Controllable Tunneling Triboelectrification of Two-Dimensional Chemical Vapor Deposited MoS2

Tunneling triboelectrification of chemical vapor deposited monolayer MoS2 has been characterized at nanoscale with contact-mode atomic force microscopy (AFM) and Kelvin force microscopy (KFM). Although charges can be trapped on insulators like SiO2 by conventional triboelectrification, triboelectric charges tunneling through MoS2 and localized at the underlying substrate exhibit more than two orders of magnitude longer lifetime. Their polarity and density can be modified by triboelectric process with various bias voltages applied to Pt-coated AFM tips, and the saturated density is almost 30 times higher than the reported result of SiO2. Thus, the controllable tunneling triboelectric properties of MoS2 on insulating substrates can provide guidance to build a new class of two-dimensional (2D) MoS2-based nanoelectronic devices.


Controllable Tunneling
Tunneling triboelectrification of chemical vapor deposited monolayer MoS 2 has been characterized at nanoscale with contact-mode atomic force microscopy (AFM) and Kelvin force microscopy (KFM). Although charges can be trapped on insulators like SiO 2 by conventional triboelectrification, triboelectric charges tunneling through MoS 2 and localized at the underlying substrate exhibit more than two orders of magnitude longer lifetime. Their polarity and density can be modified by triboelectric process with various bias voltages applied to Pt-coated AFM tips, and the saturated density is almost 30 times higher than the reported result of SiO 2 . Thus, the controllable tunneling triboelectric properties of MoS 2 on insulating substrates can provide guidance to build a new class of two-dimensional (2D) MoS 2based nanoelectronic devices.
Triboelectrification, a contact-induced electrification when a material becomes charged after brought into contact with another one through friction, is a common phenomenon in our daily life 1,2 . For example, charges formed on airplane by air friction when flying can interfere with the radio frequency communication or even cause the plane to be hit by lightning. Although triboelectrification has been known for centuries, it is only until recently that this phenomenon has been utilized as energy convertors and self-powered mechanical sensors [3][4][5][6][7] , particularly electronics like transistors, whose gates shapes, positions and on/off states can be triboelectrically modified on demand 8 .
To date, research about the triboelectrification of insulators 9-12 as well as cutting-edge zero-gap graphene 8,13 has been extensively reported, but for other non-zero gap semiconductors like transition metal dichalcogenides, which are good candidates for transistors, there are still a limited number of publications.
As a member of transition metal dichalcogenides family, MoS 2 shares many attributes of graphene but its non-zero bandgap offers opportunities unattainable for graphene, so it has been widely investigated in the past decade. According to its natural abundance and distinctive optical 14,15 , electrical 16,17 , and mechanical 18,19 properties, MoS 2 has become the driving force behind a series of applications, including optoelectronics 20,21 , sensors 22,23 , electronics 24,25 , and energy storage devices 26,27 . For instance, monolayer MoS 2 -based field effect transistors exhibits a remarkably high on/off ratio at room temperature. Gas sensors realized by few-layered MoS 2 show high sensitivity to the detection of nitrogen monoxide. In the meantime, chemically exfoliated nanosheets of MoS 2 have been utilized to fabricate batteries and capacitors. However, studies concerning the triboelectric properties of MoS 2 are poorly reported 28 . As 2D materials exhibit high charge density 29 , it is of great essence to investigate the triboelectric properties of single-layer MoS 2 , so that the results can lay the foundation for its applications in triboelectric nanodevices.
In this work, high-quality 2D MoS 2 films were synthesized by chemical vapor deposition (CVD) method. The concept of tunneling triboelectrification was introduced to define the tunneling of conventionally friction-induced charges (between AFM and MoS 2 ) through the 2D MoS 2 and their storage on the underlying insulating substrates. The diffusion, manipulation and saturation of tunneling triboelectric charges were investigated with the aid of AFM working in contact and KFM modes. Taking advantage of the capability of accurately controlled triboelectric process, surface charges can be patterned with various shapes and areas in nano-range. And importantly, their polarity and density can also be dynamically altered over time by applying various bias voltages to the conducting AFM tip during rubbing process. Given that the AFM system can be completely integrated on a single chip [30][31][32] , the realization of time-variant electronic systems and devices based on MoS 2 is within reach.

Materials and Methods
Synthesis of MoS 2 and transfer method. All the MoS 2 samples were grown on Si substrate coated with 300 nm SiO 2 . As displayed in Fig. S1, Supporting Information, the atmospheric pressure CVD reaction was carried out in a 3 cm diameter quartz tube, which was surrounded by a furnace and a heating tape. One end of the tube was linked to the gas inlet, through which gases such as nitrogen (N 2 ), hydrogen (H 2 ), argon (Ar), and hydrogen disulfide (H 2 S) can be introduced, the other end was connected to the exhaust extract component. Before the reaction, boats and substrates were cleaned with acetone, isopropanol and deionized water, and then blow-dried by the nitrogen gun. Subsequently, SiO 2 /Si substrates were placed on a boat containing 10 mg MoO 3 (Alfa Aescar 99.9995%), and the boat was then pushed into the tube center. After that, another boat containing 100 mg sulfur (Alfa Aescar 99.9995%) was loaded at the upstream, whose temperature was separately controlled by the heating tape instead of the furnace. Later on, 300 standard cubic centimeters per minute (sccm) Ar gas (BOC, 99.999% pure with additional purifications) was introduced to purge the tube for 10 minutes. Afterwards, Ar flow was decreased to 30 sccm, and the furnace temperature was programmed to heat substrates and vaporize MoO 3 to 550 °C at 25 °C/min ramping rate. After holding this temperature for 5 minutes, it was set to 700 °C at 10 °C/min ramping rate, and sulfur was heated to 200 °C so its vapor could be transported to the tube center, where CVD reaction occurred and MoS 2 was deposited. After 15 minutes deposition at 700 °C, the heating system was turned off to cool down naturally. The temperature profile of the whole process is plotted in Fig. S2. The CVD-grown MoS 2 on SiO 2 /Si was spin-coated with polystyrene (PS, Sigma-Aldrich) at the speed of 1000 rpm for 50 s, and then cured on a hot plate at 90 °C for 15 min. The coated sample was subsequently covered with thermal release tape (from Nitto) and floated in deionized water. Once the MoS 2 /PS/tape stack detached from the substrate, it was transferred to the target substrates (gold, sapphire and polyimide), and the tape was peeled off when dry. The transferred sample was then immersed in CHCl 3 solvent until the PS film was fully dissolved. Finally, solvent residues of the successfully transferred MoS 2 samples were rinsed off by deionized water.

Characterization of synthesized MoS 2 . Scanning electrons microscopy (SEM, JEOL JSM-6500F) and
AFM (Scanning Probe Microscopy 5500, Agilent Technologies) were used to investigate morphology of the as-deposited MoS 2 sample. Raman spectroscopy (InVia Raman Spectrometer, with 532 nm excitation laser) and X-ray photoelectron spectroscopy (XPS, Thermo Scientific Theta Probe XPS System MC03) were utilized to characterize the vibrational modes and elemental composition of as-fabricated nanomaterial.
Initiation and measurement of triboelectric charges. Contact-mode AFM was performed under a 25 nN applied force at a 1 Hz scan rate to generate triboelectric charges on MoS 2 with a Pt-coated conductive probe (OMCL-AC240TM-R3 from Olympus). Different bias voltages from −10 to 10 V were applied on the tip to induce a friction pattern on the sample in some experiments. The surface charge was then characterized by KFM mode at ambient environment. Figure 1a displays the SEM image of MoS 2 synthesized on SiO 2 /Si substrate by CVD method. It is noteworthy that the fabricated MoS 2 sample exhibits a triangular shape, and the lateral size can reach up to ~120 µm, whose limitation may result from the lattice mismatch between the MoS 2 and SiO 2 /Si substrate. As evident from the surface topography measured by tapping-mode AFM in Fig. 1b, the step height of monolayer MoS 2 from substrate is ~0.8 nm, which is consistent with the reported thickness of mechanically exfoliated MoS 2 monolayers [33][34][35][36][37][38] . Considering the 0.3 nm thickness of S-Mo-S sandwich structure of MoS 2 single-layer [39][40][41][42] , the air gap between the MoS 2 monolayer and SiO 2 substrate is estimated as 0.5 nm.

Results and Discussion
As can be seen in Fig. 1c, there are two characteristic peaks in the Raman spectrum: 383.2 cm −1 peak (E 2g mode) and 403.8 cm −1 peak (A 1g mode) 43,44 . The frequency difference between E 2g and A 1g modes is calculated to be 20.6 cm −1 in this study, corresponding to monolayer MoS 2 23,45,46 . Besides, the obvious emission peak at 676 nm in the photoluminescence (PL) spectrum (Fig. 1d) is also indicative of the single-layer nature of the as-fabricated MoS 2 47,48 . XPS was performed with Al Kα source to determine the chemical composition of the film. It can be noted from Fig. 1e that the Mo 3d peaks at 229.4 and 232.5 eV are the 3d 3/2 and 3d 5/2 orbitals of MoS 2 , respectively; the peak at 226.7 eV belongs to the S 2s orbital. Figure 1f shows the S 2p peaks at 162.1 and 163.3 eV corresponding to the 2p 1/2 and 2p 3/2 doublets of S 49 . All these binding energies are in agreement with the reported values for MoS 2 crystal 50,51 , and the atomic ratio of Mo and S is close to the stoichiometric 1:2. In addition, the XPS spectrum of C 1s orbital is provided in Fig. S3. The peak at 284.6 eV corresponds to the C-C bond, which is expected due to the utilization of carbon tapes for fixing the sample. However, there is no obvious C-S bond (in the 285-287.5 eV range) found in the spectrum 52 , indicating that there is no significant chemisorption of C. Although physisorption could happen, contact-mode AFM would remove adsorbed molecules at the side of tested areas or even wear mateirals 53 . In the meantime, the surface topographic images in Fig. S5 show no accumulation of materials at the side, so it is not expected to have significant contamination on the sample surface.
The surface potential before and after triboelectrification was monitored by combining contact-mode AFM and KFM, as illustrated in Fig. 2a carried out to image the tip-sample contact potential difference of a larger area (5 × 5 µm 2 ) with the rubbed area centered. As evident in Fig. S4, the MoS 2 surface is equipotential except for random fluctuations (probably induced by the absorbed charges from air or contamination) before triboelectrification. After rubbing, the topographic change is undetectable (Fig. S5) but the distinction in the surface potential image is obvious between the rubbed and unrubbed regions (Fig. 2b). The cross-section profile in Fig. 2c demonstrates that the surface potential of the rubbed section in the center is ~20 mV lower than that of the other region. Generally, contact potential difference is determined by two elements: electrostatic potential difference and effective work functions of two materials. The former one is dependent on the surface charge and applied bias, and the latter one is governed by the surface properties of materials 54 . Since the effective work function is almost identical across the sample surface, the drop of surface potential in the rubbed region is induced by the triboelectric charging. In this case, a fraction of negative charges were transferred to the sample surface via triboelectrification, which lowered the surface potential of the central region. Figure 3a,b show the surface potential maps and their cross-section profiles (corresponding to the stripy region confined by the two red lines in Fig. 3a, the surface potentials were averaged to reduce the influence of random fluctuations) after triboelectrification for up to 48 hours. It is noteworthy that the potential difference remains detectable even after two days. As charges can be kept for a long time only on insulating materials, such variation cannot originate from the charges on MoS 2 . Here we suggest that some triboelectric charges generated by triboelectrification tunnel through the single-layer MoS 2 and localize on the insulator underneath.
Additional experiments have also been conducted with other substrates, consisting of conductive gold as well as insulators like sapphire and polyimide (Figs S6-S8). Similar to the case of SiO 2 , triboelectric charges are stored at the air-insulator interface after tunneling through MoS 2 for sapphire and polyimide. However, there is no  detectable localized charges in the rubbed section with gold as the substrate. As the surface potential of the whole area is slightly lower after triboelectrification, it is suggested that the transferred charges spread across the whole MoS 2 or gold substrate due to the excellent conductivity of gold. The surface potential differences of the rubbed versus unrubbed regions on sapphire and polyimide decreased by ~10 mV after 12 hours, but in the case of SiO 2 substrate, ~5 mV decrease was observed (Fig. 3c), which indicates a better preservation of tunneling triboelectric charges. Therefore, we used MoS 2 monolayers on SiO 2 substrate for further investigation.
As charges cannot diffuse through insulators like air and SiO 2 , these tunneling triboelectric charges are localized on the air-SiO 2 interface, and attract opposite charges from MoS 2 and/or Si underneath SiO 2 , as the schematic diagram shown in Fig. 3d. Considering that the thickness of air (0.5 nm) is far smaller than that of the SiO 2 (300 nm), the capacitance of air is even larger, so almost all the opposite charges are donated by MoS 2 and accumulate across the air gap, while the stored charges on SiO 2 layer are almost unvaried. Consequently, the measured surface potential difference across MoS 2 is almost identical to the voltage change through the air gap.
After triboelectrification, no current flow will be detected in MoS 2 once the equilibrium is reached, and leakage currents will be the only reason for the discharge of both air and SiO 2 , so two time constants will be developed. Figure 3c shows the relationship between surface potential difference (ΔV) with time and its fitted function is depicted as follows: As evident, apart from a small term with a shorter time constant (τ short ~ 2.17 h), the other term owns a larger initial amplitude and a longer time constant (τ long ~ 207.8 h), which is significantly longer than the ~1 h decay time of triboelectric charges on SiO 2 with identical thickness of 300 nm 55 . The dominant term is attributed to the small-area air gap: the initial amplitude is greater since most of the charges are attracted from MoS 2 and trapped across the air gap; the longer time constant originates from the great insulating nature of air and the protection of MoS 2 film.
The polarity and density of tunneling triboelectric charges on MoS 2 can be controlled by applying different bias voltages to the conducting Pt-coated AFM tip during triboelectric process. As can be seen from the surface potential images after rubbing with biases ranging from −10 to +10 V in Fig. 4a, positive charges are trapped at the interface and electrons are attracted in the central region of MoS 2 at positive bias voltages, resulting in an n-type MoS 2 ; in contrast, negative charges are injected across the air gap and holes are attracted in the rubbed section with negative biases, so a p-type MoS 2 is formed. As surface potential difference between the intact and rubbed sections is almost identical to the voltage drop through the air gap, charge density σ of tunneling triboelectrification can be calculated based on capacitor model as:  As evident from the surface potential differences and corresponding charge densities in Fig. 4b, the surface potential difference increases with the applied voltage non-linearly, which results from the different energy state densities within the bandgap. Now we explore the ability of AFM system to provide multiple scans of selected area in order to increase the accumulated charge. The as-deposited MoS 2 sample was rubbed for multiple cycles with a constant contact force of 25 nN and bias voltage of +10 V; corresponding surface potential images are shown in Fig. 5a. As the whole measurement was completed within 30 minutes, the influence of charge diffusion can be neglected considering the good preservation of tunneling triboelectric charges. Figure 5b displays the averaged surface potential differences and relevant charge densities within these 12 cycles, and an obvious accumulation and saturation trend for the triboelectric charge can be seen.
In previous reports 56,57 , the mechanism of triboelectric process can be described by the assumption of effective work function: the amount of transferred charges in each cycle is linked with the difference of effective work functions between the tip and sample. The difference will reduce with the charge accumulation process until a saturation is reached. A phenomenological model can be used to fit the experimental data as illustrated in equation 3: where σ and n are the surface charge density and the number of rubbing cycles, respectively, k is the charge efficiency coefficient, V c is representative of the work function difference between the tip and sample, p is the charge impedance coefficient, and V e is the charge-induced potential on measured surface. Considering the boundary conditions for equation 3, the following relationship can be obtained: where σ 0 and σ ∞ are the surface charge densities at n = 0 and n = ∞, respectively. n 0 is a constant indicating the speed of charge saturation. By fitting the experimental data for different rubbing cycles (Fig. 5c), all the parameters can be extracted: σ 0 = (8.01 ± 0.01) µC/m 2 , σ ∞ = (4.46 ± 0.01) mC/m 2 , n 0 = (5.14 ± 0.01). Compared with the recently published results, the saturated tunneling triboelectric charge density of MoS 2 (~4.46 mC/m 2 ) is significantly higher than that of SiO 2 (~150 µC/m 2 ) 55 , and it can lay the foundation for triboelectric applications 28 . As for the enhancement of charge density, it mainly results from the tunneling of triboelectric charges, which are kept at the interlayer between the MoS 2 monolayer and the underlying SiO 2 /Si substrate. Under this circumstance, the MoS 2 layer as well as the thin air gap protect these charges from being neutralized by charges or ions in the air, thereby improving the charge density compared with the case of SiO 2 .

Conclusions
To conclude, nanoscale tunneling triboelectrification of chemical vapor deposited MoS 2 film was demonstrated by linking contact-mode AFM with KFM techniques at ambient environment. The charge transfer and the following diffusion process, as well as the accumulation from multi-friction effect were observed systematically. We show that the tunneling triboelectric charges can be preserved for an impressively long time, which is 100 times longer than that of the charges induced by conventional triboelectrification. Moreover, both positive and negative tunneling triboelectric charges can be induced without changing the surface topography by applying different bias voltages to the conductive tips via frictional process, and their density is controllable. Besides, a saturation of localized charges can be reached through multi-friction. These unique properties can pave the way for the development of novel 2D MoS 2 -based nanodevices, especially the triboelectrically controlled transistors, which can reduce the energy consumption to a large extent compared with the currently used voltage-controlled one.