Introduction
Let q q q be a power of two and F q 2 \mathbb{F}_{q^2} F q 2 be the finite field with q 2 q^2 q 2 elements. Niho in his thesis [1 ] intensively studied the cross-correlation between binary m-sequences of length q 2 − 1 q^2-1 q 2 − 1 and their d d d -decimated m-sequences when d d d is of the form d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 for certain positive integer s s s . It turned out this type of integers involved interesting algebraic structures over finite fields and they were later referred to as normalized Niho exponents in the literature. In the last decades researchers have explored their applications in different areas, including low-correlation sequences [1 , 2 , 3 , 4 , 5 , 6 ] , bent functions [7 , 8 ] , highly nonlinear functions [9 ] , permutation polynomials [10 , 11 ] , cyclic codes [12 , 13 ] , Weil sums [14 , 15 ] , Kloosterman sums [4 ] , etc. Interested readers may refer to the survey on Niho exponents and their applications in [16 ] .
When a Niho exponent d d d is co-prime to q 2 − 1 q^2-1 q 2 − 1 , the following three problems are equivalent [4 ] : the value distribution of cross-correlation between a binary m-sequence of length q 2 − 1 q^2-1 q 2 − 1 and its d d d -decimated sequence, the value distribution of Walsh transforms of the power function x d x^d x d over the finite field F q 2 \mathbb{F}_{q^2} F q 2 , and the weight distribution of the cyclic code of length q 2 − 1 q^2-1 q 2 − 1 associated with α , α d \alpha,\alpha^d α , α d , where α \alpha α is a primitive element of F q 2 \mathbb{F}_{q^2} F q 2 . More concretely, these three problems reduce to the value distribution of the following Weil sum
W d ( a ) = ∑ x ∈ F q 2 ( − 1 ) T r 2 m ( x d + a x ) , a ∈ F q 2 , (1) W_d(a)=\sum_{x\in\mathbb{F}_{q^2}}(-1)^{\mathrm{Tr}_{2m}(x^d+ax)},\qquad a\in\mathbb{F}_{q^2}, \tag*{(1)} W d ( a ) = x ∈ F q 2 ∑ ( − 1 ) Tr 2 m ( x d + a x ) , a ∈ F q 2 , (1)
where T r 2 m ( ⋅ ) \mathrm{Tr}_{2m}(\cdot) Tr 2 m ( ⋅ ) denotes the trace function from F q 2 = F 2 2 m \mathbb{F}_{q^2}=\mathbb{F}_{2^{2m}} F q 2 = F 2 2 m to F 2 \mathbb{F}_2 F 2 . Actually the cross-correlation function C d ( τ ) C_d(\tau) C d ( τ ) between a binary m-sequence of length q 2 − 1 q^2-1 q 2 − 1 and its d d d -decimated sequence is given by
C d ( τ ) = W d ( α τ ) − 1 , 0 ≤ τ < q 2 − 1. C_d(\tau)=W_d(\alpha^\tau)-1,\qquad0\leq\tau<q^2-1. C d ( τ ) = W d ( α τ ) − 1 , 0 ≤ τ < q 2 − 1.
To determine the value distribution of W d ( a ) W_d(a) W d ( a ) for a normalized Niho exponent d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 , one typically transforms this problem to counting roots of an associated polynomial f s , a ( x ) = x 2 s − 1 + a x s + a ˉ x s − 1 + 1 f_{s,a}(x)=x^{2s-1}+ax^s+\bar{a}x^{s-1}+1 f s , a ( x ) = x 2 s − 1 + a x s + a ˉ x s − 1 + 1 in the unit circle U = { u ∈ F q 2 : u q + 1 = 1 } U=\{u\in\mathbb{F}_{q^2}:u^{q+1}=1\} U = { u ∈ F q 2 : u q + 1 = 1 } , as the element a a a runs through F q 2 \mathbb{F}_{q^2} F q 2 . It is worth noting that the problem has been investigated more than fifty years. Back to 1972 Niho in his PhD thesis [1 ] tried different choices of s s s , provided a complete 4-valued distribution of cross-correlation for s = 2 s=2 s = 2 , and left a lot of problems unsolved. Other values of s s s were considered later [17 ] . In 2006 Dobbertin, Felke, Helleseth and Rosendahl [4 ] considered a general form of Niho exponents d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 with s ≡ 2 r / ( 2 r ± 1 ) ( m o d q + 1 ) s\equiv2^r/(2^r\pm1)\pmod{q+1} s ≡ 2 r / ( 2 r ± 1 ) ( mod q + 1 ) , which includes all known Niho-type decimations for which the cross-correlation function C d ( τ ) C_d(\tau) C d ( τ ) takes 4 possible values. For s = 3 s=3 s = 3 , the polynomial f s , a ( x ) f_{s,a}(x) f s , a ( x ) is quintic and can have at most 5 roots in U U U , implying that the cross-correlation function may take 6 possible values; the cross-correlation distribution for odd m m m was determined in [4 ] via Dickson polynomials and the cases of even m m m were unsettled. Ten years later Xia, Li, Zeng and Helleseth [5 ] resolved the problem for both even and odd m m m via an interesting connection to the number of weight-five codewords in Zetterberg codes. For s = 4 s=4 s = 4 , while the polynomial f s , a ( x ) f_{s,a}(x) f s , a ( x ) has degree seven, Niho in his thesis conjectured that it actually had 0, 1, 2, 3, 5 roots in U U U for even m m m based on numerical results; this conjecture wasn’t resolved until the work of Helleseth, Katz and Li [15 ] in 2021, where they connected this problem to an orbit pattern of roots of self-conjugate-reciprocal polynomials under a conjugate-reciprocal action; and very recently Xiong and Yan [18 ] determined the complete distribution of the cross-correlation function for s = 4 s=4 s = 4 by a detailed study of the case when f s , a ( x ) f_{s,a}(x) f s , a ( x ) has five roots in U U U . The developments of this research line, which traced back to Niho’s work in 1972, indicate the problem’s intractability as s s s increase. More importantly, those developments uncovered some intrinsic connections of Niho-exponents and interesting objects in other branches of discrete mathematics and coding theory [7 , 12 , 5 , 19 ] .
In this paper, we will consider the following Niho exponent
d = s ( q − 1 ) + 1 , 4 s ≡ 5 ( m o d q + 1 ) , (2) d=s(q-1)+1,\qquad4s\equiv5\pmod{q+1}, \tag*{(2)} d = s ( q − 1 ) + 1 , 4 s ≡ 5 ( mod q + 1 ) , (2)
which covers some instances in [1 Table 3-3] . In this case d d d is co-prime to q 2 − 1 q^2-1 q 2 − 1 when m m m is even. For the associated polynomial f s , a ( x ) f_{s,a}(x) f s , a ( x ) , one can equivalently consider a hexic polynomial P a ( x ) = x 6 + a x 5 + a ˉ x + 1 P_a(x)=x^6+ax^5+\bar{a}x+1 P a ( x ) = x 6 + a x 5 + a ˉ x + 1 , indicating that the cross-correlation function takes at most 7 values of the form q ( i − 1 ) − 1 q(i-1)-1 q ( i − 1 ) − 1 for 0 ≤ i ≤ 6 0\leq i\leq6 0 ≤ i ≤ 6 . In order to determine the cross-correlation distribution for this Niho exponent, we start by developing some general results on self-conjugate-reciprocal (SCR) polynomials [15 ] , namely polynomials f ( x ) = ∑ i = 0 d f i x i ∈ F q 2 [ x ] f(x)=\sum_{i=0}^{d}f_i x^i\in\mathbb{F}_{q^2}[x] f ( x ) = ∑ i = 0 d f i x i ∈ F q 2 [ x ] whose coefficients satisfy f i = ( f d − i ) q f_i=(f_{d-i})^q f i = ( f d − i ) q , which include polynomials f s , a ( x ) f_{s,a}(x) f s , a ( x ) as special cases. In this part we provide one general way to transform SCR polynomials over F q 2 \mathbb{F}_{q^2} F q 2 to polynomials over F q \mathbb{F}_q F q (see Proposition 3.1 ) and one way to transform quartic SCR polynomials over F q 2 \mathbb{F}_{q^2} F q 2 to cubic resolvent polynomials over F q \mathbb{F}_q F q (see Proposition 3.4 ).
This helps us characterize the cases where P a ( x ) P_a(x) P a ( x ) has exactly 4, 5, or 6 roots in U U U , which are described mainly in terms of trace conditions on involved parameters. Furthermore, we consider the enumeration of those cases by calculating character sums for those trace conditions. Based on the correspondence between ( x , y ) (x,y) ( x , y ) and ( x + y , x y ) (x+y,xy) ( x + y , x y ) in F q 2 \mathbb{F}_{q^2} F q 2 , we transform all involved character sums and express them in terms of Kloosterman sums K m ( 1 ) K_m(1) K m ( 1 ) and K m ( ω ) K_m(\omega) K m ( ω ) , where ω \omega ω is a primitive element of F 4 \mathbb{F}_4 F 4 . Finally the complete correlation distribution is determined and is given as follows.
Theorem 1.1. Let q = 2 m q=2^m q = 2 m with m ≥ 2 m\ge2 m ≥ 2 being even and ω \omega ω be a primitive element of F 4 \mathbb{F}_4 F 4 . For the Niho exponent d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 with s s s satisfying 4 s ≡ 5 ( m o d q + 1 ) 4s\equiv5\pmod{q+1} 4 s ≡ 5 ( mod q + 1 ) , the possible values of C d ( τ ) C_d(\tau) C d ( τ ) are q ( i − 1 ) − 1 q(i-1)-1 q ( i − 1 ) − 1 , 0 ≤ i ≤ 6 0\le i\le6 0 ≤ i ≤ 6 and their respective frequencies N i N_i N i are
N 0 = 53 q 2 + ( 52 ( − 1 ) m / 2 − 135 ) q − 8 K m ( 1 ) 2 − 9 K m ( 1 ) + 9 K m ( ω ) 2 − 10 144 , N_0=\frac{53q^2+\left(52(-1)^{m/2}-135\right)q-8K_m(1)^2-9K_m(1)+9K_m(\omega)^2-10}{144}, N 0 = 144 53 q 2 + ( 52 ( − 1 ) m /2 − 135 ) q − 8 K m ( 1 ) 2 − 9 K m ( 1 ) + 9 K m ( ω ) 2 − 10 , N 1 = 44 q 2 + ( 131 − 88 ( − 1 ) m / 2 ) q + 20 K m ( 1 ) 2 + 15 K m ( 1 ) − 24 K m ( ω ) 2 − 101 120 , N_1=\frac{44q^2+\left(131-88(-1)^{m/2}\right)q+20K_m(1)^2+15K_m(1)-24K_m(\omega)^2-101}{120}, N 1 = 120 44 q 2 + ( 131 − 88 ( − 1 ) m /2 ) q + 20 K m ( 1 ) 2 + 15 K m ( 1 ) − 24 K m ( ω ) 2 − 101 , N 2 = 3 q 2 + ( 4 ( − 1 ) m / 2 − 5 ) q − 2 K m ( 1 ) 2 + K m ( 1 ) + 3 K m ( ω ) 2 16 , N_2=\frac{3q^2+\left(4(-1)^{m/2}-5\right)q-2K_m(1)^2+K_m(1)+3K_m(\omega)^2}{16}, N 2 = 16 3 q 2 + ( 4 ( − 1 ) m /2 − 5 ) q − 2 K m ( 1 ) 2 + K m ( 1 ) + 3 K m ( ω ) 2 , N 3 = 2 q 2 + ( 4 ( − 1 ) m / 2 + 9 ) q − 2 K m ( 1 ) 2 − 9 K m ( 1 ) − 7 36 , N_3=\frac{2q^2+\left(4(-1)^{m/2}+9\right)q-2K_m(1)^2-9K_m(1)-7}{36}, N 3 = 36 2 q 2 + ( 4 ( − 1 ) m /2 + 9 ) q − 2 K m ( 1 ) 2 − 9 K m ( 1 ) − 7 , N 4 = q 2 + ( 4 ( − 1 ) m / 2 − 11 ) q + 4 K m ( 1 ) 2 + 3 K m ( 1 ) − 3 K m ( ω ) 2 + 2 48 , N_4=\frac{q^2+\left(4(-1)^{m/2}-11\right)q+4K_m(1)^2+3K_m(1)-3K_m(\omega)^2+2}{48}, N 4 = 48 q 2 + ( 4 ( − 1 ) m /2 − 11 ) q + 4 K m ( 1 ) 2 + 3 K m ( 1 ) − 3 K m ( ω ) 2 + 2 , N 5 = q + 1 + K m ( 1 ) 8 , N_5=\frac{q+1+K_m(1)}{8}, N 5 = 8 q + 1 + K m ( 1 ) , N 6 = q 2 + ( 9 − 52 ( − 1 ) m / 2 ) q − 10 K m ( 1 ) 2 − 45 K m ( 1 ) + 9 K m ( ω ) 2 − 44 720 , N_6=\frac{q^2+\left(9-52(-1)^{m/2}\right)q-10K_m(1)^2-45K_m(1)+9K_m(\omega)^2-44}{720}, N 6 = 720 q 2 + ( 9 − 52 ( − 1 ) m /2 ) q − 10 K m ( 1 ) 2 − 45 K m ( 1 ) + 9 K m ( ω ) 2 − 44 , where K m ( c ) = ∑ x ∈ F q ∗ ( − 1 ) Tr m ( x + c / x ) K_m(c)=\sum_{x\in\mathbb{F}_q^*}(-1)^{\operatorname{Tr}_m(x+c/x)} K m ( c ) = ∑ x ∈ F q ∗ ( − 1 ) Tr m ( x + c / x ) is the Kloosterman sum.
The remainder of this paper is organized as follows. § 2 introduces basics of Niho exponents, Kloosterman sums and self-conjugate-reciprocal (SCR) polynomials. § 3 presents two ways to transform SCR polynomials over F q 2 \mathbb{F}_{q^2} F q 2 to new polynomials over F q \mathbb{F}_q F q and discusses the relation of their roots. § 4 studies the correlation distribution for the Niho exponent d = 5 q − 1 d=5q-1 d = 5 q − 1 and gives three identities for the number of a a a ’s for which P a P_a P a has exactly 4, 5 or 6 roots in the unit circle U U U . § 5 completely determines the correlation distribution by an in-depth investigation of the involved character sums, which are finally expressed in terms of Kloosterman sums K m ( 1 ) K_m(1) K m ( 1 ) and K m ( ω ) K_m(\omega) K m ( ω ) . § 6 summerizes main contributions of this work.
Preliminaries
Throughout this paper we assume q = 2 m q=2^m q = 2 m and denote by F q \mathbb{F}_q F q and F q 2 \mathbb{F}_{q^2} F q 2 the finite fields with q q q and q 2 q^2 q 2 elements, respectively. We use Tr m ( ⋅ ) \operatorname{Tr}_m(\cdot) Tr m ( ⋅ ) for the absolute trace function from F q \mathbb{F}_q F q to F 2 \mathbb{F}_2 F 2 , and Tr 2 m ( ⋅ ) \operatorname{Tr}_{2m}(\cdot) Tr 2 m ( ⋅ ) for the absolute trace function from F q 2 \mathbb{F}_{q^2} F q 2 to F 2 \mathbb{F}_2 F 2 . The conjugate of an element a ∈ F q 2 a\in\mathbb{F}_{q^2} a ∈ F q 2 is defined as a ˉ = a q \bar{a}=a^q a ˉ = a q and the unit circle in F q 2 \mathbb{F}_{q^2} F q 2 is defined as U = { u ∈ F q 2 : u u ˉ = 1 } U=\{u\in\mathbb{F}_{q^2}:u\bar{u}=1\} U = { u ∈ F q 2 : u u ˉ = 1 } , which is a cyclic subgroup in the multiplicative group F q 2 ∗ \mathbb{F}_{q^2}^* F q 2 ∗
Kloosterman sums over finite fields
For a , b ∈ F q a,b \in\mathbb{F}_q a , b ∈ F q , the Kloosterman sum K m ( a , b ) K_m(a,b) K m ( a , b ) over F q \mathbb{F}_q F q is defined by
K m ( a , b ) = ∑ x ∈ F q ∗ ( − 1 ) Tr m ( a x + b / x ) = ∑ x ∈ F q ∗ χ ( a x + b / x ) . (3) K_m(a,b)=\sum_{x\in\mathbb{F}_q^*}(-1)^{\operatorname{Tr}_m(ax+b/x)}=\sum_{x\in\mathbb{F}_q^*}\chi(ax+b/x). \tag*{(3)} K m ( a , b ) = x ∈ F q ∗ ∑ ( − 1 ) Tr m ( a x + b / x ) = x ∈ F q ∗ ∑ χ ( a x + b / x ) . (3)
where χ \chi χ is the canonical additive character of F q \mathbb{F}_q F q , defined by χ ( x ) = ( − 1 ) Tr m ( x ) \chi(x)=(-1)^{\operatorname{Tr}_m(x)} χ ( x ) = ( − 1 ) Tr m ( x ) . Kloosterman sums were introduced in 1926 [20 ] to study a problem on the representation of integers by quadratic forms. Later they found applications in coding theory and cryptography and have been extensively studied in last decades; see, for example, [21 , 22 , 23 , 24 ] . For further basic properties of Kloosterman sums over finite fields, we refer the reader to [21 , 25 ] .
Throughout this paper, we abbreviate K m ( 1 , b ) K_m(1,b) K m ( 1 , b ) as K m ( b ) K_m(b) K m ( b ) . For a ≠ 0 a\ne0 a = 0 , the substitution y = a x y=ax y = a x gives K m ( a , b ) = K m ( a b ) K_m(a,b)=K_m(ab) K m ( a , b ) = K m ( ab ) . The substitution x ↦ x − 1 x\mapsto x^{-1} x ↦ x − 1 also gives K m ( a , b ) = K m ( b , a ) K_m(a,b)=K_m(b,a) K m ( a , b ) = K m ( b , a ) any a , b ∈ F q a,b\in\mathbb{F}_q a , b ∈ F q . We also have K m ( 0 ) = − 1 K_m(0)=-1 K m ( 0 ) = − 1 and K m ( c 2 ) = K m ( c ) K_m(c^2)=K_m(c) K m ( c 2 ) = K m ( c ) for any c ∈ F q c\in\mathbb{F}_q c ∈ F q . Below we recall some identities and recurrence relations that will be used in subsequent sections. We start with one basic identity:
K m ( a , b ) = ∑ x ≠ 0 χ ( a x + b / x ) = { q − 1 , if a = b = 0 , K m ( a b ) , otherwise. (4) K_m(a,b)=\sum_{x\ne0}\chi(ax+b/x)=
\begin{cases}
q-1, & \text{if }a=b=0,\\
K_m(ab), & \text{otherwise.}
\end{cases} \tag*{(4)} K m ( a , b ) = x = 0 ∑ χ ( a x + b / x ) = { q − 1 , K m ( ab ) , if a = b = 0 , otherwise. (4)
The following lemmas give some useful identities of Kloosterman sums over F q \mathbb{F}_q F q .
Lemma 2.1. [22 ] Let q = 2 m q=2^m q = 2 m and T i = { c ∈ F q : Tr m ( c ) = i } T_i=\{c\in\mathbb{F}_q:\operatorname{Tr}_m(c)=i\} T i = { c ∈ F q : Tr m ( c ) = i } , i = 0 , 1 i=0,1 i = 0 , 1 . The Kloosterman sums K m ( c ) K_m(c) K m ( c ) over F q \mathbb{F}_q F q satisfy the following identities
(i)
∑ c ∈ F q K m ( c ) = 0 , \sum_{c\in\mathbb{F}_q}K_m(c)=0, c ∈ F q ∑ K m ( c ) = 0 , (ii)
∑ c ∈ T i K m ( c ) = ( − 1 ) i + Tr m ( 1 ) q 2 , \sum_{c\in T_i}K_m(c)=(-1)^{i+\operatorname{Tr}_m(1)}\frac{q}{2}, c ∈ T i ∑ K m ( c ) = ( − 1 ) i + Tr m ( 1 ) 2 q , (iii)
∑ c ∈ F q χ ( c ) K m ( c ) = ( − 1 ) Tr m ( 1 ) q . \sum_{c\in\mathbb{F}_q}\chi(c)K_m(c)=(-1)^{\operatorname{Tr}_m(1)}q. c ∈ F q ∑ χ ( c ) K m ( c ) = ( − 1 ) Tr m ( 1 ) q .
Lemma 2.2. Let q = 2 m q=2^m q = 2 m for even m m m and ω \omega ω be a primitive root of F 4 \mathbb{F}_4 F 4 , namely, ω 2 + ω + 1 = 0 \omega^2+\omega+1=0 ω 2 + ω + 1 = 0 . Then the Kloosterman sums K m ( c ) K_m(c) K m ( c ) over F q \mathbb{F}_q F q satisfy the following identities:
(i)
∑ c ∈ F q ∗ K m ( 1 + c + c − 1 ) = K m ( ω ) 2 ; \sum_{c\in\mathbb{F}_q^*}K_m(1+c+c^{-1})=K_m(\omega)^2; c ∈ F q ∗ ∑ K m ( 1 + c + c − 1 ) = K m ( ω ) 2 ; (ii)
∑ c ∈ F q ∗ χ ( c − 1 ) K m ( c ) = K m ( 1 ) 2 − q ; \sum_{c\in\mathbb{F}_q^*}\chi(c^{-1})K_m(c)=K_m(1)^2-q; c ∈ F q ∗ ∑ χ ( c − 1 ) K m ( c ) = K m ( 1 ) 2 − q ; (iii)
∑ c ∈ F q ∗ χ ( c − 1 ) K m ( c + 1 ) = K m ( ω ) 2 ; \sum_{c\in\mathbb{F}_q^*}\chi(c^{-1})K_m(c+1)=K_m(\omega)^2; c ∈ F q ∗ ∑ χ ( c − 1 ) K m ( c + 1 ) = K m ( ω ) 2 ; (iv)
∑ c ∈ F q ∗ K m ( c 3 + c ) = ( 2 σ + 1 ) q + 1. \sum_{c\in\mathbb{F}_q^*}K_m(c^3+c)=(2\sigma+1)q+1. c ∈ F q ∗ ∑ K m ( c 3 + c ) = ( 2 σ + 1 ) q + 1.
Proof. These identities can be shown by routine calculations, and the proof is given in Appendix B for completeness.
Below we also recall some classical results about Kloosterman sums over F q \mathbb{F}_q F q [26 , 27 , 21 ] .
Lemma 2.3. [26 Sec. 6-7] Let q = 2 m q=2^m q = 2 m , k m = K m ( 1 ) k_m=K_m(1) k m = K m ( 1 ) , and for even m m m , let ℓ m = K m ( ω ) \ell_m=K_m(\omega) ℓ m = K m ( ω ) , where ω ∈ F q \omega\in\mathbb{F}_q ω ∈ F q satisfies ω 2 + ω + 1 = 0 \omega^2+\omega+1=0 ω 2 + ω + 1 = 0 . Then we have the following recurrences:
k 1 = 1 , k 2 = 3 , k j + 2 = − k j + 1 − 2 k j , j ≥ 1 ; k_1=1,\quad k_2=3,\quad k_{j+2}=-k_{j+1}-2k_j,\quad j\geq1; k 1 = 1 , k 2 = 3 , k j + 2 = − k j + 1 − 2 k j , j ≥ 1 ; ℓ 2 = − 1 , ℓ 4 = 7 , ℓ 2 ( j + 2 ) = ℓ 2 j − 4 ℓ 2 ( j + 1 ) , j ≥ 1. \ell_2=-1,\quad\ell_4=7,\quad\ell_{2(j+2)}=\ell_{2j}-4\ell_{2(j+1)},\quad j\geq1. ℓ 2 = − 1 , ℓ 4 = 7 , ℓ 2 ( j + 2 ) = ℓ 2 j − 4 ℓ 2 ( j + 1 ) , j ≥ 1.
Lemma 2.4. [27 ] For every c ∈ F q c\in\mathbb{F}_q c ∈ F q , one has ∣ K m ( c ) ∣ ≤ 2 q \lvert K_m(c)\rvert\leq2\sqrt{q} ∣ K m ( c )∣ ≤ 2 q .
Niho exponents and their associated key polynomials
Below we recall the well-known result on Niho exponents over finite fields of characteristic two, which will be referred to as Niho’s theorem [1 , 28 ] .
Lemma 2.5. [1 , 28 ] For a Niho exponent d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 with s s s modulo q + 1 q+1 q + 1 , the Walsh transform of the function Tr 2 m ( x d ) \operatorname{Tr}_{2^m}(x^d) Tr 2 m ( x d ) at a point a ∈ F q 2 a\in\mathbb{F}_{q^2} a ∈ F q 2 satisfies
W d ( a ) : = ∑ x ∈ F q 2 ( − 1 ) Tr 2 m ( x d + a x ) = q ( N s ( a ) − 1 ) . (5) W_d(a):=\sum_{x\in\mathbb{F}_{q^2}}(-1)^{\operatorname{Tr}_{2^m}(x^d+ax)}=q\left(N_s(a)-1\right). \tag*{(5)} W d ( a ) := x ∈ F q 2 ∑ ( − 1 ) Tr 2 m ( x d + a x ) = q ( N s ( a ) − 1 ) . (5) where N s ( a ) N_s(a) N s ( a ) is the number of distinct roots in U U U of the polynomial
f s , a ( x ) = x 2 s − 1 + x z s + a ‾ x s − 1 + 1. (6) f_{s,a}(x)=x^{2s-1}+xz^s+\overline{a}x^{s-1}+1. \tag*{(6)} f s , a ( x ) = x 2 s − 1 + x z s + a x s − 1 + 1. (6)
The polynomial f s , a ( x ) f_{s,a}(x) f s , a ( x ) will be termed the key polynomial associated with a normalized Niho exponent d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 . By Niho’s theorem, determining the value distribution of Walsh transforms of the Niho-type function Tr 2 m ( x d ) \operatorname{Tr}_{2^m}(x^d) Tr 2 m ( x d ) is equivalent to determining the possible numbers of roots of f s , a ( x ) f_{s,a}(x) f s , a ( x ) in U U U and the frequency of each number as a a a runs through F q 2 ∗ \mathbb{F}_{q^2}^{*} F q 2 ∗ . Notice that the polynomial f s , a ( x ) f_{s,a}(x) f s , a ( x ) is an instance of self-conjugate-reciprocal polynomials over F q 2 \mathbb{F}_{q^2} F q 2 [15 ] . Below we recall some basics of self-conjugate-reciprocal polynomials.
Self-conjugate-reciprocal polynomials over finite fields
The reciprocal polynomial of a polynomial f ( x ) = ∑ i = 0 d f i x i f(x)=\sum_{i=0}^{d}f_i x^i f ( x ) = ∑ i = 0 d f i x i in F q 2 [ x ] \mathbb{F}_{q^2}[x] F q 2 [ x ] is defined as f ∗ ( x ) : = x d f ( x − 1 ) f^*(x):=x^d f(x^{-1}) f ∗ ( x ) := x d f ( x − 1 ) and the conjugate polynomial of f ( x ) f(x) f ( x ) is defined as f ‾ ( x ) : = ∑ i = 0 d f i ‾ x i \overline{f}(x):=\sum_{i=0}^{d}\overline{f_i}x^i f ( x ) := ∑ i = 0 d f i x i , where f i ‾ \overline{f_i} f i is the conjugate of f i f_i f i in F q 2 \mathbb{F}_{q^2} F q 2 . The conjugate-reciprocal polynomial of f ( x ) f(x) f ( x ) is then defined as
f † ( x ) : = x d ∑ i = 0 d f i ‾ x − i = ∑ i = 0 d f i ‾ x d − i = ∑ i = 0 d f d − i ‾ x i . f^\dagger(x):=x^d\sum_{i=0}^{d}\overline{f_i}x^{-i}=\sum_{i=0}^{d}\overline{f_i}x^{d-i}=\sum_{i=0}^{d}\overline{f_{d-i}}x^i. f † ( x ) := x d i = 0 ∑ d f i x − i = i = 0 ∑ d f i x d − i = i = 0 ∑ d f d − i x i .
Definition 2.6. [15 ] A polynomial f ( x ) = ∑ i = 0 d f i x i f(x)=\sum_{i=0}^{d}f_i x^i f ( x ) = ∑ i = 0 d f i x i in F q 2 [ x ] \mathbb{F}_{q^2}[x] F q 2 [ x ] is self-conjugate-reciprocal (SCR) if f ( x ) = f † ( x ) f(x)=f^\dagger(x) f ( x ) = f † ( x ) , namely, its coefficients satisfy f i = f d − i ‾ f_i=\overline{f_{d-i}} f i = f d − i for 0 ≤ i ≤ d 0\leq i\leq d 0 ≤ i ≤ d . In particular, a monic SCR polynomial necessarily satisfies f d = f 0 = 1 f_d=f_0=1 f d = f 0 = 1 , and will be termed a normalized SCR polynomial .
For instance, normalized SCR polynomials of degrees two, three, and four have the following forms, respectively:
x 2 + α x + 1 , α ∈ F q , x^2+\alpha x+1,\qquad\alpha\in\mathbb{F}_q, x 2 + α x + 1 , α ∈ F q ,
x 3 + α x 2 + α ‾ x + 1 , α ∈ F q 2 , x^3+\alpha x^2+\overline{\alpha}x+1,\qquad\alpha\in\mathbb{F}_{q^2}, x 3 + α x 2 + α x + 1 , α ∈ F q 2 ,
x 4 + α x 3 + β x 2 + α ‾ x + 1 , α ∈ F q 2 , β ∈ F q . x^4+\alpha x^3+\beta x^2+\overline{\alpha}x+1,\qquad\alpha\in\mathbb{F}_{q^2},\ \beta\in\mathbb{F}_q. x 4 + α x 3 + β x 2 + α x + 1 , α ∈ F q 2 , β ∈ F q .
Note that SCR polynomials have some interesting properties. For an SCR polynomial f ( x ) f(x) f ( x ) and an element λ ∈ F q 2 ∗ \lambda\in\mathbb{F}_{q^2}^{*} λ ∈ F q 2 ∗ , it is readily seen that λ f ( x ) \lambda f(x) λ f ( x ) is SCR if and only if λ ∈ F q ∗ \lambda\in\mathbb{F}_{q}^{*} λ ∈ F q ∗ . The operators of reciprocal and conjugate on polynomials in F q 2 [ x ] \mathbb{F}_{q^2}[x] F q 2 [ x ] are involutions. The SCR property is preserved by polynomial multiplications. Namely, if f ( x ) f(x) f ( x ) and h ( x ) h(x) h ( x ) are SCR polynomials, then f ( x ) h ( x ) f(x)h(x) f ( x ) h ( x ) is an SCR polynomial, which is due to the following equalities:
( f h ) ∗ ( x ) = f ∗ ( x ) h ∗ ( x ) , f h ‾ ( x ) = f ‾ ( x ) h ‾ ( x ) , ( f ∗ ) ‾ ( x ) = ( f ‾ ) ∗ ( x ) . (fh)^{*}(x)=f^{*}(x)h^{*}(x),\qquad\overline{fh}(x)=\overline{f}(x)\,\overline{h}(x),\qquad\overline{(f^{*})}(x)=(\overline{f})^{*}(x). ( f h ) ∗ ( x ) = f ∗ ( x ) h ∗ ( x ) , f h ( x ) = f ( x ) h ( x ) , ( f ∗ ) ( x ) = ( f ) ∗ ( x ) .
where f h f h f h denotes the multiplication of f ( x ) f(x) f ( x ) and g ( x ) g(x) g ( x ) . If f ( x ) ∈ F q 2 [ x ] f(x) \in\mathbb{F}_{q^2}[x] f ( x ) ∈ F q 2 [ x ] is an SCR polynomial, then the following properties hold:
(i) f ∗ ( x ) = f ‾ ( x ) f^{*}(x)=\overline{f}(x) f ∗ ( x ) = f ( x ) , and both f ∗ ( x ) f^{*}(x) f ∗ ( x ) and f ‾ ( x ) \overline{f}(x) f ( x ) are SCR polynomials.
(ii) f ∗ ( x ) = f ( x ) f^{*}(x)=f(x) f ∗ ( x ) = f ( x ) if and only if f ‾ ( x ) = f ( x ) \overline{f}(x)=f(x) f ( x ) = f ( x ) , or equivalently, f ( x ) ∈ F q [ x ] f(x) \in\mathbb{F}_{q}[x] f ( x ) ∈ F q [ x ] .
For a given SCR polynomial, its roots in the algebraic closure has certain interesting structure.
Lemma 2.7. [15 ] Let f ( x ) ∈ F q 2 [ x ] f(x) \in\mathbb{F}_{q^2}[x] f ( x ) ∈ F q 2 [ x ] be an SCR polynomial of degree d ≥ 2 d \ge2 d ≥ 2 and let R R R be the set of all its roots in F q 2 \mathbb{F}_{q^2} F q 2 . Let Π \Pi Π be the map from F q 2 ∗ \mathbb{F}_{q^2}^{*} F q 2 ∗ to itself defined by Π ( a ) = a − q \Pi(a)=a^{-q} Π ( a ) = a − q . For a ∈ F q 2 ∗ a \in\mathbb{F}_{q^2}^{*} a ∈ F q 2 ∗ , let O Π ( a ) = { Π i ( a ) : i = 0 , 1 , … , k − 1 } O_{\Pi}(a)=\{\Pi^{i}(a):i=0,1,\ldots,k-1\} O Π ( a ) = { Π i ( a ) : i = 0 , 1 , … , k − 1 } be the orbit of a a a under Π \Pi Π , where k k k is the least positive integer such that Π k ( a ) = a \Pi^{k}(a)=a Π k ( a ) = a . Then we have the following statements:
(i) An element a ∈ F q 2 a \in\mathbb{F}_{q^2} a ∈ F q 2 is a root of f ( x ) f(x) f ( x ) if only if Π ( a ) \Pi(a) Π ( a ) is a root of f ( x ) f(x) f ( x ) ;
(ii) The root set R R R of f ( x ) f(x) f ( x ) can be represented as a disjoint union of distinct orbits O Π ( a ) O_{\Pi}(a) O Π ( a ) where each a a a is a root of f ( x ) f(x) f ( x ) , i.e., R = O Π ( a 1 ) ⊔ ⋯ ⊔ O Π ( a t ) R=O_{\Pi}(a_{1})\sqcup\cdots\sqcup O_{\Pi}(a_{t}) R = O Π ( a 1 ) ⊔ ⋯ ⊔ O Π ( a t ) , where a 1 , … , a t a_{1},\ldots,a_{t} a 1 , … , a t are distinct of f ( x ) f(x) f ( x ) and lie in different orbits;
(iii) The degree d d d of f ( x ) f(x) f ( x ) and the number t t t of distinct orbits in its root set R R R satisfy
( d + 1 2 ) + t ≡ Tr m ( S ) ( m o d 2 ) where S = ∑ u ≠ v u , v ∈ R u v ( u + v ) 2 . \binom{d+1}{2}+t\equiv\operatorname{Tr}_{m}(S)\pmod{2}\quad\text{where}\quad S=\sum_{\substack{u\ne v\\u,v\in R}}\frac{uv}{(u+v)^{2}}. ( 2 d + 1 ) + t ≡ Tr m ( S ) ( mod 2 ) where S = u = v u , v ∈ R ∑ ( u + v ) 2 uv .
Under an action Π \Pi Π defined above, it is clear that the orbit O Π ( a ) O_{\Pi}(a) O Π ( a ) has size one for a ∈ U a \in U a ∈ U and has size two a ∈ F q 2 ∗ ∖ U a \in\mathbb{F}_{q^2}^{*}\setminus U a ∈ F q 2 ∗ ∖ U . As a result, Lemma 2.7 can be used to characterize possible number of roots of f ( x ) f(x) f ( x ) in U U U for some cases, such as f ( x ) = x 7 + a x 4 + a ‾ x 3 + 1 f(x)=x^{7}+ax^{4}+\overline{a}x^{3}+1 f ( x ) = x 7 + a x 4 + a x 3 + 1 for the Niho decimation d = 4 q − 3 d=4q-3 d = 4 q − 3 [15 ] .
Recall that a polynomial f ( x ) ∈ F q 2 [ x ] f(x) \in\mathbb{F}_{q^2}[x] f ( x ) ∈ F q 2 [ x ] is said to split in a subset of F q 2 \mathbb{F}_{q^2} F q 2 if all the roots of f ( x ) f(x) f ( x ) belong to the set, where repeated roots are allowed. Here we are interested in the cases where polynomials in F q 2 [ x ] \mathbb{F}_{q^2}[x] F q 2 [ x ] split in the unit circle U U U . Suppose that a monic polynomial f ( x ) = x d + ∑ i = 0 d − 1 f i x i ∈ F q 2 [ x ] f(x)=x^{d}+\sum_{i=0}^{d-1}f_{i}x^{i}\in\mathbb{F}_{q^2}[x] f ( x ) = x d + ∑ i = 0 d − 1 f i x i ∈ F q 2 [ x ] splits in U U U , i.e., f ( x ) = ∏ i = 1 d ( x + u i ) f(x)=\prod_{i=1}^{d}(x+u_{i}) f ( x ) = ∏ i = 1 d ( x + u i ) with u i u_{i} u i ’s in U U U . Since u i ‾ = u i − 1 \overline{u_{i}}=u_{i}^{-1} u i = u i − 1 , we have
f † ( x ) = x d ∏ i = 1 d ( x − 1 + u i − 1 ) = ( ∏ i = 1 d u i − 1 ) ∏ i = 1 d ( x + u i ) = f 0 − 1 f ( x ) . f^{\dagger}(x)=x^{d}\prod_{i=1}^{d}(x^{-1}+u_{i}^{-1})=\left(\prod_{i=1}^{d}u_{i}^{-1}\right)\prod_{i=1}^{d}(x+u_{i})=f_{0}^{-1}f(x). f † ( x ) = x d i = 1 ∏ d ( x − 1 + u i − 1 ) = ( i = 1 ∏ d u i − 1 ) i = 1 ∏ d ( x + u i ) = f 0 − 1 f ( x ) .
Consequently, f ( x ) f(x) f ( x ) is a normalized SCR polynomial if and only if f 0 = 1 f_{0}=1 f 0 = 1 , or equivalently, ∏ i = 1 d u i = 1 \prod_{i=1}^{d}u_{i}=1 ∏ i = 1 d u i = 1 . More generally, when f ( 0 ) ≠ 1 f(0)\ne1 f ( 0 ) = 1 , if there exists an element μ ∈ U \mu\in U μ ∈ U satisfying μ d = f ( 0 ) \mu^{d}=f(0) μ d = f ( 0 ) , e.g., when gcd ( d , q + 1 ) = 1 \gcd(d,q+1)=1 g cd( d , q + 1 ) = 1 , one can transform f ( x ) f(x) f ( x ) to a normalized SCR polynomial as follows:
f ( x ) ↦ μ − d f ( μ x ) = x d + ∑ 1 ≤ i < d μ i f i x i + 1. f(x)\mapsto\mu^{-d}f(\mu x)=x^{d}+\sum_{1\le i<d}\mu^{i}f_{i}x^{i}+1. f ( x ) ↦ μ − d f ( μx ) = x d + 1 ≤ i < d ∑ μ i f i x i + 1.
Thus we have the following result.
Lemma 2.8. Suppose a monic polynomial f ( x ) = x d + f d − 1 x d − 1 + ⋯ + f 1 x + f 0 ∈ F q 2 [ x ] f(x)=x^d+f_{d-1}x^{d-1}+\cdots+f_1x+f_0\in\mathbb{F}_{q^2}[x] f ( x ) = x d + f d − 1 x d − 1 + ⋯ + f 1 x + f 0 ∈ F q 2 [ x ] splits in U U U and there exists an element μ ∈ U \mu\in U μ ∈ U satisfying μ d = f ( 0 ) \mu^d=f(0) μ d = f ( 0 ) . Then the polynomial μ − d f ( μ x ) \mu^{-d}f(\mu x) μ − d f ( μx ) is a normalized SCR polynomial.
A natural question arises here: when does a normalized SCR polynomial split in U U U ? The next section is devoted to the study of low-degree normalized SCR polynomials that split in U U U . The resulting criteria will be used to analyze the hexic polynomial P a ( x ) P_a(x) P a ( x ) derived from the Niho exponent d = 5 q − 1 d=5q-1 d = 5 q − 1 .
Low-degree self-conjugate-reciprocal polynomials splitting in the unit circle
This section characterizes the condition under which a low-degree SCR polynomial f ( x ) f(x) f ( x ) in F q 2 [ x ] \mathbb{F}_{q^2}[x] F q 2 [ x ] splits in the unit circle U U U of F q 2 \mathbb{F}_{q^2} F q 2 . We will frequently use the notation
H 1 = { w ∈ F q ∗ : Tr m ( 1 / w ) = 1 } . (7) H_1=\{w\in\mathbb{F}_q^*:\operatorname{Tr}_m(1/w)=1\}. \tag*{(7)} H 1 = { w ∈ F q ∗ : Tr m ( 1/ w ) = 1 } . (7)
It was pointed out [4 Prop. 7] that H 1 = { u + u ˉ : u ∈ U ∖ { 1 } } H_1=\{u+\bar{u}:u\in U\setminus\{1\}\} H 1 = { u + u ˉ : u ∈ U ∖ { 1 }} , which can easily verified by considering the roots of the quadratic polynomial X 2 + w X + 1 X^2+wX+1 X 2 + w X + 1 over F q \mathbb{F}_q F q .
Below we start with a method of transforming a normalized SCR polynomial in F q 2 [ x ] \mathbb{F}_{q^2}[x] F q 2 [ x ] to a polynomial in F q [ x ] \mathbb{F}_q[x] F q [ x ] and the relation between their roots.
Proposition 3.1. Let f ( x ) = ∑ i = 0 d f i x i ∈ F q 2 [ x ] f(x)=\sum_{i=0}^{d}f_i x^i\in\mathbb{F}_{q^2}[x] f ( x ) = ∑ i = 0 d f i x i ∈ F q 2 [ x ] be a normalized SCR polynomial of degree d ≥ 2 d\ge2 d ≥ 2 and denote
s k = ∑ i = k d f i f ˉ i − k , 0 ≤ k ≤ d . (8) s_k=\sum_{i=k}^{d}f_i\bar{f}_{i-k},\qquad0\le k\le d. \tag*{(8)} s k = i = k ∑ d f i f ˉ i − k , 0 ≤ k ≤ d . (8) Then each s k s_k s k lies in F q \mathbb{F}_q F q and the polynomial g ( y ) : = s 0 + ∑ k = 1 d s k D k ( y ) g(y):=s_0+\sum_{k=1}^{d}s_kD_k(y) g ( y ) := s 0 + ∑ k = 1 d s k D k ( y ) , where D k ( y ) D_k(y) D k ( y ) is the first-kind Dickson polynomial, satisfies
x d g ( x + x − 1 ) = f ( x ) f ( x ) ‾ . (9) x^d g(x+x^{-1})=f(x)\overline{f(x)}. \tag*{(9)} x d g ( x + x − 1 ) = f ( x ) f ( x ) . (9) Moreover, the following properties hold.
(i) If x 1 , … , x d x_1,\ldots,x_d x 1 , … , x d are the roots of f ( x ) f(x) f ( x ) in F ‾ q 2 \overline{\mathbb{F}}_{q^2} F q 2 , counted with multiplicity, then
g ( y ) = ∏ i = 1 d ( y + x i + x i − 1 ) . g(y)=\prod_{i=1}^{d}(y+x_i+x_i^{-1}). g ( y ) = i = 1 ∏ d ( y + x i + x i − 1 ) . (ii) Denote by m f ( u ) m_f(u) m f ( u ) the multiplicity of u u u as a root of f ( x ) f(x) f ( x ) in U U U with m f ( u ) = 0 m_f(u)=0 m f ( u ) = 0 if f ( u ) ≠ 0 f(u)\ne0 f ( u ) = 0 , and m g ( t ) m_g(t) m g ( t ) the multiplicity of a root t t t of g ( y ) g(y) g ( y ) in F q \mathbb{F}_q F q . Then the map u ↦ u + u − 1 = t u\mapsto u+u^{-1}=t u ↦ u + u − 1 = t gives the following correspondence between the roots of f ( x ) f(x) f ( x ) in U U U and those of g ( y ) g(y) g ( y ) in { 0 } ∪ H 1 \{0\}\cup H_1 { 0 } ∪ H 1 .
(a) f ( 1 ) = 0 f(1)=0 f ( 1 ) = 0 if and only if g ( 0 ) = 0 g(0)=0 g ( 0 ) = 0 ; moreover, m g ( 0 ) = m f ( 1 ) m_g(0)=m_f(1) m g ( 0 ) = m f ( 1 ) .
(b) Each root u ∈ U ∖ { 1 } u\in U\setminus\{1\} u ∈ U ∖ { 1 } of f ( x ) f(x) f ( x ) gives a root t = u + u − 1 ∈ H 1 t=u+u^{-1}\in H_1 t = u + u − 1 ∈ H 1 of g ( y ) g(y) g ( y ) . Conversely, for each root w ∈ H 1 w\in H_1 w ∈ H 1 of g g g , let u , u − 1 ∈ U ∖ { 1 } u,u^{-1}\in U\setminus\{1\} u , u − 1 ∈ U ∖ { 1 } be the two distinct roots of X 2 + w X + 1 X^2+wX+1 X 2 + w X + 1 ; then u , u − 1 u,u^{-1} u , u − 1 are roots of f ( x ) f(x) f ( x ) , and m g ( t ) = m f ( u ) + m f ( u − 1 ) m_g(t)=m_f(u)+m_f(u^{-1}) m g ( t ) = m f ( u ) + m f ( u − 1 ) .
Proof. By the SCR condition of f ( x ) f(x) f ( x ) , namely, f i = f ‾ d − i f_i=\overline{f}_{d-i} f i = f d − i for 0 ≤ i ≤ d 0\le i\le d 0 ≤ i ≤ d , one has
s ‾ k = ∑ i = k d f ‾ i f ‾ i − k = ∑ i = k d f d − i f d − i + k = ∑ j = 0 d − k f j f j + k = s k , \overline{s}_k=\sum_{i=k}^{d}\overline{f}_i\overline{f}_{i-k}=\sum_{i=k}^{d}f_{d-i}f_{d-i+k}=\sum_{j=0}^{d-k}f_jf_{j+k}=s_k, s k = i = k ∑ d f i f i − k = i = k ∑ d f d − i f d − i + k = j = 0 ∑ d − k f j f j + k = s k ,
implying that s k ∈ F q s_k\in\mathbb{F}_q s k ∈ F q . Note that the first-kind Dickson polynomial satisfy D k ( x + x − 1 ) = x k + x − k D_k(x+x^{-1})=x^k+x^{-k} D k ( x + x − 1 ) = x k + x − k for k ≥ 0 k\ge0 k ≥ 0 . Thus,
x d g ( x + x − 1 ) = x d ( s 0 + ∑ k = 1 d s k ( x k + x − k ) ) . x^d g(x+x^{-1})=x^d\left(s_0+\sum_{k=1}^{d}s_k(x^k+x^{-k})\right). x d g ( x + x − 1 ) = x d ( s 0 + k = 1 ∑ d s k ( x k + x − k ) ) .
On the other hand, for the normalized SCR polynomial f ( x ) f(x) f ( x ) , one has
f ‾ ( x ) = f ∗ ( x ) = x d f ( 1 / x ) , \overline{f}(x)=f^*(x)=x^d f(1/x), f ( x ) = f ∗ ( x ) = x d f ( 1/ x ) ,
and then
f ( x ) f ‾ ( x ) = x d f ( x ) f ( x − 1 ) = x d ( s 0 + ∑ k = 1 d s k ( x k + x − k ) ) = x d g ( x + x − 1 ) . f(x)\overline{f}(x)=x^d f(x)f(x^{-1})=x^d\left(s_0+\sum_{k=1}^{d}s_k(x^k+x^{-k})\right)=x^d g(x+x^{-1}). f ( x ) f ( x ) = x d f ( x ) f ( x − 1 ) = x d ( s 0 + k = 1 ∑ d s k ( x k + x − k ) ) = x d g ( x + x − 1 ) .
(i). Let x 1 , … , x d x_1,\ldots,x_d x 1 , … , x d are the roots of f ( x ) f(x) f ( x ) in an algebraic closure of F q 2 \mathbb{F}_{q^2} F q 2 , counted with multiplicity. Then x i x_i x i , 1 ≤ i ≤ d 1\le i\le d 1 ≤ i ≤ d , are nonzero and satisfy x 1 ⋯ x d = 1 x_1\cdots x_d=1 x 1 ⋯ x d = 1 , and 1 / x i 1/x_i 1/ x i , 1 ≤ i ≤ d 1\le i\le d 1 ≤ i ≤ d , are root of f ∗ f^* f ∗ . Consequently,
f ‾ ( x ) = x d f ( x − 1 ) = ∏ i = 1 d ( x + x i − 1 ) . \overline{f}(x)=x^d f(x^{-1})=\prod_{i=1}^{d}(x+x_i^{-1}). f ( x ) = x d f ( x − 1 ) = i = 1 ∏ d ( x + x i − 1 ) .
It follows that
f ( x ) f ‾ ( x ) = ∏ i = 1 d ( x + x i ) ( x + x i − 1 ) = ∏ i = 1 d ( x 2 + ( x i + x i − 1 ) x + 1 ) = x d ∏ i = 1 d ( x + x − 1 + x i + x i − 1 ) . \begin{aligned}
f(x)\overline{f}(x)=\prod_{i=1}^{d}(x+x_i)(x+x_i^{-1}) \\
&=\prod_{i=1}^{d}\left(x^2+(x_i+x_i^{-1})x+1\right) \\
&=x^d\prod_{i=1}^{d}\left(x+x^{-1}+x_i+x_i^{-1}\right).
\end{aligned} f ( x ) f ( x ) = i = 1 ∏ d ( x + x i ) ( x + x i − 1 ) = i = 1 ∏ d ( x 2 + ( x i + x i − 1 ) x + 1 ) = x d i = 1 ∏ d ( x + x − 1 + x i + x i − 1 ) .
Thus, over an algebraic closure of F q 2 \mathbb{F}_{q^2} F q 2 , we have
g ( y ) = ∏ i = 1 d ( y + x i + x i − 1 ) . g(y)=\prod_{i=1}^{d}(y+x_i+x_i^{-1}). g ( y ) = i = 1 ∏ d ( y + x i + x i − 1 ) .
(ii). Suppose first that f ( x ) f(x) f ( x ) has a root x i ∈ U x_i\in U x i ∈ U . Then x i q = x i − 1 x_i^q=x_i^{-1} x i q = x i − 1 , so t i = x i + x i − 1 ∈ F q t_i=x_i+x_i^{-1}\in\mathbb{F}_q t i = x i + x i − 1 ∈ F q . If x i = 1 x_i=1 x i = 1 , then t i = 0 t_i=0 t i = 0 . If x i ∈ U ∖ { 1 } x_i\in U\setminus\{1\} x i ∈ U ∖ { 1 } , then t i = x i + x i − 1 ∈ F q ∗ t_i=x_i+x_i^{-1}\in\mathbb{F}_q^* t i = x i + x i − 1 ∈ F q ∗ . Using x i q = x i − 1 x_i^q=x_i^{-1} x i q = x i − 1 , we obtain
Tr m ( t i − 1 ) = Tr m ( x i ( 1 + x i ) 2 ) = ∑ j = 0 m − 1 ( 1 1 + x i + 1 ( 1 + x i ) 2 ) 2 j = 1 1 + x i + 1 ( 1 + x i ) q = 1 1 + x i + 1 1 + x i − 1 = 1. \begin{aligned}
\operatorname{Tr}_m(t_i^{-1})=\operatorname{Tr}_m\left(\frac{x_i}{(1+x_i)^2}\right) \\
&=\sum_{j=0}^{m-1}\left(\frac{1}{1+x_i}+\frac{1}{(1+x_i)^2}\right)^{2^j} \\
&=\frac{1}{1+x_i}+\frac{1}{(1+x_i)^q} \\
&=\frac{1}{1+x_i}+\frac{1}{1+x_i^{-1}}=1.
\end{aligned} Tr m ( t i − 1 ) = Tr m ( ( 1 + x i ) 2 x i ) = j = 0 ∑ m − 1 ( 1 + x i 1 + ( 1 + x i ) 2 1 ) 2 j = 1 + x i 1 + ( 1 + x i ) q 1 = 1 + x i 1 + 1 + x i − 1 1 = 1.
Thus, t i ∈ H 1 t_i \in H_1 t i ∈ H 1 and g g g has a root in H 1 H_1 H 1 .
Conversely, suppose that g ( y ) g(y) g ( y ) has a root t i ∈ { 0 } ∪ H 1 ⊂ F q t_i \in\{0\} \cup H_1 \subset\mathbb{F}_q t i ∈ { 0 } ∪ H 1 ⊂ F q . Part (ii) shows that in the algebraic closure of F q 2 \mathbb{F}_{q^2} F q 2 there exists a root x i x_i x i of f f f such that t i = x i + x i − 1 t_i=x_i+x_i^{-1} t i = x i + x i − 1 . If t i = 0 t_i=0 t i = 0 , then x i = 1 ∈ U x_i=1\in U x i = 1 ∈ U . If t i ≠ 0 t_i\ne0 t i = 0 , the trace condition Tr m ( t i − 1 ) = 1 \operatorname{Tr}_m(t_i^{-1})=1 Tr m ( t i − 1 ) = 1 implies that X 2 + t i X + 1 X^2+t_iX+1 X 2 + t i X + 1 has two distinct roots u , u − 1 ∈ U ∖ { 1 } u,u^{-1}\in U\setminus\{1\} u , u − 1 ∈ U ∖ { 1 } . Since x i x_i x i satisfies this quadratic equation, we have x i ∈ { u , u − 1 } ⊂ U ∖ { 1 } x_i\in\{u,u^{-1}\}\subset U\setminus\{1\} x i ∈ { u , u − 1 } ⊂ U ∖ { 1 } . Thus, in this case f f f has a root in U U U , which is different from 1. This proves (iii).
The multiplicity statement follows from the fact that the correspondence t = u + u − 1 t=u+u^{-1} t = u + u − 1 preserves multiplicities between the roots u ∈ U u\in U u ∈ U of f f f and the corresponding roots t ∈ H 1 ∪ { 0 } t\in H_1\cup\{0\} t ∈ H 1 ∪ { 0 } of g g g . □ \square □
Proposition 3.1 provides a handy tool to transform the roots of a normalized SCR polynomial f ( x ) f(x) f ( x ) in U U U to that of g ( y ) g(y) g ( y ) in F q \mathbb{F}_q F q , which was well studied for low-degree polynomials. By Proposition 3.1, we immediately have the following criterion for a quadratic polynomial to split in U U U .
Corollary 3.2. A quadratic normalized SCR polynomial f ( x ) = x 2 + α x + 1 ∈ F q 2 [ x ] f(x)=x^2+\alpha x+1\in\mathbb{F}_{q^2}[x] f ( x ) = x 2 + α x + 1 ∈ F q 2 [ x ] splits in U U U if and only if either α = 0 \alpha=0 α = 0 or α ∈ H 1 \alpha\in H_1 α ∈ H 1 . In the latter case, its roots are two distinct elements u , u − 1 ∈ U ∖ { 1 } u,u^{-1}\in U\setminus\{1\} u , u − 1 ∈ U ∖ { 1 } .
More generally, a quadratic polynomial f ( x ) = x 2 + α x + β ∈ F q 2 [ x ] f(x)=x^2+\alpha x+\beta\in\mathbb{F}_{q^2}[x] f ( x ) = x 2 + α x + β ∈ F q 2 [ x ] with α ≠ 0 \alpha\ne0 α = 0 has two distinct roots in U U U if and only if β = α α ˉ \beta=\frac{\alpha}{\bar{\alpha}} β = α ˉ α and α α ˉ ∈ H 1 \alpha\bar{\alpha}\in H_1 α α ˉ ∈ H 1 . Both roots belong to U ∖ { 1 } U\setminus\{1\} U ∖ { 1 } if and only if, in addition, 1 + α + β ≠ 0 1+\alpha+\beta\ne0 1 + α + β = 0 .
Proof. For a normalized SCR polynomial f ( x ) = x 2 + α x + 1 f(x)=x^2+\alpha x+1 f ( x ) = x 2 + α x + 1 , it is clear that α ∈ F q \alpha\in\mathbb{F}_q α ∈ F q , and the associated polynomial of f ( x ) f(x) f ( x ) is g ( y ) = y 2 + α 2 g(y)=y^2+\alpha^2 g ( y ) = y 2 + α 2 , which is computed by (8) and (9) . The root of g ( y ) g(y) g ( y ) is α \alpha α with multiplicity two, which belongs to { 0 } ∪ H 1 \{0\}\cup H_1 { 0 } ∪ H 1 if and only if α = 0 \alpha=0 α = 0 or α ∈ H 1 \alpha\in H_1 α ∈ H 1 . By Proposition 3.1, we get the desired result.
For a general quadratic polynomial f ( x ) = x 2 + α x + β ∈ F q 2 [ x ] f(x)=x^2+\alpha x+\beta\in\mathbb{F}_{q^2}[x] f ( x ) = x 2 + α x + β ∈ F q 2 [ x ] with α ≠ 0 \alpha\ne0 α = 0 , it has two distinct roots in U U U if and only if μ − 2 f ( μ x ) = x 2 + α μ x + 1 \mu^{-2}f(\mu x)=x^2+\frac{\alpha}{\mu}x+1 μ − 2 f ( μx ) = x 2 + μ α x + 1 is a normalized SCR polynomial having two roots in U U U , where μ 2 = β ∈ U \mu^2=\beta\in U μ 2 = β ∈ U . The latter holds if and only if α μ ∈ H 1 \frac{\alpha}{\mu}\in H_1 μ α ∈ H 1 . This implies that ( α μ ) q = α ˉ μ 2 \left(\frac{\alpha}{\mu}\right)^q=\frac{\bar{\alpha}}{\mu^2} ( μ α ) q = μ 2 α ˉ , i.e., μ 2 = β = α α ˉ \mu^2=\beta=\frac{\alpha}{\bar{\alpha}} μ 2 = β = α ˉ α . Note that α μ ∈ H 1 \frac{\alpha}{\mu}\in H_1 μ α ∈ H 1 if and only if α μ 2 = α α ˉ ∈ H 1 \frac{\alpha}{\mu^2}=\alpha\bar{\alpha}\in H_1 μ 2 α = α α ˉ ∈ H 1 . This completes the proof. □ \square □
Corollary 3.3. Let f ( x ) = x 3 + α x 2 + α ˉ x + 1 f(x)=x^3+\alpha x^2+\bar{\alpha}x+1 f ( x ) = x 3 + α x 2 + α ˉ x + 1 in F q 2 [ x ] \mathbb{F}_{q^2}[x] F q 2 [ x ] with α ≠ 0 \alpha\ne0 α = 0 . Then f ( x ) f(x) f ( x ) splits in U U U if and only if the cubic polynomial g ( y ) = y 3 + ( α + α ˉ ) y 2 + ( α + 1 ) ( α ˉ + 1 ) y + ( α + α ˉ ) 2 ∈ F q [ y ] g(y)=y^3+(\alpha+\bar{\alpha})y^2+(\alpha+1)(\bar{\alpha}+1)y+(\alpha+\bar{\alpha})^2\in\mathbb{F}_q[y] g ( y ) = y 3 + ( α + α ˉ ) y 2 + ( α + 1 ) ( α ˉ + 1 ) y + ( α + α ˉ ) 2 ∈ F q [ y ] splits in F q \mathbb{F}_q F q and all nonzero roots of g ( y ) g(y) g ( y ) lie in H 1 H_1 H 1 . Furthermore, f ( x ) f(x) f ( x ) has exactly three distinct roots in U U U if and only if one of the following holds:
α ∈ F q \alpha\in\mathbb{F}_q α ∈ F q and α + 1 ∈ H 1 \alpha+1\in H_1 α + 1 ∈ H 1 . In this case, f ( x ) = ( x + 1 ) ( x + u ) ( x + u − 1 ) f(x)=(x+1)(x+u)(x+u^{-1}) f ( x ) = ( x + 1 ) ( x + u ) ( x + u − 1 ) for some u ∈ U ∖ { 1 } u\in U\setminus\{1\} u ∈ U ∖ { 1 } satisfying α + 1 = u + u − 1 \alpha+1=u+u^{-1} α + 1 = u + u − 1 .
α ∉ F q ∪ U \alpha\notin\mathbb{F}_q\cup U α ∈ / F q ∪ U and g ( y ) g(y) g ( y ) has three distinct roots in H 1 H_1 H 1 .
Proof. We only need to calculate s k s_k s k and D k ( y ) D_k(y) D k ( y ) for k = 0 , 1 , 2 , 3 k=0,1,2,3 k = 0 , 1 , 2 , 3 as follows: s 0 = ( α + α ˉ ) 2 s_0=(\alpha+\bar{\alpha})^2 s 0 = ( α + α ˉ ) 2 , s 1 = α ˉ + α α ˉ + α s_1=\bar{\alpha}+\alpha\bar{\alpha}+\alpha s 1 = α ˉ + α α ˉ + α , s 2 = α + α ˉ s_2=\alpha+\bar{\alpha} s 2 = α + α ˉ , s 3 = 1 s_3=1 s 3 = 1 ; D 1 ( y ) = y D_1(y)=y D 1 ( y ) = y , D 2 ( y ) = y 2 D_2(y)=y^2 D 2 ( y ) = y 2 , D 3 ( y ) = y 3 + y D_3(y)=y^3+y D 3 ( y ) = y 3 + y . Thus,
g ( y ) = s 0 + ∑ k = 1 3 s k D k ( y ) = y 3 + T y 2 + ( T + N + 1 ) y + T 2 , (10) g(y)=s_0+\sum_{k=1}^{3}s_kD_k(y)=y^3+Ty^2+(T+N+1)y+T^2, \tag*{(10)} g ( y ) = s 0 + k = 1 ∑ 3 s k D k ( y ) = y 3 + T y 2 + ( T + N + 1 ) y + T 2 , (10)
where T = α + α ˉ T=\alpha+\bar{\alpha} T = α + α ˉ and N = α α ˉ N=\alpha\bar{\alpha} N = α α ˉ . It follows from Proposition 3.1 that f ( x ) f(x) f ( x ) splits in U U U if and only if g ( y ) g(y) g ( y ) splits in F q \mathbb{F}_q F q . If f ( x ) f(x) f ( x ) has three distinct roots in U U U , then it has no repeated roots. Thus, f ( x ) f(x) f ( x ) and f ′ ( x ) = x 2 + α ˉ f'(x)=x^2+\bar{\alpha} f ′ ( x ) = x 2 + α ˉ have no common root, which implies that f ( α ˉ 2 ) = α α ˉ + 1 = N + 1 ≠ 0 f(\bar{\alpha}^2)=\alpha\bar{\alpha}+1=N+1\ne0 f ( α ˉ 2 ) = α α ˉ + 1 = N + 1 = 0 . In this situation, g ( x ) g(x) g ( x ) can split in F q \mathbb{F}_q F q . Let w 1 , w 2 , w 3 w_1,w_2,w_3 w 1 , w 2 , w 3 be the roots of g ( y ) g(y) g ( y ) given in (10) . When α ∈ F q \alpha\in\mathbb{F}_q α ∈ F q , we have f ( x ) = ( x + 1 ) ( x 2 + ( α + 1 ) x + 1 ) f(x)=(x+1)(x^2+(\alpha+1)x+1) f ( x ) = ( x + 1 ) ( x 2 + ( α + 1 ) x + 1 ) , requiring that α + 1 ∈ H 1 \alpha+1\in H_1 α + 1 ∈ H 1 . To ensure T ≠ 0 T\ne0 T = 0 , it requires that α ∉ U \alpha\notin U α ∈ / U . When α ∉ F q ∪ U \alpha\notin\mathbb{F}_q\cup U α ∈ / F q ∪ U , we have T ≠ 0 T\ne0 T = 0 . Then, w 1 , w 2 , w 3 w_1,w_2,w_3 w 1 , w 2 , w 3 are nonzero, and it follows from
∏ 1 ≤ i < j ≤ 3 ( w i + w j ) = w 1 w 2 w 3 + ( ∑ i = 1 3 w i ) ( ∑ 1 ≤ i < j ≤ 3 w i w j ) = T ( N + 1 ) ≠ 0 \prod_{1\le i<j\le3}(w_i+w_j)=w_1w_2w_3+\left(\sum_{i=1}^{3}w_i\right)\left(\sum_{1\le i<j\le3}w_iw_j\right)=T(N+1)\ne0 1 ≤ i < j ≤ 3 ∏ ( w i + w j ) = w 1 w 2 w 3 + ( i = 1 ∑ 3 w i ) ( 1 ≤ i < j ≤ 3 ∑ w i w j ) = T ( N + 1 ) = 0
that w 1 , w 2 , w 3 w_1,w_2,w_3 w 1 , w 2 , w 3 are distinct, with each w i = u i + u i − 1 w_i=u_i+u_i^{-1} w i = u i + u i − 1 for some u i ∈ U ∖ { 1 } u_i\in U\setminus\{1\} u i ∈ U ∖ { 1 } that is a root of f ( x ) f(x) f ( x ) . Thus, w i w_i w i lies in H 1 H_1 H 1 , i ∈ { 1 , 2 , 3 } i\in\{1,2,3\} i ∈ { 1 , 2 , 3 } .
Given a normalized SCR polynomial in F q 2 [ x ] \mathbb{F}_{q^2}[x] F q 2 [ x ] with higher degree that split in U U U , one can similarly associate it with a polynomial g ( y ) g(y) g ( y ) in F q [ y ] \mathbb{F}_q[y] F q [ y ] , of which all nonzero roots lie in H 1 H_1 H 1 . For example, for the quartic normalized SCR polynomial f ( x ) = x 4 + α x 3 + β x 2 + α ˉ x + 1 f(x)=x^4+\alpha x^3+\beta x^2+\bar{\alpha}x+1 f ( x ) = x 4 + α x 3 + β x 2 + α ˉ x + 1 , where β ∈ F q \beta\in\mathbb{F}_q β ∈ F q , by Proposition 3.1, the associated polynomial g ( y ) g(y) g ( y ) is given by
g ( y ) = y 4 + T y 3 + N y 2 + T β y + T 2 + β 2 , with T = α + α ˉ , N = α α ˉ . (11) g(y)=y^4+Ty^3+Ny^2+T\beta y+T^2+\beta^2,\qquad\text{with }T=\alpha+\bar{\alpha},\quad N=\alpha\bar{\alpha}. \tag*{(11)} g ( y ) = y 4 + T y 3 + N y 2 + T β y + T 2 + β 2 , with T = α + α ˉ , N = α α ˉ . (11)
Consequently, f ( x ) f(x) f ( x ) splits in U U U if and only if g ( y ) g(y) g ( y ) splits in F q \mathbb{F}_q F q and all its nonzero roots lies in H 1 H_1 H 1 . Nevertheless, investigating whether a quartic polynomial splits in H 1 H_1 H 1 is a nontrivial task. Below we consider another transformation for this case, namely, the cubic resolvent of a quartic normalized SCR polynomial, which facilitates the study of key polynomials associated with the Niho exponent d = 5 q − 1 d=5q-1 d = 5 q − 1 .
Proposition 3.4. Consider a quartic polynomial f ( x ) = x 4 + α x 3 + β x 2 + α ˉ x + 1 ∈ F q 2 [ x ] f(x)=x^4+\alpha x^3+\beta x^2+\bar{\alpha}x+1\in\mathbb{F}_{q^2}[x] f ( x ) = x 4 + α x 3 + β x 2 + α ˉ x + 1 ∈ F q 2 [ x ] , where β ∈ F q \beta\in\mathbb{F}_q β ∈ F q . Define its cubic resolvent by
h ( y ) = y 3 + β y 2 + α α ˉ y + ( α + α ˉ ) 2 ∈ F q [ y ] . h(y)=y^3+\beta y^2+\alpha\bar{\alpha}y+(\alpha+\bar{\alpha})^2\in\mathbb{F}_q[y]. h ( y ) = y 3 + β y 2 + α α ˉ y + ( α + α ˉ ) 2 ∈ F q [ y ] . Then the following results hold:
(i) f ( x ) f(x) f ( x ) splits in U U U if and only if all roots of h ( y ) h(y) h ( y ) belong to { 0 } ∪ H 1 \{0\}\cup H_1 { 0 } ∪ H 1 .
(ii) f ( x ) f(x) f ( x ) has no repeated roots if and only if h ( y ) h(y) h ( y ) has no repeated roots, and these equivalent conditions hold precisely when β α α ˉ + ( α + α ˉ ) 2 ≠ 0 \beta\alpha\bar{\alpha}+(\alpha+\bar{\alpha})^2\ne0 β α α ˉ + ( α + α ˉ ) 2 = 0 .
Consequently, f ( x ) f(x) f ( x ) has four distinct roots in U U U if and only if h ( y ) h(y) h ( y ) has three distinct roots in { 0 } ∪ H 1 \{0\}\cup H_1 { 0 } ∪ H 1 , where h ( 0 ) ≠ 0 h(0)\ne0 h ( 0 ) = 0 when α ∉ F q \alpha\notin\mathbb{F}_q α ∈ / F q .
Proof. Suppose f ( x ) = ∏ i = 1 4 ( x + u i ) f(x)=\prod_{i=1}^{4}(x+u_i) f ( x ) = ∏ i = 1 4 ( x + u i ) with u i u_i u i in an algebraic closure of F q 2 \mathbb{F}_{q^2} F q 2 , counted with multiplicity. It is clear that
∑ 1 ≤ i ≤ 4 u i = α , ∑ 1 ≤ i < j ≤ 4 u i u j = β , ∑ 1 ≤ i < j < k ≤ 4 u i u j u k = α ˉ , u 1 u 2 u 3 u 4 = 1. \sum_{1\le i\le4}u_i=\alpha,\qquad\sum_{1\le i<j\le4}u_iu_j=\beta,\qquad\sum_{1\le i<j<k\le4}u_iu_ju_k=\bar{\alpha},\qquad u_1u_2u_3u_4=1. 1 ≤ i ≤ 4 ∑ u i = α , 1 ≤ i < j ≤ 4 ∑ u i u j = β , 1 ≤ i < j < k ≤ 4 ∑ u i u j u k = α ˉ , u 1 u 2 u 3 u 4 = 1.
Set w 1 = u 1 u 2 + u 3 u 4 w_1=u_1u_2+u_3u_4 w 1 = u 1 u 2 + u 3 u 4 , w 2 = u 1 u 3 + u 2 u 4 w_2=u_1u_3+u_2u_4 w 2 = u 1 u 3 + u 2 u 4 , w 3 = u 1 u 4 + u 2 u 3 w_3=u_1u_4+u_2u_3 w 3 = u 1 u 4 + u 2 u 3 . Then we have
w 1 + w 2 + w 3 = ∑ 1 ≤ i < j ≤ 4 u i u j = β , w_1+w_2+w_3=\sum_{1\le i<j\le4}u_iu_j=\beta, w 1 + w 2 + w 3 = 1 ≤ i < j ≤ 4 ∑ u i u j = β ,
∑ 1 ≤ i < j ≤ 3 w i w j = ( ∑ i = 1 4 u i ) ( ∑ 1 ≤ i < j < k ≤ 4 u i u j u k ) = α α ˉ , \sum_{1\le i<j\le3}w_iw_j=\left(\sum_{i=1}^{4}u_i\right)\left(\sum_{1\le i<j<k\le4}u_iu_ju_k\right)=\alpha\bar{\alpha}, 1 ≤ i < j ≤ 3 ∑ w i w j = ( i = 1 ∑ 4 u i ) 1 ≤ i < j < k ≤ 4 ∑ u i u j u k = α α ˉ ,
w 1 w 2 w 3 = ( ∏ i = 1 4 u i ) ( ∑ i = 1 4 u i 2 ) + ( ∑ 1 ≤ j < k < ℓ ≤ 4 u j u k u ℓ ) 2 = α 2 + α ˉ 2 . w_1w_2w_3=\left(\prod_{i=1}^{4}u_i\right)\left(\sum_{i=1}^{4}u_i^2\right)+\left(\sum_{1\le j<k<\ell\le4}u_ju_ku_\ell\right)^2=\alpha^2+\bar{\alpha}^{2}. w 1 w 2 w 3 = ( i = 1 ∏ 4 u i ) ( i = 1 ∑ 4 u i 2 ) + 1 ≤ j < k < ℓ ≤ 4 ∑ u j u k u ℓ 2 = α 2 + α ˉ 2 .
This gives the following cubic resolvent of f ( x ) f(x) f ( x ) :
h ( y ) = ( y + w 1 ) ( y + w 2 ) ( y + w 3 ) h(y)=(y+w_1)(y+w_2)(y+w_3) h ( y ) = ( y + w 1 ) ( y + w 2 ) ( y + w 3 )
Since u 1 u 2 u 3 u 4 = 1 u_1u_2u_3u_4=1 u 1 u 2 u 3 u 4 = 1 , each w k w_k w k has the form of w k = v k + v k − 1 w_k=v_k+v_k^{-1} w k = v k + v k − 1 with v k = u 1 u 1 + k v_k=u_1u_{1+k} v k = u 1 u 1 + k for k ∈ { 1 , 2 , 3 } k\in\{1,2,3\} k ∈ { 1 , 2 , 3 } . We will show the relation between roots of f ( x ) f(x) f ( x ) in U U U and roots of h ( y ) h(y) h ( y ) in F q \mathbb{F}_q F q . Suppose that f ( x ) f(x) f ( x ) splits in U U U . Then each v k v_k v k belongs to U U U , so w k = v k + v k − 1 = v k + v ‾ k ∈ F q w_k=v_k+v_k^{-1}=v_k+\overline{v}_k\in\mathbb{F}_q w k = v k + v k − 1 = v k + v k ∈ F q . This implies that h ( y ) h(y) h ( y ) splits in F q \mathbb{F}_q F q . If w k ≠ 0 w_k\ne0 w k = 0 , then v k ∈ U ∖ { 1 } v_k\in U\setminus\{1\} v k ∈ U ∖ { 1 } , and then w k ∈ H 1 w_k\in H_1 w k ∈ H 1 . On the contrary, suppose that each w k w_k w k lies in F q \mathbb{F}_q F q , and each nonzero w k w_k w k lies in H 1 H_1 H 1 . It is clear that for each k = 1 , 2 , 3 k=1,2,3 k = 1 , 2 , 3 , the roots v k , v k − 1 v_k,v_k^{-1} v k , v k − 1 of x 2 + w k x + 1 = 0 x^2+w_kx+1=0 x 2 + w k x + 1 = 0 satisfy v k = v k − 1 = 1 v_k=v_k^{-1}=1 v k = v k − 1 = 1 when w k = 0 w_k=0 w k = 0 or v k ∈ U ∖ { 1 } v_k\in U\setminus\{1\} v k ∈ U ∖ { 1 } when w k ∈ H 1 w_k\in H_1 w k ∈ H 1 . Assume v 1 = u 1 u 2 v_1=u_1u_2 v 1 = u 1 u 2 , v 2 = u 1 u 3 v_2=u_1u_3 v 2 = u 1 u 3 and v 3 = u 1 u 4 v_3=u_1u_4 v 3 = u 1 u 4 . Then we have
v 1 v 2 v 3 = u 1 3 u 2 u 3 u 4 = u 1 2 , v 1 v ‾ 2 v ‾ 3 = u 2 2 , v ‾ 1 v 2 v ‾ 3 = u 3 2 , v ‾ 1 v ‾ 2 v 3 = u 4 2 . v_1v_2v_3=u_1^3u_2u_3u_4=u_1^2,\qquad v_1\overline{v}_2\overline{v}_3=u_2^2,\qquad\overline{v}_1v_2\overline{v}_3=u_3^2,\qquad\overline{v}_1\overline{v}_2v_3=u_4^2. v 1 v 2 v 3 = u 1 3 u 2 u 3 u 4 = u 1 2 , v 1 v 2 v 3 = u 2 2 , v 1 v 2 v 3 = u 3 2 , v 1 v 2 v 3 = u 4 2 .
This shows that from { v 1 , v 2 , v 3 } \{v_1,v_2,v_3\} { v 1 , v 2 , v 3 } , the roots u i u_i u i , i = 1 , 2 , 3 , 4 i=1,2,3,4 i = 1 , 2 , 3 , 4 , of f ( x ) f(x) f ( x ) can be uniquely determined and they must lie in U U U . Therefore f ( x ) f(x) f ( x ) splits in U U U , proving (i). For (ii), by the definitions of the w k w_k w k , we have
∏ 1 ≤ i < j ≤ 4 ( u i + u j ) = ∏ 1 ≤ i < j ≤ 3 ( w i + w j ) = β α α ‾ + ( α + α ‾ ) 2 . \prod_{1\le i<j\le4}(u_i+u_j)=\prod_{1\le i<j\le3}(w_i+w_j)=\beta\alpha\overline{\alpha}+(\alpha+\overline{\alpha})^2. 1 ≤ i < j ≤ 4 ∏ ( u i + u j ) = 1 ≤ i < j ≤ 3 ∏ ( w i + w j ) = β α α + ( α + α ) 2 .
Hence f ( x ) f(x) f ( x ) has no repeated roots if and only if h ( y ) h(y) h ( y ) has no repeated roots, and these equivalent conditions hold precisely when β α α ‾ + ( α + α ‾ ) 2 ≠ 0 \beta\alpha\overline{\alpha}+(\alpha+\overline{\alpha})^2\ne0 β α α + ( α + α ) 2 = 0 . This proves (ii). Finally, it is clear that h ( 0 ) = 0 h(0)=0 h ( 0 ) = 0 if and only if α = α ‾ \alpha=\overline{\alpha} α = α , equivalently α ∈ F q \alpha\in\mathbb{F}_q α ∈ F q . The assertion about four distinct roots in U U U follows by combining (i) and (ii). □
Remark 1. For a Niho decimation d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 with q = 2 m q=2^m q = 2 m , its key polynomial f s , a ( x ) = x 2 s − 1 + a x s + a x ‾ s − 1 + 1 f_{s,a}(x)=x^{2s-1}+ax^s+a\overline{x}^{s-1}+1 f s , a ( x ) = x 2 s − 1 + a x s + a x s − 1 + 1 is a normalized SCR polynomial. The above discussion on when a low-degree normalized SCR polynomial splits in the unit circle U U U will be used to investigate the values of N s ( a ) = # { x ∈ U : f s , a ( x ) = 0 } N_s(a)=\#\{x\in U:f_{s,a}(x)=0\} N s ( a ) = # { x ∈ U : f s , a ( x ) = 0 } for small s s s . Below we discuss the cases of s = 2 s=2 s = 2 and s = 3 s=3 s = 3 . In the next subsection, we consider the case s ≡ 5 4 ( m o d q + 1 ) s\equiv\frac{5}{4}\pmod{q+1} s ≡ 4 5 ( mod q + 1 ) .
When s = 2 s=2 s = 2 , the polynomial f s , a ( x ) f_{s,a}(x) f s , a ( x ) becomes a cubic polynomial f a ( x ) = x 3 + a x 2 + a ‾ x + 1 f_a(x)=x^3+ax^2+\overline{a}x+1 f a ( x ) = x 3 + a x 2 + a x + 1 , which has been discussed in Corollary 3.3. More concretely, suppose it splits in U U U , then one of the following cases holds:
a = 0 a=0 a = 0 , where f ( x ) = ( x + 1 ) ( x + ω ) ( x + ω 2 ) f(x)=(x+1)(x+\omega)(x+\omega^2) f ( x ) = ( x + 1 ) ( x + ω ) ( x + ω 2 ) and F 4 = { 0 , 1 , ω , ω 2 } \mathbb{F}_4=\{0,1,\omega,\omega^2\} F 4 = { 0 , 1 , ω , ω 2 } ; in this case ω ∉ U \omega\notin U ω ∈ / U when m m m is even;
a = 1 a=1 a = 1 , where f ( x ) = ( x + 1 ) 3 f(x)=(x+1)^3 f ( x ) = ( x + 1 ) 3 ;
a ∈ U ∖ { 1 } a\in U\setminus\{1\} a ∈ U ∖ { 1 } , where f ( x ) = ( x + a ) ( x + a ‾ − 1 ) 2 f(x)=(x+a)(x+\overline{a}^{-1})^2 f ( x ) = ( x + a ) ( x + a − 1 ) 2 ; iii) a ∈ F q ∗ ∖ { 1 } a\in\mathbb{F}^*_q\setminus\{1\} a ∈ F q ∗ ∖ { 1 } and a + 1 ∈ H 1 a+1\in H_1 a + 1 ∈ H 1 , where f ( x + 1 ) ( x + u ) ( x + u − 1 ) f(x+1)(x+u)(x+u^{-1}) f ( x + 1 ) ( x + u ) ( x + u − 1 ) with u ∈ U ∖ { 1 } u\in U\setminus\{1\} u ∈ U ∖ { 1 } ;
a ∈ F q 2 ∖ ( F q ∪ U ) a\in\mathbb{F}_{q^2}\setminus(\mathbb{F}_q\cup U) a ∈ F q 2 ∖ ( F q ∪ U ) , as shown in Proposition 3.3, f ( x ) f(x) f ( x ) has three distinct roots in U ∖ { 1 } U\setminus\{1\} U ∖ { 1 } iff g ( y ) = y 3 + T y 2 + ( N + T + 1 ) y + T 2 g(y)=y^3+Ty^2+(N+T+1)y+T^2 g ( y ) = y 3 + T y 2 + ( N + T + 1 ) y + T 2 has three distinct roots in H 1 H_1 H 1 , where T = α + α ‾ T=\alpha+\overline{\alpha} T = α + α and N = α α ‾ N=\alpha\overline{\alpha} N = α α .
The above characterizations are also useful for the case of s = 3 s=3 s = 3 below.
When s = 3 s=3 s = 3 , the key polynomial f s , a ( x ) f_{s,a}(x) f s , a ( x ) becomes a cubic polynomial f a ( x ) = x 5 + a x 3 + a ‾ x 2 + 1 f_a(x)=x^5+ax^3+\overline{a}x^2+1 f a ( x ) = x 5 + a x 3 + a x 2 + 1 . Based on the results on the quadratic and cubic normalized SCR polynomials in Corollaries 3.2 and 3.3, we have the following result.
Proposition 3.5. Let q = 2 m q=2^m q = 2 m with m ≡ 0 ( m o d 4 ) m\equiv0\pmod{4} m ≡ 0 ( mod 4 ) , H 1 = { w ∈ F q ∗ : Tr m ( 1 / w ) = 1 } H_1=\{w\in\mathbb{F}_q^*: \operatorname{Tr}_m(1/w)=1\} H 1 = { w ∈ F q ∗ : Tr m ( 1/ w ) = 1 } , and denote
A = H 1 ∪ { α ∈ F q 2 ∖ ( F q ∪ U ) ∣ g ( y ) = y 3 + ( α + α ˉ ) y 2 + ( α + 1 ) ( α ˉ + 1 ) y + ( α + α ˉ ) 2 has three distinct roots in H 1 } . \mathcal{A}=H_1\cup\left\{\alpha\in\mathbb{F}_{q^2}\setminus(\mathbb{F}_q\cup U)\mathrel{\Big|}g(y)=y^3+(\alpha+\bar{\alpha})y^2+(\alpha+1)(\bar{\alpha}+1)y+(\alpha+\bar{\alpha})^2\text{ has three distinct roots in }H_1\right\}. A = H 1 ∪ { α ∈ F q 2 ∖ ( F q ∪ U ) g ( y ) = y 3 + ( α + α ˉ ) y 2 + ( α + 1 ) ( α ˉ + 1 ) y + ( α + α ˉ ) 2 has three distinct roots in H 1 } . For α ∈ A \alpha\in\mathcal{A} α ∈ A , let μ 5 = α ˉ / α \mu^5=\bar{\alpha}/\alpha μ 5 = α ˉ / α and define h ( x ) = x 3 + α x 2 + α ˉ x + 1 h(x)=x^3+\alpha x^2+\bar{\alpha}x+1 h ( x ) = x 3 + α x 2 + α ˉ x + 1 . Then the polynomial f a ( x ) = x 5 + a x 3 + a ˉ x 2 + 1 f_a(x)=x^5+ax^3+\bar{a}x^2+1 f a ( x ) = x 5 + a x 3 + a ˉ x 2 + 1 satisfies
μ − 5 f a ( μ x ) = h ( x ) ( x 2 + α x + α α ˉ ) if and only if a = ( α ˉ 2 + α 2 α ˉ + α ) μ 2 α ˉ . \mu^{-5}f_a(\mu x)=h(x)\left(x^2+\alpha x+\frac{\alpha}{\bar{\alpha}}\right)\quad\text{if and only if}\quad a=\frac{(\bar{\alpha}^2+\alpha^2\bar{\alpha}+\alpha)\mu^2}{\bar{\alpha}}. μ − 5 f a ( μx ) = h ( x ) ( x 2 + α x + α ˉ α ) if and only if a = α ˉ ( α ˉ 2 + α 2 α ˉ + α ) μ 2 . Furthermore, (i) f a ( x ) f_a(x) f a ( x ) has exactly three roots in U U U if a = ( α ˉ 2 + α 2 α ˉ + α ) μ 2 α ˉ a=\frac{(\bar{\alpha}^2+\alpha^2\bar{\alpha}+\alpha)\mu^2}{\bar{\alpha}} a = α ˉ ( α ˉ 2 + α 2 α ˉ + α ) μ 2 , where α ∈ A \alpha\in\mathcal{A} α ∈ A with Tr m ( 1 α α ˉ ) = 0 \operatorname{Tr}_m\left(\frac{1}{\alpha\bar{\alpha}}\right)=0 Tr m ( α α ˉ 1 ) = 0 ; (ii) f a ( x ) f_a(x) f a ( x ) has exactly five roots in U U U if a = ( α ˉ 2 + α 2 α ˉ + α ) μ 2 α ˉ a=\frac{(\bar{\alpha}^2+\alpha^2\bar{\alpha}+\alpha)\mu^2}{\bar{\alpha}} a = α ˉ ( α ˉ 2 + α 2 α ˉ + α ) μ 2 where α ∈ A \alpha\in\mathcal{A} α ∈ A with Tr m ( 1 α α ˉ ) = 1 \operatorname{Tr}_m\left(\frac{1}{\alpha\bar{\alpha}}\right)=1 Tr m ( α α ˉ 1 ) = 1 .
Proof. Take α ∈ F q 2 ∗ \alpha\in\mathbb{F}_{q^2}^* α ∈ F q 2 ∗ and μ ∈ U \mu\in U μ ∈ U satisfying μ 5 = α ˉ / α \mu^5=\bar{\alpha}/\alpha μ 5 = α ˉ / α , for which the existence is ensured by gcd ( 5 , 2 m + 1 ) = 1 \gcd(5,2^m+1)=1 g cd( 5 , 2 m + 1 ) = 1 when m ≡ 0 ( m o d 4 ) m\equiv0\pmod{4} m ≡ 0 ( mod 4 ) . Then it is routine to verify
μ − 5 f a ( μ x ) = h ( x ) ( x 2 + α x + α α ˉ ) if and only if a = ( α ˉ 2 + α 2 α ˉ + α ) μ 2 α ˉ . \mu^{-5}f_a(\mu x)=h(x)\left(x^2+\alpha x+\frac{\alpha}{\bar{\alpha}}\right)\quad\text{if and only if}\quad a=\frac{(\bar{\alpha}^2+\alpha^2\bar{\alpha}+\alpha)\mu^2}{\bar{\alpha}}. μ − 5 f a ( μx ) = h ( x ) ( x 2 + α x + α ˉ α ) if and only if a = α ˉ ( α ˉ 2 + α 2 α ˉ + α ) μ 2 .
For any root u i u_i u i of h ( x ) h(x) h ( x ) in U U U , the element u i μ u_i\mu u i μ is a root of f a ( x ) f_a(x) f a ( x ) and it lies in U U U . According to Corollary 3.3, h ( x ) = x 3 + α x 2 + α ˉ x + 1 h(x)=x^3+\alpha x^2+\bar{\alpha}x+1 h ( x ) = x 3 + α x 2 + α ˉ x + 1 has exactly three roots if and only either α ∈ F q ∩ ( 1 + H 1 ) \alpha\in\mathbb{F}_q\cap(1+H_1) α ∈ F q ∩ ( 1 + H 1 ) or α ∉ ( F q ∪ U ) \alpha\notin(\mathbb{F}_q\cup U) α ∈ / ( F q ∪ U ) and g ( y ) = y 3 + T y 2 + ( N + T + 1 ) y + T 2 g(y)=y^3+Ty^2+(N+T+1)y+T^2 g ( y ) = y 3 + T y 2 + ( N + T + 1 ) y + T 2 , where T = α + α ˉ T=\alpha+\bar{\alpha} T = α + α ˉ and N = α α ˉ N=\alpha\bar{\alpha} N = α α ˉ , has three distinct roots in H 1 H_1 H 1 . The statements in (i), (ii) directly from Corollary 3.2. □ \square □
Remark 2. Proposition 3.5 provides characterizations on the elements a a a for which f a ( x ) = x 5 + a x 3 + a ˉ x 2 + 1 f_a(x)=x^5+ax^3+\bar{a}x^2+1 f a ( x ) = x 5 + a x 3 + a ˉ x 2 + 1 has exactly three, five roots in U U U explicitly when m ≡ 0 ( m o d 4 ) m\equiv0\pmod{4} m ≡ 0 ( mod 4 ) . Here the condition m ≡ 0 ( m o d 4 ) m\equiv0\pmod{4} m ≡ 0 ( mod 4 ) is required to ensure there exists a unique element μ \mu μ in U U U satisfying μ 3 = 1 \mu^3=1 μ 3 = 1 and μ 5 = 1 \mu^5=1 μ 5 = 1 . This complements the enumeration results on such a a a ’s determined in [5 ] .
Correlation distribution for the Niho exponent 5 q − 1 5q-1 5 q − 1
With the above preparations, we will investigate the correlation distribution for the Niho exponent d d d in (2) in this section. From now on we always assume m m m is a positive even integer, which ensures gcd ( d , q 2 − 1 ) = 1 \gcd(d,q^2-1)=1 g cd( d , q 2 − 1 ) = 1 . Note that 4 s ≡ 5 ( m o d q + 1 ) 4s\equiv5\pmod{q+1} 4 s ≡ 5 ( mod q + 1 ) implies 4 d ≡ 5 q − 1 ( m o d q 2 − 1 ) 4d\equiv5q-1\pmod{q^2-1} 4 d ≡ 5 q − 1 ( mod q 2 − 1 ) , indicating that d d d and 5 q − 1 5q-1 5 q − 1 lie in the same 2-cyclotomic coset modulo q 2 − 1 q^2-1 q 2 − 1 , and have the same cross-correlation distribution and Walsh spectrum. Thus they will be used interchangeably in our discussion. In the sequel we will first express the correlation distribution in terms of three frequencies, N 4 N_4 N 4 , N 5 N_5 N 5 and N 6 N_6 N 6 (see Proposition 4.2 ), by using the first four power moments, and then derive three identities for these frequencies from a careful analysis of the hexic polynomial P a ( x ) = x 6 + a x 5 + a ˉ x + 1 P_a(x)=x^6+ax^5+\bar{a}x+1 P a ( x ) = x 6 + a x 5 + a ˉ x + 1 .
Seven-valued correlation distribution in terms of N 4 N_4 N 4 , N 5 N_5 N 5 , N 6 N_6 N 6
Recall that, for the Niho exponent d d d in (2) , the cross-correlation function C d ( τ ) C_d(\tau) C d ( τ ) can be rewritten as C d ( τ ) = W d ( α τ ) − 1 C_d(\tau)=W_d(\alpha^\tau)-1 C d ( τ ) = W d ( α τ ) − 1 , where
W d ( a ) = ∑ x ∈ F q 2 ( − 1 ) Tr 2 m ( x d + a x ) . W_d(a)=\sum_{x\in\mathbb{F}_{q^2}}(-1)^{\operatorname{Tr}_{2m}(x^d+ax)}. W d ( a ) = x ∈ F q 2 ∑ ( − 1 ) Tr 2 m ( x d + a x ) .
As τ \tau τ ranges over { 0 , 1 , ⋯ , q 2 − 2 } \{0,1,\cdots,q^2-2\} { 0 , 1 , ⋯ , q 2 − 2 } , α τ \alpha^\tau α τ runs through F q 2 ∗ \mathbb{F}_{q^2}^{*} F q 2 ∗ . Hence, determining the correlation distribution is equivalent to determining the value distribution of W d ( a ) W_d(a) W d ( a ) over a ∈ F q 2 ∗ a\in\mathbb{F}_{q^2}^{*} a ∈ F q 2 ∗ . By Niho’s theorem (Lemma 2.5), this reduces to counting the roots in U U U of the key polynomial f s , a ( x ) = x 2 s − 1 + a x s + a ‾ x s − 1 + 1 f_{s,a}(x)=x^{2^s-1}+ax^s+\overline{a}x^{s-1}+1 f s , a ( x ) = x 2 s − 1 + a x s + a x s − 1 + 1 as a a a varies over F q 2 ∗ \mathbb{F}_{q^2}^{*} F q 2 ∗ . Raising the polynomial the 4-th power and making a substitution a 4 a^4 a 4 as a a a yields the following hexic polynomial
P a ( x ) = x 6 + a x 5 + a ‾ x + 1. (12) P_a(x)=x^6+ax^5+\overline{a}x+1. \tag*{(12)} P a ( x ) = x 6 + a x 5 + a x + 1. (12)
Clearly for any a ∈ F q 2 ∗ a\in\mathbb{F}_{q^2}^{*} a ∈ F q 2 ∗ the key polynomial f s , a f_{s,a} f s , a has the same number of roots as P a ( x ) P_a(x) P a ( x ) in the unit circle U U U , which is at most 6. Our main focus is to study the polynomial P a ( x ) P_a(x) P a ( x ) and to determine the number of a ∈ F q 2 ∗ a\in\mathbb{F}_{q^2}^{*} a ∈ F q 2 ∗ such that P a ( x ) P_a(x) P a ( x ) has i i i distinct roots in U U U for i = 0 , 1 , … , 6 i=0,1,\ldots,6 i = 0 , 1 , … , 6 . Accordingly, we define
N ( a ) : = ∣ { u ∈ U : P a ( u ) = 0 } ∣ , and N i : = ∣ { a ∈ F q 2 ∗ : N ( a ) = i } ∣ . (13) N(a):=\left|\{u\in U:P_a(u)=0\}\right|,\quad\text{and}\quad N_i:=\left|\{a\in\mathbb{F}_{q^2}^{*}:N(a)=i\}\right|. \tag*{(13)} N ( a ) := ∣ { u ∈ U : P a ( u ) = 0 } ∣ , and N i := { a ∈ F q 2 ∗ : N ( a ) = i } . (13)
The values N i N_i N i ’s have a close connection with power moments of the Walsh transform W d ( a ) W_d(a) W d ( a ) [1 , 29 ] . The k k k -th power moment of Walsh transforms of the function Tr 2 m ( x d ) \operatorname{Tr}_{2m}(x^d) Tr 2 m ( x d ) is defined as:
M k ( d ) : = ∑ a ∈ F q 2 ( W d ( a ) ) k , M_k(d):=\sum_{a\in\mathbb{F}_{q^2}}(W_d(a))^k, M k ( d ) := a ∈ F q 2 ∑ ( W d ( a ) ) k ,
and it can be transformed as
M k ( d ) = ∑ a ∈ F q 2 ( W d ( a ) ) k = q 2 ∑ x 1 + ⋯ + x k = 0 ( x 1 , … , x k ) ∈ F q 2 k ( − 1 ) Tr 2 m ( x 1 d + ⋯ + x k d ) . M_k(d)=\sum_{a\in\mathbb{F}_{q^2}}(W_d(a))^k=q^2\sum_{\substack{x_1+\cdots+x_k=0\\(x_1,\ldots,x_k)\in\mathbb{F}_{q^2}^k}}(-1)^{\operatorname{Tr}_{2m}(x_1^d+\cdots+x_k^d)}. M k ( d ) = a ∈ F q 2 ∑ ( W d ( a ) ) k = q 2 x 1 + ⋯ + x k = 0 ( x 1 , … , x k ) ∈ F q 2 k ∑ ( − 1 ) Tr 2 m ( x 1 d + ⋯ + x k d ) .
For any exponent d d d , one easily has
M 0 ( d ) = M 1 ( d ) = q 2 and M 2 ( d ) = q 4 , M_0(d)=M_1(d)=q^2\quad\text{and}\quad M_2(d)=q^4, M 0 ( d ) = M 1 ( d ) = q 2 and M 2 ( d ) = q 4 ,
and for k = 3 k=3 k = 3 , if gcd ( d , q 2 − 1 ) = 1 \gcd(d,q^2-1)=1 g cd( d , q 2 − 1 ) = 1 , then
M 3 ( d ) = q 2 ∑ x 1 , x 2 ∈ F q 2 ( − 1 ) Tr 2 m ( x 1 d + x 2 d + ( x 1 + x 2 ) d ) = q 2 ∑ x ∈ F q 2 ∑ y ∈ F q 2 ( − 1 ) Tr 2 m ( y d ( ( x + 1 ) d + x d + 1 ) ) = q 4 b 3 . \begin{aligned}
M_3(d)&=q^2\sum_{x_1,x_2\in\mathbb{F}_{q^2}}(-1)^{\operatorname{Tr}_{2m}(x_1^d+x_2^d+(x_1+x_2)^d)}\\
&=q^2\sum_{x\in\mathbb{F}_{q^2}}\sum_{y\in\mathbb{F}_{q^2}}(-1)^{\operatorname{Tr}_{2m}(y^d((x+1)^d+x^d+1))}\\
&=q^4b_3.
\end{aligned} M 3 ( d ) = q 2 x 1 , x 2 ∈ F q 2 ∑ ( − 1 ) Tr 2 m ( x 1 d + x 2 d + ( x 1 + x 2 ) d ) = q 2 x ∈ F q 2 ∑ y ∈ F q 2 ∑ ( − 1 ) Tr 2 m ( y d (( x + 1 ) d + x d + 1 )) = q 4 b 3 .
where b 3 = ∣ { x ∈ F q 2 : ( x + 1 ) d + x d + 1 = 0 } ∣ b_3=\left|\{x\in\mathbb{F}_{q^2}:(x+1)^d+x^d+1=0\}\right| b 3 = { x ∈ F q 2 : ( x + 1 ) d + x d + 1 = 0 } .
These identities are well-known and can be found in [14 Proposition 3.1] . The third identity requires the assumption gcd ( d , q 2 − 1 ) = 1 \gcd(d,q^2-1)=1 g cd( d , q 2 − 1 ) = 1 . Under this assumption, the map x ↦ x d x\mapsto x^d x ↦ x d permutes F q 2 \mathbb{F}_{q^2} F q 2 , so W d ( 0 ) = 0 W_d(0)=0 W d ( 0 ) = 0 . Hence the sum in these identities can be restricted to a ∈ F q 2 ∗ a\in\mathbb{F}_{q^2}^{*} a ∈ F q 2 ∗ . Note that the Niho exponent d d d in (2) is co-prime to 2 2 m − 1 2^{2m}-1 2 2 m − 1 when m m m is even. The following lemma gives the value of b 3 b_3 b 3 for this Niho exponent.
Lemma 4.1. Let q = 2 m q = 2^{m} q = 2 m with m m m being a positive integer. For d = s ( q − 1 ) + 1 d = s(q - 1) + 1 d = s ( q − 1 ) + 1 with 4 s ≡ 5 ( m o d q + 1 ) 4s \equiv5 \pmod{q + 1} 4 s ≡ 5 ( mod q + 1 ) , one has b 3 = q + 12 ε b_{3} = q + 12\varepsilon b 3 = q + 12 ε , where ε = 1 \varepsilon= 1 ε = 1 if m ≡ 2 ( m o d 4 ) m \equiv2 \pmod{4} m ≡ 2 ( mod 4 ) and ε = 0 \varepsilon= 0 ε = 0 otherwise.
Proof. The elements 0 , 1 0,1 0 , 1 satisfy ( x + 1 ) d + x d + 1 = 0 (x+1)^{d}+x^{d}+1=0 ( x + 1 ) d + x d + 1 = 0 . Raise this equation to the fourth power gives
( x + 1 ) 5 q − 1 + x 5 q − 1 + 1 = 0 (x+1)^{5q-1}+x^{5q-1}+1=0 ( x + 1 ) 5 q − 1 + x 5 q − 1 + 1 = 0
For x ≠ 0 , 1 x \ne0,1 x = 0 , 1 , put y = x q y=x^{q} y = x q and multiply by x ( x + 1 ) x(x+1) x ( x + 1 ) the above equation. Then we have
x ( y + 1 ) 5 + ( x + 1 ) y 5 + x ( x + 1 ) = 0 , x(y+1)^{5}+(x+1)y^{5}+x(x+1)=0, x ( y + 1 ) 5 + ( x + 1 ) y 5 + x ( x + 1 ) = 0 ,
reducing to
y 5 + x y 4 + x y + x 2 = ( y + x ) ( y 4 + x ) = 0. y^{5}+xy^{4}+xy+x^{2}=(y+x)(y^{4}+x)=0. y 5 + x y 4 + x y + x 2 = ( y + x ) ( y 4 + x ) = 0.
Thus the complete solution set S d S_{d} S d is given by
S d = { x : x q = x } ∪ { x : x 2 m + 2 = x } = F q ∪ F 2 gcd ( m + 2 , 2 m ) . S_{d}=\{x:x^{q}=x\}\cup\{x:x^{2^{m+2}}=x\}=\mathbb{F}_{q}\cup\mathbb{F}_{2^{\gcd(m+2,2m)}}. S d = { x : x q = x } ∪ { x : x 2 m + 2 = x } = F q ∪ F 2 g c d ( m + 2 , 2 m ) .
If m m m is odd, the second field is F 2 ⊆ F q \mathbb{F}_{2}\subseteq\mathbb{F}_{q} F 2 ⊆ F q . If m ≡ 0 ( m o d 4 ) m\equiv0\pmod{4} m ≡ 0 ( mod 4 ) , the second field is F 4 ⊆ F q \mathbb{F}_{4}\subseteq\mathbb{F}_{q} F 4 ⊆ F q . If m ≡ 2 ( m o d 4 ) m\equiv2\pmod{4} m ≡ 2 ( mod 4 ) , the second field is F 2 4 \mathbb{F}_{2^{4}} F 2 4 and its intersection with F q \mathbb{F}_{q} F q is F 4 \mathbb{F}_{4} F 4 . □ \square □
It follows from Lemma 4.1 that M 3 ( d ) = q 4 ( q + 12 ε ) M_{3}(d)=q^{4}(q+12\varepsilon) M 3 ( d ) = q 4 ( q + 12 ε ) for the Niho exponent d d d in (2) . Niho’s theorem gives the connection W d ( a ) = q ( i − 1 ) W_{d}(a)=q(i-1) W d ( a ) = q ( i − 1 ) in (5) with 0 ≤ i ≤ 6 0\leq i\leq6 0 ≤ i ≤ 6 . That is to say, there are seven possible values in the Walsh spectrum of x d x^{d} x d , and the frequency is given by N 0 , N 1 , … , N 6 N_{0},N_{1},\ldots,N_{6} N 0 , N 1 , … , N 6 . The first four power moments of W d ( a ) W_{d}(a) W d ( a ) provides four equations in N i N_{i} N i as follows:
M 0 ( d ) = ∑ i = 0 6 N i = q 2 − 1 , M 1 ( d ) = ∑ i = 0 6 q ( i − 1 ) N i = q 2 , M 2 ( d ) = ∑ i = 0 6 ( q ( i − 1 ) ) 2 N i = q 4 , M 3 ( d ) = ∑ i = 0 6 ( q ( i − 1 ) ) 3 N i = q 4 ( q + 12 ε ) . \begin{aligned}
M_{0}(d)&=\sum_{i=0}^{6}N_{i}=q^{2}-1,\\
M_{1}(d)&=\sum_{i=0}^{6}q(i-1)N_{i}=q^{2},\\
M_{2}(d)&=\sum_{i=0}^{6}(q(i-1))^{2}N_{i}=q^{4},\\
M_{3}(d)&=\sum_{i=0}^{6}(q(i-1))^{3}N_{i}=q^{4}(q+12\varepsilon).
\end{aligned} M 0 ( d ) M 1 ( d ) M 2 ( d ) M 3 ( d ) = i = 0 ∑ 6 N i = q 2 − 1 , = i = 0 ∑ 6 q ( i − 1 ) N i = q 2 , = i = 0 ∑ 6 ( q ( i − 1 ) ) 2 N i = q 4 , = i = 0 ∑ 6 ( q ( i − 1 ) ) 3 N i = q 4 ( q + 12 ε ) .
By the above equalities we have
Proposition 4.2. For N i N_{i} N i ’s defined in (13) , one has
N 0 = 1 3 ( q − 1 − 6 ε ) q + N 4 + 4 N 5 + 10 N 6 , N 1 = 1 2 ( q − 1 + 12 ε ) q − 4 N 4 − 15 N 5 − 36 N 6 − 1 , N 2 = ( 1 − 6 ε ) q + 6 N 4 + 20 N 5 + 45 N 6 , N 3 = 1 6 ( q − 1 + 12 ε ) q − 4 N 4 − 10 N 5 − 20 N 6 . (14) \begin{aligned}
N_{0}&=\frac{1}{3}(q-1-6\varepsilon)q+N_{4}+4N_{5}+10N_{6},\\
N_{1}&=\frac{1}{2}(q-1+12\varepsilon)q-4N_{4}-15N_{5}-36N_{6}-1,\\
N_{2}&=(1-6\varepsilon)q+6N_{4}+20N_{5}+45N_{6},\\
N_{3}&=\frac{1}{6}(q-1+12\varepsilon)q-4N_{4}-10N_{5}-20N_{6}. \tag*{(14)}
\end{aligned} N 0 N 1 N 2 N 3 = 3 1 ( q − 1 − 6 ε ) q + N 4 + 4 N 5 + 10 N 6 , = 2 1 ( q − 1 + 12 ε ) q − 4 N 4 − 15 N 5 − 36 N 6 − 1 , = ( 1 − 6 ε ) q + 6 N 4 + 20 N 5 + 45 N 6 , = 6 1 ( q − 1 + 12 ε ) q − 4 N 4 − 10 N 5 − 20 N 6 . (14)
In what follows, we aim to determine N 4 N_{4} N 4 , N 5 N_{5} N 5 , and N 6 N_{6} N 6 . To this end, we first analyze the hexic polynomial P a ( x ) P_{a}(x) P a ( x ) in (12) to derive some identities for the frequencies N 4 , N 5 , N 6 N_{4},N_{5},N_{6} N 4 , N 5 , N 6 .
Three identities for N 4 , N 5 , N 6 N_{4},N_{5},N_{6} N 4 , N 5 , N 6 from characterizations of P a ( x ) P_{a}(x) P a ( x )
The hexic polynomial P a ( x ) = x 6 + a x 5 + a ˉ x + 1 P_{a}(x)=x^{6}+ax^{5}+\bar{a}x+1 P a ( x ) = x 6 + a x 5 + a ˉ x + 1 is a normalized SCR polynomial and has the same number of roots in U U U as the key polynomial f s , a ( x ) f_{s,a}(x) f s , a ( x ) for every a ∈ F q 2 a\in\mathbb{F}_{q^{2}} a ∈ F q 2 . In this subsection, we investigate the cases where P a ( x ) P_{a}(x) P a ( x ) has 4, 5 or 6 roots in U U U as a a a runs through F q 2 \mathbb{F}_{q^{2}} F q 2 , where q = 2 m q=2^{m} q = 2 m for even m ≥ 2 m\geq2 m ≥ 2 and will then obtain three identities for N 4 N_{4} N 4 , N 5 N_{5} N 5 and N 6 N_{6} N 6 .
We first discuss the cases where P a ( x ) P_{a}(x) P a ( x ) has at least 4 distinct roots in U U U .
Proposition 4.3. Let q = 2 m q = 2^m q = 2 m for an even m ≥ 2 m \ge2 m ≥ 2 . Let S = { u 1 , u 2 , u 3 , u 4 } S = \{u_1,u_2,u_3,u_4\} S = { u 1 , u 2 , u 3 , u 4 } be a 4-subset of U U U and define elements
μ : = ( u 1 u 2 u 3 u 4 ) 1 4 , α : = μ − 1 ∑ i = 1 4 u i , β : = μ − 2 ∑ 1 ≤ i < j ≤ 4 u i u j . (15) \mu:= (u_1u_2u_3u_4)^{\frac{1}{4}}, \qquad\alpha:= \mu^{-1}\sum_{i=1}^{4}u_i, \qquad\beta:= \mu^{-2}\sum_{1\le i<j\le4}u_i u_j. \tag*{(15)} μ := ( u 1 u 2 u 3 u 4 ) 4 1 , α := μ − 1 i = 1 ∑ 4 u i , β := μ − 2 1 ≤ i < j ≤ 4 ∑ u i u j . (15) Given an element a ∈ F q 2 a \in\mathbb{F}_{q^2} a ∈ F q 2 , the polynomial P a ( x ) = x 6 + a x 5 + a x ˉ + 1 P_a(x) = x^6 + ax^5 + a\bar{x} + 1 P a ( x ) = x 6 + a x 5 + a x ˉ + 1 evaluates to zero on all four elements u i u_i u i in S S S if and only if the following two hold:
(i) the elements μ , α , β \mu,\alpha,\beta μ , α , β defined by u i u_i u i 's in S S S satisfy
( α + α ˉ ) 2 = β 3 + β , ( α + β ) μ − 6 = α ˉ + β (16) (\alpha+\bar{\alpha})^2=\beta^3+\beta,\qquad(\alpha+\sqrt{\beta})\mu^{-6}=\bar{\alpha}+\sqrt{\beta} \tag*{(16)} ( α + α ˉ ) 2 = β 3 + β , ( α + β ) μ − 6 = α ˉ + β (16) (ii) the element a a a can be expressed by them as follows
a = μ ( α 2 + β + μ − 6 ) / α . (17) a=\mu(\alpha^2+\beta+\mu^{-6})/\alpha. \tag*{(17)} a = μ ( α 2 + β + μ − 6 ) / α . (17)
Proof. We start with the necessity of the statement. Suppose P a ( x ) P_a(x) P a ( x ) evaluates to zero on the 4-subset S = { u 1 , u 2 , u 3 , u 4 } S=\{u_1,u_2,u_3,u_4\} S = { u 1 , u 2 , u 3 , u 4 } of U U U , i.e., P a ( u i ) = 0 P_a(u_i)=0 P a ( u i ) = 0 for i = 1 , 2 , 3 , 4 i=1,2,3,4 i = 1 , 2 , 3 , 4 . Consider two polynomials
Q ~ ( x ) : = ∏ i = 1 4 ( x + u i ) and Q ( x ) : = μ − 4 Q ~ ( μ x ) = ∏ i = 1 4 ( x + u i / μ ) . \widetilde{Q}(x):=\prod_{i=1}^{4}(x+u_i)\text{ and }Q(x):=\mu^{-4}\widetilde{Q}(\mu x)=\prod_{i=1}^{4}(x+u_i/\mu). Q ( x ) := i = 1 ∏ 4 ( x + u i ) and Q ( x ) := μ − 4 Q ( μx ) = i = 1 ∏ 4 ( x + u i / μ ) .
By the definition of μ , α , β \mu,\alpha,\beta μ , α , β in (15) , we have
Q ( x ) = ∏ i = 1 4 ( x + u i / μ ) = x 4 + α x 3 + β x 2 + α ˉ x + 1 , Q(x)=\prod_{i=1}^{4}(x+u_i/\mu)=x^4+\alpha x^3+\beta x^2+\bar{\alpha}x+1, Q ( x ) = i = 1 ∏ 4 ( x + u i / μ ) = x 4 + α x 3 + β x 2 + α ˉ x + 1 ,
where the coefficient of x x x in Q ( x ) Q(x) Q ( x ) is given by
μ − 3 ( u 1 u 2 u 3 + u 1 u 2 u 4 + u 1 u 3 u 4 + u 2 u 3 u 4 ) = μ ∑ i = 1 4 u i − 1 = μ − 1 ∑ i = 1 4 u i ˉ = α ˉ \mu^{-3}(u_1u_2u_3+u_1u_2u_4+u_1u_3u_4+u_2u_3u_4)=\mu\sum_{i=1}^{4}u_i^{-1}=\mu^{-1}\sum_{i=1}^{4}\bar{u_i}=\bar{\alpha} μ − 3 ( u 1 u 2 u 3 + u 1 u 2 u 4 + u 1 u 3 u 4 + u 2 u 3 u 4 ) = μ i = 1 ∑ 4 u i − 1 = μ − 1 i = 1 ∑ 4 u i ˉ = α ˉ
and the coefficient β \beta β satisfies
β ˉ = μ 2 ∑ 1 ≤ i < j ≤ 4 u i ˉ u j ˉ = μ − 2 ∑ 1 ≤ i < j ≤ 4 ( u i u j ) − 1 ( u 1 u 2 u 3 u 4 ) = μ − 2 ∑ 1 ≤ i < j ≤ 4 u i u j = β . \bar{\beta}=\mu^2\sum_{1\le i<j\le4}\bar{u_i}\bar{u_j}=\mu^{-2}\sum_{1\le i<j\le4}(u_i u_j)^{-1}(u_1u_2u_3u_4)=\mu^{-2}\sum_{1\le i<j\le4}u_i u_j=\beta. β ˉ = μ 2 1 ≤ i < j ≤ 4 ∑ u i ˉ u j ˉ = μ − 2 1 ≤ i < j ≤ 4 ∑ ( u i u j ) − 1 ( u 1 u 2 u 3 u 4 ) = μ − 2 1 ≤ i < j ≤ 4 ∑ u i u j = β .
According to the definitions of Q ~ ( x ) \widetilde{Q}(x) Q ( x ) and Q ( x ) Q(x) Q ( x ) , we see that Q ~ ( x ) \widetilde{Q}(x) Q ( x ) divides P a ( x ) P_a(x) P a ( x ) , which is equivalent to Q ( x ) ∣ μ − 6 P a ( μ x ) Q(x)\mid\mu^{-6}P_a(\mu x) Q ( x ) ∣ μ − 6 P a ( μx ) . Then we have the following factorization
μ − 6 P a ( μ x ) = ( x 4 + α x 3 + β x 2 + α ˉ x + 1 ) ( x 2 + γ x + μ − 6 ) , \mu^{-6}P_a(\mu x)=(x^4+\alpha x^3+\beta x^2+\bar{\alpha}x+1)(x^2+\gamma x+\mu^{-6}), μ − 6 P a ( μx ) = ( x 4 + α x 3 + β x 2 + α ˉ x + 1 ) ( x 2 + γ x + μ − 6 ) ,
where the constant μ − 6 \mu^{-6} μ − 6 in the quadratic polynomial is derived from that of μ − 6 P a ( μ x ) \mu^{-6}P_a(\mu x) μ − 6 P a ( μx ) immediately. Meanwhile, comparing coefficients of x 2 x^2 x 2 , x 3 x^3 x 3 , x 4 x^4 x 4 and x x x , x 5 x^5 x 5 on two sides of the above equality, respectively, gives the following relations:
β μ − 6 + α ˉ γ + 1 = 0 , α μ − 6 + β γ + α ˉ = 0 , μ − 6 + α γ + β = 0 , (18) \beta\mu^{-6}+\bar{\alpha}\gamma+1=0,\qquad\alpha\mu^{-6}+\beta\gamma+\bar{\alpha}=0,\qquad\mu^{-6}+\alpha\gamma+\beta=0, \tag*{(18)} β μ − 6 + α ˉ γ + 1 = 0 , α μ − 6 + β γ + α ˉ = 0 , μ − 6 + α γ + β = 0 , (18)
and
α + γ = a / μ , α ˉ μ − 6 + γ = a μ − 5 . (19) \alpha+\gamma=a/\mu,\qquad\bar{\alpha}\mu^{-6}+\gamma=a\mu^{-5}. \tag*{(19)} α + γ = a / μ , α ˉ μ − 6 + γ = a μ − 5 . (19)
Note that α ≠ 0 \alpha\ne0 α = 0 , otherwise Q ( x ) = x 4 + β x 2 + 1 = ( x 2 + β x + 1 ) 2 Q(x)=x^4+\beta x^2+1=(x^2+\sqrt{\beta}x+1)^2 Q ( x ) = x 4 + β x 2 + 1 = ( x 2 + β x + 1 ) 2 has repeated roots. Then equalities in (18) imply β 3 + β = ( α + α ˉ ) 2 \beta^3+\beta=(\alpha+\bar{\alpha})^2 β 3 + β = ( α + α ˉ ) 2 and ( α + β ) μ − 6 = α ˉ + β (\alpha+\sqrt{\beta})\mu^{-6}=\bar{\alpha}+\sqrt{\beta} ( α + β ) μ − 6 = α ˉ + β in (i); and (19) with the last equality in (18) implies the equality ( μ ( α 2 + β + μ − 6 ) ) / α = a (\mu(\alpha^2+\beta+\mu^{-6}))/\alpha=a ( μ ( α 2 + β + μ − 6 )) / α = a in (ii).
Conversely, suppose for a 4-subset S = { u 1 , u 2 , u 3 , u 4 } S=\{u_1,u_2,u_3,u_4\} S = { u 1 , u 2 , u 3 , u 4 } of U U U the properties (i)-(ii) are satisfied. Then we can define a polynomial
Q ( x ) = ∏ i = 1 4 ( x + u i / μ ) = x 4 + α x 3 + β x 2 + α ˉ x + 1. Q(x)=\prod_{i=1}^{4}(x+u_i/\mu)=x^4+\alpha x^3+\beta x^2+\bar{\alpha}x+1. Q ( x ) = i = 1 ∏ 4 ( x + u i / μ ) = x 4 + α x 3 + β x 2 + α ˉ x + 1.
Following from (i) and (ii) we have
Q ( x ) ( x 2 + μ − 6 + β α x + μ − 6 ) = x 6 + μ − 1 a x 5 + μ − 5 a ˉ x + 1. Q(x)\left(x^2+\frac{\mu^{-6}+\beta}{\alpha}x+\mu^{-6}\right)=x^6+\mu^{-1}ax^5+\mu^{-5}\bar{a}x+1. Q ( x ) ( x 2 + α μ − 6 + β x + μ − 6 ) = x 6 + μ − 1 a x 5 + μ − 5 a ˉ x + 1.
where a = ( μ ( α 2 + β + μ − 6 ) ) / α a=(\mu(\alpha^2+\beta+\mu^{-6}))/\alpha a = ( μ ( α 2 + β + μ − 6 )) / α . This implies that Q ( x ) Q(x) Q ( x ) divides x 6 + μ − 1 a x 5 + μ − 5 a ˉ x + 1 x^6+\mu^{-1}ax^5+\mu^{-5}\bar{a}x+1 x 6 + μ − 1 a x 5 + μ − 5 a ˉ x + 1 , equivalently, Q ~ ( x ) = μ 6 Q ( x / μ ) \widetilde{Q}(x)=\mu^6Q(x/\mu) Q ( x ) = μ 6 Q ( x / μ ) divides the polynomial P a ( x ) = x 6 + a x 5 + a ˉ x + 1 P_a(x)=x^6+ax^5+\bar{a}x+1 P a ( x ) = x 6 + a x 5 + a ˉ x + 1 . That is to say, the elements u 1 , u 2 , u 3 , u 4 u_1,u_2,u_3,u_4 u 1 , u 2 , u 3 , u 4 in S S S are four distinct roots of P a ( x ) P_a(x) P a ( x ) in U U U .
Remark 3. According to Proposition 4.3, given a 4-subset S S S of U U U , when μ , α , β \mu,\alpha,\beta μ , α , β defined in (15) satisfy the equalities in (16) , they completely determines an element a a a as in (17) , for which the polynomial P a ( x ) = x 6 + a x 5 + a ˉ x + 1 P_a(x)=x^6+ax^5+\bar{a}x+1 P a ( x ) = x 6 + a x 5 + a ˉ x + 1 evaluates to zero on each element of S S S . That is to say, the elements a a a for which P a ( x ) P_a(x) P a ( x ) has at least 4 roots in U U U are completely determined by those 4-subsets S S S of U U U that satisfy (16) via the relation in (17) . Define
S = { S ⊂ U : ∣ S ∣ = 4 , ∏ u ∈ S ( x + u ) ∣ P a ( x ) for certain a ∈ F q 2 } . (20) \mathbb{S}=\left\{S\subset U:|S|=4,\ \prod_{u\in S}(x+u)\mid P_a(x)\text{ for certain }a\in\mathbb{F}_{q^2}\right\}. \tag*{(20)} S = { S ⊂ U : ∣ S ∣ = 4 , u ∈ S ∏ ( x + u ) ∣ P a ( x ) for certain a ∈ F q 2 } . (20)
For i ∈ { 4 , 5 , 6 } i\in\{4,5,6\} i ∈ { 4 , 5 , 6 } , if P a ( x ) P_a(x) P a ( x ) has exactly i i i distinct roots in U U U , these roots determine ( i 4 ) \binom{i}{4} ( 4 i ) distinct 4-subsets of U U U , each of which belongs to S \mathbb{S} S by Proposition 4.3. Moreover, no 4-subset arises from two distinct values of a a a . Hence we have
∣ S ∣ = N 4 + 5 N 5 + 15 N 6 , (21) |\mathbb{S}|=N_4+5N_5+15N_6, \tag*{(21)} ∣ S ∣ = N 4 + 5 N 5 + 15 N 6 , (21)
where N i N_i N i are defined in (13) .
Next we consider another perspective to count the 4-subsets in S \mathbb{S} S . First, we give some auxiliary results.
Lemma 4.4. Let q = 2 m q=2^m q = 2 m , where m m m is even, and let a ∈ F q 2 a\in\mathbb{F}_{q^2} a ∈ F q 2 . The polynomial P a ( x ) P_a(x) P a ( x ) in (12) has repeated roots if and only if a ∈ F q a\in\mathbb{F}_q a ∈ F q , in which case it factors as
P a ( x ) = ( x + 1 ) 2 ( x 4 + a x 3 + x 2 + a x + 1 ) . (22) P_a(x)=(x+1)^2(x^4+ax^3+x^2+ax+1). \tag*{(22)} P a ( x ) = ( x + 1 ) 2 ( x 4 + a x 3 + x 2 + a x + 1 ) . (22) Furthermore,
(i) if a = 0 a=0 a = 0 , then P 0 P_0 P 0 has three distinct roots 1 , ω , ω 2 1,\omega,\omega^2 1 , ω , ω 2 in F 4 \mathbb{F}_4 F 4 , each with multiplicity two, and 1 1 1 is its only root in U U U ;
(ii) if a ∈ F q ∗ a\in\mathbb{F}_q^* a ∈ F q ∗ , then 1 1 1 is the unique repeated root of P a P_a P a and it has multiplicity exactly two. If Tr m ( a − 1 ) = 1 \operatorname{Tr}_m(a^{-1})=1 Tr m ( a − 1 ) = 1 , then 1 1 1 is its only root in U U U . If Tr m ( a − 1 ) = 0 \operatorname{Tr}_m(a^{-1})=0 Tr m ( a − 1 ) = 0 , there exists t ∈ F q ∖ { 0 , 1 } t\in\mathbb{F}_q\setminus\{0,1\} t ∈ F q ∖ { 0 , 1 } such that a = t + t − 1 a=t+t^{-1} a = t + t − 1 , and then
N ( a ) = { 1 , Tr m ( t ) = Tr m ( t − 1 ) = 0 , 3 , Tr m ( t ) ≠ Tr m ( t − 1 ) , 5 , Tr m ( t ) = Tr m ( t − 1 ) = 1. N(a)=
\begin{cases}
1, & \operatorname{Tr}_m(t)=\operatorname{Tr}_m(t^{-1})=0,\\
3, & \operatorname{Tr}_m(t)\ne\operatorname{Tr}_m(t^{-1}),\\
5, & \operatorname{Tr}_m(t)=\operatorname{Tr}_m(t^{-1})=1.
\end{cases} N ( a ) = ⎩ ⎨ ⎧ 1 , 3 , 5 , Tr m ( t ) = Tr m ( t − 1 ) = 0 , Tr m ( t ) = Tr m ( t − 1 ) , Tr m ( t ) = Tr m ( t − 1 ) = 1. :::`
Proof. The proof is straightforward and placed in Appendix A for completeness.
The above lemma allows us to determine N 5 N_5 N 5 explicitly.
Proposition 4.5. Let P a ( x ) P_a(x) P a ( x ) be the key polynomial defined in (12) and N 5 N_5 N 5 be defined in (13) . If P a ( x ) P_a(x) P a ( x ) has exactly five distinct roots in U U U , then a a a must lie in F q \mathbb{F}_q F q . Consequently, we have
N 5 = q + 1 + K m ( 1 ) 8 . N_5 = \frac{q+1+K_m(1)}{8}. N 5 = 8 q + 1 + K m ( 1 ) .
Proof. The proof is relatively simple and placed in Appendix A.
Lemma 4.6. For a ∈ U a \in U a ∈ U , the polynomial P a ( x ) P_a(x) P a ( x ) can be factorized as
P a ( X ) = x 6 + a x 5 + a ˉ x + 1 = ( X + a ) ( X 5 + a − 1 ) . (23) P_a(X) = x^6 + ax^5 + \bar{a}x + 1 = (X+a)(X^5+a^{-1}). \tag*{(23)} P a ( X ) = x 6 + a x 5 + a ˉ x + 1 = ( X + a ) ( X 5 + a − 1 ) . (23) (i) If m ≡ 0 ( m o d 4 ) m \equiv0 \pmod{4} m ≡ 0 ( mod 4 ) , then
N ( a ) = { 1 , a = 1 , 2 , a ∈ U ∖ { 1 } . N(a) =
\begin{cases}
1, & a=1,\\
2, & a \in U \setminus\{1\}.
\end{cases} N ( a ) = { 1 , 2 , a = 1 , a ∈ U ∖ { 1 } . (ii) If m ≡ 2 ( m o d 4 ) m \equiv2 \pmod{4} m ≡ 2 ( mod 4 ) , let η \eta η be a generator of U U U and define U 5 = { η 5 i : i = 0 , 1 , 2 , … , q − 4 5 } U_5 = \{\eta^{5i}: i=0,1,2,\ldots,\frac{q-4}{5}\} U 5 = { η 5 i : i = 0 , 1 , 2 , … , 5 q − 4 } . Then,
N ( a ) = { 1 , if a ∈ U ∖ U 5 , 5 , if a = 1 , 6 , if a ∈ U 5 ∖ { 1 } . N(a) =
\begin{cases}
1, & \text{if } a \in U \setminus U_5,\\
5, & \text{if } a=1,\\
6, & \text{if } a \in U_5 \setminus\{1\}.
\end{cases} N ( a ) = ⎩ ⎨ ⎧ 1 , 5 , 6 , if a ∈ U ∖ U 5 , if a = 1 , if a ∈ U 5 ∖ { 1 } .
Proof. Similarly we place the proof in Appendix A.
With the above preparations, we can establish two identities for N 4 N_4 N 4 and N 6 N_6 N 6 .
Theorem 4.7. Let q = 2 m q=2^m q = 2 m for an even m m m and N i N_i N i , for i = 0 , 1 , … , 6 i=0,1,\ldots,6 i = 0 , 1 , … , 6 , be defined in (13) . Set ε = 1 \varepsilon=1 ε = 1 if m ≡ 2 ( m o d 4 ) m \equiv2 \pmod{4} m ≡ 2 ( mod 4 ) , and ε = 0 \varepsilon=0 ε = 0 otherwise. Let H 1 = { w ∈ F q ∗ : Tr m ( 1 / w ) = 1 } H_1=\{w \in\mathbb{F}_q^*: \operatorname{Tr}_m(1/w)=1\} H 1 = { w ∈ F q ∗ : Tr m ( 1/ w ) = 1 } and W \mathcal{W} W be a collection of three-element subsets of H 1 H_1 H 1 defined by
W = { { w 1 , w 2 , w 3 } ⊂ H 1 : w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 } . (24) \mathcal{W}=\left\{\{w_1,w_2,w_3\}\subset H_1:w_1w_2w_3=(w_1+w_2+w_3)(w_1+w_2+w_3+1)^2\right\}. \tag*{(24)} W = { { w 1 , w 2 , w 3 } ⊂ H 1 : w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 } . (24) Then we have
2 ∣ W ∣ = N 4 + 4 N 5 + 15 N 6 − q ε , 2 ∣ W 0 ∣ = N 4 . (25) \begin{aligned}
2|\mathcal{W}| &= N_4+4N_5+15N_6-q\varepsilon,\\
2|\mathcal{W}_0| &= N_4. \tag*{(25)}
\end{aligned} 2∣ W ∣ 2∣ W 0 ∣ = N 4 + 4 N 5 + 15 N 6 − q ε , = N 4 . (25) where W t \mathcal{W}_t W t for t = 0 , 1 t=0,1 t = 0 , 1 are subsets of W \mathcal{W} W defined by
W t = { { w 1 , w 2 , w 3 } ∈ W : Tr m ( κ ) = t } . (26) \mathcal{W}_t=\left\{\{w_1,w_2,w_3\}\in\mathcal{W}:\operatorname{Tr}_m(\kappa)=t\right\}. \tag*{(26)} W t = { { w 1 , w 2 , w 3 } ∈ W : Tr m ( κ ) = t } . (26) with
κ = w 1 w 2 + w 1 w 3 + w 2 w 3 + ( w 1 + w 2 + w 3 ) 2 ( w 1 + w 2 + w 3 + 1 ) 2 . (27) \kappa=\frac{w_1w_2+w_1w_3+w_2w_3+(w_1+w_2+w_3)^2}{(w_1+w_2+w_3+1)^2}. \tag*{(27)} κ = ( w 1 + w 2 + w 3 + 1 ) 2 w 1 w 2 + w 1 w 3 + w 2 w 3 + ( w 1 + w 2 + w 3 ) 2 . (27)
Proof. According to Proposition 4.3 and Remark 3, we see that 4-subsets S = { u 1 , u 2 , u 3 , u 4 } S = \{u_1,u_2,u_3,u_4\} S = { u 1 , u 2 , u 3 , u 4 } of U U U satisfying (16) determine all elements a ∈ F q 2 ∗ a \in\mathbb{F}_{q^2}^{*} a ∈ F q 2 ∗ for which P a ( x ) = x 6 + a x 5 + a ‾ x + 1 P_a(x) = x^6 + ax^5 + \overline{a}x + 1 P a ( x ) = x 6 + a x 5 + a x + 1 has at least 4 distinct roots in U U U through (17) . Then we have the relation between the number of such 4-subsets S S S and N 4 , N 5 , N 6 N_4,N_5,N_6 N 4 , N 5 , N 6 as in (21) .
Meanwhile, the above 4-subsets S S S can be determined by elements ( μ , α , β ) ∈ U × F q 2 ∗ × F q (\mu,\alpha,\beta) \in U \times\mathbb{F}_{q^2}^{*} \times\mathbb{F}_q ( μ , α , β ) ∈ U × F q 2 ∗ × F q satisfying ( α + α ‾ ) 2 = β 3 + β (\alpha+\overline{\alpha})^2 = \beta^3+\beta ( α + α ) 2 = β 3 + β and ( α + β ) μ − 6 = α ‾ + β (\alpha+\sqrt{\beta})\mu^{-6} = \overline{\alpha}+\sqrt{\beta} ( α + β ) μ − 6 = α + β as in (15) , provided that the associated Q ( x ) Q(x) Q ( x ) has four distinct roots in U U U . Below we discuss the cases that give rise to such 4-subsets S S S .
Case 1: β = 0 \beta=0 β = 0 . By ( α + α ‾ ) 2 = β 3 + β (\alpha+\overline{\alpha})^2=\beta^3+\beta ( α + α ) 2 = β 3 + β we have α ∈ F q ∗ \alpha\in\mathbb{F}_q^{*} α ∈ F q ∗ . In this case,
Q ( x ) = x 4 + α x 3 + α x + 1 = ( x + 1 ) 2 ( x 2 + α x + 1 ) . Q(x)=x^4+\alpha x^3+\alpha x+1=(x+1)^2(x^2+\alpha x+1). Q ( x ) = x 4 + α x 3 + α x + 1 = ( x + 1 ) 2 ( x 2 + α x + 1 ) .
implying that 1 1 1 is a repeated root of Q ( x ) Q(x) Q ( x ) and that Q ( x ) Q(x) Q ( x ) cannot have four distinct roots in U U U . This case is thus excluded.
Case 2: β = 1 \beta=1 β = 1 . Again we have α ∈ F q ∗ \alpha\in\mathbb{F}_q^{*} α ∈ F q ∗ and then ( α + β ) μ − 6 = α ‾ + β (\alpha+\sqrt{\beta})\mu^{-6}=\overline{\alpha}+\sqrt{\beta} ( α + β ) μ − 6 = α + β reduces to
( α + 1 ) ( μ − 6 + 1 ) = 0. (\alpha+1)(\mu^{-6}+1)=0. ( α + 1 ) ( μ − 6 + 1 ) = 0.
If α = 1 \alpha=1 α = 1 , then α + β = 0 \alpha+\sqrt{\beta}=0 α + β = 0 , so the second condition imposes no restriction on μ ∈ U \mu\in U μ ∈ U . Otherwise, μ 6 = 1 \mu^6=1 μ 6 = 1 , which forces μ = 1 \mu=1 μ = 1 since gcd ( 6 , q + 1 ) = 1 \gcd(6,q+1)=1 g cd( 6 , q + 1 ) = 1 for even m m m . We consider the following two subcases.
Case 2.1: α = β = 1 \alpha=\beta=1 α = β = 1 . In this case we have
Q ( x ) = x 4 + x 3 + x 2 + x + 1 = x 5 + 1 x + 1 , Q(x)=x^4+x^3+x^2+x+1=\frac{x^5+1}{x+1}, Q ( x ) = x 4 + x 3 + x 2 + x + 1 = x + 1 x 5 + 1 ,
and
μ − 6 P a ( μ x ) = ( x 5 + 1 ) ( x + μ − 6 ) , a = μ − 5 \mu^{-6}P_a(\mu x)=(x^5+1)(x+\mu^{-6}),\qquad a=\mu^{-5} μ − 6 P a ( μx ) = ( x 5 + 1 ) ( x + μ − 6 ) , a = μ − 5
where μ \mu μ is a free element in U U U . In this case, a a a belongs to U U U . By Lemma 4.6 , when m ≡ 2 ( m o d 4 ) m\equiv2\pmod{4} m ≡ 2 ( mod 4 ) , the normalized SCR polynomial Q ( x ) = x 4 + x 3 + x 2 + x + 1 Q(x)=x^4+x^3+x^2+x+1 Q ( x ) = x 4 + x 3 + x 2 + x + 1 has four distinct roots in U U U . Then, P a ( x ) P_a(x) P a ( x ) has exactly 5 distinct roots in U U U if a = 1 a=1 a = 1 (equivalently, μ 5 = 1 \mu^5=1 μ 5 = 1 ) and exactly 6 distinct roots in U U U if a ∈ U 5 ∖ { 1 } a\in U_5\setminus\{1\} a ∈ U 5 ∖ { 1 } , where U 5 U_5 U 5 is defined in Lemma 4.6 . Recall that ε = 1 \varepsilon=1 ε = 1 if m ≡ 2 ( m o d 4 ) m\equiv2\pmod{4} m ≡ 2 ( mod 4 ) and ε = 0 \varepsilon=0 ε = 0 if m ≡ 0 ( m o d 4 ) m\equiv0\pmod{4} m ≡ 0 ( mod 4 ) . It follows that there are 5 ε 5\varepsilon 5 ε elements μ ∈ U \mu\in U μ ∈ U with μ 5 = 1 \mu^5=1 μ 5 = 1 such that N ( a ) = 5 N(a)=5 N ( a ) = 5 and ( q − 4 ) ε (q-4)\varepsilon ( q − 4 ) ε elements μ ∈ U \mu\in U μ ∈ U with μ − 5 ∈ U 5 ∖ { 1 } \mu^{-5}\in U_5\setminus\{1\} μ − 5 ∈ U 5 ∖ { 1 } such that N ( a ) = 6 N(a)=6 N ( a ) = 6 . This case contributes ( q + 1 ) ε (q+1)\varepsilon ( q + 1 ) ε 4-subsets to (21) .
Case 2.2: α ∈ F q ∗ ∖ { 1 } \alpha\in\mathbb{F}_q^{*}\setminus\{1\} α ∈ F q ∗ ∖ { 1 } . Then μ = 1 \mu=1 μ = 1 , a = α a=\alpha a = α , and
Q ( x ) = x 4 + α x 3 + x 2 + α x + 1 , P a ( x ) = ( x + 1 ) 2 Q ( x ) . Q(x)=x^4+\alpha x^3+x^2+\alpha x+1,\qquad P_a(x)=(x+1)^2Q(x). Q ( x ) = x 4 + α x 3 + x 2 + α x + 1 , P a ( x ) = ( x + 1 ) 2 Q ( x ) .
By Lemma 4.4 and Corollary 4.5, there are exactly N 5 = ( q + 1 + K m ( 1 ) ) / 8 N_5=(q+1+K_m(1))/8 N 5 = ( q + 1 + K m ( 1 )) /8 different α \alpha α ’s in F q ∗ \mathbb{F}_q^{*} F q ∗ satisfying α = a = t + t − 1 \alpha=a=t+t^{-1} α = a = t + t − 1 where Tr m ( t ) = Tr m ( t − 1 ) = 1 \operatorname{Tr}_m(t)=\operatorname{Tr}_m(t^{-1})=1 Tr m ( t ) = Tr m ( t − 1 ) = 1 . Moreover, α = 1 \alpha=1 α = 1 satisfies this condition if and only if m ≡ 2 ( m o d 4 ) m\equiv2\pmod{4} m ≡ 2 ( mod 4 ) , as discussed in Lemma 4.6 . Thus, this case contributes in total ( N 5 − ε ) (N_5-\varepsilon) ( N 5 − ε ) 4-subsets to (21) , and all elements a a a in this case satisfy N ( a ) = 5 N(a)=5 N ( a ) = 5 .
Case 3: β ∈ F q ∖ { 0 , 1 } \beta\in\mathbb{F}_q\setminus\{0,1\} β ∈ F q ∖ { 0 , 1 } . In this case, by ( α + α ‾ ) 2 = β 3 + β (\alpha+\overline{\alpha})^2=\beta^3+\beta ( α + α ) 2 = β 3 + β we have α ∉ F q \alpha\notin\mathbb{F}_q α ∈ / F q , and then by ( α + β ) μ − 6 = α ‾ + β (\alpha+\sqrt{\beta})\mu^{-6}=\overline{\alpha}+\sqrt{\beta} ( α + β ) μ − 6 = α + β we have
μ = ( α + β α ‾ + β ) 1 6 . (28) \mu=\left(\frac{\alpha+\sqrt{\beta}}{\overline{\alpha}+\sqrt{\beta}}\right)^{\frac{1}{6}}. \tag*{(28)} μ = ( α + β α + β ) 6 1 . (28)
Here μ \mu μ is uniquely determined by α , β \alpha,\beta α , β . Furthermore, if the quartic SCR polynomial Q ( x ) = x 4 + α x 3 + β x 2 + α ‾ x + 1 Q(x)=x^4+\alpha x^3+\beta x^2+\overline{\alpha}x+1 Q ( x ) = x 4 + α x 3 + β x 2 + α x + 1 has 4 distinct roots in U U U , denoted by u 1 , u 2 , u 3 , u 4 u_1,u_2,u_3,u_4 u 1 , u 2 , u 3 , u 4 , then the set S = { μ u i : i = 1 , 2 , 3 , 4 } S=\{\mu u_i:i=1,2,3,4\} S = { μ u i : i = 1 , 2 , 3 , 4 } is a 4-subset of U U U such that P a ( x ) P_a(x) P a ( x ) for a = μ ( α 2 + β + μ − 6 ) / α a=\mu(\alpha^2+\beta+\mu^{-6})/\alpha a = μ ( α 2 + β + μ − 6 ) / α evaluates to zero on S S S , contributing to the count of 4-subsets on the left-hand side of (21) .
According to Proposition 3.4 , the quartic SCR polynomial Q ( x ) = x 4 + α x 3 + β x 2 + α ‾ x + 1 Q(x)=x^4+\alpha x^3+\beta x^2+\overline{\alpha}x+1 Q ( x ) = x 4 + α x 3 + β x 2 + α x + 1 with α ∉ F q ∗ \alpha\notin\mathbb{F}_q^* α ∈ / F q ∗ has 4 distinct roots u i u_i u i ’s in U U U if and only if its cubic resolvent h ( y ) = y 3 + β y 2 + α α ‾ y + ( α + α ‾ ) 2 h(y)=y^3+\beta y^2+\alpha\overline{\alpha}y+(\alpha+\overline{\alpha})^2 h ( y ) = y 3 + β y 2 + α α y + ( α + α ) 2 has exactly 3 distinct roots w 1 , w 2 , w 3 w_1,w_2,w_3 w 1 , w 2 , w 3 in H 1 H_1 H 1 , which can be given by w 1 = u 1 u 2 + ( u 1 u 2 ) − 1 w_1=u_1u_2+(u_1u_2)^{-1} w 1 = u 1 u 2 + ( u 1 u 2 ) − 1 , w 2 = u 1 u 3 + ( u 1 u 3 ) − 1 w_2=u_1u_3+(u_1u_3)^{-1} w 2 = u 1 u 3 + ( u 1 u 3 ) − 1 , and w 3 = u 1 u 4 + ( u 1 u 4 ) − 1 w_3=u_1u_4+(u_1u_4)^{-1} w 3 = u 1 u 4 + ( u 1 u 4 ) − 1 . Observe that the roots w 1 , w 2 , w 3 w_1,w_2,w_3 w 1 , w 2 , w 3 of h ( y ) = y 3 + β y 2 + α α ‾ y + ( α + α ‾ ) 2 h(y)=y^3+\beta y^2+\alpha\overline{\alpha}y+(\alpha+\overline{\alpha})^2 h ( y ) = y 3 + β y 2 + α α y + ( α + α ) 2 uniquely determines the pair α , α ‾ \alpha,\overline{\alpha} α , α (via the coefficients α α ‾ \alpha\overline{\alpha} α α , α + α ‾ \alpha+\overline{\alpha} α + α ) and the coefficient β \beta β . This means that given three distinct elements w i w_i w i ’s in H 1 H_1 H 1 , they can uniquely determine { α , α ‾ } \{\alpha,\overline{\alpha}\} { α , α } and β \beta β for which h ( y ) = y 3 + β y 2 + α α ‾ y + ( α + α ‾ ) 2 h(y)=y^3+\beta y^2+\alpha\overline{\alpha}y+(\alpha+\overline{\alpha})^2 h ( y ) = y 3 + β y 2 + α α y + ( α + α ) 2 satisfies h ( w i ) = 0 h(w_i)=0 h ( w i ) = 0 for $i=1,2,3, thereby Q ( x ) = x 4 + α x 3 + β x 2 + α ‾ x + 1 Q(x)=x^4+\alpha x^3+\beta x^2+\overline{\alpha}x+1 Q ( x ) = x 4 + α x 3 + β x 2 + α x + 1 has 4 distinct roots in U U U . If the elements α , β \alpha,\beta α , β further satisfy the relation ( α + α ‾ ) 2 = β 3 + β (\alpha+\overline{\alpha})^2=\beta^3+\beta ( α + α ) 2 = β 3 + β which corresponds to the equality w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 w_1w_2w_3=(w_1+w_2+w_3)(w_1+w_2+w_3+1)^2 w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 in the definition of W \mathcal{W} W , then they together with μ \mu μ given by (28) contribute two 4-subsets S S S , namely, one by α , β \alpha,\beta α , β and the other one by α ‾ , β \overline{\alpha},\beta α , β , in (20) . That is to say, by the definition of W \mathcal{W} W , this case contributes $2|\mathcal{W}|$ 4-subsets.
Moreover, as shown in the proof of Proposition 4.3 , in this case we have the factorization
μ − 6 P a ( μ x ) = Q ( x ) ( x 2 + γ x + μ − 6 ) = ( x 4 + α x 3 + β x 2 + α ‾ x + 1 ) ( x 2 + γ x + μ − 6 ) , \mu^{-6}P_a(\mu x)=Q(x)(x^2+\gamma x+\mu^{-6})=(x^4+\alpha x^3+\beta x^2+\overline{\alpha}x+1)(x^2+\gamma x+\mu^{-6}), μ − 6 P a ( μx ) = Q ( x ) ( x 2 + γ x + μ − 6 ) = ( x 4 + α x 3 + β x 2 + α x + 1 ) ( x 2 + γ x + μ − 6 ) ,
where γ = μ − 6 + β α \gamma=\frac{\mu^{-6}+\beta}{\alpha} γ = α μ − 6 + β and μ \mu μ is given in (28) . To determine the number of roots of P a ( x ) P_a(x) P a ( x ) in U U U , it remains to examine the quadratic factor x 2 + γ x + μ − 6 x^2+\gamma x+\mu^{-6} x 2 + γ x + μ − 6 . Under the assumptions of this case, it can be directly verified that μ − 6 = γ / γ ‾ \mu^{-6}=\gamma/\overline{\gamma} μ − 6 = γ / γ . Then, by Corollary 3.2, this quadratic equation has two distinct roots in U U U if Tr m ( α α ‾ + β 2 1 + β 2 ) = 1 \operatorname{Tr}_m\left(\frac{\alpha\overline{\alpha}+\beta^2}{1+\beta^2}\right)=1 Tr m ( 1 + β 2 α α + β 2 ) = 1 , and has no root in U U U otherwise.
Note that α α ‾ + β 2 1 + β 2 \frac{\alpha\overline{\alpha}+\beta^2}{1+\beta^2} 1 + β 2 α α + β 2 can be expressed in terms of w 1 , w 2 , w 3 w_1,w_2,w_3 w 1 , w 2 , w 3 and is exactly the parameter κ \kappa κ defined in (27) . Thus, in this case, each triple ( α , β , μ ) (\alpha,\beta,\mu) ( α , β , μ ) (equivalently, each pair ( α , β ) (\alpha,\beta) ( α , β ) ) determined by { w 1 , w 2 , w 3 } ∈ W 0 \{w_1,w_2,w_3\}\in\mathcal{W}_0 { w 1 , w 2 , w 3 } ∈ W 0 yields a polynomial P a ( x ) P_a(x) P a ( x ) with exactly four distinct roots in U U U .
For each pair ( α , β ) (\alpha,\beta) ( α , β ) determined by { w 1 , w 2 , w 3 } ∈ W 1 \{w_1,w_2,w_3\}\in\mathcal{W}_1 { w 1 , w 2 , w 3 } ∈ W 1 , the factors Q ( x ) Q(x) Q ( x ) and x 2 + γ x + μ − 6 x^2+\gamma x+\mu^{-6} x 2 + γ x + μ − 6 have four and two distinct roots in U U U , respectively. By Lemma 4.4 , these two factors have at most one common root. Hence, the corresponding polynomial P a ( x ) P_a(x) P a ( x ) has either five or six distinct roots in U U U . More precisely, under the above conditions, P a ( x ) P_a(x) P a ( x ) has exactly five distinct roots in U U U if a ∈ F q a\in\mathbb{F}_q a ∈ F q , by Corollary 4.5, and exactly six if a ∉ F q a\notin\mathbb{F}_q a ∈ / F q , by Lemma 4.4 .
Combining the above arguments, Case 1 contributes no 4-subset in (20) , while Cases 2 contributes 5 ϵ + ( q − 4 ) ϵ + N 5 − ϵ = q ϵ + N 5 5\epsilon+(q-4)\epsilon+N_5-\epsilon=q\epsilon+N_5 5 ϵ + ( q − 4 ) ϵ + N 5 − ϵ = q ϵ + N 5 valid 4-subsets in (20) , and Case 3 contributes 2 ∣ W ∣ 2|\mathcal{W}| 2∣ W ∣ valid 4-subsets in (20) . This gives
N 5 + q ϵ + 2 ∣ W ∣ = ∣ S ∣ = N 4 + 5 N 5 + 15 N 6 , N_5+q\epsilon+2|\mathcal{W}|=|S|=N_4+5N_5+15N_6, N 5 + q ϵ + 2∣ W ∣ = ∣ S ∣ = N 4 + 5 N 5 + 15 N 6 ,
which implies the desired equality in (25) .
The analysis in Cases 1–3 further shows that, the triples ( α , β , μ ) (\alpha,\beta,\mu) ( α , β , μ ) that yield polynomials P a ( x ) P_a(x) P a ( x ) having exactly four distinct roots in U U U are precisely the ones determined by sets { w 1 , w 2 , w 3 } ∈ W 0 \{w_1,w_2,w_3\}\in\mathcal{W}_0 { w 1 , w 2 , w 3 } ∈ W 0 . Thus, N 4 = 2 ∣ W 0 ∣ N_4=2|\mathcal{W}_0| N 4 = 2∣ W 0 ∣ .
From the proof of Theorem 4.7 , we have the following characterization of elements a a a ’s such that P a ( x ) P_a(x) P a ( x ) has exactly 4, 5, or 6 roots in U U U , respectively.
Corollary 4.8. Let q = 2 m q=2^m q = 2 m for an even m ≥ 2 m\ge2 m ≥ 2 . Let W \mathcal{W} W , κ \kappa κ and W t \mathcal{W}_t W t be defined as in (24) , (27) and (26) , respectively. For t = 0 , 1 t=0,1 t = 0 , 1 , define
A t = { ( α + β α ‾ + β ) 1 6 ⋅ α 2 + β + α β + 1 α + β : ( α , β ) ∈ Ω t } . (29) \mathcal{A}_t=\left\{\left(\frac{\alpha+\sqrt{\beta}}{\overline{\alpha}+\sqrt{\beta}}\right)^{\frac{1}{6}}\cdot\frac{\alpha^2+\beta+\alpha\sqrt{\beta}+1}{\alpha+\sqrt{\beta}}:(\alpha,\beta)\in\Omega_t\right\}. \tag*{(29)} A t = { ( α + β α + β ) 6 1 ⋅ α + β α 2 + β + α β + 1 : ( α , β ) ∈ Ω t } . (29)
where Ω t \Omega_t Ω t consists of all pairs ( α , β ) (\alpha,\beta) ( α , β ) obtained from { w 1 , w 2 , w 3 } ∈ W t \{w_1,w_2,w_3\}\in\mathcal{W}_t { w 1 , w 2 , w 3 } ∈ W t by setting β = w 1 + w 2 + w 3 \beta=w_1+w_2+w_3 β = w 1 + w 2 + w 3 and taking α \alpha α to be a root of x 2 + ( w 1 w 2 w 3 ) x + ( ∑ 1 ≤ i < j ≤ 3 w i w j ) = 0 x^2+\left(\sqrt{w_1w_2w_3}\right)x+\left(\sum_{1\leq i<j\leq3}w_iw_j\right)=0 x 2 + ( w 1 w 2 w 3 ) x + ( ∑ 1 ≤ i < j ≤ 3 w i w j ) = 0 in F q 2 \mathbb{F}_{q^2} F q 2 . Then the following statements holds:
(i) when m ≡ 0 ( m o d 4 ) m\equiv0\pmod{4} m ≡ 0 ( mod 4 ) ,
(a) P a ( x ) P_a(x) P a ( x ) has exactly four distinct roots in U U U iff a ∈ A 0 a\in\mathcal{A}_0 a ∈ A 0 , and here ∣ A 0 ∣ = 2 ∣ W 0 ∣ |\mathcal{A}_0|=2|\mathcal{W}_0| ∣ A 0 ∣ = 2∣ W 0 ∣ ;
(b) P a ( x ) P_a(x) P a ( x ) has exactly five distinct roots in U U U iff a ∈ { t + t − 1 : t , t − 1 ∈ H 1 } a\in\{t+t^{-1}:t,t^{-1}\in H_1\} a ∈ { t + t − 1 : t , t − 1 ∈ H 1 } ;
(c) P a ( x ) P_a(x) P a ( x ) has exactly six distinct roots in U U U iff a ∈ A 1 ∖ F q a\in\mathcal{A}_1\setminus\mathbb{F}_q a ∈ A 1 ∖ F q ;
(ii) when m ≡ 2 ( m o d 4 ) m\equiv2\pmod{4} m ≡ 2 ( mod 4 ) , besides the above cases, P a ( x ) P_a(x) P a ( x ) also has five distinct roots if a = 1 a=1 a = 1 and has six distinct roots if a ∈ { η 5 i : i = 1 , … , q − 4 5 } a\in\{\eta^{5i}:i=1,\ldots,\frac{q-4}{5}\} a ∈ { η 5 i : i = 1 , … , 5 q − 4 } , where η \eta η is a generator of the unit circle U U U .
Proof. The proof is placed in Appendix A.
By the results of this section, determining N 4 N_4 N 4 and N 6 N_6 N 6 now reduces to computing the cardinalities of W 0 \mathcal{W}_0 W 0 and W 1 \mathcal{W}_1 W 1 . The next section is devoted to this task.
Determination of the correlation distribution
This section will determine the cardinality of W t \mathcal{W}_t W t for t = 0 , 1 t=0,1 t = 0 , 1 , thereby obtaining N 4 N_4 N 4 and N 6 N_6 N 6 and completing the proof of Theorem 1.1 on the correlation distribution for the Niho exponent d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 with 4 s ≡ 5 ( m o d q + 1 ) 4s\equiv5\pmod{q+1} 4 s ≡ 5 ( mod q + 1 ) .
We shall derive the cardinality of W t \mathcal{W}_t W t from calculations of some character sums. To this end, for each t ∈ { 0 , 1 } t\in\{0,1\} t ∈ { 0 , 1 } , we relate W t \mathcal{W}_t W t to a set of ordered pairs D t \mathcal{D}_t D t , of which the simplified relations facilitate our computations. Consider a set
D = { ( x , y ) ∈ ( F q ∖ { 0 , 1 } ) 2 : x ≠ y , x y ≠ 1 } . (30) \mathcal{D}=\{(x,y)\in(\mathbb{F}_q\setminus\{0,1\})^2:x\ne y,\ xy\ne1\}. \tag*{(30)} D = {( x , y ) ∈ ( F q ∖ { 0 , 1 } ) 2 : x = y , x y = 1 } . (30)
For each t ∈ { 0 , 1 } t\in\{0,1\} t ∈ { 0 , 1 } , define
D t = { ( x , y ) ∈ D : Tr m ( 1 + r x ) = Tr m ( 1 + r y ) = Tr m ( 1 + r z ) = 1 , Tr m ( κ ) = t } . (31) \mathcal{D}_t=\left\{(x,y)\in\mathcal{D}:\operatorname{Tr}_m\left(\frac{1+r}{x}\right)=\operatorname{Tr}_m\left(\frac{1+r}{y}\right)=\operatorname{Tr}_m\left(\frac{1+r}{z}\right)=1,\ \operatorname{Tr}_m(\kappa)=t\right\}. \tag*{(31)} D t = { ( x , y ) ∈ D : Tr m ( x 1 + r ) = Tr m ( y 1 + r ) = Tr m ( z 1 + r ) = 1 , Tr m ( κ ) = t } . (31)
with
z = x + y 1 + x y , r = x y z , and κ = ( x y + x z + y z + r 2 ) 2 . (32) z=\frac{x+y}{1+xy},\quad r=xyz,\quad\text{and}\quad\kappa=(xy+xz+yz+r^2)^2. \tag*{(32)} z = 1 + x y x + y , r = x y z , and κ = ( x y + x z + y z + r 2 ) 2 . (32)
The following lemma shows the relation between the cardinalities of W t \mathcal{W}_t W t and D t \mathcal{D}_t D t .
Lemma 5.1. For t ∈ { 0 , 1 } t\in\{0,1\} t ∈ { 0 , 1 } , let W t \mathcal{W}_t W t be defined by (24) and (26) , and D t \mathcal{D}_t D t be defined by (30) and (31) . Then, ∣ W t ∣ = ∣ D t ∣ / 6 |\mathcal{W}_t|=|\mathcal{D}_t|/6 ∣ W t ∣ = ∣ D t ∣/6 for t = 0 , 1 t=0,1 t = 0 , 1 .
Proof. The proof proceeds with the defining conditions for W t \mathcal{W}_t W t and D t \mathcal{D}_t D t , and is placed in Appendix A.
In the sequel we will determine the cardinalities of D t \mathcal{D}_t D t for t = 0 , 1 t=0,1 t = 0 , 1 .
Calculation of the cardinality of D t D_t D t by Kloosterman sums
In this subsection, we compute the cardinalities of D 0 D_0 D 0 and D 1 D_1 D 1 . For each ( x , y ) ∈ D (x,y) \in D ( x , y ) ∈ D , define
F 1 ( x , y ) = r + 1 x , F 2 ( x , y ) = r + 1 y , F 3 ( x , y ) = r + 1 z , F_1(x,y)=\frac{r+1}{x},\qquad F_2(x,y)=\frac{r+1}{y},\qquad F_3(x,y)=\frac{r+1}{z}, F 1 ( x , y ) = x r + 1 , F 2 ( x , y ) = y r + 1 , F 3 ( x , y ) = z r + 1 ,
and
F 4 ( x , y ) = ( x y + x z + y z + r 2 ) 2 = κ . (33) F_4(x,y)=\left(xy+xz+yz+r^2\right)^2=\kappa. \tag*{(33)} F 4 ( x , y ) = ( x y + x z + y z + r 2 ) 2 = κ . (33)
where z = x + y 1 + x y z=\frac{x+y}{1+xy} z = 1 + x y x + y . According to (31) , ∣ D t ∣ |D_t| ∣ D t ∣ can be expressed as
∣ D t ∣ = 1 16 ∑ ( x , y ) ∈ D ( 1 − χ ( F 1 ) ) ( 1 − χ ( F 2 ) ) ( 1 − χ ( F 3 ) ) ( 1 + ( − 1 ) t χ ( F 4 ) ) . (34) |D_t|=\frac{1}{16}\sum_{(x,y)\in D}\left(1-\chi(F_1)\right)\left(1-\chi(F_2)\right)\left(1-\chi(F_3)\right)\left(1+(-1)^t\chi(F_4)\right). \tag*{(34)} ∣ D t ∣ = 16 1 ( x , y ) ∈ D ∑ ( 1 − χ ( F 1 ) ) ( 1 − χ ( F 2 ) ) ( 1 − χ ( F 3 ) ) ( 1 + ( − 1 ) t χ ( F 4 ) ) . (34)
where F i = F i ( x , y ) F_i=F_i(x,y) F i = F i ( x , y ) and t ∈ { 0 , 1 } t\in\{0,1\} t ∈ { 0 , 1 } and χ ( ⋅ ) = ( − 1 ) Tr m ( ⋅ ) \chi(\cdot)=(-1)^{\operatorname{Tr}_m(\cdot)} χ ( ⋅ ) = ( − 1 ) Tr m ( ⋅ ) is the canonical additive character on F q \mathbb{F}_q F q . Expanding the product in (34) yields sixteen character sums. Symmetry in x x x , y y y , z z z groups them into eight classes, and their relations with ∣ D t ∣ |D_t| ∣ D t ∣ is given below.
Proposition 5.2. Let F i ( x , y ) F_i(x,y) F i ( x , y ) be the functions defined in (33) , 1 ≤ i ≤ 4 1\leq i\leq4 1 ≤ i ≤ 4 . For 0 ≤ j ≤ 3 0\leq j\leq3 0 ≤ j ≤ 3 and t ∈ { 0 , 1 } t\in\{0,1\} t ∈ { 0 , 1 } , denote
S j , t = ∑ ( x , y ) ∈ D χ ( ∑ i = 1 j F i ( x , y ) + t F 4 ( x , y ) ) , (35) S_{j,t}=\sum_{(x,y)\in D}\chi\left(\sum_{i=1}^{j}F_i(x,y)+tF_4(x,y)\right), \tag*{(35)} S j , t = ( x , y ) ∈ D ∑ χ ( i = 1 ∑ j F i ( x , y ) + t F 4 ( x , y ) ) , (35) where the notation S 0 , t S_{0,t} S 0 , t is used by assuming ∑ i = 1 0 F i ( x , y ) = 0 \sum_{i=1}^{0}F_i(x,y)=0 ∑ i = 1 0 F i ( x , y ) = 0 . Then we have
∣ D t ∣ = 1 16 ∑ j = 0 3 ( − 1 ) j ( 3 j ) ( S j , 0 + ( − 1 ) t S j , 1 ) , t ∈ { 0 , 1 } . (36) |D_t|=\frac{1}{16}\sum_{j=0}^{3}(-1)^j\binom{3}{j}\left(S_{j,0}+(-1)^tS_{j,1}\right),\qquad t\in\{0,1\}. \tag*{(36)} ∣ D t ∣ = 16 1 j = 0 ∑ 3 ( − 1 ) j ( j 3 ) ( S j , 0 + ( − 1 ) t S j , 1 ) , t ∈ { 0 , 1 } . (36)
Proof. Expand the product in (34) using χ ( u ) χ ( v ) = χ ( u + v ) \chi(u)\chi(v)=\chi(u+v) χ ( u ) χ ( v ) = χ ( u + v ) . As shown in the proof of Lemma 5.1 , permutations of x , y , z x,y,z x , y , z induce bijections of V t V_t V t , permute F 1 , F 2 , F 3 F_1,F_2,F_3 F 1 , F 2 , F 3 , and preserve F 4 F_4 F 4 . Thus, we have the following identities
(i) S 1 , t = ∑ ( x , y ) ∈ D χ ( F 1 + t F 4 ) = ∑ ( x , y ) ∈ D χ ( F 2 + t F 4 ) = ∑ ( x , y ) ∈ D χ ( F 3 + t F 4 ) ; \text{(i)}\quad S_{1,t}=\sum_{(x,y)\in D}\chi(F_1+tF_4)=\sum_{(x,y)\in D}\chi(F_2+tF_4)=\sum_{(x,y)\in D}\chi(F_3+tF_4); (i) S 1 , t = ( x , y ) ∈ D ∑ χ ( F 1 + t F 4 ) = ( x , y ) ∈ D ∑ χ ( F 2 + t F 4 ) = ( x , y ) ∈ D ∑ χ ( F 3 + t F 4 ) ;
(ii) S 2 , t = ∑ ( x , y ) ∈ D χ ( F 1 + F 2 + t F 4 ) = ∑ ( x , y ) ∈ D χ ( F 1 + F 3 + t F 4 ) = ∑ ( x , y ) ∈ D χ ( F 2 + F 3 + t F 4 ) , \text{(ii)}\quad S_{2,t}=\sum_{(x,y)\in D}\chi(F_1+F_2+tF_4)=\sum_{(x,y)\in D}\chi(F_1+F_3+tF_4)=\sum_{(x,y)\in D}\chi(F_2+F_3+tF_4), (ii) S 2 , t = ( x , y ) ∈ D ∑ χ ( F 1 + F 2 + t F 4 ) = ( x , y ) ∈ D ∑ χ ( F 1 + F 3 + t F 4 ) = ( x , y ) ∈ D ∑ χ ( F 2 + F 3 + t F 4 ) ,
where t ∈ { 0 , 1 } t\in\{0,1\} t ∈ { 0 , 1 } . This shows that for each 0 ≤ j ≤ 3 0\leq j\leq3 0 ≤ j ≤ 3 , j j j functions out of F 1 , F 2 , F 3 F_1,F_2,F_3 F 1 , F 2 , F 3 give character sums with the same value, namely, S j , 0 S_{j,0} S j , 0 without F 4 F_4 F 4 and S j , 1 S_{j,1} S j , 1 with F 4 F_4 F 4 , and there are ( 3 j ) \binom{3}{j} ( j 3 ) such choices. Each term has coefficient ( − 1 ) j (-1)^j ( − 1 ) j , with an additional factor ( − 1 ) t (-1)^t ( − 1 ) t when F 4 F_4 F 4 is included. Consequently,
16 ∣ D t ∣ = ( S 0 , 0 + ( − 1 ) t S 0 , 1 ) − 3 ( S 1 , 0 + ( − 1 ) t S 1 , 1 ) + 3 ( S 2 , 0 + ( − 1 ) t S 2 , 1 ) − ( S 3 , 0 + ( − 1 ) t S 3 , 1 ) . 16|D_t|=\left(S_{0,0}+(-1)^tS_{0,1}\right)-3\left(S_{1,0}+(-1)^tS_{1,1}\right)+3\left(S_{2,0}+(-1)^tS_{2,1}\right)-\left(S_{3,0}+(-1)^tS_{3,1}\right). 16∣ D t ∣ = ( S 0 , 0 + ( − 1 ) t S 0 , 1 ) − 3 ( S 1 , 0 + ( − 1 ) t S 1 , 1 ) + 3 ( S 2 , 0 + ( − 1 ) t S 2 , 1 ) − ( S 3 , 0 + ( − 1 ) t S 3 , 1 ) .
This yields the desired formula (36) .
In the sequel we shall evaluate the eight character sums
S 0 , 0 , S 1 , 0 , S 2 , 0 , S 1 , 1 , S 0 , 1 , S 2 , 1 , S 3 , 1 , S 3 , 0 . S_{0,0},\ S_{1,0},\ S_{2,0},\ S_{1,1},\ S_{0,1},\ S_{2,1},\ S_{3,1},\ S_{3,0}. S 0 , 0 , S 1 , 0 , S 2 , 0 , S 1 , 1 , S 0 , 1 , S 2 , 1 , S 3 , 1 , S 3 , 0 .
Four of them will be evaluated directly, while the other four will be expressed by three mixed Kloosterman sums. We observe that the expressions of F 1 F_1 F 1 , F 2 F_2 F 2 , F 3 F_3 F 3 and F 4 F_4 F 4 above commonly have x + y x+y x + y , x y xy x y . This motivates us to apply the following sum-product substitution :
( x , y ) ↔ ( s , v ) with s = x + y , v 2 = x y , (37) (x,y) \leftrightarrow(s,v) \text{ with } s=x+y,\ v^2=xy, \tag*{(37)} ( x , y ) ↔ ( s , v ) with s = x + y , v 2 = x y , (37)
to facilitate the calculation of S j , t S_{j,t} S j , t . Meanwhile, the aforementioned symmetric property allows us to choose any j j j of functions F 1 F_1 F 1 , F 2 F_2 F 2 , F 3 F_3 F 3 when calculating S j , t S_{j,t} S j , t . As the functions F 3 F_3 F 3 , F 1 + F 2 F_1+F_2 F 1 + F 2 can be completely expressed in terms of x + y x+y x + y , x y xy x y , we will use them as representative functions in calculating S j , t S_{j,t} S j , t . For instance we will consider
S 1 , 0 = ∑ ( x , y ) ∈ D χ ( F 3 ) , S 2 , 0 = ∑ ( x , y ) ∈ D χ ( F 1 + F 2 ) and S 1 , 1 = ∑ ( x , y ) ∈ D χ ( F 3 + F 4 ) . S_{1,0}=\sum_{(x,y)\in D}\chi(F_3),\quad S_{2,0}=\sum_{(x,y)\in D}\chi(F_1+F_2)\text{ and }S_{1,1}=\sum_{(x,y)\in D}\chi(F_3+F_4). S 1 , 0 = ( x , y ) ∈ D ∑ χ ( F 3 ) , S 2 , 0 = ( x , y ) ∈ D ∑ χ ( F 1 + F 2 ) and S 1 , 1 = ( x , y ) ∈ D ∑ χ ( F 3 + F 4 ) .
In addition, we need to consider character sums of the following four functions
F 4 , F 1 + F 2 + F 3 , F 1 + F 2 + F 4 , F 1 + F 2 + F 3 + F 4 F_4,\ F_1+F_2+F_3,\ F_1+F_2+F_4,\ F_1+F_2+F_3+F_4 F 4 , F 1 + F 2 + F 3 , F 1 + F 2 + F 4 , F 1 + F 2 + F 3 + F 4
which can be expressed in terms of rational functions in s s s and v v v .
Let F q ( X , Y ) \mathbb{F}_q(X,Y) F q ( X , Y ) denote the rational function field in two indeterminates X X X and Y Y Y over F q \mathbb{F}_q F q . The following lemma provides a general method for evaluating those character sums.
Lemma 5.3. Let R ( X , Y ) ∈ F q ( X , Y ) R(X,Y)\in\mathbb{F}_q(X,Y) R ( X , Y ) ∈ F q ( X , Y ) and let D D D be defined in (30) . Suppose that, for all v ∈ F q ∖ { 0 , 1 } v\in\mathbb{F}_q\setminus\{0,1\} v ∈ F q ∖ { 0 , 1 } and s ∈ F q ∗ s\in\mathbb{F}_q^* s ∈ F q ∗ , R ( s , v 2 ) R(s,v^2) R ( s , v 2 ) is well defined and satisfies
χ ( R ( s , v 2 ) ) = χ ( a s + b s + c ) , \chi\left(R(s,v^2)\right)=\chi\left(as+\frac{b}{s}+c\right), χ ( R ( s , v 2 ) ) = χ ( a s + s b + c ) , where a a a , b b b , and c c c are functions of v v v . Then,
∑ ( x , y ) ∈ D χ ( R ( x + y , x y ) ) = ∑ v ≠ 0 , 1 χ ( c ) [ K m ( a , b ) + K m ( a , b + v ) ] − 2 ∑ v ≠ 0 , 1 χ ( a ( v 2 + 1 ) + b v 2 + 1 + c ) . (38) \sum_{(x,y)\in D}\chi\left(R(x+y,xy)\right)
=
\sum_{v\ne0,1}\chi(c)\left[K_m(a,b)+K_m(a,b+v)\right]
-2\sum_{v\ne0,1}\chi\left(a(v^2+1)+\frac{b}{v^2+1}+c\right). \tag*{(38)} ( x , y ) ∈ D ∑ χ ( R ( x + y , x y ) ) = v = 0 , 1 ∑ χ ( c ) [ K m ( a , b ) + K m ( a , b + v ) ] − 2 v = 0 , 1 ∑ χ ( a ( v 2 + 1 ) + v 2 + 1 b + c ) . (38) where K m ( ⋅ , ⋅ ) K_m(\cdot,\cdot) K m ( ⋅ , ⋅ ) denotes the Kloosterman sum defined in (3) .
Proof. By the relation s = x + y s=x+y s = x + y and v 2 = x y v^2=xy v 2 = x y , each pair ( x , y ) ∈ D (x,y)\in D ( x , y ) ∈ D uniquely determines ( s , v ) (s,v) ( s , v ) . Since x + y ≠ 0 x+y\ne0 x + y = 0 , we have s ≠ 0 s\ne0 s = 0 . Conversely, for given s ≠ 0 s\ne0 s = 0 and v v v , the coordinates x , y x,y x , y are the roots of
Z 2 + s Z + v 2 = 0. Z^2+sZ+v^2=0. Z 2 + s Z + v 2 = 0.
Thus the number of pairs ( x , y ) (x,y) ( x , y ) given by this quadratic equation is 1 + χ ( v 2 s 2 ) = 1 + χ ( v s ) 1+\chi\left(\frac{v^2}{s^2}\right)=1+\chi\left(\frac{v}{s}\right) 1 + χ ( s 2 v 2 ) = 1 + χ ( s v ) . Under this correspondence, the conditions on x , y x,y x , y defining D D D become
v ≠ 0 , 1 , s ≠ 0 , v 2 + 1. (39) v\ne0,1,\quad s\ne0,\quad v^2+1. \tag*{(39)} v = 0 , 1 , s = 0 , v 2 + 1. (39)
Indeed, v ≠ 0 v\ne0 v = 0 is equivalent to x y ≠ 0 xy\ne0 x y = 0 , v ≠ 1 v\ne1 v = 1 is equivalent to x y ≠ 1 xy\ne1 x y = 1 , and s ≠ 0 s\ne0 s = 0 is equivalent to x ≠ y x\ne y x = y . Moreover, one of x , y x,y x , y equals 1 1 1 occurs when s = v 2 + 1 s=v^2+1 s = v 2 + 1 . Therefore,
∑ ( x , y ) ∈ D χ ( R ( x + y , x y ) ) = ∑ v ≠ 0 , 1 ∑ s ≠ 0 , v 2 + 1 ( 1 + χ ( v s ) ) χ ( a s + b s + c ) . \sum_{(x,y)\in D}\chi\left(R(x+y,xy)\right)
=
\sum_{v\ne0,1}\sum_{s\ne0,v^2+1}
\left(1+\chi\left(\frac{v}{s}\right)\right)
\chi\left(as+\frac{b}{s}+c\right). ( x , y ) ∈ D ∑ χ ( R ( x + y , x y ) ) = v = 0 , 1 ∑ s = 0 , v 2 + 1 ∑ ( 1 + χ ( s v ) ) χ ( a s + s b + c ) .
For fixed v ≠ 0 , 1 v \ne0,1 v = 0 , 1 , extending the inner sum to all s ≠ 0 s \ne0 s = 0 yields
∑ s ≠ 0 ( 1 + χ ( v s ) ) χ ( a s + b s + c ) = χ ( c ) [ ∑ s ≠ 0 χ ( a s + b s ) + ∑ s ≠ 0 χ ( a s + b + v s ) ] = χ ( c ) [ K m ( a , b ) + K m ( a , b + v ) ] . \sum_{s\ne0}\left(1+\chi\left(\frac{v}{s}\right)\right)\chi\left(as+\frac{b}{s}+c\right)
=
\chi(c)\left[\sum_{s\ne0}\chi\left(as+\frac{b}{s}\right)+\sum_{s\ne0}\chi\left(as+\frac{b+v}{s}\right)\right]
=
\chi(c)\left[K_m(a,b)+K_m(a,b+v)\right]. s = 0 ∑ ( 1 + χ ( s v ) ) χ ( a s + s b + c ) = χ ( c ) s = 0 ∑ χ ( a s + s b ) + s = 0 ∑ χ ( a s + s b + v ) = χ ( c ) [ K m ( a , b ) + K m ( a , b + v ) ] .
It remains to subtract the contribution from s = v 2 + 1 s=v^2+1 s = v 2 + 1 . When s = v 2 + 1 s=v^2+1 s = v 2 + 1 , χ ( v s ) = χ ( v v 2 + 1 ) = 1 \chi\left(\frac{v}{s}\right)=\chi\left(\frac{v}{v^2+1}\right)=1 χ ( s v ) = χ ( v 2 + 1 v ) = 1 since Tr m ( v v 2 + 1 ) = Tr m ( 1 v + 1 + 1 ( v + 1 ) 2 ) = 0 \operatorname{Tr}_m\left(\frac{v}{v^2+1}\right)=\operatorname{Tr}_m\left(\frac{1}{v+1}+\frac{1}{(v+1)^2}\right)=0 Tr m ( v 2 + 1 v ) = Tr m ( v + 1 1 + ( v + 1 ) 2 1 ) = 0 . Their total contribution is 2 χ ( a ( v 2 + 1 ) + b v 2 + 1 + c ) 2\chi\left(a(v^2+1)+\frac{b}{v^2+1}+c\right) 2 χ ( a ( v 2 + 1 ) + v 2 + 1 b + c ) . Subtracting this contribution and summing over v ≠ 0 , 1 v\ne0,1 v = 0 , 1 , we get the desired identity.
Proposition 5.4. For even m ≥ 2 m\ge2 m ≥ 2 and q = 2 m q=2^m q = 2 m , we have
S 0 , 0 = ( q − 2 ) ( q − 4 ) , S 1 , 0 = 2 σ q + 8 , S 2 , 0 = S 1 , 1 = q + 4 − 3 K m ( 1 ) + K m ( ω ) 2 , (40) \begin{aligned}
S_{0,0}&=(q-2)(q-4),\\
S_{1,0}&=2\sigma q+8,\\
S_{2,0}&=S_{1,1}=q+4-3K_m(1)+K_m(\omega)^2, \tag*{(40)}
\end{aligned} S 0 , 0 S 1 , 0 S 2 , 0 = ( q − 2 ) ( q − 4 ) , = 2 σ q + 8 , = S 1 , 1 = q + 4 − 3 K m ( 1 ) + K m ( ω ) 2 , (40) where σ = ( − 1 ) m / 2 \sigma=(-1)^{m/2} σ = ( − 1 ) m /2 and ω ∈ F q \omega\in\mathbb{F}_q ω ∈ F q satisfies ω 2 + ω + 1 = 0 \omega^2+\omega+1=0 ω 2 + ω + 1 = 0 .
Proof. The first identity counts elements in D \mathcal{D} D . The remaining evaluations can be obtained by Lemma 5.3 and the results in Lemmas 2.1 and 2.2 . We prove the formula for S 1 , 0 S_{1,0} S 1 , 0 here and give the analogous proofs for S 1 , 1 S_{1,1} S 1 , 1 and S 2 , 0 S_{2,0} S 2 , 0 in Appendix B.
For S 1 , 0 S_{1,0} S 1 , 0 , substituting z = ( x + y ) / ( x y + 1 ) z=(x+y)/(xy+1) z = ( x + y ) / ( x y + 1 ) into F 3 ( x , y ) F_3(x,y) F 3 ( x , y ) in (33) gives
F 3 = 1 + x y x + y + x y . F_3=\frac{1+xy}{x+y}+xy. F 3 = x + y 1 + x y + x y .
Using the sum-product substitution in (37) , we have
F 3 = 1 + v 2 s + v 2 and χ ( F 3 ) = χ ( v 2 + 1 s + v ) . F_3=\frac{1+v^2}{s}+v^2
\qquad\text{and}\qquad
\chi(F_3)=\chi\left(\frac{v^2+1}{s}+v\right). F 3 = s 1 + v 2 + v 2 and χ ( F 3 ) = χ ( s v 2 + 1 + v ) .
Utilizing Lemma 5.3 , we get a = 0 a=0 a = 0 , b = v 2 + 1 b=v^2+1 b = v 2 + 1 and c = v c=v c = v and
S 1 , 0 = ∑ v ≠ 0 , 1 χ ( v ) [ K m ( 0 , v 2 + 1 ) + K m ( 0 , v 2 + v + 1 ) ] − 2 ∑ v ≠ 0 , 1 χ ( 1 + v ) = ∑ v ≠ 0 , 1 χ ( v ) K m ( 0 ) + ∑ v ≠ 0 , 1 χ ( v ) K m ( 0 ) + q χ ( ω ) + q χ ( ω 2 ) + 4 = 2 × ( − 1 ) × ( − 2 ) + 2 σ q + 4 = 2 σ q + 8 , \begin{aligned}
S_{1,0}
&=\sum_{v\ne0,1}\chi(v)\left[K_m(0,v^2+1)+K_m(0,v^2+v+1)\right]-2\sum_{v\ne0,1}\chi(1+v)\\
&=\sum_{v\ne0,1}\chi(v)K_m(0)+\sum_{v\ne0,1}\chi(v)K_m(0)+q\chi(\omega)+q\chi(\omega^2)+4\\
&=2\times(-1)\times(-2)+2\sigma q+4\\
&=2\sigma q+8,
\end{aligned} S 1 , 0 = v = 0 , 1 ∑ χ ( v ) [ K m ( 0 , v 2 + 1 ) + K m ( 0 , v 2 + v + 1 ) ] − 2 v = 0 , 1 ∑ χ ( 1 + v ) = v = 0 , 1 ∑ χ ( v ) K m ( 0 ) + v = 0 , 1 ∑ χ ( v ) K m ( 0 ) + q χ ( ω ) + q χ ( ω 2 ) + 4 = 2 × ( − 1 ) × ( − 2 ) + 2 σ q + 4 = 2 σ q + 8 ,
where we use the basic identity in (4) , and χ ( ω ) = χ ( ω 2 ) = σ \chi(\omega)=\chi(\omega^2)=\sigma χ ( ω ) = χ ( ω 2 ) = σ .
To simplify the remaining four classes, we continue using the technique presented in Lemma 5.3 . The resulting expressions will be given in terms of the following three mixed Kloosterman sums over F q \mathbb{F}_q F q :
K 1 = ∑ u ≠ 0 χ ( u ) K m ( u + u − 2 ) , (41) K_1=\sum_{u\ne0}\chi(u)K_m\left(u+u^{-2}\right), \tag*{(41)} K 1 = u = 0 ∑ χ ( u ) K m ( u + u − 2 ) , (41)
K 2 = ∑ u ≠ 0 , 1 χ ( u ) K m ( ( u 2 + u + 1 ) 2 u ( u + 1 ) 2 ) , (42) K_2=\sum_{u\ne0,1}\chi(u)K_m\left(\frac{(u^2+u+1)^2}{u(u+1)^2}\right), \tag*{(42)} K 2 = u = 0 , 1 ∑ χ ( u ) K m ( u ( u + 1 ) 2 ( u 2 + u + 1 ) 2 ) , (42)
K 3 = ∑ u ≠ 0 , 1 K m ( ( u 2 + u + 1 ) 3 u 2 ( u + 1 ) 2 ) . (43) K_3=\sum_{u\ne0,1}K_m\left(\frac{(u^2+u+1)^3}{u^2(u+1)^2}\right). \tag*{(43)} K 3 = u = 0 , 1 ∑ K m ( u 2 ( u + 1 ) 2 ( u 2 + u + 1 ) 3 ) . (43)
Proposition 5.5. Let S j , t S_{j,t} S j , t be the character sums given in (35) , where j ∈ { 1 , 2 , 3 } j\in\{1,2,3\} j ∈ { 1 , 2 , 3 } and t ∈ { 0 , 1 } t\in\{0,1\} t ∈ { 0 , 1 } . For even m ≥ 2 m\ge2 m ≥ 2 , we have
S 0 , 1 = 2 σ q + 7 + K 1 , S 2 , 1 = 2 σ q + 6 + K 2 , S 3 , 1 = 2 q + 2 − 3 K m ( 1 ) + K m ( ω ) 2 + K 3 , S 3 , 0 = 2 σ q + 5 + K 1 + K 2 , (44) \begin{aligned}
S_{0,1}&=2\sigma q+7+K_1,\\
S_{2,1}&=2\sigma q+6+K_2,\\
S_{3,1}&=2q+2-3K_m(1)+K_m(\omega)^2+K_3,\\
S_{3,0}&=2\sigma q+5+K_1+K_2, \tag*{(44)}
\end{aligned} S 0 , 1 S 2 , 1 S 3 , 1 S 3 , 0 = 2 σ q + 7 + K 1 , = 2 σ q + 6 + K 2 , = 2 q + 2 − 3 K m ( 1 ) + K m ( ω ) 2 + K 3 , = 2 σ q + 5 + K 1 + K 2 , (44) where σ = ( − 1 ) m / 2 \sigma=(-1)^{m/2} σ = ( − 1 ) m /2 and ω 2 + ω + 1 = 0 \omega^2+\omega+1=0 ω 2 + ω + 1 = 0 .
Proof. The equalities are derived by routine calculations. See details in Appendix B. □
Proposition 5.5 reduces the evaluation of S 0 , 1 S_{0,1} S 0 , 1 , S 2 , 1 S_{2,1} S 2 , 1 , S 3 , 1 S_{3,1} S 3 , 1 , and S 3 , 0 S_{3,0} S 3 , 0 to that of the three mixed Kloosterman sums K 1 \mathbf{K}_1 K 1 , K 2 \mathbf{K}_2 K 2 , and K 3 \mathbf{K}_3 K 3 . The value of K 3 \mathbf{K}_3 K 3 was already given in [18 Lemmas B.3 and B.9] , namely,
K 3 = 2 K m ( ω ) 2 − q . (45) K_3=2K_m(\omega)^2-q. \tag*{(45)} K 3 = 2 K m ( ω ) 2 − q . (45)
In the following subsection we will consider K 1 \mathbf{K}_1 K 1 and K 2 \mathbf{K}_2 K 2 .
Calculations of the mixed Kloosterman sums K 1 K_1 K 1 and K 2 K_2 K 2
This subsection will derive the following two identities:
K 1 = K m ( 1 ) 2 + 2 σ q and K 2 = 2 K m ( 1 ) 2 + 2 ( σ − 1 ) q , (46) K_1=K_m(1)^2+2\sigma q\quad\text{and}\quad K_2=2K_m(1)^2+2(\sigma-1)q, \tag*{(46)} K 1 = K m ( 1 ) 2 + 2 σ q and K 2 = 2 K m ( 1 ) 2 + 2 ( σ − 1 ) q , (46)
where σ = ( − 1 ) m / 2 \sigma=(-1)^{m/2} σ = ( − 1 ) m /2 . The following result is useful for calculating K 1 \mathbf{K}_1 K 1 .
Lemma 5.6. [30 Proposition 3.2] Let q = 2 m q=2^m q = 2 m with m ≥ 1 m\ge1 m ≥ 1 , H 1 = { t ∈ F q ∗ : Tr m ( t − 1 ) = 1 } H_1=\{t\in\mathbb{F}_q^*: \operatorname{Tr}_m(t^{-1})=1\} H 1 = { t ∈ F q ∗ : Tr m ( t − 1 ) = 1 } . and let D k ( x ) D_k(x) D k ( x ) be the first-kind Dickson polynomial. Then D k ( x ) D_k(x) D k ( x ) permutes H 1 H_1 H 1 if and only if gcd ( k , q + 1 ) = 1 \gcd(k,q+1)=1 g cd( k , q + 1 ) = 1 . In particular, when m m m is even, D 3 ( x ) = x 3 + x D_3(x)=x^3+x D 3 ( x ) = x 3 + x permutes H 1 H_1 H 1 .
Proposition 5.7. Let K 1 \mathbf{K}_1 K 1 be the mixed Kloosterman sum defined in (41) . For even m ≥ 2 m\ge2 m ≥ 2 ,
K 1 = K m ( 1 ) 2 + 2 σ q , (47) K_1=K_m(1)^2+2\sigma q, \tag*{(47)} K 1 = K m ( 1 ) 2 + 2 σ q , (47) where σ = ( − 1 ) m / 2 \sigma=(-1)^{m/2} σ = ( − 1 ) m /2 .
Proof. Expanding the Kloosterman sum in (41) yields
K 1 = ∑ u ≠ 0 ∑ c ≠ 0 χ ( u + c + u + u − 2 c ) . K_1=\sum_{u\ne0}\sum_{c\ne0}\chi\left(u+c+\frac{u+u^{-2}}{c}\right). K 1 = u = 0 ∑ c = 0 ∑ χ ( u + c + c u + u − 2 ) .
Let x = v 2 x=v^2 x = v 2 . Then
χ ( u + v 2 + u v 2 + ( u v ) − 2 ) = χ ( v + ( 1 + v − 2 ) u + v − 1 / u ) . \chi\left(u+v^2+\frac{u}{v^2}+(uv)^{-2}\right)=\chi\left(v+(1+v^{-2})u+v^{-1}/u\right). χ ( u + v 2 + v 2 u + ( uv ) − 2 ) = χ ( v + ( 1 + v − 2 ) u + v − 1 / u ) .
Interchanging the sums and using identity (4) give
K 1 = ∑ v ≠ 0 χ ( v ) ∑ u ≠ 0 χ ( ( 1 + v − 2 ) u + v − 1 / u ) = ∑ v ≠ 0 χ ( v ) K m ( v − 1 + v − 3 ) = ∑ t ≠ 0 χ ( t − 1 ) K m ( t 3 + t ) = ∑ t ≠ 0 K m ( t 3 + t ) − 2 ∑ t ∈ H 1 K m ( D 3 ( t ) ) , (48) \begin{aligned}
\mathbf{K}_1&=\sum_{v\ne0}\chi(v)\sum_{u\ne0}\chi\left((1+v^{-2})u+v^{-1}/u\right)\\
&=\sum_{v\ne0}\chi(v)K_m\left(v^{-1}+v^{-3}\right)\\
&=\sum_{t\ne0}\chi(t^{-1})K_m\left(t^3+t\right)\\
&=\sum_{t\ne0}K_m\left(t^3+t\right)-2\sum_{t\in H_1}K_m\left(D_3(t)\right), \tag*{(48)}
\end{aligned} K 1 = v = 0 ∑ χ ( v ) u = 0 ∑ χ ( ( 1 + v − 2 ) u + v − 1 / u ) = v = 0 ∑ χ ( v ) K m ( v − 1 + v − 3 ) = t = 0 ∑ χ ( t − 1 ) K m ( t 3 + t ) = t = 0 ∑ K m ( t 3 + t ) − 2 t ∈ H 1 ∑ K m ( D 3 ( t ) ) , (48)
where t = v − 1 t=v^{-1} t = v − 1 .
By Lemma 2.1 (i) and Lemma 2.2 (ii), the latter sum can be evaluated as follows
2 ∑ t ∈ H 1 K m ( t ) = ∑ t ≠ 0 ( 1 − χ ( t − 1 ) ) K m ( t ) = ∑ t ≠ 0 K m ( t ) − ∑ t ≠ 0 χ ( t − 1 ) K m ( t ) = 1 − ( K m ( 1 ) 2 − q ) . \begin{aligned}
2\sum_{t\in H_1}K_m(t)&=\sum_{t\ne0}\left(1-\chi(t^{-1})\right)K_m(t)\\
&=\sum_{t\ne0}K_m(t)-\sum_{t\ne0}\chi(t^{-1})K_m(t)\\
&=1-\left(K_m(1)^2-q\right).
\end{aligned} 2 t ∈ H 1 ∑ K m ( t ) = t = 0 ∑ ( 1 − χ ( t − 1 ) ) K m ( t ) = t = 0 ∑ K m ( t ) − t = 0 ∑ χ ( t − 1 ) K m ( t ) = 1 − ( K m ( 1 ) 2 − q ) .
Then it follows from Lemma 2.2 (iv) that
K 1 = ( q ( 2 σ + 1 ) + 1 ) − ( q + 1 − K m ( 1 ) 2 ) = K m ( 1 ) 2 + 2 σ q . \mathbf{K}_1=\left(q(2\sigma+1)+1\right)-\left(q+1-K_m(1)^2\right)=K_m(1)^2+2\sigma q. K 1 = ( q ( 2 σ + 1 ) + 1 ) − ( q + 1 − K m ( 1 ) 2 ) = K m ( 1 ) 2 + 2 σ q .
To evaluate K 2 \mathbf{K}_2 K 2 , we need an auxiliary character sum over a cubic surface. This character sum can be expressed in terms of K 2 \mathbf{K}_2 K 2 and K 1 \mathbf{K}_1 K 1 . Meanwhile, its exact value can be calculated from a different perspective. This together with Lemma 5.7 enables us to determine the exactly value of K 2 \mathbf{K}_2 K 2 .
Let X \mathcal{X} X be the cubic surface defined by
X = { ( x , y , z ) ∈ ( F q ∗ ) 3 : x y z = x + y + z } . (49) \mathcal{X}=\left\{(x,y,z)\in(\mathbb{F}_q^*)^3:xyz=x+y+z\right\}. \tag*{(49)} X = { ( x , y , z ) ∈ ( F q ∗ ) 3 : x y z = x + y + z } . (49)
Here, ( x , y , z ) (x,y,z) ( x , y , z ) denotes an ordered triple, and x x x , y y y , and z z z can be a same element in F q ∗ \mathbb{F}_q^* F q ∗ . Define
K = ∑ ( x , y , z ) ∈ X χ ( ( 1 + r ) ( 1 x + 1 y + 1 z ) ) , \mathbf{K}=\sum_{(x,y,z)\in\mathcal{X}}\chi\left((1+r)\left(\frac{1}{x}+\frac{1}{y}+\frac{1}{z}\right)\right), K = ( x , y , z ) ∈ X ∑ χ ( ( 1 + r ) ( x 1 + y 1 + z 1 ) ) ,
where r = x y z r=xyz r = x y z . Using the notation in (33) , K \mathbf{K} K can be rewritten as
K = ∑ ( x , y , z ) ∈ X χ ( F 1 + F 2 + F 3 ) . \mathbf{K}=\sum_{(x,y,z)\in\mathcal{X}}\chi\left(F_1+F_2+F_3\right). K = ( x , y , z ) ∈ X ∑ χ ( F 1 + F 2 + F 3 ) .
Thus, the character sum K \mathbf{K} K over the cubic surface X \mathcal{X} X is closely related to S 3 , 0 S_{3,0} S 3 , 0 , but differs from it in the restrictions on x x x , y y y , and z z z . First, we establish the following relation.
Lemma 5.8. With the notation introduced above, the character sum K \mathbf{K} K over the cubic surface X \mathcal{X} X satisfies
K = K 2 + K 1 + 2 σ q , \mathbf{K}=\mathbf{K}_2+\mathbf{K}_1+2\sigma q, K = K 2 + K 1 + 2 σ q , where K 1 \mathbf{K}_1 K 1 and K 2 \mathbf{K}_2 K 2 are defined in (41) and (42) , respectively.
Proof. Let
P = { ( 1 , 1 , 1 ) } ∪ ⋃ t ≠ 0 , 1 { ( 1 , 1 , t ) , ( 1 , t , 1 ) , ( t , 1 , 1 ) } . \mathcal{P}=\{(1,1,1)\}\cup\bigcup_{t\ne0,1}\{(1,1,t),(1,t,1),(t,1,1)\}. P = {( 1 , 1 , 1 )} ∪ t = 0 , 1 ⋃ {( 1 , 1 , t ) , ( 1 , t , 1 ) , ( t , 1 , 1 )} .
Let S 3 , 0 S_{3,0} S 3 , 0 be the character sum evaluated in Proposition 5.5. Then, we have
K = S 3 , 0 + ∑ ( x , y , z ) ∈ P χ ( F 1 + F 2 + F 3 ) = S 3 , 0 + χ ( 0 ) + 3 ∑ t ≠ 0 , 1 χ ( 1 t ) = S 3 , 0 − 5. \begin{aligned}
\mathbf{K}=S_{3,0}+\sum_{(x,y,z)\in\mathcal{P}}\chi(F_1+F_2+F_3) \\
&=S_{3,0}+\chi(0)+3\sum_{t\ne0,1}\chi\left(\frac{1}{t}\right) \\
&=S_{3,0}-5.
\end{aligned} K = S 3 , 0 + ( x , y , z ) ∈ P ∑ χ ( F 1 + F 2 + F 3 ) = S 3 , 0 + χ ( 0 ) + 3 t = 0 , 1 ∑ χ ( t 1 ) = S 3 , 0 − 5.
This together with the value of S 3 , 0 S_{3,0} S 3 , 0 in Proposition 5.5 gives the desired formula. □
The relation between the character value and the permutation sign used below is the cubic case of Berlekamp’s discriminant criterion [31 ] . It is also a direct consequence of [32 Theorem 1] . It plays an important role in the evaluation of K \mathbf{K} K .
Lemma 5.9. Let q = 2 m q=2^m q = 2 m , where m m m is a positive integer, and let a ∈ F q a\in\mathbb{F}_q a ∈ F q and b ∈ F q ∗ b\in\mathbb{F}_q^* b ∈ F q ∗ . Denote by n n n the number of distinct roots of Z 3 + a Z + b Z^3+aZ+b Z 3 + a Z + b in F q \mathbb{F}_q F q . The number of ordered triples ( z 1 , z 2 , z 3 ) ∈ F q 3 (z_1,z_2,z_3)\in\mathbb{F}_q^3 ( z 1 , z 2 , z 3 ) ∈ F q 3 satisfying Z 3 + a Z + b = ( Z + z 1 ) ( Z + z 2 ) ( Z + z 3 ) Z^3+aZ+b=(Z+z_1)(Z+z_2)(Z+z_3) Z 3 + a Z + b = ( Z + z 1 ) ( Z + z 2 ) ( Z + z 3 ) is
2 n + χ ( 1 + a 3 b 2 ) − 1. 2n+\chi\left(1+\frac{a^3}{b^2}\right)-1. 2 n + χ ( 1 + b 2 a 3 ) − 1.
Using the above result, we can now prove the following result.
Proposition 5.10. Let F ( u ) = u 4 + u 2 + u + u − 1 F(u)=u^4+u^2+u+u^{-1} F ( u ) = u 4 + u 2 + u + u − 1 and K 4 = ∑ u ≠ 0 , 1 K m ( F ( u ) ) \mathbf{K}_4=\sum_{u\ne0,1}K_m(F(u)) K 4 = ∑ u = 0 , 1 K m ( F ( u )) . Then, the character sum K \mathbf{K} K over the cubic surface X \mathcal{X} X in (49) satisfies
K = 3 K 4 + q − 3. \mathbf{K}=3\mathbf{K}_4+q-3. K = 3 K 4 + q − 3.
Proof. We express K \mathbf{K} K in terms of root counts of cubic polynomials over F q \mathbb{F}_q F q . For each point ( x , y , z ) ∈ X (x,y,z)\in\mathcal{X} ( x , y , z ) ∈ X , let
r = x + y + z , b = x y + x z + y z . r=x+y+z,\qquad b=xy+xz+yz. r = x + y + z , b = x y + x z + y z .
Since x y z ≠ 0 xyz\ne0 x y z = 0 , it follows that r = x + y + z = x y z ≠ 0 r=x+y+z=xyz\ne0 r = x + y + z = x y z = 0 . The coordinates x , y , z x,y,z x , y , z are the roots, with multiplicities, of
f r , b ( X ) = X 3 + r X 2 + b X + r . f_{r,b}(X)=X^3+rX^2+bX+r. f r , b ( X ) = X 3 + r X 2 + b X + r .
Conversely, for any r ≠ 0 r\ne0 r = 0 and b ∈ F q b\in\mathbb{F}_q b ∈ F q , every ordering of all three roots of f r , b ( X ) f_{r,b}(X) f r , b ( X ) in F q \mathbb{F}_q F q gives a point of X \mathcal{X} X .
By the substitution X = Z + r X=Z+r X = Z + r , we get
f r , b ( Z + r ) = Z 3 + ( b + r 2 ) Z + r ( b + 1 ) . (50) f_{r,b}(Z+r)=Z^3+(b+r^2)Z+r(b+1). \tag*{(50)} f r , b ( Z + r ) = Z 3 + ( b + r 2 ) Z + r ( b + 1 ) . (50)
This translation preserves both root multiplicities and the number of distinct roots in F q \mathbb{F}_q F q . We can therefore count the ordered triples of roots of f r , b ( X ) f_{r,b}(X) f r , b ( X ) in F q \mathbb{F}_q F q by applying Lemma 5.9 to (50) for b ≠ 1 b \ne1 b = 1 . Note that both f r , b ( X ) f_{r,b}(X) f r , b ( X ) and its translate in (50) have no repeated roots if and only if r ( b + 1 ) ≠ 0 r(b+1) \ne0 r ( b + 1 ) = 0 .
For each r ≠ 0 r \ne0 r = 0 and b ∈ F q b \in\mathbb{F}_q b ∈ F q , let n ( r , b ) n(r,b) n ( r , b ) be the number of distinct roots of f r , b f_{r,b} f r , b in F q \mathbb{F}_q F q , and let N ( r , b ) N(r,b) N ( r , b ) be the number of ordered triples ( x , y , z ) ∈ ( F q ∗ ) 3 (x,y,z) \in(\mathbb{F}_q^*)^3 ( x , y , z ) ∈ ( F q ∗ ) 3 satisfying
f r , b ( X ) = ( X + x ) ( X + y ) ( X + z ) . f_{r,b}(X)=(X+x)(X+y)(X+z). f r , b ( X ) = ( X + x ) ( X + y ) ( X + z ) .
For each ordered triple ( x , y , z ) ∈ X (x,y,z) \in\mathcal{X} ( x , y , z ) ∈ X , define
W ( r , b ) = χ ( ( r + 1 ) ( 1 x + 1 y + 1 z ) ) = χ ( b ( r + 1 ) r ) . (51) W(r,b)=\chi\left((r+1)\left(\frac{1}{x}+\frac{1}{y}+\frac{1}{z}\right)\right)=\chi\left(\frac{b(r+1)}{r}\right). \tag*{(51)} W ( r , b ) = χ ( ( r + 1 ) ( x 1 + y 1 + z 1 ) ) = χ ( r b ( r + 1 ) ) . (51)
Then, we have
K = ∑ r ≠ 0 ∑ b ∈ F q N ( r , b ) W ( r , b ) . \mathbf{K}=\sum_{r\ne0}\sum_{b\in\mathbb{F}_q}N(r,b)W(r,b). K = r = 0 ∑ b ∈ F q ∑ N ( r , b ) W ( r , b ) .
We will consider two cases: b = 1 b=1 b = 1 and b ≠ 1 b\ne1 b = 1 .
Case 1. Suppose b = 1 b=1 b = 1 . Then the factorization f r , 1 ( X ) = ( X + r ) ( X + 1 ) 2 f_{r,1}(X)=(X+r)(X+1)^2 f r , 1 ( X ) = ( X + r ) ( X + 1 ) 2 gives
( n ( r , 1 ) , N ( r , 1 ) ) = { ( 2 , 3 ) , if r ≠ 1 , ( 1 , 1 ) , if r = 1. (52) (n(r,1),N(r,1))=
\begin{cases}
(2,3), & \text{if } r\ne1,\\
(1,1), & \text{if } r=1.
\end{cases} \tag*{(52)} ( n ( r , 1 ) , N ( r , 1 )) = { ( 2 , 3 ) , ( 1 , 1 ) , if r = 1 , if r = 1. (52)
That is to say, N ( r , 1 ) = 2 n ( r , 1 ) − 1 N(r,1)=2n(r,1)-1 N ( r , 1 ) = 2 n ( r , 1 ) − 1 . Thus
∑ r ≠ 0 N ( r , 1 ) W ( r , 1 ) = ∑ r ≠ 0 , 1 N ( r , 1 ) W ( r , 1 ) + N ( 1 , 1 ) W ( 1 , 1 ) . \sum_{r\ne0}N(r,1)W(r,1)=\sum_{r\ne0,1}N(r,1)W(r,1)+N(1,1)W(1,1). r = 0 ∑ N ( r , 1 ) W ( r , 1 ) = r = 0 , 1 ∑ N ( r , 1 ) W ( r , 1 ) + N ( 1 , 1 ) W ( 1 , 1 ) .
According to (52) and (51) , we obtain that
∑ r ≠ 0 N ( r , 1 ) W ( r , 1 ) = 3 ∑ r ≠ 0 , 1 χ ( r − 1 ) + 1 = − 5. (53) \sum_{r\ne0}N(r,1)W(r,1)=3\sum_{r\ne0,1}\chi(r^{-1})+1=-5. \tag*{(53)} r = 0 ∑ N ( r , 1 ) W ( r , 1 ) = 3 r = 0 , 1 ∑ χ ( r − 1 ) + 1 = − 5. (53)
Case 2. Suppose b ≠ 1 b\ne1 b = 1 . Then the constant term of the polynomial in (50) is nonzero, so Lemma 5.9, together with χ ( 1 ) = 1 \chi(1)=1 χ ( 1 ) = 1 , gives
N ( r , b ) = 2 n ( r , b ) − 1 + χ ( ( r 2 + b ) 3 r 2 ( b + 1 ) 2 ) . (54) N(r,b)=2n(r,b)-1+\chi\left(\frac{(r^2+b)^3}{r^2(b+1)^2}\right). \tag*{(54)} N ( r , b ) = 2 n ( r , b ) − 1 + χ ( r 2 ( b + 1 ) 2 ( r 2 + b ) 3 ) . (54)
We need to investigate the following sum
∑ r ≠ 0 ∑ b ≠ 1 N ( r , b ) W ( r , b ) = 2 ∑ r ≠ 0 ∑ b ≠ 1 W ( r , b ) n ( r , b ) + ∑ r ≠ 0 ∑ b ≠ 1 W ( r , b ) χ ( ( r 2 + b ) 3 r 2 ( b + 1 ) 2 ) − ∑ r ≠ 0 ∑ b ≠ 1 W ( r , b ) . (55) \begin{aligned}
\sum_{r\ne0}\sum_{b\ne1}N(r,b)W(r,b)
={}&2\sum_{r\ne0}\sum_{b\ne1}W(r,b)n(r,b)\\
&+\sum_{r\ne0}\sum_{b\ne1}W(r,b)\chi\left(\frac{(r^2+b)^3}{r^2(b+1)^2}\right)
-\sum_{r\ne0}\sum_{b\ne1}W(r,b). \tag*{(55)}
\end{aligned} r = 0 ∑ b = 1 ∑ N ( r , b ) W ( r , b ) = 2 r = 0 ∑ b = 1 ∑ W ( r , b ) n ( r , b ) + r = 0 ∑ b = 1 ∑ W ( r , b ) χ ( r 2 ( b + 1 ) 2 ( r 2 + b ) 3 ) − r = 0 ∑ b = 1 ∑ W ( r , b ) . (55)
It remains to calculate each term in (55) . We first evaluate the three inner sums over b ≠ 1 b\ne1 b = 1 in (55) for fixed r ≠ 0 , 1 r\ne0,1 r = 0 , 1 . We then treat r = 1 r=1 r = 1 , where W ( 1 , b ) = 1 W(1,b)=1 W ( 1 , b ) = 1 , before summing over all r ≠ 0 r\ne0 r = 0 . The computations are divided into four parts as follows.
Part (i): Calculating ∑ b ≠ 1 W ( r , b ) n ( r , b ) \sum_{b\ne1}W(r,b)n(r,b) ∑ b = 1 W ( r , b ) n ( r , b ) for given r ≠ 0 , 1 r\ne0,1 r = 0 , 1 .
Now fix r ≠ 0 , 1 r \ne0,1 r = 0 , 1 . Then, we have
∑ b ≠ 1 W ( r , b ) n ( r , b ) = ∑ b ∈ F q W ( r , b ) n ( r , b ) − W ( r , 1 ) n ( r , 1 ) = ∑ b ∈ F q W ( r , b ) n ( r , b ) − 2 χ ( 1 / r ) , (56) \begin{aligned}
\sum_{b\ne1} W(r,b)n(r,b) &= \sum_{b\in\mathbb{F}_q} W(r,b)n(r,b)-W(r,1)n(r,1) \\
&= \sum_{b\in\mathbb{F}_q} W(r,b)n(r,b)-2\chi(1/r), \tag*{(56)}
\end{aligned} b = 1 ∑ W ( r , b ) n ( r , b ) = b ∈ F q ∑ W ( r , b ) n ( r , b ) − W ( r , 1 ) n ( r , 1 ) = b ∈ F q ∑ W ( r , b ) n ( r , b ) − 2 χ ( 1/ r ) , (56)
where the second equality follows from (52) .
Since f r , b ( 0 ) = r ≠ 0 f_{r,b}(0)=r\ne0 f r , b ( 0 ) = r = 0 ,
n ( r , b ) = ∣ { x ∈ F q ∗ : f r , b ( x ) = 0 } ∣ . n(r,b)=\left|\{x\in\mathbb{F}_q^*:f_{r,b}(x)=0\}\right|. n ( r , b ) = { x ∈ F q ∗ : f r , b ( x ) = 0 } .
Expanding this root count and interchanging the finite sums gives
∑ b ∈ F q W ( r , b ) n ( r , b ) = ∑ b ∈ F q ∑ x ≠ 0 f r , b ( x ) = 0 W ( r , b ) = ∑ x ≠ 0 f r , b ( x ) = 0 ∑ b ∈ F q W ( r , b ) . (57) \sum_{b\in\mathbb{F}_q}W(r,b)n(r,b)
=\sum_{b\in\mathbb{F}_q}\sum_{\substack{x\ne0\\f_{r,b}(x)=0}}W(r,b)
=\sum_{\substack{x\ne0\\f_{r,b}(x)=0}}\sum_{b\in\mathbb{F}_q}W(r,b). \tag*{(57)} b ∈ F q ∑ W ( r , b ) n ( r , b ) = b ∈ F q ∑ x = 0 f r , b ( x ) = 0 ∑ W ( r , b ) = x = 0 f r , b ( x ) = 0 ∑ b ∈ F q ∑ W ( r , b ) . (57)
For each x ≠ 0 x\ne0 x = 0 , the equation f r , b ( x ) = 0 f_{r,b}(x)=0 f r , b ( x ) = 0 determines exactly one value of b b b , namely
b = x 2 + r x + r x . b=x^2+rx+\frac{r}{x}. b = x 2 + r x + x r .
Thus the pairs ( b , x ) (b,x) ( b , x ) with f r , b ( x ) = 0 f_{r,b}(x)=0 f r , b ( x ) = 0 are in bijection with F q ∗ \mathbb{F}_q^* F q ∗ via their second coordinate. Substitution of this value of b b b gives
W ( r , b ) = χ ( r + 1 r x 2 + ( r + 1 ) x + r + 1 x ) = χ ( a ( r ) x + r + 1 x ) , (58) W(r,b)=\chi\left(\frac{r+1}{r}x^2+(r+1)x+\frac{r+1}{x}\right)
=\chi\left(a(r)x+\frac{r+1}{x}\right), \tag*{(58)} W ( r , b ) = χ ( r r + 1 x 2 + ( r + 1 ) x + x r + 1 ) = χ ( a ( r ) x + x r + 1 ) , (58)
where
a ( r ) = r + 1 + r + 1 r . a(r)=r+1+\sqrt{\frac{r+1}{r}}. a ( r ) = r + 1 + r r + 1 .
and r + 1 r \sqrt{\frac{r+1}{r}} r r + 1 denotes the unique square root in F q \mathbb{F}_q F q . Combining (57) and (58) , we obtain
∑ b ∈ F q W ( r , b ) n ( r , b ) = ∑ x ≠ 0 χ ( a ( r ) x + r + 1 x ) = K m ( a ( r ) ( r + 1 ) ) = K m ( a ( r ) 2 ( r + 1 ) 2 ) . (59) \begin{aligned}
\sum_{b\in\mathbb{F}_q}W(r,b)n(r,b)
&=\sum_{x\ne0}\chi\left(a(r)x+\frac{r+1}{x}\right)\\
&=K_m\left(a(r)(r+1)\right)\\
&=K_m\left(a(r)^2(r+1)^2\right). \tag*{(59)}
\end{aligned} b ∈ F q ∑ W ( r , b ) n ( r , b ) = x = 0 ∑ χ ( a ( r ) x + x r + 1 ) = K m ( a ( r ) ( r + 1 ) ) = K m ( a ( r ) 2 ( r + 1 ) 2 ) . (59)
where the second equality holds since r + 1 ≠ 0 r+1\ne0 r + 1 = 0 . Note that
a ( r ) 2 ( r + 1 ) 2 = ( r + 1 ) 3 r ( r ( r + 1 ) + 1 ) = ( r 3 + 1 ) ( r + 1 ) 2 r = F ( r ) , \begin{aligned}
a(r)^2(r+1)^2
&=\frac{(r+1)^3}{r}\left(r(r+1)+1\right)\\
&=\frac{(r^3+1)(r+1)^2}{r}=F(r),
\end{aligned} a ( r ) 2 ( r + 1 ) 2 = r ( r + 1 ) 3 ( r ( r + 1 ) + 1 ) = r ( r 3 + 1 ) ( r + 1 ) 2 = F ( r ) ,
where F F F is the function appearing in the definition of K 4 K_4 K 4 . By (56) and (59) , we obtain
∑ b ≠ 1 W ( r , b ) n ( r , b ) = ∑ b ∈ F q W ( r , b ) n ( r , b ) − W ( r , 1 ) n ( r , 1 ) = K m ( F ( r ) ) − 2 χ ( 1 / r ) . (60) \begin{aligned}
\sum_{b\ne1}W(r,b)n(r,b)
&=\sum_{b\in\mathbb{F}_q}W(r,b)n(r,b)-W(r,1)n(r,1)\\
&=K_m(F(r))-2\chi(1/r). \tag*{(60)}
\end{aligned} b = 1 ∑ W ( r , b ) n ( r , b ) = b ∈ F q ∑ W ( r , b ) n ( r , b ) − W ( r , 1 ) n ( r , 1 ) = K m ( F ( r )) − 2 χ ( 1/ r ) . (60)
Part (ii): Calculating ∑ b ≠ 1 W ( r , b ) χ ( ( r 2 + b ) 3 r 2 ( b + 1 ) 2 ) \sum_{b\ne1} W(r,b)\chi\left(\frac{(r^2+b)^3}{r^2(b+1)^2}\right) ∑ b = 1 W ( r , b ) χ ( r 2 ( b + 1 ) 2 ( r 2 + b ) 3 ) for r ≠ 0 , 1 r\ne0,1 r = 0 , 1
For fixed r ≠ 0 , 1 r\ne0,1 r = 0 , 1 , the substitution w = b + 1 w=b+1 w = b + 1 is a bijection from F q ∖ { 1 } \mathbb{F}_q\setminus\{1\} F q ∖ { 1 } to F q ∗ \mathbb{F}_q^* F q ∗ . Using this substitution, the expression W ( r , b ) χ ( ( r 2 + b ) 3 r 2 ( b + 1 ) 2 ) W(r,b)\chi\left(\frac{(r^2+b)^3}{r^2(b+1)^2}\right) W ( r , b ) χ ( r 2 ( b + 1 ) 2 ( r 2 + b ) 3 ) becomes
χ ( r + 1 r ( w + 1 ) + ( r 2 + 1 + w ) 3 r 2 w 2 ) = χ ( ( r + 1 r + 1 r 2 ) w + ( r + 1 r + r 2 + 1 r 2 ) + ( r 2 + 1 ) 2 r 2 w + ( r 2 + 1 ) 3 r 2 w 2 ) = χ ( h r r 2 w + ( r − 1 + r − 2 ) + ( r + 1 ) 4 r 2 w + ( ( r + 1 ) 3 r w ) 2 ) = χ ( h r r 2 w + ( r + 1 ) 3 r 2 w ) , \begin{aligned}
\chi\left(\frac{r+1}{r}(w+1)+\frac{(r^2+1+w)^3}{r^2w^2}\right) \\
&=\chi\left(\left(\frac{r+1}{r}+\frac{1}{r^2}\right)w+\left(\frac{r+1}{r}+\frac{r^2+1}{r^2}\right)+\frac{(r^2+1)^2}{r^2w}+\frac{(r^2+1)^3}{r^2w^2}\right) \\
&=\chi\left(\frac{h_r}{r^2}w+(r^{-1}+r^{-2})+\frac{(r+1)^4}{r^2w}+\left(\frac{(r+1)^3}{rw}\right)^2\right) \\
&=\chi\left(\frac{h_r}{r^2}w+\frac{(r+1)^3}{r^2w}\right),
\end{aligned} χ ( r r + 1 ( w + 1 ) + r 2 w 2 ( r 2 + 1 + w ) 3 ) = χ ( ( r r + 1 + r 2 1 ) w + ( r r + 1 + r 2 r 2 + 1 ) + r 2 w ( r 2 + 1 ) 2 + r 2 w 2 ( r 2 + 1 ) 3 ) = χ ( r 2 h r w + ( r − 1 + r − 2 ) + r 2 w ( r + 1 ) 4 + ( r w ( r + 1 ) 3 ) 2 ) = χ ( r 2 h r w + r 2 w ( r + 1 ) 3 ) ,
where h r = r 2 + r + 1 h_r=r^2+r+1 h r = r 2 + r + 1 . Hence
∑ b ≠ 1 W ( r , b ) χ ( ( r 2 + b ) 3 r 2 ( b + 1 ) 2 ) = ∑ w ≠ 0 χ ( h r r 2 w + ( r + 1 ) 3 r 2 w ) = K m ( h r ( r + 1 ) 3 r 4 ) = K m ( F ( r − 1 ) ) , (61) \begin{aligned}
\sum_{b\ne1}W(r,b)\chi\left(\frac{(r^2+b)^3}{r^2(b+1)^2}\right)
=\sum_{w\ne0}\chi\left(\frac{h_r}{r^2}w+\frac{(r+1)^3}{r^2w}\right) \\
&=K_m\left(\frac{h_r(r+1)^3}{r^4}\right) \\
&=K_m\left(F(r^{-1})\right), \tag*{(61)}
\end{aligned} b = 1 ∑ W ( r , b ) χ ( r 2 ( b + 1 ) 2 ( r 2 + b ) 3 ) = w = 0 ∑ χ ( r 2 h r w + r 2 w ( r + 1 ) 3 ) = K m ( r 4 h r ( r + 1 ) 3 ) = K m ( F ( r − 1 ) ) , (61)
where the third equality holds since ( r + 1 ) 3 / r 2 ≠ 0 (r+1)^3/r^2\ne0 ( r + 1 ) 3 / r 2 = 0 .
Part (iii): Calculating ∑ b ≠ 1 W ( r , b ) \sum_{b\ne1}W(r,b) ∑ b = 1 W ( r , b ) for given r ≠ 0 , 1 r\ne0,1 r = 0 , 1 . Note that for r ≠ 0 , 1 r\ne0,1 r = 0 , 1
∑ b ∈ F q W ( r , b ) = ∑ b ∈ F q χ ( r + 1 r b ) = 0 , \sum_{b\in\mathbb{F}_q}W(r,b)=\sum_{b\in\mathbb{F}_q}\chi\left(\frac{r+1}{r}b\right)=0, b ∈ F q ∑ W ( r , b ) = b ∈ F q ∑ χ ( r r + 1 b ) = 0 ,
Hence
∑ b ≠ 1 W ( r , b ) = − W ( r , 1 ) = − χ ( 1 / r ) . (62) \sum_{b\ne1}W(r,b)=-W(r,1)=-\chi(1/r). \tag*{(62)} b = 1 ∑ W ( r , b ) = − W ( r , 1 ) = − χ ( 1/ r ) . (62)
For each r ≠ 0 , 1 r\ne0,1 r = 0 , 1 , combining equations (60) , (62) , and (61) yield
∑ b ≠ 1 N ( r , b ) W ( r , b ) = 2 ∑ b ≠ 1 W ( r , b ) n ( r , b ) − ∑ b ≠ 1 W ( r , b ) + ∑ b ≠ 1 W ( r , b ) χ ( ( r 2 + b ) 3 r 2 ( b + 1 ) 2 ) = 2 K m ( F ( r ) ) + K m ( F ( r − 1 ) ) − 3 χ ( r − 1 ) . (63) \begin{aligned}
\sum_{b\ne1}N(r,b)W(r,b) \\
&=2\sum_{b\ne1}W(r,b)n(r,b)-\sum_{b\ne1}W(r,b)+\sum_{b\ne1}W(r,b)\chi\left(\frac{(r^2+b)^3}{r^2(b+1)^2}\right) \\
&=2K_m\left(F(r)\right)+K_m\left(F(r^{-1})\right)-3\chi(r^{-1}). \tag*{(63)}
\end{aligned} b = 1 ∑ N ( r , b ) W ( r , b ) = 2 b = 1 ∑ W ( r , b ) n ( r , b ) − b = 1 ∑ W ( r , b ) + b = 1 ∑ W ( r , b ) χ ( r 2 ( b + 1 ) 2 ( r 2 + b ) 3 ) = 2 K m ( F ( r ) ) + K m ( F ( r − 1 ) ) − 3 χ ( r − 1 ) . (63)
Part (iv): This part is concerned with ∑ b ≠ 1 N ( r , b ) W ( r , b ) \sum_{b\ne1}N(r,b)W(r,b) ∑ b = 1 N ( r , b ) W ( r , b ) at r = 1 r=1 r = 1 . According to (55) , we evaluate the three sums in Parts (i)–(iii) at r = 1 r=1 r = 1 directly from their expressions. Note that W ( 1 , b ) = 1 W(1,b)=1 W ( 1 , b ) = 1 for every b ∈ F q b\in\mathbb{F}_q b ∈ F q . The results are given as follow:
(a) ∑ b ≠ 1 W ( 1 , b ) n ( 1 , b ) = ∑ b ∈ F q n ( 1 , b ) − n ( 1 , 1 ) = q − 2 \sum_{b\ne1}W(1,b)n(1,b)=\sum_{b\in\mathbb{F}_q}n(1,b)-n(1,1)=q-2 ∑ b = 1 W ( 1 , b ) n ( 1 , b ) = ∑ b ∈ F q n ( 1 , b ) − n ( 1 , 1 ) = q − 2 since n ( 1 , 1 ) = 1 n(1,1)=1 n ( 1 , 1 ) = 1 ; (b) ∑ b ≠ 1 W ( 1 , b ) χ ( ( 1 + b ) 3 ( b + 1 ) 2 ) = ∑ b ≠ 1 χ ( b + 1 ) = − 1 \sum_{b\ne1} W(1,b)\chi\left(\frac{(1+b)^3}{(b+1)^2}\right)=\sum_{b\ne1}\chi(b+1)=-1 ∑ b = 1 W ( 1 , b ) χ ( ( b + 1 ) 2 ( 1 + b ) 3 ) = ∑ b = 1 χ ( b + 1 ) = − 1 ;
(c) ∑ b ≠ 1 W ( 1 , b ) = q − 1 \sum_{b\ne1} W(1,b)=q-1 ∑ b = 1 W ( 1 , b ) = q − 1 .
Combining (53) , (55) , (63) , and the results in Part (iv), we obtain
K = ∑ r ≠ 0 ∑ b ∈ F q N ( r , b ) W ( r , b ) = ∑ r ≠ 0 ∑ b ≠ 1 N ( r , b ) W ( r , b ) − 5 = 2 ∑ r ≠ 0 , 1 K m ( F ( r ) ) + ∑ r ≠ 0 , 1 K m ( F ( r − 1 ) ) − 3 ∑ r ≠ 0 , 1 χ ( r − 1 ) + [ 2 ( q − 2 ) − ( q − 1 ) − 1 ] − 5 = 3 K 4 + q − 3 , \begin{aligned}
K&=\sum_{r\ne0}\sum_{b\in\mathbb{F}_q}N(r,b)W(r,b)\\
&=\sum_{r\ne0}\sum_{b\ne1}N(r,b)W(r,b)-5\\
&=2\sum_{r\ne0,1}K_m(F(r))+\sum_{r\ne0,1}K_m(F(r^{-1}))\\
&\quad-3\sum_{r\ne0,1}\chi(r^{-1})+\left[2(q-2)-(q-1)-1\right]-5\\
&=3K_4+q-3,
\end{aligned} K = r = 0 ∑ b ∈ F q ∑ N ( r , b ) W ( r , b ) = r = 0 ∑ b = 1 ∑ N ( r , b ) W ( r , b ) − 5 = 2 r = 0 , 1 ∑ K m ( F ( r )) + r = 0 , 1 ∑ K m ( F ( r − 1 )) − 3 r = 0 , 1 ∑ χ ( r − 1 ) + [ 2 ( q − 2 ) − ( q − 1 ) − 1 ] − 5 = 3 K 4 + q − 3 ,
where we use the fact ∑ r ≠ 0 , 1 K m ( F ( r − 1 ) ) = ∑ r ≠ 0 , 1 K m ( F ( r ) ) \sum_{r\ne0,1}K_m(F(r^{-1}))=\sum_{r\ne0,1}K_m(F(r)) ∑ r = 0 , 1 K m ( F ( r − 1 )) = ∑ r = 0 , 1 K m ( F ( r )) .
To evaluate K \mathbf{K} K , we need the following result.
Lemma 5.11. The character sum K 4 \mathbf{K}_4 K 4 in Proposition 5.10 satisfies
K 4 = K 1 − q + 1 = K m ( 1 ) 2 + ( 2 σ − 1 ) q + 1 , K_4=K_1-q+1=K_m(1)^2+(2\sigma-1)q+1, K 4 = K 1 − q + 1 = K m ( 1 ) 2 + ( 2 σ − 1 ) q + 1 , where K 1 \mathbf{K}_1 K 1 is defined in (41) and σ = ( − 1 ) m / 2 \sigma=(-1)^{m/2} σ = ( − 1 ) m /2 .
Proof. See Appendix B.
With the above preparations, we can get the exact value of K 2 \mathbf{K}_2 K 2
Proposition 5.12. For even m ≥ 2 m\ge2 m ≥ 2 , the character sum K 2 \mathbf{K}_2 K 2 defined in (42) can be expressed as
K 2 = 2 K m ( 1 ) 2 + 2 ( σ − 1 ) q , (64) K_2=2K_m(1)^2+2(\sigma-1)q, \tag*{(64)} K 2 = 2 K m ( 1 ) 2 + 2 ( σ − 1 ) q , (64) where σ = ( − 1 ) m / 2 \sigma=(-1)^{m/2} σ = ( − 1 ) m /2 .
Proof. The expression of K 2 \mathbf{K}_2 K 2 can be directly derived by combining Lemma 5.8 , Proposition 5.10 and Lemma 5.11 .
Proof of the main result Theorem 1.1
Proof of Theorem 1.1. Propositions 5.7 and 5.12 give
K 1 = K m ( 1 ) 2 + 2 σ q and K 2 = 2 K m ( 1 ) 2 + 2 ( σ − 1 ) q . K_1=K_m(1)^2+2\sigma q\quad\text{and}\quad K_2=2K_m(1)^2+2(\sigma-1)q. K 1 = K m ( 1 ) 2 + 2 σ q and K 2 = 2 K m ( 1 ) 2 + 2 ( σ − 1 ) q .
With the value K 3 = 2 K m ( ω ) 2 − q K_3=2K_m(\omega)^2-q K 3 = 2 K m ( ω ) 2 − q in (45) , it follows from Proposition 5.5 that
{ S 0 , 1 = K m ( 1 ) 2 + 4 σ q + 7 , S 2 , 1 = 2 K m ( 1 ) 2 + ( 4 σ − 2 ) q + 6 , S 3 , 1 = q + 2 − 3 K m ( 1 ) + 3 K m ( ω ) 2 , S 3 , 0 = 3 K m ( 1 ) 2 + ( 6 σ − 2 ) q + 5 , \left\{
\begin{aligned}
S_{0,1}&=K_m(1)^2+4\sigma q+7,\\
S_{2,1}&=2K_m(1)^2+(4\sigma-2)q+6,\\
S_{3,1}&=q+2-3K_m(1)+3K_m(\omega)^2,\\
S_{3,0}&=3K_m(1)^2+(6\sigma-2)q+5,
\end{aligned}
\right. ⎩ ⎨ ⎧ S 0 , 1 S 2 , 1 S 3 , 1 S 3 , 0 = K m ( 1 ) 2 + 4 σ q + 7 , = 2 K m ( 1 ) 2 + ( 4 σ − 2 ) q + 6 , = q + 2 − 3 K m ( 1 ) + 3 K m ( ω ) 2 , = 3 K m ( 1 ) 2 + ( 6 σ − 2 ) q + 5 ,
where σ = ( − 1 ) m / 2 \sigma= (-1)^{m/2} σ = ( − 1 ) m /2 and ω 2 + ω + 1 = 0 \omega^2+\omega+1=0 ω 2 + ω + 1 = 0 . This together with Proposition 5.4 , Proposition 5.2 , Theorem 4.7 and Lemma 5.1 gives
48 N 4 = q 2 + ( 4 σ − 11 ) q + 4 K m ( 1 ) 2 + 3 K m ( 1 ) − 3 K m ( ω ) 2 + 2 , 720 N 6 = q 2 + ( 9 − 52 σ ) q − 10 K m ( 1 ) 2 − 45 K m ( 1 ) + 9 K m ( ω ) 2 − 44. \begin{aligned}
48N_4 &= q^2+(4\sigma-11)q+4K_m(1)^2+3K_m(1)-3K_m(\omega)^2+2,\\
720N_6 &= q^2+(9-52\sigma)q-10K_m(1)^2-45K_m(1)+9K_m(\omega)^2-44.
\end{aligned} 48 N 4 720 N 6 = q 2 + ( 4 σ − 11 ) q + 4 K m ( 1 ) 2 + 3 K m ( 1 ) − 3 K m ( ω ) 2 + 2 , = q 2 + ( 9 − 52 σ ) q − 10 K m ( 1 ) 2 − 45 K m ( 1 ) + 9 K m ( ω ) 2 − 44.
These are the expressions for N 4 N_4 N 4 and N 6 N_6 N 6 . From Proposition 4.5 , we can obtain the value of N 5 N_5 N 5 . Substituting the values of N 4 N_4 N 4 , N 5 N_5 N 5 and N 6 N_6 N 6 into Proposition 4.2 gives N 0 N_0 N 0 , N 1 N_1 N 1 , N 2 N_2 N 2 and N 3 N_3 N 3 . This completes the determination of the correlation distribution in Theorem 1.1 . Note that N 1 N_1 N 1 does not include the case of a = 0 a=0 a = 0 . Since W d ( 0 ) = 0 W_d(0)=0 W d ( 0 ) = 0 , the frequency of the value 0 in the value distribution of W d ( a ) W_d(a) W d ( a ) as a a a ranges over F q 2 \mathbb{F}_{q^2} F q 2 is N 1 + 1 N_1+1 N 1 + 1 .
The following corollary shows that all seven values occur for even m ≥ 6 m\geq6 m ≥ 6 .
Corollary 5.13. All seven values in Theorem 1.1 occur for even $m\geq 6.
Proof. By the Kloosterman bound in Lemma 2.4 , we have ∣ K m ( 1 ) ∣ , ∣ K m ( ω ) ∣ ≤ 2 q |K_m(1)|,|K_m(\omega)|\leq2\sqrt{q} ∣ K m ( 1 ) ∣ , ∣ K m ( ω ) ∣ ≤ 2 q . Then, based on the formulas in Theorem 1.1 , we have
144 N 0 ≥ 53 q 2 − 219 q − 18 q − 10 , 120 N 1 ≥ 44 q 2 − 53 q − 30 q − 101 , 16 N 2 ≥ 3 q 2 − 17 q − 2 q , 36 N 3 ≥ 2 q 2 − 3 q − 18 q − 7 , 48 N 4 ≥ q 2 − 27 q − 6 q + 2 , 720 N 6 ≥ q 2 − 83 q − 90 q − 44. \begin{aligned}
144N_0 &\geq53q^2-219q-18\sqrt{q}-10, &\quad120N_1 &\geq44q^2-53q-30\sqrt{q}-101,\\
16N_2 &\geq3q^2-17q-2\sqrt{q}, &\quad36N_3 &\geq2q^2-3q-18\sqrt{q}-7,\\
48N_4 &\geq q^2-27q-6\sqrt{q}+2, &\quad720N_6 &\geq q^2-83q-90\sqrt{q}-44.
\end{aligned} 144 N 0 16 N 2 48 N 4 ≥ 53 q 2 − 219 q − 18 q − 10 , ≥ 3 q 2 − 17 q − 2 q , ≥ q 2 − 27 q − 6 q + 2 , 120 N 1 36 N 3 720 N 6 ≥ 44 q 2 − 53 q − 30 q − 101 , ≥ 2 q 2 − 3 q − 18 q − 7 , ≥ q 2 − 83 q − 90 q − 44.
For q ≥ 256 q\geq256 q ≥ 256 , we have q ≤ q / 16 \sqrt{q}\leq q/16 q ≤ q /16 , so all six lower bounds are positive. Moreover,
8 N 5 = q + 1 + K m ( 1 ) ≥ ( q − 1 ) 2 > 0. 8N_5=q+1+K_m(1)\geq(\sqrt{q}-1)^2>0. 8 N 5 = q + 1 + K m ( 1 ) ≥ ( q − 1 ) 2 > 0.
For m = 6 m=6 m = 6 , Lemma 2.3 gives K 6 ( 1 ) = − 9 K_6(1)=-9 K 6 ( 1 ) = − 9 and K 6 ( ω ) = 11 K_6(\omega)=11 K 6 ( ω ) = 11 . Since σ = − 1 \sigma=-1 σ = − 1 , we obtain
( N 0 , N 1 , N 2 , N 3 , N 4 , N 5 , N 6 ) = ( 1428 , 1606 , 744 , 234 , 64 , 7 , 12 ) . (N_0,N_1,N_2,N_3,N_4,N_5,N_6)=(1428,1606,744,234,64,7,12). ( N 0 , N 1 , N 2 , N 3 , N 4 , N 5 , N 6 ) = ( 1428 , 1606 , 744 , 234 , 64 , 7 , 12 ) .
Thus, all seven frequencies are strictly positive for even $m\geq 6, ensuring that all seven correlation values occur.
:::
Example 1. For some small m m m from 2 to 10, the frequencies N i N_i N i computed with Magma agree with those given in Theorem 1.1 .
For m = 2 m=2 m = 2 , we have q = 4 q=4 q = 4 and d = 5 q − 1 ≡ 4 ( m o d 15 ) d=5q-1\equiv4\pmod{15} d = 5 q − 1 ≡ 4 ( mod 15 ) . Since 4 and 1 lie in the same 2-cyclotomic coset modulo 15, the d d d -decimated sequence is the original m-sequence. Thus, the two correlation values are the autocorrelation values − 1 -1 − 1 and 15. For m = 4 m=4 m = 4 , the exponent 5 q − 1 = 79 5q-1=79 5 q − 1 = 79 is equivalent to the known Niho exponent 3 q − 2 = 46 3q-2=46 3 q − 2 = 46 , since 79 ⋅ 46 ≡ 2 6 ( m o d 255 ) 79\cdot46\equiv2^6\pmod{255} 79 ⋅ 46 ≡ 2 6 ( mod 255 ) . By Theorem 2 of [5 ] , the corresponding cross-correlation function is five-valued.
Conclusion
Our main contributions in this paper are twofold. First, we study when low-degree self-conjugate-reciprocal polynomials over F q 2 \mathbb{F}_{q^2} F q 2 split in the unit circle U U U and apply the result to characterize the conditions under which a hexic polynomial P a ( x ) = x 6 + a x 5 + a ˉ x + 1 P_a(x)=x^6+ax^5+\bar{a}x+1 P a ( x ) = x 6 + a x 5 + a ˉ x + 1 , associated with the Niho exponent d = s ( q − 1 ) + 1 d=s(q-1)+1 d = s ( q − 1 ) + 1 with 4 s ≡ 5 ( m o d q + 1 ) 4s\equiv5\pmod{q+1} 4 s ≡ 5 ( mod q + 1 ) , has exactly 4, 5, 6 roots. Second, we determine the cross-correlation distribution of a binary m m m -sequence and its d d d -decimated sequence for this Niho exponent. In the calculation of the frequencies N 4 N_4 N 4 and N 6 N_6 N 6 , we investigate the involved character sums and transform them into expressions in terms of Kloosterman sums on two basic elements 1 and ω \omega ω in F 4 \mathbb{F}_4 F 4 , which can be explicitly calculated by the well-known recurrence established by Carlitz in 1969. Niho in [1 p. 63] gave a table of entries in the form of d ( k ) d(k) d ( k ) , for Niho decimations d d d having k k k -valued cross-correlation functions for k ≤ 8 k\leq8 k ≤ 8 . The Niho decimation d d d considered in this paper covers two entries, 19(5) and 319(7) in the table, indicating that they are actually instances of an infinite family. In future, it would be interesting to consider the remaining entries in the table and to determine the value distribution for the decimation d = 5 q − 3 d=5q-3 d = 5 q − 3 discussed in Niho’s thesis [1 Theorem 3-11] .
Appendices
Appendix A.
Proof of Lemma 4.4 . When a = 0 a=0 a = 0 , P 0 ( x ) = x 6 + 1 = ( x 3 + 1 ) 2 P_{0}(x)=x^{6}+1=(x^{3}+1)^{2} P 0 ( x ) = x 6 + 1 = ( x 3 + 1 ) 2 , and this case is trivial. Now we assume a ≠ 0 a\ne0 a = 0 and that ξ \xi ξ is a repeated root of P a ( x ) P_{a}(x) P a ( x ) . It is clear that P a ′ ( x ) = a x 4 + a ˉ P_{a}'(x)=ax^{4}+\bar{a} P a ′ ( x ) = a x 4 + a ˉ has a unique root x 0 = ( a / a ˉ ) 1 / 4 ∈ U x_{0}=(a/\bar{a})^{1/4}\in U x 0 = ( a / a ˉ ) 1/4 ∈ U with multiplicity four. Substituting it into P a ( x 0 ) = 0 P_{a}(x_{0})=0 P a ( x 0 ) = 0 gives x 0 6 + a x 0 5 + a ˉ x 0 + 1 = x 0 6 + ( a x 0 4 + a ˉ ) x 0 + 1 = x 0 6 + 1 = 0 x_{0}^{6}+ax_{0}^{5}+\bar{a}x_{0}+1=x_{0}^{6}+(ax_{0}^{4}+\bar{a})x_{0}+1=x_{0}^{6}+1=0 x 0 6 + a x 0 5 + a ˉ x 0 + 1 = x 0 6 + ( a x 0 4 + a ˉ ) x 0 + 1 = x 0 6 + 1 = 0 . As m m m is even, q + 1 q+1 q + 1 is odd and is not divisible by three. Thus gcd ( 6 , q + 1 ) = 1 \gcd(6,q+1)=1 g cd( 6 , q + 1 ) = 1 , yielding x 0 = 1 x_{0}=1 x 0 = 1 . The derivative equation P a ( x 0 ) = 0 P_{a}(x_{0})=0 P a ( x 0 ) = 0 then gives a = a ˉ a=\bar{a} a = a ˉ . Hence P a ( x ) P_{a}(x) P a ( x ) has repeated roots if and only if a ∈ F q a\in\mathbb{F}_{q} a ∈ F q .
When a ∈ F q a\in\mathbb{F}_{q} a ∈ F q , expanding the right-hand side of (22) verifies the identity. Let a ∈ F q ∗ a\in\mathbb{F}_{q}^{*} a ∈ F q ∗ , and write the quartic factor as f a ( x ) = x 4 + a x 3 + x 2 + a x + 1 f_{a}(x)=x^{4}+ax^{3}+x^{2}+ax+1 f a ( x ) = x 4 + a x 3 + x 2 + a x + 1 . Since f a ′ ( x ) = a ( x + 1 ) 2 f_{a}'(x)=a(x+1)^{2} f a ′ ( x ) = a ( x + 1 ) 2 and f a ( 1 ) = 1 f_{a}(1)=1 f a ( 1 ) = 1 , the polynomial f a f_{a} f a has no repeated roots. Note that f a ( x ) f_{a}(x) f a ( x ) is a normalized SCR polynomial with coefficients in F q \mathbb{F}_{q} F q . By Proposition 3.1 , its roots in U U U can be determined from the roots of the associated polynomial g ( y ) g(y) g ( y ) in { 0 } ∪ H 1 \{0\}\cup H_{1} { 0 } ∪ H 1 . Using (9) , we obtain
g ( y ) = ( 1 + a y + y 2 ) 2 . g(y)=(1+ay+y^{2})^{2}. g ( y ) = ( 1 + a y + y 2 ) 2 .
The quadratic equation y 2 + a y + 1 = 0 y^{2}+ay+1=0 y 2 + a y + 1 = 0 has roots in F q \mathbb{F}_{q} F q if and only if Tr m ( a − 1 ) = 0 \operatorname{Tr}_{m}(a^{-1})=0 Tr m ( a − 1 ) = 0 , or equivalently,
a = t + t − 1 for some t ∈ F q ∖ { 0 , 1 } . a=t+t^{-1}\qquad\text{for some }t\in\mathbb{F}_{q}\setminus\{0,1\}. a = t + t − 1 for some t ∈ F q ∖ { 0 , 1 } .
In this case, its two roots are t t t and t − 1 t^{-1} t − 1 , which are distinct since a ≠ 0 a\ne0 a = 0 . Moreover,
g ( y ) = ( y + t ) 2 ( y + t − 1 ) 2 , g(y)=(y+t)^{2}(y+t^{-1})^{2}, g ( y ) = ( y + t ) 2 ( y + t − 1 ) 2 ,
so each has multiplicity exactly two as a root of g g g . Note that when these roots belong to H 1 H_{1} H 1 , they correspond to two simple roots of f a f_{a} f a in U ∖ { 1 } U\setminus\{1\} U ∖ { 1 } . According to the definition of H 1 H_{1} H 1 , we have
N ( a ) = { 1 , Tr m ( t ) = Tr m ( t − 1 ) = 0 , 3 , Tr m ( t ) ≠ Tr m ( t − 1 ) , 5 , Tr m ( t ) = Tr m ( t − 1 ) = 1. N(a)=\begin{cases}
1, & \operatorname{Tr}_{m}(t)=\operatorname{Tr}_{m}(t^{-1})=0,\\
3, & \operatorname{Tr}_{m}(t)\ne\operatorname{Tr}_{m}(t^{-1}),\\
5, & \operatorname{Tr}_{m}(t)=\operatorname{Tr}_{m}(t^{-1})=1.
\end{cases} N ( a ) = ⎩ ⎨ ⎧ 1 , 3 , 5 , Tr m ( t ) = Tr m ( t − 1 ) = 0 , Tr m ( t ) = Tr m ( t − 1 ) , Tr m ( t ) = Tr m ( t − 1 ) = 1.
If Tr m ( a − 1 ) = 1 \operatorname{Tr}_{m}(a^{-1})=1 Tr m ( a − 1 ) = 1 , then neither y 2 + a y + 1 y^2+ay+1 y 2 + a y + 1 nor g ( y ) g(y) g ( y ) has a root in F q \mathbb{F}_q F q . Hence, by Proposition 3.1 , P a P_a P a has only the root 1 1 1 in U U U . □
Proof of Proposition 4.5 . When a ∉ F q a\notin\mathbb{F}_q a ∈ / F q , the polynomial P a ( x ) P_a(x) P a ( x ) has no repeated roots. If it has five distinct roots u 1 , … , u 5 u_1,\ldots,u_5 u 1 , … , u 5 in U U U , then, since it is monic of degree six with constant term 1, its sixth root is
u 6 = ( u 1 ⋯ u 5 ) − 1 ∈ U . u_6=(u_1\cdots u_5)^{-1}\in U. u 6 = ( u 1 ⋯ u 5 ) − 1 ∈ U .
Moreover, u 6 u_6 u 6 is distinct from u 1 , … , u 5 u_1,\ldots,u_5 u 1 , … , u 5 , since P a P_a P a has no repeated roots. Thus, P a P_a P a cannot have exactly five distinct roots in U U U unless a ∈ F q a\in\mathbb{F}_q a ∈ F q .
By Lemma 4.4 , P a P_a P a has exactly five distinct roots in U U U if and only if
a = t + t − 1 for some t ∈ F q ∖ { 0 , 1 } with t , t − 1 ∈ H 1 . a=t+t^{-1}\quad\text{for some }t\in\mathbb{F}_q\setminus\{0,1\}\text{ with }t,t^{-1}\in H_1. a = t + t − 1 for some t ∈ F q ∖ { 0 , 1 } with t , t − 1 ∈ H 1 .
In this case, 1 is a root of multiplicity exactly two, and the remaining four roots are distinct and belong to U ∖ { 1 } U\setminus\{1\} U ∖ { 1 } .
When m m m is even, consider instead the set T = { t ∈ F q ∗ ∖ { 1 } : Tr m ( 1 / t ) = Tr m ( t ) = 1 } T=\{t\in\mathbb{F}_q^*\setminus\{1\}:\operatorname{Tr}_m(1/t)=\operatorname{Tr}_m(t)=1\} T = { t ∈ F q ∗ ∖ { 1 } : Tr m ( 1/ t ) = Tr m ( t ) = 1 } . Then
∣ T ∣ = 1 4 ∑ x ∈ F q ∗ ∖ { 1 } ( 1 − ( − 1 ) Tr m ( 1 / t ) ) ( 1 − ( − 1 ) Tr ( t ) ) = 1 4 ( ( q − 2 ) + 4 + k m − 1 ) = q + 1 + k m 4 . |T|=\frac{1}{4}\sum_{x\in\mathbb{F}_q^*\setminus\{1\}}\left(1-(-1)^{\operatorname{Tr}_m(1/t)}\right)\left(1-(-1)^{\operatorname{Tr}(t)}\right)=\frac{1}{4}\left((q-2)+4+k_m-1\right)=\frac{q+1+k_m}{4}. ∣ T ∣ = 4 1 x ∈ F q ∗ ∖ { 1 } ∑ ( 1 − ( − 1 ) Tr m ( 1/ t ) ) ( 1 − ( − 1 ) Tr ( t ) ) = 4 1 ( ( q − 2 ) + 4 + k m − 1 ) = 4 q + 1 + k m .
Since t , t − 1 ∈ T t,t^{-1}\in T t , t − 1 ∈ T gives a same a = t + t − 1 a=t+t^{-1} a = t + t − 1 , we have N 5 = q + 1 + k m 8 N_5=\frac{q+1+k_m}{8} N 5 = 8 q + 1 + k m , where k m = K m ( 1 ) k_m=K_m(1) k m = K m ( 1 ) . □
Proof of Lemma 4.6 . For a ∈ U a\in U a ∈ U with a ˉ = a − 1 \bar{a}=a^{-1} a ˉ = a − 1 , it is straightforward that P a ( x ) = ( x + a ) ( x 5 + a − 1 ) P_a(x)=(x+a)(x^5+a^{-1}) P a ( x ) = ( x + a ) ( x 5 + a − 1 ) . Since m m m is even, by Lemma 4.4 , the two factors have a common root if and only if a = 1 a=1 a = 1 .
If m ≡ 0 ( m o d 4 ) m\equiv0\pmod{4} m ≡ 0 ( mod 4 ) , then gcd ( 5 , q + 1 ) = 1 \gcd(5,q+1)=1 g cd( 5 , q + 1 ) = 1 . Hence, the map u ↦ u 5 u\mapsto u^5 u ↦ u 5 permutes U U U , so X 5 = a − 1 X^5=a^{-1} X 5 = a − 1 has exactly one solution in U U U . This solution equals a a a precisely when a = 1 a=1 a = 1 , which proves (i).
If m ≡ 2 ( m o d 4 ) m\equiv2\pmod{4} m ≡ 2 ( mod 4 ) , then 5 ∣ ( q + 1 ) 5\mid(q+1) 5 ∣ ( q + 1 ) . The map u ↦ u 5 u\mapsto u^5 u ↦ u 5 on U U U has kernel of size five and image U 5 U_5 U 5 . Consequently, X 5 = a − 1 X^5=a^{-1} X 5 = a − 1 has five distinct solutions in U U U if a ∈ U 5 a\in U_5 a ∈ U 5 , and none otherwise. Adding the root a a a and accounting for the common root when a = 1 a=1 a = 1 gives (ii). □
Proof of Corollary 4.8 . As discussed in the proof of Theorem 4.7 , for a set { w 1 , w 2 , w 3 } \{w_1,w_2,w_3\} { w 1 , w 2 , w 3 } in W W W , one can derive a polynomial h ( y ) = ( x + w 1 ) ( x + w 2 ) ( x + w 3 ) h(y)=(x+w_1)(x+w_2)(x+w_3) h ( y ) = ( x + w 1 ) ( x + w 2 ) ( x + w 3 ) , which by the description of Ω t \Omega_t Ω t expands to h ( y ) = x 3 + β x 2 + α α ˉ x + ( α + α ˉ ) 2 h(y)=x^3+\beta x^2+\alpha\bar{\alpha}x+(\alpha+\bar{\alpha})^2 h ( y ) = x 3 + β x 2 + α α ˉ x + ( α + α ˉ ) 2 . For an element a a a is determined by α , β \alpha,\beta α , β as in the set A 0 ∪ A 1 A_0\cup A_1 A 0 ∪ A 1 , it is routine to verify the following equality (as discussed in the proof of Proposition 4.3 )
μ − 6 P a ( μ x ) = ( x 4 + α x 3 + β x 2 + α ˉ x + 1 ) ( x 2 + γ x + μ − 6 ) \mu^{-6}P_a(\mu x)=(x^4+\alpha x^3+\beta x^2+\bar{\alpha}x+1)(x^2+\gamma x+\mu^{-6}) μ − 6 P a ( μx ) = ( x 4 + α x 3 + β x 2 + α ˉ x + 1 ) ( x 2 + γ x + μ − 6 )
where
μ = ( α + β α ˉ + β ) 1 6 and γ = β + 1 α + β , \mu=\left(\frac{\alpha+\sqrt{\beta}}{\bar{\alpha}+\sqrt{\beta}}\right)^{\frac{1}{6}}
\quad\text{and}\quad
\gamma=\frac{\beta+1}{\alpha+\sqrt{\beta}}, μ = ( α ˉ + β α + β ) 6 1 and γ = α + β β + 1 ,
where α + β ≠ 0 \alpha+\sqrt{\beta}\ne0 α + β = 0 is ensured by the fact { w 1 , w 2 , w 3 } ⊂ H 1 \{w_1,w_2,w_3\}\subset H_1 { w 1 , w 2 , w 3 } ⊂ H 1 and w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 w_1w_2w_3=(w_1+w_2+w_3)(w_1+w_2+w_3+1)^2 w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 . By Proposition 3.4 we know Q ( x ) = x 4 + α x 3 + β x 2 + α ˉ x + 1 Q(x)=x^4+\alpha x^3+\beta x^2+\bar{\alpha}x+1 Q ( x ) = x 4 + α x 3 + β x 2 + α ˉ x + 1 has four distinct roots in U U U . We need to further discuss the quadratic polynomial x 2 + γ x + μ − 6 x^2+\gamma x+\mu^{-6} x 2 + γ x + μ − 6 .
As shown in the proof of Theorem 4.7 , the quadratic factor x 2 + γ x + μ − 6 x^2+\gamma x+\mu^{-6} x 2 + γ x + μ − 6 of P a ( μ x ) P_a(\mu x) P a ( μx ) has no root in U U U when Tr m ( κ ) = 0 \operatorname{Tr}_m(\kappa)=0 Tr m ( κ ) = 0 and has two distinct roots in U U U when Tr m ( κ ) = 1 \operatorname{Tr}_m(\kappa)=1 Tr m ( κ ) = 1 . If a ∈ A 0 a\in A_0 a ∈ A 0 , the quadratic factor x 2 + γ x + μ − 6 x^2+\gamma x+\mu^{-6} x 2 + γ x + μ − 6 has no root in U U U , so thus P a ( x ) P_a(x) P a ( x ) has exactly four distinct roots in U U U . Since the case a ∈ F q ∗ a\in\mathbb{F}_q^* a ∈ F q ∗ has already been treated in Lemma 4.4 , we are only interested in the case a ∈ A 1 ∖ F q a \in A_1 \setminus\mathbb{F}_q a ∈ A 1 ∖ F q , for which P a ( x ) P_a(x) P a ( x ) has exactly six distinct roots in U U U .
Combining the above arguments with Lemmas 4.4 and 4.6 , we can derive the remaining claims. □
Proof of Lemma 5.1. Let
r = v 1 v 2 v 3 , κ = ( v 1 v 2 + v 1 v 3 + v 2 v 3 + r 2 ) 2 (65) r=v_1v_2v_3,\qquad\kappa=\left(v_1v_2+v_1v_3+v_2v_3+r^2\right)^2 \tag*{(65)} r = v 1 v 2 v 3 , κ = ( v 1 v 2 + v 1 v 3 + v 2 v 3 + r 2 ) 2 (65)
and define the following set
V t = { ( v 1 , v 2 , v 3 ) ∈ ( F q ∗ ) 3 | v 1 + v 2 + v 3 = v 1 v 2 v 3 , v i ≠ v j , Tr m ( 1 + r v i ) = 1 , 1 ≤ i < j ≤ 3 , Tr m ( κ ) = t } (66) \mathcal{V}_t=\left\{(v_1,v_2,v_3)\in(\mathbb{F}_q^*)^3\ \middle|\ \begin{aligned}
&v_1+v_2+v_3=v_1v_2v_3,\ v_i\ne v_j,\\
&\operatorname{Tr}_m\left(\frac{1+r}{v_i}\right)=1,\qquad1\le i<j\le3,\\
&\operatorname{Tr}_m(\kappa)=t
\end{aligned}\right\} \tag*{(66)} V t = ⎩ ⎨ ⎧ ( v 1 , v 2 , v 3 ) ∈ ( F q ∗ ) 3 v 1 + v 2 + v 3 = v 1 v 2 v 3 , v i = v j , Tr m ( v i 1 + r ) = 1 , 1 ≤ i < j ≤ 3 , Tr m ( κ ) = t ⎭ ⎬ ⎫ (66)
where t = 0 , 1 t=0,1 t = 0 , 1 . Here r ≠ 0 r\ne0 r = 0 and r ≠ 1 r\ne1 r = 1 since v i ∈ F q ∗ v_i\in\mathbb{F}_q^* v i ∈ F q ∗ and Tr m ( 1 + r v i ) = 1 \operatorname{Tr}_m\left(\frac{1+r}{v_i}\right)=1 Tr m ( v i 1 + r ) = 1 . We highlight two basic properties of these two sets. First, V t \mathcal{V}_t V t is invariant under permutations of the coordinates. Second, for any ( v 1 , v 2 , v 3 ) ∈ V t (v_1,v_2,v_3)\in\mathcal{V}_t ( v 1 , v 2 , v 3 ) ∈ V t , the relation v 1 + v 2 + v 3 = v 1 v 2 v 3 v_1+v_2+v_3=v_1v_2v_3 v 1 + v 2 + v 3 = v 1 v 2 v 3 and the pairwise distinctness of the coordinates imply that
v i ≠ 1 ( 1 ≤ i ≤ 3 ) , and v i v j ≠ 1 ( 1 ≤ i < j ≤ 3 ) . v_i\ne1\quad(1\le i\le3),\qquad\text{and}\qquad v_iv_j\ne1\quad(1\le i<j\le3). v i = 1 ( 1 ≤ i ≤ 3 ) , and v i v j = 1 ( 1 ≤ i < j ≤ 3 ) .
We first establish a relation between W t \mathcal{W}_t W t and the triple set V t \mathcal{V}_t V t for each t = 0 , 1 t=0,1 t = 0 , 1 , and then prove that ∣ W t ∣ = ∣ V t ∣ / 6 |\mathcal{W}_t|=|\mathcal{V}_t|/6 ∣ W t ∣ = ∣ V t ∣/6 . The defining relations of V t \mathcal{V}_t V t , t = 0 , 1 t=0,1 t = 0 , 1 , are simpler and facilitate subsequent computations.
Suppose { w 1 , w 2 , w 3 } ∈ W t \{w_1,w_2,w_3\}\in\mathcal{W}_t { w 1 , w 2 , w 3 } ∈ W t . Then, they are three distinct elements in H 1 = { x ∈ F q ∗ : Tr m ( 1 / x ) = 1 } H_1=\{x\in\mathbb{F}_q^*:\operatorname{Tr}_m(1/x)=1\} H 1 = { x ∈ F q ∗ : Tr m ( 1/ x ) = 1 } and satisfy
w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 (67) w_1w_2w_3=(w_1+w_2+w_3)(w_1+w_2+w_3+1)^2 \tag*{(67)} w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 (67)
We denote
v i = w i w 1 + w 2 + w 3 + 1 = w i β + 1 for i = 1 , 2 , 3 , (68) v_i=\sqrt{\frac{w_i}{w_1+w_2+w_3+1}}=\sqrt{\frac{w_i}{\beta+1}}\qquad\text{for }i=1,2,3, \tag*{(68)} v i = w 1 + w 2 + w 3 + 1 w i = β + 1 w i for i = 1 , 2 , 3 , (68)
where β = w 1 + w 2 + w 3 \beta=w_1+w_2+w_3 β = w 1 + w 2 + w 3 . In this way, the condition (67) in W t \mathcal{W}_t W t reduced to
( v 1 v 2 v 3 ) 2 = ( v 1 + v 2 + v 3 ) 2 ( v 1 2 + v 2 2 + v 3 2 + 1 β + 1 ) 2 = ( v 1 + v 2 + v 3 ) 2 (v_1v_2v_3)^2=(v_1+v_2+v_3)^2\left(v_1^2+v_2^2+v_3^2+\frac{1}{\beta+1}\right)^2=(v_1+v_2+v_3)^2 ( v 1 v 2 v 3 ) 2 = ( v 1 + v 2 + v 3 ) 2 ( v 1 2 + v 2 2 + v 3 2 + β + 1 1 ) 2 = ( v 1 + v 2 + v 3 ) 2
since v 1 2 + v 2 2 + v 3 2 = β β + 1 v_1^2+v_2^2+v_3^2=\frac{\beta}{\beta+1} v 1 2 + v 2 2 + v 3 2 = β + 1 β . Note that the map β ↦ β β + 1 \beta\mapsto\frac{\beta}{\beta+1} β ↦ β + 1 β is an involution on F q ∖ { 0 , 1 } \mathbb{F}_q\setminus\{0,1\} F q ∖ { 0 , 1 } . The identity v 1 2 + v 2 2 + v 3 2 = β β + 1 v_1^2+v_2^2+v_3^2=\frac{\beta}{\beta+1} v 1 2 + v 2 2 + v 3 2 = β + 1 β also implies that
β + 1 = 1 v 1 2 + v 2 2 + v 3 2 + 1 = 1 ( v 1 v 2 v 3 ) 2 + 1 . (69) \beta+1=\frac{1}{v_1^2+v_2^2+v_3^2+1}=\frac{1}{(v_1v_2v_3)^2+1}. \tag*{(69)} β + 1 = v 1 2 + v 2 2 + v 3 2 + 1 1 = ( v 1 v 2 v 3 ) 2 + 1 1 . (69)
In addition, the element κ \kappa κ in (27) can be expressed by v 1 , v 2 , v 3 v_1,v_2,v_3 v 1 , v 2 , v 3 as
κ = ( v 1 v 2 + v 1 v 3 + v 2 v 3 + ( v 1 v 2 v 3 ) 2 ) 2 . \kappa=\left(v_1v_2+v_1v_3+v_2v_3+(v_1v_2v_3)^2\right)^2. κ = ( v 1 v 2 + v 1 v 3 + v 2 v 3 + ( v 1 v 2 v 3 ) 2 ) 2 .
Following the relation in (68) between v i v_i v i and w i w_i w i for i = 1 , 2 , 3 i=1,2,3 i = 1 , 2 , 3 , we can see that v i v_i v i has one-to-one correspondence with w i w_i w i for i = 1 , 2 , 3 i=1,2,3 i = 1 , 2 , 3 . The condition w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 w_1w_2w_3=(w_1+w_2+w_3)(w_1+w_2+w_3+1)^2 w 1 w 2 w 3 = ( w 1 + w 2 + w 3 ) ( w 1 + w 2 + w 3 + 1 ) 2 corresponds to a simpler condition v 1 v 2 v 3 = v 1 + v 2 + v 3 v_1v_2v_3=v_1+v_2+v_3 v 1 v 2 v 3 = v 1 + v 2 + v 3 . Meanwhile, the elements in a set { w 1 , w 2 , w 3 } \{w_1,w_2,w_3\} { w 1 , w 2 , w 3 } are different, indicating that v i ≠ v j v_i \ne v_j v i = v j for 1 ≤ i < j ≤ 3 1 \le i < j \le3 1 ≤ i < j ≤ 3 ; and w i ∈ H 1 w_i \in H_1 w i ∈ H 1 implies that 1 = Tr m ( 1 v i ) = Tr m ( β 2 + 1 ( β 2 + 1 ) v i 2 ) = Tr m ( 1 + v 1 v 2 v 3 v i ) 1 = \operatorname{Tr}_m\left(\frac{1}{v_i}\right) = \operatorname{Tr}_m\left(\frac{\beta^2+1}{(\beta^2+1)v_i^2}\right) = \operatorname{Tr}_m\left(\frac{1+v_1v_2v_3}{v_i}\right) 1 = Tr m ( v i 1 ) = Tr m ( ( β 2 + 1 ) v i 2 β 2 + 1 ) = Tr m ( v i 1 + v 1 v 2 v 3 ) due to (69) . Hence the conditions on v i v_i v i ’s exactly correspond to those on w i w_i w i ’s. Lastly, each set { w 1 , w 2 , w 3 } \{w_1,w_2,w_3\} { w 1 , w 2 , w 3 } in W t W_t W t provides 6 different ordered triples, implying that ∣ W t ∣ = ∣ V t ∣ / 6 |W_t| = |V_t|/6 ∣ W t ∣ = ∣ V t ∣/6 .
Next we show ∣ V t ∣ = ∣ D t ∣ |V_t| = |D_t| ∣ V t ∣ = ∣ D t ∣ for t ∈ { 0 , 1 } t \in\{0,1\} t ∈ { 0 , 1 } . Note that each pair ( x , y ) (x,y) ( x , y ) in D D D uniquely determines z z z via z = x + y 1 + x y z = \frac{x+y}{1+xy} z = 1 + x y x + y , and ( x + 1 ) ( y + 1 ) ( x + y ) ( x y + 1 ) ≠ 0 (x+1)(y+1)(x+y)(xy+1) \ne0 ( x + 1 ) ( y + 1 ) ( x + y ) ( x y + 1 ) = 0 ensures that x , y , z x,y,z x , y , z are nonzero and pairwise distinct. Thus, for every pair in D D D , the resulting coordinates x x x , y y y , z = ( x + y ) / ( 1 + x y ) z = (x+y)/(1+xy) z = ( x + y ) / ( 1 + x y ) are nonzero and pairwise distinct, and satisfy x + y + z = x y z x+y+z = xyz x + y + z = x y z . Moreover, ( x , y ) ∈ D t (x,y) \in D_t ( x , y ) ∈ D t if and only if ( x , y , z ) ∈ V t (x,y,z) \in V_t ( x , y , z ) ∈ V t . Therefore, there is a one-to-one correspondence between D t D_t D t and V t V_t V t , i.e., ∣ V t ∣ = ∣ D t ∣ |V_t| = |D_t| ∣ V t ∣ = ∣ D t ∣ .
Appendix B.
Proof of Lemma 2.2. We first derive a product identity. For a given y ∈ F q ∗ y \in\mathbb{F}_q^* y ∈ F q ∗ , making the substitution x = r y x = ry x = r y in the expansion of K m ( a ) K m ( b ) K_m(a)K_m(b) K m ( a ) K m ( b ) yields
K m ( a ) K m ( b ) = ∑ x ∈ F q ∗ y ∈ F q ∗ χ ( x + a x + y + b y ) = ∑ r ∈ F q ∗ y ∈ F q ∗ χ ( ( r + 1 ) y + a / r + b y ) = ∑ y ∈ F q ∗ χ ( a + b y ) + ∑ r ∈ F q ∖ { 0 , 1 } K m ( ( r + 1 ) ( a / r + b ) ) = ∑ y ∈ F q ∗ χ ( a + b y ) + ∑ r ∈ F q ∖ { 0 , 1 } K m ( a + b + a / r + b r ) , (70) \begin{aligned}
K_m(a)K_m(b) &= \sum_{\substack{x\in\mathbb{F}_q^*\\ y\in\mathbb{F}_q^*}} \chi\left(x+\frac{a}{x}+y+\frac{b}{y}\right) \\
&= \sum_{\substack{r\in\mathbb{F}_q^*\\ y\in\mathbb{F}_q^*}} \chi\left((r+1)y+\frac{a/r+b}{y}\right) \\
&= \sum_{y\in\mathbb{F}_q^*}\chi\left(\frac{a+b}{y}\right)+\sum_{r\in\mathbb{F}_q\setminus\{0,1\}}K_m\left((r+1)(a/r+b)\right) \\
&= \sum_{y\in\mathbb{F}_q^*}\chi\left(\frac{a+b}{y}\right)+\sum_{r\in\mathbb{F}_q\setminus\{0,1\}}K_m\left(a+b+a/r+br\right), \tag*{(70)}
\end{aligned} K m ( a ) K m ( b ) = x ∈ F q ∗ y ∈ F q ∗ ∑ χ ( x + x a + y + y b ) = r ∈ F q ∗ y ∈ F q ∗ ∑ χ ( ( r + 1 ) y + y a / r + b ) = y ∈ F q ∗ ∑ χ ( y a + b ) + r ∈ F q ∖ { 0 , 1 } ∑ K m ( ( r + 1 ) ( a / r + b ) ) = y ∈ F q ∗ ∑ χ ( y a + b ) + r ∈ F q ∖ { 0 , 1 } ∑ K m ( a + b + a / r + b r ) , (70)
where the term with r = 1 r=1 r = 1 is q − 1 q-1 q − 1 when a = b a=b a = b and − 1 -1 − 1 otherwise, and for the terms for r ≠ 0 , 1 r\ne0,1 r = 0 , 1 we use the identity (4) .
Taking a = ω a=\omega a = ω and b = ω 2 b=\omega^2 b = ω 2 in (70) , and then setting v = ω 2 r v=\omega^2r v = ω 2 r , gives
K m ( ω ) 2 = − 1 + ∑ r ∈ F q ∖ { 0 , 1 } K m ( 1 + ω / r + ω 2 r ) = − 1 + ∑ v ∈ F q ∗ v ≠ ω 2 K m ( 1 + v + v − 1 ) = ∑ v ∈ F q ∗ K m ( 1 + v + v − 1 ) , \begin{aligned}
K_m(\omega)^2 &= -1+\sum_{r\in\mathbb{F}_q\setminus\{0,1\}}K_m(1+\omega/r+\omega^2r) \\
&= -1+\sum_{\substack{v\in\mathbb{F}_q^*\\v\ne\omega^2}}K_m(1+v+v^{-1}) \\
&= \sum_{v\in\mathbb{F}_q^*}K_m(1+v+v^{-1}),
\end{aligned} K m ( ω ) 2 = − 1 + r ∈ F q ∖ { 0 , 1 } ∑ K m ( 1 + ω / r + ω 2 r ) = − 1 + v ∈ F q ∗ v = ω 2 ∑ K m ( 1 + v + v − 1 ) = v ∈ F q ∗ ∑ K m ( 1 + v + v − 1 ) ,
where in the second equality we use the substitution v = ω 2 r v=\omega^2r v = ω 2 r and the last equality follows from 1 + ω 2 + ω = 0 1+\omega^2+\omega=0 1 + ω 2 + ω = 0 and K m ( 0 ) = − 1 K_m(0)=-1 K m ( 0 ) = − 1 . This proves (i).
Taking a = b = 1 a=b=1 a = b = 1 in (70) and using K m ( 0 ) = − 1 K_m(0)=-1 K m ( 0 ) = − 1 , we obtain
K m ( 1 ) 2 = q − 1 + ∑ r ∈ F q ∖ { 0 , 1 } K m ( r + r − 1 ) = q + ∑ r ∈ F q ∗ K m ( r + r − 1 ) . (71) K_m(1)^2=q-1+\sum_{r\in\mathbb{F}_q\setminus\{0,1\}}K_m(r+r^{-1})=q+\sum_{r\in\mathbb{F}_q^*}K_m(r+r^{-1}). \tag*{(71)} K m ( 1 ) 2 = q − 1 + r ∈ F q ∖ { 0 , 1 } ∑ K m ( r + r − 1 ) = q + r ∈ F q ∗ ∑ K m ( r + r − 1 ) . (71)
We next count the preimages of the map v ↦ v + v − 1 v\mapsto v+v^{-1} v ↦ v + v − 1 on F q ∗ \mathbb{F}_q^* F q ∗ . The value 0 0 0 has the unique preimage v = 1 v=1 v = 1 . For t ≠ 0 t\ne0 t = 0 , the equation v + v − 1 = t v+v^{-1}=t v + v − 1 = t is equivalent to
v 2 + t v + 1 = 0. v^2+tv+1=0. v 2 + t v + 1 = 0.
The number of its solutions in F q ∗ \mathbb{F}_q^* F q ∗ is 1 + χ ( t − 1 ) 1+\chi(t^{-1}) 1 + χ ( t − 1 ) . Thus, for every b ∈ F q b\in\mathbb{F}_q b ∈ F q ,
∑ v ∈ F q ∗ K m ( b + v + v − 1 ) = K m ( b ) + ∑ t ∈ F q ∗ ( 1 + χ ( t − 1 ) ) K m ( b + t ) = ∑ t ∈ F q ∗ χ ( t − 1 ) K m ( b + t ) . \sum_{v\in\mathbb{F}_q^*}K_m(b+v+v^{-1})=K_m(b)+\sum_{t\in\mathbb{F}_q^*}(1+\chi(t^{-1}))K_m(b+t)
=\sum_{t\in\mathbb{F}_q^*}\chi(t^{-1})K_m(b+t). v ∈ F q ∗ ∑ K m ( b + v + v − 1 ) = K m ( b ) + t ∈ F q ∗ ∑ ( 1 + χ ( t − 1 )) K m ( b + t ) = t ∈ F q ∗ ∑ χ ( t − 1 ) K m ( b + t ) .
since
K m ( b ) + ∑ t ∈ F q ∗ K m ( b + t ) = ∑ c ∈ F q K m ( c ) = 0. K_m(b)+\sum_{t\in\mathbb{F}_q^*}K_m(b+t)=\sum_{c\in\mathbb{F}_q}K_m(c)=0. K m ( b ) + t ∈ F q ∗ ∑ K m ( b + t ) = c ∈ F q ∑ K m ( c ) = 0.
Taking b = 0 b=0 b = 0 and using (71) proves (ii). Taking b = 1 b=1 b = 1 and using (i) gives (iii)
Finally, expanding the Kloosterman sum K m ( c 3 + c ) K_m(c^3+c) K m ( c 3 + c ) and substituting x = c v x=cv x = c v in the inner sum, we obtain
∑ c ∈ F q ∗ K m ( c 3 + c ) = ∑ c ≠ 0 ∑ x ≠ 0 χ ( x + c 3 + c x ) = ∑ v ≠ 0 ∑ c ≠ 0 χ ( c v + c 2 v + v − 1 ) = ∑ v ≠ 0 χ ( v − 1 ) ∑ c ≠ 0 χ ( c ( v + v − 1 / 2 ) ) = q ∑ v 3 = 1 χ ( v − 1 ) − ∑ v ≠ 0 χ ( v − 1 ) , \sum_{c\in\mathbb{F}_q^*}K_m(c^3+c)
=\sum_{c\ne0}\sum_{x\ne0}\chi\left(x+\frac{c^3+c}{x}\right)
=\sum_{v\ne0}\sum_{c\ne0}\chi\left(cv+\frac{c^2}{v}+v^{-1}\right)
=\sum_{v\ne0}\chi(v^{-1})\sum_{c\ne0}\chi\left(c(v+v^{-1/2})\right)
=q\sum_{v^3=1}\chi(v^{-1})-\sum_{v\ne0}\chi(v^{-1}), c ∈ F q ∗ ∑ K m ( c 3 + c ) = c = 0 ∑ x = 0 ∑ χ ( x + x c 3 + c ) = v = 0 ∑ c = 0 ∑ χ ( c v + v c 2 + v − 1 ) = v = 0 ∑ χ ( v − 1 ) c = 0 ∑ χ ( c ( v + v − 1/2 ) ) = q v 3 = 1 ∑ χ ( v − 1 ) − v = 0 ∑ χ ( v − 1 ) ,
where we use the fact v + v − 1 / 2 = 0 v+v^{-1/2}=0 v + v − 1/2 = 0 if and only if v 3 = 1 v^3=1 v 3 = 1 . Since m m m is even, the three roots are v = 1 , ω , ω 2 v=1,\omega,\omega^2 v = 1 , ω , ω 2 , for which χ ( v − 1 ) \chi(v^{-1}) χ ( v − 1 ) takes the values 1 , σ , σ 1,\sigma,\sigma 1 , σ , σ , respectively. Therefore,
∑ c ∈ F q ∗ K m ( c 3 + c ) = ( 2 σ + 1 ) q + 1. \sum_{c\in\mathbb{F}_q^*}K_m(c^3+c)=(2\sigma+1)q+1. c ∈ F q ∗ ∑ K m ( c 3 + c ) = ( 2 σ + 1 ) q + 1.
□ \square □
Proof of Proposition 5.4. We prove the formulas for S 1 , 1 S_{1,1} S 1 , 1 and S 2 , 0 S_{2,0} S 2 , 0 in two parts, treating S 1 , 1 S_{1,1} S 1 , 1 first and S 2 , 0 S_{2,0} S 2 , 0 second.
Part (i): For S 1 , 1 S_{1,1} S 1 , 1 , from (33) , we have
χ ( F 3 + F 4 ) = χ ( 1 + x y x + y + ( x + y ) ( x + y ) x y + 1 + x y x + y x y + 1 ) \chi(F_3+F_4)=\chi\left(\frac{1+xy}{x+y}+\frac{(x+y)(x+y)}{xy+1}+xy\frac{x+y}{xy+1}\right) χ ( F 3 + F 4 ) = χ ( x + y 1 + x y + x y + 1 ( x + y ) ( x + y ) + x y x y + 1 x + y )
Under the sum-product substitution in (37) ,
χ ( F 3 + F 4 ) = χ ( 1 + v 2 s + s 2 v 2 + 1 + v 2 s 1 + v 2 ) = χ ( 1 + v 2 s + s v + 1 + v 2 s 1 + v 2 ) = χ ( v 2 + v + 1 v 2 + 1 s + 1 + v 2 s ) . \chi(F_3+F_4)=\chi\left(\frac{1+v^2}{s}+\frac{s^2}{v^2+1}+v^2\frac{s}{1+v^2}\right)
=\chi\left(\frac{1+v^2}{s}+\frac{s}{v+1}+v^2\frac{s}{1+v^2}\right)
=\chi\left(\frac{v^2+v+1}{v^2+1}s+\frac{1+v^2}{s}\right). χ ( F 3 + F 4 ) = χ ( s 1 + v 2 + v 2 + 1 s 2 + v 2 1 + v 2 s ) = χ ( s 1 + v 2 + v + 1 s + v 2 1 + v 2 s ) = χ ( v 2 + 1 v 2 + v + 1 s + s 1 + v 2 ) .
By Lemma 5.3, we get a = v 2 + v + 1 v 2 + 1 a=\frac{v^2+v+1}{v^2+1} a = v 2 + 1 v 2 + v + 1 , b = v 2 + 1 b=v^2+1 b = v 2 + 1 and c = 0 c=0 c = 0 and
S 1 , 1 = ∑ v ≠ 0 , 1 [ K m ( v 2 + v + 1 v 2 + 1 , v 2 + 1 ) + K m ( v 2 + v + 1 v 2 + 1 , v 2 + v + 1 ) ] − 2 ∑ v ≠ 0 , 1 χ ( v 2 + v ) = ∑ v ≠ 0 , 1 K m ( v 2 + v + 1 ) + ∑ v ≠ 0 , 1 K m ( ( v 2 + v + 1 ) 2 v 2 + 1 ) + 2 q − 2 ( q − 2 ) = ∑ v ≠ 0 , 1 K m ( v 2 + v + 1 ) + ∑ v ≠ 0 , 1 K m ( v 2 + v + 1 v + 1 ) + 4 , \begin{aligned}
S_{1,1}
&=\sum_{v\ne0,1}\left[K_m\left(\frac{v^2+v+1}{v^2+1},v^2+1\right)+K_m\left(\frac{v^2+v+1}{v^2+1},v^2+v+1\right)\right]-2\sum_{v\ne0,1}\chi(v^2+v)\\
&=\sum_{v\ne0,1}K_m\left(v^2+v+1\right)+\sum_{v\ne0,1}K_m\left(\frac{(v^2+v+1)^2}{v^2+1}\right)+2q-2(q-2)\\
&=\sum_{v\ne0,1}K_m\left(v^2+v+1\right)+\sum_{v\ne0,1}K_m\left(\frac{v^2+v+1}{v+1}\right)+4,
\end{aligned} S 1 , 1 = v = 0 , 1 ∑ [ K m ( v 2 + 1 v 2 + v + 1 , v 2 + 1 ) + K m ( v 2 + 1 v 2 + v + 1 , v 2 + v + 1 ) ] − 2 v = 0 , 1 ∑ χ ( v 2 + v ) = v = 0 , 1 ∑ K m ( v 2 + v + 1 ) + v = 0 , 1 ∑ K m ( v 2 + 1 ( v 2 + v + 1 ) 2 ) + 2 q − 2 ( q − 2 ) = v = 0 , 1 ∑ K m ( v 2 + v + 1 ) + v = 0 , 1 ∑ K m ( v + 1 v 2 + v + 1 ) + 4 ,
where the second equality holds since K m ( v 2 + v + 1 v 2 + 1 , v 2 + v + 1 ) = q − 1 K_m\left(\frac{v^2+v+1}{v^2+1},v^2+v+1\right)=q-1 K m ( v 2 + 1 v 2 + v + 1 , v 2 + v + 1 ) = q − 1 if and only if v 2 + v + 1 = 0 v^2+v+1=0 v 2 + v + 1 = 0 . Furthermore, by making a substitution t = v 2 + v t=v^2+v t = v 2 + v in the first sum and replacing v v v with v + 1 v+1 v + 1 in the second sum, we get
S 1 , 1 = ∑ t ≠ 0 ( 1 + χ ( t ) ) K m ( t + 1 ) + ∑ v ≠ 0 , 1 K m ( v 2 + v + 1 v ) + 4 = ∑ t ≠ 0 K m ( t + 1 ) + ∑ t ≠ 0 χ ( t ) K m ( t + 1 ) + ∑ v ≠ 0 K m ( v + v − 1 + 1 ) − K m ( 1 ) + 4 = ∑ t ∈ F q K m ( t + 1 ) + ∑ t ≠ 0 χ ( t ) K m ( t + 1 ) − 2 K m ( 1 ) + K m ( ω ) 2 − K m ( 1 ) + 4 = ∑ t ∈ F q χ ( t ) K m ( t ) + K m ( ω ) 2 − 3 K m ( 1 ) + 4 = q + K m ( ω ) 2 − 3 K m ( 1 ) + 4 , \begin{aligned}
S_{1,1}
&=\sum_{t\ne0}(1+\chi(t))K_m(t+1)+\sum_{v\ne0,1}K_m\left(\frac{v^2+v+1}{v}\right)+4\\
&=\sum_{t\ne0}K_m(t+1)+\sum_{t\ne0}\chi(t)K_m(t+1)+\sum_{v\ne0}K_m(v+v^{-1}+1)-K_m(1)+4\\
&=\sum_{t\in\mathbb{F}_q}K_m(t+1)+\sum_{t\ne0}\chi(t)K_m(t+1)-2K_m(1)+K_m(\omega)^2-K_m(1)+4\\
&=\sum_{t\in\mathbb{F}_q}\chi(t)K_m(t)+K_m(\omega)^2-3K_m(1)+4\\
&=q+K_m(\omega)^2-3K_m(1)+4,
\end{aligned} S 1 , 1 = t = 0 ∑ ( 1 + χ ( t )) K m ( t + 1 ) + v = 0 , 1 ∑ K m ( v v 2 + v + 1 ) + 4 = t = 0 ∑ K m ( t + 1 ) + t = 0 ∑ χ ( t ) K m ( t + 1 ) + v = 0 ∑ K m ( v + v − 1 + 1 ) − K m ( 1 ) + 4 = t ∈ F q ∑ K m ( t + 1 ) + t = 0 ∑ χ ( t ) K m ( t + 1 ) − 2 K m ( 1 ) + K m ( ω ) 2 − K m ( 1 ) + 4 = t ∈ F q ∑ χ ( t ) K m ( t ) + K m ( ω ) 2 − 3 K m ( 1 ) + 4 = q + K m ( ω ) 2 − 3 K m ( 1 ) + 4 ,
where we use the identities in Lemma 2.1 (iii) and Lemma 2.2 (i) to simplify the calculation.
Part (ii): For S 2 , 0 S_{2,0} S 2 , 0 , since
F 1 + F 2 = 1 x + y ( x + y ) x y + 1 + 1 y + x ( x + y ) x y + 1 = x + y x y + ( x + y ) 2 x y + 1 . F_1+F_2=\frac{1}{x}+\frac{y(x+y)}{xy+1}+\frac{1}{y}+\frac{x(x+y)}{xy+1}=\frac{x+y}{xy}+\frac{(x+y)^2}{xy+1}. F 1 + F 2 = x 1 + x y + 1 y ( x + y ) + y 1 + x y + 1 x ( x + y ) = x y x + y + x y + 1 ( x + y ) 2 .
Using the substitution in (37) , we get
F 1 + F 2 = s 2 v 2 + 1 + s v 2 = s 2 ( v + 1 ) 2 + s v 2 . F_1+F_2=\frac{s^2}{v^2+1}+\frac{s}{v^2}=\frac{s^2}{(v+1)^2}+\frac{s}{v^2}. F 1 + F 2 = v 2 + 1 s 2 + v 2 s = ( v + 1 ) 2 s 2 + v 2 s .
and
χ ( F 1 + F 2 ) = χ ( a s ) , a = v 2 + v + 1 v 2 ( v + 1 ) . \chi(F_1+F_2)=\chi(as),\qquad a=\frac{v^2+v+1}{v^2(v+1)}. χ ( F 1 + F 2 ) = χ ( a s ) , a = v 2 ( v + 1 ) v 2 + v + 1 .
By Lemma 5.3, we have a = v 2 + v + 1 v 2 ( v + 1 ) a=\frac{v^2+v+1}{v^2(v+1)} a = v 2 ( v + 1 ) v 2 + v + 1 , b = 0 b=0 b = 0 and c = 0 c=0 c = 0 . Then,
S 2 , 0 = ∑ v ≠ 0 , 1 K m ( v 2 + v + 1 v 2 ( v + 1 ) , 0 ) + ∑ v ≠ 0 , 1 K m ( v 2 + v + 1 v 2 ( v + 1 ) , v ) − 2 ∑ v ≠ 0 , 1 χ ( v 3 + 1 v 2 ) = 2 q + ∑ v ≠ 0 , 1 K m ( 0 ) + ∑ v ≠ 0 , 1 K m ( v 2 + v + 1 v ( v + 1 ) ) − 2 ∑ v ≠ 0 , 1 χ ( v + 1 v ) = 2 q − ( q − 2 ) + ∑ v ≠ 0 , 1 K m ( 1 + 1 v 2 + v ) − 2 K m ( 1 ) + 2 = q + 4 − 2 K m ( 1 ) + ∑ v ≠ 0 , 1 K m ( 1 + 1 v 2 + v ) . (72) \begin{aligned}
S_{2,0}
&=\sum_{v\ne0,1}K_m\left(\frac{v^2+v+1}{v^2(v+1)},0\right)+\sum_{v\ne0,1}K_m\left(\frac{v^2+v+1}{v^2(v+1)},v\right)-2\sum_{v\ne0,1}\chi\left(\frac{v^3+1}{v^2}\right)\\
&=2q+\sum_{v\ne0,1}K_m(0)+\sum_{v\ne0,1}K_m\left(\frac{v^2+v+1}{v(v+1)}\right)-2\sum_{v\ne0,1}\chi\left(v+\frac{1}{v}\right)\\
&=2q-(q-2)+\sum_{v\ne0,1}K_m\left(1+\frac{1}{v^2+v}\right)-2K_m(1)+2\\
&=q+4-2K_m(1)+\sum_{v\ne0,1}K_m\left(1+\frac{1}{v^2+v}\right). \tag*{(72)}
\end{aligned} S 2 , 0 = v = 0 , 1 ∑ K m ( v 2 ( v + 1 ) v 2 + v + 1 , 0 ) + v = 0 , 1 ∑ K m ( v 2 ( v + 1 ) v 2 + v + 1 , v ) − 2 v = 0 , 1 ∑ χ ( v 2 v 3 + 1 ) = 2 q + v = 0 , 1 ∑ K m ( 0 ) + v = 0 , 1 ∑ K m ( v ( v + 1 ) v 2 + v + 1 ) − 2 v = 0 , 1 ∑ χ ( v + v 1 ) = 2 q − ( q − 2 ) + v = 0 , 1 ∑ K m ( 1 + v 2 + v 1 ) − 2 K m ( 1 ) + 2 = q + 4 − 2 K m ( 1 ) + v = 0 , 1 ∑ K m ( 1 + v 2 + v 1 ) . (72)
Finally, let t = v 2 + v t=v^{2}+v t = v 2 + v . Then,
∑ v ≠ 0 , 1 K m ( 1 + 1 v 2 + v ) = ∑ t ≠ 0 ( 1 + χ ( t ) ) K m ( 1 + t − 1 ) = ∑ t ≠ 0 K m ( 1 + t ) + ∑ t ≠ 0 χ ( t − 1 ) K m ( 1 + t ) = − K m ( 1 ) + K m ( ω ) 2 . (73) \begin{aligned}
\sum_{v\ne0,1}K_m\left(1+\frac{1}{v^{2}+v}\right)
&=\sum_{t\ne0}(1+\chi(t))K_m(1+t^{-1})\\
&=\sum_{t\ne0}K_m(1+t)+\sum_{t\ne0}\chi(t^{-1})K_m(1+t)\\
&=-K_m(1)+K_m(\omega)^{2}. \tag*{(73)}
\end{aligned} v = 0 , 1 ∑ K m ( 1 + v 2 + v 1 ) = t = 0 ∑ ( 1 + χ ( t )) K m ( 1 + t − 1 ) = t = 0 ∑ K m ( 1 + t ) + t = 0 ∑ χ ( t − 1 ) K m ( 1 + t ) = − K m ( 1 ) + K m ( ω ) 2 . (73)
The second equality follows by replacing t t t with t − 1 t^{-1} t − 1 , and the last follows from Lemma 2.1 (i) and Lemma 2.2 (iii). Combining (72) and (73) , we obtain
S 2 , 0 = q + 4 − 3 K m ( 1 ) + K m ( ω ) 2 . S_{2,0}=q+4-3K_m(1)+K_m(\omega)^{2}. S 2 , 0 = q + 4 − 3 K m ( 1 ) + K m ( ω ) 2 .
Proof of Proposition 5.5. For the remaining four classes, indexed by ( j , t ) ∈ { ( 0 , 1 ) , ( 2 , 1 ) , ( 3 , 0 ) , ( 3 , 1 ) } (j,t)\in\{(0,1),(2,1),(3,0),(3,1)\} ( j , t ) ∈ {( 0 , 1 ) , ( 2 , 1 ) , ( 3 , 0 ) , ( 3 , 1 )} , we apply Lemma 5.3 by determining the coefficients a a a , b b b , and c c c , together with the value of χ ( a s + b s + c ) \chi\left(as+\frac{b}{s}+c\right) χ ( a s + s b + c ) at s = v 2 + 1 s=v^{2}+1 s = v 2 + 1 . The results, obtained by direct calculation, are listed in Table 1 , where h v = v 2 + v + 1 h_v=v^{2}+v+1 h v = v 2 + v + 1 . Note that in each row, a = 0 a=0 a = 0 if and only if v ∈ { ω , ω 2 } v\in\{\omega,\omega^{2}\} v ∈ { ω , ω 2 } , where ω \omega ω is a primitive element of F 4 \mathbb{F}_{4} F 4 .
Table 1.
Table 1 (PDF p. 40). Coefficients in S j , t S_{j,t} S j , t ’s and the values of χ ( a s + b s + c ) \chi\left(as+\frac{b}{s}+c\right) χ ( a s + s b + c ) at s = v 2 + 1 s=v^{2}+1 s = v 2 + 1
Using the values in Table 1 , we apply Lemma 5.3 to each of the four classes in turn. In what follows, the proof are divided into four parts: Part (i) for S 0 , 1 S_{0,1} S 0 , 1 , Part (ii) for S 2 , 1 S_{2,1} S 2 , 1 , Part (iii) for S 3 , 0 S_{3,0} S 3 , 0 , and Part (iv) for S 3 , 1 S_{3,1} S 3 , 1 .
Part (i): For S 0 , 1 S_{0,1} S 0 , 1 , by Table 1 , we have
S 0 , 1 = ∑ v ≠ 0 , 1 χ ( v ) [ K m ( h v ( v + 1 ) 2 , 0 ) + K m ( h v ( v + 1 ) 2 , v ) ] − 2 ∑ v ≠ 0 , 1 χ ( v ) = 2 σ q + ∑ v ≠ 0 , 1 χ ( v ) K m ( 0 ) + ∑ v ≠ 0 , 1 χ ( v ) K m ( v h v ( v + 1 ) 2 ) + 4 = 2 σ q + 6 + ∑ v ≠ 0 , 1 χ ( v ) K m ( v + 1 + ( v + 1 ) − 2 ) = 2 σ q + 6 + ∑ u ≠ 0 , 1 χ ( u ) K m ( u + u − 2 ) , = 2 σ q + 7 + ∑ u ≠ 0 χ ( u ) K m ( u + u − 2 ) , \begin{aligned}
S_{0,1}
&=\sum_{v\ne0,1}\chi(v)\left[K_m\left(\frac{h_v}{(v+1)^2},0\right)+K_m\left(\frac{h_v}{(v+1)^2},v\right)\right]-2\sum_{v\ne0,1}\chi(v)\\
&=2\sigma_q+\sum_{v\ne0,1}\chi(v)K_m(0)+\sum_{v\ne0,1}\chi(v)K_m\left(\frac{vh_v}{(v+1)^2}\right)+4\\
&=2\sigma_q+6+\sum_{v\ne0,1}\chi(v)K_m\left(v+1+(v+1)^{-2}\right)\\
&=2\sigma_q+6+\sum_{u\ne0,1}\chi(u)K_m\left(u+u^{-2}\right),\\
&=2\sigma_q+7+\sum_{u\ne0}\chi(u)K_m\left(u+u^{-2}\right),
\end{aligned} S 0 , 1 = v = 0 , 1 ∑ χ ( v ) [ K m ( ( v + 1 ) 2 h v , 0 ) + K m ( ( v + 1 ) 2 h v , v ) ] − 2 v = 0 , 1 ∑ χ ( v ) = 2 σ q + v = 0 , 1 ∑ χ ( v ) K m ( 0 ) + v = 0 , 1 ∑ χ ( v ) K m ( ( v + 1 ) 2 v h v ) + 4 = 2 σ q + 6 + v = 0 , 1 ∑ χ ( v ) K m ( v + 1 + ( v + 1 ) − 2 ) = 2 σ q + 6 + u = 0 , 1 ∑ χ ( u ) K m ( u + u − 2 ) , = 2 σ q + 7 + u = 0 ∑ χ ( u ) K m ( u + u − 2 ) ,
where we use the substitution u = v + 1 u=v+1 u = v + 1 , which permutes F q ∖ { 0 , 1 } \mathbb{F}_q\setminus\{0,1\} F q ∖ { 0 , 1 } and χ ( v ) = χ ( u ) \chi(v)=\chi(u) χ ( v ) = χ ( u ) since χ ( 1 ) = 1 \chi(1)=1 χ ( 1 ) = 1 . Hence S 0 , 1 = 2 σ q + 7 + K 1 S_{0,1}=2\sigma_q+7+K_1 S 0 , 1 = 2 σ q + 7 + K 1 .
Part (ii): For S 2 , 1 S_{2,1} S 2 , 1 , we have
S 2 , 1 = ∑ v ≠ 0 , 1 χ ( v ) [ K m ( h v 2 v 2 ( v + 1 ) 2 , 0 ) + K m ( h v 2 v 2 ( v + 1 ) 2 , v ) ] − 2 ∑ v ≠ 0 , 1 χ ( v − 1 ) = 2 σ q + ∑ v ≠ 0 , 1 χ ( v ) K m ( 0 ) + ∑ v ≠ 0 , 1 χ ( v ) K m ( h v 2 v ( v + 1 ) 2 ) + 4 = 2 σ q + 6 + K 2 . \begin{aligned}
S_{2,1}={}&\sum_{v\ne0,1}\chi(v)\left[K_m\left(\frac{h_v^2}{v^2(v+1)^2},0\right)+K_m\left(\frac{h_v^2}{v^2(v+1)^2},v\right)\right]-2\sum_{v\ne0,1}\chi(v^{-1})\\
={}&2\sigma q+\sum_{v\ne0,1}\chi(v)K_m(0)+\sum_{v\ne0,1}\chi(v)K_m\left(\frac{h_v^2}{v(v+1)^2}\right)+4\\
={}&2\sigma q+6+K_2.
\end{aligned} S 2 , 1 = = = v = 0 , 1 ∑ χ ( v ) [ K m ( v 2 ( v + 1 ) 2 h v 2 , 0 ) + K m ( v 2 ( v + 1 ) 2 h v 2 , v ) ] − 2 v = 0 , 1 ∑ χ ( v − 1 ) 2 σ q + v = 0 , 1 ∑ χ ( v ) K m ( 0 ) + v = 0 , 1 ∑ χ ( v ) K m ( v ( v + 1 ) 2 h v 2 ) + 4 2 σ q + 6 + K 2 .
Part (iii): For S 3 , 0 S_{3,0} S 3 , 0 , we have
S 3 , 0 = ∑ v ≠ 0 , 1 χ ( v ) [ K m ( h v v 2 ( v + 1 ) , v 2 + 1 ) + K m ( h v v 2 ( v + 1 ) , v 2 + v + 1 ) ] − 2 ∑ v ≠ 0 , 1 χ ( v − 1 ) = 2 σ q + 4 + ∑ v ≠ 0 , 1 χ ( v ) [ K m ( v + v − 2 ) + K m ( h v 2 v 2 ( v + 1 ) ) ] = 2 σ q + 4 + K 1 + 1 + ∑ v ≠ 0 , 1 χ ( v ) K m ( h v 2 v 2 ( v + 1 ) ) = 2 σ q + 5 + K 1 + ∑ u ≠ 0 , 1 χ ( u ) K m ( h u 2 u ( u + 1 ) 2 ) = 2 σ q + 5 + K 1 + K 2 , \begin{aligned}
S_{3,0}={}&\sum_{v\ne0,1}\chi(v)\left[K_m\left(\frac{h_v}{v^2(v+1)},v^2+1\right)+K_m\left(\frac{h_v}{v^2(v+1)},v^2+v+1\right)\right]-2\sum_{v\ne0,1}\chi(v^{-1})\\
={}&2\sigma q+4+\sum_{v\ne0,1}\chi(v)\left[K_m(v+v^{-2})+K_m\left(\frac{h_v^2}{v^2(v+1)}\right)\right]\\
={}&2\sigma q+4+K_1+1+\sum_{v\ne0,1}\chi(v)K_m\left(\frac{h_v^2}{v^2(v+1)}\right)\\
={}&2\sigma q+5+K_1+\sum_{u\ne0,1}\chi(u)K_m\left(\frac{h_u^2}{u(u+1)^2}\right)\\
={}&2\sigma q+5+K_1+K_2,
\end{aligned} S 3 , 0 = = = = = v = 0 , 1 ∑ χ ( v ) [ K m ( v 2 ( v + 1 ) h v , v 2 + 1 ) + K m ( v 2 ( v + 1 ) h v , v 2 + v + 1 ) ] − 2 v = 0 , 1 ∑ χ ( v − 1 ) 2 σ q + 4 + v = 0 , 1 ∑ χ ( v ) [ K m ( v + v − 2 ) + K m ( v 2 ( v + 1 ) h v 2 ) ] 2 σ q + 4 + K 1 + 1 + v = 0 , 1 ∑ χ ( v ) K m ( v 2 ( v + 1 ) h v 2 ) 2 σ q + 5 + K 1 + u = 0 , 1 ∑ χ ( u ) K m ( u ( u + 1 ) 2 h u 2 ) 2 σ q + 5 + K 1 + K 2 ,
where the last second equality holds by the substitution u = v + 1 u=v+1 u = v + 1 and the fact χ ( u + 1 ) = χ ( u ) \chi(u+1)=\chi(u) χ ( u + 1 ) = χ ( u ) .
Part (iv): For S 3 , 1 S_{3,1} S 3 , 1 , by Table 1 and Lemma 5.3, we have
S 3 , 1 = ∑ v ≠ 0 , 1 [ K m ( h v 2 v 2 ( v + 1 ) 2 , v 2 + 1 ) + K m ( h v 2 v 2 ( v + 1 ) 2 , v 2 + v + 1 ) ] − 2 ∑ v ≠ 0 , 1 χ ( v + v − 1 ) = ∑ v ≠ 0 , 1 K m ( h v 2 v 2 ) + 2 q + ∑ v ≠ 0 , 1 K m ( h v 3 v 2 ( v + 1 ) 2 ) − 2 K m ( 1 ) + 2 = ∑ v ≠ 0 , 1 K m ( v + v − 1 + 1 ) + 2 q + K 3 − 2 K m ( 1 ) + 2 = ∑ v ≠ 0 K m ( v + v − 1 + 1 ) + 2 q + K 3 − 3 K m ( 1 ) + 2 = 2 q + K 3 + K m ( ω ) 2 − 3 K m ( 1 ) + 2. \begin{aligned}
S_{3,1}={}&\sum_{v\ne0,1}\left[K_m\left(\frac{h_v^2}{v^2(v+1)^2},v^2+1\right)+K_m\left(\frac{h_v^2}{v^2(v+1)^2},v^2+v+1\right)\right]-2\sum_{v\ne0,1}\chi(v+v^{-1})\\
={}&\sum_{v\ne0,1}K_m\left(\frac{h_v^2}{v^2}\right)+2q+\sum_{v\ne0,1}K_m\left(\frac{h_v^3}{v^2(v+1)^2}\right)-2K_m(1)+2\\
={}&\sum_{v\ne0,1}K_m\left(v+v^{-1}+1\right)+2q+K_3-2K_m(1)+2\\
={}&\sum_{v\ne0}K_m\left(v+v^{-1}+1\right)+2q+K_3-3K_m(1)+2\\
={}&2q+K_3+K_m(\omega)^2-3K_m(1)+2.
\end{aligned} S 3 , 1 = = = = = v = 0 , 1 ∑ [ K m ( v 2 ( v + 1 ) 2 h v 2 , v 2 + 1 ) + K m ( v 2 ( v + 1 ) 2 h v 2 , v 2 + v + 1 ) ] − 2 v = 0 , 1 ∑ χ ( v + v − 1 ) v = 0 , 1 ∑ K m ( v 2 h v 2 ) + 2 q + v = 0 , 1 ∑ K m ( v 2 ( v + 1 ) 2 h v 3 ) − 2 K m ( 1 ) + 2 v = 0 , 1 ∑ K m ( v + v − 1 + 1 ) + 2 q + K 3 − 2 K m ( 1 ) + 2 v = 0 ∑ K m ( v + v − 1 + 1 ) + 2 q + K 3 − 3 K m ( 1 ) + 2 2 q + K 3 + K m ( ω ) 2 − 3 K m ( 1 ) + 2.
where the last equality follows from Lemma 2.2 (i). □
Proof of Lemma 5.11. Note that F ( u ) F(u) F ( u ) has the following factorization
F ( u ) = ( u + 1 ) 2 ( u 2 + u − 1 ) , F(u)=(u+1)^2(u^2+u^{-1}), F ( u ) = ( u + 1 ) 2 ( u 2 + u − 1 ) ,
which suggests K m ( F ( u ) ) K_m(F(u)) K m ( F ( u )) can be written in the form = K m ( ( u + 1 ) 2 , u 2 + u − 1 ) =K_m\left((u+1)^2,u^2+u^{-1}\right) = K m ( ( u + 1 ) 2 , u 2 + u − 1 ) . For u ≠ 0 , 1 u\ne0,1 u = 0 , 1 , we make a new substitution x = ( u + 1 ) 2 t 2 x=(u+1)^2t^2 x = ( u + 1 ) 2 t 2 in K m ( F ( u ) ) K_m(F(u)) K m ( F ( u )) . This is a bijection of F q ∗ \mathbb{F}_q^* F q ∗ , and
K m ( F ( u ) ) = ∑ x ≠ 0 χ ( x + F ( u ) x ) = ∑ t ≠ 0 χ ( ( u + 1 ) 2 t 2 + u 2 + u − 1 t 2 ) = ∑ t ≠ 0 χ ( ( u + 1 ) t + u t + 1 u t 2 ) = ∑ t ≠ 0 χ ( t + u ( t + t − 1 ) + 1 u t 2 ) . \begin{aligned}
K_m(F(u)) &= \sum_{x\ne0} \chi\left(x+\frac{F(u)}{x}\right) \\
&= \sum_{t\ne0} \chi\left((u+1)^2t^2+\frac{u^2+u^{-1}}{t^2}\right) \\
&= \sum_{t\ne0} \chi\left((u+1)t+\frac{u}{t}+\frac{1}{ut^2}\right) \\
&= \sum_{t\ne0} \chi\left(t+u(t+t^{-1})+\frac{1}{ut^2}\right).
\end{aligned} K m ( F ( u )) = x = 0 ∑ χ ( x + x F ( u ) ) = t = 0 ∑ χ ( ( u + 1 ) 2 t 2 + t 2 u 2 + u − 1 ) = t = 0 ∑ χ ( ( u + 1 ) t + t u + u t 2 1 ) = t = 0 ∑ χ ( t + u ( t + t − 1 ) + u t 2 1 ) .
Thus,
K 4 = ∑ u ≠ 0 ∑ t ≠ 0 χ ( t + u ( t + t − 1 ) + 1 u t 2 ) − ∑ t ≠ 0 χ ( 0 ) = ∑ t ≠ 0 χ ( t ) ∑ u ≠ 0 χ ( u ( t + t − 1 ) + 1 u t 2 ) − ( q − 1 ) = ∑ t ≠ 0 χ ( t ) K m ( t − 1 + t − 3 ) − ( q − 1 ) = ∑ t ≠ 0 χ ( t − 1 ) K m ( t + t 3 ) − ( q − 1 ) . \begin{aligned}
\mathbf{K}_4
&= \sum_{u\ne0}\sum_{t\ne0}\chi\left(t+u(t+t^{-1})+\frac{1}{ut^2}\right)-\sum_{t\ne0}\chi(0) \\
&= \sum_{t\ne0}\chi(t)\sum_{u\ne0}\chi\left(u(t+t^{-1})+\frac{1}{ut^2}\right)-(q-1) \\
&= \sum_{t\ne0}\chi(t)K_m(t^{-1}+t^{-3})-(q-1) \\
&= \sum_{t\ne0}\chi(t^{-1})K_m(t+t^3)-(q-1).
\end{aligned} K 4 = u = 0 ∑ t = 0 ∑ χ ( t + u ( t + t − 1 ) + u t 2 1 ) − t = 0 ∑ χ ( 0 ) = t = 0 ∑ χ ( t ) u = 0 ∑ χ ( u ( t + t − 1 ) + u t 2 1 ) − ( q − 1 ) = t = 0 ∑ χ ( t ) K m ( t − 1 + t − 3 ) − ( q − 1 ) = t = 0 ∑ χ ( t − 1 ) K m ( t + t 3 ) − ( q − 1 ) .
The last expression ∑ t ≠ 0 χ ( t − 1 ) K m ( t + t 3 ) \sum_{t\ne0}\chi(t^{-1})K_m(t+t^3) ∑ t = 0 χ ( t − 1 ) K m ( t + t 3 ) for K 1 \mathbf{K}_1 K 1 was established in (48) in the proof of Proposition 5.7 . Therefore, K 4 = K 1 − q + 1 \mathbf{K}_4=\mathbf{K}_1-q+1 K 4 = K 1 − q + 1 , and (47) gives the stated value. □