Linear Algebra and its Applications 341 (2002) 23–34

www.elsevier.com/locate/laa

The Bush matrix over a Galois field and error

correcting quantum codes
K.R. Parthasarathy
Indian Statistical Institute, Delhi Centre, 7, S.J.S. Sansanwal Marg, New Delhi 110016, India
Received 31 August 2000; accepted 1 December 2000
Submitted by R. Bhatia

Dedicated to T. Ando

Abstract
Using the method of Bush in the construction of orthogonal arrays [A.S. Hedayat, N.J.A.
Sloane, J. Stufken, Orthogonal Arrays; Theory and Applications, Springer Series in Statistics,
Springer, Berlin, 1999] and the theory of characters of finite abelian groups we construct a
family of error correcting quantum codes. The trade-off between the dimension of the quantum
code and the number of errors corrected is investigated in this class. Associated with each
prime we present an explicit family of error correcting quantum codes. Our proofs depend
on the well-known Knill–Laflamme criterion [E. Knill, R. Laflamme, Phys. Rev. A 55 (1997)
900] for error correction and a basic result of A.R. Calderbank et al. [IEEE Trans. Inform.
Theory 44 (1998) 1369] modified appropriately to the context of finite abelian groups. © 2002
Elsevier Science Inc. All rights reserved.
Keywords: Error correcting quantum codes; Characters of finite abelian groups; Weyl operators; Galois
field; Bush matrix; Orthogonal arrays

1. Introduction

The theory of error correcting quantum codes as outlined in [3,7] is based on

the following hypotheses. Messages can be encoded as states of a finite level quan-
tum system and such states can be stored or transmitted through a communication
channel. The encoded states are called input states. In the process of storing or trans-
mission, errors can occur and the output state need not be the same as the input state.

E-mail address: [email protected] (K.R. Parthasarathy).

0024-3795/02/$ - see front matter  2002 Elsevier Science Inc. All rights reserved.
PII: S 0 0 2 4 - 3 7 9 5 ( 0 1 ) 0 0 2 5 8 - 0
24 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34

The transformation of any input state into an output state is a completely positive
affine linear map. The quantum system whose states are used for encoding and de-
coding messages is characterised by a finite-dimensional complex Hilbert space H
with scalar product
·, · which is linear in the second variable. A state ρ of such
a system is a nonnegative definite linear operator of unit trace. All such states con-
stitute a compact convex set whose extreme points are one-dimensional orthogonal
projection operators. By a projection in H we shall always mean an orthogonal
projection. We shall not hesitate to use the Dirac notation in dealing with operators
in H. For any input state ρ we assume that the output state ρ̂ takes the form


ρ̂ = Lj ρL†j , (1.1)
j

where {Lj } is a finite family of operators satisfying the condition j L†j Lj = I , I
denoting the identity operator and L†j denoting the adjoint of Lj . The operators Lj
in (1.1) induce the errors which corrupt ρ and produce the output ρ̂. We assume that
in the C ∗ -algebra B(H) of all operators on H there is a subspace E ⊂ B(H) to
which the error-inducing operators Lj belong. We call E the space of error opera-
tors. Note that the input–output correspondence ρ → ρ̂, when the Lj ’s are fixed, is
a completely positive trace-preserving map. Our goal is to retrieve the input state ρ
from a knowledge of the corrupted output state ρ̂, whatever the Lj ’s in E. It may not
be possible to retrieve every input state ρ. So our aim is to find a substantially large
subspace C ⊂ H possessing the property that every input state ρ with support in C
can be recovered without error. To this end we introduce a definition.

Definition 1.1. A subspace C ⊂ H is called an E-correcting quantum code if there

exists a finite family {Ri } of operators in H satisfying the following:
(i)

†
Ri Ri = I, (1.2)
i
(ii) For any state ρ with support in C and any finite family {Lj } of operators from E

satisfying j L†j Lj = I , one has


Ri Lj ρL†j Ri† = ρ. (1.3)
i.j
Here, ρ is said to have support in C if Tr ρP = 1, with P being the projection on
C.

Note that the correspondence ρ̂ → i Ri ρ̂Ri† is also a trace-preserving com-
pletely positive map. Condition (1.3) implies that this map retrieves or recovers the
input state ρ from the corrupted state ρ̂ of the form (1.1) as long as the error-inducing
operators Lj come from E and the input state ρ has its support in C. For a given E
our aim is to design an E-correcting quantum code C and use the states on C to
encode messages for transmission or storage.
 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34 25

In this context we have the following fundamental theorem due to Knill and Lafl-
amme.

Theorem 1.2 (Knill and Laflamme [3]). In order that a subspace C of H may be
an E-correcting quantum code it is necessary and sufficient that the following holds:
there exists an orthonormal basis {ψ1 , ψ2 , . . . , ψd } in C such that for any L, M ∈ E
one has
(i)
ψi , L† Mψj  = 0 if i =
/ j;
(ii)
ψi , L† Mψi  is independent of i = 1, 2, . . . , d.

Here d is the dimension of the quantum code. Another method of describing the
Knill–Laflamme criterion for C to be an E-correcting quantum code is that the map
ψ →
ψ, L† Mψ from the unit sphere in C to complex scalars is a constant for
every fixed L, M ∈ E.
An important class of error spaces occurs in the following manner. The quantum
system described by the states in H is itself made up of a large number of ‘ele-
mentary’ systems. For example H = h⊗ is an n-fold tensor product of copies of a
n

Hilbert space h. (In particular, one can consider h to be a two-dimensional Hilbert

space and a pure state in h is then said to contain one ‘qubit’ of information). We
assume that h has dimension m. A product operator A1 ⊗ A2 ⊗ · · · ⊗ An in H where
Ai ∈ B(h) is said to have ‘strength’ not exceeding t if all but t of the Ai ’s are scalar
multiples of the identity. Denote by Et the linear span of all product operators of
strength not exceeding t. Then an Et -correcting quantum code is simply called a
t-error correcting quantum code.
In order to store or transmit a large number of messages using states with support
in C it is necessary to have the dimension of C as large as possible. At the same
time we would like C to be t-error correcting for t as large as possible. These are two
conflicting interests.
Using the theory of characters of finite abelian groups and the method of con-
structing orthogonal arrays due to Bush (see [2]) in the framework of Galois fields
we shall construct here a family of t-error correcting quantum codes in h⊗ for dif-
n

ferent values of t and n when n is the power of a prime number. This also enables
us to examine in some detail the conflicting interests mentioned in the preceding
paragraph as n increases to ∞. As a by-product we obtain an explicit family of t-
error correcting quantum codes in terms of Van der Monde matrices when n varies
over the set of primes.

2. A reduction of the Knill–Laflamme criterion in terms of Weyl operators

In Theorem 1.2 of the preceding section we have a criterion for the subspace C
to be an E-correcting quantum code. We shall reduce this to a more manageable
computational form when H = h⊗ and dim h = m. To this end we identify h with
n
26 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34

L2 (A) where A is an additive abelian group with its addition operation denoted by
+ and null element by 0. Then H can be identified with L2 (An ), where An is the
n-fold cartesian product of copies of A. In L2 (A) define the unitary operators
(Ua f )(x) = f (x + a), a, x ∈ A, f ∈ L2 (A), (2.1)

(Vχ f )(x) = χ (x)f (x), χ ∈ Â, f ∈ L2 (A), (2.2)

where x varies over the character (or dual) group of A denoted by Â. Note that Â
is a multiplicative abelian group with its identity element denoted by 1, the identity
character. In (2.2), χ (x) denotes the value of the character χ at the point x. We
call Ua Vχ a Weyl operator for any a ∈ A, χ ∈ Â. We have the Weyl commutation
relations (see [6])
Ua Ub = Ua+b , Vα Vβ = Vαβ , Ua Vα = α(a)Vα Ua (2.3)
for all a, b ∈ A, α, β ∈ Â. The family {Ua Vχ , a ∈ A, χ ∈ Â} is irreducible in h.
This means that h does not have any proper subspace invariant under all the Weyl op-
erators. This implies that every operator in h can be expressed as a linear combination
of Weyl operators. Indeed, for the Hilbert space B(h) with scalar product

L, M = Tr L† M
the family {m−1/2 Ua Vα , (a, α) ∈ A × Â} is an orthonormal basis.
We now consider the abelian group An whose character group can be identi-
fied with Ân . Elements of An and Ân are of the form a = (a1 , a2 , . . . , an ) and

= (α1 , α2 , . . . , αn ), respectively, with ai ∈ A, αi ∈ Â for each i where

n

(a) = αi (ai ).
i=1
The Weyl operators Ua V
for An are given by
Ua = Ua1 ⊗ Ua2 ⊗ · · · ⊗ Uan ,

V
= Uα1 ⊗ Vα2 ⊗ · · · ⊗ Vαn ,

when L2 (An ) is identified with L2 (A)⊗ . Every operator in L2 (An ) can again be ex-
n

pressed uniquely as a linear combination of Weyl operators Ua V
, (a,
) ∈ An × Ân .

For any abelian group G, consider its n-fold cartesian product Gn . If g =
(g1 , g2 , . . . , gn ), gi ∈ G we define its weight denoted w(g) by
w(g) = #{i | gi =
/ e, 1
i
n}, (2.4)
where e is the identity element of G. In particular, when (a,
) ∈ An × Ân is consid-
ered as an element of (A × Â)n its weight w(a,
) is given by
/ {i | (ai , αi ) =
w(a,
) == / (0, 1)}. (2.5)
For later use we would need the following definition. For any subgroup C ⊂ Gn
define
 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34 27

d(C) = min w(g), (2.6)

g∈C
g=
/e

where e = (e, e, . . . , e) is the identity element of Gn . The subgroup C is called an

error correcting group code with minimum distance d(C). If t = [(d(C) − 1)/2], C
is called a t-error correcting classical group code.
After this small digression into notations and definitions we are ready to state an
elementary proposition.

Proposition 2.1. The linear space Et spanned by all product operators of strength
not exceeding t in L2 (An ) viewed as L2 (A)⊗ is also the linear span of all Weyl
n

operators Ua V
satisfying w(a,
)
t.

Proof. This is immediate from the fact that {Ua Vα , (a, α) ∈ A × Â} spans
B(L2 (A)) and therefore {Ua V
, (a,
) ∈ As × Âs } spans B(L2 (A)⊗ ) for every s =
s

1, 2, . . . , where Ua V
= Ua1 Vα1 ⊗ Ua2 Vα2 ⊗ · · · ⊗ Uas Vαs . 

Theorem 2.2. An orthonormal set {f1 , f2 , . . . , fs } in L2 (An ) viewed also as

L2 (A)⊗ spans a t-error correcting d-dimensional quantum code if and only if for
n

every (a,
) ∈ An × Ân with w(a, )
2t the following hold:

(i)
fi , Ua V
fj  = 0 if i =
/ j;
(ii)
fi , Ua V
fi  is independent of i = 1, 2, . . . , d.

Proof. Note that the Knill–Laflamme conditions (i) and (ii) in Theorem 1.2 are
sesquilinear in the pair (L, M) of error operators in E which is equal to Et in the
present theorem. By Proposition 2.1, Et is spanned by Weyl operators of the form
Ua V
with w(a,
)
t. If L = Ua V
, M = Ub V , where w(a,
)
t, w(b, )
t
then L† M is a scalar multiple of a Weyl operator Uc V with w(c, )
2t. Now the
required result is immediate from Theorem 1.2. 

In formulating and proving Theorem 2.2 we have used the Weyl operators for an
additive abelian group and its dual as a unitary error basis. For different examples of
such unitary error bases and their role in the study of error correcting quantum codes
we refer to [4,5,8].

3. An application of the reduced Knill–Laflamme criterion

For any subgroup C ⊂ A denote its annihilator subgroup by C ⊥ ⊂ Â, where

C ⊥ = {α | α ∈ Â, α(x) = 1 ∀x ∈ C}. (3.1)
Then the character group Ĉ of C is identified with the quotient group Â/C ⊥ so that
#Ĉ · #C ⊥ = #Â and therefore
28 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34

#A = #C · #C ⊥ . (3.2)
Let now C1 , C2 be two subgroups of A such that C1 ⊃ C2 so that C1⊥ ⊂ C2⊥ . Write
the coset partition

C2⊥ = χ C1⊥ , (3.3)
χ∈S

where S is contructed by picking exactly one element from each coset of the sub-
group C1⊥ in C2⊥ and putting them together. Then (3.2) implies
#S = #C2⊥ /#C1⊥ = #C1 /#C2 . (3.4)
Define the functions
1
fχ = (#C1 )− 2 χ 1C1 , χ ∈ S, (3.5)
where 1C1 denotes the indicator of C1 in A. We consider fχ as an element of the
Hilbert space L2 (A). Define
B(a, α; χ1 , χ2 ) =
fχ1 , Ua Vα fχ2 , (3.6)
where χ1 , χ2 ∈ S, (a, α) ∈ A × Â and Ua Vα is the associated Weyl operator deter-
mined by (2.1) and (2.2).

Proposition 3.1. Let (a, α) ∈ {(0, 1)} ∪ C1 × Â ∪ C1 × (C2⊥ ) where C1 and (C2⊥ )
are complements of C1 and C2⊥ in A and Â, respectively. Then
B(a, α; χ1 , χ2 ) = 0 if χ1 =
/ χ2 , χ1 , χ2 ∈ S
and B(a, α; χ , χ ) is independent of χ ∈ S.

Proof. From (3.5) and (3.6) we have

B(a, α; χ1 , χ2 ) = (#C1 )−1 χ̄1 (x − a)1C1 (x − a)α(x)χ2 (x)1C1 (x). (3.7)
x∈A
Let a = 0, α = 1. Then the right-hand side of (3.7) reduces to


(#C1 )−1 (χ̄1 χ2 )(x),
x∈C1

which vanishes by the Schur orthogonality relations for characters if χ̄1 χ2 restricted
to C1 is a nontrivial character. If χ̄1 χ2 is trivial on C1 , then χ̄1 χ2 ∈ C1⊥ , i.e., χ2 lies
in the coset χ1 C1⊥ . By the choice of S this would imply χ1 = χ2 , in which case the
right-hand side of (3.7) is equal to unity.
Now consider the case a  ∈ C1 and α is arbitrary. In particular a = / 0. Now the xth
term on the right-hand side of (3.7) is nonzero only if x − a ∈ C1 , x ∈ C1 . Since
C1 is a subgroup this would imply a ∈ C1 , a contradiction and hence the required
property holds.
Let now a ∈ C1 , α  ∈ C2⊥ . Then the right-hand side of (3.7) becomes equal to
 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34 29


χ1 (a)(#C1 )−1 (χ̄1 χ2 α)(x),
x∈C1
which vanishes if χ̄1 χ2 α restricted to C1 is a nontrivial character. Otherwise χ̄1 χ2 α ∈
C1⊥ ⊂ C2⊥ . Since χ1 and χ2 belong to C2⊥ it follows that α ∈ C2⊥ , a contradiction.
This completes the proof. 

We shall now apply Proposition 3.1 to the case of two subgroups of An and obtain
a slightly modified group theoretic version of a basic result of Calderbank et al. [1].

Theorem 3.2. Let Cn , Dn be two subgroups of An such that Cn ⊃ Dn ,

d(Cn ) = dn , d(Dn⊥ ) = dn ,
where the left-hand sides are defined by (2.6) in the groups An and Ân . Let
 
min(dn , dn ) − 1
tn = . (3.8)
2
Suppose

Dn⊥ = Cn⊥
∈Sn

is a partition of Dn⊥ into cosets of its subgroup Cn⊥ . For any  ∈ Sn define
f = (#Cn )−1/2 1Cn .
Then {f |  ∈ Sn } is an orthonormal set of cardinality #Cn /#Dn which spans a tn -
error correcting quantum code in L2 (A)⊗ .
n

Proof. Consider a Weyl operator of the form Ua V
, where w(a,
)
2tn . Then (3.8)
implies w(a) < dn . By the definition of dn (see (2.6)) either a = o or a  ∈ Cn . Simi-
larly w(
) < dn and therefore
= 1 or
 ∈ Dn⊥ . Now an application of Proposition
3.1 to the group An shows that the conditions of Theorem 2.2 are fulfilled by the
family {f |  ∈ Sn } for t = tn . This completes the proof. 

4. The Bush matrix and quantum codes

We shall now choose the abelian group A to be the additive group of a Galois
field GF(s) where s is a positive integer of the form pl , p being a prime and l a
positive integer. Enumerate the elements of A as α1 , α2 , . . . , αs in any order. Let
0 < t
s + 1 be fixed. Enumerate all the s t polynomials of degree
t − 1 by
ϕi (x) = ai0 + ai1 x + ai2 x 2 + · · · + ait−1 x t−1 , (4.1)
where aij ∈ A. Denote N = st . We define the Bush matrix of order N × s with en-
tries in A by
30 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34
 
ϕ1 (α1 ) ϕ1 (α2 ) ··· ϕ1 (αs )
 ϕ2 (α1 ) ϕ2 (α2 ) ··· ϕ2 (αs ) 
 
 ··· ··· ··· ··· 
Bt = 
 ϕi (α1 )
. (4.2)
 ϕi (α2 ) ··· ϕi (αs ) 

 ··· ··· ··· ··· 
ϕN (α1 ) ϕN (α2 ) ··· ϕN (αs )
We assume that the degree of ϕi is nondecreasing in i so that for 0 < t  < t, the

Bush matrix Bt  consists of the first N  = s t rows of Bt . Since all polynomials of
degree
t − 1 with coefficients in A constitute a vector space over A it follows that
the rows of Bt constitute an additive subgroup of As of cardinality N with its null
element being the first row in Bt . We denote this subgroup also by Bt . On matters
concerning the important role of Bush matrices in the context of orthogonal arrays
and error correcting classical codes we refer the reader to the book [2] by Hedayat et
al.
We now choose and fix two integers t, t  satisfying 0 < t  < t
s + 1 and con-
sider the Bush subgroups Bt  ⊂ Bt ⊂ As . We now state a proposition concerning
d(Bt ) and d(Bt⊥ ) determined by (2.6) when G = A.

Proposition 4.1. Let 0 < t  < t
s + 1. Then

d(Bt ) = s − t + 1, d(Bt⊥ )  t  + 1,
where Bt⊥ ⊂ Âs is defined as in (3.1).

Proof. Let ϕi be a nonzero polynomial of degree
t − 1. Then the number of its ze-
ros does not exceed t − 1. Hence w(ϕi (α1 ), ϕi (α2 ), . . . , ϕi (αs ))  s − (t − 1). Now
consider the specific polynomial ϕ(x) = (x − α1 )(x − α2 ) · · · (x − αt−1 ) of degree
t − 1. This vanishes exactly at α1 , α2 , . . . , αt−1 and therefore w(ϕ(α1 ), ϕ(α2 ), . . . ,
ϕ(αs )) = s − t + 1. This proves the first part. To prove the second part consider a
character (χ1 , χ2 , . . . , χs ) of As which is nontrivial but belongs to Bt⊥ . Suppose
w(χ1 , χ2 , . . . , χs )
t  . Then there exist nontrivial characters χi1 , χi2 , . . . , χir , r

t  such that
χi1 (ϕ(αi1 ))χi2 (ϕ(αi2 )) · · · χir (ϕ(αir )) = 1
for every polynomial ϕ of degree not exceeding t  − 1. Let (c1 , c2 , . . . , cr ) be any
sequence of r elements from A. Write βj = αij and

r  x − βi 
ϕ(x) = cj .
βj − βi
j =1 1
i
r
i =j
/
Then ϕ is a polynomial of degree r − 1 and ϕ(βj ) = cj , j = 1, 2, . . . , r. Thus
χi1 (c1 )χi2 (c2 ) · · · χir (cr ) = 1
for all c1 , c2 , . . . , cr ∈ A. This implies that χi1 , χi2 , . . . , χir are trivial, a contradic-
tion. This completes the proof. 
 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34 31

Theorem 4.2. Let 0 < t  < t
s + 1, where s = p l , p being a prime and l a pos-
itive integer. Let Bt  ⊂ Bt be the Bush subgroups of As and let

Bt⊥ = χ Bt⊥
χ∈S

be a partition of Bt⊥ into cosets of its subgroup Bt⊥ . Define

fχ = s −t/2 χ 1Bt , χ ∈ S,
as complex-valued functions belonging to L2 (As ), where A denotes the additive
group of the Galois field GF(s). Then {fχ | χ ∈ S} is the orthonormal basis of a

quantum code of dimension s t−t which can correct τ errors, where
 
(s − t + 1) ∧ (t  + 1) − 1
τ= , (4.3)
2
x ∧ y denoting the minimum of x and y and [x] the integral part of x.

Proof. This is immediate from Proposition 4.1 and Theorem 3.2. 

Remark 1. Denote the quantum code constructed in Theorem 4.2 by C(s, t  , t)

where s = p l , 0 < t  < t
s + 1. For any fixed 0 < θ < 1 choose
   
1+θ 1−θ
t= s , t = s ,
2 2
and denote the τ defined by (4.3) as τ (θ, s). Then
log dim C(s, t  , t)
lim = θ,
s →∞ log dim L2 (As )
τ (θ, s) 1−θ
lims → ∞  ,
s 4
where s → ∞ as a sequence of prime powers. This gives an idea of the conflicting
interests in maximising the dimension of the code and maximising the percentage of
errors to be corrected.
When θ = 12 we see that 50% of the available qubits could be used for encoding
with the possibility of correcting 12 12 % of the errors if we choose s large enough .

Remark 2. We shall now present a more explicit orthonormal basis for the quantum
code C(s, t  , t) when s = p is a prime. In this case the field A = GF(p) consists of
{0, 1, 2, . . . , p − 1} with addition and multiplication modulo p. The elements of Â
can be enumerated as χ (0) , χ (1) , . . . , χ (p−1) , where
χ (k) (j ) = ωkj , 0
j
p − 1, 0
k
p − 1,
with ω being the pth primitive root of unity, i.e., ω = exp 2
i/p. An element
(χ1 , χ2 , . . . , χp ) ∈ Âp then lies in Bt⊥ if and only if χi = χ (ri ) , where r1 , r2 , . . . , rp
satisfy the relations
32 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34


p
χ (rj ) (ϕ(j − 1)) = 1
j =1

for every polynomial ϕ of the form ϕ(x) = a0 + a1 x + a2 x 2 + · · · + at  −1 x t −1 .
This is equivalent to

p

rj (a0 + a1 (j − 1) + a2 (j − 1)2 + · · · + at  −1 (j − 1)t −1 ) = 0 (mod p)
j =1
for all a0 , a1 , . . . , at  −1 ∈ {0, 1, 2, . . . , p − 1}. This, in turn, is equivalent to the re-
lations

p
rj (j − 1)k = 0, 0
k
t  − 1.
j =1

In other words the vector (r1 , r2 , . . . , rp )T (T denoting transpose) has the form
 
0
 0 
   
 .. 
r1  . 
 r2   
  
−1  0 

 ..  = Vp  , (4.4)
.  1 x
 x2 
rp  
 . 
 .. 
xp−t 
where Vp is the Van der Monde matrix given by
 
1 1 ··· ··· 1
0 1 2 ··· p−1 
 
Vp =  0  1 2 2 2 ··· (p − 1)2  (4.5)

· · · ··· ··· ··· ··· 
0 1p−1 2p−1 · · · (p − 1)p−1
with entries in GF(p) and x1 , x2 , . . . , xp−t  are arbitrary elements in GF(p). The
same argument shows that (χ1 , χ2 , . . . , χp ) with χj = χ (rj ) is in Bt⊥ if and only if
(r1 , r2 , . . . , rp )T is given by
 
0
 0 
   
 .. 
r1  . 
 r2   
  
−1  0 

 ..  = Vp  , (4.6)
.  y 1 
 y2 
rp  
 . 
 .. 
yp−t
 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34 33

where y1 , y2 , . . . , yp−t are arbitrary elements of GF(p). Eqs. (4.4) and (4.6) describ-
ing Bt⊥ and Bt⊥ at once show that a choice for the cross-section S for the coset space
Bt⊥ /Bt⊥ is given by (χ1 , χ2 , . . . , χp ) with χj = χ (rj ) , where
 
0
 0 
 
 .. 
 . 
 
   0 
r1  
 x1 
 r2   
   
 ..  = Vp−1  x2  , (4.7)
.  .. 
 . 
rp  
xt−t  
 
 0 
 
 . 
 .. 
0
where on the right-hand side the first t  and the last p − t elements in (0, 0, . . . 0, x1 ,
x2 , . . . xt−t  , 0, 0, . . . , 0)T are zero. This shows that the orthonormal basis {f |  ∈
S} of the quantum code C(p, t  , t) of Theorem 4.2 can be labelled by x1 , x2 , . . . , xt−t 
with xj ∈ GF(p) for each j = 1, 2, . . . , t − t  and expressed as


fx = p −t/2 cϕ (x1 , x2 , . . . , xt−t  ) | ϕ(0), ϕ(1), . . . , ϕ(p − 1), (4.8)
ϕ

where the summation is over all polynomials ϕ of degree
t − 1 over GF(p),

 
0
 0 
 
 .. 
 . 
   
 0  ϕ(0)
  
2
i 
 x1 

−1
 ϕ(1) 
 
cϕ (x1 , x2 , . . . , xt−t  ) = exp T
 .  , (Vp )  ..  (4.9)
p  ..   . 
 
xt−t   ϕ(p − 1)
 
 0 
 
 . 
 .. 
0
and |a1 , a2 , . . . , ap  is the indicator function of the singleton consisting of (a1 ,
a2 , . . . , ap ) ∈ Ap considered as the ket vector |a1 |a2  · · · |ap  in L2 (A)⊗ , |a be-
p

ing the indicator of the singleton {a} in A. Then the quantum code described by
(4.8) and (4.9) can correct [((p − t + 1) ∧ (t  + 1) − 1)/2] errors and has dimension

p t−t .
34 K.R. Parthasarathy / Linear Algebra and its Applications 341 (2002) 23–34

Acknowledgements

The author is grateful to the Indian National Science Academy for its generous
financial support in the form of C.V. Raman Professorship. Part of this work was
done at the Nottingham Trent University and Nottingham University during July
2000 and the author thanks D.B. Applebaum, R.L. Hudson and J.M. Lindsay for the
generous and warm hospitality showered on him.

References

[1] A.R. Calderbank, E.M. Rains, P.W. Shor, N.J. Sloane, Quantum error correcting codes over GF(4),
IEEE Trans. Inform. Theory 44 (1998) 1369–1387.
[2] A.S. Hedayat, N.J.A. Sloane, J. Stufken, Orthogonal Arrays; Theory and Applications, Springer Se-
ries in Statistics, Springer, Berlin, 1999.
[3] E. Knill, R. Laflamme, A theory of quantum error correcting codes, Phys. Rev. A 55 (1997) 900–911.
[4] E. Knill, Group representations, error bases and quantum codes, quant-ph/9608049.
[5] E. Knill, Nonbinary unitary error bases and quantum codes, quant-ph/9608048.
[6] K.R. Parthasarathy, Quantum codes arising from Weyl commutation relations on finite abelian
groups, ISI Delhi Centre preprint 2000.
[7] A.O. Pittinger, An Introduction to Quantum Computing Algorithms, Progress in Computer Science
and Applied Logic, vol. 19, Birkhäuser, Basel, 1999.
[8] E.M. Rains, Nonbinary quantum codes, IEEE Trans. Inform. Theory 45 (1999) 1827–1832.