compatible with automatic dc testing equipDent available to the industry. The test pattern is designed with a 2 by 10 modular cell arrangement; five different classes of advanced test structures have been included for evaluation. In addition, a variety of conventional test structures is included to permit comparison measurements as well as detailed study of the cross-bridge structure.

Experimental work on the development of a test pattern for characterizing and monitoring a large-scale-integration, silicon-onsapphire process was completed at RCA Laboratories. Complete testing routines were established for all of the structures located in parts 1, 2, and 4 of the pattern which are intended to be tested on automatic electronic test equipment. An extensive report on this work is being prepared.

Assembly and Packaging - Detailed analysis was completed of the results of the interlaboratory evaluation of the radioisotope method for testing hermeticity of semiconductor devices being developed by ASTM Committee F-1 on Electronics. The estimated standard deviation ranged from 5 × 10-9 atm cm3/s for for a leak size of 1 × 10-8 atm cm3/s to 3.2 × 10-5 atm cm3/s for a leak size of 5 x 10-5 atm.cm3/s. Significant individual laboratory biases were observed; if these could be eliminated by use of suitable calibration artifacts, it is estimated that the standard deviation could be reduced by a factor of about 2.

Device Inspection and Test Studies of the application of the scanning acoustic microscope to the investigation of integrated circuits continued at Stanford University and Hughes Research Laboratories. The Stanford

group has been able to observe individual stacking faults in silicon epitaxial layers with a clarity similar to that obtained in high quality interference contrast photomicrographs, as well as defects underneath metallization layers. Considerable understanding of the response of the scanning acoustic microscope to various layer structures typical of those found in integrated circuits is being obtained through mathematical analysis. The Hughes group is investigating techniques for preparing reflection gray-scale bar charts for use in measuring the acoustic contrast characteristics to complement earlier measurements of spatial resolution.

The experimental investigation of the suitability of using the temperature response of the substrate diode in an integrated circuit to evaluate the integrity of the die attachment was completed. Results of tests on integrated transistor arrays with voided regions intentionally introduced into the die attachment showed correlation between voided area and increase in temperature of the substrate diode junction. The sensitivity under pulsed conditions appeared adequate to detect void areas as small as 17 percent of the die area. These results suggest that at least in some cases the temperature response of the substrate diode can provide information regarding the quality of the integrated circuit die attachment.

An improved test method for determining the saturation voltage of bipolar transistors was developed and initial laboratory tests were conducted to validate the procedure. This test is intended for use in hardness assurance applications where it is desired to compare measurements of saturation voltage made before and after irradiation.

3.1. Photovoltaic Method

The resistivity uniformity of the starting material plays a critical role in determining the characteristics and the performance of high power semiconductor devices. Variations in the resistivity of a slice of starting silicon not only cause variations in the characteristics of devices fabricated from that slice, but also contribute substantially to poor yield in the manufacturing process and adversely affect the reliability of finished devices. It is the purpose of this work to refine the photovoltaic scanning technique for measuring radial resistivity variations of circular semiconductor slices [1]. This technique is nondestructive in that no contact is required with the wafer surface area in which devices are to be fabricated; also it has better spatial resolution than the widely used four-probe method [2,3].

Instrumental Development* An automated data collection and analysis system has been designed and assembled (NBS Spec. Publ. 400-29, pp. 13-15) [4]. This system enables a user to measure the resistivity profile along a wafer diameter in about 2 min with some usersystem interaction after the measurement has begun. Additional equipment modifications are being made to eliminate the user-system interaction after initiation of the measurement and to automate the measurement of the average resistivity of the slice by the van der Pauw method [5].

Because a number of problems, mostly mechanical, had been encountered with the original specimen holder with knife-edge contacts (NBS Spec. Publ. 400-29, pp. 13-15), a second holder with point-probe contacts, shown in figure 1, was fabricated and installed. With this holder, contacts are made to the top surface of the slice, about 0.25 mm from the periphery. It is expected that this location is close enough to the edge that no correction for probe position is required. Since the peripheral ring is usually removed during device fabrication, any damage which might be introduced by the probes would not affect the finished device.

(G. J. Rogers and D. L. Blackburn)

System Characterization* - Study of the current-voltage characteristics of the pointprobe contacts showed that the contact resistance was less than that of knife-edge contacts to the same slice, that the linearity

Figure 1. Photograph of probe holder with point contacts for photovoltaic scanning apparatus.

of the point-probe contacts was better than that of the knife-edge contacts, and that th point-probe contacts were more repeatable from slice to slice.

During initial measurements of resistivity profiles, several problems were encountered in both the photovoltage and photoconductivity measurements. A "lateral" photovoltage appeared between the two contacts at the sam end of a measurement diameter; consequently, different photovoltages were measured by the two sets of contacts and the correct photovoltage could not be established. The measurements of photoconductivity, which require use of both sets of contacts, were found not to be reproducible. It was found that both problems were caused by large photo voltages generated by light scattered to the metal-silicon contacts. By carefully shielding the contacts from external stray light

Principally funded by the Energy Research and Development Administration, Division of Electric Energy Systems.

=

=

and using a strongly absorbed exciting laser beam (wavelength 632.8 nm) in place of the previously used very penetrating one (wavelength 1.15 μm) which might be scattered internally to the contact area, it was possible to effectively eliminate the lateral photovoltage in all but a few instances. In these cases, which are still under investigation, the photovoltage measured appears to be related to the surface condition of the wafer being measured.

With the elimination of the lateral photovoltage, it became possible to compare resistivity profiles measured with the photovoltaic and four-probe methods. Numerous such comparisons were made using the specimen holder with shielded point-probe contacts and the 632.8-nm laser source. Four examples are shown in figure 2. The results of the photovoltaic measurements were adjusted so that the average resistivity of the measured profile equals the resistivity of the slice as measured by the van der Pauw method. This procedure establishes the absolute resistivity at the initial point of the profile.

All of these measurements were made on specimens on which both surfaces had been lapped with 12-um alumina grit. It has been found that the most repeatable results and profiles most like those determined by the four-probe method are obtained on this surface or on a lapped surface that has been lightly etched. These are the types of surface finish usually present for resistivity profiling of thyristor grade silicon because an abraded surface is usually desired prior to and during the initial diffusion process for these devices.

Preliminary measurements were made to evaluate the repeatability of the system. Eight measurements of the resistivity gradient were made over a period of four days on a single slice. The results are plotted in figure 3 together with two four-probe measurements of the profile. The spread in the photovoltaic data in this figure is less than 3 percent of the average resistivity. This is probably sufficient repeatability for a rapid, nondestructive method for measuring incoming floatzoned slices.

(D. L. Blackburn and G. J. Rogers)

Theoretical Analysis - The previous analysis of the photovoltaic and photoconductive signals [1] was reviewed with the intent of removing as many simplifying assumptions and restrictions as possible. Basically the same conformal mapping approach was used to satis

of the scanning diameter and to include the effects of nonradial gradients in resistivity. The specimen geometry is depicted in figure 4 where the slice is aligned with its x-axis along the scanning diameter and probes are placed every 45 deg around the slice circumference. Probes 1 and 5, at the ends of the scanning diameter, are those used with the present measuring technique. Probes 3 and 7 along the y-axis could be used with the present technique to measure nonradial gradients in the y-direction. Probes 2, 4, 6, and 8 on diameters oriented at 45 deg with respect to the scanning diameter are the ones used in the 45-deg orientation geometry.

(x-y geometry) or at the ends of the diameters oriented at 45 deg with respect to the xaxis (45-deg geometry). In the table the subscripts indicate the probes between which the voltage is measured or between which the current is applied, e.g., V51 is the potential difference between probes 5 and 1 and 151 is the current between probes 5 and 1.

The symmetry of these equations is significant, although it should be pointed out that V and AR have somewhat different definitions in the two geometries. Simple symmetry arguments show that nonradial (tangential) resistivity gradients do not influence the photo

are the mobilities of the majority and minority carriers, respectively, is the average resistivity of the slice, b is the radius of the slice, and t is the thickness of the slice. The constant C1 and the quantities f(x), V(x), and AR(x), which depend on the position of the light spot x (which is assumed to have a radius a ‹‹ b), are listed in table 1 for four cases of interest: the x- and y-gradients of resistivity measured with probes at the ends of the x- and y-axes

RADIUS

SCANNING DIAMETER

2

Figure 4. Schematic diagram of specimen geometry used for analysis of photovoltaic method for measuring radial resistivity variation.