Strategic Plan Nanotechnology and Advanced Computing Technology March 21, 1996 I. Vision As foreseen by physicist Richard Feynman, miniaturization has had revolutionary impact in the Apollo program, in computer technology, and in spacecraft design. As manufacturing systems increasingly enter the nanometer (10^-9 meters) range, the need for computational simulations to understand the underlying physical phenomena is acute. We see Ames Research Center at the center of a nationwide network of research laboratories using computation to understand and ultimately control manufacturing processes at smaller and smaller sizes, culminating in atomically precise products. In the medium term, significant advances can be expected in computing technology as Ames works with Silicon Valley firms, computer vendors and universities to remove the barriers that may impede computer technology progress during the next few years. In the long term, petaflops computer systems, atomically precise manufacturing techniques, and biologically-inspired devices enabled by Ames and its partners will revolutionize space launch systems, climate modeling, spacecraft instrumentation, aeronautical design, planetary exploration and computer manufacturing, with significant spin-offs for other sectors of the U.S. economy as well. The NAS program at ARC has a proven ability to operate state-of-the-art parallel supercomputer systems, and has substantial expertise in parallel computing methods and scientific visualization. The Ames computational chemistry branch has world-class expertise in simulating physical interactions at the atomic scale. The Ames biocomputing branch is exploring biological approaches to future computing and instrumentation technology. Other organizations at ARC have additional expertise that can be tapped as well. This unique combination of talent and resources will be leveraged to create a world-class leader in computer simulations of ultra-small systems. One example of an important future NASA application of this technology is the development of autonomous intelligent vehicles for the exploration of the solar system and beyond. Consider for a moment the fact that radio signals require up to 40 minutes for a round trip from Mars to Earth. Thus we cannot rely solely on ground-based personnel or computers to control Mars-bound spacecraft or Mars-roving robots. An intelligent on-board computer is required, with sufficient power to insure reliable real-time operation. For missions to other planets or to the stars, even more powerful computer systems will be required. These systems must also be exceedingly compact in size, must use very little electrical power, must be highly resistant to radiation damage, and, most importantly, must be affordable in an increasingly tight budgetary atmosphere. Future spacecraft will need more than exceedingly powerful and compact computer systems. They will also need advanced scientific instruments, including image sensors, microwave receivers, light spectrometers, mass spectrometers, chemical analyzers, interferometers and others, which all must be more compact, reliable and economical than is feasible with current technology. Further, such instruments may require novel designs, such as designs inspired from biological systems. At the same time, it is recognized that the transportation costs of placing humans and equipment into space must be reduced by at least two or three orders of magnitude if large-scale human exploration and development of space is to begin. Thus there is a need for much stronger, lighter and much more affordable materials with which to construct space vehicles. One of the more promising future approaches is to employ molecular nanotechnology to construct diamondoid materials, which could potentially feature several times the payload to launch weight ratio achievable with titanium. "Smart material" technology, which can instantly adapt to changing environments, is another intriguing possibility. These more futuristic technologies will likely not be feasible and affordable for at least two or three decades, but given their potential they are worth looking into now. In the arena of computer technology, there are several important future NASA missions that will crucially rely on extra-high performance computer systems. These include climate modeling, processing of earth-observing satellite data, space vehicle engineering, aeronautical design, and astrophysics. In each of these areas, there are worthwhile applications that could utilize systems capable of sustained performance rates approaching one teraflops (also written as 1 Tflop/s, i.e. 10^12 floating-point operations per second). More sophisticated versions of some of these applications, which scientists hope to be able to do by the end of the next decade, are projected to require upwards of one petaflops (or 1 Pflop/s, i.e. 10^15 floating-point operations per second). It should also be emphasized that at each stage in the history of computing, valuable and often unanticipated applications of scientific supercomputers have emerged, and there is no reason not to presume that this will also be true for the systems of the next decade. Achieving the extraordinary performance levels projected for future missions, with the requisite main memory, mass storage and system software, will require breakthroughs in device technology, novel architectures, and advanced software concepts and programming facilities. Effectively utilizing such systems on NASA missions will require innovative algorithms and clever implementation techniques. Unfortunately, it is by no means certain that the U.S. computer industry by itself will be able to deliver such products in the desired time frames and at acceptable costs. For example, the semiconductor industry is facing daunting technical challenges to maintain its current rapid pace of progress. Further, due to changes in the computing marketplace, computer vendors are now emphasizing low- and mid-level systems rather than the high end required for the applications mentioned above. Commercial vendors cannot be expected to develop the algorithms, parallel techniques and software tools required for implementing NASA applications on these systems. Thus, increasingly, the responsibility for sponsoring the necessary long-range research leading to these leading-edge systems lies with NASA and its partner agencies in high performance computing. II. Objectives In light of these challenges, NASA Ames has recently instituted a new program in nanotechnology and advanced computing technologies. The overriding long-term objective of this effort is to insure that NASA will be able to meet its future requirements in the arenas of high performance computing, advanced scientific instruments, and advanced materials and manufacturing technology. A. Petaflops Computing The petaflops computing activity is exploring designs of computer systems capable of performing at a rate of 10^15 floating-point operations per second by the 2010 time frame. Issues to be addressed include device components, architectures, system software, algorithms and applications. Applications to the individual NASA enterprises include the following: * Aeronautics -- aeronautical design optimization has traditionally required the most powerful computer systems available, and this is expected to be true well into the future. * Space Science -- accurate astrophysics simulations, such as simulations of galaxies colliding, have very large computing requirements. * Mission to Planet Earth -- future high resolution climate modeling and earth-observing satellite data processing have huge computational requirements, both in raw compute power and also in mass storage and data communication. * Technology development -- engineering simulations of future space structures and space vehicles will require powerful computers; petaflops computing technology is also expected to facilitate the development of ultra-compact computers needed for unmanned space exploration. * Human space flight -- potential applications include remote intelligent robotic assistants, telemedicine systems and remote information storage and retrieval systems. B. Semiconductor Technology The second component of this plan is an activity to enhance progress in semiconductor technology by using highly parallel computer systems to simulate semiconductor devices, in partnership with researchers at area universities and corporate research labs. It is well known that the semiconductor industry is facing daunting challenges in the years ahead if it is to maintain its present rapid pace of advancement in small feature sizes, large capacity memory devices and high performance processors. In spite of the obvious need for highly parallel computing as a simulation and design tool in this arena, heretofore there has been relatively little utilization of parallel systems in the field. Among the principal impediments are lack of access to parallel testbeds, lack of software running on such platforms, and lack of expertise in parallel computing. There are of course numerous potential benefits to NASA of accelerated semiconductor technology. The petaflops computers mentioned in item B above will simply not be feasible in the desired time frame (2010) without one or more significant and fundamental breakthroughs in semiconductor technology. In addition, advanced semiconductor technology will be required for future spacecraft instruments and on-board computers for planetary exploration missions. Indeed, advances in semiconductor technology will have substantial benefits for all NASA enterprises, and will also benefit the private sector as well as the nation converts to an information economy. C. Computational Nanotechnology Looking forward further into the future, Ames is actively pursuing a program in computational nanotechnology. In this activity we are investigating, by means of computational simulations on high performance computer systems, the future production of components and materials by molecular manufacturing. Efforts are required in the areas of diamondoid synthesis, self-assembly of molecular structures, ab initio simulation, molecular mechanics, meso-scale simulation, long time scale simulation, finite element simulation, visualization techniques, and the software infrastructure required for effective use of these tools. Two long term goals have been proposed: (1) the detailed design and simulation of nanotechnology replicators, and (2) the development of a molecular manufacturing computer aided design system. Some potential benefits in the NASA enterprises are expected to include: * Aeronautics -- extremely strong, light materials with excellent thermal properties (e.g., diamondoid), and smart materials for aerospace vehicles (e.g., materials consisting of large numbers of microscopic mechanical devices). * Space Science -- ultra-compact instruments for space exploration. * Mission to Planet Earth -- ultra-fast computer components for climate simulation and satellite data storage systems. * Technology Development -- atomically precise aerospace components, extremely powerful computers (approx. 10^18 MIPS desktop computers) and extremely high capacity data storage media (approx. 10^15 bytes per cm^2). * Human Space Flight -- dramatically improved earth to orbit systems, using the materials and techniques described under the aeronautics enterprise. It should be emphasized that at the present time it is by no means certain when atomically precise molecular manufacturing will be possible. Also, the potential applications listed above are still somewhat speculative. Nonetheless, enough progress has been made, particularly in recent laboratory demonstrations of atomic manipulation, that molecular nanotechnology appears to be worth serious investigation by NASA at this time. D. Biological Nanotechnology Related to the computational nanotechnology activity, which focuses on molecular nanotechnology (sometimes called "dry nanotechnology"), is an activity in biological nanotechnology (also called "neuro-nanotechnology" or "wet nanotechnology"). This activity involves the computer-aided development of sensors and computer devices based on biological models. Recent successes in the field, such as the development of an artificial retina, suggest that there is potential for some valuable NASA applications in this arena. The long-term objective here is abstracting principles from biological systems to produce computer devices, sensors and intelligent computer systems that replicate human analytical power and versatility. Some potential benefits in the NASA enterprises are expected to include: * Aeronautics -- smart control devices for advanced aircraft. * Space Science -- sensors and scientific instruments based on biological designs. * Mission to Planet Earth -- sensitive instruments for the remote measurement of environmental conditions. * Technology Development -- computer systems inspired from neurons or other biological structures that are significantly more flexible than conventional computers. * Human Space Flight -- biologically-inspired sensors, control systems and computer systems will provide intelligent devices essential for this endeavor. III. Approach Our goal is to establish ARC as a national leader in the fields of petaflops computing, semiconductor device modeling and computational nanotechnology. This effort spans six ARC organizations: the NAS division (code IN), the artificial intelligence division (code IC), the computational chemistry branch (code STC), the biocomputing branch (code SLR), the advanced computational methods branch (code ADC) and the reacting flow environments branch (code STA). Most of the effort will be conducted or sponsored by codes IN, STC and SLR. Some additional details of plans for this program will be provided in plans to be drawn up by these individual organizations. However, a majority of the actual research work will be performed by scientists at university and industrial research laboratories, under grants, cooperative agreements and computer time allocations managed by scientists at NASA ARC. A. Petaflops computing The petaflops computing effort is an umbrella for a number of NASA research activities addressing the issues of future high end computer systems, with the objective of realizing systems capable of computing at a rate of one petaflops by the 2010 time frame. Research is being targeted in five areas: device technology, architectures, system software, algorithms and applications. The NASA petaflops effort will be conducted as part of the High Performance Computing Research (HPCR) program, a large inter-agency initiative now being organized under auspices of the HPCC program to pursue research and development aimed at future high end systems. Code IN will be the focal point of this activity at ARC, which will encompass all five of the major areas mentioned above. IN personnel are already serving on high level government committees and task forces planning activities in this area. Other IN scientists are expected to do studies in architectures, system software, algorithms and applications. A typical study might be to analyze a particular future application that is projected to require petaflops-level computing, and to determine which of various proposed petaflops system designs would be most effective for this application. It is expected that the majority of the actual research work in this activity will be performed by IN scientists, although some work will be done through research grants. B. Semiconductor technology Given its location in the heart of Silicon Valley, Ames is clearly well-positioned to make a positive contribution to semiconductor device technology. One arena where Ames can make a particular valuable contribution is in using highly parallel computers in the analysis, design and validation of advanced semiconductor devices. As mentioned above, the semiconductor industry and research community have heretofore made relatively little use of parallel computing technology, for a number of reasons. In addition to powerful highly parallel supercomputer systems, IN has substantial expertise in parallel computing techniques, scientific visualization and in 3-D physical modeling. While code IN will be the focal point of this activity, a number of other scientists at Ames are doing related work and will be working in conjunction with the NAS researchers. These include Peter Goorjian of code ADC, who is working on simulations of photonic devices, and David Olynick of code STA, who has applied techniques of aerothermodynamics to simulate the head dissipation of semiconductor devices. Ames researchers have initiated some promising collaborations with outside researchers in the field, including Prof. Robert Dutton at Stanford University and Dr. Vijay Naik at IBM's Yorktown Heights laboratory. NAS has organized a workshop to be held in March on semiconductor device modeling. The speakers scheduled for this workshop include leading experts of the field in academia, private industry (especially Silicon Valley firms) and government laboratories. It is further expected that this workshop will lead to additional productive collaborative relationships. It is expected that a large percentage of the actual research work in this activity will be performed by non-NASA scientists, under grants, cooperative agreements or computer time allocations managed by ARC scientists. C. Computational Nanotechnology ARC is well positioned to become a leader in computational nanotechnology. NAS has leading-edge parallel supercomputers (the IBM SP-2, for example, is presently the most powerful supercomputer on the West Coast). Ames also has world-class expertise in parallel computing and scientific visualization (code IN) and in computational chemistry (STC). While this activity is being coordinated by scientists in IN, Ames researchers in codes STC are a primary focus. It is expected that other scientists at Ames will also participate. For example, it is expected that scientists in the nanotechnology program will interact with scientists in code IC to develop "smart" tools for the high level control of molecular assembly processes. Ames researchers will also work closely with molecular nanotechnology researchers at university and industrial laboratories in the local area and elsewhere. Such collaborations are essential, not only to leverage on our in-house expertise, but also to validate our simulation results with physical experiments in the laboratory. Promising collaborations are already underway with Prof. Calvin Quate's group at Stanford, Dr. Stan Williams' group at Hewlett-Packard Laboratories, Dr. Ralph Merkle of the Xerox Palo Alto Research Center, Dr. Brenner at NCSU, Prof. William Goddard's group at Cal. Tech, and Prof. Todd Wipke's group at UC Santa Cruz. It is expected that more than 50% of the actual research work in this activity will be performed by non-NASA scientists, under grants, cooperative agreements or computer time allocations managed by ARC scientists. D. Biological Nanotechnology ARC is also well positioned to become a leader in computational nanotechnology. As before, NAS has leading-edge parallel supercomputers, as well as expertise in parallel computing and scientific visualization. The active program in biological computation (SLR) is the base for this activity. In addition to collaborations with scientists in IN and STC, it is expected that other scientists at Ames will also participate. For example, Tim Barth of code ADC has worked with scientists in the biocomputing branch on some computational modeling problems. Ames researchers will also work closely with neuronal-based nanotechnology researchers at university and industrial laboratories in the local area and elsewhere. Collaborations already exist with Gordon Shepherd, a premier neuroscientist at Yale University, and with Aviv Bergman and Marcus Feldman of Stanford University, who are leaders in the field of genetic algorithms. A framework for future interaction with the outside community has been established through a steering committee in Neurotechnology. The committee consists of Christof Koch (CalTech), Tomaso Poggio (M.I.T.), Gordon Shepherd (Yale University), and Michael Merzenich (U.C.S.F.). All are world-renowned. It is expected that nearly all of the actual research work in this activity will be performed by non-NASA scientists, under grants managed by ARC scientists. IV. Research Plan Overall, we expect that the majority of the research work described in this plan will be performed outside NASA, either by grants and cooperative agreements monitored by NASA scientists, or else by scientists sponsored by their own organizations or other government grants, but with computational support provided by NAS. We do not plan at the present time to do any significant experimental work or produce any materials, computer devices or commercial-grade software. Rather the focus will be on basic research work, the results of which will be transferred, through our publications and prototype computer programs, to computer and software vendors, aerospace firms and other NASA centers for subsequent usage in NASA missions. All work will be placed on the NAS world wide web site to facilitate rapid dissemination. It should be noted that the above activities constitute a balanced program in technology research, with an emphasis on working with premier area universities and Silicon Valley firms. Mid-term efforts (such as semiconductor device modeling) are balanced with long-term efforts (petaflops computing and molecular nanotechnology). Further, progress in the mid-term arenas, such as improved semiconductor devices, will promote greater progress in the long-term arenas. The success of this effort will be measured by (1) attendance at Ames-sponsored workshops in the field, (2) numbers and quality of research papers published by researchers under Ames sponsorship or collaboration, and (3) the ultimate adoption by U.S. computer vendors, aerospace firms, commercial manufacturing companies and other NASA centers of techniques and prototype software developed in this program. Significant successes to date include two workshops. The first workshop, entitled "From Neurons to Nanotechnology", which was held at Ames October 18-19, 1995, included leaders in neuro-nanotechnology. Papers were presented on neuroscience, neuromorphing (abstracting from neurons to chips and control systems), small autonomous robots, and on potential use of biological materials (DNA and bacteriorhodopsin) in computer devices. The result of this workshop was a report sent to Headquarters that outlined the next steps to be taken to implement a program in Biological Nanotechnology. A steering committee was appointed to oversee further development of the program. See the URL http://biocomp.arc.nasa.gov for more details on the workshop The second workshop, entitled "Computational Molecular Nanotechnology", which was held at Ames March 4-5, 1996, included many of the leaders in the molecular nanotechnology field. Papers were presented on self-assembly processes, molecular computer memory designs, numerical techniques, nano-robot designs and others. The workshop helped NASA scientists better understand the best role for NASA to play in this arena, and it also enabled them to form and strengthen collaborative relationships with other researchers in the field. See the URL http://www.nas.nasa.gov/NAS/Projects/nanotechnology/workshop/ for more details on the workshop. A third workshop, entitled "Semiconductor Device Modeling", is scheduled for Mar 28-29, 1996. Preliminary sign-ups indicate that its attendance may well top the two previous workshops. An impressive array of leading experts in the semiconductor device modeling field is scheduled to speak. Papers to be presented include analyses of quantum effects, potential roadblocks to future progress, the outlook for photolithography, numerical techniques, TCAD software, and the conversion of large device simulation codes to highly parallel computer systems. A proceedings will be published. V. Milestones 1996: * Convene or help organize workshops in petaflops computing, device modeling, biological nanotechnology and molecular nanotechnology. * Complete at least four journal-quality scientific papers. Here and in the following, these papers will include work done by scientists at Ames, or sponsored by Ames under grants, cooperative agreements or computer time allocations. These papers will be published on the Ames web site. * Establish a mechanism for peer review of grants, solicit proposals for grants, cooperative agreements and computer * Establish an archive and web site for papers published under the scope of this program. * Develop a computer simulation of a molecular memory device. * Acquire a software package for simulation of molecular systems; integrate as appropriate with NASA visualization software. * Create a computer simulation to permit the study of tension forces on linear molecules. * Initiate development of computational tools required to perform molecular syntheses as proposed by Drexler and Merkle using electronic structure calculations. 1997: * Convene or help organize at least two workshops. * Complete at least six journal-quality scientific papers. * Solicit proposals for grants, cooperative agreements and computer time allocations. * Conduct a study of alternative petaflops architectures. * Parallelize at least one semiconductor device codes. * Develop a computer simulation of silicon device line drawing using an atomic force microscope (AFM). * Develop prototype code to simulate, with visualization, the mechanical properties of a variety of molecular structures. * Integrate simulation and/or visualization software currently used by Prof. Brenner of NCSU and Prof. Goddard of Cal. Tech. with software used at ARC. * Complete simulation for a prototype system to study the role of solvent in the self-assembly of hydrogen-bonded supramolecular aggregates. 1998: * Convene or help organize at least two workshops. * Complete at least ten journal-quality scientific papers. * Solicit proposals for grants, cooperative agreements and computer time allocations. * Analyze the performance of at least one key NASA-oriented numerical algorithm on potential petaflops architectures. * Perform a "landmark" semiconductor device calculation on one of the NAS parallel supercomputers. * Establish the state of the art in artificial eye (retina) and chemosensor chip development; test ability to incorporate in small robots. * Establish a database of material property scaling characteristics. * Develop a computer simulation of diamondoid fabrication. * Use simulation to test various designs for molecular machinery. * Develop a computational model to simulate the behavior of bulk diamondoid. * Perform a study of software facilities needed for the design and control of molecular assembly systems. * Release a report on the status of progress in the field of semiconductor device simulation, including an outlook for the future. This report will be the basis for determining whether further activity and funding is warranted. 1999: * Convene or help organize at least two workshops. * Complete at least 15 journal-quality scientific papers. * Solicit proposals for grants, cooperative agreements and computer time allocations. * Assess several operating system software options for future petaflops computers. * Develop a code to study some key aspect of semiconductor technology that is considered to be a potential "show-stopper" for future progress. * Assess success in developing small sensors and control systems based on biology. * Determine by consensus of the scientific community the canonical microcircuits to be simulated initially for more intelligent control systems and chips. * Complete development of computational tools required to perform molecular syntheses as proposed by Drexler and Merkle using electronic structure calculations. * Simulate the effect of defects in molecular machines, with visualization. * Simulate the formation of nanotubes by self assembly in solution. * Develop a simulation of a moderate sized, software controlled molecular manufacturing component, such as a nano-robot arm. * Release a report on the status of progress in the field of petaflops computing, including an outlook for the future. This report will be the basis for determining whether further activity and funding is warranted. 2000: * Convene or help organize at least two workshops. * Complete at least 15 journal-quality scientific papers. * Solicit proposals for grants, cooperative agreements and computer time allocations. * Develop a detailed simulation program that permits a number of features of future petaflops computer systems to be simulated in the scope of some NASA numerical algorithm or application. * Develop a code to study some key aspect of semiconductor technology that is considered to be a potential "show-stopper" for future progress. * Use electronic structure calculations and molecular simulations to study the thermodynamics and kinetics of the formation of components for molecular manufacturing from molecular subunits in solution. * Summarize achievements in Neuro/Nanotechnology in sensor and control system architectures for further applications and implementation. * Use molecular simulation to test designs for molecular machinery. * Develop prototype software to simulate software-controlled nanotechnology replicators, and test critical portions of this code using simulations. * Release a report on the status of progress in the field of computational nanotechnology, including an outlook for the future. This report will be the basis for determining whether further activity and funding is warranted. VI. Resources Required The activities conducted so far in the above areas are being done with relatively little formal support. If Ames is to have a serious impact in these areas, more resources are required. In particular, a number of personnel slots are required, and significant funding is required for computer hardware and software (including commercial molecular modeling software, to be acquired by IN), grants and conference costs. These estimates do not include the cost of upgrading parallel supercomputer facilities, which we presume will be made available through NAS, probably through the HPCCP. They do however include the expected cost of some low- and mid-level graphics workstations. Here 'CS' denotes civil servants, 'SSC' denotes support-service contractors (perhaps later to be converted to either performance-based contractors or civil servants), 'NRC' denotes NRC postdoctoral fellows. These personnel figures are full-time equivalents. We expect that the civil servant slots will be for Ph.D-level researchers, ranging from post-doctoral scientists at grade 12 to senior scientists at grade 15 and 16. 'Computer' denotes funds for computer hardware (workstations) and software, and 'Grants' denotes funds for research grants and cooperative agreements. Costs of hosting conferences are also included in this last figure. Costs are in thousands of dollars. 1996 Project Org CS SSC NRC Computer Grants Petaflops IN 1 0 0 $ 0 $ 0 Semicond. IN 0 1 0 $ 0 $ 0 Nanotech. IN 1 2 0 $100 $600 Nanotech. STC 2 1 1 $ 0 $ 0 Bio. Nano. SLR 1 0 0 $ 0 $ 0 Total 5 4 1 $100 $600 1997 Project Org CS SSC NRC Computer Grants Petaflops IN 1 1 1 $ 50 $100 Semicond. IN 1 1 0 $ 50 $100 Nanotech. IN 3 2 1 $100 $400 Nanotech. STC 4 2 2 $150 $400 Bio. Nano. SLR 1 0 1 $100 $300 Total 10 6 5 $450 $1300 1998 Project Org CS SSC NRC Computer Grants Petaflops IN 2 1 1 $100 $200 Semicond. IN 1 1 1 $ 50 $100 Nanotech. IN 4 2 2 $150 $600 Nanotech. STC 5 2 3 $200 $600 Bio. Nano. SLR 2 0 2 $150 $600 Total 14 6 9 $650 $2100 1999 Project Org CS SSC NRC Computer Grants Petaflops IN 2 1 2 $100 $250 Semicond. IN 1 1 1 $ 50 $150 Nanotech. IN 4 2 2 $150 $800 Nanotech. STC 5 2 4 $200 $800 Bio. Nano. SLR 2 0 3 $150 $800 Total 14 6 12 $650 $2800 2000 Project Org CS SSC NRC Computer Grants Petaflops IN 2 1 2 $100 $300 Semicond. IN 1 1 1 $ 50 $200 Nanotech. IN 4 2 3 $150 $900 Nanotech. STC 5 2 4 $200 $900 Bio. Nano. SLR 2 0 3 $200 $900 Total 14 6 13 $700 $3200 The program will be re-evaluated in the year 2000 to see if further staffing, grant or contract funding should be pursued. It should be noted that some members of the team believe that NASA would benefit greatly from a significantly larger effort, but these figures represent a compromise with funding realities.