Schematics of the TEM sample preparation and AFM-based mechanical cleaning processes are shown in Fig. 1. In this study, graphene served as a benchmark sample and other 2D crystals can be potentially processed using a similar sample preparation procedure. We first prepared graphene samples on PDMS film by mechanical exfoliation (Fig. 1a). The exfoliated graphene samples (~ 5 layers) were identified with an optical microscope and subsequently transferred to a holey Si3N4 membrane TEM grid by stamping (Fig. 1a). The stamping process mediated by PDMS film is simple to perform and was widely adapted in many prior studies (Dean et al. 2010; Jain et al. 2018; Rosenberger et al. 2018). In particular, the PDMS-based stamping process has been primarily used to fabricate 2D vertical heterostructures (Dean et al. 2010). However, the surface of 2D crystals prepared by mechanical transfer can suffer from PDMS residues and requires special attention, especially for surface-sensitive studies. After we prepared a TEM sample, we performed AFM contact-mode scanning on the TEM grid. We anticipated that surface residues on graphene could be swept away and a residue-free surface prepared (Fig. 1c).
Figure 2a shows a graphene flake transferred onto the PDMS film. The graphene flake on the PDMS film was positioned onto the Si3N4 membrane region in the TEM grid and physical contact was established between the flake and membrane. After the release, the graphene flake was transferred onto the TEM gird as shown in Fig. 2b. Figure 2c demonstrates a close-up view of the optical microscope image. We then conducted AFM imaging of the graphene flake, which identified the suspended region as shown in Fig. 2d. We used a hole near the graphene flake’s edge, from which we were able to easily find the same location for subsequent TEM investigations.
We mechanically cleaned the graphene surface by contact-mode scanning the sample. We conducted the contact-mode scanning using a rectangular sweeping region, which is shown as the dashed rectangle in Fig. 2d. To directly investigate the efficiency of mechanical cleaning, we intentionally left some suspended sample areas uncleaned. After the contact-mode sweeping, we obtained a topographic image of the sample surface using the non-contact AFM mode. The surface residues accumulated at the rectangular boundary, confirming that AFM-based scanning did indeed mechanically displace the surface residues.
The sample area cleaned with AFM was investigated via TEM characterizations. Figure 2f shows a high-angle annular dark-field (HAADF) STEM image of the hole presented in Fig. 2e. We clearly observed the accumulated residues, which formed a line on the left part of the image (Fig. 2f). The regions on the left and right sides across the residue line had distinct contrast under STEM mode. The right side had darker contrast with less residue coverage than the left-side region, indicating that mechanical cleaning was indeed achieved.
Using EDX mapping, we analyzed the residues accumulated by AFM scanning as shown in Fig. 3. Figure 3b presents the HAADF-STEM image, oxygen K edge, silicon K edge, and carbon K edge intensity mapping data, respectively. Increased oxygen, silicon, and carbon signals occurred at the accumulated residue. The observed data were consistent with our interpretation that the surface residue was mainly PDMS accumulation (Fig. 3a); PDMS is composed of silicon, carbon, oxygen, and hydrogen.
We quantitatively investigated the effect of mechanical cleaning using TEM and STEM imaging as shown in Fig. 4. The as-prepared region without mechanical cleaning had typical graphene residue networks as demonstrated in Fig. 4a. The presumably residue-free region was approximately 10 nm wide. However, the mechanically cleaned region had a larger residue-free region that sometimes spanned an area larger than 20 nm. The close-up high-resolution TEM image clearly revealed a graphene lattice structure, demonstrating a pristine surface without residue (Fig. 4c).
STEM is more effective than TEM imaging to qualitatively analyze residue coverage The HAADF-STEM image demonstrated the clear contrast between the mechanically cleaned and as-prepared regions as shown in Fig. 4d. As expected, the mechanically cleaned region (bottom half, left) had darker contrast than the uncleaned region (top half, right). We plotted a histogram of the pixel intensity values and compared the two regions (dashed box in e and f). The cleaned region had a broad distribution and the maximum population was located at a mean pixel intensity of approximately 70. Based on the local pixel intensity, we identified two distinct contrast regions and deconvoluted the histogram as shown in Fig. 4e. The clean region (region 1) with a mean pixel intensity of 65 comprised approximately 61% of the sample area, and the relatively high-contrast region (region 2, with a mean pixel intensity of 101) comprised 39%. However, the as-prepared graphene region had a different proportion, and the clean region (region 1) with mean pixel intensity of 68 shared 28% of the sample area. This confirmed that the residue-free area more than doubled via mechanical cleaning.