C&EN Washington

Joseph J. Hanak was close to sparking a revolution in materials research some three decades ago, but the world wasn't ready for it yet.

Hanak, a Ph.D. chemist then working at RCA Laboratories in Princeton, N.J., was searching for new low-temperature superconductors. But he became increasingly impatient with the traditional, time-consuming approach of making one composition, testing its properties, then making a different composition, testing it, and so on. Why not make many different compositions at the same time, he wondered, and then rapidly measure the relevant properties in a single experiment?

Hanak's idea was the forerunner of a concept that is now taking materials science by storm: the application of the combinatorial approach, which the pharmaceutical industry uses for the rapid synthesis and screening of large collections of new drug candidates, to the discovery of useful new materials.

Hanak's approach, which he called the "multiple-sample concept," involved using sputtering to codeposit two or three elements or compounds on a substrate to create a continuum of compositions [J. Mater. Sci., 5, 964 (1970)]. The compositions were then determined by measuring the film thickness in two or three spots and plugging the numbers into equations that were solved using a mainframe computer at a remote site. In the course of screening such "multicomponent systems," Hanak uncovered a variety of new compositions, including superconductors, light-emitting materials, magnetic recording materials, and amorphous silicon semiconductors for solar cells. Using this route, he was able to find new materials much faster--up to several hundred times faster--than by using the conventional route.

RCA Labs, however, decided not to expand this research effort. And despite the publication of a series of papers and U.S. patents by Hanak and coworkers, he tells C&EN, the approach "never became popular" with researchers because of the general lack of computers. "Without a computer, you could not do the compositional analysis, automated testing of properties, and data processing, all of which are crucial for the method to work," he explains.

In retrospect, "we were a little bit too early," says Hanak, who is now a consultant on thin-film materials and renewable energy and is based in Ames, Iowa.

Flash forward to 1995. Computers had become commonplace in the lab. And scientists had learned how to generate large diverse collections, or libraries, of organic and biological molecules and screen them quickly for biological activity or other functions. These libraries were generated by the parallel synthesis of many similar compounds, or by reacting large numbers of precursors in many different combinations at once--the so-called combinatorial approach.

One of the key players in this field--chemistry professor Peter G. Schultz, who has a joint appointment at the University of California, Berkeley, and Lawrence Berkeley National Laboratory (LBNL)--wondered whether combinatorial chemistry could also be used to discover nonbiological or inorganic compounds. He got his colleague Xiao-Dong Xiang, a solid-state physicist, interested in pursuing the idea with him at LBNL. The result was a paper in Science [268, 1738 (1995)] in which they demonstrated, for the first time, that the combinatorial approach could indeed be used to discover solid-state materials with novel properties.

In that study, according to Xiang, the Berkeley researchers sought to determine whether a combinatorial approach could uncover two of the most important high-temperature cuprate superconductors that had been found using the conventional one-at-a-time synthesis method.

To answer this question, the researchers devised a method for forming thin-film arrays of solid-state compounds from multiple layers of precursors using sputtering. But unlike Hanak's approach, the precursors are deposited on-to the substrate sequentially rather than simultaneously. "Each section of the substrate is exposed to a different combination of precursors by depositing each layer through a different mask," Xiang explains. The masks direct the precursors to individual sample sites that are less than 1 mm across.

This study involved the use of seven precursors: oxides or carbonates of the elements that make up known superconductive cuprates. After a series of depositions, the library array was thermally processed--a necessary step leading to the production of crystalline cuprates. Then the resistance at each site in the array was measured as a function of temperature. Sure enough, the sites containing the known superconductors showed large resistance drops indicative of superconductivity.

The Xiang/Schultz paper was a revelation for many materials researchers because it showed that relatively complex materials such as high-temperature superconductors could be made and found in a parallel fashion. Before long, other papers began appearing in the literature, describing efforts at using combinatorial methods to search for a wide variety of materials, including magnetoresistive materials, phosphors, dielectrics, ferroelectrics, polymers and polymer composites, semiconductors, catalysts, and zeolites.

Solid-state chemist Robert C. Haushalter even changed jobs as a result of the Xiang/Schultz paper: He moved to Symyx Technologies, a high-tech company in Santa Clara, Calif., so that he could be part of the new wave that was beginning to sweep through materials research. Symyx was founded in 1995 by Schultz and biotechnology entrepreneur Alejandro C. Zaffaroni specifically to find new materials using combinatorial and high-throughput screening methods. When Haushalter joined Symyx two years ago, he tells C&EN, it was "the only show in town." Since then, other companies have started their own combinatorial materials efforts.

Haushalter sees the combinatorial approach as "a tool that you can use to accelerate the discovery process." Xiang says the discovery process could be accelerated "by a factor of thousands, possibly millions." And because materials libraries can be created with very small amounts of material, the combinatorial route promises to be less wasteful than traditional approaches and possibly also less expensive.

In a recent review article, Schultz and coauthor David R. Liu point out that the properties of many functional materials arise from complex interactions that depend on their composition and processing. In general, scientists don't know how to predict a material's properties from its structure, so they have to search for structures that have desirable properties. "Given approximately 60 elements in the periodic table that can be used to make compositions consisting of three, four, five, or even six elements," Schultz and Liu write, "the universe of possible new compounds with interesting physical and chemical properties remains largely uncharted."

And "once you get to four or five elements," Haushalter says, the number of possible compositions that can be made is so large that, if you make them one at a time, you can only scratch the surface. In that situation, he comments, you have to use the combinatorial approach--it's the only sensible, efficient way to survey the possibilities.

Combinatorial discovery of materials is still a very young field--one that has no proven commercial successes (C&EN, Dec. 8, 1997, page 24). And its practitioners face many technical problems and other hurdles on the road to developing marketable products. Yet scientists pursuing the combinatorial approach are upbeat about its potential, believing that it will most likely lead to a revolution in materials research.

Already, there have been some successes, many from the LBNL group. Four months after their groundbreaking paper on combinatorially produced superconductors, Xiang, Schultz, and their Berkeley coworkers announced the discovery--using solid-state libraries--of a new class of magnetoresistive materials based on cobalt oxide [Science, 270, 273 (1995)]. The resistance of these materials undergoes a large change in a magnetic field, making them potentially useful, for instance, for magnetic read/write heads in computer disk drives. Prior to this study, large magnetoresistances had been found only in certain manganese oxides, according to Xiang.

Using materials libraries, the Berkeley team also has discovered new red, green, and blue light-emitting phosphors that may find use in luminescent displays. One of the red phosphors, a gadolinium zinc oxide doped with europium [(Gd_1.54Zn_0.46)O_3-:Eu³⁺_0.06], is reported to have a purer red color than the commercial red phosphor Y₂O₃:Eu³⁺, an industry standard whose emission peak is actually in the orange-red region [Appl. Phys. Lett., 72, 525 (1998)]. Although the new red phosphor has a slightly lower quantum efficiency than the commercial phosphor, it may be a better candidate for projection TVs, field-emission displays, and X-ray imaging applications, according to Xiang and coworker Xiao-Dong Sun.

The Berkeley group hasn't been the sole player in the combinatorial phosphor sweepstakes, however. Symyx researchers have discovered and optimized new luminescent compounds by preparing and screening combinatorial libraries containing some 25,000 samples on a 3-inch wafer. One of their finds is a red phosphor--an yttrium aluminum lanthanum europium vanadate--that is redder than Y₂O₃:Eu³⁺ and offers a comparable quantum efficiency [Nature, 389, 944 (1997)].

Three months after that discovery was published, a Symyx team reported "a fundamentally new type of luminescent material, Sr₂CeO₄" [Science, 279, 837 (1998)]. When bulk samples of this blue-white phosphor were prepared and studied, it was found to have an unusual one-dimensional chain structure not previously seen in light-emitting rare-earth-based oxides. This structure is "intimately related to the charge-transfer mechanism by which Sr₂CeO₄ luminesces," according to chemist Earl Danielson and his colleagues at Symyx. Charge transfer is not the usual mechanism for light emission in phosphors.

Sr₂CeO₄ is but one example of a serendipitous discovery. Another example, from LBNL, is an unusual composite of a gadolinium gallium oxide and silicon dioxide (Gd₃Ga₅O₁₂/SiO₂). This material emits a blue glow when irradiated with ultraviolet light. The Berkeley researchers found it quite unexpectedly by depositing a library containing gadolinium gallium oxides on an oxidized silicon substrate. When that particular Gd-Ga-O compound was deposited on a different substrate, it didn't luminesce. This and other evidence suggest that the material's blue emission arises from the interaction of the Gd-Ga-O compound with the SiO₂ found on the substrate [Science, 279, 1712 (1998)]. Xiang believes that the application of combinatorial chemistry in materials science "has dramatically increased the probability of [such] chance discoveries."

Phosphor libraries are relatively easy to screen because devices that measure light intensity and color are either readily available or can be assembled from commercially available modules, according to Xiang. Other properties, such as electrical properties, are much more challenging to measure in a quick, quantitative, and nondestructive manner. Conventional contact measurements using electrodes are destructive, often yield misleading information, and are difficult to carry out on materials libraries, Xiang says. The solution, he believes, is to measure electrical properties at microwave frequencies so that no physical contact with the sample is necessary. His team has developed a new tool--a scanning evanescent microwave microscope (SEMM)--to do just that.

Xiang and coworkers have used the SEMM to evaluate thin-film libraries of ferroelectric and dielectric materials for microwave applications. Such materials are being sought for the next generation of integrated capacitors and dynamic random access memory devices. With its large dielectric constant, barium strontium titanate, Ba_xSr_1-xTiO₃ (BST), has emerged as one of the leading candidate materials, and it has been under intensive investigation in many laboratories.

The LBNL researchers suspected that adding calcium to BST would lead to a composition with improved electrical properties (specifically, the dielectric constant and dielectric loss). To check out this possibility, they used pulsed laser deposition and a computer-controlled shutter system to sequentially deposit layers of four precursors. The barium, strontium, and calcium layers were deposited so that each varied smoothly in thickness from a different edge of the triangular substrate to the opposite tip. Thermal processing of the sample over the course of several days yielded a continuum of barium strontium calcium titanate compositions. Xiang and coworkers used the SEMM to measure the microwave dielectric properties of this "composition spread." They found that compositions in the neighborhood of Ba_0.2Sr_0.4Ca_0.4TiO₃ had the most desirable properties for dielectric applications [Appl. Phys. Lett.,74, 1165 (1999)].

The LBNL scientists are now trying to see if they can use the SEMM and a magneto-optical imaging system at low temperatures to make noncontact electrical and magnetic measurements on libraries of potential superconductors.

Optical and electrical properties are just two segments of the enormous materials-screening pie. Hundreds of different performance screens are used in the materials industries, according to chemist Marianna F. Asaro of SRI International, Menlo Park, Calif. She described the many technical challenges of screening materials libraries at a Knowledge Foundation-sponsored conference on combinatorial approaches to materials that was held in San Jose, Calif., this past January. Some screens, such as luminescence or electrical resistance, can be miniaturized with either minor or significant modifications, she said. But about one-third of screens "are very difficult or impossible to redesign and require major innovation" to make them work as assays on the submicrometer scale.

In Xiang's view, probably the greatest challenge in screening a materials library is obtaining structural information on its members. Classic structure-determination tools such as X-ray crystallography require bulk samples; they are not amenable to the small-volume samples found in thin-film libraries. Therefore, to fully characterize combinatorial "leads," researchers have had to prepare bulk samples of these materials using standard solid-state synthesis. This step tends to slow down the discovery process.

As physicist Eric D. Isaacs of Lucent Technologies' Bell Laboratories in Murray Hill, N.J., noted at the San Jose conference, scientists want to be able to directly characterize as-grown thin films in libraries. And X-rays, he added, are "almost the ideal probe" because they are nondestructive and have great penetrating power, among other desirable characteristics.

Bell Labs' X-ray microprobe synchrotron beamline at the National Synchrotron Light Source at Brookhaven National Laboratory, Upton, N.Y., uses two elliptically bent glancing incidence mirrors to produce a 2-m x 20-m X-ray spot that can map the texture of a thin-film sample. Top: Map of a GdGaO₃ luminescent library on a silicon substrate under broad UV illumination. This library contains a new red phosphor with a color and quantum efficiency that is competitive with Y₂O₃:Eu³⁺, an industry standard.

Isaacs and physicist Gabriel Aeppli at NEC Research Institute in Princeton, N.J., in collaboration with Xiang and other scientists, have made significant progress toward this goal by using X-ray microbeams available at a number of synchrotron radiation facilities. The researchers adapted X-ray microprobe techniques to characterize a library of red, green, and blue phosphors grown at LBNL. With the aid of special X-ray mirrors, they were able to focus the microbeam to a spot measuring 2 m x 20 m--much smaller than the 1-mm x 2-mm phosphor samples they were looking at. To give one example, using a combination of X-ray fluorescence spectroscopy, X-ray diffraction, and near-edge X-ray absorption fine-structure spectroscopy, Isaacs and coworkers determined the chemical composition and crystal structure of a red Zn-Gd-Ga-O phosphor and the valence state of the europium dopant, which determines the phosphor's color [Appl. Phys. Lett., 73, 1820 (1998)].

In addition to garnering this information, the X-ray microprobe technique "even permits [the texture of] individual films to be mapped in exquisite detail, revealing the identity of secondary phases and compositional inhomogeneities within them," the researchers write.

The availability of very bright third-generation synchrotron sources, such as the Advanced Photon Source at Argonne National Laboratory, is now making it possible to focus X-ray beams to spots as small as 0.1 m, according to Isaacs. This will allow library characterizations to be performed 100 times faster than they are now, he told conferees in San Jose. As a result, scientists will be able to elucidate the structure and composition of more than 1,000 library samples in one hour, he said.

Although there would certainly be a market for new phosphors, even researchers working on phosphors admit that there is more commercial interest in catalysts. An improved olefin polymerization catalyst, for example, could be much more profitable for a company than a new phosphor for flat-panel displays, they point out.

The search for catalysts using the combinatorial approach has been under way for several years in a number of labs. Libraries of organic or organometallic molecules have been prepared and screened for catalytic activity on solid supports or in solution (in small vials or wells). And inorganic heterogeneous catalysts, which play an increasingly important role in the chemical and oil industries, have been identified in solid-state materials libraries. Scientists have devised several methods for screening catalyst libraries, and each has its own advantages and disadvantages.

Catalytic reactions are exothermic, so active catalysts reveal themselves as "hot spots" in infrared images. This technique--IR thermography--was used by chemists Steven J. Taylor and James P. Morken at the University of North Carolina, Chapel Hill, to screen a library of more than 3,000 potential catalysts on polymer beads [Science,280, 267 (1998)]. The study led to the identification of two organic compounds as effective nucleophilic acylation catalysts.

IR thermography also was applied to the screening of potential heterogeneous catalysts by professor Wilhelm F. Maier and coworkers at Max Planck Institute for Coal Research in Mülheim an der Ruhr, Germany [Angew. Chem. Int. Ed., 37, 2644 (1998)]. They examined different compositions of amorphous microporous mixed-metal oxides based on silica or titania. These were prepared by pipetting microliter amounts of precursor solutions into tiny wells on the surface of a slate substrate. (Slate was chosen for its low thermal reflectivity to avoid interference.) The solutions were then thermally processed to yield an array of solid samples, each one consisting of less than 200 g of material.

In a demonstration, Maier and coworkers assessed a library of 37 oxides for catalytic activity in the hydrogenation of 1-hexyne at 100 C. An IR image taken during the reaction (and carefully corrected for artifacts) revealed four spots that were warmer than the substrate, indicating that these samples were active catalysts in the reaction. The temperature differences between active and inactive spots were very small--no more than 0.7 C--but temperature differences as small as 0.1 C could be reliably detected.

By increasing the reaction temperature to 350 C and switching the gas flow over the library to a mixture of air and either isooctane or toluene, the Max Planck group was able to size up the same samples for their ability to catalytically oxidize these hydrocarbons. The catalysts found to be active in the oxidation reactions tended to be different from the ones that were good at hydrogenating 1-hexyne.

This study shows that thermal imaging can provide convenient and fast parallel assays of whole catalyst libraries, comments chemistry professor Thomas Bein of Purdue University, West Lafayette, Ind., in a recent article [Angew. Chem. Int. Ed., 38, 323 (1999)]. However, the technique has an obvious limitation in that it provides no chemical information about the products generated by the catalysts.

Ideally, scientists would like to have a simple, fast way of unambiguously identifying the product molecules that emanate from each individual catalyst site of an array. In an effort to address this need, Selim M. Senkan, a professor of chemical engineering at the University of California, Los Angeles, developed a laser-based method for the rapid screening of a library of solid-state catalysts that activate the dehydrogenation of cyclohexane to benzene. This reaction is of interest in the reforming of petroleum. Senkan's method involves the use of resonance-enhanced multiphoton ionization (REMPI) to detect the presence of the product molecule, benzene, in the gas stream containing the starting material, cyclohexane, that flows through the catalyst sites.

To demonstrate the principle, Senkan set up a library consisting of a row of eight sites, only half of which contained platinum or palladium catalysts. Reactant gases were forced through the individual sites, and a UV laser beam was passed through the air spaces above the sites. The beam was tuned to a wavelength that selectively ionized via a two-photon process any benzene that was formed to C₆H₆⁺ and an electron. These charged species were detected by an array of microelectrodes placed above the sites near the laser beam. The REMPI signals recorded in the experiment "clearly distinguished between the active and inactive sites on the catalyst library," Senkan states in his initial report on the method [Nature, 394, 350 (1998)].

At the Knowledge Foundation conference, Senkan announced that he and postdoctoral associate Sukru Ozturk had improved their earlier system. They developed a microreactor array and used it in conjunction with REMPI to screen a 66-member library of platinum-palladium-indium compositions for catalytic activity in the cyclohexane dehydrogenation reaction.

The microreactor array consists of 17 narrow micromachined channels that extend through a nonporous silica ceramic slab about 3 inches long. Each channel contains a small well for the catalyst pellet, a piece of porous alumina impregnated with 1% by weight of a Pt-Pd-In composition. The slab is heated to the reaction temperature while the reactant gas mixture (cyclohexane in helium) is fed into each of the 17 channels in parallel.

Near the exit ports, all 17 gas streams pass through the laser beam, which ionizes the benzene product, and the resulting ions are picked up by an array of 17 microelectrodes. In this experiment, the UCLA researchers used time-of-flight MS to check that the only ionized molecules produced were benzene. (Products other than benzene might be generated in this reaction, but the laser would have to be tuned to the appropriate wavelength to ionize them so that they could be detected.)

Each run of the microreactor array takes two to three minutes, Senkan said. To screen the entire 66-member library, the procedure had to be repeated five times, but that was accomplished in one day, he added. They are now building a larger system that will be able to screen an entire library of some 100 members at one time.

At the meeting, Senkan reported that the screening data enabled them "to pluck a winning combination"--a ternary mixture of 80% platinum, 10% palladium, and 10% indium that generated more benzene than any other library member. Senkan does not consider this composition a new lead for catalyst development. Rather, it simply demonstrates that screening a catalyst library can be done very quickly. He told meeting attendees that a report on this work will be published this month in Angewandte Chemie.

The 66-member catalyst library was prepared using completely automated equipment. The total time required for preparing and screening the library was about two-and-a- half days, Senkan says, but the entire process could probably be collapsed into a day, "with lunch breaks." The conventional approach, by contrast, requires months of work.

Another advantage of the library method, he points out, is that it's easy to make comparisons between the catalyst candidates because they are all prepared, processed, and screened in an identical manner. The usual practice of making catalyst candidates one at a time (over a long period of time) and in different labs makes such comparisons difficult, Senkan says.

Senkan's approach to catalyst screening has been called "elegant." But other scientists have noted that since benzene is the only product measured, no information is available on what other molecules may be formed as by-products.

A different approach that can provide such selectivity information involves the use of a small probe to "sniff" the reaction products and scanning MS to analyze them. Symyx researchers have been developing this method for the past three years. They have used it, for instance, to screen libraries of rhodium-palladium-platinum compositions for activity in reactions that occur in catalytic converters, such as CO + O₂ CO₂. These three metals are presently used in catalytic converters.

In a paper published last month [Angew. Chem. Int. Ed., 38, 484 (1999)], chemist Peijun Cong and his Symyx coauthors, including Chief Technology Officer W. Henry Weinberg, describe the probe as a system of concentric tubes that allows the gas stream containing the reactants (CO and O₂) to be delivered to a catalyst site and then removed by vacuum for analysis in a mass spectrometer. The 1.5-mm-diameter catalyst site is heated by a laser to the desired temperature before the sampling is done. The 136-element library sits on a stage that automatically moves each catalyst site into position under the probe. Each measurement is completed in about one minute, and only slightly more than two hours is needed to screen the entire library, the researchers say.

Of the three noble metals they looked at, rhodium and rhodium-rich compositions were most active for producing CO₂.

In one experiment, the Symyx workers sought to find out whether one of the expensive metals in the Rh-Pd-Pt catalysts could be replaced by a much less expensive metal like copper without significantly diminishing the catalyst's activity. By evaluating a Rh-Pd-Cu library, they discovered that a 1:1 Cu-Rh catalyst is as active as the pure rhodium catalyst at 400 C. Even so, Weinberg says that this binary catalyst is not of commercial interest. As he informed a colleague at the San Jose conference, Symyx has obtained catalyst results that are much more exciting but cannot as yet be disclosed.

The Symyx group also studied the oxidation of CO by nitric oxide (CO + NO CO₂ + N₂). This is a more complex and challenging reaction, in part because incomplete reduction of NO to N₂O is possible. Using ¹⁵NO in the feed gas, the researchers were able to distinguish the N₂ product from the unreacted CO, and the N₂O from CO₂. A number of interesting trends in reaction selectivity were observed, all "in complete agreement with the more limited data set reported previously in the literature," they write.

Even when combinatorial and fast screening methods pinpoint a promising new catalyst composition, there's no guarantee that it will become a commercial product, Senkan points out. Not only would bulk samples of the material have to be synthesized using conventional methods, but the material would have to pass numerous real-world performance tests. "There are lots of hurdles down the road," Senkan says.

During the past year, researchers have begun reporting efforts to apply the combinatorial approach to the synthesis of zeolites under hydrothermal conditions. Zeolites are microporous inorganic crystals such as aluminosilicates, which are widely used as adsorbents, ion exchangers, and catalysts. They are also being investigated as advanced materials for a host of additional applications. Their synthesis requires demanding conditions: temperatures above the normal boiling point of the solvent, high pressures, and high pH.

Duncan E. Akporiaye and coworkers at Sintef Applied Chemistry in Oslo, Norway, used a special multisample autoclave made out of Teflon (DuPont's polytetrafluoroethylene polymer) to carry out 100 zeolite crystallizations in parallel at temperatures up to 200 C [Angew. Chem. Int. Ed., 37, 609 (1998)]. The products had to be manually removed from the reactors and were analyzed by conventional X-ray diffraction techniques.

Maier and coworkers at the Max Planck Institute advanced the state of the art by reducing the hydrothermal reaction volumes to 2 L (from the 500 L used by the Sintef group) and by automating the analysis. After carrying out 37 parallel syntheses in a microautoclave, they heat the product crystals to make them adhere to the silicon wafer forming the bottom of the autoclave. This wafer is removed, and the attached crystals are automatically identified by X-ray microdiffraction, with the X-ray beam focused to a spot 500 m across [Angew. Chem. Int. Ed., 37, 3369 (1998)].

As Purdue's Bein explained at the San Jose conference, his group has developed a somewhat different automated system for preparing and identifying zeolites. In his system, reagents are dispensed automatically into Teflon autoclave blocks having eight or 19 reaction chambers with volumes of either 150 or 300 L. Six blocks have been processed at the same time. The products are recovered almost quantitatively using a centrifuge technique, and the resulting library can be analyzed automatically by X-ray diffraction or by scanning electron microscopy.

The Purdue group has used this system to study, for example, the influence of different amounts of organometallic and organic structure-directing agents (templates) on the resulting zeolite phases. "There are an enormous number of parameters" in zeolite syntheses that can be varied, Bein says, and dramatic effects sometimes are observed in the resulting structures. No one so far has reported the synthesis of any new zeolite structures using parallel synthesis, but that's something Bein would like to attempt in the near future.

One thing is clear: Bein, like many materials researchers, is enthused with the combinatorial approach. "It's very convenient," he says, ticking off the advantages: fast library preparation, smaller reaction volumes, lower consumption of chemicals, and more data. And according to LBNL's Xiang, "There are clear opportunities out there" for certain types of new materials.

What's unclear is the impact that combinatorial methods will have on the materials industries, Xiang says. For that, we'll have to wait at least a few years.

Chemical & Engineering News
Copyright © 1999 Ame rican Chemical Society