Holographic Quantum Error Correction

• Holographic quantum error correction is a framework that encodes low-energy bulk effective-field theory into boundary quantum codes, ensuring subregion duality.
• The approach leverages operator-algebra techniques to reformulate geometric concepts like the Ryu–Takayanagi formula, modular relations, and area operators.
• Tensor-network models and hyperbolic code families illustrate practical implementations, fault tolerance, and feasibility studies leading to experimental verifications.

Holographic quantum error correction is the proposal that the bulk–boundary map of holography, especially in AdS/CFT, is an operator-algebra quantum error-correcting code: low-energy bulk effective-field-theory degrees of freedom are encoded isometrically into boundary degrees of freedom, with bulk operators in an entanglement wedge reconstructible on the corresponding boundary subregion, and with entropy formulas governed by a central “area operator” that captures geometric data (Jahn et al., 2021). In this formulation, bulk locality, subregion duality, Ryu–Takayanagi/HRT entropy, JLMS modular relations, and quantum extremal-surface corrections are recast as structural statements about correctable algebras, commutants, and centers.

1. Origins in subregion duality and quantum error correction

The modern formulation begins from the observation that the same bulk operator can admit inequivalent boundary reconstructions supported on different boundary regions, while acting identically on a restricted code subspace. Almheiri, Dong, and Harlow made this precise by identifying the AdS/CFT code subspace with a low-energy bulk sector and expressing reconstruction as

$$V^\dagger (O_A \otimes I_{\bar A}) V = O_{\mathrm{bulk}},$$

for an encoding isometry $$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$ and a boundary factorization $$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$ (Almheiri et al., 2014). In this picture, radial depth becomes a coding-theoretic notion: bulk operators deeper in the bulk require larger boundary regions for reconstruction and are more resilient to boundary erasures (Almheiri et al., 2014).

Operator-algebra quantum error correction is the natural framework because holography does not generally involve a full matrix algebra of logical observables. Instead, one corrects a logical subalgebra $\mathcal A$, with erasures on $\bar A$ constrained by

$P E_i^\dagger E_j P \in \mathcal A',$

rather than by the ordinary Knill–Laflamme scalar condition. This allows nontrivial centers, sector decompositions, and algebraic no-cloning, all of which are intrinsic to holographic subregion duality (Almheiri et al., 2014). The broader review literature places RT/HRT, FLM bulk entropy corrections, JLMS relative entropy, and entanglement wedge reconstruction within this same OQEC/OAQEC architecture (Jahn et al., 2021).

2. Operator algebras, centers, and the code-subspace structure

In finite-dimensional OQEC, the logical Hilbert space admits a Wedderburn-type decomposition

$$\mathcal H_{\mathrm{code}} \cong \bigoplus_{\alpha}\,\mathcal H_{a_\alpha}\otimes \mathcal H_{\bar a_\alpha},$$

with a von Neumann algebra of reconstructable operators acting on one factor and its commutant on the other. The label $\alpha$ indexes superselection sectors and spans the center $Z_M$, which in holography carries geometric data such as area (Jahn et al., 2021). The JLMS relation is the operator statement that, on a boundary region $A$,

$$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$0

with $$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$1 central in the relevant boundary algebra (Jahn et al., 2021).

A constructive characterization of this structure identifies the algebra visible from $$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$2 as

$$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$3

If $$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$4 is a von Neumann algebra, it is the unique algebra satisfying complementary recovery for the code and the chosen erasure model (Pollack et al., 2021). The same framework yields the operator-algebraic RT formula

$$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$5

where $$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$6 is the algebraic entropy and $$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$7 is the area operator (Pollack et al., 2021).

Recent exact models built from the boundary rather than from bulk tensor networks sharpen this picture. In those constructions, bulk qudits emerge from boundary entanglement patterns, complementary recovery is established for arbitrary boundary bipartitions, and $$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$8 for any bipartition, giving a genuine OA-QEC realization with explicit split algebras on the entangling surface (Wang, 2024). This suggests that the algebraic content of holography can be derived directly from boundary entanglement data rather than assumed from a bulk tiling.

3. Entropy, area operators, and refined Rényi structure

At leading semiclassical order, holographic entropy is governed by the RT/HRT/QES hierarchy. For a boundary region $$V:\mathcal H_{\mathrm{code}}\to\mathcal H_{\mathrm{CFT}}$$9,

$$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$0

with HRT replacing RT in time-dependent settings, and QES defined by extremizing the generalized entropy

$$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$1

(Jahn et al., 2021). In OQEC this becomes an algebraic split between a central area term and a bulk logical entropy term.

The refined Rényi analysis adds a stronger constraint. With

$$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$2

Akers and Rath showed that the OQEC formula reproduces the Dong cosmic-brane prescription only if the edge-mode state $$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$3 in every $$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$4-sector has flat refined Rényi spectrum, equivalently iff $$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$5 is maximally mixed on its support (Akers et al., 2018). Under that necessary-and-sufficient condition, the refined Rényi entropy takes the RT-like form

$$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$6

with brane tension $$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$7 (Akers et al., 2018).

This result refines the meaning of the area operator. When $$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$8 is maximally mixed, the OQEC area operator reduces to

$$\mathcal H_{\mathrm{CFT}}=\mathcal H_A\otimes\mathcal H_{\bar A}$$9

so the RT term is interpreted as edge-mode entropy, closely analogous to the $\mathcal A$0 surface term in lattice gauge theory (Akers et al., 2018). The same analysis reinterprets a single holographic tensor network not as a smooth semiclassical geometry but as an area eigenstate; nontrivial Rényi $\mathcal A$1-dependence then arises from superpositions over distinct area sectors rather than from nonmaximal bond entanglement (Akers et al., 2018).

4. Tensor-network realizations and geometric code families

Tensor networks supply explicit, exactly solvable or approximately solvable realizations of HQEC. Perfect tensors, random tensors, CSS/block-perfect variants, Majorana-dimer formulations, and hyperinvariant constructions each capture different aspects of entanglement geometry, recovery, and correlator structure (Pastawski et al., 2015).

Model family	Salient structure	Citation
HaPPY / perfect-tensor codes	Exact bulk-to-boundary isometry, operator pushing, greedy entanglement wedge, and $\mathcal A$2 for connected regions	(Pastawski et al., 2015)
Random tensor networks	Typical RT/HRT behavior at large bond dimension $\mathcal A$3, replica-statistical-model mapping, approximate wedge reconstruction	(Jahn et al., 2021)
CSS/block-perfect heptagon code	Steane-based block-perfect tensors on $\mathcal A$4, asymptotic rate $\mathcal A$5, erasure threshold $\mathcal A$6	(Harris et al., 2018)
Majorana-dimer HyPeC	Boundary dimers coincide with discrete bulk geodesics; entropy is dimer counting; dimer states have flat Rényi spectrum	(Jahn et al., 2019)
HMERA / hyperinvariant codes	Multi-tensor constructions support power-law correlations; exact hyperinvariant codes yield nontrivial scaling spectra and state-dependent complementary-recovery breakdown	(Cao et al., 2021, Steinberg et al., 2023)

The original HaPPY analysis established exact RT-like entropy, negative tripartite information in appropriate regimes, disconnected-region entanglement-wedge reconstruction, and threshold phenomena in suitably modified hyperbolic codes (Pastawski et al., 2015). The Steane-based heptagon construction broadened this by relaxing perfection to block perfectness, preserving greedy reconstruction and RT-like scaling while enabling CSS structure and efficient erasure decoding (Harris et al., 2018). Majorana-dimer formulations made the geometry–entanglement correspondence particularly explicit: boundary dimers trace discrete geodesics, and each dimer crossing contributes $\mathcal A$7 to the entropy, yielding $\mathcal A$8 in HyPeC (Jahn et al., 2019).

A central limitation of perfect-tensor networks is their overly flat local scaling structure. HMERA proved that a completely regular, locally contractible, single-tensor-type hyperinvariant MERA cannot sustain nontrivial connected boundary two-point functions, motivating multi-tensor constructions with nontrivial transfer spectra (Cao et al., 2021). Exact hyperinvariant holographic codes then supplied power-law boundary correlators together with an exact encoding isometry and a controlled, state-dependent soft breakdown of complementary recovery (Steinberg et al., 2023).

Independent of any specific tensor network, hyperbolic geometry itself constrains the coding rate. For holographic codes grown on regular $\mathcal A$9 tessellations, the asymptotic rate obeys

$\bar A$0

and for a large class of tessellations with $\bar A$1 one has $\bar A$2, guaranteeing asymptotic QEC. Under the tile-completion rule, holographic triangle codes are exceptional, with $\bar A$3, whereas non-triangle families remain QEC-capable (Bray-Ali et al., 2019).

5. Black holes, approximate codes, and non-AdS extensions

In semiclassical gravity, QES and the generalized entropy reframe black-hole information flow as changes in correctable erasures. The island formula

$\bar A$4

describes a transition in entanglement wedges and hence in the logical algebra reconstructable from the radiation region (Jahn et al., 2021). Within HQEC this is an algebraic wedge transition rather than a purely geometric curiosity.

The projected-black-hole-interior construction makes this dynamical flexibility explicit. Starting from a subsystem erasure code for the eternal black hole/TFD and projecting one boundary, the bulk/boundary dictionary is “rewired” so that interior operators become reconstructable on the remaining side. In the general OAQEC theorem of that construction, an operator $\bar A$5 remains reconstructable after projection iff

$\bar A$6

and the surviving boundary representative is given by a transpose-channel/Petz-type formula $\bar A$7 (Almheiri, 2018). This places state dependence behind horizons in a controlled OAQEC framework.

Beyond AdS tensor networks, bilocal holography in the free $\bar A$8 vector model supplies an explicit bulk map from gauge-invariant bilocals to higher-spin AdS fields. After restricting to a code subspace, semicircular bilocal smearing implements entanglement-wedge reconstruction, and gauge-invariant entangled pairs spread over bulk semicircles in a manner explicitly suggestive of bit threads (Koch et al., 2021). At the opposite infrared and asymptotic end, celestial QEC encodes hard states with quantized BMS hair into a celestial CFT, with soft radiation modeled as correctable errors in a GKP-like framework governed by a $\bar A$9 hierarchy of soft currents (Guevara et al., 2024). Related approximate-QEC ideas also appear in quantum source-channel codes, where Gibbs states of the critical transverse-field Ising model show nontrivial protection against local erasure, giving a state-based CFT realization of approximate HQEC (Pastawski et al., 2016).

6. Coding performance, fault tolerance, and implementation

Recent work has pushed HQEC from conceptual toy models toward coding-theoretic performance and experimental design. Heterogeneous holographic codes alternate two seed codes layer by layer on hyperbolic tilings, thereby combining complementary transversal gate sets. This realizes universal fault-tolerant logic on the holographic boundary, supports black-hole deformations that encode multiple logical qubits, and can reduce physical-qubit overhead by $P E_i^\dagger E_j P \in \mathcal A',$0 in a two-layer Steane/quantum-Reed–Muller combination (Steinberg et al., 14 Apr 2025). In that framework, erasure thresholds reach $P E_i^\dagger E_j P \in \mathcal A',$1 for heterogeneous HaPPY/Steane and Steane/HaPPY families, and approach $P E_i^\dagger E_j P \in \mathcal A',$2 for HaPPY/QRM (Steinberg et al., 14 Apr 2025).

Under biased Pauli noise, asymptotically zero-rate holographic stabilizer codes with tensor-network maximum-likelihood decoders attain competitive thresholds. Across five model families, all reach the zero-rate hashing bound in some bias regime, and the holographic surface-code fragment appears to exceed the known benchmark in the pure XZ two-Pauli regime, with $P E_i^\dagger E_j P \in \mathcal A',$3 against a hashing-bound reference of approximately $P E_i^\dagger E_j P \in \mathcal A',$4 (Fan et al., 2024). Clifford-deformed seeds further allow hashing-bound performance in one-Pauli-dominated regimes (Fan et al., 2024). This suggests that hyperbolic bulk–boundary architecture can be compatible with modern biased-noise optimization rather than opposed to it.

Experimental implementation has likewise become concrete. A stabilizer-graph reformulation of the hyperbolic pentagon code yields explicit encoding and partial-decoding circuits tailored to long-range interactions and demonstrates a twelve-qubit instance in which partial decoding recovers bulk qubits from nearby boundary qubits (Munné et al., 2022). For logical-zero preparation, the reported trapped-ion benchmark gives an estimated fidelity $P E_i^\dagger E_j P \in \mathcal A',$5, compared with $P E_i^\dagger E_j P \in \mathcal A',$6 for a non-destructive stabilizer-measurement route (Munné et al., 2022). This does not settle the physical realization of full-scale HQEC, but it places small experimental verifications of bulk reconstruction within current hardware capabilities.

Taken together, these developments present holographic quantum error correction as a layered subject: an operator-algebraic formulation of subregion duality, an entropy framework centered on area operators and edge modes, a family of tensor-network and algebraic models with varying degrees of exactness and realism, and an emerging coding-theoretic program aimed at thresholds, fault tolerance, and laboratory implementation. The most persistent open problems concern the derivation of edge-mode maximal mixing from gravity, the construction of exact codes with nontrivial area operators and realistic correlators simultaneously, the treatment of approximate large-$P E_i^\dagger E_j P \in \mathcal A',$7 and infinite-dimensional effects, and the extension of the HQEC logic beyond AdS/CFT while preserving a comparably sharp algebraic notion of reconstructability (Akers et al., 2018).