Drain, whose own research interests include materials chemistry, invited seven leading figures in materials science to speak--people who he believes are shaping the future direction of the field and whose work he finds very interesting.

One of the invitees was Nobel Laureate Jean-Marie Lehn, professor of chemistry at Louis Pasteur University in Strasbourg, France. For years, Lehn and his coworkers have been designing and creating molecules that are programmed, by virtue of their structure and functional groups, to spontaneously organize themselves into larger supramolecular assemblies held together by such forces as hydrogen bonding or metal coordination. The hydrogen-bonded assemblies are built from molecular components that have a hydrogen-bond donor site, such as an amide hydrogen, and an acceptor site, such as a carbonyl oxygen, on different parts of the molecule. These components bind to each other to form a larger structure because the donor and acceptor sites recognize each other as complementary.

At the symposium, Lehn took his listeners on a guided tour of some of his wondrous supramolecular assemblies, with special emphasis on those that are organized and held together by metal ions. Using the metal-ion strategy, Lehn has built helicates, grids, cylindrical cages, and rotaxanes. Besides allowing such architectures to be reversibly assembled, metal ions also may imbue the structure with interesting redox, optical, magnetic, or other properties, he said.

In the helicate work, Lehn has used molecular strands consisting of several bidentate binding sites, such as 2,2'-bipyridine, separated by spacers. When linear strands having, say, five bipyridine units are mixed with metal ions (such as Cu⁺) in solution, the strands wrap themselves around a row of five metal ions, forming double-helical metal complexes. Strands containing sequences of bidentate and tridentate binding sites, when exposed to two metal ions having different coordination geometries, will wrap around a row of metal ions having the most suitable coordination geometry for each binding site. Lehn's group has used these principles to prepare a variety of double- and triple-helical complexes, including a complex in which two different strands form a double helix.

Another creation of Lehn's is a wreath-shaped double-helical complex consisting of five strands, each with three binding sites, intertwined about five Fe²⁺ ions arranged in a circle. The central cavity of this molecular wreath is tightly filled by a chloride ion, which is present in the reaction mixture because FeCl₂ is used as the source of the metal ions. The strong binding between the circular complex and the chloride ion makes the wreath a specific receptor for chloride or similarly sized ions, Lehn said.

This self-assembly, he believes, can be thought of as a type of dynamic combinatorial chemistry: The reaction components selectively form the five-iron wreathlike structure when chloride ions are present because that's the structure that will bind the chloride ion most strongly. When an anion significantly larger than chloride is used, a larger wreath is formed, the researchers have found.

Ligands that form the helicates must be flexible enough to intertwine. To make grid assemblies, on the other hand, Lehn has used rodlike pyridyl-pyridazine ligands containing several adjacent bidentate binding sites in a row. When these molecules are mixed in solution with metal ions such as Ag⁺, a grid complex resembling the pound symbol (#) is formed. At each point marking the overlapping lines (the rodlike ligands), a metal ion coordinates to two mutually perpendicular binding sites of the overlapping ligands.

Lehn described one such square assembly, a 3 x 3 array held together by nine Ag⁺ ions. But he also can create rectangular grid complexes by mixing rodlike ligands that make use of different numbers of binding sites. One example he showed was a 4 x 5 array consisting of nine ligands and 20 metal ions. Such nanoscale grids might one day find use "as components within a futuristic information storage and processing nanotechnology," he noted in a paper last year.

More recently, Lehn has prepared a hybrid ligand that contains a combination of the distinctive binding motifs that he has used in the helicate and grid complexes. This hybrid ligand, which has six binding sites, complexes with Cu⁺ ions to form, in almost quantitative yield, an intricate supramolecular macrocycle consisting of four ligand molecules and 12 Cu⁺ ions. The macrocycle, which is shaped like a bowl with a hole in its base, contains both the double-helical and gridlike motifs. The ligand strands are wrapped around each other, forming four linked double-helical sections with 12 crossing points. The structure has an external diameter of 2.8 nm and a central cavity 1.1 nm across that holds four PF₆^- anions as well as solvent molecules[Chem. Eur. J., 3, 99 (1997)].

In Lehn's view, the assembly of this intertwined macrocycle follows a "program" that contains two "subroutines"- one for double-helical assembly and one for grid assembly. He believes that multiple subroutines can be used to generate a wide variety of highly complex supramolecular architectures by combining two or more assembly processes.

Although these structures may seem complex, they are only a beginning and are still rather simple compared with what chemists should be able to build in the future, Lehn noted.

Drain, who previously worked in Lehn's lab as a postdoctoral associate, also is interested in using self-assembly to construct supramolecular architectures. Drain's assemblies, though, involve porphyrin molecules. During his opening talk at the symposium, he unveiled one of his latest laboratory creations: a square array of nine porphyrins tethered together by 12 palladium ions. This 21-component structure is 5 nm on a side and is shaped like a four-pane window. It is built from three different kinds of porphyrins-an X-shaped unit that coordinates to four metal ions and forms the center of the array, a T-shaped unit that coordinates to three metals and forms each side of the array, and an L-shaped unit that coordinates to only two metals and forms each corner of the array. When these components are placed in solution in the correct ratio, they form the square array within a half hour at room temperature in about 90% yield, Drain said.

These same porphyrin units, when used in different ratios, also can be induced to form wires or tapes. The properties of such arrays may be fine-tuned by choosing the appropriate metal ion linker and functionalized porphyrin unit, according to Drain.

The square array, he told C&EN, in many ways resembles the antenna complex that harvests light for photosynthesis. Drain believes that his artificial nanostructured materials offer unique photophysical and electronic properties that could be exploited in molecular photonic devices.

Another perspective on supramolecular chemistry was offered by Michael D. Ward, a professor in the department of chemical engineering and materials science at the University of Minnesota, Minneapolis. Ward, a crystal engineer, is interested in using molecular building blocks to construct crystalline frameworks with preordained architectures and useful functions. This has been difficult to do with organic building blocks, Ward explained, because they have numerous, very weak interactions that are hard to control. So crystal engineers have been seeking ways to control the assembly of molecular components into predictable crystal structures.

In pursuing this objective, Ward's group has developed a new class of crystalline clathrates-porous supramolecular frameworks whose voids are filled with guest molecules. In fact, Ward said, these frameworks aren't stable unless the guests are in the voids.

Unlike the assemblies of Lehn and Drain, Ward's structures don't involve metal coordination. Instead, they are based on extended sheets of guanidinium cations [C(NH₂)₃⁺] and organosulfonate anions (RSO₃^-) hydrogen bonded to each other in the plane of the sheet. The guanidinium cation has six protons-that is, six hydrogen-bond donor sites-while the oxygen lone pairs on each SO₃^- group provide six hydrogen-bond acceptor sites. Thus, the sheets containing these organic ions exhibit a hexagonal-type hydrogen-bonding array.

These sheets can stack in two different ways. When the organic substituent on the sulfonate anion is small enough so that all of the substituents can project to the same side of each sheet, an architecture consisting of discrete bilayers is produced. If, on the other hand, a heftier substituent is used, adjacent rows of these substituents project to opposite sides of the sheet, and a different architecture results in which the substituents from adjacent sheets are continuously interdigitated, as when two combs are pushed together so their teeth mesh.

The most interesting materials result when Ward uses disulfonates of the type ^-O₃S-R-SO₃^-, where the anionic groups are located at opposite ends of an alkyl or aryl substituent. In this case, the disulfonates serve as "pillars" that connect pairs of adjacent sheets, like the columns in a multifloor building. These pillars form solid walls that define extensive channels, which can hold many guest molecules [Science, 276, 575 (1997)].

Ward and coworkers have prepared more than 120 such crystalline clathrates, using different combinations of disulfonate pillars and guest molecules. They grow the crystals at room temperature by slow evaporation of a solution of a guanidinium salt, an organodisulfonic acid, and the desired guest molecule. A wide variety of clathrates may be prepared because many disulfonic acids are available, either commercially or through simple synthesis. The choice of the disulfonate determines the height, overall size, and chemical nature of the galleries in the resulting framework. The largest channels are roughly 15 Å wide.

The smaller guest molecules fit inside the pores of the pillared bilayer frameworks. But larger guests (such as 1,4-divinylbenzene) force the building blocks to assume a pillared bricklike architecture, Ward said, and, in fact, serve as a template for its formation. "Both architectures exhibit an astounding capability to adapt to the steric requirements of different guest molecules," Ward has written. "The pores can expand or contract about the guest molecules by puckering of the hydrogen-bonded sheet, and the pillars can rotate or change their conformation" to better accommodate the guests. This flexibility is important, as most attempts at crystal engineering have failed because the systems weren't adaptive enough, he said.

These robust clathrates have many potential applications. For example, the voids of the porous framework can be designed to selectively trap a particular guest molecule during formation (crystallization) of the framework. Other molecules in the crystallization solution-those not trapped in the clathrate-can then be removed. The guest molecules can later be recovered by dissolving the host framework. Ward believes this strategy might be applied to separations of fine chemicals, possibly even chiral separations.

The pores of these supramolecular frameworks also might serve as nanoscale reaction chambers. For instance, polymerization of monomers held as guests in the galleries could yield polymer replicas of the host lattice having precisely tailored pores and dimensions. In addition, designed clathrates might provide a new generation of engineered electronic, magnetic, or photonic materials. Ward and coworkers have already shown that these clathrates can be made to form crystals in which all the guests are oriented in the same direction. Such noncentrosymmetric crystals are capable of second harmonic generation, a nonlinear optical effect in which the frequency of light is doubled as it passes through the crystal.

While Ward has been learning how organic molecules assemble themselves into crystalline superstructures, chemistry professor Geoffrey A. Ozin of the University of Toronto has been investigating how inorganic materials can be made to form intricate solid-state structures whose curved shapes arise from forces other than crystallization.

In the materials world, Ozin is fond of saying, "shape is everything." Humans usually fabricate rather than synthesize shape because the shapes we know how to synthesize typically are films or polyhedral. When you grow a crystal of a salt, for example, the outcome is predictable-planar faces, sharp edges and corners. The shape of the crystal is dictated by the intrinsic characteristics of the atoms or molecules forming the crystal.

Nature, by contrast, knows how to make seashells, spirals, and a variety of other beautiful curved shapes that have been beyond the capabilities of chemists. But that's beginning to change. In 1995, Ozin and coworkers reported that they had learned how to synthesize inorganic structures that resemble the lacelike microskeletal structures of diatoms, radiolaria, and other single-cell marine organisms [Nature, 378, 47 (1995)]. Since then, the Toronto chemists have extended their studies. And using the principles of supramolecular chemistry, colloid chemistry, and liquid-crystal physics, they have begun to gain an understanding of where the shapes come from.

In their initial work, Ozin and coworkers mixed together alumina, phosphoric acid, and decylamine in aqueous tetraethylene glycol solution and let the solution sit undisturbed for hours or days. Out of this initially clear solution grew millimeter-sized aluminophosphate solid spheres and hollow shells whose surfaces were "sculpted" with micrometer-sized patterns of pores, bowls, ripples, meshes, and so forth. Ozin's experiments revealed that a decylammonium dihydrogenphosphate liquid-crystal phase was forming in solution, and that this surfactant, in conjunction with the glycol, was forming bilayer vesicles. These vesicles can adhere to one another, fuse, split apart, or collapse. The result is a microemulsion with a foamlike structure that Ozin believes can serve as a template for the deposition of curved, patterned structures made of aluminophosphate.

Last year, the research team described a related surfactant template system that produces silica objects with "a remarkable array of shapes, surface patterns, and channels" that are visible by scanning electron microscopy [Nature, 386, 692 (1997)]. These shapes and patterns included toruses, disks, spirals, pinwheels, spheres, seashells, knots, twisted hexagonal ropes, and gyroids (objects consisting of conical and cylindrical sections, some of which resemble spinning tops). The objects typically are several micrometers in size and their surface features are in the 2- to 10-nm range. Just looking at the micrographs, Ozin said, "you cannot tell whether a living entity or a nonliving entity created these shapes."

According to a model developed by the Toronto group, the silica precursor (tetraethyl orthosilicate) and the surfactant in solution assemble together into micelles. These micelles, in turn, form a liquid-crystal seed that appears to grow into objects of different shape by accretion and polymerization of the surfactant-silicate micelles. When the seed grows rapidly in the axial direction, the result is a rope with minimal curvature, Ozin believes. But stresses or other forces can act to increase the axial curvature, leading to loops and toruses. Other conditions lead to the growth of disks, spirals, knots, or spheres. Odd shapes that are difficult to describe may arise from changes in the rates of growth in different directions.

Ozin and coworkers have actually fished out a seed-a 60-nm "soft, elastic silicate liquid crystal" shaped like an egg-that they believe is an early stage in the growth of larger silica shapes.

The picture emerging from their studies is that the growth of silica objects with particular shapes and patterns is sensitive to acidity and other synthesis conditions. A slight increase in pH, for example, can alter the growth process so that continuous deposition of silicate material on specific regions of the silicate liquid-crystal seed is replaced by random, nonspecific deposition leading to spherical shapes with smooth surfaces. At one point in his lecture, Ozin noted, "We can tune the shape by tuning the dielectric constant of the solvent." Such tuning is possible because Ozin's system consists of charged colloidal particles in a charged environment. Evidence also is mounting that defects in the liquid-crystal seed also help direct the growth processes.

Aside from the fundamental interest in learning how to form curvy inorganic shapes, Ozin believes these materials could have numerous practical applications. For instance, micrometer-sized silica spheres with nanometer pores and a narrow size distribution would be of interest in catalysis and chromatographic separations of biomolecules. Other possible applications include membranes, fuel cells, batteries, electrodes, sensors for large molecules, protein purification, and toxic waste cleanup.

Lieber: creating   molecular-scale   probes

Inorganic nanostructures of a different type were the subject of a talk given by chemistry professor Charles M. Lieber of Harvard University. Much of his work has focused on synthesizing one-dimensional structures-nanowires, nanorods, and nanotubes-and measuring their mechanical and electronic properties.

Earlier this year, for example, Lieber's group reported a laser-based method for preparing crystalline semiconductor wires with diameters as small as a few nanometers. His initial report on the method [Science, 279, 208 (1998)] described the synthesis of silicon and germanium nanowires from catalyst beads consisting of FeSi₂ or FeGe₂. In New York, Lieber suggested that the technique might be used to grow "almost any elemental material" as a nanowire. His group also has grown nanowires of binary compounds such as silicon carbide, gallium arsenide, and cadmium selenide. Scientists are interested in nanowires not only for use in nanoscale electronic devices and other applications, but also for fundamental studies of materials.

For many materials scientists, carbon nanotubes are particularly fascinating objects of study because of their structure and properties. As Lieber pointed out in his talk, nanotubes are the stiffest material known. And they are tough: when bent, they buckle elastically rather than break. These characteristics help make them good candidates for use as high-resolution probes in scanning probe microscopy. In 1996, Richard E. Smalley and coworkers at Rice University, Houston, attached a single nanotube to the pyramidal silicon tip of a scanning force microscope and showed that it was quite robust and could image the bottom of deep trenches inaccessible to conventional tips (C&EN, Dec. 2, 1996, page 20).

Earlier this year, Lieber and coworkers demonstrated that a nanotube attached to the silicon tip of an atomic force microscope can be used to image biological structures with better resolution than the best silicon tips, as well as offering other advantages [J. Am. Chem. Soc., 120, 603 (1998)]. The improved resolution is due primarily to the fact that a nanotube (particularly a single-wall nanotube) has a smaller tip radius than the best silicon tips.

But even single-wall nanotubes haven't yet provided molecular-scale resolution, Lieber noted. "If we could actually modify these nanotube probes by taking advantage of organic chemistry, we could create very selective probes that could both sense and manipulate matter on the molecular scale." This is indeed what Lieber has demonstrated.

When a nanotube tip is prepared in an oxidizing atmosphere, the end of the tube is opened, and there is good reason to believe that the end is tipped with one or more carboxyl (-COOH) groups. If such a functionalized tip is immersed in aqueous solution and brought very close to a surface covered with hydroxyl groups, there should be a finite hydrogen-bonding interaction between the COOH and OH groups at low pH, Lieber explained. At higher pH, the carboxyl group would be deprotonated and sheathed by water molecules, which would dramatically reduce the carboxyl group's interaction with the hydroxyl surface. Lieber and coworkers performed this experiment in chemical force microscopy and found that at low pH there was some adhesion force between the carboxyl on the tip and the hydroxyls on the surface. But when the pH was raised above about 4.5 (the point at which COOH is expected to be deprotonated), the adhesion force quickly dropped to zero, in line with expectations. "This looks very much like a titration curve," he commented.

The Harvard researchers also have shown that the carboxyl group on the nanotube tip can be derivatized with acidic, basic, or hydrophobic groups for probing different chemical environments. "One can even go a step further" and create biological probes, Lieber said. For example, he and his coworkers covalently attached biotin (a carboxylic acid found in enzymes) to the nanotube tip. They then measured the tip's interactions with a surface covered with streptavidin, a protein that has a high affinity for biotin. Adhesion measurements revealed the occurrence of binding and unbinding of the biotin-streptavidin complex. When an excess of free biotin in solution was used to block all the receptor sites of streptavidin, or when the nanotube tip was not modified with biotin, the researchers observed no binding interactions between the nanotube tip and the surface. These and other findings in his lab have led Lieber to believe that molecular probes created by functionalizing nanotube tips could revolutionize the use of probe microscopy in chemistry and biology.

The electronic properties of carbon nanotubes also have wowed scientists. As Lieber explained, subtle structural changes in a nanotube can lead to major changes in its electronic properties. For example, whether a nanotube is metallic or semiconducting depends on its diameter and the arrangement (helicity) of the graphite rings of its walls. Thus, nanotubes have enormous potential for use in molecular electronic devices.

Two weeks ago, Smalley and coworkers at Rice reported that they had found a way to chemically cut single-wall carbon nanotubes into relatively short pipes that behave as individual macromolecules [Science, 280, 1253 (1998)]. Their idea is to use these fullerene pipes as connectors and components for nanoelectronic devices.

Zettl displays a model of   a nanotube

But nanotubes themselves can contain nanodevices, as was shown last October[ Science, 278, 100 (1997)] by a team headed by another symposium speaker, physics professor Alex Zettl of the University of California, Berkeley, who also is a researcher at Lawrence Berkeley National Laboratory. Two of Zettl's colleagues at UC Berkeley- physics professors Marvin L. Cohen and Steven G. Louie-had proposed in 1996 that the junction between two nanotube sections having different electronic properties would serve as a device. A metallic nanotube almost seamlessly joined to a semiconducting nanotube, for instance, would create a" Schottky barrier" that allows current to flow only in one direction- from the semiconductor to the metal. The junction would be marked by a pentagon-heptagon defect in the otherwise perfectly hexagonal framework of the nanotube wall.

Zettl and coworkers confirmed that Schottky barriers do indeed occur along carbon nanotubes. They used the metallic tip of a scanning tunneling microscope (STM) to reach into a tangle of nanotubes and slowly withdraw the end of one nanotube bundle or "rope" and pull it part way out of the tangled mass. This was possible because van der Waals forces caused the rope to stick to the STM tip as the tip was retracted. Once the rope was stretched out, the researchers slid the STM tip along the length of one of the nanotubes in increments of a few nanometers, measuring its current/voltage characteristics. Their measurements indicated that different segments of the nanotube exhibit different electronic behaviors, such as a Schottky barrier's diodelike rectification, suggesting the presence of "on-tube nanodevices."

Scientists currently do not know how to synthesize in a controlled way nanotubes having specific electronic properties. Nor can they wire on-tube nanodevices for specific purposes. A simpler approach to nanoelectronics, Zettl suggested at the symposium, might be to make a" tube cube," a small block of randomly, densely packed nanotubes that are interconnected in some unknown way. Zettl and coworkers prepared such a cube and stuck 10 pins into one side of it, and 10 pins into the opposite side. Then they attached leads to different sets of pins and measured the electrical characteristics of each circuit. What they found were a variety of devices, including transistors. Such a tube cube has a large number of possible electronic functions, and the most efficient way to discover them is to let a computer algorithm do the exploring, Zettl said.

Drain found this proposal quite exciting, telling C&EN, "It's the complete opposite of what everyone else is talking about." The standard approach is to design the molecular architecture so that the material will perform in a specific way, he explained. Zettl, on the other hand, is saying: Why not form the material and then learn what it can do? "It's a brand new way of thinking about such things that is extremely clever," Drain commented.

Besides exploring potential applications of carbon nanotubes, Zettl also has investigated their structural stability. Although nanotubes are strong and resilient, they can be deformed and even fully collapsed along their length, he and his coworkers have found. If one were to squeeze a nanotube between two parallel plates to the point that the flattened sections of the tube come very close together, according to Zettl, van der Waals attraction between the walls could contribute to making the total energy of the collapsed tube as low as or lower than the energy of the inflated circular tube. In this case, the fully collapsed nanotube would be stable.

Zettl's group has observed fully collapsed ribbonlike nanotubes in their samples. The researchers also have induced complete collapse by irradiating multiwall nanotubes with high-energy electrons, which selectively weaken the carbon-carbon bonds oriented at right angles to the beam [Chem. Phys. Lett., 256, 241 (1996)].

Carbon nanotubes, though, are not the only nanostructures Zettl is interested in. For example, he and his Berkeley collaborators predicted the existence of nanotubes made of pure boron nitride (BN) and then synthesized them a few years ago[Science, 269, 966 (1995)]. The synthesis involved arcing a boron nitride-filled tungsten rod against a water-cooled copper electrode in an arc-discharge chamber. The resulting multiwall BN nanotubes have the strength and elastic deformation characteristics of carbon nanotubes, Zettl noted in his symposium talk. And theoretical calculations have suggested that all boron nitride nanotubes are semiconductors, regardless of their diameter, chirality, or number of concentric tubes. Thus, BN nanotubes may turn out to be particularly useful for applications where a high-strength fiber of uniform electronic structure is desired, according to Zettl.

The Berkeley group also has prepared boron carbon nitride (BC₂N) and boron carbide (BC₃) nanotubes using arc-discharge methods [Phys. Rev. B, 51, 11229 (1995)]. Louie and Cohen have predicted that one type of BC₂N nanotube could carry current in a helical path along the carbon atoms, making it the world's smallest solenoid-a nanocoil. Such a nanocoil might be used in miniature inductors, transformers, receivers, and magnetic storage, according to Zettl.

The most recent unpublished work that Zettl presented at the symposium was not about nanotubes at all, but concerned the results of his search for a stable carbon-based cage molecule smaller than C₆₀. When C₆₀ is combined with certain alkali metals, the result is a superconductor-for example, Rb₃C₆₀ becomes superconducting at about 30 K. Fullerenes with less curvature lead to materials with a lower superconducting transition temperature (T_c). So, Zettl reasoned, a fullerene with greater curvature (and therefore smaller size) might provide a superconductor that has a higher T_c.

If one looks at nature, molecules of the protein clathrin are known to associate to form fullerene-like structures, some smaller than C₆₀. So 20-, 28-, and 36-vertex carbon-cage molecules should be possible as well, Zettl said. Indeed, when his group analyzed the carbonaceous material produced during a modified carbon-arc synthesis of C₆₀ and C₇₀, they also found C₃₆. A smaller cage like this would have adjacent pentagons, leading to a higher strain energy and perhaps less stability than C₆₀, Zettl noted. Their findings indicate that C₃₆ certainly is more reactive than C₆₀, but it does appear to be stable.

Now for the key question: Is it a superconductor? Theoretical calculations carried out by Cohen, Louie, and coworkers suggest that one of the two possible crystal structures of C₃₆ is a superconductor with a T_c much higher than 30 K (without requiring alkali-metal doping), while the other crystal structure should be an insulator. Unfortunately, the crystal structure they made happens to be the insulator. Nevertheless, Zettl was optimistic, pointing out, "We've got the right molecule-we just have to put it in the right crystal structure."

The formation of carbon-cage molecules and nanotubes occurs under high-temperature conditions where nature usually dictates the products. The same also can be said for metal-organic chemical vapor deposition (MOCVD), a process widely used in the electronics industry to deposit thin films of semiconductors, insulators, and metals. Gallium arsenide (GaAs) films, for example, are grown by mixing trimethylgallium [(CH₃)₃Ga] and arsine (AsH₃) in the gas phase and allowing them to thermally decompose near a hot substrate surface. The atoms of gallium and arsenic deposited on the surface move around until they find their thermodynamically controlled positions on the surface, explained another symposium speaker, chemistry and materials science professor Andrew R. Barron of Rice University. "The problem with that is that you can only make structures that nature wants to make-the thermodynamic phases at the temperatures at which you are growing the films."

Barron, on the other hand, isn't interested in growing nature's preferred materials. He's looking for new materials--specifically, metastable phases, which aren't stable under normal conditions and may have interesting properties. In his Hunter College lecture, he focused on the MOCVD growth of thin films of three semiconductors: gallium sulfide (GaS), gallium selenide (GaSe), and gallium telluride (GaTe). In all of these, gallium is in an unusual oxidation state- Ga(II), rather than the typical Ga(III). To prepare these "subvalent" gallium chalcogenides, Barron uses a single-source precursor, a volatile organometallic molecule that contains gallium and the chalcogenide in the same 1:1 ratio as is found in the desired thin-film product.

Industrial users are keen for single-source precursors that are volatile, nontoxic, stable to air and moisture, and that remain intact during storage and decompose at moderate temperatures. Barron imposes an additional requirement: "We want the structure of the [precursor] molecule to control the structure of the film you produce," he told his listeners. So he chose a series of precursors having a cubical structure that is unlike the structure of the thermodynamically stable phase of the thin film. The precursor in each case is a Ga₄E₄ cubane (where E represents the chalcogenide), and each gallium carries an organic substituent (R), such as a tert-butyl group.

In the synthesis of GaS thin films, the cubane precursor does not lead to the thermodynamic hexagonal phase, which has Ga- Ga bonds. Rather, it produces a face-centered cubic structure with only Ga-S bonds-a structure "that cannot be made by any other route," according to Barron. The cubic GaS phase, which even grows on amorphous substrates, has already been shown to be a promising material for electronic devices.

To be sure that the R₄Ga₄S₄ precursor they start with is the same molecule responsible for the film's crystal structure, Barron and his collaborators have performed experiments to prove each step of their proposed mechanism[ Organometallics, 14, 690 (1995)]. As a result of these experiments, Barron said, they know that when they volatilize the precursor at about 200 °C, its cubical structure remains intact. They also have confirmed that when the precursor is decomposed at MOCVD temperatures of about 350 °C, it loses four tert-butyl groups. The resulting "bare" Ga₄S₄ cubes were shown to assemble into a cubic lattice on the substrate surface. Thus the cubic structure of the thin film, he concluded, depends only on the structure of the molecular precursor that was used.

In the case of gallium selenide, however, the Rice researchers get the hexagonal phase, which is the thermodynamically stable phase. This result may be due to the fact that the Ga-Se bond is weaker than the Ga-S bond. In the gas phase, Barron explained, the R₄Ga₄Se₄ precursor decomposes not by loss of alkyl groups but by breaking Ga-Se bonds to lose an RGaSe moiety, which is one corner of the cube. That leaves an R₃Ga₃Se₃ fragment-a six-membered ring that might have a Dewar benzene or cyclohexane structure. This hexagonal fragment alights on the surface, stacking on top of other six-membered rings and forming a hexagonal lattice [Chem. Mater., 9, 3037 (1997)].

In the case of gallium telluride, a similar mechanism appears to operate, leading to hexagonal GaTe. But, as Barron pointed out, this is a new structure for this compound because the thermodynamic phase has a monoclinic structure. (The monoclinic structure is related to hexagonal in that it includes five-membered as well as six-membered rings.) By heating the hexagonal GaTe phase, Barron noted, one can convert it into the monoclinic phase. This transformation occurs readily because it requires only that certain Ga-Ga bonds reorient themselves.

Parkin: magnetic   multilayer magic

MOCVD certainly is not the only process that is used industrially to form thin films of materials for electronic applications. A simpler process called sputter deposition is being used to deposit thin, multilayer magnetic films. Because these films excel at detecting small magnetic fields, they have been incorporated into a new generation of magnetic read/write heads for hard-disk drives in computers. The physicist/materials scientist who pioneered the use of sputter deposition to prepare such magnetic multilayers is Stuart S. P. Parkin of IBM Almaden Research Center in San Jose, Calif. In his presentation at the symposium, he focused on the almost magical physical properties of magnetic multilayers that make them so useful in information technology.

In 1988, two European groups discovered that when certain films consisting of alternating layers of a magnetic and a nonmagnetic metal are placed in a magnetic field, the film's resistance changes by a large amount-a phenomenon known as giant magnetoresistance (GMR). This discovery reenergized the magnetic materials field because the best magnetoresistive materials known up till then showed much smaller changes in resistance.

According to Parkin, most of the magnetic read/write heads in use today typically are made of a nickel-iron alloy whose resistance changes by only 1 to 2% when the head senses the magnetic field emanating from data bits (tiny magnetized regions) on a magnetic disk. Scientists are continually trying to squeeze more and more information on these disks. And that means the magnetic bits are shrinking in size. Since 1990, the areal density of bits on disks has increased by more than 60% per year, Parkin said. As the bits get smaller, their magnetic field becomes less intense. Therefore, to read the information back, one needs a head that is more sensitive.

IBM scientists recently achieved the requisite sensitivity using magnetic multilayers whose resistance changes by about 10%. Although this figure is quite modest compared with the higher GMR values observed in some other materials, what's important is that the 10% resistance change is achieved in the weak magnetic fields encountered on disks. Earlier this year, IBM introduced disk-drive products incorporating these GMR read heads that offer" record-breaking storage capacities," according to Parkin.

One of the key features of multilayers is that they can be atomically engineered." By changing the layer thicknesses by just one or two atomic layers, we can dramatically change the properties" of the multilayers, Parkin pointed out.

For example, in multilayers of cobalt (a magnetic metal) and copper (a nonmagnetic metal), the thickness of the copper layers determines whether the magnetic spins in neighboring cobalt layers are aligned parallel or antiparallel. As the copper layer thickness is increased in increments of 5 Å-increasing the separation of the cobalt layers-the spin orientation of neighboring cobalt layers changes back and forth between parallel and antiparallel (that is, between ferromagnetic and antiferromagnetic coupling).

The resistance falls to a minimum when the orientation of each cobalt layer is parallel to that of its neighbors and rises to a maximum when all the cobalt layers' magnetizations are antiparallel. This oscillating pattern typically repeats every 10 Å or so of copper layer thickness. Parkin and his coworkers were the first to observe this unexpected phenomenon and demonstrate that it is a general feature of magnetic multilayers [Phys. Rev. Lett., 64, 2304 (1990)].

Such multilayers, he told his listeners, will form the heart of the $25 billion-per-year magnetic information storage industry within one or two years.

Looking even further into the future, Parkin described a newer project at IBM that is aimed at building a magnetic nonvolatile random access memory (RAM) chip. The key component of this chip's memory cell would be a magnetic tunnel junction-a sandwich of two magnetic layers separated by a thin insulating layer. Electrons would be able to tunnel from one magnetic layer to the other. When the magnetic moments of these two layers are parallel, the tunneling current is much higher than when the magnetic moments are antiparallel, he explained. Thus, the memory cell has two states defined by the relative magnetic orientation of the two magnetic layers. The states are read by sensing the resistance of the cell.

Parkin proposed that the architecture of this memory cell be very similar to Lehn's grid structures, except that the cell's features would be lithographically defined. The grid would consist of intersecting "word" and "bit" lines, and a magnetic tunneling junction sandwich would be located at each intersection. By passing currents through selected word and bit lines, information could be written and read.

This will be an extremely difficult device to build, Parkin noted, but if it works, it will make a computer's memory available for use immediately after the computer is turned on. At the same time, the device is expected to offer write and read speeds approaching that of conventional static RAM and a density comparable to conventional dynamic RAM.

"We are planning to tunnel into the future with this magnetic random access memory," Parkin said at the conclusion of his presentation. "We hope to have a prototype device [ready] toward the end of this year."

Parkin's lecture made it clear that, at least in the case of magnetic multilayers, some materials envisioned for the 21st century are already here.