Plasmonic nanoparticle clusters have attracted great attention due to the unique capability to manipulate electromagnetic fields at the sub-wavelength scale1-5. Ensembles of metallic nanoparticles generate various electromagnetisms at optical frequencies such as artificial magnetism6-10 and Fano-like interference11-13 and a strong field localization in the structure14-16. These unique properties are geometry-dependent and lead to a broad range of applications in sensing16,17, surface-enhanced spectroscopies18-22, nonlinear integrated photonics23,24, and light harvesting25,26.
Traditionally, plasmonic clusters with tailored size and geometry are fabricated on substrates by top-down processes such as electron-beam lithography4,5 or focused-ion beam milling27,28. These approaches suffer from low throughput and are generally limited to in-plane fabrication. Alternatively, the self-assembly of colloids has been proposed as a versatile, high-throughput, and cost-effective route. A number of clever methods based on chemical linking (e.g., DNA origami)29-30 and/or convective assembly on lithographically structured templates25,26,31 have been devised to construct 2D or 3D plasmonic clusters. The shape formation, however, is mostly constrained by the thermodynamic impetus and/or template geometry. A significant challenge would be overcome these restrictions and expand structural design freedom in the fabrication of plasmonic cluster architectures with symmetry-breaking geometries.
In this work, we develop a freeform, programmable 3D assembly of of hierarchical plasmonic clusters (HPCs). By exploiting micronozzle 3D printing, we demonstrate highly localized, omnidirectional meniscus-guided assembly of metallic nanoparticles, constructing a freestanding HPC with a tailored geometry that can control the far-field character. Our approach also allows versatile manipulation and exploitation of the near-field interaction in the HPC by a facile heterogeneous nanoparticle mixing. We demonstrate that 3D-printed HPCs can be utilized as an ultracompact surface-enhanced Raman spectroscopy (SERS) platform to detect M13 viruses and their mutations from femtolitre volume, sub-100pM analytes.
3D printing of HPCs
A key approach that enables the fabrication of freeform hierarchical plasmonic clusters (HPCs) is the exploitation of a femtolitre (fL) liquid meniscus32,33 to confine and guide the self-assembly of functionalized nanoparticles in three dimensions, i.e., the meniscus-guided 3D nanoassembly process. Figure 1 depicts the core of our experimental arrangement. Gold nanoparticles (AuNPs) having 100 nm diameters and functionalized with 50 kDa polystyrene (PS) were suspended in toluene and used as the primary building blocks to promote strong localized surface plasmon resonance (LSPR). The role of PS was to ensure an interparticle distance of tens of nanometres among the AuNPs after the self-assembly. A glass micropipette (having an aperture diameter of 3 µm, Supplementary Fig. S1) was filled with the AuNP suspension and used to 3D print HPCs. As depicted in Fig. 1a, a fL-volume, AuNP-laden solution meniscus is formed at a localized area by the micropipette. The continuous supply and assembly of AuNPs are driven by fast evaporation of the toluene solvent at the meniscus surface. We note that this highly localized AuNP assembly can omnidirectionally grow with the guidance of pipette movement, resulting in a freestanding, freeform AuNP cluster with controlled interparticle spacing. This approach can also be used to print freeform heterogeneous clusters, as shown in Figure 1b. Various hybrid inks prepared by mixing AuNPs and functional nanomaterials (e.g., dielectric particles, quantum emitters, or biomolecules) can be directly used to fabricate plasmonic-based heterogeneous 3D architectures. The series of optical micrographs in Fig. 1c show the printing process consisting of (i) micropipette approach toward a silicon (Si) substrate, (ii) meniscus formation by pipette-substrate contact, (iii, iv) continuous 3D assembly of the HPC by meniscus guiding, and (v) termination. Continuous HPC assembly results from maintaining an adequate pipette moving speed; a speed of 3 µm/s was used herein. Termination of the assembly process is conducted by abruptly and sufficiently increasing the speed so as to disrupt the meniscus. Figure 1d schematically depicts the multiscale features of an HPC. At the tens of nm scale, near-field coupling among the AuNPs, which have an interparticle spacing controlled by PS, determines the plasmonic characteristics of the HPC. At the µm scale, the 3D structure, consisting of thousands of assembled AuNPs, confers its far-field characteristics and allows the HPC itself to act as a free-space optical system. Figure 1e shows field-emission scanning electron microscopy (FE-SEM) and transmission electron microscopy (TEM) images for a resulting 3D-printed HPC with controlled geometry and interparticle spacing. The result shows that the meniscus-guided nanoassembly approach provides unprecedented simplicity and structural design freedom in construction of 3D HPCs.
Plasmonic characteristics of HPCs: experimental and theoretical studies
A notable feature of the 3D printed HPC is that its plasmonic characteristics are highly consistent and uniform throughout the entire structure. To examine the AuNP distributions and corresponding LSPR spectra, we fabricated a straight HPC micropillar that was ~2 µm in diameter and ~20 µm in height (aspect ratio: 10) as a model specimen (Fig. 2a). A FE-SEM image taken after focused ion beam (FIB) milling (inset of Fig. 2a) clearly shows that the AuNPs are uniformly distributed, having a 20-30 nm interparticle spacing, within the fabricated micropillar. The LSPR of the micropillar was investigated with dark-field spectroscopy. The LSPR spectrum is composed of a superradiant bright mode with a broad peak and a subradiant dark mode with a narrow dip (Fig. 2b). The LSPR is consistent along the vertical pillar axis. The spectra obtained from the bottom (i) and the top (v) of the micropillar do not exhibit noticeable differences. No significant shift is observed in the broad peak, and the subradiant dip wavelength (lsub,exp = 729 ± 7 nm) is maintained. The LSPR spectra of the HPC depend on the quantity of AuNPs in the measured region. To investigate this, LSPR spectra for AuNP clusters with different particle numbers (N) were simulated by the finite-difference time-domain (FDTD) method. A detailed description of the simulation is provided in the Methods and Supplementary Information (Supplementary Fig. S2 and Fig. S3). In this study, the particle diameter and interparticle spacing were set to 100 nm and 32 nm, respectively. The designated geometries of AuNP clusters with N from 2 to 5 for the simulation are depicted in Fig. 2c. Figure 2d clearly shows the LSPR dependence on N. Specifically, as N increases from 2 to 5, the simulated LSPR spectra sequentially redshift and broaden due to the enhancement of the superradiant mode with increasing radiative damping5. A further 3D model was obtained by vertically stacking N = 5 clusters. The simulated LSPR spectrum for N = 25 (solid line in grey, Fig. 2d) exhibits a broad superradiant peak with a subradiant dip at λsub,cal = 730 nm, consistent with the experimental data (λsub,exp = 729 nm). The surface charge maps clearly show characteristic surface charge distribution for the superradiant mode and the subradiant mode5,13. (Supplementary Fig. S4) The peak in the absorbance spectrum for the N = 25 cluster (λabs,cal = 741 nm; solid line in blue) is also located near λsub,cal and λsub,exp, consistent with the existence of a subradiant mode. These results show that the simulation performed for the N = 25 cluster well predicts the plasmonic characteristics of a printed HPC pillar.
Freestanding, freeform HPCs
The meniscus-guided method enables omnidirectional printing of HPCs with tailored geometries, providing a facile platform for manipulating light propagation at the nanoscale. In other words, the far-field properties of an HPC can be controlled by its 3D geometry. Figs. 2e-2h show the array of 3D shapes that were prepared using our method. These 3D HPC structures were readily fabricated by guiding the meniscus, with control of the printing direction being rapid and programmable. Figs. 2e, 2f, and 2g show 2-fold, 3-fold and 4-fold freestanding HPC pillars, respectively. A top-view SEM image of the 4-fold freestanding HPC pillar is shown in Fig. 2h. Additional model 3D structures are provided in Supplementary Fig. S5. We compared the light guiding direction in the 2-fold HPC pillar (having a folding angle of 90°; Fig. 2i) to that in the straight pillar previously discussed (Fig. 2j). Fig. 2k shows the directional dependences of the scattering signals from these two HPC pillars, collected by an objective lens located above the samples under side light illumination. Note that the scattering signal generated by LSPR is partially guided along the pillar direction. As a result, the 2-fold HPC pillar, for which the guided mode is excluded from the objective lens (Fig. 2i), shows a smaller collected scattering intensity than that of the straight HPC pillar. The scattering signal for the 2-fold HPC pillar has directionality according to the light illumination; the intensity is greater when the incident light is perpendicular to the length direction of the structure (90° and 270°)34.
Heterogeneous HPCs
In addition to the control over the far-field properties by tailoring 3D geometry, the use of hybrid printing ink provides a versatile route to tune and exploit the near-field optical properties of heterogeneous HPCs. The interparticle spacing of AuNPs in the HPC can be precisely controlled by adding secondary spacer particles such as 20 nm silica (SiO2) (Fig. 3a). This approach does not require a complicated lithography process or rigorous chemical synthesis. The interparticle spacing was controlled in this study by adjusting the ratio of the SiO2 NPs to AuNPs from NSiO2/NAu = 0 to NSiO2/NAu = 0.005 or NSiO2/NAu = 0.01. The FDTD model was also utilized to predict the dependence of the LSPR spectrum on the interparticle spacing within the HPC. As shown in Fig. 3b, when NSiO2/NAu = 0, the subradiant mode of the HPC is observed at λsub,exp = 733 nm, which matches well with the simulated spectrum for a 32 nm interparticle spacing (Fig. 3c). Upon increasing NSiO2/NAu from 0.005 to 0.01, the LSPR spectrum is shifted from λ = 684 nm to 632 nm (Fig. 3b), consistent with the simulated spectra obtained for 52 nm and 72 nm interparticle spacings (Fig. 3c). The change in the interparticle spacing with the NSiO2/NAu ratio was confirmed experimentally by the FE-SEM images in Supplementary Fig. S6.
A quantum dot (QD)-embedded HPC was also prepared and demonstrated outstanding near-field fluorescence enhancement and far-field radiation. As schematically depicted in Fig. 3d, a 3D QD-embedded HPC was fabricated by mixing AuNPs and cadmium selenide (CdSe) QDs in an ink solution. This approach offers a simpler strategy to embed QDs within AuNP clusters than other existing methods35-37. Time-resolved (TR) photoluminescence (PL) measurements were performed to confirm the enhancement of the QD fluorescence. Figure 3e compares the TRPL spectra for CdSe QDs on a glass substrate and CdSe QDs within the HPC. The decay time for QD fluorescence coupled to the HPC is 3 ns, which is shorter than the observed 14 ns for QDs on glass. This drastic decrease in decay time is evidence of the strong near-field enhancement of the HPC.
The far-field characteristics of the QD-embedded HPC can be manipulated by its 3D geometry at the microscale. Figure 3f compares the PL spectra of a vertical QD-embedded HPC pillar and a collapsed pillar (Supplementary Fig. S7) having the same QD concentration. The PL signals were collected by an objective lens located above the sample. The PL signal collected from the vertical pillar is 19 times brighter than that from the collapsed sample. These results show the ability of the freestanding HPC pillar to confine and guide the propagation of an electromagnetic wave along a specific direction. The far-field property of the QD-embedded freestanding HPC pillar was simulated by the finite element method using a symmetric N = 95 cluster model schematically illustrated in Fig. 3g. The results show more directed, confined radiation characteristics than the QD on a dielectric film (Fig. 3h and Supplementary Fig. S8), resulting in increased QD emission collection efficiency. We expect that this heterogeneous HPC approach will offer a facile strategy to design the far-field radiation of quantum emitters in three-dimensions.
HPC-based SERS platform for virus detection
Our heterogeneous HPC approach allows for the placement of small molecule analytes at plasmonic hot spots during the self-assembly process by simple mixing, directly printing an ultracompact surface-enhanced Raman spectroscopy (SERS) platform, as depicted in Fig. 4a. The 3D-printed SERS microplatform enables high-sensitivity diagnosis of diseases or pathogens from a femtolitre-volume analyte), which can be easily collected from blood, tear, or urine samples. For this study, AuNPs having 75 nm diameter and functionalized with polyvinylpyrrolidone (PVP) were used. The PVP-coated AuNPs had smaller interparticle spacing than the PS-coated one, leading to increased electromagnetic field enhancement. The performance of the 3D-printed SERS microplatform (HPC) was tested first using a thiophenol analyte, which showed a clear Raman signal enhancement over that embedded in the control PS pillar (Supplementary Fig. S9). To demonstrate the practical use of this microplatform for bio-sensing, we performed SERS detection of M13 bacteriophages, which are filamentous viruses (width: ~ 6.6 nm, length: ~ 880 nm) that have a well-established mutagenesis protocol38. M13 bacteriophages were effectively placed in the nanogaps between AuNPs during 3D printing of the HPC due to their nanometre-width. The printed HPC can be used as a SERS microplatform to detect M13 functional peptides. To prove the detection capability, we tested the wild-type M13 bacteriophage (WT phage) and three types of mutant M13 bacteriophage – 4E phage (Glu-Glu-Glu-Glu), 3H phage (His-His-His), and 3Q phage (Gln-Gln-Gln) –prepared by site-directed mutagenesis polymerase chain reaction (PCR) at the +81 nucleotide sequence position of the N-terminus of the major coat protein (pVIII), as illustrated in Fig. 4b. Each M13s were simply mixed with the AuNPs and used to produce an array of M13-embedded HPC micropillars, as shown in the top-view dark-field scattering images of Fig. 4c. Figure 4d shows the SERS spectra for WT-phage-embedded HPC pillars, having WT phage concentrations ranging from 0 to 300 pM, under 633 nm laser illumination. The obtained Raman spectra arising from the capsid protein enabled WT phage M13 sensing at tens of picomolar concentration. Figures 4e-h show SERS spectra for the WT-phage, 4E-phage, 3H-phage and 3Q-phage-embedded HPCs, which clearly exhibit their characteristic SERS spectra. Figure 4i-l show the fingerprint patterns by selected Raman responses for each mutant phage. The differentiated intensity of SERS spectra was used for the fingerprint patterns to prevent distortion due to baseline differences. The fingerprint patterns clearly visualize differences in the SERS peak positions for different phage genetic types. Principal component analysis was also able to discriminate the various SERS spectra (Supplementary Fig. S10). The SERS spectra obtained from different HPC locations showed high consistency, suggesting that reliable diagnostic results may be obtained. The results of this study indicate the potential for HPCs to not only diagnose specific viruses, but also detect their mutations in a facile and flexible manner. We expect that the application of HPCs will be extended to the diagnosis of pathogens and diseases in the near future.