Manage W-CVD Process Effluents to Boost Uptime

		At a Glance
	In tungsten CVD, lost production time and the toxicity of the tungsten hexafluoride precursor residue make frequent pump line cleaning undesirable. The solution is to eliminate the need for frequent exhaust maintenance by heating the exhaust.

Tungsten (W) is a refractory metal that has a melting point of 3370ºC and a bulk resistivity of 52.8 µW-cm at room temperature. It is used as a contact plug and a vertical interconnection between successive metal layers in most integrated circuits. The contact plug connects the active and passive devices in the silicon to the first layer of metalization. It also acts as a barrier, inhibiting interdiffusion and reaction between silicon and the connecting aluminum or copper metalization layers.

Low-resistivity tungsten films often are deposited using low-pressure chemical vapor deposition (LPCVD). Tungsten can be deposited selectively on reactive silicon surfaces or as blanket films. For blanket deposition of W films, the reaction most frequently used is the reduction of tungsten hexafluoride (WF₆) by hydrogen (H) or silane (SiH₄). The volatile-reaction byproducts are hydrogen fluoride (HF) or silicon tetrafluoride (SiF₄) and hydrogen:

WF₆ + 3H₂ R W + 6HF
or
2WF₆ + 3SiH₄ R 2W + 3SiF₄ + 3H₂

WF₆ also is used for the selective deposition of tungsten and reacts on the silicon surface to deposit W:
2WF₆ + 3Si R 2W + 3SiF₄

This is a self-limiting reaction, and a thin film ~20 nm thick is rapidly formed.

Unfortunately, WF₆ is toxic, corrosive and highly reactive, and readily hydrolyzes in moist air to form HF.

Downstream problems

In a W-CVD process, the major byproducts, such as HF and SiF₄, are gaseous. The pump line will remain clean as long as it is kept dry. Unless there is a significant air leak, there is a negligible amount of moisture in the vacuum foreline between the deposition chamber and pump. However, because wet scrubbers often are used for scrubbing and treating the unreacted WF₆ and other W-CVD byproducts, the concentration of water vapor downstream in the exhaust line after the pump is usually very high. The hydrolysis of WF₆ produces a tungsten oxyfluoride and HF:

WF₆ + H₂O RWOF₄ + 2HF

WOF₄ is a white solid at room temperature. We have found that yellow W₂OF₂ may also form by the further hydrolysis of WOF₄ if the residence time for WOF₄ in the exhaust line is long enough:

WOF₄ + H₂O R WO₂F₂ + 2HF

Fortunately, both WOF₄ and WO₂F₂ have a relatively high vapor pressure and will sublimate if appropriate heating is applied. However, WO₂F₂ may continue to be hydrolyzed to form tungsten oxide (WO₃), which has a high melting point (1473degC) and a relatively low vapor pressure:

WO₂F₂ + H₂O R WO₃ + 2HF

1. Clogged CVD exhaust line at a wet scrubber inlet.

If WO₃ is formed in the exhaust line, mechanical cleaning of the exhaust line is necessary, particularly if the water concentration is very high. Even with a properly heated exhaust line, the inlet of the wet scrubber still may get clogged due to the high moisture concentration at this location. It is not easy to apply sufficient heat at the scrubber entrance for efficient sublimation, especially if WO₃ is formed. The scrubber inlet usually will clog in less than two weeks of operation, as shown in Figure 1. Once this happens, the W-CVD tool has to be shut down, and the exhaust line has to be opened and cleaned. A thorough nitrogen purge must be performed whenever the exhaust line is opened to remove any residual toxic WF₆.

If SiH₄ is used as a precursor, silicon dioxide (SiO₂) powder can deposit in the exhaust line by the reaction of SiH₄ with water. When the exhaust line surface is moist, such as at the scrubber inlet, the combination of SiO₂ formation and WOF₄ deposition will accelerate the clogging of the line.

Uniform heating

WOF₄ has a high vapor pressure and can be readily sublimated at an elevated temperature. During sublimation, the partial pressure of WOF₄ is quite low due to the large amounts of diluting inert gas in the exhaust. Heating the exhaust line is sufficient to keep both the WOF₄ and WO₂F₂ in the vapor phase, keeping the exhaust line free of deposits. In addition, because of shorter residence time for WOF₄ in the exhaust line, the amount of WO₂F₂ and WO₃ formed in the exhaust is significantly less than when the line is unheated. Even when silane is used, a heated line moves the silicon dioxide deposition further downstream and eventually into the scrubber.

Sublimation is a physical process, and a condensable vapor will re-condense if there is any cold spot in the line. The entire exhaust line must be heated above a certain minimum temperature to prevent re-condensation. Potential cold spots are at flanges and valves, where a heater is difficult to install, or at a valve with poor heat transfer characteristics. Often, inefficient heating will make the problem even worse by promoting localized condensation in the exhaust line.

Why heat tape shouldn't be used

Heating tapes often are used because of their perceived lower cost and relatively quick installation. But expensive temperature controllers and unwieldy thermocouples are required. Because of the relatively low thermal conductivity of the stainless steel exhaust lines, large temperature gradients may exist along the pump line. The temperature measured at a single thermocouple location may not represent the pump line's true temperature. Potential cold spots can occur in the exhaust line, depending on factors such as the number and location of the thermocouples and manner in which the heater tape is installed. For example, spiral depositions have been reported in an exhaust line heated by heating tape. Condensation occurs in the cooler spaces between the turns of the wrapped heater tape. But if the tape is wrapped too closely, or overlapped, overheating will significantly reduce the heater tape's life. It is time consuming to remove and re-wrap the tape, and it may lead to other problems due to inconsistencies in the installation. Finally, for safety, outer insulation must be installed, adding complexity to the installation procedure.

The best way to control the temperature of a heated line is to use insulated, conformal heaters with embedded thermostats. Good temperature uniformity over the entire pump line is assured since the heaters are relatively short (maximum length ~2 ft.) and can be installed with no gaps. Individually controlled heaters also eliminate a potential fire hazard. In heating tape installations controlled by multiple temperature controllers, one controller often is used to control several heaters simultaneously. This not only increases temperature non-uniformity but also can pose a fire hazard if a bad heater is located directly over a thermocouple.

Preventing surface reactions

Condensation at the scrubber inlet is hard to prevent because of the high water concentration and formation of byproducts by chemical reaction. The area also is difficult to heat efficiently. Clogging is caused by solid buildup due to the surface chemical reactions between water and the W-CVD byproducts. The scrubber inlet often must be cleaned bi-weekly on W-CVD tools.

2. Nitrogen enters the plenum near the wall and is directed into the exhaust gas stream through overlapping chevron openings.

To eliminate buildup of solids, a Virtual Wall (Fig. 2) was installed at the scrubber inlet on a W-CVD tool (Fig. 3). This device creates annular nitrogen jets that form an inert boundary layer that can be thought of as a "nitrogen curtain" along the pump line's inner wall. The "virtual wall" provides a barrier separating the reactive effluents from the physical wall. This prevents solid-generating surface chemical reactions from occurring on the wall of the pump line. If the solid-generating reaction occurs in the gas phase, the nitrogen gas flow carries the fine powder that forms further downstream. The Virtual Wall is designed to maintain a laminar flow along its entire length. The overlapping chevron design makes it easy to adjust the boundary layer's active length.

The nitrogen curtain concept was developed to eliminate the solid buildup near the furnace exit in a TEOS LPCVD process, which operates at pressures from several hundred mTorr to several Torr. However, this technology has been proven to be very effective under ambient conditions.

3. A Virtual Wall installed at the scrubber inlet on a W-CVD tool.

The nitrogen curtain's efficiency is improved if the nitrogen is heated. Surface-adsorbed water increases the surface chemical reaction rate. Heating the nitrogen gas significantly reduces the amount of adsorbed water, thus reducing the surface reaction. This can be accomplished simply by heating the nitrogen curtain itself with an external heater similar to those used on the pumping line.

Benefits

The combination of using a heated line and nitrogen curtain has solved the exhaust line problems associated with the W-CVD tools. Condensation of WOF₄ on the exhaust line's inner wall has been eliminated. Formations of high melting point byproducts such as WO₂F₂ and WO₃ have been cut dramatically, especially on the exhaust line's inner wall. This will eliminate 8.5 hours/month downtime of the W-CVD tool, equivalent to ~$500,000/year savings on each tool. Also, because maintenance of the exhaust line is virtually eliminated, potential contacts to the toxic materials are avoided, which is beneficial for the fab's maintenance/service engineers. Furthermore, because pressure at the exhaust side of the dry pump is stable, the pump life will be extended as there is no need to boost the pump to overcome the pressure loss at the clogged scrubber inlet.

Acknowledgement

We thank Mr. Roger Mead of MKS Instruments for initiating and helping implement this project.

Youfan Gu is manager of research at MKS Instruments’ Vacuum Products Group. He joined MKS in 1993 after receiving his Ph.D. in chemical engineering from the University of Colorado in Boulder.
Phone: 1-303-449-9861
Fax: 1-303-449-2003
Youfan_Gu@mksinst.com

Kevin Grout is a field application engineer at MKS Instruments’ Vacuum Products Group. He joined MKS in 1989.
Phone: 1-303-449-9861
Fax: 1-303-449-2003
Kevin_Grout@mksinst.com

Steve Haupt is an equipment engineer at IBM’s Burlington, Vt., integrated circuit production facility.
Phone: 1-802-769-8341
shaupt@us.ibm.com

Talkback

We would love your feedback!

Post a comment

» VIEW ALL TALKBACK THREADS

Related Content

• Topics
• Author

Related Content

 

By This Author

There are no other articles written by this author.

SPONSORED LINKS