Surface passivation of germanium by atomic layer deposited Al2O3 nanolayers

Surfaces of semiconductors are notorious for the presence of electronic defects such that passivation approaches are required for optimal performance of (opto)electronic devices. For Ge, thin films of Al2O3 prepared by atomic layer deposition (ALD) can induce surface passivation; however, no extensive study on the effect of the Al2O3 process parameters has been reported. In this work we have investigated the influence of the Al2O3 thickness (1–44 nm), substrate temperature (50–350 °C), and post-deposition anneal (in N2, up to 600 °C). We demonstrated that an effective surface recombination velocity as low as 170 cm s−1 can be achieved. The role of the GeOx interlayer as well as the presence of interface charges was addressed and a fixed charge density Qf=\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Q}_{\mathrm{f}}=$$\end{document} −(1.8 ± 0.5) × 1012 cm−2 has been found. The similarities and differences between the passivation of Ge and Si surfaces by ALD Al2O3 prepared under the same conditions are discussed.


Introduction
Germanium (Ge) is a versatile semiconductor which currently attracts attention in several research fields. In microelectronics for example, it is considered one of the most attractive alternative channel materials to Si for next generation metal-oxide-semiconductor field-effect transistors (MOS-FETs) [1]. Recent research in photonics has demonstrated direct-bandgap Ge and SiGe by realizing these materials in their hexagonal lattice structure [2,3]. The potential compatibility of these materials with Si electronics makes nanoleds and nanolasers of hexagonal Ge and SiGe attractive options for monolithic optoelectronic integrated circuits. The relatively small bandgap of Ge makes this material also a favorable choice as bottom sub-cell in multi-junction space solar cells [4,5].
The potential of Ge in these applications is hampered by a high density of electronic defects at its surface. In MOSFETs, these defects can lower the sub-threshold slope [6][7][8] and lead to a compromised carrier channel mobility [9,10], while in nanolasers [11][12][13] and solar cells [14,15] these defects can act as non-radiative recombination centers for electrons and holes which decrease the conversion efficiency. Surface passivation of Ge is essential to mitigate these effects. Atomic layer deposition (ALD) is a deposition technique that allows the preparation of ultrathin films that can passivate the surface of nanolaser cavities and solar cell's absorber layers. In MOSFETs dielectric films can passivate the top surface of the channel and simultaneously act as gate-oxide. ALD is especially suited for these purposes since it enables sub-nanometer thickness control combined with excellent conformality on high aspect ratio structures.
Most research on ALD films for surface passivation is conducted in the context of Si photovoltaics. These studies have demonstrated that ALD Al 2 O 3 on Si provides excellent passivation [15][16][17]. The nature of this surface passivation is twofold. First, ALD of Al 2 O 3 on Si leads to the formation of an ultrathin (~ 1.5 nm) high-quality silicon oxide (SiO x ) layer between Si and Al 2 O 3 [16]. The remaining defects at this Si/SiO x interface are predominantly silicon dangling bonds (Pb-type defects) [15,18], which are effectively passivated by atomic hydrogen [19][20][21] provided to and/or kept at this Si/ SiO x interface by the Al 2 O 3 passivation layer during a postdeposition anneal (PDA) [22]. This mechanism leads to an interface defect density ( D it ) as low as D it ≤ 1 × 10 11 cm −2 eV −1 Invited Paper [15]. Secondly, Al 2 O 3 yields a high negative fixed charge density ( Q f ) when deposited on Si ( Q f = 4-13 × 10 12 cm −2 ) [15]. This fixed charge forms an electrical field that leads to a space charge region in the silicon and hence enhanced surface passivation [23]. This mechanism is referred to as field-effect passivation. The combination of a low D it and a high Q f leads to a low surface recombination velocity ( S eff ), a common metric of surface losses in solar cells and nanolasers. For Si/Al 2 O 3 , surface recombination velocities can be as lows as S eff ≤ 10 cm s −1 [16,24,25].
ALD Al 2 O 3 is a known passivation layer for germanium (Ge) where it is mainly explored for application in MOSFETs [9,[26][27][28][29][30]. ALD of Al 2 O 3 as passivation layer on Ge has been shown to lead to the formation of a germanium oxide (GeO x ) interlayer [27,31,32], somewhat similar to the SiO x interlayer formed by ALD Al 2 O 3 on Si. It is, however, debated whether such an interfacial GeO x layer is beneficial or not. In contrast to SiO 2 , GeO 2 is traditionally regarded as undesired due to its limited thermal stability [9,[33][34][35] and water solubility [9,36]. Yet recent DFT calculations [30,37] and several experimental studies have shown that a Ge/GeO 2 interface with a low defect density is possible ( D it ≤ 10 11 cm −2 eV −1 ) [38][39][40]. Most research with respect to ALD Al 2 O 3 on Ge was therefore aimed at understanding and engineering the GeO x layer between Ge and Al 2 O 3 [29-31, 39, 41-43]. It has been reported that D it can be lowered by a factor 10 to 100 by addressing this interlayer [26,29,30]. Especially plasma postoxidation of the Ge/Al 2 O 3 stack has proven successful ( D it ≤ 10 11 cm −2 eV −1 ) [26]. Research regarding the ALD Al 2 O 3 process parameters themselves has remained very limited. Such parameters include the Al 2 O 3 film thickness, substrate temperature, and post-deposition anneal. This is remarkable since these parameters have proven to be very important for the passivation of Si by ALD Al 2 O 3 [25,44]. Also, the role of field-effect passivation in the Ge/Al 2 O 3 system appears to be unexplored and reports on the surface recombination velocity of this interface are rare.
In this work, we have systematically investigated the effect of the ALD Al 2 O 3 process parameters on the surface passivation of Ge. The passivation quality is determined using the upper limit of the surface recombination velocity S eff ,max , which means that the crude assumption is made that the Ge bulk lifetime is infinite, i.e., not limiting. The process parameters investigated include the Al 2 O 3 film thickness (1-44 nm), deposition temperature (50-350 °C), and post-deposition anneal temperature (300-600 °C). For the synthesis of the Al 2 O 3 films, we used a plasma-enhanced ALD process since it allows for a lot of freedom in processing conditions [45]. Additionally, we monitored the temporal stability of the Ge surface passivation and determined the magnitude of the fixed charge density of Al 2 O 3 to shed light on the role of field-effect passivation. To better understand our results, we utilized the very extensive knowledge about the Al 2 O 3 passivation of Si. We did this by making a 1-to-1 comparison with the surface passivation of Si by ALD Al 2 O 3 using earlier results from our research group that were obtained using the same methodologies and experimental tools as for the Ge surface passivation.

Results
Influence of Al 2 O 3 thickness, substrate temperature, and annealing temperature on S eff The surface recombination velocity for germanium at room temperature has been investigated as a function of the Al 2 O 3 film thickness, ALD substrate temperature, post-deposition annealing temperature, and storage time. The results are shown in Figs. 1, 2, 3 and 4. In Fig. 1, the surface recombination velocity is displayed for Al 2 O 3 films up to 44 nm thick. The figure shows the surface recombination velocity right after the post-deposition anneal and 3 months later. Right after annealing, two trends seem to be present: an increasing trend of S eff ,max for Al 2 O 3 thicknesses up to 6 nm and a decreasing trend for Al 2 O 3 thicknesses > 6 nm. The two trends seem to behave differently upon aging. For Al 2 O 3 > 6 nm, aging results in a similar decrease of S eff ,max . Below 6 nm we observed only a slight improvement for the 1 nm Al 2 O 3 film and no change for the 3 nm film upon annealing. Note that since the 3 nm sample deviates from the trend, we repeated the experiment with an independently prepared sample. This gave the same result indicating that the local maximum in S eff ,max at 3 nm after aging is not an artifact. With regard to the magnitude of the surface recombination we find a minimum of S eff ,max ≈ 200 cm s −1 for a film thickness of 44 nm and a maximum of S eff ,max ≈ 1800 cm s −1 for 3 nm Al 2 O 3 . These observations for Ge are distinct from those of Si, where a monotonically decreasing trend for S eff ,max is observed [41,46] and surface recombination velocities drop to values as low as S eff ,max ≤ 10 cm s −1 for thicknesses above 10 nm [16,24,25].
The impact of the germanium substrate temperature during ALD on the passivation quality is shown in Fig. 2. Films deposited at lower substrate temperatures seem to passivate the germanium surface better ( S eff ,max ≈ 170 cm s −1 ). The value of S eff ,max increases up to 350 °C. This is in contrast with the case of Si, where an optimum was found at T sub = 150-200 °C.
In Fig. 3 the effect of the post-deposition annealing temperature on the passivation quality is presented. In the asdeposited state, a very high surface recombination velocity was found ( S eff ,max ≈ 1.8 × 10 4 cm s −1 ); i.e., the Al 2 O 3 films provide hardly any passivation. Annealing of the wafers yielded a substantial reduction in S eff ,max . The optimal anneal temperature www.mrs.org/jmr Invited Paper was found to be around 425 °C, at which S eff ,max was reduced to S eff ,max ≈ 1.0 × 10 3 cm s −1 . Substantial improvements in surface passivation around this annealing temperature were also found by others investigating Ge/Al 2 O 3 stacks [9,27,38]. Interestingly, the optimum for germanium coincides with the optimum anneal temperature of 400-450 °C for Si [25].
Besides annealing, aging of the Al 2 O 3 passivated wafers showed remarkable improvements in surface passivation as seen in Figs. 1 and 2. In Fig. 4 we report on this improvement by plotting S eff ,max versus the number of storage days after the PDA. For this experiment, a wafer with 22 nm Al 2 O 3 was stored in the dark at room temperature. In the first 30 days the surface recombination velocity drops from S eff ,max ≈ 1770 cm s −1 to S eff ,max ≈ 270 cm s −1 . After 30 days, the surface recombination velocity stabilizes. In some cases, such an effect has also been observed to a limited extent for Si wafers passivated with Al 2 O 3 [47].

Determination of fixed charge density
In the previous section, we showed the effect of various process parameters on the surface recombination velocity. In this section we discuss the fixed charge density of the Ge/Al 2 O 3 stack, as determined by corona charging experiments. In Fig. 5 the surface recombination velocity is plotted versus the deposited corona charge Q c for Ge passivated with 22 nm Al 2 O 3 deposited at substrate temperature T Sub = 200 °C and annealed at 425 °C for 10 min in N 2 ambient. First S eff ,max increases and subsequently it decreases. A maximum in S eff ,max is reached for a positive corona charge density of (1.8 ± 0.5) × 10 12 cm −2 . This means that the fixed charge density of the Al 2 O 3 is  3 Ω cm 100 p-type germanium annealed at 425 °C for 10 min in N 2 ambient (this work) and 30 nm Al 2 O 3 on ~ 2 Ω cm p-type silicon annealed at 400 °C for 10 min in N 2 ambient (data from Dingemans et al. [24]). The Ge samples were stored in air at room temperature.

Figure 3:
Effective surface recombination velocity S eff,max as a function of the post-deposition anneal temperature (N 2 ambient). A single sample was annealed in consecutive steps of 10 min. The anneal temperature '25 °C' refers to the as-deposited state; i.e., without anneal. The wafers concern germanium (~ 400 μm, ~ 0.3 Ω cm, p-type, 100 ) and silicon (~ 275 μm, ~ 2.2 Ω cm, p-type, 100 ) covered with, respectively, 22 nm and 30 nm Al 2 O 3 . The data concern measurements performed directly after PDA. The data for Si have been taken from Dingemans et al. [25]. The maximum S eff ,max in Fig. 5 is related to the D it and indicates the level of chemical passivation [49]. For Ge this value is about 8 times larger than for Si ( S eff ,max ≈ 117 cm s −1 ), which implies a substantial difference in the chemical passivation. Note that for Ge a double peak appears to be present in Fig. 5. Such a feature can arise due to several reasons including the presence of two distinct defect states and slight asymmetry of the wafer surfaces (note that 2 separate ALD runs are required to cover both sides of the wafer).

Presence of interfacial oxide
We investigated the interfacial oxide by using a combination of XPS and cross-sectional TEM. XPS was performed on Ge substrates covered with 4 nm Al 2 O 3 . Figure 6a displays the spectra of the Ge 3d peak for Ge covered with Al 2 O 3 deposited at 200 °C before and after annealing at 425 °C for 10 min. In these spectra, two distinct peaks arise. The first peak arises at a binding energy of about 30 eV which corresponds to Ge-Ge bonds from the Ge substrate. The second peak arises at about 32.8 eV and corresponds to Ge-O bonds. The presence of this peak indicates that a GeO x interlayer exists between the Ge substrate and the 4 nm Al 2 O 3 . This interlayer is already present before annealing, which means that the GeO x forms during the ALD process itself.
The subsequent annealing did not induce notable changes in the interlayer as observable by XPS. Figure 6b shows the spectra of the Ge 3d peak for annealed Al 2 O 3 deposited at various temperatures (50-350 °C). For all conditions a clear peak around 32.8 eV is observed, which discloses that a GeO x interlayer is formed. The area of such a peak relates to the amount of detected Ge-O bonds, while the peak position is associated with the stoichiometry of the GeO x (i.e., the oxidation state of Ge). The area of the GeO x peak increases with deposition temperature suggesting the growth of a thicker interlayer at elevated deposition temperatures. The Ge 3d peak position at 32.8 eV remains virtually the same which implies no major differences in stoichiometry of the GeO x .
A cross-sectional TEM image of a Ge/Al 2 O 3 stack is shown in Fig. 7a along with the earlier results on Si (Fig. 7b). For the case of Si, we observed a clear SiO x interlayer, which is about 1.5 nm thick. The TEM on Ge yielded unfortunately no clear image of the interlayer due to a low contrast between the GeO x and Al 2 O 3 [48]. Yet, from integrated image brightness contrast profiles across the layer stack, acquired both using bright-field TEM as well as high-angle annular dark field STEM, we estimate the GeO x interlayer to be 2.2 ± 0.3 nm thick. A confirmation of the layer thickness using an EDX elemental profile was not successful due to the sensitivity of the Al 2 O 3 layer for the electron beam. On the other hand, the beam sensitivity of the Al 2 O 3 layer  only created a contrast between the two oxide layers, again yielding a GeO x layer thickness of ~ 2.3 nm.

Discussion
In this work, we have presented the role of several process parameters on the surface passivation of Ge by ALD Al 2 O 3 together with the determination of the fixed charge density and temporal stability of the passivation.
In the introduction, we mentioned that post-oxidation techniques of the Ge/GeO x /Al 2 O 3 stacks have proven to be very beneficial for the surface passivation of Ge. The parameter study performed in this work shows various process aspects of the Ge/GeO x /Al 2 O 3 stack (anneal temperature, film thickness, etc.) yield comparable effects on the surface passivation. The post-deposition anneal treatment was found to have the strongest effect on the surface passivation. The latter showed an improvement by a factor 18 for S eff ,max reached after an optimum PDA (10 min at 400 °C) and no PDA. Film thickness showed an important effect as well, since an increase of the Al 2 O 3 film from 1 to 44 nm lowered S eff ,max by a factor 8. Aging of the Ge/GeO x / Al 2 O 3 stack for 30 days resulted in a similar improvement (about a factor of 6.5). Lastly, we found that the deposition temperature has only a relatively mild influence on the surface passivation: decreasing the deposition temperature from 350 to 50 °C results in an improvement of S eff ,max by only a factor of 2.5. The importance of these findings is that they enlarge the possibilities to fabricate well passivated Ge/GeO x /Al 2 O 3 interfaces which can improve the performance of Ge-based (opto)electronic devices.
The peak heights of S eff ,max in Fig. 5 are proportional to the level of chemical passivation [49]. For the Ge surface passivated by Al 2 O 3 this S eff ,max peak is substantially higher than for the Si surface passivated by Al 2 O 3 , indicating a lower level of chemical passivation of the former. This finding is in accordance with literature where interface defect densities for Si/Al 2 O 3 are commonly reported around D it ∼ 1 × 10 11 cm −2 eV −1 [15], while for Ge/Al 2 O 3 values are reported between D it = 5 × 10 11 cm −2 eV −1 and 1 × 10 13 cm −2 eV −1 [10,26,27,42] (at least without intentionally increasing the thickness of the GeO x interlayer by for example an additional post-oxidation plasma step [26], see also introduction). A different nature of the interface defects involved could lay at the heart of this difference between Si and Ge. For the Si/SiO 2 interface, it is well known that stretched Si bonds and Si dangling bonds [43,[50][51][52] form the dominant defects. For the Ge/GeO x interface, there seems less consensus. Various possibilities are proposed including: Ge-M bonds (with M = Hf, Zr, etc.) [53], sub-stoichiometric GeO x [30], and Gedangling bonds [37,54]. Especially the latter is topic of debate [37,[54][55][56][57][58], which indicates a key difference between the Ge/ GeO x /Al 2 O 3 and the Si/SiO x /Al 2 O 3 interface. Also, the role of hydrogen as passivating agent for Ge-dangling bonds appears controversial [55][56][57][58].
In the debate about Ge-dangling bonds, an extensive study by Stesmans et al. [58] concluded that Ge-dangling bonds at the Ge surface can be passivated by hydrogen, but only partially (~ 60% max). Assuming a less prominent role of hydrogen in the chemical passivation of the Ge/GeO x interface, the differences between Si and Ge in the dependence of S eff ,max upon Al 2 O 3 www.mrs.org/jmr substrate temperature as observed in Fig. 2 may not be so surprising. The effectiveness of the Si dangling bonds hydrogenation process during annealing depends on a complex interplay between the Al 2 O 3 microstructure and the hydrogen content of the film, which both vary with substrate temperature [22]. At lower substrate temperatures (below 200 °C) [22] a fairly high hydrogen content is obtained, while higher substrate temperatures (especially above 200 °C) [22] result in denser films which effuse less hydrogen during annealing. The trade-off between these properties leads to an optimum substrate temperature of T sub ≈ 200 °C [22]. If this hydrogenation process is indeed less effective for passivation of remaining defects at the Ge/GeO x interface, it can be understood that the dependence on substrate temperature may be very different.
The post-deposition annealing temperature results indicate an optimum annealing temperature around 425 °C (Fig. 3) which is similar to the optimum annealing temperature for Si [44]. The knowledge that for Si optimal hydrogenation occurs around 400 °C, and that the activation energies for the hydrogenation reactions of Ge and Si dangling bonds are similar (1.44 eV and 1.51 eV respectively [58]), seem to favor the argument for hydrogenation as a passivation mechanism for Ge. The role of hydrogenation seems, however, less than for Si, because Fig. 3 shows that the improvement of S eff ,max with this optimum annealing treatment is substantially less for Ge/Al 2 O 3 than for Si/Al 2 O 3 . These two conclusions resonate with the results of Stesmans et al. [58] who states that: hydrogenation passivates only a certain fraction of the Ge-dangling bonds and that this process is optimal around 375 °C.
The origin of the positive effect of aging on S eff ,max is not yet understood (Fig. 4). Multiple possibilities exist. A first reason could be filling of oxide traps at the GeO x /Al 2 O 3 interface which contribute to a higher Q f . For Si/Al 2 O 3 this was observed when exposed to ultraviolet light for prolonged time [47]. Our samples, however, were stored in the dark. Alternatively, the O 2 plasma in the ALD process could have induced interface and/ or bulk defects by the energetic photons it emits [59]. Post-deposition curing of these defects might cause S eff ,max to decrease over time. This process may already occur during annealing but can continue to proceed at a slower pace during storage at room temperature.
For Si/SiO x /Al 2 O 3 , the fixed charge density is reported to be virtually constant with Al 2 O 3 film thickness [41]. The measured values for plasma-enhanced ALD Al 2 O 3 range from Q f = 4 × 10 12 cm −2 to about Q f = 13 × 10 12 cm −2 [23,24,60,61]. We determined the fixed charge density to be Q f = −(1.8 ± 0.5) × 10 12 cm −2 (Fig. 5). As for Si, the nature of the charge is found to be negative. Moreover, we observed only a mild increase in Q f with film thickness, which indicates that most of this charge resides near the interface with Ge. A difference with Si is the magnitude of Q f , which is found to be lower. The significance of these findings is that they shed light on the field-effect passivation of Ge induced by Al 2 O 3 . This field-effect passivation is most relevant for devices like nanolaser and solar cells. Due to the electron-repelling effect of the negative fixed charge in the Al 2 O 3 , the induced field-effect passivation works more effectively on p-type doped Ge.
A GeO x interlayer was found to consistently form during the ALD process (Fig. 6a, b). The oxygen plasma used in our ALD process can contribute to its formation. We found by XPS that the interlayer grows thicker for higher substrate temperatures (Fig. 6b), an effect that can be attributed to enhanced diffusion of the oxygen species through the GeO x . In contrast to some earlier reports [27,29,62], this increasing GeO x interlayer with www.mrs.org/jmr substrate temperature did not result in better surface passivation (Fig. 2).

Conclusion
In this study, we elucidated the effect of the Al 2 O 3 film thickness, deposition temperature and anneal temperature on the surface passivation of Ge by ALD Al 2 O 3 . It has been found that optimization of these process parameters can enable a reduction of S eff ,max up to an order of magnitude, resulting in surface recombination velocities as low as S eff ,max = 170 cm s −1 . This information provides guidance for designing well passivated Ge/GeO x / Al 2 O 3 interfaces. We also determined the fixed charge density of the Ge/GeO x /Al 2 O 3 stack to be Q f = −(1.8 ± 0.5) × 10 12 cm −2 . This value implies that S eff ,max benefits from a certain degree of field-effect passivation, which can be especially beneficial for p-type Ge surfaces. Finally we found that the passivation quality improves during storage in air after which it finally becomes stable.

Experimental details
For the experiments, we used double side polished p-type Ge wafers (~ 400 μm, 100 , ρ = 0.2-0.4 Ω cm, dopant: gallium) procured from Umicore. The native oxide was removed from the germanium substrates by dipping in diluted hydrofluoric acid (1%, 90 s) followed by deionized water rinsing (90 s) and N 2 blow-drying. The wafers were coated on both sides with ALD Al 2 O 3 in a FlexAL™ system from Oxford Instruments, which is described in detail elsewhere [63]. The first half cycle consists of vapor drawn dosing the Al-precursor trimethylaluminum [TMA, Al(CH 3 ) 3 , 99.999% pure, Dockweiler Chemicals]. A remote oxygen plasma is used as co-reactant in the second half cycle of the ALD process. The substrate temperature was kept constant during the deposition runs at a chosen value between 50 and 350 °C. At the start of each deposition, the wafers were subjected to a warmup step of 10 min in N 2 ambient. After deposition, the wafers received a post-deposition anneal (PDA) to activate the passivation [25]. The annealing was performed for 10 min in N 2 ambient (N 2 gas purity > 99.999%) at 425 °C using a Jipelec Rapid Thermal Annealer unless stated differently. The thickness of the ALD Al 2 O 3 films was measured by spectroscopic ellipsometry using a J.A. Woollam Co., Inc. M2000 rotating compensator spectroscopic ellipsometer.
To measure the effective excess carrier lifetime [ τ eff (s)] in Al 2 O 3 passivated Ge wafers, we used the commercially available WCT-120 Photoconductance Lifetime Tester. This tool is well known in the research field of silicon solar cells and was originally developed to measure carrier lifetime in silicon wafers. The technique is referred to as quasi-steady-state photoconductance (QSSPC) and its operating principle can be found in the work of Cuevas and colleagues [64]. To make the tool suitable for Ge wafers we followed a similar approach as proposed by Cornagliotti et al. [65]. From the effective excess carrier lifetime one can derive an upper limit of the surface recombination velocity [ S eff ,max (cm s −1 )] by assuming that the bulk of the wafer has a negligible effect on τ eff compared to the surface; i.e., the bulk lifetime is assumed to be infinite. The maximum surface recombination velocity is a well-established metric to express the surface passivation [66]: where n (cm −3 ) is the average excess carrier density in the wafer and W (cm) is the wafer thickness. For this expression we thus assume that τ eff is surface limited and that S eff ,max is not diffusion limited. We report S eff ,max for n = 10 15 cm −3 , the standard for Si. The measurement uncertainty in S eff ,max is determined from the uncertainties in the parameters used to calculate the lifetime and surface recombination velocity from the measured quantities of the WCT-120 Photoconductance Lifetime Tester.
The stability of the Ge passivation was monitored over the course of 240 days by regularly performing QSSPC measurements. The wafers were stored in the dark under ambient conditions.
Surface passivation can be established by two different mechanisms [23]. The first is a reduction of the interface defect density (D it ) at the semiconductor surface. This mechanism is commonly referred to as chemical passivation. The second mechanism is a significant reduction of either the electron or hole concentration at the surface. This can be achieved by an electric field which can be induced by fixed charge ( Q f ) in the dielectric film deposited on the semiconductor surface. This type of passivation is referred to as field-effect passivation. To determine the contribution of field-effect passivation to the passivation of the Ge surface by Al 2 O 3 we conducted corona charge experiments [67][68][69]. For this purpose, incremental positive or negative charge was deposited on the surface of the passivated Ge wafer using ionized air molecules created by a needle with a DC voltage of ± 10 kV with respect to the sample using the Corona Charging System of Delft Spectral Technologies. After each corona charge step, both the S eff ,max and the deposited surface charge density were measured. The former was obtained with the QSSPC method and the latter with Kelvin probe potential measurements that were conducted in the Corona Charging System. Repetition of these steps allowed us to plot the deposited corona charge ( Q c ) versus S eff ,max . The point of maximum S eff ,max provides the fixed charge density. At this point applies: X-ray photoelectron spectroscopy (XPS) provided information about the chemical composition of the films and the Ge-Al 2 O 3 interface. A Thermo Scientific KA1066 spectrometer www.mrs.org/jmr employing monochromatic Al Kα (hν = 1486.6 eV) X-rays radiation was used. The background subtraction method for the XPS data was a Shirley background with the additional constraint that the background should not be of greater intensity than the actual data at any point in the region. The adventitious carbon C-C peak at 284.8 eV was used as charge correction reference. The transmission electron microscopy (TEM) study was performed using a probe-corrected JEOL ARM operated at 200 kV and equipped with a 100 mm 2 Centurio SDD energy dispersive X-ray spectroscopy (EDS) detector. Cross-sectional TEM samples were made using a standard lift-out focused ion beam (FIB) preparation protocol [70].

Acknowledgments
This work was supported by the Gravitation Program "Research Centre for Integrated Nanophotonics" (Grant Number 024.002.033) of The Netherlands Organization for Scientific Research (NWO). The work of J. Melskens

Open access
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/ licen ses/by/4.0/.