(520h) Teaching Machines to Design Self-Assembling Materials

Directed self-assembly has furnished a wealth of advanced materials with applications in microelectronics, optoelectronics, drug delivery, and antimicrobials. However, the inherently many-body nature of assembly makes it a challenge to identify the underlying collective mechanisms and assembly pathways, and has hindered the rational design of building blocks to produce aggregates with desired structure and function. Fundamentally, there exists a predictive gulf between individual building block architecture and chemistry, and the thermodynamics and kinetics of collective assembly behavior. By combining tools from statistical mechanics, molecular simulation, and machine learning, we have bridged this divide to systematically identify emergent mechanisms and assembly pathways from molecular simulations or experimental particle tracking data¹. Our approach furnishes low-dimensional "assembly landscapes" that naturally integrate thermodynamics (what stable aggregates can assemble?) and mechanisms (how do they assemble?) as a function of building block properties. These landscapes present a powerful means to inform inverse design of building blocks to assemble materials with desired structure and function.

In an application to simulations of the self-assembly of anisotropic patchy colloids into icosahedra as an abstracted mimic of viral capsid assembly, we resolved two parallel assembly mechanisms leading to the same terminal icosahedral aggregate and optimized the building block interactions to maximize the assembly flux. In a second application to experimental particle tracking of the non-equilibrium self-assembly of Janus colloids in an applied AC electric field, we quantitatively linked experimental conditions – electric field strength, AC frequency, salt concentration – to the stability and accessibility of different aggregate morphologies – pinwheels, clusters, archipelagos – providing empirical precepts to tailor experimental conditions to assemble desired structures.

¹ Long and Ferguson J. Phys. Chem. B 118 15 4228-4244 (2014)