1. Introduction

Plasmonic structures are attractive due to their capability to manipulate light in the near-field and far-field. In the near-field, the enhanced field associated with surface plasmon resonance strongly interacts with molecules in close proximity to the metallic surface, which leads to surface enhanced technology such as surface-enhanced Raman scattering (SERS), surface-enhanced IR absorption (SEIRA), surface-enhanced fluorescence, etc [1–3]. Furthermore, this enhanced field can be localized in an extremely small volume by a plasmonic nanocavity. Under specific conditions, the localized optical mode and electronic mode of the molecules within the nanocavity can reach the quantum strong coupling regime to produce mixed states which are half-matter and half-light [4]. In the far-field, the large scattering cross section of plasmonic nanoparticles provides better contrast for imaging [5, 6]. In addition, the strong absorption due to the surface plasmon resonance generates substantial heat for photothermal therapy [5, 6] and photoacoustic imaging [7].

These promising optical properties of plasmonic materials are partially derived from various morphologies of individual plasmonic nanoparticles [8, 9] such as core-shell nanoparticles, nanorods, nanocubes, nanostars, etc. which are fabricated using a variety of chemical synthesis methods. The alternative way to tailor optical properties is assembling existing particles to form coupled plasmonic structures. The coupled structures can be divided into substrate-supported and non-substrate supported (3D) structures. For the substrate-supported coupled structure, e-beam lithography or focused ion beam (FIB) processes can easily define size, position and orientation of constituent particles and the gap size among them. These planar coupled structures have various applications. For example, the nanoantennas consisting of different sizes of nanoparticles can control the emission direction and rates of the quantum emitters nearby [10, 11] or can convert the optical mode in the silicon waveguide underneath the nanoantennas [12]. Furthermore, the metasurface consisting of various particles on the substrate can shape the wavefront of the light beam to generate planar optical elements [13].

However, compared to the fabrication of the planar coupled structures, assembling nanoparticles for stable 3D coupled structures in the solution phase is more challenging. The nanoscale gap between particles usually relies on molecular linkers. Addition of the linker molecules may decrease or sometimes even break the stability of the original colloidal suspensions, causing random aggregates. In addition, the formation of the linkage between molecule/molecule or molecule/particle may be inefficient and time-consuming, which depends on the bonding types among them and/or the rate of ligand exchange occurring on the surface of the nanoparticles [14–17].

The core satellite assembly (CSA) is a 3D structure consisting of a core particle surrounded by several satellite particles. The plasmonic coupling between core and satellite particles produces high gap field regions which are usually called hot spots. Due to the relatively smooth surface curvature, the hot spots formed by the spherical particles provide more stable and predictable field enhancement. In addition, the optical response of the CSA is not seriously affected by the polarization of the incident wave in contrast to the linear structures such as dimers. This structure has been fabricated by several methods [18–22] with different limitations such as difficulties in decreasing gap size, demand for ionic strength tuning, or complicated linker preparation.

The Raman tag (or label) is a 3D composite nanostructure emitting the characteristic Raman signal. Recently, it has been developed and applied to labeling different levels of the bio-hierarchy, from organelles, individual cells to cellular tissue [23–30]. Raman tags are typically composed of Raman reporter molecules adsorbed on the surface of single plasmonic nanoparticles or embedded within the gaps of nanoparticle clusters. The localized field on the surface of nanoparticles or within the interparticle gaps enormously increase the signals of the Raman reporters due to the SERS effect and can even possibly augment the brightness of the tags comparable to the fluorophores. The conventional fluorophores used in biomedical imaging usually suffer from photobleaching and monotonous spectral features. The Raman tags are not limited by these two disadvantages. The electromagnetic SERS enhancement is roughly proportional to (E/E_inc)⁴. Generally, the gap field provides much higher enhancement than the localized field on the surface of single particles. In addition, the gap field is superlinearly increased with the decreasing of the gap size [18]. The single star-shaped nanoparticles provide a highly localized field with large variation because of non-uniform morphology. Therefore, a stable plasmonic structure with nanoscale internal gaps such as CSA provides an ideal backbone for the bright Raman tags.

In this work, an efficient and simple fabrication process of 3D stable CSA by polyelectrolyte coating is demonstrated. The optical properties and structure morphology is characterized and compared with the simulation results. By integrating Raman reporter molecules into the CSA backbone, three distinct Raman tags are shown. The robustness of the CSA backbone is also verified by monitoring the Raman signals. The intensity stability of Raman tags is investigated and compared with the fluorophore.

2. Methods

2.1 Sample preparation

A 50 nm gold core particle was linked to several 20 nm gold satellite particles through a polyelectrolyte layer to form a core-satellite assembly (CSA). First, citrate-stabilized core particles which have negative zeta-potential (−47 mV) were coated with positively charged polyelectrolyte, polyallylamine hydrochloride (PAH, 58 kDa, Sigma-Aldrich). The 720 uL DI water and 80 uL of 6 mM PAH aqueous solution were sequentially added to 800 uL of 50 nm core nanoparticle solution (7.48*10⁻¹¹ mole of particles/L, BBI). The mixed solution was incubated at 35 °C in the shaker at a shaking speed of 400 rpm for 30 min. Then, the solution was centrifuged at 2040 g for 15 min to remove the unattached PAH molecules. The 160 uL supernatant was removed and the PAH coated particles were resuspended by adding DI water to a final volume of 933 uL. The centrifugation and resuspension steps were repeated three times to insure all free PAH molecules were removed. After the coating of the PAH layer, the zeta potential of the core particles were inverted to + 36 mV. In the second step, 20 nm gold particles (1.16*10⁻⁹ mole of particles/L, BBI) also with negative zeta-potential were attached to the PAH coated core particles. The concentration of core particles was diluted with 2484 uL DI water to 1.75x10⁻¹¹ M and then injected into a 100 mL beaker which had a 17 cm² bottom surface area to increase the reaction surface. The 733 uL of 20 nm gold nanoparticles was divided into ten aliquots and evenly injected on the surface of the core solution. The color of the mixed solution immediately changed from red to blue, which clearly indicated a plasmonic resonance shift due to the linkage between core and satellite particles. The 1 mL solution with CSA and unattached satellite particles was drawn and centrifuged at the lower speed, 700 g for 15 min and then 800 uL red supernatant containing unlinked satellite particles was removed and replaced by 856 uL DI water. Due to the coverage of the satellite particles, the zeta potential of CSA was inverted again to −40 mV.

The silica shell for the CSA can be used as a protection layer for the internal linker molecules and also as the interface for further functionalization, e.g. coating of a specific antibody. For the silica coating, the DI water for the resuspension in the last step of the CSA fabrication was replaced by the same amount of ethanol instead. The 23 uL of 5 mM (3-Aminopropyl)trimethothoxysilane (APTMS, Alfa Aesar) in ethanol was added to the CSA solution and then the solution was shaken for 15 min to form the anchor layer. The 58 uL of 4 mM tetraethyl orthosilicate (TEOS, Showa) dissolved in ethanol was added and then the solution was shaken for another 15 min for full dispersion of the TEOS. After this step, 23 uL of 24 mM NaOH was added to increase the pH value for speeding up the formation of silica. For the growth of the silica layer, the final solution was placed in the shaker at 800 rpm and 35°C. After 24 hr, the solution was centrifuged twice to extract the silica coated CSA (CSA@SiO₂). The acceleration force of the first centrifugation was set at 300 g for 15 min to deposit the larger aggregated particles. The 1 mL supernatant was extracted and centrifuged at 700 g for 15 min to remove residues of the silica by replacing 900 uL supernatant with 100 uL DI water. The fabrication process of CSA and CSA@SiO₂ is illustrated in Fig. 1(a).

 

Fig. 1 (A) Fabrication process of a CSA backbone and Raman tags based on the backbone. The zeta potential (ζ) of particles at each step of forming intrinsic CSA (without Raman reporters) is shown. (B) TEM images of CSA with silica coating (CSA@SiO₂). (C) TEM images of desiccated CSA. The average size of core particles in (B)(C) is 50 nm.

Download Full Size | PDF

For the fabrication of three Raman tags, three Raman reporters: Cy5 modified oligonucleotide, DMcT(2,5-dimercapto-1,3,4-thiadiazole, Sigma-Aldrich), and BDT (Benzene-1,4-dithiol, Sigma-Aldrich), were integrated into the CSA backbone respectively. The fabrication of these Raman tags were the same as in the Fig. 1(a) except that the Raman reporters coated the core particles through the thiol bonding before the adsorption of the PAH layer. For the tags with the Cy5 modified oligonucleotide, the functionalization was described in the previous report [18]. For the Raman tags with DMcT and BDT, the cores particles were immersed in the 56 uM of DMcT for 9 hrs and 2.2 uM of BDT for 1 hr respectively.

2.2 TEM characterization

The morphology of the assemblies was characterized by TEM (200 kV, JEM-2010, JEOL or 80 kV, H7500, Hitachi). To prepare a TEM sample, a 10 uL drop of solution containing assemblies was dripped on a TEM grid (Ted Pella) and then kept in the vacuum dryer to evaporate the excess water.

2.3 Extinction and scattering spectrum

The extinction spectrum was used to identify the resonance wavelength of the assemblies in the solution and acquired by the UV-Vis-NIR spectrometer (U-4100, Hitachi), which had a dual beam design consisting of a signal beam and a reference beam. The spectral non-uniformity and time-dependent intensity fluctuation of the light source (halogen lamp) were eliminated by the reference beam. The background signal (baseline) was acquired by measuring the blank sample (solvent only). The transmitted light through the assemblies suspended in water was collected by the PMT attached on an integrating sphere. The elastic scattering spectrum of the assemblies in a bulk solution were collected through a custom optical scattering system. The white light from a halogen lamp was dispersed by the excitation grating and the light with the selected wavelength was illuminated on the sample in the micro-cuvette. The scattered light was collected by the lens set at 90° from the incident beam and relayed to the collection grating whose operating wavelength was synchronized with the excitation grating.

2.4 Zeta potential

The zeta potential provided the polarity and strength of the surface charge carried by the colloidal particles. Thus, it indicated the stability of the particle suspension. The zeta potential of the particle suspension was measured in each fabrication steps by an electrophoretic light scattering system (NanoPlus-3, Particulate Systems).

2.5 Raman signals

The darkfield scattering images and Raman images/spectra of individual clusters of assemblies was acquired by a custom micro-Raman system. The darkfield scattering images were acquired when the clusters were illuminated by white light from 100 W mercury lamp. For Raman images/spectra, the 13 mW 632.8 nm HeNe laser beam was focused by the 100x objective lens (MPLFLN 100xBDP, Olympus) and illuminated on the assemblies deposited on the quartz wafer which can minimize the background fluorescence. The elastic and Raman scattered light was collected by the same objective lens. The elastic scattered light was removed by the long pass filters (BLP01-633R-25, Semrock) and the remaining Stokes Raman signals were relayed to the spectrometer (SR303i, Andor) equipped with EMCCD (iXon3 888, Andor).

2.6 Electromagnetic simulation

The simulated extinction spectrum was calculated by the finite element method based software (COMSOL 5.2a) in the frequency domain. A CSA or CSA@SiO₂ was embedded in a computational space which was a rectangular parallelepiped with excitation and collection ports assigned on a pair of boundary surfaces. The propagation of the incident wave was perpendicular to the excitation port. Eight satellites are located in each octant respectively.

3. Results and Discussion

3.1 CSA Characterization

The fabrication process of CSA and Raman tags based on CSA is illustrated in Fig. 1(a) and detailed steps are described in the Methods section. For CSA, citrate anions provide substantial negative charge both on the core and satellite particles. Therefore, positively charged polyelectrolyte, polyallylamine hydrochloride (PAH), links the satellite particles immediately after being adsorbed on core particles. For fabrication of the Raman tags, the Raman reporters partially replace citrate anions and sustain negative charge on the core particles. In Fig. 1(b), the TEM image of the desiccated CSA shows that the number of the satellite particles per CSA is around eight. The bare core particles or CSA with few satellite particles are not found, which indicates the high adsorption efficiency of the satellite particles. Nearly all excess 20 nm particles are removed from the solution by the centrifugation step, which implies that the CSA solution with high purity is obtained and suitable for further processing. The gaps between core and satellite particles are the hot spots in which the Raman signal of the linker layer or Raman reporters are highly enhanced. For the silica coated CSA in Fig. 1(c), the thickness of the silica shell is ~20 nm. This thickness can be tuned through controlling the concentration of TEOS and reaction time.

Although the TEM images of the desiccated assemblies in Fig. 1(b) and 1(c) provide the basic morphological information of CSA/CSA@SiO₂ and the quality of the fabrication, they cannot fully elucidate the properties of 3D structures stretching in the solution, such as the core-satellite gap size which substantially affects the resonance wavelength and enhanced gap field. Therefore, the extinction and scattering spectra are utilized to probe the structure of the CSA and CSA@SiO₂ suspended in the solution. The extinction and scattering spectra of the bare core particles, CSA and CSA@SiO₂ in Fig. 2 both show the clear plasmonic resonance shift relative to core particles. For the solution with core particles only, the extinction peak is ~530 nm whereas for CSA structure, the main peak is ~620 nm with the shoulder at around 530 nm. The main CSA peak results from the gap mode located in the junction between core and satellite particles. The 90 nm wavelength shift implies that the junction is extremely small such that the coupling can be substantially strong. The shoulder peak is derived from the dipole mode of the satellite particles. Since the satellite particles are three dimensionally distributed around core particles, the incident wave with single incident angle cannot excite all the gaps at the same time. The satellite particles with line of center connecting the core particle away from the plane of the incident wave have low coupling to the core particles. Therefore, these particles respond to the excitation as if they were free individual particles resonant around 530 nm where the shoulder peak is observed.

 

Fig. 2 Experimental extinction (A) and scattering (B) spectra of core particles, CSA and CSA@SiO₂.

Download Full Size | PDF

The shift of the main extinction peak from 530 nm to 620 nm elucidates the obvious color change from red to blue at the moment when core and satellite particle solutions are mixed. After the silica coating, the resonance of the CSA is further red-shifted to 635 nm due to the slightly higher refractive index of silica (~1.475) compared to water (~1.33). In addition, we notice that the peak of the CSA@SiO₂ is slightly broader than CSA alone. The broadening may result from the formation of the silica shell which causes the inhomogeneous gap size. The scattering spectra of the CSA and CSA@SiO₂ both have main peaks at around 680 nm whereas the scattering peak of core particles is located at 550 nm.

3.2 Simulation Results

The comparison of simulated and experimental extinction spectra can help clarify the detailed structure of the assembly. In Fig. 3(A), the extinction spectrum of 50 nm core particle shows a clear peak at 534 nm, which is consistent with the experimental spectrum in Fig. 2(a). This single particle extinction spectrum is utilized to verify the correctness of the simulation model. To simulate the CSA and CSA@SiO₂ precisely, the accurate determination of the interparticle gap size is necessary. However, it is challenging to estimate the gap size simply from the TEM images of desiccated samples so that the CSA with several gap sizes is first simulated to narrow down the range of the gap size. The refractive index of the PAH layer is assumed to be the same as water for this preliminary simulation. As shown in Fig. 3(b), the coupling between core and satellite particles causes red-shifting of the resonance wavelength, and the amount of the shift is superlinearly increased with the decreasing of the gap size. The resonance shift between single core and CSA with a 10 nm gap is only 13 nm, which shows even though the gap size is only ~1/50 of the wavelength of the incident wave, the coupling is still negligible. The moderate shift (18 nm) occurs between CSA with 10 nm and 5 nm gap size. However, once the gap is below 5 nm, the red-shifting of the resonance become significant. There is another 25 nm resonance red-shift from CSA with 5 nm gap to 2 nm gap. The CSA with 1 nm gap has an additional 23 nm shift to reach the resonance at 614 nm, which is close to the experimental value (620 nm) in Fig. 2(a).

 

Fig. 3 (A) Simulated extinction spectrum of core, CSA and CSA@SiO₂ (B) The CSA resonance wavelength and resonance shift relative to 50 nm core particles versus different gap sizes (C) Calculated field amplitude enhancement of CSA with 1 nm interparticle gap. The |E_inc| is 1.373x10⁸ V/m.

Download Full Size | PDF

According to the simulation in Fig. 3(B) and experimental spectrum in Fig. 2(A), the gap size of CSA would probably be less than 2 nm. From experimental spectrum, the gap mode dominates the extinction spectrum. Therefore, the refractive index of the PAH layer within the interparticle gap substantially affects the resonance wavelength and needs to be considered precisely. According to the literature [31–34], the refractive index of PAH thin film in air varies between 1.45 and 1.65. However, the index of PAH adsorbed on the core particles in solution may have some variation because of the incorporation of water. Thus, In order to probe the CSA structures more accurately, the extinction spectrum of CSA is calculated under the combination of three indices for the PAH layer (n_water~1.33, n_silica~1.475, n_UL = 1.65) and the three gap sizes (1 nm, 1.5 nm, 2 nm). Table 1 lists the peak wavelength of the simulated extinction spectrum of CSA under different parameters. For the 1.5 nm and 2 nm gap size, resonance wavelength of CSA under three refractive indices are less than the experimental resonance (620 nm), which indicates the upper bound of the gap size may be less than 1.5 nm. For the 1 nm gap, the experimental value of the resonance wavelength is between the simulated one under n_water and n_silica. This indicates the index of PAH is probably higher than water but slightly lower than silica. Therefore, in Fig. 3(a), for CSA and CSA@SiO₂, the refractive index of the PAH layer is assumed to be 0.02 lower than silica such that the simulated resonance of CSA is at 619.6 nm, which is consistent with the experimental value. Under the same condition, for CSA@SiO₂, the simulated peak is at 642 nm, which is slightly longer than the experimental value (635 nm). This suggests that the formation of the silica layer may slightly increase the gap size. If the gap size is slightly increased to 1.1 nm, the corresponding wavelengths would be 634 nm, which is consistent with the experimental wavelength. The comparison of simulated and experimental spectra shows that the estimated gap size is ~1 nm and the index of the PAH layer is slightly lower than silica.

Table 1. Simulated resonance wavelength of CSA under three gap sizes and refractive indices of the PAH layer. The experimental resonance wavelength is 620 nm.

View Table

3.3 CSA as the backbone of Raman tags

Comparing the experimental to simulated extinction spectrum of CSA indicates that the estimated gap size is ~1 nm. This gap size is on the edge of the range which classical electrodynamics is still considered to be valid to fully describe the gap field without quantum correction [35]. According to the simulated field around the gap region in Fig. 3(c), the electric field enhancement |E/E_inc| is on the order of ~200, which means the SERS enhancement is on the order of 10⁹. For the single 50 nm gold particle, |E/E_inc| is only ~4. Therefore, the CSA are superior as the backbone of Raman tags.

The dark-field images, Raman scattering images, and Raman spectra of three Raman tags based on the CSA backbone and the intrinsic CSA@SiO₂ are shown in Fig. 4. For Raman Tag1, the Raman fingerprint of Cy5 is clearly demonstrated due to SERS and the resonant Raman effect. For Raman Tag2, the peaks at 656 cm⁻¹ and 1038 cm⁻¹ clearly verify the feature of DMcT. These two peaks represent the vibration of the C-S-C endocyclic bond and N-N bond of DMcT respectively [36]. For Raman Tag3, the most prominent peaks at 1072 cm⁻¹ and 1566 cm⁻¹ correspond to the ring breathing mode and the C = C stretching vibration mode of BDT respectively [37]. For the CSA backbone, the major peak between 1540 cm⁻¹ and 1600 cm⁻¹ results from the symmetric and asymmetric stretching of COO- of the citrate molecules [38]. The minor peak region around 1300 cm⁻¹ includes 1294 cm⁻¹ from the carboxylate deformation of citrate [39] and 1353 cm⁻¹ from C-H deformation of PAH [40]. Therefore, even without any Raman reporter molecules located within the junction of the CSA backbone, the Raman signals from the citrate molecule and PAH can be detected. However, embedding Raman reporter molecules produces a more distinct Raman feature which is suitable for Raman imaging.

 

Fig. 4 From left to right: dark-field images, Raman images and the corresponding Raman spectra of Raman Tag1 (red) and Raman Tag2 (blue), Raman Tag3 (green) and the intrinsic CSA backbone (black).

Download Full Size | PDF

3.4 Stability of the CSA Backbone and Raman Tags

To investigate the stability of the CSA backbone, the Raman intensity of the CSA@SiO₂ clusters was monitored under continuous laser illumination because the Raman intensity of linker molecules highly depends on the gap size. The intensity of the incident laser is ~2000 W/cm², which is two orders higher than the one normally used in photothermal therapy. The Raman spectra of four different clusters of CSA@SiO₂ are acquired at 0 min, 5 min and 10 min. The Raman intensities at 1300 cm⁻¹ and 1550 cm⁻¹ are extracted. For each cluster, the Raman intensities at 5 min and 10 min are normalized by the intensity at 0 min. As shown in Fig. 5, the average intensity of CSA@SiO₂ cluster from 1300 cm⁻¹ is 10% lower than the initial value. The standard deviation of the intensity at 10 min is slightly larger than at 5 min. The intensity at 1550 cm⁻¹ shows the same tendency. Thus, stable Raman spectra emitted from the gap indicate the robustness of the CSA@SiO₂.

 

Fig. 5 Normalized Raman intensity, (A) 1300 cm⁻¹ (B) 1550 cm⁻¹, of four CSA@SiO₂ clusters acquired at 0 min, 5 min and 10 min over the period of continuous laser illumination. (C) Normalized Raman intensity of the three Raman Tags and normalized fluorescence intensity of free Cy5 acquired over the period of continuous laser illumination.

Download Full Size | PDF

The stability of the Raman intensity from three Raman tags is also tested and compared with the fluorescence of free Cy5 molecules. As shown in Fig. 5(c), during the 10 min illumination, Raman Tag2 and Tag3 show comparable stability as CSA backbone. The Raman signals from Tag2 (658.9 cm⁻¹) and Tag3 (1566.4 cm⁻¹) have only a 10% intensity fluctuation. However, for the Raman Tag1, because part of signal resulting from resonant Raman effect decays due to the photobleaching of Cy5, the total Raman signal intensity of Raman tag1 rapidly decreases to 18% in the first minute and then slowly decreases to ~5% after 5 min of illumination. For the free Cy5, the fluorescence immediately decays to <1% in the first minute. The Raman tags embedded with non-fluorescent reporters (DMcT, BDT) show much higher stability than with fluorescent reporters (Cy5) and free Cy5.

4. Conclusion

The plasmonic core-satellite assembly with a ~1 nm gap and extremely high gap field is efficiently fabricated simply through polyelectrolyte coating without complex surface modification of particles. Three distinct Raman tags consisting of this CSA backbone and embedded Raman reporters are demonstrated. The CSA backbone is robust under strong laser illumination. The Raman tags embedded with non-fluorescent reporters (DMcT, BDT) show excellent stability.

This polyelectrolyte induced assembly method cannot only be applied to plasmonic nanoparticles but also non-metallic nanoparticles as long as they carry stable surface charge. The plasmonic CSA demonstrated here can be extended to various combinations of core and satellite particles with differing size/shape/composition for tailoring the desired resonance wavelength and gap field. The stable high field greatly increases the brightness of Raman tags and make them more comparable to the fluorescent molecules widely used in the biomedical field. Therefore, our method will greatly facilitate the development of various Raman tags for multiplexed imaging and sensing. Furthermore, the CSA with an internal high field may be used for the application relying on internal high field such as quantum strong coupling.