Structure Evolution of Graphitic Surface upon Oxidation: Insights by Scanning Tunneling Microscopy.

Oxidation of graphitic materials has been studied for more than a century to synthesize materials such as graphene oxide, nanoporous graphene, and to cut or unzip carbon nanotubes. However, the understanding of the early stages of oxidation is limited to theoretical studies, and experimental validation has been elusive. This is due to (i) challenging sample preparation for characterization because of the presence of highly mobile and reactive epoxy groups formed during oxidation, and (ii) gasification of the functional groups during imaging with atomic resolution, e.g., by transmission electron microscopy. Herein, we utilize a low-temperature scanning tunneling microscope (LT-STM) operating at 4 K to solve the structure of epoxy clusters form upon oxidation. Three distinct nanostructures corresponding to three stages of evolution of vacancy defects are found by quantitatively verifying the experimental data by the van der Waals density functional theory. The smallest cluster is a cyclic epoxy trimer. Their observation validates the theoretical prediction that epoxy trimers minimize the energy in the cyclic structure. The trimers grow into honeycomb superstructures to form larger clusters (1-3 nm). Vacancy defects evolve only in the larger clusters (2-3 nm) in the middle of the cluster, highlighting the role of lattice strain in the generation of vacancies. Semiquinone groups are also present and are assigned at the carbon edge in the vacancy defects. Upon heating to 800 °C, we observe cluster-free vacancy defects resulting from the loss of the entire epoxy population, indicating a reversible functionalization of epoxy groups.

Introduction

Oxidation of the graphitic materials has been used to synthesize graphene oxide[1,2] with a broad range of applications.[3−7] Controlled oxidation has been shown to yield angstrom-scale vacancy defects in graphene, making it promising for molecular separation.[8−11] Oxidation has been also successfully used for thinning, opening, and unzipping carbon nanotubes.[12,13] However, despite such a broad interest in the oxidized graphitic materials, studies on characterizing the structure of oxidized graphitic materials with atomic-scale insights have been limited. To the best of our knowledge, there are no experimental studies focused on elucidating the early stages of evolution of the oxidized graphitic materials.

Theory predicts a dominant role of epoxy groups (O adatom connected to a bridging C–C site) in the initial stages of oxidation.[14−16] Ab initio calculations indicate that the epoxy groups organize into an energy-minimizing oxygen cluster attributed to a low barrier for their diffusion (∼0.74 eV) on the surface of graphene.[14,15] Linear chains of epoxy/ether groups have been predicted to be responsible for the unzipping phenomenon that leads to fault lines in graphite.[17,18] However, none of this is validated by experimental studies. This is mainly because of the challenges in characterizing atomic-scale organization in these clusters which is further complicated by the mobile and reactive nature of epoxy groups at room temperature. The clusters are difficult to visualize by high-resolution transmission electron microscopy (HRTEM) because the energy supplied by the electron beam to the cluster under typical imaging conditions[19] (80 keV beam) exceeds the energy barrier for gasification of clusters into CO or CO2 (∼1.1 eV).[20] As a result, even a short beam exposure will modify the cluster structure.

Scanning tunneling microscopy (STM) is a powerful tool for the characterization of epoxy groups and their clusters on graphite. One of the first attempts to study oxidation of graphite by STM was made by Chang and Bard, where they studied formation of large (50–100 nm sized) etched pits (vacancies) upon heating highly oriented pyrolytic graphite (HOPG) in air.[21] Beebe et al. studied the kinetics of formation of 100 nm-sized monolayer pits by measuring the expansion rate of these pits.[22] Tandon et al. reported the formation of circular pits that elongated upon prolong oxidation.[23] However, oxygen clusters and their structure, the subject of this study, were never investigated. Recently, Hersam et al. observed formation of isolated epoxy groups upon a mild oxidation of graphene;[24] however, evolution of these groups into larger clusters and eventually into vacancies was not pursued.

Herein, we report the detailed structure and morphology of the oxygen clusters on graphitic surface by using a low-temperature scanning tunneling microscope (LT-STM) operating at 4 K. The imaging near absolute zero temperature helped circumvent the issue of cluster growth or reorganization during imaging. Although we reported the observation of oxygen clusters on a graphitic surface in a recent study,[25] the structure of these clusters, including their evolution, the main topic of this study, was not discussed.

We observed three distinct oxygen clusters: (i) subnanometer-sized clusters which we attribute to cyclic epoxy trimers; (ii) 1–3 nm sized clusters which evolve from the assembly of epoxy trimers in a honeycomb superstructure; and (iii) donut-shaped clusters with carbon vacancies in the middle of the cluster. The formation of vacancy defects is attributed to lattice strain in the presence of ether groups, also corroborated by the fact that the vacancies form in the middle of the cluster. We show that the clusters surrounding the vacancies can be removed by heating to 800 °C, leaving predominantly semiquinone functional groups that uniquely attach to the carbon edge in the vacancy defects.

Experimental Section

Oxidation

The samples were prepared by a short O3 treatment of highly oriented pyrolytic graphite (HOPG) at 250 °C.[25−27] A freshly cleaved HOPG (mosaic spread 0.3–0.5 degree, ScanSens) was treated by O3 at 250 °C. Details of this oxidation protocol can be found in recent studies.[25,26] Briefly, HOPG was heated in an Ar flow at a pressure of 0.8 Torr. Upon stabilization of the temperature, a millisecond pulse for O3 was introduced. The reaction was stopped after 500 ms of O3 exposure by purging O3 with Ar. The very short oxidation helps to study initial stages of evolution as pore expansion becomes dominant at longer exposure times. After cooling down, the sample was quickly transferred (within 15 min) to the cooled STM chamber to ensure no substantial reorganization of clusters.

Imaging

STM imaging was carried out using an LT-STM system (CreaTec Fischer & Co. GmbH). The STM chamber was kept at 4 K and a base pressure of 1 × 10–11 mbar. The STM probe was prepared by mechanical cutting method from a commercial Pt/Ir wire (Pt: 90 wt % and diameter of 0.25 mm, Alfa Aesar). The bias voltage used for scanning the sample’s surface was −0.05 V, and the tunneling current was 0.5 nA. The lack of thermal drift was confirmed by scanning the same area three times (Figure S2). Atomic resolution imaging was conducted to confirm the tip quality and to avoid tip contamination and double tip issues. The scanning direction was changed from 0 to 180° to further check and confirm the tip quality (Figure S3). The tilt in the acquired STM data was reduced by flattening by using WSxM software.[28]

XPS

XPS measurements on the freshly cleaved and O3-treated HOPG were carried out on an Axis Supra (Kratos Analytical) using the monochromated Kα X-ray line of an aluminum anode. The pass energy and the step size were set to 40 and 0.1 eV, respectively. To avoid charge buildup, we grounded the samples. The binding energy data were used without any corrections. The data were processed by using CasaXPS software. The background was subtracted using the Shirley method.

STM Simulation

To develop the STM images, we performed van der Waals (vdW) density functional theory (DFT) calculations using the Quantum ESPRESSO package.[29,30] A supercell made of 7 × 7 periodic unit cells of bilayer graphene was used to ensure the decoupling of in-plane molecule–molecule interactions. The Brillouin was sampled with uniform 4 × 4 × 1 unshifted k-point grids. A vacuum region of 20 Å was used in the z-direction to avoid interactions among the periodic replicas. An energy cutoff of 90 Ry was used for the plane wave expansion of the wave functions. A kinetic energy cutoff of 720 Ry on the charge was used together with ultrasoft pseudopotentials.[31,32] To include the contributions of dispersion interactions, we used the vdW-DF2[33] approximation.

First, the system was relaxed to the lowest energy configuration of oxygens atoms on the graphene plane. The surface geometry was optimized with the convergence thresholds of 2 × 10–6 Ry and 1 × 10–4 Ry/Bohr for the total energy and forces, respectively. The simulated STM images were then generated within the Tersoff–Hamman approximations.[34]

Results and Discussion

We observed a high density (4 × 1012 cm–2) of bright protrusions (clusters) with distinct morphologies in the oxidized HOPG sample (Figure A, B). Four distinct nanostructures could be identified and are indicated by colored arrows in Figure B; small clusters (blue arrow), large clusters (red arrow), donut-shaped features composed of multivacancies (nanopores) surrounded by clusters (yellow arrow), and nanopores without a clusterlike feature (black arrow, the cross-sectional profile is shown in Figure S1). The clusters, apart from their size, also vary in shape where some clusters are circular, whereas others are elongated. The donut-shaped clusters hosting a nanopore in their center are the most abundant feature. We further zoomed in on the scanning area and obtained lattice-resolved images (Figure C) where the cluster’s nanostructure could be confirmed.

Figure 1

STM images of O3-treated HOPG with a scanning area of (A) 300 nm × 300 nm, (B) 75 nm × 75 nm, and (C) 10 nm × 10 nm. Bias voltage: −0.05 V. Tunneling current: 0.5 nA. (D) Cross-sectional profile of the line A–A′ drawn in C. (E) Three-dimensional STM images of C. Statistical analysis of the dimension of the donut-shaped cluster for (F) cluster width and (G) cluster length.

We note that the atomic-resolution images in this study proves that the observed clusters are not the artifacts. It was shown that the epoxy group on the graphitic surface can be removed by scanning with a large bias voltage (+4 V).[24] Therefore, in this study, all images were acquired with a small bias voltage (−0.05 V). These clusters do not change after repeated scans (Figure S2). The reproducible repeated scanning of the same set of features proves that the tip current does not influence the cluster structure.

Figure D shows the cross-sectional profile of the donut-shaped cluster in Figure C along the A–A′ line. The height profile reveals three zones; graphite basal plane (black arrows), oxygen clusters (blue arrows), and nanopore (red arrow) in the middle of oxygen clusters. The assignment of zone with oxygen cluster and nanopores is based on the fact that the functionalized oxygen atoms (e.g., epoxies) protrude out of the two-dimensional plane, whereas the vacancies (pits) will appear as a valley.

The height of the zone assigned as vacancies is significantly lower than that of the oxygen cluster area and in fact is lower than that of the basal plane of the HOPG. This is further illustrated by the side- and top-views of the three-dimensional (3D) profile of the cluster (Figure E).

To understand the size distribution of donut-shaped clusters, we conducted a statistical analysis on the cluster dimension by sampling data from 60 donut-shaped features. The resulting histograms revealing cluster width and length distributions are shown in panels F and G in Figure , respectively. The width of the donut cluster (Figure F) was measured by drawing the cross-sectional profile vertical to the elongation direction of the donut cluster, whereas the length (Figure G) was measured by drawing the cross-sectional profile parallel to the elongation direction. The cluster width and length have a Gaussian distribution. The width varies from ∼1.9 to ∼2.5 nm with a mean size of 2.1 ± 0.2 nm (Figure F), and the length varies from ∼1.9 to ∼2.9 nm with a mean size of 2.4 ± 0.3 nm (Figure G). The elongated shape of the clusters likely has its origin in the way these clusters are formed, i.e., linear cooperative alignment of the epoxy groups.[18,35]

The chemical composition of the clusters was studied by high-resolution XPS. As a control, a freshly cleaved HOPG substrate was also analyzed, yielding line shape references for the C1s and O1s photoelectron peaks. XPS analysis confirmed the absence of carbonaceous species on the freshly cleaved HOPG control sample and only adsorbed water at 532.8 eV was detected (Figure S4C).[36] The XPS spectrum of the oxidized sample is shown in Figure S5, where apart from C1s (Figure A), a significant O1s peak (Figure B) could also be observed. The C1s spectrum has an asymmetric peak-shape centered at 284.5 eV and π to π* transition (shakeup) at around 290.9 eV, which is typical of the graphite samples.[37] A small shoulder at 287.6 eV corresponds to the oxygen functional groups.[38] They are confirmed in the O1s spectrum, which could be deconvoluted into four peaks. These are ether groups[39] at 533 eV (∼51%), epoxy groups[39] at 531.6 eV (∼31.5%) and semiquinone groups[39] (C=O) at 530.7 eV (10.4%). A small peak at 532.8 eV (7.1%) could be attributed to adsorbed water, which was also observed on the control sample, given that they were exposed to the ambient atmosphere for a brief period during the sample transfer to the XPS chamber. The above analysis proves that clusters are primarily comprised of ether and epoxy groups, whereas semiquinone groups are present at the edge of the vacancy defects. The latter is because of the bonding requirements, i.e., carbon atom with semiquinone functional group can only be bonded to two other carbon atoms, a constraint that is satisfied at the pore edge.[35,40] This confirms the theoretical prediction that semiquinone groups should form during the oxidative cleavage of the C–O–C bond.[35]

Figure 2

XPS spectra of O3-treated HOPG. (A) C1s spectrum. (B) O1s spectrum.

To understand the cluster structure, we first investigated the structure of the smallest cluster. These clusters were triangular-shaped (Figure A) as against circular-shaped protrusions reported for a single O functionalization.[24] The 3-fold symmetry in these clusters point toward cyclic epoxy trimers, reported by theory to be one of the most stable configurations for three epoxy groups on the same honeycomb ring of graphene.[16,35] To confirm this, we simulated the STM image of cyclic epoxy trimer using van der Waals (vdW) density functional theory (DFT) calculations. Indeed, the simulation predicts (Figure B, C and Figure S6A) quantitatively similar-sized protrusions with a 3-fold symmetry, confirming that these are indeed cyclic epoxy trimers, validating the theoretical prediction of the existence of these clusters.

Figure 3

Nanostructure analysis for the oxygen clusters. (A) STM image of a single epoxy cyclic trimer. (B) Structure of the single epoxy cyclic trimer. (C) Simulated STM image for the single epoxy cyclic trimer. (D) STM image of a large oxygen cluster. (E) High-magnification STM image of the area circled in (D). (F) Configuration of the epoxy cyclic trimers in the oxygen cluster shown in E. (G) Structure of the epoxy cyclic trimers formed an infinite oxygen cluster. (H) Simulated STM image for the epoxy cyclic trimers formed infinite oxygen cluster as shown in G. Bias voltage: −0.05 V. Tunneling current: 0.5 nA.

With the confirmation of cyclic epoxy trimers, we looked to resolve the structure in the larger clusters. For this, we carried out lattice-resolution imaging of one of the clusters. STM image in Figure E reveals the structure of the cluster away from the vacancy defect (Figure D). We observe a honeycomb superstructure with an average diagonal length of 4.8 ± 0.07 Å (Figure E and Figure S7), much larger than the diagonal length of a graphene honeycomb (2.84 Å). A proposed chemical structure of these superstructures, formed by adjoining linear chains of epoxy cyclic trimers, is shown in Figure F. Each epoxy cyclic trimer in the proposed structure is numbered, and the same numbering is listed in the STM image to help visualization.

The proposed honeycomb network of epoxy cyclic trimers is quantitatively confirmed by the vdW DFT calculations. For this, we built a honeycomb network of epoxy cyclic trimers on the top layer of bilayer graphene (Figure G). The simulated STM image (Figure H) agrees well with the STM image, confirming the honeycomb superstructure of cyclic epoxy trimers. The formation of honeycomb superstructure is consistent with energy minimization of the cluster as it uniquely minimizes the disruption of the benzene-like rings in graphene (Clar’s aromatic π-sextet rule[41]).

We also carried out vdW DFT simulation for a honeycomb network of ether trimers (the relaxed structure of a cyclic ether trimer is shown in Figure S6B). The simulations reveal a much larger honeycomb with a large diagonal (∼6.17 Å, Figure S8). This is expected because the O atom in ether group pushes apart the two neighboring C atoms in the honeycomb, and therefore, expands the lattice. In any case, ether groups are expected to be present in the clusters as indicated by XPS. We hypothesize that to minimize the strain the ether groups are likely present near the center of the cluster closer to the vacancy defects; however, structural imaging near the center of the cluster extremely challenging because the structure in this area is highly distorted because of the following two contributions:

vacancy defects form at the center of the cluster, which are known to distort the lattice;[40]

the large strain presented by the ether group within a 2 nm size cluster will further distort the lattice.[35,42]

As discussed before based on the XPS data, the cluster also hosts a semiquinone group, which can only be hosted at the pore edge. To prove this further, and to understand the thermal behavior of these clusters, we heated the sample to 800 °C in the ultrahigh vacuum chamber of the STM. We note that this temperature is below the stability limit of the semiquinone groups (920 °C).[43,44] After heating, we observed a significant loss of the oxygen functional groups (Figure A). The leftover functional group is ether and semiquinone with semiquinone now constituting a much larger percentage of total functional group (26.0%), compared to only a minor percentage (10.4%) before heating. There were no epoxy groups after heating attributing to limited stability of epoxy groups.[24,43,44] This is also indicated by STM image of the specimen (Figure B) where vacancy defects are observed but majority of the oxygen functional groups are removed. The size of the vacancy defects was comparable to that of before heating, indicating that the removal of functional group did not result in a significant expansion of vacancies. High-magnification STM images (Figure C, D) clearly indicate a vacancy defect; however, the majority of oxygen functional groups surrounding the defects are missing. The functional groups seem to be present only close to the edge of the vacancy defects (Figure D–F).

Figure 4

(A) XPS O1s spectrum of O3-treated HOPG after 800 °C heating treatment. The STM images of the O3-treated HOPG surface after annealing at 800 °C with a scanning area of (B) 30 nm × 30 nm, (C) 6 nm × 6 nm, and (D) 3.2 nm × 2.6 nm. (E) 3D STM images of the nanopore shown in panel D. (F) Cross-sectional profile of line B–B′ in C. Bias voltage: −0.05 V. Tunneling current: 0.5 nA.

On the basis of the above analyses, we present the following evolution insights for the vacancy defects (Figure A–C). These are (i) formation of an epoxy group followed by the formation of cyclic epoxy trimers (Figure A, Figure A); (ii) organization of the cyclic epoxy trimers in a honeycomb network (Figure E), leading to larger clusters (Figure B); and (iii) strain-related cleavage of the C–O–C bond, leading to the formation of semiquinone-functionalized vacancy defects (Figure C). It has been shown that the cyclic epoxy trimers are the most stable epoxy clusters of the three epoxy groups, as they minimize the net energy of the system.[16] As the oxidation reaction proceeds, cyclic trimers add to the existing trimer (nuclei), leading to the formation of an extended honeycomb network of trimers. The honeycomb superstructure of cyclic trimers uniquely minimizes the disruption of the aromatic rings in graphene (Clar’s aromatic π-sextet rule[41]) and therefore are favored. The generation of strain in these cluster takes place because of the evolution of epoxy (a C–O–C group where the C–C bond is intact) superstructures into ether (a C–O–C group where the C–C bond is broken) superstructures, given a low energy barrier of cyclic epoxy trimers to convert into cyclic ether trimers.[35] This conversion takes place by following the cleavage of the C–C bond in the epoxy group. Subsequently, the O atom in the ether group pushes apart the two C atoms, reducing the two-dimensional character of the lattice (out-of-plane protrusion of 2.15 and 2.0 Å for ether and epoxy, respectively, Figure S6). The strained ether superstructures have been shown to yield a pair of semiquinone group following the cleavage of C–O–C bond of the ether group.[35] This then promotes gasification of semiquinone groups in CO as the energy barrier for the gasification step is quite low (1.1 eV).[20] The gasification event removes carbon from the lattice leading to a vacancy defect. The process of formation of semiquinone groups and gasification then continues from the further structural rearrangement depending on the size of the clusters (number of oxygen groups).

Figure 5

Three-dimensional STM images (5 nm × 5 nm) of the oxidized HOPG surface to illustrate the different etching stages. (A) Several images of small epoxy clusters. (B) Several examples of larger clusters where carbon vacancies are not yet formed. (C) Several examples of donut-shaped clusters where carbon vacancies form in the middle of the cluster. Bias voltage: −0.05 V. Tunneling current: 0.5 nA.

Conclusions

In summary, by imaging and analyzing the surface structures on oxidized graphite by low-temperature STM aided by XPS and vdW DFT calculations, we provided insights into structure evolution of oxygen clusters which ultimately yield carbon vacancies. Oxidation starts with the formation of epoxy groups, which form cyclic epoxy trimers. The observation of these trimers validates the theoretical predictions for the first time. Cyclic trimers subsequently aggregate to form an extended honeycomb of cyclic trimers. Ultimately, strain in the larger clusters give way to vacancy defects at the center of the cluster. Only large clusters yield vacancy defects, indicating the key role of strain buildup inside the larger oxygen clusters. The majority of the oxygen functional groups desorb upon heat treatment with the exception of the semiquinone groups at the edge of the vacancy defects. The insights presented here are expected to guide future theoretical and experimental studies on the functionalization of graphitic materials.