Anda di halaman 1dari 6

Math NYB Partial Fractions J. E.

Klassen

NOTE There is no need for you to read and fully understand this. You can just accept
the results you learned in class and use them. If, however, your curiosity gets the better
of you, then by all means push on.

When we deal with the integration of rational functions in Call II, I am often asked why the partial
P( x)
fractions decomposition of works the way it does. Usually I tell students to accept the
Q( x)
result because it is beyond the scope of a Call II course to prove. Well, that is not exactly true.
There is nothing particularly difficult about proving the result, especially for a motivated student,
but it does involve some polynomial theory as well as a knowledge of the Division Algorithm and
the Euclidean Algorithm for polynomials. The Division Algorithm for the integers is something
that high school students should all have seen, except that it was probably not introduced as
such and, certainly, never proved. However, the algorithm for polynomials is just an extension of
the Division Algorithm for the integers to polynomials. Granted, it takes a bit of proving, but the
result can easily be understood if one knows this algorithm for the integers. So, lets prove that
P( x)
we can actually decompose a rational function into partial fractions the way we described
Q( x)
it in class.

Lets start with the Division Algorithm for the integers. This is nothing more than the usual
division of one number by another, which says that we can keep dividing until the remainder is
smaller than the divisor, at which point we stop.

Division Algorithm Let a and b be two integers, with b 0 . Then there are integers d and
r such that a db r where 0 r b

This just says that when we are diving the number a by the number b , that a is the product of
some integer d with b plus a remainder r where the remainder is smaller than the divisor ( b ).

This is something you learned way back in grade school. For example, given the two numbers
59 and 7, then 59 8(7) 3 . Here, d = 8 and r = 3 (the remainder when we divide 59 by 7).

O.K. we wont prove this algorithm, but I think everyone is confident that this can be done. Now,
the Euclidean Algorithm.

Euclidean Algorithm

The Euclidean Algorithm is actually a beautiful result that is used repeatedly in Number Theory
and deals with the greatest common divisor of two integers. In high school you were probably
taught that the way to find the greatest common divisor of two integers is to factor each one into
a product of prime powers and then to choose the largest power of each prime that occurs in
both integers and the product of these is the gcd (greatest common divisor)

Ex. Since 132 22 (3)(11) and 286 (2)(13)(11) then the gcd of 132 and 286 is d = (2)(11)=22

Granted, this works fairly well for two small numbers like 132 and 154 which we can easily
factor into a product of prime powers. How about the gcd of 3934061 and 357761. Hmmm
not so easy now. What are the prime factors of these two numbers? Unless you have a
computer algebra system, this could be a long day. However, how about this (we are here going
to just use repeated long division the Division Algorithm)

3934061 = (10)( 357761) + 356451

357761 = (1)( 356451) + 1310

356451 = (272)(1310) + 131

1310 = (10)(131) + 0 ( the remainder is now 0). Now, the gcd of these two numbers is

d = 131, the last non-zero remainder. Easy, no?

If we perform this operation with the two numbers 132 and 286 we have

286 = 2(132) + 22

132 = 6(22) + 0 and so the gcd is 22, again the last non-zero remainder

Notice that we keep dividing the divisor by the remainder at each step.

This is a part of the Euclidean Algorithm that allows us to find the gcd of two integers by
repeated division. Now the other part of the Euclidean Algorithm says that we can now work
back from the last non-zero remainder in the following way

131 = 356451 - (272)(1310)

= 356451 - (272)(357761 - 356451) = (273)(356451) (272)(357761)

= (273)(3934061 (10)(357761)) (272)(357761)

= (273)(3934061) (3002)(357761)

So, the gcd can be written as some multiple of the first of the two numbers plus some multiple of
the second

Similarly, 22 = (1)(286) (2)(132).

Lets try another one.

Find the gcd of the two numbers 287 and 84.

We have 287 = 3(84) + 35

84 = 2(35) + 14

35 = 2(14) + 7

14 = 2(7) + 0
So the gcd is d = 7

Now, 7 = 35 2(14)

= 35 2[84 2(35)] = 5(35) 2(84)

= 5{287-3(84)] 2(84) = (5)(287) (17)(84)

The Division Algorithm and Euclidean Algorithm for polynomials over the reals.

When we consider the collection of all polynomials with integer coefficients, or rational
coefficients, or real coefficients, together with the usual addition of polynomials and
multiplication of polynomials, they have properties that are strikingly similar to the integers

These properties are

1. The sum of any two polynomials is again a polynomial (called closure)


2. p( x) q( x) q( x) p( x) for any two polynomials (called commutativity)
3. p( x) (q( x) r ( x)) ( p( x) q( x)) r ( x) (called associativity)
4. There is unique polynomial the 0 polynomial which has the value 0 everywhere such
that p( x) 0 0 p( x) p( x) for all polynomials p( x) ( an additive identity)
5. For each polynomial p( x) there is a unique polynomial p( x) , called the additive
inverse of p( x) such that p( x) ( p( x)) p( x) p( x) 0
6. The product of any two polynomials is again a polynomial. (again, closure)
7. p( x)q( x) q( x) p( x) for all polynomials p( x) and q( x) (commutative)
8. p( x)[q( x)r ( x)] [ p( x)q( x)]r ( x) for all polynomials p( x), q( x) and r ( x) (associative)
9. There is a unique polynomial , the polynomial p( x) 1 , the polynomial having the value
1 everywhere for which 1q( x) q( x) for all polynomials q( x) (a multiplicative identity)
10. p( x)[q( x) r ( x)] p( x)q( x) p( x)r ( x) (the distributive property)
11. If p( x)q( x) 0 then either p( x) 0 or q( x) 0 (we say that there are no divisors of 0)
.

Compare these properties to the ones for the integers under addition and multiplication to see
the similarities.

Note that a polynomial is said to be of degree n if the highest power of x in the polynomial is n.
Constant polynomials are said to have degree 0 except for the polynomial p( x) 0 to which we
do not assign a degree. Then we have the additional property that if p( x) and q( x) are two
polynomials (neither one the 0 polynomial) with deg p( x) n and deg q( x) m then
deg p( x)q( x) n m and for any two polynomials p( x) and q( x) that deg( p( x) q( x))
larger of n and m

Now, the Division Algorithm for polynomials says the following: Given two polynomials p(x)
and q(x), there are polynomials d(x) and r(x) such that

p(x) = d(x)q(x) + r(x)

where r(x) = 0 or the degree of r(x) is less than the degree of q(x) (the divisor)
In other words, when we divide one polynomial by another we keep dividing until the remainder
has a smaller degree than the divisor, at which point we stop.

Ex: If p( x) x5 2 x 4 x3 3x 2 5x 9 and q( x) x 2 2 x 5 then

p( x) x5 2 x 4 x3 3x2 5x 9 = ( x3 4 x2 2 x 21)( x 2 2 x 5) (27 x 86)

So the remainder is a polynomial of degree 1.

Now, when we are dealing with the integers we know when one integer is larger than another
and the gcd of two integers is a unique positive integer. The gcd of two numbers is the largest
number that divides both of them and the gcd has the property that any other divisor of the two
numbers divides the gcd.

With polynomials and the gcd, we determine size by the degree . So the gcd of two polynomials
is the polynomial of largest degree that divides both polynomials. However, if, for example, a
polynomial p(x) is divisible by x 2 , then it is also divisible by 2(x 2) = 2x 4. In fact, by any
constant multiple of x 2. So which one is the gcd.? Well, a polynomial is said to be monic if
the coefficient of the highest power of x that appears in the polynomial is a 1 and the greatest
common divisor of two polynomials will then be the polynomial of largest degree that is monic.
In other words, there is not just one gcd of two polynomials, but an infinite number all differing
only by a constant multiple and we take as the gcd the one that is monic.

Now the Euclidean algorithm for polynomials works exactly the same as for the integers. We
keep using repeated division until the remainder is 0 (the 0 polynomial). Then, the last non-zero
remainder, where we multiply the remainder by the appropriate constant so that it is monic, is
the gcd..

To find the gcd of two polynomials we can (in theory) factor each one into a product of
irreducible factors - these correspond to the prime numbers for the integers and then choose
the gcd as we did for the integers. But, as for the integers, that can be a difficult task.

However, we can easily use the Euclidean Algorithm.

Ex. Find the gcd of the two polynomials p( x) x5 x 4 x 1 and q( x) x3 x 2 x 1 . We


have

x5 x4 x 1 ( x2 2 x 1)( x3 x2 x 1) (2 x 2)

1 1
x3 x 2 x 1 ( x 2 x )(2 x 2) 0
2 2

And so the gcd is the polynomial d ( x) x 1 (the last non-zero remainder just as for integers)
where we multiplied 2x+2 by to make it monic.

Now, as for the integers, this allows us to express the gcd as some multiple of the polynomial
p( x) plus some other multiple of the polynomial q( x) . In the above example

(2 x 2) ( x5 x4 x 1) ( x 2 2 x 1)( x3 x2 x 1) and so
1 1
x 1 ( x5 x 4 x 1) ( x 2 2 x 1)( x3 x 2 x 1)
2 2
Finally we are in a position to prove the Partial Fractions Result.

P( x)
Consider a rational function where P( x) has a smaller degree than Q( x) (otherwise we
Q( x)
divide). We also assume that the P( x) and Q( x) have no common factors. Otherwise we could
just cancel these common factors.

Since Q(x) can always be factored (in theory anyway) into a product of irreducible factors then
we consider the case where Q(x) is the product of two factors which are relatively prime
meaning that the greatest common divisor of the two is 1 (they have no common factors)

That is, the polynomial Q( x) is the product of two factors c( x) and d ( x) where these two
factors are relatively prime;

Then by the Euclidean Algorithm we can write

1 n( x)c( x) m( x)d ( x) for some polynomials n( x) and m( x) .

Therefore

P( x) P( x)(1) P( x)[n( x)c( x) m( x)d ( x)] P( x)n( x)c( x) P( x )m( x )d ( x )



Q( x ) Q( x ) c( x)d ( x ) c ( x )d ( x ) c ( x )d ( x )

P ( x ) n ( x ) P ( x ) m( x )
=
d ( x) c( x)

P( x) A( x) B( x)
Which shows that can be written as .
Q( x) c( x) d ( x)

Now, deg P( x)n( x) deg P( x) and deg P( x)m( x) deg P( x) so it may be that
deg A( x) deg c( x) and/or deg B( x) deg d ( x) . So we divide to obtain

P( x) A( x) B( x) E ( x) F ( x) E ( x) F ( x)
= = [ h( x ) ] [ g ( x) ] = [h( x) g ( x)] [ ]
Q( x) c( x) d ( x) c( x) d ( x) c( x) d ( x)
where now,

deg E( x) deg c( x) and deg F ( x) deg d ( x)

Let h( x) g ( x) r ( x)

We claim that h( x) g ( x) r ( x) 0 . Suppose not. Then putting everything over the same
denominator we have that
P( x) E ( x) F ( x)
[h( x) g ( x)] [ ]
Q( x) c( x) d ( x)

E ( x) F ( x) r ( x)c( x)d ( x ) E ( x )d ( x ) F ( x )c (x )
= r ( x)
c( x) d ( x) c( x) d ( x)

However, both E ( x)d ( x) and F ( x)c( x) have degree < deg Q(x) while
deg r ( x)c( x)d ( x) deg Q( x) (since Q( x) c( x)d ( x) ) and so the numerator has degree deg
Q(x). But the numerator must be equal to P( x) and deg P( x) deg Q( x) , a contradiction.
Therefore h( x) g ( x) r ( x) 0 and this part of the proof is established.

Now suppose that one of the factors of Q(x), say c(x) is a power of a linear term, so

c(x) = ( x a)n (here x = a, where a can be either positive or negative, is a repeated root of
multiplicity n). Then we have

P( x) A( x) B( x)

Q( x) ( x a) n
d ( x)

Now, just as the number (as an example)


3942567 3(10)6 9(10)5 4(10)4 2(10)3 5(10)2 6(10)1 7 and so

3942567 3(10)6 9(10)5 4(10)4 2(10)3 5(10)2 6(10)1 7



107 107

2 9 4 2 5 6 7
= 2 3 4 5 6 7
10 10 10 10 10 10 10

So we can express A( x) An1 ( x a)n1 An2 ( x a)n2 An3 ( x a)n3) An1 ( x a) A0


where all the As are constant and then

A( x) An1 ( x a)n1 An2 ( x a)n2 An3 ( x a)n3) A1 ( x a) A0



( x a) n ( x a) n

An1 An2 An3 A1 A0


n 1
. Voila!!
( x a) ( x a ) ( x a )
2 3
(x a) ( x a )n

A similar argument shows that the partial fractions decomposition for repeated quadratic factors
is as was stated in class. Now, with this burning question answered, you can rest easy and
happily integrate rational functions using partial fractions.

Anda mungkin juga menyukai