbitrary
integers, then show that there exists an integer x such that

x ≡ a (mod m)
x ≡ b (mod n)

(b) If m1 , m2 , . . . , mk are pairwise relatively prime integers greater than one, and a1 , a2 , . . . , ak
are arbitrary integers, then show that there exists an integer x such that

x ≡ ai (mod mi ), i = 1, 2, . . . , k.

5. Euler’s totient function. We use φ(n) to denote the number of elements in {1, 2, . . . , n}
that are relatively prime to n. That is, φ(n) = |Z∗n |.

(a) Compute φ(7) and φ(24).

(b) Compute φ(pn ), where p is a prime and n is a positive integer.
(c) Show that if m and n are relatively prime integers, then φ(mn) = φ(m)φ(n).
(d) Find a formula for computing φ(n) in terms of the prime factorization of n.

6. (a) Let p be a prime such that p = x2 + y 2 for some integers x and y. Show that p = 2 or
p ≡ 1 (mod 4).
(b) Let p be a prime such that p ≡ 1 (mod 4). Then there exist positive integers x and y
such that p = x2 + y 2 .
(Hint: let a be an integer such that a2 ≡ −1 (mod p), and then consider the set of
√
integers of the form ax − y, where 0 ≤ x, y < p. Use the pigeonhole principle.)

6
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Problem Solving Session

July 26, 2007

1. Let n be a positive integer. Find gcd(n! + 1, (n + 1)!).

2. Let x, y be integers. Show that 2x + 3y is divisible by 7 if and only if 5x + 4y is divisible by

7.

3. Let p ≥ 3 be a prime, and let

{a1 , a2 , . . . , ap } and {b1 , b2 , . . . , bp }

be two sets of complete residue classes modulo p. Prove that

{a1 b1 , a2 b2 , . . . , ap bp }

is not a complete set of residue classes modulo p.

4. Find all non-negative integer solutions to 4ab − a − b = c2 .

5. For any prime p, if ap ≡ bp (mod p), prove that ap ≡ bp (mod p2 ).

6. Let p be a prime number, and suppose that a is an integer such that a2 ≡ −2 (mod p). Show
that at least one of the equations x2 + 2y 2 = p, x2 + 2y 2 = 2p has a solution.

Q of integers x, 1 ≤ x ≤ n, such that both x

7. For an odd positive integer n > 1, let S be the set
and x + 1 are relatively prime to n. Show that x∈S x ≡ 1 (mod n).

7
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Lecture 9 : Equations and Polynomials

Yufei Zhao
July 27, 2007

1. (a) Let p be a prime such that p = x2 + y 2 for some integers x and y. Show that p = 2 or
p ≡ 1 (mod 4).
(b) Let p be a prime such that p ≡ 1 (mod 4). Then there exist positive integers x and y
such that p = x2 + y 2 .
(Hint: let a be an integer such that a2 ≡ −1 (mod p), and then consider the set of
√
integers of the form ax − y, where 0 ≤ x, y < p. Use the pigeonhole principle.)

2. Show that 15x2 − 7y 2 = 9 has no integer solutions.

3. Show that the only integer solution to

x2 + y 2 + z 2 = 2xyz

is x = y = z = 0.

4. Let p be a prime number. Let f (x) be a polynomial with integer coefficients, such that the
leading coefficient of f is nonzero. Prove that f (x) ≡ 0 (mod p) has at most deg f solutions
modulos p.

5. Let p be a prime number.

(a) Show that all the coefficients of the polynomial

(x + 1)(x + 2) · · · (x + p − 1) − xp−1 + 1

are divisible by p.

P of 1, 2, . . . , p − 1 taken
(b) For all positive integers i, let σi denote the sum of the products
i at a time. For example, σ1 = 1 + 2 + · · · + (p − 1), σ2 = 1≤i<j≤p−1 ij. Show that
σ1 , σ2 , . . . , σp−2 are all divisible by p.
(c) Prove Wilson’s Theorem: (p − 1)! ≡ −1 (mod p).

6. Wolstenholme’s theorem. Let p > 3 be a prime number.

1 1 1
(a) Show that the numerator of 1 + + + ··· + is divisible by p.
22 32 (p − 1)2
1 1 1
(b) Show that the numerator of 1 + + + · · · + is divisible by p2 .
2 3 (p − 1)

8
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Problem Solving Session

July 27, 2007

1. Prove that the only solution in rational numbers of the equation

x3 + 3y 3 + 9z 3 − 9xyz = 0

is x = y = z = 0.

2. Find all triples of integers (x, y, z) such that

x2 + y 2 + z 2 = 2007.

3. Find all integer solutions to x2 + y 2 + z 2 = 7w2 .

4. Let p be a prime number, and suppose that there exists an integer a such that a2 ≡ −2
(mod p). Show that at least one of the equations x2 + 2y 2 = p, x2 + 2y 2 = 2p has a solution.

5. Let n be a positive integer. Show that there exist integers x and y such that n = x2 + y 2 if
and only if each prime factor of n of the form 4k + 3 appears an even number of times.

2p − 1
6. Let p ≥ 5 be a prime number. Show that ≡ 1 (mod p3 ).
p−1
7. (APMO 2006) Let p ≥ 5 be a prime and let r be the number of ways of placing p checkers
on a p × p checkerboard so that not all checkers are in the same row (but they may all be in
the same column). Show that r is divisible by p5 . Here, we assume that all the checkers are
identical.

9
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Lecture 10 : Order of an Element

Yufei Zhao
July 28, 2007

1. Let m > 1 be a positive integer, and let a be an integer relatively prime to m. Show that
there is a least positive integer d for which ad ≡ 1 (mod m).
We say that d is the order of a modulo m, denoted by ordm (a) or simply ord(a) is the modulus
m is understood.

2. Let m be a positive integer, and a an integer relatively prime to m.

(a) Show that an ≡ 1 (mod m) if and only if ordm (a) | n.

(b) Furthermore, show that an0 ≡ an1 (mod m) if and only if ordm (a) | n0 − n1 .
(c) Show that ordm (a) | φ(m).

3. Show that the order of 2 modulo 101 is 100.

4. Prove that for all positive integers a > 1 and n, we have n | φ(an − 1).

5. Prove that if p is a prime, then every prime divisor of 2p − 1 is greater than p.

6. Prove that if p is a prime, then pp − 1 has a prime factor of the form kp + 1.

10
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Evaluation Test 2
Yufei Zhao
July 28, 2007

1. (a) [3] State Fermat’s little theorem.

(b) [7] Prove Fermat’s little theorem.

2. [10] Show that if a and b are relatively prime positive integers, then there exist integers m
and n such that am + bn ≡ 1 (mod ab).

3. [10] Let a and b > 2 be positive integers. Show that 2a + 1 is not divisible by 2b − 1.

4. [10] Show that for every prime number p, we can find some positive integer n so that

2n + 3n + 6n − 1

is divisible by p.

11
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

Evaluation Test 2
Solutions
Yufei Zhao
July 28, 2007

1. (a) [3] State Fermat’s little theorem.

Solution: If p is a prime number, and a is an integer not divisible by p, then
ap−1 ≡ 1 (mod p).

(b) [7] Prove Fermat’s little theorem.

Solution: Consider the two sets
{1, 2, 3, . . . , p − 1} and {a, 2a, 3a, . . . , (p − 1)a}.
We claim that the elements in the second set are simply a permutation of the elements in
the first set. This is because no two i, j gets taken to the same residue when multiplied
by a, as ia ≡ ja implies that p | (i − j)a, which implies that p|i − j, which implies that
i = j (as 1 ≤ i, j < p). Furthermore, none of the elements in the second set is divisible
by p. Since there are only p − 1 nonzero residues in mod p, the second set must then be
the same as the first set.
Then, by multiplying together all the elements in each set, we obtain that
1 · 2 · 3 · · · · · (p − 1) ≡ a · 2a · 3a · · · · · (p − 1)a ≡ 1 · 2 · 3 · · · · (p − 1) · ap−1 (mod p).
Since 1 · 2 · 3 · · · · · (p − 1) is not divisible by p, it follows that ap−1 ≡ 1 (mod p).
Alternate solution: We will prove that ap ≡ a for all a ∈ {0, 1, 2, . . . , p − 1} by
induction on a. For a = 0, the result is clear. Assume ap ≡ a (mod p). Then
p−1
p p
X p k
(a + 1) = a + 1 + a ≡ ap + 1 ≡ a + 1 (mod p).
k
k=1

Here we used the fact that kp is divisible by p for 1 ≤ k ≤ p − 1 (this is true since in

p!
the expansion kp = (p−k)!k!

, the factor p in the numerator cannot get canceled by any
factors in the denominator). Therefore, ap ≡ a (mod p) implies that (a + 1)p ≡ a + 1,
and the induction is complete.
Since ap ≡ a (mod p) is true for all residue classes mod p, it must be true for all integers
a. Furthermore, if a is not divisible by p, then ap ≡ a (mod p) implies that ap−1 ≡ 1,
as desired.

2. [10] Show that if a and b are relatively prime positive integers, then there exist integers m
and n such that am + bn ≡ 1 (mod ab).

Solution: Set m = φ(b) and n = φ(a). Then am ≡ 1 (mod b) and bn ≡ 1 (mod a) by Euler’
theorem. It follows that am + bn − 1 is divisible by both a and b, and so it’s divisible by ab
(since a and b are relatively prime). Therefore, am + bn ≡ 1 (mod ab).

12
AwesomeMath 2007 Track 4 — Modulo Arithmetic Week 2

3. [10] Let a and b > 2 be positive integers. Show that 2a + 1 is not divisible by 2b − 1.

Solution: Suppose that there exists such a pair a, b so that 2a + 1 is divisible by 2b − 1. Let
a = qb + r, where q, r are integers and 0 ≤ r < b. Note that 2b ≡ 1 (mod 2b − 1). So

2a + 1 = 2qb+r + 1 = (2b )q · 2r + 1 ≡ (1)q · 2r + 1 = 2r + 1 (mod 2b − 1).

It follows that 2b − 1 divides 2r + 1, so 2b − 1 ≤ 2r + 1. However, since r < b, we have

2r + 1 ≤ 2b−1 + 1. Combining the two inequalities, we get 2b − 1 ≤ 2b−1 + 1, and thus
2b−1 ≤ 2, so b ≤ 2, which contradicts the hypothesis that b > 2.

4. [10] Show that for every prime number p, we can find some positive integer n so that

2n + 3 n + 6 n − 1

is divisible by p.

Solution: If p = 2 or p = 3, then we can choose n = 2.

Otherwise, choose n = p − 2. Since none of 2, 3, 6 is divisible by p, using Fermat’s little
theorem, we have

6(2p−2 + 3p−2 + 6p−2 − 1) = 3 · 2p−1 + 2 · 3p−1 + 6p−1 − 6 ≡ 3 + 2 + 1 − 6 = 0 (mod p)

It follows that 6(2p−2 + 3p−2 + 6p−2 − 1) is divisible by p, and therefore 2p−2 + 3p−2 + 6p−2 − 1
is divisible by p.

Remarks: This problem is related to the identity 21 + 31 + 16 − 1 = 0. Fermat’s little theorem

tells us that the inverse of an element can be found by a−1 ≡ ap−2 (mod p).

13