Mitigating e-beam-induced hydrocarbon deposition on graphene for atomic-scale scanning transmission electron microscopy studies

Atomically resolved imaging via scanning probe and electron microscopy has opened the doors to the nanoworld by providing a pathway to visualize atomic structure and explore functional properties on a single atom and molecule level.¹ Multiple examples of these studies include the order parameter fields in ferroic materials,^2,3 octahedra tilts,^4–7 chemical strains,⁸ and local chemical properties^9,10 in (scanning) transmission electron microscopy (STEM) and energy loss spectroscopy.

However, the success of atomically resolved imaging often hinges on the availability of the well-prepared samples. For (S)TEM, typically the requirements include the stability and ability to form thin foils, whereas the surface stability on the order of 1–2 nm is often irrelevant. However, in studies on graphene, which is only one atom thick, any thickness of contamination significantly obfuscates the sample. Nowadays, it is recognized that STEM can also be used as a tool for single atom fabrication, where the electron beam is employed to induce controllable chemical transformation including vacancy,^11,12 adatom,¹³ and interstitial motion,^14,15 bond formation,^14,16–18 vacancy ordering,¹⁹ phase changes,²⁰ etc.,^21,22 that can further be atomically resolved. Thus, the requirements for the sample preparation are becoming much more stringent on transition from imaging to fabrication.

Combination of STEM with controlled beam motion and image-based feedback further enables atom-by-atom fabrication in STEM. The fabrication of atomic structures via STEM necessitates high-quality sample preparation. In many cases, chemical vapor deposition (CVD) grown graphene samples are prepared for (S)TEM investigation using a poly(methyl-methacrylate) (PMMA)-mediated approach for transfer from the growth substrate to the TEM grid.^23–25 With this technique, after the graphene is grown on a metal foil substrate, it is mechanically stabilized with a coating of PMMA, after which the metal foil is etched away. The graphene, attached to the layer of PMMA, may then be transferred to an arbitrary substrate and the PMMA removed with solvents. Although this approach appears to be quite common, it leaves behind PMMA and other organic residue so that the graphene is almost completely obscured from view when examined in a (S)TEM (see Fig. 1). In addition, samples prepared in this way also exhibit significant e-beam induced hydrocarbon deposition, so that the small areas of pristine graphene will become covered in mobile contaminants upon imaging and thus obscured. Many attempts have been made to address these issues with varied success and they generally involve a thermal annealing of some kind. Van Dorp et al.²⁶ show that heating exfoliated few-layer graphene to 500 °C for 10 min in the microscope is sufficient to remove the contaminant materials. These results are supported by the investigations of Xie et al.²⁷ who performed x-ray photoelectron spectroscopy and time-of-flight secondary ion mass spectrometry. Lin et al.^28,29 advance an air and H₂/Ar 200 °C annealing procedure prior to examination in the TEM, which appears to be quite effective. Liang et al.³⁰ introduce an alternate PMMA transfer method that appears to clean graphene but they fail to demonstrate cleanliness at the atomic level. Finally, Li et al.³¹ give evidence that exposing clean CVD-grown graphene to air results in significant hydrocarbon contamination within just a few minutes. This may result in conflicting reports where graphene that has undergone a cleaning treatment, is no longer clean when investigated in the (S)TEM and the cleaning treatment is assumed to have failed. Very recently, Tripathi et al.³² performed a series of experiments demonstrating various successes with thermal cleaning of graphene in air and vacuum. However, what none of these studies directly address is the e-beam-deposited hydrocarbon contamination that typically occurs under high magnification in the STEM.

Here, we investigate two methods for cleaning graphene to address both visible contaminant materials present on the graphene surface and particularly focus on contamination resulting from e-beam-induced hydrocarbon deposition and discuss our observations. The first method we investigate is an ex situ Ar/O₂ anneal suggested by Garcia et al.³³ The second method we investigate is in situ annealing.

CVD-grown graphene was transferred from the Cu foil growth substrate to a TEM sample grid followed by an Ar/O₂ anneal for the removal of volatile adsorbents. The Cu foil was spin-coated with PMMA to stabilize the graphene, and the Cu foil was etched away in a bath of ammonium persulfate-deionized (DI) water solution. The graphene/PMMA layer was transferred first to a hydrogen chloride (HCl) diluted in DI water and then to a DI water bath to remove residues of ammonium persulfate. The graphene was transferred to the final TEM substrate by scooping it from the bath and heated on a hot plate at 150 °C for ∼20 min to improve adhesion of the graphene to the grid. The PMMA was subsequently dissolved in an acetone bath followed by rinsing in isopropyl alcohol. Finally, the samples were baked in an oven under an Ar/O₂ (90%/10%) environment to remove residual PMMA and volatile organic compounds.

For the in situ heating experiments, graphene was transferred to a Protochips Aduro heating holder and heated at a rate of 1000 °C/ms in the microscope.

Imaging of the samples was performed in a Nion UltraSTEM U100 at an accelerating voltage of 60 kV in high angle annular dark field (HAADF) imaging mode. Operating beam current was in the range of 60–70 pA, and vacuum at the sample was in the range of $2×10−9−2×10−8$ Torr. The samples were loaded into the microscope using our standard loading procedure, where the microscope magazine, cartridges, and samples were baked in a vacuum chamber at 160 °C for 8 h prior to insertion into the microscope, except where explicitly indicated in the text [i.e., discussion related to the results presented in Figs. 5(a)–5(c)].

To compare the cleaning techniques to a standard sample, images were taken of a graphene sample prior to any cleaning. Figure 1 shows an as deposited sample at a variety of magnification levels, exhibiting heavy contamination. The observable clean areas are extremely small and not even visible at the lowest magnification shown. Even when these small, clean areas are found, hydrocarbon e-beam deposition is often an intractable problem which acts to quickly cover the region of interest with amorphous carbon. This problem can be mitigated to some extent by performing a so-called “beam shower” by illuminating a large area for ∼30 min. It is thought that this procedure deposits volatile contamination onto the sample across the exposed area leaving the vacuum slightly cleaner. In our experience, this technique appears to work to some degree for a period of time, sometimes as long as 1–2 h, before it must be repeated. Nevertheless, this situation is not ideal and here we explore other options.

To produce more favorable graphene samples for STEM studies that exhibit no e-beam hydrocarbon deposition we adopted the cleaning procedure of Garcia et al.³³ This procedure involves heating the graphene sample to 500 °C in an environment of Ar (90%) and O₂ (10%). The studies of Garcia et al. investigated sample cleanliness through the use of Raman spectroscopy and were primarily targeted at removing adhesive residue from exfoliated h-BN. We applied this technique to CVD grown graphene, transferred to a TEM grid through a PMMA-stabilized transfer process as described in Sec. II.

Figure 2 shows the contamination morphology after the Ar/O₂ cleaning procedure. Figures 2(a)–2(d) show the observed morphology at a variety of magnifications after microscope alignment. The central area in each image displays pristine areas of graphene due to exposure to the electron beam during alignment. These areas appear dark as they are only a single atom thick. Initially, there appears to be a significant amount of contamination covering the graphene (the areas away from the center). It is unclear, quantitatively, whether it is truly cleaner than the as-transferred graphene as far as contamination coverage is concerned because the contamination appears to contract on itself or move away from the illuminated area under e-beam influence. In other words, simply observing the sample acts to change it. Lattice resolved images produce large (10–20 nm) clean areas of graphene within a few seconds which were not present previously. To illustrate this phenomenon on a mesoscopic scale, a large area was selected and illuminated in parallel (not scanned) for 20–30 min. The before and after images are shown in Figs. 2(e) and 2(f), where we observe a significant increase in pristine graphene areas over the entire illumination area (dashed circles). The insets in Figs. 2(e) and 2(f) show a zoomed-in portion of the same area before and after illumination. We can see that the exposed contamination is revealing pristine graphene similar to the central portion of (b). Additional images and a video time-lapse series of this phenomenon are provided in the supplementary material.³⁵ Of note: smaller areas scanned with the beam become clean more quickly so that, after microscope alignment, imaging and experimentation can begin immediately with ample areas of pristine graphene.

While this result was rather unexpected and fortuitous, the truly remarkable property of such samples (remarkable based on our previous experience with graphene samples) is that we have not observed any e-beam induced hydrocarbon deposition at any magnification on samples cleaned with this method. Lattice resolved images may be taken at leisure for many hours without fear of encroaching contamination. Indeed, the contamination contracts away from the irradiated area, exposing more pristine graphene. This is important, as graphene sample quality has historically made atomic level studies difficult and sometimes impossible. The mechanisms of motion and chemical make-up of the contamination observed are beyond the scope of this article. We simply wish to highlight this recipe as a highly effective way to produce atomically clean areas of CVD grown graphene for (S)TEM studies.

While the above described method for cleaning graphene produces highly agreeable samples for (S)TEM studies, the samples are nevertheless, still covered in contaminant material. In order to fabricate a graphene sample that is atomically clean on the micron length scale we used a Protochips Aduro heater chip to heat the sample to 1200 °C at a rate of 1000 °C/ms. The results are summarized in Fig. 3. In Figs. 3(a)–3(c), we see the typical contamination of an as-prepared sample at several magnifications. After heating the sample, the images in Figs. 3(d)–3(f) were acquired and we see that the suspended graphene film is mostly atomically clean graphene over the entire observable window. This result appears similar to that observed by Tripathi et al.³² using vacuum laser heating and appears slightly superior to their result using radiative heating. This may indicate that a rapid heating rate may also improve the effectiveness of the procedure, though further experimentation is required to investigate this possibility. The location that remained dirtiest is boxed in Figs. 3(d) and 3(e) and corresponds to the previously irradiated area, discussed later. This rapid thermal cleaning procedure was repeated with several different samples with similar results.

Several observations are noteworthy:

1. After returning the sample to room temperature, the sample remained clean while in the microscope.

2. After returning the sample to room temperature, high e-beam fluence, such as that produced when performing lattice resolved images or under a stationary beam, produced heavy e-beam-induced hydrocarbon deposition. This deposition is shown in Figs. 4(a) and 4(b) and may be controllably deposited with the e-beam, Fig. 4(b). We posit that this contamination comes from areas of the heater chip that remain cool even while the chip is heated (the heated area is concentrated to within a few tens of microns around the sample).

3. This amorphous deposited carbon contamination can be converted to graphitic carbon upon heating again to 1200 °C (see supplementary material).

4. Heating the sample again to 500 °C still shows e-beam deposited contamination but heating to 800 °C prevents this deposition (see supplementary material). This is important in practice because, although the heater chip has the capability to ramp to 1200 °C, we find that the sample usually exhibits a slight mechanical instability (vibration) at this temperature, reducing resolution. Backing off from this limit appears to be more mechanically stable. As a result, the lattice resolved image in Fig. 3(f) was acquired at 800 °C since lattice resolution was not possible at 1200 °C [Figs. 4(d) and 4(e) were acquired at 1200 °C]. Although there is much to be explored here, we limit this paper to these cursory observations.

5. We note that any areas that had previously undergone e-beam irradiation or deposition did not become clean with the described heating procedure (see also the observations of van Dorp et al.²⁶] but can be made graphitic upon heating. Figure 4(c) shows a suspended sheet of graphene that had been previously imaged in an SEM to check for the success of the sample transfer. The entire region remains mostly covered in contamination, save for a few smaller patches, and areas exposed to higher e-beam fluence by increasing magnification in the SEM are clearly visible as areas with higher contaminant coverage [i.e., the brighter patches in Fig. 4(c)]. This appears to be similar to the adherence of the contamination noted in Figs. 3(d) and 3(e) which had been exposed briefly to the STEM beam before heating. Although undesirable for cleaning graphene, this is interesting because it immediately suggests possible patterning in a lower magnification instrument (SEM, FIB, etc.) and subsequent thermal treatment to remove the unexposed materials. This technique may be amenable to producing conductive carbon nanowires on h-BN, for example.

Finally, we wanted to explore what happens upon removal of a clean graphene surface from the STEM vacuum and exposure to the laboratory air. We thus, cleaned a graphene sample with the above described heating procedure, cooled the sample back to room temperature, removed the sample from the microscope vacuum, and let it sit on a “gloves only” sample exchange table under a plastic (nonairtight) cover to prevent environmental dust from settling on it. After approximately 4 h the sample was reintroduced into the microscope and imaged again.

Figures 5(a)–5(c) show the results of this procedure. The image in Fig. 5(a) was acquired while the sample was being held at 1200 °C. We observe the same atomically clean graphene over extended lengths. The heater chip support material failed during heating and can be seen laying over the graphene (labeled). After removing the sample from the microscope for 4 h and reintroducing it, the image shown in Fig. 5(b) was acquired. We observe severe e-beam deposition of hydrocarbon contamination and what appears to be a continuous layer of contamination over the entire sheet of graphene, Fig. 5(c) shows a magnified view of the contamination layer and deposition. Lattice resolved images were not possible due to the e-beam deposition, so it is possible there were small, nanometer-sized, areas of pristine graphene remaining. It is unclear whether the heavy contamination is sourced from the air or from adjacent areas of the sample which had not been heated; nevertheless, we feel it is sufficient to conclude that pristine graphene is highly susceptible to the adsorption of hydrocarbon contamination in agreement with the investigations of Li et al.³¹ What is interesting, however, is that loading a sample following our standard loading procedure, where we heat the magazine and cartridge with the sample to 160 °C for 8 h, is sufficient to remove the volatile hydrocarbon contamination observed in (b). Images of this observation are shown in Figs. 5(d)–5(f). The sample shown in Figs. 3 and 4 was removed from the microscope and stored for a week before being imaged again. We observe modest additional contamination following this procedure. The sample still exhibited e-beam-induced hydrocarbon contamination (see supplementary material) but it is not nearly as severe as that shown in Figs. 5(b), 5(c) and visible contaminants observed on the surface were limited to a few dendritic structures that appear to have grown from the edge of the support substrate as well as speckles of contamination over the entire surface of the graphene. Upon heating again to 1200 °C, Fig. 5(f), most of this additional contamination was removed.

This result suggests that, although pristine graphene becomes heavily contaminated in air [Figs. 5(a)–5(c)], a vast majority of this contamination may be removed by heating to 160 °C in vacuum for 8 h [Fig. 5(d)] which is starkly different from the kind of contamination observed in Fig. 1 on the control sample which had also undergone the same heating procedure, yet remained heavily contaminated.

We have explored two effective graphene cleaning procedures which we have been using to clean graphene for our STEM studies and we have provided our observations regarding their effectiveness for this purpose. In particular, we note that the Ar/O₂ cleaning procedure produces very agreeable samples for (S)TEM investigation at room temperature without risk of e-beam-induced hydrocarbon deposition. From a physics standpoint these samples are most interesting because there are other “contaminant” atoms sitting on the surface and frequently found in the lattice, or can even be put into the lattice, as we have recently demonstrated.³⁴ This provides a wealth of physical phenomenon to explore on a single sample. From an engineering perspective (building devices) such a sample may not be ideal since it is still technically covered with contamination. In order to produce atomically clean graphene on a mesoscopic scale (microns) for device fabrication, the Ar/O₂ cleaning procedure may be insufficient. In order address this, we demonstrated the effectiveness of a rapid thermal annealing procedure where the sample is heated to 1200 °C at a rate of 1000 °C/ms, which immediately produces atomically clean graphene. [We were unable to determine how quickly this process occurs. In the time it took to open the gun valve and examine the sample again, the graphene was clean. So, for practical purposes in (S)TEM investigations, this procedure appears to “immediately” produce clean graphene.] We also detail several observations regarding what occurred at various temperatures following the cleaning, namely: returning to room temperature retains the clean graphene but significant e-beam hydrocarbon deposition is observed, reheating of the amorphous deposited carbon converts it to graphite, and heating to 800 °C is sufficient to prevent hydrocarbon deposition under the e-beam. In addition, we also observe that clean graphene readily contaminates with volatile species in air which can mostly be removed by a 160 °C anneal for 8 h in a vacuum chamber. While the observations noted here are wide ranging, and each result is not extensively investigated from a physical perspective, we hope that, from a practical perspective, these results will be of help to the microscopy community in preparing contamination free graphene samples for atomic-scale studies.

The authors would like to thank Ivan Vlassiouk for the provision of the graphene samples and Francois Amet for performing the argon-oxygen cleaning procedure. This research was conducted at the Center for Nanophase Materials Sciences, which is a DOE Office of Science User Facility (S.V.K.) and was supported by the Laboratory Directed Research and Development Program of the Oak Ridge National Laboratory, managed by UT-Battelle, LLC, for the U.S. DOE (O.D., S.K., S.J.).