**A simple case:**
We let $p=2017$ and look for a "modulus" polynomial $f$ of degree $m=5$ which is irreducible over $F=\mathbb F_p$, associate
its companion matrix $U$ as in *loc. cit.*

https://rua.ua.es/dspace/bitstream/10045/27320/1/Tesis_Diaz_Cardell.pdf ,

and the number field $K=F(a)$ generated by $a$ with $f(a)=0$ in $K$,
and define the map
$$ \psi:K\to M_{m\times m}(F)\ ,$$
to the full space of matrices, which takes $a$ to $U$. So the codomain is bigger in the implementation, but the true image is
the one generated by the powers of $U$. We will test that $\psi$ is a map of rings on some random elements.

The code realizing this situation is as follows:

```
p = 2017
F = GF(p)
R.<X> = PolynomialRing( F )
m = 5
f = R.irreducible_element(m, algorithm='first_lexicographic')
U = companion_matrix(f, format='right')
K.<a> = GF( p**m, modulus=f )
M = MatrixSpace(F, m, m)
psi = K.Hom(M, Rings())(U)
```

This is all needed to define $\psi$ in sage.
Let us investigate the constructed structure in a dialog with the sage interpreter.

```
sage: p
2017
sage: F
Finite Field of size 2017
sage: f
X^5 + X + 11
sage: f.is_irreducible()
True
sage: U
[ 0 0 0 0 2006]
[ 1 0 0 0 2016]
[ 0 1 0 0 0]
[ 0 0 1 0 0]
[ 0 0 0 1 0]
sage: K
Finite Field in a of size 2017^5
sage: K.modulus()
x^5 + x + 11
sage: psi
Ring morphism:
From: Finite Field in a of size 2017^5
To: Full MatrixSpace of 5 by 5 dense matrices over Finite Field of size 2017
Defn: a |--> [ 0 0 0 0 2006]
[ 1 0 0 0 2016]
[ 0 1 0 0 0]
[ 0 0 1 0 0]
[ 0 0 0 1 0]
sage:
```

Now we will pick some random elements in $K$ and test the compatibility of $F$ w.r.t. the structures of
an $F$-algebra on $K$ and $M$.

```
s = F.random_element() # some element in 0, 1, ... , 2016
k1 = K.random_element() # some random element in K
k2 = K.random_element() # an other random element in K... maybe
print "s = %s" % s
print "k1 = %s" % k1
print "k2 = %s" % k2
print "Is psi(s*k1) == s*psi(k1) ? %s" % bool( psi(s*k1) == s*psi(k1) )
print "Is psi(s*k2) == s*psi(k2) ? %s" % bool( psi(s*k2) == s*psi(k2) )
print "Is psi(k1+k2) == psi(k1) + psi(k2) ? %s" % bool( psi(k1+k2) == psi(k1) + psi(k2) )
print "Is psi(k1*k2) == psi(k1) * psi(k2) ? %s" % bool( psi(k1*k2) == psi(k1) * psi(k2) )
```

This gives (this time):

```
s = 422
k1 = 1679*a^4 + 611*a^3 + 639*a^2 + 228*a + 1823
k2 = 242*a^4 + 968*a^3 + 1389*a^2 + 1838*a + 84
Is psi(s*k1) == s*psi(k1) ? True
Is psi(s*k2) == s*psi(k2) ? True
Is psi(k1+k2) == psi(k1) + psi(k2) ? True
Is psi(k1*k2) == psi(k1) * psi(k2) ? True
```

Above, we have for instance:

```
sage: psi(k1)
[1823 1701 1347 1039 1526]
[ 228 144 1090 708 811]
[ 639 228 144 1090 708]
[ 611 639 228 144 1090]
[1679 611 639 228 144]
```

Note that the space $F[U]$ is a subspace of dimension $m$ inside the full matrix space, in our case it is generated
by the $m$ matrices:

```
sage: for power in range(m):
....: print "U^%s is\n%s\n" % ( power, U^power )
....:
U^0 is
[1 0 0 0 0]
[0 1 0 0 0]
[0 0 1 0 0]
[0 0 0 1 0]
[0 0 0 0 1]
U^1 is
[ 0 0 0 0 2006]
[ 1 0 0 0 2016]
[ 0 1 0 0 0]
[ 0 0 1 0 0]
[ 0 0 0 1 0]
U^2 is
[ 0 0 0 2006 0]
[ 0 0 0 2016 2006]
[ 1 0 0 0 2016]
[ 0 1 0 0 0]
[ 0 0 1 0 0]
U^3 is
[ 0 0 2006 0 0]
[ 0 0 2016 2006 0]
[ 0 0 0 2016 2006]
[ 1 0 0 0 2016]
[ 0 1 0 0 0]
U^4 is
[ 0 2006 0 0 0]
[ 0 2016 2006 0 0]
[ 0 0 2016 2006 0]
[ 0 0 0 2016 2006]
[ 1 0 0 0 2016]
sage:
```

**Add-on:**
A real problem may be to realize a situation as the given one in sage for $q$ not a prime number.
Please do not downgrade the answer for the reason of dilution or deviation.
(Mathematically, there is one field of dimension $q$ up to isomorphism, so we pick the one. Practically, we need the isomorphism(s)...)

We have to fix the "alpha", which will become again an $a$ and in code an `a`

in my case.
And we have to construct it in such a way that we know its minimal polynomial over a choice of a
field $\mathbb F_q$. (The situation is not completely trivial when $q$ is not a prime. So my choice will be with $q=2^5=32$.)

We introduce $F=\mathbb F_q=\mathbb F_2(s)$ for an element $s$ of a small degree, $3$ or $5$ say, over $\mathbb F_2$.

Then we pick an irreducible polynomial of degree $m$ over $F$.

```
m = 3
q = 2**3 # feel free to take an other q
F.<s> = GF(q)
R.<X> = PolynomialRing( F )
f = R.irreducible_element(m, algorithm='first_lexicographic')
```

This gives so far:

```
sage: f
X^3 + s*X + s^2
```

We also want the minimal polynomial of $s$, the "modulus" of $F$, we will maybe need it:

```
sage: s.minpoly()
x^3 + x + 1
```

Now we have to introduce the field $K$ with $q^m$ elements. There are two ways to do this in sage:

- Either as
`K.<a> = F.extension( modulus )`

, we know its modulus, but in this case we do not have a method `vector_space`

for K! - Or as
`K.<a> = GF( q^m, modulus=? )`

, and we have to know the modulus of "the same $a$" over $\mathbb F_2$. In this case we can easily get `K.vector_space()`

, this makes it easy to proceed in our case.

(Real problems when dealing with real situations.)
So we will give a solution via `K.<a> = GF( q^m, modulus=? )`

, my choice since the `vector_space`

method makes for me things easier.
Which is the modulus, the minimal polynomial over $\mathbb F_q$ of $a$? The following code finds it, using the first method to find the initial data for the second one.

```
q, m = 2**3, 3
F.<s> = GF(q)
R8.<X> = PolynomialRing( F )
modulus = R.irreducible_element(m, algorithm='first_lexicographic')
K.<a> = F.extension( modulus )
R2.<y> = PolynomialRing( GF(2) )
for f, multiplicity in (y^(2^9)-y).factor():
if f.degree() != 9:
continue
if f(a) == 0:
print f
break
```

This gives the polynomial `y^9 + y^5 + y^4 + y + 1`

very quickly. But in case of a slightly bigger $q,m$ situation, e.g. $q=2^5, m=5$, this
is already a test of patience. This is no longer the question, so i will take this case, we still have to find the place of $s$
in the new incarnation of the one field $K=\mathbb F_{2^9}$.

```
R2.<y> = PolynomialRing( GF(2) )
K.<a> = GF( 2**9, modulus = y^9 + y^5 + y^4 + y + 1 )
R512.<S> = PolynomialRing( K )
factor( a^3 + S*a + S^2 )
```

This gives:

```
(S + a^7 + a^6 + a^5 + 1) * (S + a^7 + a^6 + a^5 + a + 1)
```

Note that:

```
sage: (a^7 + a^6 + a^5 + 1).minpoly()
x^9 + x^7 + x^6 + x^4 + 1
sage: (a^7 + a^6 + a^5 + a + 1).minpoly()
x^3 + x + 1
```

so the second root has to be taken "as a pendant for $s$".

Now we can realize the morphism $\psi$

```
R.<y> = PolynomialRing( GF(2) )
K.<a> = GF( 2**9, modulus = y^9 + y^5 + y^4 + y + 1 )
F.<s> = GF( 2**3, modulus = y^3 + y + 1 )
ks = a^7 + a^6 + a^5 + a + 1
R.<X> = PolynomialRing( F )
f = X^3 + s*X + s^2
U = companion_matrix(f, format='right')
M = MatrixSpace(F, 3, 3)
psi = K.Hom(M, Rings())(U)
```

We can test again the compatibility with the $F$--algebra structure.

```
incl = F.Hom(K, Fields())( ks ) # F -> K, s -> ks
t = F.random_element() # some element in F
k1 = K.random_element() # some random element in K
k2 = K.random_element() # an other random element in K... maybe
kt = incl(t)
print "t = %s" % t
print "kt = %s" % kt
print "k1 = %s" % k1
print "k2 = %s" % k2
print "Is psi(kt*k1) == t*psi(k1) ? %s" % bool( psi(kt*k1) == t*psi(k1) )
print "Is psi(kt*k2) == t*psi(k2) ? %s" % bool( psi(kt*k2) == t*psi(k2) )
print "Is psi(k1+k2) == psi(k1) + psi(k2) ? %s" % bool( psi(k1+k2) == psi(k1) + psi(k2) )
print "Is psi(k1*k2) == psi(k1) * psi(k2) ? %s" % bool( psi(k1*k2) == psi(k1) * psi(k2) )
```

getting:

```
t = s^2 + s + 1
kt = a^8 + a^5 + a^3 + a^2
k1 = a^6 + a^5 + a^4 + a^3 + a^2
k2 = a^8 + a^7 + a^3
Is psi(kt*k1) == t*psi(k1) ? True
Is psi(kt*k2) == t*psi(k2) ? True
Is psi(k1+k2) == psi(k1) + psi(k2) ? True
Is psi(k1*k2) == psi(k1) * psi(k2) ? True
```

Here, again, the codomain is full matrix space, but the image of $\psi$ is generated over $F=\mathbb F_2(s)$ by

```
sage: U^0, U^1, U^2
(
[1 0 0] [ 0 0 s^2] [ 0 s^2 0]
[0 1 0] [ 1 0 s] [ 0 s s^2]
[0 0 1], [ 0 1 0], [ 1 0 s]
)
```

Here, what is $\mathbb F_q[U]$? Is it a subspace of the matrices of type $m\times m$ over $\mathbb F_q$? Can we take the whole space instead?

An explicit example of $\alpha$ and $U$ (that match) would make things easier.

https://rua.ua.es/dspace/bitstream/10045/27320/1/Tesis_Diaz_Cardell.pdf (link text) can you please check Theorem 3.3 in page 51