reduction occurs in the Mn cluster upon the S3-to-Sotransition, presuming that the va- lence change of Mn is the main cause for this K-edge energy downshift. The simulat- ed data also indicate that the K-edge energy upshifts by 1.0 eV upon the So-to-Sl tran- sition. The extent of the upshift of the simulated K-edge energy is coincident with those upon the Sl-to-S, and S,-to-S3 tran- sitions, suggesting that an oxidation of Mn(II1) to Mn(1V) might be involved in the So-to-S, transition as well. The present study demonstrates that the valence or structure or both of the Mn cluster oscillates period-four during the cy- cling of the Joliot-Kok's oxygen clock, show- ing directly that the so-called S-states reflect mainly the differences in redox states or configurations or both of the Mn cluster during the four-electron cyclic reaction. REFERENCESANDNOTES 1. P. Joliot, G. Barbieri, R. Chabaud, Photochem. Photobiol. 10, 309 (1 969). 2. B. Kok, B. Forbush, M. McGloin, ibid. 11, 457 (1 970). 3. A. W. Rutherford, Trends Biochem. Sci. 14, 227 (1 989). 4. Govindjee and W. J. Coleman, Sci. Am. 262, 50 (February 1990). 5. D. B. Goodin et a/.,Biochim. Biophys. Acta 767, 209 (1 984). 6. V. K. Yachandra et a/., Biochemistry 26, 5974 (1 987). 7. G. N. George, R. C. Prince, S. P. Cramer, Science 243, 789 (1989). 8. M. Kusunoki et a/., J. Biochem. 108, 560 (1 990). 9. R. D. Guiles etal., Biochemistry29, 486 (1990). 10. G. C. Dismukes and Y. Siderer, Proc. Natl. Acad. Sci. U.S.A.78, 274 (1981). 11. J. P. Dekker, H. J. Van Gorkom, J. Wensink, L. Ouwehand, Biochim. Biophys.Acta 767, 1 (1984). 12. 0 . Saygin and H. T. Witt, ibid. 893, 452 (1987). 13. T. Ono etal., Biochemistry30, 6836 (1991). 14. M. Kusunoki, in Progress in Photosynthesis Re- search, J. Biggins, Ed. (Nijhoff, Dordrecht, 1987), vol. 1, pp. 729-732. 15. G. W. Brudvig and R. H. Crabtree, Proc. Natl. Acad. Sci. U.S.A.83, 4586 (1986). 16. R. D. Guiles etal., Biochemistry29, 471 (1990). 17. C. F. Fowler, Biochim. Biophys. Acta 462, 414 (1 977). 18. S. Saphon and A. R. Crofts, Z. ~aturforsch: Teil C 32, 617 (1 977). 19. P. Marchand and L. Marmet, Rev. Sci. Instrum. 54, 1034 (1983). 20. T. Ono and Y. Inoue, Biochim. Biophys. Acta 806, 331 (1985). 21. W. F. J. Vermass et a/., ibid. 764, 194 (1984). 22. S. Styring and A. W. Rutherford, Biochemistry26, 2401 (1987). 23. G. T. ~abcbck and K. Sauer, Biochim. Biophys. Acta 325, 483 (1 973). 24. Supported by a grant for Photosynthetic Science and Biodesign Research Program (to T.O.) at the Institute of Physical and Chemical Research (RIKEN) by the Science and Technology Agency of Japan and by a grant-in-aid for Scientific Re- search on Priority Areas (04225234) (to T.O.) from the Ministry of Education, Science and Culture. We thank the staff of the Photon Factory for their kind help during our experiments. 22 June 1992; accepted 29 September 1992 Linear Metal Nanostructures and Size Effects of Supported Metal Catalysts loannis Zuburtikudis* and Howard Saltsburgt Nickel metal catalysts composed of nanometer by micrometer strips have been produced with solid-state microfabrication techniques. Thestrips are actually the edges of nickel- catalyst thin films, which are sandwiched between separating support layers, which are also nanometers thick. These linear nanostructuresconstitute well-defined and well-con- trolled catalytic entities that reproduce the size of traditional supported metal clusters in one dimension, thus separating size from total number of atoms in the catalyst. Examination of their catalytic activity showed that they duplicate the behavior of conventionalsupported clusters. A specific rate maximum was observed for the hydrogenolysis of ethane at a nanoscale dimension similar to the cluster size at which the rate is maximum in the case of the supported cluster studies, whereas the hydrogenation of ethylene shows no such size dependency. The results suggest that the surface-to-volume ratio or the number of atoms in the catalytic entity cannot be the source of these size effects and that either support effects or nonequilibrium surface structures are the determining factors. T h e need for maximizing the effective when normalized to unit surface area of the surface area of catalytically active metals in clusters, as well as the selectivities of some heterogeneous systems led to the extensive of them, depend on the overall metal clus- use of highly dispersed metals in the form of ter size under otherwise similar reaction clusters supported on a variety of carrier conditions (1, 2). If one characterizes the materials. The rates of certain reactions, size of the supported metal cluster by an equivalent average diameter, clusters small- Department of Chemical Engineer~ng,University of er than -10 nm exhibit size depen- Rochester, Rochester, NY 14627-0166 dence. The phenomenon is of considerable 'Present address, Eastman Kodak Research Labs, Rochester, NY 14650-2022. technological and scientific . . importance: (i) tTo whom correspondence should be addressed. it suggests that one could produce catalyst structures that show greater normalized (specific) reaction rates or affect the selec- tivity in a favorable way or both (2), (ii) it furnishes data for testing electronic and geometrical models of catalysis (2), and (iii) it can be used to probe chemically the transition from atoms to bulk solids (3). Traditionally, supported metal catalysts are produced through chemical routes (such as impregnation and coprecipitation) . The catalytic activity of the resulting supported clusters is examined, and the reaction re- sults are correlated with theoretical calcu- lations that provide the structural and elec- tronic features of idealized and unsupported metal clusters as a function of their total number of atoms. This number is taken to be representative of the measured average size of the produced clusters and serves as a metaphor for structure. The ~roblems with using supported clusters (nonuniform shapes and sizes, possible existence of ad- ventitious surface species, very difficult di- rect characterization of their surfaces, and failure to distinguish between size and total number of atoms in the cluster) have led to studies with simpler model systems: single crystals and unsupported metal clusters. However, the results from these studies are at best indicative of what may happen on supported clusters. They address only cer- tain issues of the whole problem; for exam- ple, the single crystal studies investigate the role of specific surface structures on the origin of the phenomenon. There is a need for developing a scheme that allows the investigation of the size effects without so many limitations. We describe a new class of supported metal catalysts that are much more carefully defined, illustrate their use in addressing the origin of the size effects, and demonstrate the inappropriateness of the metaphor of the "number of atoms." By vacuum-depositing alternating layers of metal catalyst and support material on a suitable substrate, one can fabricate a lay- ered synthetic microstructure (LSM) whose edge exhibits a compositionally modulated surface that consists of catalyst and support (Fig. 1). The topmost layer is made out of support material so that the only exposed metal catalyst is in the form of strips be- tween those of the support. The thickness of the metal layers can be controlled, is in the nanometer scale, and is comparable to the size of supported metal clusters. Al- though the width of the catalyst strips is of the order of nanometers, the length of the strips is macroscopic (micrometer scale or larger). In effect, the catalyst is created as a linear microstructure that preserves only one dimension in the nanometer range. As a result, the number of connected metal atoms in this linear structure is orders of magnitude greater than in the typical clus- ter and separation of dimension from the SCIENCE VOL. 258 20NOVEMBER 1992