Parallel assembly of neutral atom arrays with an SLM using linear phase interpolation

TL;DR

This work addresses the challenge of scaling up defect-free neutral-atom arrays for quantum simulation and computation. It introduces Linear Phase Interpolation (LPI), a GPU-accelerated method that updates holograms on a high-speed spatial light modulator to control both tweezer positions and phases, enabling flicker-free, parallel transport. The authors demonstrate high per-atom rearrangement success ($ oughly 0.997$) and rapid cycle times ($2.736\,\text{ms}$) in a $6\times6$ array, with robust performance scalable to thousands of tweezers. This approach offers a path to fast, large-scale assembly of neutral-atom arrays with flexible geometries and potential applications in quantum simulation, computation, and coherent qubit transport.

Abstract

We present fast parallel rearrangement of single atoms in optical tweezers into arbitrary geometries by updating holograms displayed by an ultra fast spatial light modulator. Using linear interpolation of the tweezer position and the optical phase between the start and end arrays, we can calculate and display holograms every few ms, limited by technology. To show the versatility of our method, we sort the same atomic sample into multiple geometries with success probabilities of 0.996(2) per rearrangement cycle. This makes the method a useful tool for rearranging large atom arrays for quantum computation and quantum simulation.

Figures (6)

• Figure 1: Schematic overview of our experimental sequence for fully parallel rearrangement into arbitrary geometries. At the start of the sequence, single atoms are stochastically loaded into an initial tweezer geometry with approximately $45\,\%$ probability per tweezer. Empty tweezers and tweezers with excess atoms are extinguished. Remaining atoms are mapped to target positions and trajectories are calculated that move the other atoms into the desired geometry. Using linear interpolation of position and optical phase of the tweezers, the trajectories are divided into multiple steps. For each step a hologram is calculated in real time and displayed on a high-speed SLM, moving all atoms at the same time.
• Figure 2: (a): Fluorescence images of single atoms stochastically loaded into a square $6\times 6$ array (left panel). The green (red) circles denote the presence (absence) of an atom in a tweezer. The arrows show trajectories that sort 16 atoms into a defect-free $4\times 4$ array. The right panel shows the verification image after rearrangement. (b): The average filling fraction per tweezer in the verification image for 1000 experimental realizations. The error bars are the standard deviation of the mean. The average filling fraction over all tweezers is 0.988(4). Corrected for the image survival, this corresponds to a rearrangement success of $0.997^{+0.003}_{-0.006}$ per atom.
• Figure 3: An example experimental realization of the same atoms being sequentially rearranged to various geometries. From left to right: image of the initially loaded $6\times 6$ array, a circular geometry, a kagome lattice, a triangular lattice and a $4\times 4$ array.
• Figure 4: (a): The relative intensity recorded with an 11-kHz frames-per-second camera of a moving spot made by holograms generated with the WGS algorithm (blue trace) and made with LPI for the special case of identical initial and final tweezer phase (orange trace). The WGS-generated spot has multiple dips with almost no intensity left, while the LPI-generated spot stays above $70\,\%$ throughout the moves. (b): The calculated optical phase at the location of the moving spots for each hologram displayed during the movement.
• Figure 5: (a): The average survival rate of atoms in a $6\times 6$ array after having moved 10 steps of one Fourier unit in total with a tweezer phase slip of $\Delta\psi$ per step. The blue trace is the result while taking as computational center (see Appendix \ref{['appendixA']}) the center pixel of the SLM, which leads to a maximum loss at $\Delta\psi \approx1.05\,\pi$. To measure the effect of alignment, a second scan was performed (orange trace) with a shift in computational center of 250 pixels. This results in an additional tweezer phase slip per step of $\xi\approx0.5\,\pi$. Dashed lines are a guide to the eye. (b) and (c): Translating a spot using the SLM is equivalent to changing the slope of a phase gradient hologram (solid line) displayed on the SLM (dashed line). The left and right panels correspond to consecutive holograms moving the tweezer by one Fourier unit, which increases the hologram phase value at the edge of the SLM by $\pi$. When the optical axis (dash dotted line at center of laser beams) crosses the SLM at the position at which the phase gradient crosses zero (black dot), no phase shift is acquired by a tweezer when moving it by one Fourier unit, as shown in (b). When these positions do not match, each translation introduces an additional tweezer phase shift $\xi$, as shown in (c). The reflection angles of the outgoing beams are not to scale.