---
title: 'Skewed Quantum Codes: A Concise Overview'
url: https://www.emergentmind.com/topics/skewed-quantum-codes
type: topic
---

# Skewed Quantum Codes: A Concise Overview

“Skewed quantum codes” is not a single standardized object in the literature. The term is used in at least three technically distinct senses: quantum stabilizer codes obtained from skew cyclic or skew constacyclic algebra over noncommutative skew-polynomial rings and finite non-chain rings; asymmetric or bias-tailored codes designed for channels with unequal \(X\)- and \(Z\)-error rates; and approximate subsystem erasure-correcting codes obtained by small nonlocal perturbations of exact encoders in holographic quantum error correction [2305.10404], [2507.02239], [2603.13475]. Across these settings, “skewed” refers either to algebraic twisting by an automorphism, to a deliberate imbalance in error protection, or to a controlled deformation away from exact recovery.

## 1. Terminological scope and conceptual uses

The literature uses the adjective “skewed” in structurally different ways. In ring-theoretic coding theory, the relevant object is the skew polynomial ring \(S[x;\sigma]\), whose multiplication satisfies \(x a = \sigma(a) x\). Quantum codes then arise from dual-containing skew cyclic or skew constacyclic codes, often after a Gray map and a CSS step [2305.10404], [2010.07175]. In asymmetric quantum coding, “skewed” means tailored to unequal Pauli error rates, with distinct distances \(d_X\) and \(d_Z\), or with explicit correction capability for a larger number of one Pauli type than of generic Pauli errors [1307.4532], [2104.04365]. In recent LDPC work, “bias-tailored” plays the same role: the code geometry is matched to a bias parameter \(\eta = p_Z/p_X\) or \(\eta_Z = p_Z/(p_X+p_Y)\) while retaining single-shot properties [2507.02239]. In holographic QEC, “skewed quantum codes” are approximate subsystem erasure-correcting codes of the form \(V^{(\epsilon)} = e^{i\epsilon W} V^{(0)}\), where a small nonlocal perturbation of an exact encoder produces a state-dependent proto-area term [2603.13475].

| Usage | Core mechanism | Representative papers |
|---|---|---|
| Algebraic skew codes | Ore-type skew polynomial rings, Gray maps, CSS | [2305.10404], [2010.07175], [1005.0879] |
| Asymmetric or bias-tailored codes | Unequal \(X/Z\) protection under biased Pauli noise | [1307.4532], [2104.04365], [2507.02239], [1102.3605] |
| Approximate holographic skewing | Small nonlocal deformation of exact subsystem codes | [2603.13475] |

A plausible implication is that the term should always be read relative to context. In algebraic papers it denotes noncommutative twisting; in fault-tolerance papers it denotes noise asymmetry; in holographic code theory it denotes approximate, state-dependent deformation.

## 2. Skew cyclic and skew constacyclic constructions over rings

A major strand of the subject constructs non-binary quantum codes from skew cyclic or skew constacyclic codes over finite non-chain rings. In \(\mathbb{F}_q\mathcal{R}\)-skew cyclic constructions, the ring is
\[
\mathcal{R}=\mathbb{F}_q[u]/\langle u^2-u\rangle \cong \mathbb{F}_q+u\mathbb{F}_q,
\]
with idempotents \(e_1=u\) and \(e_2=1-u\), so that
\[
\mathcal{R}=e_1\mathcal{R}\oplus e_2\mathcal{R}\cong \mathbb{F}_q\times \mathbb{F}_q.
\]
The automorphism on \(\mathbb{F}_q\) is \(\Theta(a)=a^{p^i}\), extended to \(\theta(a+ub)=a^{p^i}+u b^{p^i}\). The skew ring \(S[x;\sigma]\) is defined by \(x a=\sigma(a)x\), and a skew cyclic code is closed under the corresponding skew shift [2305.10404].

For mixed alphabet codes of length \(n=(\alpha,\beta)\), the ambient module is
\[
R_{\alpha,\beta}=\mathbb{F}_q[x;\Theta]/\langle x^\alpha-1\rangle \times \mathcal{R}[x;\theta]/\langle x^\beta-1\rangle.
\]
An \(\mathbb{F}_q\mathcal{R}\)-skew cyclic code is an \(\mathcal{R}\)-submodule of \(\mathbb{F}_q^\alpha\times \mathcal{R}^\beta\) closed under the combined skew shift, and such a code admits generators of the form
\[
\mathcal{C}=\langle (f(x),0),(m(x),g(x))\rangle.
\]
In the separable case, \(\mathcal{C}=\mathcal{C}'\otimes \mathcal{S}\), with \(\mathcal{C}'=\langle f(x)\rangle\subseteq \mathbb{F}_q^\alpha\) and \(\mathcal{S}=\langle g(x)\rangle\subseteq \mathcal{R}^\beta\), where \(g(x)=\xi_1 g_1(x)+\xi_2 g_2(x)\) under the idempotent decomposition [2305.10404].

Duality is controlled by a non-degenerate inner product
\[
\langle l,l'\rangle = u\sum_{i=0}^{\alpha-1}x_i x_i' + \sum_{j=0}^{\beta-1} y_j y_j' \in \mathcal{R},
\]
and the dual-containing criterion is expressed through skew reciprocals. In the field skew cyclic case, if \(x^n-1=h(x)g(x)\), then
\[
\mathcal{C}^\perp \subset \mathcal{C}
\iff
h^\dagger(x)h(x)\ \text{is right-divisible by}\ x^n-1.
\]
In the mixed-alphabet separable case, one needs the analogous right-divisibility conditions simultaneously on the \(\alpha\)-block and the two \(\beta\)-components [2305.10404].

The decisive step toward quantum codes is the Gray map. For \(M\in GL_2(\mathbb{F}_q)\) with \(MM^T=\gamma I_2\), define
\[
\phi(r_1+u r_2)=(r_1,r_2)M,
\]
and extend it to
\[
\varphi:\mathbb{F}_q^\alpha\times \mathcal{R}^\beta \to \mathbb{F}_q^{\alpha+2\beta}.
\]
This map is \(\mathbb{F}_q\)-linear, distance-preserving, and orthogonality-preserving:
\[
w_L(c)=w_H(\varphi(c)), \qquad \varphi(\mathcal{C})^\perp=\varphi(\mathcal{C}^\perp).
\]
If \(\mathcal{C}^\perp\subset \mathcal{C}\), then CSS yields
\[
[[\alpha+2\beta,\ 2k-(\alpha+2\beta),\ d_H]]_q,
\]
where \(k=\dim_{\mathbb{F}_q}(\varphi(\mathcal{C}))\) [2305.10404].

A closely related construction uses the ring
\[
\mathcal{R}=\mathbb{F}_{p^m}[v]/\langle v^3-v\rangle \cong \mathbb{F}_{p^m}+v\mathbb{F}_{p^m}+v^2\mathbb{F}_{p^m},
\]
with three primitive orthogonal idempotents \(\eta_0,\eta_1,\eta_2\), hence \(\mathcal{R}\cong \mathbb{F}_{p^m}^3\). A skew \((\sigma,\delta)\)-constacyclic code decomposes as
\[
C=\eta_0 A_0\oplus \eta_1 A_1\oplus \eta_2 A_2,
\]
where each \(A_i\) is a skew \((\Theta,\lambda_i)\)-constacyclic code over the field. If \(x^n-\lambda_i=h_i f_i\), then dual containment is equivalent to
\[
(x^n-\lambda_i)\ \text{right-divides}\ h_i^*(x)h_i(x)
\]
for each \(i=0,1,2\), under \(\lambda_i=\pm 1\). The Gray image \(\psi(C)\subseteq \mathbb{F}_{p^m}^{3n}\) then gives a quantum code
\[
[[3n,\ 2k-3n,\ d_G]]_{p^m}
\]
whenever \(C^\perp\subseteq C\) [2010.07175].

Over \(\mathbb{F}_4\), an additive rather than \(\mathbb{F}_4\)-linear route is also used. The map
\[
S(v_0,\ldots,v_{n-1})=(v_0,\overline{v_0},v_1,\overline{v_1},\ldots,v_{n-1},\overline{v_{n-1}})
\]
is \(\mathbb{F}_2\)-linear, injective, and doubles Hamming weight:
\[
wt_H(S(u))=2\,wt_H(u).
\]
For any additive code \(C\), the image \(S(C)\) is self-orthogonal under the trace-Hermitian inner product. If \(C\subseteq D\) and \(d(C^\perp_{\mathrm{TrH}})\ge 2\), then one obtains an asymmetric quantum code
\[
[[2n,\ \log_4(M_2/M_1),\ 2d(D)/2]]_4.
\]
If the base code is module \(\theta\)-cyclic, then the image under \(S\) is permutation-equivalent to an additive cyclic code when \(n\) is odd and to an additive \(2\)-quasi-cyclic code when \(n\) is even [1005.0879].

These constructions share a common mechanism: noncommutative factorization enlarges the supply of candidate generator and parity-check polynomials, idempotent decomposition reduces ring calculations to field components, and Gray-type maps transport dual-containing structure into CSS-compatible linear or additive codes.

## 3. Asymmetric stabilizer codes for unequal \(X\)- and \(Z\)-error rates

A second major meaning of skewed quantum coding is asymmetric quantum error correction. Here the channel does not satisfy \(p_X=p_Y=p_Z\), and the code is designed with two different distances,
\[
d_X = \min\{wt(c): c\in C_X\setminus C_Z^\perp\},\qquad
d_Z = \min\{wt(c): c\in C_Z\setminus C_X^\perp\},
\]
under the CSS nesting condition
\[
C_Z^\perp \subseteq C_X.
\]
If \(C_X\) and \(C_Z\) have dimensions \(k_X\) and \(k_Z\), the encoded dimension is
\[
k = k_X + k_Z - n.
\]
Such a code corrects up to \(\lfloor (d_X-1)/2\rfloor\) bit-flip errors and up to \(\lfloor (d_Z-1)/2\rfloor\) phase-flip errors [1307.4532].

One systematic family uses Xing–Ling evaluation codes over \(\mathbb{F}_q\). With
\[
n=t+r,\qquad r=\frac{q^2-q}{2},
\]
and a polynomial space \(V_{m,\ell}\), the code \(C_q(t,m,\ell)\) has
\[
k = \binom{m}{2}+\ell+1
\]
and designed distance
\[
d\ge \delta := n - \frac{1}{2}\big(q(m-1)+\ell+g\big),
\]
with the parity-dependent quantity \(g\) specified by the construction. By taking \(C_Z=C_q(t,m,\ell)\) and \(C_X\) as the dual of carefully chosen low-row subcodes, one obtains pure asymmetric CSS codes with \(d_X\in \{2,3,4,5\}\) or \(d_X\ge 5\). The encoded dimension is
\[
k = h+\ell,\ h+\ell-2,\ h+\ell-4,\ \text{or}\ h+\ell-7,
\]
depending on whether the \(X\)-side uses the repetition code, the three-row code \(D\), the five-row code \(E\), or the eight-row code \(F\), where \(h=\binom{m}{2}\) [1307.4532].

A second algebraic-geometric family uses two-point divisors on the Hermitian curve \(y^q+y=x^{q+1}\) over \(\mathbb{F}_{q^2}\). With evaluation set \(D=R-P-Q\), the length is
\[
n=q^3-1,
\]
and the genus is \(g=q(q-1)/2\). For \(0<r<q(q+1)\), the code
\[
C_L(D,(q^3-r+1)P-2Q)
\]
has dimension
\[
k(r)=q^3-g-r
\]
and explicit distance \(d(r)\) determined by the decomposition \(r+q=c(q+1)-a\). From two nested two-point inputs one obtains a pure AQECC
\[
[[q^3-1,\ q^3-q(q-1)-(r_1+r_2)+1,\ d(r_2)/d(r_1)]]_{q^2},
\]
and the paper states strict improvements over one-point constructions for the range \(0<r<q(q+1)\) [1102.3605].

A more targeted asymmetric setting asks for correction of up to \(e_g\) generic Pauli errors plus up to \(e_Z\) additional errors of a specified type, usually \(Z\). The corresponding generalized quantum Hamming bound is
\[
2^{n-k}\ge \sum_{j=0}^{e_g+e_Z}\binom{n}{j}\sum_{i=0}^{e_g}\binom{j}{i}2^i.
\]
For \((e_g,e_Z)=(1,1)\), the paper constructs a non-degenerate \([[9,1]]\) stabilizer code and shows that it is the shortest code with that correction capability. For \((e_g,e_Z)=(1,2)\), it gives an explicit \([[13,1]]\) code. The construction proceeds by syndrome assignment: first fixing all \(Z_i\) syndromes, then assigning \(X_i\) syndromes to avoid collisions among all correctable operators of the form \(I\), \(X_i\), \(Y_i\), \(Z_i\), \(X_i Z_j\), \(Y_i Z_j\), and \(Z_i Z_j\) [2104.04365].

These asymmetric constructions treat skewness as a design target rather than an algebraic mechanism. The common objective is to reallocate redundancy toward the dominant physical error process.

## 4. Bias-tailored single-shot quantum LDPC codes

Recent LDPC work combines skewed noise adaptation with single-shot fault tolerance. The physical model is the independent Pauli channel
\[
\mathcal{E}(\rho)=(1-p)\rho+p_X X\rho X+p_Y Y\rho Y+p_Z Z\rho Z,
\qquad p=p_X+p_Y+p_Z,
\]
with bias parameter
\[
\eta=\frac{p_Z}{p_X}
\quad\text{or}\quad
\eta_Z=\frac{p_Z}{p_X+p_Y}.
\]
The limit \(\eta\to \infty\) corresponds to pure \(Z\) errors, whereas \(\eta=1\) is depolarizing noise [2507.02239].

The starting point is the syndrome-encoded hypergraph product (SEHGP) code. If \(H_1\in \mathbb{F}_2^{m_1\times n_1}\) and \(H_2\in \mathbb{F}_2^{m_2\times n_2}\) are parity-check matrices, the standard HGP stabilizers are
\[
H_X=
\begin{bmatrix}
I_{n_1}\otimes H_2^\top\ \ H_1\otimes I_{n_2}
\end{bmatrix},
\qquad
H_Z=
\begin{bmatrix}
H_1^\top\otimes I_{n_2}\ \ I_{n_1}\otimes H_2
\end{bmatrix},
\]
with \(H_X H_Z^\top=0\) whenever \(H_1 H_2^\top=0\). Syndrome encoding is implemented at the chain-complex level, and the resulting product syndrome maps are automatically \((t,f)\)-sound with
\[
t=\min\{d(\text{base}_1),d(\text{base}_2)\},
\qquad
f(x)=\frac{x^2}{4}.
\]
This soundness is the basis for single-shot error correction [2507.02239].

Bias tailoring is introduced through commutation-preserving Hadamard rotations (CPHR), specifically swap patterns T1 and T2 that exchange selected \(X\)- and \(Z\)-type stabilizer blocks while preserving commutation. Applied to SEHGP, these rotations yield the full bias-tailored syndrome-encoded HGP family (BSH), which under identical bases has parameters
\[
[[6n^4,6k^4,d]]
\]
and remains \((d,f)\)-sound with \(f(x)=x^2/4\). The paper states that BSH keeps good single-shot properties under depolarizing and pure \(X/Z\) noise, while improving thresholds as \(\eta\to\infty\) [2507.02239].

Two trimmed families trade redundancy against protection guarantees. The simplified family SSH/BSSH has
\[
5n^4\ \text{physical qubits},\qquad 4n^4\ \text{stabilizer checks},
\]
compared to \(6n^4\) and \(8n^4\) for SEHGP/BSH. Its logical dimension is \(2k^4\), and its distance grows from \(d\) to \(d^2\). The paper states that SSH retains single-shot protection for every noise model, while BSSH inherits the same depolarizing threshold and improves under strong bias. The reduced family RSH/BRSH has the same hardware savings, logical dimension \(4k^4\), and distance \(d\), but preserves single-shot only under depolarizing noise; it trades away single-shot protection under purely \(X\) or purely \(Z\) noise [2507.02239].

A concrete explicit member of the simplified family is the three-dimensional XZZX code obtained by lifting the two-dimensional XZZX surface code to a cubic lattice. With periodic boundary conditions it has parameters
\[
[[5n^4,2,n^2]],
\]
and inherits the bias-friendly XZZX stabilizer pattern in a 3D LDPC layout [2507.02239].

This suggests that in modern LDPC usage, skewed quantum coding is not principally about CSS asymmetry parameters \(d_Z/d_X\), but about tailoring stabilizer geometry so that the dominant error type decouples into simpler classical constituents while the code still supports one-round correction of data and measurement errors.

## 5. Skewed codes as approximate subsystem erasure-correcting codes

In holographic and information-theoretic work, “skewed quantum code” denotes an approximate subsystem erasure-correcting code obtained by perturbing an exact code. Let \(V^{(0)}:H_L\to H_P\) be an exact subsystem encoder for a bipartition \(A|\bar A\). A skewed code is a one-parameter family
\[
V^{(\epsilon)} := e^{i\epsilon W}V^{(0)},
\]
with \(W\) Hermitian and \(\epsilon\) small [2603.13475].

The exact case admits state-independent local decoders and a factorization
\[
R_A R_{\bar A}|\tilde\psi\rangle
=
|\psi\rangle_{A_1\bar A_1}\otimes |\chi\rangle_{A_2\bar A_2},
\]
which implies a fixed area term \(S(\chi)\) in an FLM-like relation. The approximate case satisfies only approximate Knill–Laflamme conditions. The paper proves an equivalence between approximate KL data and an approximate skew-recovery form:
\[
| \tilde i\rangle \otimes |0\rangle_{E\bar E}
=
e^{i\epsilon' W}
(R_{AE}^\dagger\otimes R_{\bar A\bar E}^\dagger)
\big(
|i\rangle_{A_1\bar A_1}\otimes |\chi\rangle_{A_2\bar A_2}
\big),
\]
with explicit norm bounds relating \(\epsilon'\|W\|_2\) to the sesquilinear error functionals appearing in the approximate KL statement [2603.13475].

The central entropic objects are the boundary entropy
\[
S_{bdy}(A)=S(\rho_A)
\]
and the optimally recoverable bulk entropy
\[
S^{opt}_{bulk}(A)=S(\sigma^{(R^*)}_{A_1}),
\]
where \(R^*\) maximizes coherent information. The proto-area is then defined by
\[
S_{PA}(V,\sigma^{(L)},A) := S(\rho_A)-S(\sigma^{(R^*)}_{A_1}).
\]
In exact codes this reduces to \(S(\chi)\); in skewed approximate codes it becomes state dependent [2603.13475].

For flat \(\chi\), the paper proves a relative-entropy identity:
\[
S_{PA}(V^{(\epsilon)},\sigma^{(L)},A)
=
S(\chi)-
\Big[
D(\sigma_{A_1A_2}^{(R^{(\epsilon)})}\,\|\,\sigma_{A_1A_2}^{(R^{(0)})})
-
D(\sigma_{A_1}^{(R^{(\epsilon)})}\,\|\,\sigma_{A_1}^{(R^{(0)})})
\Big].
\]
It also proves leading-order monotonicity statements. In the mixed-bulk case, after recovery optimization one has
\[
\frac{d\langle S_{PA}\rangle}{dS(\lambda)}\ge 0
\]
to \(O(\epsilon^2)\). In the pure-bulk case, for \(W\) drawn from GUE, the paper states that with probability \(1-O(e^{-d^2})\) the decreasing spectral terms dominate, so \(\langle S_{PA}\rangle\) increases monotonically with bulk entanglement to \(O(\epsilon^2)\) [2603.13475].

The mechanism responsible for this state dependence is tripartite non-local magic in the Choi state of the encoder. For stabilizer encoders, the stabilizer Rényi entropy vanishes, and the paper states that stabilizer codes and local-unitary deformations thereof have only trivial, state-independent area operators. In skewed stabilizer deformations, the leading proto-area correction is proportional to perturbative tripartite non-local magic \(M^{NL}_\alpha\), so non-Clifford, irreducibly tripartite control becomes the operational source of matter-geometry coupling [2603.13475].

In this sense, skewing is neither algebraic twisting nor noise bias. It is a controlled departure from exact recovery that converts a fixed geometric entropy term into a state-dependent quantity.

## 6. Representative constructions, parameters, and current limitations

Several papers report explicit parameter gains or structural advantages.

| Construction | Classical-to-quantum output | Reported comparison |
|---|---|---|
| \(\mathbb{F}_9\mathcal{R}\)-skew cyclic Gray image | \([121,114,4]_9 \to [[121,107,4]]_9\) | Improves \([[121,106,4]]_9\) [2305.10404] |
| Skew constacyclic over \(\mathbb{F}_{25}+v\mathbb{F}_{25}+v^2\mathbb{F}_{25}\) | \([36,32,3]_{25} \to [[36,28,3]]_{25}\) | Within \(4\) of the quantum Singleton bound [2010.07175] |
| \(S\)-map from module \(\theta\)-cyclic code | \([4,2,3]_4\) base gives \([[8,1,6/2]]_4\) | \(d_Z\) doubled while \(d_X=2\) [1005.0879] |
| Xing–Ling asymmetric CSS code | \([[36,1,\{32,2\}]]_9\) | Pure, strongly \(Z\)-biased protection [1307.4532] |
| Short stabilizer with one prevalent Pauli type | \([[9,1]]\) for \((e_g,e_Z)=(1,1)\) | Shown shortest by the generalized quantum Hamming bound [2104.04365] |
| Two-point Hermitian AQECC | \([[63,42,9/5]]_{16}\) | Improves \([[63,39,9/5]]_{16}\) from one-point inputs [1102.3605] |
| Bias-tailored 3D XZZX member of BSSH | \([[5n^4,2,n^2]]\) | Explicit simplified-family example [2507.02239] |

The algebraic skew literature emphasizes construction rather than decoding. In the \(\mathbb{F}_q\mathcal{R}\)-skew cyclic work, the paper states that it does not develop BCH-type or Hartmann–Tzeng bounds, determines distances computationally via Magma on Gray images, and does not address decoding algorithms or syndrome structures [2305.10404]. The skew-constacyclic ring construction over \(v^3=v\) identifies alternative Gray maps, BCH-type bounds, and entanglement-assisted variants as open directions [2010.07175]. In the \(\mathbb{F}_4\) additive \(S\)-map framework, the main limitation is that \(S\) is additive but not \(\mathbb{F}_4\)-linear, and the resulting AQCs typically have \(d_X=2\) [1005.0879].

The asymmetric stabilizer literature also has explicit scope constraints. The short-code syndrome-assignment method is presented for \(e_g=1\) and imposes the constructive condition
\[
n-4(e_g+e_Z)\ge 1,
\]
which the paper notes can be stricter than the generalized quantum Hamming bound [2104.04365]. The Xing–Ling and Hermitian two-point constructions are parameter-rich and often pure, but they are still CSS constructions built from classical nested families; their performance is therefore tied to the available algebraic-geometric distance control [1307.4532], [1102.3605].

The bias-tailored single-shot LDPC hierarchy opens a different set of problems. The paper states that full numerical threshold curves \(p_{\mathrm{th}}(\eta)\) for BSH/BSSH/BRSH remain to be benchmarked, finite-size effects and scaling constants are unknown, and decoder comparisons among MWPM, BP+OSD, flip, and hybrid methods under measurement noise remain to be carried out [2507.02239].

The holographic skewed-code framework is explicitly perturbative. Its monotonicity results are controlled at \(O(\epsilon^2)\), the pure-state theorem is probabilistic rather than universal, and a direct equivalence between optimal-recovery extremization and quantum-extremal-surface extremization remains open [2603.13475].

Taken together, these literatures show that skewed quantum coding is best understood as a family of strategies for exploiting asymmetry: asymmetry in algebraic multiplication, asymmetry in physical noise, or asymmetry between exact and approximate recoverability. The technical details differ sharply, but each setting uses skewing to enlarge the feasible design space beyond standard symmetric stabilizer constructions.

Source: https://www.emergentmind.com/topics/skewed-quantum-codes