Quantum MDS Convolutional Codes

• Quantum MDS convolutional codes are quantum stabilizer codes that achieve the quantum Singleton bound for free distance, ensuring optimal error protection over memory channels.
• They are constructed using Hermitian dual-containing block codes, such as constacyclic, BCH, and GRS codes, to support efficient online error correction.
• Recent advancements have expanded parameter ranges and removed field restrictions, enhancing the feasibility of real-time quantum communication.

A quantum MDS (maximally distance-separable) convolutional code is a quantum convolutional stabilizer code that attains the quantum Singleton bound for free distance, maximizing the code’s ability to protect quantum information over memory channels. These codes generalize the MDS property from block codes to the convolutional setting and are constructed to support efficient, online, framewise quantum error correction with optimal error tolerance. Recent research has produced infinite families of such codes using constructions based on Hermitian dual-containing constacyclic, BCH, and generalized Reed–Solomon (GRS) block codes, with parameters explicitly given and proofs that they attain the quantum Singleton bound (Zhang et al., 2014, Guardia, 2012, Ding et al., 2015).

1. Formalism and Quantum Singleton Bound

A quantum convolutional code of frame size $n$, $k$ logical qudits per frame, memory $\mu$, overall degree $\delta$, and free distance $d_f$ is specified by a full-rank polynomial stabilizer matrix

$$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$

satisfying the symplectic orthogonality

$X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$

The free distance $d_f$ is the minimum weight of any undetectable error, generalizing the block-code minimum distance to the convolutional context.

For any pure $[(n, k, \mu; \delta, d_f)]_q$ quantum convolutional code, the free distance must satisfy the quantum Singleton bound: $$d_f \leq \frac{n-k}{2} \left( \left\lfloor \frac{2\delta}{n+k} \right\rfloor + 1 \right) + \delta + 1.$$ When equality is achieved, the code is quantum MDS (Zhang et al., 2014), Lemma 2.7.

2. Quantum Convolutional Code Constructions

Modern constructions of quantum MDS convolutional codes employ a CSS-like framework using classical convolutional codes derived from Hermitian dual-containing block codes.

Key Construction Steps:

1. Select a classical block code $k$0 over $k$1 that is Hermitian dual-containing ($k$2), such as constacyclic, BCH, or GRS codes.
2. Split the parity-check matrix $k$3 of $k$4 appropriately (e.g., $k$5).
3. Form a classical convolutional generator $k$6, typically as $k$7 (unit-memory) or $k$8 (multi-memory).
4. If the resulting convolutional code $k$9 satisfies $\mu$0, construct the quantum convolutional code via the CSS method, achieving the desired parameters (Zhang et al., 2014, Guardia, 2012, Ding et al., 2015).

3. Explicit Families and Parameters

Three major frameworks dominate recent constructions.

(A) Constacyclic–based Families (Zhang et al., 2014):

• Family I: $\mu$1 for odd $\mu$2, $\mu$3.
• Family II: $\mu$4 for $\mu$5 or $\mu$6 ($\mu$7), $\mu$8.

(B) BCH–based Unit-Memory Codes (Guardia, 2012):

• For $\mu$9 and $\delta$0,

$\delta$1

where $\delta$2.

(C) Dual-Containing GRS–Based Constructions (Ding et al., 2015):

• Memory $\delta$3: $\delta$4, for $\delta$5 between $\delta$6 and $\delta$7.
• Memory $\delta$8: $\delta$9, for suitable $d_f$0.

These families systematically cover a broad range of lengths, rates, and free distances, and always saturate the Singleton bound.

Family	Parameters	Memory	Degree	Free Distance (MDS) Bound
(Zhang et al., 2014) Family I	$d_f$1	1	2	$d_f$2
(Zhang et al., 2014) Family II	$d_f$3	1	2	$d_f$4
(Guardia, 2012) BCH-based	$d_f$5	1	2	$d_f$6
(Ding et al., 2015) GRS MDS (p=1)	$d_f$7	1	$d_f$8	$d_f$9
(Ding et al., 2015) GRS MDS (p=2)	$$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$0	2	2	$$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$1

4. Proofs of the MDS Property

All these families achieve the quantum Singleton bound by rigorous arguments:

• The base block code ($$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$2) is constructed to be MDS and Hermitian dual-containing (e.g., via the defining set $$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$3 in constacyclic/BCH cases or the evaluation structure of GRS codes).
• The convolutional code $$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$4 inherits MDS properties by splitting the block code parity-check matrix and forming a suitable generator $$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$5.
• The Hermitian dual-containment ensures purity and self-orthogonality, pivotal for quantum stabilizer code construction.
• The final quantum code’s free distance directly saturates the Singleton bound by parameter analysis and explicit computation (see, e.g., $$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$6 in the BCH case) (Zhang et al., 2014, Guardia, 2012, Ding et al., 2015).

5. Generalizations and Advances Over Previous Work

Recent constructions generalize and improve upon earlier results:

• The constacyclic construction in (Zhang et al., 2014) removes the prior requirement $$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$7 imposed in earlier work, allowing $$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$8 to be any odd prime power $$S(D) = [X(D) \mid Z(D)] \in \mathbb{F}_q[D]^{(n-k) \times 2n}$$9 (Family I) and $X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$0 or $X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$1 (Family II).
• The BCH framework (Guardia, 2012) produces quantum MDS convolutional codes for all even $X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$2, for a broad range of parameters, using an entirely algebraic approach with no computer search.
• The GRS-based method (Ding et al., 2015) generates eighteen new families—including both memory one and memory two codes—covering a diverse spectrum of code lengths and rates, with explicit worked examples.

Compared to other approaches (e.g., Forney–Guha–Grassl, Almeida–Palazzo, Wilde–Brun), these algebraic methods yield codes with larger free distances relative to frame size and support memory one or two, which is optimal for streaming and framewise error-correction in quantum communication channels (Guardia, 2012, Ding et al., 2015).

6. Applications and Open Directions

Quantum MDS convolutional codes are integral to real-time, long-distance quantum communication, particularly in scenarios where memory and online error correction are critical (e.g., quantum repeaters and memory-noisy channels). Current lines of investigation address:

• Extending to higher memory $X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$3 and use of non-GRS (e.g., algebraic-geometry) underlying block codes (Ding et al., 2015).
• Optimization of circuit complexity for encoding and syndrome measurement.
• Development of entanglement-assisted variants to further expand achievable code parameters.

A plausible implication is that, as these algebraic constructions continue to improve, they will fill gaps in the quantum convolutional code tables and facilitate scalable quantum communication infrastructure.

7. Summary Table: Major Quantum MDS Convolutional Code Families

Reference	Underlying Block Code Type	Quantum Parameters Example	Memory	Notable Features
(Zhang et al., 2014)	Constacyclic (Hermitian)	$X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$4 (Family I)	1	Removes $X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$5 restriction
(Guardia, 2012)	BCH (Hermitian)	$X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$6	1	Algebraic, no computer search
(Ding et al., 2015)	Generalized Reed–Solomon	$X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$7, $X(D)Z(1/D)^T - Z(D)X(1/D)^T = 0.$8	1, 2	18 new families, broad parameter coverage