Вы находитесь на странице: 1из 4

Five ways of finding a closed forumula for the Fibonacci Numbers

Leonardo Ignacio Martnez Sandoval


ssbmplayer@gmail.com
April 9, 2011

The Fibonacci numbers are defined recursively as follows:

F0 = 0
F1 = 1
Fn+2 = Fn+1 + Fn for n 0

One of the things we wish to do when we have a recursive sequence is try to find a closed
form. Closing the Fibonacci recursion (or for that matter, any linear recursion) is a well-known
problem, and the conclusion is stated
in the following
proposition.

1+ 5
1 5
Binets Formula Let = 2 , = 2 . Then, the n-th Fibonacci number is given by the
following formula:
Fn =

n n

There are several ways of proving Binets Formula. In this article we will look at five different
proofs. For some of them, I will assume familiarity with some concepts of Calculus and Linear
Algebra. We state some lemmas without proof, but they are easy to check.

Induction

We will start with a proof that does not require conjecturing the formula. Simply, we will assume we
already know it and proceed by induction. In the next sections we will see some ways of discovering
the formula.
The key observation for proving Binets Formula and avoid getting stuck with the algebra is
the following. Using Binets formula notation, we have + = 1, = 1 and thus, and are
the roots of the equation x2 x 1 = 0. This means that 2 = + 1 and 2 = + 1. The proof
now proceeds smoothly as follows.
= 0 = F0 and for n = 1 the formula is
= 1 = F1 , so in this
For n = 0 the formula is 11
5
5
cases the formula works. For n 0 we have:

n+1 n+1 n n

+
5
5
n ( + 1) n ( + 1)

=
5
n
2
n
2
n+2 n+2

=
=
5
5

Fn+2 = Fn+1 + Fn =

And thus we have proved the formula by induction.

I believe n might work

The contents in this section can be used to find a closed formula for any linear recursion of order
2 (even higher with some small changes), that is, any recursion of the form an+2 = ban+1 + can .
The main idea is that maybe there is a number that satisfies an = n . Most of the time,
such number will not work directly, but by making an extra observation, we will be able to find
many candidate functions and then select one of them that works. Lets see what all this means
by rediscovering Binets formula.
Imagine there exists a number such that Fn = n . This clearly cannot be 0. If this works,
then we would need that n+2 = Fn+2 = Fn+1 + Fn = n+1 + n for every n 0, but since 6= 0,
this happens if and only if 2 = + 1, which happens if and only if is a root of the equation
x2 x 1 = 0. We have already seen that the roots of this equation are and .
Unfortunately, neither Fn = n or Fn = n seem to work directly, since, 6= 1 and 6= 1.
However, to get much more options we will use the following lemma.
Lemma 2.1. If {an } and {bn } (n 0) are sequences that satisfy the recursion xn+2 = rxn+1 + sxn
and a and c and d are real numbers, then {can +dbn } also satisfies the recursion xn+2 = rxn+1 +sxn .
This is great. We already know that n and n satisfy xn+2 = xn+1 + xn . So now the problem
is changed to finding numbers c and d such that Fn = cn d n for every n. But a sequence that
satisfy the Fibonacci recursion is completely determined by giving its first two terms. So we can find
c and d that work for every n if and only if we can find c and d such that c+d = c0 +d0 = F0 = 0
1
1
and c + d = F1 = 1. By solving the system of equations, we have c =
= 15 and d =
.
5
This completes another proof of Binets formula.

Generating Functions: Ending 1

In this section we consider a proof that uses generating function, that is, we look at a formal power
series and manipulate it. We start by defining the following function:
2

f (x) = F0 + F1 x + F2 x + . . . =

X
n=0

Fn xn

When we deal with this sort of functions when doing combinatorics, we usually forget convergence issues and stay focused on the algebraic properties of the serie. That is what we are going
to do. The idea will be to find f (x) in another way that explicitely shows the coefficients of xn . So
we start by noticing that:

f (x) =
xf (x) =
x2 f (x) =

X
n=0

X
n=0

Fn xn
Fn x

n+1

Fn xn+2 =

n=0

X
n=1

Fn1 xn
Fn2 xn

n=2

By adding the first equations and substracting the third one, we get that f satisfies x2 f (x) +
xf (x) f (x) = x (here is the only place where we use F0 = 0 and F1 = 1). So we have made a
x
great advance in finding f , since we can write f (x) = x2 +x1
. But remember that the idea is that
we get another power series to compare f with, say, a geometric power series. So we do standard
partial fractions stuff. Notice
that the roots of x2 + x 1 are and (Yes! our previous and
). Remember that = 5. So we get after some algebraic manipulations that:
x
1
1
1
=

x2 + x 1
( ) 1 x ( )


1
1
1
=

5 1 x 1 x
!

1 X n n X n n
=
x
x
5 n=0
n=0

f (x) =

1
x

X
n n n

x
5
n=0

P
1
n
In the third equality we used the well known geometric series formula 1x
=
n=0 x .
So by comparing the coefficients of f in both of its power series we get yet again Binets formula.

Generating Functions: Ending 2

P
1
n
Suppose that while using the previous argument we completely forgot that 1x
=
n=0 x (that
has actually happened to me). We can still complete the proof by making use of the following
lemma.
P
f (n) (0)
(n) denotes the n-th derivative of f .
n
Lemma 4.1. If f (x) =
n=0 an x , then an =
n! , where f

We will apply the lemma knowing that f (x) =


can be proved that the n-th derivative of
n-th derivative of

1
1x

is

n! n
.
(1x)n+1

1
1x

1
1x

. With a simple induction it

and similarily we can prove that the

Therefore, we conclude as follows.

f (n) (0)
1
=
n!
n! 5
n n
=
5

Fn =

is

1
1
5 1x
n!n
,
(1x)n+1

n!n
n! n

n+1
(1 0)
(1 0)n+1

The Fibonacci Matrix

We were told at Linear Algebra that if A is a matrix, then we can easily find the powers of A
if we can diagonalize A. More explicitly, if A is a matrix diagonalized as A = M DM 1 , then
An = M Dn M 1 , and Dn is easy to calculate because D is a diagonal matrix. We would like to
make use of this fact to find Binets



 formula.
a b
1 1
n
, then An+1 = An A =
. Notice that if A =
Lets consider the matrix A =
c d
1 0




Fn+1 Fn
a+c b+d
n
for all n 1. We will
, so we can prove inductively that A =
Fn Fn1
a
b
now find another way to calculate An by diagonalizing A.
It should be no surprise here that the characteristic polynomial of A is precisely x2 x 1. So
its
 that (, 1) is an eigenvector for , since
 . We
 notice
  and
 eigenvalues
  are,
 well,
 you know,
2
1 1

+1
, and similarily, (, 1) is an eigenvector for
=
=
=
1

1 0
1
. Therefore:


1 1
1 0

n


 n


1
1 1

0
1
=

0 n
1



1
a n n
=
b
c
5

We only computed the top right entry. By comparing that entry with our previous observations
n
n
about An , we get that Fn =
.
5

Conclusions
Thank you for reading this small article. These are just some of the ways of proving Binets
formula. These techniques can be generalized to other linear recursions, even of higher order.
Finding a closed formula for a linear recursion is definitely one of the most useful resources (it
seems that the people who makes the Putnam exams loves linear recursions). I would love to hear
about other ways of proving and conjecturing Binets formula.

Вам также может понравиться