HAL Id: hal-02411123
https://hal.archives-ouvertes.fr/hal-02411123
Preprint submitted on 14 Dec 2019
HAL
is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire
HAL, estdestinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
A New Approach to the Kasami Codes of Type 2
Minjia Shi, Denis Krotov, Patrick Sole
To cite this version:
Minjia Shi, Denis Krotov, Patrick Sole. A New Approach to the Kasami Codes of Type 2. 2019.
�hal-02411123�
IEEE Proof
AUTHOR QUERIES
AUTHOR PLEASE ANSWER ALL QUERIES
PLEASE NOTE: We cannot accept new source files as corrections for your article. If possible, please annotate the PDF proof we have sent you with your corrections and upload it via the Author Gateway. Alternatively, you may send us your corrections in list format. You may also upload revised graphics via the Author Gateway.
AQ:1 = Please confirm or add details for any funding or financial support for the research of this article.
AQ:2 = Please confirm whether the edits made in the current affiliation of the author Patrick Solé are OK.
AQ:3 = Please confirm the postal code for Aix Marseille University, CNRS.
AQ:4 = Please provide the issue no. or month for Ref. [2].
AQ:5 = Please provide the publisher name and publisher location for Ref. [7].
AQ:6 = Please provide the publisher location for Refs. [8], [14], and [16].
IEEE TRANSACTIONS ON INFORMATION THEORY 1
A New Approach to the Kasami Codes of Type 2
Minjia Shi , Denis S. Krotov , Member, IEEE, and Patrick Solé
Abstract— The dual of the Kasami code of lengthq2−1, with
1
q a power of 2, is constructed by concatenating a cyclic MDS
2
code of length q + 1 over Fq with a Simplex code of length
3
q −1. This yields a new derivation of the weight distribution
4
of the Kasami code, a new description of its coset graph, and
5
a new proof that the Kasami code is completely regular. The
6
automorphism groups of the Kasami code and the relatedq-ary
7
MDS code are determined. New cyclic completely regular codes
8
over finite fields a power of 2, generalized Kasami codes, are
9
constructed; they have coset graphs isomorphic to that of the
10
Kasami codes. Another wide class of completely regular codes,
11
including additive codes, as well as unrestricted codes, is obtained
12
by combining cosets of the Kasami or generalized Kasami code.
13
Index Terms— Kasami codes, completely regular codes, cyclic
14
codes, concatenation, automorphism group.
15
I. INTRODUCTION
16
D
ISTANCE-REGULAR graphs form the most extensively17
studied class of structured graphs due to their many con-
18
nections with codes, designs, groups and orthogonal polyno-
19
mials [1], [6]. Since the times of Delsarte [7], a powerful way
20
to create distance-regular graphs, especially in low diameters,
21
has been to use the coset graph of completely regular codes.
22
A linear or additive code is completely regular if the weight
23
distribution of each coset solely depends on the weight of its
24
coset leader. Many such examples from Golay codes, Kasami
25
codes, and others, can be found in [6]. A recent survey is [5].
26
In this note we focus on the Kasami code of type (ii) in the
27
sense of [6, § 11.2]. This code was also studied in [3]. It is a
28
cyclic code of length 22m−1 with the two zeros α,α1+2m,
29
where α denotes a primitive root of F22m. We call such a
30
code a classical Kasami code. We give a new construction
31
by concatenating a cyclic MDS code with a binary Simplex
32
code. This gives a new derivation of its weight distribution,
33
a non trivial calculation in [13], and a characterization of its
34
automorphism group.
35
AQ:1 Manuscript received December 28, 2018; revised August 2, 2019; accepted October 14, 2019. This work was supported in part by the National Natural Science Foundation of China under Grant 61672036, in part by the Excellent Youth Foundation of Natural Science Foundation of Anhui Province under Grant 1808085J20, in part by the Academic Fund for Outstanding Talents in Universities under Grant gxbjZD03, and in part by the Program of Fundamental Scientific Researches of the Siberian Branch of the Russian Academy of Sciences under Project 0314-2019-0016.
M. Shi is with the Key Laboratory of Intelligent Computing and Signal Processing of Ministry of Education, School of Mathematical Sciences, Anhui University, Hefei 230601, China (e-mail: smjwcl.good@163.com).
D. S. Krotov is with the Sobolev Institute of Mathematics, 630090 Novosi- birsk, Russia (e-mail: krotov@math.nsc.ru).
AQ:2 P. Solé is with I2M, Centrale Marseille, Aix Marseille University, CNRS, AQ:3 13397 Marseille, France (e-mail: sole@enst.fr).
Communicated by V. Sidorenko, Associate Editor for Coding Theory.
Digital Object Identifier 10.1109/TIT.2019.2949609
More generally, we construct generalized Kasami codes 36 of lengths qp−12−1, when q is an even power of p, and p 37 itself a power of 2. This is obtained by concatenation of 38 a q-ary cyclic MDS code with a p-ary Simplex code. The 39 resulting code is cyclic with two zerosα,α(p−1)(1+q), where 40 α denotes a primitive root of Fq2. When p > 2, we obtain 41 in that way infinitely many new cyclic completely regular 42 codes with a coset graph isomorphic to the coset graph of a 43
Kasami code. 44
We emphasize that one of the main contributions of our 45 research is in applying the concatenation with the Simplex 46 code to cyclic codes, with the result being cyclic too. In 47 general, the concatenation approach in the construction of 48
linear completely regular codes is known, see [16], [17], [4], 49
[5]. In particular, in [4], the concatenation of dual-distance-3 50
MDS codes with the Simplex code was considered (note that a 51 distance-3 MDS code, even non-linear, is always completely 52 regular [12, Cor. 6]); the result is a two-weight q-ary code, 53 whose dual is a covering-radius-2 completely regular code. 54 We use a similar approach, but start with a cyclic dual- 55 distance-4 MDS code, and prove also the cyclicity of the 56 resultingp-ary (pis a power of two) completely regular code 57 of covering radius 3, which is known as the Kasami code 58 if p = 2, and is a new cyclic completely regular code if 59 p > 2. In fact, concatenation with the Simplex code can be 60 applied as well to produce a completely regular p-ary code 61 from a linear completely regular q-ary code, q = pk, with 62 the same intersection array, independently on the concrete 63 parameters. Moreover, even more general approach guarantees 64 that from any unrestricted (not necessarily linear or additive)q- 65 ary completely regular code (and even an equitable partition) 66 we can construct a completely regular code with the same 67 intersection matrix over any alphabetpsuch thatqis a power 68 of p (the alphabet sizes p and q are not required to be 69 prime powers). This approach is based on the existence of 70 a locally bijective homomorphism, known as a covering, see, 71 e.g., [8, Sect. 6.8], between the Hamming graphsH(n, q)(in 72 the role of the cover) and H(p−1q−1n, p) (in the role of the 73 target) of the same degree. With the inverse homomorphism, 74 an equitable partition or completely regular code of the target 75 graph automatically maps to an equitable partition or com- 76 pletely regular code, respectively, of the cover graph, with 77
the same intersection matrix. This combinatorial approach is 78
very powerful and allows to connect equitable partitions of 79
different graphs, especially translation graphs (Cayley graphs 80 over abelian groups), such as, for example, bilinear forms 81 graphs and Hamming graphs. However, it does not always 82 allow to track some nice algebraic properties of the codes 83
such as cyclicity. 84
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
The material is arranged as follows. The next section
85
collects the notions and notations needed for the rest of the
86
paper. Section III describes the concatenation approach and
87
the inner code used. Section IV gives a new description of the
88
classical Kasami codes, including their weight distributions,
89
and their automorphism groups. Section V describes the (new)
90
generalized Kasami codes obtained by allowing a more general
91
alphabet in the outer code. This includes a determination of
92
the weight distribution, and of the coset graphs. Section VI
93
provides a combinatorial way to derive new completely regular
94
codes from old, and applies it to the generalized Kasami codes.
95
Section VII recapitulates the path followed, and points out new
96
directions.
97
II. DEFINITIONS ANDNOTATION
98
A. Graphs
99
All graphs in this note are finite, undirected, connected,
100
without loops or multiple edges. In a graph Γ, the neighbor-
101
hoodΓ(¯v)of the vertexv¯is the set of vertices connected tov.¯
102
Thedegreeof a vertex¯vis the size ofΓ(¯v). A graph isregular
103
if every vertex has the same degree. Thei-neighborhoodΓi(¯v)
104
is the set of vertices at geodetic distance i tov. A graph is¯
105
distance regular(DR) if for every pair or vertices u¯andv¯at
106
distance iapart the quantities
107
bi=|Γi+1(¯u)∩Γ(¯v)|, ci=|Γi−1(¯u)∩Γ(¯v)|,
108
which are referred to as the intersection numbers of the
109
graph, solely depend on i and not on the special choice of
110
the pair (¯u,v). The¯ automorphism group of a graph is the
111
set of permutations of the vertices that preserve adjacency.
112
The Hamming graph H(n, q) (we consider only the case
113
of prime power q) is a distance-regular graph on Fnq, two
114
vectors being connected if they are at Hamming distance
115
one. It is well known [14] that any automorphism of the
116
Hamming graph acts as a permutation π of coordinates and,
117
for every coordinate, a permutation σi of Fq; i.e., σ◦ π:
118
(v1, . . . , vn)→(σ1(vπ−1(1)), . . . , σn(vπ−1(n))).
119
B. Equitable Partitions, Completely Regular Codes
120
A partition (P0, . . . , Pk) of the vertex set of a graph Γ is
121
called anequitable partition(also known as regular partition,
122
partition design, perfect coloring) if there are constants Sij
123
such that
124
|Γ(¯v)∩Pj|=Sij for everyv¯∈Pi, i, j= 0, . . . , k.
125
The numbers Sij are referred to as the intersection
126
numbers, and the matrix (Sij)ki,j=0 as the intersec-
127
tion (or quotient) matrix. If the intersection matrix
128
is tridiagonal, then P0 is called a completely regu-
129
lar code of covering radius k and intersection array
130
(S01, S12, . . . , Sk−1k;S10, S21, . . . , Sk k−1). That is to say,
131
a set of verticesCis a completely regular code if the distance
132
partition with respect to C is equitable. As was proven by
133
Neumaier [15], for a distance regular graph, this definition
134
is equivalent to the original Delsarte definition [7]: a code
135
C is completely regular if the outer distance distribution
136
(|C∩Γi(¯v)|)i=0,1,2,... ofCwith respect to a vertexv¯depends
137
only on the distance fromv¯toC.
138
C. Linear and Additive Codes 139
A linear code (that is, a linear subspace ofFnq) of lengthn, 140 dimensionk, minimum distanced over the fieldFq is called 141 an [n, k, d]q code. An[n, k, d]q code is called an MDS code 142 if d = n−k+ 1. The duality is understood with respect 143 to the standard inner product. Let Aw denote the number of 144
codewords of weightw. 145
Linear codes are partial cases of the additive codes, which 146 are, by definition, the subgroups of the additive group ofFnq. A 147 cosetof an additive codeCis any translate ofCby a constant 148 vector. A coset leader is any coset element that minimizes 149 the weight. Theweight of a coset is the weight of any of its 150 leaders. Thecoset graphΓC of an additive codeC is defined 151 on the cosets of C, two cosets being connected if they differ 152
by a coset of weight one. 153
Lemma 1 (see, e.g., [6]). An additive code with dis- 154 tance at least 3 is completely regular with intersec- 155 tion array {b0, . . . , bρ−1;c1, . . . , cρ} if and only if the 156 coset graph is distance-regular with intersection numbers 157 b0, . . . , bρ−1, c1, . . . , cρ. 158 The weight distribution of a code is displayed as 159
[0,1,· · ·,w, Aw,· · ·], where Aw denotes the number of 160
codewords of weightw. 161
D. Cyclic Codes 162
A cyclic codeC of lengthn overFq is, up to polynomial 163 representation of vectors, an ideal of the ringFq[x]/(xn−1). 164 All ideals are principal with a unique monic generator, called 165 the generator polynomialofC, and denoted by g(x). If this 166 polynomial hastirreducible factors overFq, then the codeC 167 is called a cyclic codewithtzeros. In that case, the dualC⊥ 168 admits a trace representation witht terms of the form 169
c(a1, . . . , at) = t
i=1
Tr(aiαji) n
j=1, 170
where 171
• ki is the degree of the factor indexediofg(x), 172
• Tr = Trqki/q is the trace fromFqki down toFq, 173
• ai is arbitrary inFqki, 174
• αi is a primitive root inF×qki, 175 In general the relative trace from Fqr to Fq, where q is a 176
power of a prime, is defined as 177
Trqr/q(z) =z+zq+· · ·+zqr−1. 178 If q is prime, then Trqr/q(z) is the absolute trace. See the 179 theory of the Mattson–Solomon polynomial in [13] for details 180
and a proof. 181
E. Automorphisms 182
The automorphism group Aut(C) of a q-ary code C of 183 lengthnis the stabilizer ofC in the automorphism group of 184 the Hamming graph H(n, q). Note that every automorphism 185 acts as the composition of a coordinate permutation and a 186 permutation of the alphabet in every coordinate. 187
SHIet al.: NEW APPROACH TO THE KASAMI CODES OF TYPE 2 3
We will say that a subgroupA of the automorphism group
188
of the Hamming graph H(n, q) acts imprimitively on the
189
coordinates if the set of coordinates can be partitioned into
190
subsets, called blocks, and denoted by N1, …, Nt, 1 <
191
t < n, such that the coordinate-permutation part of every
192
automorphism from A sends every block to a block. If there
193
is no such a partition, then the action of the group on the
194
coordinates is primitive (for example, this is the case for
195
Aut(C)of any cyclic code C of prime length).
196
We denote byAut0(C)the stabilizer of the all-zero word in
197
Aut(C). Themonomial automorphism group AutM(C) of a
198
q-ary codeC is defined as the set of allmonomial transforms
199
that leave the code wholly invariant [11]. A monomial trans-
200
form acting onFnq is the product of a permutation matrix by
201
an invertible diagonal matrix, both in Fn×nq . The subgroup
202
consisting of all such transforms with trivial diagonal part
203
is called the permutation part of the automorphism group.
204
The groupAutS(C)consists of thesemilinear automorphisms
205
of a q-ary code C, where each semilinear automorphism is
206
the composition τ ◦μ of a monomial transform μ and an
207
automorphismτ of the fieldFq (acting simultaneously on the
208
values of all coordinates) such that τ◦μ(C) =C. Note that
209
for any linear codeC, the groupsAutM(C)andAutM(C⊥)
210
are always isomorphic, as well as the groups AutS(C) and
211
AutS(C⊥); the same is not necessarily true forAut orAut0.
212
If q is not prime and Fq has a subfield Fp, we also use
213
the notation AutLp(C) to denote the subgroup of Aut0(C)
214
consisting of only transformations that are linear overFp. Any
215
automorphism from AutLp(C) acts as the composition of
216
a coordinate permutation and a Fp-linear permutation (from
217
GL(logpq, p)) of Fq in every coordinate. In particular, if q
218
is a power of 2, then we have AutM(C) ≤ AutS(C) ≤
219
AutL2(C) ≤ Aut0(C) ≤ Aut(C), and if q = 2, then the
220
first four of these groups are the same (the last group, for an
221
additive code, is the product of Aut0(C) and the group of
222
translations by codewords).
223
III. CONCATENATIONWITH THESIMPLEXCODE
224
Concatenation is the replacement of the symbols of the
225
codewords of some code, called the outer code, by the
226
codewords of some other code, called theinner code. The role
227
of the inner code in our study is played by a Simplex code;
228
moreover, we mainly focused on the case when this code is
229
cyclic.
230
Let p be a prime power; let q = pk. Let ξ be a root of
231
unity of order q−1p−1 in Fq. We assume that k andp−1 are
232
coprime; hence, so are q−1p−1 andp−1 (indeed, by induction,
233
pk−1
p−1 ≡kmodp−1). It follows that
234
F×q =
bξl|b∈F×p, l∈ {1, . . . ,q−1p−1}
(1)
235
and, in particular, all powers ofξare mutually linear indepen-
236
dent overFp. Then, the matrix(ξ1, . . . , ξq−1p−1), considered as
237
ak×q−1p−1 matrix overFp, is a generator matrix of a Simplex
238
code of sizeq, where the distance between any two different
239
codewords is pk−1 [11, Th. 2.7.5].
240
Encoding the Simplex code. Let L be a nondegenerate 241 Fp-linear function fromFq toFp. Any such function can be 242 represented asL(·)≡Trq/p(a·) for somea∈Fq, and in our 243
context we can always assume that L is Trq/p. Define φ : 244
Fq →Fpq−1p−1 by 245
φ(z) =
L(zξ1), . . . , L(zξq−1p−1)
; 246
expand its action toFnq coordinatewise, and to any set of words 247
inFnq wordwise. 248
Lemma 2. For any two words y,¯ z¯ in Fnq, we have 249 d(φ(¯y), φ(¯z)) =pk−1d(¯y,z). I.e.,¯ φis a scaled isometry. 250 Proof: Immediately from the fact that the Simplex code 251 is a one-weight code with nonzero weightpk−1, we have 252 d(φ(y1, . . . , yn), φ(z1, . . . , zn)) 253
=d
(φ(y1), . . . , φ(yn)),(φ(z1), . . . , φ(zn))
254
= n i=1
d(φ(yi), φ(yi)) = n i=1
pk−1d(yi, yi) 255
=pk−1d((y1, . . . , yn),(z1, . . . , zn)). 256
257
Lemma 3. Assume that M = (Mij)mi=1nj=1 is an m×n 258 generator matrix of a codeC inFnq. Then the matrix 259
Φ(M) =
⎛
⎜⎝
M11ξ1 ... M11ξq−1p−1 M12ξ1 . . . M1nξp−1q−1 ... ... ... ... . . . ...
Mm1ξ1 ... Mm1ξq−1p−1 Mm2ξ1 . . . Mmnξq−1p−1
⎞
⎟⎠, 260
considered as a km× q−1p−1n matrix over Fp, is a generator 261 matrix of the codeφ(C) inFpp−1q−1n. 262 Proof: The proof is straightforward. Any codeword ¯c of 263 the code generated by Φ(M) is a linear combination of its 264
rows, and so is of the form 265
¯ c=
m i=1
Li(Mi1ξ1), . . . , Li(Minξq−1p−1)
266
for some Fp-linear functions Li : Fq → Fp. SinceLi(·) ≡ 267 L(ai·)for someai∈Fq, we have 268
¯ c =
m i=1
L(aiMi1ξ1), . . . , L(aiMinξp−1q−1)
269
= m i=1
φ(aiMi1), . . . , φ(aiMin)
270
= m i=1
φ(aiMi1, . . . , aiMin)
271
= φ m
i=1
ai(Mi1, . . . , Min)
. 272
That is,c¯is aφ-image of some codeword of C. 273 The following lemma provides an alternative explanation 274 to the fact that completely regular codes with identical coset 275 graphs can exist over different alphabets [17]. 276
Lemma 4. The coset graph ΓC⊥ of the dualC⊥ of a linear
277
codeC inFnq and the coset graphΓφ(C)⊥ of the dualφ(C)⊥
278
of its imageφ(C) inFpp−1q−1n are isomorphic.
279
Proof: We first observe that the number of cosets ofC⊥
280
coincides with size of the codeC, which is the same as the size
281
of the codeφ(C), which is, in its turn, the number of cosets of
282
φ(C)⊥. That is, the considered graphs have the same number
283
of vertices.
284
LetCbe a code overFq with a generator matrixM, which
285
is a check matrix of the dual code C⊥. The cosets of C⊥
286
naturally correspond to the syndromes under the action of
287
the check matrix M. By the definition of the coset graph
288
ΓC⊥, two cosets are adjacent in the coset graph if and only
289
if the difference between the corresponding syndromes is the
290
syndrome of a weight-1 word. We call the syndromes of the
291
weight-1 words the connecting syndromes because they play
292
a connecting role for the coset graph (this terminology is
293
in agree with that for a more general class of graphs called
294
Cayley graphs). For the check matrixM = (Mij)mi=1nj=1 over
295
Fq, the connecting syndromes are
296
a(M1j, . . . , Mmj)T, j∈ {1, . . . , n}, a∈F×q (2)
297
(corresponding to the weight-1 words with a in the j-th
298
position).
299
Now, consider the matrixΦ(M), which, by Lemma 3, is a
300
generator matrix forφ(C)and a check matrix forφ(C)⊥. The
301
connecting syndromes are
302
b(ξlM1j, . . . , ξlMmj)T, (3)
303
j∈ {1, . . . , n}, l∈
1, . . . ,q−1 p−1
, b∈F×p,
304
corresponding to the weight-1 words with b in the position
305
(j −1)p−1q−1 +l. Since {bξl | b ∈ F×p, l ∈ {1, . . . ,q−1p−1}}
306
coincides withF×q (see (1)), we find that the sets of connecting
307
syndromes (2), (3) for both matrices coincide; call this setS.
308
Two cosets of C⊥ are adjacent in ΓC⊥ if and only if
309
the corresponding syndromes have the difference in S. Two
310
cosets of φ(C)⊥ are adjacent in Γφ(C)⊥ if and only if
311
the corresponding syndromes have the difference in S. So,
312
indexing the cosets by the syndromes induces an isomorphism
313
between the ΓC⊥ andΓφ(C)⊥.
314
Corollary 1. If the code C⊥ is completely regular, then the
315
code φ(C)⊥ is completely regular with the same intersection
316
array.
317
In the rest of this section, we establish important relations
318
between the automorphism groups of a code and its concate-
319
nated image.
320
A. Automorphism Group
321
Lemma 5. The groupAutLp(C)is isomorphic to a subgroup
322
ofAutM(φ(C)) such that the action of this subgroup on the
323
coordinates is imprimitive with the blocks
324
1, . . . ,q−1 p−1
, . . . ,
q−1
p−1(n−1)+1, . . . ,q−1 p−1n
.
325
Proof: The action of the monomial automorphism group 326 on a codewordcofC can be represented as a permutation of 327 the coordinates and, for every ifrom 1 tonthe action of an 328
element of GL(m, p) on of the value in the i-th coordinate, 329
understood as an element ofFmp . We first consider these two 330
actions separately. 331
(i) A permutation of coordinates corresponds to a permuta- 332
tion of blocks forφ(C): we see that 333
φ(zπ−1(1), . . . , zπ−n(n)) = (φ(zπ−1(1)), . . . , φ(zπ−n(n))) 334 is obtained fromφ(z1, . . . , zn)by sending each i-th block of 335 coordinates to the place of theφ(i)-th block. 336 (ii) Next, we show that the action of aFp-linear permutation 337 on the value zi in the i-th coordinate corresponds to a 338 monomial transform inside the i-th block of φ(z1, . . . , zn). 339 Here, a crucial observation if the following well-known fact. 340 (*) Any bijective linear transform of the Simplex code 341 is equivalent to the action of some of its monomial auto- 342 morphisms. This fact is straightforward from the form of a 343 generator matrix of the Simplex code, which consists of a 344 complete collection of pair-wise linearly independent columns 345 of heightm. Applying a bijective linear transform to the rows, 346 we obtain another generator matrix with the same property. It 347 is obvious that one such matrix can be obtained from another 348 one by permuting columns and multiplying each column by a 349 nonzero coefficient, that is, by a monomial transform. 350 Consider the value zi in the i-th position and some per- 351 mutation τi of Fq that is linear over Fp. Since φ is also 352 linear over Fp and invertible, we have a bijective linear 353 transform φ(τi(φ−1(·))) of the Simplex code. By (*), there 354 is a monomial automorphismσi of the Simplex code whose 355 action on this code is identical toφ(τi(φ−1(·))). Then, we have 356 φ(τi(zi)) =σi(φ(zi)), 357 i.e., the action of a Fp-linear permutation τi in the i-th 358 coordinate of a word z¯ from Fnq corresponds to the action 359 of the monomial transformσi in thei-th block of coordinates 360
ofφ(¯z). 361
Summarizing (i) and (ii), we get the following. If the 362
mapping 363
(z1, . . . , zn)→
τi(zπ−1(1)), . . . , τi(zπ−n(n))
364
belongs to AutLp(C), then 365
(φ(z1), . . . , φ(zn))→
σi(φ(zπ−1(1))), . . . , σi(φ(zπ−n(n)))
366
mapsφ(C)to itself, i.e., belongs toAutM(φ(C)). This proves 367
the statement. 368
Lemma 6. Assume that C is a linear code overFq andC⊥ 369
satisfies the following property: 370
(i) the minimum distance of C⊥ is at least4. 371
Then the groups AutLp(C), AutM(φ(C)), and 372
AutM(φ(C)⊥)are isomorphic. 373
Proof: We will first deduce from (i) that 374 (ii) Aut0(φ(C)⊥) (and hence, also AutM(φ(C)⊥)) acts 375 imprimitively on the coordinates with the blocks 376
1, . . . ,q−1 p−1
, . . . ,
q−1
p−1(n−1)+1, . . . , q−1 p−1n
. 377
SHIet al.: NEW APPROACH TO THE KASAMI CODES OF TYPE 2 5
(As we see from Lemma 5, this property is indeed necessary,
378
while (i) is only sufficient to guarantee (ii).) To show (ii),
379
we note two other facts. From the code distance ≥4 ofC⊥,
380
we see that
381
(iii) for every codeword of weight 3 in φ(C)⊥, the three
382
nonzero coordinates belong to the same block. Indeed,
383
assume that φ(C)⊥ has a codeword with exactly three
384
nonzero coordinates, with valuesβ,β andβ respec-
385
tively. This means that for somei,i,i,j,j,j such
386
that(i, j)= (i, j)= (i, j)= (i, j), the equation
387
βL(ziξj) +βL(ziξj) +βL(ziξj) = 0
388
is satisfied for every (z1, . . . , zn) from C. Since L is
389
linear and nondegenerate, this implies
390
(βξj)zi+ (βξj)zi+ (βξj)zi= 0
391
for all(z1, . . . , zn)fromC. The last equation contradicts
392
to the code distance ofC⊥, unless i=i=i, i.e., the
393
three coordinates are from the same block.
394
On the other hand, because the dual of a Simplex code is a
395
1-perfect Hamming code, we have
396
(iv) every two coordinates from the same block are nonzero
397
coordinates for some weight-3 codeword in φ(C)⊥.
398
Indeed, in the generator matrix of the Simplex code,
399
the sum of any two columns is proportional to some
400
other column. The same relation takes place between the
401
corresponding coordinates from the same block of the
402
concatenated code, which implies the required property.
403
From (iii) and (iv), we see that the code properties of φ(C)⊥
404
distinguish the pairs of coordinates in the same block and the
405
pairs of coordinates in different blocks. Hence, the automor-
406
phism group does not break blocks and we have (ii).
407
Next, we observe that property (i) excludes some degenerate
408
cases with vanishing coordinates:
409
(v) every codeword of the Simplex code occurs in every
410
block, over the codewords of the concatenated code
411
φ(C).Indeed, this obviously equivalent to the fact that
412
for every symbol of Fq and every position, there is
413
a codeword of C with the given symbol in the given
414
position. SinceC is linear, it is the same as to say that
415
there is no vanishing coordinate inC. Equivalently,C⊥
416
does not contain a weight-1 codeword, which follows
417
immediately from (i).
418
Now, we have that every automorphism from
419
AutM(φ(C)⊥) permutes blocks, permutes coordinates
420
inside each block, and permutes the values of Fp for every
421
coordinate (fixing 0). The same is true for AutM(φ(C)),
422
because it is isomorphic to AutM(φ(C)⊥) with the
423
isomorphism keeping the permutation part. According to
424
(v), the last two steps (permuting coordinates inside each
425
block, and permuting the values of Fp for every coordinate)
426
must be compatible with the automorphism group of the
427
Simplex code, and result in a permutation of its codewords.
428
A permutation of the codewords of the Simplex code in each
429
block, for the codeφ(C), corresponds to a permutation of the
430
elements of Fq in the corresponding coordinate for C; and
431
a permutation of the blocks corresponds to a permutation of
432
the coordinates for C. So, we see that every automorphism 433 fromAutM(φ(C))corresponds to some automorphism ofC: 434
AutM(φ(C))Aut0(C). 435
Moreover, because all considered transformations are linear 436
overFp, we have 437
AutM(φ(C))AutLp(C). 438 From Lemma 5, we have the inverse correspondence. It 439 remains to note that AutM of a code and its dual are 440
isomorphic. 441
IV. CLASSICALKASAMICODES 442
Let q = 2m for some integerm≥ 1. Consider the cyclic 443 code Mq⊥ of length q+ 1 defined over Fq by the check 444 polynomial h(x) = (x−1)(x−ζ)(x−ζq), for the root 445 ζ = αq−1 of order q+ 1 over Fq2, where α is a primitive 446 root of Fq2 (for some notation reasons, we first define Mq⊥ 447 and thenMq as the dual toMq⊥). Note thatζq2=ζ, so that 448 (x−ζ)(x−ζq) =x2−(ζ+ζq)x+ζq+1 =x2−(Trq2/qζ)x+1 449
has its coefficients inFq. 450
Theorem 1. The codeMq⊥is MDS of parameters[q+1,3, q− 451
1]q. Its weight distribution is 452
0,1 ,
q−1,q3−q 2
,
q, q2−1 ,
q+1,q3−2q2+q 2
. 453 (4) 454
Proof: The BCH bound (see e.g. [13, § 7.6]) applied to 455 Mq shows that Mq, hence Mq⊥ is MDS. The frequency of 456 weightq−1is derived on applying [13, Corollary 5, p. 320]. 457
Aq−1= (q−1) q+ 1
q−1
= (q−1) q+ 1
2
. 458
The frequency of weightqis derived on applying [13, p. 320]. 459
Aq= q+ 1
3−2
(q2−1)−q(q−1)
= (q+ 1)(q−1). 460
The frequency of weightq+ 1is derived by complementation 461
Aq+1=q3−1−Aq−1−Aq. 462
Denote by Sq the binary Simplex code of parameters 463 [q−1, m,q2]2. The main result of this section is the following 464
theorem. 465
Theorem 2. The concatenation of Mq⊥ with Sq is a binary 466 cyclic codeKq⊥ of parameters[q2−1,3m], with nonzerosα, 467 α1+q, withαa primitive root ofFq2. 468 Proof: As ζ=αq−1 is a root of orderq+ 1of Fq2, by 469
the facts of Section 2.4, we have 470
Mq⊥ = c+ Trq2/q(δζj)q+1
j=1 c∈Fq, δ∈Fq2
. 471
Let θ=α1+q. Clearly θ is a root of orderq−1 ofFq. The 472
concatenation mapφfromFq toSq is conveniently described 473
by using the trace functionTrq/2. For allβ∈Fq let 474 φ(β) =
Trq/2(βθ),Trq/2(βθ2), . . . ,Trq/2(βθq−1)
475
=
Trq/2(βθi)q−1
i=1. 476