• No results found

Interface Engineering Routes for a Future CMOS Ge-based Technology

58 3.1 Introduction

This chapter will discuss two possible routes for Ge interface engineering: (i) using high- materials that have a good interface with Ge, such as La2O3 and Y2O3, and (ii) introducing a robust ultra-thin high- interfacial layer (IL) barrier, such as Al2O3 or Tm2O3. Concerning the first route, the high reactivity of Ge with high- allows for germanate IL formation, which role is two-fold: to reduce the interface states and to suppress the GeO desorption at the interface [1, 2, 3]. The second route involves the use of ultra-thin barrier layers, Al2O3 and Tm2O3, as oxides highly resistant to oxygen diffusion and to reaction with Ge. The rare-earth metals (La, Y, Tm) tend to possess multiple valencies, such as + 2 and + 3 oxidation states, that can provide effective passivation of electrically active defects [4, 5, 6, 7]. The purpose is achieving a GeOx -free gate stack with effective Ge surface passivation.

3.2 Samples Fabrication and Characterization

The 2 nm (nominal) La2O3/Ge and 5 nm (nominal) Y2O3/Ge gate stacks were deposited by Molecular Beam Epitaxy (MBE) at 400 C on n- and p-type Ge substrates. Prior to deposition, the Ge surface was cleaned by a mild degreasing with trichloroethylene, acetone and methanol for 5 minutes in each solvent to remove the organics. Then the GeOx native oxide was thermally desorbed in-situ, by annealing at 450–500 C for 30 minutes. Y2O3 films were prepared by co-deposition of Y and atomic oxygen. The reference GeO2 film of a nominal thickness of 5 nm was prepared by ex-situ furnace anneal at 450 C for 5 minutes. The Al2O3 layers were prepared in-situ by co-deposition of Al and atomic oxygen. It is worth mentioning that we have studied the effect of deposition temperature on La2O3/Ge and Y2O3/Ge stacks [8] and found that 400 C is optimal in terms of Ge interface passivation; hence, the stacks deposited at 400 C will only be considered in this chapter.

The Tm2O3 films were prepared by Atomic Layer Deposition (ALD) on Ge epitaxial layer (35 nm nominal) grown on Si (100). Prior to the gate oxide deposition on epi Ge/Si (100), samples were cleaned in a HF 0.5%/Isopropanol 1% /H2O mixture to remove the native Ge oxide layer.

The ALD was performed using a Beneq TFS 200 deposition system, heated to 250°C to deposit Tm2O3 layers of nominal thicknesses 5 and 10 nm. The layers were deposited using Tris(cyclopentadienyl)thulium, heated to 140°C, and water vapor (H2O) as precursor gases.

After the oxide deposition, a post-deposition annealing (PDA) treatment was used in order to

59

investigate the influence of post-processing temperatures of 350°C to 450°C and annealing atmospheres of O2 and N2/H2 (10% H2 in N2) to as-deposited gate stacks.

The X-ray photoelectron (XPS) spectra for La2O3/Ge stacks were recorded at the Daresbury NCESS facility using an ESCA300 spectrometer with monochromated Al K X-rays of energy 1486.6 eV and electron take-off angles (TOA) of 15-90°. The spectrometer was calibrated so that the Ag 3d5/2 photoelectron line had a binding energy (BE) of 368.35 eV, and a full width at half maximum (FWHM) of 0.5 eV. The X-ray source power was 2.8 kW and the spectrometer pass energy was 150 eV with the entrance-slit width of the hemispherical analyzer set to 1.9 mm. Under these conditions, the overall spectrometer resolution was ~ 0.5 eV [9]. Charge compensation was achieved using a VG Scienta FG300 low energy electron flood gun, with the gun settings adjusted for optimal spectral resolution. The electron BEs were then corrected by setting the C1s peak in the spectra (due to stray carbon impurities) at 284.6 eV for all samples [10]. The core-level positions are defined as the FWHM and determined to be within 0.05 eV by fitting a Voigt curve to the measured peaks. A Shirley-type background [11] is used during the fitting of all the spectra. The angle resolved (AR)-XPS and measurements of Y2O3/Ge, Al2O3/Ge and Tm2O3/Ge stacks were made in a separate ultra-high vacuum system consisting of an Al K X-ray source and a PSP Vacuum Technology electron energy analyzer. This spectrometer was operated with an overall resolution of about 0.8 eV. The VUV-VASE measurements were performed using a spectral range from 0.5 – 8.8 eV (referring to wavelength range = 140-2500 nm), and angles of incidence of 55-75°, by 10° as a step, to maximize the accuracy. The XRD measurements were done using the Philips Xpert XRD system. The high-resolution transmission electron microscopy (HRTEM) was performed on a field emission TEM, FEI TecnaiTM F20, and on a JEOL 2100F TEM operating in STEM mode, with an operating voltage of 200 kV.

3.3 Formation of Germanate Interfacial Layers using La2O3 and Y2O3

Fig. 3.1 shows the Ge 3d core level spectra for the La2O3/Ge and Y2O3/Ge gate stacks. The data were fitted using a doublet of Voigt functions corresponding to Ge 3d5/2 and Ge 3d3/2

components. The spin-orbit splitting and area ratio values of 0.6 eV and 2:3 were fixed for the fit. The spin-orbital splitting for Ge 3d substrate peak (Ge 3d0) can be seen in Fig. 3.1 at energies of 28.6 and 29.2 eV. A high BE shoulder to the Ge 3d0 substrate peak can be seen for both La2O3 and Y2O3 samples. The rising edge in Fig. 3.1 (top) at a BE lower than ~ 28 eV originates from Y 4p to O 2s peaks at ~ 25 eV. The formation of the interfacial layer will be

60

reflected in the Ge 3d spectra as positive shifts (with respect to the substrate Ge 3d0 peak) when Ge reacts to form germanate layer.

36 35 34 33 32 31 30 29 28 27

Y2O

3/Ge

La2O

3/Ge

GeO2

Binding Energy (eV)

GeO2

GeO2/Ge

Ge 3d0 YGeOx

Ge 3d Intensity (arb. units)

La 5s

LaGeOx

GeOx

Fig. 3.1 Ge 3d XPS core level for La2O3/Ge and Y2O3/Ge stacks. The GeO2/Ge is shown at the bottom as a reference. The spin-orbit splitting is visible for Ge substrate peak for the middle

spectrum since the data were taken on higher resolution instrument.

No presence of GeO2 at the interface for La2O3/Ge stack is evident (see the reference GeO2/Ge spectrum at the bottom of Fig. 3.1 for comparison); the Ge +4 oxidation state has been reported to occur above 3 eV; @ 3.2 eV [7, 12] and 3.4 eV [13, 14] from the Ge 3d0. Considering the Gibbs free energy of formation of GeO2 ( 387 kJ/mol at 1000 K), the GeO2 is thermodynamically unstable so that a GeO2 layer is unlikely to form at the La2O3/Ge interface.

Taking into account the electronegativity of Ge (2.01 using Pauling’s scale), LaGeOx is expected to appear between the chemical shifts of GeO (Ge+2) and Ge2O3 (Ge+3), i.e. between 1.7 eV and 2.8 eV [13, 15]. The energy shift of 2.2 eV for LaGeOx has been reported [7, 16]. In our work, the presence of LaGeOx (3/2 and 5/2) can be de-convoluted at the chemical shift of +2.5 eV. Further evidence of LaGeOx formation comes from the observed shift of O 1s peak towards higher BE in Fig. 3.2(a) in comparison to the pure La2O3 at ~528.7 eV [17].

61

Fig. 3.2 (a)-(b) O 1s and (c) Y 3d XPS core levels for La2O3/Ge and Y2O3/Ge stacks with pure Y2O3 and GeO2/Ge reference spectra. The spectra in (a) refer to take off angles 0 – 70 ; the ones

in (b)-(c) were taken at normal incidence angle of 0 .

In case of Y2O3/Ge, it has been reported that the Y-Ge-O bonding configuration gives rise to a BE shift within the range of + 2.2 to 2.5 eV due to a second nearest-neighbor effect, which is distinctly different from an O-Ge-O type bonding (+3.4 eV shift) [13,18]. In our data in Fig. 3.1 (top), the chemical shift for YGeOx layer is visible at + 2.7 eV from the substrate peak. Note the difference in the interfacial layer between the two samples. The La2O3/Ge stack features GeOx layer at the interface, with a chemical shift of 1.7 eV consistent with +2 Ge oxidation state [14].

The Y2O3/Ge stack has sub-oxide fully eliminated, and GeO2 appears at the interface. The angle-resolved XPS of O 1s core level is shown in Fig. 3.2(a). As the angle is increased, the broad centroid peak is transformed, showing sub-peaks as a signature of La-O-La, La-O-Ge, La-OH and the Ge-O-Ge bonds. The O 1s and Y 3d core level spectra for Y2O3/Ge were also measured to study the additional bonding and are shown in Figs. 3.2(b)-(c). A positive shift

535 534 533 532 531 530 529 528 527

O 1s Intensity (arb. units)

62 electropositive and tends to strongly attract the neighboring O atoms [7]. The influence of La is considered to regulate the distribution of O in such a way that oxygen density is maximized in the final compound [16]. Furthermore, La on Ge in the presence of oxygen has been found to produce only La–O bonds [21], with no gap states, and the formation of stable LaGeOx layers [5, 7, 16, 22]. A penetration of Ge into the La2O3 layer, observed in this work and from the Medium Energy Ion Scattering (MEIS) experiments [8], is in agreement with the previous study [5] by energy dispersive X-ray spectroscopy where LaGeOx layer has been formed across the entire film at the temperature of 360 C. The thickness of the La2O3 has been found to be 2.6 nm from the MEIS calculated La and Ge depth profiles [8].

Fig. 3.3 (a) Ge 3d XPS core-levels taken after in-situ annealing from 425-750 C and (b) the fitting shown for two characteristic temperatures. The interfacial GeO2 layer is not present after

the annealing at 550 C.