A characteristic property of self-orthogonal

codes and its application to lattices

Zhe-Xian Wan

Abstract
2πi/p
Let p be an odd prime, ζ = e , D be the ring of algebraic integers in
the field Q(ζ), and P = (1 − ζ) be the principal ideal of D generated by 1 − ζ.
For a p-ary linear code C of length n, define the lattice ΛC = {p−1/2(c + z) |
c ∈ C, z ∈ P n }. It is proved that ΛC is even if and only if C is self-orthogonal
and that ΛC is even unimodular if and only if C is self-dual. The proof rests
on the following remark that for an odd prime power q a q-ary linear code C
is self-orthogonal if and only if c · c = 0 for all c ∈ C. Finally, irreducible root
lattices arising as ΛC from p-ary linear codes C are completely determined.

1 Introduction
Let q be a prime power, n be a positive integer, and Fqn be the n-dimensional row
vector space over the finite field Fq with q elements. A k-dimensional subspace of Fqn
is called a q-ary linear [n, k]-code. For any x = (x1 , . . . , xn ), y = (y1 , . . . , yn ) ∈ Fqn ,
define

x · y = x1 y1 + · · · + xn yn .

Let C be a q-ary linear [n, k]-code. Define

C ⊥ = {x ∈ Fqn | x · y = 0 for all y ∈ C}.

Then C ⊥ is an (n − k)-dimensional subspace of Fqn and called the dual code of C. If

C ⊆ C ⊥ , then C is called self-orthogonal. If C = C ⊥ , then C is called self-dual.
Received by the editors August 1997.
Communicated by James Hirschfeld.
1991 Mathematics Subject Classification. 05, 94.
Key words and phrases. self-orthogonal code, cyclotomic field, lattice, irreducible root lattice.

Bull. Belg. Math. Soc. 5 (1998), 477–482

478 Z.-X. Wan

In the present paper it is remarked that when q is a power of an odd prime, a

q-ary linear code C is self-orthogonal if and only if c · c = 0 for all c ∈ C. Then this
remark is applied to the study of lattices.
Let p be an odd prime, ζ = e2πi/p, Q(ζ) be the cyclotomic field of pth roots of
unity, D be its ring of algebraic integers, and P = (1 − ζ) be the principal ideal of
D generated by 1 − ζ.
For a p-ary linear [n, k]-code C, define the lattice
ΛC = {p−1/2 (c + z) | c ∈ C, z ∈ P n },
where c is regarded as a vector whose components are integers 0, 1, . . . , p − 1. Then
it is proved that ΛC is even if and only if C is self-orthogonal and that ΛC is even
unimodular if and only if C is self-dual. This improves a proposition of [1].
Finally, let Λ be an irreducible root lattice in Rn . Then it is proved that Λ ' ΛC
for a p-ary code C of length n, where p is an odd prime, if and only if Λ is of type
Ap−1, E6 (when p = 3 and n = 3), or E8 (when p = 3 and n = 4, or p = 5 and
n = 2).

2 A characteristic property of self-orthogonal codes

Proposition 1
Let q be a power of an odd prime and C be a q-ary linear code. Then C is self-
orthogonal if and only if c · c = 0 for all c ∈ C.
Proof. Assume that c · c = 0 for all c ∈ C. For any c, c0 ∈ C, since C is linear,
c + c0 ∈ C. Then
c · c = c0 · c0 = (c + c0) · (c + c0 ) = 0,
which implies 2c · c0 = 0. Since q is odd, we have c · c0 = 0 for all c, c0 ∈ C. Therefore
C ⊆ C ⊥.
The converse part is trivial. 

Proposition 1 should be known, but the author could not find a reference, so let
its proof be here.
The following example shows that Proposition 1 does not always hold when q is
even.

Example

Let
C = {(0, 0, 0), (1, 1, 0), (1, 0, 1), (0, 1, 1)} ⊆ F23 .
Clearly, C is a binary linear [3, 2]-code with the property that c · c = 0 for all c ∈ C,
but C 6⊆ C ⊥ .
For the following proposition, see, for example, [1], p. 9 or [4], p. 26.
Proposition 2
Let q be a prime power and C be a q-ary linear [n, k]-code. Then C is self-dual if
and only if n is even, k = n/2, and C ⊆ C ⊥ .
A characteristic property of self-orthogonal codes 479

3 Application to lattices
Let p be an odd prime, ζ = e2πi/p, Q(ζ) be the cyclotomic field of pth roots of unity,
and D be its ring of algebraic integers. It may be shown (see, for example, [3],
Chapter 13, §2) that

Q(ζ) = Q + Qζ + · · · + Qζ p−2 ,
D = Z + Zζ + · · · + Zζ p−2 ,

where both sums are direct. Also the ideal P = (1− ζ) is a prime ideal of D, P̄ = P ,
and D/P ' Fp . Define a bilinear form on Q(ζ) by

(x, y) = TrQ(ζ)/Q(xȳ) for all x, y ∈ Q(ζ),

where ȳ denotes the complex conjugate of y. It was proved (see, for example, [1],
§5.1 or [2]) that it is a positive definite symmetric bilinear form on Q(ζ), that D
is a (p − 1)-dimensional lattice with disc D = pp−2 , and that p−1/2 P is a (p − 1)-
dimensional lattice of type Ap−1.
Let n be an integer ≥ 2,

Q(ζ)n = {x = (x1 , . . . , xn ) | xi ∈ Q(ζ)}

and define
X
n
(x, y) = (xi , yi ) for all x, y ∈ Q(ζ).
i=1

Then (x, y) is a positive definite symmetric bilinear form on Q(ζ)n . Moreover, Dn

is an n(p− 1)-dimensional lattice with disc Dn = pn(p−2) , and p−1/2 P n is an n(p− 1)-
dimensional lattice of type nAp−1.
Define a map ρ : Dn → (D/P )n ' Fpn by

ρ(x1 , . . . , xn ) = (x1 + P, . . . , xn + P ) for all (x1, . . . , xn ) ∈ Dn .

Clearly, ρ is a surjective homomorphism of additive groups. Let C be a p-ary linear

[n, k]-code. Define

ΛC = p−1/2 ρ−1 (C)

= {p−1/2 (c + z) | c ∈ C, z ∈ P n },

where c is regarded as a vector whose components are integers 0, 1, . . . , p − 1. Then

we have
Proposition 3
Let p be an odd prime and C be a p-ary linear [n, k]-code. Then ΛC is an n(p − 1)-
dimensional lattice containing the lattice p−1/2 P n of type nAp−1 and with disc ΛC =
pn−2k . Moreover,
(i) ΛC is even if and only if C is self-orthogonal.
(ii) ΛC is even unimodular if and only if C is self-dual.
480 Z.-X. Wan

Proof. We have | Fpn /C |= pn−k . By the 2nd isomorphism theorem (see [5], p. 150)

Dn /ρ−1 (C) ' Fpn /C.

Therefore | Dn /ρ−1 (C) |= pn−k . Since Dn is an n(p − 1)-dimensional lattice, so is

ρ−1 (C). It follows that ΛC = p−1/2ρ−1 (C) is also an n(p − 1)-dimensional lattice.
We have

disc ΛC = ((p−1/2 )n(p−1))2 disc ρ−1 (C)

= p−n(p−1) disc Dn | Dn /ρ−1 (C) |2
= p−n(p−1) pn(p−2) p2(n−k)
= pn−2k . (1)

(i) For any x ∈ ΛC , x can be expressed as

x = p−1/2 (c + z), where c ∈ C, z ∈ P n .

Then

(x, x) = p−1 (Tr(c · c) + Tr(c(z + z̄)) + Tr(zz̄)) ,

where Tr = TrQ(ζ)/Q. It is easy to verify that

Tr(c · c) = (p − 1)(c · c)

and

Tr(y + ȳ), Tr(y ȳ) ∈ 2pZ for all y ∈ P.

Thus,

(x, x) = p−1 ((p − 1)(c · c) + 2pr), where r ∈ Z

= p−1 (p − 1)(c · c) + 2r.

Therefore,

(x, x) ∈ 2Z ⇔ p | c · c
⇔ c · c = 0 in Fp .

Hence, ΛC is even if and only if c · c = 0 for all c ∈ C. By Proposition 1, ΛC is

even if and only if C is self-orthogonal.
(ii) By (1), disc ΛC = 1 if and only if n = 2k, i.e., n is even and k = n/2. By
Proposition 2 and (i),

C is self-dual ⇔ n is even, k = n/2, and C ⊆ C ⊥

⇔ disc ΛC = 1 and ΛC is even
⇔ ΛC is even unimodular.


A characteristic property of self-orthogonal codes 481

The “if” parts of Proposition 3 can be found in [1], i.e., Proposition 5.2 of [1],
p. 135.
Let Λ be a lattice in Rn . Define

Λ∗ = {x ∈ Rn | x · y ∈ Z for all y ∈ Λ}.

Then Λ∗ is also a lattice in Rn , called the dual lattice of Λ. Λ is called integral if

x · y ∈ Z for all x, y ∈ Λ. For an integral lattice Λ, Λ ⊆ Λ∗ and Λ∗ /Λ is a finite
abelian group. Λ is called even if x · x ∈ 2Z for all x ∈ Λ. If Λ is even then it is
integral.
Let Λ be an even lattice. A vector of square length 2 in Λ is called a root of Λ.
If Λ is generated by all its roots, Λ is called a root lattice. If Λ cannot be written as
the direct sum of two sublattices Λ1 and Λ2 such that (x1, x2) = 0 for all x1 ∈ Λ1
and x2 ∈ Λ2, Λ is called irreducible. It is known that irreducible root lattices are
of types An (n ≥ 1), Dn (n even and ≥ 4), En (n = 6, 7, 8), (cf. Theorem 1.2 of [1],
p. 20). If irreducible root lattices Λ and Λ0 are of the same type, we write Λ ' Λ0 .
As in the binary case we can study which irreducible root lattices arise as lattices
ΛC from p-ary codes C.
Lemma 1
Let C be a p-ary linear code, then Λ∗C = ΛC ⊥ .
Proof. Let x = p−1/2 (c + z) ∈ ΛC and y = p−1/2 (c0 + z0) ∈ ΛC ⊥ , where c ∈ C,
c0 ∈ C ⊥ , and z, z0 ∈ P n . Then

(x, y) = p−1 Tr(c · c0 + c · z0 + z · c0 + z · z0 ).

For c ∈ C and c0 ∈ C ⊥ we have c · c0 = 0 in Fp . Computed in C, c · c0 ≡ 0

(mod p). Since z, z0 ∈ P n and P n = P n , we have c · z0 , z · c0 , z · z0 ∈ P . Thus
Tr(c · c0 + c · z0 + z · c0 + z · z0 ) ∈ pZ. Therefore (x, y) ∈ Z. This proves ΛC ⊥ ⊆ Λ∗C .
Let dim C = k. By Proposition 3, disc ΛC = pn−2k and disc ΛC ⊥ = p2k−n . But
disc Λ∗C = (disc ΛC )−1 = p2k−n . Therefore disc ΛC ⊥ = disc Λ∗C . Hence ΛC ⊥ = Λ∗C . 

Proposition 4
Let Λ be an irreducible root lattice in Rn . Then Λ ' ΛC for a p-ary linear code C
of length n, where p is an odd prime if and only if Λ is of type Ap−1 , E6 (when p = 3
and n = 3), or E8 (when p = 3 and n = 4, or p = 5 and n = 2).
Proof. Assume that Λ ' ΛC for a p-ary linear code C of length n, where p is an odd
prime. For any p-ary linear code C 0 of length n, let x ∈ LC 0 , then x = p−1/2 (c0 + z),
where c0 ∈ C and z ∈ P n . Thus px = p−1/2 (p(c0 + z)) and p(c0 + z) ∈ P n . Therefore
pΛC 0 ⊆ p−1/2 P n ⊆ ΛC . Since Λ∗C = ΛC ⊥ , we have, in particular, pΛ∗C ⊆ ΛC . But
Λ∗C /ΛC is a finite abelian group. So

Λ∗C /ΛC ' (Z/pZ)` for some ` ≥ 0. (2)

By inspecting the irreducible root lattices one by one we find that only Ap−1 , E6
(when p = 3), and E8 satisfy the condition (2). Moreover, if E6 ' ΛC for a p-ary
linear code C of length n, then 6 = (p − 1)n, which implies p = 3 and n = 3. If
E8 ' ΛC for a p-ary linear code C of length n, then 8 = (p − 1)n. It follows that
482 Z.-X. Wan

p = 3 and n = 4 or p = 5 and n = 2. Therefore Λ is of type Ap−1 , E6 (when p = 3

and n = 3), or E8 (when p = 3 and n = 4, or p = 5 and n = 2).
Conversely, assume that Λ is of type Ap−1, E6 , or E8 . If Λ is of type Ap−1 , let C
be the 1-dimensional code {0} consisting of 0 only; then ΛC = p−1/2 P , which is of
type Ap−1 . If Λ is of type E6 , let C = F3 (1, 1, 1); then Λ ' ΛC . If Λ is of type E8 ,
let C = F3 (1, 0, 1, 1) + F3 (0, 1, 1, 2) or F5 (1, 2); then Λ ' ΛC . 

References
[1] W. Ebeling. Lattices and Codes. Vieweg, Wiesbaden, 1994.

[2] F. Hirzebruch. A letter to N. J. A. Sloane on 19 August, 1986. In Gesammelte

Abhandlungen, Collected Papers, pages 796–798. Band II. Springer, Berlin, 1987.

[3] K. Ireland and M. Rosen. A Classical Introduction to Modern Number Theory.

Springer, New York, 2nd edition, 1990.

[4] F. J. MacWilliams and N. J. A. Sloane. The Theory of Error-Correcting Codes.

North-Holland, Amsterdam, 1977.

[5] B. L. van der Waerden. Algebra I. Springer, Berlin, 6 Auflage, 1964.

Zhe-Xian Wan
Department of Information Technology
Lund University
P.O. Box 118
S-221 00 Lund
Sweden
email: [email protected]