Al2O3 Layers Grown by Atomic Layer Deposition as Gate Insulator in 3C-SiC MOS Devices

Metal-oxide-semiconductor (MOS) capacitors with Al2O3 as a gate insulator are fabricated on cubic silicon carbide (3C-SiC). Al2O3 is deposited both by thermal and plasma-enhanced Atomic Layer Deposition (ALD) on a thermally grown 5 nm SiO2 interlayer to improve the ALD nucleation and guarantee a better band offset with the SiC. The deposited Al2O3/SiO2 stacks show lower negative shifts of the flat band voltage VFB (in the range of about −3 V) compared with the conventional single SiO2 layer (in the range of −9 V). This lower negative shift is due to the combined effect of the Al2O3 higher permittivity (ε = 8) and to the reduced amount of carbon defects generated during the short thermal oxidation process for the thin SiO2. Moreover, the comparison between thermal and plasma-enhanced ALD suggests that this latter approach produces Al2O3 layers possessing better insulating behavior in terms of distribution of the leakage current breakdown. In fact, despite both possessing a breakdown voltage of 26 V, the T-ALD Al2O3 sample is characterised by a higher current density starting from 15 V. This can be attributable to the slightly inferior quality (in terms of density and defects) of Al2O3 obtained by the thermal approach and, which also explains its non-uniform dC/dV distribution arising by SCM maps.


Introduction
The cubic polytype of silicon carbide (3C-SiC) has a smaller energy gap (E g = 2.36 eV) [1,2] compared to the hexagonal 4H-SiC (E g = 3.26 eV) [3], but it possesses a higher electron mobility and saturation velocity [4][5][6][7][8]. Moreover, it exhibits a larger conduction band offset (3.7 eV) [9] with SiO 2 than 4H-SiC (2.7 eV). Hence, differently from the 4H-SiC/SiO 2 system where they are aligned with the conduction band edge of 4H-SiC, the near-interface-oxide-traps (NIOTs) inside the insulator in the 3C-SiC/SiO 2 system lie above the Fermi level and hence they are electrically inactive [10,11]. Furthermore, the lower position of the 3C-SiC conduction band edge with respect to the SiO 2 conduction band edge results immune to the interface states that are peculiar of the SiO 2 /4H-SiC interface [6]. This can lead to a higher inversion electron channel mobility (>200 cm 2 V −1 s −1 [12]) in 3C-SiC Metal-Oxide-Semiconductor Field-Effect Transistors (MOSFETs) compared to those fabricated using the 4H-poly-type.
Silicon dioxide (SiO 2 ) is the native oxide of SiC that can be obtained by a thermal oxidation process of the material [13,14]. However, its electrical behavior is adversely affected by the large number of defects [9,15] (e.g., carbon clusters and dangling bonds produced during oxidation), which results in a large negative shift of the flat band voltage (V FB ) [8,16,17]. Another issue is the response of the MOS system to the application of high voltages. In particular, in blocking configuration, the distribution of the electric field inside the insulator (E ins ) and the semiconductor (E s ) can be expressed by the Gauss' law, E ins = (κ s /κ ins )E s , where κ ins and κ s are the insulator and semiconductor permittivity values. Considering the permittivity values of SiO 2 (3.9) and 3C-SiC (9.7), the SiO 2 layer is subjected to an electric field 2.5 higher than 3C-SiC. Hence, the SiO 2 gate insulator reliability is seriously compromised under high electric field. Moreover, using SiO 2 does not enable full exploitation of the high critical field of the underlying 3C-SiC substrate. Consequently, thicker drift layer must be used, which in turn increases the total device on-resistance [18].
Insulators with high permittivity (the so-called "high-κ") can be a solution to overcome this limitation due to the better distribution of the electric field in the MOS system, which offers safer operating conditions in high voltage applications. Al 2 O 3 is a suitable high-κ oxide due to its permittivity value (κ~9), good thermal stability and relatively large band gap (~7 eV) [19][20][21][22][23]. The Atomic Layer Deposition (ALD) [24,25] is the best technique for the deposition of Al 2 O 3 thin layers with optimal thickness control, uniformity on large area, and high-quality interface [26][27][28]. The ALD growth of Al 2 O 3 thin films on SiC can be improved by the insertion of a nanometric SiO 2 interlayer (IL), which provides a larger amount of active nucleation sites than the bare SiC surface. Moreover, the introduction of SiO 2 IL between Al 2 O 3 and SiC is also convenient to guarantee a larger conduction band offset and finally to better prevent leakage phenomena [29,30]. To date, the Al 2 O 3 deposited by ALD as gate dielectric on 3C-SiC is completely unexplored. Actually, it has been adopted by R. Oka et al. [16] only as a thin interlayer between SiO 2 and 3C-SiC to improve the structural quality of their interface.
In this work, we report on the Al 2 O 3 thin film growth by ALD as an alternative insulator layer for 3C-SiC MOS capacitors using a very thin SiO 2 film as IL. In particular, the structural properties of Al 2 O 3 /SiO 2 stacks, of their interfaces on the underlying 3C-SiC but also their electrical behavior have been investigated by comparing the two different ALD approaches, namely the thermal (T-) and plasma-enhanced (PE-) ALD processes [31,32]. Both approaches allow obtaining good quality high-κ dielectrics. However, some literature works [16,22,23,31,33,34] report on slight differences both in the quality of the grown highκ and in its interfacial properties, directly related to the different oxidation mechanism between the two methods. In particular, the studies conducted on other semiconductor materials [16,32] suggest that the more reactive action of the O 2 -plasma produces Al 2 O 3 layers characterised by a higher mass density and lower amounts of the undesired carbon contaminations and unreacted OH-groups, which could act as active centres for electron trapping [35].
Furthermore, the combination of several characterisation techniques, i.e., morphologicalstructural and electrical-either at a macroscopic scale or at a nano-scale-allowed the full comprehension of the insulating properties of the differently ALD deposited Al 2 O 3 films.

Materials and Methods
A 10.2 µm thick 3C-SiC grown by chemical vapour deposition on Si (100) was employed as substrate [36]. Prior to the oxidation process, the 3C-SiC substrates were cleaned for ten minutes in an H 2 SO 4 :H 2 O 2 = 3:1 solution followed by ten minutes of etching in an HF:H 2 O = 1:5 solution. A 5 nm SiO 2 IL was grown by a controlled dry oxidation process at 1150 • C for 5 min. Successively, the Al 2 O 3 layers were deposited on SiO 2 /3C-SiC by either thermal-or plasma-enhanced ALD using trimethylaluminum (TMA) as an aluminium precursor and H 2 O or O 2 -plasma as co-reactants. Both processes were carried out at the deposition temperature of 250 • C. Meanwhile, the different growth rates of the T-ALD (~0.9 Å/cycle) and the PE-ALD (~1.2 Å/cycle) involve the use of a different number of deposition cycles (350 and 250 cycles for T-and PE-, respectively) to grow an A 2 O 3 layer with the same thickness of 30 nm.
The structural quality of Al 2 O 3 /SiO 2 /3C-SiC stacks and their morphology were investigated by transmission electron microscopy (TEM) using a FEG-TEM JEOL 2010F (Tokyo, Japan) microscope and by atomic force microscopy (AFM) using a DI3100 equipment by Bruker (Billerica, MA, USA) with Nanoscope V controller, respectively. In particular, the TEM analysis was carried out in cross-section in order to visualize the properties of the Al 2 O 3 /SiO 2 stack layer and their interfaces. For this purpose, cross-sectional specimens were properly prepared both for T-and PE-Al 2 O 3 /SiO 2 /3C-SiC samples by conventional mechanical preparation techniques, i.e., including polishing and dimple grinding, followed by a final thinning with ion milling.
The electrical behaviour of the insulating stacks was evaluated by capacitance-voltage (C-V) and current-voltage (I-V) measurements carried out on lateral metal-oxide-semicondu ctor (MOS) capacitors using a Microtech Cascade probe station equipped with a Keysight B1505 parameter analyser (Santa Rosa, CA, USA). Finally, the nanoscale electrical behaviour of the systems was monitored my means of scanning capacitance microscopy (SCM).

Discussion
The cross-section TEM image reported in Figure 1 illustrates a uniform and amorphous The structural quality of Al2O3/SiO2/3C-SiC stacks and their morphology were investigated by transmission electron microscopy (TEM) using a FEG-TEM JEOL 2010F (Tokyo, Japan) microscope and by atomic force microscopy (AFM) using a DI3100 equipment by Bruker (Billerica, MA, USA) with Nanoscope V controller, respectively. In particular, the TEM analysis was carried out in cross-section in order to visualize the properties of the Al2O3/SiO2 stack layer and their interfaces. For this purpose, crosssectional specimens were properly prepared both for T-and PE-Al2O3/SiO2/3C-SiC samples by conventional mechanical preparation techniques, i.e., including polishing and dimple grinding, followed by a final thinning with ion milling.
The electrical behaviour of the insulating stacks was evaluated by capacitancevoltage (C-V) and current-voltage (I-V) measurements carried out on lateral metal-oxidesemiconductor (MOS) capacitors using a Microtech Cascade probe station equipped with a Keysight B1505 parameter analyser (Santa Rosa, CA, USA). Finally, the nanoscale electrical behaviour of the systems was monitored my means of scanning capacitance microscopy (SCM).

Discussion
The cross-section TEM image reported in Figure 1 illustrates a uniform and amorphous Al2O3 layer with a thickness of ~30 nm and a sharp interface with the underlying SiO2/3C-SiC. The SiO2-IL is clearly distinguishable and has a thickness of about 4.5 nm. The structural properties of Al2O3 and of its interfaces are similar on both T-Al2O3/SiO2/3C-SiC and PE-Al2O3/SiO2/3C-SiC systems; thus, only the first is reported representatively. Lateral MOS capacitors schematically depicted in Figure 2a were fabricated using photolithography, metal deposition and lift-off processes. The anode of the MOS capacitors was surrounded by a large-area metal cathode so that its capacitance could be neglected ( Figure 2b). Ni/Au was used as metal electrode. The MOS structures fabricated on T-Al2O3/SiO2/3C-SiC and PE-Al2O3/SiO2/3C-SiC were probed by C-V measurements, which are shown in Figure 2c. Both samples provide C-V curves negatively shifted compared to the ideal value VFB = +0.9 V. In particular, the experimental flat band voltage values were −0.6 V and −3 V for the T-ALD and PE-ALD stacks, respectively. However, as can be observed in Figure 2c, such negative shifts were smaller compared to that of MOS capacitor where the insulator was only a thick (40 nm) thermal SiO2 [37]. This experimental finding is related to the higher dielectric constant of the Al2O3 (κ = 8) with respect to that Lateral MOS capacitors schematically depicted in Figure 2a were fabricated using photolithography, metal deposition and lift-off processes. The anode of the MOS capacitors was surrounded by a large-area metal cathode so that its capacitance could be neglected (Figure 2b). Ni/Au was used as metal electrode. The MOS structures fabricated on T-Al 2 O 3 /SiO 2 /3C-SiC and PE-Al 2 O 3 /SiO 2 /3C-SiC were probed by C-V measurements, which are shown in Figure 2c. Both samples provide C-V curves negatively shifted compared to the ideal value V FB = +0.9 V. In particular, the experimental flat band voltage values were −0.6 V and −3 V for the T-ALD and PE-ALD stacks, respectively. However, as can be observed in Figure 2c, such negative shifts were smaller compared to that of MOS capacitor where the insulator was only a thick (40 nm) thermal SiO 2 [37]. This experimental finding is related to the higher dielectric constant of the Al 2 O 3 (κ = 8) with respect to that of SiO 2 (κ = 3.9). In fact, even though the SiO 2 /3C-SiC interface is similar, in both cases resulting in analogous amount of effective charge (N eff ), this can cause a variation of the experimental V FB value, moving it toward the ideal one, as expressed by the following equations: where N eff is the effective trapped charge density, C OX is the accumulation capacitance, q is the electron charge, ε 0 is the vacuum dielectric constant, and t OX is the oxide thickness.
the accumulation capacitance, the dielectric constant κ of the insulating films was estimated to be ~8 both for T-and PE-Al2O3. As can be seen in Figure 2b, the C-V curves of the PE-and T-ALD Al2O3/SiO2/3C-SiC are characterised by a different electrical behavior. In fact, besides the negative flat band voltage shift occurring in both cases, it can be noticed that a bump was visible in the depletion region of the C-V curve of the PE-ALD sample. Plausibly, this bump was caused by the occurrence of charge trapping at deep interface states when increasing the bias [12]. On the other hand, the thermal Al2O3/SiO2/3C-SiC sample is characterised by a more pronounced stretch-out of the C-V curve. Evidently, the different nature of the oxidation process (plasma enhanced-PE, and thermal-T) used during the ALD growth of Al2O3 was responsible for the different electrical quality of the two interfaces.  According to Equations (1) and (2), for a constant N eff and an insulating layer thickness, an increased dielectric constant results in a smaller flat band voltage shift ∆V FB . Furthermore, the lower negative V FB shift of the Al 2 O 3 /SiO 2 /3C-SiC stack can be also explained by the shorter time for the thermally oxidation process needed to grow a 4.5 nm SiO 2 IL than that needed to grow a 30 nm thick SiO 2 . In fact, a shorter oxidation time produces a lower amount of carbon clusters responsible of the negative V FB shift [38]. From the accumulation capacitance, the dielectric constant κ of the insulating films was estimated to be~8 both for T-and PE-Al 2 O 3 . As can be seen in Figure 2b, the C-V curves of the PEand T-ALD Al 2 O 3 /SiO 2 /3C-SiC are characterised by a different electrical behavior. In fact, besides the negative flat band voltage shift occurring in both cases, it can be noticed that a bump was visible in the depletion region of the C-V curve of the PE-ALD sample. Plausibly, this bump was caused by the occurrence of charge trapping at deep interface states when increasing the bias [12]. On the other hand, the thermal Al 2 O 3 /SiO 2 /3C-SiC sample is characterised by a more pronounced stretch-out of the C-V curve. Evidently, the different nature of the oxidation process (plasma enhanced-PE, and thermal-T) used during the ALD growth of Al 2 O 3 was responsible for the different electrical quality of the two interfaces.
From the C-V curves, by applying the Terman's method [39], the interface states (D it ) distributions were calculated for both T-Al 2 O 3 /SiO 2 /3C-SiC and PE-Al 2 O 3 /SiO 2 /3C-SiC stacks, which are also reported in Figure 3 in comparison to that of the SiO 2 /3C-SiC system. The SiO 2 /3C-SiC and Al 2 O 3 /SiO 2 /3C-SiC samples exhibited a comparable D it distribution in the order of 2 × 10 12 cm −2 eV −1 [8]. On the other hand, the PE-Al 2 O 3 /SiO 2 /3C-SiC sample showed a lower D it distribution close to the 3C-SiC conduction band edge in the order of 5 × 10 11 cm −2 eV −1 , which can be due to the beneficial effect of the O 2 -plasma on the defects amount at SiO 2 /SiC interface [40]. In fact, Kim et al. [30] demonstrated that for the SiO 2 /SiC-based devices, the use of a SiO 2 growth process assisted by the highly reactive O 2 plasma guarantees the formation of an interface characterised by a lower amount of defects and more stable SiO bonds. Analogously, in our case, the PE-approach used to deposit the Al 2 O 3 could play a similar beneficial effect on the underlying SiO 2 /SiC interface.
in the order of 2 × 10 12 cm −2 eV −1 [8]. On the other hand, the PE-Al2O3/SiO2/3C-SiC sample showed a lower Dit distribution close to the 3C-SiC conduction band edge in the order of 5 × 10 11 cm −2 eV −1 , which can be due to the beneficial effect of the O2-plasma on the defects amount at SiO2/SiC interface [40]. In fact, Kim et al. [30] demonstrated that for the SiO2/SiC-based devices, the use of a SiO2 growth process assisted by the highly reactive O2 plasma guarantees the formation of an interface characterised by a lower amount of defects and more stable SiO bonds. Analogously, in our case, the PE-approach used to deposit the Al2O3 could play a similar beneficial effect on the underlying SiO2/SiC interface.  Figure 4. As can be seen, in both systems, the electrical breakdown occurred at a gate bias of over 26 V. However, while the PE-Al2O3/SiO2/3C-SiC sample maintained a constant current value of 10 −12 A up to the breakdown, the T-Al2O3/SiO2/3C-SiC sample exhibited a fast raise of the current starting from 15 V. The leakage current trend occurring across the Al2O3 layer deposited by thermal mode could be explained by a slightly lower mass density and a higher amount of -OH and/or -CH3 groups than that deposited by the plasma-enhanced mode due to the less efficacious oxidation process by the H2O-precursor [41][42][43]. Moreover, in comparison to the I-V curve typical of the 3C-SiC capacitor with a 40 nm thick SiO2 as a dielectric layer (also reported in Figure 4), which exhibited a breakdown voltage of about 20 V, both T-Al2O3/SiO2 and PE-Al2O3/SiO2 stacks were able to shift the breakdown phenomena toward higher voltages, over 26 V. The early breakdown of a thick thermal grown SiO2 on 3C-SiC has already been explained by F. Li et al. [44] as a consequence of the large amount of carbon left during the thermal oxidation process. However, in our case, the use of a short oxidation process to obtain only a thin IL probably resulted in a smaller amount of carbon defects to cause the early breakdown. layer deposited by thermal mode could be explained by a slightly lower mass density and a higher amount of -OH and/or -CH 3 groups than that deposited by the plasma-enhanced mode due to the less efficacious oxidation process by the H 2 O-precursor [41][42][43]. Moreover, in comparison to the I-V curve typical of the 3C-SiC capacitor with a 40 nm thick SiO 2 as a dielectric layer (also reported in Figure 4), which exhibited a breakdown voltage of about 20 V, both T-Al 2 O 3 /SiO 2 and PE-Al 2 O 3 /SiO 2 stacks were able to shift the breakdown phenomena toward higher voltages, over 26 V. The early breakdown of a thick thermal grown SiO 2 on 3C-SiC has already been explained by F. Li et al. [44] as a consequence of the large amount of carbon left during the thermal oxidation process. However, in our case, the use of a short oxidation process to obtain only a thin IL probably resulted in a smaller amount of carbon defects to cause the early breakdown.
The electrical behavior of both T-and PE-Al 2 O 3 /SiO 2 /3C-SiC stacks was studied at the nanoscale by SCM measurement. A schematic representation of the SCM experimental setup is illustrated in Figure 5a. During the surface scan with a diamond tip, an AC modulating bias at 100 kHz frequency and with amplitude ∆V = 2 V (below the conduction regime through the insulator) was applied to the sample, and the capacitance variation ∆C in response to this modulation was recorded with the SCM sensor. Figure 5b,c show the AFM morphology of the T-and PE-Al 2 O 3 /SiO 2 /3C-SiC stacks on the portion of the samples where the SCM maps were acquired. The highly irregular morphology is peculiar of the 3C-SiC material [18], which is characterised by terraces separated by anti-phase boundaries. The SCM maps of T-and PE-samples are reported in Figure 5d,e. The SCM signal is a result of the capacitance change (dC/dV) in the local metal-insulator-semiconductor capacitor, where the metal is the conductive AFM tip. Hence, the SCM response depends on the semiconductor characteristics (i.e., doping type and concentration) but also on the insulator properties (including thickness, interface state density, oxide traps, and permittivity) [45,46]. Considering that both the T-and PE-Al 2 O 3 layers were deposited on the same 3C-SiC substrate and that they are characterised by an equivalent interface with SiO 2 as IL, the different SCM maps (Figure 5d,e) obtained for the two cases can be correlated to the different insulator quality. In particular, the SCM map of the T-Al 2 O 3 /SiO 2 /3C-SiC stack, reported in Figure 5d, shows a non-uniform dC/dV signal distribution visible as the change in the color gradient from one spot to another. In contrast, the PE-Al 2 O 3 /SiO 2 /3C-SiC stack reported in Figure 5e exhibits a well-uniform SCM map, with only a small deviating region. The different SCM responses between T-and PE-systems could be due to the different structural quality of the Al 2 O 3 layers deposited by the two approaches. In fact, the different Al 2 O 3 quality, in terms of mass density and/or -OH/-CH 3 contaminations, which can arise by using the different oxidation processes (T-or PE-), determines its charge trapping behavior and permittivity and, ultimately, the SCM signal. Similar results have been previously reported for the growth of Al 2 O 3 thin layers on AlGaN/GaN heterostructures by the two different T-ALD and PE-ALD approaches, where the evolution of the insulating behavior investigated at the nanoscale upon increasing film thickness clearly indicated a different nucleation mechanism [16]. Hence, the present investigation at the nano-scale also confirms the better electrical performance of the PE-Al 2 O 3 layer already observed by the electrical measurements acquired on the macroscopic capacitors.  The electrical behavior of both T-and PE-Al2O3/SiO2/3C-SiC stacks was studied at the nanoscale by SCM measurement. A schematic representation of the SCM experimental setup is illustrated in Figure 5a. During the surface scan with a diamond tip, an AC modulating bias at 100 kHz frequency and with amplitude ΔV = 2 V (below the conduction regime through the insulator) was applied to the sample, and the capacitance variation ΔC in response to this modulation was recorded with the SCM sensor. Figure 5b,c show the AFM morphology of the T-and PE-Al2O3/SiO2/3C-SiC stacks on the portion of the samples where the SCM maps were acquired. The highly irregular morphology is peculiar of the 3C-SiC material [18], which is characterised by terraces separated by anti-phase boundaries. The SCM maps of T-and PE-samples are reported in Figure 5d,e. The SCM signal is a result of the capacitance change (dC/dV) in the local metal-insulatorsemiconductor capacitor, where the metal is the conductive AFM tip. Hence, the SCM response depends on the semiconductor characteristics (i.e., doping type and concentration) but also on the insulator properties (including thickness, interface state density, oxide traps, and permi ivity) [45,46]. Considering that both the T-and PE-Al2O3 layers were deposited on the same 3C-SiC substrate and that they are characterised by an

Conclusions
The insulating properties of the Al2O3 layers deposited on 3C-SiC both by the thermal-and plasma-enhanced ALD approaches were investigated. Our results demonstrated that:


A thin (5 nm) SiO2 IL between the Al2O3 and the 3C-SiC is useful to ensure the quality of ALD growth and to maximize the insulator/semiconductor band offset;  The Al2O3 is a valid alternative to the conventional thermally grown single SiO2 as gate insulator for 3C-SiC MOS-based devices. In fact, the Al2O3 layers showed a high permi ivity (~8), which produced a significant reduction in the negative flat band voltage shift that is usually observed with SiO2;  A different electrical behavior was found between thermal-and plasma-enhanced Al2O3 both by investigations on macroscopic MOS capacitors and at the nanoscale using SCM analysis. In fact, although both systems ensure an electrical breakdown over 26 V, the T-Al2O3/SiO2/3C-SiC stack exhibits early leakage phenomena already from 15 V. Moreover, the T-Al2O3/SiO2/3C-SiC is characterised by a non-uniform SCM map compared to the PE-Al2O3/SiO2/3C-SiC. This difference can be correlated to a different Al2O3 quality obtained through the two different oxidation processes (T-or PE-), resulting in an inhomogeneous charge trapping behavior and permi ivity.
These results can be important for the fabrication of 3C-SiC MOSFETs with a positive turn-on voltage with improved channel conduction properties.
Funding: This work has been partially supported by the European project CHALLENGE (Grant Agreement 720827). Moreover, the research received funding from the European Union (NextGeneration EU), through the MUR-PNRR projects SAMOTHRACE (ECS00000022) and iEntrance@ENL (IR0000027). Part of the experiments reported in this paper have been carried out in the Italian Infrastructure Beyond-Nano.

Conclusions
The insulating properties of the Al 2 O 3 layers deposited on 3C-SiC both by the thermaland plasma-enhanced ALD approaches were investigated. Our results demonstrated that: • A thin (5 nm) SiO 2 IL between the Al 2 O 3 and the 3C-SiC is useful to ensure the quality of ALD growth and to maximize the insulator/semiconductor band offset; • The Al 2 O 3 is a valid alternative to the conventional thermally grown single SiO 2 as gate insulator for 3C-SiC MOS-based devices. In fact, the Al 2 O 3 layers showed a high permittivity (~8), which produced a significant reduction in the negative flat band voltage shift that is usually observed with SiO 2 ; • A different electrical behavior was found between thermal-and plasma-enhanced Al 2 O 3 both by investigations on macroscopic MOS capacitors and at the nanoscale using SCM analysis. In fact, although both systems ensure an electrical breakdown over 26 V, the T-Al 2 O 3 /SiO 2 /3C-SiC stack exhibits early leakage phenomena already from 15 V. Moreover, the T-Al 2 O 3 /SiO 2 /3C-SiC is characterised by a non-uniform SCM map compared to the PE-Al 2 O 3 /SiO 2 /3C-SiC. This difference can be correlated to a different Al 2 O 3 quality obtained through the two different oxidation processes (Tor PE-), resulting in an inhomogeneous charge trapping behavior and permittivity.