Data-integrated simulation of magnetic gels

Polymer gels form the basis for many functional materials: they can be tailored to react to external stimuli such as pH and temperature change. By incorporating magnetic nanoparticles, gels are formed that deform or change elasticity in a magnetic field. This is of particular interest for biomedical applications including actuation and drug targeting, as biological matter is not adversely affected by the magnetic fields controlling the material.

We aim to bridge between the microscopic gel architecture and macroscopic properties such as viscoelastic moduli or the deformation of a magnetic gel in a magnetic field. The latter are easy to access experimentally, whereas microscopic detail such as the mesh size distribution or the magnetic-particle polymer coupling is not directly observable.

Using a coarse-grained model in which aspects of the network architecture can be varied, a data set connecting microscopic and macroscopic properties will be established. Using uncertainty quantification and sensitivity analysis, connections and relative importance of aspects of the network architecture with respect to macroscopic observables are determined. This will inform fitting of a machine learning based surrogate model, suitable for inverse uncertainty quantification. Using Bayesian inference, the most likely network architecture will be predicted from experimental data.