Periodic Reporting for period 2 - SoftML (Rational Design of Soft Hierarchical Materials with Responsive Functionalities: Machine learning Soft Matter to create Soft Machines)
Reporting period: 2022-07-01 to 2023-12-31
Nature displays fascinating examples of self-assembled materials that reconfigure and respond to external stimuli, e.g. chameleons change color for camouflage, pine cones release seeds upon a change in humidity. Advances in colloid synthesis have resulted in a diversity of self-assembled nanostructures with interesting functional properties. These nanostructures are however passive! In this project, I will focus on novel dynamic materials that can respond and undergo shape changes by elastically coupling the nanostructures to a soft elastic matrix or hydrogel, e.g. nanoparticles stabilized by (cross-linked) ligands or (core-shell) microgel particles. These hydrogels or polymer networks can be actuated by pH, temperature, humidity, salt, or light, resulting into an influx or efflux of water that swells or shrinks the gel. The aim of this project is to investigate the non-trivial structural and dynamic behavior of soft reconfigurable materials.
To this end, I have developed coarse-grained models at multiple levels to study the structural and dynamic behavior of these soft materials. We have developed a general theoretical framework to develop thermodynamic consistent coarse-grained models for anisotropic colloids (with non-adsorbing polymer), ligand-stabilized and DNA-coated nanoparticles, and microgel-like particles. To this end, we employ a machine-learning approach to describe effective coarse-grained many-body potentials as well as forces based on atomistic/molecular/particle-based simulations.
Designing dynamic materials and machines that respond, reconfigure, and change shape on demand, requires a better fundamental understanding of not only the link between the microscopic atomistic details of a system and the self-assembled structure at the macroscale, but also on the phase transformation kinetics or shape changes upon actuation.
In order to make progress in the investigation and design of shape-morphing materials, I have developed machine-learning tools to project the high-dimensional phase space containing the particle configurations or trajectories onto a low-dimensional order parameter manifold using dimensionality reduction methods. This mapping onto a low-dimensional manifold enables finding the global (meta)stable states and the relevant set of order parameters that discriminates these (meta)stable states, but allows the investigation of the slow and fast collective modes, and the phase transformation kinetics and pathways upon actuation. I have applied this approach to study the nucleation kinetics of a suspension of charged colloids, and a system of colloidal hard spheres.
I have applied the effective machine-learned potentials and the machine-learning tools to characterize simulation configurations to investigate the self-assembly behavior of plate-like particles on a substrate and in spherical confinement in collaboration with experimental groups. In addition, I have developed various theoretical approaches to predict the self-assembled structures for cellulose nanocrystal suspensions, bent silica rods, tetrahedral-sphere particles, DNA-coated nanocubes.
Designing dynamic materials and machines that respond, reconfigure, and change shape on demand, requires a better fundamental understanding of not only the link between the microscopic atomistic details of a system and the self-assembled structure at the macroscale, but also on the phase transformation kinetics or shape changes upon actuation.
In order to make progress in the investigation and design of shape-morphing materials, I have developed machine-learning tools to project the high-dimensional phase space containing the particle configurations or trajectories onto a low-dimensional order parameter manifold using dimensionality reduction methods. This mapping onto a low-dimensional manifold enables finding the global (meta)stable states and the relevant set of order parameters that discriminates these (meta)stable states, but allows the investigation of the slow and fast collective modes, and the phase transformation kinetics and pathways upon actuation. I have applied this approach to study the nucleation kinetics of a suspension of charged colloids, and a system of colloidal hard spheres.
I have applied the effective machine-learned potentials and the machine-learning tools to characterize simulation configurations to investigate the self-assembly behavior of plate-like particles on a substrate and in spherical confinement in collaboration with experimental groups. In addition, I have developed various theoretical approaches to predict the self-assembled structures for cellulose nanocrystal suspensions, bent silica rods, tetrahedral-sphere particles, DNA-coated nanocubes.
To make colloidal machines at the nanoscale, we will design routes to self-assemble colloidal molecules with controlled valency and high selectivity, as well as directional bonds with controlled flexibility. More importantly, the flexibility of these bonds will be restricted and controlled by the DNA and the temperature. These colloidal molecules will serve as promising stepping stones towards colloidal machines and colloidal metamaterials.
The large variety of colloidal building blocks, materials, structural arrangements, and actuation mechanisms that can be used for the design of soft dynamic materials poses a major challenge. In this project, I will reverse the question and resort to the inverse design problem. I will develop inverse design evolutionary algorithms to reverse-engineer novel classes of soft materials by designing colloidal building blocks that self-assemble into nanostructures at the mesoscale with a targeted set of properties, to the macroscopic scale by designing composite materials with the desired responsive properties.
The large variety of colloidal building blocks, materials, structural arrangements, and actuation mechanisms that can be used for the design of soft dynamic materials poses a major challenge. In this project, I will reverse the question and resort to the inverse design problem. I will develop inverse design evolutionary algorithms to reverse-engineer novel classes of soft materials by designing colloidal building blocks that self-assemble into nanostructures at the mesoscale with a targeted set of properties, to the macroscopic scale by designing composite materials with the desired responsive properties.