QUANTUM BCH CODES

Markus Grassl and Thomas Beth

Institut für Algorithmen und Kognitive Systeme (IAKS)
Universität Karlsruhe, Am Fasanengarten 5, 76 128 Karlsruhe, Germany
Email: [email protected], EISS [email protected]

Abstract Mathematically, the state of the basic unit of quantum in-

formation, a quantum bit (or short qubit), is represented by
After a brief introduction to both quantum computation and
the normalized linear combination
quantum error correction, we show how to construct quan-
tum error-correcting codes based on classical BCH codes.
|qi = α |0i + β |1i where α, β ∈ C, |α|2 + |β|2 = 1.
With these codes, decoding can exploit additional informa-
arXiv:quant-ph/9910060v1 14 Oct 1999

tion about the position of errors. This error model—the

The normalization condition stems from the fact that when
quantum erasure channel—is discussed. Finally, parame-
extracting classical information from the quantum system
ters of quantum BCH codes are provided.
by a measurement, the values 0 and 1 occur with probabil-
ity |α|2 and |β|2 , resp.

1 Introduction Similar to classical registers, a quantum register is built

by combining several qubits. Mathematically, this cor-
Motivated by the statement responds to the tensor product of two-dimensional vector
“BCH codes are among the best (classical) codes we know” spaces1 . Hence the state of a quantum register of length n
can be any normalized complex linear combination of the
(cited from Ch. 9, §1, p. 258 of [14]), we present the trans- 2n mutually orthogonal basis states
lation of classical Bose-Chaudhuri-Hocquenghem (BCH)
codes into quantum quantum error-correcting codes. With- |b1 i ⊗ . . . ⊗ |bn i =: |b1 . . . bn i = |bi where bi ∈ {0, 1}.
out error correction, the promising new field of quantum
computing (see, e. g., [16, 13]) would be mainly of theoret-
ical nature. A main ingredient of quantum computation is 2.2 Quantum Gates
constructive and destructive interference of different com-
The laws of quantum mechanics say that any transforma-
putation paths which is only possible when using quantum
tion on quantum systems is linear. Furthermore, in order to
states. But on the other hand, any possible computing de-
preserve the normalization any operation has to be unitary.
vice exploiting quantum mechanics has to cope with un-
Let us first consider operations involving only one qubit,
controllable interactions with the environment, e. g., single
i. e., one subsystem. Similar to the classical N OT gate,
photons. Quantum error-correcting codes help to actively
there is a quantum operation exchanging the states |0i and
reduce the decoherence due to coupling to the environment.
|1i given by the matrix
 
0 1
2 Background N OT :=
1 0
.

2.1 Quantum Registers

But on a single qubit, there is not only this “classical” op-
Classically, information is often represented by bits. A sin- eration. Examples for non-classical operations on single
gle bit takes either the value 0 or 1. In physical systems, qubits are given by
0 and 1 are represented by two different states of the sys-    
tem. These could be two different voltages, signals with 1 1 1 1 0
H := √ and σz := .
two different frequencies, but also states on the quantum 2 1 −1 0 −1
mechanical level, e. g., ground state and excited state of an (1)
electron of an atom or ion, the spin of a nucleus, or the po- Besides single qubit operations, the so-called controlled
larization of photons. In Dirac notation [7], the two states N OT gate (CN OT ) plays an important rôle since any
are written as unitary operation on a 2n -dimensional space can be im-

1
 plemented using only single qubit operations and CN OT
“0” =ˆ |0i = ∈ C2 gates (see [1]). As a classical gate, the CN OT gate corre-
0
sponds to a gate with two inputs and two outputs. One of
and   the inputs is copied to the first output, the second output is
0
“1” =
ˆ |1i = ∈ C2 . the XOR of the inputs. The transformation matrix of the
1
1 In quantum mechanics, the underlying structures are Hilbert spaces.
In quantum mechanics, the principle of superposition al- We do not stress this fact since here all vector spaces are finite dimensional
lows a system to be simultaneously in different states. and thus complete w. r. t. the standard Hermitian inner product.

207
CN OT gate is given by: The state of our quantum system is now, in general, a mixed
state given by the density operator ρsys . One interpreta-
1

0 0 0
 tion of a mixed quantum state is that we have an ensemble
 0 |ai • |ai of pure quantum states chosen according to a probability
1 0 0 
CN OT := 
 0

distribution. In our case, one can think of a measurement
0 0 1  |bi i |a ⊕ bi
0 0 1 0 performed on the environment. Due to entanglement with
the system, this may lead to different states of the system
On the right hand side, the notation for the CN OT gate depending on the measurement outcome—but we do not
as a quantum circuit is given. Each of the horizontal lines know which one since the result of the measurement is dis-
(wires) corresponds to a qubit of the whole quantum regis- carded.
ter. The dot on the upper wire indicates that the transfor- In order to model a quantum channel, we make use of equa-
mation on the lower qubit (the target)—a N OT gate—is tion (2). The disturbed quantum state ρsys can be expressed
only applied when the state of the upper qubit (the control) only in terms of the initial state |ψi hψ|sys of the system
is |1i. More examples for quantum circuits can be found in and some interaction operators Aj which completely spec-
[15]. ify the channel.
For a single qubit, i. e., a two-dimensional quantum system,
the operators Aj can be chosen to be proportional to the
3 Quantum Error Correction identity operator and the Pauli matrices
Classically, a major technique for protecting information      
0 1 0 −i 1 0
against channel errors is to add redundant information. The σx := , σy := , σz :=
1 0 i 0 0 −1
simplest example is a repetition code where information is
replicated by the sender. At the receiver’s end of the chan- where (i2 = −1). Surprisingly, in order to correct an arbi-
nel, the most likely information is chosen based on com- trary error it is sufficient to be able to correct any of these
paring all received messages and taking a majority vote. four errors.
This technique cannot be translated directly to quantum For more than one qubit, an error basis can be formed by
systems since it is not possible to copy unknown quantum tensor products of the one qubit interaction operators. A
information (no-cloning theorem [19]), and comparison of common assumption is that the errors act independently
quantum states is only possible statistically. Nevertheless, on each qubit. Furthermore, errors are assumed to be
quantum states can be protected against errors. The main small, i. e., near identity (with respect to a suitable operator
idea is to embed quantum information represented by k norm). Then errors with a small number of tensor factors
qubits into a larger Hilbert space of n qubits where n > k. different from identity are more likely than those errors with
For the construction of quantum error-correcting codes, we a large number of tensor factors different from identity.
have to model which types of errors occur during the trans-
mission over a quantum channel. This topic will be ad- 3.1.2 Depolarizing and Erasure Channel
dressed next.
To illustrate the preceding, we consider two important
quantum channels. Over a depolarizing channel [2], quan-
3.1 Error Models tum information is transmitted undisturbed with probabil-
ity 1 − ε, and it is replaced by a completely randomized
3.1.1 Open Quantum Systems
quantum state with probability ε. In this case, equation (2)
We assume that our quantum system interacts with an en- reads
vironment which is not or only partially accessible. Nev-
ertheless, we can model the interaction by a unitary trans- ρsys = (1 − ε) · |ψi hψ|sys + ǫ · 1
formation Uinteraction = Uint on the Hilbert space formed by
 
= (1 − 3/4 · ε) id |ψi hψ|sys id
the system and its environment. Assuming that there is no X  
prior entanglement of the system with the environment, the +ε/4 σj |ψi hψ|sys σj .
interaction operator reads as j=x,y,z

A related quantum channel is the quantum erasure channel

 
|ψisys ⊗ |Ψienv 7−→ Uinteraction |ψisys ⊗ |Ψienv .
[11]. Again, the quantum state is transmitted undisturbed
with probability 1 − ε. In case of an error, the quantum
After this interaction, the state need no longer be a tensor state is replaced by a quantum state |ei that is orthogonal
product. Since we cannot control the environment, we have to all other quantum states. Equation (2) now reads
to discard any information about the environment. This is
mathematically reflected by tracing out the environment: ρsys = (1 − ε) · |ψi hψ|sys + ǫ · |ei he| .

Similar to classical erasures, the state |ei indicates that an

   
†
ρsys = Trenv Uint |ψi hψ|sys ⊗ |Ψi hΨ|env Uint
  error occurred, i. e., side-information about positions of er-
Aj |ψi hψ|sys A†j .
X
= (2) rors is available for the decoding process. Note that we
j have increased the dimension of the Hilbert space of the

208
system by one adding the state |ei. Alternatively, we may the remaining phase-flip errors into sign-flip errors. The
use any state of the original space instead of |ei and de- Hadamard transform changes the code state (3) into
scribe the positions of errors by other means.
1 X
H ⊗n |ψj i = p (−1)c·wj |ci . (5)
⊥
|C | c∈C ⊥
3.2 Code Constructions
P
In this section, we will briefly describe several construc- Here c · wj is the standard inner product x · y = i xi yi .
tions of quantum error-correcting codes based on classical Again, any superposition of states (5) is a superposition
linear error-correcting codes. As discussed above, for qubit of quantum states corresponding to codewords of the dual
systems it is sufficient to be able to correct any error that is code C ⊥ . Hence the errors can be corrected in the same
a tensor product of identity and Pauli matrices. The weight manner. The last step is another Hadamard transform re-
of such an error (or the number of errors) is defined as the turning to the original basis.
number of tensor factors different from identity. Moreover, A generalization of this construction was given in [8] and
as σy = iσx σz we can restrict ourselves to no-error, σx - [5]. It is based on the algebraic properties of the group
errors, σz -errors, and combinations of them. The opera- generated by tensor products of Pauli matrices (see also
tor σx interchanges the states |0i and |1i. Hence, it corre- [3]). Here we will only present the prerequisites and the
sponds to a classical bit-flip error. The operator σz changes parameters of the resulting quantum codes. Furthermore,
the relative phase of |0i and |1i and has no classical coun- we restrict ourselves to linear codes (in contrast to additive
terpart. But the operator
√ σz interchanges
√ the orthogonal codes).
states (|0i + |1i)/ 2 and (|0i − |1i)/ 2, i. e., it acts as a
bit-flip with respect to this basis. Hence, the corresponding Construction 3.2 (Quaternary Codes)
change of basis—the Hadamard transform H (see equation By x we denote the conjugation x 7→ x2 =: x in the field
(1))—interchanges bit-flip and phase-flip errors: F4 = GF (4) = {0, 1, ω, ω = ω 2 = ω + 1}. Furthermore,
for a linear space C ≤ Fn4 , by C ∗ we denote the linear
Hσx H = σz and Hσz H = σx . space that
Pis orthogonal with respect to the inner product
x · y := j xj yj .
In summary, this enables us to use certain classical linear Let C = [n, k, d] be a self-orthogonal linear quaternary
binary codes for the construction of quantum codes. code, i. e., C is contained in C ∗ = [n, n − k, d∗ ].
The following construction is due to [17, 18] and [6]. More Then a quantum error-correcting code C = [[n, n − 2k]] of
details (and proofs) can also be found in [3, 10]. length n and dimension 2n−2k exists. Based on classical
Construction 3.1 (Binary Codes) decoding algorithms for the code C ∗ , up to (d∗ − 1)/2
Let C = [n, k, d] be a weakly self-dual linear binary code, errors can be corrected. Moreover, the code can correct
i. e., C is contained in its dual C ⊥ = [n, n − k, d⊥ ]. Fur- errors up to weight (d′ − 1)/2 where
thermore, let {wj : 0 ≤ j ≤ 2n−2k } be a system of coset
d′ = min{wgt c : c ∈ C ∗ \ C} ≥ d∗ . (6)
representatives of C ⊥ /C.
Then the 2n−2k mutually orthogonal states Note that C ⊥ and C ∗ are related by conjugation and thus
1 X d∗ = d⊥ .
|ψj i = p |c + w j i (3)
|C| c∈C Recently, it has been shown how to use linear codes over
any finite field of characteristic two, i. e., fields F2ℓ with 2ℓ
span a quantum error-correcting code C = [[n, n − 2k]] of elements for the construction of quantum error-correcting
length n and dimension 2n−2k (The notation is similar to codes [12]. Again, we only present the main parameters of
that for classical linear block codes.) Based on classical the construction.
decoding algorithms for the code C ⊥ , up to (d⊥ − 1)/2 Construction 3.3 (Codes from Extension Fields)
errors can be corrected. Moreover, the code can correct Let C = [n, k, d] be a weakly self-dual code over F2ℓ , i. e.,
errors up to weight (d′ − 1)/2 where C is contained in its dual C ⊥ = [n, n−k, d⊥ ] (with respect
to the standard inner product). Furthermore, let B be a
d′ = min{wgt c : c ∈ C ⊥ \ C} ≥ d⊥ . (4)
self-dual basis of F2ℓ over F2 .
Expanding each element of F2ℓ with respect to the ba-
The outline of the decoding process is as follows: Any su-
sis B yields a weakly self-dual linear binary code C2 =
perposition of code states |ψj i is a superposition of quan-
[ℓn, ℓk, d2 ≥ d]. Its dual C2⊥ = [ℓn, ℓ(n − k), d⊥ ⊥
2 ≥ d ] is
tum states corresponding to codewords of the dual code
obtained in the same manner.
C ⊥ . A (correctable) bit-flip error takes the superposition of
codewords into a superposition of the corresponding coset. Based on the classical codes C2 and C2⊥ , a quantum error-
Similar to classical decoding algorithms, this coset can be correcting code can be obtained using Construction 3.1.
identified by computing an error syndrome using auxiliary The resulting quantum code can be decoded as a binary
qubits. Measuring this syndrome reveals information about code or as a code over the field F2ℓ . In the latter case, ℓ
the error, but not about the original superposition. After qubits are grouped into one block, and errors can be cor-
correction of the bit-flip errors, a Hadamard transform turns rected if they are restricted to up to (d⊥ − 1)/2 blocks.

209
4 Quantum BCH Codes i. e., the zero set of the orthogonal code is given by
The quantum version of binary BCH codes was introduced ZC ∗ = {−2z mod n : z ∈ {0, . . . , n − 1} \ ZC } .
in [11]. In [5], the term quantum BCH code was used for
quaternary quantum BCH codes (see Construction 3.2). In
Corollary 4.3 A BCH code is weakly self-dual if and only
the context of [11], for the quantum erasure channel, it
if ZC ⊥ ⊆ ZC or, equivalently,
is important to use codes that allow the use of the side-
information on the positions of the errors provided by the ∀z : (z ∈ ZC ⊥ ⇒ (−z mod n) ∈
/ ZC ) .
channel. For BCH codes, a variety of such decoding algo-
rithms exists. Being cyclic codes, BCH codes allow also A BCH code over F4 is self-orthogonal if and only if
decoding based on spectral techniques. This is in particu- ZC ∗ ⊆ ZC or, equivalently,
lar true for Reed-Solomon (RS) codes where no field ex-
∀z : z ∈ ZC ∗ ⇒ (−2−1 z mod n) ∈


tension is needed to implement the Fourier transform. The / ZC .
quantum version of RS codes and their spectral decoding
is discussed in [12]. Another technique for encoding and A lower bound for the minimum distance of a BCH code—
decoding cyclic codes is based on linear shift registers (see and in turn for the corresponding QBCH code—can be
[9]). derived from its zero set.
In the sequel, we focus on the definition and the compu- Theorem 4.4 (BCH bound) If the zero set ZC ⊥ of the
tation of the parameters of quantum BCH codes, supple- dual of a weakly self-dual BCH code C contains dBCH − 1
mented by examples in Section 5. A good reference for the consecutive numbers, i. e.,
theory of classical error-correcting codes is [14]. All the-
z0 +d BCH −2
orems below can be found in a similar version in [11] and [
Cz ⊆ ZC ⊥ , (7)
[5], we will omit the proofs.
z=z0
Definition 4.1 (QBCH Codes)
A quantum BCH code (QBCH code) is a quantum error- then the minimum distance d⊥ of C ⊥ is at least dBCH .
correcting code that is derived from a classical, weakly On the other hand, if a BCH code is specified by the left
self-dual (respectively self-orthogonal) BCH code using hand side of equation (7), dBCH is called the designed dis-
Construction 3.1, 3.2, or 3.3. tance.
Usually, BCH codes are specified by the zero sets, i. e., the The actual minimum distance of a BCH code may be larger
exponents of the roots αz of their generator polynomial than dBCH . This yields another lower bound for the error
g(X)|X n − 1 where α is a primitive n-th root of unity. correcting capability of the QBCH code.
For a BCH code over the field Fq , the zero set is a union
of cyclotomic cosets modulo n closed under multiplication Theorem 4.5 (Code bound) The minimum distance of a
by q, i. e., QBCH code is at least the minimum distance d⊥ of the
dual C ⊥ = [n, n − k, d⊥ ] of the underlying BCH code.
[
ZC = Cz where Cz = {q i z mod n : i ≥ 0}. According to equations (4) and (6), the true minimum dis-
z tance of a QBCH code may be even larger, see the exam-
ples in the next section.
The zero sets of a code and its dual are related as follows.
Theorem 4.2 Let ZC denote the zero set of a BCH code
C over the field Fq , i. e., the generator polynomial of C is 5 Examples
given by Finally, we present the main results of this paper. Using
Y
g(X) = (X − αz ). the computer algebra system MAGMA [4], we have com-
z∈ZC puted the parameters for QBCH codes derived from classi-
cal BCH codes over various fields (see Tables 1–6).
Then the generator polynomial of the dual code C ⊥ is
given by In Table 1 parameters of binary QBCH codes are given. A
Y noteable code is the one with parameters C = [[49, 1, 9]].
h(X) = (X − α−z ), The corresponding BCH code is C ⊥ = [49, 25, 4] and C =
z∈{0,...,n−1}\ZC [49, 24, 4] is the even weight subcode of C ⊥ . Therefore,
d′ = min{wgt c : c ∈ C ⊥ \ C} must be odd. Computing
i. e., the zero set of the dual code is given by the weight distribution of C ⊥ , we obtain d′ = 9.

ZC ⊥ = {−z mod n : z ∈ {0, . . . , n − 1} \ ZC } . Similarly, for the code C = [[89, 1, 17]] the BCH bound
yields d′ ≥ 7, whereas the actual minimum distance of the
For codes over F4 , the generator polynomial of the orthog- BCH code C ⊥ is d⊥ = 12. Again, C = [89, 44, 12] is
onal code C ∗ is given by the even weight subcode of C ⊥ , hence d′ ≥ 13 and d′ is
Y odd. Sampling codewords at random, we find d′ ≤ 17.
h(X) = (X − α−2z ), Moreover, using MAGMA we were able to show that in-
z∈{0,...,n−1}\ZC deed d′ = 17.

210
Quaternary QBCH codes are listed in Table 2. Here are [[5, 1, 3]] [[29, 1, 11]] [[45, 41, 2]]
the codes C = [[25, 1, 9]] and C = [[35, 1, 9]] of special [[7, 1, 3]] [[31, 1, 7]] [[47, 1, 11]]
interest. For the first code, the BCH bound yields d′ ≥ 4, [[31, 11, 5]]
the minimum distance of both C and C ∗ is d = 8, but the [[13, 1, 5]] [[49, 1, 9]]c
[[31, 21, 3]]
minimum distance of the quantum code is d′ = 9. For [[15, 3, 5]] [[49, 7, 3]]
[[35, 1, 9]]b [[49, 43, 2]]
C = [[35, 1, 9]], we obtain dBCH = 5, d = 8, and d′ = 9. [[15, 7, 3]] [[35, 13, 7]]
[[15, 11, 2]] [[35, 25, 4]] [[51, 3, 11]]
Finally, in Tables 3–6 we present QBCH codes constructed
from BCH codes over fields of size 8, 16, 32, and 64. [[17, 1, 7]] [[35, 31, 2]] [[51, 19, 9]]
[[17, 9, 4]] [[51, 27, 6]]
The corresponding binary codes are obtained by expand- [[37, 1, 11]]
[[51, 35, 3]]
ing each element of the extension field with respect to a [[21, 3, 5]] [[39, 3, 9]] [[51, 43, 2]]
fixed self-dual basis. For these codes, we have listed both [[21, 9, 3]] [[39, 15, 3]]
the minimum distance d2 as binary code and the minimum [[21, 15, 2]] [[39, 27, 2]] [[53, 1, 15]]
distance dq as code over the field Fq which is relevant for [[55, 31, 5]]
[[23, 1, 7]] [[41, 1, 11]]
blockwise decoding. [[55, 35, 3]]
[[25, 1, 9]] a [[41, 21, 6]]
[[55, 51, 2]]
[[7, 1, 3]] [[69, 3, 11]] [[95, 23, 5]] [[25, 5, 3]] [[45, 17, 5]]
[[69, 25, 3]] [[25, 21, 2]] [[45, 29, 3]] [[61, 1, 17]]
[[15, 7, 3]] [[69, 47, 2]] [[103, 1, 19]]
Table 2: Parameters of some quaternary QBCH codes
[[21, 3, 5]] [[71, 1, 11]] [[105, 37, 9]] given in the form [[n, k, d]].
[[21, 9, 3]] [[105, 45, 7]]
[[21, 15, 2]] [[73, 19, 9]]
[[105, 61, 5]]
[[73, 37, 6]] [[21, 15, 2|2]] [[69, 3, 7|7]] [[117, 93, 3|3]]
[[105, 75, 4]]
[[23, 1, 7]] [[73, 55, 3]] [[21, 9, 3|3]] [[117, 69, 3|3]]
[[105, 91, 3]] [[93, 3, 7|7]]
[[31, 1, 7]] [[75, 35, 3]] [[105, 99, 2]] [[45, 21, 3|3]] [[93, 33, 5|5]] [[135, 87, 5|5]]
[[31, 11, 5]] [[75, 67, 2]] [[93, 63, 3|3]] [[135, 111, 3|3]]
[[111, 39, 3]] [[63, 27, 6|5]]
[[31, 21, 3]] [[77, 11, 6]] [[63, 21, 5|5]] [[105, 27, 6|5]]
[[35, 5, 6]] [[77, 17, 3]] [[115, 5, 14]] [[63, 33, 4|4]] [[105, 51, 5|4]]
[[35, 11, 3]] [[77, 71, 2]] [[115, 27, 5]] [[63, 45, 3|3]] [[105, 75, 3|3]]
[[35, 29, 2]] [[115, 93, 2]]] [[63, 57, 2|2]] [[105, 99, 2|2]]
[[79, 1, 15]]
[[39, 15, 3]] [[85, 53, 5]] [[117, 45, 9]] Table 3: Parameters of some binary quantum codes derived
[[85, 69, 3]] [[117, 69, 7]] from BCH codes over F8 given in the form [[n, k, d2 |d8 ]].
[[45, 13, 5]] [[117, 93, 3]] Binary expansion with respect to the self-dual basis B8 =
[[45, 21, 3]] [[87, 31, 3]]
[[119, 23, 7]] (u3 , u6 , u5 ) where u3 = u + 1.
[[45, 37, 2]] [[89, 1, 17]]b
[[119, 65, 6]]
[[47, 1, 11]] [[89, 23, 11]] [[20, 12, 2|2]] [[60, 28, 6|5]] [[92, 4, 7|7]]
[[119, 71, 3]]
[[89, 45, 7]] [[60, 36, 4|4]]
[[119, 113, 2]]
[[49, 1, 9]]a [[89, 67, 4]] [[28, 4, 3|3]]
[[60, 44, 3|3]] [[100, 12, 9|6]]a
[[49, 7, 3]] [[123, 83, 3]] [[100, 52, 4|3]]
[[91, 43, 7]] [[36, 4, 4|3]] [[60, 52, 2|2]]
[[49, 43, 2]] [[100, 92, 2|2]]
[[91, 67, 3]] [[127, 1, 19]] [[36, 28, 2|2]] [[76, 4, 7|7]]
[[51, 35, 3]] [[91, 85, 2]] [[127, 15, 16]] [[44, 4, 6|5]] [[84, 4, 6|6]] [[108, 28, 4|4]]b
[[55, 15, 5]] [[93, 13, 12]] [[127, 29, 15]] [[108, 100, 2|2]]
[[52, 4, 7|6]] [[84, 28, 5|5]]
[[93, 23, 9]] [[127, 43, 13]]
[[52, 28, 4|4]] [[84, 36, 4|3]]
[[63, 27, 7]] [[93, 33, 8]] [[127, 57, 11]] [[116, 4, 15|11]]c
[[84, 52, 3|3]]
[[63, 39, 5]] [[93, 43, 7]] [[127, 71, 9]] [[60, 12, 8|7]] [[116, 60, 7|6]]
[[84, 76, 2|2]]
[[63, 45, 4]] [[93, 63, 5]] [[127, 85, 7]]
[[63, 51, 3]] [[93, 73, 3]] [[127, 99, 5]] Table 4: Parameters of some binary quantum codes derived
[[63, 57, 2]] [[93, 83, 2]] [[127, 113, 3]] from BCH codes over F16 given in the form [[n, k, d2 |d16 ]].
Binary expansion with respect to the self-dual basis B16 =
Table 1: Parameters of some binary QBCH codes given in
(v 3 , v 7 , v 13 , v 12 ) where v 4 = v + 1.
the form [[n, k, d]].

6 Acknowledgments
The authors would like to thank Rainer Steinwandt for
his comments. This work was supported by Deutsche
Forschungsgemeinschaft (DFG), Schwerpunktprogramm
Quanten-Informationsverarbeitung (SPP 1078), Projekt
AQUA (Be 887/13-1).

211
 [[35, 5, 3|3]] [[105, 45, 3|3]] [[175, 25, 6|6]] [[42, 18, 3|3]] [[66, 6, 6|5]] [[126, 54, 8|7]]
[[105, 75, 2|2]] [[175, 55, 3|3]] [[42, 30, 2|2]] [[126, 78, 6|5]]
[[75, 35, 3|3]] [[90, 54, 3|3]]
[[175, 145, 2|2]] [[126, 90, 4|4]]
[[115, 5, 7|7]] [[54, 18, 6|4]] [[90, 78, 2|2]]
[[105, 15, 5|5]] [[126, 102, 3|3]]
[[54, 30, 4|3]] [[114, 42, 8|6]] [[126, 114, 2|2]]
Table 5: Parameters of some binary quantum codes derived [[54, 42, 2|2]] [[114, 78, 6|4]]
from BCH codes over F32 given in the form [[n, k, d2 |d32 ]].
Binary expansion with respect to the self-dual basis B32 = Table 6: Parameters of some binary quantum codes derived
(w9 , w18 , w5 , w10 , w20 ) where w5 = w2 + 1. from BCH codes over F64 given in the form [[n, k, d2 |d64 ]].
Binary expansion with respect to the self-dual basis B64 =
(z 12 , z 24 , z 48 , z 33 , z 3 , z 6 ) where z 6 = z 4 + z 3 + z + 1.

7 References
[1] Barenco, A., Bennett, C. H., Cleve, R., DiVincenzo, [11] Grassl, M., Beth, T., and Pellizzari, T. Codes for
D. P., Margolus, N., Shor, P., Sleator, T., Smolin, the Quantum Erasure Channel. Physical Review A,
J. A., and Weinfurter, H. Elementary gates for quan- 56(1):33–38, July 1997. See also LANL preprint
tum computation. Physical Review A, 52(5):3457– quant-ph/9610042.
3467, Nov. 1995. See also LANL preprint quant-
ph/9503016. [12] Grassl, M., Geiselmann, W., and Beth, T. Quantum
Reed-Solomon Codes. In Proceedings AAECC-13,
[2] Bennett, C. H., DiVincenzo, D. P., Smolin, J. A., 1999. To appear.
and Wootters, W. K. Mixed State Entanglement
and Quantum Error Correction. Physical Review A, [13] Grover, L. K. A fast quantum mechanical algo-
54(5):3824–3851, Nov. 1996. rithm for database search. In Proceedings of the 28th
Annual ACM Symposium on Theory of Computing
[3] Beth, T. and Grassl, M. The Quantum Hamming and (STOC), pp. 212–219, New York, 1996. ACM. See
Hexacodes. Fortschritte der Physik, 46(4–5):459– also LANL preprint quant-ph/9605043.
491, 1998.
[14] MacWilliams, F. J. and Sloane, N. J. A. The Theory
[4] Bosma, W., Cannon, J., and Playoust, C. The Magma of Error-Correcting Codes. North-Holland, Amster-
Algebra System I: The User Language. Journal of dam, 1977.
Symbolic Computation, 24(3–4):235–266, 1997.
[15] Rötteler, M. and Beth, T. Efficient Realisation of Dis-
[5] Calderbank, A. R., Rains, E. M., Shor, P. W., and crete Cosine Transforms on a Quantum Computer. In
Sloane, N. J. A. Quantum Error Correction Via Codes Proceedings ISTET 99, Magdeburg, 1999.
over GF(4). IEEE Transactions on Information The-
[16] Shor, P. W. Polynomial-Time Algorithms for Prime
ory, IT–44(4):1369–1387, July 1998. See also LANL
Factorization and Discrete Logarithms on a Quantum
preprint quant-ph/9608006.
Computer. SIAM Journal on Computing, 26(5):1484–
[6] Calderbank, A. R. and Shor, P. W. Good quan- 1509, 1997.
tum error-correcting codes exist. Physical Review A,
[17] Steane, A. Error Correcting Codes in Quantum The-
54(2):1098–1105, Aug. 1996. See also LANL pre-
ory. Physical Review Letters, 77(5):793–797, July
print quant-ph/9512032.
1996.
[7] Dirac, P. M. A. The Principles of Quantum Mechan-
[18] Steane, A. Multiple Particle Interference and Quan-
ics. Clarendon Press, Oxford, 4th edition, 1958.
tum Error Correction. Proceedings of the Royal Soci-
[8] Gottesman, D. A Class of Quantum Error-Correcting ety London Series A, 452:2551–2577, 1996. See also
Codes Saturating the Quantum Hamming Bound. LANL preprint quant-ph/9601029.
Physical Review A, 54(3):1862–1868, Sept. 1996. See
[19] Wootters, W. K. and Zurek, W. H. A single quantum
also LANL preprint quant-ph/9604038.
cannot be cloned. Nature, 299(5886):802–803, Oct.
[9] Grassl, M. and Beth, T. Codierung und Decodie- 1982.
rung zyklischer Quantencodes. In Michaelis, B. and
Holub, H., editors, Fachtagung Informations- und
Mikrosystemtechnik, pp. 137–144, Magdeburg, Mar.
1998.

[10] Grassl, M. and Beth, T. Relations between Classical

and Quantum Error-Correcting Codes. In Kluge, W.,
editor, Workshop “Physik und Informatik”, pp. 45–
58, DPG-Frühjahrstagung, Heidelberg, 1999.

212