This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Inverses of rr-primitive kk-normal elements over finite fields

Mamta Rani mamta11singla@gmail.com Avnish K. Sharma avkush94@gmail.com Sharwan K. Tiwari shrawant@gmail.com Anupama Panigrahi anupama.panigrahi@gmail.com Department of Mathematics, University of Delhi, New Delhi, 110007, India Scientific Analysis Group, Defence Research &\& Development Organization, Metcalfe House, Delhi, 110054, India
Abstract

Let rr, nn be positive integers, kk be a non-negative integer and qq be any prime power such that rqn1.r\mid q^{n}-1. An element α\alpha of the finite field 𝔽qn\mathbb{F}_{q^{n}} is called an rr-primitive element, if its multiplicative order is (qn1)/r(q^{n}-1)/r, and it is called a kk-normal element over 𝔽q\mathbb{F}_{q}, if the greatest common divisor of the polynomials mα(x)=i=1nαqi1xnim_{\alpha}(x)=\sum_{i=1}^{n}\alpha^{q^{i-1}}x^{n-i} and xn1x^{n}-1 is of degree k.k. In this article, we define the characteristic function for the set of kk-normal elements, and with the help of this, we establish a sufficient condition for the existence of an element α\alpha in 𝔽qn\mathbb{F}_{q^{n}}, such that α\alpha and α1\alpha^{-1} both are simultaneously rr-primitive and kk-normal over 𝔽q\mathbb{F}_{q}. Moreover, for n>6kn>6k, we show that there always exists an rr-primitive and kk-normal element α\alpha such that α1\alpha^{-1} is also rr-primitive and kk-normal in all but finitely many fields 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, where qq and nn are such that rqn1r\mid q^{n}-1 and there exists a kk-degree polynomial g(x)xn1g(x)\mid x^{n}-1 over 𝔽q\mathbb{F}_{q}. In particular, we discuss the existence of an element α\alpha in 𝔽qn\mathbb{F}_{q^{n}} such that α\alpha and α1\alpha^{-1} both are simultaneously 11-primitive and 11-normal over 𝔽q\mathbb{F}_{q}.

keywords:
Finite fields, rr-Primitive elements, kk-Normal elements, Additive and multiplicative characters.
MSC:
[20202020] 12E20, 11T23.

1 Introduction

Let 𝔽qn\mathbb{F}_{q^{n}} be the field extension of degree nn over 𝔽q,\mathbb{F}_{q}, where qq be a prime power and nn\in\mathbb{N}. We recall that, the multiplicative group 𝔽qn\mathbb{F}_{q^{n}}^{*} is cyclic, and an element α𝔽qn\alpha\in\mathbb{F}_{q^{n}}^{*} is called primitive, if its multiplicative order is qn1.q^{n}-1. Let rr be a divisor of qn1q^{n}-1, then an element α𝔽qn\alpha\in\mathbb{F}_{q^{n}}^{*} is called rr-primitive, if its multiplicative order is (qn1)/r(q^{n}-1)/r. Clearly, a 1-primitive element in 𝔽qn\mathbb{F}_{q^{n}} is actually a primitive element. Moreover, the existence of rr-primitive elements is obvious, because the multiplicative order of the rthr^{\mathrm{th}} power of a primitive element is (qn1)/r(q^{n}-1)/r. Primitive elements have many applications in the field of cryptography [5, 12, 2] due to their highest multiplicative order. Moreover, rr-primitive elements can also be considered as high ordered elements for small values of rr. Therefore, in many applications, rr-primitive elements may replace primitive elements.

Again, recall that, the finite field 𝔽qn\mathbb{F}_{q^{n}} is also an 𝔽q\mathbb{F}_{q}-vector space, and an element α𝔽qn\alpha\in\mathbb{F}_{q^{n}} is said to be normal over 𝔽q\mathbb{F}_{q}, if the set {α,αq,,αqn1}\{\alpha,\alpha^{q},\ldots,\alpha^{q^{n-1}}\} forms a basis (called normal basis) of 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}. Normal bases have many applications in the computational theory due to their efficient implementation in finite field arithmetic. For more detail on normal elements one may refer [7] and the references therein.

From [11, Theorem 2.39], we know that an element α𝔽qn\alpha\in\mathbb{F}_{q^{n}} is normal over 𝔽q\mathbb{F}_{q} if and only if the polynomials mα(x)=αxn1+αqxn2++αqn2x+αqn1m_{\alpha}(x)=\alpha x^{n-1}+\alpha^{q}x^{n-2}+\ldots+\alpha^{q^{n-2}}x+\alpha^{q^{n-1}} and xn1x^{n}-1 are relatively prime in 𝔽qn[x].\mathbb{F}_{q^{n}}[x]. Motivated by this definition, Huczynska et al. [8] introduced the concept of kk-normal elements and gave the following definition.

Definition 1.1.

An element α𝔽qn\alpha\in\mathbb{F}_{q^{n}} is said to be kk-normal over 𝔽q,\mathbb{F}_{q}, if the polynomials mα(x)=αxn1+αqxn2++αqn2x+αqn1m_{\alpha}(x)=\alpha x^{n-1}+\alpha^{q}x^{n-2}+\ldots+\alpha^{q^{n-2}}x+\alpha^{q^{n-1}} and xn1x^{n}-1 has greatest common divisor of degree kk over 𝔽q\mathbb{F}_{q}.

From the above definition, it is clear that 0kn1.0\leq k\leq n-1. We call an element α𝔽qn\alpha\in\mathbb{F}_{q^{n}}, rr-primitive kk-normal, if it is both rr-primitive and kk-normal over 𝔽q.\mathbb{F}_{q}. From the above definition, if k=0k=0, then α\alpha is normal over 𝔽q\mathbb{F}_{q}. Moreover, Huczynska et al. [8] characterized the kk-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} and established bounds for the number of such elements, and proved the existence of an element α\alpha which is both 11-primitive and 11-normal over 𝔽q\mathbb{F}_{q} for n6n\geq 6 if q11q\geq 11, and for n3n\geq 3 if 3q9,3\leq q\leq 9, where gcd(q,n)=1.\mathrm{gcd}(q,n)=1. In 1987, Lenstra and Schoof [10] proved the existence of primitive normal elements (i.e. 11-primitive 0-normal elements) for arbitrary finite fields. In 2014, Anderson and Mullen [13] conjectured that, for a prime p5p\geq 5 and an integer n3n\geq 3, there always exists an element α𝔽pn\alpha\in\mathbb{F}_{p^{n}} of order (pn1)/r(p^{n}-1)/r, which is kk-normal for (r,k)=(1,1),(2,0),(2,1)(r,k)=(1,1),(2,0),(2,1). In 2018, Reis and Thomson in [17] settled their conjecture for (r,k)=(1,1)(r,k)=(1,1), for arbitrary qq and n3n\geq 3. Finally, in 2019, Kapetanakis and Reis [9] completely settled the above conjecture for arbitrary prime power qq.

Recently, Aguirre and Neumann [1] showed that, there exists a 11-primitive 22-normal element in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} if and only if n5n\geq 5 and gcd(q3q,n)1\mathrm{gcd}(q^{3}-q,n)\neq 1 or n=4n=4 and q1(mod4).q\equiv 1\hskip 1.42262pt(\mathrm{mod}\hskip 1.42262pt4). In this direction, we [14] generalized the above existence problems to arbitrary rr and kk, and established a sufficient condition for the existence of an element α\alpha in 𝔽qn\mathbb{F}_{q^{n}} which is simultaneously rr-primitive and kk-normal over 𝔽q.\mathbb{F}_{q}. Moreover, in particular, we gave the following complete existence of 22-primitive 22-normal elements.

Theorem 1.1.

Let qq be a prime power and nn be a positive integer. Then, there always exists a 22-primitive 22-normal element in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} if and only if qq is odd, and either n5n\geq 5 and gcd(q3q,n)1\mathrm{gcd}(q^{3}-q,n)\neq 1 or n=4n=4 and q1(mod4)q\equiv 1(\mathrm{mod}\hskip 1.42262pt4).

Notice that, the inverse of an rr-primitive element is also rr-primitive in 𝔽qn\mathbb{F}_{q^{n}} for rqn1.r\mid q^{n}-1. But this may not be true in case of kk-normal elements. For example, if k=0k=0, then there does not exist any normal element α𝔽8\alpha\in\mathbb{F}_{8} over 𝔽2\mathbb{F}_{2} for which α1\alpha^{-1} is also normal element in 𝔽23\mathbb{F}_{2^{3}} over 𝔽2\mathbb{F}_{2}. In [20], Tian and Qi showed the existence of a primitive normal element α𝔽qn\alpha\in\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} whose inverse is also primitive normal over 𝔽q\mathbb{F}_{q} for all qq and n32.n\geq 32. Later, Cohen and Huczynska [3] completely settled this problem by using sieving technique and gave the following result.

Theorem 1.2 (Strong Primitive Normal Basis Theorem).

There exists a (1)(1-) primitive (0)(0-) normal element α𝔽qn;\alpha\in\mathbb{F}_{q^{n}}; n2,n\geq 2, such that α1\alpha^{-1} is also a primitive normal element over 𝔽q\mathbb{F}_{q} unless (q,n)(q,n) is one of the pair (2,3),(2,4),(3,4),(4,3),(5,4).(2,3),\ (2,4),\ (3,4),\ (4,3),\ (5,4).

Motivated by the above theorem, in this article, we discuss the existence of a pair (α,α1)(\alpha,\alpha^{-1}) of rr-primitive kk-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}. For this, we first define the characteristic function in Section 3 for the set of kk-normal elements, which plays a crucial role in establishing a sufficient condition for the existence of such pairs in Section 4. Moreover, we conclude that, for n>6kn>6k, there exists a pair (α,α1)(\alpha,\alpha^{-1}) of rr-primitive kk-normal elements in all but finitely many fields 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q.\mathbb{F}_{q}. In particular, we gave the following result in Section 5.

Theorem 1.3.

Let qq be a prime power and n5n\geq 5 be an integer. Then

  1. (i)(i)

    For n7n\geq 7 and q2q\geq 2, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}.

  2. (ii)(ii)

    For n=5n=5, 66 and q2q\geq 2 such that gcd(q,n)=1\mathrm{gcd}(q,n)=1, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} with the sole genuine exception (4,5).(4,5).

2 Preliminaries

In this section, we recall some definitions and lemmas that are crucial for proving the results of this article.

2.1 Freeness

We begin with the following definition.

Definition 2.1 (ee-free element).

Let eqn1e\mid q^{n}-1 and β𝔽qn.\beta\in\mathbb{F}_{q^{n}}. Then β\beta is said to be ee-free, if β\beta is not a dthd^{th} power in 𝔽qn\mathbb{F}_{q^{n}} for any divisor d>1d>1 of ee. In other words, β\beta is ee-free if and only if ee and qn1ord(β)\frac{q^{n}-1}{\mathrm{ord}(\beta)} are co-prime, where ord(β)\mathrm{ord}(\beta) is the multiplicative order of β\beta in 𝔽qn.\mathbb{F}_{q^{n}}^{*}. Clearly, an element β𝔽qn\beta\in\mathbb{F}_{q^{n}} is primitive if and only if it is (qn1)(q^{n}-1)-free.

Let h(x)=i=0maixi𝔽q[x]h(x)=\sum_{i=0}^{m}a_{i}x^{i}\in\mathbb{F}_{q}[x] with am0a_{m}\neq 0. Then, the additive group 𝔽qn\mathbb{F}_{q^{n}} forms an 𝔽q[x]\mathbb{F}_{q}[x]- module under the action hβ=i=0maiβqih\circ\beta=\sum_{i=0}^{m}a_{i}\beta^{q^{i}}. Notice that, (xn1)β=0(x^{n}-1)\circ\beta=0 for all β𝔽qn,\beta\in\mathbb{F}_{q^{n}}, which leads to the following definition.

Definition 2.2.

The 𝔽q\mathbb{F}_{q}-order ((denoted by Ordq())\mathrm{Ord}_{q}(\cdot)) of an element β𝔽qn,\beta\in\mathbb{F}_{q^{n}}, is the least degree monic divisor h(x)h(x) of xn1x^{n}-1 such that hβ=0h\circ\beta=0.

Similar to the ee-free elements, we have the following definition for ff-free elements.

Definition 2.3.

Let h(x)xn1h(x)\mid x^{n}-1 and β𝔽qn.\beta\in\mathbb{F}_{q^{n}}. Then, β\beta is called hh-free, if βgγ\beta\neq g\circ\gamma for any γ𝔽qn\gamma\in\mathbb{F}_{q^{n}} and for any divisor g(x)g(x) of h(x)h(x) of positive degree. In other words, β\beta is hh-free if and only if h(x)h(x) and xn1Ordq(β)\frac{x^{n}-1}{\mathrm{Ord}_{q}(\beta)} are co-prime. Clearly, an element β𝔽qn\beta\in\mathbb{F}_{q^{n}} is normal if and only if it is (xn1)(x^{n}-1)-free.

In this article, we will use the following definition of kk-normal elements.

Definition 2.4.

[8, Theorem 3.2] An element β𝔽qn\beta\in\mathbb{F}_{q^{n}} is kk-normal over 𝔽q\mathbb{F}_{q}, if and only if the 𝔽q\mathbb{F}_{q}-order of β\beta is of degree nkn-k.

The following lemma provides a way to construct the kk-normal elements from a given normal element.

Lemma 2.1.

[16, Lemma 3.1] Let β𝔽qn\beta\in\mathbb{F}_{q^{n}} be a normal element over 𝔽q\mathbb{F}_{q} and g𝔽q[x]g\in\mathbb{F}_{q}[x] be a polynomial of degree kk such that gg divides xn1x^{n}-1. Then α=gβ\alpha=g\circ\beta is kk-normal.

2.2 Characters

Let 𝔊\mathfrak{G} be a finite abelian group. A character χ\chi of 𝔊\mathfrak{G} is a homomorphism from 𝔊\mathfrak{G} into the multiplicative group of complex numbers of unit modulus. The set of all such characters of 𝔊\mathfrak{G}, denoted by 𝔊^\widehat{\mathfrak{G}}, forms a multiplicative group and 𝔊𝔊^\mathfrak{G}\cong\widehat{\mathfrak{G}}. A character χ\chi is called the trivial character if χ(α)=1\chi(\alpha)=1 for all α𝔊\alpha\in\mathfrak{G}, otherwise it is a non-trivial character.

Lemma 2.2.

[11, Theorem 5.4] Let χ\chi be any non-trivial character of a finite abelian group 𝔊\mathfrak{G} and α𝔊\alpha\in\mathfrak{G} be any non-trivial element, then

α𝔊χ(α)=0andχ𝔊^χ(α)=0.\sum_{\alpha\in\mathfrak{G}}\chi(\alpha)=0\ \text{and}\ \sum_{\chi\in\widehat{\mathfrak{G}}}\chi(\alpha)=0.

Let ψ\psi denote the additive character for the additive group 𝔽qn\mathbb{F}_{q^{n}}, and χ\chi denote the multiplicative character for the multiplicative group 𝔽qn.\mathbb{F}_{q^{n}}^{*}. The additive character ψ0\psi_{0} defined by ψ0(β)=e2πiTr(β)/p,for allβ𝔽qn,\psi_{0}(\beta)=e^{2\pi i\mathrm{Tr}(\beta)/p},\ \text{for all}\ \beta\in\mathbb{F}_{q^{n}}, where pp is the characteristic of 𝔽qn\mathbb{F}_{q^{n}} and Tr\mathrm{Tr} is the absolute trace function from 𝔽qn\mathbb{F}_{q^{n}} to 𝔽p\mathbb{F}_{p}, is called the canonical additive character of 𝔽qn\mathbb{F}_{q^{n}}. Moreover, every additive character ψβ\psi_{\beta} for β𝔽qn\beta\in\mathbb{F}_{q^{n}} can be expressed in terms of the canonical additive character ψ0\psi_{0} as ψβ(γ)=ψ0(βγ),for allγ𝔽qn.\psi_{\beta}(\gamma)=\psi_{0}(\beta\gamma),\ \text{for all}\ \gamma\in\mathbb{F}_{q^{n}}.

For any ψ𝔽^qn\psi\in\widehat{\mathbb{F}}_{q^{n}}, α𝔽qn\alpha\in\mathbb{F}_{q^{n}} and g(x)𝔽q[x]g(x)\in\mathbb{F}_{q}[x], 𝔽^qn\widehat{\mathbb{F}}_{q^{n}} is an 𝔽q[x]\mathbb{F}_{q}[x]-module under the action ψg(α)=ψ(gα).\psi\circ g(\alpha)=\psi(g\circ\alpha). The 𝔽q\mathbb{F}_{q}-order of an additive character ψ𝔽^qn\psi\in\widehat{\mathbb{F}}_{q^{n}}, denoted by Ordq(ψ)\mathrm{Ord}_{q}(\psi), is the least degree monic divisor g(x)g(x) of xn1x^{n}-1 such that ψg\psi\circ g is the trivial character and there are precisely Φq(g)\Phi_{q}(g) characters of 𝔽q\mathbb{F}_{q}-order g(x)g(x), where Φq(g)=|(𝔽q[x]/<g>)|\Phi_{q}(g)=|(\mathbb{F}_{q}[x]/<g>)^{*}|. Moreover, hgΦq(h)=qdeg(g).\sum_{h\mid g}\Phi_{q}(h)=q^{\mathrm{deg}(g)}.

We shall need the following lemma to prove our sufficient condition.

Lemma 2.3.

[6, Theorem 5.5] Let F(x)=j=1kfj(x)njF(x)=\prod_{j=1}^{k}f_{j}(x)^{n_{j}} be a rational function in 𝔽qn(x)\mathbb{F}_{q^{n}}(x), where fj(x)𝔽qn[x]f_{j}(x)\in\mathbb{F}_{q^{n}}[x] are irreducible polynomials and njn_{j} are non-zero integers. Let χ\chi be a multiplicative character of order dd of 𝔽qn.\mathbb{F}_{q^{n}}^{*}. Suppose that the rational function F(x)F(x) is not of the form L(x)dL(x)^{d} for any L(x)𝔽(x),L(x)\in\mathbb{F}(x), where 𝔽\mathbb{F} is the algebraic closure of 𝔽qn.\mathbb{F}_{q^{n}}. Then we have

|α𝔽qnF(α)0,χ(F(α))|(j=1kdeg(fj)1)qn/2.\Big{|}\sum_{\begin{subarray}{c}\alpha\in\mathbb{F}_{q^{n}}\\ F(\alpha)\neq 0,\infty\end{subarray}}\chi(F(\alpha))\Big{|}\leq\Big{(}\sum_{j=1}^{k}\mathrm{deg}(f_{j})-1\Big{)}q^{n/2}.
Lemma 2.4.

[6, Theorem 5.6] Let F(x),G(x)𝔽qn(x)F(x),\ G(x)\in\mathbb{F}_{q^{n}}(x) be rational functions. Write F(x)=j=1kfj(x)njF(x)=\prod_{j=1}^{k}f_{j}(x)^{n_{j}}, where fj(x)𝔽qn[x]f_{j}(x)\in\mathbb{F}_{q^{n}}[x] are irreducible polynomials and njn_{j} are non-zero integers. Let d1=j=1kdeg(fj(x))d_{1}=\sum_{j=1}^{k}\mathrm{deg}(f_{j}(x)), d2=max{deg(G(x)),0}d_{2}=max\{\mathrm{deg}(G(x)),0\}, d3d_{3} be the degree of the denominator of G(x)G(x) and d4d_{4} be the sum of degrees of those irreducible polynomials dividing the denominator of G(x)G(x), but distinct from fj; 1jk.f_{j}\ ;\ 1\leq j\leq k. Let χ\chi be the multiplicative character of 𝔽qn\mathbb{F}_{q^{n}}^{*} and ψ\psi be the non-trivial additive character of 𝔽qn.\mathbb{F}_{q^{n}}. Suppose that G(x)L(x)qnL(x)G(x)\neq L(x)^{q^{n}}-L(x) in 𝔽(x),\mathbb{F}(x), where 𝔽\mathbb{F} is the algebraic closure of 𝔽qn.\mathbb{F}_{q^{n}}. Then we have

|α𝔽qn,F(α)0,G(α)χ(F(α))ψ(G(α))|(d1+d2+d3+d41)qn/2.\Big{|}\sum_{\begin{subarray}{c}\alpha\in\mathbb{F}_{q^{n}},F(\alpha)\neq 0,\infty\\ G(\alpha)\neq\infty\end{subarray}}\chi(F(\alpha))\psi(G(\alpha))\Big{|}\leq(d_{1}+d_{2}+d_{3}+d_{4}-1)q^{n/2}.

2.3 Some characteristic functions

For the set of ee-free elements of 𝔽qn\mathbb{F}_{q^{n}}^{*}, we have the following characteristic function ρe:𝔽qn{0,1}.\rho_{e}:\mathbb{F}_{q^{n}}^{*}\to\{0,1\}.

ρe(β)=ϕ(e)ed|eμ(d)ϕ(d)χdχd(β),\rho_{e}(\beta)=\frac{\phi(e)}{e}\sum_{d|e}\frac{\mu(d)}{\phi(d)}\sum_{\chi_{d}}\chi_{d}(\beta), (1)

where μ\mu is the Möbius function and the internal sum runs over all the multiplicative characters χd\chi_{d} of order dd. Similar to the characteristic function for the set of ee-free elements, we have the following characteristic function Υg:𝔽qn{0,1}\Upsilon_{g}:\mathbb{F}_{q^{n}}\to\{0,1\} for the set of gg-free elements of 𝔽qn\mathbb{F}_{q^{n}}.

Υg(α)=Φq(g)qdeg(g)h|gμ(h)Φq(h)ψhψh(α),\Upsilon_{g}(\alpha)=\dfrac{\Phi_{q}(g)}{q^{\mathrm{deg}(g)}}\sum_{h|g}\dfrac{\mu^{\prime}(h)}{\Phi_{q}(h)}\sum_{\psi_{h}}\psi_{h}(\alpha), (2)

where μ\mu^{\prime} is the analog of the Möbius function, which is defined as μ(h)=(1)t\mu^{\prime}(h)=(-1)^{t}, if h(x)h(x) is a product of tt distinct monic irreducible polynomials, otherwise 0, and the internal sum runs over the additive characters ψh\psi_{h} of the 𝔽q\mathbb{F}_{q}-order h(x)h(x).

3 Characterization of rr-primitive and kk-normal elements

3.1 Characteristic function for the set of rr-primitive elements

Let rqn1r\mid q^{n}-1 be a positive integer, and write r=up1b1p2b2psbsr=up_{1}^{b_{1}}p_{2}^{b_{2}}\cdots p_{s}^{b_{s}}, where uu and (qn1)/u(q^{n}-1)/u are co-prime and pjp_{j}’s are distinct primes such that bj1b_{j}\geq 1 and pjbj+1qn1p_{j}^{b_{j}+1}\mid q^{n}-1, for all j=1,2,,s.j=1,2,\ldots,s. In this article, we denote the product of distinct irreducible factors of an integer or a polynomial mm by rad(m)\mathrm{rad}(m). Now, let R=rad(qn1)/rad(r)R=\mathrm{rad}(q^{n}-1)/\mathrm{rad}(r), and set δj:=pjbj\delta_{j}:=p_{j}^{b_{j}} and λj:=pjbj+1\lambda_{j}:=p_{j}^{b_{j}+1} for all j=1,2,,sj=1,2,\ldots,s. In [4], Cohen and Kapetanakis showed that an element α𝔽qn\alpha\in\mathbb{F}_{q^{n}} is rr-primitive if and only if α\alpha is RR-free, α=βr\alpha=\beta^{r} for some β𝔽qn\beta\in\mathbb{F}_{q^{n}}, but αβλj\alpha\neq\beta^{\lambda_{j}} for any β𝔽qn\beta\in\mathbb{F}_{q^{n}} and any j=1,2,,sj=1,2,\ldots,s, and using this fact, they defined the characteristic function for rr-primitive elements. In [14], following them, we defined the following characteristic function Γrd\Gamma_{r}^{d} for the set QrdQ_{r}^{d} of elements α𝔽qn\alpha\in\mathbb{F}_{q^{n}} such that α\alpha is dd-free for some divisor dd of RR, α=βr\alpha=\beta^{r} for some β𝔽qn\beta\in\mathbb{F}_{q^{n}}, but αβλj\alpha\neq\beta^{\lambda_{j}} for any β𝔽qn\beta\in\mathbb{F}_{q^{n}} and any j=1,2,,s.j=1,2,\ldots,s.

Γrd(α)=ϕ(d)rdd1d,d2uejλj,1jsμ(d1)ϕ(d1)(j=1spj,ej)χd1,χd2,χe1,,χes(χd1χd2χe1χes)(α),\Gamma_{r}^{d}(\alpha)=\frac{\phi(d)}{rd}\sum_{\begin{subarray}{c}d_{1}\mid d,d_{2}\mid u\end{subarray}}\sum_{\begin{subarray}{c}e_{j}\mid\lambda_{j},\\ 1\leq j\leq s\end{subarray}}\frac{\mu(d_{1})}{\phi(d_{1})}\big{(}\prod_{j=1}^{s}\ell_{p_{j},e_{j}}\big{)}\sum_{\begin{subarray}{c}\chi_{d_{1}},\chi_{d_{2}},\\ \chi_{e_{1}},\ldots,\chi_{e_{s}}\end{subarray}}(\chi_{d_{1}}\chi_{d_{2}}\chi_{e_{1}}\cdots\chi_{e_{s}})(\alpha), (3)

where pj,ej={11/pjifejλj,1/pjifej=λj.\ell_{p_{j},e_{j}}=\left\{\begin{array}[]{ll}1-1/p_{j}&\mathrm{if}\ e_{j}\neq\lambda_{j},\\ -1/p_{j}&\mathrm{if}\ e_{j}=\lambda_{j}\end{array}\right.. In particular, for d=Rd=R, the above characteristic function provides the characteristic function given by Cohen and Kapetanakis [4] for the set QrRQ_{r}^{R} of rr-primitive elements.

3.2 Characteristic function for the set of kk-normal elements

For the existence of those elements in a finite field 𝔽qn\mathbb{F}_{q^{n}} that simultaneously satisfy certain properties, such as primitivity, normality, trace, norm, etc., a general approach is to use the characteristic functions for the set of elements with those properties. In this article, we also use this approach to discuss the existence of a pair (α,α1)(\alpha,\alpha^{-1}) of rr-primitive kk-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}. For this, we shall need the characteristic function for the set of kk-normal elements that we define in this subsection.

Let SkS_{k} be the collection of all kk-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} and PkP_{k} be the collection of polynomials of degree kk that divides xn1x^{n}-1. By Lemma 2.1, if β\beta is a normal element of 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, then for any polynomial gPkg\in P_{k}, gβg\circ\beta is kk-normal over 𝔽q\mathbb{F}_{q}. So, for every gPkg\in P_{k}, we consider a set Sg,k:={αSk|α=gβ for some elementβ𝔽qn}S_{g,k}:=\{\alpha\in S_{k}\ |\ \alpha=g\circ\beta\text{ for some element}\ \beta\in\mathbb{F}_{q^{n}}\} of kk-normal elements. We have the following lemma.

Lemma 3.1.

If g,gg,\ g^{\prime} are two distinct polynomials in PkP_{k}, then Sg,kSg,k=S_{g,k}\bigcap S_{g^{\prime},k}=\emptyset, and gPkSg,k=Sk\bigcup_{g\in P_{k}}S_{g,k}=S_{k}.

Proof.

Suppose that, Sg,kSg,kS_{g,k}\bigcap S_{g^{\prime},k}\neq\emptyset, then there exist elements β\beta and β\beta^{\prime} in 𝔽qn\mathbb{F}_{q^{n}} such that gβ=gβg\circ\beta=g^{\prime}\circ\beta^{\prime}, and by uniqueness of the 𝔽q\mathbb{F}_{q}-order of an element, we get g=gg=g^{\prime}, a contradiction. Hence, Sg,kSg,k=S_{g,k}\bigcap S_{g^{\prime},k}=\emptyset.

Clearly, gPkSg,kSk.\bigcup_{g\in P_{k}}S_{g,k}\subseteq S_{k}. Now, let αSk\alpha\in S_{k}, then α=hβ\alpha=h\circ\beta, for some normal element β\beta and a polynomial h𝔽q[x]h\in\mathbb{F}_{q}[x] of degree at most n1n-1. Let ff be the 𝔽q\mathbb{F}_{q}-order of α\alpha, then fα=fhβ=0f\circ\alpha=fh\circ\beta=0, and hence, xn1fhx^{n}-1\mid fh, i.e. h=(xn1)h/fh=(x^{n}-1)h^{\prime}/f, for some h𝔽q[x].h^{\prime}\in\mathbb{F}_{q}[x]. Thus, α=xn1f(hβ)\alpha=\frac{x^{n}-1}{f}\circ(h^{\prime}\circ\beta), which implies α=gγSg,k\alpha=g\circ\gamma\in S_{g,k}, where g=xn1fPkg=\frac{x^{n}-1}{f}\in P_{k} and γ=hβ𝔽qn\gamma=h^{\prime}\circ\beta\in\mathbb{F}_{q^{n}}. Hence, gPkSg,k=Sk\bigcup_{g\in P_{k}}S_{g,k}=S_{k}. ∎

From the above lemma, it is clear that, the set {Sg,k|gPk}\{S_{g,k}\ |\ g\in P_{k}\} provides a partition of SkS_{k}. Therefore, the characteristic function for the set SkS_{k} of kk-normal elements is equal to the sum of the characteristic functions for the sets Sg,k;gPk.S_{g,k}\hskip 1.42262pt;\ g\in P_{k}. Hence, it is enough to define the characteristic function for the set Sg,kS_{g,k}.

Let gPkg\in P_{k} be an arbitrary but fixed polynomial. Write g=πf1b1f2b2ftbtg=\pi f_{1}^{b_{1}}f_{2}^{b_{2}}\cdots f_{t}^{b_{t}}, where π\pi and (xn1)/π(x^{n}-1)/\pi are co-prime, and fif_{i}’s are distinct irreducible polynomials such that bi1b_{i}\geq 1 and fibi+1xn1f_{i}^{b_{i}+1}\mid x^{n}-1, for all i=1,2,,t.i=1,2,\ldots,t. Further, let G=rad(xn1)/rad(g)G=\mathrm{rad}(x^{n}-1)/\mathrm{rad}(g), and set Δi:=fibi\Delta_{i}:=f_{i}^{b_{i}} and Λi:=fibi+1\Lambda_{i}:=f_{i}^{b_{i}+1} for all i=1,2,,t.i=1,2,\ldots,t. Now, we prove the following lemma.

Lemma 3.2.

Let kk be a non-negative integer and gPkg\in P_{k}. Then, the following are equivalent.

  1. (i)(i)

    αSg,k\alpha\in S_{g,k}.

  2. (ii)(ii)

    α\alpha is GG-free, α=gβ\alpha=g\circ\beta for some β𝔽qn\beta\in\mathbb{F}_{q^{n}}, but αΛiβ\alpha\neq\Lambda_{i}\circ\beta for any β𝔽qn\beta\in\mathbb{F}_{q^{n}} and for any i=1,2,,t.i=1,2,\ldots,t.

Proof.

First, let αSg,k\alpha\in S_{g,k}. Then by definition of Sg,kS_{g,k}, α=gβ\alpha=g\circ\beta for some β𝔽qn\beta\in\mathbb{F}_{q^{n}}, and 𝔽q\mathbb{F}_{q}-order of α\alpha is (xn1)/g(x^{n}-1)/g. Notice that, GG and gg are co-prime, therefore, gcd(G,xn1Ordq(α))=1\mathrm{gcd}\big{(}G,\frac{x^{n}-1}{\mathrm{Ord}_{q}(\alpha)}\big{)}=1, which implies α\alpha is GG-free. Now, suppose that α=Λiβ\alpha=\Lambda_{i}\circ\beta for some β𝔽qn\beta\in\mathbb{F}_{q^{n}} and some i=1,2,,ti=1,2,\ldots,t. Then, xn1Λiα=0\frac{x^{n}-1}{\Lambda_{i}}\circ\alpha=0, which implies xn1gxn1Λi\frac{x^{n}-1}{g}\mid\frac{x^{n}-1}{\Lambda_{i}}, i.e. Λig,\Lambda_{i}\mid g, which is not possible. Hence, (i)(i) implies (ii)(ii).

Now, let α\alpha is GG-free, α=gβ\alpha=g\circ\beta for some β𝔽qn\beta\in\mathbb{F}_{q^{n}}, but αΛiβ\alpha\neq\Lambda_{i}\circ\beta for any β𝔽qn\beta\in\mathbb{F}_{q^{n}} and for any i=1,2,,t.i=1,2,\ldots,t. Then, α=g(hγ)\alpha=g\circ(h\circ\gamma) for some normal element γ𝔽qn\gamma\in\mathbb{F}_{q^{n}} and a polynomial h𝔽q[x]h\in\mathbb{F}_{q}[x] of degree at most n1n-1, and Ordq(α)=xn1ggcd(h,(xn1)/g)\mathrm{Ord}_{q}(\alpha)=\frac{x^{n}-1}{g\hskip 1.42262pt\mathrm{gcd}(h,(x^{n}-1)/g)}. Suppose that, hh^{\prime} be an irreducible factor of the gcd(h,(xn1)/g)\mathrm{gcd}(h,(x^{n}-1)/g), then hhh^{\prime}\mid h and hxn1gh^{\prime}\mid\frac{x^{n}-1}{g}. This implies, either hGh^{\prime}\mid G or Λigh.\Lambda_{i}\mid gh. In any case, we get a contradiction. This means Ordq(α)=xn1g\mathrm{Ord}_{q}(\alpha)=\frac{x^{n}-1}{g}, i.e. α\alpha is a kk-normal element. Hence, (ii)(ii) implies (i)(i). ∎

Now, we define the characteristic function Ψg\Psi_{g} for the set of elements of the form gβ.g\circ\beta. Consider the set MgM_{g} of the elements of the form gβg\circ\beta, where gxn1.g\mid x^{n}-1. Clearly, MgM_{g} is a subgroup of the additive group 𝔽qn\mathbb{F}_{q^{n}}. Let A𝔽^qnA\subseteq\widehat{\mathbb{F}}_{q^{n}} be the annihilator of MgM_{g}, i.e. the collection of all additive characters ψ𝔽^qn\psi\in\widehat{\mathbb{F}}_{q^{n}} such that ψ(α)=1\psi(\alpha)=1 for all αMg.\alpha\in M_{g}. Clearly, the set AA consists of all the additive characters of 𝔽q\mathbb{F}_{q}-order hh dividing gg, and from [11, Theorem 5.6], it is isomorphic to the group of the characters of the quotient group 𝔽qn/Mg\mathbb{F}_{q^{n}}/M_{g}, i.e. A(𝔽qn/Mg)^A\cong\widehat{(\mathbb{F}_{q^{n}}/M_{g})}. Therefore, we can define a character 𝒴h\mathcal{Y}_{h} on (𝔽qn/Mg)^\widehat{(\mathbb{F}_{q^{n}}/M_{g})} by 𝒴h(α+Mg)=ψh(α),\mathcal{Y}_{h}(\alpha+M_{g})=\psi_{h}(\alpha), where ψhA.\psi_{h}\in A. Then,

hg𝒴h𝒴h(α+Mg)={qdeg(g),ifαMg,0,otherwise.\sum_{h\mid g}\sum_{\mathcal{Y}_{h}}\mathcal{Y}_{h}(\alpha+M_{g})=\left\{\begin{array}[]{lll}q^{\mathrm{deg}(g)}&,&\mathrm{if}\ \alpha\in M_{g},\\ 0&,&\mathrm{otherwise}\end{array}\right..

Thus, we can define a characteristic function Ψg:𝔽qn{0,1}\Psi_{g}:\mathbb{F}_{q^{n}}\to\{0,1\} for the set MgM_{g} as follows.

Ψg(α):=1qdeg(g)hgψhψh(α).\Psi_{g}(\alpha):=\frac{1}{q^{\mathrm{deg}(g)}}\sum_{h\mid g}\sum_{\psi_{h}}\psi_{h}(\alpha). (4)

Now, we are ready to define a characteristic function 𝒬gG\mathcal{Q}_{g}^{G} for the set Sg,k.S_{g,k}. From the characteristic functions (2) and (4) and from Lemma 3.2, we have

𝒬gG(α)=ΥG(α)Ψg(α)i=1t(1ΨΛi(α)).\displaystyle\mathcal{Q}_{g}^{G}(\alpha)=\Upsilon_{G}(\alpha)\Psi_{g}(\alpha)\prod_{i=1}^{t}\big{(}1-\Psi_{\Lambda_{i}}(\alpha)\big{)}.

Since, g=πi=1tΔig=\pi\prod_{i=1}^{t}\Delta_{i}, and π\pi, Δi;i=1,2,,t\Delta_{i}\ ;\ i=1,2,\ldots,t are mutually co-prime, Ψg(α)=Ψπ(α)i=1tΨΔi(α)\Psi_{g}(\alpha)=\Psi_{\pi}(\alpha)\prod_{i=1}^{t}\Psi_{\Delta_{i}}(\alpha), and hence,

𝒬gG(α)=ΥG(α)Ψπ(α)i=1tΨΔi(α)(1ΨΛi(α)).\displaystyle\mathcal{Q}_{g}^{G}(\alpha)=\Upsilon_{G}(\alpha)\Psi_{\pi}(\alpha)\prod_{i=1}^{t}\Psi_{\Delta_{i}}(\alpha)\big{(}1-\Psi_{\Lambda_{i}}(\alpha)\big{)}.

Moreover, if α=Λiβ\alpha=\Lambda_{i}\circ\beta for some β\beta, then α=Δiγ\alpha=\Delta_{i}\circ\gamma for some γ\gamma in 𝔽qn\mathbb{F}_{q^{n}}, and ΨΔiΨΛi=ΨΛi\Psi_{\Delta_{i}}\Psi_{\Lambda_{i}}=\Psi_{\Lambda_{i}} for all i=1,2,,t.i=1,2,\ldots,t. Hence,

𝒬gG(α)=ΥG(α)Ψπ(α)i=1t(ΨΔi(α)ΨΛi(α)).\mathcal{Q}_{g}^{G}(\alpha)=\Upsilon_{G}(\alpha)\Psi_{\pi}(\alpha)\prod_{i=1}^{t}\big{(}\Psi_{\Delta_{i}}(\alpha)-\Psi_{\Lambda_{i}}(\alpha)\big{)}. (5)

Now,

ΨΔi(α)ΨΛi(α)=\displaystyle\Psi_{\Delta_{i}}(\alpha)-\Psi_{\Lambda_{i}}(\alpha)= 1qdeg(Δi)hΔiψhψh(α)1qdeg(Λi)hΛiψhψh(α)\displaystyle\frac{1}{q^{\mathrm{deg}(\Delta_{i})}}\sum_{h\mid\Delta_{i}}\sum_{\psi_{h}}\psi_{h}(\alpha)-\frac{1}{q^{\mathrm{deg}(\Lambda_{i})}}\sum_{h\mid\Lambda_{i}}\sum_{\psi_{h}}\psi_{h}(\alpha)
=\displaystyle= 1qdeg(Δi)hΛiψhfi,hψh(α),\displaystyle\frac{1}{q^{\mathrm{deg}(\Delta_{i})}}\sum_{h\mid\Lambda_{i}}\sum_{\psi_{h}}\ell_{f_{i},h}^{\prime}\psi_{h}(\alpha),

where fi,h={11/qdeg(fi)ifhΛi,1/qdeg(fi)ifh=Λi.\ell_{f_{i},h}^{\prime}=\left\{\begin{array}[]{ll}1-1/q^{\mathrm{deg}(f_{i})}&\mathrm{if}\ h\neq\Lambda_{i},\\ -1/q^{\mathrm{deg}(f_{i})}&\mathrm{if}\ h=\Lambda_{i}\end{array}\right.. Notice that, |fi,h|fi,1|\ell_{f_{i},h}^{\prime}|\leq\ell_{f_{i},1}^{\prime} for all i=1,2,,ti=1,2,\ldots,t. Now, substituting the values of ΥG(α)\Upsilon_{G}(\alpha), Ψπ(α)\Psi_{\pi}(\alpha) and ΨΔi(α)ΨΛi(α)\Psi_{\Delta_{i}}(\alpha)-\Psi_{\Lambda_{i}}(\alpha) in Equation (5), we get

𝒬gG(α)=\displaystyle\mathcal{Q}_{g}^{G}(\alpha)= Φq(G)qdeg(G)+deg(g)g1G,g2πhiΛi,1itμ(g1)Φq(g1)(i=1tfi,hi)ψg1,ψg2,ψhi; 1it(ψg1ψg2ψh1ψht)(α).\displaystyle\frac{\Phi_{q}(G)}{q^{\mathrm{deg}(G)+\mathrm{deg}(g)}}\sum_{\begin{subarray}{c}g_{1}\mid G,\\ g_{2}\mid\pi\end{subarray}}\sum_{\begin{subarray}{c}h_{i}\mid\Lambda_{i},\\ 1\leq i\leq t\end{subarray}}\frac{\mu^{\prime}(g_{1})}{\Phi_{q}(g_{1})}\big{(}\prod_{i=1}^{t}\ell_{f_{i},h_{i}}^{\prime}\big{)}\sum_{\begin{subarray}{c}\psi_{g_{1}},\psi_{g_{2}},\\ \psi_{h_{i}};\ 1\leq i\leq t\end{subarray}}(\psi_{g_{1}}\psi_{g_{2}}\psi_{h_{1}}\cdots\psi_{h_{t}})(\alpha). (6)

Thus, we define the characteristic function 𝒬\mathcal{Q} for the set SkS_{k} of all kk-normal elements as 𝒬=gPk𝒬gG\mathcal{Q}=\sum_{g\in P_{k}}\mathcal{Q}_{g}^{G}.

Now, for a divisor HH of GG, we define a collection Tg,kHT_{g,k}^{H} of elements α\alpha of 𝔽qn\mathbb{F}_{q^{n}} such that α\alpha is HH-free, α=gβ\alpha=g\circ\beta for some β𝔽qn\beta\in\mathbb{F}_{q^{n}}, but αΛiβ\alpha\neq\Lambda_{i}\circ\beta for any β𝔽qn\beta\in\mathbb{F}_{q^{n}} and for any i=1,2,,t.i=1,2,\ldots,t. We define the characteristic function 𝒬gH\mathcal{Q}_{g}^{H} for the set Tg,kHT_{g,k}^{H} similar to that for the set Sg,kS_{g,k} as follows

𝒬gH(α)=\displaystyle\mathcal{Q}_{g}^{H}(\alpha)= Φq(H)qdeg(H)+deg(g)g1H,g2πhiΛi,1itμ(g1)Φq(g1)(i=1tfi,hi)ψg1,ψg2,ψhi; 1it(ψg1ψg2ψh1ψht)(α).\displaystyle\frac{\Phi_{q}(H)}{q^{\mathrm{deg}(H)+\mathrm{deg}(g)}}\sum_{\begin{subarray}{c}g_{1}\mid H,\\ g_{2}\mid\pi\end{subarray}}\sum_{\begin{subarray}{c}h_{i}\mid\Lambda_{i},\\ 1\leq i\leq t\end{subarray}}\frac{\mu^{\prime}(g_{1})}{\Phi_{q}(g_{1})}\big{(}\prod_{i=1}^{t}\ell_{f_{i},h_{i}}^{\prime}\big{)}\sum_{\begin{subarray}{c}\psi_{g_{1}},\psi_{g_{2}},\\ \psi_{h_{i}};\ 1\leq i\leq t\end{subarray}}(\psi_{g_{1}}\psi_{g_{2}}\psi_{h_{1}}\cdots\psi_{h_{t}})(\alpha). (7)

Observe that, in particular, if H=GH=G, then Tg,kH=Sg,kT_{g,k}^{H}=S_{g,k} and 𝒬gH=𝒬gG.\mathcal{Q}_{g}^{H}=\mathcal{Q}_{g}^{G}.

Henceforth, the notation used in this section shall have the same meaning throughout the article unless otherwise stated.

4 Existence of an rr-primitive kk-normal pair (α,α1)(\alpha,\alpha^{-1})

In this section, our aim is to show the existence of a pair (α,α1)(\alpha,\alpha^{-1}) of rr-primitive kk-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q.\mathbb{F}_{q}. From Lemma 2.1, it is clear that, if γ𝔽qn\gamma\in\mathbb{F}_{q^{n}} is a normal element and gPkg\in P_{k}, then α=gγ\alpha=g\circ\gamma is a kk-normal element over 𝔽q.\mathbb{F}_{q}. Therefore, for the existence of a desired pair, we only need to check the rr-primitivity of α\alpha and the kk-normality of α1\alpha^{-1}. This means for the existence of a desired pair, it is enough to find an (xn1)(x^{n}-1)-free element γ\gamma, such that α=gγQrR\alpha=g\circ\gamma\in Q_{r}^{R} and α1Tg,kG\alpha^{-1}\in T_{g,k}^{G}, for some arbitrary but fixed polynomial gPk.g\in P_{k}. Instead of this, more generally, we will show the existence of an hh-free element γ\gamma, for some hxn1h\mid x^{n}-1, such that α=gγQrd\alpha=g\circ\gamma\in Q_{r}^{d} and α1Tg,kH\alpha^{-1}\in T_{g,k}^{H}, where dRd\mid R and HGH\mid G. Let Nr,k,g(h,d,H)N_{r,k,g}(h,d,H) denote the number of such elements γ\gamma, then we prove the following result.

Theorem 4.1.

Let r>0r>0 and k0k\geq 0 be integers. Further, let qq be a prime power and nn be a positive integer such that rqn1r\mid q^{n}-1 and there exists a polynomial gPkg\in P_{k}. Then, Nr,k,g(h,d,H)>0N_{r,k,g}(h,d,H)>0, if

qn/2ϑ>2rrad(r)W(h)W(d)W(H),q^{n/2-\vartheta}>2r\hskip 1.42262pt\mathrm{rad}(r)W(h)W(d)W(H),

where, ϑ=2k\vartheta=2k, if gcd(q,n)=1\mathrm{gcd}(q,n)=1, and ϑ=3k\vartheta=3k, otherwise.

Proof.

Let g=i=0kaixiPkg=\sum_{i=0}^{k}a_{i}x^{i}\in P_{k} with ak=1a_{k}=1, and define gx=i=0kaixqig\circ x=\sum_{i=0}^{k}a_{i}x^{q^{i}}. Further, let 𝒵\mathcal{Z} be the set of zeroes of the polynomial gxg\circ x. Clearly, 𝒵\mathcal{Z} consists of all the elements β\beta, whose 𝔽q\mathbb{F}_{q}-order divides gg, and hence, |𝒵|=ggΦq(g)=qk|\mathcal{Z}|=\sum_{g^{\prime}\mid g}\Phi_{q}(g^{\prime})=q^{k}. Now, by using the definitions of the characteristic functions Υh\Upsilon_{h}, Γrd\Gamma_{r}^{d} and 𝒬gH\mathcal{Q}_{g}^{H}, we have

Nr,k,g(h,d,H)=\displaystyle N_{r,k,g}(h,d,H)= β𝔽qn𝒵Υh(β)Γrd(gβ)𝒬gH((gβ)1)\displaystyle\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\Upsilon_{h}(\beta)\Gamma_{r}^{d}(g\circ\beta)\mathcal{Q}_{g}^{H}((g\circ\beta)^{-1}) (8)
=\displaystyle= hhd1d,d2u,ejλj;1jsg1H,g2π,hi,Λi;1it𝝁𝚽(h,d1,g1)((ej)j=1s,(hi)i=1t)ψh×\displaystyle\hskip 2.84526pt\mathcal{H}\sum_{\begin{subarray}{c}h^{\prime}\mid h\end{subarray}}\sum_{\begin{subarray}{c}d_{1}\mid d,d_{2}\mid u,\\ e_{j}\mid\lambda_{j};1\leq j\leq s\end{subarray}}\sum_{\begin{subarray}{c}g_{1}\mid H,g_{2}\mid\pi,\\ h_{i},\mid\Lambda_{i};1\leq i\leq t\end{subarray}}\frac{\bm{\mu}}{\bm{\Phi}}(h^{\prime},d_{1},g_{1})\bm{\ell}\big{(}(e_{j})_{j=1}^{s},(h_{i})_{i=1}^{t}\big{)}\sum_{\begin{subarray}{c}\psi_{h^{\prime}}\end{subarray}}\times
χd1,χd2,χej;1jsψg1,ψg2,ψhi; 1itβ𝔽qn𝒵ψh(β)χD(gβ)ψF((gβ)1),\displaystyle\sum_{\begin{subarray}{c}\chi_{d_{1}},\chi_{d_{2}},\\ \chi_{e_{j}};1\leq j\leq s\end{subarray}}\sum_{\begin{subarray}{c}\psi_{g_{1}},\psi_{g_{2}},\\ \psi_{h_{i}};\ 1\leq i\leq t\end{subarray}}\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\psi_{h^{\prime}}(\beta)\chi_{D}(g\circ\beta)\psi_{F}((g\circ\beta)^{-1}),

where =Φq(h)ϕ(d)Φq(H)rdqdeg(h)+deg(H)+deg(g)\mathcal{H}=\dfrac{\Phi_{q}(h)\phi(d)\Phi_{q}(H)}{rdq^{\mathrm{deg}(h)+\mathrm{deg}(H)+\mathrm{deg}(g)}}, 𝝁𝚽(h,d1,g1)=μ(h)μ(d1)μ(g1)Φq(h)ϕ(d1)Φq(g1)\dfrac{\bm{\mu}}{\bm{\Phi}}(h^{\prime},d_{1},g_{1})=\dfrac{\mu^{\prime}(h^{\prime})\mu(d_{1})\mu^{\prime}(g_{1})}{\Phi_{q}(h^{\prime})\phi(d_{1})\Phi_{q}(g_{1})}, ((ej)j=1s,(hi)i=1t)=j=1spj,eji=1tfi,hi\bm{\ell}\big{(}(e_{j})_{j=1}^{s},(h_{i})_{i=1}^{t}\big{)}=\prod_{j=1}^{s}\ell_{p_{j},e_{j}}\prod_{i=1}^{t}\ell_{f_{i},h_{i}}^{\prime}, χD=χd1χd2χe1χes\chi_{D}=\chi_{d_{1}}\chi_{d_{2}}\chi_{e_{1}}\cdots\chi_{e_{s}} is a multiplicative character of order D=d1d2e1esD=d_{1}d_{2}e_{1}\cdots e_{s}, since d1,d2,e1,,esd_{1},d_{2},e_{1},\ldots,e_{s} are mutually co-prime, and similarly, ψF=ψg1ψg2ψh1ψht\psi_{F}=\psi_{g_{1}}\psi_{g_{2}}\psi_{h_{1}}\cdots\psi_{h_{t}} is the additive character of 𝔽q\mathbb{F}_{q}-order F=g1g2h1htF=g_{1}g_{2}h_{1}\cdots h_{t}.

Now, let ψ0\psi_{0} be the canonical additive character of 𝔽qn\mathbb{F}_{q^{n}}, then there exist y1,y2𝔽qny_{1},y_{2}\in\mathbb{F}_{q^{n}}, such that ψh(β)=ψ0(y1β)\psi_{h^{\prime}}(\beta)=\psi_{0}(y_{1}\beta) and ψF(gβ)=ψ0(y2(gβ))\psi_{F}(g\circ\beta)=\psi_{0}(y_{2}(g\circ\beta)) for all β𝔽qn\beta\in\mathbb{F}_{q^{n}}. Thus,

β𝔽qn𝒵ψh(β)χD(gβ)ψF((gβ)1)=β𝔽qn𝒵χD(gβ)ψ0(y1β+y2(gβ)1).\displaystyle\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\psi_{h^{\prime}}(\beta)\chi_{D}(g\circ\beta)\psi_{F}((g\circ\beta)^{-1})=\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\chi_{D}(g\circ\beta)\psi_{0}(y_{1}\beta+y_{2}(g\circ\beta)^{-1}).

Now, we estimate |β𝔽qn𝒵χD(gβ)ψ0(y1β+y2(gβ)1)|\big{|}\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\chi_{D}(g\circ\beta)\psi_{0}(y_{1}\beta+y_{2}(g\circ\beta)^{-1})\big{|} using Lemma 2.4, for which we first show that the rational function y1x+y2(gx)1y_{1}x+y_{2}(g\circ x)^{-1} is not of the form L(x)qnL(x)L(x)^{q^{n}}-L(x) for any rational function L(x)=L1L2L(x)=\frac{L_{1}}{L_{2}} (in the simplest form) over the algebraic closure 𝔽\mathbb{F} of 𝔽qn\mathbb{F}_{q^{n}}, unless y1=y2=0y_{1}=y_{2}=0.

Assume that, y1x+y2(gx)1=(L1L2)qnL1L2y_{1}x+y_{2}(g\circ x)^{-1}=(\frac{L_{1}}{L_{2}})^{q^{n}}-\frac{L_{1}}{L_{2}} for some L1,L2𝔽[x]L_{1},L_{2}\in\mathbb{F}[x] i.e. L2qn(y1i=0kaixqi+1+y2)=i=0kaixqi(L1qnL1L2qn1).L_{2}^{q^{n}}(y_{1}\sum_{i=0}^{k}a_{i}x^{q^{i}+1}+y_{2})=\sum_{i=0}^{k}a_{i}x^{q^{i}}(L_{1}^{q^{n}}-L_{1}L_{2}^{q^{n}-1}). First, let y20y_{2}\neq 0, then i=0kaixqi=L2qn,\sum_{i=0}^{k}a_{i}x^{q^{i}}=L_{2}^{q^{n}}, which is not possible because k<nk<n. So, y2y_{2} must be zero. Additionally, let y10y_{1}\neq 0, then L2qny1x=L1qnL1L2qn1L_{2}^{q^{n}}y_{1}x=L_{1}^{q^{n}}-L_{1}L_{2}^{q^{n}-1}, which implies L2L_{2} is a constant polynomial and qn1q^{n}\mid 1, which is again not possible. Hence, y1=y2=0y_{1}=y_{2}=0. Therefore, if one of the y1y_{1} or y2y_{2} is non-zero, then by Lemma 2.4,

|β𝔽qn𝒵χD(gβ)ψ0(y1β+y2(gβ)1)|2qn/2+k.\big{|}\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\chi_{D}(g\circ\beta)\psi_{0}(y_{1}\beta+y_{2}(g\circ\beta)^{-1})\big{|}\leq 2q^{n/2+k}. (I)

Now, let y1=y2=0y_{1}=y_{2}=0, then ψh\psi_{h^{\prime}} and ψF\psi_{F} are trivial additive characters, i.e. h=1=Fh^{\prime}=1=F, and the latter one is possible if and only if each factor of F=g1g2h1htF=g_{1}g_{2}h_{1}\cdots h_{t} is 11, since g1,g2,h1,,htg_{1},g_{2},h_{1},\ldots,h_{t} are mutually co-prime. Hence, we have

β𝔽qn𝒵χD(gβ)ψ0(y1β+y2(gβ)1)=β𝔽qn𝒵χD(gβ).\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\chi_{D}(g\circ\beta)\psi_{0}(y_{1}\beta+y_{2}(g\circ\beta)^{-1})=\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\chi_{D}(g\circ\beta).

Notice that, if the polynomial gx=L(x)Dg\circ x=L(x)^{D} for some L(x)𝔽[x]L(x)\in\mathbb{F}[x], then Dgcd(qk,qn1)D\mid\mathrm{gcd}(q^{k},q^{n}-1), i.e D=1D=1. Hence, by Lemma 2.3, for D1D\neq 1, we get

|β𝔽qn𝒵χD(gβ)|(qk1)qn/2<2qn/2+k.\big{|}\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\chi_{D}(g\circ\beta)\big{|}\leq(q^{k}-1)q^{n/2}<2q^{n/2+k}. (II)

Finally, let h=F=1h^{\prime}=F=1 and D=1D=1, then

β𝔽qn𝒵χD(gβ)ψ0(y1β+y2(gβ)1)=qnqk.\sum_{\beta\in\mathbb{F}_{q^{n}}\setminus\mathcal{Z}}\chi_{D}(g\circ\beta)\psi_{0}(y_{1}\beta+y_{2}(g\circ\beta)^{-1})=q^{n}-q^{k}. (III)

Using the estimates, obtained in (I), (II) and (III), in Equation (8), we get

|Nr,k,g(h,d,H)\displaystyle\big{|}N_{r,k,g}(h,d,H)- (qnqk)((1)j=1s,(1)i=1t)|\displaystyle(q^{n}-q^{k})\bm{\ell}\big{(}(1)_{j=1}^{s},(1)_{i=1}^{t}\big{)}\mathcal{H}\big{|}\leq
hhd1d,d2u,ejλj;1jsg1H,g2π,hi,Λi;1it(all are not 1 simultaneously)|𝝁𝚽(h,d1,g1)||((ej)j=1s,(hi)i=1t)|×\displaystyle\mathcal{H}\underbrace{\sum_{\begin{subarray}{c}h^{\prime}\mid h\end{subarray}}\sum_{\begin{subarray}{c}d_{1}\mid d,d_{2}\mid u,\\ e_{j}\mid\lambda_{j};1\leq j\leq s\end{subarray}}\sum_{\begin{subarray}{c}g_{1}\mid H,g_{2}\mid\pi,\\ h_{i},\mid\Lambda_{i};1\leq i\leq t\end{subarray}}}_{\text{(all are not 1 simultaneously)}}\big{|}\frac{\bm{\mu}}{\bm{\Phi}}(h^{\prime},d_{1},g_{1})\big{|}\big{|}\bm{\ell}\big{(}(e_{j})_{j=1}^{s},(h_{i})_{i=1}^{t}\big{)}\big{|}\times
ψhχd1,χd2,χej;1jsψg1,ψg2,ψhi; 1it2qn/2+k.\displaystyle\sum_{\begin{subarray}{c}\psi_{h}^{\prime}\end{subarray}}\sum_{\begin{subarray}{c}\chi_{d_{1}},\chi_{d_{2}},\\ \chi_{e_{j}};1\leq j\leq s\end{subarray}}\sum_{\begin{subarray}{c}\psi_{g_{1}},\psi_{g_{2}},\\ \psi_{h_{i}};\ 1\leq i\leq t\end{subarray}}2q^{n/2+k}.

Since |((ej)j=1s,(hi)i=1t)|((1)j=1s,(1)i=1t)\big{|}\bm{\ell}\big{(}(e_{j})_{j=1}^{s},(h_{i})_{i=1}^{t}\big{)}\big{|}\leq\bm{\ell}\big{(}(1)_{j=1}^{s},(1)_{i=1}^{t}\big{)}, then

|\displaystyle\big{|} Nr,k,g(h,d,H)(qnqk)((1)j=1s,(1)i=1t)|\displaystyle N_{r,k,g}(h,d,H)-(q^{n}-q^{k})\bm{\ell}\big{(}(1)_{j=1}^{s},(1)_{i=1}^{t}\big{)}\mathcal{H}\big{|}\leq
2qn/2+k((1)j=1s,(1)i=1t)(uj=1sλjqdeg(π)+i=1tdeg(Λi)W(h)W(d)W(H)1).\displaystyle\hskip 1.42262pt2q^{n/2+k}\bm{\ell}\big{(}(1)_{j=1}^{s},(1)_{i=1}^{t}\big{)}\mathcal{H}\big{(}u\prod_{j=1}^{s}\lambda_{j}q^{\mathrm{deg}(\pi)+\sum_{i=1}^{t}\mathrm{deg}(\Lambda_{i})}W(h)W(d)W(H)-1\big{)}.

From the above equation, we get

Nr,k,g(h,d,H)((1)j=1s,(1)i=1t)\displaystyle\frac{N_{r,k,g}(h,d,H)}{\bm{\ell}\big{(}(1)_{j=1}^{s},(1)_{i=1}^{t}\big{)}\mathcal{H}} (qnqk)2qn/2+k×\displaystyle\geq(q^{n}-q^{k})-2q^{n/2+k}\times
(uj=1sλjqdeg(π)+i=1tdeg(Λi)W(h)W(d)W(H)1).\displaystyle\big{(}u\prod_{j=1}^{s}\lambda_{j}q^{\mathrm{deg}(\pi)+\sum_{i=1}^{t}\mathrm{deg}(\Lambda_{i})}W(h)W(d)W(H)-1\big{)}.

Clearly, Nr,k,g(h,d,H)>0N_{r,k,g}(h,d,H)>0, if

qn/2k>2uj=1sλjqdeg(π)+i=1tdeg(Λi)W(h)W(d)W(H).q^{n/2-k}>2u\prod\limits_{j=1}^{s}\lambda_{j}q^{\mathrm{deg}(\pi)+\sum_{i=1}^{t}\mathrm{deg}(\Lambda_{i})}W(h)W(d)W(H). (9)

Now, if gcd(q,n)=1\mathrm{gcd}(q,n)=1, then xn1x^{n}-1 has no repeated factor over 𝔽q\mathbb{F}_{q}, and hence, g=πg=\pi and deg(Λi)=0\mathrm{deg}(\Lambda_{i})=0 for each i=1,2,,t.i=1,2,\ldots,t. Thus, qdeg(π)+i=1tdeg(Λi)=qkq^{\mathrm{deg}(\pi)+\sum_{i=1}^{t}\mathrm{deg}(\Lambda_{i})}=q^{k}, otherwise q2k.\leq q^{2k}. Moreover, uj=1sλj=rj=1spjrrad(r)u\prod\limits_{j=1}^{s}\lambda_{j}=r\prod_{j=1}^{s}p_{j}\leq r\hskip 1.42262pt\mathrm{rad}(r). Hence, from Equation (9), Nr,k,g(h,d,H)>0N_{r,k,g}(h,d,H)>0, if qn/2ϑ>2rrad(r)W(h)W(d)W(H)q^{n/2-\vartheta}>2r\hskip 1.42262pt\mathrm{rad}(r)W(h)W(d)W(H), where, ϑ=2k\vartheta=2k, if gcd(q,n)=1\mathrm{gcd}(q,n)=1, and ϑ=3k\vartheta=3k, otherwise. ∎

The following corollary provides the desired existence of rr-primitive kk-normal pairs (α,α1)(\alpha,\alpha^{-1}) in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q.\mathbb{F}_{q}.

Corollary 4.1.1.

With rr and kk as in Theorem 4.1, there exists a pair (α,α1)(\alpha,\alpha^{-1}) of rr-primitive kk-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, i.e. Nr,k,g(xn1,R,G)>0N_{r,k,g}(x^{n}-1,R,G)>0, if

qn/2ϑ>2rrad(r)W(xn1)W(R)W(G),q^{n/2-\vartheta}>2r\hskip 1.42262pt\mathrm{rad}(r)W(x^{n}-1)W(R)W(G), (10)

where, ϑ=2k\vartheta=2k, if gcd(q,n)=1\mathrm{gcd}(q,n)=1, and ϑ=3k\vartheta=3k, otherwise.

Proof.

Let h=xn1h=x^{n}-1, d=Rd=R and H=GH=G in the above theorem, then the result follows immediately. ∎

Observe that, Inequalities (10) can never hold for n2ϑn\leq 2\vartheta, and as the value of kk increases, the existence of a pair (α,α1)(\alpha,\alpha^{-1}) of rr-primitive kk-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} becomes more challenging for all n.n. However, we can show that such a pair exists in all but finitely many fields 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, for any fixed integers rr and kk. For this, we need the following two lemmas.

Lemma 4.1.

[10, Lemma 2.9] Let qq be a prime power and nn be a positive integer. Then, we have W(xn1)212(n+gcd(n,q1))W(x^{n}-1)\leq 2^{\frac{1}{2}(n+\mathrm{gcd}(n,q-1))}. In particular, W(xn1)2nW(x^{n}-1)\leq 2^{n} and W(xn1)=2nW(x^{n}-1)=2^{n} if and only if n(q1)n\mid(q-1). Furthermore, W(xn1)23n/4W(x^{n}-1)\leq 2^{3n/4} if n(q1)n\nmid(q-1), since in this case, gcd(n,q1)n/2.\mathrm{gcd}(n,q-1)\leq n/2.

Lemma 4.2.

[3, Lemma 3.7] For any MM\in\mathbb{N} and a positive real number ν\nu, W(M)𝒞νM1/νW(M)\leq\mathcal{C}_{\nu}M^{1/\nu}, where 𝒞ν=i=1t2pi1/ν\mathcal{C}_{\nu}=\prod\limits_{i=1}^{t}\frac{2}{p_{i}^{1/\nu}} and p1,p2,,ptp_{1},p_{2},\ldots,p_{t} are the primes 2ν\leq 2^{\nu} that divide M.M.

Following the idea of [18, Proposition 3.1], we have the following result.

Proposition 4.1.

Let r1r\geq 1 and k0k\geq 0 be fixed integers. Let qq be a prime power and n>2ϑn>2\vartheta, where ϑ\vartheta is as defined in Theorem 4.1, such that rqn1r\mid q^{n}-1 and there exists a polynomial gPkg\in P_{k}. Then, there exists a pair (α,α1)(\alpha,\alpha^{-1}) of rr-primitive kk-normal elements in all but finitely many fields 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q.\mathbb{F}_{q}.

Proof.

Notice that, W(R)W(qn1)W(R)\leq W(q^{n}-1) and W(G)W(xn1).W(G)\leq W(x^{n}-1). Hence, from Lemmas 4.1 and 4.2, Inequality (10) holds if

qn/2ϑ>2rrad(r)𝒞νqn/ν22n.q^{n/2-\vartheta}>2r\hskip 1.42262pt\mathrm{rad}(r)\mathcal{C}_{\nu}q^{n/\nu}2^{2n}. (11)

Set A=2rrad(r)𝒞νA=2r\hskip 1.42262pt\mathrm{rad}(r)\mathcal{C}_{\nu} and ν=max{4ϑ+2,93.46}.\nu=\mathrm{max}\hskip 1.42262pt\{4\vartheta+2,93.46\}. Then, taking logarithm on both sides, we get the following equivalent inequality.

logq>logA+2nlog2n/2ϑn/ν.\mathrm{log}\hskip 1.42262ptq>\frac{\mathrm{log}\hskip 1.42262ptA+2n\mathrm{log}\hskip 1.42262pt2}{n/2-\vartheta-n/\nu}.

The right hand side of the above inequality is a decreasing function of nn, if n2ϑnν>0\frac{n}{2}-\vartheta-\frac{n}{\nu}>0, i.e. ν2ν>2ϑn\frac{\nu-2}{\nu}>\frac{2\vartheta}{n}, which holds for n>2ϑn>2\vartheta. Therefore, there exists a number q0q_{0} such that the above inequality holds for qq0q\geq q_{0} and n2ϑ+1n\geq 2\vartheta+1.

Now, for q<q0q<q_{0}, we consider the following inequality, which is equivalent to Inequality (11).

n>logA+ϑlogq(1/21/ν)logq2log2.n>\frac{\mathrm{log}\hskip 1.42262ptA+\vartheta\hskip 1.42262pt\mathrm{log}\hskip 1.42262ptq}{(1/{2}-1/{\nu})\mathrm{log}\hskip 1.42262ptq-2\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2}.

Clearly, the denominator of the right hand side of the above inequality is positive, if q>16q>16. Hence, corresponding to each 16<q<q016<q<q_{0}, there exists a natural number nqn_{q} such that the above inequality holds for nnq.n\geq n_{q}.

Now, for 3q163\leq q\leq 16, we use the bound W(xn1)<2n/3+cqW(x^{n}-1)<2^{n/3+c_{q}} (from [10, Lemmas 2.9, 2.11]), where cqc_{q} is a constant corresponding to each qq. Then, Inequality (11) holds, if

n>logA+ϑlogq+2cqlog2(1/21/ν)logq(2log2)/3.n>\frac{\mathrm{log}\hskip 1.42262ptA+\vartheta\hskip 1.42262pt\mathrm{log}\hskip 1.42262ptq+2c_{q}\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2}{(1/{2}-1/{\nu})\mathrm{log}\hskip 1.42262ptq-(2\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2)/{3}}.

The above inequality holds for nnqn\geq n_{q}, where nqn_{q} is a natural number corresponding to each 3q163\leq q\leq 16.

Finally, let q=2q=2, and from [10, Lemma 2.11], we have W(xn1)<2n45W(x^{n}-1)<2^{\frac{n-4}{5}} for all nn except n{1,2,3,4,5,7,9,15,21}n\in\{1,2,3,4,5,7,9,15,21\}, and Inequality (11) holds, if

n>logA+ϑlogq(8/5)log2(1/21/ν)logq(2log2)/5.n>\frac{\mathrm{log}\hskip 1.42262ptA+\vartheta\hskip 1.42262pt\mathrm{log}\hskip 1.42262ptq-(8/5)\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2}{(1/{2}-1/{\nu})\mathrm{log}\hskip 1.42262ptq-(2\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2)/{5}}.

There exists a natural number n2n_{2} such that the last inequality holds for nn2.n\geq n_{2}.

Hence, from the above discussion, we conclude that, there exists a pair (α,α1)(\alpha,\alpha^{-1}) of rr-primitive kk-normal elements in all but finitely many 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q.\mathbb{F}_{q}.

4.1 The sieving technique

In this section, we give a sieving technique to improve Inequality (10) by following the methods of Cohen and Huczynska [3]. The proofs of the next two lemmas are similar to that of [14, Lemmas 5.1, 5.2], hence omitted.

Lemma 4.3.

Let rqn1r\mid q^{n}-1, and let kk be a non-negative integer and gPkg\in P_{k}. Further, let dd be a divisor of RR and {p1,p2,,pl}\{p_{1},p_{2},\ldots,p_{l}\} be the set of remaining distinct primes dividing RR. Furthermore, let hh, HH be the divisors of xn1x^{n}-1, GG respectively, and {h1,h2,,hm}\{h_{1},h_{2},\ldots,h_{m}\}, {H1,H2,,Hm}\{H_{1},H_{2},\ldots,H_{m^{\prime}}\} be the sets of remaining distinct irreducible factors of xn1x^{n}-1, GG respectively. Then

Nr,k,g(xn1,R,G)\displaystyle N_{r,k,g}(x^{n}-1,R,G)\geq i=1mNr,k,g(hhi,d,H)+i=1lNr,k,g(h,dpi,H)+\displaystyle\sum_{i=1}^{m}N_{r,k,g}(hh_{i},d,H)+\sum_{i=1}^{l}N_{r,k,g}(h,dp_{i},H)+
i=1mNr,k,g(h,d,HHi)(m+l+m1)Nr,k,g(h,d,H).\displaystyle\sum_{i=1}^{m^{\prime}}N_{r,k,g}(h,d,HH_{i})-(m+l+m^{\prime}-1)N_{r,k,g}(h,d,H).
Lemma 4.4.

With the notations of the above lemma, define

𝒟:=1i=1m1qdeg(hi)i=1l1pii=1m1qdeg(Hi) and 𝒮:=m+l+m1𝒟+2.\mathcal{D}:=1-\sum_{i=1}^{m}\frac{1}{q^{deg(h_{i})}}-\sum_{i=1}^{l}\frac{1}{p_{i}}-\sum_{i=1}^{m^{\prime}}\frac{1}{q^{deg(H_{i})}}\text{ and }\mathcal{S}:=\frac{m+l+m^{\prime}-1}{\mathcal{D}}+2.

Suppose 𝒟>0,\mathcal{D}>0, then Nr,k,g(xn1,R,G)>0N_{r,k,g}(x^{n}-1,R,G)>0, if

qn/2ϑ>2rrad(r)W(h)W(d)W(H)𝒮,q^{n/2-\vartheta}>2r\hskip 1.42262pt\mathrm{rad}(r)W(h)W(d)W(H)\mathcal{S}, (12)

where, ϑ=2k\vartheta=2k, if gcd(q,n)=1\mathrm{gcd}(q,n)=1, and ϑ=3k\vartheta=3k, otherwise.

5 Existence of a primitive 11-normal pair (α,α1)(\alpha,\alpha^{-1})

In [3], Cohen and Huczynska proved the existence of a pair (α,α1)(\alpha,\alpha^{-1}) of primitive normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}. Here, we present an application of the sufficient condition and the sieving inequality, obtained in the previous section, by showing the existence of a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}. All the non-trivial calculations wherever needed in this section are done using SageMath [19].

From Inequality (10), there exist pairs (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, if

qn/2ϑ>2W(xn1)W(R)W(G),q^{n/2-\vartheta}>2W(x^{n}-1)W(R)W(G), (13)

where ϑ=2\vartheta=2, if gcd(q,n)=1\mathrm{gcd}(q,n)=1, otherwise, ϑ=3.\vartheta=3. Notice that, if a pair (q,n)(q,n) satisfies Inequality (13) in the case of ϑ=3\vartheta=3, then it satisfies the same in the case of ϑ=2\vartheta=2 as well. Moreover, Inequality (13) can never hold for n4n\leq 4, if ϑ=2\vartheta=2, and for n6n\leq 6, if ϑ=3.\vartheta=3. Therefore, first we assume that ϑ=3\vartheta=3 and settle the cases n14n\geq 14 in Subsection 5.1, and then discuss the cases 5n135\leq n\leq 13 in Subsection 5.2.

5.1 The cases n14n\geq 14

From Lemmas 4.1 and 4.2, we have W(R)W(qn1)<𝒞νqn/νW(R)\leq W({q^{n}-1})<\mathcal{C}_{\nu}q^{n/\nu}, W(xn1)2nW(x^{n}-1)\leq 2^{n} and W(G)2n1W(G)\leq 2^{n-1}, and hence, Inequality (13) holds, if

qn/23>𝒞νqn/ν22n.q^{n/2-3}>\mathcal{C}_{\nu}\hskip 1.42262ptq^{n/\nu}2^{2n}. (14)

First, we study the cases q>3q>3 and n14n\geq 14 in the following Lemma and then the cases q=2,3q=2,3 and n14n\geq 14 in Lemma 5.3.

Lemma 5.1.

Let q>3q>3 be a prime power and n14n\geq 14 be a positive integer. Then, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}.

Proof.

For n14n\geq 14, if we choose ν=7.6\nu=7.6, Inequality (14) holds for q214183q\geq 214183. Now, for q<214183q<214183, we rewrite Inequality (14) as follows

n>log(𝒞ν)+3log(q)(1/21/ν)logq2log2,n>\frac{\mathrm{log}\hskip 1.42262pt(\mathcal{C}_{\nu})+3\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt(q)}{(1/{2}-1/{\nu})\mathrm{log}\hskip 1.42262ptq-2\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2}, (15)

provided ν>2logqlogq4log2\nu>\frac{2\hskip 1.42262pt\mathrm{log}\hskip 1.42262ptq}{\mathrm{log}\hskip 1.42262ptq-4\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2} and q>16.q>16. Notice that, the value of ν\nu is large for small values of qq, and it is difficult to calculate 𝒞ν\mathcal{C}_{\nu} in our computer. So, for simplicity, we assume 37q<214183,37\leq q<214183, and for each such qq, we find a natural number nqn_{q} (see Table 1) such that Inequality (15) holds for nnqn\geq n_{q}.

Table 1: Values of nqn_{q} for 37q<21418337\leq q<214183 using W(xn1)2nW(x^{n}-1)\leq 2^{n}.
ν\nu qq nqn_{q} ν\nu qq nqn_{q} ν\nu qq nqn_{q}
7.6 214183\geq 214183 14 7.8 307, 311, 313, 317 34 8.3 107 62
7.5 61747 to 214182 15 7.8 277, 281, 283, 289, 293 35 8.3 103 64
7.5 24428 to 61746 16 7.8 257, 263, 269, 271 36 8.3 101 65
7.5 11926 to 24427 17 7.8 243, 251, 256 37 8.4 97 67
7.5 6738 to 11925 18 7.8 229, 233, 239, 241 38 8.5 89 73
7.5 4231 to 6737 19 7.9 223, 227 39 8.5 83 79
7.5 2875 to 4230 20 7.9 211 40 8.6 81 81
7.5 2074 to 2874 21 7.9 197, 199 41 8.6 79 83
7.5 1569 to 2073 22 7.9 191, 193 42 8.7 73 92
7.5 1233 to 1568 23 7.9 179, 181 43 8.8 71 96
7.5 999 to 1232 24 7.9 173 44 8.8 67 104
7.5 829 to 998 25 8 167, 169 45 8.9 64 111
7.5 703 to 828 26 8 157, 163 47 9 61 120
7.6 606 to 702 27 8 149, 151 48 9.1 59 128
7.6 531 to 605 28 8.1 137, 139 51 9.3 53 157
7.6 471 to 530 29 8.1 128, 131 53 9.6 49 188
7.6 422 to 470 30 8.1 125, 127 54 9.7 47 210
7.6 382 to 421 31 8.1 121 56 10.1 43 272
7.7 348 to 381 32 8.2 113 59 10.3 41 320
7.7 331, 337, 343, 347 33 8.3 109 61 10.8 37 489

Now, for 3<q<373<q<37, we use the bound W(xn1)2n/3+2(q21)/3W(x^{n}-1)\leq 2^{n/3+2(q^{2}-1)/3}, which we obtain from [10, Lemma 2.9], and find that, Inequality (14) holds, if

n>log(24(q21)/3𝒞ν)(1/21/ν)logq(2log2)/3,n>\frac{\mathrm{log}\hskip 1.42262pt(2^{4(q^{2}-1)/3}\mathcal{C}_{\nu})}{(1/{2}-1/{\nu})\mathrm{log}\hskip 1.42262ptq-(2\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2)/3}, (16)

provided ν>6logq3logq4log2\nu>\frac{6\hskip 1.42262pt\mathrm{log}\hskip 1.42262ptq}{3\hskip 1.42262pt\mathrm{log}\hskip 1.42262ptq-4\hskip 1.42262pt\mathrm{log}\hskip 1.42262pt2} and q>2q>2. For these values of qq, we obtain a natural number nqn_{q} (see Table 2) such that Inequality (16) holds for nnqn\geq n_{q}.

Now, for n<nqn<n_{q} and qq’s listed in Tables 1 and 2, we test Inequality (13) with ϑ=3\vartheta=3, and list the values of qq and nn in Table 3, which fail to satisfy this inequality.

Table 2: Values of nqn_{q} for 3<q<373<q<37 using W(xn1)2n/3+2(q21)/3W(x^{n}-1)\leq 2^{n/3+2(q^{2}-1)/3}.
ν\nu 11.8 11.7 11.6 11.4 11.2 11 10.6 10.3
qq 32 31 29 27 25 23 19 17
nqn_{q} 1019 973 883 797 714 636 493 428
ν\nu 10.1 9.7 9.4 9 8.8 8.7 8.5 9
qq 16 13 11 9 8 7 5 4
nqn_{q} 397 312 263 222 206 195 214 351
Table 3: Pairs (q,n)(q,n) fail to satisfy Inequality (13) with ϑ=3\vartheta=3.
nn qq
14 4, 5, 8, 9, 11, 13, 23, 25, 27, 29, 41, 43, 64, 71, 113, 125, 127, 169, 197, 211, 239, 281, 337, 379, 421, 449, 463, 491, 547, 617, 631, 659, 673, 701, 729, 743, 883, 911, 953
15 4, 7, 8, 11, 13, 16, 19, 29, 31, 41, 49, 61, 64, 71, 121, 151, 181, 211, 241, 256, 271, 331, 361, 421, 571, 631, 751, 841
16 5, 7, 9, 11, 13, 17, 19, 23, 25, 27, 29, 31, 41, 49, 73, 81, 89, 97, 113, 193, 241, 257, 289, 337, 401
17 4, 16, 103, 256
18 4, 5, 7, 13, 17, 19, 25, 31, 37, 43, 73, 109, 127, 163, 181, 199, 289, 361
20 7, 9, 11, 13, 19, 29, 31, 41, 61, 81, 101, 121
21 4, 8, 13, 16, 43, 64, 169
22 23, 67, 89
23 47
24 5, 7, 11, 13, 17, 19, 25, 37, 49, 73, 97, 121
26 27, 53
28 13, 29
30 4, 7, 11, 19, 31, 61
31 32
32 17
36 5, 19, 37
40 9, 11, 41
42 43
45 4
48 5, 7
63 5

Now, for the values of qq and nn, listed in Table 3, we test the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=3\vartheta=3, in which we choose the suitable values of hh, dd and HH, and test Inequality (12) (whose pseudocode is given in A), which returns True for all the pairs (q,n)(q,\ n) except (4, 15), (5, 16), (5, 24), (8, 14), (9, 16), (16, 15), (17, 16), (19, 18). For these remaining pairs of (q,n)(q,\ n), we explicitly search for a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} using SageMath procedure DIRECT_SEARCH(q,n) (pseudocode is given in A), and find that such a pair always exists. ∎

Now, for the cases q=2, 3q=2,\ 3, first we recall some results. Let qq be a prime power and n=qinn=q^{i}\cdot n^{\prime} be a positive integer with gcd(n,q)=1\mathrm{gcd}(n^{\prime},q)=1; i0i\geq 0. Further, let ee be the multiplicative order of qq modulo nn^{\prime}, then xn1x^{n^{\prime}}-1 factors into irreducible polynomials of degree e.\leq e. Let ρ(q,n)\rho(q,n^{\prime}) be the ratio of the number of irreducible factors of xn1x^{n^{\prime}}-1 of degree <e<e to nn^{\prime}, then nρ(q,n)=nρ(q,n)n\rho(q,n)=n^{\prime}\rho(q,n^{\prime}), and we have the following lemma.

Lemma 5.2.

[3, Lemma 7.1] Assume that n>4n>4 (pn)(p\nmid n), then the following holds:

  1. (i)(i)

    Suppose q=2q=2, then ρ(2,5)=1/5;ρ(2,9)=2/9;ρ(2,21)=4/21;\rho(2,5)=1/5;\ \rho(2,9)=2/9;\ \rho(2,21)=4/21; otherwise ρ(2,n)1/6.\rho(2,n)\leq 1/6.

  2. (ii)(ii)

    Suppose q=3q=3, then ρ(3,16)=5/16;\rho(3,16)=5/16; otherwise ρ(3,n)1/4.\rho(3,n)\leq 1/4.

If we choose hh as the product of all irreducible factors of xn1x^{n^{\prime}}-1 of degree <e<e, H=h/(x1)H=h/(x-1) and d=Rd=R in Lemma 4.4, then following Lemma 10 of [15], we get that, 𝒟>12/e>0\mathcal{D}>1-2/e>0 and 𝒮2n2n\mathcal{S}\leq 2n^{\prime}\leq 2n, provided e>2e>2 (i.e. nq21n^{\prime}\nmid q^{2}-1). Using these results, we prove the following lemma.

Lemma 5.3.

Let q=2,3q=2,3 and n14n\geq 14 be a positive integer. Then, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}.

Proof.

Let n=qinn=q^{i}\cdot n^{\prime} with gcd(q,n)=1\mathrm{gcd}(q,n^{\prime})=1 and i0.i\geq 0. First, assume that nq21,n^{\prime}\nmid q^{2}-1, then e>2e>2 and n>4n^{\prime}>4. If we choose d=Rd=R, hh as the product of all irreducible factors of xn1x^{n^{\prime}}-1 of degree <e<e, and H=h/(x1)H=h/(x-1), then from the above discussion and Inequality (12), there exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, if

qn/23>2𝒞νqn/νW(h)W(H)𝒮.q^{n/2-3}>2\mathcal{C}_{\nu}q^{n/\nu}W(h)W(H)\mathcal{S}. (17)

Notice that, W(H)=W(h)/2W(H)=W(h)/2 and W(h)=2nρ(q,n)W(h)=2^{n^{\prime}\rho(q,n^{\prime})}. From Lemma 5.2 nρ(q,n)n/6n/6,n^{\prime}\rho(q,n^{\prime})\leq n^{\prime}/6\leq n/6, when q=2q=2, and nρ(q,n)n/4n/4,n^{\prime}\rho(q,n^{\prime})\leq n^{\prime}/4\leq n/4, when q=3q=3. Thus, the above inequality holds, if

{qn/23>𝒞νqn/ν2n/32n, for q=2,qn/23>𝒞νqn/ν2n/22n, for q=3.\left\{\begin{array}[]{ll}q^{n/2-3}>\mathcal{C}_{\nu}q^{n/\nu}2^{n/3}2n,&\text{ for }q=2,\\ q^{n/2-3}>\mathcal{C}_{\nu}q^{n/\nu}2^{n/2}2n,&\text{ for }q=3.\end{array}\right.

The above inequalities hold for n557n\geq 557, when q=2q=2 and ν=8.7\nu=8.7, and for n265n\geq 265, when q=3q=3 and ν=8.3.\nu=8.3. For the remaining qq and nn, we test Inequality (13), and find that this fails for n=n= 14, 15, 17, 18, 20, 21, 22, 27, 28, 30, 31, 33, 35, 42, 45, 63, when q=2q=2, and for n=n= 14, 15, 16, 20, 22, 26, 28, 32, 40, when q=3.q=3.

Now, assume that nq21n^{\prime}\mid q^{2}-1, and select d=Rd=R, hh as the product of all the linear factors of xn1x^{n^{\prime}}-1, and H=h/(x1)H=h/(x-1) in Lemma 4.4. Then, for q=2q=2, there exists exactly one linear and at most 1 quadratic factor of xn1x^{n^{\prime}}-1 over 𝔽2\mathbb{F}_{2}. Therefore, in this case, 𝒟1/2\mathcal{D}\geq 1/2 and 𝒮4\mathcal{S}\leq 4. Similarly, for q=3q=3, there exists at most 2 linear and at most 2 quadratic factors of xn1x^{n^{\prime}}-1 over 𝔽3\mathbb{F}_{3}. Thus, in this case, 𝒟1/3\mathcal{D}\geq 1/3 and 𝒮17\mathcal{S}\leq 17. If we take ν=6\nu=6, then Inequality (17) holds for n33n\geq 33 in both the cases. For 14n<3314\leq n<33, we test Inequality (13), which does not hold for n=n= 16, when q=2q=2, and for n=n= 18, 24, when q=3.q=3.

Finally, for q=2,3q=2,3 and the remaining values of nn, we apply the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=3\vartheta=3, which returns False only for q=2q=2 and n=n= 14, 15, 16, 18, 20, 21, 24, 30, and for q=3q=3 and n=n= 14, 16. For these remaining qq and nn, we explicitly search for a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} using SageMath procedure DIRECT_SEARCH(q,n), and find that such a pair always exists. ∎

5.2 The cases 5n135\leq n\leq 13

Observe that, Inequality (13) is applicable for the cases, n=5n=5, 6, only if gcd(n,q)=1\mathrm{gcd}(n,q)=1. Therefore, first we settle the cases 7n137\leq n\leq 13 for all q2q\geq 2, and then the cases n=5n=5, 6 with the assumption that gcd(q,n)=1.\mathrm{gcd}(q,n)=1. Before moving ahead, first we state the following lemma, which is a direct consequence of Lemma 4.4.

Lemma 5.4.

Let dd be a divisor of qn1q^{n}-1 and n0n_{0} be a fixed positive integer such that all the prime divisors of (qn1)/d(q^{n}-1)/d, co-prime to dd, are of the form n0s+1n_{0}s+1 for some positive integer ss. Further, let p1,p2,,plp_{1},p_{2},\ldots,p_{l} be the first ll primes of the form n0s+1n_{0}s+1 such that the product Pl=i=1lpi(qn1)/dP_{l}=\prod_{i=1}^{l}p_{i}\leq(q^{n}-1)/d, and denote the sum of their reciprocals by Sl=i=1l1/piS_{l}=\sum_{i=1}^{l}{1}/{p_{i}}. If h=1h=1 and H=1H=1 are the divisors of xn1x^{n}-1 and (xn1)/(x1)(x^{n}-1)/(x-1), respectively, then 𝒟1Sl(2n1)/q\mathcal{D}\geq 1-S_{l}-(2n-1)/q and 𝒮(l+2n2)/𝒟+2.\mathcal{S}\leq(l+2n-2)/\mathcal{D}+2. Moreover, if 𝒟>0\mathcal{D}>0 and qn/2ϑ>2W(d)𝒮q^{n/2-\vartheta}>2W(d)\mathcal{S}, then there exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, where 𝒟\mathcal{D}, 𝒮\mathcal{S} and ϑ\vartheta are the same as defined in Lemma 4.4 and Inequality (13).

Lemma 5.5.

Let 7n137\leq n\leq 13 and qq be a prime power such that gcd(q,n)=1\mathrm{gcd}(q,n)=1. Then, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}.

Proof.

Since, gcd(q,n)=1\mathrm{gcd}(q,n)=1, Inequality (13) holds, if

qn/22>𝒞νqn/ν22n.q^{n/2-2}>\mathcal{C}_{\nu}q^{n/\nu}2^{2n}. (18)

For each 7n137\leq n\leq 13 and suitable choices of ν\nu, the above inequality holds for qMnq\geq M_{n}, where the values of MnM_{n} are listed in Table 4.

Table 4: Values of MnM_{n} such that Inequality (18) holds for qMnq\geq M_{n}.
MnM_{n} 3.60×10123.60\times 10^{12} 2.10×1082.10\times 10^{8} 2.60×1062.60\times 10^{6} 222280222280 4689246892 1604916049 73317331
nn 7 8 9 10 11 12 13
ν\nu 7.8 7.4 7.2 7.1 7.1 7.1 7.1

Now, for each 10n13,10\leq n\leq 13, and q<Mnq<M_{n}, by calculating the exact values of W(xn1)W(x^{n}-1), W(R)W(R) and W(G)W(G), and testing Inequality (13) with ϑ=2\vartheta=2, we obtain that, there are 136136 pairs of (q,n)(q,n), which fail to satisfy this inequality. For these pairs, we use the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=2\vartheta=2, which returns False for the pairs (3,10)(3,10), (11,10)(11,10), (2,11)(2,11), (3,11)(3,11), (5,12)(5,12), (7,12)(7,12), and (13,12)(13,12). Finally, for these pairs, we explicitly search for a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} using SageMath procedure DIRECT_SEARCH(q,n), and find that such a pair always exists.

Now, let n=7n=7, then for q<M7q<M_{7}, we use Lemma 5.4 with d=q1d=q-1, h=1h=1, H=1H=1 and ϑ=2.\vartheta=2. In this case, the primes pp dividing (q71)/d(q^{7}-1)/d, co-prime to dd, are of the form 7s+17s+1, and Pli=06M7iP_{l}\leq\sum_{i=0}^{6}M_{7}^{i} for l29l\leq 29. Further, if q65q\geq 65, then 𝒟>0.6608\mathcal{D}>0.6608 and 𝒮64.042\mathcal{S}\leq 64.042, and there exists a desired pair (α,α1)(\alpha,\alpha^{-1}), if q3/2>2𝒞νq1/ν𝒮q^{3/2}>2\mathcal{C}_{\nu}q^{1/\nu}\mathcal{S}, which is true for q149q\geq 149 and ν=2.8.\nu=2.8. Proceeding in this way for 65q<14965\leq q<149, we get that, the inequality q3/2>2𝒞νq1/ν𝒮q^{3/2}>2\mathcal{C}_{\nu}q^{1/\nu}\mathcal{S} holds for all 68q149.68\leq q\leq 149. Now, for q<68q<68, we use the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=2\vartheta=2, which returns False for q=q= 2, 3, 4, 5, 8, 9, 11, 13, and n=n= 7. In the end, for these pairs, we explicitly search for a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} using SageMath procedure DIRECT_SEARCH(q,n), and find that such a pair always exists.

Finally, for the cases n=8n=8, 9, and q<Mnq<M_{n}, again we use Lemma 5.4 with d=q41d=q^{4}-1, q31q^{3}-1, respectively. In these cases, the primes pp dividing (qn1)/d(q^{n}-1)/d, co-prime to dd, are of the form 8s+18s+1, when n=8n=8, and of the form 9s+19s+1, when n=9n=9. Now, following the same line as in the case n=7n=7, we have reduced the bounds on qq from M8M_{8} to 203, and from M9M_{9} to 41.Now, for q<68q<68, we use the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=2\vartheta=2, which returns False for q=q= 3, 5, 7, 9, 11, 13, 17 and n=n= 8, and q=q= 2, 4, 7, 19, and n=9n=9. Finally, for these pairs, we explicitly search for a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} using SageMath procedure DIRECT_SEARCH(q,n), and find that such a pair always exists. ∎

Lemma 5.6.

Let 7n137\leq n\leq 13 and qq be a prime power such that gcd(q,n)1\mathrm{gcd}(q,n)\neq 1. Then, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}.

Proof.

Since, gcd(q,n)1\mathrm{gcd}(q,n)\neq 1, q=ptq=p^{t}, for each prime pp dividing nn and some positive integer tt, and Inequality (13) holds, if

pnt/23>𝒞νpnt/ν22np^{nt/2-3}>\mathcal{C}_{\nu}p^{nt/\nu}2^{2n} (19)

For each 7n137\leq n\leq 13, the above inequality holds for ttn,pt\geq t_{n,p} and the suitable choices of ν\nu, listed in Table 5.

Table 5: Values of tn,pt_{n,p} such that Inequality (19) holds for ttn,pt\geq t_{n,p}.
tn,pt_{n,p} 37973797 223223 4646 42,1942,19 99 24,1524,15 66
(n,p)(n,p) (7,7)(7,7) (8,2)(8,2) (9,3)(9,3) (10,2),(10,5)(10,2),(10,5) (11,11)(11,11) (12,2),(12,3)(12,2),(12,3) (13,13)(13,13)
ν\nu 15.7 10.3 9 8 7.5 7.5 7

For the cases 9n139\leq n\leq 13, and t<tn,pt<t_{n,p}, by calculating the exact values of W(xn1)W(x^{n}-1), W(R)W(R) and W(G)W(G), and testing Inequality (13) with ϑ=3\vartheta=3, we obtain that, there are 2525 pairs of (q,n)(q,n), which fail to satisfy this inequality. For these pairs, we use the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=3\vartheta=3, which returns False for the pairs (3,9)(3,9), (9,9)(9,9), (2,10)(2,10), (4,10)(4,10), (16,10)(16,10), (5,10)(5,10), (2,12)(2,12), (4,12)(4,12), (3,12)(3,12) and (9,12)(9,12). Finally, for these pairs, we explicitly search for a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} using SageMath procedure DIRECT_SEARCH(q,n), and find that such a pair always exists.

Now, let n=7n=7, q=7tq=7^{t} and t<t7,7t<t_{7,7}. Clearly, x71x^{7}-1 has only one linear factor over 𝔽q\mathbb{F}_{q} and 7q717\nmid q^{7}-1, therefore, using Lemma 4.4 with d=q1d=q-1, h=1h=1 and H=1H=1, we get that there exists a pair (α,α1)(\alpha,\alpha^{-1}) of desired properties, if 7t/2>2𝒞ν7t/ν𝒮7^{t/2}>2\mathcal{C}_{\nu}7^{t/\nu}\mathcal{S}. Since, the primes dividing (q71)/d(q^{7}-1)/d are of the form 7s+17s+1, so let PlP_{l} and SlS_{l} be the product and sum of reciprocals of first ll primes of the form 7s+17s+1 respectively, such that Pli=067itP_{l}\leq\sum_{i=0}^{6}7^{it}, then l3891l\leq 3891, and from Lemma 4.4, 𝒟1Sl1/7t\mathcal{D}\geq 1-S_{l}-1/7^{t} and 𝒮l/D+2.\mathcal{S}\leq l/D+2. Now, For t8t\geq 8, we get 𝒟>0.7661\mathcal{D}>0.7661 and 𝒮5080.82\mathcal{S}\leq 5080.82, and inequality 7t/2>2𝒞ν7t/ν𝒮7^{t/2}>2\mathcal{C}_{\nu}7^{t/\nu}\mathcal{S} holds for t20t\geq 20 and ν=5\nu=5. Proceeding in this way for 8t<208\leq t<20, we get that, the inequality 7t/2>2𝒞ν7t/ν𝒮7^{t/2}>2\mathcal{C}_{\nu}7^{t/\nu}\mathcal{S} always holds for t11.t\geq 11. In the end, for 1t101\leq t\leq 10, we first perform the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=3\vartheta=3, which returns False only for t=t= 1, 2, 3, 4, and then for these values, we explicitly search for a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} using SageMath procedure DIRECT_SEARCH(q,n), and find that such a pair always exists.

Now, in the case n=8n=8, we have q=2tq=2^{t} and x81=(x1)8x^{8}-1=(x-1)^{8} over 𝔽q.\mathbb{F}_{q}. If d=q41d=q^{4}-1 and pp is a prime divisor of q4+1q^{4}+1 co-prime to dd, then q41(modp)q^{4}\not\equiv 1(\mathrm{mod}\hskip 1.42262ptp) and q41(modp)q^{4}\equiv-1(\mathrm{mod}\hskip 1.42262ptp), which implies that, the multiplicative order of qq modulo pp is 88. Thus, pp is of the form 8s+1.8s+1. Again, we use Lemma 4.4 with d=q41d=q^{4}-1, h=H=1h=H=1. Let PlP_{l} and SlS_{l} be the product and sum of reciprocals of first ll primes of the form 8s+18s+1 respectively, such that Pl24t+1P_{l}\leq 2^{4t}+1, then l90l\leq 90 for t<t8,2t<t_{8,2}, and from Lemma 4.4, 𝒟1Sl1/2t\mathcal{D}\geq 1-S_{l}-1/2^{t} and 𝒮l/D+2.\mathcal{S}\leq l/D+2. Now, For t8t\geq 8, we get 𝒟>0.764\mathcal{D}>0.764 and 𝒮119.75\mathcal{S}\leq 119.75, and inequality 2t>2𝒞ν24t/ν𝒮2^{t}>2\mathcal{C}_{\nu}2^{4t/\nu}\mathcal{S} holds for t35t\geq 35 and ν=7\nu=7. Proceeding in this way for 8t<358\leq t<35, we get that, the inequality 2t>2𝒞νq4t/ν𝒮2^{t}>2\mathcal{C}_{\nu}q^{4t/\nu}\mathcal{S} always holds for t30.t\geq 30. In the end, for 1t291\leq t\leq 29, we obtain that, there are 14 values of tt (1t12;t=151\leq t\leq 12;\ t=15, 18) which fails to satisfy Inequality (13). For these values, we use the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=3\vartheta=3, which returns False for 1t61\leq t\leq 6. Finally, we explicitly search for a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} using SageMath procedure DIRECT_SEARCH(q,n), and find that such a pair always exists. ∎

Lemma 5.7.

Let n=n= 55, 66, and qq be a prime power such that gcd(q,n)=1\mathrm{gcd}(q,n)=1. Then, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} with the sole genuine exception (4,5)(4,5).

Proof.

First, we test Inequality (18) in both the cases n=n= 5, 6, and get that, this holds for qMnq\geq M_{n}, where M5=6.97×10323M_{5}=6.97\times 10^{323}, if we set ν=11.9\nu=11.9, and M6=4.74×1027M_{6}=4.74\times 10^{27}, if we set ν=8.7.\nu=8.7. Now, for q<Mnq<M_{n}, in each case, we use Lemma 5.4 with d=q1d=q-1 for n=5n=5, and d=q21d=q^{2}-1 for n=6.n=6. Notice that, primes dividing (qn1)/d(q^{n}-1)/d, co-prime to dd, are of the form 5s+15s+1 for n=5n=5, and 3s+13s+1 for n=6.n=6. Then, for n=5n=5 and q105q\geq 10^{5}, we get l360l\leq 360, 𝒟>0.648\mathcal{D}>0.648, 𝒮569.9\mathcal{S}\leq 569.9, and for n=6n=6 and q104q\geq 10^{4}, we get l48l\leq 48, 𝒟>0.429\mathcal{D}>0.429, 𝒮137.11.\mathcal{S}\leq 137.11. With these, we test the inequalities q1/2>2𝒞νq1/ν𝒮q^{1/2}>2\mathcal{C}_{\nu}q^{1/\nu}\mathcal{S} for n=5;ν=5.5n=5;\ \nu=5.5, and q>2𝒞νq2/ν𝒮q>2\mathcal{C}_{\nu}q^{2/\nu}\mathcal{S} for n=6;ν=5.2n=6;\ \nu=5.2, and get that, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of desired properties for q4.395×1013q\geq 4.395\times 10^{13} and n=5n=5, and q646882q\geq 646882 and n=6.n=6. By repeating the process for remaining values of qq in each case, we significantly reduce the bound on qq from M5M_{5} to 4.96×1094.96\times 10^{9}, and from M6M_{6} to 94531.94531.

Now, for n=5;n=5; 66 and q<4.96×109;q<4.96\times 10^{9}; 9453194531, respectively, we use Lemma 5.4 with d=gcd(235,qn1)d=\mathrm{gcd}(2\cdot 3\cdot 5,q^{n}-1). Notice that, the primes dividing (qn1)/d(q^{n}-1)/d co-prime to dd are of the form 2s+12s+1 in each case. Then, for 105q<4.96×10910^{5}\leq q<4.96\times 10^{9} and n=5n=5, we get l28l\leq 28, 𝒟>0.175\mathcal{D}>0.175, 𝒮207.05\mathcal{S}\leq 207.05, and the inequality q1/2>24𝒮q^{1/2}>2^{4}\mathcal{S} holds for q10973928q\geq 10973928. Similarly, for 103q<9453110^{3}\leq q<94531 and n=6n=6, we get l19l\leq 19, 𝒟>0.253\mathcal{D}>0.253, 𝒮116.58\mathcal{S}\leq 116.58, and the inequality q>24𝒮q>2^{4}\mathcal{S} holds for q1866q\geq 1866. Proceeding in this way, we get that, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of desired properties for q3839356q\geq 3839356 and n=5n=5, and q1109q\geq 1109 and n=6n=6. For the remaining values of qq in each case, we run the SageMath procedure TEST_SIEVE(q,n,ϑ\vartheta) with ϑ=2\vartheta=2 and get that, there are 871871 values of qq (listed in Table LABEL:Table6) in the case of n=5n=5, and 4747 values of qq (listed in Table LABEL:Table6) in the case of n=6n=6, for which this procedure returns False. Finally, we perform the procedure DIRECT_SEARCH(q,n) for the remaining values of tt, and find that, there always exists the desired pairs (α,α1)(\alpha,\alpha^{-1}) with the sole genuine exception (4,5).(4,5).

Summing up Lemmas 5.1, 5.3, 5.5, 5.6, and 5.7, we conclude our Theorem 1.3.

5.3 A note on the cases n=1,2,3,4n=1,2,3,4

From the definition of kk-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, it is clear that 0kn10\leq k\leq n-1, therefore, it is sensible to talk about the existence of a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} for n>1n>1. Now, for the case n=2n=2, if α\alpha is a 11-normal element in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, then either αqα=0\alpha^{q}-\alpha=0 or αq+α=0\alpha^{q}+\alpha=0, which implies ord(α)2(q1)<q21\mathrm{ord}(\alpha)\leq 2(q-1)<q^{2}-1. Thus, there does not exist any pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽q2\mathbb{F}_{q^{2}} over 𝔽q.\mathbb{F}_{q}. Further, we show that, such a pair does not exists in the case of n=3n=3 also.

Lemma 5.8.

Let n=3n=3 and q2q\geq 2 be any prime power. Then, there does not exist any pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q.\mathbb{F}_{q}.

Proof.

Suppose that, there exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽q3\mathbb{F}_{q^{3}} over 𝔽q\mathbb{F}_{q}. Since, x31=(x1)3x^{3}-1=(x-1)^{3} or (x1)(x2+x+1)(x-1)(x^{2}+x+1) or (x1)(xa)(xa1)(x-1)(x-a)(x-a^{-1}) for some a𝔽q{0,1}a\in\mathbb{F}_{q}\setminus\{0,1\}. Therefore, we have the following cases on the 𝔽q\mathbb{F}_{q}-orders of α\alpha and α1\alpha^{-1}.

Case 1. If Ordq(α)=Ordq(α1)=x2+x+1\mathrm{Ord}_{q}(\alpha)=\mathrm{Ord}_{q}(\alpha^{-1})=x^{2}+x+1, then, the trace of α\alpha and α1\alpha^{-1} over 𝔽q\mathbb{F}_{q} is equal to zero. Consequently, the primitive polynomial of α\alpha will be of the form x3+ax^{3}+a for some a𝔽q,a\in\mathbb{F}_{q}, and hence α3𝔽q.\alpha^{3}\in\mathbb{F}_{q}. This means q313(q1)q^{3}-1\mid 3(q-1), which is never possible for any q>1q>1.

Case 2. If (x1)gcd(Ordq(α),Ordq(α1))(x-1)\mid\mathrm{gcd}(\mathrm{Ord}_{q}(\alpha),\mathrm{Ord}_{q}(\alpha^{-1})), then we get that, (αqα)q1(\alpha^{q}-\alpha)^{q-1} and (αqα1)q1(\alpha^{-q}-\alpha^{-1})^{q-1} both belong to 𝔽q.\mathbb{F}_{q}. This implies αq21𝔽q\alpha^{q^{2}-1}\in\mathbb{F}_{q}, i.e. q31(q21)(q1)q^{3}-1\mid(q^{2}-1)(q-1), which is not possible.

Case 3. If Ordq(α)=(x1)(xa)\mathrm{Ord}_{q}(\alpha)=(x-1)(x-a) and Ordq(α1)=(xa)(xa1)\mathrm{Ord}_{q}(\alpha^{-1})=(x-a)(x-a^{-1}), then we get that, (αqaα)q1=1(\alpha^{q}-a\alpha)^{q-1}=1 and (αqa1α1)q1𝔽q.(\alpha^{-q}-a^{-1}\alpha^{-1})^{q-1}\in\mathbb{F}_{q}. This implies αq21𝔽q\alpha^{q^{2}-1}\in\mathbb{F}_{q}, i.e. q31(q21)(q1)q^{3}-1\mid(q^{2}-1)(q-1), which is not possible.

Case 4. If Ordq(α)=(x1)(xa1)\mathrm{Ord}_{q}(\alpha)=(x-1)(x-a^{-1}) and Ordq(α1)=(xa)(xa)\mathrm{Ord}_{q}(\alpha^{-1})=(x-a)(x-a),then by interchanging the role of α\alpha and α1\alpha^{-1} in Case 3, we get that, q31(q21)(q1)q^{3}-1\mid(q^{2}-1)(q-1), which is not possible.

This completes the proof of the lemma. ∎

We also provide a necessary condition for the existence of a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽q4\mathbb{F}_{q^{4}} over 𝔽q.\mathbb{F}_{q}.

Lemma 5.9.

If there exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽q4\mathbb{F}_{q^{4}} over 𝔽q\mathbb{F}_{q}, then q1(mod).q\equiv 1\hskip 1.42262pt(\mathrm{mod}\hskip 1.42262pt).

Proof.

Let’s assume that q1(mod)q\not\equiv 1\hskip 1.42262pt(\mathrm{mod}\hskip 1.42262pt), then the polynomial x41x^{4}-1 factorizes either as (x1)4(x-1)^{4}, or as (x1)(x+1)(x2+1)(x-1)(x+1)(x^{2}+1) over 𝔽q.\mathbb{F}_{q}. Further, let (α,α1)(\alpha,\alpha^{-1}) be a pair of 11-normal elements in 𝔽q4\mathbb{F}_{q^{4}} over 𝔽q\mathbb{F}_{q}. Then, we have the following cases.

Case 1: (x1)gcd(Ordq(α),Ordq(α1)),(x-1)\mid\mathrm{gcd}(\mathrm{Ord}_{q}(\alpha),\mathrm{Ord}_{q}(\alpha^{-1})), then αq2α\alpha^{q^{2}}-\alpha and αq2α1\alpha^{-q^{2}}-\alpha^{-1} both will belong to 𝔽q\mathbb{F}_{q}. This implies αq2+1𝔽q\alpha^{q^{2}+1}\in\mathbb{F}_{q}, i.e. ord(α)(q2+1)(q1)<q41.\mathrm{ord}(\alpha)\leq(q^{2}+1)(q-1)<q^{4}-1.

Case 2: (x+1)gcd(Ordq(α),Ordq(α1)),(x+1)\mid\mathrm{gcd}(\mathrm{Ord}_{q}(\alpha),\mathrm{Ord}_{q}(\alpha^{-1})), then (αq2α)2(\alpha^{q^{2}}-\alpha)^{2} and (αq2α1)2(\alpha^{-q^{2}}-\alpha^{-1})^{2} both will belong to 𝔽q\mathbb{F}_{q}. This implies α2(q2+1)𝔽q\alpha^{2(q^{2}+1)}\in\mathbb{F}_{q}, i.e. ord(α)2(q2+1)(q1)<q41.\mathrm{ord}(\alpha)\leq 2(q^{2}+1)(q-1)<q^{4}-1.

Case 3: gcd(Ordq(α),Ordq(α1))=x2+1\mathrm{gcd}(\mathrm{Ord}_{q}(\alpha),\mathrm{Ord}_{q}(\alpha^{-1}))=x^{2}+1, then either αq2α\alpha^{q^{2}}-\alpha and (αq2α1)2(\alpha^{-q^{2}}-\alpha^{-1})^{2} belong to 𝔽q\mathbb{F}_{q}, or (αq2α)2(\alpha^{q^{2}}-\alpha)^{2} and αq2α1\alpha^{-q^{2}}-\alpha^{-1} will belong to 𝔽q\mathbb{F}_{q} simultaneously. In either case, we get that ord(α)2(q2+1)(q1)<q41.\mathrm{ord}(\alpha)\leq 2(q^{2}+1)(q-1)<q^{4}-1.

In each case, we get that ord(α)<q41\mathrm{ord}(\alpha)<q^{4}-1, which implies α\alpha is never a primitive element in 𝔽q.\mathbb{F}_{q}. Thus, a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements in 𝔽q4\mathbb{F}_{q^{4}} over 𝔽q\mathbb{F}_{q} can exist, only if q1(mod4).q\equiv 1\hskip 1.42262pt(\mathrm{mod}\hskip 1.42262pt4).

For the cases, n=4n=4 and q=4t+1q=4t+1; n=5n=5 and q=5tq=5^{t}; and n=6n=6 and q=2t,3tq=2^{t},3^{t}, where tt is a positive integer, we performed the extensive experiments, and obtained that, there always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} unless n=6n=6 and q=2,4q=2,4. On the basis of these experiments, we conjecture the following.

Conjecture 5.1.

Let tt be a positive integer. There always exists a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q} if n=4n=4 and q=4t+1;q=4t+1; n=5n=5 and q=5t;q=5^{t}; and n=6n=6 and q=2t,3tq=2^{t},3^{t} with the sole genuine exceptions (2,6)(2,6), (4,6)(4,6).

6 Acknowledgements

This work was funded by Council of Scientific &\& Industrial Research, under Grant F. No. 09/045(1674)/2019-EMR-I and University Grant Commission, under Grant Ref. No. 1042/CSIR-UGC NET DEC-2018.

References

  • Aguirre and Neumann [2021] Aguirre, J.J., Neumann, V.G., 2021. Existence of primitive 2-normal elements in finite fields. Finite Fields and Their Applications 73, 101864.
  • Blum and Micali [1984] Blum, M., Micali, S., 1984. How to generate cryptographically strong sequences of pseudorandom bits. SIAM journal on Computing 13, 850–864.
  • Cohen and Huczynska [2010] Cohen, S.D., Huczynska, S., 2010. The strong primitive normal basis theorem. Acta Arith. 143, 299–332. URL: https://doi.org/10.4064/aa143-4-1, doi:10.4064/aa143-4-1.
  • Cohen and Kapetanakis [2021] Cohen, S.D., Kapetanakis, G., 2021. Finite field extensions with the line or translate property for-primitive elements. Journal of the Australian Mathematical Society 111, 313–319.
  • Diffie and Hellman [1976] Diffie, W., Hellman, M., 1976. New directions in cryptography. IEEE transactions on Information Theory 22, 644–654.
  • Fu and Wan [2014] Fu, L., Wan, D., 2014. A class of incomplete character sums. Quarterly Journal of Mathematics 65, 1195–1211.
  • Gao [1999] Gao, S., 1999. Elements of provable high orders in finite fields. Proceedings of the American Mathematical Society 127, 1615–1623.
  • Huczynska et al. [2013] Huczynska, S., Mullen, G.L., Panario, D., Thomson, D., 2013. Existence and properties of k-normal elements over finite fields. Finite Fields and Their Applications 24, 170–183.
  • Kapetanakis and Reis [2019] Kapetanakis, G., Reis, L., 2019. Variations of the primitive normal basis theorem. Designs, Codes and Cryptography 87, 1459–1480.
  • Lenstra and Schoof [1987] Lenstra, Jr., H.W., Schoof, R.J., 1987. Primitive normal bases for finite fields. Math. Comp. 48, 217–231. URL: https://doi.org/10.2307/2007886, doi:10.2307/2007886.
  • Lidl and Niederreiter [1997] Lidl, R., Niederreiter, H., 1997. Finite fields. volume 20. Second ed., Cambridge University Press, Cambridge.
  • Meletiou and Mullen [1992] Meletiou, G., Mullen, G.L., 1992. A note on discrete logarithms in finite fields. Applicable Algebra in Engineering, Communication and Computing 3, 75–78.
  • Mullen [2016] Mullen, G.L., 2016. Some open problems arising from my recent finite field research, in: Contemporary developments in finite fields and applications. World Scientific, pp. 254–269.
  • Rani et al. [2021a] Rani, M., Sharma, A.K., Tiwari, S.K., 2021a. On rr-primitive kk-normal elements over finite fields. Communicated .
  • Rani et al. [2021b] Rani, M., Sharma, A.K., Tiwari, S.K., Gupta, I., 2021b. On the existence of pairs of primitive normal elements over finite fields. São Paulo Journal of Mathematical Sciences , 1–18.
  • Reis [2019] Reis, L., 2019. Existence results on kk-normal elements over finite fields. Revista Matemática Iberoamericana 35, 805–822.
  • Reis and Thomson [2018] Reis, L., Thomson, D., 2018. Existence of primitive 1-normal elements in finite fields. Finite Fields and Their Applications 51, 238–269.
  • Sharma et al. [2022] Sharma, A.K., Rani, M., Tiwari, S.K., 2022. Primitive normal pairs with prescribed norm and trace. Finite Fields and Their Applications 78, 101976.
  • The Sage Developers [2020] The Sage Developers, 2020. SageMath, the Sage Mathematics Software System (Version 9.0). URL: https://www.sagemath.org.
  • Tian and Qi [2006] Tian, T., Qi, W.F., 2006. Primitive normal element and its inverse in finite fields. Acta Math. Sinica (Chin. Ser.) 49, 657–668.

Appendix A

Table 6: Pairs (q,n)(q,n) for which TEST_SIEVE(q,n) returns False in the cases n=n= 5, 6.
nn qq
5 2, 3, 4, 7, 8, 9, 11, 13, 16, 17, 19, 23, 27, 29, 31, 32, 37, 41, 43, 47, 49, 53, 59, 61, 64, 67, 71, 73, 79, 81, 83, 89, 97, 101, 103, 107, 109, 113, 121, 127, 128, 131, 137, 139, 149, 151, 157, 163, 167, 169, 173, 179, 181, 191, 193, 197, 199, 211, 223, 227, 229, 233, 239, 241, 243, 251, 256, 257, 263, 269, 271, 277, 281, 283, 289, 293, 307, 311, 313, 317, 331, 337, 343, 347, 349, 353, 359, 361, 367, 373, 379, 383, 389, 397, 401, 409, 419, 421, 431, 433, 439, 443, 449, 457, 461, 463, 467, 479, 487, 491, 499, 503, 509, 512, 521, 523, 529, 541, 547, 557, 563, 569, 571, 577, 587, 593, 599, 601, 607, 613, 617, 619, 631, 641, 643, 647, 653, 659, 661, 673, 677, 683, 691, 701, 709, 719, 727, 729, 733, 739, 743, 751, 757, 761, 769, 773, 787, 797, 809, 811, 821, 823, 827, 829, 839, 841, 853, 857, 859, 863, 877, 881, 883, 887, 907, 911, 919, 929, 937, 941, 947, 953, 961, 967, 971, 977, 991, 997, 1009, 1013, 1019, 1021, 1024, 1031, 1033, 1039, 1049, 1051, 1061, 1063, 1069, 1087, 1091, 1093, 1103, 1109, 1117, 1123, 1129, 1151, 1153, 1163, 1171, 1181, 1193, 1201, 1213, 1229, 1231, 1237, 1249, 1259, 1279, 1289, 1291, 1301, 1303, 1321, 1327, 1331, 1361, 1367, 1369, 1373, 1381, 1399, 1409, 1423, 1429, 1439, 1447, 1451, 1453, 1459, 1471, 1481, 1483, 1489, 1499, 1511, 1531, 1543, 1549, 1559, 1567, 1571, 1579, 1583, 1597, 1601, 1609, 1619, 1621, 1627, 1667, 1669, 1681, 1693, 1697, 1699, 1709, 1721, 1723, 1741, 1747, 1753, 1759, 1777, 1783, 1789, 1801, 1811, 1831, 1849, 1861, 1867, 1871, 1873, 1879, 1901, 1931, 1933, 1949, 1951, 1993, 1999, 2003, 2011, 2017, 2029, 2053, 2081, 2083, 2087, 2089, 2111, 2113, 2129, 2131, 2137, 2141, 2143, 2161, 2179, 2197, 2203, 2209, 2221, 2239, 2251, 2269, 2281, 2293, 2297, 2311, 2341, 2347, 2351, 2371, 2377, 2381, 2389, 2401, 2411, 2437, 2441, 2467, 2473, 2503, 2521, 2531, 2539, 2551, 2557, 2591, 2593, 2621, 2647, 2659, 2671, 2683, 2689, 2699, 2707, 2711, 2713, 2719, 2731, 2741, 2749, 2767, 2791, 2797, 2801, 2803, 2809, 2851, 2857, 2861, 2887, 2927, 2953, 2971, 3001, 3011, 3019, 3037, 3041, 3049, 3061, 3067, 3079, 3109, 3121, 3169, 3181, 3191, 3217, 3221, 3251, 3259, 3271, 3301, 3319, 3331, 3361, 3371, 3391, 3433, 3457, 3461, 3463, 3469, 3481, 3491, 3499, 3511, 3529, 3541, 3547, 3571, 3581, 3613, 3631, 3671, 3691, 3697, 3701, 3721, 3727, 3733, 3739, 3761, 3821, 3823, 3851, 3853, 3877, 3881, 3907, 3911, 3919, 3931, 4001, 4003, 4019, 4021, 4051, 4057, 4096, 4111, 4129, 4159, 4201, 4211, 4219, 4229, 4231, 4241, 4243, 4261, 4271, 4327, 4391, 4421, 4441, 4447, 4451, 4481, 4489, 4519, 4561, 4591, 4603, 4621, 4651, 4691, 4751, 4759, 4789, 4801, 4831, 4861, 4871, 4909, 4931, 4951, 4957, 4999, 5011, 5021, 5041, 5101, 5119, 5167, 5171, 5179, 5209, 5261, 5281, 5329, 5351, 5381, 5419, 5431, 5441, 5449, 5471, 5501, 5503, 5521, 5531, 5581, 5591, 5641, 5659, 5701, 5741, 5743, 5779, 5791, 5801, 5821, 5839, 5851, 5861, 5881, 5923, 5981, 6007, 6011, 6043, 6091, 6121, 6131, 6151, 6163, 6211, 6221, 6241, 6271, 6301, 6361, 6421, 6427, 6451, 6469, 6481, 6521, 6561, 6571, 6581, 6637, 6661, 6679, 6691, 6761, 6763, 6781, 6791, 6841, 6871, 6889, 6961, 6991, 7001, 7039, 7121, 7151, 7177, 7309, 7321, 7351, 7411, 7417, 7451, 7481, 7541, 7549, 7561, 7591, 7621, 7681, 7741, 7841, 7921, 7951, 8011, 8101, 8161, 8171, 8191, 8221, 8231, 8311, 8317, 8419, 8431, 8461, 8501, 8521, 8581, 8641, 8731, 8737, 8761, 8779, 8821, 8941, 8971, 9001, 9091, 9151, 9161, 9181, 9199, 9241, 9311, 9391, 9409, 9421, 9491, 9511, 9521, 9601, 9619, 9631, 9661, 9721, 9769, 9781, 9811, 9829, 9871, 9901, 9931, 10111, 10141, 10151, 10201, 10321, 10459, 10501, 10531, 10609, 10651, 10711, 10771, 10831, 10861, 10891, 11047, 11071, 11131, 11161, 11251, 11257, 11311, 11411, 11491, 11551, 11621, 11701, 11719, 11731, 11821, 11881, 11941, 11971, 12211, 12241, 12301, 12391, 12421, 12433, 12451, 12511, 12541, 12601, 12721, 12769, 12781, 12841, 13171, 13291, 13381, 13399, 13411, 13441, 13567, 13681, 13711, 13831, 13921, 14071, 14221, 14251, 14281, 14341, 14401, 14431, 14551, 14641, 14731, 14771, 14821, 14851, 15031, 15121, 15271, 15331, 15361, 15391, 15451, 15511, 15541, 15601, 15661, 15901, 15991, 16111, 16141, 16231, 16339, 16381, 16651, 16831, 16921, 17011, 17161, 17191, 17341, 17431, 17491, 17581, 17791, 17851, 17911, 18061, 18121, 18181, 18301, 18451, 18481, 18661, 18691, 19081, 19141, 19231, 19321, 19381, 19441, 19471, 19501, 19531, 19801, 19891, 20011, 20101, 20161, 20431, 20551, 20641, 20749, 21001, 21121, 21211, 21481, 21751, 21841, 21871, 21961, 22051, 22111, 22201, 22291, 22441, 22531, 22621, 22741, 22801, 22861, 23011, 23071, 23131, 23311, 23371, 23431, 23761, 23971, 24061, 24091, 24121, 24151, 24181, 24391, 24421, 24481, 24571, 24781, 24841, 25111, 25261, 25411, 25621, 25741, 26041, 26251, 26701, 26821, 26881, 27031, 27091, 27361, 27691, 27751, 27901, 28051, 28081, 28351, 28561, 28711, 28771, 29131, 29191, 29401, 29581, 29611, 29641, 29671, 29881, 30211, 30241, 30391, 30661, 30871, 31081, 31321, 31531, 31981, 32191, 32341, 32371, 32761, 32971, 33091, 33181, 33391, 33811, 34171, 34651, 35281, 35311, 35491, 36061, 36481, 36541, 37171, 37591, 37951, 38611, 38851, 39901, 40111, 40531, 41611, 42331, 42841, 43261, 43891, 44521, 46411, 47881, 51871, 53551, 57121, 60901, 62791, 63211, 65521, 70981
6 5, 7, 11, 13, 17, 19, 23, 25, 29, 31, 37, 41, 43, 47, 49, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 121, 127, 131, 137, 139, 149, 151, 157, 163, 169, 179, 181, 191, 199, 211, 229, 241.
Algorithm 1 Testing the sieving inequality for the existence of primitive 11-normal pair (α,α1)(\alpha,\alpha^{-1})
1:qq, nn, ϑ\vartheta
2:True, if qn/2ϑ>2W(h)W(d)W(H)𝒮q^{n/2-\vartheta}>2W(h)W(d)W(H)\mathcal{S} for some hxn1h\mid x^{n}-1, dRd\mid R and HGH\mid G otherwise, False.
3:procedure  test_sieve(q,n,ϑ\vartheta)
4:    Construct the set AA of divisors of rad(qn1)\mathrm{rad}(q^{n}-1) and set BB of divisors of rad(xn1)x1\frac{\mathrm{rad}(x^{n}-1)}{x-1}.
5:    for dd in AA do
6:       for HH in BB do
7:          h=Hh=H
8:          Construct the set L1L_{1} of prime divisors rad(qn1)d\frac{\mathrm{rad}(q^{n}-1)}{d}
9:          Construct the set L2L_{2} of irreducible factors of rad(xn1)H(x1)\frac{\mathrm{rad}(x^{n}-1)}{H(x-1)} and the set L3=L2{x1}L_{3}=L_{2}\cup\{x-1\}
10:          Compute 𝒟=1p1L11p1p2L21qdeg(p2)p3L31qdeg(p3)\mathcal{D}=1-\sum\limits_{p_{1}\in L_{1}}\frac{1}{p_{1}}-\sum\limits_{p_{2}\in L_{2}}\frac{1}{q^{\mathrm{deg}(p_{2})}}-\sum\limits_{p_{3}\in L_{3}}\frac{1}{q^{\mathrm{deg}(p_{3})}}
11:          if 𝒟>0\mathcal{D}>0 then
12:             compute 𝒮=|L1|+|L2|+|L3|1𝒟+2\mathcal{S}=\frac{\ |L_{1}|+|L_{2}|+|L_{3}|-1}{\mathcal{D}}+2
13:             if qn/2ϑ>2W(d)W(H)2𝒮q^{n/2-\vartheta}>2W(d)W(H)^{2}\mathcal{S} then\triangleright Sieving inequality
14:                return True
15:             end if
16:          end if
17:       end for
18:    end for
19:    return False
20:end procedure
Algorithm 2 Explicit search for a primitive 11-normal pair (α,α1)(\alpha,\alpha^{-1})
1:qq, nn
2:True, if a pair (α,α1)(\alpha,\alpha^{-1}) of primitive 11-normal elements exists in 𝔽qn\mathbb{F}_{q^{n}} over 𝔽q\mathbb{F}_{q}, otherwise, False.
3:procedure direct_search(q,n)
4:    for β\beta in 𝔽qn\mathbb{F}_{q^{n}} do
5:       α=βqβ\alpha=\beta^{q}-\beta
6:       if α0\alpha\neq 0 then
7:          construct mα(x)m_{\alpha}(x) and mα1(x)m_{\alpha^{-1}}(x) \triangleright mα(x)=i=0n1αqixni1m_{\alpha}(x)=\sum_{i=0}^{n-1}\alpha^{q^{i}}x^{n-i-1}
8:          if deg(gcd(mα(x),xn1))=1\mathrm{deg}(\mathrm{gcd}(m_{\alpha}(x),x^{n}-1))=1 & deg(gcd(mα1(x),xn1))=1\mathrm{deg}(\mathrm{gcd}(m_{\alpha^{-1}}(x),x^{n}-1))=1 & ord(α)=qn1\mathrm{ord}(\alpha)=q^{n}-1 then
9:             return True
10:          end if
11:       end if
12:    end for
13:    return False
14:end procedure