Nanos And Petas And FLOPS — Oh, My!

At U of I, researchers are working with the tiny, the enormous and the infinities in between. Here's a primer for the rest of us.

By Mary Timmins

As scientific research explores the very tiny and the very large, increasingly powerful and compact computing ability is needed. Above, four computers at the National Center for Supercomputing Applications on the University of Illinois campus can together produce 155 trillion calculations per second. At far left, among the very small projects at the U of I is a 12-millimeter-long transistor made of gold nanoparticles, coincidentally looking similar to the University's I Mark. NCSA/Blake Harvey Image

The first thing to understand about nanotechnology is what "nano" is, and that's easy. It's a prefix. "Nano," placed in front of certain words, divides those words by 109, which is 10 multiplied by itself nine times, or 1 billion. A nanometer is thus a billionth of a meter, 1/100,000th the diameter of a human hair.

Pluck out one of your hairs and stare down its barrel for a bit and think about how scientists are physically manipulating elements and compounds in a space of, say, 10 nanometers, which equals 1/10,000th the size of the end of the hair. For nanotechnologists, such diminutive dimensions are just part of life at the office. For us civilians of the scientific world, though, something that not only exists but can be physically manipulated at 1/100,000th of a hair across is something that's going to bring on deep angst – or at least the need to lie down for a few moments – if we contemplate it too long. That's why this article is not for nanotechnologists.

Nor is it for chemists or engineers or biophysicists or cosmologists or others who already understand what it means to work with the huge beyond huge and the teeny beyond tiny. This article is for those of us who are just beginning to realize that these scales exist and who want to come to better terms with what's going on beyond the not-so-big middle we all inhabit and what sorts of benefits and understanding are drifting in from those far places, where scientists at the University of Illinois and other institutions wander, as explorers in the time of Queen Victoria wandered Africa and Asia.

Consider, again, the nano realm. Atoms and molecules there behave, well, differently. When materials get down to scant billionths of a meter, the few atoms and molecules still around bond with one another in new ways and evince what people in the press call "unusual properties." The bonding and properties aren't accidental – chemists create nanomaterials from the atoms up, while engineers refine them from the mother substances down. Research at the nanoscale is mega-hot. Studies at the U of I alone range from a way to sequence the human genome molecule by molecule to fuel cells that power themselves, as well as improved panels for collection of solar energy, self-healing materials and drug delivery by "smart" molecules that send up little flares to show they're doing their job.

Nothing, though, has more nanotechnological allure than nanotubes. These are skinny, molecular structures formed, under laboratory-induced duress, from materials such as gold and carbon. Since 1991, researchers have been "growing" nanotubes and trying to get them to work as semiconductors – which is to say, as material (usually silicon) that is made into transistors and microprocessor chips and handles the electricity flowing through computers and other electronic devices.

Last year, John Rogers, a much-honored researcher who is Founder Professor of Materials Science and Engineering at Illinois (where he is affiliated with several departments as well as the Beckman Institute for Advanced Science and Technology), made something that no one had ever made before – a radio with working carbon nanotube transistors. When members of his team first tuned it in and listened on headphones, they got a radio station in Baltimore playing "Safety Dance" by Men Without Hats. The song sounded pretty good. It sounded really good, actually, considering that the radio's transistors could fit on a grain of sand.                         
Insanely tiny devices that provoke retorts about angels dancing on the head of a pin are not about to show up behind the photo counters of big-box retail stores any time soon. Rogers made the radio to show – as it did, most dramatically – that nanomanufacturing is just about ready for the marketplace. Rogers wants to use nanomaterials "to do things that silicon wafers can't do because [they are] rigid and brittle. And so we make circuits out of nanotubes and also very, very, very small, thin pieces of silicon, nanoribbons of silicon that allow you to build circuits on sheets of plastic or pieces of rubber so that you can deform them and wrap them in ways that are inconceivable if you're tied to a silicon wafer."

Such flexible circuits could – indeed, will – be used in materials such as fabric and paper, introducing computer technology in circles where it has not socialized before, like "smart" clothing that monitors your health or reads the ambient temperature and warms you accordingly.

Very, very small and very, very fast

While some computer technology is growing down into ever tinier components, other computer technology is getting bigger and bigger – by also using ever tinier components. This is, in part, because computers really aren't all that smart. They only understand yes/no – a bit like toddlers, if rather more obliging. Computers work in binary code – clumps of zeros and ones, nested within clumps of zeros and ones, nested within clumps of zeros and ones, etc. Dating back to an invention in the 1600s by the German mathematician Leibniz, through 1958 when IBM built a "stretch" computer 33 feet long, to the Pentium 4 laptop (which has 280 times more power than that 1958 IBM), computers have always been, in the most grossly simplistic terms, arrays of on-off switches. The smaller the switches the better, because designers can (a) build smaller computers and (b) build computers that contain enormous amounts of power relative to their size. It is (b) that drives the work across the desk of Thom Dunning, the affable, white-bearded chemist who is director of the National Center for Supercomputing Applications on the UI campus.

At NCSA, computing power is expressed in flops – a clumsy-sounding acronym for floating point operations per second. A simple calculator that can do the math at a rate of 10 calculations per second has a capacity of 10 flops. The current NCSA stable of four supercomputers (including machines named "Abe" and "Lincoln") have, among them, a ripsnorting 155 teraflops to bring to bear on scientific problems.

"Tera" being a prefix that multiplies by a factor of 1012 or 1 trillion, 155 teraflops means that the NCSA computers can carry out 155 trillion calculations per second. What's worse for those of us who can't conceive of this kind of computing power – much less the needs that drive it – is that there's something even bigger coming at us, courtesy of Blue Waters, the new supercomputer that NCSA is building with IBM and other partners at Illinois and around the country. Blue Waters, which will go online in 2011, will have more than one petaflops of computing power, the capacity to do more than 1015 – a quadrillion – calculations per second.

Incomprehensible? Getting there.
 
NCSA has even worked up a time scale. Imagine a researcher working continuously on a calculator. That unfortunate person would need a loooong time (31 million years) to do the calculations that Blue Waters will be able to carry out in one second.

Robert Wilhelmson and Matthew Gilmore, University of Illinois; Advanced Visualization Laboratory, NCSA; Lou Wicker, National Oceanic and Atmospheric Administration The National Center for Supercomputing Applications on the UI campus produces 3-D mathematical models of scientific phenomena such as the swirling formation of a tornado, left. View video: www.ncsa.uiuc.edu/vislab

The infinite task of rendering the world

This is talking gi-normous power – poised to roar into the void of gi-normous questions. For what sucks up computing capacity today is mathematical modeling, a process that has arisen to assume its place in the pantheon of scientific values, alongside theory and experimentation.

"As time has gone on, scientists have developed mathematical models to describe whatever phenomena it is that they're looking at," Dunning explained. "Whether it's the laws for these very small things that are embedded in something called the Schrödinger Equation or in the formation of the universe, where you're talking about laws that were formulated by Albert Einstein, you've got these mathematical models of what is happening. … What the computer allows you to do is to … get a numerical solution." Such numerical solutions allow scientists to describe the specific behavior of phenomena, but such descriptions can require trillions and quadrillions of calculations – even for tiny phenomena.

The U of I's Klaus Schulten is a veteran German researcher whose 30-year career as a molecular biologist is filled with achievements that include a map of the human brain and the sussing out of key secrets of photosynthesis. In 2006, he led a team that created the first full computer simulation of all the atoms in an entire living organism. This organism was neither animal nor plant nor insect but virus – the satellite tobacco mosaic virus At one-tenth of a micrometer (that is, a 10-millionth of a meter) across, it's among the smallest known life forms.

In tandem with the computing power of NCSA, Schulten and his team used a program (which they themselves had previously developed) to visualize the organism, simulating its million atoms over a nanosecond, which is a billionth of a second. The group has since expanded the scope of their simulations to 1/100,000th of a second.

But cells function in what Schulten calls "beats," which he likens to the beat of a piece of music. A full cellular beat consumes a millisecond – a relatively lengthy 1/1,000th of a second. And if the scale is still small, the stakes are not. "We are getting closer to describing living cells the way engineers describe a new Boeing when they compute it before it is built," said Schulten. "And that, of course, has great implications not only for the understanding of human nature – we are made of living cells – but also, of course, for medicine." Modeling what a cell does over the beat of a millisecond is what Schulten calls "the holy grail" of his field.

"The petascale computer," he said of Blue Waters, "promises to get to this goal."

While Schulten's goal may be in sight – and on site at the south edge of campus, where ground broke in November on the building that will house Blue Waters – the chain of future benchmarks stretches to infinity. A full minute of the virus. Its full life. The seconds and the minutes and the lives of the plants and the animals and the humans. And that's just biology. There's physics to consider and chemistry and meteorology – not to mention economics, the social sciences, education, law, business and the humanities. Models of systems are being created by researchers in many – perhaps all – fields that lend themselves to mathematical expression of theory and data, from meteorology (one NCSA 3-D visualization, which you can watch through those funny glasses while touring the building, shows how winds and atmosphere marry to birth a tornado) to Chicago traffic (also rendered in 3-D for a study of disaster evacuation plans for the city). Among the most stunning simulations achieved, in part, on NCSA supercomputers is a rendering of a section of the universe 1.5 billion light years across. The visualization, shown on this page, is a dark cube enclosing brilliant webs of galaxies, shimmering as sunlight shimmers in a lake on a summer day. It was created by a team from the University of California San Diego, Michigan State University and NCSA.

Thus, behemothly big as Blue Waters will be, it is no ultimate solution. In research, as in the rest of life, ultimate solutions tend to be few and very final. Computers have less in common with truth, beauty, sex and death than they do with highways, addiction, income and closet space – the need, ever-insatiable, rises to overwhelm capacity. Already, Dunning and others at NCSA and throughout the scientific world are talking about exaflops – 1,000,000,000,000,000,000 or 1018 or a quintillion. For to computers falls the infinite task of absorbing, reprising and saving (literally) the world, space and the universe.      

And in the end, even infinity has its limits – as a human concept that can be inhumanly hard to work with.

How many decimal places can a computer carry?

For Susan Lamb, a UI physicist who speaks in the clipped, luscious syllables of her native London and who models galaxies – using, as it happens, the Turing supercomputer in the College of Engineering, as well as a small supercomputer she built herself – matter and energy simply aren't all that predisposed to arrange themselves in research-friendly, interlocking ways. No matter what size the phenomena a researcher chooses to model, said Lamb, there will be associated phenomena that are too small or too large to fully describe in the model and which may yet have important effects on the model.

"The scale even in astrophysics between the things which are important – like the size of a star or even of a galaxy – the scale difference between that and that of the whole universe is so vast that you cannot build a model in the computer which encompasses the level of detail you might need on the small scale as well as embedding that on the large scale," Lamb said. Moreover, "there are always errors in the computer manipulation of numbers," she added. "You can keep dividing, but there are only so many times you can do it before you're into noise. It comes down to how many decimal places your computer can carry."

"Noise," in the tech world, means information that arrives too small or too fractured to be coherent and that may interfere with other information. Like radio static. Or typographical errors. Or a document from a printer with a spent or broken ink cartridge. In science, as in life, noise can mean big headaches. Tim McCarthy, a UI philosopher who specializes in logic and mathematics, observes that there appear to be limits to scale – meta-points where information recedes and noise takes over. "On an ordinary, macroscopic scale … the laws of classical physics seem to hold – approximately. But they break down completely at very large scales of speed and time and space. And they break down on very small scales, too. And they're replaced with something else."

Beyond John Rogers' nano radio, for example, hisses the conundrum of quantum noise – the physics of how particles behave at the atomic and subatomic levels, zinging around at speed, immeasurable and unobservable, not at all the way good particles behave here at street level. Ever intrepid, Rogers doesn't regard quantum noise as necessarily posing fundamental limits. But he observed that "designs will have to change to deal with these phenomena. Simply shrinking large-scale devices down to tiny dimensions won't work."

At the other end of the scale, the study of the cosmos for Lamb entails the elemental, extraordinary fact that the carbon atoms in our bodies come from the formation of the earliest stars. "We are stardust," she smiled, quoting folk singer Joni Mitchell. Lamb notes, moreover, that the Big Bang is thought to have happened on an extremely minute scale, with the universe expanding out of materials at the subatomic level. "The energy stored in atoms is what powers the stars," she observed.

Objects in the mirror may be closer than they appear. In the short film "Powers of Ten" – made more than 30 years ago by Charles and Ray Eames – an area of 10 square meters, iterated from 10-16 through 1024 (a scant 40 calculations) moves the observer from the nucleus of an atom to the outermost limit of the universe. Moreover, as McCarthy points out, between every point and every other point – from zero to one, for example – is a continuum that can be endlessly divided. Infinities are everywhere.

As concepts change and the research that they drive rises and challenges our understanding, as in the maelstrom of the tornado and Dorothy's nightmare about Oz, there's a desire to lean into the familiar, murmuring that there's no place like the middle ground of home. But in the very act of exploring the extremes of the very great and the very small and the infinities in between, scientists at Illinois are expanding the dimensions of what we are and our need to know who we are and what we think about the things we can see and the prefixes we give to the things that we cannot.