November 2001
Vol. 31, No. 11, pp 33–39.
Developing Technology

Table of Contents

Norman De Lue

Combinatorial chemistry moves beyond pharmaceuticals

A new technology is redefining the way materials and catalysts are discovered and developed.

Even today, industrial catalysts, inorganic materials, polymers, and chemicals are being developed using extensive compound preparation, followed by evaluation in reactors or test equipment. Usually, candidates are prepared and evaluated one at a time; on average, one compound is prepared and evaluated per day. This makes the overall process labor-intensive, time-consuming, and expensive.

Table 1. Potential cost and time savings of combinatorial chemistry—A hypothetical case

One-at-a-time approach
Combinatorial approach
Number of employees
Annual budget
$2 million
$2 million
Catalysts evaluated/year
Average preparation and evaluation cost/catalyst
Development time
2–10 years
0.5–2 years
Combinatorial chemistry, a new strategy for greatly reducing the time for the discovery of chemical substances, has been developed over the past few years (Table 1). Combinatorial chemistry is a method for creating vast numbers of molecular substances, then rapidly testing them for desirable properties. Almost every pharmaceutical company uses this technique because of the potential for immense savings in time and money, and numerous drug candidates currently in clinical trials are products of this methodology. With few exceptions (1), most combinatorial chemistry discussed in the literature and at conferences is devoted to the pharmaceutical- and life-science industries, but this enabling technology may also be applied to nonpharmaceutical chemicals.

Combinatorial chemistry concepts
The first step in combinatorial chemistry is to produce a large collection of chemical substances known as the combinatorial library. Next, the combinatorial library is rapidly evaluated to find a desirable property using a technique known as high-throughput screening (2, 3). Credit for the concept is often given to Mario Geyson, who published the first report of combinatorial chemistry in 1984; in Geyson’s work, a combinatorial library of peptides was synthesized and then evaluated en masse as viral antigens (4).

The concept can be illustrated as follows: Starting with three pools of amino acids with each pool containing 30 amino acids, it is possible to carry out couplings in which a statistical mixture of 27,000 tripeptides is obtained (Figure 1). The entire combinatorial library is tested with a high-throughput screening method to determine whether the desired biological activity is present. If no activity is observed, all 27,000 possible tripeptides are considered inactive. If there is a biological response, it is conceivable that only one tripeptide is active, and the job is to identify which one it is. A strategy for identifying the active component is called deconvolution (3).

In positional scanning, a new library is created in which one of the known amino acids from Pool 1 is omitted. The new combinatorial library will contain 29 × 30 × 30 = 26,100 tripeptides. If the new library is inactive, the omitted amino acid must have been located in position A of the active tripeptide; otherwise, the omitted amino acid is not in position A. Omission of one amino acid at a time requires a maximum of 30 experiments to define the amino acid in position A of the tripeptide. Similarly, it will take 30 additional experiments to define position B and another 30 to define position C. A maximum of 90 omission experiments is required to define the complete tripeptide sequence of the active compound. Hence, in fewer than 100 experiments, 27,000 tripeptides can be prepared and assayed—significant savings compared with synthesizing and evaluating each tripeptide individually.

Great strides have been made in refining the combinatorial chemistry approaches. Robotics and other forms of automation have greatly increased the speed and economies of generating large libraries. Generally, two strategies are used for generating combinatorial libraries (2, 3). Parallel synthesis (Figure 2) is similar to traditional one-step-at-a-time synthesis, but robotics allows a certain reagent to be placed at specified locations at the same time or sequentially. Note that 18 separate reagent additions were required in the parallel synthesis scheme of Figure 2. The power of the parallel synthesis method is that the composition of each member of the library is known.

In split–mix synthesis, a mixture of compounds is made concurrently in the same reactor, and the chemical identities are lost. In the split–mix scheme of Figure 3, eight separate additions were required. As the library becomes larger, the split–mix synthesis scheme requires far fewer steps to generate the library. For example, in a library consisting of calcium, palladium, and six other metals at 10 concentrations, the parallel approach would require 6000 additions, but the split–mix methodology would require only 80. The penalty for the split–mix method is that the identity of the metals on each catalyst bead is lost. Deconvolution would be required to identify the best catalyst after evaluating the entire library. The utility of the split–mix library is that it can give a quick “yes or no” answer for a more easily generated large number of chemical compositions.

Encoding is a technique for maintaining identification of the active components in combinatorial libraries. During the library synthesis, each site is tagged using techniques that allow unique identification, and therefore the chemical composition, at the site. In one example, individual microchips that emit different radio-frequency signals are encapsulated with a polymer, to which potential chemical candidates may be attached (2). In another example, nanocrystals that fluoresce in various colors are incorporated into the synthesis (5). When the library is screened, the identities of the chemical components of active sites can be traced back by their unique radio frequencies or fluorescent colors.

Moving beyond pharmacology
In the materials science area, combinatorial methods have been used to generate large libraries that led to the discovery of superconductors, phosphors, and electronic materials (6–12). The first use of combinatorial methods for discovery of a cobalt oxide magnetoresistance material was reported in 1995 (12). Radio-frequency sputtering and photolithography masking techniques were used to deposit metal oxides on inorganic substrates. Eventually, the method was used to produce metal combinatorial libraries with up to 25,500 compositions on 3-in. silicon wafers. Irradiating the entire library of phosphorescent materials with UV light made screening rapid and easy. A luminescent photograph of the entire library showed the locations of the active materials.

F. M. Menger reported the first application of combina torial chemistry to catalyst discovery in 1995 (13). Poly(allylamine) was functionalized using various ratios of eight carboxylic acids. The polymers were then used as ligands for the Fe(III)-catalyzed hydrolysis of bis(4-nitrophenyl)phosphate. Although significant rate enhancements were found with different ratios and combinations of carboxylic acids, the random synthesis scheme made it impossible to identify the most active catalyst using deconvolution. Other early applications of combinatorial chemistry to homogeneous catalysis are more akin to techniques used in the life sciences (1, 14). For example, optimal combinations of chiral ligands and metal salts in various solvents have been determined for catalyzing a distereoselective carbene insertion reaction to produce chiral indoyl derivatives. Also, a combinatorial approach has been used to identify trimeric ligands, made from amino acids and aldehydes, which produce optically active cyanohydrins using a titanium-catalyzed reaction with trimethylsilylcyanide.

R. C. Willson and co-workers are credited with the earliest publication, in 1996, about the application of combinatorial chemistry to heterogeneous catalysis (15). They presented a paper that was essentially a “proof of concept” for hydrogen oxidation (Figure 4). Sixteen γ-Al2O3 pellets were impregnated individually with 16 metal solutions. For the combinatorial library, the individual pellets were arranged in a 4 × 4 array on a support and placed in a reactor. The reactor was equipped with an IR-transparent sapphire window, and an IR imaging camera was used to measure the temperatures of the pellets. At the start of the experiment, all 16 pellets were held at 35 °C. A stream of hydrogen was introduced, and the pellets remained at the same temperature. When oxygen was introduced, the three pellets containing iridium, palladium, and platinum ignited. The temperature of these three pellets rose to ~125 °C, as indicated by a color change. A fourth pellet containing rhodium was less active and burned at a temperature of 82 °C under the stream of hydrogen and oxygen. The authors recognized and discussed the limitations of this simplified combinatorial chemistry approach. The use of the thermographic method provides information on heat release, but it does not provide much information on selectivities where competing reactions could be taking place.

In a 1997 patent, Willson described the use of robotic techniques to produce libraries of catalysts on the walls of honeycomb structures containing dozens of channels per square inch (16). High-throughput screening methods were suggested for running the catalysts all at once and analyzing the results with a variety of sensor techniques. Detectors such as IR could provide information on the identities and concentrations of reaction products. The method described in this patent could be used to screen large numbers of catalytic materials with much less labor and time than conventional methods require.

J. P. Morken and S. J. Taylor used beads to catalyze an acylation reaction in solution, and described a similar thermographic technique (17). A library containing 3150 statistically distinct catalysts was used en masse to catalyze reaction of ethanol and acetic acid to produce ethyl acetate. The reaction mixture was monitored with an IR camera that was able to distinguish “hot” beads by observing the increase in temperature.

Developing the instrumentation
T. E. Mallouk and co-workers used a novel combinatorial chemistry approach to discover an optimized quaternary alloy for use as an anode in a methanol–air fuel cell (18). They modified an ink-jet printer by replacing the ink reservoirs with solutions of platinum, ruthenium, osmium, rhodium, and iridium and “printing” metal salts on a piece of conducting carbon paper (Figure 5). The “printed” catalyst spots comprised a 645-member combinatorial library containing the 5 individual metals, as well as 80 binary, 280 ternary, and 280 quaternary combinations of metals. When the entire library was immersed in a methanol solution and current was applied, the most active areas of the library could be identified by an optical detection method using a fluorescent acid–base indicator. A quaternary alloy consisting of 44% Pt, 41% Ru, 10% Os, and 5% Ir was identified that permitted ~40% higher current density than the best commercial material (50% Pt, 50% Ru).

S. M. Senkan described another demonstration of high-throughput screening for dehydrogenating cyclohexane to benzene (19). Platinum, palladium, and rhodium catalysts were placed in an 8 × 9 array on a 7.5 × 7.5 cm2 square support. The catalyst library was put in a heated reaction chamber into which cyclohexane was introduced. A laser beam, tuned to a frequency that would ionize the benzene product, was directed with mirrors so that it passed over all the catalyst sites. A microelectrode at each catalyst site detected a photoionization spectrum that allowed the most active catalyst to be pinpointed.

W. H. Weinberg, chief technical officer at Symyx Technologies (Santa Clara, CA), has led a major effort to develop and commercialize combinatorial chemistry in nonpharmaceutical technologies. Symyx researchers have patented methods using the combinatorial approach to synthesize and screen libraries of supported and unsupported catalysts (20). The method was demonstrated by preparing organometallic catalysts and using them to polymerize olefins such as ethylene.

The company claims that libraries containing from 10 to >1 million components are possible. Its patent claims the use of high-speed parallel and sequential synthesis and high-speed screening of the catalysts. Symyx has coined the term VLSIPS for “very large scale immobilized polymer synthesis”. It has also developed the technology for high-throughput characterization of polymers by techniques such as molecular weight distribution. Such methods promise to make it possible to screen hundreds or even thousands of potential polymer catalysts rapidly.

Symyx described technology for automated synthesis of thin-film inorganic libraries of up to 26,000 compositions on a 3-in. silicon wafer. The company also described a fully automated scanning mass spectrometer and has demonstrated the uses of this technology for heterogeneous catalysis (21, 22). In one example, molybdenum, niobium, and vanadium were dispersed on quartz substrates with an automated dispensing machine. An arrangement of quartz capillary tubing, controlled using computer software, scanned the library to feed ethane sequentially to each catalyst element while samples of the ethylene product were transported to the mass spectrometer (Figure 6).

Weinberg reported that up to 10,000 catalysts could be evaluated per month using this method (22). To further demonstrate the utility of the combinatorial chemistry techniques, syntheses for catalyst compositions that showed promise in the thin-film tests were scaled up. A 48-barrel fixed-bed reactor was used with ~20 mg each of bulk catalysts to confirm and optimize the results from the thin-film studies.

Where do we go from here?
Combinatorial chemistry is just beginning to be used to discover nonpharmaceutical materials, catalysts, and chemicals. The maximum size and speed of evaluation for combinatorial libraries are two of the major technical considerations involved.

One challenge to preparing combinatorial libraries of inorganic materials and catalysts rapidly is trying to automate complicated, multistep procedures. These procedures often require choices of several chemical elements, a wide variety of supports, washing, reduction, oxidation, and calcination steps that may be difficult to carry out in a fast, automated sequence.

For some techniques described in this article, thin films are deposited on substrates. This brings up the question of whether materials generated in micro quantities or as thin films on special substrates will perform the same way when scaled up to where bulk properties become important. It is generally accepted with catalysts and certain inorganic materials that surface structure, chemical composition, defect sites, metal–support interactions, mass transport, and mode of synthesis play critical roles in material properties.

Perhaps one of the most challenging aspects in applying combinatorial chemistry to nonpharmaceutical chemicals is the scarcity of current methods to assay a library of inorganic materials or catalysts rapidly. In assaying biological activity, many examples exist in which libraries in the thousands can be evaluated all at once and activity deduced at the micro- or nanogram level. Corresponding techniques have not been well developed for catalytic, polymeric, and inorganic materials.

It is not clear how a library of catalysts could be screened all at once for such properties as selectivity or conversion to a desirable product. For example, if propylene and oxygen were passed over a library of 1000 potential catalysts for propylene oxide formation, how would it be ascertained which catalyst or catalysts show selectivity to propylene oxide? Which ones show the best conversion? Even a relatively fast 1-min GC technique would need almost 17 h to evaluate a 1000-member library. Moreover, the best catalyst under one set of reaction conditions (temperature, pressure, and time) might not be the best catalyst under another set of reaction conditions.

Bayer, Hoechst, Ciba Specialties, and Goodrich have paid millions of dollars to form alliances with Symyx. SRI, Combi Chem, and Avantium are also seeking out industrial collaboration for their combinatorial programs. UOP, DuPont, Kodak, Lucent Technologies, Dow, Celanese, and Shell reportedly are beginning in-house programs. These major chemical companies appear to have concluded that, although there may be unknowns in the current ability to synthesize and evaluate large nonpharmaceutical libraries, there are no intrinsic reasons why the success in life sciences cannot be duplicated.


  1. De Lue, N. R.; McGuffey, A. M. Arabian J. Sci. Eng. 2000, 25 (2A), 73–88.
  2. Wilson, S. R; Czarnik, A. W. Combinatorial Chemistry—Synthesis and Application; Wiley & Sons: New York, 1997.
  3. Czarnik, A. W.; DeWitt, S. H. A Practical Guide to Combinatorial Chemistry; American Chemical Society: Washington, DC, 1997.
  4. Geyson, H. M.; Meloen, R. H.; Barteling, S. J. Proc. Natl. Acad. Sci. U.S.A. 1984, 91, 3998–4005.
  5. Henry, C. M. Chem. Eng. News 2001, 79 (29), 24–25.
  6. Schultz, P. G.; Xiang, X.-D. Curr. Opin. Solid State Mater. Sci. 1998, 3, 153–158.
  7. Xiang, X.-D.; Sun, X.; Briceno, G.; Chang, H.; Schultz, P. G. Science 1995, 270, 273–275.
  8. Schultz, P. G., Xiang, X.-D.; Goldwasser, I. U.S. Patent 5,776,359, 1998.
  9. Xiang, X.-D.; Sun, X.; Briceno, G.; Lou, Y.; Wang, K.-A.; Chang, H.; Wallace-Freedman, W. G.; Chen, S.-W.; Schultz, P. G. Science 1995, 268, 1738–1740.
  10. Danielson, E.; Golden, J. H.; McFarland, E. W.; Reaves, C. M.; Weinberg, W. H.; Wu, X. D. Nature 1997, 389, 944–948.
  11. Sun, X.-D.; Gao, C.; Wang, J.; Xiang, X.-D. Appl. Phys. Lett. 1997, 70, 3353–3355.
  12. Wang, J.; Yoo, Y.; Gao, C.; Takeuchi, I.; Sun, X.; Chang, H.; Xiang, X.-D.; Schultz, P. G. Science 1998, 279, 1712–1714.
  13. Menger, F. M.; Eliseev, A. V.; Migulin, V. A. J. Org. Chem. 1995, 60, 6666–6667.
  14. Weinberg, W. H.; Jandeleit, B.; Self, K.; Turner, H. Curr. Opin. Solid State Mater. Sci. 1998, 3, 104–110.
  15. Moates, F. C.; Somani, M.; Annamalai, J.; Richardson, J. T.; Luss, D.; Willson, R. C. Ind. Eng. Chem. Res. 1996, 35, 4801–4803.
  16. Willson, R. C. World Patent PCT WO 97/32208, 1997.
  17. Taylor, S. J.; Morken, J. P. Science 1998, 280, 267–270.
  18. Reddington, E.; Sapienza, A.; Gurau, B.; Viswanathan, R.; Saranga pani, S.; Smotkin, E. S.; Mallouk, T. E. Science 1998, 280, 1735–1737.
  19. Senkan, S. M. Nature 1998, 394, 350–353.
  20. Weinberg, W. H.; McFarland, E.; Goldwasser, I.; Boussie, T.; Turner, H.; Van Beek, J.A.M.; Murphy, V.; Powers, T. U.S. Patent 6,248,540, 2001.
  21. Cong, P.; Doolen, R. D.; Fan, Q.; Giaquinta, D.; Guan, S.; McFarland, E.; Poojary, D. M.; Turner, H. W.; Weinberg, W. H. Angew. Chem., Int. Ed. 1999, 38, 484–488.
  22. Weinberg, W. H.; McFarland, E. W.; Cong, P.; Guan, S. U.S. Patent 5,959,297, 1999.

Norman De Lue is a freelance writer based in Houston, TX (281-492-8042; He received his B.A. in chemistry at the State University of New York at Buffalo and his Ph.D. in chemistry from Purdue University. He has worked in R&D for a number of companies in the fields of new product and process development. His current professional interests include the use of technical and managerial approaches to increase efficiency in R&D.

cartoon: "Hank just loves concocting new mixtures of cleaning products. I like to think of him as a combinajanitorial chemist."
"Hank just loves concocting new mixtures of cleaning products. I like to think of him as a combinajanitorial chemist."
Return to Top || Table of Contents