Theoretical prediction of dental composites yield stress and flexural modulus based on filler volume ratio

•Turcsányi model is suitable to dental composites and 3-point bending test.•SEM observations showed a size limit of breakable particles in accordance with Turcsányi model parameter.•Models for flexural modulus deviate for high-filled composites.•Describing the filler-matrix interface quality in predictive models is required for accuracy with high-filled composites. A costly advantageous approach in composites development process is to limit experimental tests by predicting mechanical properties with respect to their filler ratio. Models exist for other fields than dentistry. They have been compared to 3-point bending test experimental results for yield stress, flexural modulus and flexural strength. Five formulations of the same experimental material were made. They were composed of an organic matrix and different ratios of silanated barium glass particles. The samples were stored in distilled water for 24h at 37°C prior to the 3-point bending test. The Turcsányi model for yield stress was notably investigated, and SEM was used to complete data analysis. The yield stress showed reproducible results and a good fit with Turcsányi model with respect to filler ratio. The flexural modulus data are not scattered but did not fit with the existing models. No trend could emerge for flexural strength and strain because of scattering; these properties are more unpredictable. The SEM observations of fracture areas confirm a good matrix-filler interface quality. SEM pictures validated the numerical parameter obtained from Turcsányi model. The latter therefore seems to be applicable to dental composites. Firstly, it enables to predict the evolution of the material yield stress without testing all filler ratios. Secondly, this model provides a good way to get micro-information on the matrix-filler interface from macroscopic tests. The discrepancy between flexural modulus results and theory highlighted the necessity to include an “interface quality” parameter in accurate predictive models.