arXiv:2610.21609
Fast, parallel, three-dimensional rearrangement of atom arrays using superpositions of elementary holograms
D. J. Hoare, P. Dunniece, J. Rodewald, K. K. Voges, M. R. Tarbutt
Abstract
Assembling defect-free arrays of neutral atoms requires real-time rearrangement of stochastically loaded optical tweezers. Conventional rearrangement methods shuttle atoms sequentially, whereas high-speed spatial light modulators (SLMs) can move all atoms in parallel, which scales more favorably as the array size grows. For fast parallel rearrangement, the sequence of holograms displayed on the SLM should be computed more rapidly than the SLM refresh rate, and there should be phase continuity between successive holograms to minimize transient intensity fluctuations. We calculate holograms analytically by adding together the complex fields needed to produce each individual trap, and we ramp the phase of each trap linearly as it moves to its target location. Trap positions enter as continuous variables, rather than tied to points on a grid, so the method provides high positioning accuracy. Using a single GPU, we compute each intermediate hologram in 0.284 ms, considerably faster than the SLM refresh time of 0.696 ms, so the rearrangement speed is hardware-limited. In each step, we use the free processing time to iteratively optimize the holograms for equal intensity across the traps, decreasing trap non-uniformity to 1.24% during rearrangement at no cost in total time. Starting from an array of 256 sites loaded stochastically with $^{87}$Rb atoms, we assemble a 100-atom array in under 20 ms with a rearrangement survival probability of 99.31(12)% per atom. This rearrangement increases the temperature of the atoms by 9(3) $μ$K, which is small compared to the trap depth of 480 $μ$K. Our method can also move individual atoms axially, enabling transport in different planes. We demonstrate this capability with a rearrangement protocol where one set of atoms leapfrogs over another.
AI Summary
Neutral-atom array assembly requires moving stochastically loaded atoms while avoiding transient trap-depth fluctuations. This paper combines linear phase interpolation with superpositions of elementary holograms to implement parallel, three-dimensional transport using a high-speed spatial light modulator (SLM). Trap positions enter as continuous phase-gradient and quadratic-phase parameters, avoiding the positional grid and zero-padding requirements of FFT-based hologram generation. Experiments rearrange a stochastically loaded 256-site $^{87}$Rb array into a 100-site target in 18.23 ms, measured from image receipt to final hologram upload. Background-corrected rearrangement survival is $99.31(12)\%$ per atom; measured final filling is $96.33(9)\%$. Release-and-recapture thermometry gives heating of $9(3)\,\mu\mathrm K$ relative to a $480\,\mu\mathrm K$ trap depth. Axial round-trip transport achieves $98.7(2)\%$ survival, while a three-dimensional leapfrog sequence yields maximum survivals of $99.4(3)\%$ for stationary atoms and $95.1(4)\%$ for moving atoms. Single-GPU hologram computation takes 0.284 ms, below the 0.696 ms SLM refresh period. Numerical characterization finds approximately 0.045 rad phase error and 1.24% intensity nonuniformity with one equalization iteration per frame. Computation remains refresh-limited up to approximately 2,100 traps without equalization, or 435 with one iteration. The method costs $\mathcal O(MN)$ for $M$ pixels and $N$ traps, eventually losing its computational advantage over FFT methods. Transport duration also depends on trajectory length, scaling approximately as $\sqrt N$ for the specified geometry. Qubit-coherence preservation is not measured; collision avoidance and optimization of multiplane transport remain open directions.
Error found: The abstract does not identify the quoted 1.24% trap-intensity non-uniformity as a simulation result obtained with one equalization iteration, whereas the reported 100-atom assembly experiment uses zero equalization iterations. The numerical phase-control section quotes a spatial-overlap intensity factor of 0.879 for a 0.25 μm step, but its stated 1.00 μm waist and Eq. (dip2) give 0.9692.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 14 pages, 6 figures