Resource Theory of Stabilizer Computation

• Resource Theory of Stabilizer Computation is a framework that rigorously quantifies magic as a non-classical resource enabling universal quantum computation beyond the Clifford group.
• Magic monotones like Wigner negativity, Mana, and robustness of magic provide precise measures for conversion rates, simulation costs, and operational significance in quantum circuits.
• The theory links contextuality with quasiprobability representations to outline classical simulation boundaries and informs scalable algorithms for near-term quantum technologies.

The resource theory of stabilizer computation rigorously quantifies the non-classical resources—collectively termed "magic"—required for universal quantum computation beyond the Clifford group. This formalism provides a unified framework for classifying free vs. resource states and operations, introducing monotones that measure nonstabilizerness and contextuality, and establishing conversion rates and simulation costs for quantum circuits augmented with magic. Key elements include the stabilizer polytope, monotones such as Wigner negativity, robustness of magic, stabilizer Rényi entropies, and the operational role of contextuality. Recent advances give precise operational and field-theoretic interpretations, as well as scalable algorithms for resource quantification.

1. Foundational Structure: Free Operations, States, and Magic Monotones

The stabilizer resource theory is a subtheory of quantum computation defined by the Clifford group acting on $n$ qubits, together with preparation and measurement of stabilizer states (i.e., the common +1 eigenstates of an abelian subgroup of the $n$-qubit Pauli group), and classical feed-forward. The set of free states $\mathcal{F}_n$ is the convex hull of all $n$-qubit pure stabilizer states, and free operations are quantum channels composed of Clifford unitaries, Pauli measurements, stabilizer state preparation, ancilla addition/discarding, and classical randomness or conditioning (Veitch et al., 2013, Veitch et al., 2012, Schmid et al., 2021). Explicitly,

• Free pure states: $$S_\text{pure} = \{U|0\rangle^{\otimes n}: U \in \text{Clifford}\}$$,
• Full stabilizer polytope: $$\text{STAB}(\mathcal{H}_{d^n}) = \text{conv}(S_\text{pure})$$.

States $\rho \notin \mathcal{F}_n$ are termed magic states and are necessary input resources for implementing non-Clifford gates via state injection, magic-state distillation, or gate teleportation, yielding universality (Veitch et al., 2013).

Crucial to the resource theory are magic monotones: real-valued, convex measures $M(\rho)$ that do not increase under free operations, quantifying "distance from free." The primary monotones for stabilizer computation are:

• Wigner sum-negativity $M_{\text{neg}}$ (Veitch et al., 2012, Veitch et al., 2013, Schmid et al., 2021): $$M_{\text{neg}}(\rho) = \sum_{u: W_\rho(u) < 0} |W_\rho(u)|$$ for the discrete Wigner function $W_\rho(u)$.
• Mana: $M(\rho) = \log \|W_\rho\|_1$.
• Robustness of magic $R(\rho)$ (Veitch et al., 2013): minimal $t$ such that $(\rho + t \sigma)/(1 + t) \in \mathcal{F}_n$ for some stabilizer state $\sigma$.
• Stabilizer Rényi entropies (SRE), including the linear stabilizer entropy, e.g., $H_2(\psi) = -\log \sum_P p_P(\psi)^2$ with $p_P(\psi) = |\langle \psi|P|\psi\rangle|^2/2^n$ (Leone et al., 2024, Hoshino et al., 17 Mar 2025).

Each of these measures is additive on tensor products, non-increasing under free operations, and vanishes if and only if the state is stabilizer.

2. Contextuality and Quasiprobability: The Role of the Discrete Wigner Function

In odd dimensions, the unique nonnegative, diagram-preserving quasiprobability representation for the stabilizer subtheory is Gross's discrete Wigner function (Schmid et al., 2021). The Wigner representation assigns to any state $\rho$ the function

$$W_\rho(u) = \frac{1}{d} \mathrm{Tr}[A_u \rho], \qquad \sum_u W_\rho(u) = 1,$$

where $\{A_u\}$ are phase-space point operators, covariant under the Clifford group. Stabilizer states are those with $W_\rho(u) \geq 0$ for all $u$. Negativity $N(\rho)$ of the representation is monotonic under all free operations and signals a non-classical resource.

Crucially, positivity of the Wigner function is both necessary and sufficient for efficient classical simulation of Clifford circuits with stabilizer inputs (Veitch et al., 2012, Veitch et al., 2013). Negativity hence precisely marks the onset of classical intractability and quantum speed-up, and is equivalent to the emergence of generalized contextuality (Schmid et al., 2021).

In even dimension, no nonnegative, diagram-preserving quasiprobability representation exists; thus, even the pure stabilizer subtheory is inherently contextual in this sense (Schmid et al., 2021).

3. Stabilizer Rényi Entropies: Definition, Monotonicity, and Operational Role

The stabilizer Rényi entropy of order $\alpha$ is defined for pure states as: $$H_\alpha(\psi) = \frac{1}{1-\alpha} \log \left( \sum_P p_P(\psi)^\alpha \right),$$ where the sum runs over the $n$-qubit Pauli group, and $p_P(\psi) = |\langle \psi|P|\psi\rangle|^2 / 2^n$ (Leone et al., 2024, Hoshino et al., 17 Mar 2025, Bittel et al., 30 Jul 2025).

For $\alpha \geq 2$, $H_\alpha$ is a magic monotone—provably non-increasing under all stabilizer protocols, including Clifford unitaries, Pauli measurements (with post-selection or feed-forward), and stabilizer ancilla manipulations. The linear entropy $H_\text{lin} = 1 - \sum_P p_P(\psi)^2$ is a strong monotone: for any protocol with probabilistic branches $\{p_i, |\psi_i\rangle\}$,

$$H_\text{lin}(\psi) \geq \sum_i p_i H_\text{lin}(\psi_i).$$

Convex roof extensions to mixed states $$H_\alpha(\rho) = \min_{\{p_i, \psi_i\}} \sum_i p_i H_\alpha(\psi_i)$$ preserve monotonicity and permit efficient numerical evaluation for low-rank matrices (Leone et al., 2024).

SREs directly bound conversion rates and success probabilities in magic-state protocols, e.g., the number $n$ of $|C_{m-1}Z\rangle$ states required to produce $m$ $|C_2Z\rangle$ states via stabilizer operations is bounded by $n \geq m \cdot H_2(C_2Z)/H_2(C_{m-1}Z)$ (Leone et al., 2024).

Table: Faithful Magic Monotones and Their Properties

Monotone	Monotonicity	Additivity	Computability	Operational Meaning
Wigner negativity $N(\rho)$, mana $M(\rho)$	all stabilizer protocols	Yes	Easy (qudit)	Onset of simulability, necessary for universality
Robustness $R(\rho)$	all stabilizer protocols	Yes	Optimization	Mixing cost to reach free set
Stabilizer Rényi entropy $H_\alpha$	all stabilizer protocols	Yes	Scalable	Asymptotic rates, operational in property testing

Monotones such as SREs are experimentally accessible and computationally tractable compared to robustness-based monotones that require optimization over exponentially large state spaces (Leone et al., 2024, Bittel et al., 30 Jul 2025).

4. Contextuality and Resource Theory: Generalized Notions

Resource theories of stabilizer computation are deeply connected to contextuality. Generalized contextuality arises wherever there is no preparation-, transformation-, and measurement-noncontextual ontological model (Lillystone et al., 2018). Even the single-qubit stabilizer subtheory is either preparation contextual or transformation contextual, with the latter occurring for Clifford unitaries (Lillystone et al., 2018). Prepare-and-measure scenarios may admit a non-contextual (Spekkens toy) model, but inclusion of transformations reveals operational equivalences not classically simulable.

Magic monotones such as Wigner negativity and SREs can be interpreted as measures of computational contextuality—quantifying how "far" a state or channel is from being simulateable by noncontextual hidden variable models. In odd $d$, the uniqueness of the Gross representation ensures that negativity directly corresponds to contextuality and quantum advantage (Schmid et al., 2021).

5. Operational Interpretations: Distillation, Conversion, and Simulation

Magic monotones—particularly stabilizer entropy—have sharp operational meanings:

• Conversion bounds: For protocols aiming to convert one magic resource to another, additivity of e.g., SRE sets upper/lower bounds on conversion rates and success probabilities. There is a preferred direction in conversions, as shown e.g. for multi-controlled $Z$ states where $r(m) \leq H_2(C_{m-1}Z)/H_2(C_2Z)$ decays exponentially in $m$ (Leone et al., 2024).
• Distillation no-go theorems: Any state with Wigner-positive representation cannot, by stabilizer protocols, be distilled to a pure non-stabilizer state; bound magic states exist analogously to bound entangled states (Veitch et al., 2012).
• Efficient simulation: Stabilizer circuits with positive-Wigner input admit efficient classical Monte Carlo simulation via phase-space sampling (Veitch et al., 2012). The simulation cost for non-stabilizer states is governed by monotones such as the stabilizer extent, which can now be computed efficiently up to 9–10 qubits (Hamaguchi et al., 2024).
• Clifford+$k T$ free resource theory: Generalizations of robustness to "Clifford+$kT$-robustness" track resource requirements in restricted fault-tolerant settings, showing explicitly how classical or early-FTQC cost scales as $O(R_k(\rho)^2)$ for estimating expectation values, and how sampling cost drops as $k$ increases (Nakagawa et al., 20 Aug 2025).

6. Stabilizer Entropy in Many-Body Systems and Field Theory

The stabilizer Rényi entropy admits a field-theoretic interpretation as the participation entropy of Born probabilities in the Bell basis, and can be computed as a partition function in a replicated boundary conformal field theory (BCFT) with an interlayer defect (Hoshino et al., 17 Mar 2025). Universal features such as the size-independent $g$-factor term and logarithmic scaling of mutual SRE in critical 1D systems have been analytically derived and numerically confirmed (e.g., at the Ising critical point).

• Universal size-independent shift: $c_\alpha = \ln g_1/(\alpha-1)$, where $g_1$ is the Affleck-Ludwig $g$-factor.
• Mutual SRE scaling: $$W_\alpha(\ell) \sim (4 \Delta_{2\alpha}/(\alpha-1)) \ln \ell$$ for intervals $\ell$ and scaling dimension $\Delta_{2\alpha}$.

These results establish SRE as a bona fide "magic monotone" with universal signatures in quantum many-body systems (Hoshino et al., 17 Mar 2025).

7. Simulation, Algorithmics, and the Magic Gap

Scalable algorithms allow efficient computation of stabilizer monotones:

• Purity-encoding protocols: $\alpha$-stabilizer Rényi entropies can be extracted from the purity of a Pauli-twirled channel acting on $\alpha$ copies; swap-test or random measurement variants achieve copy-efficient estimation and are fault-tolerant (Stratton, 3 Jul 2025).
• Stabilizer extent: New algorithms exploit column generation and efficient overlap calculation with stabilizer states, pushing exact resource quantification for Haar random or structured states up to 10 qubits (real case), thereby benchmarking classical simulators and informing gate synthesis resource lower bounds (Hamaguchi et al., 2024).
• Embedding subspaces and the magic gap: Judicious embedding of a small system in a larger Hilbert space (e.g., via quantum error correction or symmetry subspaces) can minimize or even reduce the average stabilizer entropy—creating a "magic gap" which quantifies the change in resourcefulness (Cepollaro et al., 28 Dec 2025). For example, certain stabilizer codes exhibit zero or negative gaps, which can facilitate more efficient magic-state distillation and lower classical simulation cost.

8. Catalytic and Asymptotic Resource Theory

The regularized relative entropy of magic is the unique asymptotic monotone for catalytic resource manipulations under stabilizer operations (Anshu et al., 2017). For any magic state $\rho$: $$E_\infty(\rho) = \lim_{n \to \infty} \frac{1}{n} \min_{\sigma \in \mathcal{F}_n} D(\rho^{\otimes n} \| \sigma)$$ governs the optimal rate at which $\rho$ can be converted to standard magic states (e.g., $|T\rangle$) catalytically, and also the one-shot resource cost via the smooth max-relative entropy.

9. Classical Rewriting and Contextuality in Circuit Simulation

For CSS-preserving stabilizer circuits (i.e., circuits constructed from prep $|0\rangle,|+\rangle$, Pauli gates, CNOT, global Hadamard, Pauli measurements, and classical control), there is an exact, zero-overhead rewriting as classical probabilistic circuits, with corresponding hidden-variable models (noncontextual) (Yashin et al., 7 Nov 2025). General stabilizer (Clifford) circuits require frame-tracking or quadratic form expansions; non-CSS operations introduce contextuality, necessitating an $O(n)$ computational overhead per non-CSS gate and reflecting the resource nature of contextuality.

---

In summary, the resource theory of stabilizer computation provides a rigorous mathematical and operational framework for quantifying and manipulating non-Clifford resources in quantum computation. Its structure underpins major results in classical simulation, fault tolerance, magic-state distillation, and complexity theory, with recent advances yielding scalable quantification, field-theoretic insights, and algorithmic techniques critical for near-term quantum technologies (Veitch et al., 2012, Veitch et al., 2013, Leone et al., 2024, Hoshino et al., 17 Mar 2025, Stratton, 3 Jul 2025, Cepollaro et al., 28 Dec 2025, Yashin et al., 7 Nov 2025, Bittel et al., 30 Jul 2025, Nakagawa et al., 20 Aug 2025, Anshu et al., 2017, Lillystone et al., 2018, Schmid et al., 2021).