Fatigue of Glass-Ceramics: Obtaining v-K Diagrams From Semi-Elliptical Precracked Specimens

Objectives: The fatigue behavior of brittle materials is commonly studied using the relationship between the crack-velocity (v) vs. the applied stress intensity factor (K_appl), i.e. v-K diagrams. From those, one can derive important subcritical crack growth (fatigue) parameters, such as the threshold K_th for crack initiation and the n-fatigue exponent of the Paris’ Law. Because real-time crack growth experiments are complex to perform and common fatigue approaches must deal with the scatter of defect sizes in large datasets, new methodologies for efficient testing are overdue.
Methods: Entire sets of glass and glass-ceramic square specimens (n=30) for biaxial bending were precracked using the surface-crack-in-flexure method to obtain geometrically valid truncated semi-elliptical precracks. By pre-polishing the set to the same surface plane and posterior indentation and polishing, an entire set of specimens was obtained with identical precrack size and geometry. When subjected to static and cyclic fatigue experiments in decreasing K_appl, fracture toughness (K_Ic)-normalized K_appl/K_Ic vs. lifetime (t_c) plots were obtained as ∂ln(K_appl/K_Ic)/∂ln(t_c) and used to convert the dataset in v-K diagrams using fracture mechanics relations. This was conducted in static and cyclic loading for up to 100 h for several synthesized SiO₂-xLi₂O + P₂O₅ glass systems and a lithium disilicate glass-ceramic.
Results: We establish the accuracy of this method by demonstrating the reduced scatter in the v-K diagrams as compared to datasets in polished specimens, while obtaining comparable K_th and n-exponents. We show no effect of SiO₂/Li₂O ratio or P₂O₅-induced phase separation in the n-exponents. The “true fatigue effect” of R-curve degradation is shown for the glass-ceramic by comparing v-K diagrams under static and cyclic loading.
Conclusions: Our method of constructing v-K diagrams for glasses and glass-ceramic materials using limited datasets having same-size precracks was shown to be efficient and reproducible, allowing for faster and more accurate testing.