In situ topographical chemical and electrical imaging of carboxyl graphene oxide at the nanoscale

Visualising the distribution of structural defects and functional groups present on the surface of two-dimensional (2D) materials such as graphene oxide challenges the sensitivity and spatial resolution of the most advanced analytical techniques. Here we demonstrate mapping of functional groups on a carboxyl-modified graphene oxide (GO–COOH) surface with a spatial resolution of ≈10 nm using tip-enhanced Raman spectroscopy (TERS). Furthermore, we extend the capability of TERS by measuring local electronic properties in situ, in addition to the surface topography and chemical composition. Our results reveal that the Fermi level at the GO–COOH surface decreases as the ID/IG ratio increases, correlating the local defect density with the Fermi level at nanometre length-scales. The in situ multi-parameter microscopy demonstrated in this work significantly improves the accuracy of nanoscale surface characterisation, eliminates measurement artefacts, and opens up the possibilities for characterising optoelectronic devices based on 2D materials under operational conditions.

In the Raman spectrum of the GO-COOH flake ( Supplementary Fig. 2b) prominent bands are observed at 1352 cm -1 and 1583 cm -1 , which are assigned to the D and G bands of GO-COOH 9 . A very weak 2D band 10 is observed at ≈2700 cm -1 . The broad band in the 2900 cm -1 -3000 cm -1 region can be assigned to the stretching mode of C-H (ν S, C-H ) 11 .
Because the E 2g phonon is infrared (IR) inactive, this mode cannot be observed in the FTIR absorption spectrum 12 . Moreover, D and 2D Raman vibrational modes are also IR inactive, and hence cannot be observed in the FTIR spectrum 13 . In the Raman spectrum of GO-COOH, the very strong intensity of D and G bands makes the observation of Raman bands of other functional groups (which are present in relatively small concentration) difficult. Note that in the TERS maps measured from this sample ( Fig. 2 and 4), due to the short integration time of 0.1 s -0.4 s used at each pixel, the weak 2D band is immersed in the background noise and hence is not observed. Fig. 2 Table 1. Additionally, we observe bands at 1420 cm -1 and 1330 cm -1 , which can be assigned to the Raman vibrational modes of C-H and C-CH 3 15, 16 . Furthermore, the assignment of these bands is also consistent with the  18,19,20,21 . Therefore, we speculate that the Raman signals of the functional groups observed in the TERS maps shown in Supplementary Fig. 4 most likely result from orientational ordering rather than spatial clustering. A limitation of the present measurements is that it is not possible to determine if the absence of Raman signal at a particular location results from an absence of functional groups or a different orientation. However, this is beyond the scope of the current study and we aim to address it in our future research work.

Supplementary Note 3: Spatial resolution of the TERS map
It should be noted that the spatial resolution of the TERS map obtained in Supplementary Fig. 5 is much smaller than the radius of Au coated TERS tip-apex. The typical size of the Au grains present at the TERS tip-apex ranges from 24 nm -47 nm as shown in Supplementary Fig. 6, indicating that single Au grains rather than the entire tip-apex is involved in LSPR enhancement of Raman signals 22,23,24 . This TERS resolution is 45 × better than the highest confocal spatial resolution that can be achieved in our system, which is calculated using 0.44λ/NA 25 (λ = excitation laser wavelength; NA = numerical aperture of the objective lens) to be 447 nm. Furthermore, this spatial resolution is also higher than the previously reported value for TERS maps of graphene (12 nm -20 nm) 26 and twodimensional transition metal dichalcogenides (TMDs) such as MoS 2 (20 nm) 27 and WSe 2 Supplementary Fig. 5 a -e TERS intensity profiles (blue) along with the fitted Gaussian curves (red) from five different locations in the TERS map shown in Fig. 2c. Average spatial resolution of the TERS map is estimated from the FWHM of the fitted Gaussian curves to be 10.5 ± 1.7 nm. 15 nm) 28 measured using either bottom or side illumination TERS configurations (see Supplementary Fig. 7). The spatial resolution of the TERS map is primarily limited by the size of the plasmonic nanostructures present at the TERS tip-apex. However, we believe that in our TERS measurements the spatial resolution could also be limited by the relatively large step size of 10 nm used for TERS mapping. For example, TERS imaging of 2D polymers with a spatial resolution of < 10 nm was recently reported by Shao et al. 18 and Müller et al. 29 , using a step size of 5 nm. Therefore, TERS mapping with a smaller step size is required to accurately determine the highest spatial resolution possible with the TERS probe used in our measurement. Supplementary Fig. 6 a SEM image of a representative Au coated TERS tip used in this work. Scale bar: 100 nm. b Table listing the size of the Au grains present at the TERS tip-apex at the locations marked in a. The size of the Au grains at the TERS tip-apex ranges from 24 nm -47 nm. Supplementary Fig. 7 Spatial resolution obtained in TERS images of a graphene and b TMDs reported in literature 11,26,27,28,30,31,32,33,34,35,36,37,38,39 . For comparison, the reported spatial resolution of scanning near-field optical microscopy (SNOM) performed on TMDs is also presented in b. Black dashed lines marked in a and b signify the spatial resolution obtained in this work.

Supplementary Note 4: Reproducibility of high-resolution TERS mapping
In Figure 2 and 4, we have shown high-resolution TERS mapping of GO-COOH using two different TERS tips on two different flakes of the same sample. This indicates that different TERS tips can be used to obtain spatially resolved chemical mapping of a GO-COOH sample at the nanoscale. However, in order to further demonstrate the reproducibility of highresolution TERS mapping, we carried out TERS mapping of the same GO-COOH flake using three different TERS tips. Results of these measurements are presented in Supplementary   Fig. 8. Supplementary Fig. 8a shows the AFM topography image of the GO-COOH flake obtained whilst TERS mapping. Supplementary Fig. 8b and 8c; 8d and 8e; 8f and 8g show the TERS maps of D and G band intensity obtained using TERS tips 1 -3, respectively. These results demonstrate that high-resolution TERS mapping of the same GO-COOH flake can be reproducibly obtained using different TERS tips.
The spatial resolution of the TERS maps obtained using tips 1 -3 is estimated from the FWHM of a Gaussian fit to the line profiles across a sharp feature marked in Supplementary Fig. 8b, 8d and 8f, respectively. A similar TERS spatial resolution of 20.1 nm, 21.5 nm and 17. 5 nm is obtained using TERS tips 1 -3, respectively, as shown in Supplementary Fig. 8h. Note that a step size of 15 nm was used in these measurements, which is comparable to the spatial resolution to the TERS maps. However, a higher spatial resolution was obtained in the TERS map shown in Fig. 2, where a step size of 10 nm was used, indicating that the spatial resolution of TERS mapping could be limited by the step size.
Furthermore, three TERS spectra measured at the same position on the GO-COOH flake marked in Supplementary Fig. 8b, 8d and 8f are shown in Supplementary Fig. 8i. The three TERS tips measure almost the same TERS spectra at these locations, which indicates the high reproducibility of chemical imaging in these measurements.

Supplementary Note 5: Analysis of discontinuities within GO-COOH flake
The discontinuities in the GO-COOH flake were located by checking individual TERS spectra from the areas of low signal intensity in the TERS map shown in Fig. 3a. However, in this map, due to the very high signal intensity at certain locations, the low intensity areas are not clearly visible. Therefore, in order to highlight these areas, we remade Fig. 3a with a logarithmic scale, which is shown in Supplementary Fig. 9a. In this map, the very low intensity areas within the GO-COOH flake can be visualised more clearly. Averaged TERS Intensity (a.u.) Supplementary Fig. 9 a Zoomed-in image of TERS D band intensity from the region marked with a dashed square in Fig. 2c, presented with a logarithmic intensity scale. Note that this image represents the same TERS map shown in Fig. 3a, which has a linear intensity scale. Scale bar: 50 nm. b Averaged TERS spectra from very low intensity areas marked as 1 -5 in a. A TERS spectrum from a high signal intensity area at position 6 is also shown for comparison. spectra from five very low intensity areas marked 1-5 in Supplementary Fig. 9a are shown in Supplementary Fig. 9b, where no D or G bands are visible indicating that these locations represent discontinuities within the GO-COOH lattice. The areas of the discontinuities were found to vary from 100 nm 2 to 600 nm 2 (1 -6 pixels), which is of the order of step size used in the TERS map (10 nm) signifying that the discontinuities are very small in size. In the TERS map shown in Supplementary Fig. 9a, discontinuities were found to occupy 6.4 % of the total surface area. Supplementary Fig. 10 a Zoomed-in TERS map of the D band intensity from the region marked using a dashed square in Figure 2b. Scale bar: 50 nm. b -f TERS spectra similar to the spectra measured at positions A -E marked in a. Supplementary Fig. 11 Histogram showing the statistical distribution of the type (A-E) of TERS spectra shown in Supplementary Fig.10b -10f measured at 38 high intensity pixels of the TERS map in Supplementary Fig. 10a.   Supplementary Fig. 12 Variation of I D /I G ratio calculated from the TERS spectra at the pixels across position D shown in Fig. 3c. In this plot, pixel position 3 represents the position D marked in Fig. 3a. spatial resolution of KPFM is also apparent from the relatively gradual variation of CPD inside the GO-COOH flake as shown in Supplementary Fig. 15b, where the CPD is observed to vary significantly over a length-scale of hundreds of nanometres. Therefore, in order to have a meaningful correlation of CPD variation with the local defect density, we calculated the I D /I G ratio from the TERS spectrum averaged over an area of 0.012 µm 2 (50 pixels) from eight different locations over the GO-COOH flake shown in Supplementary Fig. 15 and correlated it with the average CPD measured in the same areas.