Machine Learning and Structural Design to Optimize the Flame Retardancy of Polymer Nanocomposites with Graphene Oxide Hydrogen Bonded Zinc Hydroxystannate.

Abstract

Designing flame-retardant polymers with high performance is a long-standing challenge, partly because of the time-consuming traditional approaches based on experiential intuition and trial-and-error screenings. Inspired by the effective new paradigm of data-driven material discovery, we used machine learning to analyze experimental data to accelerate the development of new flame-retardant polymers. To explore the relationship between limit oxygen index (LOI) and components, we prepared 20 composites and then trained a simple equation for the LOI using the method sure independence screening and sparsifying operator (SISSO). The data analysis allows us for a better understanding of the flame-retardant mechanism and components, and the equation has good accuracy in guiding the design of composites with high flame-retardant performance. Meanwhile, the increasing structural design of flame retardants is crucial to flame-retardant polymer composites. We proposed a structure of nano graphene oxide (GO) wrapped micro zinc hydroxystannate (ZHS) in a simple but effective way as a novel flame-retardant agent to enhance the flame retardancy and mechanical properties of polypropylene (PP) composites. The GO sheets were like "light yarns" wrapped onto the ZHS via hydrogen bonding in an ethanol solution. The selected samples were analyzed to confirm the predictive LOI model. The resultant composites with the substitution of intumescent flame retardant (IFR) by 1.0, 2.0, and 4.0 wt % ZHS@GO conferred better flame retardancy compared with PP composite containing only IFR, reflected by the efficient increase of LOI value and V0 rating of UL-94 vertical tests. The analysis principles and facile fabrication strategies proposed in this work could be important for developing highly flame retardant composites.