Temporal quantum interference in many-body programmable atom arrays

TL;DR

The paper addresses how temporal interference from Stückelberg dynamics manifests in large, interacting quantum many-body systems realized with programmable Rydberg arrays. It combines experiments using single- and bi-frequency driving to induce interference-based vacuum-state freezing with a Floquet perturbation framework that incorporates finite-range interactions beyond idealized blockade. Key findings include interference visibility >70% and near 1% residual excitation, with geometry, interatomic distance, and drive parameters shaping the interference; leading-order freezing follows zeros of $J_0(rac{\, riangle_0}{\omega})$, while finite-range tails introduce higher-order channels that shift resonances. The work demonstrates that multi-harmonic Floquet control is essential in 2D geometries and provides design principles for scalable, robust Floquet engineering and state preparation in large-scale quantum simulators.

Abstract

Quantum superposition famously manifests as spatial interference, epitomized by the double-slit experiment. Its less explored temporal analogue, Stückelberg interference, arises in driven systems where phases accumulated along distinct time-domain pathways recombine. Extending this phenomenon to large interacting systems introduces a new complexity as delicate phase relationships are disrupted by many-body interactions. Here we experimentally achieve controllable vacuum-state freezing in programmable Rydberg arrays of up to 100 atoms through many-body Stückelberg interference, with visibility exceeding $70\%$ and excitation suppression to $1\%$ despite periodic driving that would typically induce heating. Comparing single and dual-frequency protocols across multiple geometries, we show that simultaneous modulation of detuning and Rabi frequency dramatically enhances interference-driven freezing. Finite-range interaction tails play a decisive role, producing interference patterns which constrained $PXP$ models cannot capture. Our results establish temporal interference as a scalable microscopic mechanism for Floquet control, enabling predictive many-body state engineering in large-scale platforms.

Figures (4)

• Figure 1: Many-body Stückelberg interference and vacuum freezing in one-dimensional chains. Frequency-dependent Rydberg population $n(T)$ in 100-atom chains ($d=4.7~\mu$m). Top: Snake geometry (positions in $\mu m$ units) and drive protocols ( in rad/$\mu s$ units); shading shows atom loading probability. a,b: Single-frequency driving ($\Omega$ constant). c,d: Bi-frequency driving ($\Omega$ modulated). a,c:$\Delta_0>0$ (vacuum as instantaneous excited state). b,d:$\Delta_0<0$ (vacuum as instantaneous ground state). Blue circles: Experimental data obtained from Aquila quantum processor ($L=100$); purple: experimental data from a smaller system ($L=14$, where shown); green/yellow: constrained models (PXP/PPXPP); red: simulations from Bloqade (Classical simulator which simulates the full Rydberg Hamiltonian). Shaded regions indicate statistical uncertainty. Drive frequency $\omega$ (in rad$/ \mu s$) controls interference fringes: minima correspond to destructive interference and vacuum freezing.
• Figure 2: Tuning many-body interference by controlling interatomic distance and detuning. Frequency-dependent Rydberg population $n(T)$ in 100 atom chains reveals how interaction strength controls temporal interference. a, b: Varying interatomic spacing to $d = 5.0$ and $5.3~\mu$m systematically shifts interference minima and modifies visibility, exposing the role of finite-range interaction tails in phase accumulation. c: Reducing detuning amplitude to $\Delta_0 = 16.5~\text{rad}/\mu\text{s}$ shifts minima to lower frequencies, consistent with Floquet perturbation theory. Blue circles: Experimental data obtained from Aquila quantum processor ($L = 100$); purple: experimental data obtained from Aquila quantum processor on a smaller chain ($L=14$, where shown); red: simulations from Bloqade (Classical simulator which simulates the full Rydberg Hamiltonian, $L=14$ ). Shaded regions indicate statistical uncertainty.
• Figure 3: Temporal interference in two-dimensional geometries. Frequency-dependent Rydberg population $n(T)$ after one drive cycle in $2D$ arrays. Top: Atomic arrangements, drive protocols, and loading probabilities. a,b: Square lattice. c,d: Honeycomb lattice. Panels (a,c): Single-frequency driving produces weakened interference compared to 1D chains, with reduced visibility even in Bloqade simulations ($L=14$, red) and further degradation in Aquila experiments ($L=100$, blue), Green continuous curves denote $PXP$ results. Panels (b,d): Dual-frequency modulation restores interference contrast, achieving visibility $V>0.9$ in square lattices and significant recovery in honeycomb. The superior square-lattice performance reflects simpler next-nearest-neighbor connectivity compared to the more complex honeycomb interaction network. Blue circles: Experimental data obtained from Aquila quantum processor ($L = 100$); red: simulations from Bloqade (Classical simulator which simulates the full Rydberg Hamiltonian, $L=16$); green: numerical simulation of the constrained model (PXP).
• Figure 4: Microscopic dynamics within a drive cycle. Time-resolved measurements reveal how Stückelberg interference produces vacuum-state freezing. Top: Experimental geometry and drive protocol. The purple line indicates the half-cycle control sequence, where the detuning is held fixed beyond $t=T/2$, $\Delta(t>T/2)=\Delta(T/2)$. Left: Bloqade simulations ($L=14$) comparing constructive ($\omega=3.0$, yellow) and destructive ($\omega=3.83$, red) interference. Brown circles denote the half-cycle protocol ($C_{1/2}$), which shows only residual oscillations and no interference signature. Right: Experimental verification on Aquila ($L=100$) at the same frequencies. The clear contrast between full-cycle (red/green and yellow/blue) and half-cycle (brown/purple) dynamics demonstrates that vacuum freezing arises from coherent amplitude recombination via second passage through the avoided crossing, rather than from an individual passage.