---
title: Quantum Singleton Bound for QECC
url: https://www.emergentmind.com/topics/quantum-singleton-bound
type: topic
---

# Quantum Singleton Bound for QECC

The **Quantum Singleton Bound** is the standard Singleton-type dimension–distance constraint for quantum error-correcting codes. For a standard quantum error-correcting code with parameters \([\![n,k,d]\!]_q\), it is stated as
\[
2d \le n-k+2,
\]
equivalently
\[
k \le n-2(d-1),
\]
and, in the \(((n,K,d))_q\) notation,
\[
K \le q^{\,n-2(d-1)}.
\]
It is the quantum analogue of the classical Singleton bound, but it is stricter by a factor of two in the distance term. In the formulations summarized in the literature, the bound applies to both pure and impure QECCs, and codes meeting it with equality are quantum MDS codes [1005.4758].

## 1. Standard formulation and operational meaning

For a quantum code \([\![n,k,d]\!]_q\), \(n\) is the block length, \(k\) is the number of logical qudits, \(q\) is the local dimension, and \(d\) is the minimum distance. A code of distance \(d\) can correct up to \(\lfloor (d-1)/2 \rfloor\) arbitrary errors, or up to \(d-1\) erasures or located errors [2007.01249]. In the alternative \(((n,K,d))_q\) notation, \(K\) is the dimension of the code space, and the same bound becomes \(K \le q^{\,n-2(d-1)}\) [1005.4758].

In rate form, if \(R=k/n\), the standard quantum Singleton bound gives
\[
\frac{d}{n} \le \frac{1-R}{2} + O(1/n).
\]
This identifies \((1-R)/2\) as the asymptotic relative-distance ceiling for exact quantum codes. Since unique decoding from arbitrary adversarial errors uses roughly half the distance, the usual exact adversarial decoding radius is bounded by
\[
\text{decoding radius} \le \frac{1-R}{4},
\]
which is the “half-distance” barrier emphasized in the approximate-coding literature [2212.09935].

The bound is also naturally interpreted through erasure correction. In the entropic formulation, a code of minimum distance \(d\) corrects erasures of up to \(d-1\) carriers. This makes the parameter \(2(d-1)\) the effective amount of block length that must be sacrificed to protect quantum information against both \(X\)- and \(Z\)-type logical obstruction, which is why the quantum constraint is stricter than its classical counterpart [2010.07902].

## 2. Equality cases, quantum MDS codes, and strengthened analytic bounds

Codes attaining
\[
K \le p^{\,n-2(d-1)}
\]
with equality are quantum MDS codes [1005.4758]. In this sense the Quantum Singleton Bound plays the same extremal role for quantum codes that the classical Singleton bound plays for MDS codes. The literature summarized here does not merely treat equality as a nomenclature issue; it uses the bound as the baseline from which sharper analytic obstructions are derived.

A particularly explicit strengthening is the **quantum Hamming–Singleton bound** introduced for pure codes with
\[
d = 2t+1+o,\qquad o\in\{0,1\}.
\]
The quantum Hamming bound has the form
\[
K\,H_{t,o}^{\,n}\le p^n,
\]
while the interpolating quantum Hamming–Singleton bound is
\[
K\,H_{t-e,o}^{\,n-2e}\le p^{\,n-2e},
\]
for every integer \(0\le e\le t\) [1005.4758]. When \(e=0\), this reduces to the quantum Hamming bound; when \(e=t\), it reduces to the quantum Singleton bound; and for intermediate \(e\), it can be strictly stronger than both. The optimal choice of \(e\) is given in that work by
\[
e=\frac{1}{2}\left(n-d+1-\left\lfloor\frac{n-d+1}{p^2-2}\right\rfloor\right),
\]
with \(e=0\) when \(n>t(p^2-2)+1+o\) [1005.4758].

The same paper strengthens the combined bound further using Lloyd’s theorem. For a pure code with \(d=2t+1+o\), it defines a polynomial \(A_{t,e}(x)\) from the zeros \(x_j\) of the Lloyd polynomial \(L_{t-e,o}(x)\), and proves a strengthened inequality
\[
S_{t,e}\,K \le p^n,\qquad 0\le e<t.
\]
If at least one zero \(x_j\) of the Lloyd polynomial is non-integer, then
\[
S_{t,e} > H_{t-e,o}^{\,n-2e},
\]
so the strengthened bound is strictly better than the quantum Hamming–Singleton bound, and therefore also better than the quantum Hamming bound [1005.4758]. The paper proves this strengthened form for impure codes when \(d=3,4\), conjectures that it should hold for impure codes at arbitrary distance, and notes that there is no strengthening for qubits (\(p=2\)).

For stabilizer codes, where \(K=p^k\), these bounds translate into integer constraints on the number of logical qudits. The paper writes
\[
n-k \ge S_p(n,d)\ge h_p(n,d),
\]
and when \(S_p(n,d)>h_p(n,d)\), integrality implies a **1-logical-qudit improvement** over the ordinary quantum Hamming bound [1005.4758]. An infinite family highlighted there is
\[
n_a=\frac{4^a-1}{3},\qquad a\ge 3,
\]
for binary stabilizer codes with \(d=5\).

## 3. Proof techniques: entropy inequalities and symplectic linear algebra

One modern route to the Quantum Singleton Bound is purely entropic. For an \(n\)-party code state with reference \(R\), and any partition
\[
[n]=I \,\dot\cup\, J,\qquad |I|=n-d+1,\quad |J|=d-1,
\]
perfect correction of a \(d-1\) erasure gives the entropy identity
\[
S(X_I)=S(RX_J)=S(R)+S(X_J).
\]
Averaging over all such partitions yields
\[
S(R)=\mathbb{E}_{|I|=n-d+1}S(X_I)-\mathbb{E}_{|J|=d-1}S(X_J).
\]
Using an averaging lemma based on strong subadditivity,
\[
\mathbb{E}_{|I|=m} S(X_I)\le \frac{m}{\mu}\,\mathbb{E}_{|J|=\mu}S(X_J), \qquad \mu<m\le n,
\]
one obtains
\[
\log_2 K \le (n-2d+2)\log_2 q,
\]
which is the standard quantum Singleton bound in logarithmic form [2010.07902]. The same work emphasizes that these entropic proofs are robust: if a code corrects most erasure patterns only approximately, the same inequalities persist up to continuity corrections.

A distinct recent route is a symplectic proof for stabilizer codes. In this formulation one works over a prime field \(F=F_p\) with
\[
V := F^n \times F^n \cong F^{2n},
\]
equipped with the standard nondegenerate symplectic bilinear form
\[
\langle u,v\rangle := \sum_{i=1}^n \bigl(u_X(i)v_Z(i) - u_Z(i)v_X(i)\bigr).
\]
A stabilizer code is specified by an isotropic subspace
\[
S\subseteq V,\qquad \dim(S)=n-k,
\]
and the distance is
\[
d:=\min\{wt(v): v\in S^\perp\setminus S,\ v\neq 0\}.
\]
For a subset \(E\subseteq[n]\), correctable erasure is expressed as
\[
S^\perp \cap V_E \subseteq S.
\]
The key dimension identity is the cleaning relation
\[
g(M)+g(M^c)=2k,
\]
where
\[
g(M):=\dim\bigl((S^\perp\cap V_M)/(S\cap V_M)\bigr).
\]
From this one proves that if two disjoint sets \(A,B\) are both correctable and \(C=[n]\setminus(A\cup B)\), then
\[
k\le |C|.
\]
Choosing \(|A|=|B|=d-1\) gives
\[
k\le n-2(d-1),
\]
hence
\[
k+2(d-1)\le n
\]
for any \([[n,k,d]]\) stabilizer code [2602.20186]. That paper also provides a Lean4 formalisation of the argument.

## 4. Entanglement-assisted and hybrid generalizations

For entanglement-assisted quantum error-correcting codes with parameters \([\![n,k,d;c]\!]_q\), the familiar entanglement-assisted Singleton bound quoted in the literature is
\[
2d \le n-k+2+c.
\]
Here \(c\) is the number of shared maximally entangled pairs. The same source notes a stronger validity condition from later work: this form is valid for
\[
d \le \frac{n+2}{2}
\]
[2007.01249]. The entropic converse literature subsequently showed that the generalized quantum Singleton story for EAQECC and CQECC is not captured by a single naive extension. Instead, the correct result is a tradeoff region between transmitted qubits and net entanglement consumption, and it changes qualitatively depending on whether \(d-1<n/2\) or \(d-1\ge n/2\) [2010.07902].

For EAQECC, writing \(\ell\) for net entanglement consumption, the entropic bounds include
\[
\log_2 K \le \max\{0,n-2d+2\}\log_2 q + \ell,
\]
\[
\log_2 K \le (n-d+1)\log_2 q,
\]
and, in the regime \(d-1\ge n/2\),
\[
\log_2 K \le \frac{n-d+1}{3d-3-n} \bigl(\ell+(2d-2-n)\log_2 q\bigr)
\]
[2010.07902]. The same work proves a propagation rule: any pure QECC \([\![n,k,d]\!]_q\) gives rise to EAQECC codes
\[
[\![n-c,k,d;c]\!]_q
\qquad\text{for all } c<d.
\]

The hybrid entanglement-assisted classical–quantum setting is described by net rates \(C,Q,E\), corresponding to cbits, qubits, and ebits. For a block erasure channel erasing \(d-1\) of \(n\) subsystems, the achievable triple-rate region is constrained by
\[
C + 2Q \le (n-d+1)(\log q + t),
\]
\[
Q - E \le (n-2d+2)\, t,
\]
\[
C + Q - E \le (n-d+1)\log q - (d-1)t,
\]
for some
\[
t \in [0,\log q]
\]
[2202.02184]. Setting \(C=0\) and \(E=0\) recovers the ordinary quantum Singleton bound, while \(Q=0\) and \(E=0\) recovers the classical Singleton bound.

In the related setting of entanglement-assisted classical coding, the standard EACC Singleton bound is
\[
k \le \left(1+\frac{c}{n}\right)(n-d+1),
\]
and a 2026 result shows this bound is tight by a space-sharing construction. In the separate-encoder model, where entanglement is distributed across a subset of encoders and only local quantum operations are allowed, the tight entropic Singleton bound becomes
\[
k \le \max\{n+c-2d+2,\; n-d+1\}
\]
[2603.08563]. These are not statements about standard QECCs, but they locate the Quantum Singleton Bound within a larger family of entropic Singleton-type constraints in quantum Shannon theory.

## 5. Approximate error correction and the near-Singleton regime

The exact Quantum Singleton Bound does not disappear in the approximate setting, but its operational implications change. For exact quantum codes of rate \(R\), the maximum possible relative distance is approximately \((1-R)/2\), while exact adversarial unique decoding is limited to \((1-R)/4\) [2212.09935]. The question posed in approximate quantum coding is whether approximate recovery can move adversarial correction closer to the full erasure-style Singleton limit.

An approximate quantum error-correcting code \((Enc,Dec)\) against a \(\delta\) fraction of adversarially corrupted blocks is defined by the diamond-norm condition
\[
\bigl\| Dec \circ \mathcal A \circ Enc - \mathbb{I}_\diamond \bigr\| \le \varepsilon.
\]
A more general Singleton-type bound quoted in that setting is
\[
\delta \le \frac{1-R}{2} + O(\varepsilon \cdot c), \qquad c = 1 + \frac{\log_q(1/\varepsilon)}{n},
\]
so approximate correction still cannot exceed the erasure-style ceiling \((1-R)/2\), up to small additive terms [2212.09935].

What changes is achievability. For any \(0<R<1\) and \(\gamma>0\), there exist families of approximate quantum \([[n,Rn]]_q\) codes over a constant alphabet size \(q=q(\gamma)\) that correct errors on
\[
\frac{1-R-\gamma}{2}\,n
\]
registers, with recovery error
\[
2^{-\Omega(n)}
\]
[2212.09935]. The construction is efficiently decodable and combines three ingredients: quantum list decoding, folded quantum Reed–Solomon codes, and an Alon–Edmonds–Luby expander-based distance amplification step to obtain constant alphabet size. The list-decoding stage is then converted into approximate unique decoding via a purity testing code, yielding a private AQECC with error roughly \(2L\varepsilon\), and a robust secret sharing scheme removes the private side channel [2212.09935].

The significance is that, in the approximate regime, the **distance** and the **adversarial decoding radius** can be “nearly identical.” Exact quantum coding remains limited by the half-distance barrier, but approximate quantum coding can correct adversarial errors up to essentially the full Singleton limit \((1-R)/2\) [2212.09935].

## 6. Scope, limitations, and common misreadings

A common misreading is to treat every entanglement-assisted communication protocol as if it were governed by the standard EAQECC Singleton bound
\[
2d \le n-k+2+c.
\]
A counterexample to that broader reading is a teleportation-based communication scheme in which shared maximally entangled states are used to teleport the logical qudits, while the resulting classical Bell-measurement outcomes are protected by a classical \([n,2k,d]_q\) code. The relevant bound for that scheme is the classical Singleton bound
\[
d \le n-2k+1.
\]
In the special case \(c=k\), comparison with the EAQECC bound gives
\[
d \le \frac{n}{2}+1,
\]
and the teleportation-based scheme exceeds that whenever
\[
k<\frac{n}{4}.
\]
The paper gives the explicit family
\[
[\![\ell(q+1),1,\ell q;1]\!]_q,
\]
obtained by repeating the MDS code \([q+1,2,q]_q\), with normalized distance
\[
\delta=\frac{q}{q+1}\to 1
\qquad\text{as } q\to\infty
\]
[2007.01249]. The point is not that the ordinary Quantum Singleton Bound for standard QECCs fails, but that the usual entanglement-assisted Singleton bound does not capture all possible entanglement-assisted communication strategies.

Another important limitation is that the standard Quantum Singleton Bound is fundamentally an **exact** coding statement. In exact QECCs it implies a relative-distance ceiling near \((1-R)/2\) and an exact adversarial unique-decoding ceiling near \((1-R)/4)\); in approximate QECCs, the same Singleton ceiling remains the asymptotic obstruction, but explicit constructions can approach it algorithmically [2212.09935]. This separation is central to current work on approximate quantum coding.

Taken together, the literature presents the Quantum Singleton Bound not as a single isolated inequality, but as a structural organizing principle. It governs standard QECCs, admits sharper analytic refinements in the Hamming–Singleton and Lloyd-polynomial framework, has entropic and symplectic proofs, extends to assisted and hybrid settings in more elaborate piecewise forms, and remains the asymptotic frontier even when approximate recovery nearly closes the exact half-distance gap [1005.4758].

Source: https://www.emergentmind.com/topics/quantum-singleton-bound