Can Mixed-Metal Surfaces Provide an Additional Enhancement to SERS?

We explore the chemical contribution to surface-enhanced Raman scattering (SERS) in mixed-metal substrates, both experimentally and by computer simulation. These substrates are composed of a chemically active, transition-metal overlayer deposited on an effective SERS substrate. We report improved analytical enhancement factors obtained by using a small surface coverage of palladium or platinum over nanostructured silver substrates. Theoretical predictions of the chemical contribution to the surface enhancement using density functional theory support the experimental results. In addition, these approaches show that the increased enhancement is due not only to an increase in surface coverage of the analyte but also to a higher Raman scattering cross section per molecule. The additional chemical enhancement in mixed-metal SERS substrates correlates with the binding energy of the analyte on the surface and includes both static and dynamical effects. SERS using mixed-metal substrates has the potential to improve sensing for a large group of analyte molecules and to aid the development of chemically speciﬁc SERS-based sensors.


Introduction
][8][9][10] The enormous enhancement of the ordinarily weak Raman scattering cross sections in SERS is attributed to two main effects: electromagnetic and chemical enhancement.Electromagnetic enhancement is due to the presence of strong electromagnetic fields at the surface of nanoscale metal particles and is responsible for enhancements of Raman scattering cross sections of 10 or more orders of magnitude. 6,72][13][14][15][16] While being much smaller in absolute magnitude than the electromagnetic effect, chemical enhancement has important implications for SERS as an analytical tool, and can provide up to 2-3 orders of magnitude of enhancement in addition to the electromagnetic contribution.Furthermore, it affects Raman-active vibrations selectively and thus alters the shapes of Raman spectra in addition to providing enhancement. 17Lastly, the chemical effect is chemically specific and has potential for improved performance of SERS in analytical applications.8][19][20][21] Chemical enhancement can be modeled as a combination of a static contribution connected to the redistribution of electron density between the metal and the molecule and a dynamical contribution arising from resonance enhancement due to surface excited states of mixed metal-molecular character. 16ld and silver are the most commonly used substrate materials for SERS applications.Ag has been found to give highest enhancements of Raman cross sections for an excitation in the visible range, while Au is often used because of simpler processing and resistance to oxidation.Recently, Wang et al., have created nanostructured silicon and germanium-based SERS substrates. 22,23For these structures, the enhancement is in the order of 10 2 -10 4 .Its enhancement mechanism relies on a photon-induced charge transfer process instead of the typical plasmon-based approach.Although this effect is by no means as intense as with Ag and Au, it provides evidence that these materials can provide a SERS signal despite the lack of a plasmonic properties.
Several other metals have been explored with regard to surface enhancement. 24Differences in the localized surface-plasmon resonances (LSPR) in scattering and absorption spectra of these metals relative to the frequencies of the incident and scattered radiation result in vastly different electromagnetic enhancements. 4On the other hand, the richness of chemical reactivity of transition metals opens the prospect of chemically selective binding and detection.
0][31][32][33][34][35][36][37][38] Their work was motivated by the concept of SERS borrowing, originally put forward by Van Duyne and Haushalter. 39ERS borrowing refers to the fact that dopant metals are poor SERS substrates by themselves.However, when deposited as an overlayer atop effective SERS substrates to probe catalytic reactions, or as a protective overlayer, they acquire some surface enhancement "borrowed" from the coinage metal.If only the electromagnetic enhancement is taken into account, one expects the enhancement factors for these mixed-metal substrates to be lower than for pure coinage metals.The attenuation of the surface enhancement has been described in terms of the distance dependence of the electromagnetic fields and has been broadly explored.29,[40][41][42][43] The above picture neglects the effect of the strong coordinative bonds formed between the dopant metal and specific analytes, e. g., pyridine.Therefore, the objective of this work is to explore, both experimentally and using simulation, the chemical contribution to surface enhancement in mixed-metal SERS substrates.
We expect that stronger chemical bonding in suitably chosen mixed-metal substrates, coupled with existing advanced nanofabrication techniques, can help to balance or in ideal cases even overcome the attenuation of electromagnetic enhancement.In this case, we propose and coin the term of an additional SERS lending effect, where the transition-metal overlayer could contribute to the overall surface enhancement due to the strong chemical bonding.SERS lending would open an avenue for new analytical applications of SERS.Well-known and recent results from surface chemistry [44][45][46][47][48] may serve as an inspiration for designing new mixed-metal SERS substrates.In particular, transition metals of the nickel group are known to form strong coordinative bonds with nitrogen-containing molecules and are promising candidates for their detection with SERS.In contrast, gold and silver substrates tend to bind more strongly to thiols.[32][35][36][37] Furthermore, molybdenum-covered SERS substrates could possibly be used for improved chemical bonding to oxygen-containing molecules.
The effect of stronger chemical bonding between the analyte and the metal surface on the Raman signal is two-fold.A stronger binding affinity results in a higher concentration of the analyte on the surface, which yields higher analytical enhancements.In addition, stronger coupling between the electronic structure of metal surface and vibrations of the molecule leads to an en-hancement of the Raman cross section per molecule.These effects have to be balanced with the damping of the plasmonic resonance due to the presence of the dopant metal as well as changes in surface morphology in mixed-metal SERS substrates.A complete investigation of all these effects presents a number of challenges.In this work, we explore the chemical enhancement in mixedmetal clusters using theoretical models to describe the coupling between the analyte and the metal surface.As was shown in previous studies, chemical enhancement in SERS depends mainly on the immediate chemical environment of the binding site and can be successfully modeled using cluster models. 16,49 the following, we present Raman scattering calculations of trans-1,2-bis(4-pyridyl) ethylene (BPE) on silver-based mixed-metal clusters to theoretically characterize the chemical contribution to surface enhancement.We report experimental measurements of BPE on Ag substrates with nanometer-thick overlayers of Pd or Pt.Further computational results predict that similar effects are possible with Au as majority metal and Pd or Pt as dopant metals.Finally, we study the electronic effects of mixed-metal SERS by analyzing the Raman excitation profile of pyridine (Py) bound to silver in order to explore the dynamical contribution to the chemical enhancement in mixed-metal SERS substrates.

Experimental Methods
SERS Substrate Fabrication.Experimental Raman spectra are collected from BPE adsorbed onto a femtosecond laser nanostructured SERS substrate, using the setup previously reported on Ref.50.This type of substrate has been previously shown to provide SERS enhancement factors of 10 9 , relative to pure analyte. 50,51All substrates were fabricated using a femtosecond laser structuring process on a n-type silicon (100) wafer (r=0.005-0.020W•cm).A pulse train from a regeneratively amplified titanium:sapphire laser was used to generate 800 nm center wavelength, 100 fs pulses at a repetition rate of 1 kHz.This pulse train was frequency-doubled to a center wavelength of 400 nm using a thin BiBO 3 crystal.The second harmonic pulse width exiting the crystal was less than 200 fs.These laser pulses were loosely focused with a plano-convex lens to achieve an average fluence of 10 kJ m 2 at the surface of a silicon wafer fastened to the inside of a 10 mm deep cuvette filled with deionized water.The cuvette was mounted on a computer-controlled two-axis translation stage and rastered at an appropriate speed such that each point on the silicon wafer was subjected to approximately 500 pulses.
To render the surfaces SERS-active, silver was thermally evaporated onto the structured silicon.
Using a quartz crystal microbalance to measure the thickness of a planar continuous film, substrates were fabricated with 80 nm of deposited silver.All films were thermally evaporated at a rate of 0.15 nm s 1 , with no heating or cooling applied to the substrate during deposition.A subsequent layer of 1-4 nm of platinum or palladium is added to probe the mixed-metal effects.The metalcoated substrates were submerged in 10-µM solutions of BPE made with ethanol for 1 hour and then gently rinsed in ethanol for 1 minute, followed by drying under a stream of nitrogen.
Raman Spectroscopy.Using a 5 mW, s-polarized HeNe laser, spectra were recorded through a 10⇥ microscope objective (0.25 NA) and projected onto a thermoelectrically cooled chargedcouple device array using 1200 mm 1 diffraction grating.Individual spectra were recorded from both single spots (1.6 µm diameter) on the substrate, and from a 500 µm thick cell of neat BPE for normalization.
Electron Microscopy.Secondary electron microscope images of the substrates were collected using a field-emission gun scanning electron microscope, utilizing an accelerating voltage of 5 kV.
No additional sample preparation was performed prior to imaging the substrates.

Theoretical Methods
Computational Details.Density-functional calculations were performed using the PBE0 functional. 52Triple-z valence basis sets with one set of polarization functions (def2-TZVP) and split-z basis valence sets with polarization (def2-SVP) were used for the main group and transition metal elements, respectively. 53Relativistic effective core potentials (ECP) comprising 28 core electrons for Ag and Pd and 60 core electrons were employed for Au and Pt, respectively. 54The PBE0 functional was chosen because it has proven to be accurate for polarizabilities 55,56 and Raman intensities, 57,58 although it is known to overestimate vibrational frequencies 59 and electronic excitation energies. 60The TURBOMOLE package 61 was used to perform all calculations.
Raman Scattering Calculations.Raman intensities were computed using the static polarizability limit, except for the Raman intensities calculated for the Raman excitation profiles. 57No scaling of vibrational frequencies was applied.The Raman spectra were simulated by Lorentzian broadening of the line spectra using an empirical linewidth of 5 cm 1 .A scattering angle of 90 degrees and perpendicular polarization of both incident and scattered radiation was employed.
Geometry Optimization and Energy Calculations.The molecular systems were relaxed using the PBE0 functional.Multiple spin geometries were studied and singlets (doublets in the open-shell case) were most stable.The molecule-cluster systems presented an imaginary mode, even after repeated attempts to relax the molecule.This mode was prevalent at the 3-15 cm 1 region.None of the bare clusters or molecules presented imaginary frequencies.Cartesian coordinates of the minimized structures are provided in the Supporting Information.We performed binding energy calculations.We neglected the basis set superposition error as it has been previously estimated to be 0.03 eV for the binding of pyridine with a Ag 20 cluster. 49

I. Enhancement effects of BPE using mixed-metal substrates
We prepared silver-based SERS substrates including a thin overlayer of Pd or Pt.Details regarding the nanofabrication of these surfaces can be found in the Experimental Methods Section.Silicon surfaces were roughened by a femtosecond laser nanostructuring process. 50An 80 nm thick layer of silver was deposited on the roughened surfaces by thermal evaporation; subsequently, an overlayer of Pd or Pt of 1-4 nm thickness was created by the same technique.In Figure 1a-b, we show scanning electron microscope (SEM) images of the silver substrate with 2 nm Pt coverage.
We notice a similar quasi-uniform distribution of raised features as shown in the work by Diebold et al.; 50 even in the presence of an additional transition metal.
Thermal evaporation is a highly directional process that preferentially coats the top surface of the sample and may leave shadows exposed to Ag.There have been previous efforts to ensure a pinhole free surface in an electrochemical setting by Weaver et al. 30 We currently do not explore these or other possible chemical methods. 33The current setup to identify elemental composition of surface coverage employs the SEM setup coupled with energy-dispersive X-ray spectrometry (EDS).The latter has a micron-sized resolution, so other ways of determining surface coverage, such as chemical probes are needed. 29 elucidate the role of the chemical contribution to surface enhancement in bi-metal substrates, we calculated the Raman cross sections of BPE bound to small-cluster models of SERS substrates.We use seven-atom (Ag 6 M•BPE) and a larger 18-atom cluster (Ag 14 M 4 •BPE) models, see Figure 1c.Chemical enhancement effects appear to be well localized around the site of the chemical bond between metal atoms and the analyte molecule.In previous work, these have been shown to provide realistic models for chemical enhancement. 16,17e experimental Raman spectrum of BPE at 633 nm excitation wavelength on a silver substrate and the computed Raman spectra for the Ag 7 •BPE and Ag 18 •BPE complexes are compared in Figure 2. The vibrational frequencies and relative Raman cross sections are in good agreement between the experimental and the computed spectra.In Table 1 we present the five most prominent normal modes of the bare BPE molecule as shown in the Supporting Information and reported elsewhere.62,63 As BPE interacts with the metal, there is an shift of its normal modes.Also, the peak present at 1232 cm 1 , related to scissor C-H vibration in the ring, has a large decrease of its relative intensity.The four most prominent vibrational modes in the surface-enhanced setup are clearly identified as the ring-breathing mode of the pyridine rings at 1010 cm 1 , the in-plane bending of the pyridine rings at 1200 cm 1 , the ring stretching mode at 1610 cm 1 and the ethylene stretching mode at 1640 cm 1 .62,64 The differences between the two cluster models are comparatively small, with the largest deviations found in the relative Raman cross sections for the 1610 and 1640 cm 1 modes.In the following, we will therefore focus only on presenting the results for small (i.e., 7 atom) clusters.The computed vibrational frequencies are overestimated by about 4% by the PBE0 functional used in this work, in accord with previous benchmark studies.59 For consistency we do not rescale the calculated frequencies, and the mode assignment in Table 1 should serve as guide to relate experimental and calculated frequencies.
In Figure 3, we show the experimental Raman spectra of BPE absorbed on the silver SERS substrate and mixed-metal SERS substrates using 1 nm coverage of Pd or Pt at 633 nm excitation wavelength.We observed a stronger Raman signal for mixed-metal SERS substrates compared to the silver SERS substrate.Also, Pt coverage of 1 nm thickness is more effective in enhancing the Raman signal than Pd coverage of the same thickness.The relative Raman signals from different modes and the shapes of measured Raman spectra remain virtually unchanged.Measurements of Raman cross sections of BPE on SERS substrates are complicated by the fact that its surface coverage is not exactly known and might vary depending on the composition of the SERS substrate.Therefore, we consider analytical enhancement factors (AEF) as defined by LeRu, Etchegoin, and co-workers and used extensively in SERS research, [65][66][67] where I is the Raman intensity and c represents the concentration of the analyte.The significance of the observed analytical enhancement factors is discussed extensively below.
At close to 1 nm thickness, Pd-and Pt-covered substrates show analytical enhancement factors for the 1200 cm 1 mode that are 1.4 and 1.8 times higher than that for the silver SERS substrate, respectively.At 2 nm thickness, this ratio drops to 0.75 when using a Pd-covered substrate.The analytical enhancement factor for the 2 nm Pt-covered substrate is 1.9 times higher than for the silver substrate; indicating that there is a sustained enhancement as the coverage thickness increases.We observe no SERS enhancement beyond a layer thickness of 3 nm layer of Pd or Pt For the in-plane bending mode at 1200 cm 1 , we find that replacement of the proximal Ag atom by a Pd atom leads to an 3.8-fold increase in Raman cross section.For the Ag 6 Pt•BPE complex, the Raman cross section is highest, 5.1 times higher compared to the Ag 7 •BPE complex.The Raman spectra of the larger complexes of the form, Ag 14 M 4 •BPE, show a similar trend and are given in the Supporting Information.The computed results are summarized in Table 2.
Nitrogen-containing heterocycles such as BPE are known to form strong coordinative bonds to transition metals of the platinum group.This strong bonding can be expected to have a significant effect on the chemical contribution to surface enhancement.Therefore, it is instructive to study the binding energies of BPE to the mixed-metal clusters as a simple measure of the metal-molecular bonding.The binding energies DE are computed as electronic reaction energies (without zero-point correction) of the Ag x M y •BPE complex formation, The binding energies are given in Table 2 and follow the trend Ag < Pd < Pt.This observation is in line with complex stability of transition metals [68][69][70] and suggests that the mixing between electronic levels in the molecule and the metal cluster is correlated to the increase in Raman scattering cross sections.A simple explanation is that stronger bonding leads to a stronger coupling of vibrational modes in the analyte molecule to the large polarizability of the metal cluster.
Further, we computed Raman spectra of the Au 6 M•BPE complexes (M = Au, Pd, Pt) and their binding energies according to eq. 2 that are shown in Figure 5 and Table 3.The chemical enhancement and the Raman cross sections increase in the sequence Au < Pd < Pt.The effect of Pd and Pt is, however, smaller than for the Ag 6 M•BPE complexes, while the binding energies are similar for both Ag 6 M•BPE and Au 6 M•BPE series of complexes.It is therefore apparent that other electronic-structure considerations need to be taken into account to fully understand the origin of the mixed-metal effect.
Theory and experiment are complementary in the understanding of the molecule-metal and mixed-metal interface.In this work, we are able to provide analytical enhancement factors, but the surface coverage of BPE on SERS substrates is unknown and its measurement is far from trivial. 71eoretical results predict an increase in Raman cross sections per molecule when the analyte binds to a Pd or Pt atom, an effect that can be described by SERS lending.The enhanced Raman cross sections are concomitant with an increase in binding energy between the analyte and the metallic surface.Higher binding energies will however also lead to increased surface coverage of the analyte under equilibrium conditions, potentially yielding higher analytical enhancement factors.Estimating the surface coverage is therefore necessary in order to interpret the experimental findings.
[74] These measurements were done for a similar bulk concentration of pyridine, though each was determined with a different method.In the case of BPE, estimations of the surface coverage on Ag have relied on a geometric area estimation of 1.16⇥10 9 mol cm 2 , 75-77 though we note this is likely an overestimate.Measurements of BPE on Pt have reported to give a coverage of 4.75⇥10 10 mol cm 2 . 78Although the variety of methods shrouds the comparison, the difference in surface coverage between Pt and Ag is not over a factor of two greater.Therefore, coverage cannot be the sole factor for the 1.9-fold enhancement we find experimentally.Along with robust theoretical predictions of increased Raman cross sections per molecule, we tentatively conclude that a SERS lending effect is present in mixed-metal substrates considered in this work.

II. Theoretical Study of Pyridine: Raman Excitation Profile
We proceed by investigating the importance of a dynamical contribution to chemical enhancement in mixed-metal clusters using complexes Ag 6 M•Py as an example.Table 4 shows the relative Raman scattering cross sections for the 1230 cm 1 in-plane bending mode of pyridine as well as binding energies computed according to eq. 2. This vibration corresponds to the 1200 cm 1 mode in BPE (spectra available in Supporting Information).The binding energies follow the same trends as in the case of BPE and are within 0.04 eV of the values obtained for Ag 6 M•BPE, compare Table 2.The Raman cross section of the 1230 cm 1 mode is 6 times higher in Ag 6 Pd•Py and 9 times higher in Ag 6 Pt•Py compared to the Ag 7 •Py complex.However, the absolute Raman cross-section for this mode is 20 times smaller than in complexes containing BPE (shown in the Supporting Information).
Raman excitation profiles (REPs) are a useful tool in elucidating the dynamical contribution surface enhancement and describe the dependence of Raman cross sections on the excitation frequency. 16In Figure 6, we show REPs for the ring-breathing mode (experimental frequency 990 cm 1 ) and the in-plane bending mode (1200 cm 1 ) of Ag 6 Pt•Py and Au 6 Pt•Py complexes.
The silver-platinum complex has its lowest-energy electronic excitation at 607 nm, while the lowest excitation occurs at 616 nm in the gold-platinum complex.For both vibrational modes, Ag 6 Pt•Py exhibits a large increase in Raman cross sections close to the electronic excitation at 607 nm, showing almost an order of magnitude increase in Raman cross section for the ring-breathing mode.In contrast, Au 6 Pt•Py complex fails to show a significant increase for both of these modes at its excitation wavelength of 616 nm.
Isosurface plots of transition electron densities for the lowest-energy excitations in Ag 6 Pt•Py and Au 6 Pt•Py are displayed in Figure 6.In the case of Ag 6 Pt•Py (Figure 7a), the transition density lies exactly at the metal-molecule interface.The Au 6 Pt•Py complex (Figure 7b) is characterized by a transition density that is spread out throughout the metal cluster.These observations provide a simple explanation for the computed REPs of Figure 6.We note that when an excitation transition density is largely localized at the metal-molecule interface, the Raman cross sections for excitation energies close to it are strongly enhanced, as has been reported previously for Ag n -SPh complexes. 16In contrast, excitations in the metal cluster have a much smaller effect on Raman cross sections.The differences between Au and Ag as majority metals illustrate that excited state properties are also important to understand the effects in mixed-metal systems.

Conclusions
The chemical contribution to surface enhancement in mixed-metal SERS substrates was investigated by experimental and theoretical techniques.We report an increase by up to 1.9-fold in analytical enhancement factors obtained experimentally for BPE on silver-based SERS substrates including an overlayer of Pd or Pt of 1-2 nm thickness.The analytical enhancement factors include the effect of surface coverage of the analyte, the attenuation of the electromagnetic enhancement by the transition-metal overlayer as well as changes in the chemical enhancement.The observed increase in analytical enhancement factors is unlikely to arise solely from a higher surface coverage on Pd-or Pt-containing SERS substrates.Therefore, we conclude that a SERS lending effect is present in mixed-metal SERS substrates due to an increase in chemical enhancement.
These findings are further corroborated by theoretical results on small-cluster models, which consistently predict up to a 5-fold increase in Raman cross sections for BPE on Pd-or Pt-containing mixed-metal clusters.The differences in enhancement are understandable since the theoretical model does not take into account the distance-dependent decrease of electromagnetic enhancement (i.e., plasmon decay).It is likely the issue of the attenuation of the electromagnetic effect can be alleviated. 41,42,79Overall, the enhancement is a combination of the attenuation and the chemical enhancement effect.As mentioned above, the chemical enhancement can lead to observable SERS spectra without necessarily having a substrate with plasmonic properties. 22,23RS lending could open an avenue for new analytical applications for monitoring and iden-      The analytical enhancement factors are normalized with respect to the silver SERS substrates (0 nm thickness).For Pd coverage there is only enhancement at 1.2 nm thickness, while the Pt signal is close to 1.9 at both 1.4 nm and 2.4 nm coverage.

(
not shown).The dependence of the measured analytical enhancement factors on the thickness of the transition-metal overlayer is summarized in Figure 4.The computed Raman spectra of the Ag 6 M•BPE complexes (M = Ag, Pd, Pt) are shown in Figure 3. Increased Raman cross sections are observed in transition-metal complexes, while relative cross sections are preserved, in line with experimental results.

Figure 1 :Figure 2 :
Figure 1: a-b) SEM pictures of a femtosecond laser nanostructured SERS substrate and 2 nm Pt overlayer thermally evaporated on top of the substrate.c) Structure of a Ag 6 M•BPE complex.

Figure 4 :
Figure4: Analytical enhancement factors for the integrated signals of the in-plane bending mode of BPE on mixed-metal SERS substrates as a function of the transition-metal thickness.The analytical enhancement factors are normalized with respect to the silver SERS substrates (0 nm thickness).For Pd coverage there is only enhancement at 1.2 nm thickness, while the Pt signal is close to 1.9 at both 1.4 nm and 2.4 nm coverage.

Figure 7 :
Figure 7: Isosurface plots of computed transition densities of lowest-energy electronic excitations in Ag 6 Pt•Py (excitation wavelength 607 nm, a) and Au 6 Pt•Py complex (excitation wavelength 616 nm, b) The red color indicates a transition density of -0.005 a.u., while the blue corresponds to a transition density of 0.005 a.u.

Table 1 :
Mode assignment of the 5 most prominent normal modes of bare BPE.

Table 2 :
Enhancement factors of the in-plane bending mode of Ag 6 M•BPE complexes (relative to Ag 7 •BPE) and Ag 14 M 4 •BPE complexes (relative to Ag 18 •BPE).Binding energies DE of Ag 6 M•BPE complexes according to eq. 2.

Table 3 :
Enhancement factors of the in-plane bending mode of Au 6 M•BPE complexes (relative to Au 7 •BPE).Binding energies DE of Au 6 M•BPE complexes according to eq. 2.

Table 4 :
Enhancement factors of the in-plane bending mode of Ag 6 M•Py complexes (relative to Ag 7 •Py).Binding energies DE of Ag 6 M•BPE complexes according to eq. 2.