4-2-1 Physical characteristics of MoNx
The atomic ratio of MoNx could be detected by RBS. Fig.4-1 shows the RBS spectrums of MoNx with three different N/Mo ratios. In order to clearly distinguish the signals of C, N, and Mo, carbon substrate was used as the sample substrate. Resolution of atomic ratio from RBS is about 0.05. Theoretic values of N/Mo atomic ratios for MoN-1, MoN-2, and MoN-3 are 0.85, 1.0, and 1.45, respectively. MoN-0 stands for the pure metal of Mo.
Fig.4-2(a)-(c) were grazing-angle XRD spectrums of MoN-1, MoN-2 and MoN-3 after annealing with various temperature. After 400℃annealing, spectrum of MoN-1 showed peaks of MoN(200) and MoO3(110). The peak intensity of MoN(200) strongly increased with the annealing temperature, but that of MoO3(110) relatively decreased. Besides, there was no change of phase and orientation for MoN-1 as the annealing temperature increased. Compared
crystallization enhanced after 600℃ annealing, so MoN(200) peak appeared clearly. For MoN-3 whose nitrogen ration was 20/20, the diffraction peak of MoN(200) come out apparently until 800℃ annealing. The MoO3(110) peak at 2θ =23.34° was signal of surface oxidation during sputtering. Its intensity was very weak and increased slightly with the nitrogen ratio.
The stress of MoNx after annealing was about 1.2GPa tensile (Fig.4-3). And after the adhesion test of MoNx film, the crack of Si substrate indicated that MoNx film had superior adhesion with SiO2 and HfO2 dielectrics.
4-2-2 Electrical Analysis of MoNx
A. Sheet-Resistance Measurement
Fig.4-4(a) was the sheet resistance of as-deposit MoNx films with different nitrogen content. And Fig.4-4(b) was the sheet resistance with different nitrogen contents and various RTA temperature from 400℃ to 800℃. The resistivity was normalized to the value at 400℃.
From Fig.4-4(a) resistance increased apparently as the ratio of Ar/N2 was larger than 20/10 owing to the apparent increase of nitrogen content in MONx. And Fig.4-4(b) showed that resistivity decreased with the annealing temperature below 600℃. From XRD data, this could be due to the enhancement of crystallization caused by the increase of annealing temperature.
As annealing temperature was higher than 800℃, resistivity abruptly increased without other bonds production and color change. Further study is needed for he fact that samples with over saturated nitrogen showed high resistivity after 700℃ annealing.
B. Sputtering Damage Analysis
As shown in Fig.4-5(a), films deposited with higher power showed larger SiO2 dielectric leakage current. K. Nakajima described that high energy metal atom during sputtering would bump into the dielectric and caused sputter damage [6]. Compared with the I-V curve [Fig.4-5(b)], it seems that film with higher sputtering power shows severe sputter damage. We deposited films with three different powers upon two kinds of dielectric and investigate films by ICP-MS to find the suitable sputtering power.
Fig.4-6 shows the results of ICP-MS analysis. At sputtering power of 100W the atomic
decreases. Because the atomic resolution of Mo for ICP-MS is 0.2ppb [3x1017(atom/cm3)], the measured concentration below the resolution is considered lower Mo concentration. The detected concentration is below the resolution at the power of 25W and this means no sputter damage existed on the SiO2 surface. For HfO2 dielectric, the surface atomic concentration of Mo is below the resolution. So sputter damage doesn’t exist on HfO2 films. This ability of anti-sputter damage could be due to that HfO2 has higher density and mass than SiO2.
Sputtering process would cause damage of SiO2 in traditional process, and increase leakage current and surface trap charge. But HfO2 dielectric used in the next generation has the ability of anti-sputter damage. So it’s worthy to discuss that if sputtering process could be taken into new transistor process. Of course, further reliability analysis is still needed to be investigated.
C. C-V characteristics and thermal stability of MoNx
Fig.4-7 (a)-(d) are C-V curves of MoN-0、MoN-1、MoN-2 and MoN-3 deposited on 40nm SiO2 versus various different annealing temperature. Each C-V curve was averaged by measuring at least 10 capacitances. MoN-0 which was capacitance with pure Mo metal gate was very stable and not affected by annealing temperature (Fig.4-7). Slight distortion in accumulation mode was observed for MoN-1 which annealing below 600℃ and this was probably due to the increase of sputter damage induced interface trap. After 600℃ annealing, these interface traps were recovered. Then the measured CV curve was more normal.
The CV curves of MoN-2 after 400℃ and 500℃ annealing had slight flat-band voltage shift of about 0.1V. From XRD spectrums, this could be due to the different crystallization extent of MoNx. The flat-band voltage shift was only about 40mV when annealing temperature was higher than 500℃. This was due to that the crystallization reached stable.
The slight distortion like MoN-1 was happened when annealing temperature was lower than 600℃, and this situation could be improved by high temperature annealing. The CV curve of MoN-3 showed that the shift of CV curve was not obvious as increasing annealing temperature. According to XRD analysis results, it would have obviously crystallization as annealing temperature was up to 800℃.
between HfO2 and SiO2. Fig.4-10 was the plot of flat-band voltage shifts compared with that of 400℃. The Vfb shift of MoNx on HfO2 was very stable with annealing temperature (Fig.4-10). From above we could know that MoNx film has a good thermal stability whether on SiO2 or on HfO2 dielectrics.
D. Work-Function Modulation of MoNx
For the sample of 500℃ annealing, the curve of average Vfb versus capacitance effective thickness (CET) with various SiO2 thickness was shown in Fig.4-11 and Fig.4-12 was curve of Vfb versus CET with various SiO2 thickness for HfO2/SiO2 structure. From Fig.4-11, the slopes of MoN-0, MoN-1, MoN-2 and MoN-3 are nearly parallel and conform to constant total charge. When ratio of nitrogen flow rate increases, CV curves shift rightward and Vfb
shift upward. And it seems to get saturate when the ratio is over 20/10, thus Vfb of MoN-3 is a bit higher than that of MoN-2. The slopes of CV curve for MoN-2 and MoN-3 from Fig.4-12 are different from two others. H. Kattelus stated that nitrogen could release the stress of MoN, lower interfacial trap density and get larger slope [7]. The shift of CV curves seems to get affected by saturation of nitrogen and tends to a constant value finally. Compared Fig.4-11 with Fig.4-12, the latter get 0.3V higher Vfb than the former.
The work function could be extracted from Vfb of different oxide thickness. Fig.4-13 and Fig.4-14 are figures of work function versus different thickness of SiO2 and HfO2/SiO2 with different annealing temperature, respectively. The Vfb of each thickness of the same annealing temperature was taken the average of several values. From Fig.4-13, the work functions of MoN-0, MoN-1, MoN-2 and MoN-3 at 500℃ annealing were 4.6V, 4.97V, 5.03V and 5.11V respectively. The work function increases with the ratio of nitrogen flow rate, and finally saturates to a constant value until the ratio of flow rate larger than 20/10. The modulation of work function changes from 4.97V to 5.11V as the ratio of N/Mo is from 0.85 to 1.45. Range of modulation is 0.14V. Besides, work function slightly increases about 40mV with annealing temperature. This is due to the higher crystallization structure of MoNx from the prior results of XRD and CV analysis.
For structure of HfO2/SiO2 stack, work functions of MoN-0, MoN-1, MoN-2 and MoN-3 have a value about 4.89V, 5.31V, 5.41V and 5.37V respectively. These values are higher than those of SiO2. The work function tends to a constant value as nitrogen saturates. The
It’s worthy to discuss that the work function of MoNx on HfO2 is 0.3V higher than that on SiO2. The Fermi level pinning caused by dipole layer screening effect of High-κ materials like HfO2 limits metal work function to the mid-gap of Si. Then work function of metal gate on P substrate should reduce. But our results didn’t fit the fact. From the comparison of MoN-0 and MoN-3, the difference was independent of the kinds of oxide dielectric. So MoN on HfO2 dielectric should have no problem about Fermi level pinning.
,
The difference of 0.3V might come from the methods of work function extraction. The assumptions are HfO2/SiO2 interfacial trap density of 5*1012cm-2 and very thin HfO2
thickness. But actually the trap density is higher than that of our assumption. According to Eq 4-1, R. Jha presented that Qit High-? is negative when Qit SiO2 is positive [8]. When work function is -0.3V and ? value is 5, the value of Qit High-? equals 1.65*1013cm-2, higher than the value of our assumption.