a.

10.3. Correlation of Moisture Infusion, Leak Size, and Device Reliability

The first phase of the study to derive a quantitative relationship between leak size in hermetic packages and moisture infusion (NBS Spec. Publ. 400-19, pp. 52,54) was completed with analysis of the moisture sensor evaluation and refinement of procedures for microvent fabrication.

Two moisture sensors, one an interdigitated thick film finger structure and the other an electrode system between the die attach area and the package leads, were fabricated on each of 25 open ceramic, 14 pin, dual-inline packages. To reduce extraneous leakage, all exposed surfaces except those associated with the sensors themselves were coated with ероху.

To test the sensors, the package was mounted on a temperature controlled probe in an environmental chamber which was first stabilized in a dry mode and then set to give the desired relative humidity in the range from 5 to 50 percent; it was necessary to use high purity deionized water (7 to 10 M⚫cm) to control the humidity in the chamber. The thermoprobe was stabilized at an initial reference temperature and then its temperature was reduced in small increments while the leakage current which resulted from an applied potential of 50 V was monitored. The dew point was detected by observation of the abrupt change in current as shown in the current-temperature characteristics of figure 50a [92]. The change in current is caused by conduction through the water condensed on the sensor surface. In an integrated circuit package the normal contamination is a complex electrolyte mixture of reasonably high conductivity. However, the moisture in the environmental chamber was much less conductive, and the observed currents were orders of magnitude lower than found typically in sealed packages which contain moisture. To establish the position of the dew point under

these conditions it was required that three consecutive data points in the flat region have currents within 10 percent of each other and that the peak current be more than twice the average current in the flat region; the dew point was taken as the temperature at which the current was 120 percent of the average current in the flat region. Of the total of 80 test runs, dew points could be determined for 63.

Besides the problems of low currents, anomalies occurred in some cases because of moisture absorption by the epoxy which was used to coat the specimen and printed wiring board. The curve in figure 50b shows the appearance of the current temperature characteristic obtained in test runs that were affected by this phenomenon. These characteristics showed only a monotonic rise without a flat band. A few runs, in which the temperature range did not include the dew point resulted in a flat current-temperature characteristic as shown in figure 50c.

Analysis of the data indicated that there was no systematic error in the use of either dew point sensor but there was a random variation of about 5°C. Most of this variation appears to be related to the techniques required to combine a macro-environment and a microsensor: psychrometric error and gradients in the environmental chamber, the hydroscopic nature of the epoxy used to seal the printed wiring board, temperature differential between the thermoprobe and sensor, and perturbation of the air in the chamber by the radiating surfaces of the thermoprobe. Evaluation of procedures for calibrating and using the sensors are continuing.

A combination of laser drilling and various multiple electrochemical processes were used to fabricate 100 sample microvents. Both scanning electron microscope analysis and helium leak detector measurements were used to evaluate the fabrication procedures. Leak rates ranged from about 5 x 10-8 to over

TEMPERATURE

1 x 10-5 atm-cm3/s. The lids were sealed to headers with a yield of 98 percent. The units with gross leaks were subjected to the weight gain test [93] while the remainder were measured by the radioisotope leak test [94]. A number of units were stored in 85 percent relative humidity at 85°C for 168 h and leak tested again. The test data indicated that the microvents were highly unstable under conditions of repeated leak test,

because of distortion of the lids by the test pressurization. When stiffeners were added to the lids, satisfactory agreement was obtained between leak test results before and after exposure to 96 h storage in 85 percent relative humidity at 85°C. The microvent design has been modified to include integral stiffeners.

(S. Zatz and S. Ruthberg)

Work performed at Martin-Marietta Aerospace,

Orlando Division under NBS Contract No. 535880.

11.1. Dual-Laser, Flying-Spot Scanner

The photoresponse of the substrate diode of an integrated circuit was calibrated in terms of device temperature as an additional example of the usefulness of the electronic thermal mapping technique described previously (NBS Spec. Publ. 400-19, pp. 60-61) in connection with mapping the temperature distribution of a discrete UHF transistor. The integrated circuit studied was an array of five 750 mW silicon npn transistors on a common substrate with the substrate p-n junction located about 10 um below the top surface. The relative photoresponse to low-power 1.15 μm laser irradiation of this junction was calibrated against temperature by mounting the device in a controlled-temperature heat sink and measuring the photocurrent at a constant reverse bias of 30 V over the temperature range from 35° to 150°C.

The results are compared in figure 51 with the results of a calculation based on published values [95] of the variation of the optical absorption with temperature. For the calculation it was assumed that all absorbed radiation produced hole-electron pairs that were collected at the junction; with this assumption the photoresponse exponentially increases with temperature at a rate of about 2.6%/°C. The departure of the datum points from the line at the higher temperatures can probably be attributed to the fact that other mechanisms must be taken into account. Even

around 150°C the rate of increase is about 1.6%/°C which is large enough for the phenomenon to be a practical temperature indicator. (D. E. Sawyer, D. W. Berning, H. P. Lanyon*, J. D. Farina, and D. L. Blackburn)

11.2. Automated Scanning Low Energy Electron Probe

Initial investigations of wafer defects were begun by means of the automated scanning low energy electron probe (ASLEEP) (NBS Spec. Publ. 400-19, pp. 55-56) on a 2-1/2 in. (63 mm) diameter silicon wafer which contained process induced defects as indicated in the x-ray topograph of figure 52. For ease of mounting in the ASLEEP system and to provide samples which could be subjected to different processing steps the wafer was divided into quarters.

The first quarter was given an HF dip to remove the oxide and placed in the ASLEEP system. Curved lines concentric with the circular edge of the slice were observed. These are suspected as being due to the polishing operation when the wafer was manufactured. Figure 53 is an ASLEEP photograph showing these features. The scale of the x-ray topograph is not large enough to determine unambiguously whether these lines appear or not. However, examination in other regions of the quarter showed faint lines which intersect at a 60 deg angle as expected for dislocations. The specimen charges readily and hence the faintness of the lines may be due to an oxide layer covering the specimen.

The second quarter was examined in an electron diffraction camera and was found to have a thick oxide coating. The oxide was stripped from the specimen and then it was cleaned using a spin cleaning process where deionized water and solvents are applied to the center of the spinning specimen and centrifugal force slings the contaminated material off the edge of the wafer. The specimen was returned to the electron diffraction camera and the surface was observed to be single crystal silicon. The specimen was then installed in the ASLEEP system. One of the first things observed was a series of concentric rings centered on the center of the quarter rather than the center of the slice.

men.

The cylindrical secondary electron detector (NBS Spec. Publs. 400-4, pp. 54, 56, and 400-8, p. 36) was installed in the specimen chamber of the scanning electron microscope and its response tested. To install the new detector it was necessary to move the standard Everhart-Thornley type detector to provide added space within the specimen chamber and to eliminate the possibility that the two detectors would come into contact. It was also necessary to replace the original single conductor leads to the detector with coaxial leads. An adapter providing nine coaxial feed-throughs was made to fit the specimen chamber airlock. The use of coaxial leads removed 60 Hz pickup almost entirely and reduced to some extent other components of noise.

Various combinations of bias conditions were established on the detector and, for each, the collected current was measured as a function of specimen voltage. Figure 56

is a plot of the curves obtained for the The four conditions listed in table 10. shapes of curves such as B and C in the figure have been attributed [96] to the presence within the detector of tertiary electrons when secondary electrons strike the interior surfaces of the detector. These curves have the steepest slopes in the region about zero bias, ~70 pA/V, so that a 1 mV change in specimen otential would result in a 0.07 pA (or 0.1 percent) shift in collector current.

When the copper stub was replaced by a silicon chip, 0.8 mm square, containing seven diffused resistors and associated bonding pads mounted on a TO-5 header, the results were disappointing. The collected current was very low and did not evidence significant change as the voltage was varied on the resistive patterns. These effects were due, at least in part, to the reduced size of the specimen and lower secondary emission of silicon as compared with copper. Several techniques were tried to improve the response including increased bias voltage on portions of the detector, a positive bias on the collector electrode, changes in accelerating voltage, and increased beam current. None gave the desired improvement; in fact, in many cases the noise level increased so as to completely mask any signal.

On the basis of this evaluation, it has been concluded that the spatial resolution of this detector, as it is presently constructed, is inadequate for examining integrated circuits. Further studies of this type of detector are not planned at the present time.

(W. J. Keery