Controlling maintenance on aluminum etch system exhaust
10/01/2001
Steve Sexton, Youfan Gu, MKS HPS Products, Boulder, Colorado
overview
Maintenance of aluminum etch systems can be reduced and uptime significantly increased by keeping exhaust byproducts in a vapor phase and then trapping them upstream of the vacuum pump, thus preventing chemical reactions at the scrubber. This can be accomplished with a properly designed exhaust management system that uses heated lines and a unique nitrogen boundary at the scrubber inlet.
Aluminum chloride (AlCl3 or Al2Cl6) is a major byproduct of aluminum (Al) plasma etch systems that use chlorine (Cl2) and boron trichloride (BCl3) etch gases. AlCl3 and photoresist etch byproducts readily condense and clog downstream vacuum lines and reduce effective pumping speed. In addition, on the atmospheric-pressure side of the vacuum pump, excess BCl3 reacts with water vapor and oxygen to form solid boron oxides and boric acid that can clog the entrance to downstream scrubbers. These problems require frequent preventative maintenance.
We have adopted an integrated subsystem approach to solve these interrelated problems [1]. A similar approach has been used for exhaust control for low-pressure chemical vapor deposition (CVD) [2]. The components of our subsystem have been designed to:
- maintain etch byproducts in a vapor phase downstream of the process,
- trap and condense byproducts ahead of the vacuum pump, and
- prevent reaction of BCl3 with water and oxygen.
The reactions
Along with Cl2 and BCl3 for plasma etching Al, nitrogen trifluoride (NF3), or fluorocarbon gases are often used for chamber cleaning. Fluorocarbon gases are also used as additives to increase etch anisotropy and passivate against Al corrosion. Typical process pressures are 10-100mtorr in an Al etch chamber and 50-500mtorr in a vacuum foreline between the process chamber turbomolecular and mechanical dry pumps. Gas flow rates are in the range of 200-500sccm.
During plasma etching of Al, AlCl3 vapor is formed and subsequently condenses as a solid on all cool surfaces, including the walls of the vacuum pumping line. The melting point of AlCl3 is 190°C; at room temperature its vapor pressure is ~10-5torr.
 Figure 1. Efficient Al etch exhaust line control where heated segments are in red. |
At the atmospheric pressure exhaust side of the dry pump, AlCl3 vapor reacts with residual oxygen, BCl3, and water vapor to form HCl and boric oxide (B2O3) or boric acid (H3BO3), which are low-vapor-pressure solids near room temperature. H3BO3 melts at ~170°C and at ~300°C decomposes to B2O3, which melts at ~450°C. These materials condense in a vacuum exhaust line, especially at the wet scrubber inlet, where water vapor concentration is very high, and at any air leak.
Similarly, AlCl3 reacts with oxygen and water to form solid aluminum oxide, aluminum hydroxide (Al(OH)3), and aluminum oxychloride. Al(OH)3 decomposes at 300°C to water and Al2O3, a low-vapor-pressure refractory material.
Aluminum fluoride (AlF3) is generated during chamber cleaning with fluorine-bearing gases that enter into similar reactions as AlCl3, forming solid, low-vapor-pressure byproducts and HF. The vapor pressure of AlF3 at 1250°C is 1torr and it will condense on relatively hot reactor walls.
 Figure 2. Metal etch trap with filter element removed after six months of operation, showing that the trap body did not need to be cleaned. (Photo courtesy of AMD) |
Preventing condensation
The conventional method of managing the process described above is to prevent condensation of AlCl3 and photoresist byproducts by eliminating cold spots and keeping the exhaust line and all its components at a temperature where AlCl3 remains a vapor into the dry vacuum pump. However, this method is not efficient because it requires many heated components, and it delivers corrosive and reactive AlCl3 vapor to the pump, which significantly shortens pump life. In addition, AlCl3 exhausted at a scrubber entrance reacts with oxygen and water vapor, clogging the entrance.
An alternative approach heats only critical downstream portions of the foreline and condenses AlCl3 with a trap upstream of the dry pump (Fig. 1). We have found that the optimum foreline temperature is 105°C, where the vapor pressure of AlCl3 is >1torr. Though the minimum required temperature is around 70°C, 105°C is better; it eliminates potential cold spots at flanges and connections. Heating the foreline keeps photoresist byproducts in the vapor phase and moves AlCl3 to a location where a trap can be installed. Foreline heating can be provided by standard modular heater components with snap-on fittings; these fittings are insulated to limit external surface temperature to <65°C during 105°C operation.
Trapping AlCl3
Trapping AlCl3 before it reaches the dry pump will increase pump life by reducing abrasive and corrosive particles. The preferred location for the trap is in the pumping foreline, close to the turbomolecular pump to minimize the foreline length that must be heated, thus reducing costs.
We found that foreline pressure is very important for efficient operation of an etch trap. Foreline pressure must be >100mtorr; if this pressure is too low, it reduces trap efficiency significantly. However, AlCl3 condensed under vacuum is easiest to contain. At too high a pressure, the material condenses as a powder instead of a solid "chunk."
 |
Since gas flow in the foreline and through the trap does not typically exceed ~500sccm, lower gas flow leads to less trap cooling. Ambient cooling can be used instead of water cooling, eliminating potential water leaks that could cause severe chemical reactions with AlCl3 (a serious safety concern). Also, the trap has been designed to achieve high flow conductance while maintaining high trapping capacity.
Trap design
A disposable trap element simplifies handling extremely reactive AlCl3. In addition, a removable stainless steel shield prevents the deposition of AlCl3 on the trap's inner wall, so the trap body does not need to be cleaned (Fig. 2). The element in a well-designed trap will only need to be replaced twice a year. Trapping efficiency actually increases as material accumulates on the filter element.
An efficient trap prevents deposition of AlCl3 downstream and eliminates any need to heat pumping lines beyond the trap. We optimized the length of the trap inner core to ensure uniform deposition on the trap element and maximum capacity. Further, we extended the entrance of the trap to accommodate a 3-in.-long heater that prevents clogging of the trap inlet when foreline pressure is quite high (~300mtorr). However, with lower foreline pressure (~100mtorr), this heater can be eliminated.
When we compared installation and operational costs for our alternative design, heating the foreline and adding a trap, with the cost of a traditional heating-only method, we found a $23,000 savings in installation and annual operations savings of $1559 (see table).
Preventing surface reactions
Our final step was to prevent excess BCl3 from reacting with oxygen or water and the subsequent depositing of B(OH)3 at the scrubber. We used a device that creates annular jets of nitrogen to form an inert boundary layer — a Virtual Wall — along the inner wall of the pump line (Fig. 3). This created a barrier that separates reactive exhaust from the inner wall [3] and prevents solid-generating surface chemical reactions from occurring on the wall of the vacuum line. If a solid-generating reaction does occur in the gas phase, the nitrogen gas flow carries the fine powder that forms further downstream.
We found that the efficiency of our design was improved when the nitrogen was heated. Surface-adsorbed water increases the surface chemical reaction rate. Heating the nitrogen to 100-120°C significantly reduces the amount of adsorbed water, thus reducing the surface reaction. We did this simply by heating the vacuum line with an external heater similar to those used on the pumping line.
Nitrogen flow rate in the boundary layer is a function of process pressure, pump line size, and the number of Virtual Wall sections. Nitrogen flow used in a vacuum foreline is typically 200-1000sccm. Used at a scrubber inlet or pump exit (where pressure is close to ambient), a typical nitrogen flow is 5-20slm.
Overall, we found the boundary layer reduced surface chemical reactions and the formation and deposition of B(OH)3, Al(OH)3, and Al3O3 at the scrubber inlet, and significantly increased the time between servicing. Our method has also been successfully implemented on CVD tools used for tungsten deposition to eliminate WOF4, WO2F2, and WO3 buildup at the scrubber inlet [4].
Practical results
We installed our exhaust line control system on a production plasma etcher at Vitesse Semiconductor in Colorado and at AMD in Texas. At AMD, the heated boundary layer has not yet been implemented. Evaluating the Vitesse system, we assumed that 30% of each 200mm wafer etched was masked, so that ~0.0227cm3 of Al was removed/wafer. This is equal to the formation of 0.042g of AlCl3/wafer. The trap capacity is ~3 lbs of AlCl3, equal to processing ~33,000 wafers.
In this installation, the Al etch trap was installed to solve an existing foreline clogging problem. After processing over 19,000 wafers (792 hr of etch operation), we had used 25-40% of the trap capacity and found little buildup of AlCl3 or photoresist downstream from the trap. However, the wet scrubber inlet was still clogging several times a week, until the heated boundary layer was installed. Then, even after two month's operation, no solid depositions were found.
The AMD installation was on an oxide etch system converted to metal etch at the same time; we added our exhaust treatment without the heated boundary layer before the wet scrubber. On this tool, maintenance was done on an as-needed basis (i.e., when the throttle valve position showed pump conductance was unacceptably low). With our installation of the etch trap, there was no reduction in conductance. Six months and 19,000 processed wafers later, this had not changed and trap maintenance had still not been required.
Conclusion
Prior to the installation of an integrated exhaust management subsystem, the vacuum pumping subsystems on tested plasma etch tools required frequent maintenance and cleaning on an irregular basis. Our new approach to etch system exhaust management involves heating the foreline, trapping ahead of the dry pump, and preventing surface reactions at the scrubber with a nitrogen barrier. We can now estimate that adopting this technique will routinely cut maintenance requirements on etch systems to around every 33,000 wafers.
Acknowledgments
We thank Paul Melvin of AMD and Mike Bouvy of Vitesse Semiconductor (now with Hitachi America) for data, comments, and photographs. Effluent Management Subsystem, Virtual Wall, and HPS are trademarks or registered trademarks of MKS Instruments Inc.
References
- The MKS Effluent Management Subsystem.
- Y. Gu, D. Hauschulz, "Comprehensive Downstream Effluent Management," Solid State Technology, pp. 89-96, May 1998.
- The MKS Virtual Wall (by MKS Instruments).
- Y. Gu, K. Grout, S. Haupt, "Manage W-CVD Process Effluents to Boost Uptime," Semiconductor International, pp. 263-266, July 2000.
Steve Sexton received his BS in marketing from San Jose State University. He is director of marketing for MKS Instruments HPS Products, 5330 Sterling Dr., Boulder, CO 80301; ph 303/449-9861, fax 303/449-2003, e-mail [email protected].
Youfan Gu received his PhD in chemical engineering from the University of Colorado in Boulder. He is manager of research at MKS Instruments HPS Products.