Reactivity mapping of nanoscale defect chemistry under electrochemical reaction conditions

Electrocatalysts often show increased conversion at nanoscale chemical or topographic surface inhomogeneities, resulting in spatially heterogeneous reactivity. Identifying reacting species locally with nanometer precision during chemical conversion is one of the biggest quests in electrochemical surface science to advance (electro)catalysis and related fields. Here, we demonstrate that electrochemical tip-enhanced Raman spectroscopy can be used for combined topography and reactivity imaging of electro-active surface sites under reaction conditions. We map the electrochemical oxidation of Au nanodefects, a showcase energy conversion and corrosion reaction, with a chemical spatial sensitivity of about 10 nm. The results indicate the reversible, concurrent formation of spatially separated Au2O3 and Au2O species at defect-terrace and protrusion sites on the defect, respectively. Active-site chemical nano-imaging under realistic working conditions is expected to be pivotal in a broad range of disciplines where quasi-atomistic reactivity understanding could enable strategic engineering of active sites to rationally tune (electro)chemical device properties.

Theoretically, one monolayer (ML) AuOx (or ½ ML O2) amounts to 222 µC cm -2 on Au (111) for 1-electron transfer 2 . A complete surface oxidation reaching 6 ML AuOx would amount to 1332 µC cm -2 , roughly twice the amount of total charge we measure. Therefore, terrace oxidation likely is limited to the topmost one or two surface (mostly terrace) layers, in agreement with literature 3 . In contrast, the oxidation of the defect sites, however, reaches about 6 ML (see main manuscript text) as inferred from EC-TERS and EC-STM data. The ratio between defect charge density of ca. 60 µC cm -2 and total surface charge density thus indicates that the Au(111) single crystal has a defect coverage of ca. 4.5 %.

EC-TER spectra for ON and OFF states
The EC-TER spectra from 100 cm -1 (dichroic edge filter cut-off at 156 cm -1 ) to 1700 cm -1 corresponding to Fig. 1c in the main text are shown in Supplementary Fig. 2. When defectcatalyzed water splitting is ON, the AuOx stretching mode around 560 to 580 cm -1 is detected 4,5 .
Small, ill-defined bands can be discerned at ca. 250 cm -1 and at ca. 980 cm -1 for both ON/OFF states and are tentatively assigned to the 0.1 M sulfuric acid aqueous electrolyte (H2O and/or SO4 2modes, respectively). Note that the spectral background does not change when switching between ON and OFF states, indicating that the EC-TERS gap resonance is unaffected by the potential switch.

Supplementary Note 3:
Maintaining the Au-tip oxide free Within our study, we have taken great care to ensure that the Au-tips stay oxide free and that the obtained gold oxide (AuOx) signal originates from the electro-oxidation at the Au(111) surface. Supplementary Figure 3 shows two cyclic voltammograms (CVs) from the Au (111) electrode (blue) and the Au-tip (black) recorded simultaneously with 0 V bias and a scan rate of 100 mV s -1 . As the Au-tip is in principle a defective Au structure, the oxidation starts around the defect oxidation region above 1.3 V vs. Pd-H and exhibits its main peak around 1.45 V vs.
Pd-H. On the other hand, the Au(111) shows its strongest anodic peak around 1.55 V vs. Pd-H corresponding to (111) terrace oxidation with a shoulder extending to the defect oxidation region. Importantly, the onset of the reduction of the AuOx at approximately 1.2 V vs. Pd-H clearly overlap for both the Au(111) electrode and the Au-tip. Maintaining the Au-tip potential -14 -at a maximum potential of 1.2 V vs. Pd-H or below thus ensures that the Au-tip resides in the AuOx reduction region and remains AuOx free.
We have achieved this requirement either by applying a constant bias or by keeping the tip potential fixed. For the constant bias, we have utilized a constant bias of 400 mV or higher. For a maximum upper sample potential of 1.6 V vs. Pd-H, this corresponds to 1.2 V vs. Pd-H for the Au-tip (and 400 mV bias). The second approach fixes the Au-tip potential at 1.2 V vs. Pd-H or below, e.g. at 1.0 V vs. Pd-H as for Fig. 2 in the manuscript. It is important to note that fixing the tip potential to a specific potential value inherently changes the voltage bias when switching the working electrode. We have recently investigated that in general a larger bias will lead to an increased tip-sample distance and therefore to a decreased TERS intensity in air, but that this effect is negligible for in-liquid experiments 6 . In any case, when the sample potential is switched from 1.1 to 1.45 V vs. Pd-H, the sample bias voltage increases from 100 to 450 mV, which would (if at all) result in a larger tip-sample distance and a lower EC-TERS enhancement for the ON state compared to the OFF state. Under these conditions, we are able to detect AuOx at around 560 to 580 cm -1 , and we do not observe any AuOx intensity at 1.1 V vs. Pd-H.

EC-TER spectrum under terrace oxidation conditions
As highlighted in Supplementary Fig. 1, at 1.6 V vs. Pd-H also the (111) terraces are oxidized and not only the defects. Supplementary Fig. 4 shows an example EC-TER spectrum recorded at 1.6 V vs. Pd-H while the Au(111) surface was oxidized at 1.6 V vs. Pd-H. When the potential is changed to 1.15 V vs. Pd-H, the AuOx peak around 560 to 580 cm -1 vanishes. Note that conventional Raman (without tip enhancement) cannot resolve the small amount of AuOx surface species formed in the ON state (blue curve).

EC-TERS reactivity mapping
The EC-TERS reactivity maps presented in Fig. 2 of the manuscript were obtained in the scanto-point mode. This scan-to-point mode records the EC-STM image while scanning to the points at which EC-TER spectra are recorded. Due to thermal drift and a spectral acquisition time of 1-3 s, slight shifts in the surface morphology before and after spectral acquisition are apparent. In Supplementary Fig. 5c, this phenomenon is further highlighted. The EC-STM image recorded under reaction conditions at 1.45 V vs. Pd-H exhibits "lines", at which the EC-TER spectra were acquired. Since the spectral acquisition time is larger than the time required for the EC-STM to record a single line (32 pixels x 1-3 s vs. 1 line s -1 ), thermal drift is more apparent. For Fig. 2a,c,e and Supplementary Fig. 5a,c, the recorded EC-STM images were flattened according to the standard procedure in Gwyddion (5 th polynomial degree). The images were recorded with 1 line s -1 scan speed and 512 points per 300 nm distance. The tunneling current setpoint was 1 nA.
The EC-TERS reactivity maps were created as follows: First, the raw EC-TER spectra were Fourier-filtered in matlab using the interactive Fourier Filter (iFilter.m; matlab file exchange) to remove high frequency noise. Supplementary Figure 6a shows example filtered and unfiltered EC-TER spectra. The filter parameters were as follows: frequency = 0.00060677; period = 1648.0761; width = 0.005-0.015; shape = 5. A linear baseline was determined within the boundaries of 440 to 680 cm -1 for each individual Fourier-filtered EC-TER spectrum and then subtracted from the Fourier-filtered EC-TER spectrum ( Supplementary Fig. 6).
Afterwards, the AuOx peak area with the boundaries (440 to 680 cm -1 ) was determined by simple integration (trapz, matlab). The obtained peak area of the 440 to 680 cm -1 peak region was then plotted in OriginPro 9.1 (contour plot). In Fig. 2b, (Fig. 3a,b and Supplementary Fig. 9). Moreover, Figure 3c demonstrates that for the different step structures the EC-TERS intensity changes sharply within the 9.4 nm pixel size, suggesting that the spatial chemical feature sensitivity is even lower than the pixel size used in this work.

Fundamentals of tip-enhanced Raman spectroscopy (TERS)
TERS is based on the excitation of localised surface plasmons (LSPs) in the apex of a metal (or metallized) SPM tip. In our configuration, we employ a commercially available electrochemical scanning tunnelling microscope (EC-STM) coupled to a Raman optical platform as described -18 -in the Methods section in the main text. The working principle of TERS is based on the efficient coupling of a focused laser beam with the tip to excite LSPs, which generates a strong electromagnetic near-field at the very apex of the STM tip. In a way, the tip acts as an antenna and converts, or "concentrates", far-field radiation into a nm-confined field underneath the tip.
The near-field character of the Raman excitation provides the extreme spatial optical resolution, typically of a few nm for in-air experiments, which depends essentially on the size, curvature and surface (atomic) topography of the tip apex.
The created near-field typically is a factor 10 to 100 larger than the excitation far-field and can induce Raman scattering in the species or molecules located at nm-close distance below the STM tip. The Raman scattering intensity, in a first approximation, scales with the fourth power of the magnitude of the excitation field. The strong field enhancement at the tip compared to the far-field intensity enables the detection of very few surface scatterers down to single molecules. When the substrate is a metal or exhibits metal optical properties similar to the ones of the tip, LSP excitation in the tip is accompanied by a corresponding image dipole formation in the substrate which leads to even higher field enhancements in the tip-sample gap.
Accordingly, such a TERS configuration is called gap-mode TERS and provides typically even higher sensitivity than TERS on a dielectric substrate.
The interested reader is referred to recent reviews about TERS, its fundamentals and applications 8,9 .

TERS intensities and plasmonic gap properties
In general, Raman intensities are governed by the local field strength -for TERS, this corresponds to the local field enhancement, or near-field strength, in the tip-sample gap -and by the amount of scatterers for a given experimental configuration, detection sensitivity and -19 -scattering cross section of the system under study. The effect of the formation of a plasmonic tip-sample gap on the field (enhancement) can be seen (albeit not quantified as there is no signal in the conventional Raman spectrum) from the comparison of the conventional (no tip-sample gap) and TERS signals as shown in Supplementary Fig. 4.
For self-assembled monolayers of organic molecules adsorbed at metal substrates, sometimes a local increase in TERS intensity between a factor five to ten at step edges or surface protrusions compared to TERS at neighbouring flat Au regions, partly also accompanied by frequency shifts of Raman modes, has been reported in the literature [10][11][12][13][14][15] . Such edge effects can be attributed to local heterogeneities in the plasmonic properties of the formed tip-sample gap that differ depending on the actual (atomic) gap geometry and field localisation and polarisation. As a result, the LSP or gap resonance shows nanoscale spatial heterogeneities both in intensity and in location of the resonance maximum causing differences in coupling efficiency between excitation far-field laser and gap mode. Strongest near-fields are created under resonance conditions when the excitation wavelength approximately matches the LSP resonance maximum. Therefore, shifts in the gap plasmonic resonance maximum (intensity and position) can lead to significant differences in TER scattering intensities. Furthermore, highly localised strong fields can lead to a local Stark effect that causes shifts in vibrational frequencies 12 . Also different adsorption geometries of molecules at step edge sites compared to terrace sites can account for differences in TER shifts 13 .
In general, the plasmonic (gap) properties can be deduced from the shape (plasmon resonance energy) and the intensity (field strength, or enhancement) of the spectral background 16 .
Supplementary Figure 2 shows a comparison between TER signals recorded in ON and OFF conditions. Evidently, the background does not differ, neither in shape nor in intensity. As discussed in the main text of the manuscript, changes in the tip-sample distances and thus in the field enhancement upon variation of the tip-sample bias (due to the variation in applied -20 -potential), have previously been shown to be negligible in water 26 , in line with the conclusion drawn from Supplementary Fig. 2 about the unaltered field enhancement in air compared to in water. As such, it is fair to assume that the field strength and enhancement do not change upon potential change and that the obvious change in peak intensity is due to a change in the amount of scatterers present in the nearfield, i.e. the amount of AuOx formed.
Furthermore, Supplementary Figure 10 shows raw data from an example line scan. From these spectra it is evident that the TERS background recorded at oxide-free terrace sites and the background recorded on the oxidized defect are identical, i.e there is no measurable difference between the plasmon resonance energy or field enhancement above terrace or defect sites. As such, we have attributed the oxide intensities recorded on defect sites and their spatial variation across defects to a variation in the local number of AuOx scatterers, or the thickness of the formed defect oxide layer.
Regarding the spatial uniformity of the background signal, we also refer to Supplementary Fig.   5e. Here, the EC-TERS map without background subtraction (field enhancement and number of scatterers determine intensity) is plotted -and is essentially the same as the one presented in the main text in Fig. 2d after background correction to present a "pure" chemical map corrected for (however small) near-field spatial variations (number of scatterers determine change in peak intensity). Comparing Supplementary Fig. 6e and Supplementary Fig. 2d, it is evident that, independent of whether with or without background correction, the quantitative result of selective AuOx formation on the defects remains the same.
In summary, while edge effects due to spatially heterogeneous plasmonic gap properties were observed in TERS imaging of self-assembled monolayers, our data does not provide any indication for measurable differences in tip-sample coupling neither as a function of tip position nor of applied potential within the employed potential window. Therefore, we attribute the TER intensity differences to spatial-and/or potential-dependent variations in the amount of AuOx