Surface Science of Single-Site Heterogeneous Olefin Polymerization Catalysts

PCA ETR:REVIEW FEATURE: SPECIAL

Surface science of single-site heterogeneous olefin polymerization catalysts

Seong H. Kim* and Gabor A. Somorjai†‡

*Department of Chemical Engineering, Pennsylvania State University, University Park, PA 16802; and †Department of Chemistry, University of California, Berkeley, CA 94720

Edited by Tobin J. Marks, Northwestern University, Evanston, IL, and approved June 7, 2006 (received for review April 3, 2006)

This article reviews the surface science of the heterogeneous olefin polymerization catalysts. The specific focus is on how to prepare and characterize stereochemically specific heterogeneous model catalysts for the Ziegler–Natta polymerization. Under clean, ultra- high vacuum conditions, low-energy electron irradiation during the chemical vapor deposition of model Ziegler–Natta catalysts can be used to create a ‘‘single-site’’ catalyst film with a surface structure that produces only isotactic polypropylene. The polymerization activities of the ultra-high vacuum-prepared model heterogeneous catalysts compare well with those of conventional Ziegler–Natta catalysts. X-ray photoelectron spectroscopic analy- Fig. 1. Schematic description of the polymerization process for a metal- ses identify the oxidation states of the Ti ions at the active sites. locene catalyst system (a) and a Ziegler–Natta catalyst system (b). Cp, cyclo- Temperature-programmed desorption distinguishes the binding pentadiene; Me, methyl; Et, ethyl; MAO, methylaluminoxane. strength of a probe molecule to the active sites that produce polypropylenes having different tacticities. These findings dem- side groups with respect to the polymer backbone. If the side onstrate that a surface science approach to the preparation and groups are ordered in a single orientation with respect to the characterization of model heterogeneous catalysts can improve polyolefin backbone, the polymer structure is called isotactic; if the catalyst design and provide fundamental understanding of the they are randomly distributed along the polymer chain backbone single-site olefin polymerization process. structure, it is called atactic (Fig. 2). There are two orders of magnitude difference in the hardness (or elastic modulus) Ziegler–Natta between these two polymers at the same molecular weight (Table 2). Thus, the challenge in polyolefin synthesis is to prepare a atalysts can function by being dispersed on solid sur- polymerization catalyst that produces polyolefin with only one faces (heterogeneous catalysis) or dissolved in reaction type of structure and precise control of the molecular weight. media (homogeneous catalysis). Heterogeneous cata- These are called single-site catalysts. lysts are widely used in the chemical industry because In the case of homogeneous metallocene catalysts, the mo- theyC are in general easy to handle, separate, and recycle. lecular engineering of the organic ligands attached to the active Homogeneous catalysts are used in synthesis of specialty chem- metal ion is used for control of the stereochemistry during the icals, offering precise control of molecular structure and reac- polymerization (5, 6). For certain symmetric arrangements of the tivity. Combining the merits of these two different catalyst large cyclic ligands attached to the active metal ion, the mono- systems is of great interest for production of next-generation catalysts (1). In this article, we review the surface science of the mer molecules approaching the reaction center are added to the heterogeneous olefin polymerization. polymer chain in a specific geometry. In general, catalysts In the polymerization of small olefins such as ethylene and exhibiting certain Cs symmetries frequently produce syndiotactic propylene, both heterogeneous and homogeneous catalytic sys- polymers, whereas C2-symmetric catalysts typically produce iso- tems are operational (Fig. 1). Heterogeneous olefin polymer- tactic polymers. ization catalysts (so-called Ziegler–Natta catalysts) are used for In the case of heterogeneous Ziegler–Natta catalysts, it is very production of more than two-thirds of the commodity polyole- challenging to produce single-site catalysts. The current gener- fins consumed in the world (2, 3). Recently, a large number of ation of Ziegler–Natta catalysts is composed of TiCl4 species specialty polyolefins have been produced with homogeneous supported on high-surface-area magnesium chloride or magne- metallocene catalysts (4). Whereas most industrial heteroge- sium ethoxide. This is the most productive form of the catalyst, neous catalysts producing polyethylene and polypropylene are evolved after several generations of formulations (7–10). The titanium chloride-based catalysts, the homogeneous metal- monomer molecule adsorbs on the catalyst and reacts at the locene catalysts are organometallic compounds of Ti, Zn, and Hf active catalyst site, preformed by reaction with Et3Al. However, metals in organic solvents. In both types of catalysts, the catalytic the chlorine and growing polymer chain ligands attached to the species are activated with alkyl aluminum cocatalysts to create active site cannot control the orientation of monomer molecules the active sites for carbon–carbon bond formation. Triethyl added to the polymer chain. Thus the polymer produced con- aluminum (AlEt3) is widely used for heterogeneous Ziegler– tains a mixture of atactic and isotactic components. The stereo- Natta catalysts, whereas methylaluminoxane is typically used for chemical control of the heterogeneous Ziegler–Natta catalyst homogeneous metallocene catalysts. The polymers produced with these catalysts can have a wide range of mechanical properties depending on how many mono- Conflict of interest statement: No conflicts declared. mers are connected (molecular weight) and how they are con- This article is a PNAS direct submission. nected (microstructure). The mechanical properties of polyeth- Abbreviations: UHV, ultra-high vacuum; XPS, x-ray photoelectron spectroscopy; TPD, ylene significantly vary with the linear and branching ratio of the temperature-programmed desorption; Cp, cyclopentadienyl. polymer backbone (Table 1). In the case of polypropylene, the ‡To whom correspondence should be addressed. E-mail: [email protected]. mechanical properties depend strongly on the ordering of methyl © 2006 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0602346103 PNAS ͉ October 17, 2006 ͉ vol. 103 ͉ no. 42 ͉ 15289–15294 Downloaded by guest on October 1, 2021 Table 1. Comparison of density and tensile strengths of various Table 2. Comparison of mechanical properties of atactic and polyethylene grades isotactic polypropylenes Tensile Elastic modulus, Hardness, Polyethylene strength, Polypropylene GPa MPa grade Density, g/cm3 MPa Isotactic 1.09 125 High density 0.941–0.965 43 Atactic 0.15 1.4 Low density 0.910–0.925 24 Linear low density 0.900–0.939 37 the solid surface. This approach is taken in this surface science study. The homogeneous metallocene catalysts typically consist can be significantly improved by adding proper Lewis bases of two cyclopentadienyl (Cp) ligands and two chloride ligands (organic ‘‘modifiers’’) during the catalyst preparation or poly- (Fig. 1a). The Cp ligands provide a rigid framework housing the merization. However, their precise roles and the structure of the metal ion and the chloride ligands to create a space for monomer modified active sites are not fully understood at the molecular approach and polymer chain growth. One chloride ligand is level. This imposes difficulties in further improvement of the removed after the activation͞alkylation with the cocatalyst. The isotacticity of heterogeneous Ziegler–Natta catalysts. stereospecificity needed for isotactic polypropylene synthesis is In this article, we describe the use of surface science to rendered by the specific symmetry in the arrangement of the Cp fabricate and characterize single-site model Ziegler–Natta cat- ligands and the side groups attached to the Cp ligands. The bulky alysts for polymerization of ethylene and propylene. The mo- counteranions can amplify the effects of the framework sym- lecular structure of the homogeneous metallocene catalyst in- metry character housing the active cation. dicates that the stereospecific single-site heterogeneous catalyst Compared with the structures of metallocene catalysts, the should have an open and particularly symmetrical arrangement metal ions at the MgCl2, TiCl2, and TiCl3 solid surfaces are of ligands around the active metal ion at the catalyst surface. densely packed with chloride ions (Fig. 1b). For these com- Irradiation by the low-energy electron during the chemical vapor pounds, the thermodynamically most stable surface is the (001) deposition of model Ziegler–Natta catalyst appears to produce basal plane of the rhombohedral crystal structure (9, 11). This a ‘‘single-site’’ catalyst film with a surface structure that forms surface is terminated with a close-packed chloride layer and the only isotactic polymer chains. The polymerization activities of metal ion is shielded beneath this chloride layer (12–14). Be- the ultra-high vacuum (UHV)-prepared model heterogeneous cause of this structure, the (001) surface has a very low reactivity catalysts compare well with those of the conventional Ziegler– and is difficult to activate for polymerization (15, 16). The next Natta catalysts. most stable surface structures are the (100) and (110) planes of A combination of various surface science techniques are used the rhombohedral crystal structure. Although the metal ion is for characterization of the surface composition, structure, and partially exposed in these planes, the chloride ion arrangement oxidation state of these model catalysts. X-ray photoelectron around the metal ion does not have the correct symmetry to spectroscopy (XPS) can identify the oxidation states of Ti ions produce isotactic polypropylene (11). The thermodynamics of at the active sites. Temperature-programmed desorption (TPD) the chloride compound structure therefore render it difficult to can distinguish the binding strength of probe molecules to the obtain the open and specific ligand structure around the active active sites that produce polypropylene with different tacticities. metal ion needed for preparation of the single-site polymeriza- These findings are of great benefit to polymerization science and tion site. technology. Although the grafting of homogenous single-site catalysts on high surface area supports is a powerful way of Synthesis of Heterogeneous Ziegler–Natta Model Catalysts Under UHV heterogenizing metallocene catalysts (1), we suggest that single- Conditions. To create a solid surface with metal ions surrounded site Ziegler–Natta systems can be prepared by modifying the by more loosely packed anions in a low-symmetry arrangement, surface structure and composition of the heterogeneous cata- we used charged particles such as electrons to overcome this lysts that are used at present in large-scale olefin polymerization. difficulty (17–19). These charged particles can be generated in a Methods separate source in UHV and impinged on the catalyst surface. Because of their high cross-sections in interactions with the Strategy for Preparation of Single-Site Heterogeneous Ziegler–Natta condensed phase, these charged particles interact mostly with Catalysts. For preparation of single-site heterogeneous catalysts, surface species. In the case of electrons, the surface chloride ions a widely studied approach is to covalently attach homogeneous are removed by electron-induced desorption (20). polymerization catalyst species to surface functional groups such We have prepared two Ziegler–Natta model catalysts with and as hydroxyl groups on solid supports (1). An alternative way is without electron irradiation and compared the polymerization to use the molecular structure of the homogeneous polymeriza- activity and product stereochemistry. The first model system is tion catalyst as a guide and produce similar structures directly on a thin film of TiClx͞MgCl2, which mimics the structure of typical heterogeneous Ziegler–Natta catalyst. The TiClx͞MgCl2 thin film can be produced by simultaneous deposition of metallic Mg and TiCl4 on an Au substrate (Fig. 3 Left). Partial reduction of TiCl4 to TiClx and oxidation of Mg to MgCl2 takes place during this reaction (21, 22). The low solubility of TiClx in MgCl2 leads to the formation of a TiClx monolayer on top of MgCl2 multi- layers. The oxidation states of the titanium species have a distribution of 4ϩ,3ϩ, and 2ϩ. When these films are irradiated with low-energy electrons and ions, some fraction of surface chlorine ions are desorbed into vacuum, leaving undercoordinated metal ions at the surface (17, 20). However, these Fig. 2. Molecular structures of isotactic (Upper) and atactic (Lower) defective surfaces are readily converted to the stable structure by polypropylene. diffusion of chloride ions from the underlayers (20).

15290 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0602346103 Kim and Somorjai Downloaded by guest on October 1, 2021 molecules per cm2 catalyst per s for propylene and Ϸ5–9 ϫ 10Ϫ7 2 gC2H4 molecules per cm catalyst per s for ethylene. These polymerization rates correspond to nominal turnover frequen- 14 2 cies of Ϸ3–6 ϫ 10 C3H6 molecules per cm catalyst per s and 16 2 Ϸ1–1.8 ϫ 10 C2H4 molecules per cm catalyst per s, respec- tively. The ethylene polymerization rate is Ϸ30 times faster than the propylene polymerization rate on the same catalyst. If the number of the active sites is assumed to be only 10% of the surface Ti ions (Ϸ1 ϫ 1013 sites per cm2), then the turnover frequency (the number of monomers reacted per site) is esti- mated to be Ϸ30–60 C3H6 molecules per s and Ϸ1,000–1,800 C2H4 molecules per s. These polymerization activities of the model catalysts can be

Fig. 3. TiClx͞MgCl2 (Left) and TiCly (Right) model catalyst films. The TiClx͞ compared directly with those of industrial catalysts. Typical 12 2 MgCl2 film is produced by simultaneous dose of Mg (flux ϭϷ6 ϫ 10 atoms industrial catalysts have a surface area of Ϸ50 m ͞g. The 2⅐ ϭ ϫ Ϫ7 per cm s) and TiCl4 (pressure 2 10 Torr) on an Au substrate held at 300 polymerization activity of the model catalyst calculated for a ϭ ϫ 14 K. The TiCly film is produced by electrons irradiation (flux 1 10 electrons 1-cm2 surface area would correspond to Ϸ75 g polypropylene per 2⅐ ϭ per cm s) at an Au substrate (100 K) during exposure to TiCl4 vapor (pressure g catalyst per hr⅐atm and Ϸ1,400 g polyethylene per g catalyst per ϫ Ϫ7 1 10 Torr). hr⅐atm for catalysts with a surface area of 50 m2͞g. Industrial catalysts produce Ϸ100–500 g polypropylene per g catalyst per ⅐ Ϸ The second model system uses electrons, instead of Mg, to hr atm and 2,000–10,000 g polyethylene per g catalyst per ⅐ produce more chloride deficiency (Fig. 3 Right). In this method, hr atm (9). The fact that the model catalysts exhibit activities the substrate is continuously irradiated by an electron beam with similar to the industrial catalysts suggests that the surface properties of the model catalysts prepared in UHV are relevant an energy of Ϸ500–1,000 eV during the TiCl4 exposure. The to those of industrial catalysts (25, 26). transiently adsorbed TiCl4 molecules on Au are dissociated upon interaction with electrons and form chloride-deficient species. Single-site propylene polymerization. The stereochemistry of the two ͞ These species are eventually accumulated into a TiCly film. model catalysts (TiClx MgCl2 and TiCly) is compared for pro- Angle-resolved XPS analysis indicates that the TiCly film consists pylene polymerization. The former represents the conventional 4ϩ of a monolayer of Ti species (chemisorbed TiCl4) on top of Ziegler–Natta catalyst. The latter is produced by the electron TiCl2 layers (19). beam irradiation method to mimic the open ligand structure of the metallocene catalysts. When used for propylene polymer- Results and Discussion ization, the TiClx͞MgCl2 model catalyst produces a mixture of Polymerization Activity and Single-Site Propylene Polymerization atactic and isotactic polypropylenes, whereas the electron- Test. Polymerization activity. The polymerization activity of thin film irradiated TiCly model catalyst produces exclusively isotactic catalysts having a nominal surface area of Ϸ1cm2 is monitored polypropylene (27). Fig. 5 shows the topographic images and the nondestructively and continuously by using a laser reflection C-C chain helix vibrational peaks of the as-grown polypropylene interferometry technique (Fig. 4) (23, 24). In this method, a laser films for TiClx͞MgCl2 and TiCly. The polypropylene film on beam is reflected from the catalyst surface. As the polymer layer TiCly is much rougher than that on TiClx͞MgCl2, implying the grows, the beam reflected from the polymer͞gas interface presence of crystalline domains of isotactic polypropylene. In the interferes with the beam reflected from the catalyst͞polymer infrared spectroscopic analysis, the crystalline isotactic polypro- interface. The periodic interference fringes can be used to pylene shows a characteristic isotactic helix vibration peak at 998 calculate the thickness of the growing polymer film, which can cmϪ1. This peak is much more prominent for the polypropylene then be converted to the monomer consumption rate. Once film grown on TiCly than that on TiClx͞MgCl2. In solvent activated by a brief exposure to AlEt3 vapor, both TiClx͞MgCl2 extraction experiments, the atactic polypropylene fraction was and TiCly model catalysts can polymerize ethylene and pro- negligible for films grown on TiCly, whereas the film grown on pylene. The initial polymerization rates of the TiClx͞MgCl2 TiClx͞MgCl2 contains a large amount of atactic polypropylene. Ϫ8 model catalysts synthesized in UHV are Ϸ2–4 ϫ 10 gC3H6 These results indicate that the TiCly catalyst produced by the

Fig. 4. Laser reflection interferometry experiment showing growth of polyethylene film. (Left) Schematic description of a laser reflection interferometry experiment. (Center) Recorded data for a laser reflection interferometry experiment. (Right) Growth of a polyethylene film as a function of time during the ethylene polymerization on a TiClx͞MgCl2 model catalyst. Polymerization was performed with 900 Torr of ethylene. The reactor temperature was kept at 340 K.

Kim and Somorjai PNAS ͉ October 17, 2006 ͉ vol. 103 ͉ no. 42 ͉ 15291 Downloaded by guest on October 1, 2021 Fig. 5. Atomic force microscopy (AFM) and infrared spectroscopy analysis Fig. 6. Ti 2p3͞2 XPS spectra of the TiCl ͞MgCl (Left) and TiCl (Right) results for the polypropylene ﬁlms produced with TiCl ͞MgCl ͞Au (Left) and x 2 y x 2 catalysts that produced the polypropylene of Fig. 5. TiCly͞Au (Right) catalysts. The AFM image size is 20 ϫ 20 ␮m. The height full scale is 2.1 ␮m. found that the polymerization reactivity is not correlated with the Ti3ϩ concentration. electron-induced TiCl4 deposition is a single site from a stereochemistry point of view, producing only isotactic polypro- Surface structure of the model catalysts. The adsorption-site distribution on the model Ziegler–Natta catalyst surfaces can be pylene, whereas the TiClx͞MgCl2 catalyst produced by codepo- determined from TPD of an inert probe molecule (34, 35). A sition of Mg and TiCl4 contains multiple active sites and produces a mixture of atactic and isotactic polypropylene. good probe molecule is mesitylene, which weakly adsorbs on the catalyst surface and desorbs at Ϸ190–300 K without altering the Relationship Between Active-Site Properties and Polymerization Be- catalyst surface. Fig. 7 displays the mesitylene TPD results for a haviors. Being able to synthesize the well characterized model well characterized MgCl2 support film produced by thermal catalyst gives an unprecedented opportunity to study the corre- evaporation of MgCl2 on Au (20). From structural information Ϸ lation between the surface properties of the catalyst and the on the MgCl2 film (12–14), the 200-K desorption site is polymerization activity and stereochemistry. What are the oxi- attributed to the (001) basal plane structure where the chloride dation states of the active sites? What controls the stereochem- ions at the outermost layer are close-packed and the metal ions istry of the propylene polymerization? These questions are under the chloride layer are coordinated to six chloride ions. The studied with XPS and probe-molecule TPD. high-temperature desorption peak can be attributed to the Oxidation state of the activated Ti species. The electronic state of the nonbasal planes or defects in the ionic lattice of the basal plane polymerization-active Ti species can be found in the changes in where the surface chloride ions are not close-packed and the the oxidation-state distribution of the model catalysts before and metal ions beneath the chloride layer are undercoordinated. after the AlEt3 activation (27). Fig. 6 compares the high- The mesitylene TPD can also be used to examine the struc- resolution XPS of the Ti 2p3/2 peak for the TiClx͞MgCl2 and tural distribution of surface site on the model catalyst before and TiCl films. The former represents the conventional Ziegler– after the AlEt3 activation (Fig. 8) (27). The mesitylene TPD for y ͞ Natta system, and the latter is the single-site model catalyst. Both the TiClx MgCl2 model catalysts reveals a surface-site distribu- systems show the same changes after the activation. Upon tion similar to the MgCl2 film–basal plane sites (198 K) and 4ϩ reaction with AlEt3, the Ti peak intensity decreases and the nonbasal plane sites (245 K). In the case of the electron- Ti2ϩ intensity increases significantly. The Ti3ϩ peak intensity shows only a minor change. Because both catalysts have similar oxidation-state distributions but give much different tacticity in propylene polymerization, the Ti oxidation-state distribution does not seem to be an important factor in stereoregularity. The Al peak is not detected in the XPS of the activated catalyst surface in both systems. This result rules out the bimetallic mechanism in which the Al-containing species bonded to the Ti active site is claimed to be responsible for the stereochemistry control. Another important aspect of the XPS results is that the Ti2ϩ species appear to be the active species for polymerization. This result is somewhat contrary to the previous belief that the only Ti3ϩ species are catalytically active (28–32). The Ti3ϩ active species was inferred from ESR studies. However, it should be noted that the Ti2ϩ and Ti4ϩ ions are not ESR active because they are not paramagnetic and Ͼ80% of total Ti3ϩ ions in the Ziegler–Natta catalysts are ESR silent because of interactions with the adjacent Ti3ϩ ions (28, 29). Freund and coworkers (33)

recently used ESR to analyze model Ziegler–Natta catalysts and Fig. 7. TPD of mesitylene adsorbed on an MgCl2 ﬁlm deposited on Au.

15292 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0602346103 Kim and Somorjai Downloaded by guest on October 1, 2021 polymerization (36, 38–45). The active sites originating from alkylation of the undercoordinated Ti2ϩ sites are stereochemically specific, whereas those originating from the close-packed basal plane are stereochemically nonspecific. The alkylation of the open-structure metal ion on the Ziegler–Natta catalyst seems to create the polymerization site structure similar to the activated metallocene catalyst. Conclusions Process optimization based on empirical data for heterogeneous olefin polymerization without a fundamental understanding of the molecular processes for the polymer formation has reached the limit. Further improvements of the olefin polymerization system will require catalyst design and investigation of the molecular mechanism of the polymerization. The surface science results reported here prove the potential of such catalyst design. Although the surface science approach proves the concept of this molecular engineering of catalytic materials, it cannot be used for preparation of bulk materials. Therefore, more intense research on producing molecularly engineered bulk materials is required. The heterogenization of metallocene catalysts on the exterior of inactive solid materials has been a major research direction in this field. Recently, significant progress has been made in anchoring these metallocene catalysts in the meso- porous oxide materials (1). Another direction would be to synthesize an organo-chloride solid materials that combine the Fig. 8. Mesitylene TPD proﬁles for the TiClx͞MgCl2 (Left) and TiCly (Right) catalysts that produced the polypropylene of Fig. 5. Mesitylene exposures function of organic modifiers with 3D chloride networks. An were 0.2, 0.6, 1.0, and 1.4 liters. example is MgCl2–C2H5OH complexes (46). The bonding of olefins at the AlEt3-activated catalytic sites is still unclear. What are the structures of the ethylene and bombarded TiCly catalyst, only one mesitylene TPD peak at 247 propylene molecules initially adsorbed at these activated sites? K is observed. This finding indicates that all Ti species at the What arrangement of the various sites control their activities and catalyst surface are undercoordinated with chloride ions, which stereospecificities? These questions might be answered with is expected to be the desired structure from the comparison with real-space atomic-scale microscopy and surface-specific spec- homogeneous metallocene catalysts. troscopy under reaction conditions. These include high-pressure After activation with the AlEt3 cocatalyst, the desorption scanning tunneling microscopy (STM), atomic force microscopy temperature of the mesitylene probe molecule decreases by Ϸ4 (AFM), sum-frequency-generation (SFG) vibrational spectros- K for the basal plane sites and Ϸ15 K for the undercoordinated copy, extended x-ray absorption fine-structure (EXAFS) spec- sites. These changes of the mesitylene desorption temperature troscopy, etc. STM and AFM can give site-specific information imply chemical and structural changes on the catalyst surface. of various surface sites under the reaction conditions (47). The Because the basal-plane sites are fully coordinated with chloride SFG vibrational spectroscopy has a unique advantage of being ions, the alkylation reaction will be a replacement of one chloride able to probe surface species without interference from gas- ion with a C2H5 group (36, 37). This replacement reaction phase and bulk species (48). EXAFS spectroscopy can provide induces only a minor change in the mesitylene-surface interac- the structural information with precise distance and arrange- ͞ tions. For the undercoordinated sites of TiClx MgCl2 and TiCly, ment of ligands around the metal ion (49). If these techniques are alkylation by AlEt3 can occur by addition of one C2H5 group to combined with the single-site catalyst preparation, it should be the active metal ions, causing a larger decrease in the mesitylene possible to detect reaction intermediates leading to polymer desorption temperature (36, 37). chain growth and chiral orientation. The comparison of the tacticity of the polypropylene produced and the mesitylene TPD of the activated model catalysts pro- This work was supported by the Director, Office of Energy Research, vides direct evidence for a correlation between the structures of Office of Basic Energy Sciences, Materials Science Division of the U.S. the catalyst surfaces and the stereospecificity of propylene Department of Energy under Contract DE-AC03-76SF00098.

1. Hlatky GG (2000) Chem Rev 100:1347–1376. 13. Fairbrother DH, Roberts JG, Somorjai GA (1998) Surf Sci 399:109–122. 2. Chien JCW (1999) Top Catal 7:23–36. 14. Roberts JG, Gierer M, Fairbrother DH, van Hove MA, Somorjai GA (1998) 3. Fink F, Steinmetz B, Zechlin J, Przybyla C, Tesche B (2000) Chem Rev Surf Sci 399:123–128. 100:1377–1390. 15. Magni E, Somorjai GA (1995) Appl Surf Sci 89:187–195. 4. Imanishi Y, Naga N (2001) Prog Polym Sci 26:1147–1198. 16. Magni E, Somorjai GA (1995) Surf Sci 341:L1078–L1084. 5. Coates GW (2000) Chem Rev 100:1223–1252. 17. Magni E, Somorjai GA (1996) Surf Sci 345:1–16. 6. Kaminsky W, Laban A (2001) Appl Catal A 222:47–61. 18. Magni E, Somorjai GA (1996) J Phys Chem 100:14786–14793. 7. Soga K, Shiono T (1997) Prog Polym Sci 22:1503–1546. 19. Kim SH, Somorjai GA (2002) J Phys Chem B 106:1386–1391. 8. Dusseault JJA, Hsu CC (1993) J Macromol Sci Rev Macromol Chem Phys 20. Kim SH, Tewell CR, Somorjai GA (2001) Langmuir 16:9414–9420. C33:103–145. 21. Magni E, Somorjai GA (1998) J Phys Chem B 102:8788–8795. 9. Barbe`PC, Cecchin G, Noristi L (1986) Adv Polym Sci 81:1–81. 22. Kim SH, Somorjai GA (2000) J Phys Chem B 104:5519–5526. 10. Xie T, McAuley KB, Hsu JCC, Bacon, DW (1994) Ind Eng Chem Res 23. Kim SH, Vurens G, Somorjai GA (2000) J Catal 193:171–175. 33:449–479. 24. Kim SH, Somorjai GA (2000) Catal Lett 68:7–11. 11. Colbourn EA, Cox PA, Carruthers B, Jones PJV (1994) J Mater Chem 4:805–812. 25. Soga K, Chen SI, Ohnishi R (1981) Polym Bull 8:473–478. 12. Fairbrother DH, Roberts JG, Rizzi S, Somorjai GA (1997) Langmuir 13:2090– 26. Kashiwa N, Yoshitake J (1984) Makromol Chem 185:1133–1138. 2096. 27. Kim SH, Somorjai GA (2001) J Phys Chem B 105:3922–3927.

Kim and Somorjai PNAS ͉ October 17, 2006 ͉ vol. 103 ͉ no. 42 ͉ 15293 Downloaded by guest on October 1, 2021 28. Fuhrmann H, Herrmann W (1994) Macromol Chem Phys 195: 3509–3521. 40. Guerra G, Pucciariello R, Villani V, Corradini P (1987) Polym Commun 29. Brant P, Speca AN (1987) Macromolecules 20:2740–2744. 28:100–102. 30. Chien JCW, Wu J-C (1982) J Polym Sci Polym Chem Ed 20:2461–2476. 41. Venditto V, Guerra G, Corradini P (1991) Eur Polym J 27:45–54. 31. Chien JCW, Hu Y (1989) J Polym Sci A 27:897–913. 42. Shiga A, Kawamura-Kuribayashi H, Sasaki T (1995) J Mol Catal 98:15–24. 32. Chien JCW, Weber S, Hu Y (1989) J Polym Sci A 27:1499–1514. 43. Puhakka E, Pakkanen TT, Pakkanen TA (1995) Surf Sci 334:289–294. 33. Schmidt J, Risse T, Hamann H, Freund HJ (2002) J Chem Phys 116:10861–10868. 44. Cavallo L, Guerra G, Corradini P (1998) J Am Chem Soc I 120:2428–2436. 34. Kim SH, Somorjai GA (2000) Appl Surf Sci 161:333–339. 45. Boero M, Parrinello M, Terakura K (1998) J Am Chem Soc I 120:2746–2752. 35. Kim SH, Somorjai GA (2001) Surf Interface Anal 31:701–710. 46. Tewell CR, Malizia F, Ager JW, Somorjai GA (2002) J Phys Chem B 36. Arlman E, Cossee PJ (1964) J Catal 3:99–104. 106:2946–2949. 37. Rodriguez LAM, Van Looy HM, Gabant JA (1966) J Polym Sci 47. Somorjai GA, Hwang KS, Parker JS (2003) Top Catal 26:87–99. 4:1905–1916. 48. Somorjai GA, Chou KC, Yang M (2004) e-J Surf Sci Nanotechnol 2:106–118. 38. Cossee PJ (1964) J Catal 3:80–88. 49. Scott SL, Church TL, Nguyen DH, Mader EA, Moran J (2005) Top Catal 39. Arlman EJ (1964) J Catal 3:89–98. 34:109–120.

15294 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0602346103 Kim and Somorjai Downloaded by guest on October 1, 2021