photocell varies linearly with incident light. It was found that the current output varied inversely with subjective judgments of device quality made by counting pinholes and estimating crack lengths.

Based on these encouraging preliminary results, a photocellmicroscope system was assembled to take direct readings of reflected light from decorated wafers. The arrangement used is shown schematically in figure 44. By placing a plano-convex lens of 57-mm focal length at the end of the camera tube, an image of the light source field stop was obtained about 6 cm from the end of the camera tube. This image, of uniform luminance, was placed on the photocell faceplate. The photocell employed an S-11 cathode and was operated at 100 V. Light intensity was adjusted to give convenient current levels, and the overall magnification was chosen so that a large portion of the circuit was visible, but bond pads and grid lines were excluded. Shielding prevented introduction of ambient light. Current readings were taken on 20 circuits of each of two wafers. For the "good" wafer, the lowest individual current was 2.9 x 10-9 A, the highest was 5.5 x 10-9 A, and the average was 3.5 x 10-9 A. For the "bad" wafer, the lowest current was 6.7 x 10-9 A, the -8 highest was 1.5 x 10-8 A, and the average was 1.0 x 10-A. Pinhole counts on these wafers resulted in a difference of a factor of two in defect density. Thus, the photocell current measurement can easily detect a difference of a factor of two in pinhole density and is probably more sensitive than this.

4.6

Advantages of the Reverse Decoration

Carbon Black Method

In this section the relative advantages of the carbon black reverse decoration technique are summarized. Based on our results, the carbon black method is recommended for defect detection in most cases. In the discussion several unique advantages of the corona-charging technique as used in the reverse decoration will be pointed out. The use of the high-voltage discharge, which initially appears to violate the normal rules for treatment of devices (particularly, MOS type), will be seen to be a definite advantage and to render the process actually nondestructive.

4.6.1 Speed

In the standardized process to be described later, the sample is charged for 18 s, and after a 10-s delay, is immersed in the carbon black suspension for 6 or 12 s. Thus, the total decoration process time is, at most, about 40 s. We do not include in this time the recommended 5 min, 200°C bake since this can be done in batch. This process time of 40 s may be compared with the 5 min needed for metal etching, the several minutes needed for electrolytic techniques [14,15] or for the conventional electrophoretic method previously described here.

4.6.2 Simplicity

The process is very simple and does not require precise positioning or delicate manipulations, as is sometimes the case in, for

Figure 44.

Schematic of microscope use for obtaining image of source field stop at photocell for reflected light measurements. The sample is both illuminated by the light source and observed by the photocell from the same side.

example, the liquid crystal technique [8]. The sample is placed on a platform for charging and then dipped into the carbon black suspension. The process thus shares the simplicity of chemical etching, with the added relaxation that the time of immersion in the liquid is constant for a wide range of samples, and does not depend on sample parameters as in the etching method, where the aluminum thickness must be considered.

4.6.3 Permanence of Decoration

Liquid crystal and bubble formation techniques are real-time methods in that the decoration exists only during the process implementation. A photograph must be made for later examination. The carbon black decorated sample, however, has a permanent decoration and it may be set aside for future examination. This feature of permanence is an important one which the process shares with etching, electrophoretic, and copper decoration methods.

4.6.4 Nondestructive Nature

Unlike the etching or electrolytic copper techniques, the carbon black decoration can be removed. Tests by wafer probing showed no device loss after cleaning, as described previously. In methods which make use of voltages applied between electrodes and the sample in an electrolyte, as used in copper decoration for example, there is the possibility of sample damage by high currents at dielectric weak spots. The corona-charging process in a dry ambient (RH 30%) is inherently nondestructive since corona potentials are well below sparking voltage and the corona has a very high impedance. Furthermore, any dielectric weak spot will discharge only the surface charge immediately surrounding it, and the stored energy is not great enough to cause any dielectric damage.

4.6.5 Avoidance of Junction Effects

Decoration methods using electrolytes or other low-resistivity liquids, such as alcohol or acetone, are limited by device junction effects. For example, a dielectric defect over a reverse-biased junction may not be detected or the deposition rate of decorating material may be slow. This effect can be overcome or moderated by illuminating the sample to induce photocurrents in reverse-biased junctions, but this may be difficult in some cases. For example, in the case of a large sample, as a wafer immersed in a decorating suspension of light-scattering particles, insufficient light may reach certain portions of the sample. On the other hand, the corona current impedance is so great that even reverse-biased junctions are sufficiently conducting that they do not interfere with charging effects.

4.6.6 High Contrast

The reverse decoration with carbon black provides a very high contrast when imaged in a microscope in reflection. This high contrast contributes to rapid evaluation of the relative quality of samples by

the observer and, in addition, permits automated assessment by use of a photocell attachment to the microscope.

4.6.7 Ease of Observation

The reverse decoration process leaves the defect uncovered for easy examination by optical or electron microscopy. All other known processes result in direct decoration and the defect is covered or altered in some way.

4.6.8 Limitation

The only fundamental limitation encountered thus far has been devices with regions of metal not connected to the semiconductor. floating metal is brought to a high potential by the corona so that the metal is outlined with carbon black, but no defects can be detected over it. Thus, floating gates on some memory devices would not decorate properly if the device were examined at a stage of processing before the overlying metal is deposited. The other regions of the sample could be successfully decorated, however. After the device is completed and the second metal covers the floating gate, then it is possible, of course, to decorate the device in the standard manner.

5. ADVANTAGES AND DISADVANTAGES OF THE METHODS INVESTIGATED

In this section a comparison of the five processes investigated under this contract is presented in tabular form. The comparison is made on the basis of simplicity of procedure, process time, permanence of the decoration (adhesion), destructiveness or nondestructiveness, susceptibility to reverse-biased junction effects, contrast in microscopic observation, sensitivity of detection, ability to detect partial or latent defects and to obtain a depth profile of partial defects, and ease of microscopic observation and limitations. For each of these characteristics, the five processes are compared in table 6. The comparison assumes that typical, aluminum-metallized IC wafers with PSG passivation overcoats are being tested.