Computational Phases of Quantum Matter

• Computational phases of quantum matter are regimes where every ground state consistently functions as a resource for quantum computation using methods like MBQC and error-correcting codes.
• They leverage symmetry representations, tensor network structures, and cohomological invariants to stabilize logical operations and protect against perturbations.
• Practical realizations include MBQC resource states, robust quantum LDPC codes, and machine learning approaches for effective phase classification and diagnostics.

A computational phase of quantum matter is a distinct regime within the landscape of many-body quantum systems in which the underlying ground states or stationary states possess uniform computational properties, particularly as resource states for quantum information processing tasks such as measurement-based quantum computation (MBQC), fault-tolerant coding, or complexity-theoretic criteria. This concept generalizes and unifies familiar notions of symmetry-protected topological (SPT) phases, error-correcting code phases, and symmetry-enriched/topological phases by endowing them with explicit, operational computational power throughout the phase. The emergence of computational phases is tightly interwoven with symmetry representations, tensor-network structure, code-theoretic invariants, circuit complexity, and the structure of entanglement.

1. Conceptual Framework and Definitions

Computational phases of quantum matter are rigorously defined as gapped quantum phases wherein every ground state (or a robust manifold of ground states) serves uniformly as a resource for quantum computational tasks. The defining property is uniform MBQC power: for any ground state $|\psi(\lambda)\rangle$ in the phase $\mathcal{P}$, a fixed measurement protocol enables implementation of the same family of logical gates (i.e. Lie group $G_{\rm comp}$) with computational overhead bounded polynomially in system size or inverse precision (Raussendorf et al., 2018, Weil et al., 6 Jan 2026, Herringer et al., 2024). In error-correcting code phases, such as quantum LDPC code phases, the unifying feature is the stability of extensive ground state degeneracy and logical information storage under local perturbations (Yin et al., 2024). The formalism extends to dissipative and non-equilibrium phases via Lindbladian connectability (Onorati et al., 2023).

Computational phases often—but not necessarily—arise in the presence of symmetry protection (on-site, subsystem, or higher-form symmetry G), where the entanglement and logical operation structure is captured by projective or cohomological invariants (Stephen et al., 2016, Stephen et al., 2018). In higher dimensions, symmetry-enriched or subsystem symmetry-enriched topological (SSET) phases can host computational power protected by nontrivial fractionalization patterns (Herringer et al., 2024).

2. Computational Power: MBQC, Code Phases, and Complexity Criteria

2.1 MBQC Resource Phases

MBQC resource phases are characterized by the ability to perform quantum computation via local measurements on the ground state manifold. In 1D SPT phases, projective representations arising from group cohomology ($H^2(G,U(1))$) determine $G_{\rm comp}$, ensuring full single-qubit universality when the projective symmetry actions are maximally non-commutative (Stephen et al., 2016, Weil et al., 6 Jan 2026). In 2D, subsystem symmetries protect extensive logical qubits, and tensor-network invariants enforce uniform byproduct structures (Raussendorf et al., 2018, Herringer et al., 2024, Stephen et al., 2018). Universality extends to certain SSET phases, while other phases may be nonuniversal or classically simulable (Herringer et al., 2024, Yin et al., 2024).

2.2 LDPC and Code-theoretic Phases

Quantum LDPC code phases with code distance $d(N) = \Omega(\log N)$ and check soundness properties form robust computational phases for error correction; their entire ground-state manifold retains encoding and protection power against arbitrary few-body perturbations even in the thermodynamic limit (Yin et al., 2024). Constant-rate LDPC expander codes exhibit extensive degeneracy and violate the Third Law of thermodynamics, yet remain stable to local noise. This code-theoretic notion generalizes familiar topological order beyond spatial locality and connects to fault-tolerant quantum memory.

2.3 Complexity-Theoretic and Sampling Signatures

A distinct computational criterion arises from complexity-theoretic phase transitions in sampling and phase recognition. For generic states with large correlation length $\xi$, quantum phase recognition is exponentially hard: distinguishing between phases requires at least $2^{\Omega(\xi)}$ quantum time, rendering brute-force tomography optimal in this regime (Schuster et al., 9 Oct 2025). The emergence of Porter–Thomas statistics in dynamical or thermalized phases is used as a computational "order parameter," connecting quantum supremacy and complexity theory directly to physical phase transitions (Thanasilp et al., 2020, Golan, 2021).

3. Symmetry, Tensor Network, and Subsystem Protection

Symmetry-protected computational phases are classified by cohomological data or projective symmetry actions. In 1D, the MPS form and projective group action determine the logical gate group and computational uniformity (Stephen et al., 2016, Weil et al., 6 Jan 2026). In 2D, PEPS representations and tensor-network symmetries guarantee the propagation of logical byproduct operators, protected by global or subsystem symmetries—stripe, line-like, or fractal (Raussendorf et al., 2018, Stephen et al., 2018). Subsystem symmetries enable extensive logical degeneracy and computational power, as in the XZ-star model or anisotropic toric code, where string order correlators dualize to computational order parameters that protect MBQC capability (Herringer et al., 2024).

Duality between string and computational order (SOP ↔ COP) is a conceptual advance for subsystem and SET phases, underpinning the existence and protection of computational phases in topologically ordered regimes (Herringer et al., 2024).

4. Machine Learning and Algorithmic Phase Recognition

Machine learning—both classical and quantum—now plays a central role in computational phase identification and classification. Supervised, unsupervised, and hybrid pipelines deploy CNNs, QCNNs, spectral clustering, and kernel methods to learn order parameters and distinguish phases using data from numerics or experiment (Carrasquilla, 2020, Miles et al., 2021, Liu et al., 2022, Ye et al., 6 Aug 2025, Khosrojerdi et al., 16 Oct 2025, Sander et al., 2024).

QCNNs are used to extract robust multi-site order parameters which remain invariant across the phase (Liu et al., 2022); shadow tomography-based approaches enable phase classification without reference to any local order parameter (Ye et al., 6 Aug 2025). Fidelity-based kernels and support vector machines extract critical exponents and boundaries directly from wavefunction overlaps (Khosrojerdi et al., 16 Oct 2025, Sancho-Lorente et al., 2021). Algorithmic complexity results bound the sample complexity for learning local observables throughout a phase that is Lindbladian-connected: $$O\big(\log(n/\delta)2^{\mathrm{polylog}(1/\epsilon)}\big)$$ samples suffices under rapid mixing (Onorati et al., 2023).

Experimental advances directly study computational phase uniformity, decoherence scaling, and packing effects in MBQC resource states with symmetry (Weil et al., 6 Jan 2026).

5. Hardness Results, Obstructions, and Limitations

Important technical developments elucidate the hardness of phase recognition. Under standard cryptographic assumptions, recognizing quantum phases with large correlation length $\xi$ is exponentially hard for both quantum and classical algorithms (Schuster et al., 9 Oct 2025). The construction of symmetric pseudorandom unitaries (PRUs) produces states indistinguishable from Haar-random by any sub-exponential adversary, establishing that brute-force state learning matches the complexity of phase recognition in worst-case scenarios.

Chiral topological phases manifest intrinsic sign problems, precluding the existence of sign-free Hamiltonians or efficient classical simulation whenever $e^{2\pi i c/24}$ fails to match any topological spin $\theta_a$ (Golan, 2021). This complexity-theoretic obstruction separates computational phases from those that allow efficient classical simulation; many topological quantum-computational phases are thus sign-problematic by construction.

Open questions persist regarding the scalability of computational phases to $k$-local Hamiltonians, extension to continuous symmetry groups, generalization to mixed-state or non-equilibrium resource phases, and the classification of computational order in higher dimensions or beyond group cohomology.

6. Practical Realizations and Applications

Practical exploitation of computational phases has advanced across multiple paradigms:

• MBQC on SPT/SET resource phases: Direct implementation on quantum hardware using symmetry-respecting protocols with polynomial overhead (Weil et al., 6 Jan 2026, Raussendorf et al., 2018).
• LDPC code phases as fault-tolerant quantum memory: Stability under arbitrary local perturbations and non-geometric layouts.
• Machine learning-based phase classification: Application on both synthetic data and experimental quantum simulators, enabling automatic phase discovery and reconstruction of complex phase diagrams, including unconventional orders and edge phenomena (Miles et al., 2021).
• Algorithmic phase recognition and learning: Provably efficient learning in Lindbladian-connected phases, with theoretical sample complexity bounds (Onorati et al., 2023).
• Quantum supremacy diagnostics: Use of computational order parameters connected to Porter–Thomas statistics and hardness criteria (Thanasilp et al., 2020).

7. Classification, Dualities, and Future Directions

Classification of computational phases now draws on algebraic, symmetry, code-theoretic, and complexity-theoretic invariants. Group cohomology and subsystem symmetry representations structure MBQC resource phases; code distance and check soundness underpin LDPC code phases; sampling hardness and intrinsic sign problems specify complexity-theoretic boundaries (Herringer et al., 2024, Yin et al., 2024, Golan, 2021).

Duality between string and computational order is emerging as a universal organizing principle for SPT, SET, and SSET phases, facilitating the identification of uniform computational power throughout topologically ordered regions (Herringer et al., 2024, Stephen et al., 2018).

Challenges remain in the explicit construction of computational order parameters for exotic phases, the fine-grained classification of nonlocal or entanglement-based computational order, extension to mixed and driven phases, and the delineation of computational phase boundaries via algorithmic hardness. A full dictionary of "resource phases" and their operational quantum computational properties is an active and foundational topic in the quantum physics community.