Optical tweezers have become essential tools for dynamically manipulating objects, ranging from microspheres or biological molecules to neutral atoms. In this study, we demonstrate the creation of tweezer arrays using a generative neural network, which allows for the trapping of neutral atoms with tunable atom arrays. We have successfully loaded cold strontium atoms into various optical tweezer patterns generated by a spatial light modulator (SLM) integrated with generative models. Our approach shortens the process time to control the SLM wtih minimal time delay, eliminating the need for repeated re-optimization of the hologram for the SLM.
Comment: 5 pages, 4 figures