1. Introduction
Ge has attracted much attention as a potential channel material with high mobility to replace conventional Si and continue the scaling down of the metal–oxide–semiconductor field effect transistor (MOSFET). In spite of the higher bulk mobility of electron (3900 cm2/(V∙s)) and hole (1900 cm2/(V∙s)) than Si[1, 2], it still faces obstacles in obtaining superior mobility in practical devices because of the carrier mobility degradation. A high interface quality and optimized gate structure are necessary to boost the Ge MOSFET property. Ge surface passivation has been widely investigated including the nitride passivation layer[3-5], Si cap layer[6, 7], and Ge oxide layer[8-10]. Among all these surface passivation candidates, GeOx is promising as an interlayer and provides high interface quality. The interface state density Dit is at about 1010–1011 cm–2 eV–1 level[8, 10-12]. A proper GeOx passivation layer can be formed by different approaches, such as thermal oxidation[8], plasma treatment[9], and ozone oxidation[10]. Lee et al. reported a peak electron mobility of about 1100 cm2/(V∙s) using a 20 nm GeO2 layer[13]. However, a large equivalent oxide thickness (EOT) with a thick GeOx layer is not suitable for the requirement of scaled and high-performance devices. When GeOx thickness was reduced at about 1 nm, the apparent increase of Dit was observed. Zhang et al. has observed Dit increment from 1.7 × 1011 to 8 × 1011 cm–2 eV–1 at 0.2 eV below the intrinsic level when GeOx thickness decreased from 1.2 to 0.23 nm for Ge pMOSFET[14]. In addition, the hole mobility degraded, with thinner GeOx attributable to the stronger Coulomb scattering from Dit. By plasma post-oxidation, Zhang proposed a structure of HfO2/Al2O3/GeOx/Ge stack with an equivalent oxide thickness (EOT) of 0.7 nm[15]. A low interface state density about 1011 cm–2 eV–1 and promoted electron and hole peak mobility value of 754 and 596 cm2/(V∙s) were obtained.
Except for the Dit effect on the mobility degradation, surface roughness scattering[16, 17], oxidation atoms in the Ge substrate[18], and remote Coulomb scattering (RCS)[19, 20] from the gate stack are also reported to be possible factors causing mobility degradation. The surface roughness characterized by root mean square of the Ge surface has been reduced at 0.11 nm[21]. Further promoting the mobility by improving the surface roughness is quite limited. Out-diffusion of oxygen in the Ge substrate by hydrogen annealing is another way to increase the mobility[18]. Similar to the effect of Dit on mobility degradation, RCS from the charges in the Ge gate stack also plays an important role on mobility degradation[22-24]. For the Si MOSFET, different interface fixed charges were reported in the gate stack[25]. Si/interlayer and interlayer/high-k dielectric interface had the fixed charge density estimated to be 1011–1012 cm–2 level[25-27]. The dipole charge density at the SiO2/high-k dielectric interface is experimentally estimated at 1013 cm–2 level[26, 28, 29]. The mobility suffered severe degradation from the RCS induced by these fixed interface charges and dipole charges[23, 26, 30, 31]. The Ge MOSFET has a similar interlayer/high k dielectric gate stack structure. The interface fixed charge and dipole charge possibly exist in the Ge high k gate stack. The positive fixed charge was found at the GeO2/HfO2 interface about (2–3) × 1012 cm–2 in Bellenger’s report[32]. Deng et al. reported a fixed charge about 4.5 × 1012 cm–2 at the Ge/GeO2 interface and it could be reduced through O2 annealing[33]. The charge distribution of the Ge high-k stack still needs to be systematically characterized, especially for the dipole charge.
Energy band alignment of high k/Ge stack is a key parameter to affect the electrical characteristics, such as the gate leakage current and interface dipole. Al2O3 has been reported to serve as a barrier to restrain the oxygen diffusion and realize a thin GeOx layer and low EOT[14, 21]. The Ge MOSFET using the GeOx/Al2O3 structure usually showed high performance with enhanced mobility[14, 15, 21]. The band lineup of the Ge/Al2O3 and Ge/GeOx/Al2O3 stack is discussed in this paper. In addition, the interface passivation, gate charge distribution, and mobility degradation by RCS of the gate charges of Ge/GeOx/Al2O3 MOSFET are presented and reviewed from our previous work. The composition of the paper is organized as follows. First, the passivation mechanism based on ozone oxidation is described in Sec. 2. The effect of GeOx thickness formed by ozone oxidation on Dit is given. It is also a demonstration that whether the changing regular of Dit with GeOx thickness is the oxidation method related or a common feature regardless of the oxidation method. Second, the band alignment of the Al2O3/Ge stack with different Al2O3 or GeOx interlayer thickness is discussed in Sec. 3. Third, the charge distribution in the Al2O3/GeOx/Ge gate stack is characterized in Sec. 4. It is found that an interface fixed charge of about 1012 cm–2 exists as well as an interface dipole charge about 1013 cm–2 in the gate stack. The mobility degradation induced by the remote Coulomb scattering from the gate charges is discussed in Sec. 5. Sec. 6 is the part of the conclusion.
2. Ozone passivation mechanism
Obtaining a superior interface is a precondition to realizing a high-performance Ge MOSFET. The dependence of Dit on GeOx thickness was experimentally found both in post thermal oxidation and plasma oxidation. Kuzum et al. reported a suitable temperature control was beneficial to reduce Dit because of an increase of the Ge4+ composition oxidation state[10]. We discussed the passivation mechanism of GeOx based on ozone oxidation, and clarified whether it had a similar dependence relationship of Dit on GeOx thickness compared with thermal or plasma oxidation.
An attracting feature of ozone oxidation is that GeOx can grow at low temperature which effectively avoids GeOx desorption (>420 °C)[34, 35]. Fig. 1(a) shows the relationship between GeOx thickness and oxidation time for different deposition temperature. The temperature of ozone oxidation varied from 80 to 400 °C. The GeOx thickness is about 2.8, 5.28, 7.8, 9.7, and 12.3 Å after 25 min ozone oxidation at 80, 250, 300, 350, and 400 °C, respectively. The GeOx growth presents two different trends with initial linear and subsequent parabolic mode. The GeOx growth phenomenon observed here is similar to Si oxidation interpreted by the Deal-Grove mode[36]. The GeOx growth in the linear region is supposed to be determined by reaction rate at the Ge/GeOx interface. While it is limited by diffusion rate of oxygen through GeOx in the parabolic region. Fig. 1(b) shows the Arrhenius plot for the linear and parabolic region. The activation energy in the linear growth is calculated to be 0.06 eV, which is nearly consistent with that of Si oxidation[37-39]. Such small activation energy of Ge oxidation indicates a nearly barrier-less growth for the initial oxidation stage. The activation energy in the parabolic region is estimated to be 0.54 eV. Compared with thermal oxidation, the activation energy is much smaller in ozone oxidation which may have a higher reactivity of oxygen atoms to create the Ge–O–Ge bond.
The electrical property of the Ge/GeOx/Al2O3 gate stack by ozone oxidation is presented in Fig. 2. Fig. 2(a) shows the capacitance–voltage (C–V) property of W/TiN/Al2O3/GeOx/Ge MOS capacitor with 1.06 nm GeOx. Well-shaped C–V curves reflect a good Ge/GeOx interface. Dit is measured by the conduction method at low temperature and drew in Fig. 2(b). Fig. 2(b) shows the Dit apparently decreases with GeOx thickness increasing. The inset of Fig. 2(b) is the energy distribution of Dit with different thicknesses of the GeOx layer. When GeOx thickness is larger than 0.6 nm, Dit is nearly unchanged and reaches a minimum saturation value of about 3 × 1011 cm–2 eV–1. This result demonstrates that a GeOx thickness larger than 0.6 nm is necessary to ensure good passivation quality. Compared with thermal or plasma oxidation, ozone oxidation also induces the same Dit dependence trend on GeOx thickness. Therefore, the surface passivation of Ge using GeOx seems to have the same mechanism regardless of the oxidation method.
To understand the behavior of Dit decreasing with an increasing GeOx thickness, the oxidation state of Ge is analyzed by X-ray photoelectron spectroscopy (XPS) technology. Fig. 3 shows the XPS spectra of Ge 3d for GeOx/Ge for different GeOx thickness. In order to quantitatively evaluate each oxidation state of the Ge, the Ge 3d spectra are divided by five peaks related to the Ge substrate (Ge0), Ge suboxide (Ge1+, Ge2+, Ge3+) and GeO2 (Ge4+) components. The chemical shifts of Ge1+, Ge2+, Ge3+ and Ge4+ relative to the Ge0 are taken as 0.8, 1.8, 2.75 and 3.4 eV, respectively[40-43]. It can be seen that the peak corresponding to the GeOx increases with thicker GeOx, and the GeOx peak energy shifts toward higher binding energy, suggesting that the oxidation state of Ge and the GeOx/Ge interfacial structure are dependent on the GeOx thickness.
The peak area intensity ratio of Ge1+, Ge2+, Ge3+, and Ge4+ component to Ge0 component as a function of GeOx thickness is shown in Fig. 4(a). The Ge1+ component has an extremely small change with GeOx thickness. The Ge2+ component has a slight increase when GeOx thickness is below 0.4 nm. While it is nearly unchanged when GeOx thickness is larger than 0.4 nm. It can be seen from Fig. 2(b) that Dit still obviously decreases for GeOx thickness above 0.4 nm. At the same time, the Ge4+ component is not detected for the GeOx thickness of less than 0.52 nm. The Ge3+ component shows a monotonical increase with increasing GeOx thickness. Therefore, Ge1+, Ge2+, and Ge4+ components are less relevant with Dit passivation. The Ge3+ component increment with GeOx thickness plays a main role on passivating Dit. Fig. 4(b) shows the area intensity ratio values of Ge2+/Ge1+, Ge3+/Ge1+, Ge4+/Ge1+, and Ge3+/Ge2+ as a function of takeoff angle for 1.06 nm GeOx. The values of Ge2+/Ge1+, Ge3+/Ge1+ and Ge4+/Ge1+ decrease with an increased takeoff angle. It reflects the Ge2+, Ge3+, and Ge4+ components are located in the upper of Ge1+. The value of Ge3+/Ge2+ shows no variation when the takeoff angle changes from 35° to 90°, which means that the Ge3+ and Ge2+ are uniformly mixed. From the above discussion, it can be concluded that a suboxide of Ge composed of Ge1+, Ge2+, and Ge3+ is formed. The Ge3+ component is beneficial to passivate the Ge surface and decrease Dit.
3. Band alignment of Ge/Al2O3 stack
Energy band alignment at the hetero-interface has always been an attractive and intensive topic in semiconductor science and industry. To improve the injection current and further scale the supply voltage, the high k/Ge stack is very advantageous and preferable[44]. One of the key parameters is the energy band alignment which is related with the gate leakage current and interfacial effect such as dipole and charges[45, 46]. The valence band offset (VBO) has been found to be inconsistent among different research experiments and varied in the range of 3.17–3.4 eV[47-49]. The photoemission spectroscopy method measures the signals a few nanometers underneath the surface due to the exponential attenuation of photoelectrons and thus the obtained spectrum is a weighted average of the detected signals but not the signal just on the surface[50]. Therefore, the peak position will be affected by the electric potential distribution across the Al2O3/Ge stack induced by surface or bulk charges. In addition, the influence of an interlayer on band alignment has been investigated at the hetero-junction such as HfO2/Si, HfO2/Ge, HfO2/GaN, and TiO2/Si etc.[51-54]. For the Al2O3/Ge stack, the effect of the Al2O3 and interlayer GeOx thickness on the band alignment is described by the XPS measurement technique.
3.1 Effect of Al2O3 thickness on band offset
The VBO at the Al2O3/Ge interface is described using the method proposed by Kraut et al.[55]. The expression of VBO is shown as follows,
ΔEv=(EAl,2p−EGe,3d)Al2O3/Ge−(EAl,2p−VBMAl2O3)Al2O3+(EGe,3d−VBMGe)Ge, | (1) |
where the (EAl,2p − EGe,3d
The XPS measurements of core-level spectra for different Al2O3 thicknesses are given in Fig. 6. To compare the difference of binding energy between Ge 3d and Al 2p, the peak position of Ge 3d core-level spectra is fixed at the same binding energy. It can be seen that the Al 2p to Ge 3d separation gradually increases with larger Al2O3 thickness. The VBO at Al2O3/Ge interface can be calculated based on Eq. (1). Fig. 7 shows the dependence of VBO on Al2O3 thickness. The measured band offset at the Al2O3/Ge interface increases with Al2O3 thickness increasing. It indicates that the electrostatic potential exists across the Al2O3 film and induces energy band bending. The core level spectrum, which is the weighted average of the detected photoemission, reflects the potential at a position displaced away from the center of the film toward the outer surface.
The gap state model is widely accepted to explain the physical nature of band alignment at the hetero-junction[56-58]. When the charge neutrality level (CNL) of the gap states in two adjacent materials is not consistent, electrons then flow from one material with higher CNL to the other with lower CNL. An electric dipole forms to align their Fermi level during the process. The effect of surface gap states on the Al2O3 surface is considered to interpret the phenomenon of VBO changing with Al2O3 thickness. Due to the hetero-structure at the air/Al2O3 interface, the solutions of Schrödinger's equation with complex wave vectors become of physical relevance for energies within band gaps, resulting in gap states on the Al2O3 surface. CNL is a defined energy level at which the characteristics of these gap states change from predominately donor- to acceptor-like closer to the valence band top and the conduction band bottom, respectively.
Fig. 8 is the schematic of the energy band before and after Al2O3 and Ge contact.
3.2 Effect of GeOx thickness on band offset
The gap states model was also validated by inserting an interlayer at the hetero-junction. The different thickness GeOx is introduced to investigate the band alignment at the Al2O3/Ge interface. Al2O3 thickness remain at 2 nm. GeOx thickness varied from 0.2 to 1.2 nm. Fig. 9 shows the dependence of VBO at the Al2O3/Ge interface on the GeOx thickness. It is found that the VBO is unacted on GeOx thickness. As discussed in Sec. 3.1, the band alignment at the Ge/Al2O3 interface is driven by a matching of CNL at the adjacent materials. When introducing an interlayer, it modifies the matching of CNL by a Schottky barrier pining parameters S[59-61], which is related with the thickness, relative permittivity of GeOx and the Dit. For a GeOx interlayer with thickness of 0.2–1.2 nm, the calculated result just coincides with the experimental data as shown in Fig. 9. The VBO is not affected by GeOx thickness. The gap states model is successfully employed to explain the band alignment.
4. Interface charge in Ge/GeOx/Al2O3 gate stack
4.1 Charge distribution in gate stack
The interfacial charges in the gate stack is one of the critical issues to enhance the performance of the Ge MOSFET. These charges significantly degrade the electrical characteristics such as threshold voltage shifts[62], reliability[63], and mobility degradation[14, 17]. Although the high electron and hole bulk mobility, the mobility of the Ge MOSFET still suffers severe degradation. The remote Coulomb scattering from gate charge plays a key role on mobility degradation. Therefore, it is very instructive to quantitatively characterize the charge distribution in the gate stack of Ge MOSFET. The charge distribution is obtained by employing the relationship between flatband voltage (VFB) and EOT. The VFB and EOT relationship is measured through a Ge/GeOx/Al2O3 capacitor combining different GeOx and Al2O3 thickness as shown in Fig. 10. Figs. 10(a) and 10(b) show the VFB and EOT relationship for capacitors with varied GeOx and Al2O3 thickness, respectively. The theoretic expression for structure in Fig. 10(a) is shown as
VFB=−Q1ε0εrEOT−ε1ρ12ε0ε2rEOT2−Q2ε0εrEOT2+(ε1ρ1−ε2ρ2)2ε0ε2rEOT22+Δ+ϕms, | (2) |
where Q1 is the areal charge at the Ge/GeOx interface, Q2 is the areal charge at the GeOx/Al2O3 interface, ρ1 is bulk charge density in GeOx, ρ2 is bulk charge density in Al2O3, ε0 is the vacuum permittivity, εr is the relative permittivity of SiO2, ε1 is the relative permittivity of GeOx, ε2 is the relative permittivity of Al2O3, EOT is the equivalent oxide thickness of the whole gate stacks, EOT2 is the equivalent oxide thickness of the Al2O3 dielectric, Δ means VFB shift due to the electric dipole at the GeOx/Al2O3 interface, and ϕms is the vacuum work function difference between the Al gate electrode and Ge substrate.
The theoretic expression for structure in Fig. 10(b) is shown as
VFB=−Q1+Q2ε0εrEOT+ε2ρ2EOT1ε0ε2rEOT−ε2ρ22ε0ε2rEOT2+Q2EOT1ε0εr−ε2ρ22ε0ε2rEOT21+Δ+ϕms, | (3) |
where EOT1 is the equivalent oxide thickness of the GeOx dielectric. The linear fitting is obtained, as shown in Figs. 10(a) and 10(b). Then the fixed charges Q1 at the Ge/GeOx interface, fixed charges Q2 at the GeOx/Al2O3 interface and dipole charge at the GeOx/Al2O3 interface can be extracted from the fitting line combining the Eqs. (2) and (3). The space distribution of the interface fixed charge and dipole is shown in Fig. 11(a). Fig. 11(b) gives the charge density of these charges. The fixed charge Q1 at the Ge/GeOx interface, fixed charge Q2 at the GeOx/Al2O3 interface, dipole at the GeOx/Al2O3 interface are estimated to be 7 × 1012, –3.06 × 1012, and 1.98 × 1013 cm–2, respectively. The density of the dipole charge is larger than the fixed charge.
4.2 Origin of gate charge
To effectively modulate the charge distribution in the gate stack, understanding the origin of interface fixed charge and dipole remains a necessity. As discussed in Sec. 2, the Ge suboxide is formed at the initial stage of ozone oxidation. The Ge4+ oxidation state is not detected when the GeOx thickness is less than 0.52 nm. In other words, the Ge surface is not sufficiently oxidized. In addition, we investigated the areal intensity ration of Ge and O atom in different thicknesses of GeOx. The experimental result shows that the ratio of Ge/O decreases in a thicker GeOx. It also demonstrates that oxygen is relatively deficient at the beginning of ozone oxidation and results in Ge2O formation at the Ge/GeOx interface. Oxygen deficiency results in oxygen vacancy at the Ge/GeOx interface. To confirm the oxygen vacancy, post-deposition annealing in oxygen atmosphere is used to investigate the interface property. Fig. 12(a) compares the VFB and EOT relationship for Ge/GeOx/Al2O3 capacitors with and without O2 annealing. The annealing temperature is 400 °C to avoid the GeOx decomposition. The slope of the VFB–EOT fitting line after O2 annealing is smaller than that without O2 annealing. The slope described as Eq. (1) represents the charge density at the Ge/GeOx interface. The specific charge density at Ge/GeOx interface extracted is shown as Fig. 12(b). The positive charge density at Ge/GeOx interface is 7.34 × 1012 and 2.37 × 1012 cm–2 for the as-grown and O2-annealed samples, respectively. O2 annealing significantly decreases the fixed charge density at the Ge/GeOx interface. Extra oxygen is introduced to passivate the oxygen vacancy during the O2 PDA.
As for the origin of fixed charges at the GeOx/Al2O3 interface, the property of Al2O3 with different thickness is investigated by XPS measurement. Fig. 13 gives the intensity ratio of aluminum to oxygen for different thicknesses of Al2O3. An increasing ratio of aluminum to oxygen content is obtained with Al2O3 thickness increasing. This means aluminum is relatively deficient at the initial stage of Al2O3 deposition and an oxygen-rich area exists at the GeOx/Al2O3 interface. The high oxygen content results in oxygen dangling bond at the GeOx/Al2O3 interface. Based on a first-principles calculation, the transition level of oxygen dangling bonds is located at 0.9 eV above the valence band maximum (VBM) of Al2O3[64]. The VBO between Ge and Al2O3 was reported about 3.1–3.4 eV[47, 48, 65]. The transition level of oxygen dangling bond is localized below Ge VBM and negatively charged. This is consistent with the experimental result of negative charge extracted at the GeOx/Al2O3 interface. Therefore, it is rational to conclude that the oxygen dangling bond results in the negative charge at the GeOx/Al2O3 interface.
The dangling bond can be passivated by combining other atoms and removing the transition level from the energy gap. It has been observed that oxygen annealing can passivate the oxygen vacancy or Ge dangling bond and decrease the fixed charge density at the Ge/GeOx interface. To investigate the change of fixed charge at Ge/GeOx and GeOx/Al2O3 interface, N2, O2, NH3 and H2 ambients are used to anneal the Ge/GeOx/Al2O3 capacitor samples. Figs. 14(a) and 14(b) show fixed charge density at the Ge/GeOx and GeOx/Al2O3 interface for the samples without and with PDA in N2, O2, NH3, and H2 ambient. It can be seen that the fixed charges at the Ge/GeOx and GeOx/Al2O3 interfaces can be obviously changed by PDA in different ambients. Both the quantity and polarity of these interfacial charges can be modulated. O2 and NH3 PDA is beneficial to suppress the interfacial charges. Figs. 14(c) and 14(d) are dipole distribution and corresponding areal charge density of dipole at the GeOx/Al2O3 interface for the samples with and without PDA. The results show that a positive dipole is obtained only for the as-grown sample. The polarity of the dipole changes to negative after PDA.
To understand the change origin of dipole after different ambient PDA, the XPS measurement was employed to analyse the band structure of the Ge/GeOx/Al2O3 stack. Fig. 15 depicts the Ge 3d and Al 2p spectra for the Ge/GeOx/Al2O3 structure with and without PDA. Ge 3d has two peak positions: the higher peak position from Ge substrate is located at 29.6 eV, and the other from GeOx is near 32.6 eV. The energy scale is calibrated by setting the binding energy of the Ge substrate to be 29.6 eV. The peak position of Ge 3d from GeOx and Al 2p from Al2O3 shows a left shift for the samples after PDA in N2, O2, H2, and NH3 ambients. The shift extent gradually increases for the N2, O2, H2, and NH3 PDA. We have observed a positive dipole for the sample without PDA and an increasing negative dipole in N2, O2, H2, and NH3 PDA. The XPS measurement result and electrical dipole extraction are very consistent. Because of the dipole at the GeOx/Al2O3 interface, it will lead to the band bending of GeOx and Al2O3 at the interface and a shift of core level of GeOx and Al2O3.
Fig. 16 is the energy structure of Ge/GeOx/Al2O3 stack without a dipole, with a positive, and with a negative dipole at the GeOx/Al2O3 interface. δ(
The CNL theory is still valid to explain the origin of the positive and negative dipole at the GeOx/Al2O3 interface. The CNL plays a similar role to the Fermi level. If the CNL in two adjacent materials is not consistent, the electron will transfer from one material with higher CNL to the other and it induces an electrical dipole at the interface. The positive dipole is obtained at the GeOx/Al2O3 interface for the sample without PDA. It means CNL in Al2O3 is higher than that in GeOx. Electron flows from Al2O3 to GeOx, causing negative charge accumulating on the GeOx side and equal positive charge on the Al2O3 side. Therefore, a positive dipole forms at the GeOx/Al2O3 interface. The negative dipole is obtained at the GeOx/Al2O3 interface after PDA, indicating the CNL in Al2O3 is lower than that in GeOx. Electron transfers from GeOx to Al2O3, which induces a negative dipole at the GeOx/Al2O3 interface. The larger the CNL difference between Al2O3 and GeOx, the more electrons will transfer and induce more dipole density. Therefore, the core level of Ge 3d and Al 2p shifts more with a larger dipole.
5. Remote Coulomb scattering of gate charge
We have analyzed the charge distribution in the gate stack in Sec. 3. The effect of gate charges on mobility property is discussed in the following content. Fig. 17 gives the Id–Vg characteristics and hole mobility of the Ge pMOSFET with different GeOx thickness by ozone oxidation measured at 77 K. It can be seen that Id and mobility decrease when the GeOx thickness is thinner. This phenomenon of mobility degradation will be discussed from the viewpoint of the scattering mechanism mainly including the Coulomb scattering, phonon scattering and surface roughness scattering. The root mean square (RMS) is experimentally determined to be identical (~0.09 nm) for different GeOx thickness[20]. In addition, the mobility is measured at the same temperature of 77 K. The phonon scattering is sufficiently suppressed and can be ignored. Therefore, the surface roughness scattering and phonon scattering are excluded to induce the mobility degradation for a thinner GeOx. The Coulomb scattering is explored to explain the mobility dependence on GeOx thickness.
In Sec. 3, we have discussed the distribution of gate charges. There are three kinds of charges in the Ge/GeOx/Al2O3 gate stack, including Ge/GeOx interface charge Q1, GeOx/Al2O3 interface charge Q2, and interface dipole. It is found that the dipole charge at the GeOx/Al2O3 interface is one order of magnitude larger than the fixed charge at the Ge/GeOx and GeOx/Al2O3 interface. Fig. 17 shows the mobility decreases with the GeOx thickness decreasing. The fixed charge at the Ge/GeOx interface has the same effect on mobility of the Ge pMOSFET with different GeOx thickness. The mobility degradation is from the charges at the GeOx/Al2O3 interface. The Coulomb scattering from the fixed charge and the dipole at the GeOx/Al2O3 interface increases with GeOx thickness decline and induces mobility degradation. The charge density of the dipole is higher than the fixed charge at the GeOx/Al2O3 interface. Therefore, the remote dipole scattering plays a main role in mobility degradation dependent on the GeOx thickness.
The gate charges can be modulated in different PDA ambients. To further certify the Coulomb scattering from the gate charge, the hole mobility extracted from Ge pMOSFET in different PDA ambients is shown in Fig. 18. The hole mobility shows obvious variation among different PDA ambients. N2 ambient increases the hole mobility, while the NH3 and H2 ambients decrease the hole mobility compared to mobility without PDA. The substrate material, gate stack structure, and measurement temperature for different PDA samples are all the same as those used in this study. In addition, the high-resolution transmission electron microscope (HRTEM) images shows no changes at the interface and film thickness[66]. Therefore, the phonon scattering and surface roughness is excluded to affect the mobility at different PDA ambients. The Coulomb scattering is responsible for the mobility variation after PDA. Dit has been extracted and demonstrated not to be affected by different PDA ambients. All the samples have almost the same interface quality and Dit[66]. So, the hole mobility affected by Dit is also excluded. The reason for mobility changes now points to the interface charge and dipole modulated by PDA.
First, the Ge/GeOx interface Q1 is sufficiently decreased in N2, O2, and NH3 PDA shown as in Fig. 14(a). The Coulomb scattering from Q1 is greatly decreased. However, the mobility only shows an increase in N2 ambient. NH3 and O2 PDA decrease the hole mobility. It demonstrates that there are other charges in the gate stack degrading the mobility more severely than the Ge/GeOx interface charge Q1 in NH3 and O2 PDA. Second, we consider the fixed charges and dipole at the GeOx/Al2O3 interface. The GeOx/Al2O3 interface charge Q2 is significantly decreased but still does not help in enhancing the hole mobility in O2 and NH3 ambients. In contrast, the dipole at the GeOx/Al2O3 interface increases in the O2 and NH3 ambients shown in Fig. 14(d). The dipole is one order of magnitude larger than fixed charge at the GeOx/Al2O3 interface, indicating that the dipole degrades the hole mobility much more severely than the GeOx/Al2O3 interface charge Q2. The charge density of dipole in NH3 PDA is larger than that in O2 shown in Fig. 14(d). At the same time, we observe that the mobility degrades more severely in NH3 PDA than O2 PDA. N2 PDA can increase the hole mobility because of the decrease of Ge/GeOx interface charge. O2 and NH3 PDA decrease the hole mobility because of the joint influence of the Ge/GeOx interface charge and the GeOx/Al2O3 interface dipole. Although a decrease in fixed charge at the Ge/GeOx, the mobility still shows a decrease due to the enormous increase of the dipole in O2 and NH3 ambients. A sketch of the remote Coulomb from the gate charge is shown in Fig. 19. The above analysis shows the mobility degradation due to remote Coulomb scattering is mainly from the Ge/GeOx interface charge and GeOx/Al2O3 interface dipole.
Fig. 20 shows the gate leakage current for the Ge/0.7 nm GeOx/4 nm Al2O3/Al gate stack in different PDA ambients. It can be seen that the leakage current is obviously decreased after PDA in different ambients. The leakage current reduces about one order of magnitude in N2 PDA compared with the sample without PDA. NH3 and O2 ambients are more helpful to decrease the leakage current. This may be due to the effective passivation of the interface defect at the Ge/GeOx and GeOx/Al2O3 interface. Considering the property of mobility and gate leakage current, N2 PDA is more suitable for the performance improvement.
6. Conclusion
The Ge has been examined to be a very promising material with high bulk mobility to integrate into Si CMOS technology. The interface passivation, band alignment, gate charge distribution, and mobility property with Ge/GeOx/Al2O3 stack by ozone oxidation are reviewed and discussed. Ozone oxidation is an effective way to obtain high Ge/GeOx interface. The band alignment of the Al2O3/Ge structure is well analyzed by gap states and CNL theory. The remote Coulomb scattering from gate charges plays a key role in mobility degradation. By extracting and modulating the charge distribution by PDA in various ambients, it is found that N2 is a very proper choice to enhance the hole mobility and gate leakage current of Ge pMOSFET. The study may provide a guidance on the performance improvement of the Ge MOSFET.
Acknowledgements
This work is supported by the Natural Science Foundation of Beijing Municipality (No. 4214079).