Download Application of series

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Law of large numbers wikipedia , lookup

Addition wikipedia , lookup

Functional decomposition wikipedia , lookup

Big O notation wikipedia , lookup

Approximations of π wikipedia , lookup

Mathematics of radio engineering wikipedia , lookup

Large numbers wikipedia , lookup

History of the function concept wikipedia , lookup

Proofs of Fermat's little theorem wikipedia , lookup

Non-standard calculus wikipedia , lookup

Elementary mathematics wikipedia , lookup

Factorial wikipedia , lookup

Transcript
Calculus I – Math 104
The end is near!
Application of Series
1. Limits: Series give a good idea of the behavior
of functions in the neighborhood of 0:
We know for other reasons that
sin( x)
lim
1
x 0
x
We could do this by series:

sin( x)
(1) n x 2 n
x 2 x 4 x6
lim
 lim 
 lim 1     1  0  0...  1
x 0
x 0
x 0
x
3! 5! 7!
n 0 (2n  1)!
This can be used on
complicated limits...
Calculate the limit: lim
x 0
A.
B.
C.
D.
E.
0
1/6
1
1/12
does not exist
x  sin( x)
1 e
(  x3 )
Application of series
(continued)
2. Approximate evaluation of integrals: Many integrals
that cannot be evaluated in closed form (i.e., for
which no elementary anti-derivative exists) can be
approximated using series (and we can even
estimate how far off the approximations are).
1
Example: Calculate
e
0
( x2 )
dx to the nearest 0.001.
We begin by...
substituting - x 2 for x in the series we already know for e x ,
and integratin g it. This will give us a numerical series that
1 (  x2 )
converges to the answer : e
dx is approximated by
0
4
6
1
x
x
1
1
1
2
1 x 
 dx  ...  1  

 ...
0
2! 3!
3 5  2! 7  3!


According to Maple...
The last series is an alternating series with
decreasing terms. We need to find the first one
that is less than 0.0005 to ensure that the error
will be less than 0.001. According to Maple:
evalf(1/(7*factorial(3))), evalf(1/(9*factorial(4))),evalf( 1/(11*factorial(5)));
.02380952381, .004629629630, .0007575757576
evalf(1/(13*factorial(6)));
.0001068376068
Keep going...
So it's enough to go out to the 5! term. We do this
as follows:
Sum((-1)^n/((2*n+1)*factorial(n)),n=0..5) = sum((-1)^n/((2*n+1)
*factorial(n)),n=0..5);
5

n 0
( 1) n
( 2 n 1) n!

31049
41580
evalf(%);
.7467291967=.7467291967
and finally...
1
dx  .747 to the
So we get that  e
0
nearest thousandth.
( x2 )
Again, according to Maple, the actual answer
(to 10 places) is
evalf(int(exp(-x^2),x=0..1));
.74669241330
Try this...
Sum the first four nonzero terms to approximate
1
 cos(
0
A. 0.7635
B. 0.5637
C. 0.3567
D. 0.6357
E. 0.6735
x )dx
Series approximations for
functions, integrals etc..
We've been associating series with functions
and using them to evaluate limits, integrals
and such.
We have not thought too much about how
good the approximations are. For serious
applications, it is important to do that.
Questions you can ask-1. If I use only the first three terms of the
series, how big is the error?
2. How many terms do I need to get the error
smaller than 0.0001?
To get error estimates:
Use a generalization of the Mean Value
Theorem for derivatives
Derivative MVT approach:
Recall the mean - value theorem :
f (b)  f (a )
f ' (somewhere between a and b) 
b-a
Set a  0, b  x - - get
f ( x )  f ( 0)
f ' (somewhere between 0 and x) 
x
Rearrange this to get
f( x)  f(0)  f ' (somewhere ) x .
If you know...
If you know that the absolute value of the
derivative is always less than M, then you
know that
| f(x) - f(0) | < M |x|
The derivative form of the error estimate for
series is a generalization of this.
Lagrange's form of the
remainder:
Suppose you write the approximat ion obtained
n
using the terms up to x of the series for f(x),
and let the " remainder" be Rn(x) :
f ( x)  a0  a1 x  a2 x  a3 x  ...  an x  Rn(x)
2
where ak 
f
(k )
( 0)
.
k!
3
n
Lagrange...
Lagrange's form of the remainder looks a lot like
what would be the next term of the series, except
the n+1 st derivative is evaluated at an unknown
point between 0 and x, rather than at 0:
Rn ( x) 
f
( n 1)
( somewhere) n1
x
(n  1)!
So if we know bounds on the n+1st derivative of f,
we can bound the error in the approximation.
Example: The series for sin(x) was:
sin( x)  x 
x3
3!

x5
5!

x7
7!
 ...
If we use the first two (nonzero) terms, we have
sin( x)  x 
x3
3!
4
 R4 ( x)
because the x term of the series is zero anyhow.
5th derivative
For f(x) = sin(x), the fifth derivative is f '''''(x) =
cos(x). And we know that |cos(t)| < 1 for all t
between 0 and x. We can conclude from this that:
R4 ( x) 
x
5
5!
So for instance, we can conclude that the
approximation sin(1) = 1 - 1/6 = 5/6 is accurate
to within 1/5! = 1/120 -- i.e., to two decimal
places.
Your turn...
How accurate is the approximat ion
2
3
.5 .5
e  e  1  .5 

 1.6458333 ?
2! 3!
.5
Now turn the question around - How many terms of the series do we need
to add together to get e to 10 decimal places?
Another application...
Another application of Lagrange's form of the
remainder is to prove that the series of a function
actually converges to the function. For example, for
the series for sin(x), we have (since all the
derivatives of sin(x) are always less than or equal to
1 in absolute value):
x n 1
Rn ( x) 
- - and for any value of x, this quantity
(n  1)!
will approach zero as n goes to infinity. Thus, the error
becomes arbitraril y small - and zero in the limit. So we
(1) n x 2 n 1
are now justified in writing : sin( x)  
n  0 ( 2n  1)!

Shifting the origin -Taylor vs Maclaurin
So far, we've been writing all of our series as
infinite polynomials and using values of the
function f(x) and its derivatives evaluated at
x=0. It is possible to change one's point of
view and use values of the function and
derivatives at other points.
As an example, we’ll return to
the geometric series
1
f ( x) 
 1  x  x 2  x 3  x 4  ...
1 x
If we define a new function g(x)  f(x  1), then we could write
1
1
g ( x)  f ( x  1) 

1  ( x  1)
x
 1  ( x  1)  ( x  1) 2  ( x  1) 3  ( x  1) 4  ...
This expansion would be valid for x between - 2 and 0 (since the f(x)
expansion was valid only for x between - 1 and 1).
Taylor series
By taking derivatives of the function g(x) = -1/x
and evaluating them at x=-1, we will discover
that the expansion of g(x) we have found is the
Taylor series for g(x) expanded around -1:
g(x) = g(-1) + g '(-1) (x+1) + g ''(-1)
( x 1) 2
2!
+ ....
Note:
In general, we have the Taylor expansion of f(x) around x  a :
xa
( x  a) 2
( x  a)3
f( x)  f( a )  f ' (a )
 f ' ' (a)
 f ' ' ' (a)
 ...
1!
2!
3!
Note that this specialize s to our old friend (that is now called the
Maclaurin series) when a  0.
Maclaurin
Series expansions around points other than
zero are useful when trying to approximate
function values for x far from zero, but
close to a different point where much is
known about the function.
But note that by defining a new function g(x)
= f(x+a), you can use Maclaurin expansions
for g instead of general Taylor expansions
for f.
Binomial series
An important series that arises in many applicatio ns.
It is a generaliza tion of the binomial theorem :
 p k
(1  x)     x
k 0  k 
 p
where   is the binomial coefficien t
k
 p
p!
  
. This works and gives
 k  k!( p  k )!
the expansion of (1  x) p if p is a positive integer.
p
p
If p is not a positive integer...
then the same expansion works except it doesn' t stop
(i.e., it gives a series instead of a polynomial ) and we
need a new definition for binomial(p , k).
(1  x) p  1  p x 
p ( p 1)
2!
x2 
p ( p 1)( p  2 )
3!
x 3  ...
For instance, if p  -1, this gives the alternatin g harmonic
1
series!
 1  x  x 2  x 3  ...
1-x
Fibonacci numbers
Everyone is probably familiar with the famous sequence of
Fibonacci numbers. The idea is that you start with 1 (pair of)
rabbit(s) the zeroth month. The first month you still have 1 pair.
But then in the second month you have 1+1 = 2 pairs, the third you
have 1 + 2 = 3 pairs, the fourth, 2 + 3 = 5 pairs, etc... The pattern is
that if you have an pairs in the nth month, and an+1 pairs in the
n+1st month, then you will have a n + a n+1 pairs in the n+2nd
month.
The first several terms of the sequence are thus:
1, 1, 2, 3, 5, 8, 13, 21, 34, 55, etc...
Is there a general formula for a n?
Generating functions
This is a common problem in many parts of
mathematics and science. And a powerful
method for solving such problems involves
series -- which in this case are called
generating functions for their sequences.
For the Fibonacci numbers, we will simply
define a function f(x) via the series:
f ( x)  a0  a1 x  a2 x 2  a3 x 3  ...  1  x  2 x 2  3x 3  5 x 4  ...
Now we have to get the recurrence relation an  2  an  an 1
into the game.
Recurrence relation
To do this, we'll use the fact that multiplication
by x "shifts" the series for f(x) as follows:
f ( x)  a0  a1 x  a2 x 2  a3 x 3  a4 x 4  ...
xf ( x)  a0 x  a1 x 2  a2 x 3  a3 x 4  ...
x 2 f ( x)  a0 x 2  a1 x 3  a2 x 4  ...
Now, subtract the second two from the first -almost everything will cancel because of the
recurrence relation!
The result is...
(1  x  x )f ( x)  a0  (a1  a0 ) x
2
But recall that a0  a1  1. So we have deduced that
1
f ( x) 
!
2
1 x  x
What good does this do?
Further...
If we can figure out the series for
1
1 x  x 2
then we will
have a formula for the Fibonacci numbers, since they
are the coefficien ts of the series.
Partial fractions to the rescue! Factor the denominato r
1  x  x 2  (  x)(   x), where  
(we' ll put the values in later).
5 1
2
and  
5 1
2
Then use partial fractions to write:
1
(  x )(   x )

1
(   )(  x )

1
(   )(   x )
So if we can get the series for
and
1
 x
1
 x
we will be done (almost)!
Work it out...
First
1
 x

1

And
1
 x
1
 1 x 

x
2
 (1   
x

1


x3
3
1


x
2
1


x3
3
x 3

 ...) 
 ...
    (1  
1
 1 x
  
x 2

x

 ...
      ...) 
x 2

x 3

Now, recall that...

5 1
2
and  
about  and  are
1.     5
and
2.  
1

5 1
2
. Two important facts
Our series for f(x) becomes:
f ( x) 
1
5
(     (  2   2 ) x  (  3   3 ) x 2  (  4   4 ) x 3  ...)
This tell us that
an 
 ( n1)  ( 1) n  ( n1)
5
If we put in the known valu es of  and  we will obtain a formula
for the nth Fibonacci number