REPORTS to interface the positively charged (0001)-Zn ylindrical nanoring structure, and the plane (top surface) with the negatively charged of the nanobelt are joined by chemical Penn, Science 289, 751(2000) (0001)-O plane(bottom surface), resulting 4. Z W. Pan, ZR. Dai, Z L Wang, Science 291, 1947(2001). as a single entity. The loops that were 5.x Y. Kong, Z L Wang, Nano Lett. 3, 1625(200 neutralization of the local polar charges and in coiled first remained at the growth temperature 6. Energy-dispersive x-ray spectroscopy has a detection reduced surface area, thus forming a loop with for a longer time, resulting in higher crystallinity, not detect an overlapped end(Fig. 4B). The radius of the whereas the ones that wound on later had less purities with a concentration lower than 29 See supporting material on Science Online. during its initial growth, but the size of the loop crystallinity, forming the structure in Fig. 2G, cannot be too small to reduce the elastic defor- with two contrast regions across the width of the he-dimensional oxide nanostructures once formed, it leads to the fastest growth along a mation energy. nanoring. Finally, as the growth time was ex- The total energy involved in the proc tended, the entire nanoring exhibited high-qual stacking fault comes from polar charges, surface area, and ity crystallinity, as shown by the diffraction con- lastic deformation(12). The long-range electro- trast in Fig. 4D. A uniradial and perfectly aligned Recent study has also shown that the presence of a static interaction is likely to be the initial driving coiling is energetically favorable because of the 10. Wurtzite structure has four di force for folding the nanobelt to form the first complete neutralization of the local polar charg. loop on which subsequent growth is based. Cal- es inside the nanoring and the reduced surface terface is type L. Type I and type Ill stacking faults ulations have been made to assess the possibil- area. The entire growth process may have no 11. The local deposition temperature is -200 to 400-C ity of balancing the increased elastic deformation relation to the substrate used for collecting the nergy(AEDeform)by the decreased electrostatic sample. The thinness of the nanoring also pre- energy (AEperdtm) at the initiation of the nanoring vents the determination of its polarity by conver- and are electros atically tatically effective for aligning the oppo- structure(7)(Fig. 4E). If a nanobelt is folded to gent-beam ED(14 The coiling process presented in Fig. 4 12. There are three components of energy involved in the energy gain AE(AE AEDefom AEplectro )is unifies the two types of nanoring structures ositive regardless of the size of the loop, sug. described in Figs. 2 and 3. If the(0001) polar surface area energy gesting that a single-loop complete ring is ener- surface of the Zno nanobelt is parallel to the bonding the loops, and elastic deformation energy due 8 getically unfavorable For a 10% overlapped na- ring plane, self-coiling of the nanobelt at a noring (n=1. 1), the nanoring structure is ener- radial direction of [1210] forms the type I but we separate them here for getically possible if its diameter (D)is larger structure in Fig. 2. Alternatively, if the nano- onvenience of discussion in the text. electrostatic and than 600 nm. It is thus possible to form a single- belt is tilted toward the nanoring center so looped nanoring with overlapped ends at the that the radial direction is [1213](15), self- beginning of the growth caused by fluctuation coiling of the nanobelt produces the type ll (Fig. 4B). For a 33%overlapped nanoring(n= structure in Fig 3. The tilting of the nanobelt 13 Sintening ine case for a thin ane larger than 360 nm. The diameters of the The model presented here can also be adopt are cher sintering"here, we me experimentally observed nanorings are in the ed to explain the helical nanostructure reported ith the same crystal orientation, and there may be range of 0. 8 to 4 um, and no ring has D<0.8 previously (5). If the nanobelt in xial direction as guided by the planar defect, it was um From the SEM images, we believe that flipped by g0 so that the radial direction of the the first step(Fig. 4B) occurs before the na- nanoring is [00ol which means that the polar- noring lands on the substrate ization is in the radial direction, a bending of the anostructure can be as low as one third of its bulk The presence of a planar defect within the nanobelt into a circle slightly reduces the ele nanobelt(Fig. 2F)is likely to be the key to trostatic energy, possibly in favor of forming ar ally join the loops at producing the fast growth of the nanobelt along in-plane spiral nanoring (5). Alternatively, as 200°to400°C [1010], because it lowers the energy in the result of preserved polar charges on the inner and 14. In principle, the polarity of the surfaces can be determined wurtzite-structured lattice (8). Planar defects outer are surfaces, the circular loops of the nano- or more to enhance the may be initiated by impurity atoms, such as Li belt cannot be densely packed into a single growth continues, the nanobelt may be natural- repulsion among them; instead, a helical struc. 15. Among the possible glide systems for hexagonal ly attracted to the rim of the nanoring by elec- ture would be formed, with a pitch distance of 1213 [1213 are the two possible systems, which rostatic interactions and then extend itself par- 200 to 500 nm (much larger than the width(-20 spend tse the oases presented in Figs. 2 a allel to the rim of the nanoring to neutralize the nm)or thickness(10 nm) of the nanobelt]. 16. R.A. Romer, M E Taikh, Phys Stat SoL. B 221, 535(2000). local polar charge and reduce the surface area, which is suggested to be a result of balancing the 17. V Germain, J. Li, D Pbys.Chem.B107.8717(2003) resulting in the formation of a self-coiled, co- electrostatic repulsive force between the loops 18 C. StampfL. C. G. Van de Walle, Phys. Rev. B 57. axial, uniradial, multilooped nanoring structure. and the elastic deformation force R15052(1998) The self-assembly is spontaneous, which means The polar charge-induced nanorings present- 19. F. Vigue, P. Vennegues, S. Vezian, M rim the nanobelt grows. The reduced surface area ing fundamental physical phenomena, such as 20 L, sees al solid- state sc ences a spiny and the formation of chemical bonds (a short- the Aharonov-Bohm oscillations in exciton hu- range force) between the loops stabilize the minescence(16). The piezoelectric and semicon- 21. Support was provided by NSF(DMR-9733160) coiled structure. The width of the nanoring ducting properties of ZnO predict that the nanor- eering(DDR&E) program. w increases as more loops wind along the nanor- ings could be used as nanoscale sensors, trans- R. L Snyder and j z zhang for comments. ing axis(Fig. 40), and all of them remain in the ducers, and resonators. Online Ma same crystal orientation agog/cgi/content/fulL 303/5662/1348/DC1 Because growth is carried out in the temper- 1. C.B. Murray, C.R. Kagan, M.G. Bawendi, Science 270. ature region from200°to400°C,“ epitaxial sin tering"(3)of the adjacent loops forms a single- 2. R L Whetten et al, Adv. Mater. 8. 428(1996) 8 October 2003: accepted 29 December 2003 www.sciencemag.orgSciEnceVol30327feBruAry2004 1351to interface the positively charged (0001)-Zn plane (top surface) with the negatively charged (0001)-O plane (bottom surface), resulting in neutralization of the local polar charges and in reduced surface area, thus forming a loop with an overlapped end (Fig. 4B). The radius of the loop may be a result of how the nanobelt folds during its initial growth, but the size of the loop cannot be too small to reduce the elastic defor￾mation energy. The total energy involved in the process comes from polar charges, surface area, and elastic deformation (12). The long-range electro￾static interaction is likely to be the initial driving force for folding the nanobelt to form the first loop on which subsequent growth is based. Cal￾culations have been made to assess the possibil￾ity of balancing the increased elastic deformation energy (EDeform) by the decreased electrostatic energy (EElectro) at the initiation of the nanoring structure (7) (Fig. 4E). If a nanobelt is folded to form a single-loop complete ring (n  1), the energy gain E (E  EDeform EElectro) is positive regardless of the size of the loop, sug￾gesting that a single-loop complete ring is ener￾getically unfavorable. For a 10% overlapped na￾noring (n  1.1), the nanoring structure is ener￾getically possible if its diameter (D) is larger than 600 nm. It is thus possible to form a single￾looped nanoring with overlapped ends at the beginning of the growth caused by fluctuation (Fig. 4B). For a 33% overlapped nanoring (n  4/3), the nanoring structure is possible if D is larger than 360 nm. The diameters of the experimentally observed nanorings are in the range of 0.8 to 4 m, and no ring has D  0.8 m. From the SEM images, we believe that the first step (Fig. 4B) occurs before the na￾noring lands on the substrate. The presence of a planar defect within the nanobelt (Fig. 2F) is likely to be the key to producing the fast growth of the nanobelt along [101¯0], because it lowers the energy in the wurtzite-structured lattice (8). Planar defects may be initiated by impurity atoms, such as Li and In, introduced into the raw material. As growth continues, the nanobelt may be natural￾ly attracted to the rim of the nanoring by elec￾trostatic interactions and then extend itself par￾allel to the rim of the nanoring to neutralize the local polar charge and reduce the surface area, resulting in the formation of a self-coiled, co￾axial, uniradial, multilooped nanoring structure. The self-assembly is spontaneous, which means that the self-coiling along the rim proceeds as the nanobelt grows. The reduced surface area and the formation of chemical bonds (a short￾range force) between the loops stabilize the coiled structure. The width of the nanoring increases as more loops wind along the nanor￾ing axis (Fig. 4C), and all of them remain in the same crystal orientation. Because growth is carried out in the temper￾ature region from 200° to 400°C, “epitaxial sin￾tering” (13) of the adjacent loops forms a single￾crystal cylindrical nanoring structure, and the loops of the nanobelt are joined by chemical bonds as a single entity. The loops that were coiled first remained at the growth temperature for a longer time, resulting in higher crystallinity, whereas the ones that wound on later had less time for sintering and thus had relatively poorer crystallinity, forming the structure in Fig. 2G, with two contrast regions across the width of the nanoring. Finally, as the growth time was ex￾tended, the entire nanoring exhibited high-qual￾ity crystallinity, as shown by the diffraction con￾trast in Fig. 4D. A uniradial and perfectly aligned coiling is energetically favorable because of the complete neutralization of the local polar charg￾es inside the nanoring and the reduced surface area. The entire growth process may have no relation to the substrate used for collecting the sample. The thinness of the nanoring also pre￾vents the determination of its polarity by conver￾gent-beam ED (14). The coiling process presented in Fig. 4 unifies the two types of nanoring structures described in Figs. 2 and 3. If the (0001) polar surface of the ZnO nanobelt is parallel to the ring plane, self-coiling of the nanobelt at a radial direction of [12¯10] forms the type I structure in Fig. 2. Alternatively, if the nano￾belt is tilted toward the nanoring center so that the radial direction is [12¯13¯] (15), self￾coiling of the nanobelt produces the type II structure in Fig. 3. The tilting of the nanobelt may reduce the elastic deformation energy. The model presented here can also be adopt￾ed to explain the helical nanostructure reported previously (5). If the nanobelt in Fig. 4B is flipped by 90° so that the radial direction of the nanoring is [0001], which means that the polar￾ization is in the radial direction, a bending of the nanobelt into a circle slightly reduces the elec￾trostatic energy, possibly in favor of forming an in-plane spiral nanoring (5). Alternatively, as a result of preserved polar charges on the inner and outer arc surfaces, the circular loops of the nano￾belt cannot be densely packed into a single￾crystal coil structure because of the electrostatic repulsion among them; instead, a helical struc￾ture would be formed, with a pitch distance of 200 to 500 nm [much larger than the width (20 nm) or thickness (10 nm) of the nanobelt], which is suggested to be a result of balancing the electrostatic repulsive force between the loops and the elastic deformation force. The polar charge–induced nanorings present￾ed here have potential applications in investigat￾ing fundamental physical phenomena, such as the Aharonov-Bohm oscillations in exciton lu￾minescence (16). The piezoelectric and semicon￾ducting properties of ZnO predict that the nanor￾ings could be used as nanoscale sensors, trans￾ducers, and resonators. References and Notes 1. C. B. Murray, C. R. Kagan, M. G. Bawendi, Science 270, 1335 (1995). 2. R. L. Whetten et al., Adv. Mater. 8, 428 (1996). 3. J. F. Banfield, S. A. Welch, H. Z. Zhang, T. T. Ebert, R. L. Penn, Science 289, 751 (2000). 4. Z. W. Pan, Z. R. Dai, Z. L. Wang, Science 291, 1947 (2001). 5. X. Y. Kong, Z. L. Wang, Nano Lett. 3, 1625 (2003). 6. Energy-dispersive x-ray spectroscopy has a detection limit of typically 1 to 2 atomic %. It may not detect impurities with a concentration lower than 2%. 7. See supporting material on Science Online. 8. Planar defects usually reduce the energy for the nanostructure, and it is easy to form during the growth of one-dimensional oxide nanostructures. Once formed, it leads to the fastest growth along a direction parallel to the defect plane. For a ZnO nanobelt growing along [101¯0], a single stacking fault is always present (4). 9. Recent study has also shown that the presence of a stacking fault is the key for forming Ag disks (17). 10. Wurtzite structure has four different types of stack￾ing faults (18). The stacking fault at the nanobelt interface is type I. Type I and type III stacking faults have the lowest energy. 11. The local deposition temperature is 200° to 400°C, which is high enough to prevent physical adsorption of molecules on the surface during growth. Thus, the polar charges on the surface are likely uncompensated for and are electrostatically effective for aligning the oppo￾sitely charged surfaces of the nanobelt during growth. 12. There are three components of energy involved in the formation of ring structure: electrostatic interaction energy among the polar charges, surface area energy due to the decrease in surface area after chemically bonding the loops, and elastic deformation energy due to bending. The former two are usually called the surface energy, which includes the contribution from surface tension, but we separate them here for the convenience of discussion in the text. Electrostatic and deformation forces are long-range interactions, and chemical bonding is a short-range interaction. Self￾coiling is possible if the decreased electrostatic energy surpasses the increased elastic deformation energy, which is the case for a thin and narrow nanobelt. 13. Sintering in ceramics usually involves mass transport and diffusion. By “epitaxial sintering” here, we mean that the two loops are chemically bonded epitaxially with the same crystal orientation, and there may be no diffusion involved. As the nanobelt grew along its axial direction as guided by the planar defect, it was being bonded down on the rim of the ring by elec￾trostatic interaction. Because the melting point for a nanostructure can be as low as one-third of its bulk melting point, and the temperature required for sin￾tering is usually one-third of the melting tempera￾ture, it is thus possible to chemically join the loops at 200° to 400°C. 14. In principle, the polarity of the surfaces can be determined by convergent beam ED (CBED) (19), but it requires a specimen thickness of 50 nm or more to enhance the dynamic scattering effect. The 20-nm thickness of the rim of the nanoring is insufficient for CBED analysis. 15. Among the possible glide systems for hexagonal close-packed structure, (0001), [12¯10]; and (12¯12), [12¯13¯] are the two possible systems, which corre￾spond to the cases presented in Figs. 2 and 3, respec￾tively [see (20)]. 16. R. A. Ro¨mer, M. E. Taikh, Phys. Stat. Sol. B 221, 535 (2000). 17. V. Germain, J. Li, D. Ingert, Z. L. Wang, M. P. Pileni, J. Phys. Chem. B 107, 8717 (2003). 18. C. Stampfl, C. G. Van de Walle, Phys. Rev. B 57, R15052 (1998). 19. F. Vigue, P. Vennegues, S. Vezian, M. Laugt, J.-P. Faurie, Appl. Phys. Lett. 79, 194 (2001). 20. L. A. Shuvalov, Ed., Modern Crystallography IV, Spring￾er Series in Solid-State Sciences 37 (Springer-Verlag, New York, 1988), p. 109. 21. Support was provided by NSF (DMR-9733160), the NASA Vehicle Systems Program, and the Department of Defense Research and Engineering (DDR&E) program. We thank R. L. Snyder and J. Z. Zhang for comments. Supporting Online Material www.sciencemag.org/cgi/content/full/303/5662/1348/DC1 Methods Figs. S1 to S4 8 October 2003; accepted 29 December 2003 R EPORTS www.sciencemag.org SCIENCE VOL303 27 FEBRUARY 2004 1351 on December 19, 2006 www.sciencemag.org Downloaded from