Optical tweezer generation using automated alignment and adaptive optics
Recent progress in quantum technologies with ultracold atoms has been propelled by spatially fine-tuned control of lasers and diffraction-limited imaging. The state-of-the-art precision of optical alignment to achieve this fine-tuning is reaching the limits of manual control. Here, we show how to automate this process. One of the elementary techniques of manual alignment of optics is cross-walking of laser beams. Here, we generalize this technique to multi-variable cross-walking. Mathematically, this is a variant of the well-known Alternating Minimization (AM) algorithm in convex optimization and is closely related to the Gauss-Seidel algorithm. Therefore, we refer to our multi-variable cross-walking algorithm as the modified AM algorithm. While cross-walking more than two variables manually is challenging, one can do this easily for machine-controlled variables. We apply this algorithm to mechanically align high numerical aperture (NA) objectives and show that we can produce high-quality diffraction-limited tweezers and point spread functions (PSF). After a rudimentary coarse alignment, the algorithm takes about 1 hour to align the optics to produce high-quality tweezers. Moreover, we use the same algorithm to optimize the shape of a deformable mirror along with the mechanical variables and show that it can be used to correct for optical aberrations produced, for example, by glass thickness when producing tweezers and imaging point sources. The shape of the deformable mirror is parametrized using the first 14 non-trivial Zernike polynomials, and the corresponding coefficients are optimized together with the mechanical alignment variables. We show PSF with a Strehl ratio close to 1 and tweezers with a Strehl ratio >0.8. The algorithm demonstrates exceptional robustness, effectively operating in the presence of significant mechanical fluctuations induced by a noisy environment.