Automating Bayesian inference and design to quantify acoustic particle levitation.

Area of Science:

• Acoustic manipulation
• Colloidal physics
• Bayesian inference

Background:

• Ultrasound propulsion offers remote manipulation of micro- and nanoparticles.
• Accurate acoustic field characterization is crucial for understanding particle motion.
• Experimental setup significantly influences acoustic fields.

Purpose of the Study:

• To demonstrate automated experiments using Bayesian inference and design for acoustic field characterization.
• To accurately and efficiently determine parameters of acoustic fields in resonant chambers.
• To enable precise control over acoustic nanomotors.

Main Methods:

• Utilized automated experiments with Bayesian inference and design (OID cycles).
• Employed video microscopy to observe tracer particle relaxation to the nodal plane.
• Applied sequential Monte Carlo methods for parameter inference, accounting for noise and heterogeneity.
• Used simulated outcomes to guide experimental design for maximal information gain.

Main Results:

• Achieved accurate and efficient characterization of acoustic fields.
• Demonstrated convergence to precise parameter estimates in few automated experiments.
• Showcased the ability to discriminate between competing physical hypotheses.
• Validated the use of Bayesian methods for nonlinear, hierarchical model learning.

Conclusions:

• Automated Bayesian methods provide an efficient approach for characterizing acoustic fields.
• This technique enables precise control and understanding of ultrasound-driven microparticle manipulation.
• The OID framework is effective for complex systems with measurement noise and population heterogeneity.