Sequences

Winter Camp 2020

1 Example Problems
Problem 1. (IMO Shortlist 2015) Determine all positive integers M such that the sequence a0 , a1 , a2 , · · ·
defined by
1
a0 = M + and ak+1 = ak bak c for k = 0, 1, 2, · · ·
2
contains at least one integer term.
Proof. We again start with small cases. We note that for M = 1, we have a0 = 1.5, a1 = 1.5, . . . ,
which is never an integer, but for M = 2, we have a1 = 5; for M = 3, we have a2 = 110; for M = 4,
we have a1 = 18, for M = 5, we have a2 = 742.5 and hence a3 = 371 ∗ 1485. In particular, we note
that whenever ai is of the form 2K + 0.5 for some integer K, then ai+1 is an integer. This begs
the question: what conditions on ai−1 will force ai = 2K + 0.5? Well, since we know that ai−1 is a
half-integer, let’s let ai−1 = x + 0.5, and then we have ai = x2 + 0.5x = x2 + 0.5(x − 1) + 0.5. Thus,
we have our desired condition when x is odd and x2 + 0.5(x − 1) is even; that is, when 2x2 + x − 1
is divisible by 4. Since the first term is always 2 mod 4, this occurs when x is 3 mod 4. If we
take this one step further, we can see that if ai = (4K + 3) + 0.5 for some positive integer K, then
ai−1 = x + 0.5 when x is 5 mod 8. This suggests us to consider the number of powers of 2 dividing
M − 1.
More formally, assume M − 1 = 2N · P , where N is a nonnegative integer and P is odd. Then,
1 1 1
a1 = M (M + ) = + 22N · P 2 + 5 · 2N −1 · P + 1 = + 1 + (2N −1 )(2N +1 P 2 + P ),
2 2 2
and we note that the number of powers of 2 dividing 2N −1 (2N +1 P 2 + P ) is precisely N − 1 because
P is odd by assumption. This means that we have reduced the number of powers of 2 dividing
M − 1 after going one step forward, which means that as long as M 6= 1, we will eventually reach
an integer.

We work through a few example problems about sequences, which are typically problems where
a sequence satisfies some set of conditions, and for which one desires to prove another condition
about the sequence.
Problem 2. (IMO Shortlist 2001) Let a0 , a1 , a2 , . . . be an arbitrary infinite sequence of positive
numbers. Show that the inequality 1 + an > an−1 (1 + n−1 ) holds for infinitely many positive
integers n.
Proof. For this problem, the sequence does not have any conditions at all, so it is definitely impos-
sible to force the inequality to hold for any specific number n. Indeed, given any finite subset S

1
of the natural numbers, it is also impossible to force the inequality to hold for at least one of the
indices in S. This motivates an approach by contradiction: we assume that there is some sequence
ai where the inequality only holds for finitely many integers n, and then attempt to arrive at a
contradiction.
This is now equivalent to asserting that the inequality is eventually false; that is, there is some
N for which the inequality is false for n > N . Thus, we have reduced the problem to showing that
there exists no positive sequence for which

1 + an ≤ an−1 (1 + n−1 )

for all n > N . Now note that given any such sequence ai , we may consider the sequence bi for
which aN = bN and 1 + bn = bn−1 (1 + n−1 ) for all n > N . The motivation here is that we are no
worse off than before, because we can show that bi > ai for all i > N inductively, and so if we can
force a to be eventually negative, we must be able to do the same for b.
We have now reduced the problem to analyzing the recursion bn = bn−1 (1 + n−1 ) − 1 starting
with some initial condition for bN and showing that the terms are eventually negative. We now
simply write out the closed form for bN +k :
k
Y k X
X k
bN +k = (1 + (N + i)−1 )bN − (1 + (N + j)−1 )
i=1 i=1 j=i

k
N +1+k XN +1+k
= bN − .
N +1 i=1
N +j
We will now arrive at a contradiction if we can find some k for which bN +k is negative. But this
intuitively should happen because the first term in the equation above increases linearly in k and
the second term increases by an arbitrarily large amount when k increases by 1.
Problem 3. (Turkey TST 2016) A sequence of real numbers a0 , a1 , . . . satisfies the condition
m  
X m
an (−1)n =0
n=0
n

for all large enough positive integers m. Prove that there exists a polynomial P such that an = P (n)
for all n ≥ 0.
Proof. We start by defining N such that
m  
X m
an (−1)n =0
n=0
n

for all m > N . Now, we know that these equations determine am inductively for all m > N given
a0 , a1 , . . . , aN . Intuitively, this means that we may start with a polynomial P such that an = P (n)
for n ≤ N and hope that this polynomial also satisfies the condition for n > N . However, we
know that we need some more constraints on the polynomial P : for instance, you can always find
a polynomial P that takes on any finite number of values. So we have to figure out what the other
constraints are.

2
 If we try doing the problem for N = 1, then we have

a2 = 2a1 − a0

a3 = 3a2 − 3a1 + a0 = 3a1 − 2a0 ,

a4 = 4a3 − 6a2 + 4a1 − a0 = 4a1 − 3a0 ,
which suggests that the correct polynomial is linear: P (n) = n(a1 − a0 ) + a0 . Similarly, for N = 2,
we observe
a3 = 3a2 − 3a1 + a0 ,
a4 = 4a3 − 6a2 + 4a1 − a0 = 6a2 − 8a1 + 3a0 ,
a5 = 5a4 − 10a3 + 10a2 − 5a1 + a0 = 10a2 − 15a1 + 6a0 ,
suggesting that the correct polynomial is quadratic:

(n − 1)(n − 2) n(n − 1)
P (n) = a0 − n(n − 2)a1 + a2 ,
2 2
where I have written the formula in the above form to more clearly satisfy that P (n) = an for
n = 0, 1, 2.
These two small cases suggest that we want our polynomial P to be of degree at most N in
the general case. It turns out that there is a unique polynomial of degree at most N such that
P (n) = an for n ≤ N (going through N + 1 points), but we only need the existence. Call such a
polynomial P . We now wish to show that P (N + 1) = aN +1 , and this would finish the problem
because we now have a polynomial with degree at most N + 1 going through the first N + 2 points
and we can apply the same procedure again.
Put more precisely, we wish to show that whenever P is a polynomial of degree at most N , we
have
N +1  
n N +1
X
P (n)(−1) = 0.
n=0
n
In essence, we have reduced the problem from being one about sequences with lots of conditions
to a problem about polynomials. We can now finish the proof by induction on N : for instance, we
can rewrite
N +1   X N  
X N +1 N
P (n)(−1)n = (P (n + 1) − P (n))(−1)n+1
n=0
n n=0
n

by noting that Nn+1 = n−1 N

+ N




n , and then noting that P (n + 1) − P (n) is a polynomial of
degree at most N − 1 in n.

3
2 Olympiad Problems
1. (Belarus TST 2019) Let the sequence (an ) be constructed in the following way:
1
a1 = 1, a2 = 1, an+2 = an+1 + , n = 1, 2, . . . .
an
Prove that a180 > 19.
2. (Vietnam 2018) The sequence (xn ) is defined as follows:
√ √
x1 = 2, xn+1 = xn + 8 − xn + 3

for all n ≥ 1. For every n ≥ 1, prove that

n ≤ x1 + x2 + · · · + xn ≤ n + 1.

3. (IMO Shortlist 2004) Find all functions f : N → N satisfying

2
f (m)2 + f (n) | m2 + n

for any two positive integers m and n.

4. (Taiwan 2000) Define a function f : N → N0 by f (1) = 0 and

f (n) = max{f (j) + f (n − j) + j} ∀n ≥ 2

j

Determine f (2000).
5. (Belarus 2017) Find all positive real numbers α such that there exists an infinite sequence of
positive real numbers x1 , x2 , ..., such that
p
xn+2 = αxn+1 − xn

for all n ≥ 1.
6. (IMO 2014) Let a0 < a1 < a2 . . . be an infinite sequence of positive integers. Prove that there
exists a unique integer n ≥ 1 such that
a0 + a1 + · · · + an
an < ≤ an+1 .
n

7. (Turkey 2019) Let f : {1, 2, . . . , 2019} → {−1, 1} be a function, such that for every k ∈
{1, 2, . . . , 2019}, there exists an ` ∈ {1, 2, . . . , 2019} such that
X
f (i) 6 0.
i∈Z:(`−i)(i−k)>0

Determine the maximum possible value of

X
f (i).
i∈Z:16i62019

4
 8. (Taiwan 1999) Let a1 , a2 , ..., a1999 be a sequence of nonnegative integers such that for any i, j
with i + j ≤ 1999 , ai + aj ≤ ai+j ≤ ai + aj + 1. Prove that there exists a real number x such
that an = [nx]∀n.
9. (IMO Shortlist 2008) Let a1 , a2 , . . ., an be distinct positive integers, n ≥ 3. Prove that there
exist distinct indices i and j such that ai + aj does not divide any of the numbers 3a1 , 3a2 ,
. . ., 3an .
10. (USA 2007) Let n be a positive integer. Define a sequence by setting a1 = n and, for each
k > 1, letting ak be the unique integer in the range 0 ≤ ak ≤ k − 1 for which a1 + a2 + ... + ak
is divisible by k. For instance, when n = 9 the obtained sequence is 9, 1, 2, 0, 3, 3, 3, .... Prove
that for any n the sequence a1 , a2 , ... eventually becomes constant.
11. (IMO Shortlist 2015) Suppose that a0 , a1 , · · · and b0 , b1 , · · · are two sequences of positive
integers such that a0 , b0 ≥ 2 and

an+1 = gcd (an , bn ) + 1, bn+1 = lcm (an , bn ) − 1.

Show that the sequence an is eventually periodic; in other words, there exist integers N ≥ 0
and t > 0 such that an+t = an for all n ≥ N .
12. (IMO Shortlist 2004) Let a0 , a1 , a2 , ... be an infinite sequence of real numbers satisfying the
equation an = |an+1 − an+2 | for all n ≥ 0, where a0 and a1 are two different positive reals.
Prove that the sequence is unbounded.
13. (Turkey 2005) Suppose that a sequence (an )∞n=1 of integers has the following property: For all
sufficiently large n, an equals the number of indices i, 1 ≤ i < n, such that ai + i ≥ n. Find
the maximum possible number of integers which occur infinitely many times in the sequence.
14. (IMO 2015) The sequence a1 , a2 , . . . of integers satisfies the conditions:
(i) 1 ≤ aj ≤ 2015 for all j ≥ 1, (ii) k + ak 6= ` + a` for all 1 ≤ k < `.
Prove that there exist two positive integers b and N for which
 
 n
 X 
(aj − b) ≤ 10072



j=m+1 

for all integers m and n such that n > m ≥ N .

15. (ISL 2012) Let f : N → N be a function, and let f m be f applied m times. Suppose that for
every n ∈ N there exists a k ∈ N such that f 2k (n) = n + k, and let kn be the smallest such
k. Prove that the sequence k1 , k2 , . . . is unbounded.
16. (China 2011) Given integer n ≥ 2 and real numbers x1 , x2 , · · · , xn in the interval [0, 1]. Prove
that there exist real numbers a0 , a1 , · · · , an satisfying the following conditions:
• a0 + an = 0;
• |ai | ≤ 1, for i = 0, 1, · · · , n;
• |ai − ai−1 | = xi , for i = 1, 2, · · · , n.

5
3 One-variable Linear Recurrences
A special setup is the one-variable linear recurrence, which has a closed form solution. The following
few problems are designed to derive the full solution for a one-variable linear recurrence.
Definition 1. A one-variable linear recurrence of degree j is a sequence a1 , a2 , . . . , satisfying the
following property for n ≥ j:

an = cj−1 an−1 + cj−2 xn−2 + · · · + c0 an−j ,

such that c0 , c1 , . . . , cj−1 are constants and c0 6= 0.

Our goal is to determine an in terms of aj , aj−1 , . . . , a1 in a simple manner. We will work
through some simple examples and gradually increase the generality until we can solve the problem
in the generality of Definition 1.
Exercise 1. You already know how to solve one-variable linear recurrences of degree 1. Similarly,
you know how to solve a special type of linear recurrence of degree 2: when we have the recurrence
an = c0 an−2 . Now, let’s consider the recurrence

an = 3an−1 − 2an−2 , (1)

and let’s try to do the same thing. We add the equation

can−1 = 3can−2 − 2can−3 ,

to (1) get the equation

an − (3 − c)an−1 = (3c − 2)an−2 − 2can−3 . (2)
When c = 2, can we separate the recurrence (2) into two second-order recurrences? In particular,
what if we start with the initial values a1 = 1, a2 = 2?
Exercise 2. Is there another value of c for which we can do the same thing? What initial values
are needed there?
Exercise 3. The word linear in the phrase one-variable linear recurrence comes from the fact
that the recurrence relations are linear equations in the ai . In particular, this means that given
two solutions a1 , a2 , . . . and b1 , b2 , . . . satisfying the same one-variable linear recurrence, for any
constants c and d, we have that the sequence ca1 + db1 , ca2 + db2 , . . . satisfies the same recurrence.
There is also nothing special about there being only two solutions; the same property holds for any
linear combination of solutions of the same one-variable linear recurrence.
Going back to (2), let’s try to solve the equation given general a1 and a2 . What if we let
a1 = (2a1 − a2 ) + (a2 − a1 ) and a2 = (2a1 − a2 ) + 2(a2 − a1 )? Can you find recurrences bi and ci
that you know how to solve, and that add up to ai ?
Exercise 4. The Fibonacci sequence is given by the recurrence

an = an−1 + an−2 ,

with initial values a1 = 1, a2 = 2. Use the same procedure as Exercises 1 - 3 to get the closed form
for the Fibonacci sequence.

6
Exercise 5. There is another way to derive the same closed form. Given the recurrence relation
in (1), we may rewrite it as
an − 2an−1 = an−1 − 2an−2 .
This suggests the substitution bn−1 = an − 2an−1 , such that the recurrence relation for the bi is
bn = bn−1 , and the initial value is b1 = a2 − 2a1 . We can then get a closed form for the bi . From
here, how do we get the closed form for the ai ?
Does anything change if we rewrite Equation (1) as

an − an−1 = 2(an−1 − an−2 ), (3)

and substitute bn−1 = an − an−1 ?

Exercise 6. What happens when we have the recurrence an = 4an−1 − 4an−2 ? Note that the
procedure in Exercises 1 - 3 do not generalize to this case, but we can still do it with a procedure
similar to that of 5. What is the closed form of the recurrence?

Exercise 7. To summarize the intuition from the previous exercises, we note that in Equation (1),
we have closed forms bi = 1 and ci = 2i−1 for initial conditions (b1 , b2 ) = (1, 1) and (c1 , c2 ) = (1, 2).
Using linearity, we can extend these two special solutions to a solution for all such initial conditions.
But for (3), we do not have these closed forms. Can you come up with other simple closed forms
for special initial values?

Exercise 8. Now let’s take the general recursion

an = cj−1 an−1 + cj−2 xn−2 + · · · + c0 an−j ,

and define the characteristic polynomial of this recurrence as the polynomial

p(x) = xj − cj−1 xn−1 − cj−2 xn−2 − · · · − c0 .

We let this polynomial have roots r1 , r2 , . . . , rk with multiplicity m1 , m2 , . . . , mk , so that

k
Y
p(x) = (x − ri )mi .
i=1

Can you find j distinct j-tuples of values (a1 , a2 , . . . , aj ) for which the recurrence admits a closed
form? For example, for any ri , the j-tuple (1, ri , ri2 , . . . , rij−1 ) is one such tuple.
Exercise 9. It remains to show that for any tuple of initial values (a1 , a2 , . . . , aj ), we can find some
linear combination of the j j-tuples of initial values above that sum up to (a1 , a2 , . . . , aj ). Prove
this.
Exercise 10. Given a one-variable linear recurrence, when can you say that it is bounded? (that
is, there exists a value M such that each of the elements of the sequence has absolute value less than
M)