Quantum Reed–Solomon Codes Overview

• Quantum Reed–Solomon codes are quantum error-correcting codes derived from classical Reed–Solomon constructions, offering explicit algebraic methods and MDS properties.
• They employ both CSS-type and Hermitian-stabilizer paradigms to optimize parameters for erasure, deletion, and entanglement-assisted quantum communication.
• They provide flexible lengths and high-distance performance, making them ideal for advanced quantum repeaters, concatenation schemes, and fault-tolerant architectures.

Quantum Reed–Solomon codes are quantum error-correcting codes derived from classical Reed–Solomon or generalized Reed–Solomon (GRS) codes. In the literature, the term covers at least two closely related algebraic paradigms: CSS-type constructions from classical Reed–Solomon pairs over $GF(d)$, which yield high-dimensional codes such as $[[d,2K-d,d-K+1]]_d$, and Hermitian-stabilizer constructions from GRS codes over $\mathbb F_{q^2}$, often formalized as quantum generalized Reed–Solomon (QGRS) codes (Muralidharan et al., 2018, 0812.4514). These codes occupy a central position in quantum MDS theory, in high-dimensional erasure and loss correction, in entanglement-assisted quantum coding, and in communication-oriented architectures ranging from one-way repeaters to concatenation over high-rate quantum LDPC inner codes (Muralidharan et al., 2018, Wills et al., 21 May 2026).

1. Algebraic foundations and terminology

At the classical level, the basic object is the generalized Reed–Solomon code

$$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$

where $a=(a_1,\dots,a_n)$ consists of pairwise distinct field elements and $v=(v_1,\dots,v_n)$ has nonzero entries. It is an $[n,k,n-k+1]$ MDS code. The same framework includes extended versions: an $[n+1,k,n-k+2]$ extended GRS code, and, in the special case $k=3$ with $q$ even, an $[[d,2K-d,d-K+1]]_d$0 extended GRS code (0812.4514).

The Hermitian route to quantum coding starts from a classical code $[[d,2K-d,d-K+1]]_d$1 over $[[d,2K-d,d-K+1]]_d$2 satisfying

$[[d,2K-d,d-K+1]]_d$3

If $[[d,2K-d,d-K+1]]_d$4 is an $[[d,2K-d,d-K+1]]_d$5 code and

$[[d,2K-d,d-K+1]]_d$6

then there exists a quantum code with parameters

$[[d,2K-d,d-K+1]]_d$7

In the QGRS formulation, a quantum code is called a QGRS code precisely when it is obtained from a Hermitian self-orthogonal generalized Reed–Solomon code or extended generalized Reed–Solomon code in this way (0812.4514).

A distinct but compatible line of work uses a CSS construction directly from classical Reed–Solomon codes over $[[d,2K-d,d-K+1]]_d$8. If

$[[d,2K-d,d-K+1]]_d$9

then the resulting quantum Reed–Solomon code has parameters

$\mathbb F_{q^2}$0

This family encodes $\mathbb F_{q^2}$1 logical $\mathbb F_{q^2}$2-level qudits into $\mathbb F_{q^2}$3 physical qudits and has code rate

$\mathbb F_{q^2}$4

This suggests that “Quantum Reed–Solomon code” is best understood as an algebraic family organized around Reed–Solomon evaluation structure, rather than as a single canonical stabilizer construction (Muralidharan et al., 2018).

2. QGRS codes as a unified framework for quantum MDS theory

A central structural result is the exact characterization of when a QGRS code exists. The relevant classical object is the puncture code

$\mathbb F_{q^2}$5

equivalent to the puncture code of Rains. A QGRS code $\mathbb F_{q^2}$6 exists if and only if there is a codeword of weight $\mathbb F_{q^2}$7 in

$\mathbb F_{q^2}$8

where $\mathbb F_{q^2}$9 lists all elements of $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$0 and $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$1 is the all-ones vector. This reduces the quantum existence problem to a classical weight-distribution problem for a puncture code (0812.4514).

The same analysis gives sharp parameter bounds. If $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$2 is an $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$3 QGRS code, then for $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$4, one must have $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$5. Except for the special case $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$6 and $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$7 even, one has $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$8; for $$\mathrm{CGRS}(a,v,k)=\left\{(v_1F(a_1),\dots,v_nF(a_n)):\deg F<k\right\},$$9 and $a=(a_1,\dots,a_n)$0 even, one has $a=(a_1,\dots,a_n)$1. These bounds are tight: explicit examples include

$a=(a_1,\dots,a_n)$2

for all $a=(a_1,\dots,a_n)$3, and also

$a=(a_1,\dots,a_n)$4

in the even-$a=(a_1,\dots,a_n)$5, $a=(a_1,\dots,a_n)$6 case (0812.4514).

The analytic existence theorem complements the puncture-code criterion. It produces Hermitian self-orthogonal GRS codes, and hence QGRS codes, in several regimes: $a=(a_1,\dots,a_n)$7 with $a=(a_1,\dots,a_n)$8; $a=(a_1,\dots,a_n)$9, written more precisely as $v=(v_1,\dots,v_n)$0 with $v=(v_1,\dots,v_n)$1 and $v=(v_1,\dots,v_n)$2, where $v=(v_1,\dots,v_n)$3; and $v=(v_1,\dots,v_n)$4, where $v=(v_1,\dots,v_n)$5. Within this framework, the authors show that all previously known quantum MDS codes can be realized as QGRS codes with the same parameters, including $v=(v_1,\dots,v_n)$6, $v=(v_1,\dots,v_n)$7, $v=(v_1,\dots,v_n)$8 for $v=(v_1,\dots,v_n)$9, families $[n,k,n-k+1]$0 for $[n,k,n-k+1]$1, and several longer families over $[n,k,n-k+1]$2-scale lengths. The literature therefore treats QGRS codes as a unified source of known quantum MDS constructions (0812.4514).

3. Parameter families, flexible lengths, and the limits of the GRS method

After the unified QGRS framework was established, a substantial body of work concentrated on enlarging the available length–distance spectrum while preserving MDS optimality. One early line gave three GRS-based quantum MDS families at lengths $[n,k,n-k+1]$3, $[n,k,n-k+1]$4, and $[n,k,n-k+1]$5. In particular, for any odd prime power $[n,k,n-k+1]$6, the length-$[n,k,n-k+1]$7 family contains

$[n,k,n-k+1]$8

for $[n,k,n-k+1]$9; the length-$[n+1,k,n-k+2]$0 family contains

$[n+1,k,n-k+2]$1

for $[n+1,k,n-k+2]$2 when $[n+1,k,n-k+2]$3; and the length-$[n+1,k,n-k+2]$4 family contains

$[n+1,k,n-k+2]$5

for $[n+1,k,n-k+2]$6 (Zhang et al., 2015).

Subsequent constructions emphasized long lengths and minimum distances larger than $[n+1,k,n-k+2]$7. One paper produced six new classes of $[n+1,k,n-k+2]$8-ary quantum MDS codes from Hermitian self-orthogonal GRS codes, with three of the six classes having longer lengths than earlier constructions and with minimum distances that can be larger than $[n+1,k,n-k+2]$9 (Fang et al., 2018). Another paper gave three new classes with lengths

$k=3$0

together with ranges of $k=3$1 for which $k=3$2 quantum MDS codes exist; it also provided the explicit example

$k=3$3

highlighting both flexible length and large minimum distance (Fang et al., 2019). A later construction added five new classes of $k=3$4-ary quantum MDS codes with minimum distance larger than $k=3$5, and stated that the resulting parameters cannot be obtained from previous constructions (Wan et al., 2023).

At the same time, the literature also established a sharp obstruction for the Hermitian-dual-containing GRS mechanism. It was proved that there exists a quantum MDS code

$k=3$6

for all $k=3$7, $k=3$8, but that if $k=3$9, then there is no generalized Reed–Solomon $q$0 code which contains its Hermitian dual. In other words, the standard GRS/Hermitian-duality method stops at distance $q$1. The same work constructed sporadic non-GRS quantum MDS codes

$q$2

which were presented as the first quantum MDS codes discovered for which $q$3, apart from the $q$4 code derived from Glynn’s code (Ball, 2019). A plausible implication is that the Reed–Solomon method is both unusually powerful and algebraically rigid: it dominates the known MDS landscape up to its natural boundary, but genuinely new mechanisms are required beyond that boundary.

4. Entanglement-assisted, projective, and convolutional extensions

Entanglement assistance relaxes the self-orthogonality constraint and replaces it with rank control of a Hermitian Gram matrix. In the standard entanglement-assisted theorem, if $q$5 is an $q$6 classical code with parity-check matrix $q$7, then there exists an EAQEC code

$q$8

When the underlying GRS code is MDS and the parameters satisfy the EA Singleton equality, the result is an EAQMDS code (Jin et al., 2019).

This principle has been developed in several directions. One construction over finite fields of odd characteristic produced two families of EAQMDS codes by arranging the evaluation points and multipliers of $q$9 so that $[[d,2K-d,d-K+1]]_d$00 becomes an explicit small integer $[[d,2K-d,d-K+1]]_d$01. The resulting families have the form

$[[d,2K-d,d-K+1]]_d$02

or equivalently $[[d,2K-d,d-K+1]]_d$03, with many new lengths and in some cases larger minimum distance than known results (Jin et al., 2019). A complementary paper gave a complete and explicit formula for EAQECC parameters coming from any Reed–Solomon code in the Hermitian metric. For $[[d,2K-d,d-K+1]]_d$04, it derived, for example,

$[[d,2K-d,d-K+1]]_d$05

when $[[d,2K-d,d-K+1]]_d$06, and a different closed form when $[[d,2K-d,d-K+1]]_d$07, thereby making the entanglement cost $[[d,2K-d,d-K+1]]_d$08 completely explicit for the full Hermitian RS family (Galindo et al., 2020).

A separate line constructed three classes of EAQMDS codes with lengths

$[[d,2K-d,d-K+1]]_d$09

and corresponding entanglement consumption $[[d,2K-d,d-K+1]]_d$10, $[[d,2K-d,d-K+1]]_d$11, and $[[d,2K-d,d-K+1]]_d$12. These lengths were emphasized because some are not divisors of $[[d,2K-d,d-K+1]]_d$13, unlike many earlier EAQMDS constructions (Zheng et al., 2022). Projective variants further generalized the Reed–Solomon setting by studying subfield subcodes of projective Reed–Solomon codes and their duals, leading to symmetric, asymmetric, and Hermitian EAQECCs, often with $[[d,2K-d,d-K+1]]_d$14. Explicit examples include

$[[d,2K-d,d-K+1]]_d$15

(Gimenez et al., 2023).

The Reed–Solomon paradigm also extends beyond block stabilizer codes. Two constructions of quantum MDS convolutional codes derived from Hermitian dual-containing GRS codes yielded eighteen new classes. The two principal parameter templates are

$[[d,2K-d,d-K+1]]_d$16

and

$[[d,2K-d,d-K+1]]_d$17

demonstrating that the MDS property of GRS constituents can be transferred into the streaming setting (Ding et al., 2015). More recently, generalized Monomial Cartesian Codes were introduced as a natural extension of generalized Reed–Solomon codes, constructed by combining two different generalized Reed–Solomon codes into a bivariate Cartesian evaluation code with sufficient Hermitian self-orthogonality conditions for new quantum stabilizer constructions (Campion et al., 18 Dec 2025).

5. Communication-oriented uses: erasures, deletions, repeaters, and concatenation

Quantum Reed–Solomon codes are particularly prominent in communication-dominated noise models. In one-way quantum repeater proposals, the CSS family

$[[d,2K-d,d-K+1]]_d$18

was used as the encoding layer for $[[d,2K-d,d-K+1]]_d$19-level systems. These codes can correct up to $[[d,2K-d,d-K+1]]_d$20 erasures, and for a quantum erasure channel with erasure probability $[[d,2K-d,d-K+1]]_d$21, the success probability is

$[[d,2K-d,d-K+1]]_d$22

The same work stated that quantum Reed–Solomon codes constructed from classical Reed–Solomon codes can approach the capacity on the quantum erasure channel of $[[d,2K-d,d-K+1]]_d$23-level systems for large dimension $[[d,2K-d,d-K+1]]_d$24, and reported a reduction in the repeater cost coefficient by about a factor of $[[d,2K-d,d-K+1]]_d$25 relative to quantum parity codes for communication distances up to $[[d,2K-d,d-K+1]]_d$26 in the loss-dominated regime (Muralidharan et al., 2018).

The same erasure-correction strength has been repurposed for synchronization errors. A construction of quantum deletion-correcting codes derived from quantum Reed–Solomon codes introduced two procedures, Alternating Sandwich Mapping $[[d,2K-d,d-K+1]]_d$27 and Block Error Locator $[[d,2K-d,d-K+1]]_d$28, to reduce deletion correction to erasure correction block by block. If $[[d,2K-d,d-K+1]]_d$29 denotes the underlying quantum Reed–Solomon code, then the overall encoder and decoder are

$[[d,2K-d,d-K+1]]_d$30

and the main theorem states

$[[d,2K-d,d-K+1]]_d$31

for any deletion pattern $[[d,2K-d,d-K+1]]_d$32 of size at most $[[d,2K-d,d-K+1]]_d$33. The rate can be made asymptotically flexible: for any real number $[[d,2K-d,d-K+1]]_d$34, there exist quantum deletion-correcting codes correcting $[[d,2K-d,d-K+1]]_d$35 or fewer deletions whose rates converge to $[[d,2K-d,d-K+1]]_d$36. The paper also emphasized that the decoder does not need the exact number of deletions, only an upper bound $[[d,2K-d,d-K+1]]_d$37 (Hagiwara, 2023).

A modern systems-level application uses quantum Reed–Solomon codes as outer codes in concatenation over a high-rate quantum LDPC inner code. In that setting, each 144-qubit gross code block with parameters $[[d,2K-d,d-K+1]]_d$38 is packaged as one Galois qudit of dimension

$[[d,2K-d,d-K+1]]_d$39

and the outer code is a qRS code with parameters

$[[d,2K-d,d-K+1]]_d$40

The construction exploits list decoding, a Galois-qudit Shor scheme for syndrome extraction, and “time-like” Reed–Solomon protection against measurement errors. At uniform $[[d,2K-d,d-K+1]]_d$41 physical noise, the concatenated gross code reaches the teraquop regime and achieves lower space overhead than the 288-qubit two-gross code down to a logical error rate of $[[d,2K-d,d-K+1]]_d$42 per logical qubit-round (Wills et al., 21 May 2026). This suggests that large-alphabet Reed–Solomon structure remains relevant even when the inner architecture is non-local and highly quantum LDPC in character.

6. Encoding complexity, resource reduction, and intrinsic limitations

Despite their strong coding properties, high-dimensional Quantum Reed–Solomon encoders can be expensive. For the single-logical-qudit family

$[[d,2K-d,d-K+1]]_d$43

with odd prime $[[d,2K-d,d-K+1]]_d$44, the standard encoding circuit requires

$[[d,2K-d,d-K+1]]_d$45

SUM gates. When a $[[d,2K-d,d-K+1]]_d$46-dimensional qudit is implemented with $[[d,2K-d,d-K+1]]_d$47 qubits satisfying $[[d,2K-d,d-K+1]]_d$48, each SUM gate is decomposed into a ripple-carry adder and a modulo conversion, and the modular-adder construction requires

$[[d,2K-d,d-K+1]]_d$49

auxiliary qubits. In the example $[[d,2K-d,d-K+1]]_d$50, a single SUM gate requires $[[d,2K-d,d-K+1]]_d$51 CX gates under a general decomposition, but only $[[d,2K-d,d-K+1]]_d$52 CX gates under a multiplexed decomposition based on time-bin and polarization encoding (Nishio et al., 2022).

The multiplexing result is technically specific. In that framework, a $[[d,2K-d,d-K+1]]_d$53 gate can be replaced by one CX gate if the controls are distributed over a multiplexed photon, and a $[[d,2K-d,d-K+1]]_d$54 gate can be reduced to a single $[[d,2K-d,d-K+1]]_d$55 gate. The benefit is largest when $[[d,2K-d,d-K+1]]_d$56, because the modulo logic then contains many $[[d,2K-d,d-K+1]]_d$57 and $[[d,2K-d,d-K+1]]_d$58 gates. However, the same paper identified a family

$[[d,2K-d,d-K+1]]_d$59

for which multiplexing provides no gate-count advantage, because the encoder does not require Toffoli or $[[d,2K-d,d-K+1]]_d$60 gates in the first place (Nishio et al., 2022).

Several limitations recur across the literature. The GRS/Hermitian-duality method cannot produce quantum MDS codes with $[[d,2K-d,d-K+1]]_d$61 (Ball, 2019). The deletion-code construction has small relative distance because it targets fixed $[[d,2K-d,d-K+1]]_d$62 while letting the block length grow (Hagiwara, 2023). Repeater advantages depend strongly on the balance between photon loss, coupling efficiency, and operational errors: for low coupling efficiency or sufficiently large gate error, first- or second-generation repeaters can outperform third-generation QRSC schemes (Muralidharan et al., 2018). These are not contradictions within the theory; rather, they delimit where the Reed–Solomon algebraic advantages dominate and where other architectural constraints become decisive.

Across these variants, Quantum Reed–Solomon codes remain defined by a distinctive combination of algebraic explicitness, optimal or near-optimal distance behavior, and compatibility with erasure-dominated quantum communication models. The literature suggests that their enduring importance lies not only in the original MDS constructions, but also in the way Reed–Solomon structure continues to organize modern developments in entanglement assistance, projective generalization, convolutional coding, deletion correction, and large-alphabet fault-tolerant architecture design (0812.4514, Wills et al., 21 May 2026).