Fast and high-fidelity entangled-state preparation in neutral-atom Rydberg systems is of central importance for quantum information processing, yet conventional adiabatic protocols generally rely on long evolution times and are therefore susceptible to decoherence and control imperfections. In this work, we propose a shortcut-to-adiabaticity protocol based on Lewis—Riesenfeld invariant inverse engineering for the rapid preparation of a symmetric singly excited Bell-type state in a two-atom ^171Yb Rydberg-blockade system. The target state is the ground—Rydberg entangled state |B\rangle=(|g r\rangle+|r g\rangle) / \sqrt2, which is generated from the initial state |gg\rangle.
Starting from the original two-atom three-level ladder system, we first perform adiabatic elimination of the intermediate state under the large single-photon detuning condition and further reduce the dynamics to an effective two-level model in the symmetric subspace spanned by |G\rangle=|gg\rangle and |B\rangle. In the strong Rydberg blockade regime, the doubly excited state |rr\rangle is far off resonance and contributes only as a higher-order leakage channel. On this basis, an analytical control pulse is derived by reverse engineering the Lewis—Riesenfeld invariant. By choosing a constant auxiliary phase \beta=-\pi/2, the general invariant-based control relations are simplified to a resonant shortcut protocol with vanishing effective detuning. A quintic polynomial ansatz for the auxiliary angle \theta(t) is then adopted to satisfy the boundary conditions \theta(0)=0 and \theta(T)=\pi, yielding a smooth single-peaked pulse with zero initial and final amplitudes, which is favorable for experimental implementation.
Numerical simulations show that, for a total operation time of T=1~\mu\mathrms and a Rydberg interaction strength of V=2\pi\times120~\mathrmMHz, the effective two-level model yields a final fidelity above 0.999 for the target Bell state. The required peak effective two-photon Rabi frequency is only about 2\pi\times0.66~\mathrmMHz, indicating that the protocol can be implemented with relatively weak driving fields. To further assess the validity of the reduced model, we verify the protocol in the full two-atom three-level system. In the absence of dissipation, the final-state fidelity exceeds 0.9999, confirming the accuracy of the effective-model description. When the spontaneous decay of the intermediate state and the finite lifetime of the Rydberg state are included through a Lindblad master equation, the fidelity remains above 97%, demonstrating that the protocol retains good performance in the open-system setting. During the entire evolution, the intermediate-state population stays at the 10^-3 level and the doubly excited Rydberg-state population remains at the 10^-5 level, which validates both the adiabatic elimination and the strong-blockade approximation.
For a fair comparison with a conventional adiabatic protocol under the same peak-driving constraint, we find that the traditional adiabatic scheme requires a much longer evolution time of about 8.79~\mu\mathrms to reach a fidelity of 0.99 in the ideal closed-system case, whereas the present shortcut protocol completes the state preparation within 1~\mu\mathrms. Owing to the shorter operation time, the proposed scheme significantly suppresses dissipation-induced loss and outperforms the adiabatic protocol in open-system dynamics. In addition, robustness analyses show that the protocol is tolerant to experimentally relevant imperfections, including driving-amplitude errors, detuning offsets, interaction-strength fluctuations, Doppler shifts, laser-linewidth broadening, atom—laser coupling asymmetry, and weak time-dependent modulation of the Rydberg interaction.
These results demonstrate that the proposed Lewis—Riesenfeld invariant-based shortcut protocol provides a fast, robust, and experimentally feasible route for Bell-state preparation in ^171Yb Rydberg neutral atoms. The generated ground—Rydberg entangled state can further serve as a useful resource for subsequent state mapping into long-lived logical qubit subspaces and for scalable quantum-information processing.