fluctuating single sp2 carbon clusters at single hotspots of silver nanoparticle dimers investigated by surface enhanced resonance raman scattering

Yamamoto,2,3Vasudevanpillai Biju,1 Hiroharu Tamaru,4and Shin-ichi Wakida1 1Nano-Bioanalysis Research Group, Health Research Institute, National Institute of Advanced Industrial Science a

dimers investigated by surface-enhanced resonance Raman scattering

Tamitake Itoh, Yuko S Yamamoto, Vasudevanpillai Biju, Hiroharu Tamaru, and Shin-ichi Wakida

Citation: AIP Advances 5, 127113 (2015); doi: 10.1063/1.4937936

View online: http://dx.doi.org/10.1063/1.4937936

View Table of Contents: http://aip.scitation.org/toc/adv/5/12

Published by the American Institute of Physics

AIP ADVANCES 5, 127113 (2015)

hotspots of silver nanoparticle dimers investigated

by surface-enhanced resonance Raman scattering

Tamitake Itoh,1, aYuko S Yamamoto,2,3Vasudevanpillai Biju,1

Hiroharu Tamaru,4and Shin-ichi Wakida1

1Nano-Bioanalysis Research Group, Health Research Institute, National Institute of

Advanced Industrial Science and Technology (AIST), Takamatsu, Kagawa 761-0395, Japan

2Research Fellow of the Japan Society for the Promotion of Science,

Chiyoda, Tokyo 102-8472, Japan

3Department of Chemistry, School of Science and Technology, Kagawa University,

Takamatsu, Kagawa 761-0396, Japan

4Photon Science Center, the University of Tokyo, Tokyo 113-8656, Japan

(Received 5 October 2015; accepted 2 December 2015; published online 9 December 2015)

We evaluate spectral changes in surface enhanced resonance Raman scattering (SERRS) of near-single dye molecules in hotspots of single Ag nanoparticle (NP) dimers During the laser excitation, surface enhance florescence (SEF) of dye dis-appeared and the number of SERRS lines decreased until finally ca two lines remained around 1600 and 1350 cm−1, those are evidence of G and D lines of single sp2carbon clusters Analysis of the G and D line intensity ratios reveals the temporal fluctuation in the crystallite size of the clusters within several angstroms; whereas, broadening and splitting in the lines enable us for identifying directly the dynamics of various defects in the clusters This analysis reveals that the detailed fluctuations of single sp2carbon clusters, which would be impossible to gain with other microscopic methods C 2015 Author(s) All article content, except where otherwise noted, is licensed under a Creative Commons Attribution 3.0 Unported License.[http://dx.doi.org/10.1063/1.4937936]

I INTRODUCTION

The extremely small mode volume of several cubic nanometers for an electromagnetic (EM) field coupled with plasmonic resonance at the crevasse or gap of a metallic NP dimer enhances the electronic transition rates of a molecule located at the crevasse by a factor of ca 105.1At such cre-vasses, or so-called hotspots, enhancement factors for the molecular resonance Raman cross-section are up to 1010(105× 105), enabling us single molecule vibrational spectroscopy.2This phenomenon

is well known as SERRS,3and SERRS hotspots provides a powerful platform to investigate unique behaviors of single-molecules at the sub-nanometer scale.47Indeed, the various physical and chem-ical processes of molecules in hotspots including thermal diffusion,8reorientation,9photo-induced electron transfer,10and the photo-chemical reaction11have been investigated at the hotspots SERRS analysis of conversion from analyte molecules to heterogeneous amorphous carbon has been carried out because photo-induced carbonization is one of the important phenomena for mate-rial science by many reserchers.12–14However; inhomogenous SERRS spectral broadening owing to heterogeneity of amorphous carbons has prevented us from quantitative evaluation of the carboniza-tion Thus SERRS analysis using hotspots of single silver NP dimers are useful to directly observe photo-induced carbonization excluding inhomogeneity in the phenomenon Raman spectroscopy of amorphous carbons has provided rich information on correlations with carbon sp2systems and both external and internal perturbations,15 such as strain and stress,16 disorder,17 oxidation,18,19 hydro-genation,20 and nitridization.21 However, the spatial averaging in Raman spectroscopy has made

a Corresponding author; tamitake-itou@aist.go.jp

2158-3226/2015/5(12)/127113/10 5, 127113-1 © Author(s) 2015

the correlations unclear because domains that are subject to such perturbations are confined within

ca.1 nm2 Thus, single SERRS hotspots may enable us precise clarification of these correlations for photo-induced carbon sp2systems by excluding inhomogeneous spectral broadening

In this study, we report SERRS analysis of generation of photo-induced sp2 carbon clusters from near-single dye molecules adsorbed in single hotspots of silver NP dimers and the subsequent fluctuations in the clusters During the laser irradiation, SEF of dye disappeared and the number of SERRS lines of the dye irreversibly decreased until finally the G and D lines remained around 1600 and 1350 cm−1 The intensity ratios of the G and D line, and the G line linewidth of ca 50 cm−1 indicate that the effective crystallite size of sp2carbon clusters is less than 2.0 nm Fluctuations in the intensity ratios of these lines are attributed to changes in the effective crystallite size of the delo-calized sp2system by several angstroms The various spectral changes such as spectral line splitting represent the formation and extinction of defects in the sp2carbon systems, which is consistent with DFT calculations.18 , 22 – 24

II EXPERIMENT

Colloidal silver NPs were prepared by the Lee-Meisel method.25 The average NP diameter was estimated at 40 nm from transmission electron microscopy (TEM) observations The colloidal silver NP solution had a concentration of ca 1.0 × 10−10M In parallel, aqueous solutions of rhoda-mine 6G (R6G; C28H31N2O3Cl, 1.34 × 10−8M), rhodamine B (RB; C28H31N2O3Cl, 1.30 × 10−8M) and crystal violet (CV; C25H30N3Cl, 1.13 × 10−8M) were prepared separately, which were supple-mented with 10 mM NaCl solution These dye solutions were added separately to three silver NP solutions and incubated at room temperature (ca 20◦C) for 15 min The solutions were spin-coated onto separate cover glass plates, which were followed by placing a drop of pure water on each plate for reducing possible thermal decomposition of Ag NPs during laser irradiation.26

The spectroscopic setup for measurement of SERRS and plasmon resonance spectra is reported

in details elsewhere.26In short, a green laser (532 nm, 2.33 eV, 2.0 W/cm2) was used as the excita-tion light source and the exposure time was 5.0 s The method for the selecexcita-tion of silver NP dimers

is also described in details elsewhere with their SERRS spectra, plasmon resonance spectra, and SEM images.27,28 Briefly, SERRS active silver NPs that have dipolar plasmon resonance maxima around 1.8 to 2.1 eV with a linewidth less than ca 100 meV were identified and selected for evaluation of photo-conversion of dye molecules into sp2carbon clusters

Note that the current concentrations for both dye solution and silver colloidal solution are almost identical to those in Ref.28 Under the conditions we have frequently observed that SERRS light from single silver NP dimers exhibited blinking, which means fluctuation and intermittent of emitted SERRS light The blinking has been used for indirect evidence of near single molecule detection.2 , 8The quantitative analysis of the ratios between SERRS and SEF intensity revealed that the blinking is induced by temporal fluctuations of effective distances between dye molecules and silver surfaces within several angstroms.28 Thus, we consider that SERRS signals of the current experiments are from near-single dye molecules

III RESULTS AND DISCUSSION

Figures1(a)and1(b)show temporal evolutions of SERRS spectra commonly observed in the current experiment SERRS spectra eventually change into fluorescence spectra just before quench-ing or decrease in intensity while remainquench-ing their spectral lines until quenchquench-ing The changes into fluorescence indicate the decreasing in an enhancement factor of the decay rate for a molecule in the excited state due to resonance energy transfer from a molecule to a Ag dimer.28The enhance-ment factor is quite sensitive to the distance between a molecule and a silver surface.5 , 28Thus, the changes into fluorescence can be explained as desorption of dye molecules into water from the hotspots The decrease in intensity may be due to desorption to silver surfaces outside hotspots.8

Both two phenomena are common for the final stage of temporal evolution of single molecule SERRS spectra

127113-3 Itoh et al. AIP Advances 5, 127113 (2015)

FIG 1 (a, b) Typically observed temporal changes in SERRS spectra of R6G molecules just before SERRS quenching (a) SERRS changes into fluorescence just before quenching (b) SERRS decreases in intensity while remaining their spectral lines of dye molecules until quenching.

However, in rare cases we found SERRS temporal evolution different from these common ones

by extending observation periods Figures2(a)–2(c)show the temporal evolution for the three types

of dye molecules taken every 5 s The SERRS lines in Figs 2(a1)–2(c1)are assigned to Raman active vibrational modes for the dyes.29 – 31 As the laser irradiation time increased, the number of line decreased until finally only one to two lines were remained around 1350 and 1600 cm−1 Figures2(d1)–2(d3)show the two lines are different from both SERRS and surface-enhanced hyper Raman scattering lines for the dye molecules,32which indicates that the line in Figs.2(a)–2(c)do not represent Raman inactive modes produced by a field gradient effect,4 , 6but are similar to the G and D lines of amorphous like carbon.33Disappearance of SEF of dye, which is observed as broad background around 0 to 1000 cm−1in Fig.2(d2), as shown in Fig.2(d1)also supports the generation

of amorphous like carbon The G line (ca 1600 cm−1) is due to E2g symmetry vibrations of sp2 aromatic rings The D line (ca 1350 cm−1) is due to A1gsymmetry vibrations, which is known to be the disorder-activated optical zone edge mode of graphite crystals.15,33The line infrequently emerg-ing at ca 1100 cm−1as in Fig.2(d1) is the D” line due to vibrations of trans-polyacetylene-like structural units.34It is reported that graphene composed of only 32 carbon atoms shows the G line

in the Raman spectra.18 Thus carbon clusters generated from only one or two dye molecules can exhibit such a G line SERRS blinking commonly observed in the current experiment also supports the generation of carbon clusters from a few molecules Thus, we consider that the spectral changes are induced by conversion from near-single dye molecules to sp2carbon clusters Several groups have reported the emergence of the broad G line in the SERRS spectra of amorphous carbon.12 – 14

The current studies show large variations in the full width at half maximum (FWHM) of the G line ΓG, from 20 to 260 cm−1 Regarding homogeneous ΓG∼ 13 cm−1,35the observed ΓGmay be homogeneous line-widths of single carbon clusters and their large variations may be induced by some perturbation including generation of defects in the sp2carbon systems.18 , 22 – 24

The observation of SERRS spectra of single carbon clusters is needed further explanation, because their Raman cross-session of sp2carbon systems are much smaller than that of the used dye molecules by ca 6000 times under the current excitation photon energy.36 , 37Whereas, the observed SERRS intensities at 1650 cm−1of dye molecules are only ca 30 times larger than those of G line

FIG 2 (a, b, c) SERRS spectral changes showing conversion of dye molecules into carbon clusters: R6G (a1-a4), Rb (b1-b4) and CV (c1-c4) (c) Typical SERRS spectra of a carbon cluster (d1), R6G (d2), and surface-enhanced hyper Raman scattering (SEHRS) spectrum of R6G (d3).

intensities of carbon clusters as shown in Figs 2(a)–2(c) Thus, the observed SERRS intensities

of single carbon clusters are much larger than the expected intensities by ca 200 times One of the reasons for the too large SERRS intensities of carbon clusters may be due to closing plas-mon resonance to the laser line by the blue-shifts in plasplas-mon resonance maxima during temporal evolution of SERRS, resulting in an increase in an EM enhancement factor.28Indeed, the strong G lines were observed only when the G lines were completely overlapped with plasmon resonance maxima as shown in Fig.3(a) The blue-shifts in plasmon resonance are frequently observed and has been explained as breaking up of hybridized resonance, which is induced by strong coupling between plasmonic and molecular electronic resonance, by destabilization or decomposition of molecules located at SERRS hotspots.7However, the blue-shifts usually enlarge SERRS intensity

by ca 20 times.28 , 32 Thus, even counting the plasmonic blue-shifts, we cannot explain residual

ca.10 times We do not have clear reason for the large SERRS intensity of single carbon clusters However regarding SERRS enhancement factors ca 1010of the current systems,27 , 28 , 32the 10 times may be within the error range We have a plan to check the possibility that carbon atoms of citric acid molecules covering silver surfaces incorporate to the carbon cluster during laser excitation for several ten minutes

Figure3(a)shows SERRS of single carbon cluster and plasmon resonance of single silver NP dimer Their maxima overlap each other The overlap generates the largest EM enhancement,28

which makes the G line and its side lines clear, allowing us for discussion of detailed spectral features We discuss here various types of fluctuations in the SERRS spectra for the carbon cluster The quite unclear G’ line at ca 2700 cm−1, which is the result of an inter-valley double resonance process involving inequivalent K and K’ points in the first Brillouin zone, indicates that the carbon cluster is a defect rich sp2 system.17 , 38 The D’ line at ca 1620 cm−1, which is the result of an

127113-5 Itoh et al. AIP Advances 5, 127113 (2015)

FIG 3 (a) SERRS spectrum of a carbon cluster (red line) and plasmon resonance spectrum of single silver dimer (blue line) (b) Temporal changes in SERRS spectra for carbon clusters formed from R6G The side panels (A) - (D) show SERRS spectra at the times indicated by arrows (c) EM field distribution around a silver dimer calculated by the finite-di fference time-domain method, and enlarged image of a crevasse in a silver dimer.

intra-valley double resonance process,33,38 is observed at the right side of the G line Figure3(b)

shows temporal changes in the SERRS spectra over a 400 s period The SERRS intensities have large variations, referred to as blinking, which is indirect evidence of single molecule detection induced by fluctuation of the molecular position and orientation within the hotspot, as illustrated in Fig.3(c).8,28Thus, the observation of intermitted SERRS light spots as shown in Fig.3(b)indicates detection of single carbon cluster In the case of simultaneous detection of a few clusters in a hotspot, such intermittence of SERRS light is not likely We commonly observed such intermittence during SERRS measurement The four panels on the right in Fig.3(b)depict SERRS spectra show-ing relatively large intensities Panels (A) and (D) show the D line around 1350 cm−1, whereas the

Dline in panel (C) seems to have disappeared, and the SERRS spectrum in panel (B) is completely

different from the others These spectral changes cannot be explained simply by the spatial fluctu-ation of the sp2carbon clusters To clarify these spectral changes, we classify the changes into two types; the first incorporates minor changes represented by the G and D line intensities shown in panels (C) and (D), and the second involves major changes such as the broadening and splitting of spectra, which is seen in panels (A) and (B)

Firstly, we focus on the minor changes in the SERRS spectra of photo-converted carbon clus-ters Figures4(a)–4(c)show that the G line intensity I(G) and the D line intensity I(D) vary inde-pendently In Fig.4(a1), I(G) is three times larger than I(D), whereas in Fig.4(a3), both intensities are comparable In Fig 4(b1), I(D) is comparable to I(G), but in Fig 4(b4), I(D) is three times larger than I(G) In Fig.4(c1), I(D) is clearly observed, while in Fig.4(c4)I(D) is negligible The

FIG 4 (a-c) Temporal changes in G and D line intensities in SERRS spectra for carbon clusters converted from R6G (a1–a4), RB (b1–b4), and CV (c1–c4) (d–f) Temporal changes in L a for carbon clusters converted from (d) R6G, (e) RB, and (f) CV (g–i) Temporal changes in Γ G for carbon clusters converted from (g) R6G, (h) RB, and (i) CV.

relationship between the I(D)/I(G) intensity ratio and the “effective” crystallite size of the carbon cluster Lahas been experimentally studied as a degree of disorder in graphite microcrystals.39When

La> 2 nm, I(D)/I(G) scales as 1/La, because I(D), which is due to a forbidden mode in perfect graphite and is mainly activated by disorder at the graphite crystal edges, scales as La,40 whereas

I(G) at 1580 cm−1 is activated by whole the area of the cluster, scales as La.40 Conversely, for

La< 2 nm, I(D) is determined by the number of aromatic rings in the crystallite area, whereas

I(G), which is determined by the motion of C sp2atoms, maintains its intensity because the motion

of C sp2atoms is activated even without aromatic rings, but the position of G line fluctuates as a function of crystallite size.41Thus, I(D) decreases with decreasing Laand I(D)/I(G) scales as La

In the present experiments, scaling of I(D)/I(G) with La was adopted because of the following three reasons (1) The volume of hotspots generating single molecule SERRS signals is of the order

of several cubic nanometers.27(2) From the relationship between ΓGand La,41the typical ΓGvalue

of ca 50 cm−1approximately correspond to Laof 1 to 4 nm Note that ΓGvalue of microscale zero defect graphene is ca 13 cm−1,35and that of amorphous carbon is ca 200 cm−1.41 (3) From the relationship between 2D line at ca 2700 cm−1, which is due to an inter-valley double resonance Raman process, and La, the unclear 2D line is evidence of La< 2 nm.17(c)Figures4(d)–4(f)show the laser irradiation time vs Launder the condition I(D)/I(G) = CL2a with C= 0.55 nm−2.40The independent fluctuation of I(D) and I(G) in Figs.4(a)–4(c)can be explained as reversible changes in

Laof several angstroms for the clusters We consider that photo-induced rearrangement of clusters may be the origin of the reversible changes in La The estimated correlation between Laand ΓG

is consistent with that previously reported.41Figures4(g)–4(i)shows laser irradiation time vs Γ

127113-7 Itoh et al. AIP Advances 5, 127113 (2015)

It has been reported that Lais inversely proportional to ΓG;41however, such a correlation was not observed here; therefore, Fig.4indicates that the values of ΓGare determined by other factors, such

as the variation of defects in the cluster

Secondly, we focus on drastic temporal changes in the SERRS including spectral broadening and splitting Figures5(a)–5(c)present typical examples of SERRS spectra with splitting and unifi-cation of both the G and D lines In Fig.5(a), the G line at 1585 cm−1splits into three lines at 1662,

1569, and 1522 cm−1, and finally these lines become a single line again at 1585 cm−1 The D line

at 1351 cm−1splits into two lines at 1351 and 1320 cm−1, and then the two lines finally return to a single line at 1351 cm−1 In Fig.5(b), the G line at 1570 cm−1line splits into three lines at 1614,

1566, and 1516 cm−1, which finally return to a single line at 1582 cm−1 The D line at 1365 cm−1 splits into two lines at 1382 and 1348 cm−1, and finally these lines disappear In Fig.5(c), the G line at 1570 cm−1peak splits into four lines at 1643, 1614, 1570, and 1496 cm−1, and these lines then return to a single line at 1570 cm−1 The D line at 1331 cm−1splits into two peaks at 1376 and 1346 cm−1, and finally these lines return to a single line at 1346 cm−1 These examples of line splitting and unification cannot be explained by changes in La, but may instead be related to the formation and annihilation of certain defects.18 , 22 – 24

FIG 5 (a-c) Temporal changes in SERRS for carbon clusters converted from R6G (a1–a5), RB (b1–b5), and CV (c1–c5) (d)

Γ G , Γ D , and Γ D” vs G, D, and D” line positions for R6G ( ⃝ , △, and  red), CV ( ⃝ , △, and,  green), and RB ( ⃝ , △, and , blue) The largest lines are indicated by solid color marks The solid black mark from Ref 35 indicates the minimum value of

Γ G Solid black marks from Ref 18 correspond to Raman lines for a graphene sheet including epoxy groups (♦, ), hydroxyl groups (▽, ▼), Stone–Wales defects (+), and C 2 vacancy defects (×) (e1-e5) Temporal spectral changes in SERRS for carbon clusters converted from CV Γ G is indefinable because the spectral splitting is too large, as indicated by the inverted solid triangles.

Let us discuss the type of defects in carbon clusters that could cause the drastic spectral changes observed in Figs.5(a)–5(c)based on the theoretical work.18,22–24,42To clarify the splitting and unifi-cation, the relationship between the spectral widths and the positions for the split lines of G, D, and D” lines are plotted in Fig.5(d) Note that SERRS spectra with a definable FWHM were selected

to exclude those with too large splitting SERRS spectra with an indefinable FWHM exhibit more drastic changes, as shown in Fig.5(e) Figure5(d)shows ΓG, ΓDand ΓD”versus their corresponding line positions In the case of ΓG< ca 50 cm−1, the G, D, and D” lines appear as single lines around

1570, 1350, 1090 cm−1, respectively As ΓG, ΓD, and ΓD” increase, the G and D lines split into several lines from 1450 to 1680 cm−1for the G mode, and from 1260 to 1410 cm−1for the D mode, while the D” lines stay around 1090 cm−1 In the case of ΓG< ca 200 cm−1, the G and D lines split from 1260 to 1730 cm−1, and these lines can no longer be classified into G, D, and D” lines The upper energy limit of the vibrational density of states (VDOS) for graphite is calculated to be lower than 1620 cm−1,42which indicates that the lines located higher than the VDOS limit are not Raman mode peaks due to six-fold sp2carbon atoms The presence of hydrogen, oxygen, nitrogen atoms in dye molecules suggests that generated carbon clusters are functionalized by these kinds of atoms Thus these atoms may be the origin of the line splitting Raman analysis of graphite oxide has been theoretically conducted using DFT methods.18 , 22 – 24In particular, Kudin et al.18calculated Raman spectra for graphite sheets, of which the size of ca 4 × 4 nm2 is similar to the clusters observed in the present study The pure graphite sheets composed of 32 atoms well reproduce the

Gline at ca 1586 cm−1, which confirms that their DFT calculations are valid for the examination

of delocalized carbon sp2systems.18The addition of epoxy and hydroxyl groups onto the graphene sheet causes the G line to split into the lower vibrational energy region and several D-like lines appear, as shown in the right-hand panel insert of Fig.5(d).18The calculations partially reproduce the SERRS spectra obtained here at <1580 cm−1, but do not explain the splitting into the higher vibrational energy region >1730 cm−1 Thus, the splitting may be mainly induced by topological defects that significantly modify the delocalization of π electrons in the sp2carbon system.43Kudin

et al calculated Raman spectra by adding Stone–Wales and C2vacancy defects, and by displacing carbon atoms from the graphene sheets.18These topological defects induce splitting of the G line into more than 10 lines with positions distributed from 950 to 1750 cm−1, as shown in the insert of Fig.5(d).18The results are consistent with the distribution of split lines in Fig.5(d), which indicates the observed splitting and unifying is induced by the destruction and re-construction of six-fold car-bon sp2systems by losing and recovering of chemical bonding among carbon atoms Photo-induced electronic interaction between clusters and silver surfaces may be the origin of the destruction and re-construction of the chemical bonding.10 In Raman spectra of defective graphite, an increase in

ΓGand an increase or decrease in I(D) were experimentally observed However, line splitting has not been evaluated in such measurements owing to the spectral averaging due to the large Raman detection area of ca 0.3 µm2 In the present study, SERRS spectra were measured from a volume

of several cubic nanometers, where homogeneous spectral changes of single carbon clusters could

be directly observed, as predicted by DFT methods, by the exclusion of inhomogeneous spectral broadening

Finally we discuss the formation mechanism of carbon clusters at hot spots and size of the clusters by referring previous reports.12 , 17(c) , 18 , 41 Bjerneld et al investigated the mechanism of formation of carbon clusters.12They found the luck of threshold of excitation laser intensity for the formation of carbon clusters, indicating the mechanism of formation is a photochemical process, not photothermal one The DFT analysis of SERRS spectra in Fig.2(a)-2(c)of the transient species may make clear the detailed photochemical process We estimated the size of cluster using be Refs.17(c),18and41 The typical ΓGca.50 cm−1and the unclear 2D lines of the current exper-iments indicate that the size of the cluster is within several nanometers Indeed, DFT calculation

of such size of carbon graphite sheets (4 × 4 nm2

) exhibit G line around 1580 cm−1, supporting the estimation Such graphite sheet is composed of 32 carbon atoms and a dye molecule contains

28 carbon atoms Thus, we consider the size of cluster is several nanometers and one or two dye molecules can form the carbon clusters Indeed, such observation of carbon clusters is very rear under single molecule conditions

127113-9 Itoh et al. AIP Advances 5, 127113 (2015)

IV SUMMARY

In the present work, photo-conversion of dye molecules into sp2carbon clusters in hotspots

of Ag NP dimers is evaluated The appearance of G and D lines in the SERRS spectra shows the photo-generation of sp2carbon clusters from near-single dye molecules Fluctuations of the G and D lines suggest changes in the crystallite size of clusters within several angstrom units The spectral broadening with line splitting of sp2 carbon clusters is explained in terms of generation and elimination of various defects, including topological defects SERRS is the only experimental way to directly evaluate such local spectral properties, which can be theoretically predicted by i.e DFT calculation Thus, the present work demonstrates the usefulness of SERRS spectros-copy for the directly correlating dynamics in sp2carbon systems with vibrational properties at the sub-nanometer scale at SERRS hotspots

ACKNOWLEDGEMENTS

The current study was supported in part by Wakate B (No 26810013) and Kiban B (No 26286066) grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and by an A-STEP Grant (No AS2525017J) from the Japan Science and Technology Agency (JST)

1 M Inoue and K Ohtaka, J Phys Soc Jpn 52, 3853 (1983); E Hao and G C Schatz, J Chem Phys 120, 357 (2004); P Johansson, H Xu, and M Käll, Phys Rev B 72, 035427 (2005); E C Le Ru, E Blackie, M Meyer, and P G Etchegoin, J Phys Chem C 111, 13794 (2007); K J Savage, M M Hawkeye, R Esteban, A G Borisov, J Aizpurua, and J J Baumberg,

Nature 491, 574 (2012).

2 S M Nie and S R Emery, Science 275, 1102 (1997); K Kneipp, Y Wang, H Kneipp, L Perelman, I Itzkan, R Dasari, and M Feld, Phys Rev Lett 78, 1667 (1997).

3 M Moskovits, Rev Mod Phys 57, 783 (1985).

4 Y S Yamamoto, Y Ozaki, and T Itoh, J Photochem Photobio C 21, 81 (2014).

5 C M Galloway, P G Etchegoin, and E C Le Ru, Phys Rev Lett 103, 063003 (2009); T Itoh, Y S Yamamoto, H Tamaru,

V Biju, N Murase, and Y Ozaki, Phys Rev B 87, 235408 (2013).

6 M Moskovits and D P DiLella, J Chem Phys 73, 6068 (1980); M Takase, H Ajiki, Y Mizumoto, K Komeda, M Nara,

H Nabika, S Yasuda, H Ishihara, and K Murakoshi, Nat Photonics 7, 550 (2013).

7 A E Schlather, N Large, A S Urban, P Nordlander, and N J Halas, Nano Lett 13, 3281 (2013); F Nagasawa, M Takase, and K Murakoshi, J Phys Chem Lett 5, 14 (2014); T Itoh, Y S Yamamoto, H Tamaru, V Biju, S Wakida, and Y Ozaki,

Phys Rev B 89, 195436 (2014).

8 For example; Y Maruyama, M Ishikawa, and M Futamata, J Phys Chem B 108, 673 (2004); S M Stranahan and K A Willets, Nano Lett 10, 3777 (2010); Y Kitahama, Y Tanaka, T Itoh, and Y Ozaki, Phys Chem Chem Phys 12, 7457 (2010).

9 For example; Z Wang and L J Rothberg, J Phys Chem B 109, 3387 (2005); T Chen, H Wang, G Chen, Y Wang, Y Feng, W S Teo, T Wu, and H Chen, ACS Nano 4, 3087 (2010).

10 For example; A Weiss and G Haran, J Phys Chem B 105, 12348 (2001); E Cortes, P G Etchegoin, E C Le Ru, A Fainstein, Maria E Vela, and R C Salvarezza, J Am Chem Soc 132, 18034 (2010).

11 For example; P Xu, L Kang, N H Mack, K S Schanze, X Han, and H.-L Wang, Sci Rep 3, 2997 (2013); K Kim, J.-Y Choi, and K S Shin, J Phys Chem C 118, 11397 (2014).

12 E J Bjerneld, F Svedberg, P Johansson, and M Kall, J Phys Chem A 108, 4187 (2004).

13 A Kudelski and B Pettinger, Chem Phys Lett 321, 356 (2000).

14 K F Domke, D Zhang, and B Pettinger, J Phys Chem C 111, 8611 (2007).

15 A C Ferrari and D M Basko, Nat Nanotechnol 8, 235 (2013).

16 T M G Mohiuddin, A Lombardo, R R Nair, A Bonetti, G Savini, R Jalil, N Bonini, D M Basko, C Galiotis, N Marzari, K S Novoselov, A K Geim, and A C Ferrari, Phys Rev B 79, 205433 (2009).

17 (a) M A Pimenta, G Dresselhaus, M S Dresselhaus, L G Cancado, A Jorio, and R Saito, Phys Chem Chem Phys.

9, 1276 (2007); (b) J H Chen, W G Cullen, C Jang, M S Fuhrer, and E D Williams, Phys Rev Lett 102, 236805 (2009); (c) L G Cancado, A Jorio, E H Martins Ferreira, F Stavale, C A Achete, R B Capaz, M V O Moutinho, A Lombardo, T S Kulmala, and A C Ferrari, Nano Lett 11, 3190 (2011); (d) A Eckmann, A Felten, A Mishchenko, L Britnell, R Krupke, K S Novoselov, and C Casiraghi, Nano Lett 12, 3925 (2012); (e) D L Matz, H Sojoudi, S Graham, and J E Pemberton, J Phys Chem Lett 6, 964 (2015).

18 K N Kudin, B Ozbas, H C Schniepp, R K Prud’homme, Ilhan A Aksay, and R Car, Nano Lett 8, 36 (2008).

19 S Eigler, F Hof, M Enzelberger-Heim, S Grimm, P Müller, and A.s Hirsch, J Phys Chem C 118, 7698 (2014).

20 C Casiraghi, A C Ferrari, and J Robertson, Phys Rev B 72, 085401 (2005); Z Luo, T Yu, Z Ni, S Lim, H Hu, J Shang,

L Liu, Z Shen, and J Lin, J Phys Chem C 115, 1422 (2011).

21 G Abrasonis, R Gago, M Vinnichenko, U Kreissig, A Kolitsch, and W Möller, Phys Rev B 73, 125427 (2006).

22 L Wang, J Zhao, Y-.Y Sun, and S B Zhang, J Chem Phys 135, 184503 (2011); A J Page, C P Chou, B Q Pham, H.

A Witek, S Irle, and K Morokuma, Phys Chem Chem Phys 15, 3725 (2013).