Correlating electron trapping and structural defects in Al₂O₃ thin films deposited by plasma enhanced atomic layer deposition

Aluminum oxide (Al₂O₃) is an attractive material for a variety of applications in electronic devices.¹ As an example, thin Al₂O₃ layers can be used as a gate insulator to reduce the leakage current in gallium nitride (GaN) high electron mobility transistors (HEMTs) or as a passivation layer to reduce the current collapse phenomena in these devices.^2,3 The great interest toward Al₂O₃ in the GaN technology arises from the excellent physical properties of this insulator, i.e., high dielectric constant κ (∼9), high critical electric field (10 MV/cm), large bandgap (∼8.9), and favorable band alignment with GaN.⁴ 

In this context, atomic layer deposition (ALD) is the most widely used technique today to deposit Al₂O₃ thin films on GaN. In fact, ALD is a self-limited growth mechanism, enabling an accurate control of thickness and interface abruptness, and a uniform coverage at low deposition temperatures, i.e., compatible with the GaN thermal stability range and typical device processing.

In general, the quality of amorphous thin films grown by ALD can be extremely variable.⁵ In particular, theoretical calculations show that Al₂O₃ layers are characterized by the presence of native defects^6–8 that can also influence their electrical behavior (i.e., leakage current) by introducing electronic levels in the dielectric bandgap.^9,10

In the last decade, the vast majority of the literature on GaN-based devices focused on the electrical behavior of thermal ALD Al₂O₃ films.^{2–4,11–13} Despite in some cases charge trapping phenomena have been observed, which lead to an instability of the Al₂O₃/AlGaN device characteristics,^11,13 a clear correlation of these effects with the structural quality of the film has not been reported.

In this context, plasma enhanced atomic layer deposition (PE-ALD) is a promising alternative to the standard thermal ALD approach, possessing the advantage of higher growth rate and film density and low deposition temperature.^14,15 However, the origin of charge trapping at interface states and bulk oxide traps in PE-ALD Al₂O₃ films on GaN-based heterostructures remains under discussion,^16,17 thus hindering the full technological exploitation of these films.

In this paper, the electron trapping phenomena in Al₂O₃ thin films deposited by plasma enhanced atomic layer deposition on AlGaN/GaN heterostructures have been investigated and correlated with the interface and bulk structural quality of the film. In particular, the correlation of capacitance measurements upon stress and structural analyses allowed us to attribute the observed electron trapping to the presence of low-coordinated Al cations, i.e., oxygen vacancies in the Al₂O₃ film. A practical implication of this physical effect on the insulated gate AlGaN/GaN HEMT technology has been discussed, opening the possibility to fabricate thermally stable enhancement mode transistors.

Al₂O₃ films have been deposited by Plasma Enhanced ALD (PE-ALD) onto AlGaN/GaN heterostructures grown on Si(111) substrates. Prior to the Al₂O₃ deposition, the AlGaN surface has been treated by a sequential cleaning process using piranha (H₂O₂:H₂SO₄ = 1:5) and diluted hydrofluoric acid (H₂O:HF = 10:1) solutions, for 10 min and 5 min, respectively, in order to remove the carbon contamination and native oxide.¹⁸ These chemical treatments, before the deposition process, were selected among different possibilities, which were explored in earlier research and whose effects on the electrical properties were evaluated.¹⁹ The Al₂O₃ deposition has been carried out on a PE-ALD LL reactor from SENTECH Instruments GmbH, using trimethyl-aluminum (TMA) and O₂-plasma as aluminum and oxygen sources, respectively. The O₂-plasma has been generated by a capacitively coupled plasma (CCP) source, through a 13.56 MHz RF-generator with a power of 200 W. For the plasma run, a 100 sccm O₂ flow was released to the plasma source. Nitrogen (N₂) gas, with a flow rate of 40 sccm, has been used as a carrier for the TMA precursor. During the ALD cycle, pulse periods of 60 ms and 1 s for TMA and O₂-plasma, respectively, have been coupled with the purging pulse of N₂ gas for 2 s. The deposition process has been carried out at 250 °C, with a chamber pressure of 20 Pa. According to the nominal growth rate of 1.2 Å/cycle, the number of cycles was chosen to be 250 in order to obtain an Al₂O₃ film thickness of about 30 nm.

Circular Al₂O₃/AlGaN/GaN metal–insulator–semiconductor (MIS) capacitors were fabricated to monitor the electrical quality of the bulk oxide and of the interface by capacitance–voltage (C–V) measurements under different bias stress conditions. In these control devices, the Ohmic contact was formed by an annealed Ti/Al/Ni/Au multilayer,²⁰ while the gate contact consisted of a Ni/Au bilayer.

Morphological, structural, and chemical analyses of the deposited Al₂O₃ films were carried out employing a variety of techniques on blanket al₂O₃/AlGaN/GaN samples. In particular, atomic force microscopy (AFM) was carried out to investigate the surface morphology of the films, using a Veeco Dimension 3100 AFM with a NanoScope V controller. High-resolution transmission electron microscopy (HR-TEM) on cross-section samples enabled us to evaluate the film thickness as well as microstructural properties of the Al₂O₃/AlGaN interface. The chemical nature of the Al₂O₃ thin films was investigated by Electron Energy-Loss Spectroscopy (EELS) in the scanning transmission electron microscopy (STEM) mode with a nanometer electron probe. Both kinds of analyses were carried out using a Field Emission Gun-TEM JEOL 2010F microscope.

Finally, Al₂O₃/AlGaN/GaN metal–insulator–semiconductor high electron mobility transistors (MISHEMTs) have been fabricated and characterized by current–voltage (I–V) measurements, monitoring the stability of the transfer characteristics at high temperatures (up to 100 °C). The C–V and I–V measurements on capacitors and transistors were carried out using a Microtech Cascade probe station equipped with a Keysight B1505 parameter analyzer.

The growth of the Al₂O₃ insulating layer on AlGaN/GaN heterostructures has been obtained by the PE-ALD procedure described in detail elsewhere.¹⁹ The role of surface preparation before deposition has been fully exploited, and the effects on the formation of interface defects have already been discussed in previously reported papers.^17–19 Here, the dielectric properties and charge trapping phenomena of Al₂O₃ layers, grown by the PE-ALD optimized procedure,^18,19 have been investigated on Al₂O₃/AlGaN/GaN MIS capacitors, by the C–V measurements, as reported in Fig. 1(a). These measurements have been carried out by sweeping the gate bias from negative toward positive values and backward. Different C–V curves have been sequentially collected by increasing the final gate bias value from 0 V to +15 V on “fresh” (non-stressed) MIS capacitors, and the hysteresis between the forward and backward C–V curves has been determined for each stress condition. In particular, these measurements allowed us to discriminate among the different trapping contributions at the Al₂O₃/AlGaN and AlGaN/GaN interfaces. It should be underlined that the sweeping rate did not affect the value of the hysteresis. The reported C–V curves have been acquired at 0.05 V/s, and each point has been averaged with four measurements. The first C–V measurement [the blue curve in Fig. 1(a)], collected from V_G = −10 V to 0 V and backward, shows no hysteresis. The depletion of the two-dimensional electron gas (2DEG)²¹ at the AlGaN/GaN interface is clearly visible with a “pinch-off” of the curves at V_PO = −7.5 V. The absence of hysteresis demonstrates that no charge trapping occurs by stressing the MIS capacitor in the negative gate bias range, i.e., at the AlGaN/GaN interface where the 2DEG is located.

Furthermore, C–V curves have been collected by increasing the final positive gate bias stress up to +10 and +15 V [red and black curves in Fig. 1(a)]. Summarizing, the gate stress is applied by extending the V_G bias toward larger positive values. Hence, initially, all the C–V measurements have been carried out at the gate bias ranging from −10 V to 0 V and backward, and then, the final bias value has been progressively increased up to +15 V. Under these conditions, the C–V curves first rise at the “flat band voltage” V_FB and then exhibit a second step due to the accumulation of electrons in the AlGaN layer at the Al₂O₃ interface.

The C–V curves exhibit two capacitance saturation levels that correspond to the capacitance of the Al₂O₃ layer (at a high positive gate bias) and to the series capacitance of the Al₂O₃ and AlGaN layer (in the V_G range of −5 V/+5 V). Hence, from the accumulation capacitance, the relative permittivity of the Al₂O₃ layer has been determined to be $κAl2O3$ = 8.4, which is close to the theoretical Al₂O₃ bulk value (≈9), thus demonstrating the good dielectric quality of deposited films.

It is interesting to note that the hysteresis of the C–V curves is always observed under positive bias stress conditions, such as close to the pinch-off V_PO and to the flat band voltage V_FB. These hystereses can be associated with different charge trapping effects occurring in the MIS system.

To quantify these effects, the flat band voltage and pinch-off voltage shifts (ΔV_FB and ΔV_PO) have been monitored as a function of the positive gate bias stress. In particular, these shifts are associated with the charging of the Al₂O₃/AlGaN interface states (ΔV_FB) and of the traps in the bulk Al₂O₃ (ΔV_PO).²² In fact, to provide a quantitative description of the observed charge trapping effects, the following assumptions are needed: (i) the insulator bulk traps are located at the Al₂O₃/AlGaN interface, as often described in the text books,²³ and are responsible for the shift of V_PO (ΔV_PO) in the C–V curves and (ii) the interface traps at the Al₂O₃/AlGaN interface are completely de-trapped under the application of negative V_G values.

Following the planar capacitor approximation, the total amount of charge trapped at the Al₂O₃/AlGaN interface (N_it) can be obtained by

where $CAl2O3$ is the insulator layer capacitance in accumulation and q is the electron charge.

Based on these assumptions, the capacitance plateaux C_PO observed close to V_G = 0 V are the series of the insulator (⁠$CAl2O3$⁠) and AlGaN layer (C_AlGaN) capacitances,

and are related to the V_PO shift (ΔV_PO) through the following equation [Eq. (3)]:

which has been used to calculate the total amount of trapped charges in the Al₂O₃ film (N_OT),

where $tAl2O3$ and t_AlGaN are the Al₂O₃ and AlGaN layer thicknesses, while $κAl2O3$ and κ_AlGaN are the Al₂O₃ and AlGaN relative permittivities.

Figure 1(b) reports the total amount of the trapped charges at the Al₂O₃/AlGaN interface (N_it) and in the bulk Al₂O₃ (N_OT) as a function of the positive gate bias stress. The values have been extracted from the C/V curves from the shifts ΔV_FB and ΔV_PO, using Eqs. (1) and (4), respectively.

As can be seen, the value of N_it increases with the increase in the final positive bias stress value, until a saturation is observed at high stress bias, corresponding to a maximum of the trapped charge, N_it = 7 × 10¹² cm⁻². In fact, by increasing the positive bias stress, in the MIS capacitor, on the AlGaN/GaN heterostructure, electrons are spilled-over from the 2DEG and reach the Al₂O₃/AlGaN interface, where they can fill the available free traps until reaching saturation. Once the gate bias is swept backward, the Al₂O₃/AlGaN interface states are fully discharged.

On the other hand, the total amount of trapped charges in the bulk Al₂O₃ does not show such a saturation with the increase in the bias stress [Fig. 1(b)]. The occurrence of a positive ΔV_PO shift in the C–V curves is associated with the presence of slow traps in Al₂O₃, which retains the negative charge even after strong negative backward gate biasing. Hence, the positive shift occurs from the initial value (non-stressed MIS capacitor) and the ΔV_PO in Eq. (2) is calculated from that initial V_PO value [see Fig. 1(a)]. In this case, the maximum of the ΔV_PO shift measured after a +15 V stress corresponds to a maximum of N_OT = 5 × 10¹² cm⁻².

The charge trapping/detrapping phenomena occurring in our system are graphically reported in Fig. 2, which shows a schematic band diagram of the Al₂O₃/AlGaN/GaN MIS system.

In particular, when the MIS capacitor is in accumulation (V_G ≫ 0), the electrons at the Al₂O₃/AlGaN interface can also fill the oxide traps inside the insulator layer (N_OT). Once the capacitor is biased in the opposite direction (V_G ≪ 0), the interface traps are discharged, but the oxide traps retain their charged state. In fact, electrons in bulk oxide traps located near the oxide/nitride interfaces can be retained during reverse bias sweeps due to their long detrapping time even at high temperatures.¹⁷ 

The charge trapping phenomena observed in the C/V measurements have been correlated with the morphological and structural properties of the Al₂O₃ films.

The surface morphology of the deposited Al₂O₃ layers has been monitored by using an AFM [Fig. 3(a)] and is characterized by the presence of nanometric rounded grains. The root mean square roughness (RMS) of the film is 0.52 nm, i.e., only slightly higher compared to that of the AlGaN substrate (RMS = 0.32 nm). This latter indicates that the conformal coverage of the AlGaN surface has been achieved.

Then, TEM analysis has been carried out to assess the structural properties of the deposited Al₂O₃ layer and to image its interface with the AlGaN substrate. In particular, the high resolution cross-sectional image (HR-TEM), shown in Fig. 3(b), revealed the formation of an amorphous and uniform layer of Al₂O₃ having a thickness of about 30 nm, with an abrupt Al₂O₃/AlGaN interface and no evidence of an interfacial layer.²⁴ The good interface quality demonstrates the effectiveness of the surface treatment adopted prior to the PE-ALD and the absence of damage due to the remote plasma.

Further information on the microstructure of the Al₂O₃ film has been gained by the STEM–EELS technique. In particular, Energy Loss Near-Edge Structure (ELNES) investigation has been performed on the sample, focusing on the Al L_2–3-edge whose shape is related to Al coordination.²⁵ In order to have a proper interpretation of the data, the Al L_2–3-edge acquired on the PE-ALD deposited Al₂O₃ film has been compared to those of the tabulated defect-free crystalline phases (α-, γ-) Al₂O₃ with known Al-coordination [Fig. 3(c)].²⁵ The tabulated spectrum of the Al L_2–3 edge in the α-Al₂O₃ phase [Fig. 3(c), black line], consisting of only octahedrally coordinated Al cations (i.e., six O-coordinated Al), is identified by a high intensity peak centered at 79 eV and a much less intense peak at 80 eV.^25,26 On the other hand, the tabulated Al L_2–3 edge of the γ-Al₂O₃ phase [Fig. 3(c), red line], consisting of both tetrahedrally (four-coordinated) and octahedrally coordinated Al cations, shows two signals at 78 eV and 79.5 eV, with inverted intensity compared to α-Al₂O₃.^25–27 For both α-phase and γ-phase, the broad peak at about 84 eV is an indication of a medium-range crystalline structure. The experimental spectrum of the Al L_2-3 edge obtained on our PE-ALD deposited Al₂O₃ film [Fig. 3(c), green line] shows the absence of the peak at 84 eV due to the lack of the medium range order.²⁸ Moreover, it exhibits a split broad peak at about 79 eV and 78 eV, having the shape and energy position intermediate between the α- and γ-phase and a second low-energy signal at 76.7 eV. All the features of this edge structure can be related either to the coexistence of four- and six-coordinated aluminum or also to other coordinated aluminum cations (3-, 5-) and, consequently, to the presence of distorted bonds.^25,28–30 The presence of low-coordinated aluminum cations, revealed by our analysis, can be associated with an inherent oxygen deficiency in the Al₂O₃ film. In fact, ab initio calculations on low-coordinated aluminum atoms have indicated the presence of transition energy levels in the bandgap of the insulator, which can act as electron traps in the Al₂O₃/AlGaN system.⁶ 

It is well-known that, besides oxygen vacancies, also other defects, such as residual H and/or C species, could act as charge trapping centers within the dielectric material.³¹ For this reason, the quantitative comparison between the experimental density of trapped charges and the amount of oxygen vacancies can be useful. From the ΔV_PO of the C–V curves, the amount of trapped charges in the Al₂O₃ of N_OT = 6 × 10¹² cm⁻² has been estimated. Assuming the filling of all the bulk traps in the Al₂O₃ layer, their volume concentration (for a thickness of 30 nm) results to be about 2 × 10¹⁸ cm⁻³. In a more realistic scenario, under our bias stress conditions, electron trapping occurs in a limited portion of the insulator, i.e., within about 1 nm from the semiconductor interface, thus resulting in a volume trap charge density of about 6 × 10¹⁹ cm⁻³. Unfortunately, the quantification of the oxygen vacancies in amorphous Al₂O₃ thin films only by our ELNES spectrum is not straightforward, as it would require standard amorphous Al₂O₃ samples with a known oxygen vacancy content, which are not easily available. In fact, as proposed in the literature for other materials, such a quantification is done mainly by fitting the EELS spectra using theoretical calculations of the amorphous supercell with an appropriate amount of oxygen vacancies.^32,33 The electronic properties of oxygen vacancies in amorphous Al₂O₃ have been discussed in the literature.^9,34 Perevalov et al. estimated an oxygen vacancy content for amorphous Al₂O₃ thin films grown by ALD in the order of 7 × 10²⁰ cm⁻³.⁷ Based on the above considerations, since the experimental trapped charge density is lower than the expected oxygen vacancy density, it is plausible to argue that not all these intrinsic defects are electrically active in our system.

Based on the above considerations, since the experimental trapped charge density is lower than the expected oxygen vacancy density, it is plausible to argue that not all these intrinsic defects are electrically active in our system.

Moreover, the presence of negative charges trapped inside the bulk Al₂O₃ gate insulator can induce Coulombic screening of the underlying 2DEG at the AlGaN/GaN interface. Hence, it can be, in principle, used to achieve the enhancement mode operation in AlGaN/GaN HEMT devices, provided that the amount of negative charges trapped in the oxide is of the same order of magnitude of the 2DEG density. The achievement of enhancement mode operation in AlGaN/GaN HEMTs remains one of the most critical aspect in the GaN technology for future power electronics applications.³⁵ 

Figure 4(a) reports the transfer characteristics (I_D–V_G) of the Al₂O₃/AlGaN/GaN MISHEMT before and after a 15 V positive gate bias. As can be seen, after the positive gate bias, the MISHEMT shows an enhancement mode behavior, with a positive pinch-off bias V_PO = 1.3 V. At the same time, the transistor still exhibits a low gate leakage current (I_G ∼ 2 × 10⁻¹¹ A/mm at V_G = +5 V). Johnson et al.²² and Hou et al.³⁶ observed an analogous trapping effect in HfO₂ and Al₂O₃ AlGaN/GaN MISHEMTs and attributed it to the presence of a GaON interfacial layer formed during the thermal ALD process. However, in our PE-ALD film, the abrupt interface showed by HR-TEM analysis [Fig. 3(b)] confirms that the electrical behavior is ruled by the presence of structural defects within the Al₂O₃ layer acting as charge traps. Those traps, in turn, seem to have no significant effect on the leakage current.

It is important to point out that the effect of the trapped charges in the Al₂O₃ is stable with the temperature. In fact, Fig. 4(b) reports a sequence of five transfer characteristics of the device collected at 100 °C between V_G = −20 V and +5 V. Evidently, these curves are completely overlapped, thus demonstrating the stability of the film to retain the trapped charge under high temperature operation that requires several days to be relaxed toward the initial condition.

Certainly, the thermal stability of the charge trapping in the bulk traps deserves to be fine controlled in order to achieve fully reliable enhancement mode AlGaN/GaN HEMTs.

In summary, electron trapping occurring in Al₂O₃ films grown by plasma enhanced atomic layer deposition on AlGaN/GaN heterostructures has been monitored and correlated with the structural properties of the films. In particular, C–V analyses allowed us to monitor the trapping/detrapping effects both in the bulk Al₂O₃ and at the interface, providing an amount of Al₂O₃ bulk traps in the order of N_OT = 5 × 10¹² cm⁻². These defects can be filled with electrons under appropriate positive bias conditions. The ELNES spectra of the aluminum L_2–3-edge demonstrated the occurrence of locally low coordinated aluminum cations in Al₂O₃, which are associated with an oxygen deficiency in the film. The local oxygen deficiency can be correlated with the presence of the electron traps in the insulator and can explain the electrical results. This charge trapping phenomenon can be intentionally used to control the depletion of the 2DEG at an AlGaN/GaN interface, obtaining an enhancement mode operation stable up to 100 °C. Hence, the results can be particularly useful for device manufacturers and open new routes for achieving enhancement mode AlGaN/GaN HEMTs.

The authors would like to thank F. Giannazzo (CNR-IMM) for the stimulating scientific discussions. This work was partially funded by the Italian Ministry for Education, University and Research (MIUR) in the framework of the National Project PON EleGaNTe (Electronics on GaN-based Technologies), Grant No. ARS01_01007.

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