Virus-templated plasmonic nanoclusters with icosahedral symmetry via directed self-assembly.

The assembly of plasmonic nanoparticles with precise spatial and orientational order may lead to structures with new electromagnetic properties at optical frequencies. The directed self-assembly method presented controls the interparticle-spacing and symmetry of the resulting nanometer-sized elements in solution. The self-assembly of three-dimensional (3D), icosahedral plasmonic nanosclusters (NCs) with resonances at visible wavelengths is demonstrated experimentally. The ideal NCs consist of twelve gold (Au) nanospheres (NSs) attached to thiol groups at predefined locations on the surface of a genetically engineered cowpea mosaic virus with icosahedral symmetry. In situ dynamic light scattering (DLS) measurements confirm the NSs assembly on the virus. Transmission electron micrographs (TEM) demonstrate the ability of the self-assembly method to control the nanoscopic symmetry of the bound NSs, which reflects the icosahedral symmetry of the virus. Both, TEM and DLS show that the NCs comprise of a distribution of capsids mostly covered (i.e., 6-12 NS/capsid) with NSs. 3D finite-element simulations of aqueous suspensions of NCs reproduce the experimental bulk absorbance measurements and major features of the spectra. Simulations results show that the fully assembled NCs give rise to a 10-fold surface-averaged enhancement of the local electromagnetic field.

1 Introduction

The ability for metamaterials to tailor the interactions of electromagnetic fields and matter, enabling unique optical properties,1,2 makes them exciting candidates for disruptive technologies, such as sub-wavelength imaging and transformational optics.3,4 It has been predicted that a collection of six nanospheres (NSs) arranged with a cubic symmetry may constitute an optical magnetic metamolecule.5,6 The effectively isotropic, spherical geometry of the six NSs is equivalent to six split-ring resonators,5,7,8 providing the magnetic dipole moment at optical frequencies. However, significant fundamental assembly challenges exist to develop materials exhibiting these exotic properties.9–12 Traditionally, various top-down lithographic techniques1,13–15 have been used to produce nanostructures. Typically, the processes are expensive, producing predominantly 2D fixed structures with limited feature resolution. Conversely, bottom-up self-assembly approaches offer a powerful and simpler route for creating metamaterials. Developing novel methods to assemble the nanoscale constituents into the desired architecture remains a rich area of research.

Recent efforts to self-assemble plasmonic NCs have utilized silica or polystyrene spheres as scaffolds for self-assembling gold nanoparticles on their surface, or by flocculating polymer coated NSs.12,16,17 These approaches result in random, disordered assembly of the NSs with no precise control over the interparticle distances. On the other hand, DNA-assisted assembly18,19 enables one to control the interparticle symmetry of nanometer sized elements, but has low-throughput, is temperature sensitive, and there is a significant limitation on the size of cluster that can be created. Our strategy is to utilize the underlying symmetry of viruses to define the location of the NSs on the scaffold. The availability of viruses with different sizes and symmetries, as well as the ability to genetically-engineer the position of the NS attachment makes this a powerful approach to create the plasmonic NCs.

Cowpea mosaic virus (CPMV) is a well-characterized plant virus produced in grams amounts20 and previously used by us21,22 and others23 for material applications. The crystal structure of the approximately 30 nm diameter protein capsid is available at 2.8 Å resolution.24 Its spectroscopic characteristics25 and bioconjugation via its native groups26–28 and engineered thiols29 is well characterized. In addition, stability over broad ranges of temperature, pH, buffer, and organic solvents makes it amenable for its utilization for assembling nanoparticles.30,31 Finally, its icosahedral symmetry and well-established biochemistry makes it a great candidate for the fabrication of an icosahedral nanocluster, where the NSs have a well-defined inter-particle spacing within the nanocluster.

Previous reports using viruses32 and protein viral components33 as scaffolds to organize nanoparticles showed promising results for assembling small size nanoparticles (<10 nm) on the surface of the capsid.21,30 However, the NSs were not of the adequate size to significantly couple the near-fields. In order to create the desired optical properties the NSs need to be comparable to the size of the virus (30 nm). Here, we experimentally demonstrate a self-assembled, isotropic, 3D, plasmonic NC comprising of a genetically engineered cowpea mosaic virus (BC-CPMV) bearing covalently attached Au NSs, with diameters comparable to the diameter of CPMV, at predefined locations to form icosahedral symmetry between the NSs on the surface of CPMV. Using in situ DLS measurements, we confirm the NSs to be assembled to the virus. TEM images verify the nanospheres are at pre-defined locations on the virus's surface. We measured the absorbance from bulk aqueous suspensions of NCs and reproduced the major features of the spectrum using 3D finite-element simulations.

2 Results and Discussion

We engineered CPMV to present cysteines groups (SH) at the BC-loop for a total of 60 thiols per capsid. The thiols are organized in groups of five each located at the 12 vertices of the icosahedron (Figure 1a). Covalently attaching Au NSs with diameters greater than ≈15 nm to the thiols allows up to twelve NSs to attach to CPMV while maintaining the icosahedral symmetry between the NSs in the resulting NC.

Figure 1

(a) BC-CPMV protein structure (PDB:1NY7). The inset shows the protein subunit; in pink is a single cysteine (thiol containing amino acid) at the BC-loop resulting in a total of 60 thiols per capsid. The dotted orange lines represent a five-fold symmetry axis. (b) Representative dynamic light scattering spectrum at completion, t = 36 hrs, for the NC self-assembly reaction with 17 nm NSs (see Supporting Information S2.2). The inset schematic depicts the NC self-assembly reaction.

A schematic for the directed self-assembly reaction of the NCs is shown in Figure 1b. Au NSs ranging from 17 nm to 34 nm in diameter were attached to the virus (refer to the Supporting Information for complete assembly details and data of all NCs built, section S1.1.2, Table S1). Briefly, Au NSs are reacted with virus in 10 mM potassium phosphate, 1mM EDTA at pH 6.0 for 36 h at room temperature. The reaction with 17 nm diameter Au NSs was monitored by DLS, Figure 1b. Initially at t = 0 there is only one peak corresponding to the isolated Au NSs since they scatter more strongly than the virus at the concentrations employed (see Supporting Information, section S2.3). As the reaction progresses the Au NSs begin to bind to the thiol groups on the virus and the scattering peak from nanospheres begins to shift toward larger diameters. At approximately t = 36 h the peak reaches the expected diameter, for the example provided in Figure 1b, ≈dCPMV + 2dNS = 64.5 nm.

Results from our self-assembly optimization experiments (Supporting Information, section S2.2) indicated that as the Au NS diameter increased, >15 nm, the amount of the NSs attached to the virus decreased (Figure S4). We found that the manner in which the Au NSs were synthesized, concentrated (20×) and aged and the quantity/type of surfactant used to stabilize the NSs are critical to attach large diameter NSs to the virus while maintaining the colloidal stability of the NS to avoid massive aggregation. Au NSs synthesized using Puntes method34 resulted in the highest reactivity and NC quality. We found no reliably large diameter commercial Au NSs capable of reacting with the virus, presumably due to unknown/disclosed surfactants or unknown NS age (Figure S2).

Post reaction the suspension is purified via agarose electrophoresis. Significant amounts of unreacted free NSs are present in the suspension as expected, since a 20× excess of NSs are used for the reaction. Coupling thioctic acid to the unreacted NSs and NSs bound to the capsid creates a negative surface charge on both the NCs and NSs, permitting separation via electrophoresis. The free NSs migrate faster than the NCs under the influence of the electric field, resulting in distinct population bands within the gel (Figure 2a). These results are in agreement with larger diameter NCs migrating slower than the smaller free NSs.35 Separated gel pieces were treated with β-agarase to recover the NCs and free NSs (Figure S7) from the agarose gel. The resulting materials were re-suspended with Milli-Q water, washed twice using 100 k molecular weight cut-off (MWCO) centrifugation units and were used for DLS, TEM, and spectroscopy measurements. DLS measurements were performed on the NC suspensions post-purification. The polydispersity = σ, where σ is the standard deviation and d is the mean diameter, was typically much less than 10% indicating modest NC monodispersity.

Figure 2

(a) Image of a post-reaction suspension in the electrophoretic agarose gel. The dark red band is free NSs (top) and the purple band is the NCs (middle). (b) Image of free NSs (left side) and NCs (right side) re-suspended in an aqueous solution after gel purification. (c) False-colored TEM image of a NC with 18 nm Au NSs attached. (d) Model of a NC with 30 nm Au NSs attached and similar orientation to (c). The dotted black lines represent a five-fold symmetry axis (e) Representative TEM images of NC with 30 nm Au NS attached.

Figure 2b shows the free Au NSs and NCs suspended in aqueous buffer, placed into a 10 mm path length cuvette, and imaged while backlit with white light. The free Au NS suspension has a characteristic red color but the NCs are dramatically different with a dark purple color. The particle number density for both suspensions is approximately 1011 particles mL−1. Figure 2b demonstrates the feasibility for this self-assembly strategy to produce bulk NC quantities. For example, based on 1011 NC mL−1 and assuming twelve 30 nm diameter Au NS attached to the virus gives approximately 4 μg mL−1 of NC material.

A false-colored TEM image of a NC (built with 18 nm diameter NSs) demonstrating five-fold rotation symmetry between the NSs is shown in Figure 2c. A model of the NC is shown in Figure 2d (discussed below) and is in good agreement with the TEM images in Figure 2e. Demonstrating that the NSs are bound at the thiol groups as postulated since they show five-fold symmetry. We were unable to image the virus inside the NC, even after staining with 2% uranyl acetate (data not shown), in the TEM images, presumably due to the large contrast between the virus and the Au NSs. If virus was not added to the self-assembly reaction no NCs were observed in the DLS, agarose gel, and TEM data. The free Au NSs after agarose purification were imaged with the TEM and no NSs out of plane (3D) or NS groups with five-fold symmetry were observed (Figure S7) and no evidence of NCs is observed in DLS (Figure S6b). Figure 2e shows a number of TEM images of NCs (built with 30 nm diameter NSs) providing a representative sample distribution.

The NCs simulations were constructed by initially having one NS attached at the north pole of the virus. The north pole was designated as the center NS in Figure 2d. The next NS was placed in a nearest neighbor position unit until the northern hemisphere had all six NS attached. The southern hemisphere was done in the reverse order with the south pole being the last NS attached. The northern and southern NSs were slightly offset, breaking the icosahedral symmetry, to better represent the experimental TEM, absorbance, and crystallographic CPMV data (Figure S6). Using the NC structure shown in Figure 2d, 3D finite-element simulations were undertaken using COMSOL Multiphysics 4.3a (section S1.5) to retrieve the absorbance spectrum for the NCs in an aqueous suspension for direct comparison to experiments. The calculated absorbance spectra for the NCs as a function of the number of 30 nm diameter Au NSs attached (1–12) to the virus is shown in Figure 3a. Once the NC has 50% (6) of the NSs attached a broad shoulder develops from 550–650 nm and an asymmetric ‘Fano-like’ peak emerges at 685 nm due to the slight icosahedral symmetry breaking.36,37 As the number of NSs attached to the CPMV increases the absorbance peak from the individual NSs at 518 nm broadens and redshifts to 537 nm for the fully assembled (12 NS) NC (Figure 3a, purple line).

Figure 3

3D finite-element simulations: (a) Calculated absorbance spectra for the NCs as a function of the number of 30 nm diameter Au NSs attached (1–12) to the virus. (b) Calculated normalized absorbance spectra for an isolated 15 nm diameter Au NS (red), NC with twelve 15 nm diameter Au NSs (cyan) and twelve 30 nm diameter Au NSs (purple).

Figure 3b shows the calculated normalized absorbance spectra for an isolated 15 nm diameter Au NS (red), NC with twelve 15 nm diameter Au NS (cyan), and NC with twelve 30 nm diameter Au NS (purple). For the case of the 15 nm NC, the interparticle distance between the NSs is 7.90 nm (approximately one radius) leading to only weak near-field coupling resulting in small shifts in the absorbance spectrum relative to the isolated 15 nm NSs.38 The 30 nm NCs however have interparticle gaps of 0.79 nm, providing significant near-field coupling and enhancements between NSs giving rise to large changes to the absorbance spectrum.

The bulk experimental absorbance spectra from isolated 30 nm diameter Au NSs (red) and NCs (purple) with 30 nm diameter Au NSs attached to the virus in aqueous suspensions are shown in Figure 4a and Figure S6. The NC absorbance changes significantly from the isolated NSs. The absorbance peak from the individual NSs at 524 nm slightly red shifts to 535 nm going from the isolated NSs to the NCs. The absorbance peak also broadens with a quality factor, λ0/Δλ, decreasing from approximately 5.8 to 1.9. From 550 nm to 675 nm a broad shoulder develops in the NC absorbance spectrum. There is also the emergence of another reproducible but small peak/shoulder at 675 nm. The broadening of the peak is attributed to the polydispersity of the NCs.

Figure 4

(a) Experimental bulk absorbance spectrum for NCs (purple) with 30 nm diameter Au NSs attached to the virus and isolated 30 nm diameter Au NSs (red; negative control) in an aqueous buffer solution. (b) Normalized absorbance spectra comparing the experimental (purple) and averaged simulation (black) from 30 nm diameter Au NSs attached to the virus forming the NCs.

To compare the experimental results with the simulations the five spectra from Figure 3a were averaged, weighting each spectrum equally, to approximate the experimental NC distribution. The normalized simulated (black) and experimental (purple) absorbance spectra for NCs with 30 nm diameter Au NS attached to the virus are presented in Figure 4b. The absorbance peak from the individual NSs occurs at 535 nm and 537 nm for the experiment and simulation, respectively. Both spectra show a broad shoulder developing around 550 nm and continuing to 675 nm for the experiment and 650 nm for the simulation. The small peak/shoulder emerging at 675 nm from the experimental spectrum is at the same position as the fully assembled simulated NC (Figure 3a) with the ‘fano-like‘resonance peak emerging at 685 nm demonstrating that a fully covered NCs is not a prerequisite to observe the predicted fano-like peaks in the visible side of the spectrum.

Based on the simulations the peak at 685 nm occurs as a result of at least 50% (6) of the 30 nm diameter NSs being assembled on the virus (Figure 3a) in agreement with our TEM (Figure 2e) and DLS measurements. The simulated spectrum also shows a rapid decrease in the absorbance from 650 nm to 1000 nm and experimentally a boarder absorbance tail continues beyond 675 nm. Overall, the quality factor for the experimental absorbance peaks decreases with respect to the simulation results. The broadening of the peaks observed in the NC absorbance measurement may be partially explained by the size distribution, inter-particle separation/placement, and number of NSs attached to the virus. Despite these issues, the major features of the experimental absorbance spectrum (535 nm peak and 675 nm peak/shoulder) are well predicted by the simple model.

Simulations were performed to determine the expected electric and magnetic fields in an ideal NC. Figure 5a shows the comparison of the magnitude of the electric field averaged over the surface of the one (red) and twelve (purple) 30 nm diameter Au NSs attached to the virus. The surfaced-averaged electric field is approximately unity for the single NS attached to the virus with a slight increase around resonance at 520 nm. For the fully assembled NC (12 NS) the close proximity between the NSs on the virus (0.79 nm) gives rise to large ≈10-fold enhancements of the local electromagnetic fields through near-field coupling, though the electric field maxima are several orders of magnitude larger than the surface average. An intensity map of the magnitude of the electric field on a plane slicing orthogonally through one of the 5-fold symmetry axis of the NC at 700 nm is shown in Figure 5b. The coupling of the near-fields produced a displacement current circulating around the 5-fold symmetry axis of the NC, inducing a magnetic response.5

Figure 5

3D finite-element simulations: (a) Calculated surface-averaged electric fields for twelve 30 nm diameter Au NSs attached to the virus (purple; fully assembled NC) and one 30 nm diameter NS attached to the virus (red). (b) A plane slicing through the fully assembled NC showing the electric field distributions. (c) Calculated surface-averaged magnetic fields for the fully assembled NC (purple) and one 30 nm diameter NS attached to the scaffold (red). (d) A plane slicing through the fully assembled NC showing the magnetic field distributions.

The magnitude of the calculated magnetic field averaged over the surface of the one (red) and twelve (purple) 30 nm diameter Au NSs attached to the virus is shown in Figure 5c. For the single NS attached to the virus there is nearly no response, although there is a non-unity offset and small dip around 520 nm that is not well understood. However, for the fully assembled NC there is a significant magnetic response to the impinging field, relative to the single NS attached to the virus, demonstrating the feasibly for these NCs to provide a magnetic response at visible frequencies. Figure 5d shows the intensity map of the magnetic field on a plane slicing through the center of the NC at 720 nm, demonstrating a magnetic ‘hot-spot’ at the center of the NC.

An open-ended question is the experimental realization of a magnetic response from these NCs at visible frequencies. Based on our simulations and recent reports39 the magnetic response of the NCs is very sensitive to the positioning of each NS. We anticipate fundamental challenges to self-assemble and measure the magnetic response from bulk NC ensembles at visible wavelengths. In pursuit of this objective, future experiments will focus on measuring the optical response from the NCs and correlate with the calculated values presented herein.

3 Conclusion

We have demonstrated a self-assembly strategy to create three-dimensional, icosahedral plasmonic nanosclusters. In situ dynamic light scattering experiments confirm the nanosphere-virus assembly. Transmission electron microscopy images demonstrate the nanospheres to be assembled at fixed locations on the icosahedral virus's surface. Our results indicate that the NCs suspension is comprised of a distribution of capsids mostly covered (i.e., 6–12 NS/capsid). We measured the bulk absorbance from aqueous suspensions of nanoclusters and reproduced the major features of the spectrum using three-dimensional finite-element simulations. Furthermore, because the viruses are easily produced in gram quantities the self-assembly process is capable of high-throughput, providing a strategy to realized macroscopic quantities for metamaterial applications.

4 Experimental Section

All experimental details: synthesis of gold nanoparticles, BC-CPMV purification, nanoclusters assembly, optimization experiments, and additional data of all the NCs built are included in the Supporting Information.