Executive Summary

The 1991 National Research Council (NRC) report Plasma Processing of Materials: Scientific Opportunities and Technological Challenges1 included a projection that worldwide semiconductor sales would double from $50 billion in 1990 to $100 billion in 1995. In fact, total sales worldwide for semiconductors passed $140 billion during 1995, nearly triple the 1990 level.2 Companies that supply plasma equipment to the semiconductor industry have experienced similar, if not greater, rates of growth. Plasmas in one form or other are used in about 30% of all semiconductor manufacturing processing steps, and about the same fraction of processing equipment is plasma-based in a typical microelectronics fabrication facility.3 An important trend accompanying this growth in the industry is the fact that the capital cost of constructing a new microelectronics fabrication facility is similarly escalating and is now on the order of $1 billion or more.4 Estimates are that over 60% of this capital cost is for processing equipment, including plasma equipment. Processing equipment design, optimization, and control therefore take on added importance, because equipment depreciation accounts for a significant part of the price of a chip.

In spite of its high cost and technical importance, plasma equipment is still largely designed empirically, with little help from computer simulation. Plasma process control is rudimentary. Optimization of plasma reactor operation, including adjustments to deal with increasingly stringent controls on plant emissions, is performed predominantly by trial and error. There is now a strong and growing economic incentive to improve on the traditional methods of plasma reactor and process design, optimization, and control. An obvious strategy for both chip manufacturers and plasma equipment suppliers is to employ large-scale modeling and simulation. The major roadblock to further development of this promising strategy is the lack of a database for the many physical and chemical processes that occur in the plasma and especially at surfaces. Although a complete set of data for all gas phase and surface processes for all species present in the plasma is not necessary for many applications, the current lack of detailed information concerning the vast majority of processes and species is the major factor limiting the effectiveness of models.

Given the reality of inevitably limited resources, and the often considerable investments that must be made to measure and/or compute collision cross sections, reaction rate coefficients for gas phase reactions, and surface chemical rates at surfaces exposed to plasma, some priorities must be established. These priorities are discussed below, and the report's recommendations on priorities constitute one of the main results of the study.

Findings

  1. The integrated circuit (IC) manufacturing industry remains in its historical pattern of rapid technological change, and this pattern has begun to seriously challenge plasma equipment suppliers to continue the trend toward ever higher performance/cost ratios. Plasma processing tools are, in most cases, designed and optimized empirically. Real-rune control of plasma processes has not been adopted by the industry. Further improvements in performance by means of empirical adjustments will soon reach a point of diminishing returns, if they have not already.
  2. Control of processes in plasma reactors must occur on length scales that range from tens of angstroms to tens of centimeters and time scales that range from seconds to tens of hours. Loss of control at any point in this spectrum of length and time scales can result in reduced yields of components and therefore significant economic losses. For example, precise control of transistor gate and metal wiring levels across the entire chip is necessary to manufacture microprocessors at the highest speeds. Loss of this control over etching precision produces slower microprocessors and a loss of hundreds of dollars per chip. Obviously, across-wafer control is equally important to maintain high yields and therefore high profitability.


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 1
--> Executive Summary The 1991 National Research Council (NRC) report Plasma Processing of Materials: Scientific Opportunities and Technological Challenges1 included a projection that worldwide semiconductor sales would double from $50 billion in 1990 to $100 billion in 1995. In fact, total sales worldwide for semiconductors passed $140 billion during 1995, nearly triple the 1990 level.2 Companies that supply plasma equipment to the semiconductor industry have experienced similar, if not greater, rates of growth. Plasmas in one form or other are used in about 30% of all semiconductor manufacturing processing steps, and about the same fraction of processing equipment is plasma-based in a typical microelectronics fabrication facility.3 An important trend accompanying this growth in the industry is the fact that the capital cost of constructing a new microelectronics fabrication facility is similarly escalating and is now on the order of $1 billion or more.4 Estimates are that over 60% of this capital cost is for processing equipment, including plasma equipment. Processing equipment design, optimization, and control therefore take on added importance, because equipment depreciation accounts for a significant part of the price of a chip. In spite of its high cost and technical importance, plasma equipment is still largely designed empirically, with little help from computer simulation. Plasma process control is rudimentary. Optimization of plasma reactor operation, including adjustments to deal with increasingly stringent controls on plant emissions, is performed predominantly by trial and error. There is now a strong and growing economic incentive to improve on the traditional methods of plasma reactor and process design, optimization, and control. An obvious strategy for both chip manufacturers and plasma equipment suppliers is to employ large-scale modeling and simulation. The major roadblock to further development of this promising strategy is the lack of a database for the many physical and chemical processes that occur in the plasma and especially at surfaces. Although a complete set of data for all gas phase and surface processes for all species present in the plasma is not necessary for many applications, the current lack of detailed information concerning the vast majority of processes and species is the major factor limiting the effectiveness of models. Given the reality of inevitably limited resources, and the often considerable investments that must be made to measure and/or compute collision cross sections, reaction rate coefficients for gas phase reactions, and surface chemical rates at surfaces exposed to plasma, some priorities must be established. These priorities are discussed below, and the report's recommendations on priorities constitute one of the main results of the study. Findings The integrated circuit (IC) manufacturing industry remains in its historical pattern of rapid technological change, and this pattern has begun to seriously challenge plasma equipment suppliers to continue the trend toward ever higher performance/cost ratios. Plasma processing tools are, in most cases, designed and optimized empirically. Real-rune control of plasma processes has not been adopted by the industry. Further improvements in performance by means of empirical adjustments will soon reach a point of diminishing returns, if they have not already. Control of processes in plasma reactors must occur on length scales that range from tens of angstroms to tens of centimeters and time scales that range from seconds to tens of hours. Loss of control at any point in this spectrum of length and time scales can result in reduced yields of components and therefore significant economic losses. For example, precise control of transistor gate and metal wiring levels across the entire chip is necessary to manufacture microprocessors at the highest speeds. Loss of this control over etching precision produces slower microprocessors and a loss of hundreds of dollars per chip. Obviously, across-wafer control is equally important to maintain high yields and therefore high profitability.

OCR for page 1
--> Models of low-temperature, nonequilibrium plasmas, especially for the description of physical phenomena, have developed rapidly in the last 5 years. Computing power per unit cost continues to increase rapidly. However, few of the currently available plasma models can be easily used by process engineers. Although attempts have been made to model plasmas with realistic chemistries, the parameter space that can be addressed is limited. Only a handful of studies have been made that attempt to validate models of plasma processes with industrially relevant chemistries. Models that attempt to link the relevant length scales (from tool scale to feature scale to atomic scale) are just now emerging. Simulations can be no more accurate than the data and assumptions on which they are based. The lack of fundamental data for the most important chemical species is the single largest factor limiting the successful application of models to problems of industrial interest. Heterogeneous (surface) processes are at the heart of plasma materials processing technology, but are in many cases much less well understood than are gas phase processes. Numerous etching and deposition profile evolution simulations are used in industry. These simulations generally use empirically derived rate coefficients that must be refit to experimental data whenever conditions change. Experimental diagnostics and modeling of plasma-surface processes based on first principles are rudimentary and require much development. Surfaces exposed to plasmas are often strongly modified by intense ion, photon, and radical species bombardment. Therefore, not only are the chemical and physical processes themselves strongly perturbed by plasma exposure, but in addition, the surfaces upon which the processes take place are unconventional in their structure and composition. Electron collision cross section data are second only to data on heterogeneous processes in their importance to plasma processing. These data are sketchy at best for most species of interest, although some key species, such as SiH4 and CF4, have received considerable attention. Little information is available for dissociation products or for species in excited states. Recent progress in computational methods based on quantum scattering offers the possibility that the costly and time-consuming experiments may be augmented or even replaced by large-scale computation. There now exists a wealth of sensitive radiative and laser-based techniques that permit species concentration and temperature measurements in processing plasmas. All spectroscopic diagnostic techniques depend on a database of atomic and molecular parameters. A comprehensive spectroscopic database is important to enabling unambiguous identification of a particular species in the plasma or on the surface. Spectroscopic measurements are usually the first step in measuring a rate coefficient or in testing a model prediction. The spectroscopic database therefore serves a dual role to provide both qualitative and quantitative information. Although the spectroscopic database is far from complete, especially for surface species, data are available in the literature that are relevant to plasma processing. However, these data are scattered throughout the technical and scientific literature, and in some cases their accuracy is in doubt. The ready availability and ease of spectroscopic database manipulation and storage will stimulate the development of new diagnostic techniques and the wider application of existing methods. The database for ion-molecule and neutral-neutral chemistry varies considerably. For some species and reactions, the data are good. This is especially true for the cases in which there is overlap with processes occurring in the upper atmosphere or in some cases in chemical vapor deposition processes. In other cases, however, most notably for etching processes, there are few data available. Thermochemical data are sketchy for many species of interest in plasma processing. These data are important in helping to establish boundaries for reaction pathways and for estimating reaction rate coefficients. Techniques, both experimental and computational, are generally available to obtain these quantities, but few efforts are under way at present to meet these needs. Conclusions The major potential benefits of plasma modeling for chip manufacturers are better control over plasma processes, minimization of resources that would otherwise be devoted to optimizing plasma

OCR for page 1
--> processes, and possibly minimization of problems associated with undesirable emissions such as greenhouse gases and ozone depleters. The major potential benefits of plasma modeling to plasma equipment suppliers include more rapid and efficient development of new tools that meet increasingly stringent process requirements, optimization of designs before fabrication of prototype tools, and development of robust, real-time process control schemes for their tools. The main roadblock to development of plasma models that will have these industrially important uses is the lack of fundamental data on collisional, reactive processes occurring in the plasma and on walls bounding the plasma. Among the most important missing data are the identities of key chemical species and the dominant kinetic pathways that determine the concentrations and reactivities of these key species, especially for the complex gas mixtures commonly used in industry. The lack of a central location to collect, analyze, and disseminate the data that are currently available, or that will be available in the future, is a serious problem. Large numbers of materials and chemistries are used in plasma processes for integrated circuit manufacturing. Given the reality of inevitably limited resources, it is necessary to establish priorities that encourage development of relevant data for only a few of the most important chemistries, both currently and for the next 5 to 10 years. Recommendations Federal funding agencies should make greater and more systematic efforts to support development of an improved database for plasma modeling. In addition, given the direct benefits an improved database would provide for both semiconductor manufacturers and plasma equipment suppliers, organizations set up by industry to promote integrated circuit manufacturing and the semiconductor equipment industry should participate in supporting targeted database development. Greatest emphasis should be placed on surface processes because of their centrality in the technology and because this is the area in which in general the least is known. However, the other database needs outlined in the main text of this report—electron-collision processes, spectroscopic/radiative processes, ion-neutral and neutral-neutral chemistry, and thermochemistry—are also important and need considerable work if the models are to achieve their potential impact industrially. Computational approaches to providing database information, using ab initio electronic structure codes as well as semiempirical methods, should be encouraged. A spectrum of plasma models should be developed, aimed at a variety of uses. One set of codes should be developed to provide a compact, relatively fast simulation that addresses plasma and surface kinetics and is useful for process engineers. Convenient user interfaces would be important for this set of codes. A second set of codes would include more sophisticated algorithms and higher dimensionality, and would be more useful for equipment design. Development and testing of models that meet these needs should be supported. Careful validation of the codes by systematic comparison to the results of experiments needs to be undertaken. As learning and resources allow, some development effort should focus on fully three-dimensional plasma, electromagnetic, and neutral transport codes. The degree of chemical complexity to be included will vary depending on the availability of data, the goals of the modeler, and the available computational resources. The following chemistries and materials should have a high priority in database development. The panel chose these systems because the applications are currently important and are anticipated to continue to be important for at least the next 5 years and quite probably beyond that: Polycrystalline silicon etching in chlorine-containing gases and bromine-containing gases. This set (of materials and chemistries) is commonly used in the etching step to define the gate electrode in field

OCR for page 1
--> effect transistors. Control of this step is crucial in maintaining optimum device performance, and since at endpoint this step involves exposure of the increasingly thin gate dielectric, concern about damage and reliability is considerable for this step. Silicon dioxide etching in fluorocarbon-containing gases mixed with gases such as O2, H2, CO, He, and Ar. In back-end-of-the-line (BEOL) processes, one is mainly concerned with making the metallic interconnects inside a silicon dioxide insulator. The number of interconnect levels is rising, and the number of steps that involve contact (to the active device regions) and vias (from one level of metalization to the next) will increase accordingly. Concern here is with anisotropy, selectivity, and uniformity, as well as with contamination. These gases are also used to clean chambers that have been coated with dielectric films (silicon dioxide and silicon nitride) from previous steps using chemical vapor deposition (CVD). In addition, since many of the gases that are used for dielectric etching are currently of environmental concern because they are greenhouse gases (e.g. C2F6), an opportunity exists to minimize their use, remediate them as effluents, or even to replace them outright, if effective models of dielectric etching can be developed. Silicon dioxide deposition through plasma-enhanced chemical vapor deposition (PECVD), using mixtures of SiH4, N2O, and O2 or SiH4, O2, Ar, and TEOS (tetraethoxysilane). For the same reasons that oxide etching will continue to play an important role in BEOL processing as the number of interconnect levels increases, deposition of the intermetal dielectric will be a key process. At least one data center should be established to archive, evaluate, and disseminate the existing and future database for models of plasma materials processing in integrated circuit manufacturing. The archived database should include kinetic pathways, mechanisms, and comparisons of models to the results of experiments. This structure would provide a framework for iterative improvement of the database. Full advantage should be taken of emerging electronic data acquisition technology exploiting rapid access through the Internet and the World Wide Web. Although individual companies will no doubt develop proprietary databases, the goal sought with the establishment of the data center is to serve the entire community interested in plasma modeling and diagnostics. References 1. National Research Council, Plasma Processing of Materials: Scientific Opportunities and Technological Challenges (National Academy Press, Washington, D.C., 1991). 2. Semiconductor International, May 1996, p. 83. 3. National Research Council, Plasma Processing of Materials: Scientific Opportunities and Technological Challenges (National Academy Press, Washington, D.C., 1991). 4. Semiconductor Industry Association, The National Technology Roadmap for Semiconductors (SEMATECH, Austin, Tex., 1994).