presence of positive mobile ion species is readily revealed by the negative shift of the flat-band voltage under positive BT stress. In this case there was a slight negative shift with the negative BT stress as well. The net flat-band voltage shift (AV) indiFB cates mobile ion contamination for these devices ranging between 6.7 and 8.7 x 1010 cm-2. Subsequent heat treatment of the wafer caused V. to drift back toward the as-proFB cessed condition but without full recovery.

Wafer 5C was subjected to electron irradiation, heat treatment, negative and positive BT stress, and a second heat treatment. There are two separate subsets of devices. One subset contained 16 devices which were measured in the as-processed condition in the usual way. The second subset contained

16 devices which were not measured in the as-processed condition; the devices in this subset, consequently, had not been probed prior to the electron irradiation. The data from the second subset are marked with an asterisk in the figure. Electron irradiation caused a large negative shift of the flatband voltage and simultaneously introduced a high density of interface states in the subset measured after irradiation. Heat treatment after irradiation appeared to remove the

interface states; the flat-band voltage for previously measured devices shifted to slightly more negative values whereas those measured for the first time after processing, irradiation, and heat treatment formed a tighter group at smaller flat-band voltages. With negative and positive BT stress, both sets of devices behaved similarly although the second subset of devices was affected less and always formed a tighter grouping. The net flat-band voltage shift resulting from negative and positive BT stress indicate ionic contamination ranging from 2.9 to 10.2 x 1010 cm-2 and from 1.7 to 5.0 x 1010 cm-2 for the first and second subsets, respectively. After subsequent heat treatment of the wafer the flat-band voltage relaxed toward the as-processed value.

Wafer 5D was treated similarly to wafer 5A, but the devices measured after BT stressing were different from those measured in the as-processed condition. Although the flatband voltage showed larger spreads, the behavior was essentially the same as that of wafer 5A.

These results are presented only to be indicative of the kind of information which might be expected from study of the BT stress test; they are too preliminary to warrant detailed

analysis and definitive conclusions. It should be noted that the use of the BT stress test as an indicator of mobile ion contamination assumes that the mobile ions merely shift the C-V curve in an amount proportional to the net charge moved to the interface. In general, however, this assumption is an oversimplification. Interface states can distort the C-V characteristic which can cause error in the determination of the flat-band voltage. This is particularly true of devices which have been electron irradiated and those which have high mobile ion contamination.

(R. Y. Koyama and D. A. Maxwell)

Among the potential applications of the modified instrument are the characterization of the silicon-glass interface under a thick (50 μm) layer of glass, the determination of the flat-band voltage under a thick field oxide on an LSI chip, and the characterization of silicon on sapphire and the siliconsapphire interface by using the sapphire substrate as the insulator in an MIS capacitor. The original modified instrument [41], although suitable for demonstration purposes and laboratory application, requires improvement and additional modification before it will be suitable for general use. This task was undertaken to develop improved bias protection circuitry to prevent damage to the measurement equipment in case of catastrophic dielectric failure of the specimen under high bias, to develop a high-voltage power supply with a linear sweep capability to facilitate use of the instrument in applications where

the capacitance is a function not only of the applied voltage but of the sweep rate as well, to develop an improved specimen holder with an integral safety interlock system, and to further demonstrate the applicability of the method.

The original modified instrument [41] had a potential problem when very high bias voltages were applied to the specimen under test. If the specimen capacitor developed a short circuit, a large part of the applied bias voltage was transferred to the capacitance meter, thus damaging it. To prevent this damage, a bias protection circuit was developed that prevents voltage excursions beyond the ±200 V limit of the capacitance meter. However, the use of the circuit limits the range of capacitance that may be measured without introducing excessive added error. The first version of the circuit limited the measured capacitance to the range 0 to 40 pF if the added error introduced by the circuit was to be kept below 1 percent. To overcome this limitation, the values of the components in the previous version of the circuit were optimized. The resulting improved bias protection circuit has been tested to +9600 V under repeated breakdown conditions. The added error in the range of measured capacitance 0 to 140 pF is less than 1 percent.

The modified technique [41] often requires the application of high bias voltage to the capacitor sample to be measured. When this bias voltage exceeds some threshold value (typically 3 to 5 kV), there exists the danger of breakdown along the insulator surface or at an electrode edge. One method for prevention of this breakdown [42] is to coat the surface and electrode edge with a greaselike silicone material; this is very effective but is messy, time-consuming, and inconvenient. It has been found that if an insulating silicone rubber toroid, having appropriate properties, is pressed over the specimen surface including the electrode edge, breakdown can be effectively prevented at applied bias voltages up to 10 kV. The surface of the rubber in contact with the specimen must be very smooth, and the rubber itself must be very resilient. It is also important

that the rubber contain no voids in the vicinity of the surface (in contact with the insulator) and that it have both a very high resistivity and a low dissipation factor.

Other instrumental improvements, including the power supply with linear sweep capability interlock, are being constructed. Initial evaluations have been made of the measurement

6.

FOR WAFER

6.1. Ion Implantation Parameters

This task was undertaken to develop and disseminate to the semiconductor device and integrated circuit industry practical information and experimental data to assist in improving the design and fabrication of ionimplanted doping profiles. Some of the important aspects of profile control still needed are accurate knowledge and control of dose and of ion range and straggle (or standard deviation). This control of implanted doping profiles in production is important because implantation is known to be capable of much greater control of doping profiles than diffusion, and it is possible to construct implanted profiles which cannot be obtained by diffusion. However, this potential superiority of implantation can be lost to the practical world of device and circuit fabrication unless adequate knowledge and control of dose and range and straggle are provided. Initial measurements have been made of range and range straggle of boron and phosphorus implanted into 100 cm silicon of the opposite conducting type in a production-type implantation system. Doses were approximately 1.5 x 1012 cm-2 and implantation energies were varied from a lower limit that gives a range of about 150 nm to an upper limit of 600 keV. After implantation, the wafers were annealed at 900°C for 30 min in argon. The results for profiles implanted with the beam tilted 7 deg from the <100> direction generally agreed with the results of calculations [29,43] based on LSS theory.

P'

The range, R and range straggle, AR were determined from measurements of the dopant profile made on reverse-biased Schottky barrier diodes by means of an automatic capacitance-voltage technique described by Gordon et al. [44]. Schottky barriers were employed because with them contact can be made to a wafer without disturbing the implanted profile. The barrier is formed by evaporating a metal film (aluminum for p-type implanted surfaces and gold for n-type) about 100 nm thick through a metal mask which defines dots varying in size from 0.1 to 0.15 mm in diameter on the front (implanted) surface of the wafer. A low resistance contact is made on the backside of the wafer by low energy implantation of a suitable dopant.

(R. G. Wilson*)

PROCESSING

6.2. Passivation Integrity

This task was undertaken to develop techniques for evaluating the integrity of passivation overcoats on metallized integrated circuits, specifically practical techniques to detect localized structural defects and to measure their population density. The techniques are intended to be suitable for routine quality control by manufacturers, to be applicable both to IC devices in wafer form and to individual pellets, and to allow estimation or quantization of the number of localized structural defects (such as pinholes and microcracks) per unit area in the oxide, glass, or nitride overcoat. Essentially nondestructive methods, which do not damage areas that are defect-free, are preferred.

7.1. Optical Imaging for Photomask Metrology

Line widths in the micrometer range are commonly measured with a microscope equipped with a micrometer eyepiece. In this measurement it is not the physical object or geometrical line width which is measured, but instead it is the magnified image of the line. Therefore, the accuracy of these measurements is strongly dependent on the quality of the line image. This quality depends on the aberrations of the imaging lens, the focus or amount of defocus present, and the spatial coherence of the light [45]. The spatial coherence is a measure of the phase correlation of the light in a plane perpendicular to the direction of propagation. The degree of correlation is a measure of the ability of the light to manifest interference effects. It ranges from a value of 1 for coherent light where the phase at each point in the perpendicular plane is correlated with the phase at any other point and interference occurs, to a value of 0 for incoherent light where the phase at any point is completely independent and interference is absent. The degree of correlation is a function of the type of illumination, the path length, and the numerical aperture of the focusing lens element in the optical system.

A theoretical investigation was begun to determine quantitatively the effects of coherence on the image of two parallel opaque lines on a transparent background and its inverse. This is the simplest statement of the problem that must be solved to determine the effects of coherence on line width measurements. The first part of this investigation considers a single line and a diffraction limited lens system. It addresses two questions:

What is the minimum size of a single opaque line on a transparent background that can be recognizably imaged as a line at any given quality level? and

How can this quality level be specified?

The first of these questions was approached by calculating the intensity profile of the image of an ideal line for both the case of spatially incoherent light and the case of spatially coherent light. Here an ideal opaque line is defined as having zero percent transmittance across its width and a 100 percent transmittance elsewhere. The intensity profiles of the image of an object are for

Expressions for these functions were derived with the assumption that the lens was diffraction-limited with an impulse response of sin x/x for the one dimensional case. The characteristic width of this function, taken to be the distance between the first node points on each side of the line x = 0, is a measure of the smallest dimension that can be resolved by the lens. The expressions for the intensity profiles for both illumination cases were derived in terms of the ratio, A, of object line width, 2b, to the width of the impulse response; the numerical aperature, NA; the wavelength of the light, A; the magnification of the lens, m; and the image coordinate, B.