Triboelectrification of Two-Dimensional Chemical Vapor Deposited WS2 at Nanoscale

Triboelectric properties of chemical vapor deposited WS2 nanoflakes have been characterized in nano-range by atomic force microscopy (AFM) and Kelvin force microscopy (KFM). The triboelectric process is dependent on the thickness of WS2 nanoflakes, and it is sensitive to the adsorbates like water molecules, as well as transferred Pt from the tip on the sample. The density of tribo-charge can be modified by applying various biases to the conductive Pt-coated tip during the frictional process. Tunneling of the tribo-charge into the gap between WS2 and the underlying substrate results in a long lifetime, which is about 100 times longer than conventional triboelectric charges. Moreover, we observe a positive correlation between the layer number and resistance to charge dissipation. Our finding can become the driving force for a new category of two-dimensional (2D) WS2 triboelectrically controllable nanodevices.


Triboelectrification of Two-Dimensional Chemical Vapor
Deposited WS 2 at Nanoscale He Wang 1 , Chung-Che Huang 2 & Tomas polcar 1,3 Triboelectric properties of chemical vapor deposited WS 2 nanoflakes have been characterized in nanorange by atomic force microscopy (AFM) and Kelvin force microscopy (KFM). The triboelectric process is dependent on the thickness of WS 2 nanoflakes, and it is sensitive to the adsorbates like water molecules, as well as transferred Pt from the tip on the sample. The density of tribo-charge can be modified by applying various biases to the conductive Pt-coated tip during the frictional process. Tunneling of the tribo-charge into the gap between WS 2 and the underlying substrate results in a long lifetime, which is about 100 times longer than conventional triboelectric charges. Moreover, we observe a positive correlation between the layer number and resistance to charge dissipation. Our finding can become the driving force for a new category of two-dimensional (2D) WS 2 triboelectrically controllable nanodevices. WS 2 , a typical member of transition metal dichalcogenides (TMDs) group, possesses a similar structure to widely studied MoS 2 : W and S atoms are arranged to a sandwich structure by covalent bonds in the sequence of S-W-S for a single layer, whereas the bonding of neighboring sheets is relatively weak van der Waals interaction.
Investigation on WS 2 just began in recent years owing to the lack of crystals in nature, but it is considered as one of the most promising TMDs, and potential applications in various fields such as catalytic 1 , lithium ion batteries 2 , solar cell active materials 3 have been reported. It is expected to own the highest electron mobility among the semiconductive TMDs thanks to a reduced effective mass -mobility up to 234 cm 2 V −1 s −1 in multilayered WS 2 contacted with gold at room temperature has been reported 4 . Besides, the valence band splitting of single-layer WS 2 is nearly three times larger than that of MoS 2 , which can facilitate the observation of valley Hall effect 5 . Also, optoelectronic devices with a relatively high quantum efficiency can be realized since the photoluminescence emission of WS 2 is ~50 times higher than MoS 2 6 . Furthermore, atomically-thin WS 2 is known as n-type semiconductor (because of the numerous S vacancies) with a direct band-gap 7-9 and field effect transistors based on vertical graphene/WS 2 heterostructures exhibit a quite large on-state current and an unprecedentedly high on/ off ratio of more than 10 6 due to the combination of tunneling and thermionic transport 10 . In the meantime, WS 2 -based lubricant shows better thermal stability and oxidation resistance when compared to other TMDs 11-14 . Triboelectrification, during which charges are transferred and accumulated on materials via frictional process, can occur during sliding, leading to a perturbation of the charge neutrality of WS 2 [15][16][17][18] . In this case, a charge may move through the material freely or be localized in specific atomic sites; thereby attractive or repulsive Columbic forces can arise, which can alter the frictional properties 19 . For low dimensional WS 2 structures, the effect of tribo-charges could be either positive (accumulated charge could be transferred in the form of useful energy) or negative (an increase in friction, i.e. more energy dissipated during sliding). Therefore, it is imperative to conduct an investigation into the triboelectric properties of WS 2 , particularly in view of the poor understanding and extremely limited publications regarding this topic 20,21 .
In this study, high-quality WS 2 nanofilms were grown by chemical vapor deposition (CVD) approach. Benefitting from the capability of accurately AFM-controlled triboelectric operation, the effect of bias voltages and the diffusion procedure were observed with AFM and KFM, which follows a similar path to previous publications 20,21 . However, the effects of layer number, moisture and applied rubbing force were also investigated; the correlation of material and charge transfer was reported for semiconductors like WS 2 as well, since relevant results are only provided for Generation and measurement of tribo-charges. Conductive probe (OMCL-AC240TM-R3 from Olympus, Pt-coated) was used for AFM and KFM. Firstly, contact-mode AFM was conducted at a 1 Hz scan rate to initiate tribo-charges on WS 2 nanofilms, and various biases (−10~10 V) and normal forces (25~100 nN) were applied to the tip in some cases. KFM was subsequently performed with the tip biased by an AC voltage (amplitude: 0.8 V; frequency: 10 kHz) to obtain the surface potential maps.
To investigate the effect of surface treatment, WS 2 nanoflakes were annealed at 100 °C for 10 minutes in some experiments. Once the heat treatment was completed, KFM measurements were conducted immediately to eliminate the environmental influence on the surface.

Results and Discussion
The WS 2 nanoflakes deposited on SiO 2 /Si substrate by CVD method were observed by optical microscopy, and their lateral size can reach up to ~100 µm as shown in Figs 1a and S2. Besides, the number of layers can be generally estimated via the variation in the color of nanoflakes on the substrate: while WS 2 monolayer is almost transparent and shows a dark cyan color, a multilayer is brighter and a yellow color can be seen for much thicker flakes. To specify the exact layer number of different areas, surface topographies have been mapped by tapping-mode AFM, and the thickness is ~0.75 nm (Fig. 1b), which agrees with the reported thickness of monolayer WS 2 27-29 . Apart from it, the surface topographic image of few layers and multilayers in Fig. S3 shows that their thicknesses are ~4.2 and ~7.1 nm corresponding to 5 layers and 9 layers, respectively. Besides, Raman spectrum was measured with 532 nm laser excitation for further characterization. As displayed in Fig. 1c, the 352.4 cm −1 2LA peak and 419.6 cm −1 A 1g peak are obvious for monolayer, while a slight red shift of 2LA and a blue shift of A 1g mode were observed with the increase of layer number 30,31 . In addition, photoluminescence (PL) spectrum in Fig. 1d shows an obvious shift of emission peak from 620.2 to 638.5 nm, indicating the layer number of WS 2 nanoflakes.
According to the XPS spectra of monolayer WS 2 nanoflakes in Fig. 1e,f, the W 4f peaks at 34.7 and 32.5 eV are assigned to the 4f 5/2 and 4f 7/2 orbitals of WS 2 , respectively 32,33 ; the W 5p core level at 37.6 eV suggests the existence of W 6+ , likely residual WO 3 on the as-deposited sample 34 . 163.5 and 162.3 eV peaks belong to the S doublets: 2p 1/2 and 2p 3/2 32 . All the binding energies are indicative of WS 2 crystal, and the atomic ratio of S and W is ~2:1. Apart from it, Fig. S4 displays the C 1 s orbital XPS spectrum: the 284.7 eV peak is the C-C bond, whose existence results from the use of carbon tapes to mount the sample on the sample-plate (The XPS spectrum was measured before the performance of AFM, so the C-C bond is not a result of AFM measurement); the C-S bond (in the range of 285-287.5 eV) cannot be detected, so the chemisorption of C is not significant 35 . Despite the fact that physisorption might happen, AFM working in contact mode could eliminate adsorbed molecules to the margins of rubbed regions 36 , so the surface topographies in Fig. S7, where there is no obvious materials accumulation on the margins of the rubbed area, further confirms the very limited contamination of the sample surface. Similar results were obtained for a few layers as well as multilayers, as displayed in Figs S5 and S6.
As demonstrated in Fig. 2a, AFM was performed in contact-mode to rub a 1 × 1 µm 2 square region of the WS 2 nanoflake by a conductive Pt-coated tip with 25-nN force applied normally, KFM mode was conducted to monitor the surface potential of a larger region (5 × 5 µm 2 ), concentered with the rubbed region. From the surface potential maps (Fig. 2b), the single-layer WS 2 is almost equipotential before triboelectrification, while the random fluctuations probably result from the adsorbed charges in the environment. After triboelectric process, the topographic variation remains unobvious (Fig. S7) but the surface potential of the rubbed region is ~0.5 V higher than that of the other region.
During the surface potential measurements, an AC voltage and a DC bias are applied between the tip and sample surface. The electrostatic forces acting on the conductive tip can be detected by a lock-in technique to monitor its oscillating amplitude within the system. The amplitude signal comprises the DC bias as well as the work function difference between the tip and sample, and the contact potential difference V CPD is cancelled by the DC bias via a feedback circuit 37 , so their relationship can be described as Eq. 1:

CPD t ip sample
where ϕ tip and ϕ sample represent the work functions of the tip and sample, respectively; e is the electronic charge. As a result, the contact potential difference between the rubbed and surrounding intact areas,ΔV CPD , can be computed as Eq. www.nature.com/scientificreports www.nature.com/scientificreports/ where ϕ rubbed and ϕ unrubbed represent the work functions of the rubbed and unrubbed areas of the sample surface, respectively. As the Fermi levels of two materials are aligned when brought into contact, their work functions (the minimum energy needed to remove an electron from Fermi level to the vacuum immediately outside the solid surface) would be changed correspondingly. In this case, some electrons were transferred from the sample surface to the AFM tip via triboelectrification, leading to the enhancement of surface potential in the central region. With the increase of layer number, the surface potential becomes smaller, which can be explained by an interlayer www.nature.com/scientificreports www.nature.com/scientificreports/ screening effect, since the surface potential of a nanofilm is screened by the external electric field induced by the defects in the SiO 2 /Si substrate 38,39 .
It has been reported that the surface potential and the charge distribution of 2D materials, such as graphene and h-BN nanoflakes 40,41 , are extremely sensitive to the ambient atmosphere. For WS 2 , the water molecules adsorbed from the atmosphere may act as carrier trappers. It is suggested that defects on the SiO 2 /Si substrate mainly involves surface-adsorbed water molecules, so efficient heat treatment can reduce the presence of undesired adsorbents, thereby the effect of the external electric field from the substrate. To eliminate the effect of water, 10 minutes annealing at 100 °C was further performed on the rubbed WS 2 nanoflakes, followed by immediate measurement of surface potential by KFM. As illustrated in Fig. 2c, while the surface potential difference still shows a similar descending tendency with the increase of layer number, the thermal treatment caused a decrease of ~0.2 mV. The reason behind is that the work function of unrubbed area (ϕ unrubbed ) is lowered due to the removal of moisture via annealing, but the water molecules were already removed during the triboelectric process for rubbed central area, so the annealing has less influence on the work function of this area (ϕ rubbed ). Based on Eq. 2, the difference of the contact potential between these two regions (ΔV CPD ) is reduced. Figure 3a shows the surface potential maps after triboelectrification within 2 days. It can be noticed that the surface potential difference is still detectable even after 48 hours. Given that only insulating materials can store charges for that a long time, such observation cannot result from the charges on WS 2 . Here we introduce tunneling triboelectrification to define the tunneling of conventionally friction-induced charges (between WS 2 and conductive tip) through WS 2 nanosheets and their localization on the insulator (SiO 2 /Si substrate in this case) underneath 20,21 .
Here, it should be emphasized that the step height of monolayer WS 2 measured by AFM (Fig. 1b), consists of a monolayer WS 2 (a sandwich structure of S-W-S) and an air-gap at the interface. The lattice constant of a single-layer WS 2 is ~0.3 nm 42 , so the air-gap between WS 2 monolayer and the underlying SiO 2 /Si substrate can be calculated as 0.75-0.3 = 0.45 nm. The air-gap exists for WS 2 nanoflakes with different layer numbers as well.
After triboelectric process, there should be no detectable current flow in WS 2 once the equilibrium is reached, therefore, the discharge will only originate from leakage currents. As the diffusion of charges happens to both air and SiO 2 , two voltage magnitudes (V 1 ,V 2 ) and two time constants (τ 1 , τ 2 ) are developed. Figure 3b shows surface potential difference (ΔV) with time for monolayer, 5 layers and 9 layers, and the fitted function is depicted below: t t 1 2 1 2 All the parameters are listed in Table 1. It is obvious that apart from a term with a small magnitude (V 1 ) and a short time constant (τ 1 ), the other one has a larger initial amplitude (V 2 ) and time constant (τ 2 ). Since the air gap is considerably thinner than the SiO 2 substrate, most charges are offered by WS 2 nanoflakes and trapped at the interlayer, and a larger initial amplitude V 2 is presented. The excellent insulation of air and the shield of WS 2 nanofilms lead to a longer time constant τ 2 , which is ~200 times larger than the decaying time of tribo-charges on bare SiO 2 /Si substrate of the same thickness (~1 h) 43 . Considering the poor conductivity (higher work function) as well as better protection (more layers to shield the tunneling charges from the atmosphere) of multilayers, a larger time constant is observed as the thickness of nanoflakes increases.
Taking advantage of the versatile AFM systems, we can tune the density of tribo-charges on WS 2 nanoflakes via various biases onto the conductive tip during the sliding process. As evident from Fig. 4a, different surface potentials can be measured for the rubbed region, indicating that the amount of tribo-charges localized on the sample surface is controllable. where ΔV is the potential difference, ε 0 is the permittivity of vacuum,t Air and ε Air are the thickness and relative permittivity of the air gap, respectively. Figure 4b demonstrates that the charge density, which is proportional to surface potential, increases linearly with the bias voltage, confirming the controllable density of tribo-charges via varied biases.
AFM system is capable of scanning with varied normal forces and the as-grown WS 2 nanoflakes were rubbed with 25, 50, 75 and 100 nN to investigate the effect of load. As evident from the surface potential difference shown in Fig. 5a, there is a slight reduction in the surface potential differences for monolayer, few layer and multilayer with the increase of applied normal force. However, the work function should be identical for the measured nanoflakes with the same thickness, so it is suggested that some materials may be transferred during the rubbing process. Then, elemental composition of the rubbed areas was analyzed by EDX, and an energy peak at 2.06 keV, belonging to Pt, is detected in Fig. 5b.
The chemical composition of rubbed WS 2 monolayer with various applied normal forces is listed in Table 2. The atomic ratio of W and S is ~1:2, which is consistent with the XPS results in Fig. 1, and Pt content increased with applied force. Note that low values of chamical composition related to Pt, W and S are due to EDX interaction volume (max. depth approx. 1 μm for acceleration voltage 20 kV).
Obviously, there is a slight rise in Pt ratio with the increase of the applied forces. Considering the negative correlation between the surface potential difference and applied force, we suppose that it is Pt that has been transferred from the tip to the WS 2 nanoflakes, since the transfer of Pt (with a larger work function) can enhance the work function of WS 2 , thereby lowering the surface potential difference as depicted by Eq. 2. Based on this property, WS 2 can be utilized as an interlayer for charge storage in triboelectric nanogenerators to promote the output performance, as reported for MoS 2 44 . Besides, WS 2 can be applied to triboelectrically controllable devices and systems, such as transistors, photonics, detectors and antennas to name a few.

Conclusions
In summary, nanoscale triboelectrification of chemical vapor deposited WS 2 nanoflakes has been investigated with the help of AFM as well as KFM. Charge transfer process and the effect of temperature were observed systematically. We show that the surface potential becomes smaller with the increase of thickness, which originates from the interlayer screening effect of external electric field induced by the defects in the underlying substrate,  and heat treatment can mitigate this effect efficiently by removing undesired adsorbents. The diffusion of triboelectric charges on WS 2 nanoflakes was then monitored, and it is found that these tribo-charges can be stored for an extremely long period of time, and the dwell time is positively correlated to the layer number. Furthermore, different amount of tribo-charges can be initiated by applying various biases onto the conducting tip during the triboelectric process. At last, the effect of applied normal force during the rubbing process was studied. Some Pt has been transferred from the tip to the WS 2 nanoflakes to tune its work function. These distinctive properties provide a guidance for the design of WS 2 -based triboelectrification-controlled nanodevices, such as nanogenerators, photonics, transistors or dividers.   Table 2. Atomic ratios of different elements on monolayer with various applied normal forces.