Spectral Properties Versus Magic Generation in $T$-doped Random Clifford Circuits

Technical Summary: Spectral Properties Versus Magic Generation in T-doped Random Clifford Circuits

Problem Statement
Quantum dynamics often generate states that are intractable for classical simulation, a phenomenon linked to both entanglement and "non-stabilizerness" (or "magic"). While Clifford circuits can generate extensive entanglement, they remain efficiently simulable classically due to their stabilizer structure and lack of magic. Universal quantum computation requires the injection of non-Clifford gates, such as T-gates. This work investigates how the introduction of T-gates into random Clifford circuits induces complexity, comparing two distinct indicators: spectral properties (eigenvalue statistics) and magic generation (quantified by Stabilizer Rényi Entropy). The central question is how the transition from integrable-like Clifford behavior to chaotic, Haar-random behavior manifests differently in these two measures.

Methodology
The authors analyze an ensemble of deep random $N$-qubit circuits constructed with a brick-wall architecture. The circuits consist of alternating layers of random 2-qubit Clifford gates and layers where T-gates are randomly injected.

• Spectral Analysis: The study examines the eigenvalue spectrum of the unitary operator $U$ and the associated operator $\mathcal{C} U \mathcal{C}^\dagger$ acting on the space of Pauli strings. The authors relate the spectrum to the distribution of periodic orbits in the Pauli string space. They compute the phase correlation function $\chi(\Theta)$ and level spacing statistics $P(\zeta)$ to compare the ensemble against Random Matrix Theory (RMT) predictions for the Circular Unitary Ensemble (CUE).
• Magic Generation: The authors quantify complexity using the Stabilizer Rényi Entropy (SRE), $M_2$, which measures non-stabilizerness. They define the "non-stabilizing power" of a circuit as the average SRE generated when the circuit acts on random stabilizer states. Numerical simulations are performed for various system sizes $N$ and T-gate counts $N_T$, averaging over both circuit ensembles and initial stabilizer states.

Key Contributions and Results

1. Spectral Properties of Pure Clifford Circuits:

   • Pure Clifford circuits exhibit a unique spectral structure distinct from both Poissonian (integrable) and CUE (chaotic) statistics.
   • This structure arises from a decomposition of the dynamics into periodic orbits in the space of Pauli strings. These orbits lead to large spectral degeneracies and a distribution of eigenphases characterized by discrete peaks at rational multiples of $\pi$.
   • The distribution of orbit lengths is sparse for large orbits, with the maximum orbit length scaling as $L_{\text{max}} = 2^{N+1}$, significantly smaller than the total number of string configurations ($4^N$).

2. Spectral Transition via T-doping:

   • Injecting T-gates disrupts the periodic orbits, causing the spectral properties to converge toward RMT (CUE) behavior.
   • Sensitivity: The spectral transition is extremely rapid. The authors find that an $O(1)$ number of T-gates is sufficient to remove spectral degeneracies and induce chaotic level spacing statistics in the thermodynamic limit ($N \to \infty$).
   • The suppression of the delta peak at $\Theta = \pi$ (a signature of periodic orbits) occurs exponentially with the number of T-gates. Specifically, the fraction of degeneracies scales as $\sim e^{-c N N_T}$, vanishing for any finite $N_T$ as $N \to \infty$.

3. Magic Generation Dynamics:

   • In contrast to the spectral transition, magic generation is a slower, more gradual process.
   • Discrete Regime ($N_T \ll N$): For a small number of T-gates, magic is generated in discrete "quanta." The distribution of magic density $m_2$ is discrete and bimodal. For a single T-gate, the possible values are $m_2 = 0$ and $m_2 = M_2^H/N$, where $M_2^H \approx 0.414$ corresponds to the magic of H-type states.
   • Linear Scaling: In the dilute limit, the average non-stabilizing power scales linearly with the number of T-gates: $\langle M_2 \rangle \approx M_2^\infty(T) N_T$, where $M_2^\infty(T) \approx 0.414$.
   • Saturation: As $N_T$ increases to the order of $N$, the distribution becomes quasi-continuous and eventually converges to the distribution of Haar-random unitaries. The average magic density saturates at a value approaching 1 (specifically $1 - 2/N$) for $N_T \gg N$.
   • Phase Transition: The results support a phase transition at a critical T-gate density $n_T^* \approx 2.41$. Below this density, the average magic density is proportional to the T-gate density ($\langle m_2 \rangle = n_T / n_T^*$); above it, the system reaches the maximal complexity of Haar-random states.

Significance and Claims
The paper concludes that spectral properties and magic generation serve as complementary but distinct indicators of quantum complexity.

• Spectral Correlations: These are highly sensitive to the presence of non-Clifford gates. The destruction of periodic orbits and the onset of chaotic spectral statistics occur with a constant number of T-gates, regardless of system size.
• Magic Generation: This is a more sensitive indicator of the degree of complexity. While the spectrum becomes chaotic almost immediately, the generation of magic requires an extensive number of T-gates ($N_T \sim O(N)$) to reach the maximal complexity of Haar-random states.
• Conclusion: The authors assert that magic is a more refined measure of the "resource-rich" nature of quantum circuits than spectral statistics alone. The work highlights that while a few T-gates are sufficient to break the integrability-like structure of Clifford circuits (spectral transition), a finite density of T-gates is required to fully unlock the computational resources (magic) necessary for universal quantum advantage.