In ultralarge-scale integrated (ULSI) semiconductor fabrication, plasma processing plays a vital role in (1) plasma etching, (2) plasma-assisted chemical vapor deposition (PECVD), and (3) physical vapor deposition (PVD). In the plasma etching area, there is a very active development of high-density plasma (HDP) sources. This work is driven primarily by the need to operate at lower pressure to reduce the feature size dependence of the etch rate, improve profile control, reduce particulate formation, reduce residues and sidewall passivation layers, and reduce surface damage by controlling the ion and electron energies. Each of these needs engenders a large array of interesting problems suitable for basic studies. The optimization and characterization of plasma sources themselves are currently under intense study and will probably continue to be subjects of interest for several years.
Although all the U.S. tool vendors are actively engaged in HDP tool development, none of the vendors is pursuing electron-cyclotron resonance (ECR) technology for etching applications. At this time this is basically a business decision, since in stringent tests Japanese ECR etching technology has been performing very well. ECR technology is viewed as a Japanese technology, with Hitachi commanding an insurmountable lead in the Japanese market. Since the Japanese semiconductor tool market is essential for U.S. tool vendors, technologies are pursued in the United States that, in their judgment, offer the potential of being superior to ECR technology, for example, the transformer-coupled plasma technology of Lam Research and Applied Materials and the helicon technology of PMT and Lucas Laboratories. Given this situation, it is unlikely that a major effort in ECR technology aimed at manufacturing would greatly benefit any of the U.S. tool vendors.
For semiconductor manufacturing technology, the feature size dependence of the etch rate and the difficulty of decoupling profile control from the pattern density are probably the most important concerns at this time. Directional etching is achieved by sidewall passivation. The amount of sidewall passivation depends on the amount of etch product and mask area, and it changes dramatically as one moves from isolated features to densely populated portions of the integrated circuit. The amount of sidewall passivation material determines the profile of the structure and thus important device parameters like the channel length of a field-effect transistor. Controlling this pattern sensitivity has become a major manufacturing issue and a challenge to plasma processing scientists.
The formation of particulates from the reactant gases is a serious problem in plasma processing because the sheath fields of the plasma trap the particles and release them onto the wafer as contaminants when the plasma is turned off. The problem of dusty plasmas, which also is of concern in space research, is already under intense study by groups at IBM and the Universities of Arizona and Iowa, but there is still a need for innovative ideas. Another key problem that has often been noted in HDP technology is charging-induced damage in these high-current devices. Although several possible mechanisms have been identified, a model that explains all the experimental observations is lacking at this time.
Numerous applications and challenges exist in the plasma-assisted growth area. Important applications that are just starting to be explored are plasma-assisted chemical vapor deposition of metals (e.g., A1), PECVD of insulators, and PECVD of photoresist materials. An immense
potential exists for the all-dry processing of photoresist materials. In PECVD there is still support for ECR technology in the United States; for instance, Lam Research offers ECR technology for the deposition of silicon dioxide.
Plasma-based surface cleaning may also become an enabling technology for cluster processing in the semiconductor industry. At this time most of the surface cleaning steps are still being performed using traditional wet methods. These methods are incompatible with the goal of complete vacuum cluster processing and with environmental considerations. A satisfactory dry-cleaning method will probably, more than any other unit-process step, change the way semiconductors are being processed.
The greatest research opportunities, however, are to be found in the field of optical diagnostics. There are three general stages in the production of microelectronic circuits in which diagnostics are important. These stages, listed without priority, are (1) development and characterization of precompetitive materials and processes, (2) comparative analysis and characterization of tools and processes in development, and (3) sensor development for process control and fingerprinting of manufacturing processes. Diagnostic techniques are central to research on semiconductor processing, and significant advances can be made in these areas.
A ROLE FOR NRL
NRL has demonstrated the necessary expertise and equipment for state-of-the art analytical measurement of chemically reactive species in plasma processes. This capability may be used to augment industrial-support programs and federally funded initiatives and for internal NRL and Department of Defense (DOD) missions. Application of this resource to the Dual Use concept is practical and of considerable benefit to NRL and U.S. industry. This section presents an outline of several avenues by which NRL can benefit U.S. competitiveness by partnering with the domestic microelectronics industry and outside programs. In formulating this discussion, the panel adopted a broad-minded, university and industrial perspective based on its members' experience, current topics in the scientific literature and professional conferences, and the announced initiatives of federally funded agencies. These suggestions do not address current defense needs; NRL can provide the best insight into this important aspect of the program. Instead, the panel hopes that NRL can mesh DOD applications and programs with the programs suggested below to provide maximum synergy and vitality for NRL and the U.S. technology base.
The capabilities and facilities provided by NRL are well aligned with the three areas listed above and provide opportunities for NRL to have an impact on U.S. industry. Activities in these technology areas also provide opportunities for NRL and outside researchers to interact and exchange ideas. A robust program of this kind will jumpstart NRL's initiative in the industrial sector by rapidly providing first-hand knowledge of and a perspective on industrial needs and problems. In addition to the technical applications detailed below, personnel programs are also needed to obtain maximum benefit for DOD and industry. One such personnel exchange program is suggested in Chapter 9 .
Development and Characterization of Precompetitive Materials and Processes
The continuing trend toward smaller and faster devices and modular packaging of devices drives a need for devices with ever-smaller critical dimensions and for producing films with low or high dielectric constants and materials With special properties, such as boron nitride and diamond films, optoelectronic components, and Si/III-V alloys. Existing instrumentation and personnel at NRL are well suited for these tasks. NRL researchers could greatly assist industry and outside consortia by providing expertise in materials characterization and in situ diagnostic measurements, especially in ventures directed toward developing new
materials and evaluating material properties for electronic technologies. Current NRL capabilities of interest include the following:
Optical emission and absorption spectroscopy (vacuum ultraviolet [VUV], ultraviolet, visible, infrared);
Attenuated total internal reflection;
Fourier transform infrared absorption spectroscopy;
Infrared diode laser absorption;
Laser-induced fluorescence (LIF) and multiphoton variations of LIF;
Resonance-enhanced multiphoton ionization spectroscopy;
High-resolution electron energy loss spectroscopy;
Low-energy electron diffraction spectroscopy;
Film interference measurements;
X-ray photoelectron spectroscopy;
Auger electron spectroscopy; and
Coherent anti-Stokes Raman scattering.
This extensive diagnostic capability places NRL in a unique position for partnership with industry and technology consortia. It also distinguishes NRL from university programs that may be more limited in equipment availability and interdisciplinary collaboration.
Programs in this arena could include analysis of film properties for microelectronic technologies and the effects of ion bombardment and radiation damage thereon, development of surface structures and novel materials, and scientific issues such as reaction mechanisms and avoidance of process problems during plasma processing. Some topics in this arena are issues relating to the following:
The mechanism of diamond film deposition;
Organometallic vapor phase epitaxy;
Stress-free dielectric films;
Influence on etch and deposition uniformity in large-area plasma processes;
Particle formation and transport during plasma processing;
Reactive ion etching and VUV damage;
Generation and consequences of ionizing radiation in plasma processing; and
Electrical defects caused by plasma processing.
A number of federally funded initiatives have been implemented recently to encourage research along these lines. Some of these programs are directed toward semiconductor technology and some involve computer display technology. In both areas, improvements in plasma processing are often cited as a national technological need. Often, these initiatives require or encourage collaboration between diverse partners. NRL should continue to compete in these programs and can further its competitive stance by partnering with industry and consortia involving university and federally funded laboratories.
The precompetitive posture of programs in this arena provides an uncomplicated opportunity for association with university groups, other national laboratories, and industrial partners. Publication of results should be unencumbered by proprietary concerns. The success of programs in this area will ensure future collaborations with other outside partners as the capabilities of NRL become more widely known through publications and presentations to workers in the microelectronics field. NRL should continue to be represented at semiconductor conferences and increase its dialog with that community. NRL could serve the national interest by offering its facilities and expertise in areas for which much of industry would be reluctant to expend resources because of the lack of direct (short-term) competitive benefit of research.
Nanolithography and plasma process techniques are also needed to fabricate advanced technology components. However, the high
cost of developing and optimizing fabrication technology and the proprietary concerns typical for this aspect of technology development suggest that NRL's role is best suited for exploratory and characterization programs rather than for dedicated or “stand-alone” process development. In-house processing facilities at NRL are convenient and timely for building one-of-a-kind test structures. These facilities should be maintained for this purpose.
Comparative Analysis and Characterization of Tools and Processes in Development
Industrial research and development workers are often faced with a selection of competing tools and processes needed to fabricate a product in a planned program. Proper selection of the best fabrication equipment is needed for cost-effective manufacturing and to compete in the global marketplace. All too often, selection of tools and processes is driven by scheduling deadlines long before any reasonable understanding of the various technologies is available. In addition, as industry strives to reduce product costs, research and development investment invariably suffers, thereby creating vulnerability to errors in tool and process strategy with possible long-lasting effects.
An alternate approach to the current one involving redundant evaluations of the various technologies by each company is to rely on a centralized facility or program for objective, side-by-side comparison. This offers clear advantages in cost-efficiency and thoroughness. Industry partners will still require some independent evaluation for their proprietary concerns; even at a fundamental level of study, however, centralized evaluation of new tools and processes benefits all parties. Results of these studies should also be published and presented at conferences. By highlighting concerns in the open literature about newly developed tools and processes, such a program would serve to accelerate optimization and debugging of new technologies. Often, industrial laboratories capable of this type of work are restrained in their interaction with tool suppliers because of proprietary interests and concern about possible inference of future plans by their competitors.
An opportunity exists for NRL to contribute to the nation's technology base by providing domestic industry with an informed evaluation of the global advantages and disadvantages of various tools and processes. This evaluation should be maintained at a scientific level, rather than advocating certain technologies. This stance is also needed to maintain objectivity and credibility with domestic industry. NRL can serve as an industry-neutral resource for evaluating technologies and tools under consideration for near-term manufacturing. The diagnostic capabilities outlined above also serve well in this program arena. A role of this kind would take several years to develop, as NRL researchers become more attuned to industry and technology needs and as their customers come to trust NRL's contribution and objectivity.
One example of this comprehensive program would be comparative evaluation of several high-density-plasma tools currently being considered for 0.35-mm (and beyond) generation technology (64-MB DRAM). A host of enhanced plasma tools and processes is now in development, including helical wave and helicon resonance reactors, magnetically enhanced reactive ion etchers, ECR plasmas, and inductively coupled plasmas. Yet, to date there has been no side-by-side comparison of all of these tools and only limited side-by-side evaluation of some tools for some process applications.
Serious questions still persist regarding fundamental aspects of these high-density-plasma tools, with important implications for the domestic microelectronics industry. Which tool offers the lowest ion temperatures? The answer affects the ultimate directionality of ion-enhanced etching. Which high-density tools can be scaled up for 200-mm and even 300-mm wafers with acceptable uniformity? Does mode formation dominate in some tools with a consequent influence on center-to-edge uniformity? How serious is ion damage for
circuit components in the various tools? How do particles form and move in the different high-density-plasma tools? Since each tool has an independent plasma source and a radio-frequency-biased wafer holder, can in situ cleaning processes be developed to strip unwanted film deposits from tool walls, including those films requiring ion bombardment for removal? Which tool is most prone to film deposits and thus will require greater maintenance activities? What is the influence of cross-chamber chemical contamination on clustered wafer processing? Does plasma generation of VUV light or x-rays cause defects or alter the photoresist in partially developed wafers? These questions are best answered by comparative evaluation of the competing tools. Answers to these fundamental issues would offer significant advantage to the U.S. microelectronics industry. NRL's objective and professional evaluation of some of these issues would be greatly valued by industry and would also attract resources from outside funding agencies.
The obvious drawback to this program is the need for multiple, advanced processing tools. These tools generally cost more than $1 million and final design tools are more suited for manufacturing applications than for research studies. These problems pose serious obstacles to the proposed program. However, it must be recognized that general scientific issues are addressed in this program arena, rather than a dedicated process development effort. Because of this, laboratory prototype tools are preferred over the manufacturing versions. In addition, laboratory prototype designs are genuinely practical simulations of the commercial tools. Repeatedly, it has been shown that generic problems identified in laboratory-type reactors are also observed in final product, commercial reactors. The use of laboratory-type mock-up reactors for the various high-density-plasma tools provides a cost-effective and acceptable alternative to the purchase of multiple commercial reactors. This approach has the endorsement of several tool vendors and would open opportunities for cooperative efforts with NRL researchers in designing lower-cost mock-up tools.
Sensor Development for Control and Fingerprinting of Manufacturing Processes
It is rapidly becoming evident that robust manufacturing processes for advanced technologies require real-time process control for cost-containment and product assurance. Properly used, sensors maintain process windows and correct for natural variances, minor impurities, and aging effects in tools. Sensors may even be used to highlight and correct for deficiencies of previous process steps and human errors. By maintaining an extensive database of sensor input, one can use archival analysis of product results to infer second-order processing effects and to optimize for imprecise fabrication line concerns such as tool maintenance and long-term drift.
NRL's experience in sensor development and in diagnostic techniques is well suited for working with outside partners and industry to develop suitable sensors for semiconductor manufacturing. Sensors are required not only for detection of gas-phase reactive species and to measure surface conditions, but also for tool inputs, such as feed gas, impurities, and radio frequency power—including arcing and spikes. Modeling may be used to help evaluate the sensitivity of processes to variations in sensor measurements of the process, thereby providing a first-order estimate of tolerances and required sensitivity of the sensor.
Sensor development can become a commercial enterprise, and some small companies dedicated to this product line have been started. However, these small operations often lack the equipment and expertise to develop the sophisticated sensors needed for reliable process control. Larger companies would probably use commercially available sensors but, in most cases, will not devote the personnel and resources to developing sensors if they are not likely to be company products. Thus, a void is formed between the
entrepreneurial small companies interested in manufacturing sensors (but lacking resources to develop them) and their customers who wish to purchase rather than develop sensors. Researchers at NRL should be encouraged to work with small companies interested in manufacturing sensors and with larger companies likely to use sensors. In this way, NRL would again provide a centralized resource to domestic industry and would be furthering its mission toward Dual Use.
One example of note, useful for future planning of activities in this arena, is the recently completed and highly successful Microelectronics Manufacturing Science and Technology (MMST) program funded by the Advanced Research Projects Agency (ARPA) and cooperatively executed between Wright-Patterson Air Force Base and Texas Instruments. This program resulted in the development and testing of a wide range of process control sensors, many of which involved optical diagnostic measurements in plasma processes. Ellipsometry, a conventional technique sensitive to thin film structures, but previously considered too slow for real-time monitoring, was greatly expanded and improved on in the MMST program, eventually developing into a real-time sensor for manufacturing tools. Follow-on programs are now being proposed by ARPA. The sensors developed in the MMST have already been licensed by other companies, thereby completing the technology infusion mechanism proposed by the MMST program planners and endorsed by ARPA.
NRL can participate in or initiate similar programs, using its expertise in sensors, optical diagnostic techniques, and in-house tools. Interservice programs involving cooperative efforts with Air Force and Army laboratories are also encouraged and would offer additional expertise and insight into sensor development.