Controlling colloidal crystals via morphing energy landscapes and reinforcement learning

Area of Science:

• Soft Matter Physics
• Materials Science
• Artificial Intelligence

Background:

• Colloidal crystals are crucial in materials science, but their formation often involves defects like grain boundaries.
• Achieving defect-free colloidal crystals typically relies on slow natural relaxation processes.

Purpose of the Study:

• To develop a feedback control method for fabricating defect-free, circular colloidal crystals.
• To accelerate the formation and repair of colloidal crystal structures.

Main Methods:

• Utilizing morphing energy landscapes with tunable shapes and orientations.
• Employing reinforcement learning to optimize control policies for actuating these landscapes.
• Demonstrating the method through optical microscopy and computer simulations of colloidal particles in AC electric fields.

Main Results:

• Discovered that specific energy landscape configurations enhance grain boundary motion and crystal relaxation.
• Developed an optimized reinforcement learning policy to actuate morphing landscapes, producing defect-free crystals significantly faster than natural relaxation.
• Showcased rapid crystal repair via anisotropic stresses without melting.

Conclusions:

• The feedback control method effectively removes grain boundaries and produces circular colloidal crystals.
• The approach is scalable to at least 10^3 particles, with process times scaling as N^0.5.
• Further scalability for large-scale hierarchical structures is achievable through parallel local energy landscape control.