[*] up [*] contents
Next: 3. as example Up: Finite Fields and Zech's Previous: 1. Introduction   Contents

Subsections

2. Finite Fields

In the present section, some elementary results are collected about finite fields, and some broad outlines of their proofs is given. A less cursive approach of these questions can be found in [Lidl, 9] or in [10].

2.1 Characteristic

Let \( \mathrm{GF}\! \left( q \right) \) be a finite field of cardinal \( q \). Let \( \gamma \) be the (unique) ring homomorphism from \( \mathbb {Z}\) into \( \mathrm{GF}\! \left( q \right) \). Since \( \mathrm{GF}\! \left( q \right) \) is finite, \( \ker \gamma \) cannot be the \( \left\{ 0\right\} \) ideal and, since \( \mathbb {Z}\left/ \ker \gamma \right. =\gamma \left( \mathbb {Z}\right) \) is an integral domain, \( \ker \gamma \) is a prime ideal. The (positive) generator of that ideal is the so-called characteristic of the given field, and it is to be noticed that any \( \mathrm{GF}\! \left( q \right) \) contains some prime field \( \mathrm{GF}\! \left( p \right) =\mathbb {Z}\! \left/ p\right. \) as subfield.

2.2 Order and degree of an element

Let \( x \) be given in \( \mathrm{GF}\! \left( q\right) ^{*}\), and \( \gamma \) denotes now, the mapping \( \mathbb {Z}\hookrightarrow \mathrm{GF}\! \left( q\right) ^{*}\) defined as \( \gamma \, n=x^{n} \). Then again \( \gamma \) is a group homomorphism, and since that mapping cannot be injective, the ideal \( \ker \gamma \) is a \( n\mathbb {Z}\) for some \( n\geq 1 \). That \( n \) is the so-called (multiplicative) order of \( x \), and will be denoted as \( \mathrm{ord}x \).

As a consequence, any \( x\in \mathrm{GF}\! \left( q\right) ^{*}\) is a root of the polynomial: \( X^{\mathrm{ord}x}-1 \), which obviously belongs to \( \mathbb {Z}\! \left/ p\right. \left[ X\right] \). Hence each and every element of \( \mathrm{GF}\! \left( q \right) \) is algebraic over \( \mathbb {Z}\! \left/ p\right. \), and thus \( \mathrm{GF}\! \left( q \right) \) is a subfield of the algebraic closure of \( \mathbb {Z}\! \left/ p\right. \). In short: \( \mathbb {Z}\! \left/ p\right. \subset \mathrm{GF}\! \left( q \right) \subset \mathrm{Alg}\left( \mathbb {Z}\! \left/ p\right. \right) \).

Let be given an \( y\in \mathrm{GF}\! \left( q \right) \). Since \( \mathbb {Z}\! \left/ p\right. \left[ X\right] \) is principal, the set \( \left\{ P\in \mathbb {Z}\! \left/ p\right. \left[ X\right] \, \left\vert \, P\left( y\right) =0\right. \right\} \) is an ideal generated by an unique monic polynomial. That polynomial is obviously irreducible and will be referred as the ``minimal polynomial for \( y \)''. We denote it by \( P_{y} \) and put \( \deg y=\deg P_{y} \).

2.3 Primitive elements

Any \( \mathrm{GF}\! \left( q\right) ^{*}\) is a cyclic group [3], i.e., it exists at least one primitive element \( \xi \) such that \( \mathrm{GF}\! \left( q\right) ^{*}=\xi ^{\mathbb {Z}} \). A sketch of the proof is as follows :

  1. The ``small'' Fermat's theorem states that \( \forall x\in \mathrm{GF}\! \left( q\right) ^{*}\, :\, \mathrm{ord}x\, \vert\, \left( q-1\right) \). This relation, that stands for any finite group \( G \), follows from the bijectivity of the translation \( \tau \, :\, y\mapsto x\, y \). We must have \( \Pi _{y\in G}\left( x\, y\right) =\Pi _{y\in G}y \), i.e. \( x^{\char93 G}=1 \) and thus the order of any element divides the cardinal of the group.
  2. Let us now define \( \theta \left( d\right) \) as the number of \( x\in \mathrm{GF}\! \left( q\right) ^{*}\) such that \( \mathrm{ord}x=d \), and prove that \( \theta \left( d\right) \) is either \( 0 \) or \( \varphi \left( d\right) \), where \( \varphi \) is the Euler's indicator. Assuming \( \theta \left( d\right) \neq 0 \), let \( \xi \) be such that \( \mathrm{ord}\xi =d \) and compare the set \( \Omega \doteq \left\{ x\, \left\vert \, x^{d}=1\right. \right\} \) with the group \( \xi ^{\mathbb {Z}} \) generated by \( \xi \). The inclusion \( \xi ^{\mathbb {Z}}\subset \Omega \) is obvious. On the other hand, \( \char93 \, \xi ^{\mathbb {Z}}=d \) while \( \char93 \, \Omega \leq d \) since the polynomial \( X^{d}-1 \) should not have more roots than its degree. And the lemma follows, since there are exactly \( \varphi \left( d\right) \) primitive elements in a cyclic group with cardinal \( d \).
  3. We have obviously \( \sum _{d\in \mathbb {N}}\theta \left( d\right) =q-1 \) (any element has an order), while Euler's theorem states that \( \sum _{d\vert q-1}\varphi \left( d\right) =q-1 \). Thus \( \theta \left( d\right) =\varphi \left( d\right) \) when \( d \) divides \( q-1 \).
As conclusion, it exists exactly \( \varphi \left( q-1\right) \) primitive elements \( \xi \in \mathrm{GF}\! \left( q\right) ^{*}\) : theses elements are quite frequent !

2.4 Descriptions of a given finite field

Collecting all these results, we get that any finite field is algebraically generated over its prime field \( \mathbb {Z}\! \left/ p\right. \) by some (proper) element \( \alpha \), and thus is isomorphic to some quotient field of \( \mathbb {Z}\! \left/ p\right. \left[ X\right] \). As a consequence, \( \mathrm{GF}\! \left( q \right) \) is a vector space over \( \mathbb {Z}\! \left/ p\right. \), proving that \( q \) is a power of a prime, i.e., \( q=p^{m} \). Moreover \( \mathrm{GF}\! \left( q\right) ^{*}\) cyclic, and is thus multiplicatively generated by some (primitive) element \( \xi \). As a summary, we get:

$\displaystyle \mathrm{GF}\! \left( q \right)$ $\textstyle =$ $\displaystyle \left\{ x\in \mathrm{Alg}\left( \mathbb {Z}\! \left/ p\right. \right) \, \left\vert \, x^{q}=x\right. \right\}$  
$\displaystyle \exists \alpha \, :\, \mathrm{GF}\! \left( q \right)$ $\textstyle \sim$ $\displaystyle \mathbb {Z}\! \left/ p\right. \left[ X\right] \, \left/ \, P_{\alpha }\left( X\right) \right.$ (1)
$\displaystyle \exists \xi \, :\, \mathrm{GF}\! \left( q \right)$ $\textstyle =$ $\displaystyle \left\{ 0\right\} \cup \left\{ \xi ^{k}\, \left\vert \, 0\leq k\leq q-2\right. \right\}$  

2.5 Frobenius automorphism

The elementary binomial formula \( \left( a+n\right) ^{n}=\sum _{0}^{n}{n \choose k}a^{k}b^{n-k} \) comes to a nice result when \( n \) is taken equal to the characteristic of \( \mathrm{GF}\! \left( q \right) \). Since any prime \( p \) as the property to divide any of the binomials \( {p \choose k} \) except for the ending ones, the formula shortens to \( \left( a+b\right) ^{p}=a^{p}+b^{p} \). Thus the mapping \( \psi \, :\, x\mapsto x^{p} \) is a morphism of \( \mathrm{GF}\! \left( q \right) \) into itself. But \( a^{p}-b^{p}=0 \) implies \( \left( a-b\right) ^{p}=\left( a-b\right) =0 \) and \( \psi \) is injective and therefore bijective.

The Frobenius mapping \( \psi \) induces an automorphism over \( \mathrm{GF}\! \left( q \right) \).

Let \( x \) be some element of \( \mathrm{Alg}\left( \mathbb {Z}\! \left/ p\right. \right) \), \( P_{x} \) as minimal polynomial, and \( n=\deg x \). Since the coefficients of \( P_{x} \) are in \( \mathbb {Z}\! \left/ p\right. \), they are \( \psi \)-invariant, and we get \( P_{x}\left( \psi \left( x\right) \right) =\left( \psi P_{x}\right) \left( \psi \left( x\right) \right) =\psi \left( P_{x}\left( x\right) \right) =0 \). Any \( \psi ^{k}\left( x\right) \) is therefore a root of \( P_{x} \), and since \( \deg P=n \), there are at most \( n \) different \( \psi ^{k}\left( x\right) \). On the other hand, \( \psi ^{k}\left( x\right) =\psi ^{j}\left( x\right) \) leads to \( \psi ^{m}\left( y\right) =y \) where \( m=\left\vert k-j\right\vert \) and \( y \) is a root of \( P_{x} \). But \( \psi ^{m}\left( y\right) =y \) is \( y^{p\: \widehat{\: }\, m}=y \), implying \( y\in \mathrm{GF}\! \left( p^{m} \right) \) and \( \deg y\leq m \). Therefore, \( n \) divides \( k-j \) : there are exactly \( n \) different \( \psi ^{k}\left( x\right) \).

The elements of a \( n \)-sized \( \psi \)-orbit are the roots of an irreducible polynomial of degree \( n \) over \( \mathbb {Z}\! \left/ p\right. \).

2.6 Subfields and subgroups

Therefore \( \mathrm{GF}\! \left( p^{m} \right) \) is the subfield of \( \mathrm{Alg}\left( \mathbb {Z}\! \left/ p\right. \right) \) that is fixed by \( \psi ^{m} \), leading to the following embedding result \( \left[ \mathrm{GF}\! \left( p^{n} \right) \subset \mathrm{GF}\! \left( p^{m} \right) \right] \Leftrightarrow \left[ n\, \vert\, m\right] \). That result can also be seen as resulting from the divisibility of one characteristic polynomial by the other.

Let us now consider all the multiplicative subgroups of a given \( \mathrm{GF}\! \left( 2^{m} \right) ^{*} \). Since that group is cyclic, it contains exactly one subgroup of cardinal \( d \) for each \( d \) that divides \( q-1 \). Some of them are a field-group \( \mathrm{GF}\! \left( 2^{n} \right) ^{*} \) where \( n\vert m \), but most of them are not. Therefore, in Fig. 1, the subfields lattice (boldfaced) is a sublattice of the multiplicative groups lattice.

Figure: Subgroups of \( \mathrm{GF}\! \left( 256 \right) ^{*}\protect \).
\resizebox* {!}{0.25\textheight}{\includegraphics{struct.eps}}

The proper elements of such a \( d \)-sized subgroup are exactly the \( \varphi \left( d\right) \) roots of \( \Phi _{d}\left( X\right) \), the cyclotomic polynomial relative to the \( d \) exponent. As an example, when \( q=256 \), we get \( q-1=3\times 5\times 17 \) and therefore the characteristic polynomial of \( \mathrm{GF}\! \left( 256 \right) \) splits into \( X^{256}-X=X\times \Phi _{1}\times \Phi _{3}\times \Phi _{5}\times \Phi _{15}\times \Phi _{17}\times \Phi _{51}\times \Phi _{85}\times \Phi _{255} \).

Moreover, since \( \Phi _{15} \) describes the primitive elements of \( \mathrm{GF}\! \left( 2^{4} \right) \), it splits into two polynomials of degree 4, and since \( \Phi _{17},\, \Phi _{51},\, \Phi _{85},\, \Phi _{255} \) describe proper elements of \( \mathrm{GF}\! \left( 2^{8} \right) \), they split over \( \mathbb {Z}\! \left/ p\right. \) into irreducible polynomials of degree 8.

2.7 Trace of an element

The so-called trace of an element in \( \mathrm{GF}\! \left( q \right) \) is defined as \( \mathrm{Tr}_{m}\left( x\right) =\sum _{0}^{m-1}\, \psi ^{k}\left( x\right) \). Thus, for a proper element \( x \) of \( \mathrm{GF}\! \left( q \right) \), \( \mathrm{Tr}_{m}\left( x\right) \) is the sum of all the roots of the minimal polynomial of \( x \). For an \( x \) of smaller degree, that sum is to be multiplied by the natural integer \( m\, /\deg x \) and then reduced modulo \( p \). Elementarily, \( \mathrm{Tr}_{m}\left( x+y\right) =\mathrm{Tr}_{m}\left( x\right) +\mathrm{Tr}_{m}\left( y\right) \).

2.8 Discrete logarithm and Zech logarithm

Given a primitive element \( \xi \), there is a one to one correspondence between the \( k\in \mathbb {Z}\left/ \, q-1\right. \) and the \( \xi ^{k}\in \mathrm{GF}\! \left( q\right) ^{*}\). With the convention \( 0=\xi ^{\infty } \), we get a bijection \( L_{\xi } \) from \( \mathrm{GF}\! \left( q \right) \) towards \( \mathbb {N}_{q}\doteq \mathbb {Z}\left/ \, q-1\right. \cup \left\{ \infty \right\} \) (since \( \mathbb {Z}\left/ \, q-1\right. \) can not be ordered as a group, the traditional notation, `` \( 0=\xi ^{-\infty } \)'', is rather confusing).

That bijection \( L_{\xi } \) can be used to carry the field structure of \( \mathrm{GF}\! \left( q \right) \) onto \( \mathbb {N}_{q}\). Obviously, the image of \( \times \) inside \( \mathrm{GF}\! \left( q \right) \) is \( + \) inside \( \mathbb {N}_{q}\). But, since integer arithmetic modulo \( q-1 \) is far easier than computations of polynomials modulo \( P_{\xi }\left( X\right) \) over \( \mathbb {Z}\! \left/ p\right. \), it is of great interest to consider the isomorphic image in \( \mathbb {N}_{q}\) of the first law (addition) in \( \mathrm{GF}\! \left( q \right) \).

Let us denote by \( \mathbb {A}\) that first law over \( \mathbb {N}_{q}\). From \( \xi ^{a}+\xi ^{b}=\xi ^{c}\Rightarrow \xi ^{a+k}+\xi ^{b+k}=\xi ^{c+k} \), we get \( \left( a+k\right) \mathbb {A}\left( b+k\right) =\left( a\mathbb {A}b\right) +k \), and therefore the whole \( \mathbb {A}\) table is known when the function \( Z\, :\, x\mapsto x\mathbb {A}0 \) is known. The so-called Zech logarithm is nothing but precisely that function \( Z\).


[*] up [*] contents
Next: 3. as example Up: Finite Fields and Zech's Previous: 1. Introduction   Contents
douillet@cnam.fr
2001-02-25