1. Introduction
The strong demand to miniaturize to improve the spatial resolution and to decrease the detection limits in optical spectroscopic devices has initiated a significant amount of research in the field of the plasmonics [
1,
2,
3,
4,
5]. The recent propositions in the literature are based on the nano-structuration of nanoparticles (Nps) of metal (Reference [
5] and references therein). Indeed, electrons on a metal surface, excited by electromagnetic fields (for instance by focusing light beams on Nps), are able to enhance the electric field locally up to several orders of magnitude. The spatial area where the field is “concentrated” is as small as just a few nanometer squares which provides the opportunity to design the electric field of light with a spatial resolution of 2–3 orders of magnitude below the diffraction limit. This effect has been at the origin of the Surface Enhancement Raman Spectroscopy SERS spectroscopy and of Tip Enhanced Raman spectroscopy (TERS) that presents an exponential decrease. The enhancement was predicted to be increased on to originate in the near-field generated in inter-particle gaps, for instance within the colloidal aggregates [
1,
2,
3,
4,
5]. However, because the experiments involved wide distributions of aggregates of different sizes and shapes or of many different nano-structured surfaces, the correlations between their near-field properties and SERS spectra were highly challenging. Then many experiments have tried to rationalize these nano-structuration by lithography, colloidal depositions, self-similar assembling [
5,
6], often in a “top-down” way. In the field of molecular fluorescence detections, the combination of the process assembly simplicity with the fluorescence enhancement made the self-assembled colloidal nanoparticle gap antennas optimal to extend a wide variety of single-molecule applications toward the micromolar concentrations [
7]. This self-assembly may be helped by a pre-step of lithography [
8]. Thus, for instance, spatially programmable Au nanorod dimers have been obtained with enhanced capacities for fluorescence applications, highlighting the opportunities for precise tunability of the plasmonic modes in larger assemblies. F. Koendernicks recently published a review on the different possibilities of this kind of approach to obtain single photon nano-antennas on large spatial areas [
9]. These kinds of approaches have also been used to build substrates for surface enhanced Raman spectroscopy uses [
10,
11]. For instance, large-area multifunctional wedge and pyramid arrays directly onto planar substrates via template stripping [
10] allowing in the same time to magnetically trap the species and to characterize them by SERS in the optical near-field. By lithography, substrates constituted of gold nano-stripes have been proposed and deeply studied in SERS applications [
11]. The SERS efficiency of probe molecules was investigated on non-annealed and annealed lithographic gold stripes. The SERS intensity is not linked to the far-field response of the non-annealed stripes. This mismatch between the far-field and near-field response comes from a significant contribution of the surface roughness of the stripes. The annealing of lithographed substrates by decreasing roughness resulted in a strong weakening of the Raman signals. Their results further suggested that the SERS effect was more pronounced in the red part of the visible range, far from the plasmon resonance of the structures and was essentially generated by the sub-20nm structures coming from the roughness than more by the organized lithographed over 20 nm. Therefore, nano-shapes, sub-20 nm, for SERS applications remains an important key to investigate. However, another bottom-up approach was proposed, in particular by Chuntonovs’ team [
12,
13]. The idea was to describe small aggregates of some metallic Nps as plasmonic molecules [
5,
12,
13]. Indeed, well-defined assemblies of Nps that sustain surface-plasmon resonances have been labelled “plasmonic molecules” [
12,
13,
14], where the individual “atoms” of these plasmonic molecules are then the individual metallic Nps, called “plasmonic atoms”. In this paper, the individual gold Nps will correspond to “plasmonic atoms”. In the dipole approximation, the “plasmonic atoms” may be described as individual Nps dipoles and the inter-strength depends then on the single-particle polarizability, which scales as the cube of the diameter of the individual Nps. Moreover, the energy of the interaction of the two dipoles induced on the surfaces of the individual Nps upon excitation scales as the inverse of the cube of the gap-distance between the particle centers. Nordlander and co-workers developed a plasmon hybridization description [
12,
13,
14,
15,
16] which predicts the modes of interacting Nps (plasmonic atoms) and calculates their corresponding energies, i.e., the resonance spectra.
In this application field of individual plasmonic molecules, we have developed since 2013 [
17] an experimental approach where molecules are built atom per atom with the help of an Atomic Force Microscope (AFM) coupled directly with an optical spectrometer in order to monitor the optical changes. Tong et al. initiated this approaching 2008 to understand the enhancement of Raman signals of nanotubes of carbon (Single wall nanotubes of carbon SWCNT) by approaching an individual Au Np to one isolated target SWCNT via AFM [
18]. The advantage of this method is to use the same Au Nps in all the experiments, thus all the physical changes are caused by changes of the shape of plasmonic molecule or to the distances between plasmonic atoms but never due to intrinsic Np changes (the individual diameters, individual shapes, surface chemistry are fixed). Thus the discussion of the experimental results are facilitated. The remaining experimental problem in this approach is to have the perfect control of distances between the plasmonic molecule and the interacting nano-objects. Indeed, due to the strong decrease of the optical near fields (as previously evoked in inverse of the cube of the gap distances between Nps) the signals strongly depend on the different relative spacings. Consequently, if the goal is to use one plasmonic molecule as a nano-sensor, its response varies with only spatial parameters. This paper proposes a shaping of plasmonic molecule to solve this problem and to avoid the strong gradient effect of the near field around metallic Nps. This original approach will be applied to follow the Raman spectra of individual nanotubes of carbon. This kind of object is now a “quasi” academic object widely described and used in the same kind of experimental conditions [
19,
20]. For instance, Reichs’ team realized the coupling of carbon nanotubes as a one-dimensional model system to lithographed gold nanodimers acting as near-field cavities. Their plasmonic cavities enhanced the Raman signal of a small nanotube bundle by three order of magnitude. Our paper will show that this enhancement may also be reached by nano-manipulating one single particle from the assembly of Au Nps constituting the cavity.
4. Discussion
The enhancement factor of the Raman signals excited at 514.53 nm, of CNT 2 by the Au Np dimer (
Figure 5) from either side of CNT 2 was estimated between 500 and 1000 from the Raman experiments. This estimation is based on the value of the ratio of the detection thresholds, collected on the same area probed on the same optical objective, with the same spectroscopic configuration. This enhancement is then produced only by the presence of the plasmonic structure. Some experiments with an excitation wavelength outside of the resonances (of CNTs and AuNps) at 785 nm have been tried to estimate the backscattering effect of the AuNp structures presence on our Raman signals. We had not been able to record any signals without damaging our structures (the Finite Element Method (FEM) simulations gave only a backscattering power outside of the resonance multiplied by around 2). Thus, we assumed that the essential effect in these experiments comes from resonance plasmonic effects. In the CNT 2 case, the hybridization of plasmonic levels of the two nanoparticles is obtained with a gap spacing of around 10 nm, between the two particles of 22 nm, without chemical contact between CNT and Au Nps. In this configuration, the numerical Comsol FEM modeling gives an enhancement of the near electric field of about 4, at 525 nm. If we assumed that the SERS signal is as the power 4 of the electric field amplitude, then the SERS will be multiplied by 256. If we compare the dimer to the monomer, the maximum of the plasmonic resonance is slightly shifted from only 5 nm and broadened by a factor 2 as described in the literature on dimer plasmonic molecules [
5]. This result is comparable in particular with those obtained by Zhu et al in 2014 [
26] that measured SERS enhancement factor for gap distances of 10 nm between two 90nm disks between 2 and 3 orders of magnitude. Then in the case dimer-CNT 2 assembly, the Raman measurement may be explained only by the enhancement of local electric field created by the dimer, without other important effect involving shift of resonance or coupling between structures or yet electronic transfer. However, this enhancement is strongly dependent on the gap spacing and on the position of the Raman scatterer, because the gradient of the electric field magnitude in the gap is strong. Therefore, we had proposed other plasmonic molecules to diminish this gradient effect and tested them with CNT 1.
The CNT 1 (
Figure 6) trapped in the gap between the diamond shape molecule and an isolated plasmonic atom gives us other pieces of information than the CNT 2. The enhanced Raman spectrum of CNT 1 allowed us to identify its chirality nature in coherence with the no direct observed resonance effect at 561 nm of the CNT 1 without Au Nps. Indeed, few CNTs on Kataura diagrams [
27,
28,
29] correspond to the RBM wavenumber observed for the CNT 1. The “closest CNT” on the Kataura plot of our observations would be the metallic (8,8) SWCNT characterized by a RBM wavenumber at 216 cm
−1. It is characterized by an electronic transition of around 18,300 cm
−1 (i.e., 545 nm), not too far of 561 nm excitation. The second kind of possible SWCNT would be the metallic (12,3) tube. This tube is characterized by its electronic transitions at around 16,500 cm
−1 (i.e., 605 nm) and 19,900 cm
−1 (502 nm) and whose RBM wavenumber is expected at 217 cm
−1. These second possibility of SWCNTs is then not characterized by an electronic transition excitable at 561 nm.
The second experimental fact concerning CNT 1, is that the G band intensity, usually not very relatively strong for metallic CNTs by comparison with RBM signal, is more enhanced, with the diamond shaped plasmonic molecule than the RBM and G’ signals. The usual Raman profile is then deformed by the plasmonic resonant molecules.
The enhancement observed around 560 nm is clearly explained by the plasmonic molecule electronic structure. As displayed in the
Figure 7, the computed scattering cross section of the plasmonic molecule studied here, shows a resonance at 555 nm with the incident field parallel to the gap direction. The local electric field in the gap at the resonance wavelength is multiplied by 10, i.e., the Raman scattering could be expected to be multiply by 10
4.
Now to understand the change of profile of the Raman spectrum, the coupling between the CNT and this resonant plasmonic molecule must be investigated. Indeed, the CNT Raman resonances (see for example Reference [
30]) may be obtained either at the wavelength corresponding at the electronic transition (Raman resonance at the excitation) or at the inelastic scattered wavelength (Raman resonance at the scattering). Raman excitation profiles display these two resonances [
30] versus laser wavelength, resonances with the incident or with scattered photons.
Taking in account both these resonance effects of CNTs gives the principle scheme shown in
Figure 8 applied to our case. In this scheme, the Raman spectra may be excited at (i) the wavelength corresponding to the energy allowing the electronic transition (blue lines in
Figure 8) in order to induce the enhancement of the whole inelastic spectrum without strong change in the relative profile or (ii) also to a wavelength that could favor, not the excitation, but the inelastic scattering processes (orange and green lines in
Figure 8). The first (i) condition could be favored in the (8,8) CNT and then the Raman signal would be dominated by the RBM signal with a weaker intensity for G, as usual for metallic CNT. However, that does not correspond to our observation. Now let us consider the second (ii) condition for (12,3) chirality. For example, to favor the G band, the wavenumber of the incident laser beam must correspond to the electronic transition wavenumber plus the vibration wavenumber characteristic of the G mode (here around 1600 cm
−1). Then if CNT 1 is assigned to a SWCNT (12,3), this last inelastic amplified process will be obtained at its maximum for 16,500 cm
−1 (electronic transition) + 1600 cm
−1 (G mode), i.e., 18,100 cm
−1 (corresponding to 552 nm). In our experimental case, we have obtained this kind of enhancement with an incident wavelength at 561nm. If we add that the Raman fingerprint is deformed with a G band more enhanced, we assign the observed enhancement to a Raman resonance effect obtained with inelastic scattered photons induced by the plasmonic enhancement at this resonant wavelength due to the peculiar shape of the diamond plasmonic molecules. If the experiment was conducted to look into the resonance at the inelastic scattered G’, then the plasmonic molecule would have to be deformed to shift its plasmonic resonance to 510 nm. The experimental Raman observation of
Figure 5 is then only interpreted by the enhancement effect produced by the diamond plasmonic molecule coupled with the single plasmonic atom dressing the (12,3) CNT, where the signal G is favored by a resonance at the inelastic scattering energy.
A second interesting point with this plasmonic molecule with a spatial gap with plasmonic structures from either side of the CNT 1, comes from the quasi absence of gradient of local electric field in this gap.
Figure 8 shows that the plasmonic molecule presents several hot points (where the electric field multiplied by 50) localized inside the structure and not in the gap between the molecule and the atom. In this spatial spacing, the local electric field at the resonance is only multiplied by 10 but at all positions in this gap. This kind of behavior is encountered when two plasmonic molecules are separated by a small distance. The
Figure 9 gives another example of one possible assembly and compares it with a gold tip usable in a TERS experiment. The exponential decrease of the amplitude of the local electric field observed for TERS configuration is then replaced by a quasi-constant electric field amplitude for a gap of 5 nm. The value of the electric amplitude obtained in this gap is equal to this one obtained at 2nm of the apex of the tip in TERS mode. Now, when the distance is increased to 10 nm the amplitude decreases and mostly a gradient of the amplitudes reappears. Of course, when there is not important gradient effect, the position of the CNT trapped in the gap does not have consequences on the measurements. This point could be important in the conception of nano-Raman sensors based on plasmonic molecules.
In our experiments, we have nano-manipulated 20nm-diameter plasmonic atoms, consequently the range of wavelength accessible by tuning of the shape and inter-distances of our plasmonic nano-structures is limited between 510 and 570 nm. To increase this range (to 700 nm), we suggest manipulating 40 or 50 nm diameter gold particles, always synthesized by a Turkevich’s way by changing synthesis conditions [
21,
22]. We are developing different nano-manipulations coupled with pre-micro manipulations to generate different natures of plasmonic molecules connected more or less with each other’s, in order to use different reserves of Au Nps around the area of interest. Thus, we hope to expand the possibilities of designing plasmonic molecules and without chemical direct bound between molecule and Au surfaces to be generate Raman enhancement. Finally, the possibility to design shapes and symmetry of plasmonic molecules opens the possibilities to control the physical coupling between them and other nano-objects, such as here CNTs to generate new nano-emitters based on the Raman effect.