Ge0.83Sn0.17 P-Channel Metal-Oxide-Semiconductor Field- Effect Transistors: Impact of Sulfur Passivation on Gate Stack Quality

The effect of room temperature sulfur passivation of the surface of Ge0.83Sn0.17 prior to high-k dielectric (HfO2) deposition is investigated. X-ray photoelectron spectroscopy (XPS) was used to examine the chemical bonding at the interface of HfO2 and Ge0.83Sn0.17. Sulfur passivation is found to be effective in suppressing the formation of both Ge oxides and Sn oxides. A comparison of XPS results for sulfur-passivated and non-passivated Ge0.83Sn0.17 samples shows that sulfur passivation of the GeSn surface could also suppress the surface segregation of Sn atoms. In addition, sulfur passivation reduces the interface trap density Dit at the high-k dielectric/Ge0.83Sn0.17 interface from the valence band edge to the midgap of Ge0.83Sn0.17, as compared with a non-passivated control. The impact of the improved Dit is demonstrated in Ge0.83Sn0.17 p-channel metal-oxide-semiconductor field-effect transistors (p-MOSFETs). Ge0.83Sn0.17 p-MOSFETs with sulfur passivation show improved subthreshold swing S, intrinsic t...

Theoretical calculation [11] shows that the light hole effective mass of GeSn decreases with increasing Sn composition. It is also demonstrated experimentally [19] that increasing Sn composition in GeSn p-MOSFETs increases the effective hole mobility. However, due to the low surface energy and large covalent radius of Sn, Sn atoms may segregate to the surface during growth of GeSn [20][21][22]. Hence, the thermal stability may be worse at a higher Sn composition. Li et al. reported that a Sn-rich surface layer would form when Ge 0.922 Sn 0.078 is annealed at 620°C [23]. A similar phenomenon occurs on Ge 0.915 Sn 0.085 surface after annealing at 500°C [24]. For Sn composition as high as 17%, self-assembled Sn wires can form at an annealing temperature as low as 280°C [25]. Severe Sn segregation may reduce carrier mobility and degrade the drive current in MOSFETs. Therefore, in order to achieve high performance GeSn p-MOSFETs with high Sn composition, a fabrication process with low thermal budget may be required to maintain a good quality of the GeSn channel material and the GeSn/high-k dielectric interface.
Various passivation techniques have been demonstrated to be effective in improving the gate stack quality of both Ge and GeSn channel p-MOSFETs, such as Si 2 H 6 passivation [12,17], Ge capping [26], GeSnO x passivation [27,28], and sulfur passivation [15,29,30]. Among these passivation techniques adopted for GeSn p-MOSFETs fabrication, Si 2 H 6 passivation and GeSnO x passivation require a process temperature higher than 370°C and 400°C, respectively. It has already been reported that Sn can segregate out to the GeSn surface during Si passivation process at a temperature of 370°C and degrades the device performance [17]. Therefore, low temperature passivation technique was investigated in this work for the fabrication of GeSn p-MOSFETs with Sn composition of 17%.
The adsorption of sulfur atoms is a promising route to chemically and electrically passivate the highly reactive Ge and GeSn surface [15,[29][30][31]. Compared with other passivation techniques, room temperature sulfur passivation using (NH 4 ) 2 S solution has several advantages: (1) GeSn surface could be effectively passivated through the formation of covalent S-Ge and S-Sn bond. This will reduce oxide formation which degrades device performance; (2) The formed sulfur passivation layer is very thin with very little increase on the effective oxide thickness (EOT); (3) Sn segregation can be suppressed during the passivation process owning to a lower thermal budget. Sulfur passivation has already been implemented into Ge 0.947 Sn 0.053 p-MOSFETs fabrication and demonstrated enhanced peak hole mobility as compared with Si 2 H 6 passivation [15]. However, the mechanism of the effect of sulfur passivation on the GeSn/HfO 2 interface quality has not been investigated. In addition, the impact of sulfur passivation on the reduction of D it was not quantified.
In this chapter, sulfur passivation of GeSn surface at room temperature was investigated and implemented in the fabrication of Ge 0.83 Sn 0.17 p-channel MOSFETs. To study the impact of sulfur passivation on the quality of high-k dielectric/GeSn interface, extensive X-ray photoelectron spectroscopy (XPS) analysis was carried out. Sulfur passivation is found to be effective in suppressing the formation of Sn oxides and Ge oxides, and Sn surface segregation. In addition, sulfur passivation helps to reduce the high-k dielectric/GeSn interface trap density D it as extracted using the conductance method. Material study of nickel stanogermanide [Ni(GeSn)] contact formation at low temperatures was also performed for low resistivity [Ni(GeSn)] S/D contact. The sulfur-passivated Ge 0.83 Sn 0.17 p-MOSFETs exhibit smaller subthreshold swing S, higher intrinsic transconductance G m,int , and higher effective hole mobility μ eff as compared to the non-passivated control. At a high inversion carrier density of 1 × 10 13 cm −2 , sulfur passivation enhances μ eff by 25% as compared with the non-passivated control.

Material growth and characterization of Ge 0.83 Sn 0.17 substrate
The high quality Ge 0.83 Sn 0.17 sample was grown using molecular beam epitaxy (MBE). 4-inch (001)-oriented Ge wafers with n-type doping concentration of 5 × 10 16 cm −3 were used as the starting substrates. After the cyclic cleaning of Ge substrates using dilute hydrofluoric acid (DHF) (HF:H 2 O = 1:50) and deionized (DI) water, the unintentionally p-type doped Ge 0.83 Sn 0.17 film was grown on the Ge substrate using the solid source low temperature MBE system [32,33]. The growth temperature was set at 100°C. 99.9999% pure Ge and 99.9999% pure Sn were used as Ge and Sn sources, respectively. The growth chamber has a base pressure of 3 × 10 −10 Torr. Ge 0.83 Sn 0.17 film with the thickness of 10 nm was grown on the Ge substrates. Figure 1 shows the 5 × 5 μm AFM scan of the as-grown Ge 0.83 Sn 0.17 surface. The surface is very smooth with a root-mean-square (RMS) roughness as small as 0.198 nm. High-resolution transmission electron microscopy (HRTEM) was employed to analyze the crystalline quality of the as-grown Ge 0.83 Sn 0.17 sample. High resolution X-ray diffraction (HRXRD) was also used to analyze the Sn composition and strain property of the as-grown Ge 0.83 Sn 0.17 substrate. Figure 3(a) shows the (004) ω-2θ rocking scan of the as-grown sample. Both Ge and GeSn peaks are well-defined. The peak at smaller 2θ value is the GeSn peak. The relative broad full-width-half-maximum (FWHM) is due to the thin GeSn layer thickness (~10 nm). (115) reciprocal space mapping (RSM) of an as-grown Ge 0.83 Sn 0.17 / Ge (001) sample is shown in Figure 3(b). The GeSn film is fully strained to the Ge substrate and the substitutional Sn composition is calculated to be 17% from XRD measurement.  introduced: one with 10 minutes sulfur passivation using (NH 4 ) 2 S solution (24% by weight) at room temperature (25°C) and the other one without sulfur passivation. After that, the samples were loaded into the atomic layer deposition (ALD) chamber immediately to avoid surface oxidation due to air exposure. Surface treatment was done using Trimethylaluminum (TMA) as precursor with pulse duration of 30 ms. This was followed by deposition of 3 nm-thick hafnium dioxide (HfO 2 ) at 250°C using Tetrakis (dimethylamido) hafnium and H 2 O as precursors. The total ALD process duration including the pumping and venting steps is ~ 15 min. After that, 110 nm-thick tantalum nitride (TaN) was deposited using a reactive sputtering system. The metal gate was then patterned by photolithography and etched using chlorine (Cl 2 )-based plasma. A 10 nm-thick nickel (Ni) was then deposited using e-beam evaporator and the self-aligned Ni(GeSn) metallic contact was formed by rapid thermal annealing (RTA) at 250°C for 30 s in the N 2 ambient. Finally, excess Ni was removed by selective wet etch using concentrated sulfuric acid (H 2 SO 4 ) (96% by weight). The maximum processing temperature of

Sulfur-passivated gate stack study
The (001) surface of a diamond-structure semiconductor has two dangling bonds per surface atom. GeSn grown on Ge (001) surface has a (001) surface as shown in the atomic structure in Figure 5(a) viewed into the [110] direction. One monolayer (ML) of a Group VI element can passivate all the dangling bonds by occupying the bridge site in a (1 × 1) geometry [34,35]. Sulfur atoms could obtain an ideal (1 × 1) termination of the bivalent (001) surfaces of silicon and Ge. Although sulfur could desorb from the Si surface at room temperature or diffuse into the Si bulk during heating [36], Weser et al. found that an ordered (1 × 1) structure with one sulfur atom bonded on a bulk-like bridge site could be achieved by introducing elemental sulfur atoms on the Ge (001) surface under ultrahigh vacuum (UHV) condition [34,35]. The formation of Ge-S-Ge bridge bonds after a treatment in (NH 4 ) 2 S solution has also been reported based on various characterization techniques, such as photoelectron spectroscopy [37], ion scattering spectroscopy [38], as well as X-ray standing wave measurements [39]. Similarly, sulfur passivation should also be able to passivate the GeSn (001) surface through the formation of S-Ge and S-Sn covalent bonds which suppress the formation of Ge and Sn oxides at the surface, as illustrated in the atomic schematic shown in Figure 5(b). In this Section, the effectiveness of sulfur passivation on the gate stack of GeSn p-MOSFETs is investigated using XPS. The interface trap density value is also extracted using conductance method and compared with the non-passivated control.

XPS study on the effect of sulfur passivation
To investigate the interfacial property between the high-k dielectric and Ge 0.83 Sn 0.17 after sulfur passivation, XPS measurement was carried out to study the change of the interfacial chemical bonds. Two blanket Ge 0.83 Sn 0.17 samples were prepared for the measurement. After the cyclic DHF (1:50) cleaning, one of the sample went through 10 minutes aqueous (NH 4 ) 2 S solution (24% by weight) and the other one did not. After that, an ultra-thin (~1 nm) HfO 2 layer was deposited by ALD on these two samples. The HfO 2 layer thickness should be smaller than the XPS information depth [40]. XPS characterization was then performed using a VG ESCALAB 220i-XL imaging XPS system. Monochromatic aluminum (Al) Kα X-ray (1486.7 eV) was used to obtain the core level spectra of these samples. Binding energy was calibrated with standard samples for some pure metals. The binding energy of Carbon (C) 1 s from adventitious hydrocarbon surface contamination was set at 285.0 eV for further charge correction.
In order to confirm the incorporation of S after the (NH 4 ) 2 S passivation, core level XPS spectra of the S 2p peak were captured for HfO 2 -capped Ge 0.83 Sn 0.17 blanket samples with and without sulfur passivation, as shown in Figure 6(a). The black curve represents the S 2p signal obtained from the non-passivated Ge 0.83 Sn 0.17 sample. The circles are the raw data points obtained from the sulfur-passivated sample. Gaussian and Lorentzian line shapes with a Shirley background subtraction were used to fit the raw data. The blue line is the overall fitting of the core level spectra and the gray lines are the fitted peak components. For the S 2p core level spectra, the well-resolved two peaks correspond to S 2p 1/2 (163.4 ± 0.02 eV) and S 2p 3/2 (162.0 ± 0.02 eV) [41]. The S 2p signal obtained from the sulfur-passivated Ge 0.83 Sn 0.17 sample indicates that S is introduced onto the GeSn surface by the (NH 4 ) 2 S treatment and is still present after the deposition of HfO 2 . Figure 6(b) shows N 1 s core level spectra for both samples. The N 1 s peak (399.0 ± 0.02 eV) [41] on the sulfur-passivated sample is not observed, indicating that nitrogen is not incorporated. The sulfur passivation layer thickness is calculated using two different photoelectron peaks of the same element at kinetic energies E 1 and E 2 (E 1 and E 2 have sufficiently large differences in λ) [42,43]. Ge 3d and Ge 2p 3 data (not shown) are used and the equation is shown below: where λ A (E 1 ) and λ A (E 2 ) are the attenuation length of Ge 3d and Ge 2p 3 , I' B (E 1 ) and I' B (E 2 ) are corrected photoelectron intensities, and θ is take-off angle. The sulfur passivation layer thickness is calculated to be 0.44 nm using the Ge 3d and Ge 2p 3 XPS signals.  The circles, blue lines and gray lines are the raw data points, the overall fitting curves, and the S 2p 1/2 or S 2p 3/2 peak components, respectively. The black curve represents the S 2p signal obtained from the non-passivated sample. (b) N 1s core level spectra for both samples.
This could generate a huge amount of interface states which degrade the gate stack quality [45]. The Sn oxide could also be detrimental to the GeSn gate stack as SnO 2 is known to exhibit metallic behaviour, which leads to high gate leakage current [46,47]. Therefore, suppressing the Ge and Sn oxides formation is important for achieving good gate quality for Ge 0.83 Sn 0.17 p-MOSFETs.
To quantify the impact of sulfur passivation on Ge oxides and SnO x formation at the high-k/ GeSn interface, angle-resolved XPS (ARXPS) was performed. Both SnO x and Ge oxide signals can be detected. The ratio of oxidized Sn (or Ge) atoms to the total Sn (or Ge) atoms can be calculated using where A Sn-O , A Sn-Sn , A Ge-O , and A Ge-Ge are the normalized Sn-O peak area, normalized Sn-Sn peak area (also including Sn-Ge bonding), normalized Ge-O peak area, and normalized Ge-Ge peak area (also including Ge-Sn bonding), respectively [48]. With consideration of Scofield photoionization cross-sections and the transmission function of the spectrometer, γ Sn-O and γ Ge-O can be plotted as a function of the photoelectron take-off angle θ and are shown in Figure 8. The inset in Figure 8 illustrates the definition of θ. θ of 0°, 30°, 45°, and 60° were used. It is observed that γ Sn-O and γ Ge-O increase with increasing θ. This is due to the fact that more Ge and Sn atoms at the surface get oxidized than those at the sub-surface. For larger θ, the information depth is smaller and ARXPS becomes more surface sensitive. For Ge oxides, the oxide percentages of the sulfur-passivated Ge 0.83 Sn 0.17 sample increase from 20 to 42% when θ increases from 0 to 60°. However, all the values are 10-20% smaller than those of the non-passivated one. A similar trend is also observed for SnO x , and the sulfurpassivated Ge 0.83 Sn 0.17 sample shows more than 50% reduction in SnO x percentage than the non-passivated one at all take-off angles. The reduction of oxide formation is more obvious in Sn atoms than Ge atoms. This reveals that sulfur passivation is more effective in suppressing Sn oxide formation than Ge oxide formation. The reduction of both Ge and Sn oxides can be ascribed to the formation of S-Ge and S-Sn bonds on the sample surface. Since both samples went through the DHF treatment, most native oxides were removed and the sample surface becomes H-terminated. As a result, the Ge 0.83 Sn 0.17 sample surface has Ge-H, Sn-H bonds, and possibly Ge-O and Sn-O bonds due to the incomplete surface oxide removal in DHF [49]. After sulfur passivation, Ge-H bond (bond energy: 263 kJ/mol [50]) and Sn-H bond (bond energy: 264 kJ/mol) are replaced by more stable Ge-S bond (bond energy: 534 kJ/ mol) and Sn-S bond (bond energy: 467 kJ/mol), respectively. The S passivation layer formed at the GeSn surface can suppress the further oxidation of sub-surface Ge and Sn atoms.
To further investigate the effect of sulfur passivation on the interface quality between the high-k dielectric and Ge 0.83 Sn 0.17 , the extent of surface segregation of Sn atom was analyzed using the obtained ARXPS data. The Ge and Sn atomic concentrations can be calculated using the stabilized Ge 3d and Sn 3d spectra. The atomic concentrations of Sn (γ Sn/(Ge + Sn) ) and Ge (γ Ge/(Ge + Sn) ) can be expressed as and plotted as a function of θ. Figure 9 shows γ Ge/(Ge + Sn) and γ Sn/(Ge + Sn) near the surface of Ge 0.83 Sn 0.17 as a function of θ for both the sulfur-passivated and non-passivated samples. The calculated γ Sn/(Ge + Sn) increases with the increase of θ, indicating that surface segregation of Sn occurred in both samples. The Sn composition of the non-passivated Ge 0.83 Sn 0.17 sample even exceeds 20% at θ of 60°. This is because Sn tends to segregate toward the surface, with the severity increasing at higher Sn compositions. Wang et al. reported that Sn segregation can occur at a temperature as low as 200°C for strained Ge 0.915 Sn 0.085 grown on Ge [24]. Since our Ge 0.83 Sn 0.17 sample went through the ALD deposition process with a temperature of 250°C, Sn segregation could also occur. Although the segregation of Sn occurs on both GeSn samples, the calculated γ Sn/(Ge + Sn) of the sulfur-passivated GeSn sample is smaller than that of the nonpassivated one at all take-off angles. The S passivation layer appears to suppress the underlying Sn atoms from segregating to the surface and from further oxidation during ALD. The good integrity of high-k dielectric/GeSn interface maintained by sulfur passivation through the prevention of Sn out-diffusion and interfacial oxidation may help to improve the carrier transport characteristics in transistors.

Extraction of interface trap density
In order to extract the D it of HfO 2 /Ge 0.83 Sn 0.17 interfaces with and without sulfur passivation, Ge 0.83 Sn 0.17 MOS capacitors (MOSCAPs) with 4 nm-thick HfO 2 were fabricated. TaN and Al were deposited as the front gate and backside metals by reactive sputtering, respectively. Low temperature C-V measurement with frequencies ranging from 10 kHz to 1 MHz was performed on the Ge 0.83 Sn 0.17 capacitors. D it was extracted using the conductance method [51]. At a particular gate bias, the peak of the G p /ω versus frequency curve can be obtained at one sweeping frequency and is referring to the maximum of per-cycle energy loss. The percycle energy loss is due to charge trapping and detrapping at certain oxide-semiconductor interface and its maximum occurs when the energy level of the trap states is aligned with the semiconductor surface Fermi-level. The value of D it can be extracted using the following equation: where (G p /ω) max is the peak energy loss value, q is the electronic charge, and A is the area of capacitor. The band-gap of fully compressively strained Ge 0.83 Sn 0.17 on Ge (100) substrate is ~0.45 eV [52]. D it values from the valence band edge to the midgap of GeSn for both sulfurpassivated and non-passivated GeSn capacitors are extracted and plotted as a function of energy in the GeSn band-gap as shown in Figure 10. For the sample with sulfur passivation, D it of 10 13 cm −2 ·eV −1 was obtained at E-E v of 0.13 eV. This is much smaller as compared with the non-passivated sample which has D it of 6 × 10 13 cm −2 ·eV −1 . In addition, sulfur passivation also leads to significant reduction in D it near the valence band edge. Sulfur passivation suppresses Ge and Sn oxide formation and Sn out-diffusion, leading to the reduction of D it . As a result, S of the sulfur-passivated GeSn p-MOSFETs is improved as compared with the non-passivated sample. In terms of the Ge 0.83 Sn 0.17 p-MOSFETs, high density interface traps near the valence band edge can be charged when the device is biased to strong inversion, and degrade the effective hole mobility. In order to further improve the effective hole mobility of the Ge 0.83 Sn 0.17 p-MOSFETs, further optimization and significant improvement are needed to reduce the D it near the valence band for the high-k/GeSn gate stack.  passivation techniques [12][13][14][15][16][17][18]53]. Despite the highest Sn composition, the Ge 0.83 Sn 0.17 p-MOSFET realized in this work shows the smallest S as compared with the other GeSn p-MOSFETs. This could be attributed to the relative low D it at the mid-gap (3.4 × 10 12 cm −2 eV −1 ) as compared with the other passivation techniques (7-9 × 10 12 cm −2 eV −1 ) [17]. This indicates the high quality of the Ge 0.83 Sn 0.17 film grown by MBE which was maintained throughout the fabrication process using low processing temperatures, as well as the Ge and Sn oxides formation between Ge 0.83 Sn 0.17 and high-k dielectric enabled by sulfur passivation. However, the D it value near the valence band is still high (~1 × 10 13 cm −2 eV −1 ) as shown in the D it plot in Figure 10. This may degrade the effective hole mobility which will be discussed in the following sections. Figure 12 shows the capacitance C as a function of gate voltage V GS for the sulfur-passivated Ge 0.83 Sn 0.17 p-MOSFET (L G = 8 μm, W G = 100 μm) measured at frequency of 50 kHz, 100 kHz, and 1 MHz. The schematic in the inset illustrates the configuration for C-V measurement. A quantum-mechanical C-V simulator [54] was used to fit the measured inversion C-V curve at 100 kHz and the simulated data were plotted using solid line in Figure 12. In the C-V simulator, the C-V characteristics are obtained through the calculation of hole and electron distributions by solving Schrödinger's and Poisson's equations self-consistently with the Fermi-Dirac distribution. In the simulation, the heavy hole effective mass of 0.27 m 0 and light hole effective mass of 0.025 m 0 (m 0 is the free electron mass) were used for Ge 0.83 Sn 0.17 channel [11]. From the simulated C-V curve, the equivalent oxide thickness (EOT) is extracted to be 7.5 Å. Figure 13 shows the forward and backward inversion C-V sweeps of one Ge 0.83 Sn 0.17 p-MOSFET (L G = 8 μm, W G = 100 μm) measured at a frequency of 100 kHz. The hysteresis is small, which indicates good dielectric quality with low density of oxide traps.
The G m,int curves versus V GS at V DS of −0.05 V for both sulfur-passivated and non-passivated devices are shown in Figure 14. L G is 4 μm. G m,int is extracted using:  where G m,ext is the measured extrinsic transconductance and R SD is the source/drain resistance. Higher peak G m,int was achieved in sulfur-passivated Ge 0.83 Sn 0.17 p-MOSFET as compared with that of the non-passivated control. Improvement in G m,int is attributed to better HfO 2 /GeSn interface achieved using sulfur passivation.
The μ eff of Ge 0.83 Sn 0.17 p-MOSFETs with and without sulfur passivation is extracted using the split C-V method: where Q inv is the inversion charge density in the GeSn channel, and ΔR Total /ΔL G is the slope of total resistance (R Total ) versus L G . Q inv can be obtained by integrating the measured inversion C-V curve as shown in Figure 12. Using this approach, the impact of R SD on extraction of hole mobility is taken out. Figure 15 shows the extracted μ eff versus the inversion carrier density (N inv ) for both the sulfur-passivated and non-passivated Ge 0.83 Sn 0.17 p-MOSFETs. The sulfur-passivated Ge 0.83 Sn 0.17 p-MOSFET shows a peak hole mobility of 478 cm 2 /V·s at N inv of ~2 × 10 12 cm −2 . At N inv of 1 × 10 13 cm −2 , 25% higher hole mobility is achieved by the sulfur-passivated Ge 0.83 Sn 0.17 p-MOSFET as compared with the non-passivated one. This is consistent with the peak intrinsic transconductance results shown in Figure 14.

Conclusion
Sulfur passivation and low temperature process modules are developed and used in the fabrication of Ge 0.83 Sn 0.17 p-MOSFET. Reduction in S and improvement in peak G m,int and μ eff are observed for the sulfur-passivated Ge 0.83 Sn 0.17 p-MOSFETs as compared with the non-passivated control. This is attributed to the effective suppression of Ge and Sn oxides formation, and suppression of Sn surface segregation by sulfur passivation. In addition, the effect of sulfur passivation on D it reduction is also investigated. It is observed that sulfur passivation reduces the D it from the valence band edge to midgap of GeSn. As a result, the lowest S of 100 mV/ decade is achieved by the sulfur-passivated Ge 0.83 Sn 0.17 p-MOSFETs. D it level of 10 13 cm −2 eV −1 in the sulfur-passivated sample is still very high. Further improvement to significantly reduce D it is needed to increase the hole mobility.

Author details
Dian Lei* and Xiao Gong *Address all correspondence to: leidian@u.nus.edu Electrical and Computer Engineering, National University of Singapore, Singapore