Floating gate memory (FGM)-based two-dimensional (2D) material heterostructures can achieve high performance and stability [1–9]. For instance, the transition-metal dichalcogenide (TMDC) semiconductor channel, hexagonal boron nitride (hBN) dielectric layer, and multilayer graphene (MLG) semimetal layer heterostructure have been demonstrated and widely used in FGM devices [1–9]. The program and erase steps in memory can be achieved by electron tunneling between the TMDC channel and MLG heterostructure through the hBN barrier. Hence, the tunneling condition of hBN in a TMDC/hBN/MLG heterostructure is important in FGM applications [1–9].

For preparing 2D material heterostructures, each layer of 2D material is stacked using micro-transfer systems in air, causing unintended residues in the stacked layer heterointerfaces [10–16]. Such residues are commonly referred to as bubbles. Previous studies on bubbles have considered their influence on tunneling [17–24], and we have previously reported the charge trap by bubble formation in a TMDC/hBN/ MLG heterointerface [9]. We observed the tunneling carrier density through the hBN area that includes bubble formation and analyzed an additional charge trapping layer for electron transfer between TMDC and MLG layers in TMDC/hBN/MLG heterostructures.

In the present study, we characterized the surface potential variation on the bubble area at an hBN/MLG heterointerface. Using a MoS₂/hBN/MLG heterostructure FGM device, we confirmed that the electron transfer ratio enhances the memory window (MW) for bubbles formed at the hBN/MLG heterointerface compared with a flat interface area. This characterization of electron tunneling behavior for bubbles in 2D material heterointerfaces may unveil properties of real electron tunneling in 2D material heterostructures for charge carrier tunneling devices with unintended residues.

Figure 1 shows the fabrication steps of the MoS₂/hBN/MLG heterostructure on a SiO₂ substrate using a micro-transfer system [25–28]. A 40-nm-thick hBN dielectric layer was stacked on MLG (22 nm in thickness), which was initially prepared on a 280-nm-thick SiO₂ substrate by mechanical exfoliation. Then, multilayer MoS₂ (26 nm in thickness) was transferred to hBN/MLG/SiO₂, as illustrated in Figs. 1(a)-(c). During transfer, we observed the formation of bubbles, as shown in Figs. 1(b) and 1(c). To verify the surface doping variation owing to bubble formation in MoS₂/hBN/MLG heterostructures, we performed scanning Kelvin probe microscopy (SKPM) using a commercial atomic force microscope (XE-100; Park Systems, Suwon, Korea) in air at room temperature. To investigate the spatial surface potentials on the surface of 2D material heterostructures, we used the feedback process (measuring contact potential difference V_CPD) under applied AC voltage of ∼1 V, DC voltage of 0.5 V, and frequency of 17 kHz to an Aucoated probe.

Figure 1. Optical images of MoS₂/hBN/MLG stacking. (a) Image of mechanically exfoliated MLG on SiO₂ substrate. (b) Image of hBN stacked on MLG/SiO₂ substrate. (c) Image of MoS₂ stacked on hBN/MLG/SiO₂ substrate (Scale bars, 10 µm).

Figure 2 shows the spatial distribution of the contact potential difference for the MoS₂/hBN/MLG surface including bubbles. Owing to the increasing influence of bubbles at the 2D heterostructure, we chose a large bubble with 60 nm in height [Figs. 2(a) and 2(b)] for analysis. To confirm bubble doping, we evaluated the normalized contact potential (△V_CPD) considering a flat hBN area as reference (i.e., △V_CPD = 0 meV), as shown in Fig. 2(c). △V_CPD ≈ 200 meV at the center of bubble region [Fig. 2(d)], indicating doping variation by the formation of bubbles in hBN/MLG heterostructures. This △V_CPD value can be attributed to molecules trapped in bubbles at the hBN/MLG interface. The tunneled carrier density has shown enhancement through hBN from MLG to MoS₂ layers in MoS₂/hBN/MLG heterostructures by residues such as water or hydrocarbon (i.e., bubbles) [9].

Figure 2. (a) Schematic diagram of SKPM applied to MoS₂/hBN/MLG heterostructure. (b) Optical image of MoS₂/hBN/MLG heterostructure (scale bar, 10 µm). Inset: enlarged image of a bubble on MoS₂ surface (scale bar, 2 µm). (c) Topography (left) and surface potential (right) images of bubble shown in inset of panel b (scale bars, 0.5 µm). (d) Height (black line) and △V_CPD (red line) profiles following dashed lines for topography and surface potential image shown in panel c.

To investigate electron tunneling related to bubble formation between hBN and MLG, we connected Cr/Au (10/100 nm in thickness) electrodes to an MoS₂ channel by electron-beam lithography, as shown in Fig. 3. We chose two regions covering bubble [A1 in Fig. 3(a)] and flat [A3 in Fig. 3(a)] areas on the hBN/MLG heterostructure and transferred a MoS₂ layer covering both areas. Employing the electrode marked by E2 as border, the electric transport characterization was investigated by measuring the drain-source current (I_DS) for V₁₂ = 0.1 V (red scheme between E1 and E2) and V₂₃ = 0.1 V (blue scheme between E2 and E3) according to V_BG, as shown in Fig. 3(b). The resulting MoS₂/hBN/MLG heterostructure was employed as a memory device with electron transfer between the MLG and MoS₂ layers, leading to hysteresis in the transport curves for the MoS₂ channel according to the forward (+) and backward (–) sweeps of V_BG. We considered the hysteresis as the MW [9,29].

Figure 3. (a) Optical image before (left) and after (right) transferring MoS₂ to hBN/MLG. Clean and bubble-containing areas at hBN/MLG interface enclosed by black dashed boxes with A3 and A1, respectively. (Scale bars, 20 µm) (b) Optical image of Au electrodes connected to MoS₂/hBN/MLG heterostructure FGM. Characterization conductance with V₁₂ (red line between E1 and E2) or V₂₃ (blue line between E2 and E3) according to V_BG. (Scale bar, 20 µm) (c) Current variation (I_DS) with bias voltage V₁₂ between E1 and E2 according to V_BG. (d) Current variation (I_DS) with bias voltage V₂₃ between E2 and E3 according to V_BG.

The hysteresis for transport curves (I_DS vs. V_BG for V₁₂) on the bubble-containing MoS₂ channel area [Fig. 3(d)] was larger than that (I_DS vs. V_BG for V₂₃) on the flat area [Fig. 3(c)]. Hysteresis in the transport curve for V₂₃ while applying forward and backward V_BG = ±10 - ±20 V was not observed (i.e., MW « 0.1 V) [Fig. 3(c)]. On the other hand, the opened MW for V_BG = ±30 V was observed with approximately 4 V. Hence, electron tunneling through hBN between the MoS₂ and MLG layers began for V_BG larger than ±30 V. For the bubblecontaining MoS₂ channel area [Fig. 3(d)], the MW was much smaller than 0.1 V for ±10 V, but the MW opened 6 V for ±20 V and 17 V for ±30 V. This indicates that tunneled electrons between MoS₂ and MLG are enhanced by bubble formation between hBN and MLG. Therefore, bubble formation on the surface interface in TMDC/hBN/MLG heterostructures can allow to change the direct surface potential variation and transport characteristics by trapped charges.

We characterized the effects of bubbles formed at hBN/MLG interfaces on the electron tunneling ratio of a MoS₂/hBN/MLG heterostructure. Evidence of charge trapping was obtained from direct observation by SKPM of the surface potential variation by doping changes on a bubble-containing area at the hBN/MLG heterointerface. Based on measurements of the MW enhancement for a MoS₂/hBN/MLG FGM device with bubbles formed at the hBN/MLG interface, we confirmed the trapped charges by bubble formation leading to improved electron tunneling ratio between the MoS₂ channel and MLG floating gate layer. This result may contribute to understanding the role of trapped molecules in bubbles formed on 2D material heterostructures for electron tunneling applications.

This work was supported by the research fund of Chungnam National University.

The authors declare no conflicts of interest.