Lecture 21 - Polynomial Spaces (4.A)

Note for today that we denote:

TT=T2

and in general:

TTTk times =Tk

But we can define a polynomial operator like:

T4+T32T2

as some new operator:

(T4+T32T2)v=T4v+T3v2T2v

We'll revisit these, but for this section, we'll care a lot about polynomials, and specifically how they factor, which will be important then.

Note

We go from right to left for our transformations. However, for polynomials since TT is commutative then for our polynomials they don't matter.

Properties of C

We talked about some properties of C, most of which you know. Here z=a+bi:

  1. z+z¯=2R(z), where R denotes extracting the real part from z.
  2. zz¯=2I(z), where I denotes extracting the imaginary part from z.
  3. zz¯=|z|2, where |z|=a2+b2.
  4. w+z=w¯+z¯,wz=w¯z¯
  5. z¯¯=z
  6. |R(z)||z|,|I(z)||z|
  7. |z¯|=|z|
  8. |wz|=|w||z|
  9. (Triangle Inequality) |w+z||w|+|z|

We'll use these sporadically as we go on.

Properties of Polynomials

Division Algorithm for Polynomials

Given polynomials p,sP(F), there exist unique polynomials q,rP(F) such that:

p=sq+r

where deg(r)<deg(s).

Notice that this is the standard division algorithm except with polynomials. Just like how we did the division algorithm to generate q,r, we can do polynomial division to generate our q,r polynomials here.

Note

If p(λ)=0 then (xλ) is a factor of p, in other words:

p(x)=(xλ)q(x)

This is actually the case from the division algorithm itself. Note that if degs=1 then we are guaranteed that deg(r)=0 so then r is a constant polynomial (constant function).

Let's do an example. We can can q,r through long division of p=x4+3x2x2+2x+2 and q=x2x+1:

x2+4xx2x+1x4+3x3x2+2x+2x4x3+x24x32x24x34x2+4x2x22x2x22x+20.

Thus here q=x2+4x,s=x2x+1 while r=0. Hence:

sq+r=(x2x+1)(x2+4x)+0

But we can now prove the division algorithm for polynomials:

Proof
Let n=deg(p) and m=deg(s). If n<m then deg(p)<deg(s) so then q=0 and r=p via the algorithm above, which works:

p=s0+r=s0+p

Now suppose nm. Define Ts:Pnm(F)×Pn1(F)Pn(F) by:

Ts(q,r)=sq+r

We can verify that it's a linear transformation pretty quickly:

Factorization of Polynomials Over F

Roots and Factors of Polynomials

  1. Consider pP(F). A number λF is called a zero, or root, of p if p(λ)=0.
  2. A polynomial sP(F) is called a factor of p if there exists a polynomial of q such that p=sq.

Theorem

λ is a zero of p iff (zλ) is a factor of p.

Proof
Consider , so suppose zλ is a factor of p. Thus p=(zλ)q for some q. But then:

p(λ)=(λλ)q=0q=0

thus λ is a zero of p.

Consider , so suppose λ is a zero of p, so p(λ)=0. Using the division algorithm for polynomials, apply this to p, where p=p and s=(zλ). What do we get? We get that there must exist some q,r where deg(r)<deg(s) such that:

p=(zλ)q+r

we need to show that r=0. We know that s=(zλ) so then deg(s)=1 so that forces that deg(r)=0 so then r=cF (ie: r is a scalar). But notice that:

p(λ)=(λλ)q(λ)+r(λ)=0+r(λ)=0

so then r(λ)=c=0 so r=0 as required. So then p=sq so then s is a factor of p.

Recall from our prior example that:

x4+3x3x2+x+1=(x2+4x)(x2x+1)+0

so here x=0,x=4 are zeroes of this original polynomial, as x2+4x=(x0)(x+4) are factors of our original polynomial.

Fundamental Theorem of Algebra

For any pP(F) with deg(p)=m0, then p has at most m distinct zeroes in F.

Theorem

Every non-constant polynomial with complex coefficients has a zero.

We don't prove the latter theorem, but it introduces us factoring in C:

Factoring in C

Let pP(C) is an non-constant polynomial with deg(m). Then p can be factored uniquely (up to order m) as:

p(z)=c(zλ1)(zλ2)(zλm)

where c,λ1,,λmC

But crazy enough, this doesn't work for R!

Factoring with R coefficients.

Let pP(C), but only with real coefficients. Then λ is a zero of p iff λ¯ is a zero of p.

The proof of this is pretty simple.

Factorization

Suppose b,cR. Then:

x2+bx+c=(xλ1)(xλ2)

with λ1,λ2R iff b24c.

Factoring in P(R)

Let pP(R) is a non-constant polynomial. Then p can be factored uniquely (up to order) as:

p(x)=c(xλ1)(xλm)(x2+b1x+c1)(x2+bmx+cm)

where c,λ1,,λm,b1,c1,,bm,cmR with bj2<4cj.