New look for process tools?
07/01/2004
As the industry congregates for Semicon West this year, some changes are in the wind. Growing collaboration between major integrated device manufacturers (IDMs) and multiple foundries could have a big impact on future process tools. With companies such as IBM, Texas Instruments, Motorola, Samsung, and even Intel contracting an increasing amount of production to foundries, there will be pressure for uniformity in the tools available so that processes can be shifted to different fabs and quickly characterized to achieve process repeatability, especially when "copy exact" or "copy smart" principles are adopted.
This collaboration provides second sources in case of any breakdowns as well as capacity buffers in case of a surge in orders. Some IDMs are qualifying a number of foundries to run their processes to achieve this. Large fabless design houses also want to have second or even multiple sources for the same reasons as well as to ensure competition on pricing.
Making the transfer of processes from fab-to-fab "quick and convenient" means that there needs to be little or no variation in the tools used to implement them. This creates great pressure for toolmakers to match tool characteristics more closely than they may have in the past. Built-in advanced process control (APC) can help to keep performance excursions well within tight limits, but tools also will need to offer open interfaces so that fabs can adapt them to their own control requirements, which may involve proprietary recipe variations. So toolmakers will have to strike a balance between built-in APC and capabilities for users to add their own methods.
At the same time, foundries, which may become dominant buyers of tools, will need flexibility so they can adapt their process lines to requirements for different IDMs as well as for fabless customers wishing to create product differentiation. Thus, tool flexibility will also be an important factor for foundries, while it might not have mattered to the old-time DRAM makers. Today, however, even the DRAM houses like Samsung are modifying processes to move some of their capacity to flash or other memory variations with higher margins and fast-growing demand.
The other factor that demands quick set-up times and tool flexibility is the trend toward system-on-a-chip (SoC) designs. Many SoCs are high-value devices, but only have short runs where quick time-to-market is essential. So tools need to be geared to a high-mix, frequently changing manufacturing environment. Japanese chipmakers see this as so critical that they are working together to devise tools that can run multiple sequential processes, but are less costly and do not have the high throughput demanded by the traditional high-volume fabs. This trend away from commodities toward specific solutions was discussed in detail by Takeshi Hattori, GM and chief research officer of UCT Laboratories at Sony in Atsugi, Japan (see SST, June 2004, pp. 85–93). These efforts may lead to some new alternatives to the costly high-volume tools now common in most major fabs.
On top of all these pressures on tool vendors is the demand for them to become partners with chipmakers in process development rather than just "shipping iron." Since each fab tends to have its own idiosyncrasies, the tool must be able to adapt easily to variations in process parameters while still offering the uniformity that growing collaboration demands, and also be easily connected to speed installation.
Surprisingly, even with these much stiffer requirements, we hear fewer complaints about process tools when we talk to chipmakers. Standards for 300mm wafer handling have helped, and tools can be hooked up and on-line much quicker now. Chipmakers feel that tool vendors have been responsive to changing needs, although they also say they will shift to alternate tools if they offer more benefits. So competition is not dead in process tools. It may be tougher for a newcomer to break into the business, but if a tool does have advantages, the potential for rapid adoption at multiple fabs is probably greater than in the past. So winning a place in the process flow may be tougher than ever, but the payoff could be much greater as well.
 |
Robert Haavind
Editorial Director