The coset weight distributions of certain BCH codes and a family of curves. (Q1419566)

The authors consider the problem of determining the weight distributions of the cosets of certain binary 3-error-correcting BCH codes of length \(2^m-1\) with \(m\) odd, which has been shown in [\textit{P. Charpin} and \textit{V. Zinoviev}, SIAM J. Discrete Math. 10, 128--145 (1997; Zbl 0870.94043)] to be related to the same problem for the extended BCH code. It was also shown in [loc. cit.] that this latter problem can be solved once the weight distributions of the cosets of weight 4 with syndrome \((0,1,A,B)\) are determined, which is in turn determined by the number \(N(A,B)\) of vectors of weight 4 in the coset concerned. The quantity \(N(A,B)\) is the number of \(S_4\)-orbits of solutions to the following system (1) of equations with distinct \(x_i \in {\mathbb F}_q\): \[ \begin{aligned} x_1 + x_2 + x_3 + x_4 &= 1 , \\ x_1^3 + x_2^3 + x_3^3 + x_4^3 &= A, \tag{1}\\ x_1^5 + x_2^5 + x_3^5 + x_4^5 &= B. \end{aligned} \] By carefully analyzing the curves defined by (1), the authors show that \(N(A,B)\) lies in an explicit interval of length approximately \(1.57 \sqrt{q}\), centered at about \(q/24\). This is done through decomposing the Jacobian variety of the curves of (1) into isogeny factors of dimension 1 and 2. Moreover, a heuristic argument is given to explain why almost all \(N(A,B)\) lie in an explicit interval of length about \(0.9 \sqrt{q}\). Some numerical results are also provided.