Geometric Complexity of Quantum Channels via Unitary Dilations

Technical Summary: Geometric Complexity of Quantum Channels via Unitary Dilations

Problem Statement
The paper addresses the conceptual gap between the well-established geometric approach to quantum circuit complexity for closed systems (pioneered by Nielsen et al.) and the lack of a rigorous complexity theory for open quantum systems. In closed systems, complexity is defined as the minimal geodesic length on $SU(N)$ required to synthesize a target unitary. However, open-system dynamics are described by families of quantum channels $\{\Lambda_t\}_{t \ge 0}$, which admit infinitely many inequivalent microscopic realizations via Stinespring dilations (differing in environment dimension, preparation, and coupling). Consequently, a straightforward extension of geometric complexity is ill-defined: the "cost" of a channel depends on which microscopic degrees of freedom are counted as resources and which are treated as gauge. The authors aim to construct a geometric complexity framework that distinguishes between implementation-dependent costs and an intrinsic channel complexity, while accounting for the non-uniqueness of dilations.

Methodology
The authors develop a framework based on unitary dilations and right-invariant Riemannian metrics on the Lie group $SU(N)$, where $N$ is the dimension of the total system-environment Hilbert space.

1. Two-Layer Framework:

   • Implementation-Dependent Complexity: Defined for a specific Stinespring dilation $D = (H_E, \rho_E, H_{tot})$. The complexity is calculated based on the geometric cost of the joint unitary evolution.
   • Intrinsic Channel Complexity: Defined as the infimum of the implementation-dependent complexity over a physically admissible class of dilations (constrained by resource models such as bounded environment dimension or interaction strength).

2. Subtractive Definition:
   To define the implementation-dependent complexity, the authors introduce a "subtractive" functional. They decompose the total generator $H_{tot}(t)$ into system-local, interaction, and purely environmental components. The complexity functional is constructed by removing the purely environmental contribution ($P_E(H_{tot})$) from the geometric cost. This subtraction is not arbitrary but derived from three structural postulates:

   • Closed-system consistency: If the dynamics are unitary, the complexity must reduce to Nielsen's unitary complexity.
   • Environment-only neutrality: If the generator acts solely on the environment, the complexity must be zero.
   • Gauge invariance: The complexity must be invariant under unitary changes of basis in the environment that leave the reduced channel unchanged.

   The resulting effective generator is $\hat{H}_{eff}(t) = \hat{H}_{tot}(t) - \hat{P}_E(\hat{H}_{tot}(t))$, and the complexity is the integrated norm of this effective generator.

3. Noise Complexity:
   A companion quantity is defined to measure the geometric deviation between a noisy implementation and a prescribed ideal closed-system evolution. This quantifies the cost of mitigating noise or the geometric distance from the ideal trajectory.

4. Technical Tools:

   • Coherence-based Bounds: The authors relate geometric complexity to the "cohering power" of unitaries relative to a fixed basis, establishing lower bounds for both unitary and noise complexity.
   • Weak-Coupling Regime: The framework is applied to the Davies-motivated weak-coupling limit, connecting the geometric functionals to GKSL (Lindblad) master equations.

Key Contributions

• Formal Definition of Channel Complexity: The paper introduces the functional $G^{\text{imp}}_{b\Omega}(\Lambda_t; D)$, which separates implementation costs from intrinsic channel costs by minimizing over admissible dilations.
• Derivation of the Subtractive Term: The authors prove that the subtraction of environmental terms is the unique solution satisfying the postulates of closed-system consistency, environmental neutrality, and gauge invariance.
• Operational Lower Bounds: They establish coherence-based lower bounds for unitary geometric complexity and noise complexity, linking geometric cost to the ability of dynamics to generate or suppress coherence.
• Structural Properties: The paper proves linear time-scaling of complexity under time-homogeneous dilation models and derives explicit dissipator-controlled bounds in the Markovian (GKSL) regime.
• Benchmark Analysis: The framework is applied to canonical single-qubit channels (dephasing, amplitude damping, and depolarizing noise), illustrating how implementation-dependent complexity and noise complexity behave in standard physical models.

Results

• Coherence Bounds: In the Hilbert-Schmidt setting, the geometric complexity of a unitary is bounded from below by its ability to produce coherence relative to a reference basis. This bound extends to noise complexity via the relative noise unitary.
• Weak-Coupling Scaling: For Davies-type dilations, the implementation-dependent channel complexity scales linearly with time $t$. The bounds are controlled by the system Hamiltonian norm and the dissipator scale $\Gamma$ (sum of squared Lindblad operator norms), with corrections of order $O(\lambda t)$ where $\lambda$ is the weak-coupling parameter.
• Benchmark Findings:
  • Dephasing: The coherent distortion complexity vanishes for standard dephasing dilations (as the generator is purely interaction), yet the full implementation-dependent complexity remains non-zero due to interaction terms.
  • Amplitude Damping & Depolarizing: These models exhibit non-zero coherent distortion and show that noise complexity grows with the square root of the effective dissipative rates.
  • Intrinsic vs. Implementation: The paper notes that computing the exact intrinsic complexity (the global minimum over all dilations) is a non-convex variational problem and generally intractable; thus, the benchmarks provide upper bounds on the intrinsic complexity.

Significance and Claims
The authors claim that their framework provides a "conceptually clean" and "technically powerful" method to quantify the cost of noisy implementations beyond simple gate counting. By distinguishing between implementation-dependent and intrinsic costs, the framework avoids conflating physically distinct realizations of the same channel.

The paper emphasizes that channel complexity is inherently resource-relative; it depends on both the admissible class of dilations (physical constraints) and the choice of the penalty tensor $b\Omega$ (geometric weighting). The subtractive structure ensures that the complexity measure is sensitive only to the degrees of freedom that genuinely affect the reduced system dynamics, removing "invisible" environmental contributions.

The authors modestly state that the results in the weak-coupling regime are "coarse but physically meaningful estimates" within a surrogate model. They acknowledge that the exact intrinsic complexity is difficult to compute and that sharper stability results would require more precise approximation theories (e.g., in the diamond norm). The work serves as a bridge between weak-coupling Markovian models and geometric cost estimates, offering a resource-sensitive perspective on the complexity of open quantum systems.