Esaki Diode Behavior in Highly Uniform MoS2/Silicon Carbide Heterojunctions

The heterogeneous integration with 2D materials enables new functionalities and novel devices in state‐of‐the‐art bulk (3D) semiconductors. In this work, highly uniform MoS2 heterostructures with silicon carbide (4H‐SiC) are obtained by a facile synthesis method, highly compatible with semiconductor fab processing, i.e., the sulfurization of predeposited very‐thin (≈1.2 nm) Mo films at a temperature of 700 °C. Current–voltage characteristics of MoS2/n+‐4H‐SiC junctions collected by conductive atomic force microscopy show a pronounced negative differential resistance even at room temperature, which is a typical manifestation of band‐to‐band tunneling between degenerately p+‐/n+‐doped semiconductors. Here, the degenerate p+‐type doping of MoS2, with Nholes ≈ 4 × 1019 cm−3 evaluated by Raman mapping, is ascribed to the significant MoO3 content in the film, as demonstrated by X‐ray photoelectron spectroscopy analyses. Furthermore, atomic resolution transmission electron microscopy analyses reveal the presence of an ultrathin (≈1 nm) SiO2 tunneling barrier between MoS2 and 4H‐SiC, formed during the sulfurization process. The observation of Esaki diode behavior in MoS2 heterojunctions with 4H‐SiC opens new perspectives for this material system as a platform for ultrafast low‐power consumption digital applications.


Introduction
In the last years, the heterogeneous integration of 2D materials with bulk (3D) semiconductors [1] has been explored to enhance powders. [26] The direct growth of p + -doped MoS 2 multilayers on n + 4H-SiC substrates has been achieved by sulfurization of thin Mo/Nb/Mo stacks at high temperature (1000 °C), where Nb atoms had the role of acceptors for MoS 2 . [25] However, the current injection across these p + MoS 2 /n + SiC heterojunctions was found to be dominated by multistep recombination tunneling through midgap states in SiC, probably associated to interface defects. [25] The band-to-band tunneling (Esaki diode behavior), which typically manifests with negative differential resistance (NDR) in the heterojunction diodes formed by degenerately p +and n + -doped semiconductors, [30][31][32] has not been reported so far for p + MoS 2 /n + SiC heterojunctions, probably due to an insufficient quality of the interface.
As a matter of fact, achieving uniform and high-quality MoS 2 /4H-SiC heterojunctions on large area is a key requirement for advanced electronic applications exploiting vertical current injection at the interface.
In our work, highly uniform and controlled coverage of 4H-SiC with ultrathin MoS 2 films (predominantly monolayer (1L)) was achieved by a facile synthesis method, [33] highly compatible with semiconductor fab processes, i.e., sulfurization of predeposited very-thin (≈1.2 nm) Mo films at a temperature of 700 °C. Interestingly, current-voltage characteristics (at room temperature) of these MoS 2 heterojunctions with n + -doped 4H-SiC showed a pronounced NDR, indicating the occurrence of band-to-band-tunneling (BTBT) at the interface. Atomic resolution structural and chemical analyses by scanning transmission electron microscopy (STEM) revealed the presence of an ultrathin (≈1 nm) SiO 2 tunneling barrier between MoS 2 and 4H-SiC, formed during the sulfurization process. The degenerate p + -type doping of MoS 2 (holes concentration N h ≈ 4 × 10 19 cm −3 ) produced by the sulfurization approach, evaluated by Raman spectroscopy, was ascribed to the significant MoO 3 content in the film, demonstrated by X-ray photoelectron spectroscopy (XPS) analyses.

Results and Discussion
MoS 2 has been grown on the surface of highly n + -doped (≈10 19 cm −3 ) 4H-SiC(0001) 4°-off substrates and of n-doped (≈10 16 cm −3 ) 4H-SiC epilayers by sulfurization of predeposited ultrathin Mo films. The as-deposited Mo films (with ≈1.2 nm thickness evaluated by atomic force microscopy (AFM) step height measurements) were found to be completely oxidized, probably due to air exposure. As schematically illustrated in Figure 1, the sulfurization process was carried out in quartz tube with two-heating zones. The samples placed in the second zone (at a temperature T 2 = 700 °C) were exposed to sulfur vapors transported by the carrier gas (Ar) from the crucible in the first zone (at T 1 = 150 °C). Figure 2 shows the results of XPS analyses performed both on the as-deposited Mo on 4H-SiC (a,b) and after the sulfurization process (c,d). Mo 3d core level spectra in Figure 2a indicate that the as-deposited ultrathin Mo film was fully oxidized (being MoO 3 the main component, with a minor MoO 2 contribution). After the sulfurization process, the expected Mo3d, S2s, and S2p contributions associated to MoS 2 can be observed in Figure 2c,d, accompanied by minor contributions associated to unreacted MoO 3 . As it will be discussed later on in this paper, this MoO 3 component can play a crucial role in the electrical behavior of the MoS 2 /4H-SiC heterostructure.
The main mechanism ruling the formation of MoS 2 during the sulfurization process is the heterogeneous vapor-solid reaction between S and MoO x , while the loss of MoO x by evaporation plays a not negligible role at the temperature of 700 °C. [34] According to other recent reports, [33,34] the sulfurization of ≈1 nm MoO x is expected to result in the formation of 1L MoS 2 . No reaction between Mo or S and 4H-SiC occurs during the sulfurization process, while a slight oxidation of 4H-SiC is observed, as indicated by Si 2p core level spectra of the as-deposited and annealed samples (reported in Figure S1, Supporting Information). Figure 3a,b shows the surface morphology of the pristine 4H-SiC n + substrate and after MoS 2 formation. The conformal coverage of the SiC surface by a nanocrystalline film, with average grain size of ≈50 nm, can be deduced from Figure 3b. The morphology of the bare n -4H-SiC epitaxial layer exhibits the typical step bunching associated to the 4°-off miscut angle (Figure 3c), and the coverage by the nanocrystalline MoS 2 film can be argued after the Mo thin film sulfurization (Figure 3d).
The MoS 2 thickness uniformity on the n + and n − 4H-SiC was quantitatively evaluated by the Raman mapping. A typical Raman spectrum collected with a laser source of 532 nm on the MoS 2 /n + -SiC sample is reported in Figure 4a, where the main vibrational features of the 4H-SiC substrate (i.e., E 2 (TA), A 2 (LA), E 2 (TO), E 1 (TO), A 1 (LO)) and of 2H-MoS 2 (E 2g and A 1g ) have been highlighted. Figure 4b shows a closer view of the in-plane (E 2g ) and out-of-plane (A 1g ) vibrational modes of MoS 2 , with the indication of the peaks wavenumber difference Δω ≈ 20.1 cm −1 , which is consistent with monolayer (1L) MoS 2 thickness. [35] Figure 4c,d displays a color map and the histogram of the Δω values extracted from an array of 50 × 50 Raman spectra collected on 10 µm × 10 µm sample area. From this statistical analysis, an average Δω ≈ 20.8 cm −1 with a standard deviation of 0.8 cm −1 has been deduced, indicating that the film produced by sulfurization is predominantly formed by 1L-MoS 2 , with a small fraction of 2L or 3L areas. The same kind of analysis (reported in Figure 4e-h) was carried out on the MoS 2 /n − -SiC sample, showing a very similar distribution of MoS 2 number of layers (average Δω ≈ 20.8 cm −1 with a standard deviation of 0.5 cm −1 ), in spite of the differences in the surface morphology between the n + -SiC substrate and the epitaxy observed in Figure 3.
After assessing the morphological properties and the thickness uniformity of the as-grown MoS 2 films, the current transport across their heterojunctions with the n + 4H-SiC and n − 4H-SiC has been investigated using conductive atomic force microscopy (C-AFM) with Pt-coated tips, according to the configurations schematically illustrated in Figure 5a,d. As compared with conventional current voltage (I-V) measurements on vertical devices with deposited macroscopic electrodes, C-AFM analyses [36] provide spatially resolved information on the injected current with nanoscale resolution, corresponding to the effective contact area of the metal tip. Furthermore, this analysis is nondestructive and allows to probe the intrinsic electrical properties of the MoS 2 /4H-SiC heterojunctions, avoiding any effect of contaminations (lithography-induced resist residuals) and modification of the monolayer MoS 2 films by metal deposition during devices fabrication. These metal deposition effects can be particularly relevant, as discussed in recent reference papers. [37,38]    (NDR) under forward polarization, with some variability in the peak voltage (V P ) and peak current (I P ) at the different tip positions. The NDR behavior is typically ascribed to BTBT between degenerately n + -and p + -doped semiconductors. Hence, the observation of this phenomenon in the I-V characteristics of the MoS 2 /n + SiC system at room temperature indicates not only the formation of a very sharp 2D/3D heterojunction but also a degenerate p-type doping of 1L-MoS 2 obtained by the  sulfurization process. The comparison of the three characteristics in Figure 5b,c shows an interesting correlation between the V P and I P values under forward bias and the reverse current level, with lower V P and higher I P corresponding to a higher reverse current. As discussed in the following, such variability in the local NDR behavior can be ascribed to local changes in the MoS 2 p-type doping.
In the following, atomic resolution electron microscopy investigation of the MoS 2 /4H-SiC interface and a detailed analyses of MoS 2 Raman spectra will provide an insight on the origin of the observed Esaki-diode behavior. Figure 6a shows a low magnification cross-sectional scanning transmission electron microscopy (STEM) image in the high-angle annular dark field (HAADF) mode of the ultrathin MoS 2 film grown on the n + 4H-SiC substrate. In this imaging mode, based on Z contrast, MoS 2 appears as a single or double layer with bright contrast. It is covered by a carbon film (with low Z contrast) with a top-most Pt film, used as protective layers during FIB cross-section preparation. Furthermore, a dark stripe is typically present at the interface between MoS 2 and 4H-SiC. Figure 6b shows an atomic resolution HAADF-STEM, from which ≈1 nm thickness of this dark stripe is evaluated.
To get information on the composition of this interfacial region, an EDS spectrum image (Figure 6c) was simultaneously collected, showing the elemental distribution of carbon, oxygen, silicon, molybdenum, and sulfur. Furthermore, a depth profile of atomic fractions extracted from the spectrum image is reported in Figure 6d. On the left, we can see a composition, which belongs to the 4H-SiC substrate. Then, in the middle, where oxygen concentration (blue line) increased, we observe a SiO 2 composition. Then, one can observe the MoS 2 stripe, covered by the protective carbon film. Furthermore, Figure 6e shows an EDS spectrum collected in the rectangular box region in (c), corresponding to the dark stripe in the Z contrast image (b). This compositional analysis clarifies that the ≈1 nm dark stripe between MoS 2 and 4H-SiC observed in the HAADF image ( Figure 6b) is SiO 2 , formed during the high-temperature sulfurization of the original metal (oxide), as indicated also by XPS analyses (see Figure S1, Supporting Information). Figure 7a shows a correlative plot of the A 1g and E 2g Raman peaks' wavenumbers (ω A1g vs ω E2g ), obtained from a large array of 2500 spectra measured on MoS 2 on n + -SiC. This plot provides quantitative information on the doping and strain status of the MoS 2 film, by comparing the distribution of experimental data (open circles) with the theoretical ω A1g -ω E2g relations for a purely strained 1L MoS 2 (strain line) and for a purely doped 1L MoS 2 (doping line). [39] The strain (doping) lines are represented by the red (black) solid lines in Figure 7a, and their crossing point corresponds to the literature values ω 0 E2g = 385 cm −1 , ω 0 A1g = 405 cm −1 (at 532 nm excitation wavelength) for a suspended 1L MoS 2 membrane, [40] which is a good approximation for an ideally unstrained and undoped film. The strain line separates the p-type and n-type doping regions in the ω A1g -ω E2g plot, whereas the doping line separates the tensile and compressive strain regions. Hence, the experimental data distribution in Figure 7a clearly shows that the MoS 2 film produced by Mo sulfurization  map in Figure 7d,e. Furthermore, the comparison of the two maps in Figure 7c,e indicates the lack of correlation between the doping and strain spatial distributions in the MoS 2 film.
The same kind of analysis has been carried out on the array of Raman spectra collected on MoS 2 grown on the n − 4H-SiC epitaxy, and similar mean values of MoS 2 doping and strain were obtained (see Figure S2, Supporting Information).
In the case of 1L MoS 2 (with a thickness t ≈ 0.65 nm [41] ), an average hole density of 2.5 × 10 12 cm −2 corresponds to an average concentration of N h ≈ 4 × 10 19 cm −3 . This value is larger than the effective density of states in the valence band N v ≈ 1.2 × 10 19 cm −3 for 1L MoS 2 , estimated by using a parabolic valence-band model and the corresponding analytical formula where k is the Boltzmann constant, h is the Planck constant, T = 300 K, and m h = 0.61 m e is the holes effective mass for 1L MoS 2 . [42] Hence, the assumption of a degenerately p-type-doped MoS 2 is justified by the results of Raman analyses. On the other hand, the histogram in Figure 7b also shows a lateral variation of the hole doping in the MoS 2 layer, which allows to explain the spatial variability of the NDR in the I-V characteristic collected at different surface positions by the C-AFM tip (see Figure 5b,c).
The origin of this p-type doping can be associated to the presence of a significant MoO 3 fraction in the ultrathin MoS 2 film produced by the sulfurization process, as indicated by XPS analysis. In fact, several recent studies reported on the p-type conduction in MoS 2 by unintentional incorporation of oxygen, e.g., during the growth, [43] or by postdeposition oxygen plasma treatments. [44,45] Furthermore, the formation of shallow acceptor levels in oxygen-doped MoS 2 has been indicated by first principle calculations. [45] In the following, the forward bias I-V characteristics of the p + -MoS 2 /SiO 2 /n + -SiC heterojunction are discussed in detail, considering the different mechanisms ruling current injection at the interfaces. The equivalent circuit of this system, schematically depicted in Figure 8a, can be described by the series combination of the contact resistance R c associated to the tip/MoS 2 Schottky barrier [46] and of the p + -MoS 2 /SiO 2 /n + -SiC tunnel (Esaki) diode. Hence, the applied tip bias V tip > 0 partially drops across the contact resistance and partially across the diode. Figure 8b shows a typical forward bias I-V characteristic measured on the heterojunction. At low forward bias values (V tip < 3 V), the current injection is limited by the Pt tip/MoS 2 contact resistance. The semilog-scale plot of the I-V curve in this low current regime, reported in insert of Figure 8b, shows that current transport is limited by direct tunneling (DT) across the Pt/MoS 2 Schottky barrier from 0 to 2 V, being I DT ∝ V tip T with T the tunnel probability. Field emission (FE) across the barrier becomes the dominant mechanism between 2 and 3 V. In particular, the FE current increases with the forward bias as I FE ∝ exp[qV tip /E 00 ], being the q the electron charge and E 00 a characteristic energy related to the holes concentration N h in the p + -MoS 2 layer: with ε 0 the vacuum dielectric constant, ε r ≈ 6 the relative permittivity, [47] and m h the holes effective mass for 1L MoS 2 . A value of E 00 ≈ 377 meV was obtained from the I-V curve fit in the insert of Figure 8b, from which a hole concentration N h ≈ 6.9 × 10 19 cm −3 (i.e., a hole density, n ≈ 4.5 × 10 12 cm −2 ), in the range of values deduced by Raman mapping. At forward bias V tip > 3 V, the BTBT at the heterojunction becomes the dominant transport mechanism, resulting first in the current increase up to the peak value (I P ) at the voltage V P , followed by the decrease at the valley current (I V ) at the bias V v as indicated in Figure 8b. To better illustrate the BTBT phenomenon, the energy band diagrams of the p + -MoS 2 /SiO 2 /n + -SiC system under different biasing conditions are reported in Figure 8c-f. Furthermore, Figure S3 (Supporting Information) shows the alignment between the bands of p + -doped 1L MoS 2 , SiO 2 , and n + 4H-SiC separated by vacuum, evaluated using the literature values of electron affinity and energy bandgap for these materials. [48][49][50] Under equilibrium conditions, i.e., at V tip = 0 V (Figure 8c), the Fermi levels of the two degenerate semiconductors are aligned. The thickness of the tunnel barrier between the two semiconductors is the sum of the SiO 2 interfacial layer (≈1 nm) and the depletion region of the n + -SiC at V tip = 0 V, corresponding to the Debye length L D = (ε 0 ε SiC kT/qN D ) 1/2 ≈ 1.1 nm for N D ≈ 10 19 cm −3 . Under equilibrium, the filled states of the n + -SiC conduction band are aligned with the filled states of p + -MoS 2 valence band and BTBT is forbidden. When high-enough positive bias is applied, filled states on the SiC side, and unoccupied states on the MoS 2 side are partially overlapped, thus the electrons can tunnel through the barrier from n-SiC to p-MoS 2 .The BTBT current reaches its maximum (peak current I p ) at the bias V tip = V P, when the overlap between the filled and empty states at the two sides of the barrier reaches its maximum, as illustrated in Figure 8d. By further increasing the V tip the overlap starts to decrease, giving rise to a decrease of the BTBT current, i.e., to the NDR part of the curve. Ideally, the BTBT current is expected to decrease to zero at the bias V tip = V V, when the conduction band edge of SiC and the valence band edge of MoS 2 are aligned (as illustrated in Figure 8e). However, the actual measured value of I V and, hence, the peak-to-valley current ratio of ≈1.5, can be explained by the occurrence of other competitive transport mechanisms, such as current tunneling mediated by defects states at the interface. [51] For larger bias values (V tip > V V ), current transport is ruled by a combined thermionic-tunneling mechanism through the tunnel barrier, as schematically depicted in Figure 8e.
The reported results demonstrate an efficient method to obtain large area and uniform p + MoS 2 films on 4H-SiC, which is compatible with semiconductor industry process flows. Furthermore, the observation of room temperature NDR behavior in MoS 2 /SiC heterojunctions paves the way to the implementation of Esaki diodes on the silicon carbide platform, extending its range of potential applications to fast switching devices and circuits.

Materials Growth
The thin (1.2 nm) Mo films on n + 4H-SiC substrates and n − 4H-SiC epilayer were obtained by DC magnetron sputtering from a Mo-target (using a Quorum Q300-TD system). The sulfurization process was carried out in a two-heating zones furnace, with the first zone (at a temperature of 150 °C) hosting a crucible with 300 mg sulfur, and the second zone (at a temperature of 700 °C) hosting the Mo/SiC samples. Starting from a base pressure of 4 × 10 −6 bar, the Ar carrier gas (with a flux of 100 sccm) transported the S vapors from the first to the second zone. The duration of the sulfurization process was 60 min.

AFM Morphology and Conductive AFM Analyses
Morphological analyses were carried out by Tapping mode Atomic Force Microscopy using a DI3100 system by Bruker with Nanoscope V electronics. Furthermore, nanoscale resolution current-voltage characterization of MoS 2 /SiC heterojunctions was performed by Conductive Atomic Force Microscopy (C-AFM) at room temperature and under ambient conditions, using Pt-coated Si tips with ≈5 nm curvature radius.

XPS Analyses
The compositional properties of the as-deposited metal films and MoS 2 formation after the sulfurization process were evaluated by XPS using an XSAM 800 XPS instrument by Kratos Analytical (Manchester, UK), with a non-monochromatic Mg Kα X-ray source (energy = 1253.6 eV). The spectra were collected at a take-off angle of 90° relative to sample surface and pass energy of 40 eV. The instrument resolution is 1.1 eV (Au 4f 7/2 , pass energy of 5 eV). The spectra were aligned using the C-Si (silicon carbide) component of C1s (283.1 eV) as reference.

STEM and EDS Analyses
High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) and energy dispersion spectroscopy (EDS) analyses of the MoS 2 /4H-SiC heterojunctions were carried out with an aberration-corrected Titan Themis 200 microscope. To this aim, cross-sectioned samples were prepared by focused ion beam (FIB), after depositing a carbon/Pt protective layer on MoS 2 surface.

Raman Analyses
Raman spectroscopy and mapping of MoS 2 vibrational peaks were carried out by a WiTec Alpha equipment, using a laser excitation at 532 nm, 1.5 mW power, and 100 × objective.

Statistical Analysis
In this paper, the statistical analysis was performed on the results of Raman mapping reported in Figures 4 and 7. Arrays of 50 × 50 Raman spectra on 10 µm × 10 µm areas were collected both on the MoS 2 /n + 4H-SiC and MoS 2 /n − 4H-SiC samples. For each array, individual Raman spectra were extracted using the Witec software. A Matlab code written by the authors was used to evaluate the E 2g and A 1g vibrational peaks wavenumbers (ω E2g and ω A1g ) for all the spectra, and to generate the color maps/hystograms of Δω in Figure 4c,d,g,h. The average value of Δω and its standard deviation were calculated from these distributions. The Matlab code was also used to build the correlative plot of ω E2g , ω A1g wavenumbers in Figure 7a and to evaluate the maps/hystograms of local strain and doping values on the MoS 2 in Figure 7b-e. The average value of the strain and doping and their standard deviation were calculated from these strain/doping distributions.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.