Surface Enhanced Raman Scattering Selectivity in Proteins Arises from Electron Capture and Resonant Enhancement of Radical Species

Abstract

Plasmon-enhanced Raman scattering is a powerful approach to detecting and characterizing proteins in live and dynamic biological systems. However, the selective detection/enhancement of specific residues as well as spectral diffusion and fluctuations have complicated the interpretation of enhanced Raman spectra and images of biological matter. In this work, we demonstrate that the amino acid tryptophan (Trp) can capture an electron from an excited plasmon, which generates a radical anion that is resonantly enhanced: a visible excited electronic state slides into resonance upon charging. This surface enhanced resonance Raman scattering (SERRS) mechanism explains the persistence of Trp signatures in the SERS and TERS spectra of proteins. Evidence for this picture includes the observation of visible resonances in the UV-Vis extinction spectrum, changes in the ground state vibrational spectrum, and plasmon-resonance dependent behavior. DFT calculations support the experimental observations. The behavior observed from the free Trp molecule is shown to explain the SERS spectrum of the Trp-cage protein. In effect, resonant Raman scattering from radicals formed through plasmonic excitation represents an under-investigated mechanism that may be exploited for chemical sensing applications.

Graphical Abstract

Introduction.

Proteins serve as the machinery of life and the ability to determine chemical information related to their function is important in diverse applications spanning the biological, environmental, and energy sciences. The ability to study proteins in their native environment and to monitor interactions on the individual protein level are increasingly desired. The enhancement of Raman signals by plasmonic structures, so called surface enhanced and tip enhanced Raman scattering (SERS and TERS, respectively),^{1, 2, 3, 4, 5, 6} are methods that have demonstrated potential to address the challenge of in vitro protein characterization.⁷ Colorimetric,⁸ electrochemical⁹ and mass spectrometry¹⁰ are common techniques applied for protein detection however they are unable to compete with the sensitivity and high spatial resolution on proteins and other biomolecules that surface enhanced Raman scattering (SERS) and tip enhanced Raman scattering (TERS) provide.¹¹ However, despite more than 40 years since the initial reports of SERS,^{12, 13} the origins of the observed signals continue to be a subject of debate.¹⁴

Raman signal enhancement from molecules in the vicinity of plasmonic nanostructures is a result of two multiplicative effects: electromagnetic and chemical. Electromagnetic enhancement is a result of the resonant interaction between light and plasmonic nanostructures, which excites surface plasmons.¹ Indeed, numerical simulations of the enhanced fields at plasmonic surfaces and junctions support experimental observations of large signal enhancements that support the detection of single molecules.¹⁵ While enhanced local optical fields are well-understood,^{15, 16, 17} reports of enhancement factors that cannot be readily explained solely by plasmonic enhancement have renewed interest in chemical effects.¹⁴

In general, chemical enhancement has been more controversial and has become – in effect - the catch-all category for effects that do not correlate with the expected electromagnetic enhancement factors. Chemical enhancement typically results from (i) changes in the polarizabilities of molecules dynamically interacting with a metal, and (ii) resonant enhancement, operative when plasmonic metal-to-molecule (or vice versa) charge transfer states are accessible at the incident photon energy.^{13, 18, 19, 20} Ab initio molecular dynamics (AIMD) simulations have explored how a fluctuating excited state manifold can alter the observed Raman response of single molecules,²¹ effects that are increasingly relevant when the excitation laser is on-resonance with charge transfer states. Chemical enhancement is also associated with first-layer effects, as the adsorbed layer will show enhancement separate from other molecules near the surface but not bound.²² The manner in which the molecule interacts and orients itself on the surface and its concentration has a great importance to the enhanced scattering.²³ Overall, chemical effects have been more difficult to quantify as the molecules also experience the electromagnetic effect, and the overall enhancement is difficult to de-convolute from contributions of varying numbers of the molecule. Interestingly, the formation of a charge-transfer complex is believed to enable otherwise forbidden optical transitions, such as b2 modes, a concept that was long used, e.g., as proof of chemical effects in the detection of p-aminothiophenol (ATP).²⁴ However, in ATP these modes were later shown to actually arise from the photoproduct di-mercaptoazobenene (DMAB), generated from plasmon-induced interfacial chemistry.²⁵ Indeed, other surface chemical reactions, such as the oxidation of pyrene on Ag,²⁶ have been shown to impact the detected SERS spectrum. It is becoming more recognized that electron transfer from plasmon resonances can promote chemical reactions at surfaces.²⁷

Vibrational modes observed in the enhanced Raman spectra of proteins have been similarly contentious. The amide I band has been reported as a marker of protein secondary structure;²⁸ however, the assignment was disputed²⁹ as the mode is oftentimes not experimentally observed. Within the electromagnetic enhancement model, the sporadic detection of amide I band was attributed to screening of the electric field in proteins with large amino acid side chains,³⁰ a conclusion that has been met with skepticism. High resolution TERS was reported to detect glycosylation status and also characterize the glycans themselves.³¹ The authors were careful to note that the Raman spectra obtained are very complex and dependent on protein orientation but with stricter control parameters characterization of glycans with TERS could be utilized in biopharmaceutical and formulation research. More recently Zenobi and coworkers have suggested that photo-fragments and radical species, similar to electron beam products in surface science, can explain the discrepancies and sporadic observation of certain Raman peaks.¹⁴ The formation of these photo-fragments was shown to be power dependent, providing an explanation for the sporadic observation of peaks.

In this report, we demonstrate that the enhanced Raman spectrum of the amino-acid tryptophan (Trp) is in effect surface enhanced resonance Raman scattering (SERRS) from the Trp radical anion. The latter is formed through plasmonic excitation followed by subsequent electron transfer. The anion features a visible optical absorption band that can be excited at the visible laser wavelength used in this study, as confirmed through time dependent density functional theory calculations (TD DFT). The described mechanism is a new form of chemical enhancement that is broadly applicable to molecules prone to plasmon-induced redox chemistry in the SERS and TERS schemes. The enhanced Raman spectra of Trp as well as the Trp-cage protein can be rationalized on the basis of our analysis.

Experimental.

Materials

All chemicals were purchased from Sigma Aldrich (USA)

Instrumentation

Extinction measurements were carried out on a VWR spectrometer (UV-1600PC). Size and zeta potential measurement were performed on a Malvern Zetasizer Nano ZPS. Nanoparticle concentrations were calculated using a Malvern NanoSight (NS300). SERS spectra were accumulated using a Snowy Range Sierra Raman spectrometer with 638 nm laser excitation and a Renishaw Qontor InVia Raman microscope with 532, 633, and 785 nm laser excitation.

Synthesis of gold nanoparticles

50 nm Au NPs were synthesized using a citrate reduction method. Gold tetrachloroaurate (16 mg) was added to 100 mL of HPLC grade water and heated to 90 °C with continuous stirring. Sodium citrate (16 mg) was then added to the solution which was stirred and heated for an additional 15 minutes until red. Au NP were characterized by extinction spectroscopy, dynamic light scattering, and zeta potential measurements. The nanoparticle concentration was determined using the NanoSight and calculated to be 6 × 10¹⁰ particles per mL for these synthesis conditions.

Characterization of MBN on nanoparticles

Au NPs (1 mL) were functionalized with 4-mercaptobenzonitrile (MBN,10 μL, 1mM) and left to shake overnight. The Au-MBN NPs were characterized using extinction and 638 nm laser excitation before and after the addition of NaCl (100 μL, 100 mg/mL).

Characterization of TRP on gold nanoparticles

The effect of aggregation on the SERS signal was investigated by adding salt to solutions of TRP and Au NPS. 100 μL of concentrated Au NP, 3.5 × 10¹¹ particles per mL, were added to 100 μL of TRP (20 mM) with 100 μL of HPLC grade water or 100 μL of 100 mg/mL NaCl. The samples were analyzed using 638 nm and 785 nm laser excitation as well as extinction spectroscopy. The experiments were repeated with MgSO₄ taking the place of NaCl.

Samples with a higher and lower number of TRP molecules per Au NP than that of the salt aggregation experiment were prepared combining TRP (20 mM) and a Au NP suspension (6 × 10¹⁰), which was modified by diluting with HLPC water or further concentrating by centrifugation. NanoSight tracking analysis determined the concentration of Au NPs. The samples were analyzed using extinction and Raman spectroscopy.

Salt titration on Au NP with TRP addition

The effect of NaCl addition on Au NP and TRP was investigated by preparing the following samples. 0, 5, 10, 25, 50 or 100 μL of NaCl (100 mg/mL) were added to samples containing 400 μL of Au NP and 100 μL of TRP (50 mM). Duplicate samples were prepared without the addition of NaCl. Sample volumes were kept constant with the addition of HPLC grade water. All samples were characterized using extinction, DLS and Raman using a 638 nm laser excitation.

Wavelength Dependent SERS Experiments

The effect of aggregation at different excitation wavelengths was investigated by mixing 5mM Trp solution with H₂O or 100mg/mL NaCl solution and AuNP (1:1:1). The extinction was acquired of each solution to determine the aggregation state of each sample. Each solution was drawn up into an acid cleaned capillary and the ends were sealed with a flame. Each capillary was placed under the microscope of the Renishaw Qontor InVia Microscope and was brought into focus on the top of the capillary. A depth study taking 10s acquisitions every 100μm using 785nm excitation was completed to determine the optimal focal point within the capillary by monitoring the decrease in the observed glass spectrum. This focus was maintained for each laser excitation wavelength. Spectra were acquired for 60s acquisitions in triplicate for each laser wavelength at comparable laser powers in the following order; 633nm (1.639mW), 785nm (0.616mW), 532nm (1.82mW). Spectra were obtained again for 633nm (1.639mW) and 785nm (0.616mW) after 532nm exposure at 9.71mW as well as 532nm at 18.4mW.

Effect of pH of TRP solution on SERS signal

To investigate the effect that pH has on the SERS signal of TRP on Au NP, the pH of the TRP solution (20 mM) was either changed to pH 1 using 1 M HCl, kept neutral or made alkaline (pH 10) with NaOH addition. 100 μL of each TRP solution was added to 400 μL of Au NP and the samples were analyzed using 638 nm laser excitation.

Density Functional Calculations.

All calculations were performed using a local development version of NWChem.³² UV-Vis and Raman spectra were calculated following full geometry optimization of the neutral and radical species using the PBE exchange-correlation functional³³ in conjunction with the def2-TZVP basis set.³⁴

Results and Discussion

SERS and TERS experiments on proteins consistently show signals that can be attributed to the aromatic amino acids: tryptophan (Trp), tyrosine (Tyr), and phenylalanine (Phe).^{29, 35, 36, 37} The spontaneous Raman response for these amino acids is slightly larger than for the aliphatic amino acids. Indeed, the conventional Raman scattered signal from proteins often contains contributions from a variety of amino acids in addition to the aromatic residues. Our lab and others have repeatedly observed spectra that are dominated by the aromatic amino acids.³⁷ In the case of αvβ3 integrin, the consistency of the spectrum strongly suggests selective detection of the protein. Indeed, this signal has been observed in cells, where a single protein receptor may generate the observed signal.^{38, 39}

Given the strong enhanced Raman scattering response observed, we adopted a minimalist approach and investigated the signal observed from Trp in solution using colloidal Au nanoparticles (AuNPs) synthesized by citrate reduction. Figure 1 shows the spontaneous Raman spectrum of tryptophan (Trp) and the SERS spectra for Trp and chemically oxidized Trp (Trp^+•), generated by oxidizing Trp with Ce^IV(SO₄)₂.⁴⁰ The spectra from the three experiments are markedly different. Both the solution spectra and the spectrum from Trp^+• are consistent with literature spectra of these species.^{40, 41, 42} While some peaks are shared between the three species, in general, the overall spectrum observed for the TRP SERS experiment with gold nanoparticles (AuNPs) is markedly different.

Figure 1.

The solution SERS spectrum of Trp mixed with AuNPs (black) is compared with the spontaneous Raman of a 50 mM Trp solution (red) and the spectrum observed from the cation Trp radical (Trp^+•) produced by mixing with Ce^IV(SO₄)₂ (blue). The intensities have been normalized and the spectra offset to enable clear differentiation of the observed Raman bands. The vertical lines provide reference markers for bands.

Trp^+• is known to have an electronic energy level transition near 450 nm. Cation-π interactions to Trp are reported to give rise to similar visible absorption band that red-shift depending the strength of interaction, suggesting a charge transfer-like effect is operative.⁴² It is important to note that ground state Trp does not have visible optical absorption bands. However, the formation of a species with an allowed electronic transition could give rise to resonant Raman enhancement, which may explain the selective enhancement/observation of Trp.

To explore this possibility, we followed the resonance Raman work of Shafaat and Kim and generated the Trp^+• species.⁴⁰ The blue SERS spectrum observed in Figure 1 agrees with the previously reported conventional resonance Raman spectrum, and shows a large SERS response from TRP^+•. The observed SERS for Trp^+• can be attributed to SERRS, where the electronic state slides into resonance with the visible 638 nm laser, thereby causing an increase in scattering activity through the resonance Raman effect. Preresonance effects have also been shown to increase Raman signals, even when the electronic transition is not exactly coincident with the Raman excitation.⁴³ While the signal observed from Trp^+• is consistent with the literature, it does not agree with SERS signals observed from Trp in Figure 1 (black spectrum).

The UV-Vis extinction spectra observed from Trp mixed with AuNPs shows evidence that the recorded enhanced vibrational spectra may be attributed to resonant SERS. Figure 2 shows the extinction spectra obtained from solutions of different concentrations of AuNPs mixed with a fixed concentration of Trp. The fixed concentration of Trp was chosen such that no aggregation was detected by dynamic light scattering (DLS, Table S1 and Figure S1). Very little peak broadening or LSPR red shifting is observed, another indication that little aggregation has occurred. As the ratio of Trp to AuNPs increases, we observe a clear shoulder at 643 nm in the extinction spectrum. This is most prominent when there were 4.8 ×10⁸ TRP molecules per molecules (red extinction spectrum) This shoulder is close to the 638 nm excitation wavelength of the laser used for SERS. The observed band at 643 nm is significantly red-shifted from previous reports of Trp radicals that were closer to 500 nm.⁴² It is well known that aggregation of nanoparticles can also give rise to red-shifted resonances. However to provide evidence, along with the DLS data and extinction spectra that indicated that the addition of TRP did not aggregate the Au NP and the SERS signal was not from hotspots, we intentionally aggregated our nanoparticles in the presence of NaCl to show the changes in the extinction and SERS spectra. We also investigated the change in extinction and SERS spectra that is commonly experienced when molecules experiencing the increased electromagnetic field associated with hotspots.

Figure 2.

a) Normalized extinction and b) SERS spectra of AuNP with varying number of TRP molecules per nanoparticle, 4.4×10⁷ (black), 8.8×10⁷ (blue) and 4.8×10⁸ (red). The spectra in (a) have been normalized to the LSPR peak maxima. The colors of the SERS spectra correspond to those of the observed extinction spectra. Arrow marks Trp related band.

Figure 3 shows the effect of aggregation with NaCl on the extinction spectrum of Au NP and TRP and the corresponding SERS spectrum. In the extinction spectrum (red), the addition of NaCl decreases the electrostatic repulsion between AuNPs and promotes aggregation as van der Waals forces become increasingly dominant.⁴⁴ Aggregation results in the formation of a coupled plasmon resonance as well described in prior analyses.⁴⁵

Figure 3.

The extinction (A) and SERS spectra (B) observed from AuNPs in the presence (red) and absence of NaCl (blue) are shown. The Extinction spectrum shows the expected formation of a second plasmon resonance. The SERS spectrum shows an unexpected decrease in signal. The SERS experiments used a 638 nm laser for excitation.

Unexpectedly, the observed SERS behavior in Figure 3 is not consistent with previous reports. The aggregation of NPs is expected to form hotspots with increased electric fields and larger electromagnetic enhancements. In our case of Trp, aggregation decreases the SERS signal with 638 nm excitation. Repeating this experiment with an aromatic thiol adsorbate on the AuNPs shows the expected increase in SERS signal for the aggregated AuNPs (see supporting information, Figure S2). To test for displacement of the Trp molecules from the NaCl addition, we repeated the experiment with MgSO₄, a salt that is reported to have less surface affinity and thus less ability to displace analytes, and observed a similar response (Figure S3) to that with NaCl.

To further validate that the TRP SERS spectra was due to the resonant peak at 638 nm and that TRP was still present after the addition of NaCl, we repeated the experiments using different laser excitation wavelengths. Where Figure 3 was obtained on a Snowy Range spectrometer with a single laser, the wavelength dependence (Figure 4) was performed using a Raman Microscope with multiple lasers to enable measurements on the same sample. Figure 4 shows the average SERS result from replicate experiments performed using 532, 633, and 785 nm excitation on two samples, one with and one without NaCl in sealed glass capillaries. Using the 785 nm laser, a weak response was observed from the unaggregated particles. Upon the addition of NaCl, a SERS spectrum similar to that observed at 633 nm excitation is observed. The behavior observed at 633 nm is the same as observed in Figure 3, in that a strong SERS response is seen in the unaggregated nanoparticles, and no Trp peaks are observed at 633 nm in the NaCl aggregated particles. No peaks that can be assigned to Trp are observed at 532nm excitation, which is explained by intra-band transitions of Au dampening the plasmon resonance, and thus no SERS, at this wavelength.⁴⁶ The data in Figure 4 confirms that NaCl does not displace surface-bound Trp molecules. Additionally, the data supports a correlation between the plasmon resonance and the observed SERS signal that cannot be solely understood on the basis of plasmonic field enhancement.

Figure 4.

The extinction spectrum of A) unaggregated AuNPs with Trp and B) AuNPs and Trp aggregated with NaCl are shown. The laser wavelengths used to generate SERS from the C) unaggregated AuNPs with Trp and D) NaCl aggregated AuNPs with Trp.

An intriguing hypothesis is that the observed Trp SERS signal arises from the radical anion species (Trp^−•), where electrons from the excited plasmon resonance are transferred to/captured by Trp molecules. As noted above, recent work by Zenobi et al has implicated species observed in low energy electron spectroscopy experiments to explain the changing peaks observed in SERS.¹⁴ Fragmentation and radical chemistry generate new species and alter the observed Raman spectrum. Our hypothesis is also consistent with a prior report by Kneipp that showed electron capture by imine polymer on silver NPs gave rise to a radical species with resonance Raman enhancements in the SERS spectrum.⁴⁷ The formation of an anion species has also been shown to change the vibrational spectrum in MBN.⁴⁸ Interestingly, the low energy electrons found in electron stimulated desorption experiments generated the Trp^−• species as the dominant product.⁴⁹ To test the hypothesis that the observed signal arises from excitation of the plasmon resonance, we measured the SERS signal at both 633 and 785 nm after exposure to different power 532 nm excitation. Figure 5 shows that exposure at 532 nm (on resonance with the unaggregated AuNPs) produces an increased SERS response at both wavelengths. This further supports the hypothesis that electrons excited by the plasmon resonance are responsible for the observed signal. The increased SERS response is consistent with a stable radical species, which increases in concentration when the localized plasmon resonance on the AuNPs is excited. The power dependence in Figure 5C shows the increasing SERS response until, at high powers, a decrease is observed. This decrease may result from fragmentation suggestive of photodamage when more electrons are available, consistent with Zenobi et al’s results and their analysis.¹⁴

Figure 5.

SERS spectrum observed from a solution of Trp and AuNPs A) before and B) after exposure at 532 nm. C) The power dependence of the 532 nm exposure is shown to increase with power before decreasing at the highest exposures, possibly due to fragmentation.

Density functional theory (DFT) calculations can predict the formation of electronic states that will give rise to resonance Raman scattering. Figure 6 shows that both the anion and cation radicals of Trp absorb light at energies associated with the plasmon resonance of AuNPs. The anion radical shows a red-shifted absorption that is closer to the 643 nm resonance observed when Trp interacts with AuNPs. The cation radical (Trp^+•) shows the previously reported resonance noted above. While Raman cross sections of molecules are notoriously small, resonance Raman can increase the Raman response by as much as 10⁶,⁵⁰ which is comparable to the electromagnetic enhancement mechanism typically associated with SERS and TERS. Indeed, the combination of resonance Raman and electromagnetic enhancement have been shown to enhance the detection sensitivity in SERS down to the single molecule limit.^{51, 52, 53, 54} The ground state vibrational spectrum of each radical can also be calculated and compared to the experimental results. The calculated Trp^+• spectrum (see supporting information Figure S4) shows good agreement with the experimental result in Figure 1. There are some differences in intensity, which are difficult to accurately calculate, but overall the agreement is noteworthy. Similarly, calculating the Trp^−• Raman spectrum qualitatively agrees with our experimental SERS results (Figure S5), further supporting our hypothesis of electron capture. Interestingly, pH does not have a significant effect on the Trp spectrum (Figure S6), except at alkaline pH where the spectroscopic changes are not as evident. At alkaline pH, electrostatics may hinder electron capture due to charge repulsion. At neutral and acidic pH similar spectroscopic features are observed, suggesting the changes reside in the indole ring which would be less sensitive to pH effects.

Figure 6.

The formation of an electronically resonant state upon formation of Trp radicals is supported by DFT calculations. Electronic transitions to the LUMO of closed shell Trp are only evident as absorption bands in the UV (~4.4 eV). Above, the energy transition to the LUMO in both the cation and anion radicals of Trp are calculated to be in the visible range of the electromagnetic spectrum, near the localized plasmon resonance of AuNPs (~2.2 eV).

To connect these observations to proteins, we recorded the SERS spectrum of Trp-Cage protein. This 20 amino-acid protein has a lone Trp (W) residue in its sequence (NLYIQ WLKDG GPSSG RPPPS).⁵⁵ This protein readily folds in water to enclose the Trp residue within a hydrophobic core. This cage conformation is expected to inhibit a direct interaction between Trp and the AuNP. Figure 7 compares the SERS spectra observed from Trp-Cage and Trp solution. The SERS spectra show many similar features, including key bands at 927, 1010, 1118, and 1454 cm⁻¹ that are different from the spontaneous Raman spectrum of the Trp solution. The shifting of some frequencies is expected in the protein environment. In addition to exhibiting many of the vibrational features we attribute to the Trp^−• species, the extinction spectrum of Trp-Cage also shows a resonance near 643 nm (Figure 7B), similar to observed in free Trp. Similar to the case with free Trp, DLS does not indicate aggregation in this sample. Additionally, electron capture dissociation mass spectrometry on Trp-Cage shows the dominant ion detected is the intact protein ion, strongly suggesting the protein does not fall apart or fragment with the addition of an electron.⁵⁶

Figure 7.

A) SERS spectrum of Trp-cage protein and AuNPs (red) is compared with SERS of Trp and AuNPs (black). B) the extinction spectrum from bare AuNPs (yellow) and with Trp-cage added (red). The Trp-cage structure is shown.

The data we have collected support the model illustrated in Figure 8 that additional resonance enhancement associated with electron capture from the plasmon excited nanostructure generates increased SERS signals observed from select amino acids in proteins. This hypothesis provides a mechanism to chemically enhance molecules near, but not adsorbed directly to the nanoparticle surface. More generally, differences in the electron capture stabilities of molecules may explain why certain small molecules are readily detected and others are not.

Figure 8.

Illustrations shows how electrons excited in AuNPs can be captured by free molecules to generate species with lower energy transitions, that can be resonantly enhanced and generate an increased surface enhanced Raman scattering signal.

The electron captured by the tryptophan could arise through several possible mechanisms. It is now well established that exciting plasmon resonances can produce energetic electrons that can result in chemical transformations.^{27, 57, 58} Recent work has addressed the relative contributions of hot electrons or thermalized electrons that arise from plasmon excitation.⁵⁹ In the present work, the energy of the lowest unoccupied molecular orbital (LUMO) of Trp is quite high, which would suggest a highly energetic electron is necessary.⁶⁰ The required energy is greater than the 2.2 eV associated with the plasmon resonance on AuNPs. Interestingly, there are reports of mechanisms that give rise to lower energy electronic levels as molecules approach surfaces.^{61, 62, 63} Additionally, there are reports of low lying triplet states from molecules on surfaces, which might be accessible to plasmon excited electrons.⁶⁴ These lower energy electronic states provide possibilities for the increased signals attributed to resonant Raman enhancement; however, they do not explain the changes observed in the Raman spectrum. It is possible that a surface complex forms between the anionic tryptophan species and the nanoparticle. The peaks observed agree with ground state calculations of the radical anion, indicating a significant amount of electron density has changed in the indole ring and around the Trp residue that is consistent with electron capture.

This mechanism will have testable observations and implications for enhanced Raman based sensing. In proteins, the captured electron will reside on the residue that can stabilize the additional charge, likely that with the lowest energy. An interesting observation reported was the TERS spectrum of Azurin, a copper containing metallo-protein with multiple Trp residues, did not show the enhancements we report.⁶⁵ We also tested azurin with AuNP based SERS and observed the same signals reported in the prior TERS report. The enhanced Raman signals observed are associated with ligation to the Cu metal ion, suggesting this center is more likely to accept the electron, or that the resonance Raman contribution from native azurin overwhelmed the signals from other species.

It is not clear how many Trp molecules capture electrons and are enhanced. The solution concentration of Trp is quite high, in the low millimolar range. Our data in Figure 5 suggest that increased energy into the LSPR can increase the observed signal. The resulting product appears to be stable on the timescale of several seconds, as we were able to increase the signal by pumping the system at the plasmon resonance (532 nm) and then subsequently measuring Raman at a different wavelength. This stability may change from molecule to molecule and may be related to SERS intensity fluctuations previously reported.⁶⁶ A short lived radical species might appear to blink, a behavior typically associated with sparse coverage and single molecule detection but also observed from monolayer coverages on single particles.⁶⁷ The enhancements observed in prior SERS/TERS work suggest that only a few, possibly one, protein is detected,^{68, 69} suggesting the signals are quite intense. Experiments examining the temporal response will further elucidate these possibilities.

Conclusions

In summary, we describe an indirect chemical enhancement mechanism in SERS, whereby plasmon-induced molecular charging produces radicals that feature excited electronic states that can be accessed using the visible wavelength laser lines that are typically used in SERS and TERS. The latter activates resonant Raman scattering, which may lead to the selective observation of resonantly excited species. The described mechanism explains the SERS spectra of Trp and Trp-cage protein, whereby resonance Raman from the radical anion of Trp predominates the recorded spectra. The generality of the mechanism we describe will need to be evaluated in further studies, particularly given the rapid advances in the field of plasmon-enhanced chemical conversion.