Figure 17. Measured pull strength of unannealed, round-wire, two-level bonds as a function of loop height above the high pad. (Solid points are for the first bond made on the high pad; open points are for the first bond made on the low pad. The data points represent the mean of 10 bonds, all of which ruptured at the heel of the first bond. Error bars indicate one sample standard deviation above and below the mean. The solid curves were calculated by resolution-of-forces.) Figure 18. Normalized pull strength of unannealed, roundwire, two-level bonds as a function of loop height above the high pad. (The values are normalized to the mean pull strength at the lowest loop height. Solid points are for the first bond made on the high pad; open points are for the first bond made on the low pad. The data points represent the mean of 10 bonds, all of which ruptured at the heel of the first bond. Error bars indicate one sample standard deviation above and below the mean. The solid curves were calculated by resolution-offorces. The bonds used in this experiment were made on a different bonding machine from that used to provide the data for figure 17.) 5. VARIABLES NOT INVOLVED IN RESOLUTION-OF-FORCES CALCULATION In this section, two of the variables which are not involved in the resolution-offorces calculation are considered. These variables are the rate of pull and the angle a (the angle between the direction of pull and the normal to the substrate in the plane perpendicular to both the substrate and the plane of the undisturbed bond loop). 5.1 RATE OF PULL As the resolution-of-forces calculation is of a static nature, an investigation was undertaken to see if there were appreciable effects on the pull strength brought about by any transient effects induced by a rate dependent pull force. This is especially important as in industrial use of the pull test the wire is frequently pulled rapidly or jerked which may give rise to large transient effects. This is of potentially greater importance as the elongation of the wire increases, since the effects of larger transients may be accentuated by greater wire elongation. In a first series of experiments, a group of 10 unannealed bonds was pulled to destruction at 11 different pull rates ranging from 1.0 gf/s (9.8 mN/s) to 12.5 gf/s (122 mN/s) The results of this experiment are presented in figure 19 where the mean pull strength (with the 95 percent confidence intervals for the mean) is plotted against the pull rate. The slope of the curve was determined to be essentially zero. It was, therefore, concluded that the pull rate has no significant effect on the bond pull strength for the range of pull rate between 1.0 and 12.5 gf/s (9.8 to 122 mN/s) [12]. In a second series of experiments, the investigation of the effect of the rate of pull on pull strength was extended to cover rates of 1.0 and 77 gf/s (9.8 and 764 mN/s). As with the first series of experiments, all bonds were made with the same bonding schedule in sequence by one operator. Also, this schedule was set up such that the failure mode was breakage at the heel of the first bond. Groups of 60 bonds each were then pulled at rates of 1.0 and 77 gf/s (9.8 and 764 mN/s) resulting in times to pull a bond of approximately 9 and 0.12 s. The date obtained indicated that there was no statistically significant difference in the measured pull strengths for either pull rate from those in figure 19.* The above experiments were performed on single-level unannealed bonds where heel breakage was the failure mode. For bonds which fail by peeling or lift-off, it is assumed that the rate of pull would have a large effect on the measured pull strength since peel failure appears to be rate sensitive. However, it is very difficult to show this statistically as a procedure to make bonds reproducibly that will fail by peeling has not been found. Attempts have resulted in pull strengths which have greater variations from bondto-bond, at a given pull rate, than between groups pulled at different rates. Therefore, a pull-rate experiment as described was not performed for bonds which fail by peel. However, it appears that the nature of the failure mode may affect comparisons of pull strengths Recent work (Kessler, H. K., in Semiconductor Measurement Technology: Progress Report, Bullis, W. M., Ed., NBS Special Publication 400-25 to appear) indicates that the pull strength of gold wire is also pull-rate independent. Figure 20. Normalized pull strength as a function of pull angle, a, for round-wire bonds with small (•) and large (▲) deformation. (Error bars represent the 95 percent confidence interval for the mean. The values are normalized to the mean pull strength at each deformation at a = 0 deg.) and rates of pull [13]. 5.2 ANGLE a 5.2.1 INTRODUCTION The resolution-of-forces calculation as presented in Section 2 and the Appendix of thi report is essentially a two-dimensional calculation. That is, only those forces operative in the plane of the bond loop are considered. A series of experiments was undertaken to in vestigate the effects of pulling the bond in a direction not in this plane on the measured pull strength. The angle a is defined as that between the direction of pull and the normal to the substrate in the plane perpendicular to both the substrate and the plane of the undisturbed bond loop. Since the principal effect of pulling at a finite value of a is to rotate the plane of the bond loop, the ratio of the force of pull and the force exerted along the wire is the same for all values of a. This would lead one to conclude that no great dependence on the measured pull strength on a would be expected. However, the rotation of the plane of the bond loop does not include rotation of the bond heel. This tends to put the bond heel under an anisotropic stress with respect to the line between the bonds This anisotropic stress may be expected to be accentuated by greater bond deformation. This reasoning points to the possibility of decreased measured bond pull strength with increased α. This decrease is expected to be most pronounced with a large deformation. 5.2.2 SINGLE-LEVEL BONDS Single-level round-wire bonds were made on three metallized substrates with 0.001-in. (25-um) diameter aluminum (1% silicon) wire and a bond-to-bond spacing of 0.04 in. (1.0 mm). All the bonds on a particular substrate were made with the same power setting, but different power settings were used in making the bonds on different substrates to obtain bonds with different degrees of bond deformation. On each substrate, about 35 bonds were pulled at the midpoint of the loop for eight values of a from 0 to 20 deg. The pull angle ☀, in the plane of the bond loop, was O deg. For each group, the mean pull strength and the 95 percent confidence interval for the mean was determined; the measured pull strengths of bonds that did not break at the heel were excluded from the calculations. The pull strengths were then normalized to the mean value obtained at a = O deg. The results for bonds on two substrates with small deformation (1.25 to 1.50 wire diameters) and large deformation (2.25 to 2.50 wire diameters), are shown in figure 20. The zero degree value for pull strength of the bonds with large bond deformation was approximately half that of the bonds with small bond deformation. Bonds made on the third substrate had excessively large bond deformation but yielded results essentially the same as those observed for bonds with large bond deformation. The results show that as a increases, the measured pull strength decreases. This appears to be a result of the twisting or tearing of the bond heel as the wire is pulled out of the plane normal to the substrate plane. These results show that, although the value of angle a has little effect on the measured pull strength of bonds with small to moderate bond deformation, the effect is more pronounced for the bonds with greater bond deformation. Although it was not experimentally demonstrated, it is hypothesized that if the failure mode were by peel significantly lower pull strengths would occur at large values of a. This reasoning follows that presented in the introductory material. For peel failure at a=0, the force tending to lift the bond acts uniformly over a large area. For a 0, the force I tends to be concentrated on one side of the bond and hence initiates lift-off at a lower pull force. It should be possible to control the position of the pulling hook so that the angle a is less than 10 deg by visual inspection, but more precise positioning may be required for bonds with large deformation [14]. 5.2.3 TWO-LEVEL BONDS The experiments on the effects of the angle a on the pull strength were carried out for small deformation two-level bonds in which the first bond was made on the high pad. Bond pairs were made on two different bonding machines designated as machine A and machine B. The measured pull strength, normalized to its value for a = 0, is presented as a function for a for both cases in figure 21. As with the analysis of the single level data, only those failures due to rupture at the first bond were included in the data. The curves in figure 21 show roughly the same general trends in both cases except for differences in the standard deviations. The results are in general agreement with the measurements of pull strength made on single-level bond pairs of small deformation [5]. Figure 21. Measured pull strength as a function of angle of |