Thermal Processing: Meeting the Challenges of 300 mm
Staff -- Semiconductor International, 4/1/1998
Though the more glamorous areas of wafer processing, like lithography,
typically steal the headlines regarding 300 mm development, the
challenges in other areas, such as thermal processing, should not be
trivialized. In fact, engineers designing 300 mm furnaces and rapid
thermal processors (RTPs) are facing the very same issues of uniformity
and temperature control that they only recently solved for 200 mm
systems. Now, however, they must demonstrate their engineering prowess
on a grander (2.25X) scale.
Interestingly, much debate exists over which tool type - vertical furnace or RTP system - will become the tool of choice for processing leading-edge dielectrics. The two approaches compete on the basis of the following:
- Reduced thermal budget
- Improved temperature control
- Thermal uniformity
- Tight ambient control
- High throughput
- Low cost-of-ownership (COO)
Of these important factors, temperature uniformity is perhaps the most critical. In fact, because of the superior temperature capabilities of vertical furnaces over early RTP systems, RTP was mostly used for annealing types of applications (i.e., silicidation, TiN diffusion barriers, implant activation, etc.).
New RTP systems, however, are demonstrating vastly improved temperature uniformity and control. For instance, rapid thermal oxidation of a 30 film requires temperature control of ±1°C. Today's 300 mm systems will have to not only meet these demanding performance specifications, but also deliver acceptable throughput. Tool productivity on larger wafers is essential if the industry is to realize the 30% reduction in device cost per square centimeter of silicon that it expects from this transition.
 |
| 1. Model SHS 3000 RTP uses 26 tungsten halogen lamps in each lamp array and four side lamps to achieve 250°C/sec ramp rates and thermal variation of 4°C across the 300 mm wafer. |
In a related area, AST has delivered two thermal donor annihilation tools to wafer manufacturers for improving the quality of 300 mm starting wafers. As in many other areas, there remains a shortage of 300 mm wafers to test and qualify the various types of wafer processing equipment.
At high IC process temperatures, key concerns include thermal and gravitational stress and possible plastic deformation of the silicon wafer. The likelihood of plastic deformation increases with faster ramp rates, so maximum thermal uniformity must be designed into the system so that high ramp rates can be employed to reduce thermal budget. In addition, even slight mishandling of 300 mm wafers can cause localized transition of the silicon from its crystalline structure to amorphous silicon. Subsequent high-temperature processing of the wafer may facilitate an amorphous/crystalline transition, with propagation of dislocations from the wafer edge, for instance, into the wafer bulk.
|
| 2. At 300 mm, fast ramp furnace design was optimized for maximum within-wafer temperature uniformity, with ramp rates varying from 10min to 100min. (Source: TEL) |
The design of 300 mm furnaces and RTP systems must consider all these factors. The fast ramp furnace shown in Figure 2, Tokyo Electron Ltd.'s (TEL, Austin, Texas) Alpha 303 FTP system, was an engineering challenge in terms of boat design, wafer support, temperature monitoring and control, and other issues. The fast ramp furnace accommodates boat sizes of 25-50 300 mm wafers. Ramp rates as high as 100°C/min have been achieved with this system, with 20 mm wafer pitch (spacing between wafers) and 850°C process temperature. Models of radial heat transfer and thermal stresses during heating and cooling were used to determine adequate wafer spacing and ramp rates needed to avoid plastic deformation during large wafer processing. Increased wafer pitch decreases within-wafer thermal variations by increasing the amount of radiation reaching the wafer's surface. Boat design and wafer support methods also influence the likelihood of slip, where the silicon crystalline planes literally slip because of stress.
It was once believed that all 300 mm processes would have to be single-wafer processes, as the value of 300 mm wafers is so high. Veterans of the industry recall similar arguments at the 150 mm to 200 mm transition. In fact, one industry observer recalled a time when nobody saw a need to process more than 35 200 mm wafers at one time. However, competitive pressures in the industry at that time convinced companies of the need to conduct the majority of thermal processes in batch furnaces to realize the lowest COO. One could say that competitive pressures have greatly increased since then, and where it is possible, 300 mm wafers too will be processed in batch furnaces. Forecasters are now expecting a blanket transition to single-wafer processing for 450 mm wafers, with concurrent needs for in situ monitoring and control methods. Will it even happen then?
