From digital twins to singularities: Unlocking acoustic holograms with automatic differentiation

The ability to precisely shape acoustic fields is important for diverse applications such as particle manipulation, material assembly, and ultrasonic haptics. Automatic differentiation (AD) offers a flexible and efficient framework for optimizing acoustic holograms, enabling precise control over acoustic fields. By integrating AD with experimental workflows via digital twin methodologies, it provides a powerful tool to bridge simulation and real-world implementation, allowing for iterative refinement with a focus on efficiency and accuracy. This talk begins by explaining the principles of AD-driven optimization, illustrating its capability through examples such as phase singularity control and hologram optimization under complex constraints. These examples demonstrate how AD can efficiently produce solutions tailored to specific requirements, serving as an effective tool for practical implementation. Beyond optimization, this talk highlights the importance of understanding what acoustic holograms represent and the broader principles they embody. AD provides a pathway to explore and manipulate specific solutions, but true insight requires analyzing the fundamental properties and constraints of holographic fields to better understand the mechanisms and limits that define acoustic holograms.