Synthesis, electron tomography and single-particle optical response of twisted gold nano-bipyramids

Noble metal nanostructures are of particular interest due to their physical and chemical properties [1], making them suitable for potential applications such as optics [2–5], catalysis [6] or surface-enhanced Raman scattering [7]. One characteristic of noble metal nanoparticles is their optical response which gives rise to a broad absorption band in the visible region. The size and shape of nano-objects such as gold nanoparticles play an essential role in their typical optical response [8–12]. Therefore, being able to design these in a very predictive and reproducible way remains a real challenge [13]. In this respect, it is important to achieve a precise control of all the parameters involved in the synthesis of such objects. Recently, it has been shown that sharp and edgy structures can be of great interest since they provide a very high sensitivity to local changes in the dielectric environment, as well as larger enhancements of the local electric field [14–17]. Thus, several synthetic procedures have been attempted to provide such anisotropic shapes (rods [18], cubes [19] and stars [20]) through the well-known seed-mediated growth approach [21–23]. Among these structures, the bipyramidal geometry is less described in the literature. This is mainly due to the fact that such objects appear essentially as by-products originating from nanorod synthesis [18]. Although great efforts have been made to increase the yield of bipyramids [24–29], relatively large amounts of other 'impurities' such as nanospheres or prisms are still observed. In this context, a process leading to high yield homogeneous batches of bipyramids with both controlled size and optical response would be of great interest, even more since these nanomaterials may find applications in optoelectronic devices or in health care for imaging or therapy processes. Moreover the understanding of their spectroscopic properties will bring important information for the design of future nanocomposite materials.

In the present work, we first propose an approach to prepare bipyramid-like gold nanoparticles (Au-BPs) in high yield, through the seed-mediated growth process [26]. In the second part, photophysics of these objects are recorded in suspension and compared with the corresponding properties when measurements are performed on a single particle. A theoretical description of the optical response of these systems is also described in order to support experimental measurements. Finally, these results are correlated with characterization by electron tomography (3D-TEM).

2.1. General experimental information

2.2. Synthesis of the seeds

2.3. Anisotropic growth of bipyramids

From a synthetic point of view, the structures were prepared using hydroquinone as a mild reducing agent and small gold nanoparticles (2–5 nm) as seeds. This method produces water-soluble bipyramidal nano-objects stabilized by a bilayer of cetyltrimethylammonium bromide (CTAB). This bilayer prevents particles from aggregation or flocculation in water. The anisotropic growth is highly dependent on the nature of the seeds and the ratio of each reactant. Next, we investigated the seed's aging effects on the bipyramid anisotropic growth as well as the influence of the reactants' respective concentrations. Once optimized, the process led to bipyramidal gold structures with controllable aspect ratio.

The plasmonic properties of these nano-objects have been studied and modeled theoretically using the discrete dipole approximation. The role of several parameters has been investigated such as the structure aspect ratio or the tip truncation. These parameters directly impact on the extinction spectrum through a significant shift of the longitudinal plasmon band. Optical properties of these nanostructures are compared both at the single-particle level and in solution. The resulting extinction spectra are correlated to the single structure through TEM. The real nature of the morphology of the particles was confirmed for the first time using electron tomography (3D-TEM) showing an original twisted bipyramidal structure.

Gold bipyramids were synthesized using a two-step synthesis process, also known as seed-mediated growth. First, CTAB-protected spherical seeds (AuNP, 2 nm) were obtained by the reduction of a gold salt with a strong reducing agent (NaBH₄) in the presence of the surfactants. Then the as-prepared gold seeds were aged for one week (see below) before being injected into a growth solution of HAuCl₄, AgNO₃, CTAB and hydroquinone. The produced Au-BPs were assumed, as for rod-like structures, to be stabilized by a bilayer of cetyltrimethylammonium bromide [30, 31]. By varying the experimental conditions (e.g. seed solution volume), we were able to control the size of the final object. The extinction spectrum and TEM pictures of gold bipyramid suspensions are presented in figure 1.

Zoom In Zoom Out Reset image size

The TEM picture reveals extremely monodispersed bipyramids (70% yield) with a general tip-to-tip distance of 110 nm and a base length of 37 nm (figure 1). The statistical counting of the gold objects (over 500 objects) recorded in the TEM pictures yields a size distribution of 110 nm ± 16 nm and 37 ± 5 nm. The by-product, identified as dog bones and spheres, are less abundant (e.g. 15% yield for each). The UV–visible spectrum shows two clearly distinct peaks: one broad absorption band in the range 500–600 nm, attributed to the base of the bipyramid (and a 15% sphere contribution) and a more intense and thin band localized at longer wavelengths (760–1040 nm) attributed to the elongation of the tips (figure 1). Compared to rod [32, 33] or dog-bone [34, 35] shape extinction spectra, the longitudinal plasmon band is less broad. It is already well known that, concerning nanorods, the longitudinal plasmon band shift depends on the aspect ratios [36]. According to our theoretical investigations in the case of bipyramids, this position is connected to the aspect ratio and the refractive index of the environment but also the tip curvature. For all our batches, the transversal plasmon band was localized at 530–560 nm when, in the case of the longitudinal plasmon band, an important redshift is observed with increasing tip-to-tip elongation.

In order to improve the growth conditions for the Au-BPs, several parameters were optimized. The impact of the concentration of each reactant (silver nitrate and hydroquinone) was studied, as well as seed suspension aging.

The as-prepared seeds were firstly aged for 24 h to ensure the complete hydrolysis of the excess NaBH₄ before injection in the growth solution. Separately, part of the same seed suspension was aged at 40 °C for 58 d. The gold bipyramid synthesis was repeated each day with the same seed suspension to investigate the aging influence. The extinction spectrum highlights the gold core size evolution from 24 h to 58 d, inducing the surface plasmon band shift from 519 to 532 nm as well as increasing the intensity. We then assumed that the seeds grow from 2 to 5 nm, according to previous reports [37–39]. Interestingly, the absorption spectra of the bipyramid solutions obtained by the addition of freshly prepared seeds (24 h aging seed) show three bands, which could be assigned to a mixture of bipyramids (530–650 nm) and rods (530–1000 nm) (figure 2, left). However, aging the seeds for one week allowed us to obtain the desired bipyramidal shapes, with only two absorption bands in the range 530–550 and 800–850 nm (figure 2, right). For a seed aging period over 15 d, bipyramid suspensions were not obtained and precipitation was observed.

Zoom In Zoom Out Reset image size

These results demonstrate that the nature of the seeds is one of the key parameters for providing efficiently homogeneous gold bipyramids in a significant yield (figure 3).

Zoom In Zoom Out Reset image size

The seed aging step induced an increase in the core size of the initial gold nanoparticles (2–5 nm) and might also affect the seed's crystallinity. Freshly prepared gold seeds are single nanocrystals and may evolve into a twinned crystalline structure after a few days of aging allowing anisotropic growth at the surface of the crystals. The consequence is that the synthesis of Au-BPs cannot be achieved using freshly prepared seeds; the seeds need to be aged for at least 7 d.

The second studied parameter is the hydroquinone concentration. Hydroquinone reductant appears to be very efficient when used in gold nanosphere synthesis, according to Perrault et al [40]. In comparison with ascorbic acid, the reduction kinetics is slower, allowing a better organization of the structure during growth. Figure 4 shows the absorption spectra of a Au-BP suspension synthesized with different reductant concentrations. The longitudinal plasmon band, corresponding to the tip-to-tip elongation, is redshifted from 680 to 720 nm when reductant concentration increased. The hydroquinone did not reduce silver ions to silver nanoparticles, even in high concentration since no surface plasmon band localized at 400 nm was observed, which is a feature of pure spherical silver nanoparticles. These results clearly show how the hydroquinone concentration directly affects the object structure. For further investigations, it was then fixed to 13.8 mM with an aging seed period of 7 d.

Zoom In Zoom Out Reset image size

Another possible parameter to play with is the silver nitrate effect. The silver salt is used to induce the anisotropic growth on the seeds as previously described [26, 27]. Figure 5 shows the effect of adding various concentrations of silver nitrate (20–430 µM) to identical growth solutions.

Zoom In Zoom Out Reset image size

It confirms that the relative silver nitrate quantity plays an important role in the Au-BP formation. For low concentrations (20 and 40 µM), no anisotropic shape is obtained. The absorption band localized at 570 nm is similar to the absorption band of large-sized spherical gold nanoparticles. The optimal silver nitrate concentration for producing a monodisperse bipyramidal shape (yield 70%) was found to be 140 µM. By increasing the silver nitrate concentration (80–160 µM), the band localized at 570 nm splits into two distinct absorption bands and also a new intense band appears in the range 700–900 nm, easily attributed to the longitudinal mode of the bipyramidal structure. It was found that the appropriate silver nitrate concentration allowed the formation of Au-BPs instead of spheres, as confirmed by TEM analysis (figure 3). The addition of silver ions does not always produce a monodisperse bipyramidal shape. When raising the silver content up to 430 µM, a new absorption band appears in the 800–1200 nm range. This can be explained by the formation of a mixture of gold nanorods and bipyramids (figure 6). As previously reported [32], gold nanorods have a broad longitudinal plasmon band in this region. The proportion of rods and bipyramids was estimated to be 45% each (estimation considering over 600 objects).

Zoom In Zoom Out Reset image size

Most important experimental conditions are gathered in table 1 in order to summarize the obtained gold nanostructures.

Optical properties of bipyramid-like nanoparticles have been characterized at the single nano-object level followed by a correlation with their detailed morphology (determined by transmission electron microscopy (TEM)). Actually, although our process leads to a quasi-monodispersed population in high yield, the Au-BP specific optical properties cannot be reliably deduced from their UV–visible spectra in suspension (figure 1). This is due to inhomogeneous broadening effects resulting from the size, shape, orientation and environmental distributions characterizing the probed sample. In particular, it should be emphasized that all the synthesized particles contribute noticeably to the absorption band located around 520 nm, which corresponds to the degenerate plasmon band of spherical particles and to the transverse plasmon band of elongated ones (rods, bipyramids and dog bones).

The optical characterization has been performed using the far-field spatial modulation spectroscopy (SMS) technique [41]. A diluted sample (typically, less than one particle per square micrometer) was prepared by spin-coating a drop of the gold nanoparticle suspension on a thin substrate. That substrate was specifically designed to perform both the optical spectra and TEM images of single particles and correlate them unambiguously through a one-to-one correspondence [42]. Briefly the SMS technique consists in irradiating a single particle by a light beam which is focused at the diffraction limit by a high-numerical-aperture reflecting objective. In our set-up, a white lamp is used as a light source, permitting easy, rapid and broad-band extinction measurements (spectral range 300–900 nm). The technique consists in periodically displacing the nano-object under the inhomogeneous focused light spot along a direction perpendicular to the beam propagation axis by mounting the sample on a high resolution piezoelectric stage. The transmitted light, collected with a second reflecting objective, is dispersed via a spectrophotometer and detected by a photomultiplier. The modulation of the measured signal, which contains the physical information, is finally extracted by lock-in detection. On condition that the intensity distribution in the focused light spot and the oscillatory particle motion are known, the SMS technique gives access to the absolute extinction cross section of the nano-object [41–44].

For our study, this correlation technique combining SMS and TEM characterizations has been applied to a large number of bipyramids. Moreover extensive calculations, based on several theoretical approaches, have been carried out for analyzing the experimental optical spectra. These investigations will be reported on in a forthcoming more comprehensive paper, so we present in this short paper only a few selected illustrative examples.

Figure 7 exhibits typical results obtained for a single bipyramid-shaped particle representative of the synthesized population shown in figure 1. The high magnification image (right) is clearly consistent with the synthesis of a slightly truncated bipyramid with high aspect ratio L/b (where L = 115 nm and b = 34 nm are the (projected) longitudinal and transverse particle dimensions deduced from the TEM image). On the left are plotted the experimental extinction cross sections for the longitudinal excitation (red curve; the incoming electric field is polarized along the long axis of the bipyramid) and the transverse excitation (blue curve; orthogonal polarization with respect to the longitudinal one).

Zoom In Zoom Out Reset image size

Roughly, these results are typical of the optical response of strongly elongated particles, such as for nano-ellipsoids or nanorods of high aspect ratios [33]. The 'longitudinal' spectrum is dominated by an intense band in the red range. Its spectral location critically depends on both (i) the (effective) dielectric index of the particle environment and (ii) the aspect ratio. These two properties are illustrated in figures 8 and 9 in the case of a truncated bipyramid-shaped gold nanoparticle with a square base (edge b = 40 nm). The noticeable increase of the weak extinction signal below 550 nm has to be ascribed to the onset of the interband transitions in gold. On the other hand, the cross section corresponding to the transverse excitation is very weak over the entire spectral range (in fact, consistent with a zero value above 650 nm, in view of the small signal-to-noise ratio in the red/IR spectral range); only a noticeable increase below 600 nm is observed. Here again, the interband transitions contribute to a large extent to this weak signal.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In order to analyze the experimental data, theoretical spectra computed within the discrete dipole approximation (DDA) have been carried out [45, 46]. Briefly, the DDA approach consists in replacing the nanoparticle (or any homogeneous or composite nanostructure of arbitrary shape) with a cubic array of polarizable points, which is then assigned in relation to the bulk dielectric functions of the involved materials (in the present context, gold and the 'effective' homogeneous embedding matrix (that is, the environment)). The dipole polarizations, the scattered and internal fields, are then determined self-consistently for an incident monochromatic plane-wavefield and the various cross sections are finally obtained. In our calculations the data from the Johnson and Christie table have been used for gold [47], a frequency-independent dielectric index (N_m) has been assumed for the particle environment and about 2 × 10⁵ dipoles have been taken for 'filling' the truncated bipyramid-shaped gold particle. In these preliminary investigations, the simplest shape, quite consistent with the TEM images, has been selected, namely a symmetric truncated bipyramid with a square basis. This shape is entirely characterized by three geometrical parameters, that is, its physical length (L), the edge of the square basis (b) and the edge of the square apexes (c). In figure 8 are shown the DDA results for the parameter set L = 115 nm, b = 34 nm, c = 7 nm and N_m = 1.17.

The effective dielectric index of the 'pseudo-matrix' mimicking the particle environment is actually very close to the one inferred from our previous experiments on single gold nanospheres or nanodimers [36, 37]. In the framework of this simple modeling, the very large difference between the respective magnitudes of the 'longitudinal' and 'transverse' bands is well reproduced, as well as the spectral location of the 'longitudinal' band. Regarding the transverse excitation spectrum, the absence of a conspicuous plasmon band at 520 nm is due to the too weak signal-to-noise ratio. Considering the approximations made concerning the particle shape and the impact of the environment (during the experiment the particles are deposited on an ultrathin Formvar film; the surface layer of the surfactant or stabilizing molecules used in the chemical synthesis may modify the local dielectric constant), the quantitative agreement between these very simple model calculations and the experimental spectra is quite good. Small changes of the geometrical length parameters may result in large changes of the cross sections since the absorption and scattering cross sections scale roughly as V and V², respectively, where V is the particle volume. This is supported by experimental works and model calculations pointing out the strong dependence of the optical response on slight distortions of perfect shapes, such as the edge or corner smoothing for sharp geometries. Besides these modeling approximations, an experimental point has to be discussed here. To ensure the largest signal-to-noise ratio for such elongated particles, the axial location optimization of the piezoelectric stage along the beam axis is performed in the red spectral range (at the longitudinal plasmon band). Unfortunately, in that range, the signal optimization corresponds to axial locations (1–2 µm apart from the nominal focus of the reflective objective) where the intensity distribution is very complex. It then leads to some uncertainty with respect to the theoretical calibration factor used for relating the lock-in signal to the extinction cross section [41]. The complex structure of the focused light spot, far away from the nominal focus, is a consequence of the geometrical aberration arising from the focusing reflective objective [44]. Theoretical modeling shows that, apart from the nominal focus, the calibration factor in the short wavelength range is dramatically sensitive to the exact axial location of the substrate. This probably explains the strong disagreement between theory and experiment in this spectral range.

Considering that the 2D-TEM images provide only projected particle shapes, additional DDA calculations have been performed to assess the relevance of the theoretical spectra displayed in figure 8. First, for a given physical shape, we have checked that the longitudinal and transverse excitation spectra are not perceptibly modified when the bipyramid is turned around the longitudinal axis. Second, for a given base area (the square base is replaced by a disc, for instance) and a similar aspect ratio, the spectra remain almost unchanged too. However, for non-axially symmetric geometries such as, for instance, a truncated bipyramid with a square base, the length of the base edge (b) cannot be unambiguously deduced from the transverse dimension of the projected 2D-TEM image. In fact, depending of the orientation of the particle around the longitudinal axis, the transverse dimension in the image may vary in the range [b, 2^1/2b]. Since, for a small c/b truncation parameter, the particle volume scales as Lb², it can be conjectured that the transverse dimension in the TEM image is very close to b (otherwise, the magnitude of the computed spectra would be weaker).

Because of this drawback, the 3D shape of a single truncated bipyramid-like particle has been determined by tomographic electron microscopy. This technique consists in reconstructing the 3D shape of the nano-object from a series of 2D-TEM images taken at different orientations of the sample. In figure 9 are displayed three images of the reconstructed 3D particle shape. Far from the expected regular geometry, the object looks like a twisted truncated bipyramid. This shape is indeed much more complex than the one assumed in the simple previous modeling. In particular, the basis is not a perfect square and the edges and apexes are strongly smoothed. These noticeable morphological differences compared with the perfect square bipyramid geometry are probably responsible for the underestimation of the 'longitudinal' band magnitude in the DDA spectrum. This complex and rather unexpected shape could be a direct consequence of the seed's morphology elaborated during the synthesis.

In figure 10 are also shown the TEM image and the experimental 'longitudinal' spectrum of this specific particle. The almost perfect similarity between the longitudinal spectra in figures 7 and 10 indicates that our new synthesis route will ensure a quasi-perfect size and morphology monodispersity in the twisted bipyramid population.

Zoom In Zoom Out Reset image size

In our efforts to prepare well-defined gold nanoparticles for specific optoelectronic applications involving plasmonic effects, we report an efficient and original process for the preparation of homogeneous gold bipyramid-like nanoparticles. The deep investigations regarding the effects of the synthesis parameters, more specifically the seed aging, allow the fine tuning of the final nanostructure. The optical response of the single particles found a good correlation with the spectroscopy of a suspension in solution. The study was deepened through theoretical approaches using DDA modeling that found consistent results with both spectroscopy on single particles and microscopy. The real morphology was finally elucidated using electron tomography and, for the first time, a twisted bipyramidal-like structure was observed. This information can now be of great interest for the further design of new nanomaterials with controlled optical properties that could find applications in optical, optoelectronical or biomedical nanosystems.

This work was supported by ANR P3N project nanoPDT ANR-09-NANO-027. The authors thank Charlène Brillard for her contribution to the single-particle experiments and the theoretical analysis, and Xavier Jaurand (Centre Technologique des Microstructures de Lyon) for the 3D tomographie analysis.