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

Re: Parallel Assembly of Neutral Atom Arrays with an SLM Using Linear Phase Interpolation
by Ivo H. A. Knottnerus, Yu Chih Tseng, Alexander Urech, Robert J. C. Spreeuw, and Florian Schreck

This study introduces a new method for the rapid rearrangement of neutral atom arrays using a spatial light modulator (SLM), leveraging direct Fourier transform calculations that account for both the intensity and optical phase of each tweezer. Compared to conventional approaches based on the Gerchberg–Saxton (GS) algorithm, this technique achieves significantly improved computational speed and rearrangement success rates. In particular, the method shares conceptual similarities with the concurrently developed work “AI-Enabled Rapid Assembly of Thousands of Defect-Free Neutral Atom Arrays with Constant-Time-Overhead” by the USTC group (Ref. [42], cited on p.11), and demonstrates experimental performance on arrays of up to 36 traps, with simulation benchmarks showing constant computational times even for systems exceeding 2400 tweezers.
The proposed method effectively overcomes key limitations of GS-based SLM rearrangement, particularly by reducing computational time and minimizing atom loss. This advancement facilitates efficient large-scale atom array rearrangement without incurring additional time overhead. Leveraging the inherent parallelism of SLMs, the approach presents a promising alternative to AOD-based techniques, particularly in high-density configurations. The authors further investigate the role of optical phase by applying linear interpolation between initial and final configurations, and assess success probabilities as a function of frame intervals—thereby identifying key factors contributing to intensity flicker and atom loss.

In that regards, we conclude that this manuscript is worthy of being published with some improvements suggested below:

Comments and Suggestions:

(1) A quantitative benchmark would strengthen the manuscript—for instance, identifying the array size or regime at which the proposed method surpasses AOD-based rearrangement in transport speed. A direct performance comparison with the GS algorithm would also be valuable in demonstrating the method’s practical benefits.

(2) In Figure 3, it would be helpful to include visual markers for trap and atom positions, similar to those in Figure 2, to enhance clarity and consistency.

(3) The argument in Chapter 3 and Figure 4(b) regarding atom loss due to phase differences could be further substantiated by including plots of the inter-frame phase differences. Correlating these with the intensity flicker observed in Figure 4(a) would reinforce the proposed interpretation.

Ask for major revision

1- SLM re-arrangement has important implications for preparing atomic arrays as it is expected to scale with atoms numbers. This work demonstrated a great improvement from previous achievements, with the smart, simple and clearly explained idea of linear interpolation of the tweezers phase.

2- It was a pleasure to read the paper and it should become a must-read for anyone working with holographic arrays of tweezers.

The technique of SLM re-arrangement is greatly improved from previous works, that required heavy computation.
The core idea of “linear phase interpolation” is very well explained (one of my master-level student could reproduce the work based on the paper explanation), and its importance is effectively demonstrated in Figure 4. The more subtle point of aligning the phase mask “center” and optical axis is also nicely highlighted in Figure 5.
The potential scalability to 1000s tweezers is well supported in Figure 6.
It would have been interesting to have a more detailed study of how to choose the minimal step size when moving a tweezers from one spot to the next. This could be left for future work though, as the main point of this paper is rather to point the importance of controlling the tweezers phase during displacement.
Overall, the work is a significant technical improvement, and I enjoyed much reading the paper as the underlying ideas are clearly conveyed to the reader.
The paper should definitely be published in SciPost Physics following the minor modification requested.

(1) I would like to read the following numbers in the paper: the size of the tweezers and the minimum step size (the expression lambda*f/mL is given in the paper, but one cannot compute it as L is not given). I feel they are important for the reader to visualize quantitatively how small are the changes between 2 holograms. Currently, only a qualitative comment is given (“overlap between consecutive tweezers”).

(2) Is there laser-cooling during the re-arrangement? If yes, this should be mentioned explicitly. If not, could this help to make larger steps and thus reduce the re-arrangement time?

(3) The re-arrangement time is dominated by the SLM refresh rate, but 30% is from the memory transfer. Could the authors comment briefly if this could be improved by implementing Direct Memory Transfer between the GPU and the SLM (bypassing the CPU)? What is the limit set by the PCIe bus bandwidth?

Publish (easily meets expectations and criteria for this Journal; among top 50%)