Neutron diffraction study of in situ-reinforced silicon nitride during creep

creep;CTE tensor;high temperature;Neutron diffraction;non-destructive testing;silicon nitride;single crystal stiffness

Swift, Geoffrey Allen ; Ustundag, Ersan

University:California Institute of Technology

Department:Engineering and Applied Science

关键词: creep;    CTE tensor;    high temperature;    Neutron diffraction;    non-destructive testing;    silicon nitride;    single crystal stiffness;   

Others  :  https://thesis.library.caltech.edu/935/1/Swift_Thesis.pdf

美国|英语

来源: Caltech THESIS

【 摘 要 】

Neutron diffraction was used to study the high temperature creep of in situ-reinforced silicon nitride (ISR Si3N4).Full pattern and single peak fitting methods were used to calculate the average and hkl-specific mechanical properties, including thermal expansion coefficient, creep exponent, Young?s modulus, and Poisson?s ratio.This is both the first in-depth study using time-of-flight neutron diffraction to examine materials at such temperatures (1473 and 1648 K) and the first in situ microstructural study of creeping silicon nitride.Two commercial grades of ISR Si3N4 were tested, AS800 and GS44.The refractory grain boundary phase of AS800 prevented the onset of creep in the vacuum environment of the SMARTS furnace.However, the high temperature stress-strain data allowed determination of the 1648 K single crystal elastic stiffness tensor, the first such calculation from neutron diffraction strain data.Also determined was the coefficient of thermal expansion (CTE) tensor.The 1648 K stiffness tensor indicated a less stiff C33 component compared to a room temperature stiffness tensor.This lesser value is due either to microstructural or thermal effects.Creep was observed for GS44.A stress step-up test and four constant stress creep tests were performed.Large strains were measured by an extensometer, though of much less magnitude than literature creep studies, with the difference attributed to the vacuum environment protecting the grain boundary phase from suffering reduced viscosity.This is supportive evidence of the long-held notion that the grain boundary phase is the primary determinate of creep behavior.The diffraction strains, though of significantly lower magnitude than the extensometer strains, were measured as non-constant, though the silicon nitride lattice parameters behaved in an unexpected manner.The two lattice parameters were seen to fork from a nearly common initial strain, with the c lattice parameter indicating tensile strain and the a lattice parameter compressive.The relative changes, however, corresponded to an essentially constant unit cell volume, as computed from simple geometry.None of the potential inelastic strain effects on diffraction peaks, such as peak broadening, were observed, further supporting the notion that the grain boundary phase is the source of strain.Neither was there any measured preferred orientation evolution due to creep.The creep exponent of GS44 was calculated as 3.18, a greater value than in literature, likely due to the same creep inhibition of the vacuum furnace.The classic Norton Equation for creep matched well with the steady state creep rates as a function of applied stress, while a newer model by Luecke and Wiederhorn, incorporating multiple facets specific to Si3N4, matched the data comparably, though with an additional empirical stress dependence incorporated.The effect of performing these experiments in a vacuum rather than in air likely prevented as accurate prediction by their model as with Norton?s.This result is based on the much-reduced creep strain measured compared to literature measurements of the same material at like temperature and stress.However, given the large disparity between the extensometer creep strain and the diffraction creep strain, it is clear that the grain boundary phase experiences the bulk of the deformation.Subsequent SEM observations of tested samples indicated no microstructural change due to the short duration of creep experiments.As with AS800, the GS44 CTE tensor was found, while the stiffness tensor was incalculable due to extreme non-linearity of single peak data.