%X-Sender: zcollins@euclid.ucsd.edu
% More problems

MYPAPERS WAVRIKSTANKUS SYSTEMS anitem.tex
or factorprob.tex


\noindent {\bf TRADITIONAL PROBLEM}  
Given $F$ a rational function on the plane. 
Factor $$F(z) = \theta (z) Q(z)$$

where $\theta (jw)$ is unitary and $Q^{\pm 1}$ is analytic in
R.H.P.

Actually one wants formulas for systems:
GIVEN
$$ A, B, C, 1 \stackrel{F.R.F}{\longrightarrow}\quad
F(z)  $$
FIND
$$1, a, b, c, \stackrel{F.R.F}{\longrightarrow}\quad
\theta(z) \quad {\rm also \ for}\ m$$
AND
$$1, e, f, g, \stackrel{F.R.F}{\longrightarrow}\quad
Q(z)$$


\noindent {\bf Definition}  The inverse system of $ e,f, g, 1$
is  defined to be $ e-fg, -f, g, 1$.
Let $Q^\#$ denote its F.R.F..  Note $QQ^\# = Q^\#Q=1$ 

\noindent 
Thus $R$ is not a pole of  $Q^{\pm 1}$ is equivalent to 
$e-R \ \ and \ \ e-fg-R$ invertible. 
\bigskip
%-------------------------------------------------


All but the last part of this can be {\bf ENCODED AS ALGEBRA }
in ways we understand.

$\theta(I\omega)$ is unitary for all $\omega$ we have done already. 
Namely, there is
a $m$ invertible so that our 3 equations hold?

\begin{verbatim}
old wrong?? order
m <  Inv[m] <  a <  Aj[a] <  Inv[a] <  
Inv[Aj[a]] <  b <  Aj[b]  <  c <  Aj[c]

Starting Relations:

     Rel[0] =  + Inv[m] ** m - ID  ;
     Rel[1] =  + m ** Inv[m] - ID  ;
     Rel[2] =  + Inv[a] ** a - ID  ;
     Rel[3] =  + a ** Inv[a] - ID  ;
     Rel[4] =  + Inv[Aj[a]] ** Aj[a] - ID  ;
     Rel[5] =  + Aj[a] ** Inv[Aj[a]] - ID  ;
     Rel[6] =  + Inv[m] ** b ** c ** m - Inv[m] ** a ** m - Aj[a] ;
     Rel[7] =  + Aj[b] ** Inv[m] - c ;
     Rel[8] =  + m ** Aj[c] - b ;

\end{verbatim}
$F=\theta Q$ is equivalent to: the system for $F$ is similar to
the system for the `product' of the systems $\theta Q$.
That is, 

$F=\theta Q$ is equivalent to: 

\begin{verbatim}

SNC[m11,m12,m21,m22,A11,A12,A21,A22,b2,b1,c1,c2,C1,C2,B1,B2];

the system for $F$ 

A={{A11,A12},{A21,A22}};
B={{b1},{b2}};
C0={{c1,c2}};

is similar , via similarity matrix  

M={{m11,m12},{m21,m22}};

to the 'product'  

AA={{a,b**g},{0,e}};
BB={{b},{f}};
CC={{c,g}};
DD=1;

of the systems $\theta Q$.

Without loss of generality (i think) we can take $A21, m21$
to equal 0.

Better yet in a system problem where $F$ is stable we do not need
to decompose $A$ into 2 parts (a stable and a antistable ) part.
We have therefore only $A_{11}$  $A12, A22, A21$ are all 0.
Also $M$ is a 1x2 matrix {{m11,m12}}, so $m21=0$ and $m22=0$


(******************************)
(* IGNORE *)
diffA=MatMult[M,AA]-MatMult[A,M];

diffC=MatMult[C0,M]-C0;

diffB=MatMult[M,BB]-B;

ru0= {A21->0, m21->0};

diffA0=Sub[diffA,ru0];
diffB0=Sub[diffB,ru0];
diffC0=Sub[diffC,ru0];

(******************************)

(*
Writing down what this similarity means yields

THE PROBLEM 

Given these lists of relations

diffA0=
  {{-A11 ** m11 + m11 ** a, -A11 ** m12 - A12 ** m22 + m12 ** e + 
     m11 ** b ** g}, {0, -A22 ** m22 + m22 ** e}}

diffB0
= {{-b1 + m11 ** b + m12 ** f}, {-b2 + m22 ** f}}

diffC0
={{-c1 + c1 ** m11, -c2 + c1 ** m12 + c2 ** m22}}


Also one wants to add RICATTI on a b c and their transposes and on m.
 

---------------------------------------------
The problem is A B C0 are given
 FIND 
 abc  efg 
maybe some of the m M 's (but I doubt many)
--------------------------------------

We want to use a GB to solve this system of equations.

So the PARTIAL ORDERING might be 
aimed at eliminating a b c 
then e f g
   
A B C0   < m < M < e f g < a tp[a] b c 

So the issue is to find a GB and backsolve for whatever we can
then run GB again then back solve

Fiddle with the order
QUESTION DO WE GET A FINITE GB?
DO WE AFTER BACK SOLVING GET A NICE SMALL SET OF EQUATIONS?
ARE THESE LIKE THE CLASSICAL ANSWER?
*)
??
\end{verbatim}

THE part which is not so clear is how to incorporate 
is $e-R , e-fg-R$ invertible for $R$ in the R.H.P.
 
The easiest case is where $\theta$ is analytic on the RHP.
So  
$$A11 - R \quad\quad{\rm is\ invertible}$$
$$A22 - R \quad\quad {\rm is\ invertible}$$
This sugests ajoining their inverses to the starting relations.
Eventually we want to derive some formula for
 $e-R$ and  $e-fg-R$ where $R$ is in the R.H.P.
from which we can determine (by hand possibly ) that they
are invertible. Or maybe reduce the defining relations 
 $n1(e-R)$ and  $n2(e-fg-R)$ where $R$ is in the R.H.P.
for their inverses with our GB to get $n1,n2$ which are
clearly invertible.

FOR LATER

COMPUTER EXERCISE 2 (later)

$$F(z) = 1+ (C_1 C_2)
\left(\left(
\begin{array}{cc} A_L& 0\\
0&A_R\end{array}\right)- z I\right)^{-1}
\left(
\begin{array}{c}B_1\\
B_2\end{array}\right)
 $$

Do not take $m21=0$.
First we ensure that $Eigs(A_K) \in  L. H. P$,  $Eigs(A_R)$
R.H.P by including in our list of symbols $L,R$ scalars.  
They commute everything and are invertible

$$A_R - L \quad\quad{\rm is\ invertible}$$
$$A_L-R \quad\quad {\rm is\ invertible}$$

JUNK

Here $1, A, B, C$ are best represented as block
matricies.
$$
\left(
\begin{array}{cc}1&0\\
0&1\end{array}\right)
\left(
\begin{array}{cc} A_{11}& A_{12}\\
A_{21}&A_{22}\end{array}\right),
\left(
\begin{array}{c}B_1\\
B_2\end{array}\right) (C_1 C_2)
 $$ 


