1 Introduction

The semiconductor/oxide interface is a basic constituent in modern electronics devices, such as metal–oxide–semiconductor field-effect transistors (MOSFETs). The semiconductor/oxide interface strongly influences various crucial electrical properties of MOSFETs, such as threshold voltage, carrier mobility, 1/f noise, radiation response, long-term reliability, and stability [1,2,3,4,5,6]. Amount of dangling bonds exit at the semiconductor/oxide interface due to the interruption of the periodic lattice structure. The dangling bonds can be charged by the gain or loss of electron. These interfacial defects induce trap energy levels in the semiconductor band gap and play a dominant role in degrading the device performance [1]. At the crystalline-Si/amorphous-SiO2 interface, the interfacial states are primarily Si dangling bonds or the Pb centers [1, 2, 7,8,9,10], such as the Pb for (111) and (110) Si orientations and Pb0 and Pb1 for (100) Si orientation [11,12,13]. In addition, a paramagnetic recombination center named Pm center [14] and SinO3−n≡Si· (n = 1, 2) defect named S center [15] were reported. With further scaling of the Si MOSFET node, performance improvement becomes increasingly difficult. Ge as channel material with better hole and electron mobility than that of Si is expected to be a significant breakthrough to achieve high performance [16, 17]. Similarly, Ge dangling bonds (Ge3≡Ge·) were reported due to the inherent mismatch between Ge and its oxide [18,19,20,21,22].

In general, the density of dangling bonds and interfacial states can be reduced by sufficient oxidation or terminated with another atom. For Si case, the dangling bonds can be effectively passivated with termination to H or D by annealing in hydrogen or deuterium [23,24,25,26]. For the Ge case, the interface state density (Dit) is still a major technical issue hindering the application of Ge-based metal–oxide–semiconductor (MOS) devices [27, 28]. Surface passivation is a critical challenge to achieve high-quality high-k stacks. Commonly used approaches include nitride [29, 30], S [31, 32], Si [33, 34], and Ge oxide passivation [35,36,37]. Among these methods, Ge oxide passivation shows super electrical property with the effective passivation at the Ge/GeOx interface having low Dit. In recent years, there have been extensive studies on the GeOx passivation and Ge/GeOx interface properties. For example, Zhang et al. [38] investigated the effect of plasma post-oxidation on the MOS interface properties of Al2O3/GeOx/Ge structures with different GeOx thicknesses. Although a relationship between Dit and GeOx thickness has been reported, the mechanism behind this relationship has not been elucidated. Yang et al. [39] investigated the capacitor performance of Al2O3/GeOx/Ge gate stack by ozone oxidation method. But it mainly focused on the electrical property comparison between cycling ozone oxidation and single-ozone oxidation passivation. Xu et al. [40] investigated the electrical performance of HfO2/Al2O3/GeOx/Ge pMOSFET with ozone post-oxidation and plasma post-oxidation, listing the Dit values following the two oxidation methods but providing no detailed explanation of the results. Although there have been numerous reports on interface passivation, especially for the GeOx/Ge system, the researches mainly focus on the electrical property of the interface and Ge MOS devices.

The low Dit and low equivalent oxide thickness (EOT) are the prerequisites for continued dimensional scaling of Ge MOSFETs. Owing to the relatively low permittivity of GeOx, decreasing the GeOx thickness is an effective method to scale down the EOT. However, Dit increases when the GeOx thickness decreases. Zhang et al. [38] reported that Dit increases as the GeOx thickness decreases by plasma post-oxidation. Shibayama et al. [41] attributed the dependence of Dit on GeOx thickness to the amount of Ge3+ oxide component by plasma thermal oxidation. Ozone oxidation is considered as an effective method to obtain superior GeOx/Ge interface at low temperature and avoid the thermal degradation of GeOx [42]. Kuzum et al. [43] reported that Dit decreases with increasing GeOx thickness owing to the Ge4+ oxidation component by ozone oxidation. The passivation mechanism seems to differ between plasma oxidation and ozone oxidation. Therefore, it is necessary to further clarify the passivation mechanism by ozone oxidation and observe whether the suboxide passivation is related to the oxidation method.

In this work, we investigate the Dit passivation mechanism of GeOx/Ge stack by ozone oxidation. We found the Dit at the Ge/GeOx interface is not be passivated by termination with another atom. The high oxidation state of GeOx is the main reason for the decrease of Dit. We propose a possible physical mechanism to explain the Dit passivation, namely remote Coulomb potential perturbation from high oxidation state of GeOx. This remote Coulomb potential shifts the eigen energy of dangling bonds, moving it from the band gap into the conduction or valence band and passivating Dit.

2 Experiment

We fabricated Ge-sub MOS capacitors and film stacks to explore the distributions of Dit and different Ge oxidation states, respectively. As for the capacitor samples, the GeOx/Al2O3 gate stack was used. After cleaning the Ge surface in HF (100:1) for 60 s, the GeOx was grown by ozone oxidation at 300 °C. The oxidation time was varied to control the GeOx thickness. Then, 10 nm Al2O3 was deposited by atomic layer deposition (ALD) using trimethylaluminum (TMA) and H2O as precursors at 300 °C. After post-deposition annealing (PDA) at 400 °C in N2 for 5 min, 30 nm TiN and 75 nm W were deposited by ALD. Al was used as the backside contact. All these samples were annealed in forming gas ambient at 400 °C for 30 min. In addition, the Ge MOS capacitors with GeOx/Al2O3 gate stack were fabricated in different PDA ambients. GeOx was grown by ozone oxidation at 300 °C for 25 min. 10 nm Al2O3 was then deposited at 300 °C. One sample was not annealed and used as the control. The others underwent PDA in N2 and O2 ambient at 400 °C for 30 min. Subsequently, the metal electrode was formed using the same process conditions as above. As for the Ge/GeOx/Al2O3 film stack samples, one group was with different GeOx thicknesses and the same thickness of Al2O3 of 1 nm, and the other group was with 0.7 nm GeOx and 2 nm Al2O3 in different PDA ambients. The X-ray photoemission spectroscopy (XPS) measurement was performed using Thermo Scientific ESCALAB 250xi equipped with a monochromatic Al Kα radiation source of 1486.8 eV. The pass energy was set as 15 eV. All data were collected at a take-off angle of 90° relative to the sample surface.

3 Results and discussion

The electrical property of MOS capacitors with different GeOx thicknesses is investigated. The Dit is measured using the low-temperature conductance method. Figure 1 shows the Dit at 0.3 eV above the valence band maximum vs. GeOx thickness. Dit rapidly decreases as the GeOx thickness increases up to 8 Å but varies minimally when the GeOx thickness increases beyond 8 Å. Therefore, Dit at the Ge/GeOx interface is dependent on the GeOx thickness. These results show that ozone oxidation leads to the same Dit trend with GeOx thickness as plasma oxidation [44, 45]. A GeOx layer with thickness larger than 8 Å is necessary to achieve good electrical property. The inset in Fig. 1 shows the capacitance–voltage (C–V) curves of MOS capacitor with 10.6 Å GeOx at multiple frequencies. Super C–V curves indicate a low Dit and high-quality Ge/GeOx interface.

Fig. 1
figure 1

Dit as a function of GeOx thickness. The inset shows the C–V curves of the MOS capacitors with 10.6 Å GeOx

Full size image

Studies have shown that Dit at the Ge/GeOx interface correlates with Ge dangling bonds or oxygen-related defects [22, 46]. Oxygen deficiency leads to oxygen vacancies and Ge dangling bonds near the Ge/GeOx interface [19, 47]. To investigate the origin of the Dit–GeOx thickness relationship, the chemical state of Ge is examined using XPS. Figure 2 shows the Ge 3d spectra of Ge/GeOx stacks with different GeOx thicknesses. The Ge 3d peaks at binding energy of 29.35 and 32 eV are from Ge substrate and GeOx, respectively. The peak signal corresponding to GeOx gradually increases with a thicker GeOx layer. The chemical shifts of Ge1+, Ge2+, Ge3+, and Ge4+ relative to that of Ge0 are taken as 0.8, 1.8, 2.75, and 3.4 eV, respectively [48, 49]. The full widths at half maximum (FWHM) for Ge0, Ge1+, Ge2+, Ge3+, and Ge4+ spectra are determined to be 0.57, 0.57, 0.9, 1.12, and 1.2 eV, respectively. Each core-level spectrum was fitted by a nonlinear Gaussian–Lorentzian line shape and Shirley background subtraction. The fixed spin orbit splitting of Ge 3d is 0.58 eV. Figure 2(a) shows the Ge 3d spectrum of the 4.1 Å GeOx. Only Ge1+ and Ge2+ signals are observed, and no Ge3+ and Ge4+ signals are detected. For the 6.6 Å GeOx, Ge3+ signal is obvious but Ge4+ signal is extremely small as shown in Fig. 2b. Figure 2c shows Ge3+ and Ge4+ signals become more evident in a thicker GeOx of 7.8 Å. In addition, we analyze the Ge(GeOx) to O(GeOx) areal intensity ratio as a function of GeOx thickness, as shown in Fig. 3. The value of Ge(GeOx)/O(GeOx) decreases with increasing GeOx thickness, which suggests that the oxygen content in GeOx is comparatively low in a thinner GeOx and increases as the GeOx thickness increases. Oxygen is relevantly deficient, and it results in the Ge dangling bonds the Ge/GeOx interface.

Fig. 2
figure 2

Ge 3d spectra of Ge/GeOx stack with GeOx thickness of a 4.1 Å, b 6.6 Å, and c 7.8 Å

Full size image
Fig. 3
figure 3

The areal intensity ratio of Ge(GeOx) vs. O(GeOx) for different GeOx thicknesses

Full size image

The composition changes of each chemical state, namely Ge1+, Ge2+, Ge3+, and Ge4+, is quantitatively calculated, as shown in Fig. 4a–d. The Ge1+ content is unaffected by the GeOx thickness. The Ge2+ content increases initially and then is unchanged when the GeOx thickness increases beyond after 4.1 Å. The Ge3+ content increases with increasing GeOx thickness. The Ge4+ signal is undetectable when the GeOx thickness is less than 8 Å and increases markedly when the GeOx thickness increases beyond 8 Å. The shaded area covering the GeOx thickness from 2 to 8 Å is the region of Dit reduction. Dit remains unchanged when the GeOx thickness is greater than 8 Å, as shown in Fig. 1.

Fig. 4
figure 4

Areal intensity of Ge chemical state of a Ge1+, b Ge2+, c Ge3+, and d Ge4+ with GeOx thickness. The shaded area is the region of Dit reduction

Full size image

We discuss the microstructure and chemical bonding states near the Ge/GeOx interface based on the above XPS results. Owning to the mismatch of Ge substrate and GeOx, the low oxidation states locate near the Ge substrate at the Ge/GeOx interface. The Ge1+ and Ge2+ states are localized mainly in the first and second Ge oxidation layers. This conclusion can be further confirmed as follows. The Ge1+ and Ge2+ signals are nearly unchanged when the GeOx thickness is greater than 4.1 Å, as shown in Fig. 4a, b. This value of 4.1 Å corresponds to oxidation of two Ge layers, which is explained as follows. The distance between Ge and O atom layers in GeOx (here we denote it as Ge–O) is ~ 1–1.6 Å [50, 51]. Thus, the GeOx thickness corresponding to the oxidation of two Ge layers, i.e., Ge–O–Ge–O, is ~ (1–1.6) × 3= ~ (3–4.8) Å. This value is consistent with the value of ~ 4.1 Å. As shown in Fig. 2a, the 4.1 Å GeOx only contains Ge1+ and Ge2+ states. Consequently, the Ge1+ and Ge2+ states are localized mainly in the first and second Ge oxidation layers. Moreover, the interfacial oxidation state and microstructure at the Ge/GeOx interface can be considered unchanged when GeOx thickness is beyond 4.1 Å, i.e., the distribution of Ge dangling bonds at the Ge/GeOx interface is unchanged during the process of ozone oxidation. While the Dit keeps to decrease when GeOx thickness is greater than 4.1 Å. These results suggest that the decrease of Dit at the Ge/GeOx interface does not arise from the passivation of Ge dangling bonds by additional oxygen atoms. Dit continues to decrease and the Ge3+ content keeps to increase as the GeOx thickness increases from 2 to 8 Å in Figs. 1 and 2. The relationship between Dit and the content ratio of Ge3+ in GeOx and Ge0 in Ge substrate is shown in Fig. 5. Dit decreases with increasing Ge3+ content, which means Ge3+ plays an important role in Dit passivation by ozone oxidation. Similar results have been reported for plasma oxidation [41]. Both for plasma and ozone oxidation, Dit similarly changes with increasing GeOx thickness. Therefore, Ge3+ plays a key role in the Dit passivation at the GeOx/Ge interface regardless of the oxidation method.

Fig. 5
figure 5

Dit as a function of the content ratio of Ge3+ in GeOx and Ge0 in Ge substrate

Full size image

One can propose another possible reason that is the mechanical stress exerted by different thicknesses of GeOx to induce Dit passivation. However, Dit decreases when GeOx thickness is less than 8 Å and is unchanged when GeOx thickness is larger than 8 Å, as shown in Fig. 1. Moreover, the Ge3+ content increases as the GeOx thickness increases from 2 to 8 Å and gradually stabilizes as the GeOx thickness is larger than 8 Å. While the stress at the GeOx/Ge interface still increases when GeOx thickness is larger than 8 Å. This is inconsistent with the change in Dit. To further investigate the effect of stress on Dit, the samples were treated in different PDA ambients to observe the changes of the gate stack and Dit. Figure 6 shows the high-resolution transmission electron microscope (HRTEM) images of the GeOx/Al2O3 structure without PDA and with PDA in N2 and O2 ambient. The thickness of the GeOx interlayer is nearly unchanged after PDA. The stress at the Ge/GeOx interface is also considered the same for different PDA samples. Figure 7 shows the energy-dispersive spectroscopy (EDS) depth profiles of Ge and O elements for Ge/GeOx/Al2O3 structure without PDA and with PDA in N2 and O2. PDA does not change the physical thickness of the gate dielectrics. The EDS results show that the depth profiles of the O and Ge elements are the same after PDA, indicating that the physical structure does not change. Therefore, the change in Dit does not arise from the physical structure or elemental interdiffusion during N2 and O2 PDA.

Fig. 6
figure 6

HRTEM images of the GeOx/Al2O3 structures without PDA and with PDA in N2 and O2 ambient

Full size image
Fig. 7
figure 7

EDS depth profiles for Ge/GeOx/Al2O3 structures without PDA and with PDA in N2 and O2 ambient

Full size image

Figure 8 shows the C–V characteristics of Ge MOS capacitors at multiple frequencies and the Dit for various PDA ambients. Well C–V plot indicates good interface quality. The Dit was obtained by the low-temperature conductance method. As shown in Fig. 8d, Dit slightly differs depending on the PDA treatment: without PDA, N2 PDA, and O2 PDA. O2 PDA is more beneficial for the Dit passivation than N2 PDA and without PDA condition. Although the GeOx thickness does not change, Dit distribution shows slight difference between the N2 and O2 PDA-treated samples. Therefore, the mechanical stress exerted by different thicknesses of GeOx as the origin of the Dit variation is excluded. Dit reduction at the Ge/GeOx interface does not arise from the passivation of dangling bonds by another atom or the stress from the different thicknesses of GeOx, but from the appearance of high Ge oxidation states. The contribution of high Ge oxidation state in GeOx to Dit reduction has been observed in other studies in which the GeOx was grown by H2O plasma oxidation in an ALD chamber [52]. A longer deposition time enhances the Ge oxidation state, which decreases the Dit.

Fig. 8
figure 8

 C–V plots of Ge MOS capacitors with GeOx/Al2O3 gate stack in different PDA ambients: a without PDA, b N2 and c O2. d Dit for different PDA ambients

Full size image

Figure 9 provides the XPS Ge 3d spectra of Ge/GeOx/Al2O3 in different PDA ambients. The corresponding intensity ratio of Ge1+, Ge2+, Ge3+, and Ge4+ in GeOx is shown in Fig. 10. The results show that the Ge3+ component in GeOx increases after O2 PDA while decreases after N2 PDA. Ge1+ component in GeOx has nearly no changes after PDA, indicating that the PDA has a negligible effect on Ge dangling bonds. Although the amount of Ge dangling bonds is similar, Dit still shows different distribution in N2 and O2 PDA as shown in Fig. 8d. The Ge2+ component in GeOx is similar between N2 and O2 PDA sample, indicating that Ge2+ is not the main sources of Dit passivation. The content of Ge4+ component decreases after PDA in N2 and O2 compared with the untreated sample. This changes of Ge4+ content after PDA is inconsistent with that of Dit. While the Ge3+ component in GeOx increases after O2 PDA and decreases after N2 PDA, which declares Ge3+ is responsible for the Dit passivation. The change of Dit distribution is related to the content of Ge3+ oxidation state component. This is consistent with the result for Dit dependence on GeOx thickness. Dit only decreases at the GeOx thickness range of 2 ~ 8 Å but remains unchanged with GeOx thickness larger than 8 Å. Moreover, the content of Ge3+ component obviously increases as the GeOx thickness increases from 2 to 8 Å, and gradually stabilizes as the GeOx thickness increases beyond 8 Å. These results demonstrate that the Ge3+ component is responsible for the Dit passivation.

Fig. 9
figure 9

Ge 3d spectra of Ge/GeOx stacks a without PDA, b with N2 PDA, and c with O2 PDA

Full size image
Fig. 10
figure 10

The intensity ratio of Ge1+, Ge2+, Ge3+, and Ge4+ in GeOx for the without PDA, and with N2 and O2 PDA samples

Full size image

To examine the above experimental results, we employed first-principles modeling to observe the remote passivation effect of Ge3+ on the gap state. The calculations were performed within density functional theory, as implemented in the code Vienna ab initio simulation package [53]. The projector augmented plane-wave method with Perdew–Burke–Ernzerhof was used [54]. The generalized gradient approximation was constructed for the exchange-correlation potential. A plane wave with 450 eV cut-off energy was used for the structural relaxation. All the structures were relaxed until the residual forces on the atoms declined to less than 0.02 eV/Å. The energy criterion in the iterative solution of the Kohn–Sham equation was set to be 10−5 eV. The interface region was modeled in a 20 Å vacuum spacing perpendicular to the slab to avoid interlaminar interactions. In this study, we mainly consider the changes of partial density of states (PDOS) by the introduction of Ge3+.

Figure 11a shows the atomic structure of GeOx with Ge1+ and Ge2+ at the Ge/GeOx interface and corresponding PDOS. The results show that Ge1+ induces defect states within the Ge band gap, while Ge2+ leaves no gap states within the Ge band gap. This means Ge1+, without oxidated sufficiently, at Ge/GeOx interface plays a key role in the Dit. Ge2+ has little effect on the Dit. In contrast, Ge3+ has the effect of removing the trap states of Ge1+ from the band gap energy range, as illustrated in Fig. 11b. The calculation result is consistent with the experimental result in that the Ge3+ reduces the trap states at the Ge/GeOx interface. In the study of investigating the dependence of Dit on GeOx thickness, Dit reduces as the GeOx thickness increases up to 8 Å. In this range of GeOx thickness, the Ge3+ content markedly increases, which passivates the trap states and reduces Dit at the Ge/GeOx interface. When GeOx thickness is larger than 8 Å, the Ge3+ content gradually saturates and Dit drops to a minimum and remains unchanged.

Fig. 11
figure 11

a Atomic structure of GeOx with Ge1+ and Ge2+ and PDOS at the Ge/GeOx interface. b Atomic structure of GeOx with Ge1+ and Ge3+ and PDOS at the Ge/GeOx interface

Full size image

We discuss a possible physical origin of Dit decrease at the Ge/GeOx interface by ozone oxidation. We consider the remote Coulomb potential perturbation from high oxidation state Ge atoms, based on the perturbation theory of quantum mechanisms. In general, the Dit originates from dangling bonds [19, 22], as shown in Fig. 12a, c. The appearance of additional Ge atoms with higher oxidation states induces a remote Coulomb potential near the dangling bonds, as show in Fig. 12b. This Coulomb potential perturbation can induce an eigen energy shift of the dangling bond based on the perturbation theory of quantum mechanisms, as shown in Fig. 12c. If this shift moves the eigen energy of the dangling bonds from the band gap into the valence band, then Dit is passivated. It should be noted that the dangling bond is passivated by the eigen energy shift into valence band induced by remote Coulomb charges, such as Ge3+, not by termination with another atom. The Dit passivation by Ge high oxidation state is verified by analyzing the samples produced under different thickness GeOx and different PDA ambient conditions. The GeOx thickness and PDA ambient affects the distribution of Ge3+ oxidation state in GeOx, which induces corresponding Dit variation. The perturbation theory of quantum mechanisms is a possibly reasonable explanation for Dit passivation.

Fig. 12
figure 12

Schematic of the Dit passivation mechanism of remote Coulomb perturbation. A dangling bond of the Ge substrate (a) induces the interfacial trap eigen energy in the Ge band gap (c). The appearance of the Ge atom with high oxidation state (b) induces a remote Coulomb potential perturbation near the dangling bond, shifting the eigen energy of the interfacial trap. This energy shift moves the interfacial trap from the band gap into the valence band, consequently passivating the Dit

Full size image

4 Conclusion

In this study, passivation of the Ge/GeOx interface by ozone oxidation is investigated, and a possible passivation mechanism is proposed. The experimental results indicate that the Dit passivation is related to the Ge3+ oxidation state in GeOx. Dit at the Ge/GeOx interface dramatically decreases and the Ge3+ content constantly increased as the GeOx thickness increased up to 8 Å. The effect of Ge3+ oxidation state is also demonstrated by different PDA ambients. The Ge3+ content differs after PDA and leads to different Dit distribution. The PDOS obtained by first-principles calculation with Ge3+ component in GeOx shows the removal of trap state from the band gap, which is in good agreement with the experiment results. The high Ge oxidation states may induce a remote Coulomb potential to move the dangling bond eigen energy from the band gap into the conduction or valance band to passivate Dit. This study provides another insight into the passivation mechanism at the semiconductor/oxide interface.