Non-split toric codes (Q2190910)

Consider a (projective, non-singular, absolutely irreducible, algebraic) variety \(X\) over a finite field \(GF(q)\), where \(q\) is a prime power. Let \(D\) be an ample sheaf on \(X\) and let \(H^0(X,D)=\Gamma(X,D)\) be the sheaf cohomology space on \(X\) with coefficients in \(D\). (In the case of curves this is a Riemann-Roch space. In the case of ``split'' toric varieties this is generated by monomials whose exponents lie in a polytope.) We recall the AG code construction: Let \(P = \{P_1,\dots,P_n\} \subset X(GF(q))\) be a set of \(n\) rational distinct points, The AG code \(C(P, D)\) is defined as \(C(P,D)=ev_P(\Gamma(X,D))\), where \(ev_P\) is the evaluation (germ) at \(P\) map, \(ev_P(f)=(f(P_1),\dots,f(P_n ))\). Codes associated to ``split'' toric varieties have been very well-investigated by a large number of mathematicians. To obtains estimates on the parameters of these (split) toric AG codes, one takes advantage of the extensive (and relatively simple) theory of toric varieties. This paper looks at a related class of varieties, so-called ``non-split'' toric varieties. As a very simple example, \(T_1(GF(q))=GF(q)^\times \times GF(q)^\times\) is a 2-dimensional split torus over \(GF(q)\), while \(T_2(GF(q))=GF(q^2)^\times\) is a 2-dimensional non-split torus over \(GF(q)\) (and note that \(T_2\) has more rational points than \(T_1\)). These non-split toric varieties form an interesting class, from the perspective of coding theory, since they tend to have more rational points than the split ones. However, the theory, which the author nicely summarized in section 2 of the paper under revew, is more complicated that the split case. None-the-less, the author is able to compute an explicit generator matrix and parity check polynomial of non-split toric AG codes. Moreover, in the case of some non-split toric surfaces the author is able to explicitly compute parameters of the corresponding non-split toric AG codes in a wide range of cases (and in other cases he gets a lower bound for the minimum distance). In several cases these parameters are best-known. See the paper itself for more precise statements of these results.