Hostname: page-component-586b7cd67f-gb8f7 Total loading time: 0 Render date: 2024-11-24T19:49:18.253Z Has data issue: false hasContentIssue false

ON GENERALISED LEGENDRE MATRICES INVOLVING ROOTS OF UNITY OVER FINITE FIELDS

Published online by Cambridge University Press:  22 April 2024

NING-LIU WEI
Affiliation:
School of Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, Jiangsu Province, PR China e-mail: [email protected]
YU-BO LI
Affiliation:
School of Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, Jiangsu Province, PR China e-mail: [email protected]
HAI-LIANG WU*
Affiliation:
School of Science, Nanjing University of Posts and Telecommunications, Nanjing 210023, Jiangsu Province, PR China
Rights & Permissions [Opens in a new window]

Abstract

Motivated by the work initiated by Chapman [‘Determinants of Legendre symbol matrices’, Acta Arith. 115 (2004), 231–244], we investigate some arithmetical properties of generalised Legendre matrices over finite fields. For example, letting $a_1,\ldots ,a_{(q-1)/2}$ be all the nonzero squares in the finite field $\mathbb {F}_q$ containing q elements with $2\nmid q$, we give the explicit value of the determinant $D_{(q-1)/2}=\det [(a_i+a_j)^{(q-3)/2}]_{1\le i,j\le (q-1)/2}$. In particular, if $q=p$ is a prime greater than $3$, then

$$ \begin{align*}\bigg(\frac{\det D_{(p-1)/2}}{p}\bigg)= \begin{cases} 1 & \mbox{if}\ p\equiv1\pmod4,\\ (-1)^{(h(-p)+1)/2} & \mbox{if}\ p\equiv 3\pmod4\ \text{and}\ p>3, \end{cases}\end{align*} $$

where $(\frac {\cdot }{p})$ is the Legendre symbol and $h(-p)$ is the class number of $\mathbb {Q}(\sqrt {-p})$.

Type
Research Article
Copyright
© The Author(s), 2024. Published by Cambridge University Press on behalf of Australian Mathematical Publishing Association Inc.

1 Introduction

1.1 Related work and motivations

Let p be an odd prime and let $(\frac {\cdot }{p})$ be the Legendre symbol. Chapman [Reference Chapman1, Reference Chapman2] investigated determinants involving Legendre matrices

$$ \begin{align*}C_1=\bigg[\bigg(\frac{i+j-1}{p}\bigg)\bigg]_{1\le i,j\le (p-1)/2}\end{align*} $$

and

$$ \begin{align*}C_2=\bigg[\bigg(\frac{i+j-1}{p}\bigg)\bigg]_{1\le i,j\le (p+1)/2}.\end{align*} $$

Surprisingly, these determinants are closely related to quadratic fields. In fact, letting $\varepsilon _p>1$ and $h(p)$ be the fundamental unit and the class number of $\mathbb {Q}(\sqrt {p})$ , and writing $\varepsilon _p=a_p+b_p\sqrt {p}$ with $a_b,b_p\in \mathbb {Q}$ , Chapman [Reference Chapman1] proved that

$$ \begin{align*}\det C_1= \begin{cases} (-1)^{(p-1)/4}2^{(p-1)/2}b_p & \mbox{if}\ p\equiv1\pmod4,\\ 0 & \mbox{otherwise,} \end{cases}\end{align*} $$

and

$$ \begin{align*}\det C_2= \begin{cases} (-1)^{(p+3)/4}2^{(p-1)/2}a_p & \mbox{if}\ p\equiv1\pmod4,\\ -2^{(p-1)/2} & \mbox{otherwise.} \end{cases}\end{align*} $$

Later, Chapman [Reference Chapman2] posed the following conjecture.

Conjecture 1.1 (Chapman).

Let p be an odd prime and write $\varepsilon _p^{(2-(2/p))h(p)}= a_p^{\prime } +b_p^{\prime }\sqrt {p}$ with $a_p^{\prime },b_p^{\prime }\in \mathbb {Q}$ . Then

$$ \begin{align*}\det \bigg[\bigg(\frac{j-i}{p}\bigg)\bigg]_{1\le i,j\le (p+1)/2} =\begin{cases} -a_p^{\prime} & \mbox{if}\ p\equiv1\pmod 4,\\ 1 & \mbox{otherwise.} \end{cases}\end{align*} $$

Due to the difficulty of the conjecture, Chapman called this determinant ‘the evil determinant’. In 2012 and 2013, Vsemirnov [Reference Vsemirnov9, Reference Vsemirnov10] confirmed the conjecture (the case $p\equiv 3\pmod 4$ in [Reference Vsemirnov9] and the case $p\equiv 1\pmod 4$ in [Reference Vsemirnov10]).

In 2019, Sun [Reference Sun8] studied some variants of Chapman’s determinants. For example, let

$$ \begin{align*} S(d,p)=\det\bigg[\bigg(\frac{i^2+dj^2}{p}\bigg)\bigg]_{1\le i,j\le (p-1)/2}.\end{align*} $$

Sun [Reference Sun8, Theorem 1.2] showed that $S(d,p)=0$ whenever $(d/p)=-1$ and that $(-S(d,p)/p)=1$ whenever $(d/p)=1$ . (See [Reference Krachun, Petrov, Sun and Vsemirnov3, Reference Li and Wei5, Reference Wu11, Reference Wu, She and Wang13] for recent progress on this topic.) Also, Sun [Reference Sun8, Theorem 1.4] proved that

(1.1) $$ \begin{align} \det\bigg[\frac{((i+j)/p)}{i+j}\bigg]_{1\le i,j\le (p-1)/2} \equiv\begin{cases} (2/p)\pmod p & \mbox{if}\ p\equiv1\pmod4,\\ ((p-1)/2)!\pmod p & \mbox{otherwise,} \end{cases} \end{align} $$

and that

$$ \begin{align*} \det\bigg[\frac{1}{i^2+j^2}\bigg]_{1\le i,j\le (p-1)/2}\equiv (-1)^{{(p+1)}/{4}}\pmod p \end{align*} $$

whenever $p\equiv 3\pmod 4$ . In 2022, the third author and Wang [Reference Wu and Wang14, Theorem 1.7] considered the determinant $\det [1/(\alpha _i+\alpha _j)]_{1\le i,j\le (p-1)/k}$ , where $0<\alpha _1,\ldots ,\alpha _{(p-1)/k}<p$ are all the kth power residues modulo p and showed that for any positive even integer k such that $k\mid p-1$ , if $-1$ is not a kth power modulo p, then

$$ \begin{align*} \det\bigg[\frac{1}{\alpha_i+\alpha_j}\bigg]_{1\le i,j\le m} \equiv \frac{(-1)^{(m+1)/2}}{(2k)^m}\pmod p, \end{align*} $$

where $m=(p-1)/k$ .

Now let $\mathbb {F}_q$ be the finite field of q elements with $\mathrm {char}(\mathbb {F}_q)=p>2$ . It is known that $\mathbb {F}_q^{\times }=\mathbb {F}_q\setminus \{0\}$ is a cyclic group of order $q-1$ and that the subgroups

$$ \begin{align*}U_k=\{x\in\mathbb{F}_q: x^k=1\}=\{a_1,\ldots,a_k\}\quad (k\ge 1, k\mid q-1)\end{align*} $$

are exactly all subgroups of $\mathbb {F}_q^{\times }$ . Let $\phi $ be the unique quadratic character of $\mathbb {F}_q$ , that is,

$$ \begin{align*}\phi(x)=\begin{cases} 1 & \mbox{if}\ x\ \text{is a nonzero square},\\ 0 & \mbox{if}\ x=0,\\ -1 & \mbox{otherwise.} \end{cases}\end{align*} $$

As $\mathrm {char}(\mathbb {F}_q)>2$ , the subset $\{\pm 1\}\subseteq \mathbb {Z}$ can be viewed as a subset of $\mathbb {F}_q$ . From now on, we always assume $\pm 1\in \mathbb {F}_q$ . Inspired by Sun’s determinant (1.1), it is natural to consider the matrix

$$ \begin{align*} \bigg[\frac{\phi(a_i+a_j)}{a_i+a_j}\bigg]_{1\le i,j\le k}. \end{align*} $$

However, if $k\mid q-1$ is even, then the denominator $a_i+a_j=0$ for some $i,j$ since ${-1\in U_k}$ in this case. To overcome this obstacle, note that for any $x\in \mathbb {F}_q$ , we have $\phi (x)=x^{(q-1)/2}$ . Hence, we first focus on the matrix

$$ \begin{align*} D_k:=[(a_i+a_j)^{(q-3)/2}]_{1\le i,j\le k}. \end{align*} $$

The main results involving $D_k$ will be given in Section 1.2.

We now consider another type of determinant. Sun [Reference Sun8, Remark 1.3] posed the following conjecture.

Conjecture 1.2 (Sun).

Let $p\equiv 2\pmod 3$ be an odd prime. Then

(1.2) $$ \begin{align} 2\det\bigg[\frac{1}{i^2-ij+j^2}\bigg]_{1\le i,j\le p-1} \end{align} $$

is a quadratic residue modulo p.

The third author, She and Ni [Reference Wu, She and Ni12] obtained the following generalised result.

Theorem 1.3 (Wu, She and Ni).

Let $q\equiv 2\pmod 3$ be an odd prime power. Let $\beta _1,\ldots ,\beta _{q-1}$ be all the nonzero elements of $\mathbb {F}_q$ . Then

$$ \begin{align*}\det\bigg[\frac{1}{\beta_i^2-\beta_i\beta_j+\beta_j^2}\bigg]_{1\le i,j\le q-1} =(-1)^{(q+1)/2}2^{(q-2)/3}\in\mathbb{F}_p, \end{align*} $$

where $p=\mathrm {char}(\mathbb {F}_q)$ .

Recently, Luo and Sun [Reference Luo and Sun6] investigated the determinant

(1.3) $$ \begin{align} \det S_p(c,d)=\det[(i^2+cij+dj^2)^{p-2}]_{1\le i,j\le p-1}. \end{align} $$

For $(c,d)=(1,1)$ or $(2,2)$ , they determined the explicit values of $({\det S_p(c,d)}/{p})$ .

Motivated by Sun’s determinants (1.1)–(1.3) and the above discussions, we also consider the matrix

$$ \begin{align*}T_k:=[(a_i^2+a_ia_j+a_j^2)^{(q-3)/2}]_{1\le i,j\le k}.\end{align*} $$

We will state our results concerning $T_k$ in Section 1.3.

1.2 The main results involving det $\,D_k$

Theorem 1.4. Let $\mathbb {F}_q$ be the finite field of q elements with $\mathrm {char}(\mathbb {F}_q)=p>2$ . Then for any integer $k\mid q-1$ with $1<k\le q-1$ ,

$$ \begin{align*}\det D_k=(-1)^{(k+1)(q-3)/2}\cdot w_k\cdot k^k\in\mathbb{F}_p,\end{align*} $$

where

$$ \begin{align*}w_k=\prod_{s=0}^{k-1}\sum_{r=0}^{\lfloor{(q-3-2s)}/{2k}\rfloor} \binom{(q-3)/2}{s+rk}\in\mathbb{F}_p.\end{align*} $$

Suppose now that $k=(q-1)/2$ , that is, $U_{(q-1)/2}$ is the set of all the nonzero squares over $\mathbb {F}_q$ . Then we can obtain the following simplified result which will be proved in Section 2.

Corollary 1.5. Let $\mathbb {F}_q$ be the finite field of q elements with $\mathrm {char}(\mathbb {F}_q)=p>2$ . Then

$$ \begin{align*}\det D_{(q-1)/2}= \begin{cases} (-1)^{({q+3})/{4}}u^2 & \mbox{if}\ q\equiv 1\pmod4,\\ (-1)^{({q+5})/{4}}\binom{(q-3)/2}{(q-3)/4}v^2 & \mbox{if}\ q\equiv 3\pmod4\ \text{and}\ q>3, \end{cases}\end{align*} $$

where $u,v\in \mathbb {F}_p$ are defined by

$$ \begin{align*}u=\prod_{s=0}^{(q-5)/4}\binom{(q-3)/2}{s}\quad \text{and}\quad v=\prod_{s=0}^{(q-7)/4}\binom{(q-3)/2}{s}.\end{align*} $$

In particular, if $q=p>3$ is an odd prime, then $D_{(p-1)/2}$ is nonsingular and

$$ \begin{align*}\bigg(\frac{\det D_{(p-1)/2}}{p}\bigg)= \begin{cases} 1 & \mbox{if}\ p\equiv1\pmod4,\\ (-1)^{(h(-p)+1)/2} & \mbox{if}\ p\equiv 3\pmod4\ \text{and}\ p>3, \end{cases}\end{align*} $$

where $h(-p)$ is the class number of $\mathbb {Q}(\sqrt {-p})$ .

From Theorem 1.4, we see that $\det D_k\in \mathbb {F}_p$ . The next result gives the explicit value of $({\det D_k}/{p})$ when k is odd.

Theorem 1.6. Let $\mathbb {F}_q$ be the finite field of q elements with $\mathrm {char}(\mathbb {F}_q)=p>2$ . Let ${1<k\le q-1}$ be an odd integer with $k\mid q-1$ . Suppose that $D_k$ is nonsingular. Then

$$ \begin{align*} \bigg(\frac{\det D_k}{p}\bigg)=\bigg(\frac{s_k}{p}\bigg), \end{align*} $$

where

$$ \begin{align*} s_k:=k\sum_{r=1}^{({q-1)}/{2k}}\binom{(q-3)/2}{((2r-1)k-1)/2}\in\mathbb{F}_p.\end{align*} $$

1.3 The main results involving $\mathrm{det}\ T_k$

To state the next results, we need to introduce some basic properties of trinomial coefficients. Let n be a positive integer. For any integer r, the trinomial coefficient $\binom {n}{r}_2$ is defined by

$$ \begin{align*}\bigg(x+\frac{1}{x}+1\bigg)^{n}= \sum_{r=-\infty}^{+\infty}\binom{n}{r}_2x^r.\end{align*} $$

This implies that $\binom {n}{r}_2=0$ whenever $|r|>n$ and that $\binom {n}{r}_2=\binom {n}{-r}_2$ for any integer r. In particular, $\binom {n}{0}_2$ is usually called the central trinomial coefficient because $\binom {n}{0}_2$ is exactly the coefficient of $x^n$ in the polynomial $(x^2+x+1)^n$ . For simplicity, $\binom {n}{0}_2$ is also denoted by $t_n$ .

Theorem 1.7. Let $\mathbb {F}_q$ be the finite field of q elements with $\mathrm {char}(\mathbb {F}_q)=p>2$ . Then for any integer $k\mid q-1$ with $1<k\le q-1$ ,

$$ \begin{align*}\det T_k=l_k\cdot k^k\in\mathbb{F}_p,\end{align*} $$

where

$$ \begin{align*}l_k=\prod_{s=0}^{k-1}\sum_{r=0}^{\lfloor{(q-3-s)}/{k}\rfloor} \binom{(q-3)/2}{(q-3)/2-s-kr}_2\in\mathbb{F}_p.\end{align*} $$

As a direct consequence of Theorem 1.7, we have the following result.

Corollary 1.8. Let $\mathbb {F}_q$ be the finite field of q elements with $\mathrm {char}(\mathbb {F}_q)=p>2$ . For any integer $k\mid q-1$ with $1<k\le q-1$ , the matrix $T_k$ is singular over $\mathbb {F}_q$ if and only if

$$ \begin{align*} \sum_{r=0}^{\lfloor{(q-3-s)}/{k}\rfloor} \binom{(q-3)/2}{(q-3)/2-s-kr}_2\equiv0\pmod p\end{align*} $$

for some s with $0\le s\le k-1$ . In particular, $T_{q-1}$ is a singular matrix over $\mathbb {F}_q$ .

In the case $k=(q-1)/2$ , similar to Corollary 1.5, by Theorem 1.7, we deduce the following simplified result.

Corollary 1.9. Let $\mathbb {F}_q$ be the finite field of q elements with $\mathrm {char}(\mathbb {F}_q)=p>2$ .

  1. (i) If $q\equiv 1\pmod 4$ , then

    $$ \begin{align*}\det T_{(q-1)/2}= \prod_{s=0}^{(q-5)/4} \bigg(\binom{(q-3)/2}{(q-3)/2-s}_2+\binom{(q-3)/2}{1+s}_2\bigg)^2.\end{align*} $$
  2. (ii) If $q\equiv 3\pmod 4$ and $q>3$ , then

    $$ \begin{align*}\det T_{(q-1)/2}= \binom{(q-3)/2}{0}_2\prod_{s=0}^{(q-7)/4} \Bigg(\binom{(q-3)/2}{(q-3)/2-s}_2+\binom{(q-3)/2}{1+s}_2\Bigg)^2.\end{align*} $$

In particular, if $T_{(q-1)/2}$ is nonsingular, then

$$ \begin{align*} \bigg(\frac{\det T_{(q-1)/2}}{p}\bigg)=\begin{cases} (-1)^{(q-1)/4} & \mbox{if}\ q\equiv1\pmod4,\\ \bigg(\dfrac{t_{(q-3)/2}}{p}\bigg)(-1)^{(q+5)/4} & \mbox{if}\ q\equiv3\pmod4\ \text{and}\ q>3. \end{cases} \end{align*} $$

2 Proofs of Theorem 1.4 and Corollary 1.5

We begin with the following result (see [Reference Krattenthaler4, Lemma 10]).

Lemma 2.1. Let R be a commutative ring. Let $P(t)=p_0+p_1t+\cdots +p_{n-1}t^{n-1}\in R[t]$ . Then

$$ \begin{align*}\det[P(X_iY_j)]_{1\le i,j\le n}=\prod_{i=0}^{n-1}p_i \cdot \prod_{1\le i<j\le n}(X_j-X_i)(Y_j-Y_i).\end{align*} $$

We also need the following result.

Lemma 2.2. Let $\mathbb {F}_q$ be the finite field of q elements with $\mathrm {char}(\mathbb {F}_q)=p$ . For any positive integer $k\mid q-1$ , if we set $U_k=\{a_1,\ldots ,a_k\}$ , then

$$ \begin{align*}\prod_{1\le i<j\le k}(a_j-a_i)\bigg(\frac{1}{a_j}-\frac{1}{a_i}\bigg)=k^k\in\mathbb{F}_p.\end{align*} $$

Proof. It is clear that

(2.1) $$ \begin{align} \prod_{1\le i<j\le k}(a_j-a_i)\bigg(\frac{1}{a_j}-\frac{1}{a_i}\bigg) =\prod_{1\le i<j\le k}\frac{(a_j-a_i)(a_i-a_j)}{a_ia_j} =\prod_{1\le i\neq j\le k}(a_j-a_i) \prod_{1\le i<j\le k}\frac{1}{a_ia_j}. \end{align} $$

Let $S_1=\prod _{1\le i\neq j\le k}(a_j-a_i)$ and let $S_2=\prod _{1\le i<j\le k}{1}/{(a_ia_j)}$ . We first consider $S_1$ . Let

$$ \begin{align*}G_k(t)=\prod_{i=1}^k(t-a_i)\in\mathbb{F}_q[t]\end{align*} $$

and let $G^{\prime }_k(t)$ be the formal derivative of $G_k(t)$ . Then by the definition of $U_k$ , we see that $G_k(t)=t^k-1$ . Thus, $G_k^{\prime }(t)=kt^{k-1}$ and $\prod _{1\le j\le k}a_j=(-1)^{k+1}$ . Now we can verify that

(2.2) $$ \begin{align} S_1 =\prod_{1\le i\neq j\le k}(a_j-a_i)=\prod_{1\le j\le k}G_k^{\prime}(a_j) =\prod_{1\le j\le k}ka_j^{k-1} =k^k(-1)^{k+1}. \end{align} $$

We turn to $S_2$ . It is clear that

(2.3) $$ \begin{align} S_2=\prod_{1\le i<j\le k}\frac{1}{a_ia_j}=\prod_{1\le j\le k}\frac{1}{a_j^{k-1}}=(-1)^{k+1}. \end{align} $$

Combining (2.1) with (2.2) and (2.3),

$$ \begin{align*} \prod_{1\le i<j\le k}(a_j-a_i)\bigg(\frac{1}{a_j}-\frac{1}{a_i}\bigg)=S_1S_2=k^k\in\mathbb{F}_p. \end{align*} $$

This completes the proof.

Proof of Theorem 1.4.

As $\mathrm {char}(\mathbb {F}_q)=p>2$ , the subset $\{1,-1\}\subseteq \mathbb {Z}$ can be naturally viewed as a subset of $\mathbb {F}_q$ . One can verify that

(2.4) $$ \begin{align} \det D_k =\det[(a_i+a_j)^{{(q-3)}/{2}}]_{1\le i,j\le k} &=\prod_{i=1}^ka_i^{{(q-3)}/{2}}\det\bigg[\bigg(1+\frac{a_j}{a_i}\bigg)^{{(q-3)}/{2}}\bigg]_{1\le i,j\le k}\nonumber\\ &=(-1)^{(k+1)(q-3)/2}\det\bigg[\bigg(1+\frac{a_j}{a_i}\bigg)^{{(q-3)}/{2}}\bigg]_{1\le i,j\le k}. \end{align} $$

The last equality follows from $\prod _{1\le j\le k}a_j=(-1)^{k+1}$ . Let

$$ \begin{align*}f_k(t) =\sum_{s=0}^{k-1} \Bigg(\sum_{r=0}^{\lfloor{(q-3-2s)}/{2k}\rfloor} \binom{(q-3)/2}{s+rk}\Bigg)t^s\in\mathbb{F}_p[t]\end{align*} $$

with $\deg (f_k)\le k-1$ . Noting that $(a_j/a_i)^{k+s}=(a_j/a_i)^s$ for any integer s, by (2.4),

$$ \begin{align*} \det D_k =(-1)^{(k+1)(q-3)/2} \cdot \det\bigg[f_k\bigg(\frac{a_j}{a_i}\bigg)\bigg]_{1\le i,j\le k}. \end{align*} $$

Let

$$ \begin{align*}w_k:=\prod_{s=0}^{k-1}\sum_{r=0}^{\lfloor{(q-3-2s)}/{2k}\rfloor} \binom{(q-3)/2}{s+rk}\in\mathbb{F}_p.\end{align*} $$

Then by Lemmas 2.1 and 2.2,

$$ \begin{align*}\det D_k=(-1)^{(k+1)(q-3)/2}\cdot w_k\cdot \prod_{1\le i<j\le k}(a_j-a_i)\bigg(\frac{1}{a_j}-\frac{1}{a_i}\bigg)=(-1)^{(k+1)(q-3)/2}\cdot w_k\cdot k^k\in\mathbb{F}_p.\end{align*} $$

This completes the proof.

Proof of Corollary 1.5.

By Theorem 1.4, if $k=(q-1)/2$ , then

(2.5) $$ \begin{align} \det D_{(q-1)/2} &=(-1)^{(q-3)/2}\cdot\prod_{s=0}^{(q-3)/2}\binom{(q-3)/2}{s}\cdot (-1)^{(q-1)/2}\bigg(\frac{1}{2}\bigg)^{(q-1)/2}\nonumber\\ &=-1\cdot\prod_{s=0}^{(q-3)/2}\binom{(q-3)/2}{s}\cdot\phi(2). \end{align} $$

The last equality follows from

$$ \begin{align*}(\tfrac{1}{2})^{(q-1)/2}=\phi(\tfrac{1}{2})=\phi(2).\end{align*} $$

We now divide the remaining part of the proof into two cases.

Case 1: $q\equiv 1\pmod 4$ .

In this case, we have $\sqrt {-1}\in \mathbb {F}_q$ , where $\sqrt {-1}$ is an element in the algebraic closure of $\mathbb {F}_q$ such that $(\sqrt {-1})^2=-1$ . Since $2=-\sqrt {-1}(1+\sqrt {-1})^2,$ we have $\phi (2)=\phi (-\sqrt {-1})$ and hence

(2.6) $$ \begin{align} \phi(2)=\phi(-\sqrt{-1})=(-\sqrt{-1})^{(q-1)/2}=(-1)^{(q-1)/4}. \end{align} $$

Combining (2.5) with (2.6) and noting that

$$ \begin{align*}\binom{(q-3)/2}{s}=\binom{(q-3)/2}{(q-3)/2-s},\end{align*} $$

we obtain

(2.7) $$ \begin{align} \det D_{(q-1)/2}=(-1)^{(q+3)/4}\prod_{s=0}^{(q-5)/4}\binom{(q-3)/2}{s}^2. \end{align} $$

This proves the case $q\equiv 1\pmod 4$ .

Case 2: $q\equiv 3\pmod 4$ and $q>3$ .

In this case, since $q\equiv 3\pmod 4$ , $(1+\sqrt {-1})^q=1+(\sqrt {-1})^q=1-\sqrt {-1}$ . This, together with $2=-\sqrt {-1}\big (1+\sqrt {-1}\big )^2$ , implies that

(2.8) $$ \begin{align} \phi(2) &=2^{(q-1)/2}=(-\sqrt{-1})^{(q-3)/2}(-\sqrt{-1})\big(1+\sqrt{-1}\big)^{q-1}\nonumber\\ &=(-1)^{(q-3)/4}(-\sqrt{-1})\frac{1-\sqrt{-1}}{1+\sqrt{-1}}\nonumber\\ &=(-1)^{(q+1)/4}. \end{align} $$

Combining (2.5) with (2.8),

(2.9) $$ \begin{align} \det D_{(q-1)/2}=(-1)^{(q+5)/4}\binom{(q-3)/2}{(q-3)/4}\prod_{s=0}^{(q-7)/4}\binom{(q-3)/2}{s}^2. \end{align} $$

This proves the case $q\equiv 3\pmod 4$ and $q>3$ .

Now we assume that $q=p$ is an odd prime. Suppose first $p\equiv 1\pmod 4$ . Then by (2.7), we see that $\det D_{(q-1)/2}$ is a nonzero square in $\mathbb {F}_p$ , that is, $({\det D_{(p-1)/2}}/{p})=1$ . In the case $p\equiv 3\pmod 4$ and $p>3$ , by (2.9) and $({-2}/{p})=({-\tfrac 12}/{p})=(-1)^{(p+5)/4}$ ,

$$ \begin{align*} \bigg(\frac{\det D_{(q-1)/2}}{p}\bigg)&=(-1)^{(p+5)/4}\Bigg(\frac{\tfrac{p-3}{2}!}{p}\Bigg) =(-1)^{(p+5)/4}\Bigg(\frac{\tfrac{p-1}{2}!}{p}\Bigg) \Bigg(\frac{\tfrac{-1}{2}}{p}\Bigg) =\Bigg(\frac{\tfrac{p-1}{2}!}{p}\Bigg) =(-1)^{(h(-p)+1)/2}. \end{align*} $$

The last equality follows from Mordell’s result [Reference Mordell7] which states that

$$ \begin{align*}\frac{p-1}{2}!\equiv (-1)^{{(h(-p)+1)}/{2}}\pmod p\end{align*} $$

whenever $p\equiv 3\pmod 4$ and $p>3$ . This completes the proof.

3 Proof of Theorem 1.6

To prove Theorem 1.6, we first need the following well-known result.

Lemma 3.1. Let $\mathbb {F}_q$ be the finite field of q elements and let r be a positive integer. Then

$$ \begin{align*} \sum_{x\in\mathbb{F}_q}x^r=\begin{cases} 0 & \mbox{if}\ q-1\nmid r,\\ -1 & \mbox{if}\ q-1\mid r. \end{cases} \end{align*} $$

We will see later in the proof that $\det D_k$ has close relations with the determinant of a circulant matrix. Hence, we now introduce the definition of circulant matrices. Let R be a commutative ring. Let $b_0,b_1,\ldots ,b_{s-1}\in R$ . We define the circulant matrix $C(b_0,\ldots ,b_{s-1})$ to be an $s\times s$ matrix whose ( $i,j$ )-entry is $b_{j-i}$ where the indices are cyclic module s, that is, $b_i=b_j$ whenever $i\equiv j\pmod s$ . The third author [Reference Wu11, Lemma 3.4] obtained the following result.

Lemma 3.2. Let R be a commutative ring. Let s be a positive integer. Let $b_0,b_1,\ldots ,b_{s-1}\in R$ such that $b_i=b_{s-i}$ for $1\leqslant i\leqslant s-1$ .

If s is even, then there exists an element $u\in R$ such that

$$ \begin{align*} \det C(b_0,\ldots,b_{s-1})=\bigg(\sum_{i=0}^{s-1}b_i\bigg)\bigg(\sum_{i=0}^{s-1}(-1)^ib_i\bigg)\cdot u^2. \end{align*} $$

If s is odd, then there exists an element $v\in R$ such that

$$ \begin{align*} \det C(b_0,\ldots,b_{s-1})=\bigg(\sum_{i=0}^{s-1}b_i\bigg)\cdot v^2. \end{align*} $$

Proof of Theorem 1.6.

As k is odd, we have $2\mid (q-1)/k$ . For simplicity, we let $q-1=nk=2mk$ . Since $k\mid (q-1)/2$ in this case, $\phi (a_i)=a_i^{(q-1)/2}=1$ for each $a_i\in U_k$ . Let g be a generator of the cyclic group $\mathbb {F}_q^{\times }$ . By the above, one can verify that

$$ \begin{align*} \det D_k =\prod_{i=1}^ka_i^{(q-3)/2}\det\bigg[\bigg(1+\frac{a_j}{a_i}\bigg)^{(q-3)/2}\bigg]_{1\le i,j\le k} =\det[(1+g^{nj-ni})^{(q-3)/2}]_{0\le i,j\le k-1}. \end{align*} $$

The last equality follows from

$$ \begin{align*} \prod_{i=1}^ka_i=(-1)^{k+1}=1. \end{align*} $$

By the above and using the properties of determinants, one can verify that

(3.1) $$ \begin{align} \det D_k=\det[(1+g^{nj-ni})^{(q-3)/2}g^{mj-mi}(-1)^{j-i}]_{0\le i,j\le k-1}. \end{align} $$

For $0\le i\le k-1$ ,

$$ \begin{align*}b_i=(1+g^{ni})^{(q-3)/2}g^{mi}(-1)^i.\end{align*} $$

We claim that $b_i=b_{k-i}$ for $1\le i\le k-1$ . In fact, for $1\le i\le k-1$ , noting that

$$ \begin{align*}g^{km}=\phi(g)=-1,\quad g^{nk}=1,\ 2\nmid k \quad \text{and}\quad \bigg(\frac{1}{g^{ni}}\bigg)^{(q-3)/2}=g^{ni},\end{align*} $$

one can verify that

$$ \begin{align*} b_{k-i}&=(1+g^{nk-ni})^{(q-3)/2}g^{mk-mi}(-1)^{k-i}\\ &=\bigg(\frac{1+g^{ni}}{g^{ni}}\bigg)^{(q-3)/2}g^{-mi}(-1)^i\\ &=(1+g^{ni})^{(q-3)/2}g^{(n-m)i}(-1)^i\\ &=b_i. \end{align*} $$

Hence, by (3.1), $\det D_k=\det C(b_0,b_1,\ldots ,b_{k-1})$ . Now by Lemma 3.2 and (3.1),

(3.2) $$ \begin{align} \det D_k=\bigg(\sum_{i=0}^{k-1}b_i\bigg)v^2 \end{align} $$

for some $v\in \mathbb {F}_q$ . Now we consider the sum $\sum _{i=0}^{k-1}b_i$ . It is easy to verify that

(3.3) $$ \begin{align} \sum_{i=0}^{k-1}b_i &=\sum_{i=0}^{k-1}(1+g^{ni})^{(q-3)/2}g^{mi}(-1)^i =\sum_{i=0}^{k-1}(1+g^{ni})^{(q-3)/2}g^{mi}g^{mki}\nonumber\\ &=\frac{1}{n}\sum_{x\in\mathbb{F}_q}(1+x^n)^{(q-3)/2}x^{m+mk} =\frac{1}{n}\sum_{r=0}^{mk-1}\binom{(q-3)/2}{r}\sum_{x\in\mathbb{F}_q}x^{m+mk+2mr}. \end{align} $$

Now by Lemma 3.1, since $2\nmid k$ ,

$$ \begin{align*}\sum_{x\in\mathbb{F}_q}x^{m+mk+2mr}=\begin{cases} 0 & \mbox{if}\ k\nmid 1+2r,\\ -1 & \mbox{if}\ k\mid 1+2r. \end{cases} \end{align*} $$

Applying this and Lemma 3.1 to (3.3) and noting that $-1/n=k$ in $\mathbb {F}_p$ ,

(3.4) $$ \begin{align} s_k:=\sum_{i=0}^{k-1}b_i=k\sum_{r=1}^m\binom{(q-3)/2}{((2r-1)k-1)/2}. \end{align} $$

Suppose that $D_k$ is nonsingular. Then by Theorem 1.4, we have $\det D_k\in \mathbb {F}_p^{\times }$ . Hence, by (3.2) and (3.4),

$$ \begin{align*}\bigg(\frac{\det D_k}{p}\bigg)=\bigg(\frac{s_k}{p}\bigg).\end{align*} $$

This completes the proof.

4 Proof of Theorem 1.7

It is clear that

(4.1) $$ \begin{align} \det T_k &=\prod_{i=1}^{k}(a_i^2)^{(q-3)/2} \cdot \det\bigg[\bigg(1+\frac{a_j}{a_i}+\bigg(\frac{a_j}{a_i}\bigg)^2\bigg)^{(q-3)/2}\bigg]_{1\le i,j\le k}\nonumber\\ &=\det\bigg[\bigg(1+\frac{a_j}{a_i}+\bigg(\frac{a_j}{a_i}\bigg)^2\bigg)^{(q-3)/2}\bigg]_{1\le i,j\le k}. \end{align} $$

The last equality follows from

$$ \begin{align*}\prod_{i=1}^ka_i=(-1)^{k+1}.\end{align*} $$

Let

$$ \begin{align*}g_k(t)= \sum_{s=0}^{k-1} \Bigg(\sum_{r=0}^{\lfloor{(q-3-s)}/{k}\rfloor} \binom{(q-3)/2}{s+rk-(q-3)/2}_2\Bigg)t^s\in\mathbb{F}_p[t]\end{align*} $$

with $\deg (g_k)\le k-1$ . Then by (4.1), Lemma 2.1 and the definition of trinomial coefficients,

$$ \begin{align*} \det T_k &=\det\bigg[g_k\bigg(\frac{a_j}{a_i}\bigg)\bigg]_{1\le i,j\le k}\\ &=\prod_{1\le i<j\le k}(a_j-a_i)\bigg(\frac{1}{a_j}-\frac{1}{a_i}\bigg)\cdot \prod_{s=0}^{k-1}\sum_{r=0}^{\lfloor{(q-3-s)}/{k}\rfloor} \binom{(q-3)/2}{s+rk-(q-3)/2}_2\\ &=l_kk^k\in\mathbb{F}_p. \end{align*} $$

The last equality follows from Lemma 2.2. This completes the proof. $\Box $

Acknowledgement

We would like to thank the referee for helpful comments.

Footnotes

This work was supported by the Natural Science Foundation of China (Grant No. 12101321).

References

Chapman, R., ‘Determinants of Legendre symbol matrices’, Acta Arith. 115 (2004), 231244.10.4064/aa115-3-4CrossRefGoogle Scholar
Chapman, R., ‘My evil determinant problem’, Online lecture notes, December 12, 2012. Available at http://empslocal.ex.ac.uk/people/staff/rjchapma/etc/evildet.pdf.Google Scholar
Krachun, D., Petrov, F., Sun, Z.-W. and Vsemirnov, M., ‘On some determinants involving Jacobi symbols’, Finite Fields Appl. 64 (2020), Article no. 101672.10.1016/j.ffa.2020.101672CrossRefGoogle Scholar
Krattenthaler, C., ‘Advanced determinant calculus: a complement’, Linear Algebra Appl. 411 (2005), 68166.10.1016/j.laa.2005.06.042CrossRefGoogle Scholar
Li, Y.-B. and Wei, N.-L., ‘A variant of some cyclotomic matrices involving trinomial coefficients’, Colloq. Math. 174 (2023), 3743.10.4064/cm9115-6-2023CrossRefGoogle Scholar
Luo, X.-Q. and Sun, Z.-W., ‘Legendre symbols related to certain determinants’, Bull. Malays. Math. Sci. Soc. 46 (2023), Article no. 119.10.1007/s40840-023-01505-2CrossRefGoogle Scholar
Mordell, L. J., ‘The congruence $\left(\left(p-1\right)/2\right)!\equiv \pm 1\;\left(\operatorname{mod}\;p\right)$ ’, Amer. Math. Monthly 68 (1961), 145146.10.2307/2312481CrossRefGoogle Scholar
Sun, Z.-W., ‘On some determinants with Legendre symbols entries’, Finite Fields Appl. 56 (2019), 285307.10.1016/j.ffa.2018.12.004CrossRefGoogle Scholar
Vsemirnov, M., ‘On the evaluation of R. Chapman’s “evil determinant”’, Linear Algebra Appl. 436 (2012), 41014106.10.1016/j.laa.2011.08.039CrossRefGoogle Scholar
Vsemirnov, M., ‘On R. Chapman’s “evil determinant”: case $p\equiv 1\;\left(\operatorname{mod}\;4\right)$ ’, Acta Arith. 159 (2013), 331344.10.4064/aa159-4-3CrossRefGoogle Scholar
Wu, H.-L., ‘Elliptic curves over ${F}_p$ and determinants of Legendre matrices’, Finite Fields Appl. 76 (2021), Article no. 101929.10.1016/j.ffa.2021.101929CrossRefGoogle Scholar
Wu, H.-L., She, Y.-F. and Ni, H.-X., ‘A conjecture of Zhi-Wei Sun on determinants over finite fields’, Bull. Malays. Math. Sci. Soc. 45 (2022), 24052412.10.1007/s40840-022-01357-2CrossRefGoogle Scholar
Wu, H.-L., She, Y.-F. and Wang, L.-Y., ‘Cyclotomic matrices and hypergeometric functions over finite fields’, Finite Fields Appl. 82 (2022), Article no. 102054.10.1016/j.ffa.2022.102054CrossRefGoogle Scholar
Wu, H.-L. and Wang, L.-Y., ‘Applications of circulant matrices to determinants involving $k$ th power residues’, Bull. Aust. Math. Soc. 106 (2022), 243253.10.1017/S0004972722000053CrossRefGoogle Scholar