*2.4. Low Work Function and Refractory Metals*

In the last years, regarding the choice of the Schottky metal for 4H-SiC-JBS, increasing interest has been devoted to metallization schemes containing metals (and their compounds) with low work function and a certain degree of stability with 4H-SiC and the environment. In the facts, these low work function materials (such as the refractory metals Mo, W, Nb, etc.) can guarantee a minimization of the on-state conduction losses, making this configuration highly aimed at industrial 4H-SiC-based Schottky device development [48]. Moreover, since these metals exhibit a high melting point, they could be indicated for harsh environment applications, requiring temperature-resistant materials [62].

In recent literature, many papers have dealt with the electrical characterization of Mo/4H-SiC Schottky contacts for power electronics [63–67], highlighting the possibility of achieving a barrier height value as low as 1.010 eV and an ideality factor of 1.045 [65]. As reported in those studies, the Mo/4H-SiC behaved as an inhomogeneous contact, with the current conduction dominated by a TE mechanism and a slight discrepancy from the ideal behavior explained either according to the Werner and Güttler [21] or the Tung model [17].

Very recently, Renz et al. [63] studied a series of surface passivation treatments to achieve an improvement of the Mo/4H-SiC Schottky diode electrical properties. In particular, after the deposition or thermal growth of an oxide layer, annealing processes, similar to those employed in metal-oxide-semiconductor field-effect transistors (4H-SiC MOSFETs) technology, were considered [68–70]. These treatments included thermal oxidation in O<sup>2</sup> or N2O environments at temperatures of 1400 and 1300 ◦C, respectively, or the deposition of a phosphorus pentoxide (P2O5) layer at 1000 ◦C for 2 h. The first two processes consume the SiC surface while the third one does not. In all the samples, the oxide on the surface was removed by cleaning in dilute HF (10%) solution prior to Mo Schottky metal deposition. Figure 8a shows this approach schematically, with the treated area of the semiconductor

depicted as a patterned blue layer. An electrical *I*–*V<sup>F</sup>* characterization at room temperature was performed on a set of equivalent Mo/4H-SiC diodes fabricated under different conditions and, for comparison, on an untreated Mo/4H-SiC contact (labelled as "control"). The *I*–*V<sup>F</sup>* curves were analyzed according to TE model, obtaining an almost ideal behavior after the treatments. The lowest barrier height value (*φ<sup>B</sup>* = 1.27 ± 0.03 eV) was observed for the contact subjected to a prior deposition of P2O<sup>5</sup> (left scale of Figure 8b). Surprisingly, although a reduced value of barrier height was obtained in the P2O5-treated contacts with respect to the control sample, this process enabled the lowest value of the leakage current (right scale of Figure 8b) to be obtained. The authors explained this effect with the capability of the oxide to homogenize the interface by filling the nanopits, as witnessed by means of morphological (AFM) and microstructural (TEM) analyses. The high density (5 <sup>×</sup> <sup>10</sup><sup>9</sup> cm−<sup>2</sup> ) of these nanopits allows one to believe that they are different from those typically related to the threading dislocations arriving on the surface of 4H-SiC and observed after removal of surface electrodes [71]: if the threading dislocation-related nanopits were demonstrated to be potential leakage paths for the current, plausibly, in the case of treated P2O<sup>5</sup> deposited on 4H-SiC surface, the nanopits were oxide-filled, with a barrier lowering due to two contributions, one associated with the phosphorous-rich region below the contact, which increases the n-type doping, and the other related to a homogenization of the barrier height after oxide termination of the surface defects. This could explain the reduction of the barrier, with a simultaneous decreasing of the leakage current. It is worth noting that the absence of silicide reaction at the Mo/SiC interface, which would otherwise consume the top few nanometers of 4H-SiC, enabled these beneficial changes in the contact subsurface. *Φ* −

− **Figure 8.** (**a**) Schematic view of Mo/4H-SiC Schottky diodes with premetallization treatments of the semiconductor surface, consisting of thermally grown oxidation in O<sup>2</sup> and N2O or oxide deposition of P2O<sup>5</sup> , followed by oxide removal prior to Mo deposition in all cases. (**b**) Schottky barrier height and reverse leakage current density at −500V values averaged over a set of *I*–*V<sup>F</sup>* curves of equivalent Mo/4H-SiC Schottky diodes with 4H-SiC surface pretreated under different conditions. Panel (**b**) is adapted with permission from Ref. [63]). Copyright 2021 AIP Publishing.

Furthermore, Mo-based Schottky contacts were also investigated as a possible route for an improved control of the Schottky contact properties. For instance, Stöber et al. [72] proposed the use of molybdenum nitride (MoNx) thin film metallization for adjusting the barrier height within a large range by varying the nitrogen N<sup>2</sup> fraction in the reactive sputtering metal deposition step in the fabrication process. The total gas flow, i.e., sum of argon and nitrogen, was kept constant at 80 sccm while the nitrogen fraction *χ* = N2/(Ar + N2) in the gas composition varied from 0 to 80%, by increasing the content of N<sup>2</sup> in the chamber. For a pure Mo contact (with *χ* = 0%), the Schottky barrier was *φ<sup>B</sup>* = 0.68 eV at roomtemperature, increasing up to 1.03 eV for *χ* = 80%. Due to the polycrystalline nature of the Mo2N thin films, the barriers showed an inhomogeneous behavior, probably arising from different microstructures with regard to the nitrogen fraction used in the processing.

In parallel to Mo-based 4H-SiC Schottky contacts, Schottky contacts based on W have also been largely studied, obtaining a Schottky barrier ranging between 0.94 to 1.29 eV [62,73,74]. According to these papers, a certain degree of inhomogeneity was observed in the W/4H-SiC Schottky contact, successfully explained by means of the Tung's model [73] or the Werner and Güttler's model [74].

χ

χ *Φ*

χ

Noteworthy, for both Mo and W metals, also the carbide compounds were considered as possible electrode material [75–78]. As an example, Knoll et al. [76] investigated a Schottky barrier based on tungsten carbide, fabricated by depositing a thin layer of W (2 nm) followed by a rapid thermal annealing in vacuum for 5 min at temperatures ranging from 600 to 1200 ◦C. Then, a 500 nm thick Al layer was deposited on the top of the structure to define the diode structures. At temperatures > 1000 ◦C, a W2C hexagonal structure layer in epitaxial relation with 4H-SiC was produced, stable up to the highest tested temperature of 1200 ◦C. In this system, they observed a barrier height of 0.94 eV extrapolated under forward I–V characterization.

Recently, we investigated 4H-SiC Schottky diodes with an 80 nm thick layer of tungsten carbide (WC) barrier metal, deposited by magnetron sputtering and defined by optical lithography and lift-off [39,78] process. The Schottky diodes were characterized both before (as-deposited) and after some annealing treatments with temperatures varying form 475 ◦C to 700 ◦C for 10 min in N<sup>2</sup> atmosphere by I–V measurements and applying the thermionic emission (TE) model to the analysis of the electrical characteristics. The Schottky barrier height *φB*, derived by fitting the linear region of the semilog forward *J*–*V<sup>F</sup>* curves reported in Figure 9, had an average value of 1.12 eV in the as-deposited contact and decreased down to about 1.05 eV after annealing at 700 ◦C. In our experimental conditions, the ideality factor was only slightly affected by the annealing treatment, with a value decreasing from 1.08 to 1.03. For sake of comparison, we also reported in the same Figure 9, a representative *J*–*V<sup>F</sup>* curve of a similar Ti/4H-SiC Schottky contact, for which a barrier height of 1.21 eV was observed [79]. *Φ*

**Figure 9.** Representative forward I–V characteristics of WC/4H-SiC Schottky diodes for the asdeposited contact and after thermal annealing at 700 ◦C. The I–V characteristic of a reference Ti/4H-SiC is also reported for comparison. The data are taken from Refs. [78,79].

In those studies, we observed a temperature-dependence of the *φ<sup>B</sup>* and *n*, extrapolated by means of I–V–T characterization, for the annealed-WC/4H-SiC contact [78], as well as for the W/4H-SiC contact fabricated and annealed under similar conditions [39]. This indicated the presence of a nanoscale lateral inhomogeneity for both Schottky contacts,

that was fully described by means of the Tung's model, with an effective barrier *φBeff* of 1.15 and 0.96 eV and a homogeneous barrier *φB*<sup>0</sup> of 1.28 and 1.11 eV for the W/4H-SiC and WC/4H-SiC contact, respectively. *Φ Φ*

*Φ*

Essentially, the promising results obtained for the 4H-SiC Schottky diodes based on these low-work function refractory materials (mainly W and Mo) enable the study of this kind of contacts to be pushed forward towards a better understanding of the Schottky properties and inhomogeneity, and a suggestion of possible solutions for a better barrier uniformity and interface quality.

## **3. Unconventional Approaches for the Control of 4H-SiC Schottky Interfaces**

Parallel to the standard metallization stacks and layouts presented in the previous Section, a variety of innovative contacts, chemical compounds or alternative metal stacks have been proposed as new routes to control the Schottky barrier height values on 4H-SiC. In the next subsections, we will discuss some of the representative papers on these unconventional methods.

#### *3.1. Manipulation of the Schottky Interface*

Since the early 2000s, some studies demonstrated the possibility of lowering the barrier height in the Schottky diode by the incorporation of nanostructures in the metal layer of the Schottky contact [80–83]. One of the first attempts in SiC was reported by Lee et al. [80], who studied the effect of Au-nanoparticle embedding in Ti/n-type 4H-SiC contact. The diodes were fabricated by first depositing Au-aerosol nanoparticles (diameter of 20 nm, density of 90 µm−<sup>2</sup> ) and then depositing 200 nm thick Ti layer in an evaporation chamber. The schematic view of the final Schottky diode structure is depicted Figure 10a. In Figure 10b, the I–V curves for these Schottky-diode embedding nanoparticles are compared to those of a control Ti/4HSiC Schottky diode. μ <sup>−</sup>

**Figure 10.** (**a**) Schematic view of Ti/4H-SiC Schottky contact with embedded Au nanoparticles on 4H-SiC. (**b**) Forward current–voltage characteristics of Au-nanoparticle embedded Ti/4H-SiC contact and particle-free Ti/4H-SiC control contact for different measurement temperatures (25, 100, 200 and 300 ◦C). Inset: comparison between the Schottky barrier height value of Au-nanoparticle-embedded Ti/4H-SiC contact and Ti/4H-SiC control contact, as a function of the measurement temperature. Figures adapted with permission from Ref. [80]. Copyright 2021 Elsevier Ltd.

From the comparison, carried out at four different measurement temperatures (25, 100, 200 and 300 ◦C), it was possible to point out that for each testing temperature, the I–V curve related to the Au-nanoparticle-embedded Ti/4H-SiC contact was shifted towards a lower voltage than the control sample (Figure 10b). The Schottky barrier height, derived by a fit in the linear region according to TE theory and reported in the inset of Figure 10b, was lowered

of 0.19 eV in the sample with nanoparticles. To explain this result, the authors invoked the enhancement of the electric field under the interface in the depletion region, due to the small size of the embedded particles and the large Schottky barrier height difference obtained by using two metals as Ti and Au. This is, in part, confirmed by theoretical calculation according to the Tung's dipole-layer approach [18].

Later on, other studies reported on Schottky contacts with Au- or Ag-nanoparticles embedded in a Ni-metal [82] or Al-metal [83] layer, obtaining a similar reduction of the Schottky barrier height, which was associated with a reduction of the metal work function induced by the presence of the interfacial nanoparticles.

Besides working on the barrier material, many efforts have been dedicated also to the preparation and treatments of the semiconductor surface, to obtain a higher degree of homogeneity of the contact.

As an example, the inhomogeneity observed in Schottky contacts to 4H-SiC could be reduced by suitable treatments of the semiconductor surface, such as passivation with the insertion of an insulating thin film between the semiconductor and the metal [84–86].

For example, Shi et al. [86] demonstrated that the presence of an ultrathin Al2O<sup>3</sup> layer, deposited on the semiconductor surface by atomic-layer deposition before the metal stack (Al 300 nm/Ti 100 nm) annealed at in Ar at 300 ◦C for 5 min (Figure 11a), enabled a reduction of the barrier height. Three different oxide-layer thicknesses were investigated in that work (0.8, 1.2 and 2 nm). The forward I–V characteristics, shown in Figure 11b for all the tested Al2O<sup>3</sup> thicknesses, indicated a reduction of the Schottky barrier height with the increase of the Al2O<sup>3</sup> thickness, down to a value lower than 1 eV (inset of Figure 11b). In particular, the cross-section TEM analyses showed that the insertion of Al2O<sup>3</sup> reduces the diffusion of Ti into 4H-SiC and, hence, the possible occurrence of solid-state reactions between metal and semiconductor. In this way, the formation of new titanium silicide and carbide phases is prevented, thus resulting in an improvement of the interface homogeneity.

**Figure 11.** (**a**) Schematic view of a Ti/4H-Si Schottky diode with the insertion of an ultrathin Al2O<sup>3</sup> layer between metal and semiconductor surface. (**b**) Forward I–V characteristics of the Ti/4H-SiC Schottky contact with increasing thickness of the inserted Al2O<sup>3</sup> -layer (0, 0.8, 1.2 and 2 nm). The trend of the Schottky barrier height as function of Al2O<sup>3</sup> thickness is reported in the inset. Panel (**b**) is adapted with permission from Ref. [86]. Copyright 2021 Elsevier Ltd.

In another case, the insertion of an ultrathin amorphous-hydrogenated SiC layer (a-SiC:H) in the Ti/4H-SiC contact has been assessed with promising results (see schematic view in Figure 12a) [87]. The amorphous layer, with a thickness between 0.7 and 4 nm, was grown on the 4H-SiC surface by means of plasma-enhanced chemical vapor deposition prior to Ti deposition. Thermal annealing in a vacuum at 600 ◦C was also performed. The value of the Schottky barrier height varied between 0.78 and 1.16 V. These values, derived from room temperature I–V measurements, are reported in Figure 12b. As one can see, the Schottky barrier height depends on the amorphous layer thickness and thermal annealing duration. Specifically, while a slight influence of the amorphous layer thickness was observed on

the Schottky barrier value, the duration of the 600 ◦C annealing, supposed to result in the formation of the Ti5Si<sup>3</sup> phase [88], had a more significant impact on the barrier height. In particular, the lowest barrier value was obtained after the longest annealing treatment.

*Φ* **Figure 12.** (**a**) Schematic view of a Ti/4H-SiC contact with an ultrathin amorphous SiC:H layer inserted between Ti and 4H-SiC. (**b**) Barrier height *φ<sup>B</sup>* for different thicknesses of the amorphous layer and duration of the annealing treatment. Panel (**b**) adapted with permission from Ref. [87]). Copyright 2021 Elsevier Ltd.

*Φ Φ* ≈ − ≈ − Another method to modify, at atomic level, the surface where the Schottky contact is formed, consisted of the graphitization of the 4H-SiC surface [89,90]. It is known that as-grown monolayer graphene (MLG) on hexagonal SiC consists of a buffer layer (BL), similar to graphene but still covalently bond to SiC, plus a graphene overlayer [91]. The as-grown MLG contact exhibits ohmic characteristics, which have been explained by a low Schottky barrier height (*φ<sup>B</sup>* = 0.36 ± 0.1 eV) with SiC [92]. Such a low *φ<sup>B</sup>* value has been ascribed to the positively charged dangling Si bonds at the BL/SiC interface, which cause a Fermi level pinning of graphene close to the SiC conduction band, as well as a high n-type doping of graphene itself (*<sup>n</sup>* <sup>≈</sup> <sup>1</sup> <sup>×</sup> <sup>10</sup><sup>13</sup> cm−<sup>2</sup> ) [92]. Differently, after an annealing treatment of the MLG in H2-atmosphere, the covalent bonds between the BL and SiC break up and H<sup>2</sup> saturates the dangling bonds, converting the electrically inactive BL into an additional real graphene layer [93]. This quasi-freestanding bilayer graphene (QFBLG) is moderately hole-doped (*<sup>p</sup>* <sup>≈</sup> <sup>8</sup> <sup>×</sup> <sup>10</sup><sup>12</sup> cm−<sup>2</sup> ) and provides a Schottky contact to 4H-SiC [94]. As discussed by Hertel et al. [94], these two different graphene/4H-SiC interfaces can be used side-by-side on the same chip in a real 4H-SiC-based MESFET device, as illustrated in Figure 13. Specifically, the MLG on 4H-SiC is used as ohmic contact for the source and drain electrodes (marked as "contact graphene"), while the QFBLG/4H-SiC–Schottky interface serves as a gate electrode (marked as "gate graphene") [94].

In this system, nanoscale conductive atomic force microscopy (C-AFM) on QFBLG showed a dependence of the Schottky barrier height on the diode area, from values in the range (0.9–1) eV obtained for large contacts, up to values approaching ~1.5 eV for the smallest contacts. The behavior of this kind of contact was explained by considering that SiC step edges and facets are preferential current paths causing the effective lowering of the barrier. The reduced barrier height in these regions can be explained in terms of a reduced doping of QFBLG from SiC substrate at (11–20) step edges with respect to the p-type doping on the (0001) terraces [93].

**Figure 13.** Schematic view of graphene/4H-SiC-based transistor, with two different interfaces, i.e., MLG and QFBLG contact acting as ohmic and Schottky gate contacts, respectively. Figure adapted with permission from Ref. [94]). Copyright 2012 Nature Portfolio.

A final example regards the work of Lin et al. [95], who explored a new way to fabricate tunable Schottky diodes with ns-pulsed excimer laser (193 nm)-modified n-type single-crystal 4H-SiC. The diodes were analyzed both by macroscopic I–V measurements by using Au-layer as electrode and by nanoscale characterization by means of atomic force microscopy in PeakForce TUNA configuration, this latter schematized in Figure 14a. Particularly, as noticed from the macroscopic I–V characterization on pristine and irradiated contacts, the most notable change in the I–V behavior was observed for the contact exposed to 2 J/cm<sup>2</sup> (not shown here). For the contact irradiated at such fluence, the nanoscale I–V characterization for different numbers of pulses (1–20 pulses), directly on the bare laser-exposed surface of the semiconductor, showed a rectifying electrical behavior of the contact, with the Schottky barrier increasing from 0.38 up to 1.82 V in the range 3–20 pulses (reported in Figure 14b). A combined analysis with Raman spectroscopy for the sample irradiated at 2 J/cm<sup>2</sup> demonstrated a graphitization of the 4H-SiC surface after laser irradiation, which is probably at the base of the barrier height increase in contact to the laser-modified 4H-SiC surface. For fluence as high as 5 J/cm<sup>2</sup> , the appearance of the peak corresponding to monocrystalline silicon (~520 cm−<sup>1</sup> ) was observed. −

**Figure 14.** Nanoscale current–voltage characterization (by PeakForce TUNA mode of AFM) of Au/4H-SiC contact to laser-irradiated semiconductor surface: (**a**) scheme of the nanoscale current– voltage measurement set-up; (**b**) Schottky barrier height values extrapolated by I–V analysis for 4H-SiC surface irradiated with different pulse numbers. Figure adapted with permission from Ref. [95]. Copyright 2021 Elsevier Ltd.

−

## *3.2. N-Type Doping of the Interface*

The capability of 4H-SiC to sustain a high electric field (if compared to conventional semiconductors, such as Si) enables the possibility of tailoring the Schottky barrier height by varying the doping concentration (and hence the electric field) below the contact. If, under reverse bias, the effect of a larger electric field has been widely investigated with an

experimentally observed larger leakage current explained by the *TFE* model [96,97] and mitigated by the use of the JBS layout, under forward bias, the effect of a modification of the electric field requires deeper understanding. For instance, ion-irradiation-induced damage below the interface in Ti/4H-SiC Schottky diodes showed the possibility of increasing the barrier height by a deactivation of the dopant and a reduction of the electric field at the interface following a re-ordering of the crystal structure [98].

On the other hand, as mentioned above, a way to increase the electric field consists of increasing the doping concentration in the semiconductor below the contact layer. In this context, it is interesting to study the effects on the barrier height and carrier transport mechanisms in a heavily-doped 4H-SiC layer [19,99]. Hara et al. [19] studied the dependence of the barrier height and forward carrier transport mechanism on the doping concentration *N<sup>D</sup>* in Ni/SiC Schottky barrier diodes with 4H-SiC epitaxial layer. In particular, they investigated a range of doping concentrations, varying from 6.8 <sup>×</sup> <sup>10</sup><sup>15</sup> up to 1.8 <sup>×</sup> <sup>10</sup><sup>19</sup> cm−<sup>3</sup> . The increase in the doping concentration entailed a shift of the forward I–V characteristics (that means larger current observed for higher doping concentration for a given voltage value). This shift corresponded to a lower turn-on voltage, increasingly stronger from the lightly-doped sample (*N<sup>D</sup>* = 6.8 <sup>×</sup> <sup>10</sup><sup>15</sup> cm−<sup>3</sup> ) to the heavily-doped sample (*N<sup>D</sup>* = 1.8 <sup>×</sup> <sup>10</sup><sup>19</sup> cm−<sup>3</sup> ). On the other hand, a modification was observed for the predominant current transport mechanism, sweeping from TE to a *TFE* mechanism for a higher doping concentration (*N<sup>D</sup>* > 2.6 <sup>×</sup> <sup>10</sup><sup>17</sup> cm−<sup>3</sup> ).

The predominance of the *TFE* mechanism for Schottky contacts on a heavily-doped 4H-SiC layer was also demonstrated for Ni/4H-SiC with a n+-type implanted layer of 4H-SiC (*N<sup>D</sup>* = 1.97 <sup>×</sup> <sup>10</sup><sup>19</sup> cm−<sup>3</sup> ) [100], whose forward *J*-*V<sup>F</sup>* characteristics are shown in Figure 15a. The inset reports the schematic energy band diagram for the metal/4H-SiC interface when a *TFE* current transport mechanism is predominant. This contact exhibited a lower value of turn-on voltage if compared to a reference Ni/4H-SiC contact formed on the 4H-SiC epilayer without implanted layer and standard epitaxial layer doping concentration, as clearly highlighted by the graph in Figure 15b. The possible increase of the leakage current could be mitigated by an appropriate choice of the device layout, as in the JBS diode. This last point was theoretically investigated in Ref. [100].

**Figure 15.** (**a**) Forward current density–voltage characteristics (open symbols) for Ni/n-type implanted-4H-SiC Schottky diode and fitting curve according to the *TFE* model (continuous line). In the inset, schematic energy band diagram for the metal/4H-SiC contact under forward bias, according to the *TFE* current transport mechanism. (**b**) Forward current density–voltage characteristics of the Ni Schottky contacts to n-type 4H-SiCwith or without a heavily doped n-type implanted layer. (Figure extracted from Ref. [100]).

#### **4. Conclusions**

In this paper, we overviewed some approaches applied in the 4H-SiC Schottky contact development in order to improve the performance of the Schottky devices.

After a short discussion on the fundamentals of the metal/4H-SiC Schottky contact formation and the typical electrical characterization by I–V measurements, we pointed out the well-established technology of Schottky diodes, using Ti or Ni-based Schottky barriers and discussed the current solutions, including the most promising low work function and highly chemically stable metallization schemes and appropriate diode layouts. Then, we presented some unconventional methods based on the manipulation of the metal/semiconductor interface and aimed at an improved control of the Schottky properties of the contact. As a matter of fact, although the metal/4H-SiC system has been studied for a long time, many aspects in the contact formation are still unclear and require a deeper understanding, both from a fundamental and a technological standpoint, in order to obtain superior control of the Schottky contact electrical properties.

Nevertheless, some solutions have shown interesting outcomes. For instance, the introduction of metal nanoparticles in the metal layer has been considered for the advantages given in terms of barrier reduction. Other solutions act on the semiconductor side, for example, with treatments before metal deposition, in order to homogenize the surface and narrow the barrier heights and ideality factor distribution. The effects on the Schottky barrier related to an increase of the doping density of the semiconductor layer have also been investigated. Although these are early studies, they are very promising for the practical implications in Schottky diode technology. In fact, in addition to an improvement of the electrical properties in terms of uniformity, these solutions addressed the superior control of the Schottky barrier height, with the ultimate capability to tailor and tune its value. Besides the possibility of obtaining insight into the physical characteristics of the Schottky contact, this aspect is of particular interest for the device makers, for the development of a new class of Schottky diodes with tailored characteristics.

**Author Contributions:** Writing—original draft preparation, experimental investigation and data analyses M.V.; writing—review and editing, M.V., F.R. and F.G.; conceptualization, supervision, project management and funding acquisition, F.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was in part funded by the ECSEL-JU project REACTION (first and euRopEAn siC eigTh Inches pilOt liNe), Grant Agreement No. 783158.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that support the findings of this study are available from the corresponding author upon reasonable request.

**Acknowledgments:** The authors would like to acknowledge their colleagues G. Greco, P. Fiorenza and R. Lo Nigro from CNR-IMM and G. Bellocchi and S. Rascunà from STMicroelectronics for fruitful discussions. S. Di Franco (CNR-IMM) is acknowledged for his technical support during device fabrication.

**Conflicts of Interest:** The authors declare no conflict of interest.
