Mean-field Pulse Adaptation for the Circularization of Interacting Rydberg Atoms

AI summary

Overview

• Research area: Quantum control and atomic physics — specifically the preparation ("circularization") of circular Rydberg states in arrays of interacting neutral atoms, using optimal-control pulse shaping.
• Technical level: Intermediate (assumes familiarity with Rydberg atoms, dipole–dipole interactions, and time-dependent Schrödinger evolution; the mean-field construction itself is explained step by step).
• Scope: The paper introduces a semi-classical mean-field model that makes the circularization of N interacting Rydberg atoms simulable in a state space of dimension Nd instead of d^N, and uses that model to adapt a single-atom-optimized RF pulse to interacting systems on the fly.

What This Paper Is About

Fast, shaped radio-frequency pulses can circularize a single ⁸⁷Rb atom, but in a real array the atoms interact via dipole–dipole coupling, which detunes and distorts the drive and ruins the transfer. Simulating and then re-optimizing the pulse for many interacting atoms is blocked by the exponential growth of the many-body Hilbert space (d^N). The authors build an effective mean-field model — each atom evolves in a classical field produced by the mean dipole moments of the others — and use it to compute amplitude and phase corrections that restore the pulse's performance.

Key Contributions

1. An effective semi-classical (mean-field / Hartree) model for circularization of arbitrary N-atom arrays, in which each atom keeps its full d-dimensional single-atom state space while interactions enter as classical fields I_i*(t)*. The total cost scales as Nd rather than d^N.
2. An on-the-fly pulse adaptation scheme that converts a pulse optimized for non-interacting atoms, F_sa*(t), into an interaction-aware pulse F_ad(t)* by adding amplitude corrections (from I^x, I^y) and a time-dependent phase modulation / chirp (from I^z) — obtained from a single time evolution, not an iterative optimization.
3. A quantitative benchmark of the mean-field error against exact two-body simulation for ⁸⁷Rb pairs in n = 52, including a map of where the model is valid in the (interatomic distance R, angle θ) plane.
4. Analysis of the method's fundamental limits: the phase modulation compensating I^z(t) breaks down when the interaction-induced anharmonicity becomes large compared to the drive strength, and non-symmetric multi-atom arrangements cannot all be brought into resonance by a single global RF pulse.

Main Findings

• Effective model accuracy below 1%: For the pair at R = 7 µm and θ = 0.1π, the effective model's final-state infidelity against the exact two-body simulation is 0.8%. For moderate interactions with R ≥ 7 µm the infidelity stays below 1.0% for all angles.
• Error scales with interaction strength: The infidelity follows the scaling (1 − 3cos²θ)/R³ set by the ⟨ẑ⟩ dipole term, growing to 46.7% for the arrangement R = 4 µm and θ = 0.
• Unadapted pulses fail: With the single-atom pulse F_sa*,* the pair at R = 7 µm, θ = 0.1π drops from p_C = 99.0% (non-interacting) to p̄_C = 65.9% in the effective model and 65.6% in the exact model. In exact simulation, p̄_C ≤ 1.0% for R ≤ 5 µm.
• Adapted pulses recover the performance: For the same pair, F_ad* gives p̄_C = 99.0% in the effective model and 98.3% in exact simulation. Adapted pulses reach p̄_C ≥ 95.0% for R ≥ 5.6 µm at any θ, and the effective model recovers a full 99.0% for R ≥ 7.3 µm, where the deviation from exact evolution is below 0.7%.
• Strong-interaction ceiling: Full recovery is impossible for small R; p̄_C = 96.2% at R = 4 µm and θ = 0. The cause is the anharmonicity from the I^z(t) detuning: the chirp outruns the Rabi cycling and coherent transfer breaks down.
• "Magic" angle removes the problem: At θ = arccos(1/√3) the leading interaction term cancels, no phase modulation is needed, and adapted pulses reach p̄_C = 99.0% in the effective model even at R = 4 µm, limited only by amplitude constraints.
• Detuning dominates over amplitude errors: Pulses adapted only to I^z(t) (no amplitude corrections) stay within 0.5% of fully adapted pulses at R = 7 µm, showing that interaction-induced frequency shifts are the main disturbance. The gap maximises near θ = arccos(1/√3), where that detuning vanishes.
• Amplitude budget: The original pulse keeps amplitudes below 92 mV/cm and frequencies below 480 MHz. Interaction compensation at R ≥ 4 µm shifts frequencies by at most 10 MHz, but amplitude corrections reach up to 72 mV/cm at R = 4 µm, θ = 0.25π — dropping to at most 57 mV/cm in both components if the interatomic axis is aligned along (e_x + e_y)/√2.
• Figures of merit for arrays: For pairs separated by R ≥ 7 µm the adapted pulses give mean circular pair-state probabilities p̄_C ≥ 98% in exact simulation, and effective-model states overlap the exact ones by at least 99%.

Methodology in Plain English

The authors keep the single-atom problem exactly as it is — the same Hamiltonian with static electric and magnetic fields (𝓔 = 2.1 V/cm, B = 14 G), the same σ₊-polarized RF pulse, the same n = 52, d-dimensional state space for ⁸⁷Rb — but replace the exact dipole–dipole coupling between atoms with an average. Each atom feels a classical field I_i*(t)* built from the expectation values of the other atoms' dipole operators, weighted by 1/(4πε₀R_ij³) and the usual dipole–dipole angular factors. The atoms are then propagated in parallel in discrete time steps, recomputing the fields from the current states at each step.

Because the interaction enters the Hamiltonian exactly the way the RF pulse does (both are fields multiplying the dipole operator), the authors can simply subtract the interaction field from the drive: the x and y components become amplitude corrections and the z component becomes a phase modulation (a chirp) accumulated step by step. The result is a corrected pulse computed during the same single time evolution used to simulate the system — no separate optimization loop. The authors validate this against full two-body exact simulations for atom pairs at many distances and angles, and against the earlier parameter-optimization approach for pairs.

Why This Matters

Circular Rydberg atoms combine long lifetimes with strong interactions, which makes them attractive for quantum technology — but only if they can be prepared quickly while interacting. This work turns an intractable N-atom control problem into something computable on modest hardware, and shows that pulse corrections can be generated essentially for free during simulation.

• Quantum simulation: interacting circular-state arrays are a proposed platform for simulating lattice models and long-range interacting spin systems; reliable, fast preparation is a prerequisite.
• Neutral-atom quantum computing: high-fidelity array initialization is required for scaling, and the adapted-pulse recipe is a lightweight calibration step.
• Sensing: circular Rydberg atoms are used for electric and microwave field sensing, where strong dipole coupling to external fields is the point.
• Dynamical architectures: the authors note the model could handle moving atoms, enabling "flying qubits" for fast array formation and defect correction.

Industry relevance: the results speak directly to neutral-atom quantum-computing hardware, to RF/arbitrary-waveform control electronics (the pulses require bandwidth beyond the 480 MHz used by the seed pulse), and to quantum-sensing instruments. The fact that the correction is a closed-form feed-forward on top of an existing pulse makes it cheap to deploy in a lab control stack.

Future Directions

• Restoring symmetry in multi-atom arrays: a global RF pulse cannot resonate with atoms whose interaction fields I_i*(t)* differ, so the authors propose arranging atoms (e.g., a square arrangement of four atoms with full permutation symmetry) or optimizing the orientation of the static fields 𝓔 and B to maximize symmetry. They state they are currently working on these directions.
• Crossing into the strong-interaction regime: using the adapted pulse as an initial guess for pulse-shaping algorithms with access to exact simulation or closed-loop experimental feedback.
• Adapting to mean versus local fields: choosing between maximizing the average p̄_C (adapting to Ī(t)) and maximizing a single atom's or a symmetric sub-lattice's probability — the trade-off is quantified only qualitatively here.
• Going beyond static arrangements: extending the model to moving atoms and dynamically reconfigurable architectures.

Target Audience

Quantum-control researchers and Rydberg-atom experimentalists preparing circular states in arrays; theorists interested in mean-field approximations for driven interacting many-body systems; and quantum-computing/simulation hardware teams who need practical, low-cost pulse calibration schemes. It presupposes comfort with Hamiltonians, dipole–dipole interactions, and pulse shaping, so it is best suited to graduate students and researchers already working in AMO physics or quantum optimal control rather than newcomers.

Authors’ abstract

Arrays of circular Rydberg atoms provide a promising platform for quantum simulation and computation; however, their preparation in the presence of interatomic interactions remains a major challenge. While optimal control methods have enabled the design of fast and accurate radio-frequency pulses for the circularization of a single atom and of an atom pair, the extension to more atoms is fundamentally limited by the exponential growth of the Hilbert space, which renders numerical simulations computationally infeasible. Here, we introduce an effective model that treats interactions within a mean-field approximation, thereby enabling the simulation of large atomic systems. Our model further enables the adaptation of pulses optimized for non-interacting atoms to interacting systems, based on the computation of a single time evolution. For two interacting $^{87}\mathrm{Rb}$ atoms, we demonstrate that the error of our method remains below $1 \, \%$ and that our adapted pulses recover the initial performance of optimal pulses in the regime of weak to moderate interaction strengths.