Abstract: In this paper, we produce new classes of MDS self-dual codes via (extended) generalized Reed-Solomon codes over finite fields of odd characteristic. Among our constructions, there are many MDS self-dual codes with new parameters which have never been reported. For odd prime power $q$ with $q$ square, the total number of lengths for MDS self-dual codes over $\mathbb{F}_q$ presented in this paper is much more than those in all the previous results.
Introduction
Maximum Distance Separable (MDS) codes represent the "gold standard" of error-correcting capability, as they achieve the theoretical Singleton bound for a given length and dimension. Within this class, self-dual codes—where the code is identical to its dual—are particularly significant due to their applications in cryptography, combinatorics, and lattice theory.
While the construction of MDS self-dual codes is largely resolved for finite fields of characteristic 2 (even $q$), the odd characteristic case remains a significant challenge. For a field of size $q$, the MDS conjecture suggests a maximum length of roughly $q+1$. However, systematically constructing codes for every possible even length $n \le q+1$ has proven difficult. For instance, in a field like $\mathbb{F}_{83^2}$, over 3,000 possible lengths might exist, yet prior to recent advancements, only approximately 700 had been successfully documented. This paper aims to fill these significant gaps by introducing new systematic construction methods.
Methodology
The researchers utilized Generalized Reed-Solomon (GRS) codes and their extended variants as the foundational building blocks. The primary technical hurdle in constructing self-dual GRS codes lies in the selection of evaluation points and multiplier vectors that satisfy the self-duality condition.
To address this, the authors moved beyond random searches, instead leveraging the inherent algebraic structures of finite fields:
- Group Homomorphism: By applying the second fundamental theorem of group homomorphism to cyclic subgroups, the authors partitioned the field into structured cosets.
- Quadratic Characters: Properties of quadratic characters (determining if an element is a "square") were employed to satisfy the necessary technical conditions for self-duality.
- Structured Evaluation Sets: Selecting evaluation points from specific subgroup cosets allowed the authors to prove self-duality for entire classes of code lengths simultaneously.
New Constructions and Key Results
The paper introduces five main construction theorems that significantly expand the known parameter space for MDS self-dual codes, particularly for fields where the size $q$ is a square ($q=r^2$). This subfield structure provides more flexibility to find elements with the specific quadratic properties needed for dual-code requirements.
Key highlights of these results include:
- Quantitative Advancement: In a comparative analysis for $q = 151^2$, the authors' methods increased the number of known code lengths from 862 to 1,228.
- Novel Parameters: New codes were successfully constructed for previously unreachable lengths, such as $n = 426$ and $n = 1006$ over $\mathbb{F}_{151^2}$, and $n = 730$ over $\mathbb{F}_{243^2}$.
- Theorem 5: This theorem provides a construction for lengths of the form $2tp^{ke}$, covering a vast range of parameters in fields of characteristic $p$.
Conclusion and Future Work
This work represents a substantial leap in the classification of MDS self-dual codes over odd characteristic fields. By relaxing previous constraints (such as $t = \sqrt{q}$), the authors have provided a much more flexible framework for code construction.
However, the classification remains incomplete. The total number of known lengths is still below the theoretical maximum of $(q+1)/2$. Future research may need to explore non-GRS MDS codes or further refine subgroup-based techniques to bridge the remaining gaps in the parameter space.
Summary updated