Machine learning-assisted design and scalable fabrication of high-performance fire-safe polycarbonate for advanced applications

[Display omitted] •Machine learning aids in designing fire-safe PC for scalable production.•Achieving LOI value of 39.3 %, 0.8 mm UL-94 V-0, and low heat/ smoke release.•Great resistance to butane torch fire > 1300 °C and GWFI of 975 °C.•Enhanced mechanical strength and low dielectric constant were realized.•Fire-safe mechanism confirmed through experiments and calculations. Polycarbonate exhibits great potential in various advanced applications, however its unsatisfied fire-safe performance has hindered its widespread application. Here, we use a structure-based machine learning algorithm and establish two random forest models using the molar group contribution method to engineer a robust and multifunctional polycarbonate copolymer. Our design contains both cardo structures and poly(dimethylsiloxane) chains as functional groups to achieve desired performance. As a result, the copolymer achieved UL-94 V-0 rating with a thickness of 0.8 mm, showed 57 % lower peak heat release rate, 70 % lower smoke density, as well as high glow-wire ignition temperature of 875 °C and glow-wire flammability index of 975 °C. It should be noted that the copolymer exhibited great resistance to butane torch fire of > 1300 °C. Experimental and calculation results confirm a condensed-phase dominant mechanism based on thermal rearrangement and cross-linking. Furthermore, enhanced mechanical strength and low dielectric constant were also observed, powering the potential applications of the obtained copolymer in various fields. The successful application of machine learning on multifunctional fire-safe polycarbonate provides a promising avenue for the design of high-performance polymeric materials.