Development of a Fully Oxidized PECVD PSG Film
Tony Lowrie, National Semiconductor Corp., Greenock, Scotland -- Semiconductor International, 8/1/2000
 |
At a Glance | | |
|
| ||
| |||
With the introduction of modern PECVD tools to production, a large push to transfer all undoped silica glass (USG) and phosphorus-doped silica glass (PSG) processes from current low-temperature oxidation (LTO) equipment onto the PECVD tool was initiated. Dielectric processes have undergone major improvement in all aspects of thin-film processing, with the introduction of the PECVD tool to production showing improved uniformity over standard low-temperature oxides, greater than 99% reduction in particulate levels and increased throughput by nearly a factor of two. Therefore, from a manufacturing, device and product standpoint, PECVD is a very desirable tool for all dielectric deposition processes.
Finding an LTO-like film
Work carried out on poly to metal 1 interlayer dielectric (ILD) on DMOS and CMOS process flows has indicated several problems occur when using the PECVD systems. In particular, electrical shifts and poor poly step coverage resulted from poor reflow and densification following deposition (Fig. 1). Initially, electrical studies compared "standard" PSG PECVD to LTO films, with specific focus on the PECVD film used as a poly to metal 1 ILD with film thickness of 8000 Å and 7.5% wt. phosphorus doping. The film then received a 1000°C wet reflow to smooth it over the poly bars and allow for good metal 1 step coverage.
|
|
The study showed the LTO film was electrically and structurally superior to the PSG PECVD film. We found that vendor-recommended PECVDfilms act essentially like barriers to O2 diffusion through the film during reflow,causing out-diffusion problems in p-type regions. In contrast, LTO films allow substantial O2 diffusion, resulting in substantial boron out-diffusion during the reflow stage. During boron out-diffusion, the oxygen diffuses to the silicon substrate through the LTO film, and a thermal oxide growth takes place. This has the effect of pulling boron from the p-type areas into the growing oxide, leaving the p-type regions depleted of boron. Standard process flows compensate for this by increasing the implant dose.
The PECVD films also showed a different phosphorus autodoping mechanism where there seemed to be more phosphorus travelling further into substrate than the LTO films. The net effect was a shift at electrical testing of almost all DMOS, CMOS and bipolar electrical test parameters. (These were the first-order electrical effects, and it can be seen that second-order leakage effects stem from different problems.)
PECVD PSG recipe
A recipe that would act physically like an LTO film would be ideal. To begin, the PECVD deposition technique is plasma enhanced; and, therefore, the deposited film is completely non-stoichiometric and may not be in its fully oxidized state. This suggests the presence of some incomplete reacted species (Si-H, OH, P-H, free Si, etc.) in the film not seen in stoichiometric LTO films. If this is true, it would explain the above problems. First, if the film is not fully oxidized, the oxygen from the preceding wet reflow process would: a) react with the unoxidized species in the film; b) not reach the substrate; and c) not reduce boron in p-type areas. Second, if the phosphorus in the film is not in its fully oxidized, stable P2O5 state, it will show differing diffusion rates in silicon when compared to the LTO film, resulting in differing autodoping characteristics. Finally, if the phosphorus in the film is not in its fully oxidized state, then we cannot expect reflow (and hence step coverage) to be anywhere as good as with the fully oxidized LTO film.
It was surmised that increasing the nitrous oxide-to-silane ratio, increasing the process pressure and making the deposition technique with high-frequency rf power only would give the film a much better chance to be fully oxidized. These parameters were adjusted and tested to see the effects on the film's structure.
Boron out-diffusion
The following short-loop experiment was carried out to determine if the new recipe had the same oxygen diffusivity effects as LTO. On a <100> p-type silicon wafer, 1000 Å thermal oxide was grown. This served as a screen oxide for a P+ boron implant of 7.43 x 1014 at 50keV.
The different films were then deposited, keeping the thickness at 8 kA and dopant level at 7.5%. Then, the wafers were wet reflowed at 1000°C, and thickness measurements were taken both pre- and post-reflow. The wafers were then stripped back using standard wet processing, and the sheet resistance was measured.
Two interesting results were seen (Table). Looking at the change in thickness (delta) through reflow, the old PECVD film shrinks in thickness by ~500 Å, whereas the new PECVD film increases by ~500 Å and the LTO film increases by 560 Å. This suggests there is extra oxide growth through reflow in the new film and LTO methods, which compensates for the thickness loss due to densification. On the other hand, the old film decreases in thickness due solely to densification and the blockage of the oxygen in the wet ambient of the reflow. The P+ sheet resistance data shows the wafers with the old PECVD film are at 122 W/sq,whereas in the other two cases wafers are well matched at ~140 W/sq. This suggests the new PECVD film does not have the same oxygen barrier problem seen on the previous recipe and is well matched to LTO film. It also suggests the new film may be free of species that are not fully oxidized.
| Table. Comparison of Films for Poly to Metal 1 ILD | ||||
| Process | Tox 1(Å) | Tox 2(Å) | Delta(Å) | P+ Rs (W/sq) |
| LTO | 8924 | 9484 | +560 | 141.04 |
| LTO | 8679 | 9259 | +580 | 139.62 |
| Old PECVD | 9066 | 8561 | -505 | 122.4 |
| Old PECVD | 9062 | 8556 | -506 | 122.2 |
| New PECVD | 9094 | 9549 | +455 | 139.25 |
| New PECVD | 9095 | 9551 | +456 | 141.25 |
Autodoping
A similar short-loop experiment looked at the phosphorus out-diffusion or autodoping effect, again comparing the new PECVD film to the standard PSG PECVD film and LTO. Wafers were prepared as in the previous section; however, no P+ implant was carried out this time.
After strip, sheet resistance measurements were taken. The wafers with the old PECVD film showed a sheet resistance of 2821W/sq;however, for the new PECVD film and the LTO film, no measurement could be taken because the values were outside the range of resolution of the metrology. This suggests the old PECVD film had additional autodoping, and the new film may be well matched to the LTO. This was confirmed with spreading resistance profiling (SRP), which determines sheet resistance and estimates carrier concentration. Here standard LTO is compared with old PECVD and the new recipe.
|
|
SRP of the new film (Fig. 2) is identical to those received from the LTO wafers, indicating full oxidization. The absence of an n-type region suggests zero autodoping into the substrate. Consistent in the LTO and new PECVD wafers is a large peak 0.05 µm into the substrate from boron depletion due to the out-diffusion discussed above. However, this effect is not deemed to be an issue. The SRPs also suggest the phosphorus incorporated into the new film acts in a similar manner to the LTO film, whereas the old PECVD film seemed to show a different autodoping mechanism that moves the phosphorus out of the film and into the underlying substrate.
Fourier transform infrared spectroscopy
Fourier transform infrared spectroscopy (FTIR) analysis was performed to define the major chemical bonds inherent in these films.
|
|
Plot of the old PECVD film (Fig. 3b) is quite different from the LTO plot. While there is still the major Si-O-Si peak at 1068 cm-1 and a minor peak at 808 cm-1, there is no evidence of a major P-O bond, just a very small peak at 605 cm-1. A major peak at 877 cm-1 can be attributed to the Si2O3 bond, and there are other peaks located at 2256 cm-1 to Si-H, 3315 cm-1 to N-H (with another peak at 1455 cm-1) and 3600 cm-1 to free O-H. This plot confirms that the old film is not in its fully oxidized state. The absence of a P-O bond suggests there is no phosphorus present in its stable P2O5 state, resulting in poor reflow and strange autodoping effects. There also is an abundance of unoxidized silicon in the film, which explains the oxygen barrier effect seen on all previous work carried out. In contrast, FTIR on a new PECVD sample (Fig. 3c) appears very similar to the LTO plot. There are the Si-O-Si peaks at 1091 and 817 cm-1, and the P-O peaks are at 600 and 1320 cm-1. With no other peaks, it can therefore be said that the new film is in fact in its fully oxidized state.
These results indicate the new PECVD film has matched the LTO film with respect to the chemical makeup and confirm that the old PECVD film is in an unoxidized state. Also supported is the theory that the incomplete oxidation of the old film is the cause of autodoping and oxygen barrier problems.
Metal step coverage
SEM was used to indicate the effectiveness of this new PECVD over topography, in particular, metal 1 tracks traversing at a 45° angle over minimum spaced poly bars, on top of both LOCOS field and gate oxides.
|
|
|
|
The SEMs indicate the new PECVD film has reflow characteristics comparable to if not better than LTO and, hence, gives excellent corresponding metal 1 step coverage. This is a further indication that the new film, unlike standard PECVD films, is in its fully oxidized state and therefore acts as LTO film would.
From these results it can be said that the new oxidized PECVD film has matched the LTO process from the point of view of the oxygen barrier. The results also suggest the new film may not contain species that are not fully oxidized. •
Talkback
Related Content
Related Content
SPONSORED LINKS
