Silicon Dioxide
4-2-1 Motivation
With the progress of high integration and modern ultra-large-scale integration (ULSI) with gate widths of less than 0.25 µm, it has become necessary to improve device operating speed, which is limited by the capacitance of inter-metal dielectric (IMD) capacitance. To reduce the IMD capacitance, fluorine-doped silicon dioxide (FSG) [81]
films deposited by high-density plasma chemical vapor deposition (HDP-CVD) have been introduced in an advanced IMD application. The HDP-CVD technique has been demonstrated to have good gap-filling capability and film stability[82,83]. Moreover, Yang et al. [84] in 1998 successfully incorporated FSG films into 0.18 µm logic devices.
As a consequence, FSG films are considered a suitable and manufacturable low-k IMD for below 0.25 µm devices.
Many researchers have studied the formation process of fluorine-doped silicon dioxide films, viewing it as having a low dielectric constant, excellent gap-filling ability due to in-situ etching by SiF4, and being void free. They pointed out that the higher fluorine concentration would lower the dielectric constant[85-88] and improve the gap-filling ability. However, these researchers also express caution that FSG films have a
thermal stability issue that affects the integration. The subsequent deposition of FSG capped silicon-dioxide (SiO2), metal film, and passivation layer has shown blistering after the alloying process[89]. As a result, new precursors, reaction methods, and optimized process conditions have been proposed to improve the thermal stability of FSG films [90-93].
In this work, we conducted a comprehensive study of the dependence on the deposition temperature of the physical properties and thermal stability of FSG films prepared by HDP-CVD using Ar, O2, SiH4, and SiF4 gas. The relevance of the deposition temperature in influencing the properties of FSG films was reported. Also, a comparative analysis of thermal desorption spectrum (TDS), annealing test and secondary-ion mass spectrometer (SIMS) results allowed us to determine the thermal stability of individual FSG films with varying deposition temperature.
4-2-2 Experimental Procedures
The FSG films were prepared in an Ultima HDP-CVD Applied Materials Centura 5200 system using Ar/O2/SiH4/SiF4 as reaction gas. The gas flow rate of Ar, O2, SiH4, and SiF4 were 50, 110, 45, and 30 cm3/min., respectively. The deposition temperature was detected at the backside of the deposited wafer by a wafer temperature monitor (WTM).
The backside He pressure was adjusted to control the deposition temperature, which varied from 350 o to 450oC. The as-deposited films were analyzed for thickness and refractive index (RI, at 633 nm) by reflectometry and/or ellipsometry using the Nano-Spec9100. This thickness of the FSG film is the net-deposition (ND) thickness,
which is the sum of deposition, sputter and etch. The sputter (S) rate was measured by sputtering the FSG films for 60 seconds on a blanket wafer using Ar gas. The pure deposition (PD) thickness (no sputter /etch effect) was calculated by setting the bias-RF as zero on a blanket wafer. Therefore, the removal thickness of sputter and etch (S+E) effects should be the PD thickness minus the ND thickness. Besides, the etch (E) rate was calculated through the removal thickness of sputter and etch (S+E) minus the thickness of S. TDS and furnace alloy samples were deposited directly on bare silicon wafers. In addition, SIMS was used to analyze the film structure based on a silicon wafer that is covered with a silicon-dioxide film, FSG film with different temperature, and capped silicon-dioxide oxide.
The furnace annealing condition reached 425°C for 60 min. in nitrogen ambient and the alloying process required seven heating/cooling cycles in total. Next, the fluorine concentration was measured before and after alloying by Fourier transform infrared spectroscopy (FT-IR) using the peak Si-F was about 930 cm-1 and also using X-ray fluorescence (XRF). A Bio-rad FT-IR spectroscope was used to collect absorption spectra and quantified the Si-F/Si-O peak-height ratios in reflectance mode. FT-IR analysis was performed at a resolution of 4 cm-1, averaging 16 scans. XRF detected the non -Si-F- bonding fluorine in 200 second scans. The FSG films with different deposition temperatures were prepared, and then the gap-filling capability of the various FSG films was determined by scanning electron microscopy (SEM). Subsequently, the structure of short-loop samples was simulated on a 0.18 µm IMD scheme, to verify and support previous TDS and alloying results.
4-2-3 Results and Discussions 1. Film Property
Figure 4-2-1 shows the refractive index of FSG thin films, measured by ellipometry, as a function of the FSG deposition temperature. For these films deposited by HDP-CVD, the refractive indices are 1.40-1.45, which are lower than the 1.46, value of thermal oxide SiO2. It is well known that the refraction index is closely related to the porosity of these SiO2-based materials, being smaller for higher porosity. Therefore, porosity might be one reason for the lower refractive indices. However, as shown in Figure 4-2-1, the refractive index decreases appreciably with decreasing FSG film deposition temperature, but with increasing fluorine content. This shows that the lower refractive index may also influenced by fluorine incorporation into SiO2. Further support for this may be found with FT-IR spectra. We find that at lower deposition temperature, the peak of the Si-O stretching band shifts to higher frequency, in the presence of a high concentration of fluorine in the structure. As a consequence, the Si-O-Si bonds are weakened so that they become less rigid and more stretchable, resulting in a cage structure (porosity) in the film. This structure causes the film’s density to decrease.
Therefore, the refractive index also decreases with decreasing deposition temperature.
Agreeing with this result, the dielectric constant of the as-deposited FSG film decreases to 3.346 at 350oC from 3.964 at 450oC. The reduction of electronic polarization by more Si-F bonds and by an increment in porosity in the FSG film, both contribute to a lower dielectric constant. Table 4-2-1 summarizes the trends for increasing deposition temperatures. The deposition rate increase as FSG deposition temperature decreases owes to fluorine having less thermal energy in which to provide plasma etching. For higher
deposition temperatures, the FSG film becomes denser as a result of the plasma etching ability to remove weak bonds, and results in a material having lower wet etch rates. It is worth noticing that the hardness of the FSG films decrease as deposition temperature decreases due to increase in porosity and decrease in cross-linking of terminal Si-F bonds.
The hardness of the low-k film is an essential property during processing with chemical mechanical polishing (CMP).
2. Gap-Filling Ability
As the minimum geometry of integrated-circuit devices becomes increasingly smaller, the separation between metal lines becomes smaller, as well. Because the deposition rate is lower at the side wall of the metal gaps than at the bottom of the gap, HDP-CVD has been introduced to enhance the gap-filling ability owing to the sputter ability of HDP. However, it still has difficulties in filling higher-aspect ratio gaps, without voids (Figure 4-2-2a). On the other hand, we can use the etching ability of fluorine to facilitate the deposition to fill the smaller gap (Figure 4-2-2b). Therefore, to investigate the effect of deposition temperature on gap-filling ability, we separated the HDP-CVD FSG deposition process into three main reactions: deposition (D), sputter (S), and etch (E). Furthermore, we also defined the parameter D/(S+E) as the gap-filling index. The better the gap-filling capability, the smaller the D/(S+E) value. The calculated results are summarized in Table 4-2-2. As seen in this table, the E/(S+E) value quantifies the contribution of the etch component of the total removal in the HDP-CVD FSG process, and this value increases with the rise of deposition temperature. The result also shows that the sputter rate is temperature independent in our experimental temperature range. Thus, the etch rate contribution to the total removal rate increases with temperature.
There is about a 5% increase in etch rate with an increase of 50oC temperature, which causes a decrease of the D/(S+E) ratio. Therefore, we infer that FSG films with higher deposition temperature prepared by HDP-CVD have high gap-filling capacity compared to those with lower deposition temperature.
The gap-filling ability was investigated by depositing FSG films with different deposition temperatures wafers pattered with different metal widths/gaps. Figure 4-2-3 reveals that 350oC-deposited FSG film can fill into the metal gap with 0.25 µm spacing, but not reach the void-free requirement for a 0.21 µm gap (Figure 4-2-3a). On the other
hand, a higher deposition temperature (~450oC) can fill a 0.21 µm metal gaps (aspect ratio was 3.6) and with no metal clipping, which implies that the etch ability of the F-atom has increased abruptly at the higher temperature on the wafer. Again, we also demonstrate that the deposition rate decreases with increased deposition temperature due to the fact that F etches SiO2 more effectively at high temperature.
3. Thermal Stability-Out gassing Issues
In real interconnect fabrication, thermal treatment is an indispensable step. Here, the interconnect medium consists of up to 8 separate layers, all of which are deposited at about 400oC. As a result, a suitable interconnect dielectric should possess excellent thermal resistance. To study the thermal stability of FSG films with varying deposition temperature, the films were annealed in N2 ambient for 1 h at 425oC. The film thickness, RI, and fluorine concentration (FT-IR value) after annealing all remained stable even after repeated annealing tests (seven times) through the entire range of deposition temperatures. This points out that the Si-O and Si-F bonds in FSG films have enough thermal resistance against a 425oC heat treatment. However, the fluorine content
measured by XRF gradually declined and reached saturation after the third annealing test.
Moreover, the magnitude of this decline is dependent upon the deposition temperature of FSG films. The FSG film with lower deposition temperature has a larger change in XRF value, as shown in Figure 4-2-4. FTIR was used to monitor Si-F/Si-O ratio, calibrated with Rutherford back-scattering (RBS), and XRF detected total fluorine content (including bonding Si-F and non -Si-F- bonding fluorine). Consequently, we can use Eqn.
(4-2-1) to calculate the amount of non -Si-F- bonding:
XRF/FTIR=1+ (non -Si-F-)/Si-F (4-2-1) The higher XRF/FTIR ratio means a higher amount of non -Si-F- bonding fluorine.
As shown in Figure 4-2-5, the lower deposition temperature of the FSG process has more non -Si-F- bonding fluorine. This result can explain the result of Figure 4-2-4. As the FSG film with lower deposition temperature was immersed into N2 ambient at elevated temperature, more non -Si-F- bonding fluorine outgassed in the thermal process.
Furthermore, after three cycles of the annealing process, for the FSG film with deposition temperature above 400°C, the outgassing reached saturation. On the contrary, the fluorine at.% (XRF value) kept decreasing for FSG films with deposition temperature below 400°C, even after seven cycles of the annealing process. This implies that FSG films deposited at temperatures lower than 400°C have greater amounts of either weak Si-F bonding or non-bonded fluorine (free fluorine). Hence, the weaker Si-F units will be broken, or free fluorine will be outgassed by the thermal process. Considering this outgas as excess fluorine, this excess could possibly be avoided by use of a lower gas ratio (SiF4/SiX4) and thereby keep the fluorine concentration stable after heat treatment. It is speculated that more SiF4 (less SiH4) would create a greater F% concentration without
being bonded with the Si atom. Of course, lowering the gas ratio might also impact the k-lowering property of fluorine. So a proper balance must be struck. A comprehensive
study will be reported in another paper.
The thermal stability of FSG films is also studied using thermal desorption spectroscopy (TDS). Gas desorption from films, especially at lower temperature, has been a concern for device reliability issues, such as the failure of pressure cooling, thermal cycle, and thermal stress tests. The TDS spectra with respect to the mass fragment 18M/e, 19M/e and 20M/e for FSG films with different deposition temperatures are checked (not show). Corresponding to these masses, H2O, F and HF desorption from FSG films at lower temperature for lower deposition temperature, which summarized in Table. 4-2-3. As shown, the desorption temperature of Ar is relatively insensitive to the FSG film deposition temperature. On the other hand, the desorption temperatures of H2O, F, and HF are strongly dependant upon the deposition temperature.
Higher-deposition-temperature FSG films have higher desorption temperature and with lower desorption pressure. Higher onset of evolution of desorption temperature means that the film has greater thermal stability during post-thermal processes. Consequently, we have demonstrated that FSG films with higher deposition temperature have less moisture content, greater thermal stability, and better suitability for the IMD application in the back-end process.
4. Fluorine Stability-Diffusion Issues
Fluorine stability with respect to diffusion within FSG films strongly influences its utility in IMD integration, with Al wiring delamination as one important area of concern.
Fluorine stability was investigated by secondary-ion mass spectrometry using a sandwich
structure, consisting of oxide/FSG/oxide, where the oxides are 2000-Å of SiO2. This sandwich structure was deposited on the Si substrate. All the samples were annealed at 425oC for 1 h in N2 atmosphere. The annealing test was performed seven times.
As observed in Figure 4-2-6(a), when the FSG film underwent thermal treatment, the fluorine atoms diffused into the adjacent oxide films. However, this diffusion was rather modest and the fluorine profile remained quite sharp at the FSG/SiO2 interface.
Therefore, we concluded that the driving force of fluorine diffusion is the overall thermal budget. Figure 4-2-6(b) compares the effect of deposition temperature on fluorine out-diffusion stability. It shows that the FSG film with lower deposition temperature exhibits a higher fluorine diffusion capability. The fluorine diffusion length for the FSG film at 350oC deposition temperature is about ~500Å, higher than that (~200Å) of the FSG film at 450oC deposition temperature. Most of the diffused F is believed to be weak bonding or free fluorine. As a consequence, a lower deposition temperature of the FSG film causes weaker Si-F bonds will be broken and diffused, along with the free fluorine, affecting film stability.
5. Interconnect Pattern Wafer Test
The thermal stability of FSG films with varying deposition temperature is also confirmed using actual Al interconnect test structures. The effect of the deposition temperature on via resistance for the design rule (0.23 µm width /0.21 µm spacing) of 0.18 µm using FSG films as IMD was also checked. Via resistance has no significant difference in any deposition temperature, ranging 350° to 450°C (not shown). This implies that FSG films with deposition temperature from 350° to 450oC can be integrated well with photo and etch process. On the other hand, reliability and defects were
frequently detected on patterned wafers, especially in the case of lower deposition temperatures (350° to 400oC). A serious bubble defect was observed after completing a 7-layer metal structure. This type of defect was found on the metal pads, shown in Figure 4-2-7(a), similar with the result of Kawashima et al. [94] These defects arise from the diffusion of unstable fluorine and the reactions with the under-layer barrier layer (TiN), metal lines (Al), and passivation layer (Si3N4). Another defect type, i.e., peeling, is also observed for 350oC FSG film, shown in Figure 4-2-7(b). Peeling is hardly ever found in higher-deposition-temperature FSG films. Additionally, it always occurred after CMP polish. Lower hardness of the FSG film with lower-deposition temperature was suspected because the high down-force of the CMP process. To solve these problems of lower-temperature FSG film, one should increase the liner and capped layer thickness and reduce the down-force in the CMP process. However, a sacrifice of the effective dielectric constant and CMP planarization properties may occur.
4-2-4 Summary
The thermal stability for fluorine-doped silicon dioxide deposited by high-density plasma chemical vapor deposition is highly influenced by deposition temperature. All analyses, including SIMS, TDS, and annealing thermal tests, have shown that FSG films deposited above 400°C have better thermal stability. However, the high deposition temperature (over 450°C) creates metal (AlCu) extrusion and melting issues. Patterned wafers with short-loop results have also demonstrated that low deposition temperature results in F-bubble formation because of greater amount of more free fluorine. Results of this study demonstrate that the deposition temperature of FSG films is extremely
important for the films thermal stability.