New WN Deposition Process Avoids Salt Contamination
New diffusion barrier research is under way.
Peter Singer Editor-in-Chief -- Semiconductor International, 4/1/1998
New WN Deposition Process Avoids Salt Contamination
A new process for depositing tungsten nitride (WNx) films has been developed by researchers at Texas Instruments (Dallas, Texas). Unlike other PECVD methods used to deposit WN, which employ a WF6 /NH3 mixture, the new process uses WF6/N2/H 2. The researchers -- Lu, Hsu, Luttmer, Magel and Tsai - said the problem with the WF6/NH3 approach is that NH3 can react with the HF byproduct to form NH4 salt, a potential source of particles. WN films are of interest primarily for use as diffusion barriers but are also promising for use as electrodes in Ta 2O5 capacitors. The work was reported in the February issue of the Journal of the Electrochemical Society.
It is also possible to deposit salt-free WN films by metalorganic chemical vapor deposition (MOCVD), but the researchers reported that this process requires high deposition temperatures(>500Å) and the deposition rate is low.
In the work at TI, the films were deposited in a PECVD reactor with an RF (13.56 MHz) powered showerhead as top electrode and wafer susceptor as bottom electrode. WF6 was used as a source gas for W; N 2 was used as a source for N, while H2 was used to remove F originating from WF6. The process was carried out at susceptor temperature in the range of 300-460Å, reactor pressure in the range of 0.5-5 Torr and plasma power in the range of 200-400 W.
An added advantage of using the WF6/H2/N2 chemistry, according to the researchers, is that it makes it convenient to use a N2/H2 plasma to in situ treat the substrate surface before the deposition, which they said is critical for depositing ultrathin uniform films.
Studies showed the deposited films to be of high purity and low resistivity, with excellent stability. The films also have desirable adhesion properties and good step coverage.
Rethinking the Wafer-Handler Interface
In conventional cluster tools, the process chamber pressure is typically changed during every wafer exchange. After a wafer is processed, the process chamber pressure is slowly lowered, and the wafer is exchanged through a gate valve that is opened when the process chamber is equal to the pressure in the transfer chamber. A new wafer is processed only after restoring the process chamber pressure. Several minutes may be needed to pump and vent the process chamber to prevent increased desorption of species from the chamber walls and to suppress particle contamination caused by rapid pumping and venting cycles. This, in turn, reduces throughput.
Hitachi (Tokyo, Japan) scientists said they have come up with a solution that avoids these pumping cycles altogether. In a concept called WHIPAC (Wafer-Handling Interface that operations under Processing Ambient Conditions), a buffer chamber in the transfer chamber is rotated to face the process chamber, and the ambient condition in the buffer chamber is the same as that in the process chamber. As soon as the processing in the process chamber is finished, the isolating gate valve is opened, and the processed wafer is replaced with a new wafer. Exhausting the process chamber and restoring its pressure becomes unnecessary. Furthermore, the process conditions should be more stable and the reliability of the equipment improved since there is less of a process chamber pressure fluctuation for every wafer exchange.
In a typical DRAM fabrication process, about 75% of the transfers are between atmosphere and a vacuum ambient, and 14% are between a low and a high vacuum ambient.
TiN Deposited from Tetraiodotitanium
TiN films can be deposited using a variety of precursors, ranging from titanium tetrachloride (TiCl4) to tetrakisdimethylaminotitanium (TDMAT) and tetrakisdiethylaminotitanium (TDEAT). Another lesser known but promising candidate is tetraiodotitanium TiI4, which is said to overcome the problems associated with TiCl4, TDMAT and TDEAT precursors.
According to researchers at the University at Albany, SUNY (Albany, N.Y.), led by professor Alain Kaloyeros, the problem with TiCl4 is that serious concerns exist regarding chlorine inclusion in the TiN films because of chlorine's corrosive nature. To minimize chlorine contamination, it is necessary to use a higher temperature, which prohibits the use of the process above the contact level. The problem with both TDMAT and TDEAT is one of step coverage and contamination -- films with the lowest levels of carbon and oxygen contamination tend to have average step coverage (only 40-50% in one set of experiments). In a recent paper (Journal of the Electrochemical Society, Feb. 1998), Kaloyeros' group stated: "While the MOCVD processes generate lower levels of particle contamination than the TiCl4 based inorganic CVD method of growth, their inclusion in emerging subquartermicron device generations appears limited by concerns over MOCVD TiN step coverage in such aggressive structures."
At the University of Albany, studies are under way to show that TiI4 is a viable alternative. In part, this means showing that iodine does not pose the same contamination threat as chlorine. It also means studying the impact of various process parameters (substrate temperature, source temperature, hydrogen carrier gas flow and reactor pressure) on key film properties, such as resistivity, step coverage, film purity, growth rate, structure and morphology.
Although further investigations are necessary, the results are quite promising. The group has shown that a wide process window exists for the deposition of nitrogen-rich, gold-colored TiN with I and O levels below 2 atom % and resistivities less than 200 Wcm.
