2.3 Polynomials and Rational Functions

P(z)=a0+a1z+a2z2++anznP(z)=a1+2a2z++nanzn1

Fundamental Theorem of Algebra

Every polynomial of degree n has n roots (counting the multiplicity)
Polynomial division:
P(z)=D(z)Q(z)+R(z)

If α1 is root (P(α1))=0

P(z)=(zα1)Q(z)+R , R must be degree 0 since we are dividing by (zα1)

P(α1)=(α1α1)Q(α)+R=0×Q(α)+R=R=0
P(z)=(zα1)P1(z)

We can repeat this process until we get

P(z)=an(zα1)(zα2)(zαn)

Another method is:

P(z)P(α1)0=anzn++a1z+a0(anα1n+a1α1+a0)=an(znα1n)++a1(zα1)

For any k1,

zkα1k=(zα1)(zk1+zk2α1++α1k1).

Note that:

logP(z)=an+log(zα1)+log(zαn)P(z)P(z)=1zα1++1zαn

Convex Hull

Convex Set:

A set is convex if the the line that joins any two points in a set is also inside the set.

Convex Hull:

Is the smallest convex set that completely encloses a given set of points
image-7.png|420x169

Convex Polygon:

When we have finite points a complex polygon is equivalent to the convex hull if its interior angles are less than π

A point is in a complex npolygon if we can write that point as

m1a1+m2a2++mnan

where a1,,an are the vertices and i=1nmi=1 , mi0

Gauss-Lucas Theorem:

Let P(z) be a complex polynomial with roots α1,α2,,αnC. Then the roots of P(z) lie in the convex hull of α1,α2,,αn

Proof:

We know that P(z)P(z)=i=1n1zαi

Assume that P(z0)=0

0=i=1n1z0αi0=i=1nz0αi|z0αi|2i=1nz0|z0αi|2=i=1nαi|z0αi|2z0i=1n1|z0αi|2=i=1nαi|z0αi|2z0=1i=1n1|z0αi|2i=1nαi|z0αi|2

let mi=1i=1n1|z0αi|21|z0αi|2

mi=(i=1n1|z0αi|2)1(i=1n1|z0αi|2)=1

Rational Functions

A rational function is a quotient of two polynomials

R(z)=P(z)Q(z)

Assuming P and Q have no common factors.
The zeros of Q(z) are the poles of R(z).

We can write:

R(z)=G(z)+P1(z)Q(z) using long division where degree of G = degree(P)-degree(Q)

P(z)=anzn+an1zn1++a0., substitute z=1/z:

P(1/z)=an(1/z)n+an1(1/z)n1++a1(1/z)+a0P(1/z)=anzn+an1zn1++a1z+a0

To clear the fractions, let's find a common denominator is zn:

P(1/z)=an+an1z+an2z2++a1zn1+a0znzn=callP1(z)

Step 2: Analyze the Denominator Q(1/z)

Q(z)=bmzm++b0.

Substitute z=1/z and find the common denominator zm:

Q(1/z)=bm+bm1z++b1zm1+b0zmzm=callQ1(z)

Step 3: Combine them to find R(1/z)

Now we just divide the two results:

R(1/z)=P(1/z)Q(1/z)=an+an1z++a0znznbm+bm1z++b0zmzm

To simplify this complex fraction,

R(1/z)=(an+an1z++a0znzn)(zmbm+bm1z++b0zm)

Now, just group the z factors:

R(1/z)=zmznan+an1z++a0znbm+bm1z++b0zmR(1/z)=zmnan+an1z++a0znbm+bm1z++b0zm{m>n, is a 0 of order mnm<n, is a pole of order nmm=n,R()=anbm0,

Pole

A pole of R(z) is a point where R(z)= or Q(z)=0 but P(z)0

Order of a Pole

The order of the corresponding zero of Q(z)

If α is a zero of Q so that Q(z)=(zα)nQ1(z) and α is not a root of Q1(z) then n is the order of the pole.

The Partial Fraction Decomposition Theorem

Theorem: Any rational function R(z)=P(z)/Q(z) can be uniquely decomposed into the form:

R(z)=G(z)+j=1nGj(1zzj)

where z1,,zn are the distinct poles of R(z), G(z) is a polynomial, and each Gj is a polynomial without a constant term.


Proof

Step 1. Isolate the Polynomial Part (Behaviour at )

Using polynomial long division, we can always divide P(z) by Q(z) to get a quotient G(z) and a remainder P1(z):

R(z)=P(z)Q(z)=G(z)+P1(z)Q(z)

By the definition of long division, deg(P1)<deg(Q).

Our new goal is to decompose the proper fraction R1(z).


Step 2. Isolate the Behaviour at a Single Pole z1

Let z1 be a pole of R1(z). We want to find the "principal part" G1 associated with this pole.

F(ζ)=R1(z)=R1(z1+1/ζ)

Since R1(z) has a pole at z1, F(ζ) must have a pole at ζ=.

F(ζ)=G1(ζ)+F1(ζ)

where G1(ζ) is a polynomial and F1(ζ) is a proper rational function (so limζF1(ζ)=0).

limζ0F(ζ)=limζ0(G1(ζ)+F1(ζ))=limzR1(z)=0

This implies G1(0)+F1(0)=0. This forces G1(ζ) to have no constant term (or F1(0) to cancel it). For simplicity, we define G1 as the polynomial in ζ without a constant term that captures the pole at .

R1(z)=F(1zz1)=G1(1zz1)+F1(1zz1)

Step 3. Analyze the New Remainder (Induction)

Let H1(z)=F1(1/(zz1)).

We have successfully "peeled off" the pole at z1. We can now write:

R1(z)=G1(1zz1)+H1(z)

where H1(z) is a new rational function that contains all the other poles (z2,,zn), but is analytic at z1.

We repeat this process (inductively) for H1(z) and the pole z2, then for H2(z) and z3, and so on.


Step 4. The Final Remainder

After n steps, we have "peeled off" all n poles:

R1(z)=G1(1zz1)++Gn(1zzn)+Hn(z)R1(z)=j=1nGj(1zzj)+Hn(z)

This final remainder Hn(z) is a rational function.

  1. It has no finite poles (we removed them all). A rational function with no poles must be a polynomial.

  2. It must be a proper fraction, since it's the sum/difference of other proper fractions.

    • limzR1(z)=0 (by definition).
    • limzGj(1zzj)=Gj(0)=0 (since Gj has no constant term).
    • Therefore, limzHn(z)=limz(R1(z)Gj(1zzj))=00=0.

A function Hn(z) that is both a polynomial and tends to 0 as z must be the zero polynomial: Hn(z)=0.


Conclusion

This leaves us with:

R1(z)=j=1nGj(1zzj)

Substituting this back into the result from Step 1 gives the final theorem:

R(z)=G(z)+R1(z)=G(z)+j=1nGj(1zzj)

Example

Decompose R(z)=z4(z1)3(z+1)

This fraction is improper because the degree of the numerator (4) is equal to the degree of the denominator (3+1=4).

Target Form:

The decomposition will look like this:

R(z)=G(z)+A(z1)3+B(z1)2+Cz1+Dz+1

1. Polynomial Part G(z)

We use long division. First, expand the denominator:

Q(z)=(z33z2+3z1)(z+1)=z42z3+2z1

Now, divide z4 by this:

1z42z3+0z2+2z1)z4+0z3+0z2+0z+0(z42z3+0z2+2z1)a2z3+0z22z+1

We now only need to decompose the remainder R1(z).


2. Find Coefficient D (Simple Pole z=1)

We can use the simple "cover-up" method for the simple pole.

D=limz1(z+1)R1(z)=limz12z32z+1(z1)3D=2(1)32(1)+1(11)3=2+2+1(2)3=18

D=1/8


3. Find Coefficients A,B,C (Order 3 Pole z=1)

We use the general "Taylor series" method.

  1. Define Φ(z): "Clear" the pole by multiplying R1(z) by (z1)3:
Φ(z)=(z1)3R1(z)=2z32z+1z+1
  1. Match Forms: We know that...
Φ(z)=A+B(z1)+C(z1)2+

..and we also know from Taylor's theorem that...

Φ(z)=Φ(1)+Φ(1)1!(z1)+Φ(1)2!(z1)2+
  1. Find Coefficients:
    • A=Φ(1)
    • B=Φ(1)/1!
    • C=Φ(1)/2!

Let's calculate these:

To find A:

A=Φ(1)=2(1)32(1)+11+1=12

A=1/2

To find B:

First, find Φ(z) using the quotient rule on 2z32z+1z+1:

Φ(z)=(6z22)(z+1)(2z32z+1)(1)(z+1)2=4z3+6z23(z+1)2

Now, plug in z=1:

B=Φ(1)=4(1)3+6(1)23(1+1)2=4+634=74

B=7/4

To find C:

First, find Φ(z) by differentiating Φ(z)=4z3+6z23(z+1)2:

Φ(z)=(12z2+12z)(z+1)2(4z3+6z23)(2(z+1))(z+1)4

Simplify by canceling a (z+1) from all terms:

Φ(z)=(12z2+12z)(z+1)2(4z3+6z23)(z+1)3Φ(z)=(12z3+24z2+12z)(8z3+12z26)(z+1)3=4z3+12z2+12z+6(z+1)3

Now, plug in z=1:

Φ(1)=4+12+12+6(1+1)3=348=174

Finally, remember C=Φ(1)/2!:

C=17/42=178

C=17/8


4. Final Result

Assemble all the pieces:

z4(z1)3(z+1)=1+1/2(z1)3+7/4(z1)2+17/8z11/8z+1R(z)=1G(z)+[1/2(z1)3+7/4(z1)2+17/8z1]Pole 1+[1/8z+1]Pole 2