Resolving surface potential variation in Ge/MoS₂ heterostructures with Kelvin probe force microscopy

Two-dimensional (2D) materials have been intensively studied and have shown great promise for several applications, such as in nanoelectronics,^1,2 sensors,^3,4 and optoelectronics,^5,6 due to their outstanding electrical, mechanical, and optical characteristics. To meet the requirements for 2D materials with respect to the technological applications, scalable growth techniques for producing wafer-scale 2D thin films are needed. Synthesis using chemical vapor deposition (CVD) can now produce large-area 2D materials.^7,8 Recently, 2D/three-dimensional (3D) heterostructures by growing 3D materials on 2D materials have been widely studied due to their remarkable carrier transport properties.^9,10 In particular, the characterization of Ge films on monolayer MoS₂ as a 2D/3D heterostructure, which have unique electrical characteristics due to charge transfer between the two layers, was reported in our previous work.¹¹ Despite the rapid progress of this field, the scalable growth of high-quality films has remained challenging due to the lack of understanding of defects^12–15 possibly introduced during the growth process. Considerable efforts have been devoted to probing and controlling the inhomogeneity of 2D thin films or defects at the interfaces of 2D/3D heterostructures in terms of defect engineering and interface engineering,^16–19 but the lack of characterization techniques inhibits further understanding of the origin of the defects/inhomogeneity.^20–22 

The techniques that are commonly used to investigate the film heterogeneities of 2D materials include transmission electron microscopy (TEM),²³ scanning tunneling microscopy (STM),²⁴ and photoluminescence (PL).^9,25 Typically, TEM and STM are nanoscale characterization techniques providing insights into the origin of the inhomogeneity. However, they have limited field of views, which are not suitable for a macroscale mapping of 2D/3D heterostructures to spatially unveil the heterogeneities and defects of the films.²⁶ Conventional PL does not provide the desired nanoscale resolution mapping because of insufficient spatial resolution. Therefore, nanoscale characterization techniques are desirable to unveil the nature of heterogeneities and defects limiting the electrical properties of 2D/3D heterostructures.

Here, we use an atomic force microscopy (AFM)-based technique, Kelvin probe force microscopy (KPFM), to characterize the heterogeneities of 2D/3D Ge/MoS₂ heterostructures with nanoscale resolution. MoS₂ and Ge are deposited on a SiO₂/Si substrate by CVD.^11,27 For the KPFM measurements, the 2D/3D heterostructures are contacted by an electrically conductive tip of a cantilever. Topography and contact potential difference (CPD) information are simultaneously obtained^28–30 in tapping mode AFM to interpret the spatial correlation between the height profiles and the surface potential signals of the 2D/3D Ge/MoS₂ heterostructures. In this work, we analyze an as-prepared sample and four different 2D/3D heterostructures with Ge CVD deposition time to demonstrate the impact of Ge CVD deposition time on the film quality of 2D/3D heterostructures.

A SiO₂/Si substrate was washed with acetone, IPA, and de-ionized (DI) water, respectively. After the washing process, the SiO₂/Si substrate was flushed with Ar gas three times. The CVD of monolayer MoS₂ was performed under an Ar atmosphere on the substrate. Sulfur and MoO₂ powders were employed as precursors of MoS₂. Monolayer MoS₂ was grown at 750 °C. The details of MoS₂ synthesis are described elsewhere.²⁷ For Ge deposition, low pressure CVD (LPCVD) was performed on a cold wall stainless steel reactor with germane (GeH₄) as the precursor of Ge.¹¹ Thermal cleaning at 500 °C under hydrogen flow was performed prior to flowing GeH₄ into the reactor. Table I describes an as-prepared sample and four different samples with Ge CVD deposition time.

Figure 1 shows a schematic diagram of KPFM and band diagrams of the tip and sample when (1) the sample and tip are separated, (2) the sample and tip are in electrical contact without applying bias voltage, and (3) the sample and tip are in electrical contact with applying bias voltage (V_CPD). An Asylum Cypher S AFM was used to image the local surface potential using a platinum iridium (PtIr) coated conductive tip. Topography images and their corresponding surface potential images are simultaneously obtained by moving a sharp tip at the end of a cantilever in tapping mode AFM across the sample. The cantilever is connected with the rear contact of the sample and grounded to avoid charging of the sample surface. In tapping mode AFM, the cantilever oscillates just above the sample surface and amplitude modulation KPFM feeds back on changes in the amplitude of the oscillating probe. At each point for KPFM measurements, an electric field between the sample and probe is detected, and it is varied by the voltage (V_CPD) applied to the sample relative to the probe. If an applied external bias DC voltage has the same magnitude as V_CPD, the surface charge in the contact area is nulled. This voltage is called the contact potential difference (CPD). The KPFM method measures the work function of the samples using the following equation:^31,32

where Φ is the work function and Δ V_CPD is the contact potential difference. For precise calculation of the sample work function, we used a pre-calculated SiO₂ film deposited on a Si substrate as a reference sample. Therefore, Eq. (1) can be recast using the following equation:

Figure 2 shows the topography and the corresponding KPFM images of an as-prepared MoS₂/SiO₂/Si sample, as described in Table I. The topography image shows a MoS₂ triangular piece on a SiO₂/Si substrate. Interestingly, grain boundaries (GBs) are observed in the topography image of MoS₂ [see yellow arrows in Fig. 2(a)]. Note that GBs can be normally observed in transition metal chalcogenides (TMDs), such as MoS₂, deposited by CVD,^27,33,34 and some types of GBs (e.g., tilt and mirror twin) on MoS₂ possess distinctive electrical properties compared to those of intra-grains.³³ For the as-prepared sample, GBs do not display any special CPD contrast as shown in the KPFM image [see Fig. 2(b)]. As shown in Fig. 2(c), the line–scan profiles extracted from the white dotted lines in the topography and KPFM images of the as-prepared sample clearly show the absence of correlation between the two images. The observation suggests that GBs in the as-prepared MoS₂/SiO₂/Si sample do not have local compositional or stoichiometric heterogeneities to show any CPD contrast compared to other MoS₂ areas.

As Ge is deposited on MoS₂, local variations of the CPD over the 2D/3D heterostructures are found. Figure 3 displays the topography and KPFM images of samples A and B with Ge deposition for 2 and 5 min, respectively. As shown in Fig. 3(a), GBs are also observed in the topography image of sample A the same as the as-prepared sample. There are some tiny particles over MoS₂. In particular, small particles deposited on the GBs are found. Note that sulfur vacancies randomly formed on MoS₂ in a hydrogen-rich atmosphere^11,35 can be nucleation centers for Ge nanoislands that have a honeycomb structure.^36,37 Ge nanoislands and MoS₂ are held together by weak Van der Waals force.³⁶ Despite the lattice mismatch between the Ge nanoislands and MoS₂, the Ge nanoislands are grown once high symmetry directions of the Ge nanoislands and MoS₂ are aligned.^36,38 In the early stage of the Ge particle formation (sample A), Ge particles are preferentially formed in GB regions due to their higher surface energy compared to intra-grain areas.³⁹ 

In the corresponding KPFM image [Fig. 3(b)], CPD contrasts along the GBs are obviously noticed.³⁹ We attribute the CPD contrasts on the GBs to the Ge particles on the GBs during the Ge growth. The line–scan profiles [Fig. 3(e)] extracted from the white dotted lines in the topography and KPFM images of sample A notice that the CPD on the GB is almost 77% higher than CPDs in intra-grain areas even though the height profile does not exist any pronounced peak on the GB. Sample B with 5 min of Ge growth has more Ge particles evenly distributed on MoS₂ than sample A [see Fig. 3(c)]. The Ge particles cover most of the MoS₂ surface, so the nucleation of Ge particles does not preferentially take place at the GBs anymore. The KPFM image in Fig. 3(d) also shows uniform CPD contrast over the MoS₂/Ge 2D/3D heterostructure. This implies that the CPDs of sample B are mostly affected by the Ge particles covering the entire 2D/3D heterostructure so that the GBs have little impact on the surface potentials. We note that sample C shows almost identical topography and KPFM images as those of sample B except the coverage of the Ge layer. For this reason, the topography and KPFM maps of sample C are not included in this work.

Interestingly, the topography and KPFM images of sample D with 35 min of Ge CVD deposition show very different features compared to those of other samples in terms of Ge particle geometry and spatial distribution of CPD. The height image in Fig. 4(a) shows Ge islands grown on MoS₂ with the size in the range of a few tens of nm to a few μm. We note that the Ge islands are most likely to be the agglomeration of Ge particles spreading over MoS₂, thereby showing GBs of the MoS₂ again. The high resolution of the red marked area in Fig. 4(a) clearly shows the features of Ge islands and GBs [see Fig. 4(c)]. Many features in the KPFM map [Fig. 4(b)] of sample D show some degree of correlation with grain structures. Grains are homogeneously dark or bright with the contrast edges markedly correlating with the GBs. The highlighted KPFM image shown in Fig. 4(d) clearly shows the correlation. The GBs marked in the topography image [Fig. 4(c)] are well matched with the CPD contrast edges (white dotted lines) in the KPFM map. Even though there exists some CPD contrast of Ge islands compared to MoS₂, the contrasts between grains are more significant. This is because the exposed MoS₂ is still covered by Ge.

Figure 4(e) displays the line profiles extracted from the yellow dotted lines in the topography [Fig. 4(a)] and KPFM [Fig. 4(b)] images. The fluctuation (maximum 170 nm) in the height profile is mainly due to the Ge islands, but any clues of grain structures are not observed. However, the CPD line profile shows strong grain-to-grain variations. The CPD varies from grain-to-grain by ≈25% in the voltage range of 0.85–1.05 V. The CPD of the grain marked between A and B in Fig. 4(b) is also noticeable in the line profile. We note that these grain-to-grain variations of the CPD in sample D are attributed to the strain induced in MoS₂ during the Ge growth. In the early stage of Ge deposition by CVD (samples A to C), sulfur vacancies play an important role in forming the Ge layer as discussed above. When the growth time is sufficiently long (35 min, sample D), sulfur vacancies randomly distributed in MoS₂ can also act as nuclei of nano-cracks⁴⁰ and migrate to form and propagate a sharp crack tip.⁴¹ The formation of Ge particles on this cracked MoS₂ layers can induce strain, which can lead to the work function heterogeneity of the MoS₂ layers.⁴⁰ 

We determined the work functions of the samples by calibrating the KPFM probe tip. A SiO₂/Si sample was used as a reference sample, which has a known work function of 5.0 eV.⁴² To ensure the tip condition, we performed KPFM for the reference sample prior to the KPFM measurement for each sample. Table II summarizes the CPDs of the as-prepared sample, four different 2D/3D Ge/MoS₂ heterostructure samples, and the SiO₂/Si reference sample. The work functions of the four different 2D/3D Ge/MoS₂ heterostructure samples were calculated using Eq. (2).²⁹ Sample A with 2 min of Ge deposition has the highest mean work function value of 5.858 ± 0.010 eV, which is similar to that of the as-prepared sample (5.887 ± 0.009 eV). This is due to the lack of time to deposit the Ge layer over MoS₂ in sample A so that the mean work function of the sample is almost identical to that of the as-prepared sample. However, a significant drop (≈0.569 eV) in the mean work function between samples A and B, which is 5.858 ± 0.010 and 5.289 ± 0.004 eV, respectively, is observed. This work function change is mainly caused by the charge transfer, which is commonly observed in heterostructures, across the interface between Ge and MoS₂ in sample B.¹¹ MoS₂ becomes more electron deficient, and Ge gets more electrons doped by the charge transfer, resulting in the reduction of the work function of the Ge layer.^43,44 However, due to the limited amount (monolayer) of MoS₂, the effect of electron doping of Ge on the reduction of the work function is limited as the thickness of the Ge layer increases. For this reason, the CPDs of samples C and D are rather gradually increased. To verify the grain-to-grain work function variation in sample D, the mean work functions of three different grains [G1 to G3 in Fig. 4(d)] were measured as shown in Table II. As discussed above, the variation of work functions is mainly caused by strain induced during the Ge deposition, and the fluctuation is about 0.231 eV, which is within the range of work function variations in a fractured MoS₂, which was reported as 0.4 eV.⁴⁰ 

In summary, we have analyzed the structural and functional changes of 2D/3D Ge/MoS₂ heterostructures during Ge CVD deposition that both drive spatially the inhomogeneous CPD over the samples. A nanoscale AFM-based measurement technique, KPFM, is applied for local surface potential measurements, which enables us to observe the heterogeneities of the 2D/3D heterostructures. The information obtained from the KPFM technique includes the evolution of the 2D/3D heterostructures with different Ge CVD deposition times. First, the CPD map in the as-prepared sample does not display any heterogeneity even though GBs are observed in the topography image. Second, as Ge is deposited on MoS₂, the CPD contrasts along the GBs are obviously noticed. We attribute the CPD contrasts on the GBs to the Ge particles deposited on GBs due to their higher surface energy compared to those of intra-grain areas. Third, with a sufficiently long Ge CVD deposition time, strong grain-to-grain CPD variations over the 2D/3D heterostructures are found. This is mainly due to the strain induced by the Ge particle formation on the cracked MoS₂ initiated by sulfur vacancies. The work function heterogeneity of the 2D/3D heterostructures can be an indicator of tuning 2D materials properties.

This work was supported by the 2020 Korea Aerospace University faculty research grant. H. P. Yoon ans D. J. Magginetti thank X. Cheng and E. Pourshaban for participating in part of KPFM analysis. We acknowledge the support of the MRSEC Program of the NSF under Award No. DMR-1121252. This was partly performed at the Center for Integrated Nanotechnologies, a U.S. Department of Energy, Office of Basic Energy Sciences User Facility, Los Alamos National Laboratory, and Sandia National Laboratories.

The authors have no conflicts to disclose.

The data that support the findings of this study are available from the corresponding author upon reasonable request.