Jiwon
Lee‡
a,
Bo
Hua‡
b,
Seungyoung
Park
a,
Minjeong
Ha
a,
Youngsu
Lee
a,
Zhiyong
Fan
*b and
Hyunhyub
Ko
*a
aSchool of Energy and Chemical Engineering, Ulsan National Institute Science and Technology (UNIST), Ulsan, Republic of Korea. E-mail: hyunhko@unist.ac.kr
bDepartment of Electronic & Computer Engineering, Hong Kong University of Science & Technology (HKUST), Hong Kong SAR, China. E-mail: eezfan@ust.hk
First published on 28th October 2013
Plasmonic systems based on metal nanoparticles on a metal film have generated great interest for surface-enhanced Raman spectroscopy (SERS) chemical sensors. In this study, we describe the fabrication of ultrasensitive SERS substrates based on high-density gold nanostar assemblies on silver films with tailored surface plasmons, where multiple field enhancements from particle–film and interparticle plasmon couplings and lightening rod effects of sharp tips of nanostars contribute to the enormous Raman enhancements. We show that the interplay between interparticle and particle–film plasmon couplings of high-density gold nanostars (GNSs) on metal and dielectric films as a function of interparticle separation can be tailored to provide maximum SERS effects. We observe that the SERS enhancement factor (EF) of GNSs on a metal film as a function of interparticle separation follows a broken power law function, where the EF increases with the interparticle separation for the strong interparticle coupling range below an interparticle separation of ∼0.8 times the GNS size, but decreases for the weak interparticle coupling range (for an interparticle separation of >0.8 times the GNS size). Finally, we demonstrate the use of tailored plasmonic substrates as ultrasensitive SERS chemical sensors with an attomole level of detection capability of 2,4-dinitrotoluene, a model compound of nitroaromatic explosives.
Recently, plasmonic systems consisting of metal nanoparticles separated from metal films by nanometer scale gaps have attracted great attention as SERS substrates, because the precise gap regions between metal nanoparticles and films can provide reproducible hot spots with large SERS enhancements.31–37 Similar to the interparticle couplings between LSPs of metal nanoparticles, the LSPs of metal nanoparticles can couple with their own image charges in the metal films or the propagating SPPs at the surface of metal films, rendering the formation of hot spots between the nanoparticles and films.38,39 Theoretical and experimental studies have shown that the plasmonic properties of particle-on-film systems are strongly dependent on the gap distances between nanoparticles and films, the nanoparticle shapes and sizes, and the dielectric properties of both substrates and their environments.40–43 Plasmonic systems, consisting of a single metal nanoparticle or a nanoparticle pair on a metal film, have been investigated extensively, resulting in a good understanding of the plasmonic coupling of the LSP of individual nanoparticles with its image charges or SPP modes on the metal film.41–46 However, few studies have been performed on plasmonic systems consisting of high-density metal nanoparticle assemblies on metal films,47–49 where the strong interparticle couplings in high-density nanoparticle assemblies are expected to affect the nanoparticle–film plasmon couplings.
For the design of highly-sensitive SERS substrates based on high-density metal nanoparticle assemblies on metal films, it is critical to understand the effects of nanoparticle surface densities on the interparticle and the particle–film plasmon couplings and the resulting ensemble-averaged E-field enhancements. Here we report on our investigation of the plasmonic properties of high-density gold nanostar (GNS) assemblies arranged on various metals (silver and gold) and dielectric (silicon and glass) substrates. In particular, by analyzing the finite-difference time-domain (FDTD) calculation of E-fields and the experimental SERS intensities, we observed that the E-fields from particle–film plasmon couplings were strongly affected by interparticle couplings, which depended on the GNS surface densities (or interparticle gap separations). Based on the understanding of the effects of GNS surface density on the interplay between interparticle and particle–film couplings, we fabricated optimally designed SERS substrates with tailored surface plasmons that can detect target molecules on the scale of an attomole.
To investigate the effects of different substrates on the SERS properties of GNS assemblies, we used 200 nm thick Au and Ag films on Si substrates, as well as Si and glass substrates. Fig. 2(a) shows the Raman spectra of 0.1% benzenethiol molecules adsorbed on SERS substrates consisting of GNS assemblies on Au, Ag, Si, and glass surfaces with a PDDA layer in between (SEM images in Fig. S3†). The intensity of the Raman spectra was strongest for GNS assemblies on Ag films, followed by Au, Si, and glass substrates. In particular, we noted that the SERS EF increased from 2.7 × 105 for GNS assemblies on the glass surface to 4.4 × 107 for GNS assemblies on the Ag surface, which is a 160-fold increase (Fig. 2(b)). This significant increase in the SERS EF indicates the critical role of plasmon couplings between GNSs and metal films in addition to the interparticle plasmon couplings in high-density GNS assemblies on metal films. The LSPs of GNS assemblies hybridize with the propagating SPPs on the noble metal (Au, Ag) films and also interact with their image charges in the metal films,38,39 resulting in a strong enhancement of the E-fields in the particle–film gap region.32,35,37,43 In contrast, for GNS assemblies on dielectric substrates (silicon and glass), there are only image charge contributions, with no evidence of the effects of hybridization between the LSPs and SPPs; this is attributed to the absence of SPP modes on the surface of the dielectric substrates,40,54 resulting in a relatively weak SERS enhancement (Fig. 2(a) and (b)). The difference in SERS EFs on metal and dielectric substrates agreed well qualitatively with the FDTD simulations of the E-field intensity for GNSs assembled on different substrates (Fig. S4†).
We also noted that the SERS EF for the Ag surface was about ∼9× stronger than that for the Au surface (Fig. 2(b)), which can be explained by the difference in optical loss between the Au and Ag surfaces. The SPP modes excited in the metal film have been known to enhance interactions between the LSPs of metal nanoparticles on a metal film.43 Because Ag has a lower optical loss and longer SPP propagation length than Au,55 Ag films can support stronger interparticle couplings in GNS assemblies and thus provide larger SERS EFs. The difference in the SERS EF between the non-conducting Si and glass substrates is attributed to the difference in the permittivity of the substrates. The substrates with a higher permittivity (ε) provide stronger image charges and thus larger interactions with the LSPs of metal nanoparticles.40,56 Therefore, Si substrates (ε ∼ 11.76 at the frequency of 0.5 MHz)57 exhibited a larger SERS EF than that of glass substrates (ε ∼ 4.6 at the frequency of 94 GHz).58
To understand the behavior of interparticle and particle–film plasmon couplings with different interparticle separations for GNSs on metal and dielectric substrates, FDTD simulations were carried out to analyze the E-field enhancements for the interparticle and particle–film gap regions. Fig. 3(a) and (b) show the FDTD simulation results for the E-field distribution of GNS assemblies on Ag and glass substrates with an interparticle tip-to-tip separation of 10 nm. Detailed inspection of the local E-fields in the gap regions clearly showed distinct field intensities at the interparticle and particle–film gaps (Fig. 3(c) and (d)). Fig. 3(c) indicates that the local fields at the gap between the GNS tip and the Ag film were significantly stronger than those between the GNS tips. On the other hand, for GNSs on a glass substrate, the local fields between the tips were similar to those between the GNS tip and the glass surface (Fig. 3(d)). Note that hot spots with strongly-enhanced E-fields were located directly between the tip and the film for the GNS–Ag gap, which was different from the hot spots located near the end of the tip for the GNS–glass gap. Additionally, for GNSs on a Ag film, the E-field from interparticle coupling was ∼30% stronger than that for GNSs on a glass substrate. All of these observations imply that the plasmon coupling behaviors are different for GNS–Ag and GNS–glass systems.
To shed light on plasmon coupling behaviors that depend on interparticle separations, the local (GNS–Ag, GNS–GNS) E-field enhancements (|E/E0|) as a function of interparticle separation were analyzed for GNS assemblies on Ag and glass substrates (Fig. 3(e) and (f)). For GNS assemblies on Ag films, the field enhancement from GNS–GNS coupling decayed exponentially as the interparticle separation (D) increased, which is typically observed for near-field interactions in nanoparticle assemblies.59,60 The field enhancement was well fit with the exponential decay curve y = a × exp(−x/l) + y0, resulting in an amplitude of a = 33.1 and a decay length of l = 7.5. Here, the amplitude (a) tells us the magnitude of the E-field dependence on the interparticle gap distance and the decay length (l) tells us the exponential damping length of the E-field intensity. However, the field enhancement from GNS–Ag film coupling exhibited an inverse exponential decay curve as a function of interparticle separation (a = −310.6, l = 51.9). This inverse exponential decay behavior can be explained by the competition between the GNS–GNS and GNS–Ag film couplings. For small interparticle separations, the strong GNS–GNS coupling leads to the screening of particle–film plasmon couplings, in particular the hybridization between the LSP and SPP modes. The opposite behavior happens for the large interparticle separations. In contrast, the field enhancements from the GNS–Ag film couplings were significantly larger (16–42× larger) than those from GNS–GNS couplings, indicating the significance of GNS–Ag film couplings in the overall E-field enhancements for systems with GNS assemblies on Ag films. For GNS assemblies on glass substrates, the field enhancements from both GNS–GNS and GNS–glass couplings exhibited exponential decay with interparticle separation (Fig. 3(f)). The dependence of the field enhancement on the interparticle separation was stronger for GNS–GNS coupling, which was indicated by the larger amplitude (a = 18.7) for GNS–GNS coupling compared with that for GNS–glass coupling (a = 2.0). It is also worth noting that the field enhancement from interparticle couplings on the Ag film showed 30–60% larger values than those on a glass substrate; this was attributed to the enhanced interparticle plasmon couplings on the Ag film compared with those on the glass substrate due to the propagating SPP modes in the Ag film.47
The above FDTD simulations of field enhancements for GNS assemblies on Ag and glass substrates indicated that the E-field enhancement was predominantly from particle–film couplings for GNS assemblies on Ag films, whereas the system of GNS assemblies on glass substrates had E-field contributions from both weak particle–film and interparticle plasmon couplings. As reported previously, both the resonant wavelength shift and the field enhancement exhibit a universal exponential or power law dependence on the interparticle separation for metal nanoparticle dimers on dielectric substrates.60,61 For metal nanoparticles on metal substrates, previous research has shown that the power law function of both the resonant wavelength shift and field enhancement is dependent on the particle–film separation.43 However, it is still unclear whether there is a general law governing the relationship between the overall SERS enhancements and the interparticle separations for GNS assemblies on metal substrates.
To investigate this relationship further, we compared FDTD simulation results with the experimental results of the SERS intensity as a function of interparticle separation. For this purpose, we prepared different surface densities of GNSs on Ag and glass substrates (Fig. S5†) and measured the SERS intensities of benzenethiol molecules (Fig. S6†). For the analysis of SERS intensity as a function of interparticle separation, we converted the experimental GNS surface density into interparticle separation by approximating the randomly distributed GNS assemblies as hexagonal packed GNS assemblies. Additionally, to compare the experimental SERS intensity with the simulated EF, we used a GNS density-normalized EF (EFN), which can be approximated as EFN = (4|E/E0|GNS–film4 + (|E/E0|GNS–GNS4)) × (GNS surface density), where E and E0 are the local and incident electric fields, respectively.62 A detailed description of EFN is given in the ESI.†Fig. 4 shows the simulated EFN and the experimental SERS intensity as a function of the interparticle separation on Ag and glass substrates.
Fig. 4(a) and (b) show that the normalized EF as a function of interparticle separation was well fit by the power law (y = axb), where a and b are fitting parameters. We also determined that the normalized EF of the GNS assemblies on Ag films, as a function of interparticle separation, followed a broken power law function or a combined power law distribution (Fig. 4(a)). At first, the normalized EF increased as the interparticle separation increased, following a power law function with an index (b) of 0.31 ± 0.05 for interparticle separation below ∼60 nm; however, the normalized EF decreased with b = −0.68 ± 0.05 for interparticle separation over ∼60 nm. This behavior can be explained by the interplay between interparticle and particle–film plasmon couplings, where the particle–film plasmon couplings become weak when there is strong interference from interparticle couplings. Therefore, when the interparticle separation was below ∼60 nm, i.e., below ∼0.8 times the GNS size (80 nm) (i.e., the strong interparticle coupling range), the E-field from particle–film couplings increased exponentially as the interparticle separation increased or as the interparticle coupling decreased exponentially (Fig. 3(e)). When the interparticle separation was over ∼60 nm (i.e., the weak interparticle coupling range), the increase in the E-field from particle–film couplings began to saturate as the interparticle separation increased. Because the particle–film couplings dominate in the weak interparticle coupling range, the normalized EF strongly depended on the number of hot spots from particle–film couplings, which in turn depended on the GNS assembly surface density. Therefore, the normalized EF decreased with an increase in the interparticle separation or decrease in the GNS surface density for an interparticle separation over ∼60 nm.
For GNS assemblies on glass substrates, the EF decreased as the interparticle separation increased, as described by the single power law index of b = −0.90 ± 0.02, in agreement with previous reports (Fig. 4(b)).61 The smaller value of the power law index (b = −0.68) for the GNS assemblies on Ag films, as compared to that on glass substrates (b = −0.90), indicated that the SERS enhancement was less sensitive to interparticle separation. To compare the simulation results with the experimental results, Fig. 4(c) and (d) show the measured SERS intensity as a function of interparticle separation. The experimental SERS intensity for GNSs on Ag films exhibited a value of b = −1.71 for interparticle separation over ∼80 nm and b = −1.70 for GNSs on glass substrates. The difference in the power law exponents for the simulated EF and experimental SERS intensity is attributed to the regular GNS assemblies used in the simulated EF as opposed to the random distribution of GNSs in the experimental SERS results, which resulted in the dispersion of interparticle separations and occasional nanostar aggregations. It is worth noting that the trends in simulation results match quite well with the experimental results although the simulation work is based on regular GNS assemblies, where the long-range effects could affect the trends. However, in our study, the long-range effects from the regular GNS assembly in the simulation results can be considered to be minimal since the periodicity is far below the excitation wavelength.
Because the GNS surface density can affect the interplay between the interparticle and particle–film plasmon couplings, the density-controlled, optimally designed SERS substrates could be advantageous for trace-level detection of analytes. Fig. 5(a) shows the Raman spectra of benzenethiol (concentrations varying from 10 nM to 1 aM) adsorbed on optimized SERS substrates (GNS assemblies on Ag films with a GNS surface density of ∼43 GNSs per μm2). Optimization of the SERS substrates with maximum SERS effects was enabled by controlling the GNS surface density. The Raman spectra of benzenethiol with C–H out-of-plane bending (998 cm−1) and C–C symmetric stretching (1021, 1071 cm−1) were observed. Compared with the Ag/PDDA/GNS substrates without benzenethiol molecules, the Raman peak at 998 cm−1 was clearly visible down to 1 attomole (aM) with the signal-to-noise ratio over ∼6 (Fig. S7†), demonstrating the ultrasensitive detection capability of the SERS substrates based on GNS assemblies on metal films. This detection capability is not achievable with typical gold nanosphere assemblies on Ag films (Fig. S8†). The ultrasensitive detection capability of our SERS substrates is attractive for the trace level detection of explosives. Since the 2,4-dinitrotoluene (2,4-DNT) is a residue of TNT-based explosives, the detection of trace amounts of 2,4-DNT is an important requirement for the landmine detection and security checkpoints in the airport. As shown in Fig. 5(b), our SERS substrates allow the detection of 10 aM of 2,4-DNT with characteristic vibration modes of NO2 out-of-plane bending modes at 834 cm−1 and NO2 symmetric stretching modes at 1355 cm−1. The detection of DNT down to 10 aM is well below the best known detection limit of DNT (∼4 fM) based on 3D SERS substrates.63
For the SERS measurement of analytes, SERS substrates were prepared by varying the concentration of benzenethiol (Sigma-Aldrich) solution from 100 μM to 1 aM in ethanol. The SERS substrates were immersed in the benzenethiol solution for 15 min, rinsed with ethanol, and then blown dry with nitrogen gas. For the SERS measurement of 2,4-DNT, the SERS substrates are immersed in the DNT solution in methanol with the concentration from 100 μM to 10 aM for 2 hours, rinsed in methanol, and blown dry. The Raman spectra of the analytes were collected with 785 nm laser excitation using a confocal Raman microscope (Alpha 300, WITec) and an objective lens (20×, numerical aperture = 0.4). The integration time was 0.5 s; the laser power was ∼15 mW.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3nr04752k |
‡ These authors contributed equally. |
This journal is © The Royal Society of Chemistry 2014 |