Overview

  • What is the difference between a field generator, a primitive element, and a normal element?
  • How are primitivity and normality tested?
  • When can both conditions hold?
  • What is known about the Morgan–Mullen conjecture?

Three properties of an element

Fix K=FqnK=\mathbb F_{q^n} over Fq\mathbb F_q. An element α\alpha is a field generator if K=Fq(α)K=\mathbb F_q(\alpha); equivalently, its minimal polynomial has degree nn. It is primitive if it generates K×K^\times, and normal if α,αq,…,αqn−1\alpha,\alpha^q,\ldots,\alpha^{q^{n-1}} form an Fq\mathbb F_q-basis of KK.

Primitivity and normality each imply degree nn, but degree nn does not imply either of them. In F3(α)\mathbb F_3(\alpha) with α2=−1\alpha^2=-1, the element α\alpha has degree two and multiplicative order four; the full multiplicative group has order eight.

References 1

Testing primitivity

Put N=qn−1N=q^n-1. An element α∈K×\alpha\in K^\times is primitive if and only if αN/ℓ≠1for every prime ℓ∣N.\alpha^{N/\ell}\ne1\quad\text{for every prime }\ell\mid N. Indeed, a proper divisor of NN divides N/ℓN/\ell for some prime ℓ∣N\ell\mid N. For an irreducible polynomial ff of degree nn, these tests can be performed with α=x mod f\alpha=x\bmod f.

The cyclic group K×K^\times has φ(N)\varphi(N) generators. Frobenius groups them into orbits of length nn, so the number of monic primitive polynomials of degree nn is φ(qn−1)/n\varphi(q^n-1)/n.

References 1

Testing normality

Choose any Fq\mathbb F_q-basis of KK. Form a matrix whose columns are the coordinates of α,αq,…,αqn−1\alpha,\alpha^q,\ldots,\alpha^{q^{n-1}}. The element is normal exactly when this matrix has rank nn.

There is also a polynomial formulation. Let σ(β)=βq\sigma(\beta)=\beta^q, and let mα(T)m_\alpha(T) be the monic polynomial of least degree such that mα(σ)(α)=0m_\alpha(\sigma)(\alpha)=0. It divides Tn−1T^n-1. Normality is equivalent to mα(T)=Tn−1m_\alpha(T)=T^n-1, or to ((Tn−1)/P(T))(σ)(α)≠0\bigl((T^n-1)/P(T)\bigr)(\sigma)(\alpha)\ne0 for every distinct monic irreducible divisor PP of Tn−1T^n-1. Here ∑biTi\sum b_iT^i acts as the linear map β↦∑biβqi\beta\mapsto\sum b_i\beta^{q^i}.

The criterion remains valid when the characteristic divides nn; the multiplicities in Tn−1T^n-1 must be retained.

References 1

Examples

Example

Let α\alpha be a root of x3+x2+1x^3+x^2+1 over F2\mathbb F_2. The polynomial has no root in F2\mathbb F_2, hence is irreducible. The nonzero element α≠1\alpha\ne1 has order seven, so it is primitive. Also α4=α2+α+1\alpha^4=\alpha^2+\alpha+1. Relative to the basis 1,α,α21,\alpha,\alpha^2, its conjugates have columns (0,1,0)t,(0,0,1)t,(1,1,1)t(0,1,0)^t,(0,0,1)^t,(1,1,1)^t, which are independent. Thus α\alpha is normal.

In contrast, a root β\beta of x4+x+1x^4+x+1 over F2\mathbb F_2 is primitive but not normal. Irreducibility is checked in the quartic example; the tests β3≠1\beta^3\ne1 and β5=β2+β≠1\beta^5=\beta^2+\beta\ne1 give order fifteen. Its trace is zero because the coefficient of x3x^3 vanishes. Hence its four conjugates are linearly dependent.

Conversely, let γ\gamma be a root of x4+x3+x2+x+1x^4+x^3+x^2+x+1 over F2\mathbb F_2. Since ord⁡5(2)=4\operatorname{ord}_5(2)=4, this polynomial is irreducible and γ\gamma has multiplicative order five. It is not primitive in F16×\mathbb F_{16}^{\times}, whose order is fifteen. Its trace is one, so it is normal by the trace criterion below. Thus neither primitivity nor normality implies the other.

References 1

The primitive normal basis theorem

Theorem

For every prime power qq and every n≥1n\ge1, the extension Fqn/Fq\mathbb F_{q^n}/\mathbb F_q contains a primitive normal element.

Lenstra and Schoof proved this in 1987. Cohen and Huczynska later gave a proof without computer calculations. Their method counts elements satisfying multiplicative and additive conditions by character sums, then uses a sieve to show that the count is positive.

A stronger theorem asks for both α\alpha and α−1\alpha^{-1} to be normal over Fq\mathbb F_q, with α\alpha primitive. Cohen and Huczynska proved that such an element exists except for (q,n)=(2,3),(2,4),(3,4),(4,3),(5,4).(q,n)=(2,3),(2,4),(3,4),(4,3),(5,4). These five pairs are genuine exceptions.

References 3

Counting normal elements

The normal basis theorem identifies KK, as an Fq[T]\mathbb F_q[T]-module with TT acting by Frobenius, with Fq[T]/(Tn−1)\mathbb F_q[T]/(T^n-1). Its generators correspond to the units of this quotient. Therefore the number of normal elements is qn∏P∣Tn−1(1−q−deg⁡P),q^n\prod_{P\mid T^n-1}\left(1-q^{-\deg P}\right), where the product is over distinct monic irreducible factors.

For q=2,n=4q=2,n=4, one has T4−1=(T+1)4T^4-1=(T+1)^4, giving eight normal elements. In this case normality is equivalent to nonzero trace: (σ−1)3=1+σ+σ2+σ3(\sigma-1)^3=1+\sigma+\sigma^2+\sigma^3, and the normality test has just the factor P=T+1P=T+1.

References 1

Complete normality

An element of Fqn\mathbb F_{q^n} is completely normal over Fq\mathbb F_q if it is normal over every intermediate field Fqd\mathbb F_{q^d}, d∣nd\mid n.

Completely normal basis theorem · Blessenohl–Johnsen

For every prime power qq and integer n≥1n\ge1, the extension Fqn/Fq\mathbb F_{q^n}/\mathbb F_q has a completely normal element.

Primitive elements, normal elements, primitive normal elements, and completely normal elements therefore exist in every finite extension. Morgan and Mullen ask for the additional simultaneous condition: an element that is both primitive and completely normal.

Write n=pamn=p^a m, with p∤mp\nmid m. A published theorem of Garefalakis and Kapetanakis proves this when m<qm<q, in particular when n≤qn\le q. Their 2026 paper develops further constructions using translates of completely normal elements.

Preprint claim. Hellwig's October 2026 preprint claims existence for every q,nq,n, using partial Lagrange resolvents, character sums, and finite computations. The author reports a Lean formalization of the new cases and computations.

Formalization and dependencies

The author’s version 1.0 archive corresponds to commit 09291f9dbd175ffdd8c22cddcbaff7152c516272. The pinned build instructions, dependency manifest, and reported axiom audit identify the computational assumptions.

The two imported mathematical axioms are the Garefalakis–Kapetanakis theorem for n≤qn\le q and a Rosser–Schoenfeld estimate. Compiled computations additionally use Lean.ofReduceBool and Lean.trustCompiler. These records were inspected; the Lean project and its proof dependencies have not been independently audited here.

References 6

Questions

ClaimedFinite fields

Does every finite extension have a primitive completely normal element?

Scope. Every prime power qq and integer n≥1n\ge1; the Morgan–Mullen conjecture.

Published results include the range in which the largest divisor of nn coprime to the characteristic is smaller than qq. Hellwig claims a proof for all extensions in arXiv:2610.36534v1 (October 2026), using character sums, finite computations, and a reported Lean formalization. The general result is listed as a preprint claim.

References 3
Discuss this question →

Related topics

References

  1. Finite Fields

    Lidl, Rudolf and Niederreiter, Harald. Edition 2. Cambridge University Press, Encyclopedia of Mathematics and its Applications, vol. 20, 1997.

    Read the source
  2. The strong primitive normal basis theorem

    Cohen, Stephen D. and Huczynska, Sophie. Acta Arithmetica, 143(4), 299–332, 2010.

    Read the source
  3. Primitive normal bases for finite fields

    Lenstra, Hendrik W., Jr. and Schoof, René J. Mathematics of Computation, 48(177), 217–231, 1987.

    Read the source
  4. The primitive normal basis theorem—without a computer

    Cohen, Stephen D. and Huczynska, Sophie. Journal of the London Mathematical Society, 67(1), 41–56, 2003.

    Read the source
  5. Translates of completely normal elements and the Morgan–Mullen conjecture

    Garefalakis, Theodoulos and Kapetanakis, Giorgos. Canadian Mathematical Bulletin, 1–21, 2026.

    Read the source
  6. Eine Verschärfung des Satzes von der Normalbasis

    Blessenohl, D. and Johnsen, K. Journal of Algebra, 103(1), 141–159, 1986.

    Read the source
  7. Completely normal primitive basis generators of finite fields

    Morgan, Ilene H. and Mullen, Gary L. Utilitas Mathematica, 49, 21–43, 1996.

    Read the source
  8. On the existence of primitive completely normal bases of finite fields

    Garefalakis, Theodoulos and Kapetanakis, Giorgos. Journal of Pure and Applied Algebra, 223(3), 909–921, 2019.

    Read the source
  9. Resolvent Components of Completely Normal Elements and the Morgan-Mullen Conjecture

    Hellwig, Florian. arXiv:2610.36534 [math.NT], 2026.

    Read the source
  10. Resolvent Components of Completely Normal Elements and the Morgan–Mullen Conjecture: Lean 4 Formalisation and Python Programs

    Hellwig, Florian. Author repository; release v1.0, 2026.

    Read the source