Tip-enhanced Raman spectroscopy for nanoscale probing of dynamic chemical systems

Understanding the nanoscale dynamics of chemical processes is of great interest both to fundamental science and for practical applications. The rates and pathways of chemical processes depend on the environment in which they occur. These environments can show large variations across a substrate, even on the atomic scale, so the ability to analyze their effects on chemical dynamics requires techniques that can provide chemical information with high spatial resolution and, if possible, high temporal resolution. Common spectroscopic techniques that rely on diffraction-limited optics, such as absorption, emission, and Raman scattering, are ideal for studying dynamic systems due to their combination of high time resolution and high chemical specificity. However, the spatial resolution of these techniques is limited to a scale equal to approximately half of the wavelength of the excitation light, so they can only reveal an ensemble average of the effects of many local environments on the chemical dynamics. On the other hand, techniques such as electron microscopy and scanning probe microscopy have sub-nanometer spatial resolution, making them ideal for observing nanoscale phenomena, but they lack the chemical specificity and time resolution of spectroscopic techniques. It is therefore challenging for such methods to elucidate chemical processes at the nanoscale.

Tip-enhanced Raman spectroscopy (TERS) is a near-field spectroscopic technique that combines the high spatial resolution of scanning probe microscopy with the chemical specificity of Raman spectroscopy,^1–3 making it well-suited to the task of examining dynamic systems with nanoscale spatial resolution. This combination is possible because nanostructures of metals that exhibit strong, localized plasmon resonances, such as Au, Ag, and Cu, can confine light below the diffraction limit, amplifying the Raman signals from materials in the vicinity of the nanostructure by factors of 10⁵–10⁶.⁴ This method of near-field excitation is readily combined with scanning probe microscopy techniques such as atomic force microscopy (AFM) and scanning tunneling microscopy (STM), which both use nanoscopic tips as probes. Using plasmonic STM/AFM tips enables measurement of Raman spectra from anywhere on a surface, with nanoscale resolution. Under optimal conditions, the spatial resolution of TERS can be better than 1 nm.^5–7 

Since the first reports,^8–11 TERS has been used in many areas of research, including the fundamental physics of small molecules,^6,12 structures of biological systems^13–17 and devices,^18–21 and electrochemical processes.^22,23 While early studies focused on using TERS to image stationary materials, improvements in TERS instrumentation and methods have expanded its capabilities, and there is growing interest in applying TERS to dynamic processes. Using TERS to access dynamic processes in different environments requires overcoming several technical hurdles. Fundamental processes can be examined under ultrahigh vacuum (UHV) or atmosphere, but most chemical reactions that are currently accessible to TERS occur in solution, where the fragility of TERS tips and weak signals hinder data acquisition. Thus, it was many years before TERS measurements were able to study the dynamics of systems more complicated than model reactions. As the capabilities of TERS grew, many of these technical challenges have been overcome, paving the way for TERS to study both fundamental processes in vacuum and complex electrocatalytic reactions.

In this Perspective, we discuss the frontiers of research in each of these areas and present the directions for future progress. Although the application of TERS to dynamic systems is still an emerging field, the range of topics that have been studied is quite broad, so we have categorized the dynamic systems based on the driving force for the dynamics and the type of process. Before describing this research, we introduce some of the fundamental theory behind plasmonic enhancement, as well as the experimental requirements for performing TERS measurements in different environments.

Based on this result, surface-enhanced Raman scattering (SERS) and TERS studies usually use the approximation that the enhancement of the Raman signal in the near field is proportional to the fourth power of the electric field enhancement factor. This statement is known as the E⁴ approximation. Note that this is an approximation that does not take into account polarization differences between the incident and scattered fields, the effects of molecular orientation, and the validity of the optical reciprocity assumption used for Eq. (6).^24,29 Hence, TERS enhancement factors frequently deviate from E⁴ behavior.^24,30 Nonetheless, the approximation is convenient and widely used in discussion, so it will be used here as well. Note that for the spherical nanostructure described above, Eq. (2) states that E_loc decays as r⁻³. In this case, application of the E⁴ approximation indicates that the Raman signal decays as r_m⁻¹². Thus, Raman signals generated from materials near a plasmonic nanostructure are much stronger than those a few nanometers away, as illustrated in Fig. 1(a). The expression of the field decay for TERS tips is more complicated and is therefore usually simulated. However, the same basic principles apply to Raman enhancement at TERS tips. Namely, the electric field enhancement factor decays within a few nanometers of the tip apex. This rapid electric field decay is the basis of the high spatial resolution of TERS. The concept for TERS tips is illustrated in Fig. 1(b).

Hot spots are local regions of a plasmonic nanostructure exhibiting larger electric field enhancement factors than the rest of the structure. Due to the E⁴ scaling of Raman enhancement, small variations in the electric field enhancement produce large variations in the enhancement of Raman signals. Since TERS always samples a very small number of molecules, the additional Raman enhancement in hot spots can be critical to producing detectable signals in TERS. The theory behind hot spots has been thoroughly studied in SERS,^24,31 and the same concepts also apply to TERS hot spots. The most important types of hot spots in TERS occur due to the lightning rod effect and the formation of nanogaps between a metal tip and the substrate.

The lightning rod effect occurs at sharply curved surfaces on nanostructures, such as edges, corners, or at the apex of a TERS tip with a high aspect ratio, where the incident electric field drives charges into a tightly confined space. Even a perfectly smooth tip that does not exhibit a plasmon resonance will show a weakly enhanced electric field near its apex that is mostly independent of excitation wavelength.^1,32,33 The hot spot is not always the tip apex because TERS tips are usually fabricated by using electrochemical deposition or vacuum evaporation to deposit a plasmonic metal onto a silicon tip. This deposits nanostructures randomly on the tip surface, and these randomly distributed structures may also act as local antennas.^32,34

Much larger enhancement factors can be found in a nanoscale gap between two metals.^24,35,36 In this gap, the electric field is enhanced by capacitive coupling between the metals.³⁷ Raman enhancement factors in the gap are typically around 10⁸–10⁹, but factors as high as 10¹² have been reported.^36,38,39 In TERS, gap mode enhancement is achieved by approaching the tip to a metallic substrate, creating a hot spot in the gap between the tip and the substrate. This is illustrated in Fig. 2, which depicts the electric field distribution at the apex of a tip near an insulating substrate [Fig. 2(a)] and a plasmonic substrate [Fig. 2(b)]. In the former, the electric field is most intense near the tip but rapidly decays with distance, whereas in the latter, the electric field is enhanced by the effective electromagnetic coupling between the tip and the metallic substrate. The enhancement factor in the gap mode increases with a decrease in gap size,^35,36 but when the gap reaches ca. 5 Å–7 Å, quantum tunneling between the tip and substrate begins, and the enhancement factor decreases.^12,40,41

Although geometrical calculations predict that the highest spatial resolution of TERS is similar to the diameter of the tip,⁴² spatial resolutions much better than this have been frequently reported.^6,12,43 When a strong field is confined in a narrow gap, there are several effects that can change the spatial distribution of the enhancement factor, resulting in higher spatial resolution. One is the field gradient effect, which exists because the strong local electric field has a large gradient. This gradient can be used to detect transitions that are normally forbidden, and it leads to high spatial resolution because its nonlinear dependence on field strength means that it is significant over a smaller region of the electric field profile.^41,44 The molecular cavity optomechanics model proposes enhancement of molecular vibrations within the cavity via dynamical backaction coupling, in which coupling between the vibrating molecule and the localized plasmon shifts the plasmon frequency, which, in turn, amplifies the vibrational mode in the cavity.⁴⁵ Another way in which the molecules may interact with the cavity is through multiple elastic scattering, which also produces a small hot spot consisting of only a few molecules.⁴⁶ A fourth model proposes that when the gap size is below 1 nm, the atomic structure of the tip–substrate interaction must be considered because an atomic protrusion in a narrow gap functions as an atomistic lightning rod, focusing the electric field around the protruding atom.^40,47–51

TERS measurements in ambient air are the most common, primarily because this is the most convenient environment to work. However, this convenience comes with some challenges for studies of chemical dynamics, foremost of which is that many interesting chemical reactions in air, such as those used for catalysis, occur at elevated temperature and pressure, and TERS has not yet been developed for these conditions. Nonetheless, within the limits of ambient conditions, one can examine the dynamics of model systems in depth. For these studies, the main technical challenges to overcome are vibrations and thermal drift of the instrument. An experimental setup for TERS measurements in air must be isolated from ambient noise and be able to compensate for thermal drift. A second problem is that, in ambient air, oxidation and contamination shorten the lifespans of TERS tips.⁵² This can be addressed either by performing TERS experiments in a controlled atmosphere^53,54 or by coating the tip with an inert oxide.^55,56 Coating the tip reduces the signal strength due to the increased distance between the tip and the sample. However, the development of tip coatings with high dielectric constants, which theory shows can compensate for attenuation of light scattering from particles,^57,58 may solve this problem.

The rapid progress in TERS measurements in air makes it well-suited to examining fundamental problems, which usually require highly stable, accurate measurements. The stability can be seen in the progress in spatial resolution achievable in air. As early as 2014, 1.7 nm resolution was reported in a TERS imaging study of a carbon nanotube,⁵⁹ and by 2019, sub-nanometer spatial resolution had been reported in imaging studies of graphene⁵⁴ and single-stranded DNA.¹⁷ The high spatial resolution has been utilized to indirectly study site-specific changes in surface properties by observing variations in the Raman spectra of a probe molecule adsorbed to different sites on a metal surface.^60,61 The result of this combination of convenience and high resolution, which comes with the drawback of having fewer practical dynamic systems to study, is that TERS in ambient air occupies a middle ground between basic and applied research.

Achieving the highest spatial resolution in TERS requires high stability and strong signals. The ideal environment for this is a low-temperature, ultrahigh vacuum. In vacuum, oxidation and contamination are inhibited, which not only slows changes to the system, allowing for long-duration measurements of weak signals, but also extends the lifetimes of silver tips and substrates, which is especially important, due to the large plasmonic enhancement factor of silver tips compared with gold. Reducing the temperature to a few degrees Kelvin minimizes thermal drift of the instrument and conformational fluctuations of the sample. At room temperature, even strongly adsorbed molecules display conformational changes on the timescale of seconds,⁶² and a typical SPM instrument drifts in the scanning plane by about 1 nm/min, making it impossible to image features on the order nanometers when the scan time is on the order of minutes.⁶ Therefore, UHV and low temperatures are essential for obtaining the highest quality, sub-nanometer images.

Even under these highly stable conditions, completing measurements in a reasonable amount of time requires TERS tips with large enhancement factors. Silver tips have high enhancement factors, and the UHV condition drastically increases their lifetimes, making silver an ideal choice for this condition. The enhancement factors of these tips can be further improved by cleaning the tip and optimizing its shape. In the most common method of preparing tips for STM-TERS, the first step is electrochemical etching of the tip. A tip prepared in this way has nano-scale defects, and it becomes oxidized and contaminated before it can be introduced into the UHV chamber.⁵² Since the signal strength and tip shape are crucial in high resolution measurements, further treatment is needed once the tip is brought into the UHV chamber. Therefore, the tip is cleaned, in vacuo, by outgassing and ion sputtering,^6,63 and the tip shape is optimized by applying high voltage pulses or by using tip indentation, in which the tip is brought into a quick contact with the substrate.^5,6,63 The tip quality is characterized in situ via the tip-induced luminescence.⁶³ An alternative approach that can be used to make high quality, reproducible tip shapes is focused ion beam (FIB) milling.^33,64 However, FIB milling is a relatively slow process. Alternatively, after electrochemically etching the tips, another method called field-directed sputter sharpening can be used, which allows batch-processing of tips with consistent, high quality shapes.^39,65

The cost and technical requirements of UHV-TERS experiments are high, and because of this, the first UHV-TERS experiment was not reported until 2007.⁶⁶ Even now, only a few groups perform these experiments. Nonetheless, the progress in the field has led to some incredible achievements. In 2013, sub-molecular imaging was reported for the first time, and the vibrational modes localized on the four lobes of a porphyrin could be clearly observed.⁶ Since that initial report of sub-molecular imaging, there has been interest in further improving the spatial resolution of TERS. Atomic-scale TERS mapping was finally demonstrated in 2019 by two reports of TERS imaging of porphyrins that demonstrated ca. 1.5 Å resolution.^12,43 UHV-TERS continues to grow in new directions, as shown by recent work examining the TERS selection rules on an insulating substrate consisting of three monolayers of NaCl coated Ag(111).⁶⁷ Our focus in this Perspective is on studies of dynamical systems, so we will not discuss the research on pushing the limits of spatial resolution. Readers interested in a detailed account of the progress of TERS in UHV are encouraged to read the review article by Mahapatra et al.⁶⁸ 

Introducing TERS into the liquid phase potentially enables high resolution studies of biological systems and catalytic processes, in situ. So far, almost all TERS studies carried out in liquids are electrochemical TERS (EC-TERS) studies. TERS was first introduced into the solution phase in 2009,⁶⁹ and gradual improvements in the experimental design led to the development of EC-STM-TERS in 2015,⁷⁰ followed shortly by the introduction of EC-AFM-TERS in the same year.⁷¹ As with any new experimental condition, the introduction of TERS into the solution phase adds several complications to the measurement. Even under optimal conditions, the TERS signals are relatively weak, resulting in long acquisition times.

Insulating TERS tips from the solution is necessary for the stability of the measurements. In addition to protecting the tip from impurities in the solution, the insulation has a different specific purpose for STM-TERS and AFM-TERS. In STM-TERS, the insulation prevents Faradaic current from overwhelming the tunneling current used to control the tip–substrate distance. This is usually done using a polymer that coats all but the apex of the tip.^70,72,73 In AFM-TERS, tips are modified to prevent delamination of the metal coating of the tip from its substrate. One way to promote adhesion between the plasmonic metal coating and the tip is to evaporate another metal, such as W or Cr, onto the tip before adding the plasmonic layer.^74,75 Alternatively, the plasmonic layer can be protected by coating an insulating layer on top of it. Monolayers of organic materials have been used as insulators, but they have short lifetimes in solution.^69,76 It is more common to use 1 nm–2 nm of an inorganic inert material, such as SiO₂⁷⁷ or ZrO₂,⁷⁸ to protect the tip. This method has the additional advantage that it can be used to protect Ag tips, extending their shelf-lives from days to weeks. With proper preparation of the tip, AFM-TERS or STM-TERS experiments in solution can be stable for at least an hour.

To obtain adequate signal strength for performing TERS measurements, the optical configuration for solution phase TERS must overcome the reflective losses and defocusing caused by the refractive index mismatch at the air–glass and glass–liquid interfaces between the microscope objective and the tip in solution. The challenges of designing a suitable setup are compounded by the space constraints of the SPM system. The ideal configuration places a high NA, water immersion microscope objective as close as possible to the tip. For a bottom-illumination configuration, the placement of the microscope objective is not constrained by the SPM system, so the implementation is fairly straightforward, as shown in Fig. 3(a).^69,71 However, this configuration is only suitable for studying transparent samples, which excludes many important surfaces. Side- and top-illumination configurations enable the study of opaque samples, but for these, there is a space conflict between the SPM system and the optics. For the top-illumination configuration, the tip must be mounted at an angle, as shown in Fig. 3(b), because the Raman excitation light cannot easily be focused on a vertically mounted tip. Accommodating the large bulk of a water immersion objective can be achieved in STM-TERS by insulating a long section of the wire that will be submerged and then bending it to bring the tip into the focus of the objective.^79,80 In EC-AFM-TERS, a tip mounted on the cantilever with an angle can achieve the same effect.⁸¹ The side-illumination configuration was used for the first report of electrochemical TERS (EC-TERS).⁷⁰ In this initial report, the beam was introduced horizontally, and the sample surface was tilted 10° to allow collection of the signal scattered from the surface. The light collection efficiency is poor when the laser and the sample surface are nearly parallel to each other, and the slight tilt of the sample is not ideal for imaging. Therefore, the more common design now uses a sample surface that is perpendicular to the tip and an optical window that is tilted to accept a beam that is tilted relative to the sample plane [Fig. 3(c)].⁷² This provides a better collection angle for the TERS signal, and it can accommodate a high NA, water-immersion objective.⁸² 

Together, these improvements to the optical configuration and tip protection have led to rapid development of solution-phase TERS in recent years. A spatial resolution better than 10 nm has been reported,⁸³ which is comparable to that achieved in air in 2005.⁸⁴ One very recent accomplishment in EC-TERS is the development of EC-AFM-TERS using a top-illumination configuration with a water-immersion objective.⁸¹ The top-illumination configuration is more useful than the bottom-illumination configuration developed earlier⁶⁹ because it can be used with opaque samples. Although EC-AFM-TERS has not been used in much of the previous research, it has some advantages over EC-STM-TERS. In particular, it can be coupled with other techniques, such as Kelvin probe force microscopy, and it allows the use of solvents that are easily oxidized or reduced by an STM tip current. EC-TERS is a fairly recent development, but because of its importance to the study of chemical reactions, it is already a major focus for many research groups. Hence, EC-TERS research occupies the largest portion of the section on dynamics.

The simplest dynamic processes on surfaces are thermally driven diffusion and conformational fluctuations of adsorbates. Due to the small sampling volume of TERS, only a few molecules are detected at one time, so diffusion of molecules through the sampling area or conformational changes that affect the orientations of the molecular polarizability tensors cause fluctuations in the Raman intensity, similar to those observed in single molecule fluorescence. The use of TERS to detect the dynamic fluctuations of molecules began in 2011, with the development of fishing-mode TERS, in which TERS spectra [Fig. 4(a)] were acquired during transient formation of a conductive junction [Fig. 4(b)] between a Au STM-TERS tip and 4,4’-bipyridine on a Au(111) substrate.⁸⁵ The relative intensities of the Raman bands obtained from molecular junctions fluctuated in time, demonstrating the observation of dynamic changes in the junction. The simultaneous measurement of TERS and conductance allows mutually verifiable single molecule conductance and Raman signals with single-molecule contributions to be acquired simultaneously at room temperature.

Obtaining deeper insight into the dynamics of molecules on the surface requires studying them under low temperature, UHV conditions because thermally driven processes are so rapid at room temperature that they produce broad Raman lines due to motional averaging of molecules fluctuating and diffusing through the TERS hot spot more rapidly than the typical TERS acquisition time of 1 s.⁸⁸ Moreover, TERS signal changes in air are more complicated due to the higher probability of sample oxidation in air.^89,90 Hence, low temperature UHV-TERS simplifies the measurement by ensuring sample stability and slowing thermal processes. This can be seen in the spectra of Fig. 4(c), which show the result of cooling a sample consisting of Rhodamine 6G on Ag(111) to 19 K for TERS measurements. Of the four samples shown, only the one examined by TERS at low temperature shows narrowed Raman bands. This occurs because molecular conformation heterogeneity is reduced, due to the small size of the population sampled by TERS, and because conformational fluctuations are reduced by the low sample temperature.⁸⁶ 

Thus, at low temperature, the properties and dynamics of small ensembles of otherwise highly mobile molecules can be studied. Molecular orientation is a popular target for such studies because TERS selectively enhances Raman signals from vibrations whose polarizability tensors are parallel to the tip axis. Using low temperature UHV-TERS, Jiang et al. identified mobile and stationary phases of a perylene-diimide-based molecule deposited on Ag(100), and they showed that the molecules in each phase had different orientations, which could explain the differences in their mobility.⁹¹ Even the orientational dynamics of molecules on a surface can be identified by examining correlations between the intensity fluctuations of Raman modes with different symmetries. For example, as shown in Fig. 4(d), at low temperature, the intensities of the A₂ and B₁ Raman modes of malachite green on gold at 90 K are anticorrelated in measurements taken on the timescale of seconds.⁸⁷ Since the anticorrelated modes represent different symmetries of the same out-of-plane C–H bend, the result suggests that the molecule is rotating on that timescale to alternately orient each of the symmetries along the tip axis.

A molecular switch is a molecule that can undergo a reversible intramolecular process. The main benefit of using TERS to study switching properties is that the simplicity of the reaction, combined with the ability of TERS to examine only a few molecules on a surface, is ideal for identifying specific substrate properties that affect the reaction efficiency. In that sense, switching reactions are one step more complicated than thermally driven diffusion and conformational changes. In fact, similar to the statistical studies of diffusion, one of the first TERS studies of molecular switching examined the isomerization of azobenzene bound to the tip by a long alkyl chain.⁹² By monitoring the TERS signal in time, the authors showed that bursts of isomerization events would occur, which they attribute to the conformational sampling of the chain bringing the azobenzene moiety into a region where its isomerization is more active. The occurrence of a switching event turned out to be connected to the bias potential of the STM, but the authors could not clearly identify the nature of the hot spot. This is an especially challenging issue in TERS, and it is even more difficult for light-driven reactions, which may be initiated by direct excitation of the molecule, SPR excitation of the tip or the substrate, or heat generated in the tip–substrate gap. A heat-driven reaction has been observed in a UHV-TERS study of a helicene, in which only the dehydrogenated form of the helicene could be detected by TERS.⁹³ Since only the part of the molecule facing the tip was dehydrogenated, the authors suggest that the reaction was driven by excess heat at the tip.

If heat can be eliminated as a factor, one can consider other things that may affect the reaction, such as substrate-mediated hole transfer from the d-band of the metal substrate after photoexcitation. There are two examples of this. In one study, variation in d-band energy between different metals was used to explain why hydrazone isomerization occurs in the presence of 340 nm excitation light on Au and Cu, but not on Ag.⁹⁴ In another, variation in d-band energy between steps and terraces was used to explain why isomerization of a SAM of azobenzene thiol on a stepped gold surface proceeded more efficiently on steps than on terraces.⁹⁵ 

Although the title of this section is generalized to refer to dimerization reactions, we are only aware of one reaction system in this category that is studied by TERS. However, it is so widely studied that it warrants its own section. The conversion of p-nitrothiophenol (pNTP) or p-aminothiophenol (pATP) to dimercaptoazobenzene (DMAB) is an important, prototypical plasmon-induced reaction that has been extensively studied by SERS and hence became a natural target for TERS as well. SERS of the precursors, pNTP and pATP, has been examined for a long time,⁹⁶ but it was not until 2010 that the plasmon-induced photo-oxidative coupling of pATP to form DMAB was demonstrated,⁹⁷ and it was two more years before a comparison of SERS spectra simulated from DFT calculations with data from the literature showed that pNTP could be photo-reduced to DMAB.⁹⁸ Interest in these reactions has continued, in SERS, with one study examining the reaction kinetics,⁹⁹ another demonstrating that photogeneration of singlet oxygen is the first step in the photo-oxidation of pATP,¹⁰⁰ and a 2018 study of the ultrafast reaction dynamics confirming that photo-induced heating does not significantly contribute to the reaction.¹⁰¹ 

While these early SERS studies have done much to elucidate the mechanism of this complex reaction, the small sampling volume of TERS enables a deeper investigation. Several experiments have been carried out to develop the capabilities of TERS for interrogating this reaction, and this body of work forms the basis of many other TERS studies of dynamic processes. In the first TERS study of the formation of DMAB from a monolayer of pNTP on a silver surface,¹⁰² a 532 nm laser was used to initiate the reaction, while a 633 nm laser was used as a probe [Fig. 5(a)]. As shown in Fig. 5(b), after the sample is irradiated at 532 nm for ∼30 s, several new Raman bands are observed, consistent with the formation of DMAB, and the monolayer of pNTP becomes unstable, as indicated by the fluctuations of the observed bands. Shortly after this work, TERS experiments carried out in UHV were used to confirm that the tip plasmon resonance, rather than heat, catalyzes the conversion of pNTP to DMAB.^103,104 The authors show that heat generated during the plasmon excitation rapidly diffuses away from the tip–substrate gap. These early single-point measurements on relatively smooth surfaces were later complimented by TERS mapping on a rough silver surface. This mapping study not only demonstrated the presence of regions of high reactivity on the surface but was also able to map the regions of high activity with 20 nm spatial resolution.⁵⁵ The correlation between edges on the surface and enhanced plasmonic activity was already apparent by this time, but the work demonstrated the capability of TERS mapping for identifying the surface features responsible for the highest activity of a reaction.

Understanding the origin of high reactivity in some regions by identifying the many contributions to the reactivity has been the goal of more recent TERS studies. One contributing factor is the higher electric field strength near the edges and corners of a metal substrate, which enhances the yield of a plasmonic reaction. The actual field strength can be measured by adsorbing to the tip or the substrate molecules whose Raman signals are sensitive to the local electric field and scanning the tip over the region of interest.^30,106–108 A comparison of the map of the local electric field to the distribution of the DMAB product generated from pNTP on a smooth gold plate in solution reveals that the regions of high reactivity do not always match the spatial distribution of the electric field intensity.¹⁰⁹ Moreover, in some regions, there was no correlation between the Raman bands associated with reactants and those associated with the product, indicating that changes in molecular orientation and diffusion out of the reaction zone may also affect the reaction yield.

This ambiguity points to the need to examine other factors that may contribute to the reaction. One contributing factor is relative orientation of the adsorbed molecule on the surface, which is expected to vary at edges and rough surfaces. TERS has a unique advantage over SERS in that it does not require a roughened substrate, so it can be used to examine molecules adsorbed on single crystal substrates, where nanoscale surface roughness does not contribute to the reaction. Examining the conversion of both pATP and pNTP to DMAB on single-crystalline Au(111) and Ag(111) surfaces shows that the reactivity on the smooth surfaces is different from that on roughened substrates of the same metals.¹⁰⁵ On Au(111), the formation of DMAB from pATP can be observed after just 5 s of laser irradiation [Fig. 5(c)], whereas on Ag(111), pATP is not converted to DMAB, even after 10 min [Fig. 5(d)]. Similar results were obtained for pNTP, except that on Ag(111), adsorbed pNTP was not detected. This is attributed to the different relative orientations of the molecules on the different metals, which were determined by DFT calculations and supported by the relative TERS intensities of the reactants. In the case of Au(111), calculations reveal that pNTP and pATP are tilted on the surface, bringing their N-containing moieties together at an angle similar to that of the CNN bond angle of DMAB, as shown in Fig. 5(e). Thus, the initial tilt angle of the precursor molecules ensures that DMAB can be formed with minimal molecular strain. In the case of Ag(111), pATP is adsorbed in a nearly vertical orientation, which would lead to the formation of DMAB with a CNN bond angle of 90°, as depicted in Fig. 5(f), resulting in prohibitively high molecular strain, so the reaction fails. The calculations also explain that pNTP cannot be detected on silver because it adsorbs via its –NO₃ group, making it nearly parallel to the surface. From these results, it is clear that TERS offers a unique way to isolate many aspects of this reaction. Although determining all factors contributing to the reaction on a complex surface remains a work in progress, we envision that the investigations on single-crystal surfaces will help to unravel the reaction mechanisms.

TERS studies of the dimerization of N-aromatic thiols to form azo-products are not limited to fundamental research. The plasmon-assisted conversion of pATP to DMAB is sensitive to pH,¹¹⁰ which means that it can potentially be used as a sensor for the local pH, if the molecule is attached to a TERS tip. A recent proof-of-concept study was carried out, in which this approach was used to image the local pH distribution over a rough, NH₂-functionalized glass substrate.⁷⁶ 

Although we have focused on using TERS to study reactions in situ, recent studies of prominent ex situ systems can point toward promising future applications. As demonstrated in studies of the formation of DMAB,⁵⁵ TERS is ideal for identifying surface regions of high reactivity. This capability was used in a study of the generation of active oxygen species on a surface consisting of a Pd monolayer deposited on a Au(111) surface and the subsequent diffusion of the active oxygen species.¹¹¹ To do this, a thiolate molecule was adsorbed onto the surface, and the surface was placed in a solution containing H₂O₂ to generate active oxygen species that could oxidize the thiolate. After the sample was removed from the solution, a TERS image of the surface showed a reduced Raman signal from the thiolate near the Pd step edges, indicating that the edges are more active and revealing the diffusion distance of the reactants from the step edge [Figs. 6(a) and 6(b)]. A similar approach was used to investigate hydrogen spillover from Pd island catalysts on a Au(111) surface.¹¹² In this case, chloro-nitrobenzene thiol (CNBT), which is converted to chloro-aminobenzene-thiol (CABT) upon hydrogenation, was adsorbed to the bimetallic surface and used as the probe of the reaction. After exposing the catalyst to H₂ gas, the CNBT on the Pd islands was converted to CABT, but the CNBT on Au was mostly unaffected since Au cannot catalyze this reaction. However, using TERS mapping, the authors found that the CNBT on Au atoms within 15 nm–30 nm of a Pd island had reacted, indicating the range of hydrogen spillover from Pd.

Ex situ TERS has also been used to examine lithium ion battery chemistry by mapping the chemical composition of the solid electrolyte interphase (SEI) accumulated by galvanostatic cycling of a Si electrode into a solution of ethylene carbonate, with LiPF₆ as the electrolyte.¹¹³ Every few cycles, a TERS image of the surface was collected in order to show the distribution of different chemical species in the SEI [Fig. 6(c)]. A mosaic of different chemical components was detected in each image of the SEI, and this mosaic changes with an increase in SEI thickness, indicating that under the reaction conditions used in the study, the SEI grows with a layered structure, in which the mosaic-like composition of individual layers changes with thickness. All three of these systems represent forays into practical applications that are currently challenging to study in situ using TERS. However, as the capabilities of TERS continue to develop, greater insights into these processes may be revealed.

With the development of EC-TERS, it finally became possible to examine electrochemical reactions with nanoscale spatial resolution.^22,23 This development provides a new parameter space to explore, and it opens the way for TERS to contribute to practical applications, such as electrocatalysis. EC-TERS introduces the possibility of examining the effects of nanoscale electrode heterogeneity on a reaction, which could be directly used to optimize the design of electroactive materials. Although EC-SERS was developed 50 years ago^114,115 and solution phase TERS was demonstrated 11 years ago,⁶⁹ EC-TERS was only recently developed. The technique introduces many new complications to these established techniques, even under ideal conditions. Therefore, most of the work in EC-TERS consists of fundamental studies of relatively simple systems.

The simplest electrochemical systems are non-Faradaic processes, such as the potential-dependent protonation and orientation of adsorbed molecules. Potential-dependent EC-TERS was first demonstrated using the protonation/deprotonation reaction of 4-PBT [(4′-(pyridin-4-yl)biphenyl-4-yl)methanethiol] molecules adsorbed on a Au(111) surface.⁷⁰ When the electrode potential is reduced from 0.4 V to −0.7 V vs Ag/AgCl, a substantial decrease in the relative intensities of the peaks at 1592 cm⁻¹ (pyridine ring) and 1605 cm⁻¹ (benzene ring) is observed, as shown in Fig. 7(a). Protonation of the nitrogen end of the 4-PBT molecule [depicted in Figs. 7(b) and 7(c)] shifts the 1592 cm⁻¹ band to higher frequency, causing it to overlap with the 1605 cm⁻¹ peak. This potential-induced change is invisible in a SERS study of 4-PBT because the SERS substrate is a more heterogeneous surface, and the enhanced Raman signal includes contributions from molecules whose conformations prevent their protonation. In contrast, the EC-TERS experiment samples a small number of molecules on a single crystal surface. Thus, compared with EC-SERS, EC-TERS can more faithfully reflect a very subtle change in the molecular configuration. The ability of EC-TERS to detect the protonation and deprotonation processes is of practical significance for studying well-known proton-coupled electron transfer reactions.

Because TERS enhancement is strongest for polarizability tensors oriented along the tip axis, TERS can also be sensitive to changes in molecular orientation, which can be determined from changes in the Raman band intensities. In electrochemical experiments, changing the applied potential can change the orientation of molecules adsorbed on the electrode. Potential-induced molecular reorientation can be studied using the surface selection rules of SERS,^116,117 but the large sample volume of SERS averages together the Raman intensities from large ensembles of molecules with different orientations, which lowers the accuracy of EC-SERS measurements of molecular orientation. With EC-TERS, a much smaller domain is sampled, allowing for more quantitative determination of the molecular orientation. The accuracy of EC-TERS for determining molecular orientation was demonstrated using the potential-dependent reorientation of adenine adsorbed on Au(111).¹¹⁸ The average orientation of an ensemble of 50–70 adenine molecules was determined by measuring potential-dependent Raman intensities of the ring breathing mode at 723 cm⁻¹ [Fig. 7(d)] and the 1523 cm⁻¹ band [Fig. 7(e)] corresponding to several vibrations within the ring (N7–C8 stretch, C8–H bend, and NH₂ scissor). The experimental data for the protonated and deprotonated forms were compared with calculated Raman intensities for different tilt angles of adsorbed adenine. The good agreement between the experimental data [points in Figs. 7(d) and 7(e)] and theoretical results (lines) demonstrates the power of EC-TERS for interrogating the electrode–solution interface.

Fundamental electrochemical reactions are those involving oxidation or reduction that has been frequently studied in the literature. These reactions are relatively simple, making them ideal processes for studying the basic components of electron transfer with nanoscale resolution. A key goal of this type of research in TERS is to better understand how the local environment can influence electron transfer processes. A series of studies using EC-AFM-TERS to examine the reduction of Nile Blue on an ITO electrode was designed to do just that. Nile Blue undergoes a two-electron, one-proton reduction, as shown in Fig. 8(a). A cyclic voltammogram of Nile Blue on ITO exhibits a single, large reduction wave [Fig. 8(b)]. EC-TERS was used to monitor the reduction locally via changes in the intensity of the Raman band at 591 cm⁻¹, which corresponds to the oxazine ring stretch.⁷¹ The significance of the local activity of the electrode can be seen in the differences between the electrochemical response, which represents the average voltammetric behavior over the entire electrode, and the TERS voltammogram, which represents reduction of Nile Blue in the local area of the tip. Plots of the TERS voltammogram from four locations on the ITO electrode are shown in Fig. 8(c). In some regions, the TERS voltammogram shows two separate reductions. The number of sites at which this occurs decreases with a decrease in surface coverage, but they are still observed at 10⁻⁵ ML, which suggests that Nile Blue preferentially adsorbs to specific kinds of sites on the electrode. By extracting the voltammograms corresponding to single redox events, the variation in the formal redox potential of isolated molecules at different sites on the electrode surface can be observed.¹¹⁹ A histogram showing the frequency of sites corresponding to each redox potential [Fig. 8(d)] shows that the distribution of reduction potentials is broader than the distribution of oxidation potentials, indicating that reduction of the cationic form of Nile Blue is more sensitive to the local environment than the oxidation of the neutral form. This method of extracting the site-dependent formal potential has also been applied to a plate of Au(111) on ITO.¹²⁰  Figure 8(e) shows the distribution of the formal potentials on the electrode surface, with a dashed line indicating the boundary between Au and ITO. Both materials show wide variation in the formal reduction potential of Nile Blue over their surfaces. As shown in Fig. 8(f), the distribution is narrower for Au(111) than it is for ITO, indicating greater heterogeneity of the polycrystalline ITO. In addition, the distribution on ITO is best fit with a bimodal Gaussian, which could be caused by contributions from molecules on different crystal grains on the ITO surface, which may have a different distribution of reactive sites with different distributions of formal redox potentials. On average, the formal potential is 4 mV more negative on Au than on ITO. These site-dependent surface properties are not detectable by pure electrochemistry, and they highlight the advantages of using TERS to examine electrochemical processes.

In addition to understanding the basics of electron transfer, fundamental electrochemical reactions are used for testing new experimental setups and understanding processes that may affect more complex reactions. An EC-TERS study of anthraquinone on Au(111) showed that the combination of laser light and applied potential causes the reduced form of anthraquinone to decompose, making a normally reversible electrochemical process quasireversible.⁸² This laser-induced effect should be carefully examined when studying light- or hot-carrier-sensitive systems. In contrast, thiobenzene was shown to be very stable prior to its oxidation,⁷³ and it was suggested as a possible EC-TERS standard.

Another configuration used to examine fundamental electrochemical processes is EC-tip-SERS, in which the molecule is adsorbed on an STM tip, and the tip bias drives the electrochemical process. This configuration was first used to study the electrochemical reduction of pNTP on a gold tip.⁷⁹ The ability to independently tune the potentials of the tip and substrate makes it possible to use the tip voltage to transfer the molecule between the tip and the substrate¹²¹ or to regulate its redox state.¹²² 

The challenge of studying problems related to practical applications, such as catalysis, is that most systems of interest cannot be examined without the detection of short-lived species generated in the small sampling volume of TERS. An effective catalyst has a high turnover rate, which means that the reactants, products, and intermediates do not reside on the electrode surface for long, leaving very little time for them to be detected. While this alone results in weak signals, the problem is exacerbated in TERS because the probability of the reactants diffusing through the sampling volume is very low.

For these reasons, direct detection of reaction intermediates is challenging, as demonstrated by efforts to detect oxygen reduction reaction (ORR) intermediates on a phthalocyanine catalyst. For the cobalt phthalocyanine (CoPc) catalyst, the reaction intermediate, a complex between oxygen and Co, is relatively stable under vacuum and can be detected by TERS.¹²³ However, intermediates in solution were not observed in solution by EC-TERS or EC-STM presumably due to the combination of rapid consumption of oxygen and its infrequent diffusion into the region below the TERS tip.¹²⁴ Alternatively, indirect evidence for oxygen adsorbed to FePc can be obtained in solution¹²⁵ because the intensities of certain Raman bands of the FePc molecule change when a reducing potential is applied for a short time in an oxygen-saturated solution. The change in band intensities is a result of the formation of a nonplanar geometry caused by the adsorption of O₂ to the metal center.

Catalytic reactions such as water splitting can produce oxides on the electrode surface that are stable under the applied potential and can be more easily detected by EC-TERS. Recently, the oxidation process was examined on a Au(111) electrode.⁸³ When a gold electrode in a solution of 0.1M H₂SO₄ is held at the non-oxidizing potential of 1.1 V vs Pd–H, in situ STM images show that the gold surface remains smooth [Fig. 9(a)], and the gold oxide band at 560–580 cm⁻¹ cannot be seen, as shown by the EC-STM-TERS image of the band integral in Fig. 9(b). When the oxidizing potential of 1.45 V vs Pd–H is applied, Fig. 9(b) shows that the Au oxide band immediately becomes visible and that it immediately disappears when the potential is returned to 1.1 V. The appearance of the oxide band is accompanied by a change in the surface structure of the electrode [Fig. 9(c)], and a TERS map of this structure reveals the distribution of different gold oxide species [Fig. 9(d)]. In order to examine the role of gold nano-defects in the formation of gold oxide, the electrode potential was switched from 1.1 V to 1.45 V while scanning over a nano-defect on the Au (111) surface [Figs. 9(e) and 9(f)]. The formation of oxides primarily at the defect confirms its role in the formation of gold oxide. Moreover, the peak of the gold oxide Raman band ranges from 550 cm⁻¹ to 580 cm⁻¹, depending on the position of the nano-defect, suggesting two different oxide species distributed unevenly throughout the nano-defect. The authors suggest that Au₂O₃ forms primarily on terraces of the defect and Au₂O forms primarily on sharp protrusions.

To observe intermediates that have short lifetimes, an indirect detection method involving measurement of a molecule that reacts with the short-lived intermediate has been the most successful. A recent experiment to image the diffusion of hot carriers [Fig. 9(g)] generated on a Au(111) substrate demonstrates the use of this scheme.¹²⁶ To generate the carriers in the metal, a silver tip positioned over the surface was irradiated with 632.8 nm light, while the potential of the Au(111) substrate was maintained near open circuit. Under this condition, the monolayer of para-mercaptobenzoic acid (pMBA) in the tip-substrate gap is oxidized to thiophenol (TP) by the photogenerated holes in the substrate. A TERS map of the thiophenol, therefore, provides a map of the diffusion of the holes. In order to image the diffusion length of the holes, the potential of the substrate was lowered to −0.7 V to prevent further photooxidation of pMBA so that the spatial distribution of reaction product TP is maintained during TERS mapping. Therefore, when the tip is scanned over the surface, the reaction profile can be obtained from the TERS signal of the converted TP, as shown in Fig. 9(h). In this way, the TERS map of TP reveals how far the holes had diffused. Although hole diffusion occurs in any environment, the mapping of hole diffusion by TERS was possible only because EC-TERS provides a means of controlling the electrode potential.

TERS has undergone impressive advances in recent years, and its usefulness as a technique for examining chemical dynamics in situ has grown considerably. In addition to several promising developments in the understanding of the fundamental dynamics of reactions, TERS is beginning to move beyond the examination of simple model systems and into studies of more complex processes with potential applications. Furthermore, several recent developments in TERS methods have made new kinds of reactions accessible to the TERS study. Nonetheless, the field is still quite young, and there is much room for further development of TERS as a means of examining dynamic systems. In particular, improving the time resolution of TERS and introducing TERS into other reaction environments will make TERS a more versatile tool for studying nanoscale dynamics. However, the low signal strength of TERS is a major limiting factor for both of these directions and represents a third path that needs further development. Sections V A–V C focus on how TERS can develop in each of these three directions and the opportunities that these developments would create.

Greater signal strength is needed to improve the time resolution of TERS and to enable the study of more complex dynamical systems. In addition to the issue of time resolution, TERS signals are often too weak to detect without gap mode operation, which limits experiments to those systems involving metal substrates. It is therefore essential to boost the signal strength of TERS experiments so that they can be more easily used to examine poorly conductive materials, such as biological materials and semiconductors. The signal strength in TERS can be increased by improvements either to the optical configuration of the experiment or to the tip.

A typical TERS experiment involves focusing low power, continuous wave laser light through a microscope objective onto the tip apex, with the laser light polarized parallel to the tip axis. There are several adjustments to this standard optical configuration that can increase the enhancement of the TERS signal. Some basic improvements include optimizing the polarization of the incident and detected light for maximum contrast between the near-field and far-field signals¹²⁷ and, with a tunable laser, choosing laser wavelengths that best optimize the overlap between the tip or the gap plasmon.¹²⁸ Several recent developments, including using radially polarized light instead of linearly polarized light,^129,130 using a Fabry–Perot cavity to support the sample,¹³¹ and optimizing the phase of the laser beam cross section using a spatial light modulator,¹³² have shown 5–10× enhancement of the TERS signal. The combined effects of these optical modifications could significantly boost the signal-to-noise ratio over that of a typical experiment.

Drastic improvement in signal strength can be achieved using stimulated Raman spectroscopy, which was previously reported to achieve 10⁹ enhancement in the TERS signal.¹³³ Although the narrowband stimulating laser used in this experiment only allowed the observation of one Raman band at a time, the experiment could be done using a broadband light source, as was demonstrated for a related experiment combining coherent anti-Stokes Raman spectroscopy (CARS) with TERS.¹³⁴ One drawback in using stimulated Raman spectroscopy to enhance the TERS signal is the strong background signal that comes from the metallic tip and substrate, which complicates interpretation of the data obtained using this method. Therefore, more data processing and methods for reducing the background are needed to make this approach more generally applicable.

The other major route for improving the signal-to-noise ratio in TERS is to improve the TERS tip by changing it in ways that increase its stability, enabling longer acquisition times, or that increase the signal strength. With precise control of the tip shape, it is possible to increase the TERS enhancement factor by optimizing the focusing of light at the tip apex or optimizing the overlap of the surface plasmon resonance (SPR) maximum with the laser and the Raman signal. Such precise control is usually achieved by focused ion beam milling, which, although effective, is also slow and expensive.^33,64 Therefore, it is necessary to develop more practical methods for controlling the tip shape. Reproducible STM tips with high enhancement factors can be produced in batches by electrochemical etching followed by field-directed sputter sharpening,^39,65 but additional methods that are suitable for AFM tips and allow for precise shape control are also needed.

An alternative to reshaping the tip is to modify its optical properties using a dielectric coating. With a sufficiently high refractive index, a dielectric coating can improve coupling of the light to the tip, making the enhancement factor higher than that of a bare metal tip.^57,58 A dielectric coating is also ideal for protecting TERS tips from the environment. It increases the shelf-lives of silver tips from days to weeks, and it makes TERS tips more durable.^77,78 Therefore, developing methods for coating TERS tips with materials that have high refractive indices will not only make it easier to introduce TERS into harsher media, which is necessary to study dynamic processes under reactor conditions, but will also increase the signal strength of TERS measurements.

Although there are several methods for improving TERS signal strength, many species generated during electrochemical and biological processes will remain difficult to detect by TERS due to their exceptionally small Raman cross sections and short dwell times in the region under the tip. These include the products and intermediates of important catalytic reactions, such as CO₂ reduction and O₂ reduction. For species such as protons, electrolytes, and products, an alternative is to modify the tip with molecules that have large Raman cross sections or fluorescence quantum yields. These molecules can function as indirect sensors by specifically interacting with the species of interest in ways that change the sensors’ Raman or fluorescence signals. This approach has already been demonstrated in a study in which the pH distribution was mapped over a roughened, functionalized surface.⁷⁶ The spatial- and pH-resolution reported in this study has yet to be improved. The usefulness of this approach can be expanded by developing sensor molecules for detecting other materials, such as ions and reaction products with better spatial resolution.

A central aspect of dynamical studies is time resolution. In typical TERS experiments, a one-second acquisition time for each data point gives reasonable signal strength. This 1 s time resolution is sufficient for observing dynamic processes under steady-state conditions and random molecular motion at low temperatures. However, better time resolution would open a plethora of new possibilities. In addition to exploring more rapid molecular fluctuations, the dynamics of a system returning to equilibrium after a laser pulse or a potential pulse can be used to understand processes such as surface diffusion, electrolyte reorientation in response to an applied potential, or electrode reconstruction. With microsecond acquisition time, it would be possible to introduce video-rate TERS, in which hyperspectral images showing the dynamics of the entire surface are obtained almost simultaneously, as is currently done with video-rate SPM.¹³⁵ The most direct way to improve time resolution is to acquire stronger signals so that shorter acquisition times can be used. Based on the methods currently available, improving the signal strength by the several orders of magnitude needed for microsecond acquisition times will likely require stimulated Raman spectroscopy. For irreversible processes, boosting the signal strength is the only option for improving the time resolution or imaging speed when monitoring TERS dynamics. For processes that are initiated by a voltage or light pulse and quickly return to their initial state, multiple acquisitions at each time delay following the pulse can be averaged together, allowing for an improved signal-to-noise ratio despite weak signals.

Observing ultrafast dynamics using TERS would combine the highest spatial resolution with the highest temporal resolution, enabling the observation of electronic transitions and structural changes on the nanoscale. This is the scheme needed to study the elementary steps of plasmon-induced reactions at specific reaction sites or to observe single-molecule photoswitching. Processes such as hot carrier or exciton diffusion could also be monitored in real time by spatially separating the nanoscale pump and probe volumes, for example, by using a dual-tip configuration.¹³⁶ At present, most efforts to combine plasmonic enhancement with ultrafast spectroscopy have been SERS studies of molecular systems.¹³⁷ Research in this area includes pump–probe SERS,¹³⁸ CARS,^139,140 femtosecond stimulated Raman spectroscopy,^141,142 and impulsive stimulated Raman spectroscopy.^143,144 Each of these methods has specific advantages, but at the time of this writing, only CARS has been combined with TERS to study dynamics.¹³⁴ Based on the ultrafast SERS research, there are several specific challenges for combining each of these techniques with plasmonic enhancement. However, the most common difficulty is that combining ultrafast spectroscopy with TERS involves exposing a small, stationary region of a sample under the tip to thousands or even millions of intense laser pulses, sometimes for several minutes. This requires high stability of the sample, the tip, and the tip position. Recent ultrafast SERS experiments have developed techniques for lowering the required pulse power for femtosecond stimulated Raman¹⁴² and impulsive stimulated Raman spectroscopies,¹⁴⁴ but these methods have not been attempted under the conditions that would be required to study ultrafast dynamics with TERS. Nonetheless, the successful demonstration of CARS-TERS shows that using TERS to examine ultrafast processes is possible, but more work is needed to establish the optimal conditions for the system and the sample.

Given that much of the literature on TERS studies of dynamics is electrochemical and concerned with catalytic reactions, a long-range goal is to be able to extrapolate a TERS study of dynamics to the performance of a catalytic reactor. The challenge in doing so is that there are many fundamental differences between systems examined using TERS and those of a functioning reactor. One of the major issues is that catalytic reactors are usually three-dimensional systems, often with porous catalysts, in which reactants and products diffuse freely, whereas TERS studies need to be performed on relatively flat surfaces. In addition, even the studies of dynamical systems rarely involve observing diffusion in situ due to the difficulty in measuring TERS signals for molecules with short dwell times. Developing video-rate TERS imaging, as discussed in Sec. V B, would enable direct observation of changes in different parts of a catalyst in real time. However, this solution has only limited ability to address the complexity of a real reactor. For this reason, more work is needed to make TERS more quantitative and to introduce TERS into more complex reacting systems.

The development of quantitative models of TERS for reacting systems would need to deconvolute intensity fluctuations caused by molecular diffusion and reorientation from those caused by other sources of instability, such as atomic diffusion on the tip.¹⁴⁵ Such development is further complicated by the fact that even under ideal conditions, many factors affect the relative intensities of Raman bands in TERS, so there is a need to build models using simpler systems operating under reactor conditions. 2D materials, such as graphene and transition metal dichalcogenides (TMDCs), make excellent testbeds for developing quantitative TERS under the conditions of a chemical reaction. Because the structures are highly planar, they can provide uniquely detailed structural and electronic information. For example, high resolution TERS mapping allows the observation of strain and doping effects in graphene^54,146 and TMDCs.¹⁴⁷ Defects and edges in 2D materials are also identifiable because of their unique phonon scattering properties, which give rise to different Raman selection rules than those of the pristine material, resulting in distinct Raman bands.^148–150 2D materials are not only ideal for quantitative TERS but are also particularly useful for the purpose of understanding chemical reactions because they are also promising catalysts in solution due to their optimal binding energies for reactants and products.^151–153 Now that performing TERS in solution and under electrochemical conditions has become routine, the effects of these nanoscale features on chemical reactions on 2D materials can be examined directly.

At the same time, TERS should be introduced to the conditions of other kinds of reactions in order to expand its capabilities. Although SPM has been used in situ to examine the nanoscale structures of nonaqueous systems, such as lithium batteries and ionic liquids,^154–156 at present, there is only one ex situ TERS study of the electrochemistry of lithium-ion battery materials.¹¹³ The results of these previous studies demonstrate the importance of understanding the nanoscale structural changes in lithium ion batteries. TERS measurement in nonaqueous liquids requires performing TERS in a controlled-atmosphere box.^53,54 This opens the way not only for lithium ion battery measurements but also for many other important air- and moisture-sensitive reactions. Non-aqueous TERS has already been demonstrated in hexadecane,¹⁵⁷ and the principles are similar to those of TERS experiments in aqueous solutions. Depending on the liquid, different coatings may be needed to protect TERS tips from the environment. With these issues resolved, important chemical reactions, such as those occurring in lithium ion batteries and ionic liquids, will be accessible for an in situ study, and our understanding of these reactions will benefit from the chemical insights provided by TERS.

The authors acknowledge the financial support from the MOST of China (Grant No. 2016YFA0200601) and NSFC (Grant Nos. 21633005 and 21790354). M.M.S. was supported by the iChEM Fellowship Program.