Data-Enabled Theoretical Understanding of the Structure and Properties of Solvent-cast Polymer Nanocomposites

NONTECHNICAL SUMMARY

From sport shoes to credit cards, materials consisting of synthetic polymers are ubiquitous in present-day society. Adding a filler material to the polymer can lead to novel properties that enable specialized applications such as extraordinarily hard coatings, fire-resistant fabrics, or modern car tires. Filler materials consist of small nano-sized particles that are intended to distribute homogeneously within the polymer. However, like water and oil, the nanoparticles and polymers are known to have an intrinsic tendency to separate from each other and form undesired nanoparticle agglomerations instead of a homogeneous nanocomposite. Interestingly, there is experimental evidence that the process of creating the nanocomposite can nevertheless lead to the formation of homogeneous dispersions, with desired properties. The goal of the proposed work is to quantitatively understand the science behind how processing determines the dispersion states and the ensuing mechanical properties of polymer-nanoparticle mixtures. To this end, the project will combine data-rich methods, theoretical calculations, and computer simulations. In a first step, a machine learning algorithm will be trained on the available, large database of polymer-nanoparticle composites, their preparation methods, and the resulting dispersion states. These tools, when implemented, will be able to identify critical regions of parameter space where interesting phenomena occur, e.g., where the material goes from well-mixed to agglomerated nanoparticles. The work will then focus on these regions and use a combination of computer simulations and theoretical calculations to delineate the physical mechanisms that control the nanoparticle dispersion states and the ensuing mechanical properties of the nanocomposite material. This integrated data science-theory-simulation-experimental data workflow will enable the determination of what the optimal nanoparticle dispersion states are for a set of desired mechanical properties and how processing can be manipulated to achieve these states. The project will train students in integrated design and modeling of next-generation materials, develop online learning modules related to nanocomposites and their applications, and implement a shared, open source repository of research data and machine learning tools for wide dissemination in the scientific community.

TECHNICAL SUMMARY

It is now well-accepted that adding nanoparticles (NPs) to commodity polymers can lead to hybrid materials with substantially improved properties. The most significant complication encountered, which frequently prevents these property improvements from being realized, is that inorganic NPs are hydrophilic while organic polymers are hydrophobic. These physical mixtures thus have a strong propensity to phase separate. In contrast to these expectations, a large body of experiments has shown that the process of creating these nanocomposites, e.g., by solvent casting, can leverage a variety of non-equilibrium phenomena to yield dramatically different but temporally stable NP dispersion states. A canonical example is the competitive sorption of the polymer in solution to the NP surface; this leads to the formation of a long-lived bound polymer layer which sterically stabilizes well-dispersed NPs. The goal of this proposed work is to quantitatively understand the poorly enunciated science underpinning solution-based processing protocols so as to obtain NP dispersion states with optimized mechanical properties at will.

The proposed work will synergistically combine data rich methods and theory/computer simulations on two inter-related tasks: (1) An ML algorithm will be trained on the available, large database of polymer/NP composites that have been experimentally cast from a range of different solvents and the NP dispersion states that result after solvent removal. After training, these ML methods will be able to identify critical regions of parameter space where interesting phenomena occur, e.g., where the material goes from well-mixed to agglomerated NPs. The work will then focus on these regions and use a combination of computer simulations and theory to delineate the physics that control the solvent casting process. (2) The role of different NP dispersion states on linear and non-linear mechanical properties will then be quantitatively enumerated. This integrated data science-theory-simulation-experimental data workflow will enable answers to several key scientific questions: (1) What is the space of polymer-NP-common solvent interactions that yield different NP dispersions? Previous work has suggested that the critical parameter is the effective solvent mediated polymer-NP interaction energy. Is this description accurate, and can effective NP-NP interactions be derived through known metrics such as solubility parameters and measured NP surface potentials? (2) What is the structure of the bound layer formed when polymer/NP interactions are more favorable than solvent/NP interactions? How does the structure of this bound layer depend on solvent quality and how does it yield good dispersion? (3) Going beyond casting from one solvent, how does the addition of a second, non-solvent allows for the precipitation of a NP-polymer composite with well-dispersed NPs? Does this process really only utilize entropic factors in effecting NP dispersion? (4) Under what conditions do kinetic issues, such as solution viscosity, become important determinants of NP dispersion? (5) How does NP dispersion state affect mechanical properties in the linear and non-linear regimes? Can the optimal NP dispersion states (and the associated solvent casting conditions) for mechanical properties be located and understood?

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.