Introduction

Arranging individual nanoparticles into specific and regular assemblies is a powerful tool for practical applications in electronics1,2,3, photonics4,5, catalysis6, plasmonics7,8,9, sensing10, mechanics11, and bioanalysis12. Several techniques have been established to develop assembly protocols, e.g., lithography patterning13, DNA-mediated assembly14,15,16, polymer-grafted assembly17,18, and colloidal self-assembly19,20,21,22,23,24, and others25. Among them, colloidal self-assembly is one of the most promising methods for developing superstructures. This approach is facile, versatile, scalable, and cost-effective, making it applicable to a wide range of nanoparticles with varying shapes, sizes, crystallinities, and compositions. In natural processes, colloidal self-assembly tends to maximize interparticle interactions by a close-packed system driven by van der Waals forces, electrostatic forces, and other physicochemical interactions26,27, leading to densely packed assemblies with highly ordered periodicity. However, superstructures aligned with tip-to-tip manner that defy the natural thermodynamic principles of nanoparticle crystallization have been difficult to achieve. In this study, we explored a multistep process to create an artificially engineered architecture: three-dimensional tip-to-tip octahedral supercrystals. These supercrystals maintain their morphology via tip-exclusive interactions, forming highly regular peculiar nanoporous networks with all-face exposed surface. We utilized template-assisted strategy to assemble colloidal nanoparticles into organized complex patterned array of superstructures. By adopting epitaxial silver (Ag) growth on colloidal gold (Au) octahedral nanoparticles, the nanoparticles evolve into nanocubes with Ag shells as hard template, extending until the inner Au vertices touch the faces of Ag cubes. Subsequently, the slow evaporation of the solvent leads to the formation of a self-assembled supercrystal powder with a long-range ordered face-to-face array. In this array, the Au octahedral cores are oriented with their vertices vertically aligned on the xy plane. These nanocube blocks transform superstructure symmetry from a hexagonal to a tetragonal lattice during colloidal crystallization, thereby creating three-dimensional cubic lattice templates on a substrate. When employing Ag etching on the substrate, the Ag outer shells are selectively etched out, leaving the Au octahedra cores anchored by Ag remnants. Notably, this orientation enables tip-induced hot-spots under laser irradiation. The ensemble of hot-spots facilitates ultrahigh-sensitivity surface-enhanced Raman scattering (SERS) detection of gas molecules.

Results

Fabrication of 3D tip-to-tip octahedral supercrystals

In this study, we present a stepwise method for fabricating three-dimensional tip-to-tip octahedral supercrystals. As depicted in the schematic illustration, the process involves the following steps: (1) Colloidal Ag shelling in a suspension, (2) templated crystallization on a substrate, and (3) Ag selective etching on the substrate (Fig. 1a). Au octahedral nanoparticles were synthesized as building blocks with a high yield (> 95%) of uniform shapes and tunable edge sizes (from 32 ± 4 to 75 ± 4 nm, Supplementary Figs. 1, 2). The shelling and assembly processes described are applicable to Au octahedra of all sizes (vide infra, Supplementary Table 1). First, Au octahedra were templated as Ag cube shells in colloidal synthesis. The six vertices of the Au octahedra were positioned at the center of the faces of Ag nanocubes by Ag shelling on the octahedral surface. Next, a concentrated solution of colloidal Au octahedra@Ag cubes was dropped onto a Si substrate and crystallized via self-assembly during evaporative steps, resulting in the formation of well-defined cubic superlattices. The size of the supercrystals was controllable from 0.9 ± 0.3 to 5.0 ± 1.3 μm (vide infra). The critical step to complete the artificial superstructures was selective etching of the Ag on the substrate. Ag was removed by an etchant, leading to the unique morphology of Au octahedra cores that were interconnected to neighbor octahedral particles maintaining their vertex-up orientation. We have realized artificially engineered superstructures in the form of a macroscopic supercrystal powder composed of three-dimensional micron-scale structures (a macroscopic powder form as shown in Fig. 1b, the magnified microsized habit as depicted in panel c, the zoomed-in image of the habit shown in panel d, and the top-view and side-view in panels e and f, respectively). Scanning electron microscopy (SEM) images of the top view (Fig. 1e) and side view (Fig. 1f) of the superlattice revealed a mosaic chessboard pattern, where all the surfaces of the nanoparticles are exposed to air. Each single Au octahedral unit was coordinated with six neighboring building blocks by tip-to-tip bonding along the Cartesian coordinate axes xyz.

Fig. 1: Schematic illustration of the stepwise process for the three-dimensional tip-to-tip octahedral supercrystal.
figure 1

a Au octahedral nanoparticle building blocks were crystallized with tip-to-tip assembly including colloidal Ag shelling (in suspension), templated crystallization (on substrate), and Ag-selective etching (on substrate). b Optical image and (cd) low-to-high magnified SEM images of three-dimensional tip-to-tip octahedral supercrystals. e Top and (f) side view of the tip-to-tip octahedral nanoparticle assembly.

Synthesis and self-assembly of Ag coated Au octahedral NPs

A batch of colloidal Au octahedral nanoparticles was dispersed in aqueous cetyltrimethylammonium chloride (CTAC) surfactant, and Ag precursors and ascorbic acid were slowly injected into a heating flask (at 80 °C). This process resulted in the development of highly homogeneous Ag shells on the core Au octahedral nanoparticles, creating Ag nanocube templates (Fig. 2a). Interestingly, the well-defined Ag cubic shells enclosing the Au octahedral core with sharp edges resulted in the isolation of the Au octahedral core in a standing-on-tip state (Fig. 2b). Here, we defined cubicity (Cn = 1-D/L, 0 ≤ n ≤ 1) as a function of relativity between the corner diameter (D) and the outer shell edge length (L) of cubes to quantify the outer shape of the Ag shell. The corner diameter (D) is defined as the diameter of the perfect circle on the two-dimensional plane that follows the corner curvature of Ag cubes (see Supplementary Fig. 3 for more details). The corner curvature was determined by transmission electron microscopy (TEM) measurements. The TEM images revealed that when Ag was grown to form Ag (100) facets on the surfaces of Au octahedra, the corners developed into sharper cubes (i.e., Cn converges toward unity), positioning the six sharp vertices of the Au octahedral core at the centers of the Ag cube faces (Figs. 2c and Supplementary Fig. 4). By contrast, if Cn deviates from unity, the corners of the Ag cubes become rounded. The Ag cubic shells can be sharpened to a uniform size and shape by precisely controlling the surfactant concentration and injection rate of Ag precursors (Fig. 2d). This process exhibits distinctive localized surface plasmon resonance (LSPR) spectral dynamics, transitioning from Ag layers to spherical growth or to cube formation stages (Supplementary Figs. 5, 6)28. Notably, the resulting shapes of Ag shells played a key role in determining the template organization of the three-dimensional superstructure. Typically, their arrangements transformed from a hexagonal spherical lattice to a disordered structure to a tetragonal cubic lattice, as depicted in the SEM images and corresponding fast Fourier transform (FFT) patterns of the assemblies of the Ag-shelled particles with cubicities of 0.31, 0.42, 0.68, and 0.80 (Fig. 2e–h and Supplementary Table 2). Consequently, highly ordered cubic lattices of three-dimensional Au octahedral nanoparticle@Ag sharp-cornered template superstructures were achieved by controlled evaporation (Figs. 2i, Supplementary Fig. 7). The Ag shell-templated three-dimensional superstructures exhibited close proximity of building blocks, with neighboring nanocubes adjoined through face-to-face interactions with narrow gaps of less than 1 nm (Supplementary Figs. 8, 9).

Fig. 2: Colloidal synthesis and self-assembly of Ag shelling on Au octahedral nanoparticles.
figure 2

a Scheme of experimental set-up for colloidal Ag shelling controlled by CTAC concentration, injection rate (ml min−1) of Ag precursors, and ascorbic acid. b Figure of cubicity (Cn) of edge length (L), corner diameter (D) for Au octahedra@Ag shell, and (c) TEM image of Ag shells on Au octahedra corresponding to cubicities of 0.31, 0.42, 0.68, and 0.80, respectively (scale bar = 20 nm). d Plots of Au octahedra@Ag outer shell morphology controlled by CTAC concentration and injection rate of Ag precursor (5 mM) and ascorbic acid (0.1 M). SEM images of close-packed Au octahedra@Ag shell assemblies with cubicities of (e) 0.31, (f) 0.42, g 0.68, and h 0.80, and corresponding FFT pattern (scale bar = 0.1 μm−1) and particle distributions of cubicity. i Low-magnified SEM images of Au octahedra@Ag nanocube supercrystals (inset shows the cubic habits). Source data are provided as a Source Data file.

Ag selective etching on the templated supercrystals

After the crystallization of Au octahedra@Ag nanocube shells as cubic templates on the substrate, tip-to-tip octahedral supercrystals were constructed by employing etching solvents. Specifically, hydrogen peroxide (H2O2) dissolved in ammonium hydroxide (NH4OH) was introduced to a batch containing a Si substrate immersed in distilled water (Fig. 3a). When Ag etchant was introduced to the supercrystals, Ag outer shells were preferentially etched away from the corner domain (Supplementary Fig. 10). During a mild Ag etching reaction, remnant Ag atoms were welded to Au octahedral vertexes, forming Ag bridges within individual Au octahedra building blocks. Notably, the three-dimensional colloidal superstructures maintained their morphological integrity even after oxidative etching (Fig. 3b and Supplementary Figs. 1113). The interconnection of Au octahedral cores with the small amounts of Ag remnants was confirmed in the high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) and energy-dispersive X-ray spectroscopy (EDS) superlattice images (Fig. 3c). The elemental composition of the surface of superlattices was analyzed by X-ray photoelectron spectroscopy (XPS) and EDS mapping spectrum, proving the existence of Au-exposed surface with thin Ag layers (Supplementary Figs. 14, 15). Although we believe that the residual Ag (4.27% based on EDS mapping analysis) is most likely located in the tip-to-tip bridging regions rather than on the Au surface, we cannot entirely exclude the possibility of a thin layer of Ag existing on the Au surface. Further, the lattice parameters of the supercrystals can be precisely manipulated from 45 ± 5 to 115 ± 7 nm by adopting various edge sizes of octahedral nanoparticles. SEM images and corresponding FFT patterns revealed the arrangements of the octahedral superlattices with the vertices pointing upward (Fig. 3d–g). Interestingly, the superlattices showed a checkered pattern in which the octahedral building blocks were represented by square orthogonal projections, diagonally adjacent to neighboring octahedral particles (Supplementary Fig. 16).

Fig. 3: Ag selective etching on three-dimensional Au octahedra@Ag cube supercrystals.
figure 3

a Schematic illustration of three-dimensional tip-to-tip octahedral supercrystal. b SEM, c STEM, and EDS image (Au/yellow, Ag/blue) of three-dimensional tip-to-tip octahedral supercrystals. dg SEM images of tip-to-tip octahedral supercrystals with different edge sizes (32, 45, 60, and 75 nm) of octahedral blocks, corresponding FFT pattern (scale bar = 0.1 μm−1) and figures of the edge size and distance of neighboring particles.

SERS analysis of gas molecules

In this study, it was essential to define the terminology used for our assemblies. The term “superpowder” refers to a powder form containing numerous supercrystals29. Supercrystals denote specific aggregates characterized by a long-range superlattice order of nanoparticle building blocks30,31. Superlattice refers to the specific pattern within these supercrystals32,33. To characterize their near-field focusing capability originating from tip-to-tip coupling of electric fields resulting from the lightning rod effect34, various sizes of supercrystal powders ranging from 0.9 ± 0.3 to 5.0 ± 1.3 μm were prepared, and the surface was cleaned with 0.1 mM NaBH4 to remove any adsorbates (Figs. 4a–l and Supplementary Fig. 17). In the zoomed-out images (500×), supercrystals exhibited homogeneous size distributions (Supplementary Fig. 18). The corresponding zoomed-in images and additional cross-sectional SEM images (Supplementary Fig. 12) revealed chessboard patterns, which remained consistent regardless of supercrystal sizes. The SERS performance of crystals of different sizes was evaluated to assess the penetration depth. Under the given experimental conditions, our preliminary characterization revealed that SERS signals were saturated with the supercrystals with crystal grain areas at 6.2 ± 2.4 µm², optimizing the number of hot-spots and minimizing background interference (Figs. 4m, n and Supplementary Fig. 19). Accordingly, the supercrystal dimensions were standardized to crystal grain area of 6.2 ± 2.4 µm² (crystal size of 3.2 ± 0.7 μm) for all experimental SERS substrates. In addition, we validated consistency and stability of SERS analysis in the tip-to-tip octahedral supercrystals as SERS substrate (Supplementary Figs. 20, 21). To predict light–matter interactions of plasmonic superstructures, the tip-to-tip octahedral assemblies were compared with analogous close-packed octahedral assemblies in the presence of thin Ag layers by surface charge distribution calculations as illustrated in Fig. 5a, c. Interestingly, the charge distribution was asymmetric in the close-packed octahedral assembly, originating from the coupling of various surface plasmon modes. By contrast, the surface charge was symmetrically distributed in the form of dipole-dominant modes in the tip-to-tip octahedral assembly, efficiently confining surface electrons at the sharp tips (i.e., coupling of the lightning rod effect). Additionally, we calculated the electric field distributions of both assemblies. The average hot-spot density of the close-packed assembly was 17.5, whereas the tip-to-tip octahedral assembly had a density of 12.6. However, the local electric field at the tip apex was strongly maximized (|E | / | E0 | = 261.4) because of the coupling of lightning rod effects induced by near-field focusing at sharp tips. This value was much higher than that of the close-packed octahedral nanogaps (|E | / | E0 | = 124.1) (Figs. 5b and d). To demonstrate the tip-enhanced near-field focusing capability of the three-dimensional tip-to-tip octahedral superpowders, bulk SERS measurement was conducted using two different types of the aforementioned three-dimensional supercrystals with analogous octahedral building blocks with the size (73 ± 4 nm) and composition of Au (95.6%) and Ag (4.4%) (Fig. 5e, f, Supplementary Fig. 22). These measurements involved exposure to 1000 ppm 2-chloroethyl phenyl sulfide (CEPS) gas molecules under 785 nm excitation laser, matched with spectral plasmonic band of both assemblies (Supplementary Figs. 23, 24). The typical close-packing arrangement of the octahedra showed a Raman signal with high intensity due to the higher hot-spot density in the inter-nanogaps as the power of the 785 nm laser was increased from 0.5 to 21 mW. However, the obtained SERS signals were accompanied by significant background interference, leading to signal degradation mainly originating from the formation of carbonaceous species at the higher laser power at 21 mW (Figs. 5g, Supplementary Fig. 25). We also would like to note that thin layers of Ag contributed to an increase in Raman intensity compared to supercrystals of pure Au octahedral building blocks (Supplementary Figs. 2627). However, the close-packed octahedral assembly with thin Ag layers also suffered from an irreversible decrease in SERS signals at high laser powers (Fig. 5g). By contrast, the SERS signals measured from tip-to-tip octahedral superstructures clearly exhibited an improved signal-to-noise ratio (SNR) as the laser power increased from 0.5 to 21 mW (Figs. 5h, Supplementary Fig. 28). Such enhancements in SNR were attributed to the fully exposed octahedral faces of the tip-to-tip octahedral superstructures, which exhibited a higher exposed surface area, significantly lower packing density (Supplementary Table 3), and efficient dissipation of plasmonic heat generated by laser excitation (Supplementary Fig. 29). These factors contributed to the structural stability under high-intensity laser irradiation (Supplementary Fig. 30). Additionally, these structural properties effectively suppressed the formation of carbonaceous species and their accompanying fluorescence backgrounds. This phenomenon was more apparent in the measurement of 2-chloroethyl ethyl sulfide (CEES, more vulnerable to heat burning), small gas molecules used in chemical warfare agent. The tip-to-tip octahedral superstructures exhibited clear Raman fingerprints with a high SNR when exposed to CEES gas molecules (Supplementary Fig. 31). Assuming that the observed SERS signals are mainly generated at the inter-nanogaps, the calculated enhancement factor is \(2.5\times {10}^{6}\). The ensemble of hot-spots and the wide void space within the supercrystals contribute to the enhancement of SERS signals to the maximum extent by suppressing background signals (Supplementary Table 4), particularly when high laser powers are adopted35. Notably, these experimental results are analogous to those obtained using tip-enhanced Raman spectroscopy (TERS), where a sharp metal tip is utilized as a scanning probe that generates a near-electric field confined to the tip by the aligned laser focal point, exhibiting extremely high spatial resolutions and single-molecule sensitivity36. Our tip-to-tip octahedral supercrystals are analogous to bulk SERS substrates comprising a myriad of TERS probes, although they are designed for molecular detection rather than spatial measurement. The resulting SERS calibration curve for the tip-to-tip octahedral superstructures showed a lower limit of detection (100 ppb) compared to the close-packed superstructures (5 ppm) for CEPS gas analyte concentration (Supplementary Figs. 32, 33; Supplementary Tables 5, 6), demonstrating the potential of SERS as a gas sensing platform.

Fig. 4: Optimizing three-dimensional tip-to-tip octahedral supercrystals as a SERS substrate.
figure 4

Low-to-high magnified SEM images (ad) 500×, (eh) 5000 ×, (il) 25000× of controlled supercrystal size (inset optical image, scale bar = 10 μm). m SERS measurement of 100 ppm 2-naphthalenethiol molecules according to supercrystal size and (n) diagram of SERS intensity at 1069 cm−1 obtained from different supercrystal size (785 nm laser, integration time = 1 s, laser power = 4 mW). The standard deviations were calculated from N = 10 independent samples. Source data are provided as a Source Data file.

Fig. 5: SERS analysis of gas molecules through close-packed and tip-to-tip octahedral supercrystals.
figure 5

Surface charge distribution and electromagnetic field distributions of (a, b) close-packed (7 × 3) and (c, d) tip-to-tip (9 × 3) assembly of octahedral building blocks with 3 nm Ag layers under a 785 nm excitation (scale bar = 100 nm). Microscopic images and zoomed-in SEM images of (e) close-packed and (f) tip-to-tip octahedral supercrystals with existence of thin Ag layers. Corresponding SERS measurements of 1000 ppm CEPS gas molecules plotted for different laser powers (0.01, 0.5, 4, 11, and 21 mW) for (g) close-packed and (h) tip-to-tip octahedral supercrystals. Source data are provided as a Source Data file.

Discussion

In this study, we have fabricated the three-dimensional tip-to-tip octahedral supercrystals, where octahedral building blocks maintained their morphology in a standing-on-tip state. The highly uniform Ag nanocube-coated Au octahedral nanoparticles self-crystalized into microsized cubic superstructure templates under the evaporation steps, fixing the Au octahedral cores in a vertex-up orientation. Importantly, the Au octahedral cores maintained their configurations with thin Ag layers, leading to the tip-to-tip octahedral supercrystals as a superpowder. These plasmonic superstructures, with all-facet-exposed octahedral building blocks, exhibited a highly ordered nanoporous network and well-organized hot-spots, providing high hot-spot accessibility to small molecules. Notably, they generated tip-enhanced hot-spots under a higher laser power, resulting in excellent SERS sensitivity for gas analytes, aided by the suppression of carbonaceous species formation due to efficient heat dissipation. Our results suggest a strategy to design an alternative type of nanoparticle assembly, harnessing their material structure properties for broad applications.

Methods

Materials

Hydrogen tetrachloroaurate(iii) hydrate (HAuCl43H2O, 99%) was purchased from Kojima. Silver nitrate (AgNO3, 99.9%) and sodium borohydride (NaBH4, 98%) were purchased from Junsei. Acetone (CH3COCH3, 99.5%) and L-ascorbic acid (C6H8O6, 99.5%) were purchased from Sigma-Aldrich. Ethyl alcohol (C2H5OH, 99.9%), ammonium hydroxide (NH4OH, 28.0%–30.0%), and hydrogen peroxide (H2O2, 34.5%) were purchased from Samchun. Hexadecyltrimethylammonium bromide (CTAB, C19H42BrN, 99%) was purchased from Acros Organics. Hexadecyltrimethylammonium chloride (CTAC, C19H42ClN, 95.0%) was purchased from Tokyo Chemical Industry. 2-naphthalenethiol (2-NTT, 95.0%), 2-chloroethyl phenyl sulfide (CEPS, 98%), and 2-chloroethyl ethyl sulfide (CEES, 97%) were purchased from Sigma-Aldrich. All chemicals were dissolved in distilled water (≥ 18.2 MΩ cm) prepared by a Milli-Q water purification system (Millipore).

Instrumentation

Field-emission scanning electron microscopy (FE-SEM) images were obtained using JSM-7100F and JSM-7800F instruments (JOEL). Talos F200X, JEM-2100F, and JEM-ARM 200 F instruments (JEOL) were used for the transmission electron microscopy (TEM) analysis. UV–visible–near-infrared absorption spectra were obtained using a spectrophotometer (Shimadzu UV-3600). Raman spectra of the samples were recorded on a Raman microscope (XperRAM-S700).

Preparation of Au octahedral nanoparticles

Au octahedral nanoparticles were synthesized by a third growth process37. A total of 7 mL of 75 mM CTAB solution, 87.5 μL of 20 mM aqueous HAuCl4·3H2O solution, and 0.6 mL of 10 mM aqueous NaBH4 solution (ice cold for 15 min) were combined under vigorous stirring. The color turned dark brown at 30 °C in 3 h of seed formation. The seed solution was diluted 100 times in distilled water.

The first growth of Au octahedral nanoparticles with a size of 32 nm was conducted directly after seed preparation. To synthesize the first growth of the Au octahedra, 480 mL of 16 mM CTAB, 0.2 mL of 20 mM aqueous HAuCl4·3H2O solution, 6 mL of 0.1 mM ascorbic acid (AA), and 6 mL of seed solution were added to a 500 mL round-bottom flask. After gentle shaking, the solution was incubated at 30 °C for 12 h. The color turned to bright pink. The residual CTAB surfactant was removed by centrifugation twice at 9358 × g for 15 min.

In the second growth of Au octahedral nanoparticles with a size of 45 nm, 210.8 mL of 100 mM CTAB, 700 μL of 20 mM aqueous HAuCl4·3H2O solution, 3.76 mL of 0.1 M AA, and 140.5 mL of the first seed solution were added to a 500 mL round-bottom flask. The flask was incubated at 30 °C for 4 h. The color turned to reddish purple. The residual CTAB surfactant was removed by centrifugation twice at 9358 × g for 15 min.

In the second growth of Au octahedral nanoparticles with a size of 60 nm, 210.8 mL of 100 mM CTAB, 1.4 mL of 20 mM aqueous HAuCl4·3H2O solution, 6.53 mL of 0.1 M AA, and 140.5 mL of the first seed solution were added to a 500 mL round-bottom flask. The flask was incubated at 30 °C for 4 h. The color turned to purple. The residual CTAB surfactant was removed by centrifugation twice at 6300 × g for 15 min.

In the third growth of Au octahedral nanoparticles with a size of 75 nm, 320.7 mL of 100 mM CTAB, 2.1 mL of 20 mM aqueous HAuCl4·3H2O solution, 9.9 mL of 0.1 M AA, and 134.7 mL of the second seed solution (45 nm size condition) were added to a 500 mL round-bottom flask. The flask was incubated at 30 °C for 4 h. The color turned to dark purple. The residual CTAB surfactant was removed by centrifugation twice at 6300 × g for 15 min.

Preparation of colloidal Ag shelling on Au octahedral nanoparticles in suspension

Octahedra@Ag nanocubes were synthesized by using a epitaxial growth of heterogeneous metal naoparticles38. A batch samples of colloidal Au octahedral nanoparticles were prepared as a aqueous solution volume of 60 mL (an optical density of 2.0) and then 30 mL of 0.2 M CTAC surfactant were added to 150 mL of deionized (DI) water in a 500 mL round-bottom flask (700 rpm with a stirring bar), and the mixture was heated to 80 °C. Then, 5 mM AgNO3 and 100 mM AA solutions were added to the solution using two-channel syringe pumps. Importantly, Ag shelling growth on Au octahedral nanoparticles was precisely controlled by the CTAC surfactant and the injection rate of AgNO3 and AA. The roundness of outer silver nanocubes was precisely controlled and optimized by the CTAC surfactant and the injection rate of AA and Ag precursors (Supplementary Fig. 2. and Supplementary Table 1). After the addition of all the reagents, the solution was left for 1 h without any treatment, being stirred at 700 rpm. Finally, the solution was washed twice (6300 × g for 20 min) with DI water. After washing, colloidal Au octahedra@Ag nanocubes were moved to an Eppendorf tube and washed at 9358 × g for 10 min. The final Au octahedra@Ag nanocube concentrate was dispersed in 100 μL aqueous solution.

Preparation of Au octahedra@Ag nanocube supercrystals on substrate

The final Au octahedra@Ag nanocubes concentrates was re-dispersed to 1 mL of the specific concentration of CTAC solution to control the size of the supercrystal. The size of the supercrystal powders was controlled from re-dispersed CTAC concentration to 100, 50, 20, and 5 mM, and the resultant supercrystals sizes could be obtained as 0.9 ± 0.3, 2.1 ± 0.7, 3.2 ± 0.7, and 5.0 ± 1.3 μm, respectively. The samples were centrifuged at 9358 × g for 5 min, leaving 50 μL of Au octahedra@Ag nanocube dispersions. Before preparing the supercrystals, a silicon or glass substrate was washed with acetone and ethanol. Then, 2.5 μL of Au octahedra@Ag nanocube dispersion was dropped on the substrate as one batch of a substrate. Around 20 (20 × 2.5 = 50 μL) droplets could be obtained from initial colloidal dispersion of Au octahedra@Ag nanocubes. The substrates were placed in a pressure- and humidity-controlled container. To increase the humidity above 80%, a small Petri dish containing 500 μL of DI water was placed in the container. Pure N2 gas was introduced to the container, raising the internal pressure to 4 bar. The container was then stored a 30 °C oven for controlled evaporation. After 12 h of evaporation, the substrates were immersed in ethanol solution for 20 min to eliminate the surfactant.

Preparation of three-dimensional tip-to-tip octahedral supercrystals on substrate

To prepare the three-dimensional tip-to-tip Au octahedral supercrystals, 5 mL of NH4OH was dropped on the Petri dish. Then, Au octahedra@Ag nanocube supercrystals on the substrate were completely immersed in the NH4OH solution. 50 μL of 10% H2O2 was carefully and slowly injected into the Petri dish using a micropipette; the outer Ag shell was selectively etched away after 30 min, and Ag residues were welded to Au octahedral vertexes, creating the tip-to-tip Au octahedral supercrystals. If excess amounts of H2O2 were injected into the Petri dish, the building blocks lost their orientation, resulting in randomly assemblies. After mild Ag selective etching, the remaining etchants were evaporated in a vacuum chamber. Then, the substrates were redispersed in ethanol for 10 min. Finally, the fabricated supercrystals were collected in powder form and stored at room temperature.

SERS measurement of gas molecules using three-dimensional tip-to-tip octahedral supercrystals

Before the SERS measurement of gas molecules, the substrates were dispersed in 0.1 mM NaBH4 solution for 30 min to clean the exposed surface. Then, the samples were washed with ethanol for 10 min and dried in the vacuum chamber. For gas molecule detection, the substrates containing tip-to-tip Au octahedral supercrystals were placed on the cap of a vial, and 10 μL of a specific concentration of analyte (2-NTT, CEPS, or CEES) solution was loaded in the vial and fully evaporated in a closed vial at 70 °C for 1 h for the quantitative SERS measurement of the gas molecule. We analyzed SERS at a laser excitation wavelength at 785 nm with an integration time of 1 s with a Raman microscope using a 100× objective lens.

Electromagnetic simulations

For the electromagnetic simulation, we calculated electric field enhancements, enhanced volume, theoretical extinction cross sections of nanoparticles, and surface-charge distribution using a finite element method (FEM) method (COMSOL Multiphysics 6.0), a radio-frequency module, and a frequency domain study. The developed models were based on TEM and SEM data and designed with AutoCAD software. We selected an assembly of Au octahedra with 3 nm Ag layers (7 × 3 particles for the close-packing model and 9 × 3 particles for the tip-to-tip model), the gap between the nanoparticles was 3 and 0 nm for the close-packing and tip-to-tip models, respectively. The maximum tetrahedral mesh size of the simulation region in single-particle simulation was 1 nm, and that in the assembly was 2 nm. The refractive index of Au was defined using the empirical dielectric function evaluated by Johnson and Christy39.

Calculation of the packing density of three-dimensional tip-to-tip octahedral supercrystal

We calculated volume fractions and surface area ratio of three-dimensional tip-to-tip Au octahedral supercrystal. Assuming the edge length of a single octahedron is L.

$${{\rm{Volume}}}\; {{\rm{of}}}\; {{\rm{the}}}\; {{\rm{unit}}}\; {{\rm{cell}}}=\sqrt{2}L\times \sqrt{2}L\times \sqrt{2}L=2\sqrt{2}{L}^{3}$$
(1)
$${{\rm{Volume}}}\; {{\rm{of}}}\; {{\rm{the}}}\; {{\rm{one}}}\; {{\rm{octahedron}}}=\frac{\sqrt{2}}{3}{L}^{3}$$
(2)

Packing density of three-dimensional standing-on-tip Au octahedral supercrystal:

$$\frac{\frac{\sqrt{2}}{3}{L}^{3}}{2\sqrt{2}{L}^{3}}=\frac{1}{6}\approx 0.167$$
(3)

Calculation of the enhancement factor

$${{\mbox{EF}}}=\frac{{I}_{{\mbox{SERS}}}}{{I}_{{\mbox{bulk}}}}\times \frac{{N}_{{\mbox{bulk}}}}{{N}_{{\mbox{SERS}}}}$$
(4)

Where ISERS and Ibulk are the intensities of a peak (1081 cm−1) in the SERS spectrum (\({I}_{{\mbox{SERS}}}=940\) in 785 nm laser, laser power = 4 mW) and conventional Raman spectrum of pure benzenethiol (Nbulk=138 in 785 nm Laser, laser power = 4 mW), and NSERS and Nbulk are the corresponding number of the analytes (benzenethiols) in the SERS measurement and the conventional normal Raman spectrum for the liquid sample solution.

\({N}_{{\mbox{SERS}}}\) is calculated by a packing density (D) of \(6.8\times {10}^{14}\) molecules per cm2 reported for benzenethiol40, and exposed surface area (A) of ensemble hotspots of the 3D standing-on-tip Au octahedral superpowder crystals in the laser spot. We assumed approximately 8000 upright octahedral building blocks (60 nm edge length) were organized as supercrystals in the laser spot (r = 1 μm), and 1% of exposed surface area was dominantly contributed to the enhancement41.

$$A=8000\times \frac{3\sqrt{3}}{2}\times 60\times 60\times 0.01{\mbox{=}}7.48\times {10}^{-9}\,{{\mbox{cm}}}^{2}$$
(5)

As a result, ≈ 5.09 × 105 molecules were probed on the SERS substrates.

$${N}_{{\mbox{SERS}}}=A\times D{\mbox{=}}5.09\times {10}^{5}$$
(6)

Nbulk is the number of molecules within the excitation volume of the laser spot for the liquid samples.

Nbulk was calculated as follows:

$${N}_{{\mbox{bulk}}}=\frac{\pi \times {r}^{2}\times h\times d}{{{\mbox{M}}}_{{\mbox{w}}}}\times {{\mbox{N}}}_{{\mbox{A}}}{\mbox{=}}1.85\times {10}^{11}$$
(7)

The Nbulk was estimated by the laser spot (r = 1 μm), the laser penetration depth (h = 10 μm) in the bulk sample, molecular weight (Mw) of benzenethiol (110 g mol−1), density (d = 1.077 g mL−1) of the benzenethiol and Avogadro’s number (\({{\mbox{N}}}_{{\mbox{A}}}\)).

Finally, the enhancement factor will be obtained by the calculated numbers.

$${{\mbox{EF}}}=\frac{{I}_{{\mbox{SERS}}}}{{I}_{{\mbox{bulk}}}}\times \frac{{N}_{{\mbox{bulk}}}}{{N}_{{\mbox{SERS}}}}=\frac{940}{138}\times \frac{1.85\times {10}^{11}}{5.09\times {10}^{5}}{\mbox{=}}2.5\times {10}^{6}$$
(8)