Understanding Processing-Structure-Mechanical Property Relationships in Sustainable Biopolymer-based Composites Using Design of Experiments and Machine Learning

Abstract

Nature uses hierarchy as a means to imbue materials with mechanical and physical properties that are greater than the sum of its parts. Molecules are combined into polymers which bundle together into fibers that make up ever-larger structural building blocks that comprise the organism. Wood exemplifies this concept, with cellulose as the main polymer backbone supported by biopolymers such as hemicellulose, pectin, and lignin. These components act as binders to toughen the tree against environmental forces. Moreover, the resulting composite exhibits functional features, such as the ability to transport nutrients throughout the structure.One of the most common approaches to making synthetic biocomposites has been through the extraction and use of select components from biomass. However, despite the use of the same materials used by nature, the properties observed in natural composites have proven difficult to replicate. As an alternative approach to making better biocomposites, there has been a research push in recent years toward exploring the use of the entire organism as a component. The recently coined term “biomatter” refers to such materials wherein the entire biological matter is used without removing any components, retaining native or minimally altered hierarchical nano- and microstructures. Biomatter materials offer a promising solution to address the need for sustainable composites; none of the components within the material go to waste or lead to waste from inefficient extraction processes, and the mechanical properties have already been optimized against stressors by the hierarchical design inherent to natural materials. In addition, by shifting away from petroleum feedstocks and harnessing abundant, renewable biological resources, we can fabricate materials that align with sustainability goals. Another approach to making better biocomposites has been to combine natural materials not normally found together in nature. By leveraging general chemical and physical principles, best-in-class materials can be combined to rationally design composite materials. However, we currently lack the knowledge to construct composites that maximize the utility of materials with intricate and dissimilar components. The work presented in this thesis tackles this issue by employing three different strategies to rationally build composites with biomatter building blocks. We started by selecting model materials of varying complexity, ranging from molecules to polymers to entire cells, as composite components. For the cell building blocks, we opted for microalgae, specifically Spirulina platensis (“spirulina”) and Chlorella sp. (“chlorella”). These algae offer several advantages, including their abundance in industries such as biofuel, pharmaceuticals, and nutraceuticals, ease of culturing and growth, absence of complex tissue-like structures, and a diverse polymer composition of proteins, short and long polysaccharides, phenolics, and small molecules, all of which serve as valuable components for constructing polymer-based composites. We selected bacterial cellulose (BC) fibers grown from bacteria-yeast cocultures as a model structural polymer. BC fibers possess a high molecular weight, a high degree of crystallinity, and exceptional mechanical properties, owing to their inherent hierarchical structure. Notably, BC fibers are the sole product of this culture system, eliminating the need for separation and extraction after harvesting. Lignin, an abundant phenolic polymer found in woody biomass and a byproduct of the pulp and paper-making industries, was chosen to serve as a binder in our composite systems. Lignin's rich aromatic backbone and diverse functional groups allow for secondary interactions with glucans, like cellulose, making it an appropriate choice as a binder. Lastly, we incorporated stearic acid as a small, functional molecule. Stearic acid is a native component of many types of biomass, and its amphiphilic structure enables it to function as a plasticizer and/or surfactant within the composite system. The three strategies chosen to guide the development of multiscale biomatter composites were a traditional trial-and-error approach, design of experiments, and machine learning. Using the trial-and-error approach, we investigated the effects of three different post-processing methods and the influence of plant-extracted micro-crystalline cellulose fibers (CFs) on the structure and mechanical properties of a spirulina-based matrix. We aimed to establish methods by which a single biomatter component could be used to create multiscale objects. The smallest scale consisted of the polymeric composite units inherent in the cell walls of spirulina. The next scale consisted of individual cells of spirulina. The subsequent scale consisted of physically-confined particles of spirulina chains. The final scale was dictated by the macroscale-controllable geometry extruded via a 3D printer. We employed a design of experiments approach to explore how processing conditions influence the ex-vivo incorporation of lignin into a BC matrix in hierarchical lignocellulose papers. We used heat, pressure, and pressing time as parameters in a hot-pressing process to facilitate a systematic understanding of the processing-property relationships. Our analysis revealed optimal processing temperature/pressure/time conditions for tuning mechanical and water-repellency properties via structural changes in the processed papers. Finally, we developed machine learning models based on a Bayesian Optimization approach to guide the search for optimal compositions of ternary composites for improved mechanical properties. Chlorella, BC, stearic acid, and a water-ethanol solution were combined, processed, and cast into films to facilitate understanding of the relationship between composition and the resulting mechanical properties of the composites. We conclude with a summary of the benefits and disadvantages of using traditional experimental approaches, statistical designs of experiments, and machine learning models to design and fabricate next-generation, sustainable biocomposites.