Analytical controls for dispersive multi-qubit interactions in mediator-coupled quantum registers

Technical Summary: Analytical Controls for Dispersive Multi-Qubit Interactions in Mediator-Coupled Quantum Registers

Problem Statement
Long-lived quantum registers are essential for quantum information storage but often suffer from slow or technically demanding direct multi-qubit control due to their isolation. Hybrid architectures address this by coupling the register to an auxiliary "mediator" (e.g., an electron spin, bosonic mode, or ancillary particle) that provides addressability. However, real excitation of the mediator introduces error channels such as decoherence, leakage, and loss, particularly when the mediator is not the coherence-optimized component. While standard numerical quantum optimal control can search for pulses to implement target gates, it typically treats synthesis as an iterative, black-box forward search over a high-dimensional landscape, obscuring the relationship between spectral responses and the resulting interactions. The challenge is to engineer selective multi-qubit interactions while suppressing real mediator population, leveraging the dispersive regime where off-resonant driving produces state-dependent energy shifts without mixing register states.

Methodology
The authors introduce Dispersive Interaction via Analytical Linear Inversion (DIAL), a control construction that exploits the configuration-preserving dispersive structure of mediator-coupled registers.

1. Theoretical Framework:

   • The system is modeled as a quantum register coupled to an effective two-level mediator transition. In the configuration-preserving regime, the register eigenbasis remains unchanged during mediator excitation, leading to a diagonal effective Hamiltonian.
   • Under off-resonant multi-tone driving, the mediator induces configuration-dependent AC Stark shifts. These shifts are decomposed into a basis of Pauli-Z strings ($Z_S = \prod_{i \in S} Z_i$).
   • Linear Control Map: For constant-amplitude multi-tone drives, the interaction rates ($K_S$) of these Pauli-Z channels depend linearly on the drive intensities ($x_k = |\Omega_k|^2/2$). This establishes a linear transfer matrix $G$ such that $\mathbf{K} = G\mathbf{x}$, where $\mathbf{K}$ represents the vector of interaction rates and $\mathbf{x}$ the vector of tone intensities. The matrix $G$ is determined solely by the register-resolved transition spectrum and the chosen tone frequencies.

2. Target-Aware DIAL Algorithm:

   • To synthesize a specific target interaction (e.g., a high-order Pauli string), the algorithm selects a finite set of off-resonant tones from a discretized frequency pool.
   • It employs a greedy iterative procedure: starting with an empty set, it tests candidate frequencies, solves a non-negative linear inversion problem to find optimal intensities for the target channel, and retains the candidate that best minimizes the residual error (spectator suppression).
   • The algorithm respects physical constraints, including non-negative intensities, minimum detuning from register transitions, and a maximum drive-to-detuning ratio ($r_{\text{disp}}$) to ensure the validity of the perturbative dispersive approximation.
   • Once the relative intensities are determined, a global scaling factor is applied to set the drive strength within the admissible dispersive regime, and the interaction time is calculated to accumulate the desired phase.

Key Contributions

• Analytical Linearization: The paper derives an explicit linear map from experimentally accessible control parameters (tone intensities) to the coefficients of the effective multi-qubit Hamiltonian. This transforms the control synthesis problem from a non-linear optimization into a constrained linear inversion.
• Target-Aware Synthesis: The development of the DIAL algorithm provides a practical, non-iterative method to select tone frequencies and determine intensities for a prescribed interaction, avoiding the need for full-dynamics numerical optimization.
• Trade-off Characterization: The work analytically establishes a direct trade-off between dispersivity (drive strength relative to detuning) and gate time. It shows that reducing mediator population (increasing dispersivity) necessitates longer interaction times, quantified by $T_{\text{gate}} \propto r_{\text{disp}}^{-2}$.
• Open-Source Implementation: The manuscript is accompanied by a Python package implementing the DIAL algorithm and the associated numerical validation workflow.

Results
The authors benchmarked the DIAL controls across 100 random register realizations for register sizes $n = 2, 3, 4, 5$. The target was the highest-order interaction $Z_1 \cdots Z_n$ with a phase of $\pi/4$, which generates an $n$-qubit GHZ state (up to local equivalence).

• Fidelity: Propagation under the full driven mediator-register dynamics (without further optimization) yielded typical gate fidelities above 0.997 for all register sizes (medians: 0.9976 for $n=2$ to 0.9984 for $n=5$).
• Mediator Excitation: The maximum transient mediator excitation remained near 1% on average (medians ranging from 0.99% to 1.36%), significantly lower than what is typically achievable with resonant Rabi drives.
• Gate Time: There is a systematic increase in the normalized gate duration ($T_{\text{gate}}/T_{\text{Rabi}}^{\text{max}}$) as register size increases, rising from a median of ~10.8 for $n=2$ to ~36.0 for $n=5$. This increase is attributed to spectral crowding, which makes selective interaction engineering more difficult and requires longer times to accumulate the target phase while maintaining low mediator population.
• Robustness: The controls maintained high accuracy even when the linear dispersive construction was propagated under the full non-perturbative dynamics, validating the analytical approximation.

Significance and Claims
The paper claims that the DIAL framework provides a direct, analytical route from spectral information to engineered multi-qubit interactions. By establishing a linear control map, it enables the design of excitation-suppressed interactions without iterative propagation of the full driven dynamics. The authors emphasize that this approach is particularly valuable for platforms where the mediator is prone to decoherence, as it systematically suppresses mediator population while achieving high-fidelity register operations. The work highlights that the performance of such controls is fundamentally limited by the spectral geometry of the system; specifically, spectral crowding necessitates longer gate times to maintain the dispersive regime. The authors position this as a complementary approach to numerical optimal control, offering transparency and analytical tractability for mediator-coupled architectures.