3. Page 12 25 4. 5. Selective Demarcation Etching of Commonly Used Metal Films Demonstration of Sequential Metal-Glass Etching Method 6. Advantages and Disadvantages of Methods Investigated 32 34 42 83 "Techniques for Measuring the Integrity of by Werner Kern and Robert B. Comizzoli Conventional test methods to evaluate the quality of glass passivation overcoats on semiconductor devices are generally inadequate and/or destructive. Three new methods have been devised that overcome these problems: (1) Sequential selective chemical etching of metal/dielectric structures to detect buried, latent, or partial defects as a function of dielectric layer depth. (2) Electrophoretic cell decoration with uv phosphor particles suspended in an insulating liquid, the sample forming one electrode of the cell. (3) Electrostatic corona charging to selectively deposit surface ions from a high voltage dc discharge on the insulating surfaces of the sample, followed by placing of the charged sample in a suspension of charged carbon black particles in an insulating liquid; depending on the polarity of the ions the particles can be deposited on the insulator surface or at the defect sites. The etching method is most suitable in process research studies, and the electrophoretic technique for demarcating relatively large defects. The corona decoration method, coupled with automated instrumental read-out based on measuring the reflected light intensity, is ideal for routine testing of devices because it is fast, simple sensitive, and nondestructive to devices such as glass passivated bipolar and MOS ICs. The practical benefits of the new test methods can be considerable in production and product control, with cost savings through early detection of production line defects and rapid corrective action. Key words: Corona charging decoration; dielectric defect detection; electrophoretic decoration; integrated circuit quality control; selective chemical etching; and semiconductor device reliability. "This research funded by the Advanced Research Projects Agency Order 2397 through the National Bureau of Standards' Semiconductor Technology Program Contract (5-35913). views and conclusions expressed are those of the authors and do not necessarily represent the official policies of the Department of Defense, Department of Commerce, or the United States Government. This publication is not subject to copyright. 1. SUMMARY Most integrated circuits now on the market are metallized with aluminum and passivated with an overcoat of phosphosilicate glass (PSG) and/or silicon dioxide. The quality of present-day overcoats is highly variable, indicating that more effective quality control measures are needed to improve the product and its reliability. However, test methods available for evaluating the quality of glass passivation overcoats have been inadequate and/or destructive. The objective of this research program has been the development of analytical test methods that do not suffer from these shortcomings. This goal has been accomplished successfully. We have demonstrated the applicability of the new methods to the specific requirements of this contract, as well as to the evaluation of insulating coatings in general. The experimental approaches chosen to solve these problems were based on the following basic methods: (1) Selective chemical etching, intended primarily as an absolute standard for comparing other test methods; (2) electrophoretic cell decoration with uv phosphor particles; and (3) electrostatic corona charging to deposit surface charge followed by decoration. Techniques for quantifying the defect density were examined in conjunction with these detection and decoration methods. The major results and conclusions for each are summarized below. The method of selective metal etching underneath the glass overcoat to demarcate localized structural defects in the overcoat has been demonstrated to be simple, fast, effective, and sensitive. It is an absolute method but has the disadvantage of being destructive. It is used primarily when the sample can be sacrificed in the test, or when an absolute standard is desired for comparison with other techniques. Microcracks in typical glass layers over aluminum can be detected down to widths of less than 500 A. Sequential etching of metal/insulator structures is useful for detecting buried, latent, or partial defects within the dielectric overcoat. Determination of the defect density after each pair of etching treatments then provides the defect density as a function of dielectric or insulator layer depth. The method is reliable but time-consuming; it is recommended for applications in process research. Numerous examples are presented, illustrating specific and general applications of the two selective etching methods. In the electrophoretic method, the sample to be decorated is placed in the suspension, and a voltage is applied between it and an opposite electrode to move the decorating particles to the defects. A rectangular stainless-steel tank is used to hold the decorating suspension. The tank is also used as one electrode. In some cases, a glass beaker is used and a stainless-steel electrode is placed opposite the sample. This method was found suitable for decoration of relatively large defects and cracks with white or uv phosphor powders. The use of insulating liquids was found to give superior results in terms of adhesion and nondestructiveness. The deposition mechanism, basically a capacitor-charging phenomenon, was investigated in some detail. This technique is useful but less attractive than the corona-charging methods. In the corona-charging method of decoration, surface ions from a corona discharge are deposited on the insulating surfaces of the sample. At a defect in the insulator, ions flow to the grounded substrate. After the charging step, the sample is placed in a suspension of charged decorating particles in an insulating liquid. If the electrical polarities of the ions on the sample insulating surfaces and of the charged decorating particles are opposite, then the particles are deposited on the insulating surfaces and not at the defects, resulting in reverse defect decoration. If the electrical polarities of ions and particles are the same, then the particles are deposited on the defects, resulting in direct decoration. For the corona charging, a plane array of 40-um-diameter wires is used, spaced 2 cm from the sample, which rests on a grounded plate. For certain samples, it is necessary to place a grounded grid over the sample to limit the surface voltage of the passivating layer during the corona charging. The principles of the corona-charging mechanism were studied and related to the process; in particular, the basis for the nondestructive nature of the process is well understood. In general, the corona-charging techniques are nondestructive, rapid, and simple. Various decorating powders were tested including silicate glasses, uv phosphors, and carbon black. The corona-charging methods are very sensitive; e.g., the carbon black reverse decoration method can detect defects not found by aluminum etching, nor by scanning electron microscopy (SEM), unless glass etching is first used to enlarge the defect. By comparison with etching, this technique has been shown to be very selective. These techniques are also capable of detecting certain partial or latent defects. The carbon black reverse decoration by corona charging results in a very high contrast sample when viewed by reflectance microscopy. This is a particular advantage for process automation and was explored using a photocell mounted on the microscope to measure reflected light intensity, which was related to sample quality. The population density of decorated defects can thus be quantified. It is possible to automate this type of read-out technique on a step-and-repeat basis, using automatic and computerized instrumental recording of the data. Post-decoration device recovery procedures have been developed and proven to be effective. It has been shown that device yield is not decreased by this procedure. A good correlation has been shown to exist among the various methods developed on this contract. Advantages, disadvantages, limitations, sensitivity limits, and applicability of each have been pointed out and demonstrated experimentally in many instances. We The described methods were developed and refined specifically for the evaluation of dielectric overcoats on aluminum-metallized ICs. have shown, however, that these methods are also applicable to analyzing other metal/dielectric structures and devices and, in fact, to evaluating insulator coatings in general. The availability of a well-defined practical test method for evaluating the integrity of IC passivating overcoats now offers device manufacturers a much needed tool for controlling their products during fabrication. Since the recommended techniques are nondestructive, a 100 percent quality control is feasible at the device wafer level. will be possible to reprocess defective wafers at this point rather than continue their processing, thus eliminating very large potential losses that would be caused by completing defective materials into finished IC devices. Even more important, these expedient, sensitive, and nondestructive test methods allow rapid information feedback of the test results to the production line and, thus, immediate correction of faulty processing conditions. In addition to being an early, rapid detection system for defects occurring on the production line, the new test methods are a valuable tool for assessing developmental studies for improving materials and processes. The methods developed and perfected during this program also make it possible to test finished IC devices by the manufacturer (or by the procurement agency or individual customer), and allow a reasonable degree of presently nonexisting standardization of the integrity of overcoat passivation layers on finished IC products. In summary, the practical benefits of the new test methods are very considerable, when applied to production and product control, in terms of both cost savings due to early detection of production line defects and rapid information feedback for corrective action. Batch removal and batch reprocessing of defective material in wafer form will be an additional cost-saving factor. |