Figure 8. Optical photomicrograph of stress cracks of a CVD silicon dioxide film of 1-um thickness deposited over oxidized silicon with squares of 105 x 105 um of evaporated aluminum. Demarcation was achieved by 10-min aluminum etching. Cracks surround most of pattern along edge of aluminum. Differences in the width of the inside area are most probably caused by differences in the width of the cracks (150X, brightfield). Selected samples were examined by SEM at various angles and magnifications to provide high-magnification micrographic records of glass defects for comparison with optical photomicrographs, to measure the width of typical glass cracks, and to demonstrate the predicted cross-sectional shape of demarcation etched structures. results are presented pictorially in figures 9 to 13. 4.1.5 Parameters of Selective Aluminum Etching Representative The results of a systematic investigation aimed at optimizing the selective chemical etching technique for detecting localized structural defects in glass layers over aluminum-metallized ICs are summarized in this subsection. Variables examined include et chant composition, addition of surfactants, etching temperature and time, use of ultrasonic agitation, and vacuum impregnation to enhance penetration. The development of demarcation areas for microcracks in silicon dioxide layers over aluminum capacitors of IC test wafers with a high density of uniformly distributed microcracks was employed as a criterion for quality of resolution. 4.1.5.1 Optimum Etchant Composition - Nine different compositions, with and without surfactants, were formulated and tested at 50°C on a comparative basis. The etch rate of each composition was first determined for pure aluminum. On this basis the etch times to be used for evaluating the various compositions were then calculated using a constant etch time factor of 1.25, the factor for just etching through the aluminum being 1.00. The experimental data presented in table 1 (section 4.1.2) show that the defect detection sensitivity is the same for all compositions. One may therefore continue to use our "standard composition" defined in table 1. 4.1.5.2 Optimum Etching Temperature Using the standard et chant, effects of 10°C above and below the normal temperature of 50°C were tested by the technique described above. The data presented in table 2 show no difference in defect detection sensitivity, so that we have continued to use the temperature of 50°C, which affords a convenient etch time. 4.1.5.3 Optimum Etching Duration and Etch Time Factor - Using the same type of IC samples as above, we tested the optimum etch time at 50°C in standard etchant with respect to completeness of detection of all structural defects in the dielectric that penetrate down to the aluminum, to sharpness and definition of demarcated defects, and to rate of undercut-etching. Etch time was normalized and expressed in terms of the "etch time factor," defined as the actual etch time used for a sample over the etch time required to just etch through the aluminum. The half-widths of the aluminum removed under the glass served as a convenient measure of the lateral etch rate. Figure 9. Scanning electron micrographs at 5000X of untreated microcrack in 1.4-um-thick PSG over 1.8-um-thick evaporated aluminum on oxidized silicon substrate wafer. Width of crack is about 0.4 μm. (a)-45° view and (b)-25° view from perpendicular. |