Bounds for Pure Disjoint $(r,\delta)$-Quantum Locally Recoverable Codes

Quantum locally recoverable codes (qLRCs) have recently attracted significant attention [9, 21, 16, 8, 24, 15, 6] as a framework for protecting quantum information under locality constraints. In an $(r,2)$-qLRC, the erasure of a single qudit can be corrected by accessing at most $r$ other qudits, thereby avoiding recovery operations that act on the entire system.

The study of qLRCs was initiated by Golowich and Guruswami [9], who introduced $(r,2)$-qLRCs and constructed explicit families via the CSS construction [5]. They also derived a quantum Singleton-like bound for these codes, which, unlike its classical counterpart [10], can be strengthened when the recovery sets are disjoint. Luo et al. [16] subsequently obtained further bounds for CSS-based $(r,2)$-qLRCs, including quantum Singleton-like and CM-type bounds. Galindo et al. [8] then introduced the more general notion of $(r,\delta)$-qLRCs in the stabilizer setting and established a correspondence with dual-containing classical $(r,\delta)$-LRCs, leading in particular to a Singleton-like bound for stabilizer $(r,\delta)$-qLRCs. More recently, Li et al. [14, 15] derived further bounds for $(r,2)$-qLRCs. With the exception of the quantum Singleton-like bounds in [9], the known qLRC bounds have largely been obtained through connections with underlying classical codes.

A key technique for deriving upper bounds on code parameters is the use of weight enumerators and the linear-programming (LP) method [7]. In the quantum setting, Shor and Laflamme [22] introduced the first quantum weight enumerators, now known as the Shor–Laflamme (SL) weight enumerators, together with the corresponding quantum MacWilliams identities. Rains [17] later introduced the unitary weight enumerators, which are invariant under code equivalence, and related them both to correctable erasure patterns and to the SL weight enumerators. Using this connection together with the properties of these enumerators, Rains [18] derived the quantum Singleton bound for nonbinary quantum codes.

The LP approach to quantum codes goes back to [22] and was further refined by Rains [19] through the introduction of the shadow enumerator, which provides additional constraints on admissible weight distributions. Ashikhmin and Litsyu [4] then formulated this approach systematically and derived several fundamental upper bounds for binary quantum codes by choosing auxiliary polynomials satisfying suitable conditions. Their method was later extended to the nonbinary stabilizer setting in [12]. More recently, Lai and Ashikhmin [13] generalized this framework to entanglement-assisted quantum codes by introducing split weight enumerators [23].

Motivated by this line of work, we derive upper bounds on the dimension of pure disjoint $(r,\delta)$-qLRCs using quantum weight enumerators. We first formulate local KL conditions and introduce blockwise SL and unitary weight enumerators that record how error weight is distributed across recovery sets. This framework yields a Singleton-like bound that strengthens the known bound for general disjoint $(r,\delta)$-qLRCs. In the special case $\delta=2$, our approach recovers the corresponding stabilizer $(r,2)$-qLRC bound [8] without assuming a stabilizer structure, under the additional assumption of disjoint recovery sets. Finally, we use the same blockwise enumerator framework to derive an LP upper bound on the code dimension.

Let $\mathcal{H}=(\mathbb{C}^{q})^{\otimes n}$ denote the Hilbert space of $n$ qudits of local dimension $q$. A quantum code $\mathcal{Q}$ of length $n$, dimension $K$, and minimum distance $d$ is a $K$-dimensional subspace of $\mathcal{H}$, and is denoted by $((n,K,d))_{q}$. If $K=q^{k}$, we write $k=\log_{q}K$ and denote the code by $[[n,k,d]]_{q}$. Let $\mathcal{D}(\mathcal{Q})$ denote the set of density operators whose support is contained in $\mathcal{Q}$. For any subset $I\subseteq[n]:=\{1,\dots,n\}$, define the subsystem $\mathcal{H}_{I}:=\bigotimes_{i\in I}\mathbb{C}^{q}$, and let $I^{c}:=[n]\setminus I$. A quantum channel $\widetilde{\mathcal{N}}$ on $\mathcal{H}$ is supported on $I$ if it admits Kraus operators of the form $\widetilde{K}_{a}=K_{a}\otimes\mathbb{I}_{I^{c}}$. Equivalently, $\widetilde{\mathcal{N}}=\mathcal{N}^{I}\otimes\mathrm{id}_{I^{c}}$, where $\mathcal{N}^{I}$ is a quantum channel on $\mathcal{H}_{I}$ with Kraus operators $\{K_{a}\}_{a}$.

That is, for each $i\in[n]$, any located error affecting at most $\delta-1$ qudits inside the local group $R_{i}$ can be corrected by a recovery operation supported on $R_{i}$.

Disjoint recovery sets induce a blockwise error decomposition that is essential for the weight enumerator arguments.

In this section, we derive local KL conditions for disjoint $(r,\delta)$-qLRCs, since the standard KL conditions do not encode the locality constraint on the recovery operation.

The local KL condition shows that correctability within a recovery block depends only on the local code support $\mathcal{Q}_{\ell}$.

The global minimum distance $d$ of $\mathcal{Q}$ is the smallest weight of an operator $E\in\mathcal{E}$ that is not detectable on $\mathcal{Q}$, that is, $\langle\psi_{i}|E|\psi_{j}\rangle\neq C_{E}\,\delta_{ij}$ for some $i,j$. Likewise, for a recovery block $J_{\ell}$ with local code support $\mathcal{Q}_{\ell}$, the local minimum distance $\delta^{(\ell)}$ is the smallest weight of an operator $E^{(\ell)}$ supported on $J_{\ell}$ that is not locally detectable on $\mathcal{Q}_{\ell}$, that is, $\langle\alpha_{t}^{(\ell)}|E^{(\ell)}|\alpha_{t^{\prime}}^{(\ell)}\rangle\neq C_{E}^{(\ell)}\,\delta_{tt^{\prime}}$ for some $t,t^{\prime}\in[a_{\ell}]$. The local distance of the code is then defined by $\delta:=\min_{\ell}\delta^{(\ell)}$.

Let $\mathcal{Q}$ be a disjoint $(r,\delta)$-qLRC with global distance $d$ and block partition $\{J_{\ell}\}_{\ell=1}^{s}$. For a basis operator $E=\bigotimes_{\ell=1}^{s}E^{(\ell)}$, let $i_{\ell}=\mathrm{wt}(E^{(\ell)})$ denote the number of affected qudits in block $J_{\ell}$, and write $\boldsymbol{i}\in\mathcal{I}_{N}^{s}$ for the corresponding erasure pattern. By locality and disjointness, each block with $i_{\ell}\leq\delta-1$ can be corrected independently, while the remaining erasures can be corrected globally provided their total weight is at most $d-1$. Therefore the set of correctable erasure patterns is

Equivalently, $\mathcal{T}_{d,\delta}$ consists of all $\boldsymbol{i}\in\mathcal{I}_{N}^{s}$ such that $\sum_{j:\,i_{j}\geq\delta}i_{j}\leq d-1$.

Here we introduce blockwise SL and unitary weight enumerators, which record how error weight is distributed across recovery blocks. Related refinements include split weight enumerators for classical codes [23] and two-block split weight enumerators for entanglement-assisted quantum codes [13].

Next, we derive a Singleton-like bound for pure disjoint $(r,\delta)$-qLRCs using blockwise weight enumerators.

This is the blockwise analogue of the usual purity condition, restricted to the set of $(r,\delta)$-qLRC-correctable error patterns.

Next, we compare Theorem 4 with the known bounds in the literature. For disjoint $(r,\delta)$-qLRCs of length $n$ and minimum distance $d$, the quantum dimension $k$ satisfies

which for $\delta=2$ reduces to the bound derived in [9]. A direct comparison shows that (5) strengthens the general disjoint $(r,\delta)$-qLRC bound (7) under the additional purity assumption.

For pure stabilizer $(r,\delta)$-qLRCs, one also has

When $\delta=2$, Theorem 4 agrees with the pure stabilizer bound (8). For $\delta>2$, the comparison with (8) is more delicate because of the floor and ceiling terms, although the two bounds coincide in several cases.

Theorem 4 also matches the dimension bound obtained in the dual-containing classical $(r,\delta)$-LRC setting. Indeed, the local-plus-global recovery argument guarantees correction of an erasure of size $m=(d-1)+f(\delta-1)$, and hence the corresponding classical dimension satisfies $k^{\prime}\leq n-m$. Substituting $k^{\prime}=\frac{n+k}{2}$ yields exactly the bound in Theorem 4 for $K=q^{k}$. Thus the theorem recovers the corresponding classical dimension bound in quantum form.

In this section, we derive an upper bound on the dimension of a pure disjoint $(r,\delta)$-qLRC using an LP approach based on the blockwise weight enumerators from Theorem 2.

Note that the set of uncorrectable blockwise erasure patterns is $\mathcal{T}_{d,\delta}^{\mathrm{c}}=\big\{\boldsymbol{i}\in\mathcal{I}_{N}^{s}\,\big|\,\sum_{j:\,i_{j}\geq\delta}i_{j}\geq d\big\}$.

To compare Theorem 6 with the bounds (7), (8), and (5), we relax the floor and ceiling terms in a way that preserves the direction of the inequalities. Relaxing the floor in (5) gives