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 5
--> 1 Industrial Perspectives Introduction The main purpose of this report is to highlight the opportunities for more rapid and effective development of plasma process modeling and simulation as an engineering tool in the semiconductor manufacturing and plasma equipment supplier industries. It is intended for a variety of audiences: academic and government laboratory researchers, industrial engineers and scientists, and program managers at federal funding agencies. The scope of the report is substantial, coveting the industrial needs for better plasma process engineering, the current state of the art in plasma modeling, and the various supporting databases and diagnostic techniques that underlie and complement modeling and simulation. The report begins with this chapter on industrial perspectives to emphasize the primary purpose of this activity: to serve the needs of industrial suppliers and users of plasma process equipment. The need to maintain this industrial perspective is a recurring theme of this report. The potential for using modeling and simulation to benefit industrial users of plasma processes and equipment has never been greater. Computational costs continue to decrease steadily, and in the last several years, considerable progress has been achieved in establishing the major modeling strategies that are necessary to achieve practical industrial objectives. Nevertheless, low-temperature plasma processing science is a relatively young field, and has not therefore received the in-depth, sustained attention that is required to have a significant, timely impact in industry. This situation is perhaps most evident in the area of the database for physical and chemical processes in plasma materials processing. The data that are currently available are often scattered throughout the scientific literature, and assessments of their reliability are usually unavailable. The goals of this report include identifying strategies to add data to the existing database, to improve access to the database, and to assess the reliability of the available data. In addition to identifying the most important needs, this report assesses the experimental and theoretical/computational techniques that can be used, or must be developed, in order to begin to satisfy these needs. A major complication in this process is the fact that industry uses a large variety of gases and materials in plasma processes and equipment. Since time and resources are always limited, one must make choices regarding which chemical systems to examine carefully. Experiments are expensive and time-consuming, and therefore it may be necessary to augment these measurements of fundamental data with theory. Computational techniques are useful, but may require careful testing since the methodologies are sometimes not fully mature. The panel has attempted to develop a compromise between the competing needs for breadth and depth in the database, recognizing that needs change as industry evolves. The recommendations presented here will therefore require updating, probably within 3 to 5 years of the publication of this report. Plasma Processing for Semiconductor Manufacturing The semiconductor industry and related industries such as flat panel display manufacturing are growing rapidly. In the semiconductor industry, worldwide revenues grew from $50 billion to $140 billion from 1990 to 1995.1 Projections are that a similar rate of growth will persist at least through the end of the decade, and probably longer. Plasma processing is one of the key technologies in this industry, accounting for ˜ 30% of all process equipment in a typical wafer fabrication facility and a similar percentage of process steps. These include etching, deposition, cleaning, and stripping. The current level of annual sales for plasma equipment is on the order of $3 billion to $4 billion, with increases projected to match the rate
OCR for page 6
--> of growth of the industries it serves.2 Projections are that by the end of the decade, sales of equipment for plasma etching, cleaning, physical vapor deposition (PVD), and chemical vapor deposition (CVD) will reach $10 billion.3 An important trend in the industry is the rapidly escalating cost of capital equipment, and the growing share of equipment costs (now about 60%) in the total cost of a state-of-the-art fabrication facility (currently about $1.5 billion).4 The projected changes in integrated circuit technology anticipated in the 1994 roadmap of the Semiconductor Industry Association (SIA) are given in Table 1.1. The historic trend of a doubling in device density every 1.5 years or so will continue. New generations of technology are introduced every 3 years, fueled by information age demands that continue to explode. Notably, the introduction of 0.25 µm technology overlaps the introduction of the next jump in wafer size to 300 mm diameter in 1999. Part of the reason for the exceptional rate of growth in the semiconductor industry has been the dramatic and steady rise in performance per unit cost. In order that this trend in the performance/cost ratio continue, the next-generation developments listed above must be accomplished with a corresponding increase in manufacturing efficiency. TABLE 1.1 Changes in Silicon Integrated Circuit Technology Projected to 1999 1990 1993 1996 1999 Critical dimension (µm) 0.8 0.5 0.35 0.25 DRAM capacity (Mbits)a 4 16 64 256 MPU/logic clock speed (MHz)b 25-40 50-200 140-350 240-500 Wafer size (mm) 150 150-200 200 200-300 Defect density (no./cm2) 0.2 0.1 0.05 0.03 Interconnect levels 2-3 34 4-5 5-6 a DRAM = dynamic random access memory. b MPU = microprocessor unit. SOURCE: A. Voshchenkov, Workshop on Database Needs in Plasma Processing, Washington, D.C., April 1-2, 1995. Increasing manufacturing efficiency will require a significant increase in the sophistication and effectiveness of process control, among other changes. For example, as critical dimension (CD) decreases, the control of the CD must be to within 0.03 µm. Even though the wafer diameter will increase, etching rate and selectivity nonuniformity across the wafer must in some cases be kept to less than 2%. During etching, an important control variable is the angle of the microfeature sidewall with respect to the surface. Control of this profile angle is sought to within 3 degrees. Another important processing variable in MOSFET (metal oxide semiconductor field effect transistor) manufacturing is related to the thin (<< 80 Å) gate oxide between the gate electrode and the active device region below it. During the etching step in which the gate electrode is defined, and when gate oxide may be exposed to the plasma, selectivity must be high enough to keep gate oxide loss to less than 15 Å. This is only 4 to 5 atomic layers. From the industrial perspective, plasma processing, especially plasma etching, is often seen as being unusually difficult to understand and control.5 Although some of the general mechanisms of the plasma are known—such as the role of chemical interactions with radicals such as F or C1 atoms, the role of sidewall passivation in preserving etch anisotropy, and the fact that positive ion bombardment of surfaces has a mechanical sputtering role—many details remain obscure. Interactions between plasma species and the walls bounding the discharge are complex and depend on surface temperature, surface and bulk composition, and other variables that are empirically observed to change with time, but are not well understood. The goals of plasma etching, including high rate, selectivity, uniformity, minimal damage to the underlying nascent electrical devices, minimal chemical residue contamination, minimal particulate deposition, and microfeature critical dimension control, sometimes depend in subtle ways on the plasma quantities. Moreover, these processing objectives are commonly difficult to achieve simultaneously. For example, there is often a conflict between etch anisotropy (enhanced by more energetic ion bombardment)
OCR for page 7
--> and selectivity (degraded by more energetic ion bombardment, since most solid materials sputter at about the same rate). Measuring fundamental plasma quantities is challenging, especially in industrial plasma tools with limited access for diagnostic instruments and rudimentary sensors. Most often, one must develop empirical correlations between observed process characteristics such as etching rate, uniformity, and selectivity, and design and operating variables such as chamber shape and materials, applied power, gas pressure, composition of the inlet gases, and substrate temperature. There are many of these parameters in a plasma etching tool. The tendency to view plasma etching as '' an art rather than a science, more black magic than engineering,''6 can be attributed largely to the murky relationship between the output of a plasma tool and its operating and design characteristics. In order to develop a manufacturable plasma process, engineers must carefully balance the competing process objectives, selecting process operating conditions (and tool type) so that the trade-offs are reasonably satisfied. In addition, this optimization must occur in a region of parameter space that is as large as possible. That is, if some set of operating conditions yields acceptable process results, but a very small change in gas pressure or wall temperature (for example) results in unacceptable process characteristics, the "operating window" is too narrow and the process will be ill suited to an industrial operation. Finding the right combination of operating conditions usually involves a tedious search of a large parameter space, unaided by theory or computation and guided only by intuition and/or experience. Partly in response to the challenges of etching high-aspect-ratio features, plasma equipment manufacturers have introduced in the last several years new plasma tools capable of operating at relatively low gas pressures, usually between several millitorr and several tens of millitorr. It is thought that the lower operating pressures improve etched feature profiles—for example, anisotropy—by minimizing collisions between bombarding ions and neutral molecules in the sheaths. However, lower operating pressures tend to reduce etching rates, lowering wafer throughput and increasing the cost of operating the equipment (part of the cost of ownership). In order to raise rates in spite of the reduced gas pressure, the newer tools have been designed to sustain a higher plasma charged-particle number density—typically on the order of 1011 to 1012 cm-3. Perhaps more importantly, these newer tools (high density, low pressure) have at least partly separated power deposition into electrons to maintain plasma density (the source power) from power deposition into ions bombarding the substrate (the bias power). It should be noted that in addition to decreasing the frequency of ion-neutral collisions in the sheath and therefore increasing ion velocity anisotropy at the processed surface, these newer tools operate with considerably different ratios of ion flux to neutral flux bombarding the substrate, and with different neutral species impacting the surfaces. The latter effect is due to higher plasma densities—both neutral and ionic species in the high-density plasmas tend to be more highly dissociated, no doubt affecting the plasma chemistry. For example, fluorocarbon gases commonly used in dielectric film etching are considerably more dissociated into atomic F than is the case in more conventional tools. This has caused problems with selectivity between silicon dioxide and silicon and has necessitated other changes in tool operation (or design) to counter this effect. Some observers have noted that the chemistry in high-density plasmas, originally thought to be simpler and "cleaner" than in conventional tools, is perhaps simply different, still requiting careful balancing to meet all objectives. Other effects are observed in high-density tools that are not observed in conventional tools. For example, in high-density plasmas, the flux of positive ions to the walls, and the return flux of neutral species, represent an internal recirculating mass flow of a magnitude similar to that of the gas flow entering and leaving the chamber due to pumping. The effects of this internal recirculation on processing are not known at present. In sum, because of the relatively new conditions experienced in high-density plasma tools, it is fair to say that modifications are needed to some of the established wisdom gleaned from years of experience with conditions of lower plasma density and higher gas pressure. The combination of rapidly changing process objectives (see, e.g., Table 1.1) and a rapidly changing technology (e.g. the shift to low-pressure, high-density plasma tools) compounds the challenges for
OCR for page 8
--> manufacturers and users of plasma processing equipment. Plasma processes have been successful in meeting the processing needs of many aspects of the semiconductor manufacturing industry. However, to keep up with the projected acceleration in the industry, plasma equipment design, optimization, and control must become more efficient. Plasma equipment suppliers that are able to meet the accelerating demands will win market share over their competitors both domestically and abroad. Chip manufacturers better able to optimize, maintain, and control their plasma tools will enjoy an advantage in productivity and therefore profitability. This advantage will in turn translate into more capable and affordable electronics for the civilian and defense sectors. In the rest of this chapter the panel outlines the opportunities that modeling and simulation offer to chip manufacturers and to plasma equipment suppliers to meet these goals. Plasma Equipment Supplier Perspectives Equipment suppliers design plasma process chambers, with the associated pumping, gas handling, wafer handling, and control software and hardware. In addition, plasma equipment suppliers develop process chemistries to meet the needs of their chip manufacturing customers, often in a collaborative effort. As silicon technology evolves toward the gigabit era in the next millennium, the requirements listed in Table 1.1 will place increased demands on the design of both new chambers and process chemistries. The choice of plasma tool design, process gas chemistries, and operating conditions (the operating window) must satisfy these criteria over a reasonable range of conditions. A major concern is the rune it takes to design a new plasma process chamber and to select the appropriate process chemistries and operating conditions. Plasma-induced contamination and damage are always a concern and must be kept to an absolute minimum, if not eliminated altogether. The major problems with plasma processing technology from the point of view of the tool supplier can be summarized as difficulties in tool design, tool optimization, and tool control. Historically, and currently, plasma process equipment has been designed and optimized largely empirically. Designers have relied on experience, intuition, and estimation to develop the next generation of tools. Process control strategies have lagged behind the standards seen in other industries. Mostly, control has focused on simply maintaining constant mass flow of reactants, constant pressure in the chamber, constant wafer temperature, and constant radio-frequency power to the discharge. The wafer-to-wafer and batch-to-batch drifts that are well known to all users of plasma process equipment have not been addressed, or they have been dealt with by more frequent (and costly) chamber cleaning and wall conditioning. The general consensus is that the traditional method of meeting the requirements listed above, namely empirical trial and error, is encountering a point of diminishing returns. One view of the overlapping roles that modeling and simulation can play is illustrated in Figure 1.1. The "modeling design engine" includes virtual prototyping of equipment, real-time process control, and process design. Virtual prototyping is the use of modeling to determine the location and size of gas inlet and pumping ports, the design of the vacuum pumping layout, the electrode and electromagnetic power coupling configurations, and the wafer clamping mechanism, among other equipment components, before constructing a prototype chamber (that is, before " cutting metal"). Process design involves the selection of process chemistries and operating conditions (e.g. pressure, power, gas flow rate) that will provide the desired processing characteristics (e.g. uniformity, rate, anisotropy in etching, film properties in deposition). Real-time process control involves the ability to make appropriate measurements of the important processing variables (or to infer these quantifies from other, more convenient measurements) and to adjust operating conditions to remain within the desired set-points. Modeling and simulation can play an important role in process control in several ways. For example, it is not always clear what operating variables (or Figure 1.1 Overlapping functions of modeling: the "modeling design engine." (Courtesy of A. Voshchenkov, Lam Research Corporation.)
OCR for page 9
--> Figure 1.2 Coupled, comprehensive models. Databases and diagnostics are important at all levels. (Courtesy of A. Voshchenkov, Lam Research Corporation.) combinations) will be most effective in controlling the process. The current approach involves the use of empirical correlation techniques such as response surface methodology, but these have limited validity outside the conditions under which the data were obtained. Also, model-based process control offers the opportunity to incorporate what is known about the process dynamics into the process control scheme. A set of integrated comprehensive models that are envisioned by plasma tool suppliers7 is shown in Figure 1.2. This set includes physical models to predict transport and electromagnetic phenomena; chemical models for both gas phase and surface chemistry; integrated system models to enable predicting what is happening at the wafer surface in terms of macroscopic phenomena at the tool scale; and finally empirical sensor-based models for real-time, adaptive process control. Process diagnostics play a key role in all of these models, from validation of the physical and chemical models to helping to identify and develop the appropriate sensors for process control applications. The models envisioned in Figure 1.2 are termed "comprehensive," but it is recognized that hardware and chemistry process models need not capture every detail in order to be useful. Convenient user interfaces, and the development of compatible modeling modules that can be integrated into evolving simulation codes, are the keys to exploiting current advances without limiting future developments as they become available. Databases are the subject of other chapters in this report, and as illustrated in Figure 1.2, they play an important role in all of the subsets of the comprehensive model. Chip Manufacturer Perspectives The semiconductor industry has established the necessity of plasma processing for both deposition and etching of materials. These applications include highly anisotropic etching of deep trenches in silicon, highly selective etching of polysilicon gates on very thin gate oxides, and blanket photoresist removal. Plasma deposition has been applied to metals, to barrier layers between films that might otherwise interdiffuse and/or react, and to insulators such as silicon dioxide and silicon nitride. In addition, plasmas are often used to clean surfaces of debris and other unwanted materials. From the point of view of the chip manufacturer, the development of plasma processes has been chiefly an experiment-driven procedure, and this has carried the industry to working with submicron features.
OCR for page 10
--> Advances in very large scale integration (VLSI) have brought the industry to the point that small differences in process precision can have a large economic impact. An example is apparent in the area of microprocessor manufacturing. After fabrication, microprocessors today are sorted based on the maximum speed at which they will operate. This sorting is determined primarily by the precision in etching the gate and metal wiring levels. Relatively small losses in the precision of these etching steps result in slower devices. (It should be noted that it can be difficult to separate the roles of lithography and etching in this loss of precision. Sometimes it is possible to compensate for problems in lithography with alterations in the etching step.) The sales price of a single chip can differ by hundreds of dollars between the fastest speed category and a slower one. Control of processes on the feature length scale, across the entire wafer, must be maintained over time scales that can range from seconds to many hours. During etching of a single wafer, short transients that might last on the order of seconds during the start or end-point of an etch can influence processing characteristics. In addition, variations on time scales of tens of minutes due to, for example, relatively slow changes in tool wall temperature, can be important. On even longer time scales, wall surface deposits slowly build up and must eventually be removed in a cleaning step. These are some of the major forces driving the industry to better understand and control plasma processes. The industry is looking to modeling and simulation to help gain this understanding. Another area of concern for chip manufacturers is in the so-called back end of the line (BEOL) where the metal interconnects are formed. As noted in Table 1.1, the number of these interconnect levels is increasing. At the same time, the defect density must be reduced. However, the danger of introducing device-threatening contamination and/or damage during these steps increases as the number of processing steps increases with the number of levels of metalization. Problems such as this have prompted interest in contamination-free manufacturing (CFM). Plasmas are acknowledged generators of particles, both from chamber walls and from processes occurring within the plasma. Particles become charged in the plasma and are often trapped above the wafer. An important opportunity exists for plasma modeling and simulation to contribute to better understanding and eventual control of particles in plasma process equipment. Improvements in plasma tool design and operation are needed to minimize particle nucleation, growth, and eventual deposition on the wafer. Models for plasma processes today exist largely at two levels, based on the two major length scales in the technology: the microfeature level (~ 0.1-10 µm) and the tool level (~ 1-100 cm). However, between the microfeature level and the tool level is a mid-scale level that involves patterns of features on wafers, the edge region of the wafer, and so on. This mid-scale level is on the order of 0.1 mm to 1 cm, and to date it has received less attention from the modeling and simulation community. However, issues associated with a dependence of etching characteristics on the local pattern (such as the density of lines) are of considerable practical importance to the chip manufacturer. Connections between the mid-level scale and the tool scale must be developed in order to address these concerns. In addition, many important phenomena occur on the molecular scale (~ 1-100 Å). Modeling and simulation are needed for developing processes that produce the desired uniform results. There is a clear need to minimize the time it takes to develop a process for a new pattern, minimizing the feature size and "adjacency" problems noted above. Cost is increasingly important. Tool productivity in terms of cost per operation is key to competitiveness. Reduction of the number and complexity of experiments through simulation is increasingly attractive since engineering experiments have become very costly. Achieving these efficiencies requires valid models, access to model data, and calibration techniques. In order to involve plasma process engineers in the use of plasma equipment models, it is important that such models be relatively fast and easy to use. Multidimensional simulations tend to be relatively time-consuming (a few to several tens of hours), even on fast workstations. This is probably too slow for the process engineer. In addition, user interfaces are now relatively crude. One option, therefore, is to develop less comprehensive but faster running models that are coupled to an etch profile simulator. With a convenient user interface, this combination of relatively simple models might be a useful aid to the plasma
OCR for page 11
--> process engineer. Acceptance of such a simulation package would set the stage for the use of more complex models by industrial engineers. Recommended Priorities for Development of an Improved Database Although there are many applications of plasmas to a wide variety of processes as noted above, it is appropriate to attempt to establish several applications that, from an industrial point of view, seem especially promising for modeling. These applications are listed in Table 1.2. The first application is polysilicon gate etching in chlorine- and bromine-containing gases. The second is SiO2 etching in hydrofluorocarbon-containing gases (i.e. CxFyHz), with various other gases added such as O2 and N2. The third is a common plasma-enhanced chemical vapor deposition (PECVD) application, SiO2 deposition. The primary gases for this application include SiH4, O2, N2O; SiH4, O2, Ar; and tetraethoxysilane (TEOS). Note that this list is not all-inclusive in that in some cases other gases are used in addition to the major gases listed in Table 1.2. TABLE 1.2 Recommended Priorities for Developing an Improved Database Application Gases Poly-Si etching Cls2, Br2, HBr, O2, N2 SiO2 etching CF4, CHF3, C2F6, O2, N2, CO, Ar SiO2 deposition SiH4, O2, N2O; SiH4, O2, Ar; TEOS The rationale for these choices is that all three are major applications with widespread interest in the industry, and they all appear to be applications that will persist for at least the next 3 to 5 years. Gate electrode etching is a key in controlling the effective channel length for complementary metal oxide semiconductor (CMOS) devices, and therefore plays a major role in the sorting of microprocessor speeds described above. This has a direct impact on chip profitability, and is in need of close attention as a result. Dielectric etching (mainly contacts and vias) is crucial because of the increasing aspect ratios (3-4 and above), coupled with a high degree of selectivity between the oxide and the silicon (50:1 is desired but difficult to achieve). Dielectric etching is also the largest segment of the plasma etching market. For PECVD, simultaneous deposition and etching offers the opportunity to fill gaps between metal interconnect lines. As the number of metalization levels increases, this application will become more important. Other PECVD oxide deposition applications include planarization layers. The list of recommended chemical systems and associated applications in Table 1.2 does not in itself constitute a "database." For each of these systems, the database consists of a choice of chemical species to include in the overall mechanism, the key reactive pathways by which the selected set of chemical species are created and destroyed, and, in addition, rate expressions and parameters describing the nature of the interactions between these selected species and with surfaces. These needs are discussed in greater detail in subsequent chapters. In this chapter, the emphasis is on identifying the general chemical systems that are related to a selected set of common industrial processes (i.e. those listed in Table 1.2), rather than on prioritizing individual collisional and/or reactive processes. It should be noted that, in addition to the recommended high-priority chemical systems listed in Table 1.2, there are many other chemical systems that are of interest in various applications of plasma processing in integrated circuit manufacture. These include physical vapor deposition techniques such as sputtering and reactive sputtering; conventional metal etching (Al/Cu alloys); photoresist stripping/ashing; plasma-enhanced chemical vapor deposition of a variety of materials; compound semiconductor etching; and emerging applications involving etching of ferroelectric materials and noble metals. These are all important applications of plasma processes, and it is likely that new applications will emerge in the future. Plasma modeling and simulation have the potential to significantly improve these applications as well as the ones listed as recommended priorities. Although the systems listed in Table 1.2 were judged to be the prime
OCR for page 12
--> candidates for database development support, it is certainly true that other applications are important and deserve support as well. Findings The integrated circuit manufacturing industry remains in its historical pattern of rapid technological change, coupled with impressive rates of growth. This pattern has begun to seriously challenge plasma equipment suppliers to continue the trend toward ever higher performance/cost ratios. Rapidly changing plasma process requirements, increasingly stringent process control requirements, and pressure to decrease the time to market for both new semiconductor products and new plasma equipment all contribute to the demand for a more effective approach to plasma process engineering. Meeting these more stringent requirements will make the U.S. semiconductor manufacturing industry more competitive and responsive to both civilian and defense markets. Plasma processing tools are, in most cases, designed and optimized empirically. Real-time control of plasma processes is limited to individual subsystem loops for variables such as flow or pressure. Feedback control of important process parameters such as etching rate, uniformity, and selectivity 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. Conclusions Plasma modeling and simulation can develop into a powerful scientific and engineering tool, but a number of obstacles are limiting the pace of progress. It is generally agreed that the primary obstacle is a lack of a suitable database for the many physical and chemical processes that make up a plasma process. This includes not only data characterizing individual collisional processes, but also the selection of the key chemical species, and the reaction pathways for these species. The most promising applications on which to focus for developing or improving the database are poly-Si etching, SiO2 etching, and SiO2 deposition. These applications are currently important and appear likely to remain important for the next 3 to 5 years at least. The remainder of this report is intended to sharpen the focus on the existing state of the art in plasma modeling and simulation, and the database that supports it. While the treatment is not exhaustive, each subsequent chapter aims to summarize the most important current issues and to point the way to the most fruitful directions for improvement. References 1. Semiconductor International, May 1996, p. 83. 2. A. Ghanbari, Workshop on Database Needs in Plasma Processing, Washington, D.C., April 1-2, 1995. 3. Ibid. 4. Semiconductor Industry Association, The National Technology Roadmap for Semiconductors (SEMATECH, Austin, Tex., 1994). 5. P Singer, "New Frontiers in Plasma Etching," Semiconductor International 19(8):152 (July 1996). 6. Ibid. 7. A. Voshchenkov, Workshop on Database Needs in Plasma Processing, Washington, D.C., April 1-2, 1995.
Representative terms from entire chapter: