REPORTS ry detected 0.013% Fe as the only micellar suspensions in the past(6), and thus supported by the fact that the weight returns ty, meaning carbon purity was more the observation of the fluorimetric peaks in to its initial value by subsequent annealing the as-grown SWNT forest strongly indi- in dry n, gas. Second, a black amorphous Pure SWNTs should allow for the cates that the SWNTs are not heavily carbon-coated quartz tube was cleaned trans- tigation of the intrinsic magnetic pro bundled but that many individual SWNTs parent by flowing water vapor at 750oC, of SWNTs and for the study of the exist. Peak locations in the spectrofluori- which provides direct evidence of water luminescent properties that usually netric contour plot were mapped out and induced oxidation of amorphous carbon. In dissolution of the SWNTs. For example, compared with that of the HiPco tubes(6), the literature, hot water on carbon nanotubes dimensional excitation-emission contour plot where a coincidence in peak location was has been used to purify amorphous carbon spectrum(Fig. 3) from preliminary spectro- found, which provides additional evidence (7, 8). Also, it has been reported that fluorimetric measurements on the as-grown that the tubes are SWNTs. However, the amorphous carbon is effectively removed SWNT forest clearly shows discrete spectral individual peak intensities differ noticeably from MWNTs by introducing water vapor peaks. Each peak corresponds to characteristic from those of the HiPco tubes [e.g, the into the CVd fumace (9). Metal particles absorption-emission wavelengths from van absence of the peak at 1105-nm emission further stimulate oxidation of carbon [for Hove optical transitions of SWNTs and can and 647-nm adsorption wavelengths with example, carbon in contact with metal be assigned to specific SWNT structures(6. assigned tube index (7.6), which indicates particles is removed at temperatures as low These characteristic peaks have been ob- a different abundance distribution of nano- as 225C (5), and thus this effect should served only in individual SWNTs in aqueous tube radii and chiralities. Moreover, a detailed further assist the role of water as a protective investigation reveals that the spectrofluori- agent against amorphous carbon coating. metric contour plot of the SWNT forest is These results suggest that water-assisted richer in structure than that of the HiPco growth could be applied to other growth tubes; we tentatively attribute this result to a systems, such as methane and acetylene wider distribution of nanotubes in our sam- CVD, or to grow other nanotubes, such as ples. Our results provide a direct route to map MWNTs. We believe that our understanding the detailed composition of the as-grown concerning the role of water represents a SWNT materials and can be adapted to reasonable basic description, although future directly study the dependence of the nanotube work is required to quantify and better a distribution on the synthesis conditions and understand the effect of water 800 catalysts without any ambiguity Realization of large-scale organized Temperature('c) Several additional points and experiments SWNT structures of desired shape and form D regarding the effect of water deserve com- is important for obtaining scaled-up func- nent. First, TGA(Fig. 2B) on pure SWNT tional devices. With the assistance of water naterial using N, gas with water shows that SWNTs grow easily from lithographically SWNT oxidization starts at about 950C, patterned catalyst islands into well-defined 89 which indicates that water does not oxidize vertical-standing organized structures, as and damage SWNTs at the growth temper- demonstrated by the large-scale arrays of ature. We believe that the small initial macroscopic cylindrical pillars(Fig. 4A) weight increase is due to physisorption, with 150-um radius, 250-um pitch and a g. 4. SEM images of WNT structures. (A) Eoco033E9=30moE age of SWNT cylindric SWNT material (A) Tric properties of the h, and -1-mm height Ins 10-C/min) of mple of the SV M image of a root of a pillar. Scale bar. 50 min) of a 9-mg sample of the SWNT material in N2 (flow rate, 100 cc/min standard temperature and D)SEM images and pressure) passed through a water bubbler sheets 10 um thick.(E)SEM sheet 5 um thick.(F)SEM image Umm of the sheet face Fig. 3. Contour plot of fluorescence intensi versus excitation and emission wavelengths for the as-grown SWNT forest sample www.sciencemag.orgScieNceVol30619NovembEr2004 1363trometry detected 0.013% Fe as the only impurity, meaning carbon purity was more than 99.98%. Pure SWNTs should allow for the inves￾tigation of the intrinsic magnetic properties of SWNTs and for the study of the photo￾luminescent properties that usually requires dissolution of the SWNTs. For example, a two￾dimensional excitation-emission contour plot spectrum (Fig. 3) from preliminary spectro￾fluorimetric measurements on the as-grown SWNT forest clearly shows discrete spectral peaks. Each peak corresponds to characteristic absorption-emission wavelengths from van Hove optical transitions of SWNTs and can be assigned to specific SWNT structures (6). These characteristic peaks have been ob￾served only in individual SWNTs in aqueous micellar suspensions in the past (6), and thus the observation of the fluorimetric peaks in the as-grown SWNT forest strongly indi￾cates that the SWNTs are not heavily bundled but that many individual SWNTs exist. Peak locations in the spectrofluori￾metric contour plot were mapped out and compared with that of the HiPco tubes (6), where a coincidence in peak location was found, which provides additional evidence that the tubes are SWNTs. However, the individual peak intensities differ noticeably from those of the HiPco tubes Ee.g., the absence of the peak at 1105-nm emission and 647-nm adsorption wavelengths with assigned tube index (7.6)^, which indicates a different abundance distribution of nano￾tube radii and chiralities. Moreover, a detailed investigation reveals that the spectrofluori￾metric contour plot of the SWNT forest is richer in structure than that of the HiPco tubes; we tentatively attribute this result to a wider distribution of nanotubes in our sam￾ples. Our results provide a direct route to map the detailed composition of the as-grown SWNT materials and can be adapted to directly study the dependence of the nanotube distribution on the synthesis conditions and catalysts without any ambiguity. Several additional points and experiments regarding the effect of water deserve com￾ment. First, TGA (Fig. 2B) on pure SWNT material using N2 gas with water shows that SWNT oxidization starts at about 950-C, which indicates that water does not oxidize and damage SWNTs at the growth temper￾ature. We believe that the small initial weight increase is due to physisorption, supported by the fact that the weight returns to its initial value by subsequent annealing in dry N2 gas. Second, a black amorphous carbon-coated quartz tube was cleaned trans￾parent by flowing water vapor at 750-C, which provides direct evidence of water￾induced oxidation of amorphous carbon. In the literature, hot water on carbon nanotubes has been used to purify amorphous carbon (7, 8). Also, it has been reported that amorphous carbon is effectively removed from MWNTs by introducing water vapor into the CVD furnace (9). Metal particles further stimulate oxidation of carbon Efor example, carbon in contact with metal particles is removed at temperatures as low as 225-C (5)^, and thus this effect should further assist the role of water as a protective agent against amorphous carbon coating. These results suggest that water-assisted growth could be applied to other growth systems, such as methane and acetylene CVD, or to grow other nanotubes, such as MWNTs. We believe that our understanding concerning the role of water represents a reasonable basic description, although future work is required to quantify and better understand the effect of water. Realization of large-scale organized SWNT structures of desired shape and form is important for obtaining scaled-up func￾tional devices. With the assistance of water, SWNTs grow easily from lithographically patterned catalyst islands into well-defined vertical-standing organized structures, as demonstrated by the large-scale arrays of macroscopic cylindrical pillars (Fig. 4A) with 150-mm radius, 250-mm pitch and a Fig. 3. Contour plot of fluorescence intensity versus excitation and emission wavelengths for the as-grown SWNT forest sample. Fig. 2. Thermogravimetric properties of the SWNT material. (A) TGA data (ramp rate, 10-C/min) of a 10-mg sample of the SWNT material in air. (B) TGA data (ramp rate, 10-C/ min) of a 9-mg sample of the SWNT material in N2 (flow rate, 100 cc/min. standard temperature and pressure) passed through a water bubbler. Fig. 4. SEM images of organized SWNT structures. (A) SEM im￾age of SWNT cylindrical pillars with 150-mm radius, 250-mm pitch, and È1-mm height. Inset, SEM image of a root of a pillar. Scale bar, 50 mm. (B) Side view of a pillar. Scale bar, 100 mm. (C and D) SEM images of SWNT sheets 10 mm thick. (E) SEM image of an isolated SWNT sheet 5 mm thick. (F) SEM image of the sheet face. www.sciencemag.org SCIENCE VOL 306 19 NOVEMBER 2004 1363 R EPORTS on January 23, 2008 www.sciencemag.org Downloaded from