2-1 Introduction
As the dimensions of complementary metal oxide semiconductor (CMOS) devices are scaled into the nanometer regime, the equivalent oxide thickness (EOT) of the gate dielectric decreases steadily to thinner than 1nm. Its leakage current under normal operation bias falls into the direct tunneling regime. For future generations of metal-oxide-semiconductor field-effect transistors (MOSFET), the current gate oxide layer (SiO2 or SiOxNy) will need to be replaced with a new material possessing a higher dielectric constant (κ > 3.9). High-κ materials are employed to increase the physical thickness of the gate insulator while maintaining the same EOT and gate capacitance, thus reduces significantly the tunneling leakage current. Although many high-κ materials are proposed to replace silicon dioxide as gate insulator, HfO2 is the most promising candidate for its excellent advantages, such as a suitable dielectric constant (~25) [2.1], high band-gap energy (~ 5.9eV), and suitable tunneling barrier height for both electron and hole (>1eV). However, HfO2 is easily crystallized during deposition or following annealing processes, and crystallization increases the leakage current via grain boundaries. In order to improve the relatively low crystallization temperature of around 600°C of pure HfO2 , alloying HfO2 with Al2O3 and SiO2 has been proposed[2.2]-[2.3]. Since their silicon or aluminum binary oxides, such as HfSiOx and HfAlOx, retain an amorphous structure after high-temperature treatment, these binary oxides are now the most promising candidates to become the gate dielectric for
next-generation MOSFETs [2.4]-[2.8].
Recently, high-κ materials have been investigated using several deposition techniques including physical vapor deposition (PVD) [2.4], atomic layer deposition (ALD) [2.9], and pulsed laser deposition [2.10]. Although PVD is a simple technique for depositing new materials for evaluation in an academic organization, it may cause severe plasma damage to the electrical devices and is not preferred by industries because of poor step coverage and thickness uniformity. Chemical vapor deposition has the advantages of uniform thickness over large substrate areas and good conformal step coverage. In contrast to ALD, it is relatively easy to dope the HfO2 using CVD, which may be necessary for future gate dielectrics.
In this chapter, we employed the new atomic-vapor deposition (AVD) system to deposit the high-κ films. The AVD system would be introduced briefly in section 2-2.
Afterward, we focused on a study in which HfO2 and HfSiOx films were deposited under different conditions using AVD system. We present both structural and electrical characterizations of the high-κ films. First of all, the deposition and evaluation of HfO2 thin films have been performed in section 2-3. The effects of important deposition parameters, including the growth temperature, the chamber pressure and the gas flow of oxygen on the physical properties of as-deposited thin films have been examined. And then, the thermal stability of HfO2 films would be tested by high temperature post-deposition annealing (PDA). In the second part, the Si incorporation into HfO2
films was investigated and the results of HfSiOx films were discussed in section 2-4.
The effects of the growth temperatures, the deposition frequencies and the composition adjustments on the physical and electrical properties of as-deposited thin films have been examined.
2-2 Overview of Atomic-Vapor Deposition (AVD) System
Figure 2-1 illustrates the schematic diagram of the AVD system. The main parts of the AVD system contain an AIXTRON horizontal reactor and a liquid-delivery TRIJET-TM vaporizer. Metal-organic precursors are used as the source of the high-κ film and kept at room temperature in liquid phase in a stainless tank. The precursor would be injected into the vaporizer via high-speed electro-mechanical valves and the injector plays the important role to control the injection amounts of the precursors. The injected amounts of the precursors can be controlled exactly by adjusting the injection numbers and opening time of individual injectors. In our experiment, the opening times of the injectors were all fixed at 0.8 msec. The injection periods and pulses can be adjusted to control the thickness and composition of the deposited films. The liquid precursor was injected to the vertical vaporizer and transferred from liquid type to gas type immediately. The temperature of vaporizer (160oC and 170oC in our experiment) could be adjusted according to the kind of precursors. Argon gas would be used as carrier gas to carry the vaporized precursor into the reactor through the showerhead. The process gas, oxygen in our experiment, would be heated first in gas-box and then mixed with vaporized precursors in the showerhead. Finally, the mixed gases flowed to the process reactor and film deposition would take place on the hot substrate. The deposition parameters, including deposition temperature, chamber pressure, oxygen gas flow, injection frequency and pulse numbers, could be fine-tuned to obtain the adaptable films in different device applications. Among all process parameters, the substrate temperature is the key issue to affect the quality of the as-deposited films.
2-3 Structural and Electrical Characterizations of HfO2 Films 2-3.1 Experimental
HfO2 films were deposited by liquid-injection atomic-vapor deposition (AVD) system and the liquid precursor was tetrakis(diethylamido)hafnium, Hf[N(C2H5)2],
which was dissolved in octane to make a 0.05 M solution. The evaporation temperature of vaporizer was 170oC. Argon was used as the carrier gas, with a flow rate of 200 sccm, and oxygen as the oxidant, with a flow rate of 100 to 500 sccm. Substrate temperatures were in the range from 340oC to 500oC, and the chamber pressures were varied from 1.5 to 5mbar. Prior to the deposition, the 6-inch silicon substrates were treated with standard RCA clean. After the cleaning process, the HF-treatment was to immerse wafers into a 100:1 diluted HF solution and then spun dry without rinse in DI water.
Subsequently, wafers were put immediately into AVD system for HfO2 deposition to prevent the native oxide formation. The thickness of HfO2 was controlled by the injection pulse numbers. The deposition rate was extracted by measuring the thickness of thick HfO2 films with N&K analyzer. Because the system was designed for 200 mm wafers, the 150 mm wafers would be placed on a quartz adaptor and transferred to the process reactor. After film deposition, post-deposition annealing (PDA) was performed on all samples to investigate its impact on material properties and electrical characteristics of HfO2 films. The fundamental physical properties of these films were analyzed by many techniques, such as grazing incidence x-ray diffraction spectrum (GI-XRD), x-ray photoelectron spectrum (XPS), transmission electron microscopy (TEM) and conductive atomic force microscopy (C-AFM). In addition, the electrical characteristics of the HfO2 films were extracted from the capacitors and MOSFET devices. For electrical analysis, a precision impedance meter (Agilent 4284) was used for C-V measurements and a semiconductor parameter analyzer (Agilent 4156C) was used for I-V measurements.
2-3.2 Thickness Dependence of HfO2 Films on Deposition Temperature
As mentioned above, the substrate temperature is the most important process parameter of as-deposited films by AVD system. The thickness of as-deposited HfO2
films versus substrate temperature is shown in Figure 2-2. It can be seen clearly that the film thickness decreases monotonously as the substrate temperature increases. For conventional CVD systems, the deposition mechanisms could be divided into two regimes: Reaction-rate-limited regime at low temperature and mass-transport-limited regime at high temperature (Figure 2-3). The CVD mechanism, in which the growth rate decreases as the deposition temperature increases, is contrary to our experimental data.
Possible reasons are described as follow: Firstly, although the liquid precursor was evaporated at the vaporizer, the vaporized precursor still contained a lot of organic elements. When the substrate temperature was higher, the precursor would be decomposed more quickly and completely during film deposition step. And then, the large amount of decomposed organic elements could not be pumped out immediately and retarded subsequent precursors to go to the surface. So the thinner films obtained at higher deposition temperature could be attributed to the reduced surface chemical reaction. Secondly, the supply of the precursor was discrete and limited in AVD system.
The desorption and flow rate would increase at higher temperature due to the higher thermal energy. For this reason, the surface reaction time became shorter and the thinner deposited film would be obtained.
2-3.3 Structural Characterizations of HfO2 Thin Films by XRD Analysis
Figure 2-4 shows the GI-XRD spectra of the HfO2 films, which were deposited at various temperatures ranging from 340°C to 500°C. The chamber pressure and oxygen flow were fixed to 1.5 mbar and 500 sccm, respectively. For the samples deposited below 400°C, the intensities of the signals are extremely low. However, a bump at the position of around 57° has been clearly observed. We speculate that thin films deposited at the temperature below 400°C consist of diverse-oriented small granules, and the
bump is the convolution of these discrete signals of the granules. As a result, it is unlikely to exactly identify the structure of HfO2 thin films with this broad and weak x-ray signal peak. The only thing can be confirmed is that films deposited at such low temperature range have poor crystallinity. This trend may imply that lower temperature will lead to the formation of amorphous structure. Concomitant with increasing temperature, more sharp peaks, which are identified to come from monoclinic crystal structure, become more visible. It means that thin film will undergo structural phase change and start to form monoclinic polycrystalline structures as the deposition temperature is higher than 460°C. The corresponding orientations in monoclinic structure are identified and shown by the labeled indices.
The effect of chamber pressure on the structure of thin films is demonstrated in Fig.
2-5. It is found that the crystallinity of the deposited film also strongly depends on the pressure conditions. Obviously, lower pressure results in better crystallinity for the samples grown at the same temperature. This is related to longer mean free path of reactive species in lower pressure ambient.
Figure 2-6 shows the XRD spectra of HfO2 thin films deposited at 340oC with various post-annealing conditions. The initial aim of depositing thin film at such low temperature of 340°C is to see if HfO2 can retain its amorphous phase even after subjecting to higher thermal cycles in subsequent processes, for example, during annealing for activation of dopant impurities in S/D region. The advantages of amorphous thin films in the device applications are the lower leakage current, superior heterogeneous interface quality, blocking capability against impurity diffusion from poly-electrode. Rapid thermal annealing in N2 ambient for 30s was employed to test the thermal stability of the deposited HfO2 films. However, the HfO2 film is not a good diffusion barrier for O atoms, HfO2 film starts to crystallize as annealing temperature is above 650°C and the crystallinity of deposited films is enhanced by increasing RTA
temperature, as shown by increasing signal intensity. These results show that the HfO2 film will depict polycrystalline structure above 600oC, no matter at film deposition or subsequent high-temperature annealing.
2-3.4 Chemical Bonding and Composition of HfO2 Thin Films by XPS
Analysis Chemical characterizations of HfO2 films were accomplished by x-ray
photoelectron spectroscopy (XPS) utilizing monochromatic and standard Al x-ray source. The results are shown in figure 2-7. Detected elements in thin films are hafnium (Hf), oxygen (O), and carbon (C). In order to avoid the undesirable carbon contamination on the sample surfaces, XPS analyses were also performed with ion milling. The low energy ion sputtering time is 15 sec, 3 min, and 8 min, respectively.
Negligible damage by low energy ions during depth profiling could be assumed since no significant shift of the binding energies is observed. The relative contents of Hf, C, and O elements, which were determined by the spectra of the non-sputtering and 8min sputtering samples, are summarized in Table 2-1. It is found that the relative intensity of C1s signals decreases drastically after sputtering. This result is reasonable due to the fact that all air-exposed materials will have a thin film deposition, composed primarily of hydroxide (i.e., alcohol-type, C-OH units). After removing this thin layer, the signals originating from purer HfO2 can be obtained. Estimating from the data of 8-min sputtered samples, the content of C elements incorporated in the bulk of HfO2 thin films during deposition process is only at the level of approximately 2.56%. This result strongly suggests that the decomposition of the employed Hf-precursor is very effective at 400oC. Therefore, the carbon atoms contained in the ligand can easily be evacuated.
As a consequence, the concern of high level C incorporation using other Hf-precursors, such as alkoxdes and β-diketonates, doesn’t apply to the use of the
tetrakis(diethylamido)-hafnium precursor. With such low C concentration level, thin films are be expected to have lower defect density, resulting in more robust thin films from the viewpoint of reliability.
Figure 2-8 shows Hf4f and O1S XPS spectra as a function of deposition temperature.
The binding energy of Hf-O bond is nearly the same among all samples deposited at different temperatures. This result also means that the Hf precursor, Hf[N(C2H5)2], is easily decomposed above 360oC to form the HfO2 films. Nevertheless, it can be seen clearly that the separation of the two peaks of Hf spectra, Hf 4f7/2 and Hf 4f 5/2, becomes more apparent at higher deposition temperature. This phenomenon shows that the composition of HfO2 becomes more stoichiometric at higher deposition temperature; in contrast, the impurities of the organic precursor are more easily incorporated into HfO2 system at lower deposition temperature. The proof that the binding energy of Hf-O bond in a silicate film is about 1eV higher than that of Hf-O bond in a pure HfO2 system will be given in section 2-4.
2-3.5 Structural Images of HfO2 Thin Films by TEM Analysis
Figure 2-9 shows the images of plane-view TEM for (a) the HfO2 sample deposited at 400°C and the samples with subsequent (b) 600°C, (c) 800°C, (d) 1000°C post-deposition annealing for 30sec in N2 ambient, respectively. There are obviously two contrast regions in all the samples: dark and bright regions; with the dark areas embedded in the bright regions. At first glance, the spherical-shaped dark regions would be easily recognized as polycrystalline grains. However, only a few of them in as-deposited sample show crystalline diffraction patterns. This poor crystallinity has been evidenced by broad x-ray peaks, which become sharper concomitant with enhancing growth temperature. The contrast is supposed to be caused by the different compositions. Dark region is more likely to be produced by Hf-rich composition while
bright zone is more close to stoichiometric composition, which are identified by TEM energy-dispersive spectroscopy (EDS). Table 2-2 lists the element ratios of O and Hf in dark and bright regions separately. The preciseness depends on the beam size. However, this table reveals important information of the composition of HfO2 thin films. This compositional inhomogeneity is hypothesized to be caused by the insufficient Hf atom surface migration due to its large mass and insufficient oxygen gas flow during the film depositions.
With 600°C annealing, it is observed that many grains start to crystallize while the change of the film structure in bright regions can hardly be detected. Noteworthy, there is dramatic structural alternation taking place around the grain boundary. High temperature annealing produces significant structural modification along the grain boundaries. The presence of the seam between the grains and adjacent regions is supposed to be the primary cause of tremendous leakage current increase, compared with as-deposited samples. After annealing at 800°C, the situation is very different from that in 600°C. The crack along the original grain boundary disappears. Instead, the crystalline patterns with diverse orientations could be seen almost everywhere in both dark and bright regions. As annealing temperature goes up to 1000°C, it is found that the grain starts to merge with each other and the size of a single grain substantially increases as a result of high temperature annealing. This structural re-arrangement reduces the number of the grain and grain boundary.
Figure 2-10 shows the images of cross-sectional TEM for (a) the HfO2 sample deposited at 400°C and the samples with subsequent (b) 600°C, (c) 800oC, and (d) 1000°C post-deposition annealing for 30sec in N2 ambient, respectively. The contrast is not only observed in plane-view TEM but also in cross-sectional TEM pictures. The lighter contrast is near the Si/film interface. This interfacial layer is thought to be a Si-rich Hf silicate according to many previous reports, even though this speculation can
be hardly identified by any compositional analysis method. The total physical thickness, the individual thickness of HfO2 films and interfacial layer are plotted in Fig. 2-11.
After 600°C annealing, total physical thickness of thin film is increased by 10 Å.
Interestingly, most of the thickness increase, approximately 8.5 Å, occurs on the upper layer. Meanwhile, the increase in thickness for interfacial layer is only slightly larger than 1 Å. This result obviously contradicts with the previous report, which stated high temperature annealing would lead to significant increase in thickness of interfacial layer.
It is hypothesized that the as-deposited thin films content more Hf than expected. In other words, it is Hf-rich as revealed by the huge dark area portion in plane-view TEM picture. In rapid thermal annealing process, there exists residual moisture, which would react with the excess Hf violently and contributes to the growth of additional upper layer. Once the excess Hf atoms are consumed, the oxygen would diffuse through metal oxide and interfacial layer and react with underlying silicon substrate. It can be confirmed by the fact the thickness of upper layer doesn’t change with further enhanced annealing temperature. Meanwhile, the interfacial layer experiences nearly 10 Å increase after 1000oC PDA treatment. With the combination effects of reduced grain-boundaries (shown in plane-view TEM images) and increased thickness of interfacial layer, a lot of leakage paths are eliminated and leakage current is also decreased significantly, which are supported by following results of conductive-AFM images. Nevertheless, the EOT of high-κ films also increases unexpectedly which is related to the interfacial layer with lower κ value. In fact, the κ values still increase as the PDA temperature increases (the κ value of interfacial layer is assumed to be 5) and close to the idea HfO2 dielectric value (~25) due to the improved film quality. The characterizations of the as-deposited HfO2 film and the samples after PDA treatments are summarized in Table 2-3.
2-3.6 Structure Changes of HfO2 Thin Films by Conductive AFM Analysis
Figure 2-12 shows the images of conductive AFM(C-AFM) for (a) the HfO2 sample deposited at 400°C and with subsequent (b) 600°C (c) 800°C (d) 1000°C post-deposition annealing for 30sec in N2 ambient, respectively. The contrast represents the relative magnitude of the leakage current; the brighter the region, the larger the leakage current. The darkest region denotes zero current flow, which means the leakage current is too small to be detected by C-AFM system. The scanning area is approximately 800×800 nm2. The parameters are summarized in Table 2-4. In general, the as-deposited thin film exhibits the lowest average leakage current. However, there are still many bright spots in its C-AFM image, shown in Fig. 2-12(a). The presences of bright dots strongly indicate the leakage current is a localized behavior. From the results of AFM, this characteristic doesn’t correlate with the surface condition of the thin film since only minor difference of roughness exists among all samples. Among the annealed
Figure 2-12 shows the images of conductive AFM(C-AFM) for (a) the HfO2 sample deposited at 400°C and with subsequent (b) 600°C (c) 800°C (d) 1000°C post-deposition annealing for 30sec in N2 ambient, respectively. The contrast represents the relative magnitude of the leakage current; the brighter the region, the larger the leakage current. The darkest region denotes zero current flow, which means the leakage current is too small to be detected by C-AFM system. The scanning area is approximately 800×800 nm2. The parameters are summarized in Table 2-4. In general, the as-deposited thin film exhibits the lowest average leakage current. However, there are still many bright spots in its C-AFM image, shown in Fig. 2-12(a). The presences of bright dots strongly indicate the leakage current is a localized behavior. From the results of AFM, this characteristic doesn’t correlate with the surface condition of the thin film since only minor difference of roughness exists among all samples. Among the annealed