Vibrational Sum Frequency Generation Spectroscopy of

Vibrational Sum Frequency Generation Spectroscopy of Dodecanethiol on Metal ... Holden T. Smith , Tony E. Karam , Louis H. Haber , and Kenneth Lopata ...
3 downloads 0 Views 855KB Size
Download PDF
J. Phys. Chem. C 2007, 111, 8925-8933

8925

Vibrational Sum Frequency Generation Spectroscopy of Dodecanethiol on Metal Nanoparticles† Andrey N. Bordenyuk, Champika Weeraman, Achani Yatawara, Himali D. Jayathilake, Igor Stiopkin, Yi Liu, and Alexander V. Benderskii* Department of Chemistry, Wayne State UniVersity, Detroit, Michigan 48202 ReceiVed: December 31, 2006; In Final Form: March 15, 2007

We report studies of metal nanoparticle surfaces using vibrational sum frequency generation (SFG) spectroscopy. SFG is demonstrated as a sensitive and information-rich probe of nanostructured surfaces. Detection of SFG spectra from a few percent of a monolayer of nanoparticles on a transparent substrate was achieved, corresponding to as few as 105 isolated particles within the laser beam spot. A new effect arises when the nanoparticle size approaches the molecular scale: the dependence of the molecular conformation on the geometry of the substrate. Conformation of the dodecane chain of the ligand shows systematic variation with the particle diameter in the 1.8-25 nm range. More gauche defects are observed on smaller particles, judged from the relative intensities of the CH2 and CH3 stretch transitions in SFG spectra. Similar behavior observed for both gold and silver particles suggests a nanoscale geometric packing effect, whence more volume is available to the chain on a curved surface than a flat surface. Drying-mediated aggregation of gold nanoparticles results in an enhancement of the SFG signal, enabling vibrational SFG spectra of single submicron size aggregates to be obtained. These are different from spectra of isolated particles, indicating a vibrational mode-selective enhancement mechanism.

1. Introduction A common motif in many emerging nanotechnology applications involves nanostructures covered with chemi- or physisorbed organic molecules which perform the desired physical, chemical, or biological function. These include chemical and biosensors,1,2 heterogeneous catalysis,3 and hybrid organicinorganic photovoltaic devices.4 As the characteristic size of the nanostructures approaches the molecular scale, a clear need is emerging for characterization techniques providing molecularlevel information on the adsorbates. Vibrational sum frequency generation (SFG) spectroscopy has been extensively developed in the past few years5-7 due to its extreme surface selectivity and the sensitivity to the molecular conformation.8-12 SFG has been previously applied to study surfaces of metal nanoparticles13,14 and regular nanoarrays15 in the 10-100 nm size range and other nanostructured surfaces.16-18 SFG scattering from bulk suspensions of nanoparticles (liposomes in the 100 nm range) has also been observed.19,20 In this paper, we present application of SFG to characterize the vibrational spectroscopy at metal nanoparticle surfaces approaching the molecular size, 1-2 nm. We demonstrate that SFG is both a sensitive probe of nanostructured surfaces and is capable of providing a wealth of molecular-level information. In particular, the second-order nonlinear SFG technique is governed by different selection rules (or, rather, propensity rules) than the linear techniques such as FTIR and spontaneous Raman, making it more sensitive to the changes in molecular conformation.8,9,21-25 We demonstrate that a new effect arises when the nanoparticles approach the molecular size limit, namely the dependence of the molecular conformation on the geometry of the nanostructured substrate (in this instance, on the average size of the †

Part of the special issue “Kenneth B. Eisenthal Festschrift”. * Corresponding author. E-mail: [email protected].

approximately spherical particles). A preliminary report for gold nanoparticles has appeared recently.26 Here, we present the complete results on both gold and silver nanoparticles in the 25-1.5 nm size range, as well as the analysis of the molecular orientation. Similar size-dependences observed on gold and silver particles suggest that the observed relationship between the molecular conformation and the nanoscale geometry may be a universal geometric effect. Enhancement of spectroscopic signals for molecular on metal nanostructures has attracted much interest, in particular in connection with lowering the detection limit down to the singlemolecule regime.1,27 Single-molecule detection was demonstrated even for such inefficient process as spontaneous Raman scattering, when full advantage is taken of the local field enhancement on a metallic nanostructure of proper geometry.27 Enhancement of the nonlinear SFG signals up to 4 orders of magnitude has been observed on regular arrays of isolated platinum nanoparticles.15 However, the largest enhancement in Raman spectroscopy (up to the single-molecule detection limit) has been achieved not for isolated particles but rather for aggregates where the field is strongly amplified in the small gaps between metal nanoparticles.27 Here we observe nonlinear spectroscopic SFG signals of gold nanoparticle clusters, which are significantly enhanced relative to the isolated nanoparticle signals and also exhibit different spectra, pointing to a modeselective enhancement mechanism. 2. Experimental Section 2.1. Sample Preparation. Dodecanethiol-capped gold and silver nanoparticles of four different sizes in the 1-25 nm size range were purchased from Meliorum Technologies. Four different size samples for gold were (1) mean diameter 〈d〉 ) 1.8 nm with standard deviation σd ) 1.3 nm; (2) 〈d〉 ) 2.9 nm

10.1021/jp069062n CCC: $37.00 © 2007 American Chemical Society Published on Web 05/16/2007

8926 J. Phys. Chem. C, Vol. 111, No. 25, 2007 σd ) 0.6 nm; (3) 〈d〉 ) 7.4 nm σd ) 1.1 nm; and (4) 〈d〉 ) 23 nm σd ) 8.1 nm, dissolved in toluene. Comparable silver nanoparticle samples are (1) 〈d〉 ) 1.8 nm σd ) 0.4 nm; (2) 〈d〉 ) 3.6 nm σd ) 0.8 nm; (3) 〈d〉 ) 7.9 nm σd ) 1.9 nm; and (4) 〈d〉 ) 24.6 nm σd ) 5.6 nm, dissolved in benzene. Spectroscopic studies were done on homogeneously dispersed submonolayer nanoparticle samples on an optical grade CaF2 window. Samples were deposited by dropping 10 µL of nanoparticle solution onto the CaF2 substrate (approximate area 2 cm2 for gold samples and 1 cm2 for silver samples) and allowing the solvent to evaporate by drying under vacuum for 15 h. For gold nanoparticle samples, the toluene solution concentrations were 3.40 × 1014 particles/mL (0.02 mg/mL) for 〈d〉 ) 1.8 nm sample, 3.40 × 1014 particles/mL (0.08 mg/ mL) for 〈d〉 ) 2.9 nm sample, 2.44 × 1013 particles/mL (0.1 mg/mL) for 〈d〉 ) 7.4 nm sample, and 1.24 × 1012 particles/ mL (0.1 mg/mL) for 〈d〉 ) 23 nm sample. Therefore, the expected surface coverage is 51.8, 51.8, 721, and 14 200 nm2/ particle for 1.8, 2.9, 7.4, and 23 nm samples, respectively. This corresponds to 5%, 14%, 7%, and 3% of a closed-packed (hexagonal) monolayer for the four particle sizes. For silver nanoparticle samples, the benzene solution concentrations were 3.12 × 1014 particles/mL (0.01 mg/mL) for the 〈d〉 ) 1.8 nm sample, 3.90 × 1013 particles/mL (0.01 mg/mL) for the 〈d〉 ) 3.6 nm sample, 3.70 × 1012 particles/mL (0.01 mg/mL) for the 〈d〉 ) 7.9 nm sample, and 1.20 × 1011 particles/mL (0.01 mg/ mL) for the 〈d〉 ) 24.6 nm sample. Therefore, the expected surface coverage of the silver nanoparticles is 25.2, 201, 2130, and 64 300 nm2/particle for 1.8, 3.6, 7.9, and 24.6 nm samples, respectively. This corresponds to 11%, 5%, 3%, and 1% of a closed-packed (hexagonal) monolayer for the four particle sizes. 2.2. Particle Size Distribution. Transmission electron microscopy (TEM) was performed to characterize the particle size distribution (to confirm the data provided by the supplier) and the clustering on the surface following drying of the spreading solvent. The TEM samples were prepared by drop-casting the nanoparticles onto a carbon coated copper grid (mesh 200) under the same conditions as in preparation of spectroscopic samples (same concentration of drop solution and evaporation conditions). TEM images were recorded by using JELO FasTEM 2010 system (1.4 Å resolution) operating at 200kV. The representative TEM images of the four gold nanoparticle samples are shown in Figure 1, along with the size histograms obtained from analyzing multiple TEM images to achieve appropriate statistics. The mean gold particle diameters and standard deviations were determined to be (a) 〈d〉 ) 2.2 nm with standard deviation σd ) 0.71 nm; (b) 〈d〉 ) 3.5 nm σd ) 0.92 nm; (c) 〈d〉 ) 7.3 nm σd ) 1.4 nm; and (d) 〈d〉 ) 21.1 nm σd ) 4.7 nm. Figure 2 shows the TEM images and the size histrograms for the four silver nanoparticle samples, whose mean diameters and standard deviations were determined to be (a) 〈d〉 ) 2.3 nm with standard deviation σd ) 0.39 nm; (b) 〈d〉 ) 3.5 nm, σd ) 0.50 nm; (c) 〈d〉 ) 7.90 nm, σd ) 1.9 nm; and (d) 〈d〉 ) 24.1 nm, σd ) 3.2 nm. Slight deviations from the manufacturer-specified particle size for the small diameter samples (both gold and silver) may be due to the insufficient resolution of our TEM instrument, which is consistent with the apparently narrower size distribution for the 2-3 nm samples. The average shape of both gold and silver particles is approximately spherical. Slight deviations from spherical shape, while they clearly exist, were not quantified in this study. 2.3. Surface Aggregation. Depending on the concentration of the drop-casting solution, the nanoparticles show a tendency to aggregate and form clusters on the substrate upon drying of

Bordenyuk et al.

Figure 1. TEM images and size distribution histograms of gold nanoparticles used in this study.

the spreading solvent.28,29 For the size-dependent spectroscopic studies, this undesirable side-effect was avoided by lowering the concentration of the precursor solution. The cluster formation was evident in the TEM images of samples prepared at high concentration (the stock solution concentration C0 ) 3.40 × 1015 particles/mL for 〈d〉 ) 1.8 nm to C0 ) 1.2 × 1013 particles/ mL for 〈d〉 ) 23 nm for gold and C0 ) 3.12 × 1016 particles/ mL to 1.2 × 1013 particles/mL for silver nanoparticles), depending on particle size. This aggregation was drastically reduced at lower concentrations (C0/10 to C0/100). The TEM images presented in Figures 1 and 2 were obtained using the lowest concentration (C0/100), and these samples were used for the size-dependent SFG spectroscopic studies.

Dodecanethiol on Metal Nanoparticles

J. Phys. Chem. C, Vol. 111, No. 25, 2007 8927

Figure 3. SEM images of aggregates of gold particles (〈d〉 ) 23 nm) on CaF2 substrate prepared using different drop-casting solution concentrations: (A) 2C0 ) 2.4 × 1013 particles/mL and (B) C0/10 (two aggregates barely visible in the latter image are indicated by arrows). (C) Cluster size distribution histogram for 〈d〉 ) 23 nm for the two different concentrations of the drop-casting solution, 2C0 (red) and C0/ 10 (blue bars, dashed): vertical axis - number of clusters per 100 µm × 100 µm; horizontal axis: cluster diameter.

Figure 2. TEM images and size distribution histograms of silver nanoparticles used in this study.

Scanning electron microscopy (SEM) was performed using a Hitachi S-2400 instrument to quantify the extent of clustering on the sub-micron scale. The representative SEM images of gold 〈d〉 ) 23 nm nanoparticle samples prepared using 2C0 and C0/ 10 solutions are shown in Figure 3A,B. The distribution of cluster sizes found in multiple SEM images is plotted in Figure 3C. The vertical axis is the number of clusters per 100 µm × 100 µm area, which is similar in the order of magnitude to the area covered by the laser beams in the SFG experiments (∼50 × 100 µm2, vide infra). The SEM images show drastic reduction in the number of clusters upon lowering of the concentration by a factor of 20. For the samples prepared at concentration below C0/10, less than one cluster is found in the laser beam spot (upon careful inspection, image in Figure 3B shows 3 sub-micron sized aggregates ∼200 µm apart).

This information was used to observe SFG signals from single clusters by scanning the sample laterally under the laser beams, as described in the Results section. 2.4. Spectroscopy. A detail description of our sum frequency generation spectroscopy (SFG) set up has been published elsewhere.24,30,31 Briefly, the vibrational sum frequency generation (VSFG) spectroscopy setup is based on a high power amplified femtosecond Ti-sapphire laser system (Spectra Physics Spitfire sub-50 fs HP). 50% of the 2 mJ fundamental output pulse (800 nm, fwhm 35 fs) is used to pump an Optical Parametric Amplifier (OPA) followed by the signal-idler difference frequency mixing in a 0.5 mm thick AgGaS2 crystal producing 75 fs IR pulses (300 cm-1 spectral fwhm) centered at 2900 cm-1. Spectra of the IR pulses were measured using an IR grating (blazed at 5 µm) in the monochromator and liquidnitrogen-cooled MCT detector (IR Associates). The broad-band VSFG scheme32,33 was employed that uses spectrally broad IR and narrow visible pulses to obtain the spectrum by frequencydispersing the SFG signal. A zero-chirp 4-f design pulse stretcher34 consisting of a grating, a collimating lens, and a mirror equipped with a tunable slit was used to narrow the spectrum of the visible (800 nm) pulse from 430 cm-1 of the Spitfire output down to the desired spectral width. For gold nanoparticles, the full width at half-maximum (fwhm) of the spectral features of the nanoparticles samples was determined to be greater than 10 cm-1 (vide infra), so the visible pulse spectral width was selected to be e8 cm-1 Gaussian fwhm. For silver nanoparticles, the linewidths are typically broader, in the 15-20 cm-1 range, so a high-finesse Etalon (TecOptics) with 14 cm-1 Gaussian fwhm (38 nm FSR, centered at 809.3 nm) was used to narrow the visible pulse spectrum. The laser beams were focused and overlapped on the sample surface (f.l. 20 cm for the IR beam and 40 cm for the visible beam) with the incidence angle 65° from surface normal. The beam diameters at the sample are ∼100 µm for the visible and ∼100 µm for the IR beams. The laser power at the sample is typically 2-4 µJ/

8928 J. Phys. Chem. C, Vol. 111, No. 25, 2007

Bordenyuk et al.

Figure 5. Amplitude ratios of the CH2 and CH3 stretch modes (symmetric d+/r+ and asymmetric d-/r-) for dodecanethiol on gold nanoparticles, extracted from the (A) SSP and (B) PPP SFG spectra, as a function of particle size. Figure 4. SFG spectra for (A) SSP and (B) PPP polarization (SFGvisible-IR) of dodecanethiol on gold nanoparticles of 4 sizes (average diameters 〈d〉 ) 1.8, 2.9, 7.4, and 23 nm). The spectra are vertically offset for clarity, the baseline levels are 0, 0.82; 1.80, and 2.8 for the four SSP spectra; 0, 0.45; 0.95, and 1.5 for the four PPP spectra. Solid lines show fits to eq 1.

pulse for IR and up to 17-18 µJ/pulse for the visible at 1 kHz repetition rate. The SFG signal reflected from the sample surface into the phase-matched direction was recollimated using a 10 cm lens, spatially and frequency filtered, passed through a 300 mm monochromator (Acton Spectra-Pro 300i), and detected using a LN2-cooled CCD (Princeton Instruments Spec-10:100B, 100 × 1340 pixels). Five pixel binning along the horizontal (wavelength) axis was performed to reduce noise. With this binning, the spectrometer resolution is 2.7 cm-1. The SFG spectrum was obtained by vertical integration of the CCD image over the region of interest. The signals were averaged over many laser shots by using the CCD as the integrator, with acquisition time typically 20-60 min per spectrum. The background spectrum for each spectrum measurement was recorded by blocking IR beam on to the sample for the same acquisition time. The background correction was performed by subtracting from the signal region of interest (e.g., horizontal strips 4045), a nonilluminated region of background spectrum of the same size. Spikes due to the cosmic X-rays were eliminated using a differential discriminator program. The IR frequencies are calculated by subtracting the central frequency of the narrow visible pulse (measured using the same monochromator) from the SFG frequency. In addition, we calibrate the IR frequency scale using a known SFG surface spectrum of dimethyl sulfoxide (DMSO). We estimate our IR frequency calibration accuracy to be (2 cm-1. Polarization of the visible beam was controlled by using a zero-order half-wave plate, whereas the IR beam polarization was vertical (P) in this study. FTIR absorption spectra of the nanoparticle samples were recorded at 1 cm-1 resolution using Digilab Excalibur Series FTS 3000MX spectrometer. Gold nanoparticle samples were prepared by drop-depositing gold nanoparticle solution onto an optically graded CaF2 window as same as for VSFG measurements but were considerably thicker (multilayer) due to its inferior detection limit compared with SFG. Silver nanoparticles samples for FTIR measurements were prepared by forming thin liquid layer of samples between two CaF2 windows since solvent benzene does not have C-H transitions in the 2800 - 3000 cm-1 region. 3. Results 3.1. SFG Spectroscopy of Isolated Particles. Vibrational SFG spectra of dodecanethiol covered gold nanoparticles in the

CH-stretch region are shown in Figure 4 for SSP and PPP polarization combinations (SFG-vis-IR) as a function of particle size. The samples were prepared at the lowest concentration of the precursor solution, same as in TEM images in Figure 1, so that the spectra correspond to isolated nanoparticles homogeneously distributed on CaF2 substrate. The observed bands can be assigned to the vibrations of the alkane chain of the dodecanethiol ligand: CH2 symmetric stretch (d+) at 2855 cm-1; symmetric stretch of the terminal CH3-group (r+) at 2880 cm-1, CH2 antisymmetric stretch (d-) at 2915 cm-1, Fermi resonance between CH3 symmetric stretch and bend overtone modes (r+FR) at 2935 cm-1, and CH3 asymmetric out-of-plane and in-plane stretches (r-op and r-ip) at 2950 and 2960 cm-1.35 An alternative assignment proposed recently for long-chain alcohols describes the 2915 cm-1 mode as the CH2 Fermi resonance (d+FR), and places d- at 2900 cm-1.25 Regardless of this uncertainty, a clear trend is observed: the relative intensity of the CH2 stretch modes increases with respect to the CH3 modes as the particle diameter decreases. Appearance of the CH2 stretch modes in SFG spectra is usually associated with the formation of gauche defects in an otherwise all-trans alkane chains.8,9,21,22 In contrast to the nanoparticle surfaces, where the CH2 modes become dominant features for the smaller sizes, SFG spectra of dodecanethiol SAM on plane gold show CH2 transitions that are at least 10 times weaker than the dominant CH3 stretch bands (r+, r+FR, and r-).23,36 Our measurements therefore can be qualitatively interpreted as increase in the fraction of the gauche defects of the ligand with decreasing particle diameter. In order to quantify the observed trend, we fit all measured vibrational SFG spectra using the standard multi-Lorentzian approximation which coherently adds Lorentzian line shapes

| | N BjΓj |2 | iφ | ISFG(ωIR) ∝ |χ (ωIR)| ) |ANRe + | | j ) 1(ωIR - ωj) + iΓj| | (1) (2)

2

∑

for each vibrational mode j, described by Bloch-type (exponential) dephasing with amplitude Bj, line width Γj, and transition frequency ωj. The first term accounts for the nonresonant part of the response with amplitude ANR and phase φ with respect to the vibrationally resonant part. The fitting parameters can be found in the Supporting Information. Because the macroscopic susceptibility χ(2) is proportional to the number of chromophore groups at the surface, we may consider the ratio of amplitudes Bj for the methylene and methyl stretch modes as a measure of the fraction of gauche defects in the dodecanethiol chain. Figure 5 shows the amplitude ratios for the

Dodecanethiol on Metal Nanoparticles

Figure 6. SFG spectra (A) SSP and (B) PPP polarization (SFGvisible-IR) of dodecanethiol on silver nanoparticles of 4 sizes (average diameters 〈d〉 ) 1.8, 3.6, 7.9, and 24.6 nm). The spectra are vertically offset for clarity, the baseline levels are 0, 1.0, 2.0, and 3.0 for the four spectra. Solid lines show fits to eq 1.

Figure 7. Amplitude ratios of the CH2 and CH3 stretch modes (symmetric d+/r+ and asymmetric d-/r-) for dodecanethiol on silver nanoparticles, extracted from the (A) SSP and (B) PPP SFG spectra, as a function of particle size.

symmetric B(d+)/B(r+) and asymmetric modes B(d-)/B(r-) as a function of gold particle size for SSP and PPP polarizations. SFG spectra of dodecanethiol on silver nanoparticles of four sizes are shown in Figure 6 for SSP and PPP polarization combinations. Although the dodecanethiol vibrational transitions are in general broader on silver nanoparticles, a similar trend is observed in the relative intensities of the CH2 and CH3 stretch modes as a function of average particle diameter. The spectra were fitted to eq 1 (the fitting parameters can be found in the Supporting Information), and the amplitude ratios for the symmetric B(d+)/B(r+) and asymmetric modes B(d-)/B(r-) in the SSP and PPP spectra are shown in Figure 7 as a function of particle size. Lorentzian line widths Γj for the dominant d+ and r+ modes of dodecanethiol on gold nanoparticles are typically in the 812 cm-1 range, whereas on silver they are in the 15-23 cm-1 range. This qualitatively indicates the less-ordered nature of the dodecanethiol monolayer on silver nanoparticles, which is consistent with a weaker thiol-silver bond compared to the thiol-gold bond resulting in a more robust self-assembly on gold surfaces compared to silver.37 We also note that, for nanoparticle samples, the overall intensities of the SSP and PPP SFG signals are of the same order of magnitude (within a factor of 2), whereas for SAMs on flat metal surfaces, the PPP signal is more intense by a factor of ∼10.36 This behavior is similar to that observed previously

J. Phys. Chem. C, Vol. 111, No. 25, 2007 8929

Figure 8. FTIR spectra of dodecanethiol on gold (A) and silver (B) nanoparticles, 4 sizes each: A1 〈d〉 ) 1.8 nm; A2 〈d〉 ) 2.9 nm; A3 〈d〉 ) 7.4 nm; A4 〈d〉 ) 23 nm; B1 〈d〉 ) 1.8 nm; B2 〈d〉 ) 3.6 nm; B3 〈d〉 ) 7.9 nm; B4 〈d〉 ) 24.6 nm. The spectra are vertically offset for clarity.

for CO on flat vs nanostructured Pt surfaces15 and can be understood in terms of the Maxwell-Garnett theory. The Fresnel factors on a flat metallic surface heavily favor the PPP signals, whereas the nanoparticles on a dielectric substrate at low coverage (e.g., 5% of monolayer on CaF2 substrate in this study) present a nearly dielectric surface which favors the SSP signal.15 3.2. FTIR Spectra. For comparison, we show the linear absorption FTIR spectra of the gold and silver nanoparticle samples in Figure 8. Although the same transitions are observed in the FTIR spectra, no systematic size dependence of the CH2 vs CH3 is observed, neither in the relative intensities nor in the peak frequencies within 2-3 cm-1. Similarly, spontaneous Raman spectra presented earlier for gold nanoparticles26 do not show systematic size dependence. This can be attributed to the different selection rules governing the nonlinear (SFG) vs linear (FTIR absorption or Raman scattering) spectroscopies, which make SFG more sensitive to the conformational state of the alkane chain. Conformational analyses of the linear IR and Raman spectra can be made sometimes using the vibrational frequency of the CH2 groups which blue-shifts by ∼4-6 cm-1 from “solid-like” (all-trans) to “liquid-like” (multiple gauche defects) alkane chains.11 Several previous FTIR studies on gold38,39 and copper40 nanoparticles reported CH2 frequencies indicative of all-trans chains, similar to those in SAMs on flat gold, thus suggesting a high degree of conformational order. As in our present study, no size-dependent spectral shift was observed.39 Our results indicate that SFG spectra are significantly more sensitive to the trans-gauche conformational changes and thus allow us to observe gauche defects not detectable by FTIR or Raman. Interpretation of the frequency shift measurements is difficult because the shift is significantly smaller than the line width for nanoparticles (15-20 cm-1 in FTIR spectra). Also, an additional frequency shift from the bulk alkane value due to the effect of the metallic nanoparticle surface may present a complicating factor. 3.3. SFG Signal Levels. The SFG signal intensity from the gold or silver nanoparticle sample at 3-5% surface coverage (relative to the close-packed hexagonal 2D monolayer of nanoparticles) on a nonresonant dielectric substrate is approximately 3-5 times weaker than from a saturated LangmuirBlodgett (LB) monolayer of heptadecanoic acid on a transparent dielectric substrate (IR-grade fused silica).24 These estimates are made using integration time on the CCD detector required to obtain the same signal level (e.g., 1000 counts per pixel at a given peak of the SFG spectrum) from different samples.

8930 J. Phys. Chem. C, Vol. 111, No. 25, 2007

Bordenyuk et al.

TABLE 1

sample/substrate

SFG intensity (a.u.)

% monolayer

SAM/gold LB/fused silica Au, Ag nanoparticles/CaF2

100 10 2-3

100% 100% 1-10%

orienta- enhancetional ment factor factor ∼1 ∼1 0.1

∼10 1 103

Because the SFG intensity scales as the square of the surface coverage, ISFG ∝ Ns2, we estimate an enhancement factor of at least 102 on metal nanoparticle surfaces. In addition, it is likely that the orientational distribution of the dodecanethiol on physisorbed nanoparticles is very broad (indeed, nearly spherically symmetric), which would result in extensive cancellation as described in the next section. When this is taken into account, the estimated enhancement factor is 103 or higher. We note that the all optical fields (IR at 3.4 µm, visible at 800 nm, and SFG at 650 nm) are fairly far away from the nanoplasmon resonance on isolated gold (∼520 nm) and silver (∼400 nm) particles. Another comparison can be made with the SFG signal levels from SAMs on plane gold substrates. The vibrationally resonant part of the SFG field from alkanethiol SAM is heterodyned against the strong nonresonant background from the gold surface, where the latter is ∼100 times stronger than the SFG signal intensity from the LB monolayer at the peak. Accordingly, the vibrationally resonant part of the SAM SFG signal is enhanced by a factor of ∼10 compared to the LB monolayer.36 The observed nanoparticle SFG spectra required 30-60 min integration, compared to ∼1 min for SAMs. This is consistent with the estimate above of the enhancement factor of at least 102 relative to alkane chains on nonresonant substrate, without taking into account the orientational factor. The signal level comparison is summarized in Table 1. From the known surface coverage (particles/cm2) and the beam size, we estimate that there are as few as ∼105 isolated metal nanoparticles (for the 23 nm gold and 24 nm silver samples) in the laser beam spot contributing to the spectra in Figures 4 and 6. 3.4. SFG from Aggregates. SFG spectra from aggregates of gold nanoparticles formed on the surface upon drying of the solvent for samples prepared using high concentration of the drop-casting solutions, such as those depicted in Figure 3A, are presented in Figure 9. Although the same dodecanethiol vibrational transitions are observed, their relative intensities are very different from those on the isolated nanoparticles or of dodecanethiol SAMs on flat surfaces. The SFG spectra in both SSP and PPP polarization combinations are dominated by the CH3 asymmetric out-of-plane stretch (r-op), with the CH3 Fermi resonance mode (r+FR) appearing as a shoulder. The average area of the aggregates of 23 nm gold nanoparticles, determined from the histogram in Figure 3C, is 0.35 µm2, so that an aggregate contains less than ∼104 nanoparticles assuming a spherical aggregate, and only ∼103 particles if it is a flat “pancake”. Using SEM images to quantify the surface coverage of the aggregates, we estimate that the SFG signal level per gold nanoparticle from aggregates is at least 10 times stronger than for isolated nanoparticles. These observations indicate additional enhancement of the SFG signal inside the aggregates and the mode-specific enhancement mechanism. SFG signal from individual aggregates was detected using the fact that, at low concentration of the drop-casting solution (C0/10 or below), there is less than one aggregate in the laser beam spot (vide supra, Figure 3B). By scanning the sample laterally under the laser beam, we observed a ∼3-5 fold increase in the SFG signal from certain spots on the sample, a few hundreds of microns apart from one another on average,

Figure 9. SFG spectra of aggregates of dodecanethiol-covered gold nanoparticles, 〈d〉 ) 23 nm (A, SSP and B, PPP), and 〈d〉 ) 1.8 nm (C, SSP and D, PPP). The spectra are vertically offset for clarity, the baseline levels are 0, 0.5, 1.0, and 1.5 for the four spectra. Solid lines show fits to eq 1.

which we interpret as the signal from single aggregates. Between these spots, the signal level was homogeneous, and the SFG spectra are those presented in Figure 4 for isolated nanoparticles. The SFG spectra from the enhanced spots are the same as the aggregate spectra shown in Figure 9. For the 23 nm gold nanoparticles prepared at low concentration, the laser beam spot covers ∼3 × 105 isolated nanoparticles, whereas a single aggregate contains less than ∼104 nanoparticles, and this observation is consistent with the estimate above of the additional enhancement upon aggregation by a factor of 10 or more. 4. Discussion The sum-frequency signal in these experiments is collected in the phase-matched direction expected for the flat surface. We therefore begin our analysis by considering the orientational distribution of the dodecanethiol with respect to the surface normal. Because the ligand is covering nearly spherical nanoparticles much smaller than the wavelength of light, a nearly isotropic distribution is expected, perturbed only by the presumably weak physisorption interactions of the dodecanethiol chains with the substrate (Figure 10). We calculate the expected nonlinear susceptibility of the CH3 and CH2 groups for a broad (but not isotropic) step-function orientational distribution of the tilt angle θ (Figure 10) assuming the other two Euler angles (azimuthal ψ and twist φ) to be isotropically distributed. Tilt angle θ is defined as the angle between the surface normal and the main symmetry axis of the molecular group (C3 axis for CH3 group and C2 axis for CH2 group). We assume that the CH2 transitions are SFG-allowed in the dipole approximation; that is, CH2 is at the gauche defect position. Details of the formulas can be found in a previous work.31 Briefly, we begin by assuming a symmetry-dictated relation between the nonvanishing elements of the molecular hyperpolarizability tensor β(2) lmn, usually rationalized in terms of the bond-additivity model.6,41-45 Calculations are performed for the CH3 and CH2 symmetric stretch modes r+ and d+, for which the assignment in the experimental SFG spectra is unambiguous and the β(2) lmn

Dodecanethiol on Metal Nanoparticles

J. Phys. Chem. C, Vol. 111, No. 25, 2007 8931 (2) XSSP ) Ly(ωSFG)Ly(ωvis)Lz(ωIR) sin(RIR) X(2) yyz

(4)

(2) X(2) PPP ) -Lx(ωSFG)Lx(ωvis)Lz(ωIR) cos(RSFG) cos(Rvis) sin(RIR) Xxxz (2) - Lx(ωSFG)Lz(ωvis)Lx(ωIR) cos(RSFG) sin(Rvis) cos(RIR) Xxzx (2) + Lz(ωSFG)Lx(ωvis)Lx(ωIR) sin(RSFG) cos(Rvis) cos(RIR) Xzxx

+ Lz(ωSFG)Lz(ωvis)Lz(ωIR) sin(RSFG) sin(Rvis) sin(RIR) X(2) zzz (5)

where Lj(ω) (j ) x,y,z) are the Fresnel factors representing the local fields at the surface for the SFG, visible, and IR beams and RSFG, Rvis, and RIR are the incidence and reflection angles of the SFG , visible, and IR beam with respect to the surface normal. The three layer model was used to calculate the Fresnel factors.46 Because of the low surface coverage of metallic nanoparticles ( 150°. We have calculated the amplitude ratios of the d+ and r+ symmetric stretch modes for the SSP and PPP polarization (2) (2) + combinations, BSSP(d+)/BSSP(r+) ) XSSP (d +)/XSSP (r ) and BPPP(2) (2) + + + + (d )/BPPP(r ) ) XPPP(d )/XPPP(r ), assuming equal number densities of the SFG-active CH2 and CH3 groups. The results are shown in Figure 12 and show nearly quantitative agreement with the experimental amplitude ratios (Figures 5 and 7) for smaller diameter particles (2-3 nm): BSSP(d+)/BSSP(r+) ≈ 1 and BPPP(d+)/BPPP(r+) ≈ 0.5. This could be interpreted as equal (within an order of magnitude) populations of the SFG-active CH2 and CH3 groups. Because SFG-active CH2 groups are usually associated with gauche defects of the alkane chain, our results suggest that a significant fraction of the chains (up to 50%) may have a single gauche defect on particles approaching the molecular size. We note that having two or more gauche defects per chain would likely result in isotropic distribution of the terminal CH3 groups and complete suppression of the r+ and r- transitions in SFG spectra. The size-dependent trend in the CH2 (gauche defect) vs CH3 transitions points to a correlation between the molecular conformation and the nanoscale geometry of the substrate. Although the differences clearly exist between gold and silver nanoparticle surfaces, similar trends observed on both gold and silver nanoparticles suggest that this is a universal geometric effect for the long alkane chains on the spherical (on-average) particles, rather than material-specific morphology of surface defects or details of chemical binding to the surface. The thiol surface-binding affinity is higher for gold than for silver; as a consequence, self-assembled monolayers (SAMs) on gold are

8932 J. Phys. Chem. C, Vol. 111, No. 25, 2007

Figure 11. Calculated SFG intensities (SSP polarization, normalized) of CH2 and CH3 symmetric stretch modes for broad step-function orientational distribution of dodecanethiol, as a function of the cutoff angle θC.

Figure 12. Calculated amplitude ratios of the CH2 to CH3 symmetric stretch modes B(d+)/B(r+), for (A) SSP and (B) PPP polarization combinations assuming equal populations of the SFG-active CH2 and CH3 groups.

usually better organized that on other metals.37 Observation of broader vibrational line shapes on silver compared to gold nanoparticles in this study is consistent with this notion. Nevertheless, the observed size dependence of the amplitude ratios of the CH2 vs CH3 transitions is very similar on Ag and Au nanoparticles. The observation can be rationalized in terms of the volume available to the dodecanethiol chain on a flat vs curved surface.26 Because of high thiol-gold and thiol-silver affinity,37 the thiol headgroups efficiently pack to fill the surface sites at ∼21.6 Å2/headgroup for flat SAMs.51-53 A recent highresolution STM study of homoligand (octanethiol) nanoparticles found only small variations (less than 10%) of the area per headgroup with the particle diameter.54 This is comparable to the cross-sectional area of the alkane chain (∼18-19 Å2/ chain),55 which explains why alkanethiol SAMs on flat surfaces have a nearly all-trans conformation, similar to LangmuirBlodgett films at high surface pressure, when the area per molecule approaches the cross-section of the chain.8,24 However, when the surface is curved on a size scale comparable to the alkane chain length (∼1.6 nm for dodecanethiol), the chain occupies a significantly larger volume. For example, conical volume per dodecanethiol adsorbed on a spherical 2.0 nm diameter nanoparticle is ∼2.5 times larger than the cylindrical volume per molecule on a flat surface. Because the free energy difference between the trans and gauche conformations is less than kBT (and the conformational isomerization barrier is few kBT), the chain is likely to explore both conformations given sufficient free volume. The orientational analysis above shows that, although the broad orientational distribution of dodecanethiol on spherical nanoparticles leads to extensive cancellation of SFG signals from

Bordenyuk et al. flat substrates, the signal does not vanish completely. The origin of the observed signals may therefore be the weak physisorption interaction of the alkane chains with CaF2 substrate which breaks the spherical symmetry. Observation of the SFG spectra from a few percent of the monolayer samples, together with the orientational averaging, implies significant enhancement of the nonlinear SFG process on metallic nanoparticles, by at least 3 orders of magnitude. The CH-stretch transitions of dodecanethiol are not coupled to any electronic transitions of the molecule/ metal system that are close to resonance with the laser fields (excited electronic states of the alkane chain are in the deep UV range; those of the thiol-metal group, in the near-UV, are not likely to be strongly coupled to the alkane chain vibrations). The enhancement mechanism therefore is likely to be electromagnetic in origin. The laser fields in the presented experiments are far off resonance with the nanoplasmon bands of isolated gold (∼520 nm) and silver (∼400 nm) particles. Therefore, even stronger enhancement may be expected by tuning the visible and/or SFG wavelength into local plasmon resonance, similar to the electromagnetic enhancement mechanism in Raman scattering.1,27 Further studies of the SFG enhancement mechanism, either involving tunable visible wavelength or nanostructures with tunable plasmon resonance frequency,56 are clearly necessary to address these questions. Further enhancement by more than 1 order of magnitude is observed in aggregates of gold nanoparticles. This situation is reminiscent of the Raman scattering of molecules trapped in narrow gaps between metal nanoparticles, where strong amplification of the local EM fields may occur, enabling singlemolecule detection.27 Moreover, the electromagnetic coupling between the particles associated with aggregation may cause in a significant red-shift of the collective plasmon resonance,57-59 resulting in a better spectral overlap with the laser wavelengths used in this study. We note that SFG, like spontaneous Raman, involves two optical fields in the visible range that may be amplified due to the local plasmon (Pump and Stokes fields in the case of Raman and visible and sum frequency beams in the case of SFG), and therefore achievable enhancement factors should be similar for the two techniques (the IR field in the SFG scheme is far off resonance and therefore unlikely to contribute significantly to the enhancement). SFG spectra from aggregates are different from those of isolated particles, suggesting that the enhancement mechanism is vibrational modespecific. This is also reminiscent of surface plasmon-enhanced Raman, where special “SERS selection rules” arise due to the anisotropic locally enhanced fields, which sometimes significantly alter the relative intensities of the vibrational bands.60 Further investigations are clearly needed to elucidate the mechanism and fully explore the detection limits of the surfaceselective nonlinear SFG spectroscopy enabled by the nanoplasmon enhancement. 5. Conclusions SFG is demonstrated as both sensitive and information-rich probe of molecular structure at nanosurfaces and interfaces. A new effect arises when the characteristic size of the nanostructures approaches the molecular scale: the conformation of the adsorbate molecules becomes dependent on the geometry of the substrate. A simple example presented here is a relationship between the trans-gauche conformation of the alkane chain of the ligand on the curvature of the spherical nanoparticles, with more gauche defects observed on smaller particles. Similar behavior was observed on gold and silver nanoparticles, suggesting that this is a geometric packing effect, rather than

Dodecanethiol on Metal Nanoparticles material-specific morphology of the surface defects or details of surface binding. Although our results pertain to only a narrow subset of possible nanoscale geometric effects on the molecular conformation, a broader conjecture may be suggested of a relationship between the conformation of adsorbate and the geometry of the nanostrictured substrate whose characteristic size approaches the molecular scale. Enhancement by more than 3 orders of magnitude (compared to similar molecules on nonresonant dielectric substrate) is observed for dodecanethiol on isolated gold or silver nanoparticles. Despite of the near spherical symmetry of the nanoparticles, SFG spectra can be readily recorded from nanoparticles on a flat substrate at surface coverage as low as a few % of a monolayer. Drying-mediated aggregation of nanoparticles on the surface results in further enhancement by more than 1 order of magnitude, such that SFG spectra from individual submicron size aggregates can be readily recorded. To our knowledge, this is the first example of an SFG spectrum recorded from a submicron size object. This approach may enable applications to vibrational spectral imaging with sub-wavelength resolution. Acknowledgment. This research is supported by NSF CAREER Grant No. CHE-0449720. Supporting Information Available: Fitting parameters. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Dieringer, J. A.; McFarland, A. D.; Shah, N. C.; Stuart, D. A.; Whitney, A. V.; Yonzon, C. R.; Young, M. A.; Zhang, X. Y.; Van Duyne, R. P. Faraday Discuss. 2006, 132, 9. (2) Sandros, M. G.; Gao, D.; Benson, D. E. J. Am. Chem. Soc. 2005, 127, 12198. (3) Daniel, M. C.; Astruc, D. Chem. ReV. 2004, 104, 293. (4) Huynh, W. U.; Dittmer, J. J.; Alivisatos, A. P. Science 2002, 295, 2425. (5) Chen, Z.; Shen, Y. R.; Somorjai, G. A. Annu. ReV. Phys. Chem. 2002, 53, 437. (6) Wang, H. F.; Gan, W.; Lu, R.; Rao, Y.; Wu, B. H. Int. ReV. Phys. Chem. 2005, 24, 191. (7) Richmond, G. L. Chem. ReV. 2002, 102, 2693. (8) Guyot-Sionnest, P.; Hunt, J. H.; Shen, Y. R. Phys. ReV. Lett. 1987, 59, 1597. (9) Bain, C. D. J. Chem. Soc.-Faraday Trans. 1995, 91, 1281. (10) Conboy, J. C.; Messmer, M. C.; Richmond, G. L. J. Phys. Chem. B 1997, 101, 6724. (11) Clarke, M. L.; Wang, J.; Chen, Z. J. Phys. Chem. B 2005, 109, 22027. (12) Chen, X. Y.; Clarke, M. L.; Wang, J.; Chen, Z. Int. J. Mod. Phys. B 2005, 19, 691. (13) Kawai, T.; Neivandt, D. J.; Davies, P. B. J. Am. Chem. Soc. 2000, 122, 12031. (14) Humbert, C.; Busson, B.; Abid, J. P.; Six, C.; Girault, H. H.; Tadjeddine, A. Electrochim. Acta 2005, 50, 3101. (15) Baldelli, S.; Eppler, A. S.; Anderson, E.; Shen, Y. R.; Somorjai, G. A. J. Chem. Phys. 2000, 113, 5432. (16) Holman, J.; Ye, S.; Neivandt, D. J.; Davies, P. B. J. Am. Chem. Soc. 2004, 126, 14322. (17) Wang, C. Y.; Groenzin, H.; Shultz, M. J. J. Phys. Chem. B 2004, 108, 265. (18) Rupprechter, G. Phys. Chem. Chem. Phys. 2001, 3, 4621. (19) Roke, S.; Berg, O.; Buitenhuis, J.; van Blaaderen, A.; Bonn, M. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 13310. (20) Roke, S.; Roeterdink, W. G.; Wijnhoven, J. E. G. J.; Petukhov, A. V.; Kleyn, A. W.; Bonn, M. Phys. ReV. Lett. 2003, 91. (21) Ward, R. N.; Duffy, D. C.; Davies, P. B.; Bain, C. D. J. Phys. Chem. 1994, 98, 8536. (22) Himmelhaus, M.; Eisert, F.; Buck, M.; Grunze, M. J. Phys. Chem. B 2000, 104, 576. (23) Nishi, N.; Hobara, D.; Yamamoto, M.; Kakiuchi, T. J. Chem. Phys. 2003, 118, 1904. (24) Bordenyuk, A. N.; Jayathilake, H.; Benderskii, A. V. J. Phys. Chem. B 2005, 109, 15941.

J. Phys. Chem. C, Vol. 111, No. 25, 2007 8933 (25) Lu, R.; Gan, W.; Wu, B. H.; Zhang, Z.; Guo, Y.; Wang, H. F. J. Phys. Chem. B 2005, 109, 14118. (26) Weeraman, C.; Yatawara, A. K.; Bordenyuk, A. N.; Benderskii, A. V. J. Am. Chem. Soc. 2006, 128, 14244. (27) Jiang, J.; Bosnick, K.; Maillard, M.; Brus, L. J. Phys. Chem. B 2003, 107, 9964. (28) Freeman, R. G.; Grabar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter, D. G.; Natan, M. J. Science 1995, 267, 1629. (29) Rabani, E.; Reichman, D. R.; Geissler, P. L.; Brus, L. E. Nature 2003, 426, 271. (30) Bordenyuk, A. N.; Benderskii, A. V. J. Chem. Phys. 2005, 122, 134713. (31) Jayathilake, H. D.; Zhu, M. H.; Rosenblatt, C.; Bordenyuk, A. N.; Weeraman, C.; Benderskii, A. V. J. Chem. Phys. 2006, 125, 064706. (32) van der Ham, E. W. M.; Vrehen, Q. H. F.; Eliel, E. R. Opt. Lett. 1996, 21, 1448. (33) Richter, L. J.; Petralli-Mallow, T. P.; Stephenson, J. C. Opt. Lett. 1998, 23, 1594. (34) Diels, J.-C.; Rudolph, W. Ultrashort Laser Pulse Phenomena; Academic Press: San Diego, CA, 1996. (35) Macphail, R. A.; Strauss, H. L.; Snyder, R. G.; Elliger, C. A. J. Phys. Chem. 1984, 88, 334. (36) Yatawara, A. K.; Weeraman, C.; Jayathilake, H.; Bordenyuk, A. N.; Benderskii, A. V. unpublished. (37) Love, J. C.; Estroff, L. A.; Kriebel, J. K.; Nuzzo, R. G.; Whitesides, G. M. Chem. ReV. 2005, 105, 1103. (38) Badia, A.; Cuccia, L.; Demers, L.; Morin, F.; Lennox, R. B. J. Am. Chem. Soc. 1997, 119, 2682. (39) Shimizu, T.; Teranishi, T.; Hasegawa, S.; Miyake, M. J. Phys. Chem. B 2003, 107, 2719. (40) Ang, T. P.; Wee, T. S. A.; Chin, W. S. J. Phys. Chem. B 2004, 108, 11001. (41) Hirose, C.; Akamatsu, N.; Domen, K. J. Chem. Phys. 1992, 96, 997. (42) Hirose, C.; Yamamoto, H.; Akamatsu, N.; Domen, K. J. Phys. Chem. 1993, 97, 10064. (43) Wei, X.; Hong, S. C.; Zhuang, X. W.; Goto, T.; Shen, Y. R. Phys. ReV. E 2000, 62, 5160. (44) Lu, R.; Gan, W.; Wu, B. H.; Chen, H.; Wang, H. F. J. Phys. Chem. B 2004, 108, 7297. (45) Wu, H.; Zhang, W. K.; Gan, W.; Cui, Z. F.; Wang, H. F. J. Chem. Phys. 2006, 125, 133203. (46) Zhuang, X.; Miranda, P. B.; Kim, D.; Shen, Y. R. Phys. ReV. B 1999, 59, 12632. (47) Wolfrum, K.; Laubereau, A. Chem. Phys. Lett. 1994, 228, 83. (48) Jayathilake, H. D.; Zhu, M. H.; Rosenblatt, C.; Bordenyuk, A. N.; Weeraman, C.; Benderskii, A. V. J. Chem. Phys. 2006, 125. (49) Hirose, C.; Akamatsu, N.; Domen, K. Appl. Spectrosc. 1992, 46, 1051. (50) Zhang, H. P.; Romero, C.; Baldelli, S. J. Phys. Chem. B 2005, 109, 15520. (51) Mar, W.; Klein, M. L. Langmuir 1994, 10, 188. (52) Muller-Meskamp, L.; Lussem, B.; Karthauser, S.; Prikhodovski, S.; Homberger, M.; Simon, U.; Waser, R. Phys. Status Solidi A 2006, 203, 1448. (53) Giz, M. J.; Duong, B.; Tao, N. J. J. Electroanal. Chem. 1999, 465, 72. (54) Jackson, A. M.; Hu, Y.; Silva, P. J.; Stellacci, F. J. Am. Chem. Soc. 2006, 128, 11135. (55) Gaines, G. L., Jr. Insoluble Monolayers at Liquid-Gas Interfaces; J. Wiley & Sons, Inc: New York, 1966. (56) Haynes, C. L.; Van Duyne, R. P. J. Phys. Chem. B 2003, 107, 7426. (57) Jensen, T.; Kelly, L.; Lazarides, A.; Schatz, G. C. J. Cluster Sci. 1999, 10, 295. (58) Zhao, L. L.; Kelly, K. L.; Schatz, G. C. J. Phys. Chem. B 2003, 107, 7343. (59) Rechberger, W.; Hohenau, A.; Leitner, A.; Krenn, J. R.; Lamprecht, B.; Aussenegg, F. R. Opt. Commun. 2003, 220, 137. (60) Otto, A.; Mrozek, I.; Grabhorn, H.; Akemann, W. J. Phys.Condensed Matter 1992, 4, 1143.