Multiscale models based on statistical mechanics and physically-based machine learning for the thermo-hygro-mechanical behavior of spider-silk-like hierarchical materials

Scientists are continuously fascinated by the high degree of sophistication found in natural materials, arising from evolutionary optimisation. In living organisms, nature provides a wide variety of materials, architectures, systems and functions, often based on weak constituents at the lower scales. One of the most extensively studied natural materials is spider silk, renowned for its outstanding mechanical properties, which include exceptional strength and toughness. Owing to its wide range of properties, which vary depending on factors such as the type of silk (up to seven) that each spider can produce, and the species of spider, it can be considered a class of semi-crystalline polymeric material. Indeed, spider silk cleverly combines, depending on the application required, the great deformability of an amorphous phase with the stiffness and strength conferred by pseudo-crystals consisting of specific secondary structures of some of the proteins constituting the material. Based on the countless studies conducted on spider silk, it is now also clear that its remarkable performance are the result of a sophisticated optimisation of the material's hierarchical structure. Nevertheless, many of the multiscale mechanisms that give rise to the striking macroscopic properties are still unclear. Many open problems are also related to the relevant effects of environmental conditions and in particular on temperature and humidity, strongly conditioning the mechanical performances. In this thesis, aimed at unveiling some of these open problems, we introduce a multiscale model for the thermo-hygro-mechanical response, starting with the influence of water molecules modifying the microstructure, up to the effects at the macroscopic scale, including softening, increase in elongation at break and supercontraction, i.e. the shortening (up to half the initial length) of the spider threads in wet environments. Thereafter, we describe how the supercontraction effect can be adopted to obtain humidity-driven actuators, and in particular, we determine the maximum actuation force depending on the silk properties at the molecular scale and on the constraining system representing other silk threads or the actuated device. The spider silk actuation properties turned out to be extraordinary, making spider silk potentially the best performing humidity-driven actuator known to date in terms of work density. As observed in many natural materials, spider silks are characterized by a strong variability in both chemical and structural organization, as for example described in the recently published experimental database of properties at different scales of about a thousand different spider silks, where evident correlations among quantities are scarce. This large variability makes the theoretical understanding of the observed material behavior, in relation of the complex hierarchical structure, particularly intriguing. To address this novel amount of experimental data without losing sight of theoretical analytical modelling, we propose a new data modelling methodology to obtain simple and interpretable relationships linking quantities at different scales. In particular, we employ a symbolic regression technique, known as 'Evolutionary Polynomial Regression', which integrates regression capabilities with the Genetic Programming paradigm, enabling the derivation of explicit analytical formulas, finally delivering a deeper comprehension of the analysed physical phenomenon. Eventually, we provide insights to improve our multiscale theoretical model accounting for the humidity effects on spider silks. This approach may represent a proof of concept for modelling in fields governed by multiscale, hierarchical differential equations. We believe that the analytical description of the macroscopic behaviour from microscale properties is of great value both for the full understanding of biological materials, as well as in the perspective of bioinspired materials and structures.