Frequency-Domain Proof of the Existence of Atomic-Scale SERS Hot

Dec 5, 2017 - Such site-specific, single-peak spectra could be successfully modeled using single-molecule SERS induced by a hot-spot with a diameter n...
0 downloads 5 Views 3MB Size
Letter Cite This: Nano Lett. 2018, 18, 262−271

pubs.acs.org/NanoLett

Frequency-Domain Proof of the Existence of Atomic-Scale SERS HotSpots Hyun-Hang Shin, Gyu Jin Yeon, Han-Kyu Choi, Sang-Min Park, Kang Sup Lee, and Zee Hwan Kim* Department of Chemistry, Seoul National University, Seoul 08826, Korea S Supporting Information *

ABSTRACT: The existence of sub-nanometer plasmonic hotspots and their relevance in spectroscopy and microscopy applications remain elusive despite a few recent theoretical and experimental evidence supporting this possibility. In this Letter, we present new spectroscopic evidence suggesting that Angstrom-sized hot-spots exist on the surfaces of plasmonexcited nanostructures. Surface-enhanced Raman scattering (SERS) spectra of 4,4′-biphenyl dithiols placed in metallic junctions show simultaneously blinking Stokes and anti-Stokes spectra, some of which exhibit only one prominent vibrational peak. The activated vibrational modes were found to vary widely between junction sites. Such site-specific, single-peak spectra could be successfully modeled using single-molecule SERS induced by a hot-spot with a diameter no larger than 3.5 Å, located at the specific molecular sites. Furthermore, the model, which assumes the stochastic creation of hot-spots on locally flat metallic surfaces, consistently reproduces the intensity distributions and occurrence statistics of the blinking SERS peaks, further confirming that the sources of the hot-spots are located on the metallic surfaces. This result not only provides compelling evidence for the existence of Angstrom-sized hot-spots but also opens up the new possibilities for the vibrational and electronic control of single-molecule photochemistry and real-space visualization of molecular vibration modes. KEYWORDS: Surface-enhanced Raman scattering, plasmonics, vibrational spectroscopy

I

with abnormally large anti-Stokes versus Stokes intensity ratios were observed. In addition, they observed subtle site-specific variations in the spectral features. These observations were attributed to the interaction of the field with an effective modevolume of 50 nm still awaits confirmation. There exist a few notable experimental observations, which, among several possibilities, could be attributed to the action of sub-nm hot-spots. For example, several groups26−29 observed sub-nm image features in tipenhanced Raman scattering (TERS) measurements, which could arise from sub-nm hot-spots at the apexes of scanning probes. In cryogenic SERS measurements,30 blinking spectra © 2017 American Chemical Society

Received: September 20, 2017 Revised: November 23, 2017 Published: December 5, 2017 262

DOI: 10.1021/acs.nanolett.7b04052 Nano Lett. 2018, 18, 262−271

Letter

Nano Letters

Figure 1. Structure of the AgNP−BPDT−AuTF junction and the modeling of single-molecule SERS signals induced by a sub-nm hot-spot. (a) AgNP−BPDT−AuTF junction and SERS measurement. (b) Coupling scheme for the SERS induced by sub-nm hot-spots: far-field excitation field (E0) induces a gap-field (Eloc), which excites a sub-nm hot-spot E(r,r0) at the end of sub-nm protrusion (r0) on AgNP or AuTF surfaces. The sub-nm hot-spot locally excites Raman radiation of a BPDT molecule, and the Raman radiation is out-coupled to the hot-spot mode E(r,r0), then to the gapmode Eloc and finally scattered to the far-field (ESERS).

Figure 2. SERS trajectories showing Stokes and anti-Stokes blinking. (a) Time-resolved Stokes and anti-Stokes SERS spectra obtained from a AgNP−BPDT−AuTF junction showing on (red arrow)/off (green arrow) blinking of a peak at 998 cm−1 (ring-deformation mode). The lowfrequency range (ν = −500 to +500 cm−1) is blocked out by a Raman notch filter used in the measurement. (b) Blinking component of the trajectory in (a): spectrum at each time delay is subtracted from an average SERS spectrum sampled during the off-period. (c) SERS spectra sampled during on (red arrow in (a)) and off (green arrow in (a)) periods. The two are displaced along the y-axis to better show the spectral features. Also shown in blue is the difference between the on and off spectra for the blinking component. (d−f) Another set of SERS trajectory data with pronounced blinking peak at 1576 cm−1 (aromatic C−C stretching). (g) Collection of blinking components (on−off) showing single-peak SERS spectra (top eight spectra, frequencies of the peaks are shown in cm−1) as well as multipeak spectra (bottom three spectra). In these spectra, relative intensities of the Stokes and anti-Stokes peaks are not rescaled unless specified. For the assignment of the peaks, see Table 1.

covalent bridges with the surfaces of AgNP (average diameter of 80 nm) and AuTF (thickness of 10 nm) and provides an average gap-distance of 1.1 ± 0.1 nm.32 A focused laser beam

Figure 1a shows the structure of a self-assembled AgNP− (monolayer of BPDT)−AuTF plasmonic junction and the experimental setup (see also Methods). The BPDT forms 263

DOI: 10.1021/acs.nanolett.7b04052 Nano Lett. 2018, 18, 262−271

Letter

Nano Letters (power density of ∼1 MW/cm2) with a wavelength λex = 633 nm was impinged on individual junction sites, and the antiStokes and Stokes spectra were simultaneously recorded by a Raman spectrometer. Assuming that a spherically shaped AgNP and a flat AuTF create a gap of 1.1 nm, the gap-field is estimated to span an area of ∼20 nm2 on the AuTF surface, thereby exciting ∼60 BPDT molecules per junction (Supporting Information-A). Figure 2a−f displays two SERS trajectories recorded from two different junction sites (the low-frequency peaks with 1.0. The positions and diameters of hot-spots in (d)−(g) are shown as black cross-markers and dashed circles. (h) Simulated SERS spectrum for a uniform field polarized along the zaxis. In the molecular models shown in (a), (d), and (f), the two C atoms in the phenyl groups are shown as black spheres to clarify the optimal positions of hot-spots. (i−p) Corresponding results for the peak at 998 cm−1.

SERS could be efficient enough to compete with rapid vibrational relaxations, which may lead to hyperthermal antiStokes signals. Below, we specifically focus on modeling the single-peak SERS spectra in the blinking components since they constitute the most extreme cases of site-specific vibrational selection and thus could provide the most clear-cut insight into the interaction between a hot-spot and a single molecule.

To search for the configurations of hot-spots that give rise to the observed single-peak spectra, the picocavity coupling model (see Figure 1b) developed by the group of Aizpurua26,30,45 was employed. A laser illumination (E⃗ 0(ν0), where ν0 is the excitation laser frequency) of a plasmonic junction generates a gap-field (E⃗ loc(ν0) = Eloc(ν0)ε⃗loc = Eloc(ν0)ẑ, with a gap-axis aligned along the ẑ-direction), which is locally uniform on 265

DOI: 10.1021/acs.nanolett.7b04052 Nano Lett. 2018, 18, 262−271

Letter

Nano Letters

normal mode. The φLn,k is the normalized displacement of the atom at qn associated with k-th normal mode vibration,48 which is expressed in lab Cartesian coordinates. The factors φnL, k / μk = ∂qn /∂Q k are the elements of the Jacobian transformation matrix, JLk̃ , which relates the mass-scaled normal mode coordinate (Qk) and the Cartesian coordinate of atoms L (qn). The (∂αρσ /∂qn)0 is an element of the Cartesian polarizability derivative tensor, α̃ ′L. The JkL̃ and α̃ ′L are dependent on the molecular orientation (θ, ϕ, χ) in the labframe. We first evaluated the two quantities in a molecule-fixed mol L̃ ̃ frame (Jmol k and α̃ ′ ) and then transformed them into Jk and L −1 mol L L̃ ̃ ̃ ̃ α̃ ′ via rotational transformation (Jk = T Jk T and α̃ ′ = T̃ −1 α̃ ′molT̃ , where T̃ = T̃ (θ, ϕ, χ) is the Euler rotation matrix). Eq 2 shows that the SERS amplitude of a k-th vibrational mode is determined by the local-field-weighted (εσ(qn)ερ(qn)) sum of changes in the polarizability that are induced by the atomic L /∂qn)0 . displacements of the vibrational mode, φnL, k / μk (∂αρσ As such, with a submolecular sized hot-spot, the SERS intensity of a vibrational peak will be highly dependent on the location of the hot-spot (r0⃗ ) with respect to the positions of atoms in a molecule, {qn}. For the simulation of SERS spectra, molecular geometries, vibrational frequencies, Raman tensors, and Jacobian matrices for the geometry-optimized Ag−BPDT−Au complex were calculated by the density functional theory (DFT) with the B3LYP (Becke, three-parameter, Lee−Yang− Parr) exchange-correlation functional (see Methods and Supporting Information-D). These molecular parameters and various configuration of ε⃗(r,⃗ r0⃗ ) were combined to model the spectra. Figure 3 compares the blinking Stokes SERS spectra and model spectra calculated with various positions and sizes of hotspots. The staggered-BPDT conformation with a dihedral angle between two phenyl rings of 36° was used for all the calculations shown. Figure 3a shows the simulated density plot (gray distribution in the logarithmic color scale) of the peak intensity of the 1275 cm−1 mode drawn as a function of the hot-spot (a diameter of 3.5 Å) position. The spatial variation of intensity ratio for the 1275 cm−1 peak versus the second strongest peak in the spectrum is shown in the colored contours. For this particular mode, both the peak intensity and the peak ratio distributions are highly confined around two C atoms (the two black spheres in Figure 3a) connecting two phenyl rings. With the hot-spot located close to one of the C atoms (marked as a red-cross in Figure 3a; the distance between C atom and hot-spot center is 2 Å), the model (Figure 3c, spectrum in red) successfully reproduces the experimental spectrum (Figure 3b). As we move the hot-spot away from the optimal position (see green and black crosses in Figure 3a and the spectra in the corresponding color in Figure 3c), both the SERS intensity and the peak ratio rapidly decrease, which results in spectra that significantly differ from the experiment. Figure 3d−g displays the results of the calculations with hotspot diameters of 5 and 7.5 Å. The ratio-optimized spectrum with 5 Å hot-spot (spectrum in Figure 3e) shows only a marginal resemblance to the experiment, and the simulation with 7.5 Å hot-spot results in the spectra (see, for example, Figure 3g) that are qualitatively different from the experiment, irrespective of the hot-spot positions. In fact, the latter spectrum is similar to the usual Raman spectrum induced by a uniform field (Figure 3h). Figure 3i−p displays the results of analogous simulations for the single-peak spectrum at 998 cm−1

single-molecule scale. The gap-field, in turn, generates an Angstrom-sized field distribution E⃗ (ν0, r,⃗ r0⃗ ) (where r ⃗ and r0⃗ are the field-point and the center of the hot-spot, respectively) around a sub-nm protrusion (at r0⃗ ) on the metallic surface. A recent atomistic quantum calculation24 on metal clusters indicates that a single atomic protrusion could generate an Angstrom-sized field that is distributed approximately isotropically around an atom. Here such a field is modeled by an isotropic Gaussian distribution with a radially diverging field vector: ⎛ | r ⃗ − r ⃗ |2 ⎞ r − r0⃗ 0 ⃗ (ν0)|· ⃗ ⎟ E ⃗(ν0 , r ⃗ , r0⃗ ) = F ·|Eloc ·exp⎜ − | r ⃗ − r0⃗ | a2 ⎠ ⎝ = E(ν0)ε ⃗( r ⃗ , r0⃗ )

where ε⃗( r ⃗ , r0⃗ ) =

(1)

(−

( r ⃗ − r0⃗ ) ·exp | r ⃗ − r0⃗ |

2

| r ⃗ − r0⃗ | a2

) and a represent the

shape (hot-spot mode) and the size (fwhm = 2 ln 2 a) of the field distribution centered at position r0⃗ , respectively. The E⃗ is independent of the vectorial direction and frequency of the driving E⃗ loc-field, and F, the relative field enhancement of |E⃗ | with respect to |E⃗ loc|, is sufficiently large so that a single molecule near the hot-spot primarily experiences E⃗ . This E⃗ excites the Raman transition of the molecule, and the resulting radiation at frequencies ν0 ± νk (where the + and − signs correspond to anti-Stokes and Stokes components with vibrational frequency νk, respectively) is out-coupled to the same hot-spot mode (E(ν0 ± νk)ε⃗(r,⃗ r0⃗ )), then to the gap-mode (Eloc(ν0 ± νk)ε⃗loc), and finally to the far-field, where it is detected as Stokes and anti-Stokes SERS signals. The in- and out-coupling between E(ν0 ± νk)ε⃗(r,⃗ r0⃗ ) and Eloc(ν0 ± νk)ε⃗loc are assumed to be polarization- and frequency-independent. Thus, the SERS spectrum of a single molecule is completely determined by the spatial configuration (r0⃗ and a) of ε⃗(r,⃗ r0⃗ ) and molecular geometry and is not dependent on the molecular orientation with respect to the junction axis. In general, the SERS signals induced by field distributions much smaller than molecular sizes cannot be described by the polarizability derivatives with respect to normal mode coordinates of the molecule. To rigorously describe the interaction of ε⃗(r,⃗ r0⃗ ) and a single molecule, matrix elements of a multipolar light−molecule interaction Hamiltonian46,47 and their partial derivatives with respect to normal mode coordinates should be numerically evaluated. However, under the assumptions that the field gradient at the length scales of atomic displacements of molecular vibration (typically 20 ms exposure time/ spectrum). A Raman notch filter was placed in front of the spectrometer to filter out both the Rayleigh scattering and lowfrequency vibrational peaks (