NSF provided $213 million in support of research infrastructure during 1993, while NIH provided $117 million and DoD contributed $106 million. Of the non-federal sources of funding, the largest single source came from the academic institutions. A sizable contribution of $105 million came from private, non-profit foundations, gifts, bonds, and other donations.
A more recent survey of academic R&D expenditures reveals that, in 1999, slightly more than $1.3 billion in current funds was spent for academic research equipment. 12 Such expenditures grew at an average annual rate of 4.2 percent (in constant 1996 dollars) between 1983 and 1999. The share of research equipment expenditures funded by the Federal Government declined from 62 percent to 58 percent between 1983 and 1999. In addition, total annual R&D equipment expenditures as a percentage of total R&D expenditures were lower in 1999 (5 percent) than it was in 1983 (6 percent).13 As a point of comparison, during the past decade NSF support of equipment within a research grant has declined from 6.9 percent to 4.4 percent of the total grant budget. 14
Biannual surveys of U.S. research-performing colleges and universities reveal how these institutions fund capital research construction (costing $100,000 or more), in contrast to research instrumentation. The Federal Government's contribution to construction funds at the Nation's research-performing colleges and universities has varied over the past decade. In 1986-87 it accounted for 6 percent of total funds for new construction and repair/renovation of research facilities at public and private universities and colleges. This percentage increased steadily to 14.1 percent in 1992-93 and then declined to 8.8 percent in 1996-97. Very recent data indicate this percentage declined to 6.2 percent in 1998-99. 15
Table 2 indicates that, in 1996-97, research-performing institutions
their S&E capital projects funds from three major sources: the
Federal Government, State and local governments, and other institutional
resources (consisting of private donations, institutional funds,
tax-exempt bonds, and other sources).
Table 2. Sources of Funds to Construct and Repair/Renovate S&E
Research Space: 1996-1997
NOTE: Only projects costing $100,000 or more
The Federal Government directly accounted for 8.7 percent of all
new construction funds ($271 million) and 9.1 percent ($121 million)
of all repair/renovation funds. Additionally, some Federal funding
was provided through indirect cost recovery on grants and/or contracts
from the Federal Government. These overhead payments are used to
defray the indirect costs of conducting federally funded research
and are counted as institutional funding.
Table 3. Academic Research Space by S&E Field, 1988-2001
Maintaining the academic research infrastructure in a modern and effective state over the past decade has been especially challenging because of the increasing cost to construct and maintain research facilities and the concomitant expansion of the research enterprise, with substantially greater numbers of faculty and students engaged in S&E research. 17
The problem is exacerbated by the recurrent Federal funding of research below full economic cost, which has made it difficult for academic institutions to set aside sufficient funds for infrastructure maintenance and replacement. A recent RAND study estimated that the true cost of facilities and administration (F&A) for research projects is about 31 percent of the total Federal grant. Because of limits placed on Federal F&A rates, the share that the Federal Government actually pays is between 24 percent and 28 percent. This share amounts to between $0.7 billion and $1.5 billion in annual costs that are not reimbursed. Moreover, the infrastructure component in negotiated F&A rates has increased since the late 1980s, from under 6 percent in 1988 to almost 9 percent in 1999.18
Determining what colleges and universities need for S&E infrastructure is a difficult and complex task. Nevertheless, over the past decade a number of diverse studies and reports have charted a growing gap between the academic research infrastructure that is needed and the infrastructure provided. For example:
While these surveys and studies provide a rough measure of the magnitude of problem, they say little about the cost of lost S&E opportunities. In a number of critical research fields, the lack of quality infrastructure is limiting S&E progress. For example:
THE IMPORTANCE OF PARTNERSHIPS
The international dimensions of research and education are increasingly essential to U.S. science and engineering. As S&E infrastructure projects grow in size, cost, and complexity, collaboration and partnerships are increasingly required to enable them. These partnerships increase both the quality of the research enterprise and its impact on the economy and society.
The very nature of the S&E enterprise is global, often requiring
access to geographically dispersed materials, phenomena, and expertise,
as well as collaborative logistical support. It also requires open
and timely communication, sharing, and validation of findings, data,
and data analysis procedures. Projects in areas such as global change,
genomics, astronomy, space exploration, and high-energy physics
have a global reach and often require expertise and resources that
no single country possesses. Further, the increasing cost of large-scale
facilities often requires nations to share the expense.
ALMA conceptual image courtesy of the European Southern Observatory
The number of government-funded infrastructure projects that entail international collaboration has increased steadily over the last decade. For example, NSF currently supports a substantial and growing number of projects with international partnering. Among them are the twin GEMINI Telescopes, the Large Hadron Collider, the IceCube neutrino observatory at the South Pole, the Laser Interferometer Gravitational Wave Observatory, the Ocean Drilling Program, and the Atacama Large Millimeter Array.
All large future infrastructure projects should be considered from the perspective of potential international partnering, or at a minimum of close cooperation regarding competing national-scale projects. An additional challenge is maintaining interest in and political support for long-term international projects. Any absence of follow through on high-profile projects could increase the danger of the U.S. becoming known as an unreliable international partner.
Interagency coordination of large infrastructure projects is also extremely important. For example, successful management of the U.S. astronomy and astrophysics research enterprise requires close coordination among NASA, NSF, DoD, DoE and many private and State-supported facilities. Likewise, implementation of the U.S. polar research program, which NSF leads, requires the coordination of many Federal agencies and nations. University access to the facilities of many of the national laboratories has been facilitated through interagency agreements. There are a number of models for effective interagency coordination, such as committees and subcommittees of the White House-led NSTC.
In the fields of high-energy and nuclear physics, NSF and DoE have developed an effective scheme that facilitates interagency coordination while simultaneously obtaining outside expert advice. The High Energy Physics Advisory Panel (HEPAP), supported by NSF and DoE, gives advice to the agencies on research priorities, funding levels, and balance, and provides a forum for DoE-NSF joint strategic planning. This scheme has facilitated joint DoE-NSF infrastructure projects. For example, the HEPAP-backed plan for U.S. participation in the European Large Hadron Collider has been credited with making that arrangement succeed. 27
Partnerships have also played an important role in developing the genomics infrastructure. For example, the Human Genome Project, the Arabidopsis Genome Project, and the International Rice Sequencing Project have made vast amounts of genomic information available to researchers in the life sciences and other fields. Each of these projects was accomplished through a strong network of interagency and international partners.
Partnerships with the private sector also play an important role in facilitating the construction and operation of S&E infrastructure. For example, industrial firms have funded much of the equipment available in the Engineering Research Centers and the National Nanofabrication Users Network (NNUN). Public-private sector partnerships have also helped to enable the Internet, the Partnerships for Advanced Computational Infrastructure (PACI), and the TeraGrid Project.
THE NEXT DIMENSION
While there have been many significant breakthroughs in infrastructure development over the last decade, nothing has come close to matching the impact of IT and microelectronics. The rapid advances in IT have dramatically changed the way S&E information is gathered, stored, analyzed, presented, and communicated. These changes have led to a qualitative, as well as quantitative, change in the way research is performed. Instead of just doing the "old things" cheaper and faster, innovations in information, sensing, and communications are creating new, unanticipated activities, analysis, and knowledge. For example:
Research tools and facilities increasingly include digital computing capabilities. For example, telescopes now produce bits from control panels rather than photographs. Particle accelerators, gene sequencers, seismic sensors, and many other modern S&E tools also produce information bits. As with IT systems generally, these tools depend heavily on hardware and software.
The exponential growth in computing power, communication bandwidth, and data storage capacity will continue for the next decade. Currently, the U.S. Accelerated Strategic Computing Initiative (ASCI) has as its target the development of machines with 100 teraflop/second capabilities 28 by 2005. Soon many researchers will be able to work in the "peta" (1015) range. 29 IT drivers - smaller, cheaper, and faster - will enable researchers in the near future to:
With the advent of networking, information, computing, and communications technologies, the time is approaching where the entire scientific community will have access to these frontier instruments and infrastructure. Many applications have been and are being developed that take advantage of network infrastructure, such as research collaboratories, interactive distributed simulations, virtual reality platforms, control of remote instruments, field work and experiments, access to and visualization of large data sets,30 and distance learning (via connection to infrastructure sites). 31
Advances in computational techniques have already radically altered the research landscape in many S&E communities. For example, the biological sciences are undergoing a profound revolution, based largely on the enormous amount of data resulting from the determination of complete genomes. Genomics is now pervading all of biology and is helping to catalyze an integration of biology with other scientific and engineering fields. In order to fully understand the vast amount of genomic information available and apply it to improve the environment, nutritional quality of food, and human and animal health and welfare, new and improved computational and analytical tools and techniques must be developed, and the next generation of scientists and engineers must be trained to use them. Central to genomic sequencing and analysis is access to high-speed computers to store and analyze the enormous amount of data. Automated methods for model search, classification, structure matching, and model estimation and evaluation already have an essential role in genomics and in other complex, data-intensive domains, and should come to play a larger role in the future.
The Nation's IT capability has acted like adrenaline to all of
S&E. The next step is to build the most advanced research computing
infrastructure while simultaneously broadening its accessibility.
NSF is presently working toward enabling such a distributed, leading-edge
computational capability. Extraordinary advances in the capacity
for visualization, simulation, data analysis and interpretation,
and robust handling of enormous sets of data are already underway
in the first decade of the 21st century. Computational resources,
both hardware and software, must be sufficiently large, sufficiently
available, and, especially, sufficiently flexible to accommodate
unanticipated scientific and engineering demands and applications
over the next few decades.