Graded Quantum Codes

• Graded quantum codes are quantum error-correcting codes that use explicit grading to organize algebraic data, homological complexes, or local subsystem dimensions.
• They refine classical quantum Singleton bounds by integrating weighted degrees, orbifold singularities, and torsion contributions into parameter tuning.
• Construction techniques span weighted polynomial evaluations on projective hypersurfaces, LDPC homological methods, and mixed-alphabet frameworks for versatile error correction.

Searching arXiv for the cited papers and related terminology. arXiv search query: id:(Shaska, 11 Aug 2025) OR title:"Graded Quantum Codes" OR title:"Weighted Algebraic Geometry" Searching for (Shaska, 11 Aug 2025) and related records. Graded quantum codes are quantum error-correcting codes equipped with an explicit grading that organizes either the algebraic data used to construct the code or the local Hilbert-space structure on which the code is defined. In one formulation, a graded quantum code is a stabilizer/CSS quantum code whose classical and/or homological ingredients share an explicit grading: on the weighted algebraic geometry side, codes come from evaluating graded polynomials on rational points of weighted projective hypersurfaces, while on the homological side, codes arise from chain complexes of graded vector spaces or graded modules, possibly with torsion (Shaska, 11 Aug 2025). In another formulation, graded quantum codes are mixed-alphabet quantum error-correcting codes, in which the physical subsystems have differing local dimensions and the grading is by local dimension (Wang et al., 2012). These two usages are not identical in construction, but both treat grading as a structural device that refines parameters, constraints, and decoding.

1. Conceptual scope and definitions

In the weighted-algebraic and homological framework, a graded quantum code is a stabilizer/CSS quantum code whose classical and/or homological ingredients share an explicit grading (Shaska, 11 Aug 2025). Two complementary instances are developed. The first is the weighted algebraic geometry side, where codes come from evaluating graded, weighted-homogeneous polynomials on rational points of hypersurfaces in weighted projective spaces over finite fields. The grading is the weighted degree of monomials and polynomials, and it interacts with orbifold singularities. The second is the homological side, where codes arise from chain complexes

$C_2 \to C_1 \to C_0$

of graded vector spaces, and more generally graded modules, possibly with torsion, where the boundary maps respect degree and yield LDPC stabilizer codes via the $\partial$-maps (Shaska, 11 Aug 2025).

The key unifying innovation in this framework is a shared grading across code components. On the algebraic geometry side, the weight grading of coordinates and polynomials controls point counts, dimensions, and self-orthogonality conditions for CSS constructions. On the homological side, the chain-complex grading, possibly multi-graded, with torsion, controls sparsity and the ranks of homology groups used to define quantum parameters. This shared grading enables a refined Singleton-type bound that tightens the standard quantum Singleton bound by an explicit positive correction term $\epsilon$ reflecting entropy or defect contributions from orbifold singularities or graded twisted sectors (Shaska, 11 Aug 2025).

In the mixed-alphabet formulation, graded quantum codes are codes defined over physical subsystems of differing local dimensions, and they are precisely the mixed-alphabet quantum error-correcting codes introduced in "Quantum Error-Correcting Codes over Mixed Alphabets" (Wang et al., 2012). A mixed-alphabet code on $n$ physical subsystems whose local dimensions are not uniform is denoted by

$((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$

or more generally by

$((n,K,d))_{p_1 p_2 \dots p_n}.$

Here $n$ is the number of physical subsystems, $K$ is the dimension of the code subspace, and $d$ is the code distance; the code detects all errors of weight $< d$ and corrects all errors of weight $\partial$0 (Wang et al., 2012).

A plausible implication is that the term "graded quantum codes" functions as an umbrella for at least two structurally different, but conceptually related, uses of grading: grading by weighted degree or homological degree in code construction, and grading by local subsystem dimension in code representation.

2. Weighted algebraic geometry constructions

The weighted algebraic geometry construction begins with a weighted projective space. Let $\partial$1 be positive integers with $\partial$2; such a weight vector is well-formed. The weighted projective space over a field $\partial$3 is

$\partial$4

where $\partial$5 acts by

$\partial$6

Its homogeneous coordinate ring is $\partial$7, graded by $\partial$8, with graded pieces $\partial$9 consisting of weighted homogeneous polynomials of total weighted degree $\epsilon$0 (Shaska, 11 Aug 2025).

A weighted homogeneous polynomial $\epsilon$1 satisfies

$\epsilon$2

and the weighted hypersurface

$\epsilon$3

is the vanishing locus of $\epsilon$4, defined over $\epsilon$5 when the coefficients lie in $\epsilon$6 (Shaska, 11 Aug 2025). Over $\epsilon$7, rational points of $\epsilon$8 are orbits of $\epsilon$9 acting by weighted scaling on $n$0. With support $n$1 equal to the indices of nonzero coordinates, the stabilizer has order $n$2 with $n$3. The point count on the ambient space is

$n$4

For hypersurfaces $n$5, one stratifies by $n$6 and restricts to the zero-locus of $n$7 on each stratum (Shaska, 11 Aug 2025).

Classical evaluation codes are constructed by evaluating a finite-dimensional subspace of functions on a set of $n$8 rational points:

$n$9

Here $((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$0 is often a graded piece $((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$1 or a Riemann–Roch space $((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$2 on a curve embedded in a weighted projective space. Weighting and orbifold singularities affect both the dimension and the zero counts, which later determine the quantum CSS parameters (Shaska, 11 Aug 2025).

For the explicit evaluation code

$((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$3

with $((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$4, the dimension is

$((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$5

where $((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$6 is the weighted vanishing ideal. The minimum distance satisfies

$((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$7

and can be bounded using weighted footprint or Gröbner techniques, adjusting zero multiplicities by stabilizers $((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$8 (Shaska, 11 Aug 2025).

The CSS lift proceeds when a classical code $((n,K,d))_{q_1^{n_1} q_2^{n_2}\dots},$9 under the Euclidean or Hermitian inner product. In that case, CSS yields a quantum code with parameters

$((n,K,d))_{p_1 p_2 \dots p_n}.$0

and

$((n,K,d))_{p_1 p_2 \dots p_n}.$1

In matrix form, for parity-check matrices $((n,K,d))_{p_1 p_2 \dots p_n}.$2, the orthogonality condition is

$((n,K,d))_{p_1 p_2 \dots p_n}.$3

Self-orthogonality can be achieved by choosing degrees or divisors respecting weighted symmetry, residues, or automorphisms; weighted degrees and the canonical class control when $((n,K,d))_{p_1 p_2 \dots p_n}.$4 holds (Shaska, 11 Aug 2025).

A heuristic weighted analogue of Riemann–Roch is

$((n,K,d))_{p_1 p_2 \dots p_n}.$5

where $((n,K,d))_{p_1 p_2 \dots p_n}.$6, $((n,K,d))_{p_1 p_2 \dots p_n}.$7 is the canonical divisor, $((n,K,d))_{p_1 p_2 \dots p_n}.$8 is the weighted degree, and $((n,K,d))_{p_1 p_2 \dots p_n}.$9 is the weighted orbifold genus. Singularities and orbifold points introduce correction terms that affect $n$0 and thus $n$1 and $n$2; in practice one uses that $n$3 when $n$4, adjusted for orbifold defects (Shaska, 11 Aug 2025).

3. Homological chain-complex and LDPC formulations

The homological formulation generalizes homological quantum codes to graded complexes and modules with torsion (Shaska, 11 Aug 2025). Let

$n$5

be a chain complex over $n$6 with boundary maps preserving grading, possibly multi-grading. One defines CSS matrices by

$n$7

so that

$n$8

This realizes the CSS orthogonality relation through functoriality of the boundary (Shaska, 11 Aug 2025).

The resulting quantum code parameters are specified homologically:

$n$9

$K$0

and $K$1 is the minimum weight of nontrivial cycles or cochains representing undetectable $K$2 errors (Shaska, 11 Aug 2025). The LDPC property arises when each boundary involves bounded numbers of cells, producing sparse rows and columns. Grading, including bigrading and orbifold sectors, induces block structure in Tanner graphs and constrains sparsity (Shaska, 11 Aug 2025).

Multiple gradings and torsion are central in this formulation. Multiple gradings, as in Khovanov-type complexes or orbifold homology, split the complex into graded sectors which can be filtered to define codes with targeted sparsity and rank. Torsion in homology, including examples associated with rotor models or orbifold fixed loci, contributes logical operators and can alter $K$3 and $K$4 via additional graded sectors (Shaska, 11 Aug 2025). The paper explicitly states that torsion homology $K$5 may carry logical operators with special algebra, and that multi-grading lets one filter complexes to tune $K$6 and $K$7 (Shaska, 11 Aug 2025).

This homological construction links graded quantum codes to topological and categorical invariants. Examples in the source material include toric surface codes, Khovanov homology codes, and rotor codes with torsion grading (Shaska, 11 Aug 2025). This suggests that grading functions not merely as book-keeping, but as a mechanism for sector decomposition, sparsity control, and parameter tuning in LDPC stabilizer design.

4. Mixed alphabets as grading by local dimension

The mixed-alphabet framework treats grading as heterogeneity in local Hilbert-space dimension (Wang et al., 2012). For a single $K$8-level system, generalized Pauli operators are defined by

$K$9

with

$d$0

On a heterogeneous register with local dimensions $d$1, the Pauli-type error set is

$d$2

The support of an error at site $d$3 is defined by $d$4, and its weight is the size of the support (Wang et al., 2012).

The Knill–Laflamme condition takes the standard form adapted to the heterogeneous Pauli basis. For an encoding projector $d$5 onto a $d$6-dimensional subspace, the code detects all errors of weight $d$7 if for every such error $d$8,

$d$9

for some scalar $< d$0. The distance $< d$1 is the smallest weight of an error for which the detection condition fails (Wang et al., 2012).

Two construction paradigms are developed. In the reducible case $< d$2, where one alphabet factors, a $< d$3-level subsystem can be represented as a composite of a $< d$4-level and an $< d$5-level subsystem. The construction then uses a composite coding clique based on graph states over $< d$6 and $< d$7 (Wang et al., 2012). For a $< d$8-weighted graph $< d$9, the graph state is

$\partial$00

with

$\partial$01

and stabilizer generators

$\partial$02

Similarly, one defines $\partial$03 and $\partial$04 on the $\partial$05-level subsystem (Wang et al., 2012).

The mixed-alphabet basis is then

$\partial$06

A composite coding clique $\partial$07 is defined through the $\partial$08-uncoverable set $\partial$09 and the $\partial$10-purity set $\partial$11, together with three conditions: $\partial$12; for all $\partial$13 and all $\partial$14, one has $\partial$15; and for all distinct $\partial$16, $\partial$17 (Wang et al., 2012). The corresponding code basis is

$\partial$18

and the resulting subspace is a mixed-alphabet code $\partial$19 (Wang et al., 2012).

When $\partial$20 and $\partial$21 are coprime, the composite coding clique approach is unavailable. The alternative is projection-based construction. One embeds the mixed-alphabet register into an ancillary $\partial$22-quqit register and defines

$\partial$23

where each $\partial$24 projects a $\partial$25-level subsystem down to a $\partial$26-level subspace and $\partial$27 (Wang et al., 2012). If an ancillary $\partial$28-dimensional subspace spanned by $\partial$29 can detect or correct the projected errors $\partial$30 for all mixed-register errors of weight $\partial$31, then the normalized projected states

$\partial$32

define a mixed-alphabet QECC $\partial$33 (Wang et al., 2012).

5. Bounds and parameter relations

The weighted/homological graded framework proposes a refined Singleton bound

$\partial$34

where $\partial$35 quantifies entropy or defect reduction due to grading-compatible orbifold singularities on the weighted algebraic geometry side or twisted homological sectors on the chain-complex side (Shaska, 11 Aug 2025). In the paper, for quotient or orbifold points $\partial$36 with stabilizer $\partial$37,

$\partial$38

On weighted AG hypersurfaces, $\partial$39 reflects Chen–Ruan orbifold cohomology contributions from twisted sectors; on homological codes with orbifold grading, $\partial$40 counts twisted-sector ranks in orbifold homology or $\partial$41-theory, tightening entropy inequalities (Shaska, 11 Aug 2025).

The validity of this refined bound is limited in scope. It is partially validated for simple $\partial$42-orbifolds, requires a shared grading alignment and mild singularities such as simple quotient singularities, and is supported by heuristic entropy derivations using TQFT defect sectors. A rigorous coding-theoretic proof remains open (Shaska, 11 Aug 2025). Other bounds discussed in the same framework include Gilbert–Varshamov-type existence bounds for AG codes with weighting, zeta-function or Weil-type bounds controlling $\partial$43, and Hamming or Plotkin variants in which polytope footprint methods are modified by grading (Shaska, 11 Aug 2025).

The mixed-alphabet framework establishes a different generalized quantum Singleton bound. For a mixed-alphabet QECC $\partial$44 and any partition $\partial$45 with $\partial$46 and $\partial$47, one has

$\partial$48

Optimizing over all subsets $\partial$49 of size $\partial$50 yields

$\partial$51

In the uniform case $\partial$52 for all $\partial$53, this reduces to

$\partial$54

and for stabilizer codes with $\partial$55, to the familiar quantum Singleton bound

$\partial$56

(Wang et al., 2012).

The two bounds address different notions of grading. The refined bound with $\partial$57 modifies the standard Singleton inequality through geometric or homological defect data (Shaska, 11 Aug 2025). The mixed-alphabet bound modifies the capacity term through the product of local subsystem dimensions (Wang et al., 2012). This suggests that grading can alter code limitations either through defect-sensitive entropy corrections or through heterogeneity in local alphabet size.

6. Representative examples and constructions

The weighted algebraic geometry pathway includes several worked examples over small fields (Shaska, 11 Aug 2025). In one example, a hypersurface in $\partial$58 over $\partial$59 is defined by

$\partial$60

of weighted degree $\partial$61. Stratifying by supports and stabilizers yields $\partial$62. Taking $\partial$63 gives $\partial$64, and choosing a self-orthogonal subspace $\partial$65 of dimension $\partial$66 produces a classical code with parameters $\partial$67. The CSS lift gives

$\partial$68

with

$\partial$69

Using local stabilizers of orders $\partial$70, the refined bound gives

$\partial$71

for positive but modest $\partial$72 (Shaska, 11 Aug 2025).

A second example uses the genus-2 curve

$\partial$73

homogenized in $\partial$74 over $\partial$75 as

$\partial$76

Using $\partial$77 rational places and a divisor $\partial$78 of degree $\partial$79 yields a classical self-orthogonal AG code with parameters $\partial$80, and the CSS lift gives

$\partial$81

(Shaska, 11 Aug 2025). A third weighted-projective example reports a CSS code $\partial$82 over $\partial$83 from a genus-$\partial$84 weighted projective curve in $\partial$85 (Shaska, 11 Aug 2025).

The homological pathway includes the toric surface code on an $\partial$86 torus, with parameters

$\partial$87

where $\partial$88, $\partial$89, and $\partial$90 is the shortest nontrivial cycle length $\partial$91 (Shaska, 11 Aug 2025). A bigraded Khovanov homology code is reported with parameters $\partial$92, obtained by filtering the complex $\partial$93 to $\partial$94 (Shaska, 11 Aug 2025). A rotor code with torsion grading on a $\partial$95 rotor lattice has $\partial$96 sites and parameters

$\partial$97

with $\partial$98 rank $\partial$99 plus torsion contributing one logical qubit; the paper gives $\epsilon$00, leading to

$\epsilon$01

(Shaska, 11 Aug 2025).

The mixed-alphabet literature provides optimal and suboptimal examples (Wang et al., 2012). The code

$\epsilon$02

is optimal because the generalized Singleton bound gives $\epsilon$03 and the construction achieves $\epsilon$04 (Wang et al., 2012). Likewise,

$\epsilon$05

is optimal with $\epsilon$06, and

$\epsilon$07

is optimal via stabilizer pasting, with the bound giving $\epsilon$08 and the construction achieving $\epsilon$09 (Wang et al., 2012). By contrast,

$\epsilon$10

constructed via projection is suboptimal, since the bound gives $\epsilon$11 but the realized code has $\epsilon$12; the suboptimality is attributed to the fact that the projected errors expand into several Pauli-type operators that must all be detected separately (Wang et al., 2012).

The paper also shows that mixed-alphabet methods inform standard uniform constructions, including the optimal code

$\epsilon$13

and the optimal distance-$\epsilon$14 families

$\epsilon$15

as well as optimal

$\epsilon$16

for any $\epsilon$17 when combined with known odd-prime-dimension families (Wang et al., 2012).

7. Decoding, applications, limitations, and open questions

For weighted AG/CSS codes, decoding is described in terms of syndrome-based algebraic methods (Shaska, 11 Aug 2025). One measures stabilizers or parity checks derived from evaluation maps or vanishing ideals and then solves for the error through algebraic relations. On curves, Riemann–Roch-based decoders, Berlekamp–Welch-like methods for general evaluation codes, Kötter–Vardy soft-decision approaches, and Sudan-style list decoding can be adapted to weighted settings by respecting weighted degrees and orbit stabilizers in Gröbner computations (Shaska, 11 Aug 2025). The Hermitian inner product over $\epsilon$18 can simplify dual-containing checks, stabilizer generation, and decoding tables (Shaska, 11 Aug 2025). Parity-check matrices inherit block structures by weighted degree, and Tanner graphs reflect orbit decompositions, potentially aiding iterative decoding (Shaska, 11 Aug 2025).

For homological LDPC codes, the listed decoding methods include iterative belief propagation on Tanner graphs derived from sparse $\epsilon$19-maps, union-find-like decoders for surface or toric generalizations, and minimum-weight perfect matching for planar-like embeddings (Shaska, 11 Aug 2025). Grading gives multi-layer graphs with structured priors, can localize defects and prioritize correction along graded sectors, and torsion sectors require careful handling because of logical degeneracy (Shaska, 11 Aug 2025). The source further states that homological LDPC structures often admit near-linear-time decoders and favorable thresholds, and that grading improves locality and message-passing efficacy by partitioning constraints (Shaska, 11 Aug 2025).

The applications identified for the weighted/homological framework are post-quantum cryptography, fault-tolerant quantum computing, and optimization via graded neural networks (Shaska, 11 Aug 2025). In post-quantum cryptography, graded AG/CSS codes provide structured, high-rate codes whose parameters can be tuned via weighted heights and zeta-function point counts, and grading aids parameter selection and key-size optimization (Shaska, 11 Aug 2025). In fault-tolerant quantum computing, graded LDPC homological codes offer sparse checks and topological protection; grading aligns with transversal operations or subsystem variants by sectoring stabilizers and logical operators, while orbifold defects $\epsilon$20 relate to enhanced syndrome distinguishability and potentially improved thresholds (Shaska, 11 Aug 2025). For optimization, graded structures enable machine-learning-based parameter search guided by height zeta functions and Hilbert functions to balance $\epsilon$21, $\epsilon$22, and $\epsilon$23 (Shaska, 11 Aug 2025).

The mixed-alphabet framework is motivated by hardware heterogeneity. Multi-level systems such as ququarts and qutrits may co-exist with qubits, higher excited states may be controllable on some sites but not others, and different modules may expose different native dimensions. Mixed-alphabet, hence graded, codes respect this heterogeneity without forcing uniformity (Wang et al., 2012). Decoding proceeds via syndrome extraction from the stabilizers of the underlying graph states or from projections; the syndrome spaces are graded by local dimensions, but the algebra of generalized Pauli operators preserves the standard Knill–Laflamme logic (Wang et al., 2012).

Several limitations and open problems remain. In the weighted/homological setting, the refined bound is only partially established for simple orbifold lines, and a general rigorous proof for higher-dimensional weighted hypersurfaces and arbitrary homological gradings is open (Shaska, 11 Aug 2025). Determining families that meet the refined bound, understanding interaction with asymptotic bounds such as Tsfasman–Vladut–Zink in the quantum setting, developing robust decoders exploiting bigrading and torsion, and building efficient pipelines for weighted point counts, height zeta functions, Gröbner bases in weighted rings, and verification of self-orthogonality are all identified as open directions (Shaska, 11 Aug 2025). In the mixed-alphabet setting, clique search is NP-complete, stabilizer pasting is shown for distance $\epsilon$24 but not extended to $\epsilon$25, and projector choice in the coprime case remains difficult because projected errors may expand into many independent Pauli constraints (Wang et al., 2012).

Taken together, these results show that graded quantum codes form a technically diverse domain in which grading can organize algebraic geometry data, homological sectors, torsion, or local subsystem dimensions. The common consequence is a refinement of code construction and parameter analysis beyond the ungraded stabilizer setting, with explicit formulas, constructions, and examples already available in both the weighted/homological framework (Shaska, 11 Aug 2025) and the mixed-alphabet framework (Wang et al., 2012).