Abstract
The purpose of this paper is to classify and enumerate self-dual codes of length 6 over finite field . First, we classify these codes into three cases: decomposable, indecomposable non-MDS and MDS codes. Then, we complete the classification of non-MDS self-dual codes of length 6 over for all primes p in terms of their automorphism group. We obtain all inequivalent classes and find the necessary and sufficient conditions for the existence of each class. Finally, we obtain the number of MDS self-dual codes of length 6.
1. Introduction
The classification problem is one of the fundamental problems in various areas of mathematics. From the time self-dual codes began to attract attention amongst coding theorists, many papers have been published to classify binary self-dual codes [1,2,3] and non-binary self-dual codes [4,5,6,7,8]. While self-dual codes of moderate lengths are classified over a finite field for a fixed prime p in these papers, the first efforts are made in [9] to classify self-dual codes of the fixed length 4 over for all primes p. In [9], it is also shown that the classification of self-dual codes over for all primes p is essential to classify self-dual codes over integer ring for arbitrary m.
Let be the number of all self-dual codes of length n and s be the number of equivalent classes. The main tool for the classification of self-dual codes is the mass formula:
where is a representative of each equivalent class and is the group of γ-monomial transformations. Therefore, calculating the total number of codes and automorphisms of each equivalence class is critical for the classification of self-dual codes.
It is shown that self-dual codes of length 6 are classified simply into the following three cases by checking the number of zero elements of the standard generator matrix: decomposable case, indecomposable non-MDS case, and MDS case. For the decomposable codes, we need to understand the complete classification of self-dual codes of smaller lengths, which is already completed in [9]. For the indecomposable non-MDS codes, we complete the classification by computing the total number of bisorted standard generator matrices and orbits of each equivalent class. Consequently, with these results, we calculate the exact number of distinct MDS self-dual codes of length 6, so that we can obtain the mass formula for MDS codes.
This paper is organized as follows: first, we introduce some preliminaries to understand self-dual codes over in Section 2. In Section 3, we investigate self-dual codes of length 6 over and equivalence relations and automorphism groups in Section 4. Our main results are given in Section 5.
All computations in this paper were done with the computer algebra system Magma and SageMath.
2. Preliminaries
Let p be a prime number and n be a positive integer. A linear code of length n and dimension k over is a k-dimensional subspace of . An element of is called a codeword. A generator matrix of is a matrix whose rows form a basis of . Thus, a generator matrix of a linear code of length n and dimension k over is a matrix over . The space is equipped with the standard inner product, , where . The dual code is defined by
A linear code is called self-orthogonal if and self-dual if .
The weight of a codeword is the number of nonzero coordinates of the codeword. The minimum distance of , denoted by , is the smallest Hamming distance between distinct codewords. In determining the error-capability of , the minimum distance is the most important. For linear codes, the minimum distance equals the minimum weight of the non-zero codewords. It is well-known (see [10] for example) that a linear code of length n and dimension k satisfy the Singleton bound,
A code which achieves the equality in the Singleton bound is called MDS (maximum distance separable) code. Thus, a self-dual code of length over is MDS if the minimum weight equals .
Let be a group of permutation matrices of length n and be a group of diagonal matrices over with the diagonal elements , where for ; that is,
We denote by the permutation matrix corresponding and the group of all γ-monomial transformations of length n, is defined by
Since a monomial transformation does not preserve the self-duality in general, we only consider -monomial transformation in this paper (see [10], Thm 1.7.6). Two self-dual codes and of length are called equivalent and denoted by if there exists an element such that , where . An automorphism of is an element such that . The set of all automorphisms of forms the automorphism group as a subgroup of .
Any element has a unique representation for and . Thus, we define groups and which are called the permutation parts and sign parts of , respectively. Moreover, we denote a code with a generator matrix G by and by . We usually abuse the notations and simply write as and as , if there is no confusion.
Let denote the transpose of a matrix A. It is well-known that a self-dual code of length over is equivalent to a code with a standard generator matrix
where A is a matrix satisfying .
Definition 1.
Let , be self-dual codes of length and whose generator matrices are and , respectively. We define the direct sum of codes (direct sum of matrices at the same time), with the direct sum of generator matrices,
A code is called decomposable if it is equivalent to a direct sum of two codes. If a code is not decomposable, it is called indecomposable.
The next proposition is derived directly from the previous definitions.
Proposition 1.
Suppose that for some codes and . Then,
and
where and .
Proposition 2.
Let be a self-dual code of length over with a standard generator matrix Then,
is a generator matrix of .
Proof.
Since is self-dual, and . Thus, is non-singular. This implies that the rows of matrix form a basis of the code and . □
Corollary 1.
Let be a codeword of with a standard generator matrix , where , . Then, if and only if
Proof.
Suppose that . By the self-orthogonality,
Thus, . The other direction follows immediately in a similar way with the previous proposition. □
Corollary 2.
Let and be generator matrices of self-dual codes and , respectively. Then, and are equivalent.
Proof.
Let and , where denotes all one vector of length n. It is clear that is equal to by Proposition 2. □
Definition 2.
A matrix A is called bisorted if the rows and columns of A are sorted in a fixed order. If the submatrix A of a standard generator matrix of a self-dual code of length over is bisorted, we call G a bisorted generator matrix of .
Definition 3.
A matrix A is called symmetric if . If the submatrix A of a standard generator matrix of a self-dual code of length over is symmetric, we call G a symmetric generator matrix of .
Proposition 3.
Let and be generator matrices of self-dual codes and , respectively. If for some , then and are equivalent.
Proof.
For ,
Therefore, and are equivalent. □
This proposition implies the following corollary, which shows that the class of self-dual codes generated by bisorted generator matrices contain representatives of all inequivalent codes.
Corollary 3.
A self-dual code of length over is equivalent to a code with a bisorted generator matrix .
3. Self-Dual Codes of Length 6 over
It is well known that, for , a self-dual of length n over exists if and only if , and, for , a self-dual code of length n over exists if and only if .
Theorem 1.
Let
be a self-dual code of length 6 over . Then,
- (i)
- is decomposable if and only if A has at least two zero elements.
- (ii)
- is indecomposable and non-MDS if and only if A has exactly one zero element.
- (iii)
- is MDS if and only if A has no zero element.
Proof.
We are only to prove (i) and (iii), since (ii) is deduced directly from (i) and (iii).
Let the matrix for and . We know that ’s are solutions of the following simultaneous quadratic equations:
Since also generates a self-dual code, ’s are solutions of the following simultaneous quadratic equations as well:
Case (i). The ‘only if’ part follows from the definition of decomposable codes. For the ‘if’ part, suppose that there are two zero elements in A and these zero elements are in a row of A. We assume that without loss of generality. Then, by Equation (2), we know that is non-zero since
This implies that since
and
Thus, is decomposable. Suppose that the two zero elements are not in a row of A. Then, the two zero elements are in a column of A, or there is no row or column of A with the two zero elements. If two zero elements are in a column of A, say . By following the same argument as above with Equation (3), we have that . This implies that is decomposable. Lastly, suppose that there is no row or column of A with two zero elements. Without loss of generality, we can assume that . Then, Equation (2) implies that
This means that or , which is contradictory to the assumption. In addition, this proves the case (i).
Case (iii). The ‘only if’ part follows from the definition of the MDS code. For the ’if’ part, suppose that is not an MDS code, i.e., there is a non-zero codeword and . By Corollary 1, we know that not all ’s and not all ’s are zero. In addition, we recall that and are both generator matrices of by Lemma 2. We note that since is self-orthogonal. Assume that . Then, exactly one and one are non-zero for some where . Say and . Then, since the inner product of the second row of G and equals zero. Thus, A has at least one zero element. Now, assume that . There are two cases: exactly two of are zero and exactly one of is zero or vice versa. Assume that and . Then, since the inner product of the first row of and equals zero. Thus, A has at least one zero element. Other cases are proved similarly and this proves case (iii). □
This theorem does not hold for since there exist non-MDS self-dual codes of length 8 over with
Proposition 4.
Let
be a generator matrix of an indecomposable non-MDS self-dual code of length 6 over . Ignoring the signs, all the elements of the matrix G are determined by b and d.
Proof.
A matrix A is obtained from a solution of following simultaneous quadratic equations over :
We note that all variables are non-zero. The first equation determines c as and, for the fourth and the fifth equations, we get and . From the other three equations, we get
Since , we have that
Equating these equations, we have
Thus, g is determined by d as
Substitute with and with into the Equation (4), we get
Since and ,
and this proves the theorem. □
Corollary 4.
Let j be a square root of –1 and be an indecomposable non-MDS self-dual code of length 6 over . Then, is equivalent to one of the following:
where and .
Proof.
By Proposition 3 and Theorem 1, we know that is equivalent to a code with a generator matrix
by applying a suitable column and row permutations. By Proposition 4, it only needs to determine the signs of and i. The result follows easily by the direct calculation. □
We point out that and are equivalent since for . In addition, and are equivalent by Corollary 2. Moreover, , , and are all equivalent. We also note that we only need to investigate the bisorted generator matrices to classify self-dual codes of length 6 by Corollary 3.
Since every standard generator matrix of an indecomposable non-MDS self-dual code has exactly one zero element in A, we can assume that an indecomposable non-MDS self-dual code length 6 over has a generator matrix
where A is a bisorted matrix with the order of , regarding and g as integers. Hence, we assume that a bisorted generator matrix of a non-MDS self-dual code of length 6 over follows the order defined as above.
The number of solutions of plays an essential role in this paper. Thus, we give the number of solutions from [11] (Chapter 6) without a proof.
Lemma 1.
Let be a finite field with elements for an odd prime p. For non-zero , the cardinality of the set
is given by
In addition, for and for .
Proposition 5.
Let be a set of indecomposable non-MDS self-dual codes length 6 over with bisorted generator matrices in the form
Satisfying the order of , regarding and g as integers, then
Proof.
Since four solutions are in , there are elements with non-zero and . If for all , then there are choices of and , where and . It is well known that –2 is square if and only if . We recall that . Therefore, there exist four elements of in if . Exactly three among these four elements, , and one of and can be chosen as and satisfying the order condition. Thus, if , there are possible choices of and . Finally, we know that there exist two distinct matrices for one pair of and by the Corollary 4, and this completes the proof. □
Next, we describe the orbits of in under the -monomial transformation.
For a tuple in with non-zero b and c, we let
Since , b determines c and the set . Thus, we also define a set
for , where and g are non zero. Then, each element corresponds to the code and , which are equivalent to . If , , and , contains 128 elements and these elements correspond to 256 codes equivalent to . Moreover, for a tuple in with non-zero b and c, and , thus are also elements in . Thus, there exist nine sets:
Each of these sets has at most 128 elements corresponding to 256 codes equivalent to . This means that there exist at most codes equivalent to .
The equivalence of these codes can be checked easily by the direct computation. For example, the generator matix of is
and this shows that is equivalent to .
We note that the number of elements in and the number of corresponding codes equivalent to depend on the values of b and d. For instance, if and , then contains only 64 elements and ,, and hold.
The following proposition shows the correspondence between a set and elements in .
Proposition 6.
- (i)
- If and are all distinct, then corresponds to 64 equivalent codes with a bisorted generator matrix.
- (ii)
- If and , then corresponds to 48 equivalent codes with bisorted generator matrices.
- (iii)
- If and , then corresponds to at 32 equivalent codes with bisorted generator matrices.
- (iv)
- If and , then corresponds to 18 equivalent codes with bisorted generator matrices.
Proof.
(i) If , then there are four elements satisfying . If , then there are only three elements satisfying . Thus, it is easy to check that, if and are all distinct, there are 32 elements in satisfying and and this corresponds to the code and with bisorted matrices, which are equivalent to . (ii), (iii) and (iv) are proved similarly. □
4. Automorphism of Self-Dual Codes of Length 6 over
Due to the results in Section 3, we continue our arguments by discussing the automorphism group of with the following proposition.
Proposition 7.
Let be a self-dual code of length 6 over with a standard generator matrix G and let be the k-th column vector of G. If , then both the matrix and the matrix are non-singular.
Proof.
Let and assume that . Then, the matrix is row equivalent to G for some Thus, is row equivalent to for some and is row equivalent to A for some . Since we know that A is non-singular, this proposition holds. □
Let be an indecomposable non-MDS self-dual code of length 6 with a bisorted generator matrix , where . Proposition 7 says what elements of cannot be in . For example, the submatrix of the first three columns of is and this has rank 2. Thus, . In this manner, we check the rank of the submatrix of for every and we conclude that 72 elements in Table 1 cannot be in .
Table 1.
Seventy-two permutations that are not in by Proposition 7.
Furthermore, by the definition of automorphism of , it holds that the row canonical form of is equal to for some . Since , by comparing the element of the row canonical form of the matrix , we check each element whether it is included in or not. For example, it is easy to check that the row canonical form of the matrix is
Since is non-zero, is not row equivalent to G for all . Thus, . In this manner, we check all elements and find out that 576 elements in Table 2 cannot be included in . Consequently, we have the next proposition.
Table 2.
576 permutations which are not in .
Proposition 8.
Let be an indecomposable non-MDS self-dual code of length 6 over with the bisorted generator matrix. Then, .
Proof.
We check all 72 elements which can be in using the computer algebra system. We obtain the necessary condition for each element to be in . We summarize the results in Table 3.
Table 3.
Possible permutations to be in and their conditions.
5. Enumeration of Self-Dual Codes of Length 6 over
The number of self-dual codes of length n over is given by the following mass formula ([12], Chaper 19).
Theorem 2.
The number of self-dual codes of length n over for odd p is given by
Thus, there are self-dual codes of length 6 over for .
Example 1.
There exists a unique self-dual code of length 6 over ,
is decomposable and
There exist two inequivalent self-dual codes of length 6 over ; one is decomposable and the other is MDS:
We note that and .
The following proposition is deduced directly from Theorem 4.5 and Theorem 4.10 in [9].
Proposition 9
(Decomposable codes). Let and and j be a root of . Then, the decomposable self-dual code of length 6 over with generator matrix
is equivalent to one of the following four classes of inequivalent codes:
| Class | d | ||
| (i) | 8.48 | ||
| (ii) | 4.24 | ||
| (iii) | 4.16 | ||
| (iv) | 4.8 |
The code from class (i) is unique, if it exists, up to equivalence and has the weight enumerator
Codes from class (ii) and (iii) are unique, if they exist, up to equivalence and have the weight enumerator
Proof.
This proposition is deduced directly from Theorem 4.5 in [9] and the definition of a decomposable code. □
Proposition 10.
Let be the number of k-th class of decomposable self-dual codes of length 6 over . These numbers are determined as in the following table:
| p (Mod 24) | ||||
| 1 | 1 | 1 | 1 | |
| 5 | 1 | 0 | 0 | |
| 13 | 1 | 1 | 0 | |
| 17 | 1 | 0 | 1 |
Proof.
This proposition is deduced directly from Theorem 4.10 in [9]. □
Proposition 11
(indecomposable non-MDS codes). An indecomposable non-MDS self-dual code of length 6 with generator matrix
over is in the one of the following seven classes:
| Class | b, c, d, g | ||
| (i) | 2.18 | ||
| (ii) | 2.6 | ||
| (iii) | 2.3 | ||
| (iv) | 2.8 | ||
| (v) | 2.2 | ||
| (vi) | 2.2 | ||
| (vii) | 2.1 |
Codes from classes (i),(ii),(iv) are unique up to equivalence and all codes have the weight enumerator
Proof.
Clearly, for each code in all classes. The permutation part of each class is obtained directly from the conditions in Table 3. □
Proposition 12.
Let be the number of self-dual codes in equivalent to a code in the k-th class. These numbers are determined as in the following table:
| 32 | 112 | 192 | 98 | 336 | 288 | 576 |
Proof.
By Proposition 6, it is routine to count and sum the number of self-dual codes in corresponding to each of the nine sets: , , , , , , , , and
Three conditions , and in class (i) imply that and , thus . Since , it also holds that . Thus, . Futhermore, implies that . This means that
and, by Proposition 6, we know that corresponds to 32 equivalent codes in . This proves the case of class (i). All of the other cases are proved similarly. □
Theorem 3.
Let be the number of k-th class of self-dual codes of length 6 over . These numbers are determined as in the following table:
| p (Mod 24) | N1 | N2 | N3 | N4 | N5 | N6 | N7 |
| 1 | 1 | 1 | 1 | ||||
| 5 | 0 | 0 | 0 | 0 | 0 | ||
| 13 | 1 | 0 | 0 | 0 | |||
| 17 | 0 | 0 | 0 | 1 |
Proof.
We note that for all prime p. The existence of (i) and (iii) follows from the fact that is a multiplicative cyclic group of order . The existence of (ii), (iv) and (v) follows from the fact that is a quadratic residue in if and only if . For the uniqueness in (i), (ii) and (iv), it is easy to show by checking the equivalence of codes corresponding to the roots of or .
(iii) It is easy to show that the number of codes with bisorted generator matrix equivalent to , where is if and if , which is equal to the total number of distinct codes in classes (i), (ii) and (iii). Among them, there are 32 codes equivalent to the code in class (i) and there are 112 codes equivalent to the code in class (ii). Since , we have that
(v) To calculate , we get the number of codes equivalent to , which is equal to the total number of distinct codes in classes (ii), (iv) and (v). The result follows by a similar argument as in case (iii).
(vi) At first, we count the number of symmetric generator matrices. Each element corresponds to two symmetric generator matrices of and , thus there exist symmetric generator matrices. In class (i), if any, there are 16 symmetric generator matrices equivalent to the unique code of class (i). Similarly, in the class (iv), there are 16 symmetric generator matrices corresponding to and there are another eight symmetric generator matrices corresponding to . Since and in the class (vi), , and are all distinct. This implies that, for each code in the class (vi), there are 48 equivalent codes with symmetric generator matrices. Thus, the following equality holds:
(vii) By Proposition 12, we have that
and we obtain . This completes the proof. □
Corollary 5.
The number of MDS self-dual codes of length 6 over for an odd prime p is
Proof.
We compute the number of non-MDS self-dual codes of length 6 over using the mass formula,
where is a representative of each equivalent class.
- (i)
- For ,
- (ii)
- For ,
- (iii)
- For ,
- (iv)
- For ,
Finally, the number of MDS self-dual codes of length 6 over for every prime is obtained as
and this completes the proof. □
Funding
This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (NRF-2018R1A6A3A01013052).
Acknowledgments
The author wishes to thank the reviewers for valuable remarks that helped to improve this article.
Conflicts of Interest
The author declares no conflict of interest.
References
- Huffman, W.C. On the classification and enumeration of self-dual codes. Finite Fields Appl. 2005, 11, 451–490. [Google Scholar] [CrossRef]
- Pless, V. A classification of self-orthogonal codes over GF (2). Discret. Math. 1972, 3, 209–246. [Google Scholar] [CrossRef]
- Pless, V.; Sloane, N.J.A. On the classification and enumeration of self-dual codes. J. Combin. Theory Ser. A 1975, 18, 313–335. [Google Scholar] [CrossRef]
- Betsumiya, K.; Georgiou, S.; Gulliver, A.; Harada, M.; Koukouvinos, C. On self-dual codes over some prime fields. Discret. Math. 2003, 262, 37–58. [Google Scholar] [CrossRef]
- Leon, J.S.; Pless, V.; Sloane, N.J.A. Self-dual codes over GF (5). J. Combin. Theory Ser. A 1982, 32, 178–194. [Google Scholar] [CrossRef]
- MacWilliams, F.J.; Odlyzko, A.M.; Sloane, N.J.A.; Ward, H.N. Self-dual codes over GF (4). J. Combin. Theory Ser. A 1978, 25, 288–318. [Google Scholar] [CrossRef]
- Pless, V.; Sloane, N.J.A.; Ward, H. Ternary codes of minimum weight 6 and the classification of the self-dual codes of length 20. IEEE Trans. Inform. Theory 1980, 26, 305–316. [Google Scholar] [CrossRef]
- Pless, V.; Tonchev, V. Self-dual codes over GF (7). IEEE Trans. Inform. Theory 1987, 33, 723–727. [Google Scholar] [CrossRef]
- Park, Y.H. The classification of self-dual modular codes. Finite Fields Appl. 2011, 17, 442–460. [Google Scholar] [CrossRef][Green Version]
- Huffman, W.C.; Pless, V. Fundamentals of Error-Correcting Codes; Cambridge University Press: Cambridge, UK, 2010. [Google Scholar]
- Lidl, R.; Niederreiter, H. Finite Fields; Cambridge University Press: Cambridge, UK, 1997. [Google Scholar]
- Mac Williams, F.J.; Sloane, N.J.A. The Theory of Error-Correcting Codes; North Holland Publishing Co.: Amsterdam, The Netherlands, 1977. [Google Scholar]
© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).