Surface‐Enhanced Electronic Raman Scattering at Various Metal Surfaces

Surface‐enhanced Raman scattering (SERS) has been extensively used to obtain vibrational information at metal/dielectric interfaces because of its extremely high chemical sensitivity and surface selectivity. Since the discovery of this effect in 1970s, however, the origin of the spectral background has remained questionable. Because the “background” signals dominate in the low‐frequency region, it is necessary to deeply understand this phenomenon for analyzing SERS spectra properly in the wide spectral range. Herein, scattering spectra for various metal surfaces are measured with and without excitation of plasmonic resonances in the wide frequency ranges covering both Stokes and anti‐Stokes branches, including very low‐frequency region. The comparison of the spectra between smooth surfaces and plasmonic rough surfaces of Au, Ag, and Pt and the theoretical evaluation of local fields at the metal/air interfaces indicate that electronic Raman scattering accounts for the background continuum in SERS spectra. This is also confirmed by the comparison of the spectra with and without ultrathin Ag films on a same Au surface. An analytical method of SERS spectra is demonstrated in which both contributions of electronic Raman scattering and vibrational Raman scattering are considered.

DOI: 10.1002/pssb.202100589 Surface-enhanced Raman scattering (SERS) has been extensively used to obtain vibrational information at metal/dielectric interfaces because of its extremely high chemical sensitivity and surface selectivity. Since the discovery of this effect in 1970s, however, the origin of the spectral background has remained questionable. Because the "background" signals dominate in the low-frequency region, it is necessary to deeply understand this phenomenon for analyzing SERS spectra properly in the wide spectral range. Herein, scattering spectra for various metal surfaces are measured with and without excitation of plasmonic resonances in the wide frequency ranges covering both Stokes and anti-Stokes branches, including very low-frequency region. The comparison of the spectra between smooth surfaces and plasmonic rough surfaces of Au, Ag, and Pt and the theoretical evaluation of local fields at the metal/air interfaces indicate that electronic Raman scattering accounts for the background continuum in SERS spectra. This is also confirmed by the comparison of the spectra with and without ultrathin Ag films on a same Au surface. An analytical method of SERS spectra is demonstrated in which both contributions of electronic Raman scattering and vibrational Raman scattering are considered.
In this work, scattering spectra for pure metal surfaces of Au, Ag, and Pt are compared with and without plasmonic enhancement. Au and Ag are the most common noble metals in the fields of SERS spectroscopy and plasmonics. [1,11] In contrast, Pt is known to be SERS-inactive because the plasmon resonance is strongly damped by localized d-electrons near the Fermi level of Pt. [12,13] Instead, owing to its d-electrons, this transition metal has been extensively used as a heterogeneous catalyst for various reactions such as the cathode of fuel cells. [14] In addition to these pure metals, atomically thin Ag films on Au surface are also studied in terms of the SERS background generation. At smooth surfaces of these metals where plasmon resonances are absent, characteristic features of ERS are clearly observed in their scattering spectra. At rough metal surfaces with broad plasmonic resonance features, the intensities of the scattered signals are enhanced in accordance with trends in local field enhancement at respective metal/air interfaces, which is a strong support that ERS is the dominant origin of the SERS background.

Sample Preparation
Polycrystalline beads of Au, Ag, and Pt were prepared by flame melting of the end of metal wires with the purity of 99.999% for Au, 99.95% for Ag, and 99.99% for Pt. Their half-cut beads were mechanically polished and extensively annealed to obtain mirror finished smooth surfaces. These surfaces were carefully cleaned using piranha solution for Au and Pt and hydrogen peroxide/ aqueous ammonia for Ag before spectroscopic use. SERS-active plasmonic surfaces of these metals were fabricated by electrochemical surface roughening procedures [12] using a function generator (WF1948, NF Corporation). The applied potential cycles are summarized in Table S1, Supporting Information. The obtained rough surfaces were extensively cleaned in boiling water. For Au, single crystalline beads were also prepared by slow cooling after flame melting of Au wires, which was followed by mechanical polishing and annealing of the half-cut beads so that (111) or (100) faces were obtained. [4] These defined surfaces were annealed in Ar-flow using an induction heater (EASYHEAT 0224, Ambrell, Ltd.) before use. Atomically thin Ag films were formed on both single-crystalline and polycrystalline surfaces of Au using the underpotential deposition (UPD) with the scan rate of 5 mV s À1 in 0.1 M H 2 SO 4 solution containing 1 mM AgNO 3 . [15,16]

Scattering Measurements
Scattering spectra for metal surfaces with and without surface roughness were measured using two different types of homebuilt inverted Raman microscope systems equipped with an objective of 0.6 N.A. and a CCD polychromator (PIXIS 400 BR & IsoPlane, Princeton Instruments). The sensitivity of the CCD polychromator was calibrated using a NIST traceable intensity calibration light source (IntelliCal, Princeton Instruments). The excitation energy dependence of scattering spectra was measured using two different light sources: 1.96 eV (632.8 nm) from a He-Ne laser (HNL100L, THORLABS) and 1.58 eV (785 nm) from a laser diode (NovaPro FIBER WL HiRes785-45 SM, RGB lasersysteme GmbH), and also using conventional Raman filters (RazorEdge longwave-pass edge filters, Semrock) to remove Rayleigh scattered light. For comparison of scattering responses in the anti-Stokes and Stokes branches, ultranarrow band notch filters (reflecting volume Bragg grating, OptiGrate Corp) was used to remove the Rayleigh scattering under excitation of 1.96 eV (NEO-30MS1, NEOARK). [8,17] 2.3. Analytical Procedure for ERS Spectra The absolute wavenumber (ν > 0 in cm À1 ) and the relative wavenumber (ν Δ in cm À1 ) were used in the following description by considering the connection to traditional wavenumber plotting. A power spectrum for ERS, I ERS ðν Δ Þ, was connected with the imaginary part of the dynamic susceptibility, χ 00 ERS ðνÞ, by the following expressions [8,9] for anti-Stokes branch and for Stokes branch of the spectrum. Here,ν 0 is the photon energy for the incident laser (ν 0 ¼ 1/632.8 nm ¼ 15802.8 cm À1 [1.96 eV] or 1/785 nm ¼ 12738.9 cm À1 [1.58 eV] in the present case).ν is the energy exchanged between photon and electron, i.e., the energy difference between the initial and final states of scattered electrons in the conduction band of the metal.ν Δ as andν Δ s were the Raman shifts relative toν 0 in the anti-Stokes and Stokes branches, i.e.,ν Δ as ¼ Àν andν Δ s ¼ν. The cubic terms of the scattered photon energies, ðν 0 þνÞ 3 and ðν 0 ÀνÞ 3 , expressed the scattering efficiency factors for the anti-Stokes and Stokes processes, respectively; the cubic weighting was valid when the signals are measured using a photon counter. K is an instrument function. χ 00 ERS is determined by integrating the thermally occupied electronic states, which are described by the Fermi-Dirac function, over the distribution of excited wave vectors, resulting in a source for the broad continuum in the ERS spectra.
[nðνÞ] and [nðνÞ þ 1] correspond to thermal factors for the anti-Stokes and Stokes processes, respectively, and nðνÞ is the Bose-Einstein distribution described as nðνÞ ¼ ½expðhcν=k B TÞ À 1 À1 , where h, c, and k B are Planck's constant, the velocity of light and the Boltzmann constant, respectively. The microscopic origin of the general Bose-Einstein weighting for ERS is related to the thermal occupation of photon states. From Equation (1) and (2), the well-known relation between Stokes and anti-Stokes branches is obtained by neglecting the small difference in the scattering efficiency factors. [18]

Estimation of Local Field Enhancement
The enhancement factor (EF) of SERS intensities due to the electromagnetic effect is known to be roughly proportional to the fourth power of local field enhancement [1] www.advancedsciencenews.com www.pss-b.com where E loc and E 0 denote the amplitude of the local and freespace electric fields, respectively. Here, we assumed both ERS and VRS gain intensity from propagating surface plasmons (SPs) at a metal/air interface for respective metals; surfaceenhanced ERS (electronic SERS) and surface-enhanced VRS (vibrational SERS) are hereinafter referred to as ESERS and VSERS, respectively. The complex wavevector of propagating SPs at a smooth metal/dielectric interface ( m 2 by assuming that the dielectric loss is small enough. [19] k 0 is the wavevector of photon in free space, ε d is the dielectric constant for the dielectric layer, and ε ' m and ε '' m are real and imaginary parts of the dielectric constant for a metal. To deal with highly damping Pt, however, the contribution of the dielectric loss must be treated rigorously using [20] The propagation length of SP is written as Here, we assumed the normal incidence and perfect coupling of the external field into the SP mode to determine the maximum possible enhancement in the calculation. Based on the conservation of energy between the incoming power and the energy loss of SP, we obtain for ESERS and for VSERS. Here, E m loc and E d loc are the local fields in the metal and dielectric phases of the interface, respectively. q 1 ≡ k 0 and q 2 ≡ k 0 Àε m ðÀε d Àε m Þ 1=2 correspond to the decay constants in the metal and dielectric phases, respectively. In general, EF VSERS is much larger than EF ESERS because of the difference in the decay constants. Besides, the longer SP propagates, the larger EF becomes, resulting that the EFs are larger in Au and Ag than in Pt.

Scattering Spectra for Smooth Surfaces of Pure Metals
In the absence of plasmon resonances, scattered signals under moderate CW excitation are expected to be generated by the ERS process when the excitation energy is smaller than the interband transition energy of metal; the contribution of nonlinear optical effect or near-field effect is excluded in this case. Figure 1a shows Stokes branches of scattering spectra for the smooth surfaces of pure Au, Ag, and Pt obtained using two different excitation energies (for the conventional energy plotting, hν in eV is used instead ofν in cm À1 ). The signal intensities are normalized by the scattering efficiency factor, ðν 0 ÀνÞ 3 , and the incident photon numbers, I 0 =hcν 0 , where I 0 denotes the incident (a) (b) Au Ag Pt Au Ag Pt Figure 1. a) Scattering spectra for smooth surfaces of pure Au, Ag, and Pt, measured using two different excitation energies: 1.58 and 1.96 eV. The signal intensities were normalized by the scattering efficiency factor, ν 3 ¼ c 3 ðν 0 ÀνÞ 3 , and the incident photon numbers, I 0 =hν 0 ¼ I 0 =hcν 0 . b) Reflectance spectra calculated for normal incidence to Au, Ag, and Pt surfaces using the dielectric constants of respective metals. [26] www.advancedsciencenews.com www.pss-b.com light intensity. Note that the appearance of fringes below 1.5 eV is due to the etaloning in the CCD detector, which was not taken into account in the instrument factor K. For the 1.58 eV excitation (785 nm), the signal intensities for Au and Ag were much smaller than that for Pt. This is because the Fermi surfaces in these noble metals are much more spherical compared with that in Pt [21] ; light-induced density fluctuations are effectively screened by the long-range Coulomb interaction in such electronic system, resulting in the reduction of the cross sections. (Although the scattering volume, governed by the penetration depth of metal, also affects the cross section, this contribution seems to be negligible as shown in Figure S1, Supporting Information.) For the 1.96 eV excitation (632.8 nm), on the other hand, the largest signal intensity is found in Au. This is apparently because the excitation energy is pre-resonant to the sp-d interband transition of Au; as shown in Figure 1b, the reflectance of Au significantly decreases above the excitation energy. (In contrast, the low reflectivity of Pt in the entire visible range is due to nonresonant dielectric loss, meaning that there is no resonant enhancement of ERS for Pt.) The correlation between the anti-Stokes and Stokes branches of scattering spectra were examined using the 1.96 eV excitation, because of the experimental difficulty in measuring the anti-Stokes branch with extremely weak signal intensities under 1.58 eV excitation. Figure 2a shows the raw scattering spectra for smooth surfaces of pure Au, Ag, and Pt measured using the ultranarrow band notch filters. The anti-Stokes intensities are rapidly decreased in all metals, which is a typical behavior in Raman scattering. As is known, the anti-Stokes and Stokes processes in Raman scattering are connected each other via the time-reversal symmetry. [18] Therefore, when the measured scattering signals are indeed originated from ERS, both the anti-Stokes and Stokes branches must provide same χ 00 ERS by reducing the Bose-Einstein weighting, according to Equation (1) and (2). [9] This is indeed confirmed in Figure 2b; the symmetry of the response between the Stokes and anti-Stokes branches is well verified in all metals when the local temperature of 300 K was assumed for calculating the Bose-Einstein weighting. It is evident that the scattering signals from smooth metal surfaces in the absence of plasmon resonances are originated from the ordinary ERS process.

Scattering Spectra for Rough Surfaces of Pure Metals
It is nontrivial whether the SERS background is dominantly originated from plasmon-enhanced ERS, although the scattering signals from smooth metal surfaces can be successfully ascribed to the ERS process. Indeed, there have been the extensive contentions about the origin of SERS background by considering the other possibilities such as plasmon-induced PL. [5,10] To examine which mechanism is dominant, SERS spectra for rough surfaces of metals were measured using the ultranarrow band notch filters under the 1.96 eV excitation (632.8 nm). In general, one of the technical difficulties for comparing SERS properties between various metals is to control the photon-SP coupling efficiency. In this work, we have employed two different approaches. One is to compare SERS spectra on a same geometric metal surface with and without atomically ultrathin foreign metal films, which will be discussed later. The other is to use a rough metal surface consisting of various nanostructures with different sizes and shapes. This system gives rise to broad plasmon resonances in the wide spectral range, which can reduce the difference in the photon-SP coupling efficiency. Figure 3 shows the surface morphology of the electrochemically roughened metal surfaces used in the SERS measurements. Although the morphological properties are different among these surfaces, the inhomogeneous nanostructures made the plasmon resonances broaden. For the electrochemically stable Au and Pt, the roughness factors were estimated to be 6.4 for Au and 13.8 for Pt using surface oxide formation in sulfuric acid solution ( Figure S2, Supporting Information). For Ag, the roughness factor was estimated to be 18.6 using the underpotential deposition of Pb ( Figure S3, Supporting Information). [22] It is noted that the results obtained by the different methods are comparable because both methods directly measure the number of metal  8 nm). b) χ 00 ERS spectra obtained from the raw scattering spectra by reducing the Bose-Einstein weighing for 300 K and the scattering efficiency factor using Equation (1) and (2).
www.advancedsciencenews.com www.pss-b.com atoms exposed to the surface using surface-specific reactions.
The plasmonic resonance properties of these surfaces are presented in Figure S4, Supporting Information. Figure 4a shows raw SERS spectra for these roughed surfaces of pure metals, which are presented in the log-scale due to the large difference in intensity. The small vibrational features indicated by the dotted circle is due to unavoidable carbonaceous contaminants in ambient conditions. The vibrational peak at 570 cm À1 in the SERS spectrum of Pt is due to residual surface oxides [23] formed in the electrochemical roughening process (as shown in Figure S2b, Supporting Information, most of the surface is still metallic). For the spectral continuum, Ag gives the largest SERS background unlike in the case of ERS on smooth surface. It is noted that the spectral background intensities decay quasiexponentially in the anti-Stokes branch, which is typically seen when the signals are originated from Raman scattering processes. [24] This implies that the origin of SERS background in Ag and Pt is also ascribed to ERS signals rather than PL, just like in the case of Au. [6][7][8] To examine this, the ratios of the SERS intensity to the ERS intensity (I SERS =I ERS ) for respective metals with and without surface roughness were calculated from Figure 2a and 4a, as shown in Figure 4b. The values of I SERS =I ERS , i.e., the apparent EFs, are roughly in the order of 10 4 for Ag, 10 3 for Au and 10 1 for Pt, except for the low-Raman shift (ν Δ ) region below 200 cm À1 , where additional enhancement is seen. These apparent EFs include not only the plasmonic enhancement but also the increase of the surface area due to the surface roughening. Indeed, the EF value obtained for nonplasmonic Pt is close to the roughness factor of the Pt surface, indicating that the plasmonic enhancement is weak in this case. In contrast, the EF values for plasmonic Au and Ag are apparently larger than the roughness factors of these surfaces, suggesting the contribution of the plasmonic enhancement of ERS signals. Accordingly, the purely plasmonic contributions to EFs for ERS, i.e., the apparent EFs divided by the roughness factor, (I SERS =I ERS =r), are estimated to be in the order of 10 3 for Ag, 10 2 for Au and 10 0 for Pt (more quantitative analysis of the plasmonic effect will be provided later). For the additional enhancement seen in the low Raman-shift region below 200 cm À1 , the plausible mechanism is nonplasmonic modification of χ 00 ERS due to the surface roughness because this feature is observed even in the highly damping Pt. There are two different mechanism for the surface roughness induced ERS: weakening of the screening of light-induced charge Au Ag Pt    8 nm). b) The ratios of the SERS intensity to the ERS intensity for respective metals with and without surface roughness, which were calculated using the raw spectra shown in Figure 2a and 4a.
www.advancedsciencenews.com www.pss-b.com fluctuation [9] and momentum transfer from the microscopic surface roughness. [2b] Both effects can relax the usual momentum conservation rule, resulting in the enhancement of cross section for ERS. Indeed, when SERS spectra are measured in situ during the surface roughening process of Au, the rapid increase of the signal intensity was observed in the low Raman-shift region prior to the evolution of SERS activity. [8] The correlation between the experimentally obtained plasmonic enhancement (I SERS =I ERS =r) and the local field enhancement was evaluated quantitatively by calculating the field enhancement at the respective metal/air interfaces. According to Equation (6), the propagation lengths of SP at a planar metal/air interface for the photon wavelength of 632.8 nm in free space are calculated to be 64.6 μm for Ag, 9.8 μm for Au, and 2.6 μm for Pt. Using Equation (7), then, EF ESERS ð% jE m loc ðν 0 Þ=E 0 ðν 0 Þj 4 ) is calculated to be 335 for Ag, 33 for Au, and 0.2 for Pt. (Similarly, EF VSERS is estimated to be 10 5 for Ag, 5000 for Au, and 180 for Pt using Equation (8).) These values can be a good indicator to evaluate trends in the relation between the dielectric loss of metal and the plasmonic effect even though the excited SPs are scattered by surface roughness in the practical SERS systems. Figure 5 shows the plots of the purely plasmonic EF (¼ I SERS =I ERS =r) versus the theoretically calculated EF ESERS . Here, the values for I SERS =I ERS were taken at 1.93 eV, where the non-plasmonic enhancement of χ 00 ERS is absent. Clearly, the observed EFs are related to the theoretical values of EFs, strongly suggesting that the ERS accounts for the SERS background.

Scattering Spectra for Atomically Thin Ag Films on Au
The contention that the SERS background is originated from plasmon-enhanced ERS signals is further confirmed by measuring scattering spectra for atomically thin Ag films formed on both smooth and rough Au surfaces. [15] This ultrathin Ag film, fabricated using the UPD technique (see Figure S3, Supporting Information), is referred to as AgUPD. In this experiment, the metal dependence of the SERS background can be evaluated for the same surface morphology.
When AgUPD is formed on a smooth Au surface (AgUPD/ Au), the ERS intensity should decrease because the cross section of ERS at 1.96 eV is much lower in Ag than in Au. This is indeed seen in Figure 6a. Such intensity decrease was similarly observed on Au(111), as shown in Figure S5, Supporting Information. According to the previous report on surface X-ray analysis, the  . a) Raw scattering spectrum for atomically thin Ag layers on smooth Au (AgUPD/Au), in addition to those for smooth surfaces of pure Au and Ag, measured using the ultra-narrow band notch filters under the 1.96 eV excitation (632.8 nm). b) Raw SERS spectrum for AgUPD on rough Au, in addition to those for rough surfaces of pure Au and Ag, measured using the ultranarrow band notch filters under the 1.96 eV excitation (632.8 nm). c) The ratios of the SERS intensity to the ERS intensity for respective metals with and without surface roughness, which were calculated using the raw spectra shown in (a,b).
www.advancedsciencenews.com www.pss-b.com AgUPD/Au(111) maintains the atomic bilayer structures even under ambient conditions. [16] Thus, the intensity decrease shown in Figure 6a is ascribed to the contribution of the atomically thin Ag films, even though the penetration depth of photons is much larger than the thickness of AgUPD (see Figure S1, Supporting Information). This indicates that ERS is highly surface-sensitive. Next, when AgUPD was formed on a SERS-active rough Au surface, the induced change was rather opposite from the result on the smooth surface; the spectral background intensities on the rough Au were enhanced by formation of AgUPD, as shown in Figure 6b. This is reasonably explained by the fact that the dielectric loss of SP was decreased by the presence of the AgUPD. Figure 6c shows that the spectral profile of the I SERS =I ERS for AgUPD/Au is rather similar to that for pure Au, suggesting that the background continuum is originated from the plasmonic enhancement of ERS signals generated from the Au underlayer. If the PL is the dominant origin of the SERS background in Ag surfaces, the spectral profile for the AgUPD/Au should be much more close to that of Ag. Only the plasmon-enhanced ERS mechanism can properly explain these behaviors observed on various metal surfaces with and without surface roughness.

Enhanced Signals of ERS and VRS in SERS Spectra
SERS is typically used to obtain chemical information at metal/ dielectric interfaces such as electrode/electrolyte interfaces. According to the results demonstrated in the above sections, SERS spectra should consist of discrete peaks of VSERS generated in the dielectric phase of the interface and a background continuum of ESERS generated in the metal phase of the interface. Given that both ERS and VRS are inelastic light scattering processes by different Raman scatterers, the relation between I ERS and χ 00 ERS in Equation (1) and (2) are applicable for the relation between I VRS and χ 00 VRS ; while χ 00 ERS is determined by integrating the thermally occupied electronic states, χ 00 VRS is given by the vibrational states of polyatomic molecules. On the other hand, the EFs are different between ESERS and VSERS as shown in Equation (7) and (8). EF VSERS is typically several orders of magnitude larger than EF ESERS due to the rapid decay of the electric field in the metal phase. Plasmonic enhancement of ERS, i.e., ESERS, is described as for anti-Stokes branch and for Stokes branch. Similarly, plasmonic enhancement of VRS, i.e., VSERS, is described as for anti-Stokes branch and for Stokes branch. Here, g m and g d denote the coupling efficiencies of Raman transitions to the plasmonic cavity in the metal phase and dielectric phase, respectively; g d explains the surface selection rules because the coupling efficiency is affected by the orientation and the position of molecular dipoles in the cavity. On the other hand, g m would be less sensitive toν because free electrons are the source of the signals. For ESERS, therefore, the experimentally obtained I ESERS =I ERS , shown in Figure 4b, corresponds to the Purcell factor as follows Now, when the Raman frequency shiftν is small, one can expect the ratio of EF VSERS to EF ESERS is nearly constant: EF VSERS =EF ESERS % C. Then, we obtain for anti-Stokes branch and for Stokes branch, where χ 00 SERS ðνÞ ≡ χ 00 ERS ðνÞ þ C ⋅ g d ⋅ χ 00 VRS ðνÞ. To examine the validity of Equation (14) and (15), we have measured SERS spectra for monolayers of 4-methylbenzenethiol (MBT) on SERS-active rough surfaces of Au and Ag. Figure 7a and 8a show raw SERS spectra of MBT measured on Au and Ag, respectively. In both surfaces, distinctive Raman bands of MBT are clearly observed together with the background continuum in the Stokes branches, and these signals rapidly drop in the anti-Stokes branches. [8,17] Their reduced spectra, i.e., χ 00 SERS spectra of MBT on Au and Ag are presented in Figure 7b and 8b, respectively, indicating that the symmetry of the response between the Stokes and anti-Stokes branches is well verified in both the vibrational peaks and the electronic continuum. (The detailed procedure for obtaining χ 00 SERS is presented in our previous work. [8a] ) As for the Au surface, the symmetry of the response becomes low in the high frequency www.advancedsciencenews.com www.pss-b.com region; for example, the stretching mode of CS at 1079 cm À1 has different intensities in the Stokes and anti-Stokes branches. This is because the assumption of EF VSERS =EF ESERS % C is not appropriate near the edge of the interband transition of Au. For the Ag surface, on the other hand, the symmetry of the response is verified in the entire frequency range. Therefore, under electronic non-resonant excitation condition, Equation (14) and (15) are widely applicable for analyzing SERS spectra. The conversion of raw SERS spectra into the χ 00 SERS spectra is particularly effective to extract low-frequency vibrational information; since the ESERS signals are extremely strong in the low frequency region, the low-frequency vibrations such as δAuS-Ph or νAg-S are not clearly observed in the raw SERS spectra, which are unveiled in the χ 00 SERS spectra.

Conclusion
The origin of the SERS background has been extensively studied on various metal nanostructures using both the moderate CW excitation and the intense pulsed laser excitation. In this work, we focused on analyzing the background generation under moderate CW excitation, which is a typical condition in conventional SERS spectroscopy. Among various models proposed for the origin of the background, only ERS must have the symmetry of the response between the Stokes and anti-Stokes branches, because the Stokes and anti-Stokes transitions are connected with the time-reversal symmetry. In this work, scattering spectra were measured on the atomically thin Ag films on Au as well as on pure metal surfaces of Au, Ag, and Pt with and without plasmonic surface roughness. All of the measured spectra clearly exhibited  www.advancedsciencenews.com www.pss-b.com such a correlation between the Stokes and anti-Stokes branches in the wide frequency range covering 10 <ν < 1500 cm À1 , indicating that ERS accounts for the background generation in SERS spectra. The theoretical calculations of the local fields were also conducted for both sides of the interface to evaluate the plasmonic enhancement of ERS and VRS. For Ag and Au, the significant plasmonic enhancement effect was confirmed for both ERS and VRS. For highly damping Pt, the plasmonic effect for enhancing ERS was negligible while that for VRS was still available. The fact that ERS is the dominant source of the SERS background means that the generation of the spectral background is an intrinsic phenomenon of SERS. Therefore, the depression of the background intensity is essentially difficult, except for the ultralow temperature spectroscopy. One possible method to suppress the background intensity without losing VSERS sensitivity would be to use the so-called long-range surface plasmon (LRSP) mode. [25] In this double-interface SP mode excited at both sides of a thin metal film, the mode field is expelled to the dielectric phase of the interfaces, resulting in the decease of the plasmonic enhancement of ERS in the metal phase. From the viewpoint of vibrational spectroscopy, the extraction of Raman susceptibility from measured SERS spectra is also the practical way to perform vibrational analysis without suffering from the background continuum, as shown in Figure 7 and 8. Since the ERS behavior is well described as resulting from the thermal population of photon states and the joint density states of electron-hole excitation in a metal, the reduction of the Bose-Einstein weighting from measured SERS spectra considerably improve the analyzability of vibrational information especially in the low frequency region below 200 cm À1 .

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.