Hollow buckyballs built of gold instead of carbon were observed in experiments and confirmed through computational modeling.
Balancing Computation and Experiment
April / May 2007
Volume 5 Number 2
How do you know when the science being performed on a supercomputer reflects what actually occurs inside a test tube, or in real life? The answer can be stated simply enough: "run the model on the computer, get the result, and then perform an experiment to test the result." The adage "easier said than done" truly applies—€”especially when the focus is on innovative science to benefit government and industry. The trick in this case is to find the right balance of science-driven computing integrated with experiment.
Balancing innovation, science, customer needs, plus an evolving state of the art in technology is a question worthy of study in and of itself, and there appears to be many valid approaches. Although the internet and modern communications have greatly enhanced our ability to interact with other people at a distance, there is still much to be said for assembling a critical mass of people and technology under one roof to solve a problem. History has repeatedly demonstrated the value of this paradigm—€”which still works well today—€”even though the scientific and technology landscape has significantly changed from even a decade ago. Long-lived organizations with a successful history of innovation can then provide valuable insight into those characteristics needed to dynamically balance computation and experiment across a variety of projects and technology landscapes.
The William R. Wiley Environmental Molecular Sciences Laboratory (EMSL), a Department of Energy national scientific user facility at Pacific Northwest National Laboratory, is celebrating its tenth year of operation. The facility combines science-driven computation with an experimental infrastructure housing a wide array of powerful high-throughput and high-resolution instruments to provide the tools scientists need to address complex questions in catalysis, materials research, energy production, hydrogen production and storage to name several of many different research areas. EMSL is home to the development team of the DOE's computational chemistry software, NWChem. As a result, this organization provides one example where software development, computational hardware and experimental capability are integrated under one roof allowing computer and experimental scientists from around the world to work together to solve large problems.
In essence, balancing computation and experiment requires maintaining a focus on science with an end application in mind while simultaneously providing an internal capability that is greater than the sum of its parts. This recognizes:
—€ That innovative science occurs when modeling (both theory and simulation) and experiment can interact with each other to converge to those solutions needed to solve real-world problems.
—€ The computational and experimental infrastructure needs to provide sufficient capability so this "convergence" process is not hindered, which highlights the need for a judicious investment policy.
—€ Through regular internal and external review processes, plus grand challenge events, the strengths and weaknesses of the existing capability are assessed in light of current and future requirements.
The interpretive power of EMSL modeling and experimental capability has recently been demonstrated by scientists from Washington State University and the University of Nebraska as they develop new materials with specific properties that are assembled from tiny clusters of atoms (e.g. finite-clusters of atoms).
In their research, they utilized the EMSL experimental facilities to create minute clusters of metal atoms and measure their unique properties.
Computational models were then used to interpret spectra measurements and analyze the shape and structure of the clusters. Their most recent work on hollow buckyball-like cages, was featured on the cover of the May 2006 issue of the Proceedings of the National Academy of Science.
The quest for environmentally friendly energy generation, storage and pollution reduction are important questions facing DOE, the United States and the world in general. Researchers are utilizing EMSL computational and experimental resources to focus on these issues.
For example, biological systems can convert sunlight into chemical energy very effectively—€”just think of the plants all around us. Researchers from Washington State University and the University of Alabama are using nuclear magnetic resonance experiments in conjunction with computational chemistry models in NWChem to gain a fundamental understanding as to how this conversion is performed by nature, which has the potential to provide insight into the development of novel solar energy devices.
At the same time, a group of researchers from Pacific Northwest National Laboratory are attempting to design new materials for solar cell collectors to harness energy from the sun. This requires advancing our fundamental understanding of material properties and their interactions with sunlight.
Their effort combines EMSL's computer simulation and analytical capabilities to characterize the composition and properties of thin films. Computational approaches are essential to guide material design while experimental analysis confirms or refutes the calculations. The ultimate goal is to find the right combination of materials and fabrication to more efficiently convert sunlight into electricity. Photosynthesis evolved over millions of years, these EMSL researchers would like to make their conversion of sunlight to energy happen much quicker.
Researchers from the University of Wisconsin-Madison used simulations to better understand how carbon monoxide can be oxidized on a catalytic surface by oxygen molecules, which can lead to better, more efficient ways to remove carbon monoxide from internal combustion engine exhaust and other sources. This oxidation process was observed experimentally but not well understood.
Computations performed at EMSL provided the first-ever quantitative analysis of this chemical reaction on a surface and confirmed experimental results. This work demonstrates how EMSL computation can participate in the quest for better experiments and theories.
These same researchers are utilizing the predictive power of computational modeling to design novel metal alloys that have superior catalytic properties for hydrogen-related reaction processes. Catalysts of this type have great potential in many applications ranging from new, highly selective catalysts for pharmaceutical production to robust hydrogen fuel cells. Their research was published in the November 2004 issue of Nature Materials.
Just as an organization must achieve a balance so everyone has the resources they need, so must the supercomputer support this philosophy in hardware through a balanced architecture. As a result, EMSL supercomputing provides some unique capabilities not found at other organizations—€”such as the storage system of the MSCF supercomputer (MPP2). Accurate modeling of chemical processes requires the generation of, and repeated access to, a large table of integrals that is computationally expensive to calculate and impractical to keep in memory. To overcome this challenge, the EMSL computer scientists came up with a method to efficiently use external disk storage. To meet this need of the scientific community and of the NWChem software, the EMSL MPP2 supercomputer was procured with a significant disk bandwidth per computational capability (Gflop/s). The next-generation EMSL supercomputer, currently being procured, will also maintain this capability.
It is worth noting that other data-intensive projects, aside from computational chemistry, benefit from the EMSL hardware capability. This includes pressing the limits of existing storage technology as demonstrated in a recent test on the EMSL MPP2 supercomputer that achieved a sustained read rate of 136 GB/s (roughly 34 DVDs per second) and a sustained write rate of 80 GB/s through the use of wide-striping files within a parallel distributed file-system.
Closing the loop with computation at EMSL are five tightly integrated experimental facilities, which are:
—€ Chemistry and Physics of Complex Systems Facility: fosters fundamental research in the natural sciences to provide the basis for new and improved energy technologies and for understanding and mitigating the environmental impacts of energy use.
—€ Environmental Spectroscopy and Biogeochemistry Facility: involves experimental and modeling studies of chemical phenomena and mechanisms on mineral and other surfaces.
—€ High-field Magnetic Resonance Facility: houses state-of-the-art nuclear magnetic resonance (NMR) and electron paramagnetic resonance (EPR) instrumentation for determining molecular structures.
—€ High-performance Mass Spectrometry Facility: provides high-throughput leading-edge mass spectrometry capabilities that allow visualization and analyses of proteins of a cell in greater detail than before.
—€ Nanoscale Science Facility: provides cutting-edge facilities and scientific expertise to design surfaces, interfaces and thin films for selective chemical and physical properties.
In summary, computation serves to interpret and understand the complex data generated by experiment. Operating as part of the discovery process rather than in isolation, it now plays an ever-increasing role in predicting and guiding experiments - and for chemistry in particular in the choice of materials analyzed. Model validation provides the essential feedback to the computational scientists and software developers to ensure the computer is correctly representing what is happening in the test tube, and to the computer architects that the computer hardware provides the right tools to solve the problems.
Copyright © 2012 | Innovation America