Spontaneous versus Stimulated Surface-Enhanced Raman Scattering of Liquid Water.

We have observed for the first time the surface-enhanced (SE) signal of water in an aqueous dispersion of silver nanoparticles in spontaneous (SERS) and femtosecond stimulated Raman (SE-FSRS) processes with different wavelengths of the Raman pump (515, 715, and 755 nm). By estimating the fraction of water molecules that interact with the metal surface, we have calculated enhancement factors (EF): 4.8 × 106 for SERS and (3.6-3.7) × 106 for SE-FSRS. Furthermore, we have tested the role of simultaneous plasmon resonance and Raman resonance conditions for the aν 1 + bν3 overtone mode of water (755 nm) in SE-FSRS signal amplification. When the wavelength of the Raman pump is within the plasmon resonance of the metal nanoparticles, the Raman resonance has a negligible effect on the EF. However, the Raman resonance with the aν 1 + bν3 mode strongly enhances the signal of the fundamental OH stretching mode of water.

Introduction

Raman spectroscopy provides rich chemical and structural information on the studied samples and is, potentially, an extremely useful tool in chemical, biomedical, and material sciences. Still, its applications in the past have been limited because the measured signal is weak due to the low conversion of photons in the Raman process (just 1 Raman photon out of 1 million pump photons for a typical scattering cross-section).[1] One of the promising methods used to amplify the Raman signal is the deposition of an analyte on a (noble) metal surface, leading to the enhancement of the Raman signal in a process that involves surface plasmons. It was estimated that the Raman signal enhancement factor (EF) in surface-enhanced Raman scattering (SERS) is on the order of 106.[2] Another approach to Raman signal amplification relies on nonlinear optical processes (i.e., coherent anti-Stokes Raman scattering (CARS) or stimulated Raman scattering (SRS)). It has been shown that photon conversion in a stimulated Raman process can be as high as 10%.[1] Both CARS and SRS require a strong pump and are usually performed with short laser pulses. Applying picosecond or femtosecond pump pulses allows for sufficiently large instantaneous intensities required for CARS and SRS with relatively low average intensities, which alleviates the sample degradation. Additionally, short laser pulses enable one to study ultrafast dynamics in the sample. In traditional SRS, two narrow-bandwidth laser lines were used to produce a single stimulated Raman line from a sample.[3] In the femtosecond stimulated Raman scattering spectroscopy, first reported by Yoshizawa et al.[4] and developed later by Richard Mathies’ group,[5] the Raman transition is induced with two pulses: a narrow-bandwidth, strong picosecond pulse to initialize the Raman transition (Raman pump) and a weak, broadband femtosecond probe pulse to stimulate the Raman transition (Raman probe). Such an FSRS scheme experiment, which in principle enables multiplex detection with a single laser shot, was used in this article.

Recently, a very attractive possibility of combining SERS and either CARS or FSRS and thus amplifying the Raman signal even more has been studied in some detail. The first experimental demonstration of surface-enhanced femtosecond stimulated Raman scattering (SE-FSRS) has allowed for the precise determination of the effective EF and its dependence on experimental parameters, although the aggregation of gold dumbbell nanostructures was not controlled.[6] A total EF of 1011 has been demonstrated in the CARS experiments with para-mercaptoaniline molecules deposited on engineered flat plasmonic structures.[7] In a similar experiment exploiting ultrafast measurements, quantum beats have been observed in a trans-1,2-bis(4-pyridyl) ethylene molecule placed in the vicinity of a gold dumbbell nanostructure.[8] The method was sensitive enough to allow for single-molecule identification. Ultrafast surface-enhanced and stimulated Raman spectroscopy proved to be a valuable tool in studies of the dynamics of plasmon-assisted chemical reactions.[9]

Here, we report experiments on the surface-enhanced femtosecond stimulated Raman scattering of water molecules in aqueous dispersion of silver nanoparticles. We compare EFs of the SE-FSRS signal at different Raman pump and Raman probe wavelengths with spontaneous surface-enhanced Raman scattering (SERS) of the same samples. Such an approach provides information about the mechanism of signal amplification in SE-FSRS.

So far, mostly electrochemical SERS signals of water at electrode/electrolyte interfaces have been registered. Many investigations have concentrated on aqueous solutions of electrolytes between metal electrodes (Ag, Au, and Cu) in the negative potential range.[10−13] It has been assumed that the SERS observed for water molecules most probably arises from surface complexes involving metal adatoms (clusters), halide ions, cations, and water rather than bulk water present at electrode/electrolyte interfaces.[12] Spectra presented in those studies are dramatically different from the regular Raman spectrum of water.[13] Very recently, the SERS signal of water was observed on several solid metallic substrates.[14]

In this work, we demonstrate that SERS and SE-FSRS spectra of liquid water can be obtained. This conclusion is supported by the fact that the spectra we register resemble those of neat water very closely. This finding is particularly interesting for SE-FSRS as it presents the possibility for time-resolved femtosecond stimulated Raman studies of the vibrational relaxation and dynamics of water present at a surface of a metal nanoparticle and water interactions with the surface plasmon of metals.

Experimental Section

Sample Preparation

Silver nanoparticles (AgNPs) were synthesized via the simple chemical reduction of silver nitrate with sodium borohydride.[15] The amount of added potassium bromide during synthesis was crucial to AgNP size control. Samples without KBr turned blue, and samples with 40 μL of KBr turned yellow. Synthesized AgNPs were studied by UV–vis and Raman spectroscopy without any purification. Water for measurements of the Raman spectra (reference) and for synthesis was purified with a Merck Millipore system.

Sample Characterization

Absorption spectra of AgNP dispersions were obtained with the use of a double-beam UV–vis–NIR spectrophotometer (Cary 5000, Varian) with superb photometric performance in the 200–800 nm range in relation to pure water as a reference sample. AgNP samples were measured in QX quartz cuvettes (Hellma) with a 1 mm optical path, while the water absorption spectrum was measured in a QX cuvette with a 1 cm optical path in relation to an empty cuvette as a reference.

The characterization of the shape and size estimation of synthesized AgNPs was performed on the basis of the transmission electron microscopy (TEM) images. Images were obtained with the use of a Jeol ARM 200F high-resolution transmission electron microscope. The metal nanoparticle size was estimated with the use of ImageJ software. The diameter of nanoparticles was estimated with a circle circumscribed on the nanoparticle. More than 100 results for each sample were analyzed to construct a histogram and to determine the average size and size distribution. Exemplary TEM images of the nanoparticles, together with the determined distribution of their sizes, are shown in Figure .

Figure 1

Exemplary TEM images of AgNPs (scale bar is 20 nm): (a) particles from AgNPs blue; the inset is a side view of a nanoprism (scale bar in the inset is 5 nm) and (b) particles from AgNPs yellow and their size distributions.

The concentration of Ag in AgNPs dispersion was determined by flame atomic absorption spectrometry (FAAS) using a GBC 932 plus instrument. Calibration was performed using the silver standard solution (Merck).

Spontaneous Raman Scattering Measurements

Spontaneous Raman spectra were collected using a T64000 (JobinYvon) triple-grating spectrometer (Ar laser excitation line at 514.5 nm) with a spectral resolution of ca. 0.5 cm–1. Measurements of samples were carried out in a macrochamber, in a spectrofluorometric QX quartz cuvette (Hellma) in a classic configuration of illumination–observation geometry (excitation and scattered beams were perpendicular). The laser power determined on the sample was in the range of 25–30 mW. The time of acquisition of the spectra was adjusted to obtain high-quality Raman spectra.

Stimulated Raman Scattering Measurements

The stimulated Raman scattering measurements were performed with the use of the setup for pump–probe femtosecond stimulated Raman scattering which is described in detail in ref (16), yet only the stationary-state (not time-resolved) measurements were performed. The setup is based on a commercial femtosecond Yb:KGW laser system (Pharos, Light Conversion) which produces 200 fs pulses centered at 1030 nm with a repetition rate of 1 kHz. Raman pump lines were generated in a home-built picosecond OPA (optical parametric amplifier) in the process of frequency mixing of two oppositely chirped copies of the same femtosecond pulse. The generated Raman pumps had narrow bandwidths (∼5 cm–1 for 515 nm and around 12–13 cm–1 for 715 and 755 nm pump) and were 1–3 ps long. (the construction of OPA and the method for preparation narrow-bandwidth ps pulses was described in ref (17)). The Raman pumps’ energies were 2 μJ for 515 nm, 1.1 μJ for 715 nm, and 0.95 μJ for 755 nm. The beam waists of the pumps in the focal points in a sample were around 25 μm (1/e2). The femtosecond Raman probe beam (white light) was generated by focusing a small portion of the laser pulse in a sapphire plate; the probe’s energy was 11 nJ. The measurements were performed with parallel polarization between the Raman pump and Raman probe. Samples were measured in a 1 mm sample path fused silica cuvette (Hellma QX quartz) at room temperature. The Raman pump and the probe were temporarily overlapped with the use of a manual optical delay line. The Raman probe spectrum with and without the presence of the Raman beam (Thorlabs chopper blocks for every second Raman pump pulse) in the sample was recorded by the spectrometer (spectrograph Andor Shamrock SR 500i with CCD camera Andor Newton U971N).

Results and Discussion

Two types of an aqueous dispersion of silver nanoparticles, AgNPs, with different average sizes named according to the colors of their dispersions: “AgNPs blue” (34 ± 14 nm) and “AgNPs yellow” (16 ± 7 nm) were studied in this work. Figure shows the spontaneous Raman spectra of pure water and water in AgNPs blue and AgNPs yellow samples in the range of the OH stretching modes (Raman pump 514.5 nm). Since AgNPs dispersions have nonzero absorption in the spectral ranges of both the Raman pump (514.5 nm) and Raman scattered light (roughly 610–635 nm range; for absorption spectra of the samples see Figure a), the spectra in Figure are baseline-corrected and corrected for losses according to the procedure described in the Supporting Information (SI), Part SI 1.1. The raw, spontaneous Raman spectra are shown in Figure S1, and their processing is presented in the Supporting Information figures. The pronounced baseline in the raw Raman spectra of AgNPs originates from, as predicted by theory, resonance surface plasmon emission.[18] It is clear from the corrected spectra that the Raman signal of water in AgNPs blue is uniformly amplified with respect to the pure water. The signal of water in AgNPs yellow has, however, nearly the same intensity as that of pure water. The absorption by metal nanoparticles at the Raman pump wavelength is essential to the presence of the nanoantenna effect, which is the enhancement of the Raman signal resulting from the enhancement of the local electromagnetic field in the vicinity of the plasmonic surface.[2] The absorbance of AgNPs blue is around 3 times higher at 515 nm than that for AgNPs yellow, which is sufficient to justify significantly stronger Raman signal enhancement for the AgNPs blue.

Figure 2

Spontaneous Raman spectra of AgNPs blue (blue line) and AgNPs yellow (red line) compared to the spectrum of pure H2O (black line) with a 514.5 nm Raman pump; spectra were baseline-subtracted and corrected for the absorption of the Raman pump and Raman signal (SI 1.1 in the Supporting Information).

Figure 3

(a) Absorption spectra of pure water (black solid line) and dispersion of AgNPs blue (blue solid line) and AgNPs yellow (red solid line) with marked wavelengths of Raman pumps (green, dark red, and pink arrows for 515, 715, and 755 nm, respectively) used in this work together with spectral ranges of Raman scattered light (green bar for the 515 nm Stokes shift and dark-red and pink bars for the 715 nm anti-Stokes shift and the 755 nm anti-Stokes shift, respectively). The absorption curve for pure H2O is magnified 100-fold. (b) Stimulated Raman spectra of pure water (black lines), AgNPs blue (blue lines), and AgNPs yellow (red lines) in the OH stretching range obtained with the 515 nm pump (long dashed lines, Stokes shift measured), 715 nm pump (short dashed lines, anti-Stokes shift), and 755 nm pump (solid lines, anti-Stokes shift). (Inset) Close-up image of the stimulated spectra obtained with the 515 nm Raman pump for the SE-FSRS spectra from Figure b represented against the Raman shift (cm–1) on the x axis; see Figure S9. (c) SE amplification for AgNPs blue and AgNPs yellow, calculated as the ratio of the maximum intensity of the OH stretching Raman peak of water with and without silver nanoparticles as a function of the wavelength of the Raman pump in the FSRS experiment. Lines are drawn to guide the eye.

The standard SERS enhancement factor for a given substrate is expressed by the formula[19]where ISERS and IRS denote the intensities of the SERS signal and the normal (non-SERS) Raman scattering signal, respectively, and Nsurf and Nvol are average number of molecules in a scattering volume under SERS and normal Raman scattering conditions. To correctly calculate the EF in the presented case, it should be noticed that water plays the role of the analyte and of the solvent simultaneously, and thus in the EF calculation only the water molecules which are in contact with AgNP surfaces should be considered. The maximal number of water molecules interacting directly with AgNP surfaces was estimated on the basis of the shape and average size of particles in a AgNPs blue dispersion determined by TEM (Figure ) and the Ag concentration in the dispersion (17.96 μg/mL, determined by flame atomic absorption spectrometry). We estimate that, at most, 1 water molecule per 10 million is in contact with AgNP surfaces. However, it should be stressed that only the first layer of water hydrating the nanoparticles is assumed to be involved in the enhancement of the Raman signal in this estimation. In reality, the plasmonic field extends further than a monolayer. Nevertheless, according to computational work, the plasmonic field is located mostly at the tips of silver nanoprisms (such as in AgNPs blue) and decays quickly with distance from a nanoparticle.[20] Moreover, the orientation of an analyte with respect to the metallic surface likely influences the magnitude of the enhancement.[21] Henceforth, only the limited number of molecules from each layer that hydrates a silver nanoparticle may contribute to the plasmonic enhancement of the Raman signal. Our assumption should therefore compensate for some contribution from other hydration layers. Taking this into account, the EF for AgNPs blue is (4.8 ± 0.8) × 106. (For details of the calculations, see section SI3 of the SI.) The EF’s error value is calculated for the smallest and the largest particle sizes of studied AgNPs from the size distribution.

We should note, however, that in the spectral range of the enhanced Raman signal (610–635 nm) the absorption of AgNPs blue is at least 16 times stronger than that of AgNPs yellow. Henceforth, such conditions allow for the radiation enhancement mechanism of SERS (modification of the radiation of Raman light by the oscillating dipole due to the presence of the metallic surface in the dipole environment).[2] Furthermore, the nonzero absorbance of water at wavelengths longer than approximately 600 nm may lead to the additional resonant amplification of the Raman signal, which was already observed by Pastorczak et al.[22] To test which is the dominant mechanism of the Raman signal enhancement in AgNPs blue, we performed a similar experiment with the use of our femtosecond stimulated Raman spectroscopy setup which enables (i) almost free choice of the wavelength of the Raman pump, (ii) measurements of both Stokes and anti-Stokes scattering, and (iii) better control of the experimental parameters (pulse energies and beam geometry) due to the coherent nature of the stimulated Raman process. The determination of these parameters allows us to calculate gain factor “g” for the stimulated Raman enhancement. Our other motivation for this study was to observe the multiplicative SE-FSRS enhancement of the water signal.

Stimulated Raman spectra were measured in the following configurations: a 515 nm Raman pump with the Stokes probe, a 715 nm Raman pump with the anti-Stokes probe, and a 755 nm Raman pump with the anti-Stokes probe. These configurations allow us to test how the wavelengths of the Raman pump and the Raman probe influence the intensity of the SE-FSRS signal with respect to absorption peaks of AgNPs and water. The wavelengths of Raman pumps and probes with respect to the absorption spectra of AgNPs blue, AgNPs yellow, and pure water are shown in Figure a. Configurations with near-infrared Raman pumps (715 and 755 nm) and the Stokes probe were not tested since the OH stretching mode for such pumps would be Stokes shifted to around 1000 nm, where the sensitivity of our detector (CCD camera) falls to below 10% of its maximal value. It must be emphasized here that measurement on the anti-Stokes side in SRS also involves a 0 → 1 transition. The only difference is that in the stimulated Stokes process (also known as stimulated Raman gain) some of a pump photons are converted to Raman photons, and in the stimulated anti-Stokes process (also known as stimulated Raman loss), the opposite occurs.[23]

We observe some surface enhancement of the OH stretching mode of water in AgNPs blue in raw FSRS spectra (Figure S5) as well as in the spectra normalized to the fused silica band (which is located at around 490 cm–1 and originates from the front window of a cuvette; Figures S6 and S7). However, to be able to directly compare Raman intensities obtained with different Raman pumps, the corrections for parameters such as the absorbance of the Raman pump and probe by a sample, pulse lengths, and pulse energy need to be made. As a result of these operations, the FSRS spectra may be expressed as an “absolute” stimulated Raman gain, g, which in the configuration used in the experiment iswhere αP describes linear losses of a Raman pump, IP and IS are the intensities of the Raman pump and the Raman signal, respectively, superscript 0 indicates the intensity of the signal measured without the pump beam (Raman probe intensity), and z ∈ [0, L] is the position within the sample of thickness L. The ± signs correspond to Stokes and anti-Stokes scattering. For the αP = 0 case, eq can be simplified toThe derivations of eqs and 3 are presented in SI2.

The FSRS spectra expressed as Raman g factor versus wavelength for all measured configurations of the Raman pump and probe are compared in Figure b. The stimulated Raman gain spectra with the 515 nm Raman pump–Stokes probe are qualitatively similar to the results obtained with spontaneous Raman using a 514.5 nm pump. We observe a significant surface enhancement of the signal for AgNPs blue and only a slight enhancement for AgNPs yellow. The data shown in Figure confirms that the SERS of water molecules may be observed in the SE-FSRS experiment. The Raman pump at 715 nm is almost at the absorption maximum of AgNPs blue and at the growing wing of the water absorption peak at around 755 nm. The absorption of AgNPs yellow is around zero in that spectral range. It is thus understandable that the Raman gain for all samples studied is higher with the 715 nm Raman pump than with the 515 nm Raman pump, which should be attributed to water preresonance conditions. Surface enhancement of the AgNPs blue is similar to that with the pump at 515 nm, while there is no SE for AgNPs yellow. It is interesting that in the raw spectra (Figure S5), surface enhancement in AgNPs blue for the 715 nm Raman pump seems to be much larger than for the 515 nm Raman pump. This observation indicates how important it is to use absolute Raman gains in the assessment of the magnitude of the SERS effect. The estimated enhancement factors (with respect to the FSRS spectrum of pure water) were (3.7 ± 0.6) × 106 (for the 515 nm Raman pump) and (3.6 ± 0.6) × 106 (at both 715 and 755 nm).

Remarkably, we observe SE in AgNPs blue only for the OH stretching mode, not for the H–O–H bending mode or the librational modes of water. This finding draws our attention again to the water absorption mode at around 755 nm, attributed to the combinational overtone of symmetric and antisymmetric OH stretching modes aν1 + bν3, where a + b = 4.[24] One possible explanation for our experimental result is that the energy of the Raman pump absorbed by the AgNPs blue could be transferred through water combinational overtone aν1 + bν3 to the fundamental stretching modes, resulting in the amplification of their Raman signal. To test this hypothesis, we performed the SRS experiment with the Raman pump at 755 nm (i.e., at the maximum in the water absorption peak, which also overlaps with the shoulder of absorption of AgNPs blue). This should constitute conditions close to the optimal for the energy transfer from the Raman pump through metal nanoparticles to water stretching modes. As is clear from Figure b, the Raman gain of water stretching mode is, in this case, a few times stronger than in the experiments with the Raman pump at 715 or 515 nm. The magnitude of the surface enhancement of the signal from AgNPs blue compared to neat water is, however, similar to these in the two earlier experiments. This result justifies two important conclusions: (i) a resonance (755 nm) or preresonance (715 nm) condition for the Raman pump with the aν1 + bν3 mode of water strongly (approximately 10 times compared to that of the 515 nm Raman pump) enhancing fundamental OH stretching modes of water and (ii) Raman resonance and plasmon resonance enhancement factors are not multiplicative. We may compare the magnitudes of the SE effect between spontaneous Raman and SRS experiments by merely calculating the ratios of the maxima of the intensity of OH stretching modes of water (in the spectra corrected for absorption losses) with and without the silver nanoparticles. Surface enhancement is the strongest in the spontaneous Raman experiment at around 50% of the amplification in AgNPs blue, and in the same sample in the SE-FSRS experiment, it varies between 12 and 16% (Figure c). This difference may come from the dielectric breakdown of water molecules at the AgNP surface in the SE-FSRS experiment, as was postulated by Keller et al.[9] One may also notice some differences among the shapes of water OH stretching modes obtained with 515, 715, and 755 nm Raman pumps. One possible explanation for this effect is a Raman resonance of some components of the OH stretching multimode band with overtones of vibrational modes located in the red to the near-infrared range. (See the water absorption in Figure a.) Such resonance has already been observed in one of our previous papers: the resonance of the component at around 3200 cm–1 with the light from the red range was observed.[22] This effect, however, demands further studies with other wavelengths of the Raman pump and with probing Raman signals on both the Stokes and anti-Stokes sides for each pump.

Conclusions

We observed both spontaneous and stimulated Raman surface enhancement of the liquid water signal in an aqueous dispersion of silver nanoparticles for the first time. We think that the observation of the SE-FSRS signal was possible due to the ultrafast resonant transfer of vibrational energy in liquid water,[16,25] which prevented the accumulation of energy in water molecules at the metal surface and to some extent also protected the water analyte from the dielectric breakdown. However, the dielectric breakdown of an analyte was a likely reason for somewhat lower magnitudes of SE in SE-FSRS (12–16% relative enhancement) as compared to the SERS (around 50% relative enhancement) experiment. The measured enhancement factors for SE-FSRS experiments were (3.7 ± 0.6) × 106 (515 nm), (3.6 ± 0.6) × 106 (715 nm), and (3.6 ± 0.6) × 106 (755 nm), while for SERS it was (4.8 ± 0.8) × 106 (514.5 nm). These values are 1 order of magnitude lower than the maximum enhanced factor obtained very recently by Shin et al. (EF < 5.42 × 107) for water in a nanomeniscus in a silver solid substrate.[14] We tested the SE-FSRS amplification of the water signal in aqueous silver nanoparticle dispersions in several experimental configurations (different wavelengths of the Raman pump and Raman probe with respect to absorption bands of AgNPs and water). We found that if the Raman pump is within the plasmon resonance of the metal nanoparticles then the configuration has a minor effect on the EF. However, the resonance (755 nm) or preresonance (715 nm) condition of the Raman pump with the aν1 + bν3 vibrational mode of water strongly (approximately 10 times as compared to that of the 515 nm Raman pump) enhanced the Raman signal of the fundamental OH stretching mode of water. Given the mere 0.014/cm absorption of water at 755 nm in comparison to the absorption of the main water band (4600/cm at 3000 nm), the magnitude of this resonance enhancement is impressive.

This work opens up exciting perspectives for studies of the vibrational relaxations and dynamics of water molecules adsorbed at the surfaces of metal nanoparticles with the use of time-resolved femtosecond stimulated Raman spectroscopy. The surface-enhanced FSRS studies of water would be particularly interesting when we take into account the possible dielectric breakdown of water molecules. The study could provide original information on the kinetics and products of such a breakdown of water at irradiated metal surfaces.