Prerequisites: field extensions and finite-field examples, matrices of linear maps and the Cayley–Hamilton theorem over a field.
An extension field is also a vector space over its base field. Fixing one field element and multiplying by it therefore defines a linear map. Writing that map as a matrix gives a systematic way to calculate products and inverses, and relates familiar matrix invariants to the arithmetic of a quadratic extension.
From multiplication to a matrix
Let be a field and let be irreducible over . Put and . Irreducibility makes a field. Polynomial division gives the -basis , with the relation
Definition 2.5.1
(Multiplication matrix)
For , let be . The multiplication matrix is the matrix of in the basis .
Proposition 2.5.2
(Multiplication matrices represent the field)
For the map is -linear and
Moreover , , , and is injective. The matrix is invertible exactly when , and then .
Proof
(Images of the basis vectors, then multiplicativity)
is -linear by distributivity and commutativity of field multiplication. To find its matrix in the basis , calculate the images of the basis vectors:
Their coordinate columns give the displayed . Thus multiplication in becomes matrix multiplication: , since . Likewise and . The correspondence is injective: if , applying the two maps to gives . For , multiplication by reverses multiplication by , so is invertible with ; conversely is not invertible.
Common mistake
(Not every invertible matrix is a multiplication matrix)
A matrix of the right size is not automatically in the image of ; its entries must have the displayed form. Over with the multiplication matrices are , only four matrices. The matrix is invertible, but it would need from the top right entry and from the bottom left, so it represents no element of .
Invertibility, trace and determinant
In these coordinates,
Both expressions lie in . The determinant is nonzero for by the invertibility argument; there is no need to guess which coefficient combinations make it vanish. Changing the -basis changes by similarity and preserves trace and determinant. In a finite extension these invariants are called the field trace and field norm of relative to .
Proposition 2.5.3
(Characteristic polynomial of a multiplication matrix)
For , the characteristic polynomial of is
and in . If , then is the minimal polynomial of over .
Proof
(Cayley–Hamilton applied to the vector 1)
By Cayley–Hamilton, . Since preserves sums, products and the identity, , and applying this operator identity to shows . If , then , so and the minimal polynomial of has degree 2. It divides the displayed monic quadratic and therefore equals it. This argument works over the stated field, including characteristic 2; no division by 2 is used.
Proof X-ray
Why the hypothesis on the coefficient of the generator matters
The proof has two layers, and the hypothesis enters only in the second. The vanishing holds for every , because a matrix identity transfers to the element through the injective homomorphism and evaluation at . Turning that vanishing into equality with the minimal polynomial is a degree count and uses once: it forces , hence and degree exactly 2. A monic polynomial of degree 2 that is divisible by the monic minimal polynomial of degree 2 must equal it. Nothing divides by 2 anywhere, which is why the argument survives characteristic 2.
The characteristic polynomial need not be the minimal polynomial of over when .
Counterexample mode
Characteristic polynomial versus minimal polynomial for scalars
Claim: is always the minimal polynomial of over . Take , and , so and . Then , , and , but the minimal polynomial of over is , of degree 1. The characteristic polynomial has degree 2 because , while a scalar satisfies a linear equation. The repair is the hypothesis .
Worked examples
Worked example 2.5.4
(A four-element field)
Take and . It has no root in and is irreducible, so has basis and . For ,
The entries are in : in particular . Since , the inverse is multiplication by , represented by
Direct multiplication gives . The matrix check and the field calculation certify the same inverse in two coordinate systems; neither requires treating as integers modulo 4.
Worked example 2.5.5
(Inverting an element of )
Take and , so , and with , . Then
so . Check: , and . The inverse matrix is , which is for . Indeed .
Worked example 2.5.6
(An element of norm 2 in a nine-element field)
Take and , which has no root in because square to . For , , , , and
Since the norm , is invertible. The characteristic polynomial is in , and gives . The inverse is , since .
Quick check
(In the four-element field with over , compute and for .)
Use , and remember .
Solution · Answer
in , and . This agrees with , which has trace and determinant .
Quick check
(For and , what is the norm of ?)
Use with , .
Solution · Answer
. Directly, has determinant .
Quick check
(For and , write the inverse of and its multiplication matrix.)
Use .
Solution · Answer
Since , we get . With , the matrix satisfies .