Results and Discussions

3.3.1 Physical Properties of MoN

Films

Fig. 3-1 shows the RBS spectrum of MoNx films deposited on graphite substrate with various Ar/N2 gas-flow ratios of 20/5, 20/10 and 20/20. Due to the use of graphite substrate instead of Si substrate, the nitrogen signals can be resolved clearly and the N/Mo atomic ratio can be determined precisely. The respective nitrogen contents are 46, 50, and 59 in atomic percentage (at. %) for MON-1, MoN-2 and MoN-3.

Fig. 3-2 shows the XRD spectra of MoNx films annealed at different temperatures in N2 ambient for 30sec. Strong MoN (200) phase was detected over all samples. With the increase of annealing temperature, the intensity of MoN (200) phase increases. The increasing intensity of MoN (200) signal implies grain growth during high temperature annealing. Looking into the XRD spectra, the higher the nitrogen content is, the weaker the MoN (200) signal intensity is. This phenomenon implies that the excess nitrogen can make the MoNx films tend to be amorphous-like. However, the drawback of high nitrogen content in the MoNx is the high resistivity of the film. As the annealing temperature increases, the sheet resistance decreases due to grain growth but the higher nitrogen content samples still show higher resistivity. Table 3-1 lists the MoNx films ID and summarizes the above material characteristics.

The Mo-N system phase diagram was reported by Jehn et al. in 1978 [14]. When

the nitrogen content is less than 28 at. %, the main phase is Mo. With the increase of nitrogen content ranged from 28.7 at. % to 35 at. %, the main phase is Mo2N including β-Mo2N and γ-Mo2N. The β-Mo2N will transfer into the γ-Mo2N phase at higher temperature anneal and higher nitrogen content. The nitrogen content about 40 at. %, the main phase is Mo3N2. Furthermore, molybdenum nitride with nitrogen content higher than 40 has the main phase of MoN [14].

3.3.2 Effective Work Function on SiO

and HfO

Fig. 3-3(a) and (b) show the flat-band voltage (Vfb) versus equivalent oxide thickness (EOT) plot of the 500 °C annealed samples with SiO2 and HfO2/SiO2

stack as gate dielectric, respectively, where the Vfb values are the average from more than ten samples. The good linearity reflects the validity of effective work function extraction by the linear extrapolation of EOT vs Vfb plot. According to the discussion of chapter 2, the Y-axis interception in the Vfb-EOT plot of SiO2

samples represents the work function difference between metal gates and Si substrate (Φms=Φm,eff-ΦSi). While the Y-axis interception in the Vfb-EOT plot of HfO2 samples represents the work function difference (Φms) plus the effect of the HfO2/SiO2 interface charges (Qhigh-k). With known Si substrate concentration, the work function of metal gate on SiO2 layer can be extracted. However, on HfO2

layer, an offset due to the Qhigh-k are inevitable. We denote the effective work function of HfO2 samples as Φm,eff’ counting for the oxide charge effect.

The Φm,eff of MoNx films on SiO2 and Φm,eff’ of MoNx films on HfO2 extracted from the Vfb-EOT plots are shown in Fig. 3-4 (a) and (b), respectively. The Φm,eff

and Φm,eff’ of MoNx films are almost independent of the annealing temperatures but are dependent on the nitrogen content. This is one of the advantages of MoNx

because the work function of several metals change after thermal annealing [11].

61

The MoNx films on SiO2 with nitrogen contents of 0, 46, 50, and 59 at. % have work function of 4.60 eV, 4.97 eV, 5.03 eV, and 5.11 eV, respectively, after 500 °C annealing. Compared with the MoN-0 sample, the increase of work function (∆Φm) is 0.37 eV, 0.43 eV, and 0.51 eV as the nitrogen contents are 46, 50, and 59 at. %, respectively. The increase of work function is not linearly dependent on the nitrogen composition. It increases rapidly with the addition of nitrogen and then gradually saturates at high nitrogen concentration. It is concluded that the work function of MoNx film on SiO2 can be adjusted by nitrogen incorporation from mid-gap to close to valance band of Si and can be applied as gate electrode of bulk PMOSFETs or fully-depleted SOI PMOSFETs.

The Φm,eff’ values of MoNx films on HfO2 layers with nitrogen contents of 0, 46, 50, and 59 at. % are 4.89eV, 5.31eV, 5.41eV, and 5.37eV, respectively after 500°C annealing. Since the HfO2/SiO2 stacks were prepared simultaneously and HfO2 film is immune to sputtering damage [15], it is reasonable to assume that the Qhigh-k values are independent of gate electrode. The increase of work function, excluding the effect of Qhigh-k, is 0.42 eV, 0.52 eV, and 0.48 eV as the nitrogen contents are 46 at. %, 50 at. %, and 59 at. %, respectively. It is important that the magnitude of work function adjustment (∆Φm,eff) on HfO2 film is nearly the same as that on SiO2 film.

The Φm,eff and Φm,eff’ of samples with the same MoNx film and annealing conditions has a difference of about 0.3 eV over all samples. It is indeed necessary to concern if the interface dipole layer at the MoNx/HfO2 interface induces the Fermi-level pinning effect. The formation of interface dipole is due to the interaction of gate material and gate dielectric. The pinned Fermi energy should depend on gate material. Furthermore, the reported Fermi-level pinning effects

always force the work function of metal gate tends to the mid-gap of silicon. Both phenomena are not observed in this work. The Φm,eff’ of MoNx films on HfO2 shifts toward the valence band of silicon. Moreover, compared with MoN-0 film, the work function increase of MoNx films incorporated with the same nitrogen on both SiO2 and HfO2 is almost the same. Therefore, it is believed that Fermi-level pinning does not occur in the MoNx/HfO2 stack. This is an important advantage for MoN films as gate electrode because it is quite possible that metal gate will integrate with high-k dielectric while Hf-based dielectrics are very promising. The 0.3 eV shift between Φm,eff and Φm,eff’ is attributed to Qhigh-k now. After a simple calculation, the Qhigh-k is -1x10¹³ cm^-2, which is consistent with that reported in literature [16].

The extracted work function of pure Mo film (MoN-0) on SiO2 is 4.6 eV in this work. This value is different from the results of some literatures. It should be noted that the work function of Mo depends on its crystalline orientation. There are several possible orientations of Mo film deposited by sputtering technique.

Among them, only the (110) orientation has work function close to valence band of Si. The other orientations show work function close to mid-gap of Si [3]. Fig.

3-5 shows the electron diffraction pattern of the 600 °C annealed Mo film.

Although (110) is the dominant orientation, several crystalline orientations including (200), (211), and (220) are also identified. Therefore, the work function of the MoN-0 sample of 4.6eV is reasonable and is consistent with some other works [2, 3]. Although some recent papers reported that the work function is reduce by implanting nitrogen into Mo(110) film. It should be noted that the nitrogen concentrations are not high enough so that the main phase is Mo2N in those works [4-8]. In this work, Mo-nitride films were prepared by reactive

63

sputtering technique and the N/Mo ratio is around unity. As shown in Fig. 3-2, the main phase of our samples is MoN(200) but not Mo2N. Therefore, the different observation between our work and the references could be explained as follows.

As the orientation of Mo film is pure (110), the work function is close to the valance band of Si. Nitrogen implantation destroys the (110) crystalline and the Mo film becomes amorphous. After annealing, the phase is Mo2N due to insufficient nitrogen concentration. The work function of Mo2N is close to the mid-gap of Si. Therefore, the higher the nitrogen implantation dose is, the lower the work function is. Once the nitrogen concentration is high enough and the phase changes to be MoN, the work function increases. The work function of MoN is close to the valence band of Si while the work function of Mo2N is close to the mid-gap of Si.

3.3.3 Electrical Effect of Thermal Annealing

Fig. 3-6 (a) and (b) shows the C-V curves of MoN-0 and MoN-3 samples, respectively, after annealing at different temperatures. The gate dielectric is a 40 nm thick SiO2 film. Slight distortion of the C-V curves at weak accumulation mode is observed for MoN-0 sample annealed at temperatures below 600°C and the distortion can be recovered after annealing at higher temperature. The distortion should be attributed to the interface states between gate dielectric and Si substrate. Similarly, the C-V curves of HfO2 samples with HfO2 (5 nm)/SiO2 (40 nm) stack gate dielectric were shown in Fig. 3-7. No distortion of C-V curves is observed. This observation confirms that HfO2 is more immure to sputtering damage than SiO2 previously [15]. The accumulation capacitance of all SiO2 and HfO2 samples is almost the same. This result confirms the thermodynamic stable of HfO2/thick SiO2 stack [17]. All of these results indicate that MoNx films on both

SiO2 and HfO2 layers are thermally stable up to 800 °C. This result is better than several metals such as TiN, TaN, and TaSiN [11,12].