5-2-1 Physical Characteristics of WNx
The atomic ratio of WNx could be detected by RBS. Fig.5-1 shows the RBS spectrums of WNx with three different N/W 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/W atomic ratios for WN-1, WN-2, and WN-3 are 0.8, 1.26, and 1.57, respectively. MoN-0 stands for the pure metal of tungsten.
Fig.4-2 (a)-(c) were grazing- angle XRD spectrums of WN-1, WN-2 and WN-3 after annealing with various temperature. WN-1 without annealing showed unapparent diffraction peak of WN(100). This weak X-ray diffraction signal meant that the film didn’t have long-term crystal arrange After 400℃ annealing, the signals from XRD remained unchanged, and peak of WN(100) showed very obviously until the annealing temperature reached 600℃.
tungsten oxide [2].
Peak intensity of WN(100) and WO3(001) remained unchanged after 800℃ annealing.
For WN-2 without annealing, peak of WN(100) is similar to WN-1. After 400℃ annealing, WN(100) peak was obvious. From RBS spectrums, we could know that it could be due to that N/W ratio of 1.26 for WN-2 is larger than that of 0.8 for WN-1. More nitrogen atoms could be supplied to bonding with tungsten. As the annealing temperature reached 600℃, WN(100) peak intensity become slightly higher and 2?(=35.24 ?) was also a little smaller than that of WN-1 (2?=35.84 ?). The situation of lattice constant expansion could happen [3].
N/W ratio should be about 1 for WN-1 but the ratio the ratio becomes 1.26 for WN-2.
We supposed that too much nitrogen filled between lattices and caused the expansion of lattice constant. Similarly WN(100) peak remained unchanged as the annealing temperature was up to 800℃. From Fig.4-2(c) WN(100) signal was hardly detected for WN-3 without annealing , and this meant the weak W-N bond. According to the research of K. J. Huber, much nitrogen atoms among WNx would cause the increase of W-N bonding energy, so it would be difficult to form WN crystal [4].
WN(100) peak was weaker after 400℃ annealing. For the same reason, the peak of WN(100) after 600℃ annealing was smaller than that peak for WN-2. The intensity would be the same as that of WN-2 until annealing temperature of 800℃.
From the XRD analysis above, we can find that overdose of nitrogen was filled the site outside the lattice. The bonding of WN could also be affected by nitrogen concentration and annealing temperature. The intensity of WN and WO3 peaks remained constant when the annealing temperature reached 800℃. This means that oxidation didn’t happen during annealing. This situation is different from that of MoNx.
WNx without annealing shows compressive stress (Fig.4-3) and this compressive stress of WNx could decrease slowly to the value of -0.2Gpa. So the annealing process could release the compressive stress and this is helpful to decrease the density of interfacial trap charge.
In the adhesion test, the results are the same as those of MoNx. The crack of Si substrate indicated that WNx film had superior adhesion with SiO2 and HfO2 dielectrics.
5-2-2 Outgas Effect of WNx
increasing annealing temperature. Although mist might be the cause of bubbles, the sputtering films were deposited at high vacuum and so mist couldn’t exist.
When annealing temperature got higher, the range of regular crystal for WNx became wider from XRD analysis. We assumed that these bubbles could be nitrogen come from WNx during annealing. And the metal films could crack due to the rapid outgassing rate of nitrogen.
TDS was used to detect the components of released gas from WNx. From Fig.5-5(a) the background ion current of nitrogen was about 10-11(A) in Ar environment. After the temperature was raised to 800℃ and maintained for 10minutes, the ion current of nitrogen enhanced with the temperature while that of oxygen and water unchanged with temperature.(Fig.5-5(b)) So we found out that the cause of bubbles was the nitrogen released from WNx. We also found that there were no cracks for WNx films with furnace annealing from the SEM photos. This could be due to that nitrogen could slowly diffuse out the sample when the temperature rising was slowly. While the rising rate of temperature was too fast, the released nitrogen couldn’t have enough time to diffuse out the sample and then formed the bubbles. Thus in the proceeding experiments, we performed furnace annealing for WNx instead of RTA for MoNx in order to avoid the bubbles.
5-2-3 Electrical Characteristics of WNx A. Sheet Resistance Measurement
Fig.5-6(a) shows the sheet resistance of as-deposit WNx films with various nitrogen contents. And Fig.5-6 (b) is sheet resistance with different nitrogen content and various annealing temperature from 400℃ to 800℃. The resistivity is normalized by that of 400
℃.
Resistivity of all the samples without annealing from Fig.5-6(a) is very high and it’s due to that high-resist (1000-4500µΩ −cm) phase WN(100) was the main phase of un-annealing WNx [4]. Resistivity increases with the increase of nitrogen atoms filled among the lattice.
When the annealing temperature was below 600℃, crystallization enhanced with the increase of annealing temperature, then resistivity had the trend of decrease. As the annealing temperature was higher than 600℃, nitrogen was released from WXN and reisistivity lowers
versus various different annealing temperatures. Each C-V curve was measured and averaged by at least 10 capacitances. Fig.5.8 is the plot of flat-band voltage derived from the CV curves of Fig.5-7 versus temperature. The flat-band voltage is normalized to that of 400℃. WN-0 sample which is the pure tungsten metal was very stable with the annealing temperature. The deviation of CV curve for WN-1 is about between 0.1V from 400℃ to 800℃ annealing.
Slightly distortion was found at 400℃and 500℃ annealing for WN-2 and WN-3, and this could be due to much nitrogen induced expansion of lattice constant.
Flat band voltage shift was related to the concentration of nitrogen. We could divide the flat-band voltage shifts of Fig.5-8 into two groups with WN-0~1 and WN-2~3. And this shift was improved after 600℃ annealing. Besides, CV shifted to the right and this meant negative charge existed in the interface as anneal temperature increased. This effect conflicted with the fact that the charge induced by the stress should be positive charge and the compressive stress would decrease after annealing. We supposed that Si-N bond formed in the interface of SiO2
and Si, caused negative charge and shifted the CV curve to the right. And this effect would decrease by the separate out of nitrogen with high enough annealing temperature.
Fig.5-9 and Fig.5-10 are CV curves and Vfb shifts of WNx deposited on HfO2. WN-0 was very stable while the deviations of other samples increased with the increase of nitrogen atoms. The distortion improved after high temperature annealing. From CV data, we knew that the deviation range of WNx on HfO2 was smaller and this might be due to the fact that HfO2 is harder to bond with nitrogen.
C. Work-Function Modulation of WN
For the sample of 600℃ annealing, the curve of average Vfb versus capacitance effective thickness (CET) with various SiO2 thickness was shown in Fig.5-11. Fig.5-12 was the same curve with various thickness of HfO2/SiO2 structure. From Fig.5-11, Vfb of WN-1 with W/N ratio of 20/5 moved upward about 0.5V than that of WN-0. And Vfb moved toward downward as the ratio of nitrogen increased. The trend of HfO2/SiO2 structure in Fig.5-12 is almost the same as that of SiO2 structure.
We extracted the work function of WNx from Vfb and plotted work functions versus various annealing temperature with SiO2 (Fig.5-13). And Fig.5-13 was the same plot with the structure of HfO2/SiO2 stack. At the annealing temperature of 600℃, the work functions of
The other group was the range from WN-1 to WN-3. This group showed the trend of decreasing work functions with the increase of nitrogen ratio.
From XRD spectrum of 600℃, when the diffraction peak of WN(100) was stronger, the distance to that of WN-0 got longer. Stronger peak of WN(100) stood for the more complete structure of WNx film. When the film structure was not complete enough, the work function would be relatively small. Peak of WN-1 showed the strongest intensity and the largest deviation. This might be that there was not so much nitrogen to hinder from bonding. The range of modulation was only 0.15V which was from 5.11V to 4.96V with N/W ratio of 0.8 to 1.57 respectively. Besides, work function of 800℃ annealing compared to that of 400℃
increased slightly. According to data of XRD and CV curves, this could be due to the complete structure of WNx. At high annealing temperature, the outgassing of nitrogen and completeness of bond caused the difference of N/W ratios between all the WN samples to get smaller. And the modulation range of nitrogen got smaller, too.
At the annealing temperature of 600℃, the work functions of WN-0~3 with HfO2/SiO2
structure were 4.77V, 5.14V, 5.08V and 4.92V respectively. These values were larger than those of SiO2 dielectric. For the deviation of work function, it increased only 0.37V which value was 0.14V lower that of SiO2. Without taking Fermi Pinning effect into consideration, the work function was irrelevant to the dielectric. So the modulation range of WNx on HfO2 dielectric was smaller than that on SiO2 dielectric. This meant that problem of work function limiting exists for WNx on HfO2 dielectric. This case is different from the situation of MoNx. From WN-1 to WN-3, the ratios of nitrogen flow rate increases, and the work function decreases. The modulation range is 0.22V which is from 4.92V to 5.14V with N/W ratio from 0.8 to 1.57.
From the work function modulation of MoNx, we discovered that high density interfacial trap between HfO2/SiO2 interface induced the work function of MoNx on HfO2 dielectric 0.3V to be higher than that on SiO2 dielectric. Because the dielectric for MoNx and WNx was deposited at the same time, the interface trap should be the same. But work function of WNx
on HfO2 was only 0.1V higher than that on SiO2. This could be proven that due to the effect of Fermi Level Pinning. And the real position of limited Fermi level still was needed to
various N/W ratios. Physical analysis was also used to instigate the physical mechanism of work function modulation of WNx. According to the RBS spectrums, the ratios of N/W are 0.8, 1.26 and 1.57. And several important characteristics were described below:
1. The main phase and orientation of WN-1, WN-2 and WN-3 is WN(100). Effect of lattice constant expansion was found in WN-2 and WN-3. As the increasing annealing temperature, the phase and orientation didn’t change, diffraction peak enhanced, and the structure of WN became more complete.
2. At the annealing temperature of 400℃, WN(100) peak of WN-2 was the most obvious peak. Over-high dose of nitrogen would increase the bonding energy of W-N, and this made the WN(100) peak intensity of WN-3 is the second highest. At the annealing temperature of 600℃, the peak signal of WN-1 is the highest because there was no over-saturate nitrogen atoms to block the bonding.
3. Nitrogen between WNx would separate out during RTA from the analysis of TDS.
This was because RTA would cause the nitrogen to separate out too quickly and caused the film to crack. This effect could be avoided when replacing RTA with furnace annealing.
4. Thermal stability of WNx on SiO2 could be affected by content of nitrogen atoms and annealing temperature. But thermal stability of WNx on HfO2 is better that that on SiO2.
5. The work function modulation of WNx by increasing nitrogen ratio is divided into two stages. First stage is from pure metal tungsten to tungsten nitride and second stage is affected by the ratio of nitrogen atoms.
6. At 600℃ annealing, the work function decreases as the ratio of nitrogen flow rate increases, and this is because the incomplete structure caused by high dose of nitrogen atoms. The modulation range of work function is about 0.2V as the ratio of N/W is from 0.8 to 0.57. And the work function increases slightly with the annealing temperature.
The range of work function modulation can be 0.4V~0.5V from pure tungsten metal to WN with the N/W ratio of 1.57. This modulation range has finite effect for the adjustment of
resistance seems to be the common problem of nitride which to be the candidate of metal gate, and this could be solved by stack a layer of low resistance metal upon nitride.
Besides, the work function of WNx on HfO2 is only 0.1V higher than that on SiO2. This effect confirmed that Fermi pinning effect occurred. It’s a problem worthy to notice and make further research for it.
References
[1]. G. Wei, “ Trandition Metal Nitride Functional Coatings”, JOM. September 2001.Y. G.
shen, Y. W. Mai, “effect of oxygen on residual stress and structural properties of tungsten nitride films grown by reactive magnetron sputtering”, MSE, B76(2000), p.
107-115.
[2]. P. C. Jiang, “Preparation of W-N thin film and its characteristics as gate electrode”, NCKU MSE 2002.
[3]. K. J. Huber and C. R. Aita, “Resistivity changes and phase evolution in W-N films sputter deposited in Ne-N2 and Ar-N2 discharges” J. Vac. Scl. Technol. A 6 (2), May/Jun 1988.
0 500 1000 1500 2000 0
500 1000 1500 2000 2500 3000
WN-1 WN-2 WN-3
O C
W
Normalize Yeild N
Channel
Fig.5-1. RBS spectrum of WNx on carbon substrate.
Ratio of Ar/N2 flow rate: WN-1=20/5，WN-2=20/10，WN-3=20/20 Ratio of N/W: WN-1=0.8，WN-2=1.26，WN-3=1.57
(a)
20 30 40 50 60
As-depo WO 3(002)
800oC
400oC 600oC
WO 3(001) WN(100)
Intensity(a.u.)
2? (deg.)
(b)
20 30 40 50 60
As-depo
WN(100)
WO 3(001)
800oC
400oC 600oC
Intensity(a.u.)
2? (deg.)
(c)
20 30 40 50 60
As-depo
WN(100)
WO 3(001)
800oC
400oC 600oC
Intensity(a.u.)
2? (deg.)
Fig.5-2. (a) XRD spectrums of WN-1 with various annealing temperature, (b) XRD
spectrums of WN-2 with various annealing temperature, and (c) XRD spectrums of WN-3 with various annealing temperature.
0 100 200 300 400 500 -0.8
-0.6 -0.4 -0.2 0.0 0.2
0.4 tensile
compressive
1st ramp 2nd ramp
Stress(GPa)
Temperature(
OC)
Fig.5-3. Film stress of WN-2 versus annealing temperature
Fig.5-4. SEM photo of cracked metal surface.
(a)
Fig.5-5. (a) background ion current of various gases in Ar environment, and (b) background
(a)
(a) (b)
Fig.5-7. CV curves of WNx/SiO2 with different annealing temperatures.
(a)WN-0; (b)WN-1; (c)WN-2; (d)WN-3.
400 500 600 700 800 0.0
0.1 0.2 0.3 0.4 0.5
WN-3 WN-2
WN-1
∆V(V) FB WN-0
Temperature(OC)
Fig.5-8. Flat-band voltage shift of WNx/SiO2 with different annealing temperatures.
(a) (b)
Fig.5-9. CV curves of WNx/HfO2/SiO2 with different annealing temperatures.
(a)WN-0; (b)WN-1; (c)WN-2; (d)WN-3.
400 500 600 700 800 -0.1
0.0 0.1 0.2 0.3
WN-3 WN-2
WN-1 WN-0
∆V FB(V)
Temperature(OC)
Fig.5-10. Flat-band voltage shift of WNx/HfO2/SiO2 with different annealing temperatures.
0 10 20 30 40 50 60 70 80 90 100 110 120 -2.0
-1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4
WN-0 WN-1 WN-2 WN-3
Flatband Voltage(V)
CET(nm)
Fig.5-11. Flat-band voltage shift of WNx/SiO2 versus CET.
0 10 20 30 40 50 60 70 80 90 100 110 120 -2.0
-1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4
WN-0 WN-1 WN-2 WN-3
Flatband Voltage(V)
CET(nm)
Fig.5-12. Flat-band voltage shift of WNx/HfO2(5nm)/SiO2(40nm) versus CET.
400 500 600 700 800 5.2
5.1 5.0 4.9 4.8 4.7 4.6 4.5
WN-0 WN-1 WN-2 WN-3
φ
ms(V)
Temperature(
OC)
Fig.5-13. Work function of WNx/SiO2(40nm) versus annealing temperature.
400 500 600 700 800 5.2
5.1 5.0 4.9 4.8 4.7 4.6
4.5 WN-0
WN-1 WN-2 WN-3
φ
ms(V)
Temperature(
OC)
Fig.5-14. Work function of WNx/HfO2(5nm)/SiO2(40nm) versus annealing temperature.
Chapter 6
Formation of Interfacial Layer during Reactive Sputtering of Hafnium Oxide
6-1 Introduction
Silicon dioxide (SiO2) have been used as gate dielectric of CMOS devices for several decades because of its superior properties such as low interface state density, large energy bandgap (8.9eV), low leakage current, and good thermal stability for Si substrate and poly-Si gate. As device dimensions scale down, the thickness of SiO2 must be reduced to keep sufficient current driving capability. But when the thickness of SiO2 was below 3.5nm, direct tunneling current increases 100 times for every 0.4~0.5 nm decrease of thickness [1]. This high gate leakage current would increase standby power consumption and induce loss of inversion layer charges. According to the 2001 ITRS roadmap, the effective oxide thickness (EOT, a number which converts thickness of dielectric thickness into thickness of SiO2) will reach 1.8nm and the maximum gate current should be lower than 2mA/cm2 for low power application in 2005 [2]. The simulated leakage current of 1.8nm SiO2 was about 1A/cm2 and couldn’t meet the specified gate leakage current.
In order to reduce gate current caused by direct tunneling, the physically thickness of dielectrics must increase such that the EOT can scale down continuously. New gate dielectrics with dielectric constant (K) higher than SiO2 must be developed. Oxynitride (SiON), with K in the range of 5~7, has been extensively studied and was found to exhibit more controllable oxidation rate, lower interface state generation, improved resistance against dopant diffusion, higher dielectric integrity, lower stress induced leakage current, and lower charge trapping characteristics [3-5]. It has been thought as an alternative dielectric for the next generation.
But it is unclear if oxynitride will meet future leakage current target because its dielectric constant is not high enough. Several alternative high dielectric constant (high-k) materials with dielectric constant higher than SiON have been studied to overcome the challenge of gate dielectric scale down.
The most commonly reported high-k materials are HfO2, ZrO2, Ta2O5, TiO2, Al2O3, and
K-value is higher than that of Si3N4 (~7) and Al2O3 (8~11.5) [6, 7]. At the same time, it is not too high to induce sever FIBL effect [13]. The energy bandgap of HfO2 is about 5.68eV, which is higher than that of the other high-K materials [14]. Band alignment determines the barrier height for electron and hole tunneling from gate or Si substrate. The calculated band offsets of HfO2 for electron and hole is 1.5ev and 3.4eV, respectively [14]. This band alignment is acceptable and better than other high-K materials such as Ta2O5 [14]. The free energy of reaction with Si is about 47.6 Kcal/mole at 727℃, which is also higher than that of TiO2 and Ta2O5. Therefore, HfO2 is a more stable material on Si substrate as compared to TiO2 and Ta2O5 [15]. Among the elements in the IVA group of the periodic table (Ti, Zr, Hf), Hf has the highest heat of formation (271 kcal/mole) [16]. Unlike other silicides, the silicide of Hf can be easily oxidized [17]. That means that Hf is easy to be oxidized to form HfO2 and the oxide of Hf is usually stable on Si substrate. Unlike ZrO2, HfO2 shows a good thermodynamic stability with poly-Si. It had been reported that HfO2 would not react with poly-Si at temperatures as high as 1000℃ [18].
Several deposition techniques have been employed to prepare HfO2 film. They are physical vapor deposition (PVD) [19, 20], chemical vapor deposition (CVD) [21, 22], atomic layer deposition (ALD) [23, 24], and jet vapor deposition (JVD) [25, 26]. Among these deposition techniques, PVD has advantages of simple process, high purity, and low cost-of-ownership. However, unusually thick interfacial layer (IL) was observed in some literatures [27]. Fig.6-1 shows the calculated impact of interfacial layer on the effective k-value of the HfO2/IL stack assuming the k-value of HfO2 is 27. A 10% SiO2-like interfacial layer results in a more than 30% degradation of effective k-value. To take the advantages of high K value of HfO2 thoroughly, the interfacial layer must be reduced as possible.
In this chapter, we focused on the formation of interfacial layer using PVD method.
Various deposition schemes were employed to form HfO2 film. Detailed experimental procedure is described in the next section. Experimental results are presented and discussed in section 6-3. The formation mechanism of interfacial layer and the guidelines for minimizing interfacial layer are proposed at last.
standard RCA clean, wafers were immersed in dilute HF solution to remove chemical oxide.
Wafers were then loaded into the chamber of a reactively DC sputtering system. During film deposition, substrate temperature was held at 100℃. Either HfO2 or Hf film was deposited by sputtering from a Hf target. The base pressure of the sputtering chamber before deposition was pumped down to 2x10-8 Torr and the pressure during deposition was kept at 2x10-3 Torr.
When Hf was deposited, the gas and flow rate is Ar and 120 sccm, respectively. For HfO2
deposition, the gas mixture and flow rate is Ar/O2 and 30sccm/10sccm, respectively. The
deposition, the gas mixture and flow rate is Ar/O2 and 30sccm/10sccm, respectively. The