Ron Dagani
C&EN Washington

"Few terms in the chemical and physical sciences have seen more use (and abuse) in recent years than 'nanoscience' or, even worse, 'nanotechnology,' " writes James R. Heath in a guest editorial published last year in a special issue of Accounts of Chemical Research that was devoted to nanoscience.

"Why all the interest and hype?" writes Heath, a chemistry professor at the University of California, Los Angeles. The interest is relatively easy to explain, he says: "The past 15 years or so have witnessed an explosion of relatively inexpensive analytical tools, such as scanning probe microscopies, for interrogating and manipulating materials on the nanometer length scale. At the same time, several previously unrelated fields [such as electrical engineering and biology] have begun to focus on understanding and controlling physical and chemical phenomena" on this length scale, typically 1 to 100 nm.

Scientists have learned how to control the size and shape of a wide variety of materials at the atomic or molecu-lar level. And in the process, they have uncovered interesting and potentially useful properties, many of them unanticipated.

The field "is going to blossom and turn into a major force in science over the next few years," comments Chad A. Mirkin , a chemistry professor who directs the Institute for Nanotechnology at Northwestern University, Evanston, Ill. "It's almost a runaway train. There's so much excitement over it."

At the same time, Mirkin points out, "there's a tremendous amount of hype in this area." Much of that hype is embodied in optimistic forecasts of futuristic nanotechnologies that, some claim, will grow out of today's rudimentary nanoscience.

Nanotechnology visionary K. Eric Drexler , chairman of the Foresight Institute in Palo Alto, Calif., and some of his associates, for instance, have proposed the construction of molecular-scale robotic arms that could be used to build a variety of objects--including other molecular robotic arms--one atom at a time. In one daring vision, programmed robotic devices smaller than 100 nm would circulate freely in a person's bloodstream, identifying cancer cells and selectively destroying them before they could form a tumor.

Such ideas are considered science fiction by many scientists who are otherwise enthusiastic about nanoscience. For example, UCLA chemistry professor J. Fraser Stoddart says, "This field got off to such a bad start because of these pictures being painted in people's minds of robots swimming through our bloodstream and killing this or that nasty."

As a result of all the interest and hype, the definition of nanotechnology has become a bit blurry--in part because so many researchers are trying to wedge themselves in under the nanotechnology umbrella, even those who are working on micrometer-scale systems. Some use nanotechnology in the Drexlerian sense to refer to the creation of molecule-size machines that can manipulate matter with atomic precision. At the other extreme, nanotechnology is sometimes taken to include all of molecular biology and chemistry--a notion that Mirkin considers "silly."

To be sure, chemists are used to working on the nanometer scale. "But making an organic compound using traditional synthetic chemistry is not an example of nanotechnology," Mirkin points out. By contrast, the use of self-assembly techniques to make small molecular components coalesce into a macrocyclic molecule having multinanometer dimensions can legitimately be considered nanotechnology, he believes.

The crucial difference is that in the second example, the nanoscale structures are being prepared, characterized, manipulated, and even visualized with tools that weren't available more than 15 years ago. Nanotechnology "is a tool-driven field," Mirkin stresses. "And the tools are going to get better and better as we go."

Stoddart (left) and Heath: catenanes, rotaxanes, and pseudorotaxanes

Another key aspect of nanotechnology is that nanoscale materials offer different chemical and physical properties than the bulk materials, and that these properties could form the basis of new technologies. For example, scientists have learned that the electronic--and hence optical--properties of nanometer-size particles can be tuned by adjusting the particle size. Thus, when gold metal is reduced to nanosize rods, its fluorescence intensity is enhanced over 10 million-fold, according to a recent study by chemistry professor Mostafa A. El-Sayed's group at Georgia Institute of Technology. The study found that the wavelength of the emitted light increases linearly with the rod length, while the light intensity increases with the square of the rod length [Chem. Phys. Lett., 317, 517 (2000)].

"These nanoparticles are considered to be a new state of matter whose properties depend not only on the chemical composition, but also on the size and shape" of the particles, El-Sayed explains. Such properties are of interest for applications in optical data storage, ultrafast data communications systems, and solar energy conversion.

Nanomaterials already play a key role in a number of commercial technologies. But this article will focus on some nanoscience research that is still years from commercial fruition. This research captures the imagination, though, because it offers the promise of dramatically changing the way electronic devices, sensors, motors, and many other items are manufactured.

Today, such devices are fabricated using a "top down" approach. In the microelectronics industry, for instance, lithographic techniques are used to etch away at a silicon crystal to form micrometer-size devices and circuitry. These techniques lately have been refined to the point that features with nanoscale dimensions can be fabricated. As device features have become finer, the number of devices that can be crammed onto a chip has been doubling every 18 to 24 months.

But chip makers will be hard-pressed to extend this miniaturization trend for another decade. As device features shrink into the low-nanometer range, the chips will not be able to perform as reliably. Moreover, the cost of constructing new fabrication lines for each new generation of chips will become prohibitive.

Nanotechnology promises an inexpensive "bottom up" alternative in which electronic or other devices will be assembled from simpler components such as molecules and other nanostructures. This approach is similar to the one nature uses to construct complex biological architectures.

Switching with molecules

Heath's lab is in the forefront of efforts to build a computer from the bottom up--what he calls a "chemically assembled electronic nanocomputer." In collaboration with chemist R. Stanley Williams and computer architect Philip J. Kuekes of Hewlett-Packard Laboratories in Palo Alto, Calif., his group has developed much of the architecture for such a machine. And more recently, collaborating with Stoddart's group at UCLA, his group has begun to build it.

"When you tell people you want to build a computer, they think you're going to take on Intel," Heath remarks. "That's not our point." His aim is to demonstrate that a rudimentary nanocomputer can indeed be built. Mindful of the challenges, he says, "We think it would be a tremendous exercise to try to learn how to make such a machine and then actually do it."

Basically, Heath explains, the project involves stringing together dozens of molecular switches and nanowires into logic circuits and memory circuits, and "getting them to talk to each other." The molecular switches the UCLA researchers are experimenting with are the redox-active catenanes, rotaxanes, and pseudorotaxanes developed in the Stoddart lab during the past decade. The simplest examples of these switches consist of a molecular ring that is mechanically interlinked with a different ring (to form a catenane) or is threaded on a molecular axle (to form a rotaxane or pseudorotaxane). In either structure, the ring can assume two different positions that represent the digital states "1" and "0," and it can be switched between those states using applied voltages.

To connect the molecular switches, the UCLA team is exploring the use of silicon nanowires and carbon nanotubes arranged in a grid "like a tic-tac-toe board," according to Heath. This architecture is derived from that used in a unique silicon-based computer known as Teramac, which was built at Hewlett-Packard a few years ago. At each junction of this grid, the nanowires are connected by a monolayer of the molecular switches.

Last year, Heath, Stoddart, and their coworkers demonstrated that rotaxane-based molecular switches could be strung together to form logic gates, although the state of the switches could only be changed one time [Science,285, 391 (1999); C&EN, July 19, 1999, page 11 ].

In August, the group took the next step and reported on catenane-based molecular switches that can be reconfigured--that is, switched on and off--many times [Science, 289, 1172 (2000)]. Although the difference between the "on" and "off" states is much too small (in terms of resistance) to be useful for logic circuits, the switches could be useful for memory, Heath says. What's important about this paper, Heath points out, is that it is the first time he's aware of that a solid-state molecular switch has been shown to work repeatedly under ambient conditions.

The UCLA researchers now have other reconfigurable molecular switches that are a big improvement over the one they reported in August. Soon they hope to use these in demonstrations of logic and memory circuits. "Then we just have to get them to talk to each other, and you have a computer," Heath says. Such a prototype computer could be just three or four years away, he adds.

Toward a nanocell strategy

Tour: training nanocells to compute

A very different approach to building a molecular computer from the bottom up is being pursued at Rice University's Center for Nanoscale Science & Technology in Houston. There, chemistry professor James M. Tour and coworkers have been synthesizing and studying molecular wires and molecular devices of a different sort. Their nanowires are conjugated chains in which, for example, functionalized benzene rings alternate with acetylene groups. The wires bear specific functional groups at either end that serve as "molecular alligator clips" for attaching the wires to gold or other electrodes. Using several techniques, Tour and his longtime collaborator Mark A. Reed , professor of electrical engineering and applied physics at Yale University, have measured small electrical currents flowing through these wires.

Tour's lab also has synthesized related molecules--likewise consisting of aromatic rings alternating with acetylenic groups--that function as molecular devices such as diodes and switches. Last year, for example, Tour and Reed reported that a monolayer of one such molecule, when cooled to 60 K, exhibits an unusual switching behavior that is not seen in conventional silicon devices [Science, 286, 1550 (1999); C&EN, Nov. 22, 1999, page 11 ]. When a steadily increasing voltage is applied to the monolayer, the molecules do not pass any significant current until a threshold voltage is reached. The current then zooms up, peaks, and turns off sharply as the voltage continues ramping up.

Reed and Tour also have observed this switching behavior, known as negative differential resistance (NDR), in a related molecule at room temperature, although in that case the magnitude of the effect is not as impressive [Appl. Phys. Lett., 77, 1224 (2000)]. Because these molecules can be switched between two stable oxidation states, they can store information in the form of a "0" (insulating state) or "1" (conducting state) and thus also serve as a molecular memory device.

Molecules with unusual device properties like NDR are "very interesting" from a scientific standpoint, Heath comments. It's exciting that "you can design a property into a molecule using traditional organic techniques and see that property emerge in a solid-state device when you put that molecule between two electrodes. That's a result no one really expected to see," he tells C&EN. "It means that you can imagine making a whole cast of devices that have unique properties."

In studies at 60 K by Tour and Reed, this molecule was shown to exhibit negative differential resistance (a type of switching behavior) and store information like a memory device.

Tour hopes to use such functional molecules to build a molecular computer. As he explained in a talk at the national meeting of the American Chemical Society, held in August in Washington, D.C., this computer would be fabricated from basic units called nanocells that would be chemically self-assembled and then programmed to do something useful.

The self-assembly processes that are at the heart of the Rice/Yale and UCLA approaches to a molecular computer are inherently imperfect--that is, they cannot guarantee that a particular molecule will always end up in the right position and in the right orientation. But that doesn't matter because both computer designs are highly tolerant of defects. That's in stark contrast to present-day computers, which can be crippled by a single defective element.

The nanocell conceived by Tour and coworkers is about 1 m² in size and contains a two-dimensional array of a few hundred metallic nanoparticles bridged by about 1,500 functional molecules (such as those exhibiting NDR). These molecules also would connect nanoparticles to input and output leads arranged around the periphery of the nanocell. Thus, different combinations of input and output leads would allow one to address different current-carrying pathways.

The arrangement of nanoparticles and bridging molecules in these pathways would be random. And the pathways probably won't be able to perform any useful logical functions at first, Tour says. But by applying voltage pulses to various combinations of input and output leads, he explains, it will be possible to set molecules (switches) "on" or "off" in groups. Which switches are on (conducting) and which are off (insulating) won't be known, but that doesn't matter. In a trial-and-error fashion, special computer algorithms would repeatedly test and tinker with a pathway (using voltage pulses of different magnitudes) until the pathway performs the desired operation, such as that of a logic gate or adder.

	A test chip (left) designed and used by Reed to study the current/voltage characteristics of organic molecules prepared by Tour. The two images on the right show magnified views of the center of two different square patterns seen on the chip. In the magnified views, the wires along the edges extend out to the macroscopic world, where test leads from instruments can be hooked up. Some of the lithographic lines visible in the magnified views are tipped with gold contact pads that are 0.3 to 1 -m apart. When the chip is momentarily dipped into a solution of a test compound, molecules assemble themselves across pairs of these pads. The electrical properties of these molecules can then be studied.

A full-scale molecular computer would consist of at least 100,000 to 1 million nanocells connected to each other through conventional lithographic wiring. After the first nanocells are trained, the trained nanocells would serve as testers and trainers of neighboring nanocells. This "bootstrapping" approach would allow the programming or training of nanocells to be carried out rapidly and automatically, Tour says.

He and his coworkers have already shown by modeling (simulation) that they could train a nanocell to perform a function. "But we haven't yet constructed a full nanocell and demonstrated the programming within it," Tour says. Such a demonstration, though, could come within six months. Beyond that, he and his team members will still have to successfully resolve many formidable challenges before they can show off a working model of a molecular computer.

Like the UCLA scientists, Tour doesn't think that such a molecule-based computer will supplant traditional silicon-based computers in the near term. Rather, molecular electronics will find their first applications in hybrid systems "where molecules work in concert with silicon."

Switching with nanotubes

Not all approaches to molecular computing necessarily rely on molecules that are accessible via stepwise organic synthesis. At Harvard University, for instance, chemistry professor Charles M. Lieber and his coworkers--including Thomas Rueckes, Kyoungha Kim, and Ernesto Joselevich--are exploiting single-walled carbon nanotubes (SWNTs) for both device elements (such as switches) and wires for reading and writing information.

Lieber's idea is to pattern an array of parallel nanotubes on a thin dielectric (insulating) layer covering a conducting substrate, and then suspend above that array--and at right angles to it--another parallel array of nanotubes. The top nanotubes would cross--but not touch--the bottom nanotubes, because they would be suspended above the bottom nanotubes on regularly spaced supporting blocks about 5 nm high. One end of each nanotube would be connected to a metal electrode.

"Each cross point in this structure corresponds to a device element," Lieber and coworkers explain in a recent paper [Science, 289, 94 (2000)]. And each device element can exist in two states: In the "off" state, the crossing tubes are well separated and the junction resistance at the cross point is very high. In the "on" state, by contrast, the top nanotube stretches toward the bottom tube just enough to contact it, leading to a dramatically lower junction resistance.

"A device element could be switched between these well-defined off and on states by transiently charging the nanotubes to produce attractive or repulsive electrostatic forces," the researchers write. This would be done by applying voltage pulses to the pair of electrodes addressing a specific cross point. The state--on or off--of each cross point could easily be read by measuring the resistance of the junction, Lieber says.

Such a crossbar array not only could be used to configure logic elements for computing, but also could serve as a nonvolatile random access memory (RAM) that would offer significant advantages over conventional semiconductor RAM in terms of size, speed, and cost. Lieber points out, for example, that 10¹² device elements would fit on a 1-cm² chip. By comparison, a 1-cm² Pentium chip contains 10⁷ to 10⁸ devices. Furthermore, each element of the nanotube-based memory can store one bit. Current silicon-based devices, by contrast, require a transistor and a capacitor to store a bit in dynamic RAM (which must be continually refreshed) or four to six transistors to store a bit in static RAM. Furthermore, experiments and calculations suggest that the nanotube-based RAM could carry out switching operations at 100 GHz (100 billion cycles per second)--more than 100 times faster than the current generation of Intel chips.

The Harvard group's experiments thus far have been conducted on single junctions of 20- to 50-nm-thick bundles of nanotubes known as ropes. In several such devices, Lieber and coworkers have observed reversible switching between well-defined on and off states. "We think that these experiments represent clear proof of concept for our proposed architecture," they write.

But relying on nanotubes alone for crossbar arrays is problematic. Ideally, the Harvard researchers would like to construct the arrays using individual nanometer-thick SWNTs--semiconducting nanotubes on the bottom and metallic nanotubes on top. That way, Lieber says, "we would always have metal/semiconductor junctions," which are rectifying--that is, they allow current to flow in only one direction. Having rectifying junctions would ensure that the state of each junction could be read independently of the others.

Unfortunately, no one knows how to make metallic or semiconducting nanotubes on demand. Researchers typically make do by using mixtures of different types of nanotubes or by making chance observations.

One way to get around this problem would be to use doped semiconductor nanowires in conjunction with nanotubes. Lieber's group has spent the past few years developing a catalytic laser-based method for preparing a wide range of nanowires, including those made of silicon, gallium arsenide, indium phosphide, and other semiconductors. The method allows "a great deal of synthetic control" over the diameter, length, and electrical characteristics of these nanowires, Lieber says.

	Lieber: nanotube/nanowire arrays [Justin Ide/Harvard News Office ]

Recently, for example, his group has shown that silicon nanowires can be doped with other elements to give n-type (electron-doped) or p-type (hole-doped) materials [ J. Phys. Chem. B , 104, 5213 (2000)]. "An n-type silicon nanowire will always form a rectifying junction with a nanotube" whether the nanotube is metallic or semiconducting, Lieber points out. Furthermore, by crossing doped nanowires with nanotubes, one can obtain device junctions with different kinds of electronic properties. And incorporating silicon components into your device makes sense if you're interested in making hybrid devices, he adds.

How would such crossed arrays of wires be fabricated? One promising strategy, according to Lieber, is to chemically pattern a surface, such as with parallel lines a few nanometers apart, and then use the flow of a liquid over that patterned surface to align the nanowires in that pattern. "Making the suspended structure is a little more tricky," he notes, but it might be done by controllably growing nanotubes from nanoscale catalyst particles.

Lieber says his group is "working like crazy" to construct crossbar arrays such as one containing 16,000 junctions "at a density that's beyond what can conceivably be done in the next few years in silicon technology." Such a chip would be a milestone--a very early one on the long road to a commercializable nanoelectronics technology, he says.

DNA assembly, computation

The concepts of arrays, crossed "wires," and computation also figure in the work of chemistry professor Nadrian C. Seeman at New York University. But in his case, the wires are DNA strands that spiral, weave, zigzag, and cross each other to form unusual DNA motifs not seen in nature. Some of these molecules could be useful in the construction of nanoscale DNA-based objects and devices or even in DNA computation.

During the past two decades, Seeman has been exploring the potential of DNA as a construction or scaffolding material for the direct fabrication of structures such as crystals and nanodevices. Using branched, double-helical DNA molecules with "sticky ends" (single-stranded overhangs that are available to bind to complementary overhangs on other DNA helices), he and his coworkers have prepared complex nanoscale objects such as a DNA cube, truncated octahedron, and various kinds of knots.

Ultimately, Seeman hopes he'll be able to make intricate structures in two and three dimensions without necessarily having to specify where every single component has to go within the array. "I believe this will eventually lead us to designer solids and smart materials," he says.

As a building material, though, branched DNA generally lacks stiffness, he points out. So in recent years, his group has been preparing assemblies of DNA strands that provide greater structural rigidity. These have been used to build two-dimensional DNA arrays and a nanomechanical device whose two rigid arms can be rotated between two fixed positions.

Seeman's latest feat along these lines involves so-called triple crossover molecules, in which four DNA strands combine to form three double-stranded helices in a planar arrangement called a tile [ J. Am. Chem. Soc., 122, 1848 (2000) ]. The helices are linked together at four points where strands cross over from one helix to a neighboring helix, exchanging their binding partners. The central helix is closed by hairpin loops at both ends, while the outer helices have sticky ends that allow the tiles to recognize one another.

The sticky ends contain information that allows the tiles to undergo self-assembly in a way that performs a logical computation, according to Seeman and his collaborator, John H. Reif, a computer science professor at Duke University [Nature, 407, 493 (2000)]. They and coworkers Chengde Mao and Thomas H. LaBean used these tiles to carry out four computational steps on a string of ones and zeros using a logical operation called cumulative XOR. The result of the XOR operation is 0 if two successive numbers are the same (0 and 0, or 1 and 1), but it is 1 if the two successive numbers are different. The value of each tile (0 or 1) is indicated by a restriction site, a sequence that is recognized and cut by restriction enzymes.

	Seeman: nanohertz computing with DNA [Photo by Ron Dagani]

Input and output tiles, which have different sticky ends, are mixed in solution along with corner tiles that initialize the values of the calculation and help set up a framework for connecting the input and output tiles. The tiles self-assemble according to an algorithm that is determined by the output tiles. The input tiles assemble first in a flat staircaselike arrangement. And depending on how these tiles arrange themselves, the output tiles insert themselves into the available slots on the staircase by matching up complementary sticky ends.

Once the assembly is complete, the answer must be extracted. Woven into each tile is a reporter strand containing the restriction site that indicates the tile's value. The reporter strands of contiguous tiles are ligated to give a longer strand, which is then removed from the assembly. After amplification, the ligated strand is cut using two restriction enzymes and the fragments are analyzed by gel electrophoresis. "It's just like sequencing DNA, except at very low resolution," Seeman notes. The answer--the values of the output tiles that self-assembled--can be read directly from the pattern of lines in the gel.

The current demonstration uses only four inputs. Longer computations could also be performed with a single self-assembly step, Seeman says. But as the number of computational steps increases, errors are likely to become more common. In the experiments described in the Nature paper, the error rate was estimated to be 2 to 5%.

Seeman points out that this algorithmic self-assembly requires greater fidelity than other examples of DNA assembly he has worked on. In his earlier work on periodic arrays, a "right" tile was competing with "wrong" tiles, "so it wasn't too hard to get the thing to work," he says. "Here, the right tile is competing with partly right tiles. You need to be a little more stringent here than you are in the assembly of periodic arrangements. You've got to be all right, not just half right," Seeman adds.

Erik Winfree , an assistant professor of computer science and of computation and neural systems at California Institute of Technology, first suggested a few years ago that DNA could be used to mimic Wang tiles--squares with colored edges that can be used to perform calculations when they are arranged in a mosaic so that each edge is flanked by the same color. The sticky ends on the DNA tiles are the logical equivalent of the colored edges of Wang tiles. Winfree is pleased to see in the Seeman-Reif paper what he calls "the first real experimental demonstration of ideas I proposed in my Ph.D. thesis."

The hard part, though, will be moving from one-dimensional assembly to two or three dimensions. "This will allow for much more sophisticated information processing by the molecules--if it can be achieved," Winfree says.

David Harlan Wood , a professor of computer science at the University of Delaware, Newark, believes that Seeman's method will be more useful for construction than computation. "When I read this paper and think of construction, I have a wonderful vision of nails and boards flying into place," Wood tells C&EN. But applying this computational technique with 10¹² distinct molecules would be quite a challenge, he thinks. A powerful electronic computer, by contrast, "would tear through problems of this size in less than a microsecond."

Seeman agrees: "We're not talking about gigahertz here--we're talking about 100 nanohertz."

In any case, Seeman says, his primary goal is not computation per se but algorithmic assembly--using DNA to make novel and potentially useful nanostructures.

Nanostructures, after all, are the key to nanotechnology, whether they are designed to perform lightning-fast calculations, detect molecules in the environment, eliminate pathogens from the body, or improve the properties of a material.

And key to the creation of nanostructures is chemistry. Indeed, Seeman once characterized nanotechnology as "a very fancy buzzword for the chemistry of the next century." That may be, but crucial contributions also are being made by physicists, biologists, materials scientists, chemical and electrical engineers, and other specialists working together.

"This is a very exciting time to be doing nanoscience," UCLA's Heath says. "The field is moving forward very, very quickly." He's ebullient that the U.S. government intends to boost funding for nanoscience research as part of its National Nanotechnology Initiative (see page 39). "My only complaint," he tells C&EN, "is that they should have named it the National Nanoscience & Technology Initiative." The reason is simple, Heath explains: Nanotechnologies arising from nanoscience "are likely to revolutionize most of our key industries. But the science needs to come first."

C&EN Special Features Editor Celia Henry contributed to this article.

Top

Chemical & Engineering News
Copyright © 2000 American Chemical Society