1. Introduction

Surface-enhanced Raman scattering (SERS) is an effective approach for high-sensitivity detection at the single-molecular level [1–4]. Seeking for better SERS-active substrates is essential for applications. In the past decades, amounts of SERS substrates have been developed for assessments and applications. These substrates could be enabled either by chemically-synthesized plasmonic nanoparticles [5–9] or lithographically defined nanoparticle arrays [10–13]. While wet-chemical approach has the advantages of low-cost high enhancement factor and extremely low detection limit down to 1fM [14–17], lithographically defined periodic plasmonic structures are promising due to their high repeatability and stability for both fundamental investigations and practical applications.

Reasonable design of the particles with appropriate size [18,19], geometry [20,21] and period [13,22] is of great importance to maximize the enhancement factor in lithographically-defined plasmonic nanostructures. While many efforts have been made to obtain ultrasmall nanogaps [23–27] and ultrasharp corners [28,29] via advanced patterning processes to achieve the highest possible enhancement factor, a systematic work to investigate the particle density effect on SERS intensity is still missing.

In this work, we specifically aim to investigate and understand how density of periodic plasmonic structures affects the SERS enhancement factor. Our results show that three different density regimes can be identified when increasing the particle density in terms of SERS enhancement factor, including a linear increase regime for very sparse structures, a gradual decrease regime for relatively dense structures, and a strong increase regime for extremely densely-packed structures with tiny gaps. The experimental results could be intuitively understood based on the extinction cross-section concept and finite-difference time-domain (FDTD) simulations. With the systematic investigation of the particle density effect on the SERS intensity, our work provides robust evidences to guide the design of SERS substrate with the maximal enhancement factor.

2. Methodology

2.1 Sample Fabrication

HSQ resist (XR-1541-006, Dow Corning) was first spin-coated onto silicon substrates to a thickness of ~90 nm. No baking was done before exposure. HSQ nanopillars with different densities were defined by a Raith 150^TWO electron-beam-lithography system with an accelerating voltage of 30 kV. No proximity-effect correction was performed for the exposure. The sample was developed by salty developer (1% NaOH + 4% NaCl in deionized water) at 24 °C for 1 min [30], rinsed by DI water for 2 min, and finally blow-dried under a steady stream of N₂.

Metal deposition was performed using an electron-beam evaporator (Labline, Kurt Lesker). A silver layer (15 nm) and a gold capping layer (5 nm) were sequentially deposited onto the samples with a chromium adhesion layer (1 nm). All metals were deposited at a rate of 1 Å s-1. The fabricated sample was imaged using a scanning electron microscope (Sigma-HD, Zeiss) with an accelerating voltage of 10 kV and a working distance of 6 mm.

2.2 SERS Measurements

To prepare the substrate for SERS, the sample was first immersed in a solution of 4-aminobenzenethiol (4-ABT) at a concentration of 10⁻⁵ M for 12 h. The sample was then rinsed several times with ethanol and blow dried using a N₂ stream to enable a monolayer of 4-ABT. The SERS measurements were carried out on a Raman microscope (WITec Alpha-300R) equipped with a 633 nm He-Ne laser. Raman signals were collected using a 100 × objective (N.A. = 0.85) under a laser power intensity of 0.5 mW and with an integration time of 0.5 s.

2.3 Numerical Simulations

Full 3D finite-difference time-domain (FDTD) simulations were performed on a workstation using the commercial software (Lumerical Solutions Inc.). Periodic boundary conditions were used for the two in-plane dimensions to simulate an infinite array of periodic plasmonic structures. Perfectly matched layer (PML) boundary conditions were used in the z-direction. The mesh size used in the simulation region was 1 nm. The complex dielectric constants for Au, Ag and Si were selected as Palik from the database, respectively. The structures were illuminated with a plane wave directed along the z-axis. The time-averaged total electric field intensity was extracted to generate the field distributions.

3. Results and Discussion

Elevated plasmonic nanostructures were used to investigate the density effect. Figure 1 shows the schematics of the fabrication flow for the nanostructures. In a typical process, HSQ nanoposts (Figs. 1(a) and 1(c)) with different pitches were first defined by electron beam lithography. The metallic plasmonic structures composed of 15 nm silver and 5 nm gold were directly evaporated onto the HSQ pillars with 1 nm Cr adhesion layer (Figs. 1(b) and 1(d)) for SERS measurements. Because the thickness of HSQ pillars was ~90 nm, the metal film was separated into top disks and back reflector. The silver film was used to enable the high SERS activity and the thin gold film was used to protect silver from sulfurization or oxidization.

 

Fig. 1 (a) Schematics of EBL-defined HSQ nanoposts with diameter of d and pitch of p. (b) Schematics of the plasmonic naonstructures after depositing metal on HSQ nanoposts. (c) SEM image of HSQ nanoposts with diameter of 120 nm and pitch of 240 nm. (d) SEM image of the plasmonic nanostructure array after metal deposition on HSQ nanoposts in (c).

Download Full Size | PDF

Such a configuration could enable clean plasmonic structures for stable SERS measurements [31,32]. Meanwhile, the elevated plasmonic nanostructures have been demonstrated with enhanced SERS response compared to their planar counterparts [33]. To simplify the investigation of the density effect, the diameter of the HSQ nanoposts was designed to be 120 nm. The periods of the structures were designed to fifteen different numbers from 125 nm to 500 nm. Due to the proximity effect in the electron-beam lithography, the actual size of the fabricated HSQ nanoposts would increase when increasing the density of the structures. Meanwhile, the resultant plasmonic nanostructures after metal deposition were also slightly larger than the HSQ nanoposts due to the clogging effect in the deposition process [34].

Figures 2(a)-2(d) show four representative SEM images of the final periodic plasmonic nanostructures with different pitches of 500 nm (Fig. 2(a)), 240 nm (Fig. 2(b)), 175 nm (Fig. 2(c)) and 135 nm (Fig. 2(d)). From the SEM images, it can be seen that the plasmonic nanostructures were well defined from the HSQ nanoposts. For the structures with a designed pitch of 135 nm, the final plasmonic nanostructures were closely packed with sub-10-nm gaps. Further decreasing the designed pitch to 125 nm led to the connection of plasmonic nanostructures.

 

Fig. 2 (a-d) Representative plasmonic nanostructure arrays with different densities. The diameter of the structures was 125 nm, and the pitch of the arrays was 500 nm (a), 240 nm (b), 175 nm (c), and 135 nm (d), respectively. (e) Single particle SERS measurement from sample (a). (f) The SERS mapping result of a corner of the 500-nm-pitch plasmonic structures.

Download Full Size | PDF

Figure 2(e) shows the single-particle Raman scattering spectrum of the defined plasmonic nanostructures shown in Fig. 2(a) (i.e. 500 nm pitch), in which strong Raman scattering signal of 4-ABT molecules was obtained. For the same metal surface without plasmonic nanostructures, no Raman signal was detected by the same measurement parameters. Further mapping of the scattering signal at 1435 cm⁻¹ confirms that the strong Raman scattering signal was originated from the top plasmonic nanostructures, instead of the bottom metal surface, as shown in Fig. 2(f).

To analyze the particle density effect on the SERS intensity, the Raman scattering spectra of all fabricated plasmonic nanostructures with different densities were measured, as shown in Fig. 3(a). The corresponding intensity of the two main Raman scattering peaks as a function of the filling factor is given in Fig. 3(b), in which filling factor is defined to be the area ratio between a plasmonic disk and a unit cell in the arrays. With a simple calculation, the filling factor equals to $\frac{\text{π}{d}^2}{4p^2}$, where d is the designed diameter of the plasmonic disks, i.e. 120 nm, and p is the pitch of the plasmonic nanostructures arrays. Larger filling factor means denser structures.

 

Fig. 3 (a) Raman scattering spectra of plasmonic nanostructures with varied density. (b) The SERS intensity as a function of the particle density.

Download Full Size | PDF

From Fig. 3(b), four regions (I, II, III and IV) could be distinguished to describe the change of SERS signals. For sparse structures, the SERS intensity firstly increases when increasing the density of plasmonic nanostructures and then achieves a maximum at a certain particle density, as indicated by region I. However, when further increasing the particle density, the SERS intensity would decrease, as indicated by region II. In the region III, the SERS intensity drastically increases when the particles are closely packed and achieve a maximum at a corresponding pitch of 135 nm. In the region IV, the SERS intensity suddenly decreases when the particles are connected each other. Note that the typical enhancement factor of our substrate in region I-III was estimated to be on the order of 10⁷ according to the commonly used calculation method [35], which was quite low compared to other SERS substrates fabricated by wet-chemically synthesized plasmonic structures [14–17]. The main reason of such a low enhancement factor was because our work specifically aimed to investigate and understand the particle density effect in lithographic periodic structures, and thus we did not put attention to achieve the highest SERS enhancement factor via optimizing gap size and geometries (e.g. bowtie structures).

The above measurement results could be intuitively explained below and the relevant schematics are shown in Fig. 4. It is well known that plasmonic nanostructures have the ability to focus light into nanoscale to achieve large near-filed enhancement. The extinction cross-section of plasmonic nanostructures could be much larger than their physical size, as indicated by Fig. 4(a). We define the effective extinction cross-section of a plasmonic nanostructure to be σ = w² for the following discussion. For sparse structures with a pitch of p larger than w, each plasmonic nanostructure can funnel the most possible incident photons same as the isolated one, enabling similar near-field enhancement, as implied by Fig. 4(b). In this case, when increasing the density of plasmonic nanostructures, the SERS intensity would linearly increase as a function of the number of structures (or filling factor), corresponding to the experimental result of region I in Fig. 3(b). The maximal SERS intensity is supposed to be achieved by the particle array with a pitch almost equaling to w.

 

Fig. 4 Schematics showing how the density of plasmonic nanostructures affects the light-matter interactions: (a) isolated or extremely sparse structures; (b) relatively sparse structures; (c) relatively dense structures; (d) extremely dense structures with tiny nanogaps.

Download Full Size | PDF

When further increasing the particle density with a pitch of p smaller than w, as shown in Fig. 4(c), in a certain extinction cross-section area w², more than one plasmonic disk would participate in sharing the incident photons, resulting in decreased near-field enhancement for each plasmonic disk. It is known that SERS intensity is proportional to the square of laser intensity. With an assumed perfect extinction by plasmonic disks, the total SERS intensity in a plasmonic disk array would be proportional to $N*{(\frac{I}{N})}^2$, where N is the number of disks in the extinction cross-sectional area, I is the intensity of excitation light in the extinction cross-sectional area, while $\frac{I}{N}$ represents the average light intensity for each disk. From this formula, we can see that when the particle density increases, the total SERS intensity would decrease to a certain degree, corresponding to the region II in Fig. 3(b).

However, when the density of plasmonic disks further increases to be very closely packed, tiny nanogaps form between the adjacent disks, enabling significantly enhanced near field for SERS, as demonstrated by Fig. 4(d), which explains the rapid increase of the SERS intensity in the region III in Fig. 3(b). Once the nanogaps disappear in the connected nanostructures, the field enhancement would suddenly decrease, leading to minimal SERS intensity, as demonstrated by the region IV in Fig. 3(b).

The above intuitive assumptions were also verified by numerical simulations, as shown in Fig. 5. From the near-field distribution of plasmonic nanoparticles with different pitches, we can see that the electric field intensity first decreases when increasing the particle density (e.g. in Figs. 5(c) and 5(d)) and then significantly increases when further increasing the particle density to form tiny nanogaps (Fig. 5(e)). Figure 5(f) shows the cross-sectional near-field distribution of plasmonic system, indicating that the field enhancement was originated from the top plasmonic nanodisks while the bottom metallic back reflector did not obviously contribute the field enhancement for SERS.

 

Fig. 5 Simulated near-field distribution of 120-nm-diameter plasmonic nanoparticles with different pitches: (a) 360 nm; (b) 240 nm; (c) 180 nm; (d) 150 nm; (e) 135 nm. (f) The cross-section view of the near-field distribution of the plasmonic system in (e), showing the field was mainly confined at the top region. The excitation wavelength in the simulation was 633 nm.

Download Full Size | PDF

4. Conclusions

We systematically investigated how the density of plasmonic nanostructures affects the intensity of SERS signal. By directly depositing of metallic layer on electron-beam-lithography defined dielectric nanoposts, plasmonic structures arrays with different densities were reliably fabricated for SERS measurements. Two main experimental phenomena were obtained: (1) the SERS intensity did not increase monotonically when increasing the density of plasmonic structures, and (2) these ultra-dense plasmonic structures resulted in the maximal SERS intensity. These results could be well explained based on finite-difference time domain (FDTD) simulations and provide robust experimental evidences to guide the design of the best possible SERS substrate.