STM-Tip-Enhanced Raman Spectroscopy toward Single Molecule Scale

Chapter 7

Downloaded by UNIV OF CALIFORNIA SAN DIEGO on December 26, 2016 | http://pubs.acs.org Publication Date (Web): December 20, 2016 | doi: 10.1021/bk-2016-1246.ch007

STM-Tip-Enhanced Raman Spectroscopy toward Single Molecule Scale Rafael Buan Jaculbia, Kuniyuki Miwa, and Norihiko Hayazawa* Surface and Interface Science Laboratory, RIKEN, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan *E-mail: [email protected]

In this chapter, we describe the route of tip enhanced Raman spectroscopy towards the single molecule scale. We pay particular attention to the use of the scanning tunneling microscope tip for enhancement of the Raman signal, a technique popularly known as STM-tip enhanced Raman spectroscopy (STM-TERS). A review of pioneering works on STM-TERS as well as recent results on the topic is given. We also provide our insights on the future direction of this topic both towards potential improvements of the methodologies to achieve even better spatial resolutions and possible applications to different materials.

Introduction Recently, there have been tremendous developments in material science and its growth technologies of nanoscale materials, which turns out to show new characteristic features, distinctly different from those of its bulk counterparts. However, in contrast to the rapid growth of fabrication techniques, the progress in analytical techniques for such nanoscale materials has been slow. In order to characterize the local properties of the material, in particular, with nanoscale heterogeneity, the spatial resolution has to go beyond the scale of the heterogeneous distribution in the nanoscale. Tip-enhanced spectroscopy has been recently recognized as one of the promising analytical tools that can access local properties of the materials both through topographic and chemical contrasts. This is based on the fact that tip-enhanced spectroscopy is the synergy of the virtues of both scanning probe microscopy (SPM) and optical spectroscopy. In this sense, © 2016 American Chemical Society Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


the high spatial resolution beyond the diffraction limit of light can be determined by SPM while the high chemical contrast is provided by a variety of optical spectroscopies. So far a number of tip-enhanced spectroscopy techniques, such as fluorescence (1–4), two-photon excited fluorescence (5–7), infrared absorption (8, 9), Raman (10–13), and nonlinear Raman (14–16) have been reported and for sure more spectroscopic techniques will be employed for tip-enhancement in the near future. Among the various tip-enhanced spectroscopic techniques, tip-enhanced Raman spectroscopy is the first reported (10–13) and the most developed spectroscopy as it is now recognized as TERS because of the analogy with the well-known surface enhanced Raman spectroscopy (SERS). At the early stage of TERS, mainly atomic force microscopy (AFM) either by cantilever based (10, 11) or tuning fork based (12, 17) feedback control as SPM is employed. One of the reasons why AFM has been initially employed is the fact that the pioneering groups of TERS originated from photonics research so that ambient conditions and no special sample treatment are preferred. Due to this, AFM characterization has been already one of the common analytical tools in combination with optical spectroscopies so that the poor spatial resolution of conventional optical microscopies can be compensated by AFM topography. Since the invention of TERS in 2000, lots of groups have joined the development of TERS and there have been a kind of competition both for higher spatial resolution and higher sensitivity towards single molecule level. Lots of efforts have been devoted such as optimization of polarization control (18, 19), tuning the tip-plasmon resonance (20–23), photon confinement by nonlinear optics (14, 15), and combination with scanning tunneling microscopy (STM) (13). However, there has been a big barrier to break the spatial resolution limit of 10 nm in TERS imaging whereas several groups have reported single molecule sensitivity (24–28). One of the major reasons why the spatial resolution is limited by 10 nm can be due to the quality of tips in the sense that the full potential of the tip-plasmon has not been fully utilized due to the limited hot spot engineering since it has been impossible to apply post-tuning of the tip-plasmon actively on-site during the experiments. The other major reason is the stability of the system. Especially in the case of AFM, which are mostly used in ambient conditions, the thermal drift of the entire system and the relatively large dynamic range of PZT scanner used (>50 μm) seriously limit the position stability. In order to tackle these two issues, in this chapter, special focus is given to the recent scientific challenges towards the extreme spatial resolution beyond the limit of 10 nm by STM-TERS followed by the pioneering work of Ertl and Pettinger (13, 29). In the case of STM, indeed the major drawback commonly accepted is that the sample and substrate has to be conductive (i.e. metal substrate is a requirement). However, the non-contact feature of the tip-substrate gap control and the higher stability of the system owing to the relatively smaller dynamic range of PZT used in STM ensure the precise control of the tip-plasmon. Moreover, the major drawback of the conductive substrate turned out to be a promising advantage for the excitation of so called gap-mode plasmon, which has been recently recognized as a hot spot for single molecule SERS. Furthermore, as is discussed below, the capability of the post-tuning of tip-plasmon is very powerful for tip-enhanced 140 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


spectroscopy and unique for STM. Figure 1 shows the schematic of typical STM-TERS configuration. The essence of STM-TERS originated from the hot spot engineering of gap mode plasmon between metallic tip and substrate that is precisely controlled below 1 nm with sub-Å resolution using the STM. Moreover the tip can be exactly positioned over the target molecule with angstrom level precision by STM as well. This active three-dimensional gap control relative to the position of the molecule makes STM-TERS an extremely unique and distinct technique from gap mode SERS. In the following sections, we start from the brief review of gap mode SERS. Then, the application of gap mode plasmon is extended to tip-substrate system for STM configuration depicted in Figure 1. Particular attention is given to STM-induced light emission for the analysis of gap mode plasmon in tip-substrate system. After discussing the analytical background of gap plasmon of tip-substrate system, we will review the recent reports of STM-TERS both in ultrahigh vacuum-low temperature (UHV-LT) and ambient conditions. At the final part of this chapter, we will summarize the discussions and add some future aspects of the related researches.

Figure 1. The schematic of STM-TERS configuration.

Gap Mode SERS The concept of TERS has been developed from SERS, which was reported more than 25 years (30–32) before the first demonstration of TERS. The first reports of SERS were on molecules adsorbed on roughened electrode surfaces as shown in Figure 2(a), where the surface roughness was brought about by the anodization of the electrodes. Moskovits (33) however suggested that these 141 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


surface enhancement effects can also be possible on molecules adsorbed on surfaces covered with metal colloids. This brought about various researches both theoretically and experimentally on the subject. Of particular interest in our review here is the case of electromagnetic field enhancement at the “gap” of two metal nanoparticles so called the “gap-mode” (34, 35) as illustrated in Figure 2(b). Aravind et al. described theoretically that the interaction between the resonances of two spheres can cause a very high field at the region between the spheres (34). The experimental verification of Aravind’s theory has been clearly reported by Xu et al. (35) more than 15 years after. In their work, they studied hemoglobin molecules adsorbed on Ag nanoparticles and immobilized on a silicon wafer. Similar to what is commonly observed in SERS experiments only a few “hot spots” show detectable SERS intensities. In order to analyze this further, they investigated the physical characteristics of their sample using AFM and SEM. In these experiments, they found that while 80% of the surface is covered by single nanoparticles, none of these regions are SERS hot spots. Indeed SERS was only detectable on regions where there are either dimers (around 15% of the area) or trimers or large aggregates (around 5% of the area) of Ag particles. And also, among these hot spots, the regions where the dimers are oriented parallel to the incident electric field produced the highest SERS intensities. While they were unable to visualize it directly, they hypothesized that these dimers or trimers are actually connected by hemoglobin molecules that is, the hemoglobin is located in the gap of two metal particles where the electric field is enhanced leading therefore to strong Raman signals. Under these observations, they concluded that the majority of the enhancement observed in SERS is actually caused by the electromagnetic enhancement at the center of two particles. As we will see in the next section, this particle-particle electromagnetic enhancement mechanism can be extended to the case of a particle-flat substrate and therefore to a metallic tip-flat substrate system which is exactly the situation in STM.

Figure 2. Substrates used for SERS experiments. (a) Roughened surfaces used for the initial discovery. (b) Surfaces deposited with metal colloids. It was found that the enhancement can only be observed for dimers of the nanoparticles connected by the molecule. 142 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Gap Mode Plasmon in Tip-Flat Substrate System While SERS and its gap mode described above showed an unprecedented advantage in sensitivity, SERS has been still a spectroscopic technique for a long time and not a microscopic technique in the sense that the relative position between the metallic nanostructures and the target molecules is not controlled. In 1985, the idea of “surface enhanced optical microscopy” was proposed by Wessel (36) as illustrated in Figure 3. The metallic tip (Figure 3(a)) can be used not only for amplification of Raman signals but also for controlling the relative position of the tip onto the target by e.g. piezoelectric translators (Figure 3(d)) providing topographic imaging of a surface at a subnanometer special resolution. This concept is directly employed in STM-TERS based on gap mode plasmon in tip-substrate system. Focusing on the physical enhancement processes of the gap mode plasmon in tip-substrate system, amplification of Raman signals from molecules positioned under the tip is sensitive to the wavelength of interface plasmon modes localized near the tip-substrate gap region. In this section, we first discuss electromagnetic properties of localized interface plasmon within a simple model where a tip and sample are assumed to have spherical and planar geometry, respectively (Figure 4). The sphere-plane model has been widely used as a first approximation to a geometry corresponding to measurements of TERS as well as STM-induced light emission (STM-LE), where light emission from the sample is induced by the tunneling current of STM (37, 38). Next, we mention a computational progress toward a more advanced simulation using a complicated geometric model and/or an improved numerical method.

Figure 3. The optical probe particle (a) intercepts an incident laser beam, of frequency ωin, concentrates the field in a region adjacent to the sample surface (b). The Raman signal from the sample surface is reradiated into the scattered field at frequency ωout. The surface is scanned by moving the optically transparent probe-tip holder (c) by piezoelectric translators (d). Reproduced from (36). Copyright (1985 Optical Society of America). 143 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 4. Geometry for the sphere-plane model. The tip and sample are modeled by a sphere of radius R and a semi-infinite substrate with a planar surface, respectively. The tip-sample distance is given by d. Each material is characterized by a frequency-dependent dielectric constant ε(i)(ω), i.e., the dielectric constant takes the value ε(1)(ω) in the substrate, ε(0)(ω) in the surrounding medium, and ε(2)(ω) in the particle. Calculation of plasmon modes using the sphere-plane model has been reported by Rendell et al. in 1978 (39). Originally, the work has conducted to explain electroluminescence from tunnel junctions composed of small metal particles deposited on an oxidized metal film (40). In the analysis, each material is characterized by a frequency-dependent dielectric constant ε(i)(ω), i.e., the dielectric constant takes the value ε(1)(ω) in the substrate, ε(0)(ω) in the surrounding medium, and ε(2)(ω) in the particle. Assuming that the radius R is small relative to the wavelength of light, retardation effects were neglected. Then, a set of discrete resonant can be determined by solving Laplace’s equation subject to a set of boundary conditions at interfaces. Rendell et al. assumed that a charge distribution associated with the inelastic electron tunneling is localized near the region of closest contact between the particle and film, and reported the calculated luminescence spectra displays the peak, the energetic position of which corresponds to the energy of localized interface plasmon mode. In Figure 5, we show calculation results of the frequency ωn of plasmon modes for various values of the geometric parameters R and d. Here, we used experimental data for the complex dielectric constants ε(1)(ω) and ε(2)(ω) for bulk Ag material (41) and assumed that ε(0) is unity (ε(0) = 1). The frequency ωn becomes lower as R and d increases and decreases, respectively. Figure 5(c) shows spatial distribution of electric field E associated with localized interface plasmon modes n = 0, 1, and 2. Here, the z and r axes are along the direction normal and parallel to the plane surface, respectively. Both z and r component of electric field, Ez and Er, have a drastically fall off within a distance of (2dR)1/2 from a point immediately below is more than ten times larger than the amplitude the particle. The amplitude . This means that polarization direction of the electric field is dominantly normal to the substrate. 144 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 5. Dependence of the frequency ωn of interface plasmon modes on (a) the tip radius R and (b) the tip-substrate distance d. (c)(d) and (e) show the spatial distribution of electric field associated with the interface plasmon mode n=0, 1, and 2, respectively. Here, R and d are set as 30 nm and 0.5 nm, respectively. The z and r axes are along the direction normal and parallel to the plane surface, respectively. Left and right rows represent the results for the z and r components of electric field Ez and Er, respectively.

In Ref. (39), analytic solution of the Laplace equation was also reported employing further approximations. For a perfectly conducting particle (the plasma frequency of metal ωp(2) ≫ ω so that ε(2) → −∞) above a free-electron metal film with ε(1)(ω) = 1 − (ωp/ω)2 surrounded by a medium with a dielectric constant ε(0), the frequencies ωn of plasmon modes are given by 145 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


with n = 0, 1, 2 …, and β0 = arcosh[1 + d/R]. Since β0 ~ (2d/R)1/2 for d/R ≪ 1, a frequency ω0 of the lowest mode is given by ωp/(ε(0))1/2(d/2R)1/4, and the upper cutoff for n → ∞ is ωp/(ε(0) + 1)1/2 which corresponds to the short-wavelength plasma frequency of the interface between the medium with ε(1)(ω) and that with ε(0). The frequency ω0 is shifted to a lower energy side as the radius R and the distance d increases and decreases, respectively, varying as d1/4R−1/4. The lowest mode correspond an interface plasmon mode localized near a region of order beneath the particle (42). As shown in Figure 5, one can confirm that the tendency of plasmon properties discussed above is consistent with the results obtained by the numerical calculations. Nearly a decade after Rendell’s report, the sphere-plane model was utilized in the theory of STM-LE from a clean metal surface, proposed by Johansson et al. in 1990 (43). Here, light emission is attributed to the radiative decay of interface plasmons localized near the tip/vacuum/substrate interface that are excited by the inelastic electron tunneling between the tip and sample (44). The intensity of emitted light was simulated by proceeding the following two steps: (i) calculation of current density by solving the one-dimensional Schrödinger equation with a trapezoidal potential (Figure 6 (a)) and (ii) calculation of an electromagnetic field to determine the strength of the coupling between tunneling electrons and the electromagnetic field. In the later step, the authors utilized the sphere-plane model to describe the tip-sample geometry. The fields associated with the excitation of a localized interface plasmon are expected to be concentrated to the region between the tip and sample surface, and therefore the sphere-plane model would be useful to describe the geometry in that region (45, 46). Experimental and calculation results were shown in Figure 6(b). Here the theory of Johansson et al. was use to simulate luminescence spectra obtained from Ag, Au, and Cu surfaces with a W tip (47). The spectral shape and signal intensity were reported to be in good agreement with the experimental results. As discussed in the previous paragraph, for a model system with free-electron metal electrodes, the frequencies ωn of plasmon modes depends on the tip shape. Since the dielectric property of Ag over the range of 2-3 eV is free-electron-like, calculation results of ωn is sensitive to the tip geometry as well as some variation in spectral structure during consecutive experimental runs for Ag were reported to be observed. The real part of the dielectric constant of Au and Cu rapidly changes near 2.4 eV and 2.1 eV, respectively, due to interband transitions. As shown in Figure 6(b), both experimental and calculation results of luminescence spectra show a sharp cutoff in these energy ranges at low wavelengths. Effects of tip and substrate materials on luminescence properties of the system were further discussed for several combinations of tip (W, Ag, Au) and substrate (Ag, Au) materials (48). Energy of plasmon modes has also been investigated through an analysis of experimentally observed luminescence spectra. Wang et al. observed STM-LE 146 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


from a clean Au(111) surface using a gold tip (49). As shown in Figure 7, the luminescence spectra were fitted to a model function composed of three Lorentzian curves reflecting the three types of plasmon modes. Energies of plasmon modes were estimated to be about 1.39, 1.57, and 1.72 eV. According to eq. (1) and taking ωp=3.46 eV and d=1.0 nm, the authors fitted the estimated values to ωn, as 26.5 nm. Observation of STMand extracted the geometric parameter LE spectra and spectral analysis using the sphere-plane model provides easily accessible sources of information about the energy of interface plasmon modes localized near the tip-substrate gap region. Qualitative and readily analysis of localized interface plasmon modes would also be useful for the optimization of TERS measurement.

Figure 6. (a) One dimensional trapezoidal potential corresponding to the potential for free electron in tip/vacuum/substrate. The work functions, bandwidths, and Fermi energy are φtip, Wtip, and EF tip for the tip material as well as φsub, Wsub, and EF sub for the substrate material. (b) Experimental and calculation results for STM-LE from Ag(111), Au(111), and Cu(111) surface obtained with an Ir tip. Spectra in the topmost row were observed in the high-voltage field emission regime. Spectra in the tunnel regime (the bias voltage Vbias=2.8, 3.0, 3.6 V, the tunneling current It=10, 10, 100 nA) are shown in the middle row. The results of theoretical calculation for the emission in the tunneling regime using experimental parameters are presented in the bottom row. A tip radius of 300 Å was assumed as suggested by scanning electron microscopic images. The sensitivity of the detection system shown as a cashed line was included in the calculation. Reproduced from (47). Copyright (1991, American Physical Society).

After the theory of STM-LE from a clean metal surface reported by Johansson et al., several attempt to improve numerical methods and models that can treat a more complicated geometry have been proposed. Downes et al. calculated excitation (induced by inelastic electron tunneling) and deexcitation (caused by radiation and dielectric loss) rates of each electromagnetic modes (50). The information about these rates is useful for investigating individual excitation and deexcitation processes. In 1998, Johansson employed exact diffraction theory to investigate retardation effects (51). The effects are rather small for 147 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Au or Cu sample with W or Ir tip, whereas inclusion of retardation effects leads to significant change in luminescence spectral profile for Ag sample. In Ref. (52), Aizpurua et al. investigated the influence of tip shape on luminescence properties of the system using a hyperbolic tip geometry where the aperture and the apex curvature of the tip can be changed independently (Figure 8(a)). The Tersoff-Hamann approach (53) and the boundary charge method (54) were utilized in the calculation of the tunneling current and the electromagnetic field distribution, respectively. Energy of interface plasmon modes was found to be sensitive to the aperture angle ψ and relatively insensitive to the curvature of the tip apex for the case of a free-electron metal tip over a free-electron metal substrate. The authors also calculated the luminescence spectra for the case of Ag tip and Ag sample as well as Au tip and Au sample. The tip aperture gave significant influence on the positions of peaks in luminescence spectra, and the radius of apex curvature was reported to be more important for the luminescence intensity (Figure 8(b)). Tao et al. investigated the effects of dielectric film grown on the metallic substrate on the plasmon mode by means of a model composed of a hyperbolic tip on a metal substrate covered with a dielectric film (55). To accurately determine the tip-sample distance for a given current and bias voltage, the effective potential along the surface normal direction was calculated by use of the density functional theory (DFT), and then the one-dimensional Schrödinger equation with the obtained potential was solved. Here, two kinds of systems were analyzed, i.e., W-tip over a monolayer of C60 molecules adsorbed on Au(111) surface and W-tip over Al2O3 films grown on NiAl(110) surface. The authors reported that the insertion of a dielectric layer leads to the reduction in the luminescence intensity due to the increase in the tip-metal separation. In addition, the insertion of dielectric layer was found to hardly change the spectral profile with any significant peak shifts with respect to a clean surface. This was explained by a compensation between a blue shift of interface plasmon modes due to the increase in the tip-metal separation and a red shift due to the screening of surface charge by dielectric layers (56). As introduced here, a variety of models and methods have been proposed and all of these analyses are in principle based on combination of Schrödinger equation employing the one-body approximation and Maxwell equation using macroscopic local dielectric constant. As discussed above, to determine optical properties of the system composed of a STM tip close to a metal substrate, it is essential to unveil the properties of interface plasmons localized near the tip-sample gap region. Luminescence spectra and related observables acquired by STM-LE measurement contain the information related to the energy of localized interface plasmon modes, electric fields associated with these modes, and so forth. Since the plasmonic characteristics strongly depend on the geometric shape of the STM tip, properties of localized interface plasmons can be varied via modification of the tip shape by applying voltage pulse, crashing to the substrate and so forth. It is thus expected that, by the repetition of STM-LE measurements and modification of the tip shape, one can tune optical response of the tip-substrate system to realize favorable conditions for STM-TERS measurements. In the next section, we review procedures to prepare the STM tip and to modify the tip shape. 148 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 7. (a) Electroluminescence spectra recorded from the Au-Au junction as a function for increasing bias voltage (bottom to top). Lorentzian line shapes fitted to the sub peaks are also indicated. (b) Three types of spectral contributions represented by different peak positions form the bias voltage dependent luminescence spectra, reflecting three radiative localized interface plasmon modes (1.39, 1.57, and 1.72 eV) that can be excited by inelastic electron tunneling in the junction. (c) Spectrally integrated luminescence intensities (solid squares) and the corresponding quantum efficiencies as a function of bias voltages. Reproduced from (49). Copyright (2015, American Chemical Society).

149 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 8. (a) Geometry for a hyperbolic metal tip over a metal plane. The aperture of the tip is given by the angle ψ, and the curvature of the apex is given by the ratio b/d. (b) Comparison between luminescence spectra for the case of an Ag tip and an Ag sample for different apertures of the tip (ψ=15°, 30°, and 45° in solid, dashed, and dotted lines, respectively). The tip-sample distance is fixed at 5 Å, and the bias is 4 V. The peaks have been normalized by the elastic current corresponding to each case in order to improve the validity of the comparison (b/d in all cases). Peak positions are strong functions of the aperture angle, but the intensity is similar after normalization, with a slight tendency to increase for more open tips. Reproduced from (52). Copyright (2000, American Physical Society).

Tip Preparation Tip preparation is one of the most important key elements for the reliable and stable tip-enhancement effect because the material and the shape of the tip relative to the incident light polarization are only the parameters governing the tip-enhancement effect. However, controlling these parameters of the tips has been the most challenging issue in TERS in the sense that it is essentially still difficult to fabricate tips with a small diameter based even on the current technologies and the subsequent analysis of the tip qualities are also elaborative. After tip fabrication, it is often recommended to be used immediately because the fabricated tips can be easily contaminated or degraded (e.g. oxidized) in air. During the TERS measurement, the tips tend to be contaminated and damaged as well, which make the qualitative analysis of tips difficult. Moreover, even when you have a successful tip with a high tip-enhancement factor, it becomes more difficult to handle the tip since the higher the tip-enhancement factor is the more sensitive the contamination issue becomes. For example, a successful tip with a single molecule sensitivity means that even a single molecule contamination is not allowed for the TERS measurement. Based on the above background of requirements for tip-fabrication, lots of tip-preparation method has been proposed both in AFM and STM based feedback control. In case of AFM based TERS, highly reproducible tip 150 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


preparation has been reported with 20 nm spatial resolution (21, 57). For a higher spatial resolution, STM is the promising option because of its compactness, high stability, non-optical feedback and the utilization of the gap mode configuration since the substrate needs to be conductive such as noble metals. In this section, we will focus on the tip-preparation for STM feedback. At the early stage of STM research, various metal such as W, Pt-Ir, and Au are sharpened mechanically by grinding, cutting or crashing. However, the reproducibility is low and the diameter of the tip is relatively large with asperities. In order to improve the reproducibility, electrochemical etching based on the anodic dissolution of the metal is one of the most commonly used methods for tip sharpening nowadays (58). W is widely used in STM research whereas in case of STM based TERS, it is important to fabricate a sharp metallic probe using noble metals such as Ag and Au. For Ag tips, both single step (59) and two steps (60) electrochemical processing in ammonia and HClO4+MeOH, respectively are reported at an early stage of STM researches. Since the latter seems to have a smaller tip diameter, which is crucial for tip-enhancement effect, the similar method is utilized for STM tip-enhanced luminescence (61) and STM-TERS (62). In these tip-enhancement works, after being fabricated by the same method as (60), the Ag tips are introduced into UHV chamber. The tips are then cleaned by Ar ion sputtering, which is a common equipment in most of the sample preparation chamber of UHV-STM setups. Since the tip status such as diameter and asperity are still not optimized for tip-enhancement effect at a certain frequency range, the further tip modification can be done by bias voltage pulse or tip indentation by STM (61). These so-called “tip-making” processes are common techniques for STM imaging and spectroscopy such as STS and STM-IETS. However, in the case of STM-TERS, it is also possible to monitor the tip-plasmon status by STM-LE as discussed in the previous section during the tip-making process by using TERS detection optics. Figure 9 shows the example of STM-LE from the same Ag tip during tip-making on Ag(111) surface, which are directly comparable to Figure 7 based on Au tip-Au(111) configuration. The tip-plasmon can be actively tuned in visible to near-infrared wavelength and the each plasmon mode could be fit with eq. (1) as discussed in the previous section. The emission is strong enough to be monitored even by a conventional digital camera as shown in the Figure. The capability of the active control of the tip status is one of the biggest advantages of STM-TERS over AFM-TERS, in which the tip status cannot be tuned once mounted in AFM. It should be noted that the tip-making process and the subsequent characterization by STM-LE requires a high stability of the STM system such as the gap distance control so that it is still challenging for STM operated in ambient, and so LT-UHV environment is preferred as of now. While Ag is the most promising plasmonic material and is widely used in UHV-STM system, it is difficult to handle the tips in ambient or in solution. In terms of chemical stability, Au is another option as a plasmonic material especially when the frequency range is VIS~NIR. Similarly, for Au, both single step and two steps electrochemical processing in KCl (63) or HCl+EtOH (64) and HCl+subsequent CaCl2/H2O/Aceton (65), respectively are reported. Figure 10 shows the comparison of the Au tips fabricated by each method. In all cases, 151 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


the tip diameter of ~50 nm or less can be readily achievable. The choice of the method depends on the users’ preference however, still in any cases for both Ag and Au tips, suppression of over-etching is critical for making a sharp tip, e.g. one method is by fast electronic switch off of the applied voltage upon drop-off of the etched wire because the tip apex of the remaining wire may be still in contact with the electrolyte. Moreover, the diameter of Ag tips reported so far is relatively larger than Au tips. In case of STM-TERS, it is possible to modify the tip status actively as shown in Figure 9, however, the tip-making in general does not make the diameter smaller so that the fabrication of the smaller diameter tips are still strongly required. Aside from the suppression of over-etching, the improvement of the crystallinity of the metallic wire by annealing prior to etching is also expected for finer tip etching (65, 66).

Figure 9. STM electro-luminescence spectra of Ag tip tuned by tip-making on Ag(111) surface and the corresponding CCD image at the tip apex. Bias voltage and the tunneling current are set to 2.5 V and 250 pA, respectively.



Figure 10. SEM images of electrochemically etched Au tips of single step (a) KCl, (b) HCl+EtOH, and two steps (c) HCl+subsequent CaCl2/HaO/Aceton electrolytes. Reproduced from (63). Copyright (2011, American Institute of Physics): Reproduced from (64). Copyright (2004, American Institute of Physics): Reproduced from (65). Copyright (1995, American Vacuum Society).

The Pioneering Work: Pettinger When TERS was first reported around 16 years ago, several researchers used an AFM feedback for their experiments (10–12). On the other hand, at the same time, Pettinger reported the use of an STM for TERS (13). A schematic of their experimental configuration is shown in Figure 11(a). Unlike AFM, STM requires the use of conducting substrates, so for Pettingers’s work, he used a metal film deposited on glass slide for his substrate. The sample was then adsorbed on this substrate and an electrochemically etched Ag tip was positioned on top of the sample for the experiments. STM-TERS provides some notable advantages compared to AFM based TERS. First, the requirement of the STM for conducting surfaces provided an added advantage of suppressing the fluorescence signal, which is a big problem in Raman studies. Moreover, the distance d between the molecule and the tip apex can be more easily controlled in the STM unlike the AFM allowing for a better understanding of the tip distance effect. Lastly, due to the short localization length of the excited plasmons compared to the radius of the tip apex, the enhancement can be larger and is localized in a smaller area (13). They used a smooth gold film of around 12 nm thickness deposited on a 1 mm glass slide as the substrate. Around one monolayer thick brilliant cresyl blue (BCB) was deposited on the substrate and its Raman and corresponding TERS signal is shown in Figure 11(b). An increase by a factor of 15 was observed when the tip is in a tunneling condition. They attribute this enhancement to the excitation of localized surface plasmons (LSP). Since the excitation of the LSP is localized at the tip apex (estimated to be 100-500nm in radius), they estimated that a factor of 15 increase in the Raman signal would correspond to an average of 60-1500 enhancement factor due to the tip. At that time, TERS enhancements have been reported for adsorbates with relatively large Raman cross sections such as dyes and carbon based molecules. However, in a follow-up work (67) the same group managed to obtain comparable enhancements on a CN- ions deposited on Au substrate highlighting the potential of STM-TERS for single molecule studies. 153 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 11. (a) Schematic of the STM-TERS experiment of (4). They used an electrochemically etched Ag wire as the STM tip and Raman signal is performed on a backscattering geometry. A 633 nm laser focused to a spot size of about 2 μm in diameter with power of 0.05 mW was used as the excitation source. (b) Raman spectrum of brilliant cresyl blue with and without the presence of the Ag tip. At tunneling condition, the tip provides a noticeable enhancement in the Raman intensity revealing multiple peaks absent in the tip retracted case. Reproduced from (13). Copyright (2000, The Electrochemical Society of Japan).

Breaking 10 nm Barrier and Towards Single Molecule Scale In this section, we describe the route of TERS toward single molecule sensitivities. We divide this section into three parts, 1st one describes the first attempts for single molecule sensitivity and the problems encountered by researchers, the 2nd describes the use of extreme environments such as UHV and LT and the last one describes the approach back to ambient conditions. Approaching Single Molecule Sensitivities: First Attempts After the 1st successful demonstration of TERS in 2000 (10–14), researchers acknowledged its potential for chemical identification at the single molecule level. Hence many researchers have started instrumental development in order to attain the highest spatial resolution possible. Moreover, several materials such as carbon nanotubes (68, 69), adenine molecules (26, 70), benzenethiol (71), BCB (27), etc. have been investigated using TERS. While STM-TERS and TERS in general showed great promise in achieving single molecule spatial resolution, researchers know that some issues have to be addressed. One issue is that with any SPM based techniques, the tip conditions has a big effect on the experiment. In topography measurements using either AFM or STM, the shape and size of the tip affect the resolution of the measurements 154 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


and the general quality of the images produced (59, 60). As with normal SPM based experiments, one can expect that the tip condition greatly affect TERS experiments. As such careful attention has been given by several researchers both in the early days of development and quite recently with regards to tip preparation (64, 66, 72–75). Since Pt and W are popular metals used for STM experiments, the methodologies for producing STM tips using these metals are well known. However, for STM-TERS measurements, more plasmon active metals such as Au and Ag are necessary. This is because of their high free electron density, strong local surface plasmon effect and better stability in the ambient compared to other materials (72). Another important issue in single molecule studies is the contamination of both the sample and the tip. As these experiments are done in ambient conditions, despite careful sample/tip preparation and handling, contaminants are totally unavoidable. Moreover, apart from enhancing the signal from the sample, the Raman signals from these contaminations are also strongly enhanced in a TERS experiment. Clearly, these unwanted signals add difficulty in analyzing the experimental results. A good illustration of this difficulty is shown in the paper published by Neacsu et al. (76) and the corresponding comment and replies (77, 78). In (76), the authors claim single molecule sensitivity in their shear force AFM-TERS measurements. For this paper, they used a side illumination and detection configuration in combination with tip radii down to about 10 nm. They studied one monolayer and submonolayer coverage of malachite green on glass substrates. A reproduction of their Figures is shown in Figure 12. With a distance dependent measurement, they were able to illustrate the enhancement of several Raman peaks including the fluorescence from the tip. In their results, the relative peak intensities are different compared to the far field Raman spectrum. They claimed that this difference is the result of a strong optical field localization which is related to the different selection rules for the tip scattered Raman response. Moreover, they highlighted that the spectra while reproducible for a certain tip, varies from tip to tip. They also showed a time series data where they kept the tip at a certain position on top of the submonolayer thick dye molecule and observed variations in the relative peak intensities and the spectral positions. This result, in combination with photobleaching, they claimed, signifies that their experiment is probing at the single molecule level. Moreover, an enhancement factor of 5x109 was obtained. However, their conclusions were challenged by Domke and Pettinger in a comment (77). The main contention of (77) is that the assigned peaks by Neacsu et al. should be attributed to carbon contaminations rather than malachite green. According to Domke and Pettinger, this is because in (76), the peaks observed in the TERS spectra do not correspond to neither the far field nor the DFT calculations. In fact, they show that the observed peaks by Neacsu are more easily attributed to carbon contamination. The carbon species might come from photodecomposition of the malachite green, which according to (77) is not surprising given the high intensities used in the experiments of Neacsu.



Figure 12. (a) TERS spectra for different tip-sample distance d. As expected, the intensity of the Raman decreases with increasing d. (b) Spectra from the sample showing the absence of any Raman peaks when the tip is far from the sample. (c) Comparison of the TERS spectra with DFT calculations. (d) Spectra from the clean Au surface (area where there is no molecule) showing the enhancement of the fluorescence background in the presence of the tip. Reproduced from (76). Copyright (2006, American Physical Society). 156 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


In a reply (78), Neacsu et al. defended their work by showing additional data that points out that at low enhancements, the TERS spectra closely resemble the far field Raman spectra. However, as the enhancement increases, the spectra show distinct changes. They also show that while photobleaching occurs in their experiments even for relatively low enhancements, no new Raman bands show and all the peaks reduce in intensity at a similar rate. This observation was also obtained when using higher enhancements, albeit showing an understandably faster reduction rate of the intensity. This further suggests that the peaks observed for higher enhancements although different from the far field case, comes from the malachite green and not from carbon products. While this reply answered some questions raised by Domke and Pettinger’s comment, this exchange goes to show how a study can be questioned easily on the basis of the contamination issue. Although this work is based on AFM-TERS, such a situation is definitely possible for single molecule studies using STM-TERS. Domke et al. provided an insight on how to potentially distinguish between contaminations and actual Raman signals from single molecules (79). The key point in their paper is that, a comparison between near-field and far-field signals should not show significant differences. Variations in the peak position, fluctuations of the band intensities, and/or drastic changes in the peak widths should not be immediately attributed to single molecule effects. More specifically, claims concerning single molecule sensitivities should be analyzed closely by getting a significant number of spectrum and making sure that the observed changes in the spectra is not similar to fluctuations observed in carbon related studies. In claims relating to statistical analysis, they mentioned that at least 10,000 samples have to be collected for the claims to be reliable. This is technically difficult for TERS experiments as the positioning of the laser focus relative to the tip can have a significant effect on the TERS signal. Moreover, under low coverage used for single molecule studies, thermal diffusion will also affect the measurements. Also, one has to carefully consider the fact that large enhancements would lead to quick photodegradation and decomposition into carbon species. Therefore, the spectra have to be checked closely if carbon related peaks appear in the TERS spectrum. Using Extreme Environments: TERS in Low Temperature and Vacuum Conditions The experiments described above are all performed in ambient conditions. While the results previously shown have been promising in doing some single molecule studies, achieving more reliable data and even smaller spatial resolutions shows the need to modify the experimental procedures. While the use of ambient conditions is simpler experimentally, it is clear that performing TERS in more controlled environments such as in vacuum and/or at low temperatures can help with issues encountered in the ambient. The first demonstration of TERS at UHV conditions was reported by Steidtner and Pettinger in 2007 (80). For their experiment, they used parabolic mirrors to focus the laser to the sample and collect the Raman signal. The use of parabolic mirrors for focusing light into an STM junction was previously demonstrated by 157 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Meixner’s group (81). Owing to the high numerical aperture of a parabolic mirror, it was demonstrated that using such in combination with radially polarized fields generate a tighter focus (82) and a higher longitudinal electric field compared to normal objective lenses (18). An additional advantage for using parabolic mirrors is that due to the reflection geometry used, the sample is no longer limited to transparent samples. For an effective UHV-TERS experiment, Steidtner and Pettinger pointed out that it is important that optical alignment can be performed without breaking vacuum. As such in their design (80), the parabolic mirrors were mounted on piezomotor stages to allow control under UHV conditions. They generated the radially polarized light inside the vacuum by using four segmented half-wave plates similar to how it is generated and used for AFM-TERS (83). To achieve UHV, their main chamber is pumped by an ion getter pump and a sublimation pump. Similar to other UHV-STM systems sample cleaning and preparation as well as tip cleaning is performed in a load-lock chamber in order to maintain UHV conditions at the measurement chamber. In order to demonstrate the potential of their setup, they used as test samples Si(100) wafers and BCB adsorbed on either Au(111) or Pt(111) single crystals. The Si wafer was cleaved at UHV to reveal an atomically clean face while the Au and Pt crystals were annealed in an Ar atmosphere for cleaning before depositing the BCB.

Figure 13. TERS experiments under UHV environments. (a) and (b) shows the non-resonant Raman spectra of Si surface both when the tip is in the tunneling condition and not. Similarly (c) and (d) shows the TERS spectra for BCB monolayer on Au(111) and Pt(111) substrates. Reproduced from (80). Copyright (2007, American Institute of Physics). A comparison of their far field and tip enhanced signals for both Si and BCB samples is shown in Figure 13 where the excitation source is a 633 nm laser. For the silicon sample, the peak at 520 cm-1 showed an increase by 152% when a Au tip is brought near compared to normal Raman signal without the tip, this corresponds 158 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


to an enhancement factor of 1.2x105. It is important to note that the signal from the 950 cm-1 peak was stronger by only around 78%. This shows clearly the importance of the matching of the plasmon frequency of the tip and the Raman signal for TERS - if the plasmon frequency of the tip too far from that of the tip, the signal enhancement is compromised. On the other hand, for the one monolayer thick BCB samples, a significantly larger increase in the Raman signal intensity is observed. For BCB on Au surface, the enhancement was found to be 1.6x106 while for the BCB on Pt, the enhancement was 5x105 - an order of magnitude lower than that of Au. This was explained in terms of the stronger tip-substrate coupling for the Au substrate compared to the Pt substrate. The power of UHV-TERS was further illustrated in the follow-up paper of Steidtner and Pettinger (28). In this paper, they show up to 15 nm spatial resolution of their setup also on BCB molecules adsorbed on Au surfaces and using an Au tip. In Figure 14(a), 14(b) and 14(c), their data used to evaluate the spatial resolution is shown. Data shows a line scan of the Raman intensity along the black arrow in the STM topography image. It can be observed that the TER scan resolves molecules separated by as small as 15 nm, highlighting the resolution of their instruments. A TERS image of the single BCB molecule is also provided (Figure 14(e) and 14(f)) showing around 15 nm resolution. The successful imaging of this molecule shows that the stability of the system is quite high and that the photobleaching effect is reduced under UHV-TERS. In order to emphasize further that photobleaching effects are reduced under UHV conditions, they performed an experiment where they monitored the TERS intensity of the BCB peak at 570 cm-1 under UHV and an oxygen environment, which is shown in 14f, 14g and 14h. The data clearly shows that under O2, the Raman intensity quickly quenches significantly after only 5s of exposure to laser. Comparing this with the sample under UHV, while the TERS intensity also shows some photobleaching effects, the rate is clearly reduced. Owing to the high potential of performing TERS at vacuum environments, many researches were also performed under vacuum conditions. Even at high vacuum (HV) levels, a number of significant results mainly reported by Sun group were reported recently (84). They investigated plasmon driven chemical reactions (85) and showed a procedure to visualize photoinduced charge transfer (86). They also investigated the underlying mechanisms of HV-STM-TERS. In a report, they showed the effect of the electric field gradient under the STM on the spatial resolution of TERS (87). They showed that this electric field gradient is affected by the shape of the tip and is one of the controlling factors for spatial resolutions of 1 nm. Apart from the spatial resolution, they also showed that this electric field gradient can excite infrared vibrational modes, which are not normally seen under their excitation conditions (88). Under UHV conditions, Van Duyne’s group performed a significant number of studies. They showed the possibility to obtain multiple vibrational modes even in flat lying molecules such as CuPc (89). They were also able to combine UHV-TERS with UHV-Tip enhanced fluorescence (90). In this report they showed the advantage of using all external optics in the sense that it allows you the possibility to change the excitation wavelength easily. This allowed them to selectively excite different Q-bands of their molecule. In a different report, they 159 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


showed that UHV-TERS is sensitive enough to identify boundaries between two phases of a molecule. They showed that even if the topographic image shows no noticeable difference between these two phases, a TERS image would provide us information on each phase thereby allowing the identification of each phase in the image as well as give some idea on the molecules orientation with respect to the surface (91).

Figure 14. (a) STM topography image of the BCB molecule on Au(111). A linescan of the topography (b) and Raman intensity (c) along the arrow depicted in (a) shows the resolution of the setup. The TERS image of a single molecule is shown in (d) and its corresponding x and y cross sections are depicted in (e). (f) and (g) shows a series of 30 spectra taken at a rate of 1 spectra/second under UHV (f) and in O2 environment (g), the time dependence of the 570cm-1 Raman peak and the background is also shown in (h) under both UHV and O2 environments. Reproduced from (28). Copyright (2008, American Physical Society). The advantage of developing a TERS system in the UHV environment is the possibility of extending the setup to low temperatures. It was believed that using at low temperatures, the stability is more improved and hence one can get better signals and better spatial resolution. Moreover, atomic resolution topographic scans of the target molecules can be performed alongside the TERS image for easier interpretation of the results. The first demonstration of TERS measurements at UHV, LT environment was reported by Zhang et al. (62). A diagram of their system is shown in Figure 15 below. In their experiment, most of the optics are external to the UHV chamber except for the aspheric lens near the STM tip junction. They utilize a single lens for both focusing of the 532 nm laser and collection of the Raman from the sample. Their system is cooled using liquid nitrogen and they reported that they were able to get stable temperatures at around 80 K. 160 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 15. Schematic of the LT - UHV TERS setup of Ref. (62). Their setup is capable of reaching 80 K through liquid nitrogen cooling. An aspheric lens is placed inside the vacuum chamber to allow high collection efficiency. Reproduced from (62). Copyright (2013, Nature Publishing Group).

It has been reported by Pettinger et al. (92) that the enhancement in TERS is strongly affected by the tip surface geometry, which has a direct effect in the nanocavity plasmon modes at the tip junction. At low temperatures, the spectral position of the nanocavity plasmon modes can be tuned by applying tip bias (so called “tip-making” discussed in “Tip preparation” section), which in turn makes changes to the tip geometry. The nanocavity plasmon peak emission can be monitored via STM-LE (93). Zhang et al. found out that in doing low temperature TERS measurements, it is important to match the peak position of the plasmon modes to the electronic transitions of the molecule under investigation (62). In Figure 16 we show the data they reported on meso-tetrakis (3, 5-di-tertiarybutylphenyl)-porphyrin (H2TBPP) molecule on the Ag(111) surface. Figure 16 shows a clear difference when the nanocavity plasmon is “on-resonance” – that is its peak position is spectrally matched to the electronic transition of the molecule or “off-resonance” which is the opposite case. First, at on-resonance condition, we can observe that the Raman peaks rides on top of the broad plasmon peak. More importantly, although some Raman peaks are present at the off-resonant condition, much more spectral features are available during on-resonance conditions suggesting the importance of tuning the plasmon emission peak position. They interpreted their these effects of the plasmon position by pointing out that LT-TERS can be considered as an analogue to broadband femtosecond stimulated Raman scattering (BB fs-SRS) where in both cases the laser acts as the pump while in STM-TERS the nanocavity plasmon serves a similar purpose as the broadband fs probe pulse in BB fs-SRS.



Figure 16. The effect of spectral matching on the LT-TERS spectra. Reproduced from (62). Copyright (2013, Nature Publishing Group).

In Zhang et al.’s work, they also showed TERS images of their molecule and in Figure 17, we reproduce their work. The capability to perform single molecule imaging shows a high stability of their setup (~0.15 nm/min). They found that the Raman intensity is stronger at the lobes of the molecule compared to the center especially if you look at the low wavenumber modes. They explained that this finding could be related to the individual molecular vibrations of the molecule. Based on their DFT calculations, low wavenumber signals correspond to vibrations at the lobes while high wavenumber ones are concentrated at the center. This is the reason why when they monitored the high wavenumber modes the center is bright while for low wavenumbers, the vibrations are near the lobes. This image also shows the remarkable spatial resolution they obtained, which is around 1 nm – comparable to the resolution of their STM topography also shown below. In a follow-up paper, they investigated the origin of the high 162 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


spatial resolution of their technique using calculations (94). Using a quantum mechanical description, they showed that on the limit when the nanocavity plasmon distribution is confined such that it is of the same size as the molecule, the optical transition matrix of the molecule becomes dependent on both the position and distribution of the plasmonic field producing a spatial resolution nearly equal to the size of the molecule itself. The same group further emphasized the potential of LT-TERS by showing the ability to differentiate two different and adjacent molecule islands (95) and demonstrating subnanometer spatial resolution using carbon nanotubes (96). In (95), they investigated two different porphyrin molecules, free-base (H2TBPP) and metal centered (zinc-5, 10, 15, 20-tetraphenyl-porphyrin, ZnTPP) on Ag (111) surfaces While these two molecules both belong to the poryphyrin family, their Raman signatures are very different. Figure 18(b) shows a TERS line scan across two molecule islands separated by 2.5 nm. It can be observed that the Raman signals from positions 1-4 are clearly different from positions 8-10. Moreover, the space between the islands (positions 5-7) does not show any Raman peak but only the broad luminescence from the Ag surface. From here, we can easily identify that positions 1-4 are from the ZnTPP island and positions 8-10 are from the H2TBPP. Also, the individual molecules at the edge of each island (positions 4 and 8) show weaker TERS intensities. In order to investigate this more, a finer scan was made along the edge of the ZnTPP island as shown in Figure 18(c). Here, the scan is taken at 0.25 nm steps and 20 spectra were taken. From these Figures, we can observe that the intensity reduction is not only reproduced but also, we can observe a reduction of the TERS intensity at the center of the molecule showing a spatial resolution in the order of 1 nm. Moreover, the TERS line scan at 700 cm-1 also show weak signals at the center of the molecule which is since this represents intermolecular vibrations, they are expected to be prominent along the edges of the molecules and weak at the center. To challenge the capability of this technique, they scanned along two molecular islands that are in contact (Figure 19(a)). In the line scan shown in Figure 19(b), it can be seen that the molecule at position 6 is ZnTPP and the molecule at 8 is H2TBPP. Here we can observe the power of LT-TERS in identifying different molecules even when they are in contact. Apart from the possibility to identify individual molecules, they also showed a way of determining the orientation of the molecule using LT-TERS. Because of the relative position of the tip and the surface, it is expected that the vibrational modes perpendicular to the surface are more strongly enhanced than those parallel to the surface. Calculated TERS spectra for different tilts of the ZnTPP molecules are shown in Figure 20(b) where the coordinate axis is shown in 20(a). We can see that the 700 cm-1 mode has the strongest intensity for a flat lying molecule. Since this peak is associated with the out of plane vibration of the porphyrin core and the phenyl ring, tilting the molecule reduces its enhancement since it becomes parallel to the surface. This is nicely reproduced in the experiment as shown in Figures 20(c) and 20(d) where the molecules along the step edge is studied. Here, apart from distinguishing alternating molecules, they were also able to reproduce the results of the calculations from molecules not flat lying on the surface. 163 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 17. TERS mapping of a single molecule. The TERS spectra at the lobe and center of the molecule as well as the Ag surface are shown in (a). (b) Shows the corresponding TERS map of representative Raman peaks shown in (a) and their associated calculated Raman maps. (c) and (d) shows a line scan of the topography (c) and TERS map (d) showing comparable spatial resolutions. Reproduced from (62). Copyright (2013, Nature Publishing Group).



Figure 18. Identifying adjacent molecular islands. Topographic (a) and TERS line scan (b) of ZnTPP and H2TBPP separated by 2.5nm showing the possibility to identify each island using TERS. (c) Shows the topographic and Raman line scan along the edge of the island. Reproduced from (95). Copyright (2015, Nature Publishing Group).



Figure 19. Identifying the contact area of two different molecular islands. Topographic (a) and TERS line scan (b) of ZnTPP and H2TBPP. Reproduced from (95). Copyright (2015, Nature Publishing Group)

The most recent report of the same group on LT-TERS was measurements on carbon nanotubes (96). Carbon nanotubes are essentially used in Raman experiments or specifically in TERS as a test sample to demonstrate its effectiveness and spatial resolution. In their experiment, they showed spatial resolutions as high as 0.7 nm for the G-band and 0.8 nm for the D-band. At present, this is the highest spatial resolution demonstrated for CNTs investigated using TERS. From Figure 21(a), we can see that the D-band signal is not uniformly distributed along the CNT. In fact, the D-band signal was observed to be very strong at the end of the CNT, as this peak is mainly attributed to defects, it can be said that the defects are concentrated along the tip of the CNT. They also studied the strain variations along bent CNTs by looking at the changes in the G-band signal. At first, they demonstrated the differences in the TERS signals for naturally bent CNTs found on the surface. The key finding was that splitting of the G-band was observed for bent portions of the CNT. They studied this further by looking at the TERS spectra of CNTs they cut and bent using the STM tip. Figure 22(a) and 22(b) shows the topography image of the CNT before and after manipulation. The corresponding Raman spectra on 3 representative points are shown in Figure 22(c) and 22(d). It can be observed that Raman spectra at the end of the CNT which was basically untouched did not show any changes while other parts on the CNT show some significant changes. To highlight this more, they took a spectral map of the bent CNT and were able to show a clear shifting of the G-band peaks as well as appearance and disappearance of other peaks. This emergence and disappearance of other peaks are interpreted as splitting of the G-band and is therefore attributed to local strain along the bent regions of the nanotubes. Moreover, as expected, the amount of spectral shift was observed on regions where the bending radius is smaller as this region has the largest amount of strain. 166 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 20. Determining the molecular orientation using TERS. (a) is a schematic of the coordinate axes used and (b) is a calculation of the spectra when the molecule is tilted by different angles. (c) and (d) Due to the highly oriented enhancement by TERS, both the identification of different molecules and their orientation is possible. Reproduced from (95). Copyright (2015, Nature Publishing Group).



Figure 21. Simultaneous topographic and TERS image of a carbon nanotubes using LT-STM TERS. The G-band shows a relatively uniform intensity along the nanotube length while the D-band shows regions of intense signals. The line scan (b) along the arrows in (a) shows a subnanometer spatial resolution regardless of whether the G or D band is scanned. Reproduced from (96). Copyright (2016, American Chemical Society).

Lastly, they also observed that the effect of the bending is different on spectra is different on the inner region of the bend and its outer region. Since the inner and outer region of the bend experiences strain on opposite direction such a difference is to be expected. These results highlight not just the chemical sensitivity of TERS but also its capability to identify localized strain effects, which can be powerful for investigation of low dimensional structures. Another noteworthy work is the report by Van Duyne’s group on their investigation of the Rhodamine 6G molecule at liquid helium (19 K) temperatures (97). To date, we believe this is the only report of low temperature TERS at liquid helium environments. In contrast to the setup of Zhang et al., all optics used in this work is outside the UHV. In this sense, they used a long focal length lens (250 mm) to focus the laser light into the tunneling junction. Also, they collect the Raman signal from the opposite side of the STM viewport. In their experiment, although they were able to cool down their sample to 8 K, to allow optical access to the STM junction, they had to open the cryostat shutter near the STM junction, causing the temperature to rise to 19 K. In their report, they paid particular attention to the peak sharpening at low temperatures as shown in Figure 23. To do this, they compared the spectra of both room and low temperature TERS and SERS. For SERS measurements, they sublimed their sample on Ag film over nanosphere (AgFON) substrates optimized for 532 nm while for TERS, the substrate is atomically smooth Ag(111). 168 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 22. Investigation of the strain using TERS. The CNT is manipulated using the STM and its corresponding topography before (a) and after (b) is shown. A distinct difference was observed as evidenced by the changes in the G-band before (c) and after (d) manipulation, this is more clearly seen in the Raman intensity map along the wire after manipulation (e). Reproduced from (96). Copyright (2016, American Chemical Society). They pointed out that in their experiments, it was difficult to obtain any information about the adsorption configuration of the molecule to weak interaction with the substrate. They therefore used the Raman data got to determine more information about the molecule-substrate interaction. Since the TERS experiments were performed on atomically smooth Ag(111) substrate, the number of possible adsorption conformation is less than that for SERS performed on AgFON. This aside from the fact that the probe area of TERS is much smaller than TERS, explains why the RT-TERS FWHM is narrower than the RT-SERS as can be seen in 23(a) and (b). Apart from the narrower linewidth of RT-TERS compared to RT-SERS, another important result they got is the reduction of the FWHM at LT for both TERS and SERS. This can be explained that at room temperature, the molecule is free to move at the surface. This motion may cause diffusion along the surface as well as conformational changes. This causes a motional averaging on a certain vibrational mode, which in turn increases the linewidth of a certain peak. On the other hand, at low temperatures, this motional averaging is greatly reduced 169 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


and therefore since the molecule is effectively stationary on the surface. This therefore would show up as a decrease in the linewidth of the Raman peak. In fact, in their result, they found that the FWHM of the LT-TERS is smaller by almost 50 % on the average compared to RT-TERS with minimum of around 6 cm-1. For similar reasons, the average linewidths of the LT-SERS spectra is reduced to 18 cm-1 compared to 20 cm-1 at RT.

Figure 23. The effect of temperature on the FWHM of the TERS and SERS spectra of Rhodamine 6G. The numbers are FWHM values of the fits for (a) RT-SERS, (b) RT-TERS, (c) LT-SERS, and (d) LT-TERS. (e) Shows the residuals after fitting the LT-TERS (d). Reproduced from (97). Copyright (2014, American Chemical Society). Apart from line narrowing, another feature they observed is peak shifts, in particular, the peaks from LT-TERS is shifted compared to RT/LT-SERS and RTTERS as shown in Figure 24. Another interesting point about peak shifts they observed is that not all peaks are shifting. As can be seen from the Figure below, only the peaks located at 1132, 1205, 1350, 1527 and 1547 cm-1 are shifting. They found that the potential energy distributions for vibrational modes corresponding to these peaks are approximately localized to the ethylamine or on the xanthene ring moieties. On the other hand, the unshifted modes correspond to the phenyl ring or to vibrational modes, which are at the most internal part of the molecule. They hypothesize that the shifting modes correspond to moieties nearest to the surface. This therefore suggest a possible adsorption configuration on the surface of the Ag(111) where the molecule is situated “edgewise” along the xanthene moiety and the ethylamine are against the substrate. 170 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 24. Spectral shifts of the TERS and SERS peaks of Rhodamine 6G. The plots shown are (a) ambient RT-SERS on Ag colloids, (b) ambient SM RT-TERS on a Ag film, (c) RT-UHV-SERS on AgFON, (d) RT-UHV-TERS on Ag(111), (e) LT-UHV-SERS on AgFON, (f) LT-UHV-TERS on Ag(111), and (g) calculated Raman spectrum of free R6G. Reproduced from (97). Copyright (2014, American Chemical Society). One Nanometer Scale Spatial Resolution in the Ambient While the works described above solved many of the problems encountered by researchers trying to get single molecule resolution, the extreme conditions required limited the materials that can be studied. In particular, low temperatures and high vacuum requirements are unattractive to biological researchers. It is therefore necessary to attain similarly high spatial resolutions without using extreme conditions of low temperature and high vacuum. With this in mind, our group developed a system capable of controlling the environmental conditions but not using low temperatures and high vacuums. Figure 25 shows an image of the setup we developed. The system is based on an STM-TERS setup consisting of a custom built STM head and an optical system mounted on a piezo-driven XYZ stage. The tip approaches the sample at an angle of approximately 10° to allow a wider optical access and an objective (NA=0.7) focuses the light at an incidence angle of 50° and the signal is collected in a backscattering geometry. The whole system is contained in an enclosure that allows us to contain the environmental conditions in the setup. Figure 25(c) shows the temperature and humidity inside of the chamber under nitrogen purge taken for 24 hours. It can be observed from the Figure that after about two hours, the humidity stabilizes to 171 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


zero while the temperature takes around 6 hours to stabilize. After stabilization, both the temperature and humidity remains constant for the rest of the duration of the measurements. These environmental controls ensure the thermal drift is down to 1 nm/min.

Figure 25. STM-TERS system in the ambient. (a) Picture of the chamber used to contain the STM-TERS setup, the chamber can be isolated from the environment and purged with gas to add to the stability of the system. (b) Zoomed in view of the STM head showing the Au tip and the objective. Inset shows a CCD image of the STM tip and its reflection illuminated with the HeNe laser. (c) Shows the relative humidity and the temperature inside the chamber monitored for 24 hours, the nitrogen purging was started at the same time as the monitoring. Using this system, the highest spatial resolution attained for TERS is 1.7 nm (98) and to date, we believe this is still the highest spatial resolution reported for a TERS experiment in ambient conditions as shown in Figure 26. In this work, single walled carbon nanotubes (SWCNT) on Au(111) surfaces were investigated. A simultaneous STM topography and TERS image of the D-band, G-band and 2D band were observed. First, the D-band signal was observed on the ends of the SWNT. The D-band is known to be related to defects and structural deformations in a nanotube. It is therefore to be expected that the D-band signal is strong along the ends of the wires. The G-band signal on the other is known to be directly affected by the diameter of the nanowire due to its resonance with the laser wavelength. In the results, the G-band signal is strongest at the region some regions of CNT-2, this is where the topography shows a diameter of around 1.2 nm, which is in good resonance with the laser wavelength used. In contrast, the topography image shows that the diameter of CNT-1 is nearly double the diameter of CNT-2 suggesting which suggests that is it not in resonance with the laser used, 172 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


therefore leading to a reduced G-band signal. Lastly, the 2D band signal can be observed in all of the nanotubes shown. In order to understand this further, the spectra along each nanotubes were taken and is shown in Figure 27.

Figure 26. STM topography (a) and TERS in the ambient taken simultaneously. The TERS image of the D-band (b), G-Band (c) and the 2D band (d) are shown. Reproduced from (98). Copyright (2014, Nature Publishing Group).

Figure 26 shows the spectra at representative points along each SWNT. As discussed above, a variation in the intensity of both the D and G-band were observed along the length of the nanowire. Of interest here is the observation that the 2D-band signal is split at the center of CNT-1. This splitting cannot be related to defects since the D-band signal is not dominant on regions where this splitting were observed. Instead, the authors attribute this splitting of the 2D-band to the presence of double walls in CNT-1. This is further corroborated by the fact the D-band signal for CNT-1 is much larger than both CNT-2 and CNT-3, this observation was also attributed previously for interactions between layers in a 2-layer graphene (99). The presence of double walls maybe caused by the processing of the SWNT wherein a nanotube was inserted inside another nanotube. While the spatial resolution showed in this work is still not at the subnanometer range described in LT-STM-TERS experiments, this shows the possibility to attain very high spatial resolution in STM-TERS enough for local investigations of nanostructured materials. 173 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Figure 27. (a) STM image of three CNTs. (b) Line scan of the TERS taken across CNT-1 while (c) shows the line scan along CNT-2 and CNT-3. The 2D band was observed to split at certain regions of the nanotubes. Reproduced from (98). Copyright (2014, Nature Publishing Group).

Summary and Future Outlook Since the invention of TERS in 2000, lots of efforts have been done mainly for improving the reproducibility of tip preparation based on AFM (21, 57). Owing to the improvements, several companies are commercializing TERS in order to expand the applications of TERS. However, the reported materials are still limited to either highly Raman active molecules or materials assisted by resonant Raman condition. This is mainly because of two factors, 1) stability and 2) background signals. It is straightforward to think that we should use either higher excitation power or longer exposure time of CCD camera for sensing a weak signal. However, it turns out to be difficult to use higher excitation power because tip-enhanced electric field easily increases the local temperature (100–102), which eventually causes the damage of either the tip itself or the sample. It is also practically difficult to use a longer exposure time due to the limited position stability of the tip relative to the target sample especially when it is not single point detection but a two-dimensional imaging that requires a long acquisition time. 174 Ozaki et al.; Frontiers of Plasmon Enhanced Spectroscopy Volume 2 ACS Symposium Series; American Chemical Society: Washington, DC, 2016.


Moreover, TERS signals have to compete with various background signals since tip-sample geometry with possible unknown contamination molecules generates lots of photons such as luminescence when irradiated by laser. This situation becomes stricter when the spatial resolution gets higher meaning the number of target molecules becomes smaller. Due to the development of the commercial TERS systems, automation of the system could be much developed whereas the above mentioned issues are still left for scientist to tackle by various ways. For example, in this chapter, the particular focus is given to gap-mode plasmon for higher sensitivity and higher spatial resolution. The other option is to employ nonlinear response of materials such as coherent anti-Stokes Raman scattering (CARS) (14–16, 103). It has been known that nonlinear response is useful for higher signal confinement as the nonlinear response is induced selectively at the high photon concentration (104) resulting in higher spatial resolution. In this sense, the matching of nonlinear response with tip-enhanced electric field should be good not only for higher sensitivity but also for higher spatial confinement. However, due to the technical difficulty of nonlinear spectroscopy, the high cost of the pulse laser optics, and the poor reproducibility of tip-enhancement effect, the combination of the two have not been well developed after the pioneering work of tip-enhanced CARS in 2004 (14) despite of the expected benefits. Owing to the development of STM-TERS with a high position stability will open up again the synergy with nonlinear spectroscopy. When successfully combined with nonlinear spectroscopy, time resolved spectroscopy in the nanoscale is also expected as an extreme spatio-temporal nanosensing technique (105–107). It should be noted that only tip position stability but also optics stability (108) are strongly required for nonlinear spectroscopy as the signal is extremely sensitive to the photon concentration. From the scientific point of view, it is important to establish the theoretical approach how to analyze the obtained date from the nanoscale spanning from mesoscopic to single molecule scale. In such an extremely confined system, it is essential to analyze the fundamental properties of the gap-mode plasmon especially when the gap distance becomes extremely small such as less than 1 nm. For example, even the selection rule could be modified in such spatially confined system due to high electric field gradient (87, 109, 110). In such a case, electron clouds both from the tip and the substrate overlap and induce plasmon hybridization (111), which is exactly the case of STM. In the classical electrodynamical approach, the transition from the separated to the overlapping nanoparticles of gap-mode is characterized by the discontinuity of the resonant energy upon contact. However, this is not true when the gap distance becomes

STM-Tip-Enhanced Raman Spectroscopy toward Single Molecule Scale

Recommend Documents