Our study (1) described a new technique that allowed for the precise targeting of chemical precursors to specific surface sites with nanometer-range accuracy. One of our aims was to control both the site and growth process of single-walled carbon nanotubes (SWCNTs) by depositing controlled mixtures of C60 and metal catalyst through a moveable shadow mask, followed by annealing to form single-walled nanotubes (SWNTs), a growth process with proven efficacy. We applied the technique to a number of substrates including molybdenum grids to allow for transmission electron microscopy (TEM) and electron energy loss studies. Although anticipating that the molybdenum would further catalyze the SWNT growth, we did not anticipate the degree to which the surface chemistry of the grids would dominate the resultant structures.
Shortly after the publication of (1), we recognized the problems associated with the Mo grids on the interpretation of our data, having carried out a detailed high-resolution energy loss and Raman study of the growth process. We reported the results of this study widely (2–5). The comment by Chisholm et al. (6) independently confirms our previous findings, and we are able to add some further detail as a result of our subsequent studies.
We followed the growth process by varying both temperature and time of annealing of the C60/nickel structures, recording Raman spectra and electron energy loss spectra. The former provided the greater insight. Briefly, Raman spectra were taken with two laser energies, 2.41 eV and 1.58 eV. During the very early stages of growth after the exfoliation of the C60, all the characteristic vibrational modes expected for SWCNTs were observed. The first-order features consisted of a G peak around 1590 cm–1, corresponding to the high-energy (∼ 0.2 eV) tangential modes (TM) of SWCNTs, and a D peak around 1300 cm–1, consistent with a disorder mode in SWCNTs. The G mode was very narrow, with a full-width at half maximum (FWHM) of about 13 cm–1. The low-energy radial breathing modes (RBM) of SWCNTs were not resolvable because the chemically complex substrates gave contributions in this frequency range of the Raman spectra. Their presence was, however, indirectly demonstrated by their involvement in combination modes in the second-order spectra.
Raman spectra consistent with SWCNTs were, however, observed only fleetingly during the growth process, so that the combination modes involving the RBM modes were rapidly lost, while the others—the D and G modes and their overtones—transformed into their nanocrystalline graphite (nc-C) counterparts. Thus, the resultant structures obtained on Mo grids after annealing at the times and temperatures reported in the original paper were based not on SWNTs but on a complex of molybdenum, as detailed by Chisholm et al. (6). The only area in which we would take issue with the Chisholm et al. comment is in the origin of the carbon edge in the energy loss spectrum as originally reported, which we would argue is at least partially residual carbon as inferred by the Raman study above.
Although the shadow mask technique remains a unique method to fabricate nanostructures at controlled surface sites, including the growth of CNTs, we accept that our conclusion of the growth of single crystals containing SWCNTs was mistaken. In particular, in the case of the Mo substrates the complex surface chemistry resulting from the manufacturing process dominates the final nanostructures observed.