1.4 and 64 gf/s (0.014 to 0.63 N/s) and was controlled by a variable speed motor. At each of seven pull rates, 20 to 25 bonds were pulled to destruction. For pull rates lower than about 10 gf/s (0.098 N/s) the breaking force was recorded on an X-Y recorder (NBS Tech. Note 560, p. 37) but, because the recorder has a low (nominally 76 cm/s) slewing speed, the breaking force at more rapid pull rates was read directly from the gram gauge. The results, plotted in figure 44, show that the pull rate can vary over this range without a significant variation in measured pull force. The slight increase at higher pull rates is probably an artifact due to the inertia of the gram gauge.

Further investigation of the characteristics of the double deep-grooved bonding tool showed that a broad range of ultrasonic power can be used without appreciably increasing the bond deformation, which suggests that the bonding tool may stall after the bond is formed and the ultrasonic energy is no longer transmitted to the bonding interface. (H. K. Kessler)

9.3. Bondability of Doped Aluminum

Metallizations

A study was undertaken to evaluate the bondability of aluminum ribbon and round wire ultrasonically bonded to copper- and silicondoped aluminum metallizations. Recently improvements have been reported relating to the addition of silicon to aluminum metallization to prevent interdiffusion [86] and the addition of copper to aluminum metallization to prevent electromigration [86,87].

A 900-nm thick film of aluminum-3.5% copper metallization was deposited on an oxidized silicon wafer by using an aluminum-12% copper mixture as the charge in a one-pot electron beam evaporation system. Films of aluminum1% silicon and aluminum-1.5% silicon were ob

tained from commercial sources. The films were etched into square pads, 0.005 in (0.13 mm) on a side on 0.010 in. (0.25 mm) centers by conventional photolithographic techniques to form single-level substrates [84] for the bonding study.

Two types of wire were employed. Both were aluminum-1% silicon with 1 to 2 percent elongation and a breaking load of 12 to 14 gf (0.118 to 0.137 N). The ribbon wire had dimensions of 1.5 by 0.5 mil (35 by 13 μm) and the round wire had a diameter of 1.0 mil (25 μm) so that the cross sectional areas were essentially the same.

Power series [85] were carried out by varying the power dial setting for the first bond. All bonds were made with a bonding force of 25 gf (0.24 N) using a tungsten carbide tool with a foot length of 4.5 mil (114 μm). The power dial setting for the second bond was kept constant at 4.5. Bonding time for both the first and second bond was 50 ms. A11 bonds were pulled to destruction with a pull rate of 3.8 gf/s (0.037 N/s).

The results are presented in figure 45. For the case of aluminum metallization doped with 1 or 1.5% silicon the pull strength-power curves generally follow the shape obtained for pure aluminum metallization which suggests that both the ribbon and round wire can be bonded satisfactorily to these alloys. Although satisfactory bonds could be made to the copper-doped aluminum metallization, it was found that at high power (settings from 7 to 10) the variability was significantly larger than is usually obtained for bonding to pure aluminum. Also, for the round wire there was a sharp drop in pull strength at low power. For both copper- and silicon-doped aluminum metallization, the ribbon wire exhibited a higher pull strength than the round wire; this is consistent with previous results obtained on pure aluminum metallization [88]. (H. K. Kessler)

10.1. Gas Infusion into Double Hermetic Enclosures

In some applications, it is customary to incorporate hermetically sealed semiconductor devices and other components within an outer hermetic case. Intuitively, this would be expected to increase seal assurance. However, a detailed analysis of the gas flow equations suggests that a significant reduction of gas infusion occurs only under certain conditions.

Consider first the infusion of dry gas into the double enclosure depicted in figure 46. The outer case, which has an internal free volume of V1, has been leak tested to a value of L1 so that the leak size is L1 or lower. Inside, there is a smaller package of internal free volume V2 which had been leak tested to a value of L2 before being placed in the outer case. Since the infusion of a noncondensable gas into semiconductor devices from a dry atmosphere appears to be primarily by free molecular flow for the leak range < 1 × 10-5 atm-cm3/s, which is the range of interest in determining the fine leak reject limit, it is assumed that infusion through the leaks L1 and L2 is by free molecular flow.

Exact solutions of the flow equations have been obtained [89]. Pressure-time curves for a typical example are shown in figure 47 in terms of normalized variables. On the vertical axis the chamber pressure is normalized to the external driving pressure, P and on b' the horizontal axis the time is normalized to the time constant of the inner volume which is given by T2 PoV2/L2 where Po is atmospheric pressure. Curve 1 represents the case where the inner package has been exposed directly to the external gas; the pressure rises exponentially and reaches a value 63 percent of the external driving pressure at t = 12. Curves 2 and 3 depict the pressure in the outer and inner packages, respectively, if the smaller package is surrounded with

=

an outer case whose volume is 10 times that of the inner package but whose leak rate is identical. Note that for very small times the pressure rise is linear in time in the outer chamber but quadratic in time in the inner package. Thus, although pressure is much lower in the inner chamber at early time, it rises rapidly and reaches that of the outer chamber within a time less than 10τ2. By comparing curves 3 and 1, one can determine the reduction in pressure obtained with a double enclosure as compared to the single package by itself.

A complete description of pressure reduction can be obtained from the exact solution by computing the pressure time characteristics for any desired situation. The long term behavior of the system can be quantified by considering as a merit factor the ratio of the pressure of the inner enclosure when isolated to that when enclosed at the time t = T2. This allows one to represent the behavior of any combination of volumes and leak sizes by a single number. Further, examination of many characteristics shows that this merit factor also provides a rough approximation to the increase in time necessary for the gas concentration in the inner enclosure to rise to the same value as would exist in the single enclosure at t = 12.

=

It

The merit factor is shown in figure 48 for a broad range of leak size (y L1/L2) and volume (6 V1/V2) ratios. It is seen that simply surrounding one hermetic enclosure by another is not assurance of hermetic improvement, although it is obviously a protection against a badly leaking inner enclosure. is further seen that significant enhancement is only obtained if outer enclosure leakage is not greater than ten times inner leakage or the outer free volume is at least ten times that of the inner free volume. For vanishingly small outer free volumes, the effective leak rate approaches that of a single enclosure with leak conductances in shunt. The curves in this figure can also be used to establish the conditions where the use of a double enclosure can be expected to significantly reduce gas infusion and to determine the effects of leak testing sensitivity and precision on the assurance of seal quality [90].

The above analysis was made for a dry, noncondensable gas. While a package with the reasonably acceptable leak rate of 107 atm•cm3/s has a time constant of the order of 1 day to 4 months, depending on its volume, the effective service life of typical devices is many times longer. This is probably because the infusion rate of water vapor, which has been identified as a principal source of device degradation, into microchannels is complicated by condensation, absorption, etc. Unfortunately, there is no good model to relate the time constant of a package for a well behaved gas to the service life. However, since moisture infusion is much slower than the infusion of a dry, noncondensable gas, one may infer that the ratios presented in figure 48 would also apply to the longer

The limiting factor in dry gas, gross leak test procedures is the rapid depletion of gas from the package interior which results in the nondetection of large leaks. The static-expansion, differential-pressure test has been suggested (NBS Spec. Publ. 400-4, p. 70, and 400-8, p. 42) as one method to circumvent the problem associated with long dwell time between pressurization and testing. The procedure employs gas expansions in two similar and parallel systems. A differential pressure measurement made after expansion provides a quantitative measure of the leak size.

same time. After pressurization, the quantity of gas in the test and reference chambers is the same provided that there is no leak in the device under test. However, if the device leaks, the quantity of gas introduced into the test chamber is greater by the amount driven into the test device interior. The gas in the small volumes is then allowed to expand into the associated larger and previously evacuated identical chambers of volume v2 and V2. A difference in quantity of gas in the two volumes causes a meter indication. The use of a parallel reference chain eliminates the need for absolute measurement, minimizes the effects of adsorption and variation in valve closure, and allows a wide range of test object volumes and materi

5 als by adjustment of the tare, and the carrier where needed. Any test gas may be used.

PUMP

where L is the leak rate, vo is the difference between v1 and the external volume of the package, Vi is the internal volume of the package, Po is atmospheric pressure, and the other symbols have been defined previously. Thus it can be seen that the rate of change of indication is a measure of the leak rate and the amplitude is a measure of the internal free volume of the package. (S. Ruthberg)