Preparation and Machine-Learning Methods of Nacre-like Composites from the Self-Assembly of Magnetic Colloids Exposed...

Composite materials designed by nature, such as nacre, can display unique mechanical properties and have therefore been often mimicked by scientists. In this work, we prepared composite materials mimicking the nacre structure in two steps. First, we synthesized a silica gel skeleton with a layered structure using a bottom-up approach by modifying a sol-gel synthesis. Magnetic colloids were added to the sol solution, and a rotating magnetic field was applied during the sol-gel transition. When exposed to a rotating magnetic field, magnetic colloids organize in layers parallel to the plane of rotation of the field and template the growing silica phase, resulting in a layered anisotropic silica network mimicking the nacre's inorganic phase. Heat treatment has been applied to further harden the silica monoliths. The final nacre-inspired composite is created by filling the porous structure with a monomer, leading to a soft elastomer upon polymerization. Compression tests of the platelet-structured composite show that the mechanical properties of the nacre-like composite material far exceed those of nonstructured composite materials with an identical chemical composition. Increased toughness and a nearly 10-fold increase in Young's modulus were achieved. The natural brittleness and low elastic deformation of silica monoliths could be overcome by mimicking the natural architecture of nacre. Pattern recognition obtained with a classification of machine learning algorithms was applied to achieve a better understanding of the physical and chemical parameters that have the highest impact on the mechanical properties of the monoliths. Multivariate statistical analysis was performed to show that the structural control and the heat treatment have a very strong influence on the mechanical properties of the monoliths.