 
 
 
 
 
   
 2. So we put in
an additional hypothesis that matrix multiplication between elements
of K is commutative.
 2. So we put in
an additional hypothesis that matrix multiplication between elements
of K is commutative.
Now, consider the map 
 
  
  . v where v is any
(fixed) non-zero column vector such as the transpose of
(1, 0..., 0). When
 . v where v is any
(fixed) non-zero column vector such as the transpose of
(1, 0..., 0). When  and
 and  are an elements of K with
 are an elements of K with
 . v =
 . v =  . v, we obtain 
(
 . v, we obtain 
( -
 -  ) . v = 0. But
we have assumed that every non-zero element of K is invertible so we
must have
) . v = 0. But
we have assumed that every non-zero element of K is invertible so we
must have 
 -
 -  = 0. In other words this map is   one-to-one on K. Thus K is actually isomorphic to a vector
space of rank at most n over the rationals. By a suitable change of
basis (and restricting to a submatrix) we may as well assume that the
space K . v contains all column vectors or equivalently
that K has rank n. Then K . w is the space of all column
vectors for any non-zero vector w.  We will henceforth make
this additional assumption as well.
 = 0. In other words this map is   one-to-one on K. Thus K is actually isomorphic to a vector
space of rank at most n over the rationals. By a suitable change of
basis (and restricting to a submatrix) we may as well assume that the
space K . v contains all column vectors or equivalently
that K has rank n. Then K . w is the space of all column
vectors for any non-zero vector w.  We will henceforth make
this additional assumption as well.
For any n×n matrix  we have (the Cayley-Hamilton
theorem) that characteristic polynomial 
ch
 we have (the Cayley-Hamilton
theorem) that characteristic polynomial 
ch (T) of
degree n and 
ch
(T) of
degree n and 
ch (
( ) = 0. (In the words of one
mathematician khudh kaa nahi satisfy karega to kiska satisfy
  karega?(Hindi); if it doesn't satisfy its' own then whose will it
satisfy?). On the other hand, we have the minimal
polynomial
) = 0. (In the words of one
mathematician khudh kaa nahi satisfy karega to kiska satisfy
  karega?(Hindi); if it doesn't satisfy its' own then whose will it
satisfy?). On the other hand, we have the minimal
polynomial 
 (T), which is the polynomial of least degree
with rational coefficients that is satisfied by
(T), which is the polynomial of least degree
with rational coefficients that is satisfied by  . If
. If
 (T) = P(T)Q(T), then 
P(
(T) = P(T)Q(T), then 
P( )Q(
)Q( ) = 0. Since,
P(
) = 0. Since,
P( ) and Q(
) and Q( ) are in K at least one of them must be
zero thus one of them must be a constant; in other words the minimal
polynomial is irreducible. It also follows as before that it
divides the characteristic polynomial. One can show that, under the
hypothesis of the previous paragraph (and the fact the we are working
over rationals; a perfect field), there is an element
) are in K at least one of them must be
zero thus one of them must be a constant; in other words the minimal
polynomial is irreducible. It also follows as before that it
divides the characteristic polynomial. One can show that, under the
hypothesis of the previous paragraph (and the fact the we are working
over rationals; a perfect field), there is an element  in K whose characteristic polynomial is irreducible, i. e. 
its characteristic polynomial equals its minimal polynomial. In
particular, the field K has a basis over the field
in K whose characteristic polynomial is irreducible, i. e. 
its characteristic polynomial equals its minimal polynomial. In
particular, the field K has a basis over the field 
 of
rationals of the form 1,
 of
rationals of the form 1,  , ...,
, ..., 
 .
.
 ,
,  , ...,
, ...,
   be a basis of K over the field
 be a basis of K over the field 
 . Consider the
  characteristic polynomial of 
T1
. Consider the
  characteristic polynomial of 
T1 + ... + Tn
 + ... + Tn as
  a function of the variables T1, ..., Tn. The condition
  that this is reducible will impose certain non-trivial polynomial
  relations between the Tk's. Thus all we need to do is to find
  rational numbers rk that do not satisfy these relations. Then the
  characteristic polynomial of
 as
  a function of the variables T1, ..., Tn. The condition
  that this is reducible will impose certain non-trivial polynomial
  relations between the Tk's. Thus all we need to do is to find
  rational numbers rk that do not satisfy these relations. Then the
  characteristic polynomial of 
 = r1
 = r1 + ... + rn
 + ... + rn will be irreducible (and of degree n. It follows
  that, the elements 1,
 will be irreducible (and of degree n. It follows
  that, the elements 1,  , ...,
, ...,
 will be
  independent over
 will be
  independent over 
 .
.
  
 whose characteristic polynomial P(T)
is equal to its minimal polynomial. An further extension of the above
argument then shows that any invertible matrix g that commutes
with every element of K is contained in K; we will use this in
later sections.
 whose characteristic polynomial P(T)
is equal to its minimal polynomial. An further extension of the above
argument then shows that any invertible matrix g that commutes
with every element of K is contained in K; we will use this in
later sections.
As an example, let us consider the ``construction'' of the field associated with an irreducible polynomial P(T) = Tn + a1Tn - 1 + ... + an. We consider the matrix
 =
 =  
 is the
required field
 is the
required field 
 (
( ), sometimes also denoted by
), sometimes also denoted by
 [T]/(P(T)) (one uses Euclid's algorithm for polynomials to show
that every non-zero element of this is invertible). The above
discussion says that any field under consideration is isomorphic
to a field of this form for some irreducible polynomial P(T).
[T]/(P(T)) (one uses Euclid's algorithm for polynomials to show
that every non-zero element of this is invertible). The above
discussion says that any field under consideration is isomorphic
to a field of this form for some irreducible polynomial P(T).
 
 
 
 
