NANO LETTERS
Subdiffraction Limited, Remote Excitation of Surface Enhanced Raman Scattering
2009 Vol. 9, No. 3 995-1001
James A. Hutchison,† Silvia P. Centeno,† Hideho Odaka,† Hiroshi Fukumura,‡ Johan Hofkens,† and Hiroshi Uji-i*,† DiVision of Molecular and Nano Materials, Department of Chemistry, Katholieke UniVersiteit LeuVen, Celestijnenlaan 200F, HeVerlee 3001, Belgium Department of Chemistry, Graduate School of Science, Tohoku UniVersity, Sendai 980-8578, Japan Received October 9, 2008; Revised Manuscript Received January 5, 2009
ABSTRACT We demonstrate that focused laser excitation at the end of silver nanowires of 50-150 nm diameter excites SERS hot-spots at points of nanoparticle adsorption many micrometers along the wire due to the plasmon waveguide effect. The total SERS intensity detected at the hot-spots following wire-end excitation correlates with the known wavelength, polarization, and distance dependences of surface plasmon polariton (SPP) propagation in nanowires. The SERS spectra obtained at the hot-spots following wire-end excitation show very little background compared to when excitation occurs directly at the hot-spot, suggesting that a much smaller SERS excitation volume is achieved by remote, waveguide excitation. The ability to transfer SERS excitation over several micrometers, through a structure with a subdiffraction limit diameter, is discussed with respect to potential high-resolution SERS imaging applications.
Surface enhanced Raman scattering (SERS) spectroscopy is a promising technique for biochemical imaging and sensing as it can provide information on chemical structure with very high sensitivity.1 Under specific conditions SERS from single molecules can be detected, requiring enhancement factors of ∼1014-1015 over the typical Raman scattering cross sections of individual molecules (∼10-29-10-31 cm2).2-5 Such huge enhancements are usually realized at coinage metal surfaces with nanoscale features (e.g., roughened electrodes, nanoparticle aggregates), which can sustain collective oscillations of surface electrons (surface plasmons) resonant with typical visible/near IR optical excitation fields.6-8 The electromagnetic (EM) enhancement of both the excitation and scattered fields by surface plasmon resonance (SPR) is thought to contribute most strongly to the SERS effect,4 while specific molecule-metal interactions such as charge transfer (CT) processes also make an important contribution, known as chemical effects.9 SERS enhancement strongly depends on the metal, size and shape of the nanoscale structures of a substrate. Since the edges and points of nanoparticles with nonisotropic shapes have been shown to be points of high concentration of EM fields, the SERS properties of nanotriangles, nanocubes, nanorods, nanowires, and so forth continue to be investi* To whom correspondence should be addressed. E-mail: Hiroshi.Ujii@ chem.kuleuven.be. † Katholieke Universiteit Leuven. ‡ Tohoku University. 10.1021/nl8030696 CCC: $40.75 Published on Web 02/03/2009
2009 American Chemical Society
gated.10-13 The highest enhancements are obtained at socalled “hot-spots” in the nanogaps between nanoparticles, leading to huge interest in the coupling of different nanoparticle types in aggregate systems.6,7 More recently ordered arrays of particles,14 and nanoapertures,15,16 have been pursued as these substrates provide better spatial distribution of SERS hot-spots formation and therefore more homogeneous and reproducible detection. A fundamental challenge particularly for SERS imaging applications is the generation and control of the position of a SERS hot-spot relative to the sample of interest. Tipenhanced Raman scattering (TERS) has been developed in response to this challenge. The essential feature of TERS is the concentration of electromagnetic fields at a sharp, metalized tip due to localized SPR.17,18 Combined with scanning probe microscopy (SPM), the position of the nearfield of the tip can be controlled very precisely relative to a sample affording high lateral resolution (down to 20 nm). In most TERS setups however, both the tip and sample must be exposed to the diffraction-limited focal spot of the exciting laser (the spot having a diameter roughly half the wavelength employed, ∼300 nm for 600 nm excitation). Since the area of illumination is much greater than the area of SERS enhancement, achieving adequate contrast between unenhanced Raman from the far-field, and SERS from the nearfield of the tip can be a challenge.18,19 Furthermore sample
Figure 2. A schematic representation of remote excitation by plasmon propagation of SERS hot-spots at points of NPs adsorption on a pATP-nanowire.
Figure 1. The transmission, wide-field illumination, and focused laser excitation images of a pATP-coated silver nanowire (a-c, respectively), and a pATP-nanowire with several NPs adsorbed (d-f, respectively). Focused laser excitation was at the left-hand side of the wires in panels c and f. (g) AFM image of two pATPnanowires showing adsorption of individual silver NPs. (h) Raman spectra of pATP. Top: Normal Raman spectrum. Bottom: SERS spectra recorded at a hot-spot on a NPs-pATP-nanowire with excitation polarization parallel (black line) and perpendicular (dashed line) to the long axis of the nanowire.
photodegradation can become an acute problem, especially in biological studies. In this paper we propose nanowire/nanoparticle aggregate systems as potential probes for SERS sensing/imaging applications, in which excitation can be delivered to a SERS hot-spot many micrometers away from the point of laser excitation via plasmon propagation along the nanowire. The plasmon waveguide effect in nanowires, the coupling of free-space photons to the conduction electron gas of the nanowire to form propagating surface plasmon polaritons (SPPs), allows the spatial confinement and transport of light through structures with subdiffraction limit diameters. The effect is of enormous interest in the field of plasmonics for the miniaturization of photonic devices.20-24 SPPs confined at the surface of single crystalline metal nanowires can propagate over tens of microns before energy is lost by Ohmic damping.25-28 Alternatively, the SPPs can be scattered into free-space photons at a disturbing point, such as the end or a sharp corner in the nanowire, or (of most interest here) by coupling with a nanoparticle adsorbed at the surface of the nanowire.29,30 We demonstrate here that light excitation focused at the end of silver nanowires 2-30 µm long and 50-150 nm in diameter can excite SERS hot-spots at points of nanoparticle (NP) adsorption on the wire many micrometers away from 996
the laser focus (Figure 1 and schematic in Figure 2). paraAminothiophenyl (pATP) acts both as a coupling agent for the attachment of NPs to the wire and as a nonresonant Raman reporter in the experiments. We show that the total SERS signal intensity detected after remote excitation of the hot-spots correlates with the well-known wavelength, polarization, and distance dependences of SPP propagation along nanowires.25-28,30 The SERS spectra of pATP obtained at hot-spots during remote excitation show very little background compared to when excitation occurs directly at the hot-spot with both EM and CT contributions observed. These excellent SERS properties, combined with the subdiffraction limit diameter of the nanowire/nanoparticle systems, are discussed with regard to their potential as probes for high-resolution SERS sensing/imaging applications. The silver nanowires used in this study were synthesized following the well-established polyol method, in which ethylene glycol acts as both the solvent and reducing agent in the presence of poly(vinyl pyrrolidone) (PVP) (Supporting Information).27,31 The resulting nanowires had diameters of 50-150 nm and lengths of 2-30 µm. The surface of the nanowires was functionalized by pATP (Supporting Information). The binding of pATP to silver occurs through the thiol group by a sSsAg bond (Figure 2).32 The functionalized pATP-nanowires were then deposited onto glass coverslips. Because of the free amino group of pATP, pATP-nanowires strongly attach to glass surfaces by electrostatic interaction and were not displaced during rinsing with ethanol and Milli-Q water. Silver NPs were prepared using a citric acid reduction method.33 The diameter of the NPs was estimated to be ∼40 nm by UV/vis absorption spectroscopy. Different concentrations of solutions of silver NPs were then spincoated onto the pATP-nanowire samples and the resulting NPs-pATP-nanowire samples were rinsed with ethanol and Milli-Q water and dried with argon. The optical properties of the nanowire/nanoparticle systems were studied using an optical microscope (IX71, Olympus) equipped with a piezoelectric stage (Physik Instrument (PI) GmbH & Co.). Unless indicated, both widefield illumination and diffraction-limited focused excitation was provided by a continuous wave 632.8 nm He-Ne laser (Model 1145P, JDS Uniphase Co.). Excitation (488 nm) was provided by a continuous wave Ar+ laser (Stabilite 2017, Nano Lett., Vol. 9, No. 3, 2009
Spectra-Physics). For wide-field illumination, collimated laser light was focused into the back focal plane of the objective (Plan Fluorite, NA 1.3, 100×, Olympus) via a prefocus lens. In order to achieve a diffraction-limited focus, collimated laser light was guided to the objective. Excitation polarization was controlled using λ/2 and/or λ/4 waveplates (Zero-order waveplates, ThorLabs Inc.). Laser light was directed to the sample via the appropriate dichroic mirror (z633rdc for 633 nm excitation or z488rdc for 488 nm excitation, Chroma Technology Co.). The scattering/emission from the sample was collected using the same objective and passed through the dichroic mirror to the detection apparatus. A long pass cutoff filter (OD > 5 at the excitation wavelength) was employed before the detector to reject the excitation light (HQ645LP for 633 nm excitation or HQ500LP for 488 nm excitation, Chroma Technology Co.). In order to realize Raman imaging, 50% of the collected light was detected with a cooled charge-coupled device (CCD) camera (Cascade 512B, Princeton Instruments). To measure scattering spectra, 50% of the collected light was passed through a 100 µm pinhole and then guided through a spectrograph (250IS/SM, Chromex) to a cooled CCD camera (LN/CCD-512B, Princeton Instruments). This allowed a spectrum of light emission/scattering from a circle of 1 µm diameter in the sample plane to be selectively detected (due to the 100× magnification of the objective). Spectra could thus be collected from individual hot-spots along the wire during remote excitation, so long as the spots were separated by more than 1 µm. To achieve this, the sample was moved by piezoelectric stage so that the SERS hot-spot of interest was at the correct region for spectral detection, then the laser focus was moved carefully to the end of the nanowire for remote excitation by adjusting two mirrors before the microscope. The focus position was adjusted to obtain the highest signal on each hot-spot by checking both the wide-field Raman image and the transmission image. The most intense signals were always obtained when the focus was located exactly at the edge of the wire end. The spectra and Raman images shown in this article were obtained with 10 s and 100 ms integration time, respectively. At the focus point, the excitation power used was in the range 1 to 50 kW/cm2. Figure 1 shows optical images of individual pATPnanowires before and after spin-casting with silver NPs. For pATP-nanowires without any NPs spin-cast, the transmission images show no features (Figure 1a) and no emission/ scattering upon wide-field illumination (Figure 1b). The pATP-nanowires exhibit the plasmon waveguide effect when one end of the nanowire is irradiated with a diffractionlimited laser focus (Figure 1c). The bright spot at the left is due to intense scattering at that end of the wire due to the focused excitation. Light can be seen to be relatively weakly scattered along the wire close to the end where excitation occurs and also to be emitting from the nonilluminated end. The weak scattering along the wire is attributed to nanowire surface roughness leading to some leakage of the SPP into far-field radiation. Similar features were observed for nanowires that were not treated with pATP (i.e., PVP-capped nanowires). Nano Lett., Vol. 9, No. 3, 2009
In contrast, when NPs were spin-cast onto the pATPnanowires, small features could be recognized in the transmission images, suggesting formation of the NPs-pATPnanowires (Figure 1d). Furthermore, bright spots were observed on these nanowires when irradiated by wide-field illumination (Figure 1e) as well as by laser focus at the left end of the wire (Figure 1f). The number of bright spots observed on the nanowires depended on the concentration of the NP solution used for spin-coating. Figure 1g shows an atomic force miscroscopy (AFM) image of two NPspATP-nanowires from the sample used in experiments. The image clearly shows the adsorption of individual silver NPs along the wires; however aggregates of NPs were also found to be adsorbed in the AFM experiments (Figure S1, Supporting Information). The bright spots observed upon illumination of the NPspATP-nanowires most likely occur at positions of NP adsorption on the wire (Figure 1e,f). It has been demonstrated previously that SPPs excited in nanowires can be scattered into far-field radiation at points of NP adsorption, where there are few NPs adsorbed per nanowire.29,30 Recently we have observed regularly spaced bright spots on nanowires with a very dense coverage of adsorbed NPs under wide-field illumination.34 In the current study, it cannot be said if it is a single NP adsorbing at a bright spot on the pATPnanowires or an aggregate of NPs. The different relative intensities of bright spots observed on the same wire under wide-field illumination (Figure 1e) suggest that local differences in the binding of NPs to the pATP-nanowire do indeed occur. The driving force for the attachment of the silver NPs to the pATP-nanowires is the electrostatic interaction between the citrate-coated NP and the amino group of the pATP.35 Figure 1h (bottom) shows the emission spectra collected at a bright spot when polarized laser light was focused at one end of a NPs-pATP-nanowire. In the case of Figure 1h, the length of the nanowire was 10 µm and it had only one bright spot, located approximately 7 µm away from the laser focus. The spectra in Figure 1h (bottom) can be assigned to SERS of pATP by comparison with the normal Raman spectrum of pATP (Figure 1h, top) and with literature reports of SERS spectra of pATP.32,35,36 The normal Raman spectrum of pATP has been studied extensively (Figure 1h, top).32,35,36 It is mainly characterized by two strong bands recorded at 1090 and 1600 cm-1, assigned to the totally symmetric modes 7a (C-S stretch) and 8a (C-C stretch), respectively. The medium intensity band at 825 cm-1 is assigned to the ring breathing mode. The spectra measured at the hot-spot on the NPs-pATPnanowire (Figure 1h, bottom) are similar to several reports of SERS spectra of pATP.32,35,36 The band with the strongest intensity is recorded at 1432 cm-1 and is assigned to the in-plane vibrational 19b mode of B2 symmetry. Additional to this the bands recorded at 1394 and 1143 cm-1 and assigned to 3;B2 and 9b;B2 modes, respectively, show similar or even higher intensities compared to the 7a mode. The medium intensity band recorded at 1583 cm-1 is assigned to the 8b;B2 mode. 997
The origins of the SERS intensities of these B2 modes was investigated by Osawa et al. and are proposed to be related to the CT mechanism since these vibrations show a strong dependence on excitation wavelength and on applied electric potential in experiments carried out at an electrode.32 Conversely the 7a mode was shown to be insensitive to the applied potential and is therefore considered to have contributions mainly from the EM mechanism. The bright spots observed on the NP-pATP-nanowires during end-excitation are therefore attributed to SERS from pATP molecules in the nanogap hot-spot between the wire and the NPs (Figure 2). SERS spectra of molecules in the gap between a NP and a nanowire have been observed recently with excitation directly at the NP.37,38 In our case, we propose that the SERS hot-spots are instead generated when a propagating SPP on the wire couples to an adsorbed NP. It should be noted that PVP-capped nanowires also showed SERS signals due to PVP at points of NP adsorption during remote excitation (see Supporting Information, Figure S2). In studies of NP aggregates, it is often found that NP junctions exhibit vastly differing efficiencies of SERS hotspot generation, primarily due to the difficulty of controlling the separation and geometry, and therefore the coupling between adjacent NPs.6,7 In the NPs-pATP-nanowires, the distance between NPs or NP aggregates and the nanowire is expected to be relatively well-controlled due to chemical linkage via pATP (samples were thoroughly rinsed with ethanol and water to remove any NPs not electrostatically bound to the pATP). Nevertheless, variation of SERS intensity at different hot-spots on the same wire is clear during wide-field illumination studies (Figure 1e) and it is possible that some adsorbed NPs and NP aggregates show very weak or no SERS activity and are not detected in the optical studies presented herein. The percentage of adsorbed NPs and NP aggregates that generate detectable SERS hotspots in the NPs-pATP-nanowire systems is currently under investigation in our laboratory using a combination of optical spectroscopy and AFM techniques. As the laser excitation was strongly scattered when focused at the end of the nanowires, it was important to rule out Rayleigh scattering as the source of the observed excitation of SERS hot-spots (the extent of this scattering is not visible in the images shown in Figure 1 due to optical filtering). To do this, Raman imaging was performed on two adjacent nanowires as shown in the transmission image in Figure 3a, where a short wire (wire 1) and a longer wire (wire 2) lay almost parallel, separated by only ∼1 µm. Figure 3b,c shows the Raman images when the left ends of wire 1 and wire 2, respectively, were excited by a focused laser. It is clearly seen that bright spots are observed only on the irradiated nanowire in each case (this effect is observed irrespective of the excitation polarization employed). Therefore, SERS hot-spot excitation is not likely due to the Rayleigh scattering of the focused laser if the distance of the hot-spot from the point of excitation is greater than 1-2 µm. The SERS signal intensity at a hot-spot following remote excitation depends strongly on the excitation polarization. 998
Figure 3. (a) Transmission image of two NPs-pATP-nanowires separated by 1-2 µm. (b) Focused laser excitation at the left end of wire 1 (excitation point indicated by yellow circle). (c) Focused laser excitation at the left end of wire 2 (excitation point indicated by yellow circle).
A much higher SERS signal was observed when the focused laser was polarized along the long axis of the nanowire (Figure 1h, bottom, solid line), compared to when the polarization was parallel (Figure 1h, bottom, dashed line). This correlates with the known polarization dependence of the coupling efficiency of incident light into nanowires to form SPPs.26,30 The SERS signal intensity at a hot-spot following remote excitation also shows a strong dependence on distance from the point of excitation. Figure 4a shows the transmission image of a NPs-pATP-nanowire with length ∼19 µm and that displayed seven SERS hot-spots when excited both by wide-field illumination and by laser focus at one end (Figures 4b,c, respectively). For Figure 4c, the first and last bright spots correspond to the laser focus point at the left end of the wire and to the scattering point at the nonilluminated end. The 3D view of the image in Figure 4c is shown in Figure 4d. The intensities of the SERS hot-spots decay in a roughly exponential fashion as a function of distance along the wire (Figure 4e). The decay was fitted as per eq 1 in which x is the distance from the SERS hot-spot closest to the end of focused excitation and L is the distance taken for SERS intensity (I) to drop to 1/e of its value at the hot-spot closest to the end of excitation (I0). I(x) ) I0e-x⁄L
(1)
For the wire in Figures 4a-c, a value for L of 3.4 ( 1.6 µm was extracted. A second example of exponential decay of SERS hot-spot intensity along a wire (L ) 1.9 µm) is Nano Lett., Vol. 9, No. 3, 2009
Figure 4. Transmission (a), wide-field illumination (b), and focused laser excitation (c) images of a NPs-pATP-nanowire exhibiting seven SERS hot-spots. Focused excitation was at the left end of the wire in panel c. (d) 3D view of the image in panel c. (e) The SERS intensity of the seven hot-spots along the wire in panel as a function of distance from the point of remote excitation. The intensity of scattered light from the illuminated and nonilluminated ends is indicated. A characteristic length L ) 3.4 ( 1.6 µm was extracted from an exponential fit to the decay of SERS hot-spot intensity over distance (solid line, see also eq 1). (f) SERS spectra collected at hot-spots 1, 4, and 5 shown in panel c, following remote excitation.
shown in Supporting Information, Figure S3. While these analyses are rudimentary and ignore damping due to SPP leakage at the multiple hot-spots, the L values extracted are similar to the characteristic propagation lengths observed for SPP propagation in silver nanowires of similar dimensions and measured with excitation at 800-830 nm (∼3 µm).26-28 Exponential decay behavior is not observed on every wire and not observed at all times (see Figures 1f and 3b,c). The uncertainty in the intensity of the SERS hot-spots indicated in Figure 4e is due to the hot-spots showing burst behavior over time. The SERS intensity trajectories of the hot-spots observed in Figure 4c are included as Supporting Information, Figure S4. Bursts of high SERS intensity occur even at hot-spots very far from the point of excitation and can last for as much as 0.5 s. Another likely reason for the nonexponential distance dependence of SERS hot-spot intensity observed on some wires following remote excitation is the inhomogeneous binding of NPs to the pATP-nanowire (e.g., individual NPs or NP aggregates). Furthermore it was observed that nonexponential behavior became more apparent as the sample aged over days, suggesting that oxidation of the wire and/or particle influenced SERS hot-spot generation by remote excitation. Dickson and Lyon observed a strong wavelength dependence for SPP propagation in gold nanowires, with 820 nm excitation light propagating much more efficiently than 532 nm excitation light.25 We observe a similar trend here for the wavelength dependence of SERS hot-spot excitation following remote excitation. When the same NPs-pATPNano Lett., Vol. 9, No. 3, 2009
Figure 5. SERS spectra collected from the same hot-spot on a NPspATP-nanowire following (top) wire-end excitation and (bottom) on-hot-spot excitation. The hot-spot was located ∼4 µm from the end of the wire where focused excitation occurred. Excitation intensity was adjusted to get a similar signal in the maximum peak in each case.
nanowire was irradiated with both 488 and 633 nm focused excitation at one end, hot-spots were generated much further across the length of the wire for 633 nm excitation (see for example Figure S5 in Supporting Information). This wavelength dependence, together with the distance and polarization dependences discussed above, strongly suggest that the plasmon waveguide effect is responsible for the generation of SERS hot-spots following remote excitation. Figure 4f shows the SERS spectra at three of the hotspots on the NPs-pATP-nanowire shown in Figure 4c following remote excitation. As discussed earlier, the bands in the spectra can be identified as SERS of pATP. It is 999
interesting to note that the bands between 1140 and 1580 cm-1, associated with CT, change their relative contribution to the spectrum as a function of distance of the hot-spot away from the excited end of the wire. Although the relative changes are small, they indicate a change in the propagating SPP and its interaction with the pATP/adsorbed NP as a function of propagation distance. The exact explanation for this phenomenon is currently under study. We note also that the relative contribution of the background continuum is reduced for spectra measured at hotspots >2 µm away from the point of focused excitation. Finally in Figure 5 we compare the SERS spectra recorded at the same hot-spot on a NPs-pATP-nanowire following focused excitation at the wire end (top) and following focused excitation directly on the hot-spot (bottom). The relative intensity of the 7a mode is substantially increased when excitation occurs directly at the hot-spot (this is clearer from background-subtracted spectra of Figure 5, shown in Supporting Information, Figure S6). The relative contribution of the background continuum is also much greater following excitation directly at the hot-spot. This background has been shown to be strongly correlated with the SERS effect and in the case of silver nanostructures is thought to be due in part to luminescence from the metal,39,40 while emission or Raman from PVP, other chemicals, and/or oxidized metal in the region of focused excitation can also contribute. The reduced contribution of this background in the case of remote excitation, seen in both Figure 5 (top) and Figure 4f, is likely due to the fact that SERS excitation of pATP occurs only in the nanogap between the wire and the NPs, rather than in a ∼300 nm diameter diffraction limited laser spot. The reduced volume of SERS excitation afforded by waveguide excitation as compared to direct focused excitation highlights the potential of these nanowire/nanoparticle systems as probes for high-resolution SERS sensing/imaging applications. The combination of these systems with SPM will allow the nanovolume SERS hot-spot region to be manipulated relative to a sample of interest, as per highresolution TERS experiments. The attachment of silver nanowires to AFM probes for use in TERS experiments has been demonstrated.41 The obvious challenge is to vary the shape of one end of the nanowire, or to vary the method and position of attachment of the NP(s) to the nanowire, so that the sample of interest can easily access the nanogap hotspot region for SERS detection. In NPs-pATP-nanowires we estimate that the separation between NPs and the nanowire is less than a few nanometers (in range of the thickness of a molecular layer of pATP); however a precise experimental determination was not obtained. By employing other bifunctional linker molecules of different lengths, the dependence of the nanogap distance on SERS efficiency can be tested and access to the hot-spot region for an external sample of interest may be possible. Recently, metal or carbon nanotube “nanoneedles” have been combined with SPM to allow probing inside single live cells.42,43 The nanoneedles are usually a few micrometers in length, with diameters of
