Sequences

Adithya B., Brian L., William W., Daniel X.

6/10

§1 Basic Sequences
We’ll start with the most basic sequences: arithmetic and geometric sequences.

Definition 1.1. An arithmetic sequence is a sequence in which the difference between

any two consecutive terms is constant. We call this constant the common difference of
the arithmetic sequence.

Thus the terms of a k-term arithmetic sequence can be written as

a, a + d, a + 2d, . . . , a + (k − 1)d

where a is the first term and d is the common difference.

Definition 1.2. A geometric sequence is a sequence in which the ratio between any two
consecutive terms is (a nonzero) constant. This constant is called the common ratio of
the sequence.

We may write the terms of a k-term geometric sequence as

a, ar, ar2 , . . . , ark−1

where a is the first term and r is the common ratio.

When faced with a problem dealing with an arithmetic or geometric sequence, a
reliable strategy is to write write all of the terms in terms of the first term and common
ratio/difference. This is because the entire sequence can be described in terms of three
numbers: the first term, number of terms, and common ratio/difference. All of the
information we know about the sequence can be encapsulated in a fairly small number
of variables, allowing us to easily solve for the variables and thus determine the entire
sequence.

Example 1.3 (2003 AIME I #8)

In an increasing sequence of four positive integers, the first three terms form an
arithmetic progression, the last three terms form a geometric progression, and the
first and fourth terms differ by 30. Find the sum of the four terms.

Solution. It is often a good idea to assign variables so that the expressions we obtain are
as clean as possible. Since the first three terms form an arithemetic progression, we could
set them to be a, a + d, a + 2d. Since the last three terms form a geometric sequence,
(a+2d)2
their common ratio is a+2d
a+d so the fourth term is a+d . But we can clean this up a

1
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

little, by instead setting the first three terms to be a − d, a, a + d. Then the fourth term
2
is (a+d)
a instead. Now we have the equation

(a + d)2
− (a − d) = 30 =⇒ (a + d)2 − a(a − d) = 30a =⇒ d2 + 3ad = 30a.
a
Let’s solve for a in terms of d; we try this because the above equation is a linear equation
in a. We get
d2
a= .
30 − 3d
Since we know that a, d > 0 (the terms of the sequence are positive, and the sequence is
increasing) this gives us that d < 10. Moreover, the denominator is divisible by 3, and as
a is an integer the number must be divisibly by 3 as well. This narrows down d to 3, 6, 9.
If d = 3 then a = 37 , which is not an integer. IF d = 6 then a = 3, but then the first term
is a − d = −3 which is negative. Thus we must have d = 9, which gives us a = 27. So
our sequence is
18, 27, 36, 48
which we see satisfies the conditions of the problem. So the answer is 18 + 27 + 36 + 48 =
129 .

Sometimes, we aren’t presented with a completely arithmetic or geometric sequence.

Nevertheless, we can still use the special properties of such sequences.

Example 1.4 (2004 AIME II #9)

A sequence of positive integers with a1 = 1 and a9 + a10 = 646 is formed so that the
first three terms are in geometric progression, the second, third, and fourth terms are
in arithmetic progression, and, in general, for all n ≥ 1, the terms a2n−1 , a2n , a2n+1
are in geometric progression, and the terms a2n , a2n+1 , and a2n+2 are in arithmetic
progression. Let an be the greatest term in this sequence that is less than 1000. Find
n + an .

Solution. Since the first three terms are in geometric progression, we can write them
as 1, r, r2 for their common ratio r (which must be an integer). Since a2 , a3 , a4 are in
arithmetic progression we can write the fourth term as the third term plus the common
difference, or

a4 = a3 + (a3 − a2 ) = 2a3 − a2 = r2 + (r2 − r) = 2r2 − r.

Now a3 , a4 , a5 are in geometric progression, so

a4 a5 a2 (2r2 − r)2
= =⇒ a5 = 4 = = (2r − 1)2 .
a3 a4 a3 r2
Now a4 , a5 , a6 form an arithmetic sequence, so

a6 = 2a5 − a4 = 2(2r − 1)2 − r(2r − 1) = (2r − 1)(3r − 2).

We can repeat this logic: since a5 , a6 , a7 form a geometric sequence we have

a26 (2r − 1)2 (3r − 2)2

a7 = = = (3r − 2)2 .
a5 (2r − 1)2

2
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

Continuing,

a8 = 2a7 − a6 = 2(3r − 2)2 − (2r − 1)(3r − 2) = (4r − 3)(3r − 2),

a28 (4r − 3)2 (3r − 2)2
a9 = = = (4r − 3)2 ,
a7 (3r − 2)2
a10 = 2a9 − a8 = 2(4r − 3)2 − (4r − 3)(3r − 2) = (4r − 3)(5r − 4).

Now that we have expressions for a9 , a1 0, we can write

646 = a9 + a10 = (4r − 3)2 + (4r − 3)(5r − 4) = (4r − 3)(9r − 7).

Since this is a quadratic we can solve for r. Alternatively, we can factor 646 = 2 · 17 · 19
and try to use this factorization to find r (since r must be an integer). Either way we
can find that r = 5.
Now we have a9 = (4r − 3)2 = 289, a10 = (4r − 3)(5r − 4) = 357. We can continue
computing the ai until we get the largest term less than 1000:

a210
a11 = = 441,
a9
a12 = 2a11 − a10 = 525,
a212
a13 = = 625,
a11
a14 = 2a13 − a12 = 725,
a214
a15 = = 841,
a13
a16 = 2a15 − a14 = 957,
a216
a17 = = 1089.
a15

So a16 = 957 is the largest term less than 1000, and the answer is 16 + 957 = 973 .

§2 Recursions
Broadly speaking, recursive sequences are those for which the nth term is determined
based on the terms before it. As this is a very broad class of sequences, there aren’t any
panaceas for dealing with them. However, some ideas may help reduce work by a lot.

Sometimes, a recursion will be periodic, so that the sequence repeats itself after some
number of terms. If this happens, it is usually extremely important to solving the
problem.

Example 2.1 (2012 AIME I #11)

Let f1 (x) = 23 − 3x+13
, and for n ≥ 2, define fn (x) = f1 (fn−1 (x)). The value of x
that satisfies f1001 (x) = x − 3 can be expressed in the form m n , where m and n are
relatively prime positive integers. Find m + n.

Solution. There’s not much that immediately stands out about the function or equation
that we are presented with. To get a better feel for the sequence, let’s write out a few of
the terms and see if we notice anything special. In general, when faced with a unfamiliar

3
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

situation, it’s a good problem solving strategy to look at some concrete examples and
piece together data into observations and conjectures.
2 3
f1 (x) = − ,
3 3x + 1
2 3 2 3x + 1
f2 (x) = − 2 3 = − ,
3 3( 3 − 3x+1 ) + 1 3 3x − 2
2 3
f3 (x) = − 2 3x+1 = x.
3 3( 3 − 3x−2 ) + 1

So the sequence is periodic! In particular, we see that f1 (x), f2 (x), f3 (x), f4 (x), . . . repeats
itself after every 3 terms. This means that f1001 (x) = f2 (x). So we only need to solve
2 3x + 1
f2 (x) = x − 3 ⇐⇒ − =x−3
3 3x − 2
5
Solving this equation gives x = 3 so the answer is 8 .

Finally, another good strategy to try is generating functions. As we learned last class,
we can often solve a summation by translating it into a generating function and using
snake oil. Actually, the same method applies for sequences, as sometimes it is easier
to first find the generating function, and backtrack to produce a general form for the
sequence. Let’s see this method in an example:

Example 2.2 (2018 PUMaC Algebra #7)

Let the sequence {an }∞
n=−2 satisfy a−1 = a−2 = 0, a0 = 1, and for all non-negative
integers n,
Xn n
X
2
n = an−k ak−1 + an−k ak−2 .
k=0 k=0

Given a2018 is rational, find the maximum integer m such that 2m divide the
denominator of the reduced form of a2018 .

Solution. At first sight, this reminds us a little of the Catalan recursion that we discussed
in the previous handout. Therefore, we are going to use a similar strategy and try to
determine a generating function for this sequence. Let A(x) = a0 + a1 x + a2 x2 + · · · .
Now, note that we can rewrite the equation we were given as
n
X
2
n = an−k (ak−1 + ak−2 ).
k=0

The term ak−1 + ak−2 seems hard to handle directly, so we are going to define a new
sequence bk = ak−1 + ak−2 . Let its generating function be B(x). Note that this is defined
for all nonnegative k due to the problem statement. Now, the equation in the problem
becomes
Xn
n2 = an−k bk .
k=0

Now, if we define a new sequence by cn = nk=0 an−k bk , then we have cn = n2 for all n.
P
Let the generating function for this sequence be C(x). Note that from the recursion we

4
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

have C(x) = A(x) · B(x). This is because when we multiply A(x) and B(x), we obtain

(a0 + a1 x2 + a2 x2 + · · · )(b0 + b1 x + b2 x2 + · · · )
= a0 b0 + (a0 b1 + a1 b0 )x + (a0 b2 + a1 b1 + a2 b0 )x2 + · · ·
= c0 + c1 x + c2 x2 + · · ·
= C(x)

from the recursion for cn . Now, we can find the thePgenerating function for C(x) because
we know the closed form of cn . We have C(x) = ∞ 2 n
n=0 n x . Now, the question might
be how to obtain a term n2 from the “known” generating functions. And the answer is
right from the binomial coefficients! We already know the following generating functions:
∞   ∞
2 X n+2 n X 2
= 2 x = (n + 3n + 2)xn .
(1 − x)3 2
n=0 n=0

∞   ∞
1 X n+1 n X
= x = (n + 1)xn
(1 − x)2 1
n=0 n=0
∞
1 X
= xn
1−x
n=0

We can combine these to find

∞
X 2 3 1 x + x2
C(x) = n2 xn = − + = = A(x)B(x).
(1 − x)3 (1 − x)2 1 − x (1 − x)3
n=0

Now, our goal is to find A(x), so we want to find a way to relate A(x) and B(x). Note
that
∞
X ∞
X ∞
X
B(x) = (an−1 + an−2 )xn = x an xn + x2 an xn = (x + x2 )A(x)
n=0 n=0 n=0

Substituting,
x + x2 1
(x + x2 )A(x)2 = 3
=⇒ A(x) = 3 .
(1 − x) (1 − x) 2
We claim that
∞  
1 X 2n + 1 2n n
3 = x .
(1 − x) 2 4n n
n=0

This can be proven either by taking a derivative of the Catalan generating
function twice
4036
or expanding with the binomial theorem. Therefore, a2018 = 44037 2018 2018 . It suffices to
compute
 
4036
ν2 = ν2 (4036!) − 2ν2 (2018!) = 4029 − 2 · 2011 = 7.
2018

There are
4036 powers of 2 in the denominator before dividing out the 7 powers of 2
from 4036
2018 , so there are 4036 − 7 = 4029 powers of 2 in the denominator.

In fact, this is a very general technique, and as we will see in the next section, it can be
extended to a very broad class of recursions, namely linear recurrences.

5
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

§3 Linear Recursions
A linear recursion is one of the form

an = c1 an−1 + c2 an−2 + . . . + ck an−k ∀n ≥ k

for some constants c1 , c2 , . . . , ck . Generally, the most important thing when dealing with
linear recursions is to find some sort of closed form for an given a set of initial conditions.
Such a task can be achieved using last week’s ideas on generating functions.

In particular, define the generating function

S = a0 + a1 x + a2 x2 + . . .

What our recursion tells us is that

c1 xS + c2 x2 S + c2 x2 S + . . . + ck xk S = S − P (x)

where P (x) is some polynomial with deg P < k. To see this, simply consider the coefficent
of xn on the LHS for some n ≥ k. Adding its coefficient over each term individually,
we get c1 an−1 +c2 an−2 +. . .+ck an−k = an where the equality is due to our recursive form.

This tells us that the coefficient of xn on the LHS matches that of P (x) for all n ≥ k.
However, we need P (x) because this equality is not guaranteed for smaller coefficients,
so P (x) acts as a correcting factor.

Now that we have this equation, we can actually solve for S. In particular,

P (x)
S=
1 − c1 x − c2 x2 − . . . − ck xk
Though this form may not seem useful. Note that the denominator can actually be
factored! In particular, if we denote it as Q(x), we can write it as

Q(x) = −ck (x − r1 )(x − r2 ) · · · (x − rk )

where the ri are the roots of Q(x) (for now, we assume that all the ri are distinct.)

Now, using partial fraction decomposition, we know that we can break this fraction up
into something of the form
m1 r1 m2 r2 mk rk
S= + + ... +
r1 − x r2 − x rk − x
Though we won’t go into detail, the general idea for why this works is that all the
Q k
i(x−r )
polynomials of the form i=1x−rj are linearly independent, and thus form a basis over
polynomials with degree ≤ k. Thus, we are guaranteed to find such a partial fraction
decomposition which will evaluate with numerator −P (x)/ck .

ri
Our last step, of course, is to note that x−r i
is actually a generating function. In
particular,
∞  j
ri 1 X x
= x =
ri − x 1 − ri ri
j=0

6
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

So, if we sum up across all i, we get that

∞ k k
X X mi X mi
S= xj =⇒ an =
j=0 i=1 rij i=1 rij
1
In practice, to find such a closed form, we just need to find the values of ri , and then use
the first k terms of the recursion in order to calculate all the mi .

1
Note that the ri are roots of the polynomial

C(x) = xk − c1 xk−1 − c2 xk−2 − . . . − ck

This is known as the ”characteristic polynomial” of the recursion.

Note that, if the roots α1 , α2 , . . . , αk of C(x) are distinct, we just get the aforemen-
tioned form, namely
an = m1 α1n + m2 α2n + . . . + mk αkn
On the other hand, if we have a root with multiplicity, we must modify this form slightly.
In particular, suppose α1 actually has multiplicity two. Then, in our partial fraction
decomposition, the numerator will be a linear expression rather than a constant, so we
must replace m1 with a + bx for some a, b. So, the term corresponding to α1 becomes
(a + bn)α1n . The idea is analogous for roots of higher multiplicity.

Now, let’s try to apply the above ideas to find the closed form of a particular re-
cursion:

Example 3.1 (Brilliant)

A sequence xn is defined by x0 = −1, x1 = 0, x2 = 1 and the recurrence relation

xn = 6xn−1 − 12xn−2 + 8xn−3 .

Find the closed form of xn .

Solution. Our first step should be to find the characteristic polynomial. We see that
C(x) = x3 − 6x2 + 12x − 8 = (x − 2)3
So the only root of this characteristic polynomial is −2, but this root has multiplicity 3.
That means the closed form of our sequence is of the form
xn = (an2 + bn + c)2n
for some constants a, b, c. Now we use the given values of x0 , x1 , x2 to solve for a, b, c:
−1 = x0 = c,
0 = x1 = 2(a + b + c),
1 = x2 = 4(4a + 2b + c).
Solving these equations gives (a, b, c) = − 83 , 11


8 , −1 , so our closed form is
 
3 2 11
xn = − n + n − 1 2n .
8 8

7
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

Example 3.2 (1990 AIME #15)

Find ax5 + by 5 if the real numbers a, b, x, and y satisfy the equations

ax + by = 3,
ax2 + by 2 = 7,
ax3 + by 3 = 16,
ax4 + by 4 = 42.

Solution. We’ve done this problem before, through tricky algebraic manipulation. But
let’s see how our knowledge of linear recursions can solve this problem extremely quickly
and efficiently!
Let xn = axn + by n ; we’re given the values of x1 , x2 , x3 , x4 , and we want to find x5 .
What do we know about a sequence of the form axn + by n ? It must satisfy a linear
recurrence relating xn to xn−1 and xn−2 ! In particular, we must have the recurrence
xn = cxn−1 + dxn−2
for some constants c and d. But we can solve for c and d with our given values. Choosing
n = 3 and n = 4, we get the equations
16 = 7c + 3d, 42 = 16c + 7d.
Solving this system of linear equations gives c = −14 and d = 38, so
xn = −14xn−1 + 38xn−2 .
Then x5 = −14x4 + 38x3 = −14 · 42 + 38 · 16 = 20 .

§4 Weird Sequences
Unfortunately, most sequences elude any specific categorization, and require an ad hoc
idea in order to solve. Here are some general ideas in order to approach novel sequences:
• Try to compute small terms if possible. Guess a pattern through engineers induction
and try to prove it
• Look at how the sequence behaves as a whole, and if there are any overarching
global patterns
• Be on the look out for manipulations, such as factorizations and substitutions which
will simplify how the sequence looks
• If initial conditions are given, see if they are special by trying the question with
your own conditions. If something nice happens regardless of the values, the given
information may be a decoy and you can use variables instead. Otherwise, try to
understand why the numbers given are actually significant
• Try to prove subresults to get a better intuition with the sequence. Jot down ideas
you may have or qualities of the sequence you see, even if they have little to do
with what we are actually trying to prove
Often times, the recursive formula given may appear random or too chaotic. However,
if how the sequence behaves as a whole is considered, then the solution becomes much
more apparent. Let’s see an example where, though the recursion isn’t nice, there is still
an underlying structure:

8
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

Example 4.1 (2017 CMIMC Algebra & Number Theory #9)

√
Define a sequence {an }∞
n=1 via a1 = 1 and an+1 = an + b an c for all n ≥ 1. What
is the smallest N such that aN > 2017?

Solution. If one tries to immediately find a closed form for {ai }, it is very easy to get
√
lost. After all, b xc is already an unwieldy function, so trying to bash out terms by
nesting it will make a complicated question even more complicated.
√
Instead, we try to see how b xc behaves. Suppose we have ak = m2 for some m.
Then, ak+1 = m2 + m, and ak+2 = m2 + 2m. So, ak+2 is just under (m + 1)2 . This means
√
that, if we start from m2 , it will take three iterations to increment b ai c. However, if
we start from a number between m2 + 1 and m2 + m, it will only take 2 increments.
Furthermore, as we are only adding 2m, if we started a away from m2 , we will be a − 1
away from (m + 1)2 after these two increments.

This gives us a very good idea about how the sequence will behave: At some point,
an will overshoot a square, and produce a number a larger than a square. Then, ev-
√
ery two turns the size of b xc will increase by 1, but the overshoot will decrease by
1, until it reaches 0 and we need three turns, which will in turn produce another overshoot.

Now, let’s rigorize these ideas. We have that a1 = 1 which is a square, so we need
3 iterations to get to a4 = 4, which is another square. Three more iterations, we get to
a7 = 10, which now has overshoot of 1, so we can get to a9 = 16 in two turns. Now,
the next overshoot is over size 3, and it will take 2 ∗ 3 more turns to diminish back to a
square, so we get a9+3+2∗3 = (4 + 4)2 =⇒ a18 = 64.

At this point, things are looking a bit fishy - in particular, we see that all perfect
squares in our sequence are powers of 4. It is not hard to prove once we notice it - if
ak = 4n , then after 3 turns we have an overshoot of 2n − 1, which will take 2 ∗ (2n − 1)
turns to correct. This means ak+3+2n+1 −2 = (2n + 2n )2 =⇒ ak+2n+1 +1 = 4n+1 . In fact,
as we always add 2n+1 + 1, this also gives us a closed form for k. See if you can show
that a2n+1 +n−1 = 4n .

The rest of the solution is surprisingly simple, especially as we now have so much
intuition. We get that a68 = 1024, and we want to get to around 452 ≈ 2017. So, we need
to increase the argument of our square by 45 − 32 = 13, which will take 3 + 2 · 12 = 27
terms. After this many increments, our overshoot is 32 − 13 = 19, so a95 = 2025 + 19.
Now, a94 = 2025 + 19 − 44 < 2017, so our answer is 95 .

Example 4.2
Romania TST 2003/1 Let (an )n≥1 be a sequence for real numbers given by a1 = 1/2
and for each positive integer n

a2n
an+1 = .
a2n − an + 1

Prove that for every positive integer n we have a1 + a2 + · · · + an < 1.

Solution. At the beginning, it seems really unclear as to how to get the desired sum. So,

9
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

rather than try to work backwards, we try to build up information about the sequence at
hand.

1 1 1 1
First, note that if we take reciprocals, we get an+1 = 1− an + a2n
, so setting bn = an
gives the much nicer recurrence of

bn+1 = b2n − bn + 1

This is still not great. At the end of the day, we want to find a nice way to relate all the
bi . Let’s try to list a few values to see what we get.

b1 = 2, so b2 = 3, b3 = 7, b4 = 43, b5 = 1807, and so on. With these terms, it is

not hard to guess that we must have bn = bn−1 bn−2 . . . b1 + 1. To prove, note that our
−1
recursion can be written as bn+1
bn −1 = bn , so if we multiply and telescope, we get

bn+1 − 1
= b1 b2 · · · bn =⇒ bn+1 = b1 b2 · · · bn + 1
b1 − 1
as desired. Note that we could have found this relation with the ai as well, however first
converting into bi greatly reduced our computation for finding the pattern.

Now, let’s convert back to as. We get that

1 a1 a2 · · · an (a1 a2 . . . an )2
an+1 = 1 = = (a1 a2 . . . an ) −
a1 ···an +1 1 + a1 a2 · · · an 1 + a1 a2 . . . an

= (a1 a2 . . . an ) − (a1 a2 . . . an+1 )

Aha! Now, if we add all the as, we get a telescoping series. In particular,

a1 + a2 + . . . + an = 1 − a1 a2 · · · an < 1

as desired.

10
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

§5 Problems
Problem 5.1 (2018 AMC 12 #22). Define a sequence recursively by x0 = 5 and

x2n + 5xn + 4
xn+1 =
xn + 6
for all nonnegative integers n. Let m be the least positive integer such that
1
xm ≤ 4 + .
220
In which of the following intervals does m lie?
(A) [9, 26] (B) [27, 80] (C) [81, 242] (D) [243, 728] (E) [729, ∞]

Problem 5.2 (Classic). Derive a closed form for the Fibonacci numbers defined by
F1 = F2 = 1 and Fn = Fn−1 + Fn−2 .

Problem 5.3 (Brilliant). A sequence xn is defined by x0 = 0, x1 = 1, and the recurrence

relation
xn = xn−1 − 4xn−2 .
Find the closed form of xn .
 √ 2020
Problem 5.4. Prove that (2 + 5) is odd.

Problem 5.5 (2013 HMMT Algebra # 2). let {an }n≥1 be an arithmetic sequence
{gn }n≥1 be a geometric sequence such that the first four terms of {an + gn } are 0,0, 1,
and 0, in that order. What is the 10th term of {an + gn }?

Problem 5.6 (1988 AIME # 2). For any positive integer k, let f1 (k) denote the square
of the sum of the digits of k. For n ≥ 2, let fn (k) = f1 (fn−1 (k)). Find f1988 (11).

Problem 5.7 (2016 PUMaC Algebra #3). Let x0 , x1 , . . . be a sequence of real numbers
such that xn = 1+x n−1
xn−2 for n ≥ 2. Find the number of ordered pairs of positive integers
1
(x0 , x1 ) such that the sequence gives x2018 = 1000 .

Problem 5.8 (2017 PUMaC Algebra #4). Let a1 , a2 , . . . be a sequence of positive real
numbers such that an = 11an−1 − n for all n > 1. The smallest possible value of a1 can
be written as pq , where p and q are relatively prime positive integers. Find p + q.

Problem 5.9 (2016 HMMT Algebra #5). An infinite sequence of real numbers a1 , a2 , . . .
satisfies the recurrence
an+3 = an+2 − 2an+1 + an
for every positive integer n. Given that a1 = a3 = 1 and a98 = a99 , compute a1 + a2 +
· · · + a100 .

Problem 5.10 (Brilliant). Let f be a function that satisfies the equation

(n + 1)2 f (n) − n3 f (n − 1) = 1

for all nonnegative integers n with f (0) = 3. Find f (n).

Problem 5.11 (2016 PUMaC Algebra #4). Define a sequence ai as follows: a1 = 1,

na2n−1
a2 = 2015, and an = an−1 +nan−2 for n > 2. What is the least k such that ak < ak−1 ?

11
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

Problem 5.12 (2009 AIME I #7). The sequence (an ) satisfies a1 = 1 and 5(an+1 −an ) −1 =
1
for n ≥ 1. Let k be the least integer greater than 1 for which ak is an integer.
n + 23
Find k.

Problem 5.13 (2016 AIME I #10). A strictly increasing sequence of positive inte-
gers a1 , a2 , a3 , . . . has the property that for every positive integer k, the subsequence
a2k−1 , a2k , a2k+1 is geometric and the subsequence a2k , a2k+1 , a2k+2 is arithmetic. Suppose
that a13 = 2016. Find a1 .

Problem 5.14 (2002 AIME I #12). Let F (z) = z+i z−i for all complex numbers z 6= i, and
1
let zn = F (zn−1 ) for all positive integers n. Given that z0 = 137 + i and z2002 = a + bi,
where a and b are real numbers, find a + b.

Problem 5.15 (2009 AIME I #13). The terms of the sequence (ai ) defined by an+2 =
an +2009
1+an+1 for n ≥ 1 are positive integers. Find the minimum possible value of a1 + a2 .

Problem 5.16 (2007 AIME I #14). Let a sequence be defined as follows: a1 = 3, a2 = 3,

and for n ≥ 2, an+1 an−1 = a2n + 2007. Find the largest integer less than or equal to
a22007 +a22006
a2007 a2006 .

Problem 5.17 (2009 AIME II # 14). The sequence (an ) satisfies a0 = 0 and an+1 =
8 6p n
an + 4 − a2n for n ≥ 0. Find the greatest integer less than or equal to a10 .
5 5
Problem 5.18 (2004 AIME I #15). For all positive integers x, let

1
 if x = 1
x
f (x) = 10 if x is divisible by 10

x + 1 otherwise


and define a sequence as follows: x1 = x and xn+1 = f (xn ) for all positive integers n.
Let d(x) be the smallest n such that xn = 1. (For example, d(100) = 3 and d(87) = 7.)
Let m be the number of positive integers x such that d(x) = 20. Find the sum of the
distinct prime factors of m.

Problem 5.19 (2014 AIME II #15). For any integer k ≥ 1, let p(k) be the smallest
prime which does not divide k. Define the integer function X(k) to be the product of all
primes less than p(k) if p(k) > 2, and X(k) = 1 if p(k) = 2. Let {xn } be the sequence
defined by x0 = 1, and xn+1 X(xn ) = xn p(xn ) for n ≥ 0. Find the smallest positive
integer, t such that xt = 2090.

Problem 5.20 (2019 PUMaC Algebra #7). A doubly-index sequence am,n , for m and
n nonnegative integers, is defined as follows.

(a) am,0 = 0 for all m > 0 and a0,0 = 1.

(b) am,1 = 0 for all m > 1, and a1,1 = 1, a0,1 = 0.

(c) a0,n = a0,n−1 + a0,n−2 for all n ≥ 2.

(d) am,n = am,n−1 + am,n−2 + am−1,n−1 − am−1,n−2 for all m > 0, n ≥ 2.

∞ P
∞
am,n xm
= 1. Find b1000x2 c.
P
Then there exists a unique value of x so 3m−n
m=0 n=0

12
Adithya B., Brian L., William W., Daniel X. (6/10) Sequences

Problem 5.21 (USAMTS 30/3/5). The sequence {an } is defined by a0 = 1, a1 = 2,

and for n ≥ 2,
an = a2n−1 + (a0 a1 · · · an−2 )2 .
Let k be a positive integer, and let p be a prime factor of ak . Show that p > 4(k − 1).

Problem 5.22 (EGMO 2020/1). The positive integers a0 , a1 , a2 , . . . , a3030 satisfy

2an+2 = an+1 + 4an for n = 0, 1, 2, . . . , 3028.

Prove that at least one of the numbers a0 , a1 , a2 , . . . , a3030 is divisible by 22020 .

Problem 5.23 (IMO 2014/1). Let a0 < a1 < a2 . . . be an infinite sequence of positive
integers. Prove that there exists a unique integer n ≥ 1 such that
a0 + a1 + a2 + · · · + an
an < ≤ an+1 .
n
Problem 5.24 (IMO 2018/2). Find all integers n ≥ 3 for which there exist real numbers
a1 , a2 , . . . an+2 satisfying an+1 = a1 , an+2 = a2 and

ai ai+1 + 1 = ai+2 ,

for i = 1, 2, . . . , n.

13