Vous êtes sur la page 1sur 5

TOPIC 10: Sequences:

This is a popular topic in competitions of all levels. It is also a very wide topic. However
some themes recur such as recursive definitions, arithmetic and geometric sequences. The
most important sequence is probably the ubiquitous Fibonacci sequence defined
recursively by f0 = 0, f1 = 1 and fn+1 = fn + fn-1. There is even a magazine, The Fibonacci
Quarterly, which is devoted solely to the study of this and related sequences. You should
immediately recognize this sequence: 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, and so on. Also read
Chapter 16 in the textbook which discusses the Fibonacci numbers.
How do we get an explicit formula for a recursively defined sequence? If the sequence is
an+1 = 2an with a1 = 3, say, then it is easy to see that the n-th term is given by the
(geometric) formula an = 32n-1. There is a general law: a sequence defined by a linear
recursion can be presented as sums of geometric sequences!
Example: If a0 = 2, a1 = 7 and an+1 = 7an 12an-1 then the so-called associated
characteristic equation is 2 = 7 12 with solutions 1 = 3 and 2 = 4. The general
solution then is an = a1n + b2n for some constants a and b. Plugging in the initial
conditions a0 = 2 and a1 = 7 yields a + b = 2 and 3a + 4b = 7 with solutions a = b = 1.
Hence the explicit formula for the given sequence is an = 3n + 4n, a sum of two geometric
sequences.
For the Fibonacci sequence we get using the same approach fn = 1/ 5 [ 1n+1 + 2n+1]
where 1 = (1+ 5 )/2 and 2 = (1- 5 )/2. This is called Binets formula. Note that 1
and 2 are the solutions to the characteristic equation 2 = + 1. This works even for
higher order linear recurrences for example the characteristic equation to an+3 = 2an+2 +
3an+1 + 4an is 3 = 22 + 3 + 4 and the sequence can be represented as a sum of three
geometric sequences.
But note that an explicit representation for the n-th term does not always make the
problem easier to deal with. You should always think about rearranging the terms (as in
example 7 below). Another important technique is to shift the index in the recursion (so
replace n by n-1 say) and combine the two equations (see example 1 or 4). One should
always compute the first few terms of the sequence to get a feel for what is going on. If
everything fails we make a table.

Example 1 (AMC12 1984):


Given the sequence an defined by a1 = 2 and an+1 = an + 2n for n 1 find a100.
Solution:
We chose this example to illustrate that even though it is a linear recursion an explicit
formula cannot be arrived at with the method of characteristic equations since the

absolute term (2n) is not a constant. In such a situation keep applying the recursion to see
what happens (keep shifting the index):
a100 = a99 + 299
= a98 + 298 + 299
= a97 + 297 + 298 + 299
= a1 + 2(1 + 2 + + 99)
= 2 + 2(99100/2)
= 9902
Example 2 (BWM 1990.1.2):
The sequence a0, a1, a2, is defined by a0 = 0, a1 = a2 = 1 and an+2 + an-1 = 2(an+1 + an) for
all positive integers n. Show that all terms of this sequence are square numbers.
Solution:
One approach would be to solve the characteristic equation of degree 4 to get an explicit
representation for the terms. This is a bid daunting so we make a table and will probably
notice soon that an = fn2 where fn is the Fibonacci sequence. We will prove this, as so
often with sequence problems, by mathematical induction.
For n=0 and n=1 we have a0 = 0 = f0 and a1 = 1 = f1. Now we assume that ak = fk2 for k =
0, 1, 2, , n. Then an+1 = 2an + 2an-1 an-2 = 2fn2 + 2fn-12 fn-22
= 2fn2 + 2fn-12 (fn fn-1)2
= 2fn2 + 2fn-12 fn2 +2fnfn-1 fn-12
= fn2 + fn-12 +2fnfn-1
= (fn + fn-1)2
= fn+12.

Example 3 (AIME 1990.1):


The increasing sequence 2, 3, 5, 6, 7, 10, 11, consists of all positive integers that are
neither the square nor the cube of a positive integer. Find the 500th term of the sequence.
Solution:
Clearly the required term n is unique and greater than 500. But not much greater than 500
as the squares and cubes thin out as they increase. In fact, the number of squares less than
500 is 22 (as 232 = 529) and the number of cubes is only 7 (as 83 = 512). Thus up to 500
not more than 29 positive integers have failed to qualify to be included in the sequence.
Actually, since 1 = 16 and 64 = 26 are both squares and cubes (and the only ones less than
500 as 36 = 729) only 27 positive integers less than 500 are terms of the sequence. Here
we applied the principle of inclusion and exclusion (see topic 8 on combinatorics). So,
500 is the 473rd term of the sequence. Advancing 27 terms from 500 takes us to 528 since
we need to skip the cube 512. Since 512 is the only square or cube between 500 and 528
the number 528 is indeed the 500th term of the sequence. This cuts things pretty close as
the next integer, 529, is a square.

Example 4 (AIME 2001II.3)


Given that x1 = 211, x2 = 375, x3 = 420, x4 = 523 and xn = xn-1 xn-2 + xn-3 xn-4 for n 5
find the value of x531 + x753 + x975.
Solution:
Looking for an explicit formula would be a bid ambitious as the characteristic equation is
of degree 4. So, instead we shift the index by 1 (replacing n by n-1) and combine the two
equations: xn = xn-1 xn-2 + xn-3 xn-4
xn-1 = xn-2 xn-3 + xn-4 xn-5
Adding these two equations yields xn = -xn-5. I think you can see now how we will
proceed. Since 531 = 1065 + 1, 753 = 1505 + 3 and 975 = 1945 + 5, x531 + x753 + x975 =
(-1)106x1 + (-1)150x3 + (-1)194x5 = 211 + 420 + 267 = 898. Note that x5 was found by
plugging in n=5 in the original recursion equation.

Example 5 (AIME 1986.7):


The increasing sequence 1, 3, 4, 9, 10, 12, 13, consists of all powers of 3 or sums of
distinct powers of 3. Find the 100th term of the sequence.
Solution:
Rewrite all of the terms in base 3 (this is suggested by the fact that we only look at sums
of powers of 3). Since the numbers are sums of distinct powers of 3, in base 3 each
number is a sequence of 1s and 0s (if there is a 2, then it is no longer the sum of distinct
powers of 3). Therefore, we can recast this into base 2 (binary) in order to determine the
100th number. 100 is equal to 64 + 32 + 4 = 26 + 25 + 22 (binary form is 110010).
However, we must change it back to base 3 for the answer, which is 36 + 35 + 32 = 729 +
243 + 9 = 981.
Example 6 (AIME 2002I.4):
Consider the sequence defined by ak = 1/(k2+k). Given that am+am-1+ + an-1 = 1/29 for
positive integers m < n find m + n.
Solution:
The sequence is already given in explicit form - crucial is the special form of the terms.
Terms of the form 1/(k2+k) or 1/(k2-k) can be written as 1/(k2+k) = 1/k 1/(k+1) (or
1/(k2-k) = 1/(k-1) 1/k). This is neat because the sum then telescopes:
am+am-1+ +an-1 = 1/m 1/(m+1) + 1/(m+1) 1/(m+2) + + 1/(n-1) 1/n = 1/m 1/n.
So the condition of the problem can be written as: (m-n)/mn = 1/29.
Since 29 is a prime number it is a factor of m or n. Lets try m and put m = 29t. We get
29t n = nt or n = 29t/(t+1). Since t+1 divides 29 we obtain t+1 = 29 or t = 28. Then m =
2928 and n = 28, so that m + n = 3028 = 840.
Had we tried n and put n = 29t similar calculations would imply m > n which is ruled out.
So, 840 is the answer.

Example 7 (AIME 2008II.6):


The sequence an is defined by a0 = 1, a1 = 1 and an = an-1 + an-12/an-2 for n 2.
The sequence bn is defined by b0 = 1, b1 = 3 and bn = bn-1 + bn-12/bn-2 for n 2.
Find b32/a32.
Solution:
The recursion is non-linear, so we try shifting the index or rearranging. Here rearranging
the definitions yields an/an-1 = an-1/an-2 + 1 and similarly bn/bn-1 = bn-1/bn-2 + 1. So we have
a telescoping effect and get an/an-1 = 1 + an-1/an-2 = 1 + (1 + an-2/an-3) = = (n-1) + a1/a0 =
n. For the second sequence we get bn/bn-1 = n+2. So, the recursions an = nan-1 and bn =
(n+2)bn-1 can be solved easily by inspection: an = n! and bn = (n+2)!/2 (after applying the
initial conditions). It follows that b32/a32 = (34!/2) / 32! = 3433/2 = 561.
Example 8 (AMC12 1990):
If Rn = (an + bn) where a = 3+ 2 and b = 3 - 2 for all n 0 show that all terms of the
sequence Rn are integers and then find the unit digit of R12345.
Solution:
Here we are given the explicit form of the terms but one cannot easily see that they are all
integers (only R0 = 1 and R1 = 3 are obvious). A recursion would (probably) easily make
it transparent. Here is what would you do to undo the explicit form. Multiply by the sum
of the roots of the characteristic polynomial, here a + b:
(a+b)Rn = (a+b)(an + bn) = (an+1 + bn+1) + (ban + abn) = Rn+1 + abRn-1.
Now a+b = 6 and ab = 1, so 6Rn = Rn+1 + Rn-1, or Rn+1 = 6Rn Rn-1. Since the first two
terms are integers from this it is immediate that all terms are integers. And this recursion
allows computing the first terms (mod 10 since we are interested in the unit digit):
1, 3, 7, 9, 7, 3, 1, 3, 7, 9, 7, 3, . The unit digits are periodic with period length 6. Now
12345 3 (mod 6) so, the unit digit of R12345 equals that of R3 which is 9.
Example 9:
Show that all the terms of the sequences given by a1 = 1, a2 = 1, a3 = 2 and
an+1 = (1+an-1an)/an-2 are integers.
Solution:
As in the previous problem we seek an easy linear recursion which would imply that the
sequence is an integer sequence. The sequence is given in a non-linear recursion. We can
shift the index and rearrange: we have an+1an-2 = 1 + an-1an. Shifting the index to n-1 and
subtracting yields that (an+1 + an-1)/an = (an-1 + an-3)/an-2. So, this eventually equals
(a3 + a1)/a2 = 3 or (a4 + a2)/a3 = 2 (since a4 = 3). So, for even n we get: an+1 = 3an an-1
and for n odd we get: an+1 = 2an an-1. This shows that all terms are indeed integers.

Practice Problems:
1)
Given g0 = 3, g1 = -1 and gn = gn-2 gn-1 find g100 in terms of the Fibonacci numbers.

2)
Let 1 = a1 < a2 < a3 < be positive integers such that a2n = an + n for n 1. It is known
that an is a prime number if n is a prime number. Find a1993.

3)
Define two sequences of rational numbers as follows: let a0 = 2 and b0 = 3 and
recursively define an = an-12 / bn-1 and bn= bn-12 / an-1. Find b8 leaving it in exponential
form mn / pq.

4)
Show that all terms of the sequence a1 = 1, a2 = 1 and an = (an-12 + 2008)/an-2 are integers.

5)
The sequence an is defined by a1 = a2 = 1, an+2 = an+1 + 2an. The sequence bn is defined by
b1 = 1, b2 = 7, bn+2 = 2bn+1 + 3bn. Show that the only integer belonging to both sequences
is 1.

Vous aimerez peut-être aussi