A wide range of methods are currently available for the preparation of gold nanorods with controlled shape, yield and aspect ratios in the solution. The synthetic procedure as well as the mechanism of the formation has also been well-summarised in many recent reports [1,3,5,29]. To mention a few examples, cetyltrimethylammonium bromide (CTAB) surfactant-assisted seed growth method by Murphy et al. appeared as a versatile approach for the synthesis of gold nanorods with a controlled aspect ratio in the solution [5,29,34]. A much improved technique was later demonstrated by El-Sayed and co-workers, in which, by modifying the seeding and adding of Ag⁺into the growth solutions that was used by Murphy, nanorods with a yield of nearly 100% could be achieved [35]. Effective facet-selective gold deposition process in the presence of silver ions in the growth solution was considered as the key driving factor for the high-yield gold nanorod formation. However, in contrast to such outstanding achievement in solution phase, their growth directly on the surface has not yet been developed. Since many applications, such as optoelectronic, catalyst and surface-enhanced Raman scattering (SERS), required that they are attached onto the surface, the technique for growing the nanorods directly on the surface should be developed. It is true that the solution phase prepared gold nanorods may be transferred onto the surface via, for example, Langmuir-Blodgett [36] or drop and spin casting assemblies [37]; however, they are likely intricate and inadvertently expose the nanostructures to possible properties disruption during the process. Therefore, to develop a method that enables the shape-controlled growth directly on the surface is crucial.

Recently, we have grown metal nanoparticles directly on the surface using a seed-mediated growth method [38-40]. Gold nanoparticles with a variety of shapes, such as nanoplates, nanocubes, etc., have been successfully prepared directly on the surface [41,42]. As for 1D morphology, particularly in nanorods, considerable efforts have been devoted by us to synthesise them directly on the surface. Unfortunately, so far their yield was relatively low so that the unique 1D optical properties that are indicated by the presence of two separate bands, i.e. transverse and longitudinal SPRs, could not be detected on their optical absorption spectra. Actually, we have also attempted many approaches to improve the yield including the use of the well-known growth solutions for gold nanorods such as prescribed by Murphy's [3,29,34,43], El-Sayed's [35] and Korgel's [44] groups combined with our method. Unfortunately, again, these attempts failed. This could be the result of unique growth characteristic of gold nanostructures on the surface, which might be influenced by many factors including substrate surface-nanoseeds interaction effect, substrate surface-solution interface characteristics (e.g. wet-ability), and the nature of the precursors transport to the growing crystallite on the surface, which produce a different crystal growth compared to that of in the solution. We note that the formation of 1D gold nanoparticles directly on the surface is a remarkably challenging task, however, since the anisotropic effect of the nanostructure in the applications are very special and even superior in many aspects over other morphologies, the synthesis of gold nanoparticles with alternative morphology directly on the surface but producing properties that exhibit those in one-dimensional nanostructures is of particular interest.

The growth process was carried out by immersing the substrate that has been treated with the nanoseeds into a growth solution that contains 0.5mL of 0.01M of HAuCl₄, 8mL of 0.1M CTAB, 12mL of 0.1M hexamethylenetetramine, 0.3mL of 0.1M ascorbic acid and 40 μL of 0.01M of AgNO₃. The sample was kept undisturbed for 4h at 28°C for the growth process. If this condition is used, the final concentration of each reagent is 0.25, 40, 60 and 1.5mM and 20μM for HAuCl₄, CTAB, hexamethyleneteramine (HMT), ascorbic acid and AgNO₃, respectively. Finally, the sample was removed, rinsed with plenty of pure water and dried with a flow of nitrogen gas. The growth solution used in the present study was a new growth solution developed by our group of which was the modification of our standard solution for preparing gold nanotripod in solution [46] as our initial expectation to also realise the formation of nanotripods on the surface if using similar growth solution. There are several modifications that have been made to the standard solution, namely the AgNO₃ was new here and the NaOH is not required for the present study as it was in the preparation of gold nanotripod.

The original growth solution for the nanotripods contained two special surfactants, namely CTAB and HMT. As an attempt for obtaining 1D gold nanoparticle growth from the nanoseeds on the surface and being inspired by the fact that the AgNO₃ promotes the formation of nanorods in solution [35], we added a minute amount of AgNO₃ into the original solution. After being immersed for 4h in the growth solution, a purple-blue colour was formed on the surface. Normally, this kind of colour only appears on the gold nanostructures with one-dimensional morphology, such as nanorods or nanowires [3,29,35]. Therefore, this result signifies that one-dimensional morphology of gold nanoparticles might have formed on the surface. Optical absorption spectrum collected from the samples surprisingly showed the presence of two plasmonic bands at 520 and 680nm. In agreement with the observed purple-blue colour, two-banded absorption spectrum is also an indication of the formation of one-dimensional gold nanostructures on the surface. We then carried out a FESEM characterisation on the as-prepared samples. To our surprise, only high density networked string-like structures of gold nanoparticles were formed on the surface covering the entirety of the substrate surface instead of nanorods or other 1D morphologies, to which such unique two plasmonic characters can be related to. The optical absorption and the FESEM results are shown in Figure ​Figure1.1. However, as revealed in the high-resolution FESEM image of the samples (see Figure ​Figure1C),1C), actually the networked structures were mostly composed of quasi-1D nanoparticles (e.g. fusiform) with size (length) approximately 10nm that aligned side-by-side with each other. The length of the nanonetworks can be up to ca. 100nm, which corresponds to the aspect ratio of ca. 10. Despite the fact that the spherical nanoparticles also formed the networked structures, the network is normally relatively shorter than those composed of the quasi-1D nanoparticles. Even in most cases, they seemed to be an aggregate instead of a long-range networked morphology. However, no nanorods, nanowires or other 1D nanostructures were observed on the image, confirming that one-dimensional optical properties are solely produced by the networked-gold nanoparticles on the surface. As revealed in the image, the quasi-1D structures that composed the networked structure are considerably small with average length and diameter in the range of ca. 2 to3 nm and 10 to 15nm, correspondingly. Such small 1D nanostructures might have prospective use in catalysis and sensing applications such as the possibility of producing peculiar properties as the results of anisotropic morphology and quantum effect.

As evident in Figure ​Figure1A,B,C,1A,B,C, no perfect nanorods or nanowires were obtained on the surface. Therefore, the presence of one-dimensional optical properties could be directly related to the networked structure (nanonetworks) that formed linear chains on the surface. Thus, the transverse surface plasmon band, the shorter wavelength band, can be easily attributed to the oscillation of the free electron system toward the short axis of the nanonetworks. Meanwhile, the longitudinal plasmonic band, i.e. the longer band, was related to a long-range plasmonic coupling [47] amongst the nanoparticles in the networked structure (see Figure ​Figure1E).1E). Actually, the appearance of a longitudinal plasmonic band from the networked nanoparticles here is quite unusual and is predicted as resulting from the unique 1D morphology of the individual nanoparticles in the nanonetworks. According to our earlier results [38-40], in most cases, such unique 1D plasmonic characteristics were not observable if the network was composed of spherical nanoparticles. Therefore, the appearance of LSPR in these nanonetworks in the present study indicates a peculiar dependence of long-range plasmonic coupling on shape at this length-scale regime [47]. Thus, more unusual properties will then be expected to arise by the quasi-1D but assembled in a network structure.

As has been noted earlier, the AgNO₃ might have played a substantial role behind the formation of quasi-1D gold nanoparticles, which is the basis for the nanonetwork structure. To obtain a detailed understanding on its role in this process, we examined the growth characteristic of gold nanostructures on the surface in the presence of several AgNO₃ concentrations, namely 10, 20, 30, 40 and 50μM, via UV-vis absorption spectroscopy and FESEM imaging. Figure ​Figure22 is the related optical absorption spectra of the samples. As judged from the curve a, neither quasi-1D nor nanonetworks were formed when the AgNO₃ was absent in the reaction. However, 1D gold nanostructures might have effectively produced when the AgNO₃ as low as 10μM was used as the appearance of two plasmonics bands at ca. 520 (TSPR) and 630nm (LSPR) (see curve b). The quasi-1D nanostructures as well as nanonetworks formation might be optimum when the AgNO₃ concentration used was 40μM. It is indicated by a maximum shifting of the LSPR to red, reflecting the improvement of the aspect ratio of the nanonetworks (curve c). If the AgNO₃ concentration was higher or lower from this value, lower LSPR peak position was obtained, suggesting a decrease in the aspect ratio of the nanonetworks. The FESEM results as shown in Figure ​Figure33 further verified such growth characteristic of the nanostructures on the surface under the present treatment. For example, effective nanonetworks formation was confirmed when AgNO₃ concentration of as low as 10μM was used in the reaction (see Figure ​Figure3A).3A). This is probably the origin of the observed LSPR band in the spectra, formed via the presence of a long-range plasmonic coupling process in the nanonetworks. As has been noted earlier, spherical nanoparticle chains did not produce such optical properties resulting from a long-range plasmonic coupling amongst the nanoparticles in the chain. Thus, special nanostructure morphology, i.e. quasi-1D, should be present in the nanonetworks in order to support effective long-range plasmonic coupling via optimum edge-to-edge quasi-1D arrangement. The available FESEM images actually did confirm the nanostructures that composed the nanonetworks are quasi-1D morphology instead of spherical nanoparticles. It is true that high-resolution FESEM images were not available to support the claim. However, the present optical absorption spectra results that show 1D optical property, produced from effective long-range plasmonic coupling, are strong evident for the formation of such nanostructures morphology in the nanonetworks. However, a complete analysis on the detailed morphology of the individual nanostructures on the nanonetworks including high-resolution electron microscopy is being pursued and will be reported in a different publication. Large-scale nanonetworks were effectively formed covering the entirety of the surface when optimum AgNO₃ concentration was used, namely 20μM (see Figure ​Figure3B).3B). One point to be noted here is that the actual length of the nanonetworks obtained using this condition was relatively the same with that shown in Figure ​Figure3A;3A; however, their shorter axis was relatively smaller. Thus, the aspect ratio of the nanonetworks increased, shifting the LSPR to red. As has been earlier probed in the absorption spectra, further increase in the AgNO₃ added decreased the aspect ratio of the nanonetworks. Judging from the FESEM results, we pointed out the following facts as the reason for the process: The density of the nanonetworks increased with the increasing of AgNO₃ concentration. As revealed in Figure ​Figure3C,3C, despite the fact that the short axis of the nanonetworks decrease due to the increase in the density and possible strong coupling along the short axis of the nanonetworks, further red-shifting in the LSPR was not achieved but blue-shifting instead. A different structure was obtained when high concentration of AgNO₃ was added, namely 50μM. As can be seen from the Figure ​Figure3D,3D, nanonetworks that are composed of relatively bigger spherical nanoparticles were typical of the nanostructures product. The increase in the individual nanoparticle size might be due to a kind of ‘steric’ hindrance of surfactant adhesion onto the nanoparticles surface in the presence of high concentration of AgNO₃, accelerating the growth of individual nanoparticles.

While AgNO₃ drives the formation of 1D crystal growth of the nanoseeds, both CTAB and HMT via their combinative effect are also crucial in this process. In a typical case, no 1D optical characteristic was obtained if one of the surfactants was absent in the reaction. Typical optical absorption spectra of gold nanostructures prepared using several concentration ratios of CTAB and HMT (millilitre to millilitre ratio) are shown in Figure ​Figure4.4. As revealed in the spectra, both surfactants must be present in the reaction to give an optimum 1D optical characteristic, which is indicated by the longest LSPR peak position when CTAB to HMT ratio is 12:8 (see curve d). By keeping the cumulative concentration of both surfactants unchanged, if the ratio was increased or decreased, the LSPR peak position was blue-shifted compared to the optimum one, a sign of a decrease in the aspect ratio of the 1D nanostructures. It needs to be noted that limited number of gold nanoparticles probably grew on the surface when only surfactant HMT presented in the reaction, as judged by its low optical absorbance (see curve h). This could be due to a weak capping nature of the HMT to Au⁺so that the Au⁺reduction occurred in the solution instead on the nanoseeds surface. Selected FESEM image of the samples as shown in Figure ​Figure55 further confirms and provides detailed pictures of gold nanostructures' growth characteristic on the surface under several CTAB to HMT concentration ratios. For example, high-yield nanonetworks were obtained when the CTAB to HMT ratio of 12:8 was used. In good agreement with the optical absorption spectrum as shown in Figure ​Figure4,4, the yield was found to decrease when the ratio was increased (see Figure ​Figure5C5C for CTAB to HMT ratio of 10:10). Bigger-sized nanoparticle network was even obtained if the ratio further decrease to 8:12. On the basis of these results, it can be worthwhile concluding that combinative function of surfactants plays a strategic role in the formation of quasi-1D morphology and then the nanonetwork structures on the surface.

It needs to be noted here that the individual nanoparticles which compose the nanonetworks were actually mainly the quasi-1D gold nanostructures formed by a unique interplay amongst the reagents under the present condition. We hypothesised that a facet-selective surfactant adhesion, which is driven by the Ag⁺via its unique underpotential deposition nature onto the nanoseed surface, could be the main factor for the formation of the structures. However, the exact mechanism is not clear at the moment, especially the nature of the surfactants adhesion as well as the interplay amongst them and the Ag⁺. Although the following facts could be considered: (a) Quasi-1D nanostructures were not formed when one of the surfactants was absent. The quasi-1D nanoparticles may be formed at any concentration ratio between the two surfactants but optimum at the CTAB to HMT ratio of 12:8. This reveals that the combinative function of surfactant here is crucial. (b) The quasi-1D nanostructures were also not formed when Ag⁺is absent in the reaction. In fact, a diluted Ag⁺concentration (ca. 10μM) projected the quasi-1D nanoparticles growth and optimum at a concentration of approximately 20μM. These indicated that the interplay amongst the surfactants and the Ag⁺is necessary for the formation of quasi-1D nanoparticles. It is widely known that the Ag⁺may adsorb onto the gold nanoseed surface and then induces a one-dimensional crystal growth in the nanoseeds, certainly in the presence of surfactant. This scheme might probably also be valid to the present condition. However, since the system was on the solid surface, the growth characteristic should be different. It is true that the bromide ions were recognised to have a key effect in the promotion of nanorod morphology growth in the solution phase via effective adsorption onto the lower-energy surface of the nanostructures, i.e. (111) and (100), and play as a steric hindrance on these planes, which lead to a nanorod shape formation with growth direction toward (110) [48]. However, since in this experiment the quasi-1D morphology were not formed in the absence of Ag ion, the role of Br ion in the formation of quasi-1D nanostructures on the surface is not relevant. It is also true that the AgBr complexes may also be formed during this process. Since these complexes are relatively inactive compared to the Br ions as well as their quantity might be considerably small as the result of effective Ag ion deposition onto the Au nanoseed via effective underpotential deposition process, this chemical was believed to have minor effect on the promotion of quasi-1D gold nanostructures.

Meanwhile, on the network structures, their formation is assumed as an attempt to minimalise the high surface energy of the quasi-1D nanostructures that formed under the present condition. Actually, the initial nanoseeds on the surface were randomly distributed without any specific orientation (see Figure ​Figure6A).6A). After being grown in the growth solution, adjacent nanoseeds grew into a quasi-1D morphology and aligned with each other forming the nanonetwork structures (see Figure ​Figure6B,C).6B,C). Meanwhile, the nanoseeds that were far from the nanonetworks may have grown into bigger spherical nanoparticles or simply dissolved into the bulk solution via the Oswald annealing process and then supported the growth of bigger nanostructures on the surface. As revealed in the FESEM images, the nanonetwork structures have been observed on the sample that was grown for 30min. In spite of that fact, their formation may be earlier. The formation of these structures via aggregation amongst the nanostructures is not applicable in this process. It is because the nanostructures were grown from the nanoseeds attached to the surface so that the possibility of the nanostructures migration on the surface could be minimalised. The aggregation of the nanostructures from the bulk solution and then attached onto the surface is also irrelevant here since the mature nanostructures are only on the nanoseeds that are attached to the surface.