We examined several different flakes to ensure that we had selected a typical specimen; electron microscopy images are given in Supporting Information File 1
. The following analysis concerns the single flake shown in . This sample has an average thickness of ~100 nm. Generally, optimal samples for carbon K-edge TXM have an optical density (OD) around 1, which for multilayered graphene samples corresponds to about this thickness. In principle, absorption signals detected by transmission are observable down to an OD of 0.01, which would correspond to a multilayered graphene sample of three layers thick. Even-thinner samples represent a significant instrumental challenge, but with sufficient stability and signal integration it should be possible to measure the carbon K-edge NEXAFS of a single layer of graphene. The thickness of the studied sample ensures a good signal-to-noise ratio allowing detailed analysis of the peak form and composition.
Figure 2 (a) Average of aligned TXM images (283–311 eV). (b) Spectra extracted from the three regions indicated by the coloured shapes in part (a): red is from the lacey carbon support, green is from the flat area of the sample, while blue is from the (more ...)
shows an average of 236 TXM transmission images (after alignment), from 283 to 311 eV. From the differences in contrast we can identify areas with a different number of sheets, and also some steps and some folds (schematically marked in Supporting Information File 1
). presents C 1s spectra from selected regions, indicated by the coloured shapes in . The most intense peak in the C 1s NEXAFS spectra at 285.3 eV is due to electronic excitation from the C 1s level to the conduction π* states. This transition should be forbidden by selection rules, in the geometry used (E
parallel to the basal plane). However, we see a residual intensity due to the folding of a few graphene layers, implying that there is a nonzero angle between the polarization vector and the basal plane in places. This is most notable in the central spectrum, confirming that this region corresponds to a fold in the flake.
The peak near 292 eV corresponds to the σ* threshold. This peak is composed of two distinct features at 291.5 eV and 292.6 eV. The sharp feature at 291.5 eV is an exciton [24
], its sharpness reflecting the strong correlation effects of electron–hole pairs within the flake. The broad peak at 292.6 eV is related to the transition from the C 1s level to the relatively nondispersing σ* states at the Γ point of the Brillouin zone (BZ) [25
For the HZB TXM the E-vector of the light is always in the horizontal plane, which is thus in the plane of the graphene sheets for the unfolded area, but approximately perpendicular to the folds. Thus the blue spectrum () of the fold exhibits a strong C 1s → π* peak at 285.3 eV, since the E-vector is orthogonal to the graphene sheets folded out of the plane of the sample. In contrast, the signal at the corresponding transition energy in the flat part of the graphene sample is negligible, as expected when the E-vector is in the plane of the sheets. In the spectrum recorded in the flat region of the sample, the peak near 285.3 eV has much lower intensity than in the folded region, because the angle between the polarization vector and the basal plane is nearly zero.
The peak at 284.2 eV, very clear in the spectrum of the flat part of the graphene, also exists at about the same intensity relative to the C 1s continuum intensity, in the spectrum of the fold. Schultz et al. [12
] assigned this pre-edge structure to dopant-induced states. As we did not observe this in the spectrum recorded on an amorphous carbon film with the sodium cholate, we suggest that this structure arises instead from metal impurities in the graphite used for exfoliation. A careful examination of the spectrum of the folded region () shows the presence of a shoulder at the same photon energy of this structure, indicating that the doping is rather uniform in the flake.
In a recent theoretical study, spectral features appearing at 287–290 eV photon energy were associated with topological defects, such as monovacancies, divacancies and Stone–Wales defects [26
]. In particular, a structure centred at 287 eV was assigned to carbon atoms beside divacancy defects [26
]. Accordingly, the structure observed at 287.4 eV in the NEXAFS spectra recorded in both the flat and folded regions may be thus associated.
The spectra in can be used as reference signals to discriminate, in the X-ray image of the sample, between the regions that are in-plane (flat regions), out-of-plane (folded region) and the lacey carbon. presents results from a three-component fit by using the spectra shown in as reference signals. In the images, flat and folded regions can be easily distinguished, as can step edges. In the in-plane mapping (σ*) the folded region appears black, i.e., the (π*) intensity is minimum while in the out-of-plane region it is maximum. At the edges of a graphene sheet the π orbitals are tilted, thus a marked increase/decrease in the π* intensity relative to σ* in the in-plane/out-of-plane component map is expected. Combining the three images in , the morphology of the sample can be well described (Figure S1, Supporting Information File 1
Figure 3 Component map of (a) in-plane (σ) and (b) out-of-plane (π) bonds; (c) colour-coded composite of the maps for lacey carbon (red), in-plane (green) and out-of-plane (blue), derived from a three-component fit of the carbon K-edge image sequence. (more ...)
We also scanned over the O 1s spectral range 528–556 eV, and the absence of signal confirms that the sample is essentially oxygen free (Figure S2, Supporting Information File 1
). In addition this confirms the absence of sodium cholate (C24
Na), which could appear, for example, due to residual intercalation in the flake.
In summary, the distinctly different X-ray spectra of the flat and folded regions of the sample, which are related to the strong dichroism of the C 1s spectrum of graphene, have been used in a fitting procedure to generate maps of the flat and folded regions of the flake. A low-lying peak at 284.2 eV is associated with doping states in the electronic structure of the flake induced by metallic impurities present in the graphite that was exfoliated to make the multilayer graphene sample. We tentatively associate a peak at 287.4eV with defects created during the exfoliation using sodium cholate (C24H39O5Na).
The combination of TXM with NEXAFS has the potential to become an important tool in nanotechnology, particularly in the imaging and analysis of nanoscale samples that are sensitive to electron irradiation. The current study shows that it is applicable to the study of carbon nanomaterials, notably large two-dimensional structures for which statistical spectroscopic analyses over large areas are to be preferred.