1630 mObERLYChAN et al: ROLES OF AMORPHOUS GRAIN BOUNDARIES 1950°c Fig. 5. Bright field TEM image of a crack path along grain boundaries in ABC-SiC hot pressed at 1950C for I h. This crack was propagated by bending a doubly-dimpled TEM specimen prior to final hicker amorphous layers in Hexoloy-SA also phase could have resulted from different sintering appeared to vary in thickness along a single grain additives in other reports boundary; however, this observation could have Techniques other than HR-TEM EDS(Energy been partially inflated by imaging conditions chan- Dispersive X-ray Spectroscopy), PEELS(Parallel ging due to the grain boundaries not being atomic- Electron Energy Loss Spectroscopy ) energy-filtered ally flat and or to subtle bending of the TEM thin TEM [36] were utilized to corroborate the presence of this grain boundary layer. Scanning auger elec- The nature of TEM sample preparation of poly- tron spectroscopy(AES) of the fracture sur crystalline materials and the requirements of HR. of ABC-Sic provided the most definitive chemical TEM imaging precluded detection of an amorphous analysis(Fig. 7). AEs detected a thin alumina-con- phase at every grain boundary. Even the grain taining pha Ise on all exposed grain boundaries boundary interface incorporating a basal plane had which was removed with <I nm ion etching occasional steps, often I nm in height. On the [Fig. 7(a)and 7(b)]. The aluminum detected on the atomic level, the grain boundaries which did not in- as fractured surface exhibited a shift in energy typi cal for an oxidized form of al such merous steps. This typically resulted in the thin The intensity of the o signal was diminished as amorphous boundary layer being neither parallel to compared to ideal sapphire [34], due to the thin the electron beam nor discrete in its projected po- nature of the amorphous layer and possibly the tential throughout the thickness of the TEM speci- amorphous phase being oxygen deficient. The Si men. Thus a(possible) thin amorphous layer often present in the spectrum of the as fractured surface could not be observed, most probably as a result of did not exhibit an oxidized signature, thereby indi nonideal imaging orientations. Such conditions are cating most of this Si signal was from the Sic grain more pronounced for smaller grain sizes and shorter below a thin surface layer. a trace sulfur signal was detected as fractured [Fig. 7(a) and after removal grain boundary facets; and this may account for of <I nm ion etching (Fig. 7(b)]. but was removed which reported no amorphous grain boundary by subsequent I nm) ion etching. Because of its layer[1, 21, 23, 31-33]. Also different observed grain came primarily from the SiC grain beneath the boundary phases or the lack of a grain boundary grain boundary layer. However, some C was believed soluble in the amorphous boundary layer, 个 A Perkin Elmer小660 was operated at3kV as many of the crystallized triple junctions conthicker amorphous layers in Hexoloy±SA also appeared to vary in thickness along a single grain boundary; however, this observation could have been partially in¯ated by imaging conditions chan￾ging due to the grain boundaries not being atomic￾ally ¯at and/or to subtle bending of the TEM thin foils. The nature of TEM sample preparation of poly￾crystalline materials and the requirements of HR￾TEM imaging precluded detection of an amorphous phase at every grain boundary. Even the grain boundary interface incorporating a basal plane had occasional steps, often 1 nm in height. On the atomic level, the grain boundaries which did not in￾corporate a basal plane were twisted and had nu￾merous steps. This typically resulted in the thin amorphous boundary layer being neither parallel to the electron beam nor discrete in its projected po￾tential throughout the thickness of the TEM speci￾men. Thus a (possible) thin amorphous layer often could not be observed, most probably as a result of nonideal imaging orientations. Such conditions are more pronounced for smaller grain sizes and shorter grain boundary facets; and this may account for many of the HR-TEM analyses in the literature which reported no amorphous grain boundary layer [1, 21, 23, 31±33]. Also di€erent observed grain boundary phases or the lack of a grain boundary phase could have resulted from di€erent sintering additives in other reports. Techniques other than HR-TEM EDS (Energy Dispersive X-ray Spectroscopy), PEELS (Parallel Electron Energy Loss Spectroscopy), energy-®ltered TEM [36] were utilized to corroborate the presence of this grain boundary layer. Scanning auger elec￾tron spectroscopy (AES){ of the fracture surfaces of ABC±SiC provided the most de®nitive chemical analysis (Fig. 7). AES detected a thin alumina-con￾taining phase on all exposed grain boundaries, which was removed with <1 nm ion etching [Fig. 7(a) and 7(b)]. The aluminum detected on the as fractured surface exhibited a shift in energy typi￾cal for an oxidized form of Al, such as alumina. The intensity of the O signal was diminished as compared to ideal sapphire [34], due to the thin nature of the amorphous layer and possibly the amorphous phase being oxygen de®cient. The Si present in the spectrum of the as fractured surface did not exhibit an oxidized signature, thereby indi￾cating most of this Si signal was from the SiC grain below a thin surface layer. A trace sulfur signal was detected as fractured [Fig. 7(a)] and after removal of <1 nm ion etching [Fig. 7(b)], but was removed by subsequent (>1 nm) ion etching. Because of its higher energy and greater escape depth, the C signal came primarily from the SiC grain beneath the grain boundary layer. However, some C was believed soluble in the amorphous boundary layer, as many of the crystallized triple junctions con￾Fig. 5. Bright ®eld TEM image of a crack path along grain boundaries in ABC±SiC hot pressed at 19508C for 1 h. This crack was propagated by bending a doubly-dimpled TEM specimen prior to ®nal ion beam thinning. {A Perkin Elmer F-660 was operated at 3 kV. 1630 MOBERLYCHAN et al.: ROLES OF AMORPHOUS GRAIN BOUNDARIES