Finite Fields and Their Applications 7, 449}467 (2001)

doi.10.1006/!ta.2000.0297, available online at http://www.idealibrary.com on

Codes over p-adic Numbers and Finite Rings Invariant

under the Full Affine Group
Kanat S. Abdukhalikov

Institute of Mathematics, Pushkin Str 125, Almaty 480100, Kazakhstan

E-mail: abdukhalikov@math.kz

Communicated by Vera Pless

Received December 21, 1998; revised April 7, 2000

Codes over p-adic numbers and over integers modulo pB of block length pK
invariant under the full a$ne group AG¸ (F ) are described.  2001 Academic
K N
Press

1. INTRODUCTION

Let p be a prime number. We consider codes of length pK whose permuta-

tion groups contain the a$ne group AG¸ (F ). They are classical generalized
K N
Reed}Muller codes if the alphabet is a "eld (of characteristic p) We study
codes in the case when the alphabet is not a "eld.
Let < be an elementary abelian group of pK elements and F be a ring. We
consider the group ring F[<] as the set of all formal linear combinations
a" a XT, where a 3F. Addition and scalar multiplication are compon-
T
TZ4 and T
ent-wise the multiplication is given by the addition in <:

a XT# b XT" (a #b )XT,

T T T T
c a XT" ca XT, c3F,
T T

( a XT) ) ( b XT)" a b XS>T" a b

T T
ST
S T
U S
S U\S
XU.


The element X is the unity of the ring F[<] and we write X"1. So F[<] is
a free module over F of rank pK and with basis +XT"v3<,. We consider
449
1071-5797/01 $35.00
Copyright  2001 by Academic Press
All rights of reproduction in any form reserved.
450 KANAT S. ABDUKHALIKOV

submodules of F[<] as codes of length pK based on the alphabet F. These

notations (for a "eld F) are introduced in [9}11]; see also [3].
Note that the elements of F[<] can be considered as functions from < to
F. As a function, an element a XT is the one that assigns a to the element
T T
v of <.
A$ne group G"AG¸(<)"< ) G¸(<) is a semidirect product of the
abelian group < and the group G¸(<), where < is considered as a vector
space over a "eld F of p elements. We make F[<] into a G-module by
N
putting

p (XT)"XS>ET, u3<, g3G¸(<).

SE
A code CLF[<] is said to be G-invariant if it is invariant under the action of
the group G (in [3] the group G is called the full a$ne group). If we "x a basis
+e ,2, e , of < over F then we can identify G¸(<) with the group G¸ (F )
 K N K N
of nondegenerated m;m-matrices.
Our goal is to study G-invariant codes in the cases F"Z (the ring of
N
p-adic integers) and F"Z/pBZ (the ring of integers modulo pB). Consider
modules

TZ4
T 
=" a XT " a 3Z ,
T N 
B
TZ4

= " a XT " a 3Z/pBZ .
T T 
The module = (resp. = ) is the ambient space for our G-invariant codes over
B
Z (resp. Z/pBZ). We have = :=/pB=, since Z /pBZ :Z/pBZ.
N B N N
Note that the invariance of a code CL= under the group < means that
C is an ideal in the group ring =. Thus one has to study ideals in = invariant
under the group G¸(<). Moreover, G-invariant codes over Z/pBZ can
be obtained from these universal codes (i.e., codes over Z ) by reduction
N
modulo pB.
At "rst we shall deal with invariant codes with zero sum, that is, codes in

M"+ a XT " a 3Z , a "0,.

T T N T
Theorem 3.4 gives a description of these codes, and Theorem 3.7 classi"es
G-invariant codes with zero sum for F"Z/pBZ.
G-invarient codes for the case F"F were considered in [3, 12]. There it
N
was shown that G-invariant codes are exactly generalized Reed}Muller codes
GRM(r, m), 04r4m(p!1), where codes with r(m(p!1) are codes with
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 451

zero sum. Generalized Reed}Muller codes are extended cyclic ones, so using
this fact one can construct (see [7, 8, 13, 15, 16]) lifted Reed}Muller codes
¸RM(r, m) over Z . Therefore one can construct an extended cyclic code
N
¸(r , r ,2, r )"¸RM(r , m)#p¸RM(r , m)#2
  K\  
#pK\¸RM(r , m)#pKM
K\
over Z , where 04r 4r 424r (m(p!1). We show that any
N   K\
G-invariant code with zero sum is equal to some code ¸(r , r ,2, r ) with
  K\
the following properties: (1) condition 0(r 4(m!1)(p!1)!1 implies
G
r 5r #(p!1); (2) condition r 5(m!1)(p!1) implies r "m(p!1)!1.
G> G G G>
Note that any such code contains pKM.
The description of codes with nonzero sum is a little more complicated and
is shown in Section 5 (Theorems 5.1 and 5.2). In Section 6 we consider the
case d"2 in detail. In particular, we compute dual codes. Duality is de"ned
with respect to the standard inner product

( a XT, b XT)" a b .
T T T T
If we consider < as the additive subgroup of a "nite "eld F K of pK elements
N
then < is a one-dimensional vector space over the "eld F K . De"ne the a$ne
N
group G "AG (F K )"< ) F*K over the "eld F K . It is a subgroup of G. Codes
  N N N
over Z and Z/pBZ invariant under this smaller group were classi"ed in [2].
N
Codes over a "nite "eld invariant with respect to the group G are studied in

[9}11, 14] in detail, and their automorphism groups are studied in [5, 6].
Our methods are similar to those used in [1] for investigation of invariant
integral lattices.

2. PRELIMINARY RESULTS

In this section we recall known results on the case F"F and prove
N
auxiliary lemmas that will help us to settle the general case F"Z .
N
For F"F the ambient space is
N
= "+ a XT " a 3F ,.
 T T N
Here subscript 1 means that we are considering codes over Z/pBZ for d"1,
and we identify F "Z/pZ. We set
N
M"+ a XT " a 3F , a "0,
T T N T
and denote by MI the ideal in = generated by elements “I (XTG!1). Note
 G
that M is a unique maximal ideal in = and it is nilpotent.

452 KANAT S. ABDUKHALIKOV

We collect here known results [3, 4, 12].

PROPOSITION 2.1. (1) ¹he only subspaces of = invariant under the group

G are powers of M and

= "MMMMM2MMKN\MMKN\>"+0,.


(2) For 04k4m(p!1) the quotient space ¸ "MI/MI> is an irredu-

I
cible G¸(<)-module over F . Furthermore, they are nonisomorphic, except for
N
¸ :¸ .
 KN\
(3) Elements (XC!1)I2(XCK!1)IK, 04k (p, 14i4m, form a basis
G
of the space = , and elements (XC!1)I2(XCK!1)IK, 04k (p, 14i4m,
 G
K k 5t, form a basis of the code MR.
G
G(4) (XCG!1)N"0.
(5) (MR)N"MKN\>\R, where CN denotes the code orthogonal to C.

Let us return to the general case. We denote by ML the ideal in = gener-

ated by elements

L
“ (XTG!1), v 3<,
G
G

and set M"M. Observe that ML is G-invariant.

LEMMA 2.2. ¹he following statements hold:

(1) (XT!1)#(XS!1),(XT>S!1) (mod M).
(2) ¹here exist integer numbers a , a ,2, a such that
  N\

(XT!1)N"!p(XT!1)!a p(XT!1)!2!a p(XT!1)N\.

 N\

(3) (XT!1)N3pM, (XT!1)N,!p(XT!1) (mod pM).

(4) If +e ,2, e , is a basis of < over F , then the elements
 K N

(XC!1)I2(XCK!1)IK, 04k 4p!1,

G

form a basis of = over Z .

N
Proof. (1) (XT!1)#(XS!1)"(XT>S!1)!(XT!1) (XS!1).
(2) In the ring of polynomials Z[t] we have

(t!1)N\"(tN\#2#t#1)!pf (t)
N
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 453

for some f (t)"1#a (t!1)#2#a (t!1)N\3Z[t], and

N  N\
(t!1)N"(tN!1)!p(t!1) f (t).
N
Therefore

(XT!1)N"!p(XT!1) f (XT).
N
(3) It follows from (2).
(4) Any element of = can be written as a polynomial in XC,2, XCK,
where the powers of XCG do not exceed p!1. Therefore, = is generated over
Z by the elements (XC!1)I2(XCK!1)IK, 04k 4p!1. 䊏
N G
The previous lemma suggests the de"nition of the following elements:

> "XCG!1, 14i4m.

G
LEMMA 2.3. ¹he following statements hold:
(1) ¹he elements >I2>IK, 04k 4p!1, 14i4m, form a basis of
 K G
= over Z .
N
(2) >N"!p> !a p>!2!a p>N\.
G G  G N\ G
(3) p> 3MN, >N,!p> (mod MN>).
G G G
(4) If element g3G¸(<) is represented by the matrix (a ) (that is,
GH
g(e )" K a e ) then p (> ), K a > (mod M).
H G GH G E H G GH G
Proof. (1) and (2) come from Lemma 2.2.
(3) From (2) we have

p> "!>N!a p>!2!a p>N\

G G  G N\ G
"!>N!a > [!>N!a p>!2!a p>N\]!2
G  G G  G N\ G
!a >N\ [!>N!a p>!2!a p>N\].
N\ G G  G N\ G
Continuing, we get the result.
(4) Part (1) of Lemma 2.2 implies

p (> )"X ?GHCG!1, a (XCG!1), a > (mod M). 䊏

E H GH GH G

3. CODES WITH ZERO SUM

In this section we are going to classify G-invariant codes with zero sum,
that is, codes in

M"+ a XT " a "0,.

T T
454 KANAT S. ABDUKHALIKOV

At "rst we are going to study the structure of the ideals MI. They will be basic
ingredients in the construction of G-invariant codes. We de"ne

NI"1>I2>IK " 04k (p, k 5k2Z .

 K G G N

Modules NI are modules over Z and the next lemma shows that they are not
N
invariant under the group G, except for M"N. In fact, MI"GNI.

LEMMA 3.1. If k"n(p!1)#s, 14s4p!1 then

MI"NI#pNI\N\#pNI\N\#2#pLNQ#pL>N.

Proof. Lemmas 2.3 and 2.2 imply that

MI"1>I2>IK " k 5k2Z .

 K G N

Hence, by Lemma 2.3(2) we have

MILNI#pNI\N\#pNI\N\#2#pLNQ#pL>N.

Conversely, one has MIMNI. Further, if k '0 and k "t5k!(p!1)

 G
then according to Lemma 2.3(3) we have

p>I>I2>IK,!>I>N\>I2>IK (mod MR>N).

  K   K
Hence MIMpNI\N\. Remaining inclusions can be proved similarly. 䊏

COROLLARY 3.2. If k"n(p!1)#s, 14s4p!1 then

MIMpMI\N\MpMI\N\M2MpLMQMpL>M. 䊏

Denote

w"(1!pK)X# XT.
TO

LEMMA 3.3. ¹he following statements hold:

(1) w3MKN\.
(2) A G-invariant code C in M, not contained in p=, contains the element w.
(3) A G-invariant code C in M, not contained in p=, contains pKM.
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 455

Proof. (1) Consider B"w#pK"w#pKX" XT. One has B ) > "0

T G
for any i. Let

B"b #b > #2#b >N\,

   N\ 
where b 3= is written as polynomial in > ,2,> (it does not depend on > ).
H  K 
Then by Lemma 2.2. we have

0"B ) > "b > #b >#2#b (!p> !a p>!2!a p>N\).

     N\    N\ 
Therefore,

b !pb "0,
 N\
b !pa b "0,
  N\
2

b !pa b "0.
N\ N\ N\
From here one has

B"(p#pa > #a p>#2#pa >N\#>N\)b .

    N\   N\
Further, let

b "c #c > #2#c >N\,

N\    N\ 
where c does not depend on > and > . Then we can repeat the previous
H  
reasoning using the equality B ) > "0. Hence, continuing this process, we get

K
B"e “ (p#pa > #a p>#2#pa >N\#>N\ ),
 G  G N\ G G
G
where e3Z . But the sum of the coe$cients of XT in B is equal to pK; thus
N
e"1. Therefore, w"B!pK3MKN\.
(2) For v3< let I -Z be the ideal of all a encountered in the
T N T
decomposition A" a XS of vectors A3C. Then I does not depend on v. If
S T
I "pQZ OZ then CLpQ=, a contradiction. Hence I "Z . Choose a vec-
T N N T N
tor B"X# O a XT3C. One has O a "!1, since B3M. Let
T  T T  T
g3G¸(<) be an element of order pK!1 permuting all nonzero vectors from
< cyclically. (This element can be constructed as follows: view a "nite "eld
456 KANAT S. ABDUKHALIKOV

F K as a vector space of dimension m over the "eld F and let u be a primitive

N N
element of F K; multiplication by u induces a required transformation of the
N
vector space <"F K.) Then
N


NK\
gG(B)"(pK!1)X# a XS"(pK!1)X! XS"!w.
T
G SO TO SO

(3) pK(XT!1)"!w ) (XT!1)3C and thus pKMLC. 䊏

Now we describe G-invariant codes in M. Consider m-tuples

(s , s ,2, s ) with property m(p!1)5s 5s 525s 51. First, if
  K\   K\
all s (m(p!1) then we de"ne
G

C(s , s ,2, s )"MQ#pMQ#2#pK\MQK\#pKM.

  K\

Further, if m(p!1)"s "s "2"s 's then we de"ne

  H\ H

C(s , s ,2, s )"Z ) w#pHMQH#2#pK\MQK\#pKM,

  K\ N

where the element w was de"ned before Lemma 3.3. It is clear that they are
G-invariant codes, but may not be di!erent; see Corollary 3.2.

THEOREM 3.4. A G-invariant code C in M not contained in p= is equal

to a code C(s , s ,2, s ) satisfying the following conditions:
  K\
(1) if m(p!1)'s 5p then s !(p!1)5s for i"0, 1,2, m!2;
G G G>
(2) if s 4p!1 then s "1 for i"0, 1,2, m!2.
G G>
We are going to prove the theorem in the next section. Conditions (1) and
(2) ensure that the codes C(s , s ,2, s ) are di!erent; see Lemma 3.1.
  K\
Observe also that

C(s , s ,2, s )" NQ# pNQ#2# pK\NQ#pKN

  K\
Q5Q Q5Q Q5QK\

if s Om(p!1), and


C(s , s ,2, s )"Z ) w# pHNQ# pH>NQ#2

  K\ N
Q5QH Q5QH>

# pK\NQ#pKN
Q5QK\
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 457

if s "2"s "m(p!1)'s . Furthermore,

 H\ H
(C(s , s ,2, s ) 5 pH=#pH>=)/pH>=:MQH:GRM(m(p!1)!s , m)
  K\ H
(1)
for j"0, 2, m!1, where GRM(r, m) is the rth order generalized Reed}
Muller code over F .
N
EXAMPLE 3.5. Let p"2 and m"4. Then G-invariant codes in M are
C(1, 1, 1, 1)"M, C(2, 1, 1, 1), C(3, 2, 1, 1), C(3, 1, 1, 1), C(4, 4, 4, 4), C(4, 4, 4, 3),
C(4, 4, 4, 2), C(4, 4, 4, 1), C(4, 4, 3, 2), C(4, 4, 3, 1), C(4, 4, 2, 1), C(4, 4, 1, 1),
C(4, 3, 2, 1), C(4, 3, 1, 1), C(4, 2, 1, 1), C(4, 1, 1, 1).
EXAMPLE 3.6. Let p"3 and m"2. Then G-invariant codes in M are
C(1, 1)"M, C(2, 1), C(3, 1), C(4, 4), C(4, 3), C(4, 2), C(4, 1).
Now we turn to codes over Z/pBZ. Let

M "+ a XT " a 3Z/pBZ, a "0,.

B T T T
Denote by MI and C (s , s ,2, s ) the images of MI and
B B   C\
C(s , s ,2, s ), respectively, under the natural homomorphism =P= ,
  K\ B
where e"min(m, d). Theorem 3.4 immediately implies
THEOREM 3.7. A G-invariant code C in M not contained in p= is equal to
B B
a code C (s , s ,2, s ), e"min(m, d), satisfying the following conditions:
B   C\
(1) if m(p!1)'s 5p then s !(p!1)5s for i"0, 1,2, e!2;
G G G>
(2) if s 4p!1 then s "1 for i"0, 1,2, e!2.
G G>
EXAMPLE 3.8. Let p"2, m"4, and d"2. Then G-invariant codes over
Z/4Z in M are C (1, 1)"M, C (2, 1), C (3, 2), C (3, 1), C (4, 4), C (4, 3),
B B B B B B B
C (4, 2), C (4, 1).
B B

4. PROOFS

In this section we will prove Theorem 3.4. We need several lemmas.

LEMMA 4.1. (1) If 0(k(m(p!1), k"n(p!1)#s, 14s4p!1, then
G¸(<)-module MI/MI> has a unique composition series

MI/MI>M(MI 5 pM#MI>)/MI>M2

M(MI 5 pLM#MI>)/MI>M+0,

with composition factors ¸ , ¸ , ,¸ .

I I\N\ 2 Q
458 KANAT S. ABDUKHALIKOV

(2) If k"m(p!1) then G¸(<)-module MI/MI> is a direct sum of two

submodules

MI/MI>"1w#MI>2 @ (MI 5 pM#MI>)/MI>,

where w"(1!pK)X# O XT. Furthermore, G¸ (<)-module (MI 5 pM#

T 
MI>)/MI> has a unique composition series

(MI 5 pM#MI>)/MI>M(MI 5 pM#MI>)/MI>M2

M(MI 5 pK\M#MI>)/MI>M+0,

with composition factors ¸ ,¸ , ,¸ .

K\N\ K\N\ 2 N\
Proof. (1) Denote Q "MI/MI>, Q "(MI 5 pM#MI>)/MI>,2,
 
Q "(MI 5 pLM#MI>)/MI>, Q "0. We recall that k"n(p!1)#s.
L L>
Lemma 3.1 and Proposition 2.1 imply

Q /Q :¸ , Q /Q :¸ , , Q /Q :¸ . (2)
  I   I\N\ 2 L L> Q
It is clear that Q MQ M2MQ MQ is a composition series. We have
  L L>
to prove that this series is unique.
Let Q be a submodule in Q and let Q contain an element A , j4n, such
 H
that A 3Q !Q . We will show that then one can "nd elements
H H H>
A ,2, A 3Q, such that A 3Q !Q . If we prove this it implies conse-
H> L G G G>
quently Q.Q , Q.Q ,2, Q.Q (see (2) and Lemma 2.1).
L L\ H
For simplicity we can assume that j"0, so A 3Q. Also we can assume

that A is congruent to >N\2>N\>Q modulo Q , since Q /Q is an
  L L>   
irreducible G¸(<)-module. Thus

Q
A ">N\2>N\>Q # b p>N\2>N\>Q\P>P #f #h ,
  L L> P  L\ L L>  
P
assuming here and later on in the proof that element f 3MI 5 pM is a sum of
G
monomials that do not contain elements of the form >N\2>N\>Q\P>P ,
 L\ L L>
and h 3Q .
G 
Let q be the element from G¸(<) such that q (> ),a> (mod M)
? ? L> L>
and q (> ),> (mod M) for iOn#1, a3F (see Lemma 2.3). Set
? G G N

B" a\Qq (A )
? 
?O
and consider two cases.
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 459

(a) Let sOp!1. Then

B"!>N\2>N\>Q !b p>N\2>N\>Q #f #h 3Q.

 L L> Q  L\ L>  
Take o 3G¸(<) such that o (> ),> #a> (mod M) and o (> ),>
? ? L L L> ? G G
(mod M) for iOn. Then

o (B)" (!>N\2(> #a> )N\>Q !b p>N\2>Q )

?  L L> L> Q  L>
?ZFN ?ZFN
# o ( f #h )
?  
?ZFN
"!p>N\2>N\>Q #f #h 3Q.
 L\ L>  
Here we have used Lemma 2.3(3). So we can assume that

D">N\2>N\>Q #d p>N\2>N\>Q #f #h 3Q
 L L> Q  L\ L>  
for some d O1. Furthermore, let o3G¸(<), o(> ),> #> (mod M)
Q L> L L>
and o(> ),> (mod M) for iOn#1. Then
G G


Q\ s
A "o(D)!D" (d !1)p>N\2>N\>Q\P>P #f #h 3Q !Q .
 r Q  L\ L L>    
P
(b) Let s"p!1. Then

B"!>N\2>N\>N\!b p>N\2>N\>N\
 L L> N\  L\ L>
!b p>N\2>N\>N\#f #h 3Q.
  L\ L  
De"ne o 3G¸(<) by o (> ),> #a> (mod M) and o (> ),>
? ? L L L> ? G G
(mod M) for iOn. Then

o (B)"(b !1)p>N\2>N\>N\#f #h 3Q.

?   L\ L>  
?ZFN
If b O1 then we can proceed as in step (a). Let b "1. Then we can also
 
assume that b "1, otherwise we may apply to B a linear map permuting
N\
> and > . Thus
L L>
B"!>N\2>N\>N\!p>N\2>N\>N\
 L L>  L\ L>
!p>N\2>N\>N\#f #h 3Q.
 L\ L  
460 KANAT S. ABDUKHALIKOV

We have n#1(m for kOm(p!1). Let h 3G¸(<), h (> ),

? ? L>
> #a> (mod M) and h (> ),> (mod M) for iOn#1. Then
L> L> ? G G

h (B)">N\2>N\>N\#p>N\2>N\>N\#f #h 3Q.
?  L L>  L\ L>  
?ZFN
Note that the element f does not contain monomials of the form

>N\2>N\>Q\P>P . Therefore, permuting > and > , we have
 L\ L L> L> L>

D">N\2>N\>N\#p>N\2>N\>N\#f #h 3Q.
 L L>  L\ L>  

Finally, we take

A #B#D"!p>N\2>N\>N\#f #h 3Q !Q .
  L\ L    

(2) According to Lemma 3.3 we have w3MKN\, w , pM. Furthermore,

w is invariant with respect to G¸(<). Thus we have the "rst assertion. The
second one can be proved as the Case (1). 䊏
LEMMA 4.2 ¸et K"K @ K be a decomposition of Z G-module K of
  N
,nite order into a direct sum of submodules and the submodules K , K have no
 
isomorphic composition factors. If NLK is a submodule then N"N @N ,
 
where N LK , N LK .
   
Proof. Let u : KPK be the projection homomorphism and let
G G
NMQ MQ M2MQ M+0, be a composition series. So Q /Q is an
L L\  G> G
irreducible module.
Consider the irreducible module Q . Let us prove that either Q -K or
  
Q -K . Indeed, if Q JK and Q JK then u (Q )O0 and u (Q )O0;
         
hence Q :u (Q ):u (Q ), a contradiction.
    
Suppose our assertion holds for Q , that is, let Q "S @ S , S -K ,
G G    
S -K . Consider the quotient module K/Q "K#K, K:K /S ,
  G  
K:K /S . By reasoning as in the previous step we have that Q /Q is
  G> G
a subset of either K or K. Therefore Q "R @ R , where R -K .
G>   G G
Induction gives the result. 䊏
Proof of the ¹heorem 3.4. According to Lemma 3.3 we have C.pKM. At
"rst, we note that for C.pM the assertion holds. Indeed, in this case one has
C/pM-M/pM, M/pM:M and by Proposition 2.1 any subcode in M is
isomorphic to MI, so C"C(k, 1,2, 1).
The idea of proof is, roughly speaking, the following. If
C U x" a 2 >I2>IK, and a 2 "pQb, p /b, k "t(m(p!1) for
I IK  K I IK G
some coe$cient a 2 , then C.pQMR. We will realize this formally.
I IK
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 461

Let (C#pM)/pM:MR, t4m(p!1). Denote D"MR 5 pM. Then we

have D"pMR\N\#pKM if t'p!1, and D"pM#pKM if t4p!1.
Let P(C) be the submodule in M, generated by elements a 2 >I2>IK,
I IK  K
such that there exists x" a 2 >I2>IK3C, p"a 2 . Then
I IK  K I IK
C-MR#P(C). Therefore,

(C#D)/D-MR/D @ (P(C)#D)/D.

Composition factors of MR/D are from the set +¸ , ¸ ,2, ¸ ,,

R R> KN\
and composition factors of (P(C)#D)/D are from the set +¸ , ¸ ,2, ¸ ,.
  R\
Thus composition factors of MRD and (P(C)#D)/D are not isomorphic;
hence by Lemma 4.2 we have (C#D)/D"N @ N , where N -MR/D and
  
N -(P(C)#D)/D. It is clear that N "MR/D; hence C#D.MR, C.MR.
 
Therefore C"E#N, where E"MR and pM.N.pKM (according to
Lemma 4.1, MR is the minimal G-module E with the property
(E#pM)/pM:MR).
In the case (C#pM)/pM:MR, t"m(p!1), we denote by E the G-
module generated by w. One has C.E by Lemma 3.3. Therefore again
C"E#N, where pM.N.pKM.
Furthermore denote N"pN, M.N.pK\M. We have reduced our
problem from the module C.pKM to the moldule N.pK\M. Arguing by
induction we have N"C(t ,2, t , 1). Then C"E#N"C(t, t!
 K\
(p!1), 2)#pC(t ,2, t , 1)"C(s ,2, s ) for some (s ,2, s ) 䊏
 K\  K\  K\
Now we are going to study connections with lifted Reed}Muller codes
¸RM(r, m). Consider < as the additive group of a "eld F K of pK elements. Let
N
a be a primitive element of F K and let T "+0, 1, m, m,2, mNK\, be the
N K
TeichmuK ller representatives of elements of <"+0, 1, a, a,2, aNK\,, where
m is the primitive (pK!1)th root of unity lying in some extension of the "eld
of p-adic numbers, and m is mapped into a under the reduction map modulo
p (see, for example, [2]). Thus the coordinates of an element of = can be
indexed by T , as well as the group <. An a$ne invariant extended cyclic
K
code over Z can be described as the set of all vectors (c , c , c ,2, c NK )3
K N   K K \
(Z )N with the property that the sum T c xI,0 (mod pG) for certain
N VZ K V
values of i and certain values of k, 04k4pK!1. In particular,
¸RM(m(p!1)!t, m) is the set of all vectors (c ) T such that
V VZ K

c xI"0
V
VZTK
for all k, 04k4pK!1, wt (k)(t (recall that if k" K\ k pH is the p-adic
N H H
expansion of k, then wt (k)" K\ k ).
N H H
Let C"C(s ,2, s ) be a G-invariant code with zero sum. Recall that
 K\
C is an extended cyclic code. Then the isomorphisms (1) imply that we can
462 KANAT S. ABDUKHALIKOV

express C as a linear combination of lifted Reed}Muller codes:

C"¸(r , r ,2, r )"¸RM(r , m)#p¸RM(r , m)#2

  K\  
#pK\¸RM(r , m)#pKM.
K\
Then (1) implies that

C(s ,2, s )"¸(m(p!1)!s ,2, m(p!1)!s ).

 K\  K\
The code C"C (s ,2, s ) consists of all vectors (c ) T such that
 K\ V VZ K
c "0
V
VZTK
and

c xI,0 (mod pG>) ∀k, wt (k)(s ,

V N G
VZTK
04i4m!1.
Also, it should be noted that, unlike the case of the "nite "eld,
MRO¸RM(m(p!1)!t, m) and NRO¸RM(m(p!1)!t, m).
Finally, let h denote the dimension of the generalized Reed}Muller code
G
GRM(m(p!1)!s , m) over the "eld F . Then the code C (s , s ,2, s )
G N B   C\
over Z/pBZ, e"min(m, d), has type

(h , h !h , h !h ,2, h !h )
     B\ B\
and size

pBF>B\F\F>2>FB\\FB\

for d4m. Similarly, it has type

(h , h !h , h !h ,2, h !h , pK!1!h )
     K\ K\ K\
and size

pBF>B\F\F>2>B\K>FK\\FK\>B\KNK\\FK\

for d'm. It follows from (1) and Theorem 3.7 (recall that
C (s , s ,2, s )-M and the rank of M is equal to pK!1).
B   C\ B B
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 463

5. G-INVARIANT CODES: THE GENERAL CASE

In this section we are going to classify G-invariant codes in =. At "rst we

introduce several codes in =. For "nite sequence (s ,2, s ) of positive
 K\
integers with property s "2"s "m(p!1), s (m(p!1) for some
 H\ H
j'0, positive integers r, t and element b3Z !pZ , pPbOpK, we de"ne:
N N

CPR@(m(p!1), s ,2, s )"Z ) (w#pPbX)#Z ) pRw

 K\ N N
#Z ) pP>RX#pHMQH#2#pK\MQK\,
N

and for code C(s ,2, s ) with t'0 we de"ne

 K\

DR(s ,2, s )"Z ) (w#pKX)#Z ) pK>RX#pRC(s ,2, s ).

 K\ N N  K\

Note that the repetition code Z ) ( xT)L= is G-invariant and Theorem

N
3.4 describes G-invariant codes in M.

THEOREM 5.1. ¸et C be a G-invariant code in = and CJp=, CJ=. ¹hen

one of the following assertions are true:
(1) C is the repetition code;
(2) C is a G-invariant code in M;
(3) C"C#pP=, where C is a G-invariant code in M;
(4) C"DR(s ,2, s );
 K\
(5) C"CPR@(m(p!1), s ,2, s ), pPbOpK, where the following holds:
 K\
(i) 14t4min(r, m);
(ii) if r(m then s "s "2"s "1;
P P> K\
(iii) if s 5p then s !(p!1)5s for i"j, j#1,2, m!2;
G G G>
(iv) if s 4p!1 then s "1 for i4m!2;
G G>
(v) t4j.

Proof. If we consider < as the additive group of a "eld F K of pK elements

N
then a primitive element of F K generates a cyclic subgroup ¹ of order pK!1
N
in G¸(<). G-invariant code in = will be invariant under the smaller group
< ) ¹. Codes invariant under < ) ¹ (i.e., a$ne invariant codes) were classi"ed
in [2], so we can use these results.

From [2], it follows that a G-invariant code is one of the following types:
(a) C is the repetition code;
(b) C is a G-invariant code in M;
(c) C"C#pP=, where C is a G-invariant code in M;
(d) C"C#C, where C is a code generated by vectors w#pPbX ,

464 KANAT S. ABDUKHALIKOV

pRw, pP>RX and C is a subcode in pM. Here we take for t the minimal
possible value. The "rst three cases correspond to assertions (1)}(3) of the
theorem; in the last case it is easy to see that either C"CPR@(m(p!1),
s ,2, s ) or C"DR(s ,2, s ). Furthermore, 5(iii)}5(v) follow from
 K\  K\
Lemma 3.1. Finally, the equality (w#pPbX)(XT!1)"(!pK#pPb)(XT!1)
follows CMpKM and CMpPM and thus p PKw3C; if t'min(r, m) then
pP> PKX3C, pP> PKw3C, a contradiction, so statements 5(i), 5(ii)
hold. 䊏

Let us consider G-invariant codes over Z/pBZ. At "rst we construct

codes by reducing CPR@(m(p!1), s ,2, s ) and DR(s ,2, s )
 K\  K\
modulo pBZ. Let e"min(m, d). For "nite sequence (s ,2, s ) of positive
 C\
integers with property s "2"s "m(p!1), s (m(p!1) for some
 H\ H
j'0, positive integers r(d, t4d!r and element b3Z/pBZ, p/b, pPbOpK,
we de"ne:

CPR@(m(p!1), s ,2, s )
B  C\

"Z/pBZ ) (w#pPb ) X)#Z/pBZ ) pRw#Z/pBZ ) pP>RX

#pHMQH#2#pC\MQC\.
B B

Note that the element b is de"ned modulo pB\PZ/pBZ. Furthermore, if d'm,

t(d then we de"ne as DR (s ,2, s ) the image of the module
B  D\
DR(s ,2, s ) under the natural homomorphism =P= , where
 K\ B
f"min(m, d!t).
The next theorem is a consequence of Theorem 5.1.

THEROEM 5.2. ¸et C be a G-invariant code in = and CJp= , CO= .

B B B
¹hen one of the following assertions are true:
(1) C is the repetition code;
(2) C is a G-invariant code in M ;
B
(3) C"C#pP= , where C is a G-invariant code in M , r(d;
B B
(4) C"DR (s ,2, s ), f"min(m, d!t);
B  D\
(5) C"CPR@(m(p!1), s ,2, s ), e"min(m, d), where the following
B  C\
holds:
(i) 14t4min(r, m);
(ii) if r(m then s "s "2"s "1;
P P> C\
(iii) if s 5p then s !(p!1)5s for i"j, j#1,2, e!2;
G G G>
(iv) if s 4p!1 then s "1 for i4e!2;
G G>
(v) t4j, r#t4d.
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 465

EXAMPLE 5.3. Let p"2, m"4, and d"2. Then a G-invariant code
CO= over Z/4Z in = not contained in p= is one of the following codes:
B B B
(1) the repetition code 1 XT2;
(2) a G-invariant code in M ;
B
(3) C (1, 1)#p= , C (2, 1)#p= , C (3, 1)#p= , C (4, 1)#p= ;
B B B B B B B B
(4) C(4, 1).
B

6. THE CASE d"2

In this section we assume that d"2, m52. Theorem 5.2 gives

COROLLARY 6.1. ¸et C be a G-invariant code in = and CJp= , CO= .
  
¹hen one of the following assertions are true:
(1) C"C (s , s ), p!1(s (m(p!1), 14s 4s !(p!1); C"C (s , 1),
       
14s 4p!1 or C"C (m(p!1), s ), 14s 4m(p!1);
   
(2) C"C (s , 1)#p= , where 14s 4m(p!1);
   
(3) C"C@(m(p!1), 1) b"1, 2, 2, p!1.

Here the code C (m(p!1), m(p!1)) is the repetition code. Recall that the

subscript in C"C (s , s ) means that we consider codes over Z/pBZ for
  
d"2.
Concerning the duality of codes we have the following
THEOREM 6.2. ¸et C be a G-invariant code in = . ¹hen the following

assertions hold:
(1) C (s , s )N"C (m(p!1)#1!s , m(p!1)#1!s );
     
(2) (C (s , 1)#p= )N"pC (m(p!1)#1!s , 1);
   N  
(3) C@(m(p!1), 1) "CN\@ (m(p!1), 1);
 
Proof. We will use the following simple property. If K MK are codes in
 
= then C(K /K )"C(KN/KN), where K /K , KN/KN are quotient rings
B        
and CK denotes the number of elements in K. Therefore, if we are given
K MK , KN then in order to "nd KN it is su$cient to "nd a code KLKN
    
with the properties C(KN/K)"C(K /K ) and (K, K )"0. Similarly for
   
codes K LK .
 
It is clear that (p= )N"p= . Hence, consequently for s "m(p!1),
  
m(p!1)!1, 2, 1, we have

(C (s , 1)#p= )N"(MQ#p= )N"pMKN\>\Q

   

"pC (m(p!1)#1!s , 1)
 
466 KANAT S. ABDUKHALIKOV

by Proposition 2.1 (5). Furthermore

C (1, 1)N"MN"1w2"C (m(p!1), m(p!1)).

 
Therefore for s "2, 3,2, m(p!1) we have consequently

C (s , 1)N"(MQ#pM)N"1w2#pMKN\>\Q
 
"C(m(p!1), m(p!1)#1!s ).

Similarly, for s "2, 3,2 we have

C (s , s )N"MKN\>\Q#pMKN\>\Q
  
"C(m(p!1)#1!s , m(p!1)#1!s ).
 
Finally,

C@(m(p!1), 1)N"(1w#pbX2#pM)N"1w!pbX2#pM

"CN\@(m(p!1), 1). 䊏

EXAMPLE 6.3. Let p"2, m"4, and d"2. Then self-dual G-invariant
codes in = are 2= , C (4, 1), C (3, 2) and C(4, 1).
B B B B B

ACKNOWLEDGMENTS

We thank the referees for their useful suggestions.

REFERENCES

1. K. S. Abdukhalikov, Integral lattices associated with a "nite a$ne group, Mat. Sb. 185,
No. 12 (1994), 3}18. English transl. in Russian Acad. Sci. Sb. Mat. 83, No. 2 (1995), 431}443.
2. K. S. Abdukhalikov, A$ne invariant and cyclic codes over p-adic numbers and "nite rings,
Des. Codes Cryptog. 23 (2001), 343}370.
3. E. F. Assmus and J. D. Key, Polynomial codes and "nite geometries, in &&Handbook of
Coding Theory'' (V. S. Pless and W. C. Hu!man, Eds.) Vol. 2, Chap. 16, pp. 1269}1343,
Elsevier, Amsterdam, 1998.
4. S. D. Berman, On the theory of group codes, Kibernetika 3 (1967), 31}39.
5. T. Berger and P. Charpin, The automorphism group of generalized Reed}Muller codes,
Discrete Math. 117 (1993), 1}17.
 CODES OVER p-ADIC NUMBERS AND FINITE RINGS 467

6. T. Berger and P. Charpin, The permutation group of a$ne-invariant extended cyclic codes,
IEEE ¹rans. Inform. ¹heory 42 (1996), 2194}2309.
7. A. R. Calderbank and N. J. A. Sloane, Modular and p-adic cyclic codes, Des. Codes Cryptogr.
6 (1995), 21}35.
8. A. R. Calderbank, W.-C. W. Li, and B. Poonen, A 2-adic approach to the analysis of cyclic
codes, IEEE ¹rans. Inform. ¹heory 43 (1997), 977}986.
9. P. Charpin, &&Codes cycliques eH tendus invariants sous le groupe a.ne,'' Thèse de Doctorat
d'ED tat, UniversiteH Paris VII, 1987.
10. P. Charpin, Une generalization de la construction de Berman des codes de Reed et Muller
p-aries, Comm. Algebra 16 (1988), 2231}2246.
11. P. Charpin. The extended Reed-Solomon codes considered as ideals of a modular algebra,
Ann. Discrete Math. 17 (1983), 171}176.
12. P. Delsarte, On cyclic codes that are invariant under the general linear group, IEEE ¹rans.
Inform. ¹heory 16 (1970), 760}769.
13. R. Hammons, P. V. Kumar, A. R. Calderbank, N. J. A. Sloane, and P. SoleH , The Z -linearity

of Kerdock, Preparata, Goethals, and related codes, IEEE ¹rans. Inform. ¹heory 40 (1994),
301}319.
14. T. Kasami, S. Lin, and W. W. Peterson, Some results on cyclic codes which are invariant
under the a$ne group and their applications, Inform. and Control 11 (1967), 475}496.
15. V. Pless, P. SoleH , and Z. Qian, Cyclic self-dual Z -codes, Finite Fields Appl. 3 (1997), 48}69.

16. V. Pless and Z. Qian, Cyclic codes and quadratic residue codes over Z , IEEE ¹rans. Inform.

¹heory 42 (1996), 1594}1600.