In-situ electrical conductance measurement of suspended ultra-narrow graphene nanoribbons observed via transmission electron microscopy

As the most well-known two-dimensional (2D) material, graphene has received significant interest due to its unique characteristics, including a remarkably high carrier mobility [1, 2], strong mechanical strength [3, 4], and excellent thermal [5–7] and electric conductivities [8–12]. In particular, graphene nanoribbons (GNRs) are expected to be useful in the fabrication of nanoelectrical devices, as their edge structure can control the electrical transport [13–17]. In the case of sub-10 nm-wide GNRs, the edge structure, which may be zigzag or armchair, plays a critical role in tuning the electrical properties, such as the band gap. Theoretically, the origin of the band gap has been reported to differ between GNRs with armchair edges (AGNRs) and those with zigzag edges (ZGNRs). The band gap of AGNRs is thought to be caused by quantum confinement, while that of ZGNRs is attributed to spin polarization around the zigzag edges [18]. Experimentally, the edge structure dependence of the band gap has been obtained by scanning tunneling microscopy (STM) [19]. This STM study suggested that the magnetic coupling between two edges changes from antiferromagnetic to a ferromagnetic configuration as the ZGNR width increases, which is different from the previous theoretical report. Recently, it has been reported that GNRs fabricated on a substrate may show significant suppression of predicted electronic properties due to substrate dielectric screening [20]. Thus, the electron transport properties of GNRs may change due to substrate dielectric screening.

To avoid the influence of the substrate, it is necessary to measure the electrical transport properties of free-standing GNRs while simultaneously observing the edge structure. To resolve this issue, in-situ transmission electron microscopy (TEM) may be a suitable approach, as TEM can enable observations and nanoscale modifications of the GNR structure with simultaneous measurements of electrical conductance. Drndic's group [21] achieved in-situ TEM observation of GNRs experimentally and found that the conductances of both monolayer and bilayer GNRs are nearly proportional to their width in the range of 1–1000 nm. Wang et al [22] reported the current–voltage (I–V) characteristics of a thin (1.6 nm-wide) GNR suspended between two electrodes. However, in these previous in-situ TEM observations, the correlation between the edge structure and electrical conductance of GNRs was not clearly obtained.

In this paper, we establish a method for fabricating a suspended GNR device and develop an in-situ TEM holder for observing the ribbon structure during electrical measurements. By combining our developed equipment and in-situ TEM technique, we can controllably sculpt the width of a suspended GNR and measure I–V curves for ultra-narrow GNRs with either zigzag or armchair edges.

2.1. Experimental setup

Figure 1(a) presents a photograph of the head of our develo-ped in-situ TEM holder. The details of design can be obtained from the supporting information SI-1 (available online at https://stacks.iop.org/NANO/32/025710/mmedia. It enables us to load a silicon (Si) chip onto the sample stage for three different GNR devices (figure 1(b)). We can establish electrical contact without breaking the Si chip by controllably connecting three tungsten probes, which are fixed on a poly-ether-ether-ketone board mounted on a spring, with the three electrodes of the GNR devices. Si chips with dimensions of approximately 2.6 × 2.6 mm were fabricated on a 200 μm-thick Si wafer covered with a 50 nm-thick Si nitride (SiN) film on both sides; these chips were produced by SiMPore Incorporated. As shown in the inset of figure 1(b), the Si chip was customized to have five square windows of 60 × 60 μm and three rectangular windows of 500 × 60 μm; these windows were transparent to electron beams.

Zoom In Zoom Out Reset image size

Figure 1(b) displays a scanning electron microscopy (SEM) image of the Si chip after the fabrication of chromium/gold (Cr/Au) electrodes. As shown by the brighter regions in the SEM image, three Cr/Au electrodes were fabricated as individual source electrodes for different GNRs, and a large electrode located at the opposite side of the square window was fabricated as the common drain electrode. The thinnest part of the electrode at the square window was 1.5 μm wide and 20 μm long, as shown in the enlarged SEM image (lower left corner of figure 1(b)). The spatial gap for suspending the GNR was created by cutting the electrode using a focused ion beam (FIB) [23, 24] at the position near the center of the thinnest electrode. Figure 1(c) shows an optical microscopy image of the Si chip after the fabrication of three GNRs. Bare SiN film is shown by the brown regions on the Si chip; thus, monolayer graphene remained on the chip, except in the brown regions. The color results confirmed that the graphene was separated among the three electrodes. As shown by the enlarged images of the square window (bottom side of figure 1(c)), a narrow ribbon was suspended over the gap between the source and drain electrodes. The size of this suspended ribbon was measured at approximately 300 nm wide and 3 μm long.

2.2. Fabrication of suspended GNR device

Figure 2 illustrates the fabrication process for the suspended GNR device on the Si chip, which primarily consists of three steps. In the first step, the source–drain electrodes were fabricated for electrical conductance measurements (figures 2(a)–(e)). After the surface of the Si/SiN chip was cleaned with acetone solution, thin methyl methacrylate (MMA, 2000 rpm) and poly methyl methacrylate (PMMA, 4000 rpm) resist layers were spin-coated onto the surface of the chip and patterned by conventional electron beam lithography (EBL, ELS-7500). The samples were developed in methylisobutylketone/isopropyl alcohol to dissolve the exposed PMMA; next, Cr and Au were deposited on the chip at thicknesses of 5 and 40 nm, respectively, by electron beam evaporation. Subsequently, N-methyl-2-pyrrolidone was used to lift off the MMA/PMMA layer while leaving the electrodes. In the second step, a spatial gap was cut for suspending the GNR (figure 2(f)). A gap with a width of 100–300 nm and a length of 2.5 μm was cut at the center of the narrowest electrode by a Ga FIB (SMI-3050), as shown in the lower right corner of figure 1(b). When suspended in the gap, the GNR can be directly observed by TEM and can also be modified in shape and/or size by an intense electron beam.

Zoom In Zoom Out Reset image size

In the final step, the graphene was transferred [25] and patterned on the chip (figures 2(g)–(i)). Large-area graphene grown by chemical vapor deposition (CVD) on a copper foil was cut to an area of 1 × 1 cm; then, a thin PMMA resist layer was spin-coated onto the surface at 4000 rpm and baked on hotplate at 180 °C for 5 min. Subsequently, the PMMA-coated graphene with the copper foil was floated facing down on a 5% ammonium persulfate solution for 8 h to remove the copper foil, and the remaining PMMA-coated graphene was cleaned by deionized water. Then, the graphene was directly transferred onto the Si chip with the electrodes, and the sample was dried for 24 h. The ribbon shape of graphene was patterned by EBL, followed by O₂ plasma etching (RIE-10NR, 25 s at 30 W) to remove the exposed graphene. Finally, the sample was soaked in acetone overnight to completely remove the PMMA.

2.3. In-situ TEM observation and measurement

The layer number and quality of the fabricated GNRs were evaluated using atomic-resolved TEM images. Figures 3(a)–(c) show three suspended ribbons with different widths. The widths were measured as 102, 395, and 798 nm, while the designed widths were 100, 400, and 800 nm, respectively. Thus, the width of the suspended ribbons was well controlled by the exposure conditions. I–V curves were measured for these GNRs, as shown in figure 3(d), and all samples exhibited metallic properties, with a linear I–V relationship. The different slopes of the I–V curves indicate different resistances for the GNR devices, depending on the ribbon width.

Zoom In Zoom Out Reset image size

Prior to the modification of the GNR by nanosculpting, a current annealing procedure is needed to remove contaminants associated with lithographic processing from the surface of the GNR (figure S3) [1, 26–28]. The cleaning procedure of GNR is inevitable in order to investigate the edge structure dependence of its conductance behavior. For this purpose, we first carefully reduced the contact resistance before the GNR was cleaned. We sometimes observed that the ribbon was broken and that the electrode had simultaneously melted at the GNR contact area. We believe that the temperature in the contact area may be sufficiently high to melt the electrode when the current increases during contaminant removal. In this case, the contact area must remain at a higher resistance and temperature than the other areas. Thus, we must be aware that the area with the highest resistance tends to reach high temperatures under constant-voltage conditions.

Figures 4(a)–(c) show typical time evolution curves of the resistance, current, and voltage during the current annealing process for a 395 nm-wide GNR, respectively. The voltage was increased to 1 V, remained constant for a few minutes, and then gradually increased to 1.2 V in 10 mV steps while the ribbon shape was observed by TEM. The resistance (current) decreased (increased) from 66 to 30 kΩ (16 to 39 μA) in approximately 7 min, which is thought to be related to the reduction in contact resistance.

Zoom In Zoom Out Reset image size

After the reduction in contact resistance, the current was sufficiently stable for annealing of the graphene. The resistance remained nearly constant for approximately 12 min at a voltage of 1.2 V and then decreased from 28 to 2 kΩ in 4 min. TEM observation confirmed that the ribbon was clean, as shown in the inset TEM image in figure 4(a). Finally, the bias voltage returned to 1.0 V. When the contaminants were removed by evaporation, the resistance decreased from 30 to 2 kΩ. In this stage, the contact area was not melted due to the low resistance, although the current dramatically increased, as shown in figure 4(b).

By applying current annealing, a clean GNR could be obtained. Almost all contaminants were removed, and a monolayer ribbon remained. A clean GNR is necessary for evaluating the electrical transport properties of GNRs because contaminants have a strong influence on electrical transport in GNRs. In addition, it is easier to control the narrowing process when using a clean ribbon because clean ribbons have a much higher resistance to electron beam irradiation.

After the cleaning step, narrower GNRs were controllably fabricated by a convergent electron beam using a process known as nanosculpting. First, the electron beam was converged to a spot for sculpting the ribbon in the TEM mode at 80 kV. This voltage, 80 kV, is slightly below the threshold for knock-off damage of carbon atoms in graphene (86 keV) [29], which is expected to suppress electron-beam-induced damage in graphene lattices. During the nanosculpting process, the bias voltage was maintained at 1.0 V to heat the ribbon, in order to avoid unnecessary damage due to electron beam irradiation and/or reattachment of contaminants onto the ribbon [30–32]. As shown in figures 5(a)–(c), the ribbon width was reduced from 358 to 30 nm by gradually nanosculpting both sides of the GNR toward the inside. However, it was difficult to make the GNR thinner than 30 nm by nanosculpting due to breaking. Instead of nanosculpting, a further thinning process was achieved by current annealing (heating). The bias voltage was ramped to ∼2.5 V to further reduce the width from 30 nm until breaking occurred, as shown by the TEM images in figures 5(c) and (d).

Zoom In Zoom Out Reset image size

We successfully obtained ultra-narrow, 1.5 nm-wide GNRs with armchair and zigzag edge structures, as shown in figures 6(a) and (d), respectively. The width of narrowest GNR was measured from the cross section between two red arrows, the edge structures are indicated by red honeycomb structures. In each TEM image, the edge structure was identified by the orientation relationship between the direction along the edge and the direction of a spot in the fast Fourier transformation (FFT) pattern, as shown in the inset of figure 6. The FFT patterns shown in figures 6(a) and (d) were both obtained at the graphene region near the narrowest ribbon in the atomic-resolved TEM images. The edge structure is determined to be an armchair structure when the direction along the edge is parallel to the connection of two diffraction spots with central symmetry in the FFT pattern, as shown in the inset of figure 6(a). Meanwhile, the edge structure is identified as a zigzag structure when the direction along the edge is tilted by 30 degrees from the connection of two diffraction spots with central symmetry in the FFT pattern, as shown in the inset of figure 6(d). The I–V characteristics of both GNRs were measured by the source meter. The bias voltage was swept in order from 0.0 to 1.0 V, 1.0 to −1.0 V, and −1.0 to 0.0 V, corresponding to one cycle, and the current was measured in 50 mV steps. To obtain an accurate current value and reduce the influence of noise, the output current was programmed to average ten cycles.

Zoom In Zoom Out Reset image size

Figures 6(b) and (e) show I–V curves for a 1.5 nm-wide AGNR and ZGNR, respectively. In addition, figures 6(c) and (f) show their differential conductance–bias voltage (dI/dV–V) curves, respectively. The associated videos of AGNR and ZGNR which shows the change of the width from ca.1.5 nm to breaking is shown in the supporting information, the I–V curves were measured simultaneously during this process. Both ultra-narrow AGNR and ZGNR exhibit semiconducting behavior, suggesting opening of the energy gap. The energy gaps of the AGNR and ZGNR were estimated to be 300 and 700 meV, respectively, which were calculated as one half of the nonlinear region △V in the I–V curve [33–36]. The energy gap for the zigzag edge is more than two-fold larger than that of the armchair edge, even though these two GNRs have the same width of 1.5 nm. We also found that the dI/dV–V curves were obviously different between the AGNR and ZGNR. The dI/dV–V curve for the AGNR (figure 6(c)) shows a parabolic pattern near the origin (0 V), while the curve for the ZGNR (figure 6(f)) shows an abrupt increase above the critical bias voltage [37]. We believe that such abrupt increase in the differential conductance can be explained by a previously reported mechanism, namely, the magnetic-insulator and nonmagnetic-metal nonequilibrium phase transition predicted by Nikolić et al in 2009 [38]. As shown in figure 6(f), the device shows insulating properties for voltages from −0.7 to 0.7 V, where the antiferromagnetic ordering of spin states is maintained between the edges of the ZGNR. Above this bias voltage, the device changes to a metallic state with the complete collapse of spin polarization between the two edges. This potential for tuning the electrical properties by controlling the edge structures makes GNRs a promising material for future nanoelectronic devices.

In the present study, we obtanined a correlation between the edge structure and electrical transport properties of GNRs by fabricating GNR devices based on Si chips and developing an in-situ TEM holder. These findings indicate that the in-situ TEM technique is a powerful tool for elucidating the electrical properties of nanomaterials, particularly ultra-narrow 2D materials, which tend to be more sensitive to the atomic configuration, including the edge structure. In the near future, we hope that the physical and chemical properties of other 2D ribbons will be clarified by this in-situ TEM technique.

In this report, we proposed a method for fabricating high-quality monolayer GNRs suspended between two electrodes, with the aim of measuring the electrical conductance while simultaneously performing TEM observations. For this purpose, we fabricated GNR devices based on Si chips and developed an in-situ TEM holder for high-resolution TEM observations of the GNR structure during electrical conductance measurement.

In the process of fabricating narrow GNRs, the contact resistance must first be reduced, and then the contaminants adsorbed on the GNR surface was removed to obtain a clean GNR. We successfully obtained 1.5 nm-wide zigzag and armchair GNRs, which exhibited the opening of energy gaps. The zigzag GNR showed a larger energy gap and a sharp increase in the differential conductance curve, which can be explained by the magnetic-insulator and nonmagnetic-metal nonequilibrium phase transition. Our in-situ TEM technique is promising for elucidating the structural dependence of electrical conduction in 2D materials.

X Zhang would like to thank the financial supports from the Iketani Science and Technology Foundation (Grant No. 0291068-A), the Izumi Science and Technology Foundation (Grant No. H28-J-058) and the Sasakawa Scientific Research Grant of The Japan Science Society (Grant No. 28-226). C Liu wish to thank the financial support from China Scholarship Council under No. 201808050001. The authors gratefully acknowledge the use of the facilities in the Center for Nano Materials and Technology of the Japan Advanced Institute of Science and Technology. The authors thank Professor Sannomiya for using R005 in Tokyo Institute of Technology.