March 24, 2006
Will Emerging Molecular Electronic Nanotechnologies Overcome the Impending Limitations of Current Lithographic Microtechnologies?
The growth of current lithographic CMOS technologies faces future limitations, and the field of molecular electronics currently appears to offer the most viable alternative. This paper attempts to determine whether molecular electronics will be able to deliver high-performance and cost-effective computer systems by the time the slowdown occurs. To answer this question, the paper examines the basic molecular electronic technologies, analyzes their advantages over lithographic approaches, investigates the chief challenges to molecular electronics, and looks at the projections of researchers and experts in the field.
According to many sources, the growth of current lithographic-based microelectronic technologies may reach physical and economic limits in the relatively near future. One projection estimates that the slowdown might occur in about twelve to fourteen years from the date of this paper (Murdock, 2004). It therefore appears that the continuance of Moore's Law into the decades ahead will require a radically new underlying technology.
Molecular electronics currently appears to be the most promising new area of technology for creating the next generation of computers. This decade has seen intense research in creating molecular electronic devices and in using these devices to build computer systems. Molecular electronics operates at the nanoscale (dimensions between 1 and 10 nanometers) and its devices often consist of a single molecule.
The purpose of this research project is to determine whether the field of molecular electronics is likely to provide a feasible, higher density, higher performance, and lower cost alternative to current lithographic microtechnology, and whether it is probable that molecular electronics will offer this alternative in time to avoid the erosion of Moore's Law.
Because molecular electronics is probably the most viable potential replacement for lithographic technologies, knowing whether it is likely to succeed is critical for an accurate prediction of the future validity of Moore's Law and of the future growth rate of the computer industry that is based on this law. Also, knowing what types of technologies will underlie computers in the future will help computer professionals get started on the task of redesigning the architecture of those systems to use the new technologies. If the next fundamental computer technology is in fact molecular electronics, the architectural redesigning effort will be extensive.
The widespread research efforts in molecular electronics have produced a large body of literature, which forms the basis for this research project. This literature is briefly described in the next section.
2. Review of the Literature
The molecular electronics literature used as the basis for this research project come from two primary sources:
- Databases of computing literature, primarily the ACM Digital Library, ACM Guide to Computing Literature, IEEE Computer Society Digital Library, and Computer Database
- Technical, business, and research sites on the web, such as NanoBusiness Alliance (http://nanobusiness.org), DARPATech2002 Symposium (http://www.darpa.mil/DARPATech2002/), and Foresight Nanotech Institute (http://www.foresight.org).
The research sources fall into the following main categories:
- General technical reports and tutorials on molecular electronic topics written by researchers in the field, primarily conference proceedings—for example, Wang, Qi, & Moritz, 2004; Butts, DeHon, & Goldstein, 2002; Lundstrom, 2002; Goldstein & Budiu, 2001
- Researchers' technical presentations of their research results—for example, Beckett & Jennings, 2002; Goldstein & Budiu, 2001
- Symposium addresses by leaders in the field of molecular electronics—for example, Kwok, 2002
- Reports on molecular electronics by technology journalists, including interviews with researchers and experts—for example, Murdock, 2004; Olson 2003a, 2003b; Tristram, 2003; Goering, 2005; Brown, 2004; "HP Reports," 2005; Weiss, 2005.
The next section describes how this literature was used in the research project.
Although the number of source articles available for this research project is large, each of them provides only a small piece of the research-question puzzle. The primary research task, therefore, was to survey the literature, analyze the articles, piece together the information, and draw conclusions based on the total information gleaned. The following is the methodology that was used:
- Obtain articles from the sources described in the previous section, and sort them according to relevance.
- Research the nature of the problem with current lithographic approaches and determine its extent. These results are covered in "4.1 Limitations of Current Lithographic Microtechnologies."
- Obtain a definition of molecular electronic devices and learn the basics of molecular electronics. These results are in "4.2 What are Molecular Electronic Devices and How can they Overcome the Limitations of Current Lithographic Technologies?" and throughout "4.3 Challenges to Molecular Electronic Devices."
- Research the positive features of molecular electronics and determine whether they offer significant advantages over current lithographic technologies. These results are in "4.2 What are Molecular Electronic Devices and How Can they Overcome the Limitations of Current Lithographic Technologies?"
- Research the challenges facing molecular electronics, ascertain which are the most significant, and determine the likelihood that each of them will be overcome. These results are in "4.3 Challenges to Molecular Electronic Devices."
- Draw conclusions based on the research results, as well as on the opinions of experts in the field, and attempt to answer the research question "Will molecular electronics be able to offer a superior alternative to today's lithographic-based technologies in time to preserve Moore's Law?" These results are given in "5. Conclusion."
4.1 Limitations of Current Lithographic Microtechnologies
Many of the sources for this research project refer to impending limitations on current lithographic microtechnologies (generally CMOS), and cite these limitations as the basis for the importance of molecular electronics. For example, Kwan Kwok, manager of the DARPA Molecular Electronics Program, states that the ability of lithographic technologies to maintain Moore's Law for many years has been a "remarkable achievement," but that this technology is now approaching the limits of miniaturization "at an exponential rate" (Kwok, 2002). These limitations are both physical-technical and economic.
In the physical and technical realm, the very laws of physics change as the nanoscale is approached and quantum mechanical effects begin to predominate. According to Goldstein and Budiu (2002), the specific physical problems with lithographic-based CMOS devices as they are further miniaturized include doping fluctuations, short channel effects, and the formation of oxides on ultrathin gates. To this list "HP Reports" (2005) adds tunneling effects, which cause electrons to leak through conventional silicon-based gates and result in the loss of binary information.
In discussing the economic limitations, Kwok (2002) states that "the continued miniaturization of microelectronics has become more and more costly for industry, reducing its ability to be truly innovative and responsive." Plants creating conventional lithographic-based devices now cost billions of dollars, and a new generation of these plants, including revamped chip masks and more expensive optical equipment, would be needed in order to continue miniaturization using the conventional lithographic approach (Goldstein & Budiu, 2001; Tristram, 2003).
How soon might these limitations begin to impede Moore's Law? That is hard to estimate, but according to the Semiconductor Industry Association's International Technology Roadmap on Semiconductors, the slowdown might occur in about twelve to fourteen years from the date of this paper (Murdock, 2004).
4.2 What are Molecular Electronic Devices and How Can they Overcome the Limitations of Current Lithographic Technologies?
In this paper, the term molecular electronic devices refers to computer components with dimensions in the range of 1 to 10 nanometers (10 to 100 atomic diameters), which is roughly the scale of individual molecules. In fact, some of these components, such as nanotubes, consist of a single molecule. Other general terms describing these devices include moletronics, nanoelectronics, electronic nanotechnology, and nanodevices. Another essential feature of molecular electronic devices is that, in contrast to lithographically manufactured devices, they are manufactured in batches using bottom-up chemical self-assembly, without lithographic patterning. (Beckett & Jennings, 2002)
Molecular electronic devices have an inherent density advantage compared to lithographic devices. According to Wang et al. (2004), nanotubes and nanowires (two specific types of molecular electronic devices to be described shortly) are a few nanometers in size and can be used to create feature densities of 1011 to 1012 switches or other devices per cm2. These sizes are approximately 105 times the density of current lithographic devices (Goldstein & Budiu, 2002). And even by 2016, the ITRS (the Semiconductor Industry Associationís International Technology Roadmap for Semiconductors) projects that the densities of lithographic devices will reach only about 3 x 109 transistors per cm2 (Butts et al., 2002).
Although it is inherently difficult and expensive to divide a lithographic-based chip into smaller and smaller features, as Kowk (2002) expresses it, "molecules are natural nanometer-scale structures that can be made by the trillions very inexpensively in a beaker or test tube." Thus, molecular electronic devices should also help overcome the economic limitations of lithographic devices. Manufacturing molecular electronics will not require multi-billion dollar factories, and their price will depend mainly on design and test costs (Butts et al., 2003).
Because very few electrons are used for switching, molecular electronic devices also have lower power requirements. Furthermore, they actually take advantage of quantum-mechanical effects such as tunneling, which (as mentioned in the previous section) are increasingly becoming obstacles to lithographic-scale devices. (Goldstein & Budiu, 2001)
Finally, according to Franz Kreupl, principal for molecular electronics at Infineon Technologies, molecular electronics (specifically, carbon nanotubes) also have the potential to exceed the speeds of silicon devices, due to their small diameters, high current densities, and high mobility (Goering, 2005).
Researchers are experimenting with a variety of different types of molecular electronic devices. Currently, the two basic devices that promise to become practical soonest are carbon nanotubes and nanowires, sometimes combined with organic molecules that serve as switches.
A carbon nanotube is a single molecule of C16 atoms formed in the shape of a tube that is approximately a nanometer in diameter and several millimeters long. Because a nanotube is a single molecule, it is very strong and flexible. An individual nanotube can serve as either a metal or a semiconductor, depending upon its lattice geometry (the two types are always mixed in a given batch). A nanowire is similar, although it consists of a solid wire composed of a single molecule of silicon (or one of several other elements or compounds). Nanowires have diameters of 6-20 nanometers and lengths of 1-30 microns. Nanotubes and nanowires can be used to create either interconnecting wires or active logic elements (diodes or transistors), and they can be arranged in crossbars to create arrays of programmable non-volatile switches. (Butts et al., 2002; Goldstein & Budiu, 2001)
Crossed nanotubes or nanowires may be used by themselves to form arrays of mechanical switches (a voltage draws the wires together and van der Waals forces keep them together), or certain organic molecules (such as catenane) can be trapped between the crosspoints to create molecular switches (a voltage shifts the organic molecule's structure, changing its conductivity). (Butts et al., 2002; Goldstein & Budiu, 2001)
This section has described how molecular electronic devices can potentially overcome the limitations of current lithographic technologies. The remaining large question is whether these devices will in fact provide a solution to these limitations—that is, whether the devices will become technically and financially feasible by the time traditional technologies begin to reach their limitations. The answer to this question depends upon whether researchers will be able to overcome each of the major challenges to molecular electronics in the reasonably near future. This is the topic of the next section.
4.3 Challenges to Molecular Electronic Devices
The primary challenges to molecular electronic devices may be divided into fabrication challenges and computer architectural challenges, plus several additional problem areas.
4.3.1 Fabrication Challenges
As mentioned previously, molecular electronic devices are typically manufactured in batches using bottom-up (sometimes described as hierarchical) chemical self-assembly and self-alignment, without lithographic patterning. The bottom-up molecular electronics manufacturing process, which contrasts sharply with lithographic processes, typically consists of the following main steps: First, individual nanoscale components such as nanowires are created. Then, self-assembled arrays of these components are made. The arrays are then interfaced with CMOS support circuitry. And finally, complete computer systems are built. (Goldstein & Budiu, 2001)
Researchers have used a variety of nano-component assembly techniques. For example, nanowires can be induced to assemble themselves into an array of parallel wires by suspending a batch of nanowires in a fluid and pouring the fluid over a silicon substrate. A second array of nanowires can then be placed on top of the first array, crossing the first array at right angles, by flowing another batch of nanowires at right angles to the direction of the first flow. (Butts et al., 2002)
Bottom-up chemical self-assembly, however, results in a lack of manufacturing precision. In other words, manufacturing is a stochastic process rather than a deterministic one. For example, when using the fluid method that was described for assembling nanowires, Butts et al. (2002) explain that the only parameter that can actually be controlled is the average separation between the wires (this is done by altering the rate and duration of the flow).
There are two fundamental challenges induced by the stochastic manufacturing process. First, the process can create only simple, regular molecular electronic structures, such as crossing arrays of nanowires. It cannot differentiate highly specific features within a nanoscale device, as can lithographic processes. And second, molecular electronic devices have inherently high defect and fault rates. The following two sections analyze the difficulty of these two problems.
18.104.22.168 Working with Regular Molecular Electronic Device Structures
The key to making use of a regular, relatively homogeneous molecular electronic device is to program complex functionality into the device after the assembly process—in other words, add logic after manufacture, as is done today with field programmable gate arrays (FPGAs). Fortunately, molecular electronic devices tend to be inherently reconfigurable. It therefore appears that a molecular electronic device's simple, regular structure is not a serious problem. (Goldstein & Budiu, 2001)
22.214.171.124 Overcoming the High Defect and Fault Rates of Molecular Electronic Devices
A potentially more serious problem is that molecular electronic devices will have high rates of both defects (permanent physical manufacturing flaws) and faults (transient errors occurring at runtime). The high defect rate is due directly to the stochastic assembly process. The high fault rate is caused by the sensitivity of these small devices—for example, an alpha particle can easily trigger a transient fault or even create a permanent defect. Also, because these devices function with so few electrons, the law-of-large-numbers behaviors of electrons will tend to fail—for example, there can be relatively large fluctuations in electron transfer time, causing erratic gate behavior. (Butts et al., 2002)
The key to overcoming high defect rates is basically the same as that for dealing with homogeneous structures: configuring the device after assembly. When dealing with defects, however, an initial discovery phase must be added. That is, the faults must first be found and then the device must be configured to work around the faults. Because new defects can appear over the life of the system (for example, from alpha particles), defect self-diagnosis and subsequent reconfiguration should occur repeatedly, at runtime. (Beckett & Jennings, 2002)
Not only are molecular electronic devices inherently reconfigurable, allowing the defective areas to be avoided, but also they include high levels of feature redundancy. That is, they have many more potential features (such as interconnects and switches) than are needed to create a fully functional device. The feature redundancy means that having a large number of unusable areas in a molecular electronic device will not impact its functionality. According to Rice University molecular electronics researcher James Tour, a molecular electronic device with only 5% of its interconnections working could still function correctly. (Olson, 2003b)
In the opinion of Butts et al. (2002), transient faults can be overcome by using the traditional fault-tolerance techniques described in the literature. One specific fault strategy is to provide logical redundancy at the circuit level; for example, the device can be designed to use dual-rail redundancy, in which all inputs arrive in both true and complemented form, and all outputs are generated in both true and complemented form (Wang et al., 2004).
Overall, it appears that the high defect and fault rates inherent in molecular electronic devices constitute a serious, but fully solvable problem and one that computer engineers are accustomed to dealing with. (The literature includes many sophisticated and highly mathematical papers on creating fault tolerance in molecular electronic devices.) In fact, a high defect rate and a simple, regular structure might even be indirect assets, because they require the implementation of runtime device reconfiguration, which could also be used to literally customize the processor for each specific application, thereby enhancing performance. (This type of reconfigurable computing would be based on a configuration file generated by the compiler.) (Goldstein & Budiu, 2001)
4.3.2 Computer Architectural Challenges
Assuming that basic programmable arrays of molecular electronic devices can successfully be manufactured, the next challenge is to assemble these components into working computer systems. Attempting to use a conventional stored-program von-Neumann computer architecture, in which memory is separated from processing and logic, would result in a paradox. With this traditional architecture, extremely dense, low latency molecular electronic devices, running at high clock speeds, would be connected by long, high-latency interconnections, creating an overall low-density system with communication bottlenecks. Thus, the computer architectural challenges will move from processing (which will be inexpensive) to communication (which will be expensive). (Beckett & Jennings, 2002)
Butts et al. (2002) believe that this dilemma can be overcome by creating a delay-independent design—for example, by optimizing the placement of logical functions so that frequently communicating elements are as close as possible. Beckett & Jennings (2002) think the problem can be solved by using an alternative architecture composed of locally interconnected meshes that merge both molecular electronic processing and memory, creating a finer-grained overall architecture that will be more compatible with the new finer-grained molecular electronic devices.
Not only can memory be merged with processing to minimize communication delays, but also because molecular electronic memory devices are naturally non-volatile, memory access delays can be greatly reduced by consolidating the memory hierarchy (that is, the registers, caches, DRAM, disk cache, and disk) into a single fast, non-volatile memory system tightly coupled with processing logic (Beckett & Jennings, 2002; Olson, 2003a).
4.3.3 Other Challenges
ICCAD challenges Because molecular electronic systems may contain billions of gates, designing and testing these systems with current automated design tools will be difficult. However, Butts et al. (2002) think that this problem can be solved by using new ICCAD (integrated circuit computer aided design) algorithms and optimizations. Also, as mentioned previously, compilers for the molecular electronics age might even output logic designs customized for each application, which can be configured into the molecular electronic logic at program runtime (Goldstein & Budiu, 2001).
Funding limitations Speakers at the Nanotech 2004 conference concluded that although academic research is essential for the success of molecular electronics, such research is currently underfunded by the federal government. However, at this same conference, Clayton Teague, director of the National Nanotechnology Coordination Office, reported that government initiatives should result in improved funding. (Brown, 2004)
Health dangers As if there were not enough physical and economic challenges to molecular electronics, there are even concerns that molecular electronic devices can pose health dangers—specifically, that nanotubes can accelerate cell death (Weiss, 2005). This concern is part of the general worry over health and environmental dangers posed by nanotechnology (dramatized in Michael Crichton's novel Prey).
4.3.4 Can the Challenges to Molecular Electronic Devices be Solved?
Although the computer architectural challenges will require extensive reengineering work, and the solutions to some of the final challenges mentioned (such as funding and health dangers) are uncertain, overall, the major challenges to implementing molecular electronics, while not trivial, seem solvable. In the words of Butts et al. (2002) "In many cases, the characteristics [of molecular electronic devices] are currently inferior, but these weakness do not appear fundamental."
The question of whether and how soon molecular electronics will provide a higher-performing and more cost-effective alternative to lithographic-based devices is a difficult one.
This paper has shown that molecular electronic devices have many inherent density, economic, power, and performance advantages. It has also shown that the challenges to realizing practical molecular electronics are neither deeply fundamental nor highly intractable. It therefore appears that molecular electronics will in fact be able to provide a viable alternative to lithographic technologies.
However, the question of whether molecular electronics will be able to provide this alternative in time to stave off the impending slowdown in lithographics is a more difficult one, due to the sheer number of molecular electronics problems that must still be solved and due to the fact that current advanced silicon lithographic devices have set very high standards of performance and economy that must be met and eventually exceeded. As Rice University molecular electronics researcher James Tour expresses it, "State-of-the-art silicon is so sophisticated that there would need to be a world of improvement and testing [of molecular electronic devices] before anything commercially viable could be manufactured." (Olson, 2003b)
In addition to the analyses given previously in this paper, another way to approach the molecular electronics timeframe question is to look at the time projections made over the years by researchers who have been working in this area. The projections of Paul Beckett, of the School of Electrical & Computer Engineering, RMIT University, are typical of the early optimism and estimated that the first practical devices would start to emerge from research labs in two to three years (from 2002) (Beckett & Jennings, 2002). Obviously, that has not happened.
Butts et al. (2002) estimated more conservatively that general molecular electronic systems may be economically feasible in about five to ten years (from 2002), although molecular electronic RAMs should be available sooner.
The IBM group that created a nanotube NOT gate estimated that general-purpose molecular electronics will not become economically viable for at least fifteen years (from 2004). However, they estimated that nanotube-based RAM may be available much sooner. (Murdock, 2004) David Tennenhouse, Intel Corporation's director of research, projected that components based on nanowires and nanotubes will become the dominant component type in exactly this same timeframe (Brown, 2004).
The most recent and perhaps most reliable of the estimates given here come from the molecular electronics researchers who attended the International Symposium on the Quality of Electronic Design (ISQED), held in San Jose, California in 2005. Vivek De, senior principal engineer at Intel's circuit research lab, stated that carbon nanotubes and nanowires are likely to hit the market around 2015, and that although working prototypes have already appeared, only a small portion of the required engineering details for commercial viability have been worked out. H. S. Philip Wong, professor of electrical engineering at Stanford University, was slightly less optimistic and felt that the industry will need to rely on silicon CMOS for about the next fifteen years. Robert Doering, Senior Fellow at Texas Instruments, agreed, estimating that nanotubes and nanowires will not appear commercially for about ten to fifteen years. (Goering, 2005)
Finally, Stanley Williams, director of the molecular electronics research group at Hewlett-Packard, honestly stated that he simply does not know when the commercialization of molecular electronic devices will happen. To Williams, the essence of the problem is that researchers do not yet understand the fundamental physics of why molecules switch. Although Williams admits that this understanding might take decades to acquire, he optimistically states that it might also be solved by tomorrow. As he explains, "I think we've picked the winner, something that will allow this thing we call Moore's Law to continue on for another 50 years. I used to think it was impossible. Now I think it's inevitable." (Tristram, 2003)
It is interesting to note that the projected dates for the commercial availability of molecular electronic systems made at various times during this decade have tended to move steadily into the future. This trend can be seen in the following table and graph:
|Year projection was made||Projected date for commercial availability of molecular electronics|
Judging by the most conservative of these estimates, it appears that molecular electronics might start becoming fully viable by approximately 2020. Interestingly, this is just about the time when the International Technical Roadmap on Semiconductors estimates that conventional lithographic technologies might start hitting their expected slowdown, as cited earlier in this paper (Murdock, 2004). Of course, if projected molecular electronics dates continue to move into the future, viable molecular electronic systems might not arrive quite in time to prevent a slowdown in Moore's Law.
Given the well known performance gap between processors and RAM, it is interesting and encouraging to observe that two of the sources quoted above estimate that high-performance molecular electronic RAM, the most needed computer component, may also be the first component to reach the market. This possibility means that molecular electronic memories might tend to support Moore's Law during the critical time before full-scale molecular electronic devices are feasible.
The primary limitation to this research project is the timing: It is simply too early in the evolution of molecular electronics to arrive at a fully conclusive answer to the feasibility question and especially to the timeframe question. To obtain more definitive answers, the best follow-up to this project would be to essentially repeat the same research process every few years. The results would certainly be different (as the above graph shows, the projections of experts have shifted steadily over the years). The results might even be radically different, because, as Stanley Williams stated, breakthroughs can happen at any time
In conclusion, because of the many unsolved challenges and engineering problems in the manufacture of molecular electronic devices and the building of computers based on them and because of the very high standards these devices must meet, full-scale molecular electronic technology might not become completely feasible by the time current lithographic miniaturization begins to hit its physical and economic limits. However, because of the inherent advantages of molecular electronic devices, the tractability of most of the problems these devices pose, and the intensity of research being conducted in this area, it is likely that these devices will ultimately help guarantee that Moore's Law, after perhaps faltering slightly, will continue into the decades ahead.
6. Reference List
Beckett, P. & Jennings, A. (2002). Towards nanocomputer architecture. In Australian Computer Science Communications Proceedings of the Seventh Asia-Pacific Conference on Computer Systems Architecture (pp. 141-150). Australian Computer Society, Inc., IEEE Computer Society Press.
Brown, C. (2004). Universities seen as nurturing nanotech growth -- Academe, not industry, will propel gains, Intel research chief says; asks more funds. In Electronic Engineering Times, p10. Retrieved February 20, 2006, from Computer Database via Thomson Gale: http://0-
Butts, M., DeHon, A., & Goldstein, S. C. (2002). Molecular electronics: Devices, systems and tools for gigagate, gigabit chips. In Proceedings of the 2002 IEEE/ACM International Conference on Computer-Aided Design (pp. 433-440). New York: ACM Press.
Goering, R. (2005). Nanotechnology won't help ICs for a decade or more. In Electronic Engineering Times, p1. Retrieved February 20, 2006, from Computer Database via Thomson Gale: http://0-
Goldstein, S. C. & Budiu, M. (2001). NanoFabrics: spatial computing using molecular electronics. In Proceedings of the 28th Annual International Symposium on Computer Architecture (pp. 178-191). New York: ACM Press.
HP Reports A Nanotech Breakthrough. (2005). In InformationWeek, pNA. Retrieved February 20, 2006, from Computer Database via Thomson Gale: http://0-
Kwok, K. (2002). Moletronics: Transforming nanotechnology and nanocomputers from vision to reality. Retrieved January 26, 2006 from DARPATech 2002 Symposium web site: http://www.darpa.mil/DARPATech2002/.
Lundstrom, M. (2002). Is nanoelectronics the future of microelectronics? In Proceedings of the 2002 International Symposium on Low Power Electronics and Design (pp. 172-177 ). New York: ACM Press.
Murdock, S. (2004). Electronics and information technology. Retrieved January 26, 2006, from the Nano Business web site: http://nanobusiness.org/info/aboutNano/applications/index_html/electronicsAndIT.
Olson, S. (2003a). All-molecular computer within a decade? James Marek touts the benefits of Chiropticene. Retrieved January 26, 2006, from the Geek.com web site: http://www.geek.com/news/geeknews/2003Sep/bch20030912021731.htm.
Olson, S. (2003b). Rice University researchers show molecular logic, memory. Retrieved January 26, 2006, from the Geek.com web site: http://www.geek.com/news/geeknews/2003Nov/bch20031103022495.htm.
Tristram, C. (2003). The transistor reinventing: molecules, not silicon, may be the workhorses of tomorrow's ultrafast, ultra-powerful computers. In Technology Review (Cambridge, Mass.), 106, p54(7). Retrieved February 20, 2006, from Computer Database via Thomson Gale: http://0-
Wang, T., Qi, Z., & Moritz, C. A. (2004). Architectures: Opportunities and challenges in application-tuned circuits and architectures based on nanodevices. In Proceedings of the 1st Conference on Computing Frontiers (pp. 503-511). New York: ACM Press.
Weiss, R. (2005). Nanotechnology Regulation Needed, Critics Say. In Washington Post (12/05/05) pA8. Retrieved February 15, 2006 from ACM TechNews web site: http://technews.acm.org/articles/2005-