REQUIREMENTS AND PROBLEMS FOR FURTHER INTEGRATION
The committee concludes that integrated processes as a manufacturing scheme are Attractive for lowering the overall cost of the final product because of the potential for providing:
greater flexibility through software modification rather than hardware modifications;
shorter lead times;
higher productivity; and
improved quality, reliability, and performance.
As the critical dimensions in integrated circuits shrink from 0.5 to 0.35/µm or less, present manufacturing practice requires increasingly cleaner environments to keep the defect density low enough to maintain adequate yields. This has required not only development of processing tools with the requisite cleanliness but also increasingly cleaner rooms in which the equipment is housed. Each incremental improvement drives up the cost of a fabrication facility to the point where a new state-of-the-art facility costs in excess of $500 million. If this trend continues and manufacturing technology does not change, the cost of a facility for manufacturing of 64 M-bit random access memory (RAMs) may exceed $1 billion (see Figure 7-1). The application of improved integrated processing has the potential of reducing the steepness of the projected cost curve.
Integration of related processing steps can lower this cost through such factors as the sharing of common functions such as load locks, wafer handlers, and central computers. As discussed in Chapter 6, this is now nearing reality in the microelectronics industry with the increasing use of cluster tools (Figure 6-1). In these, three or four process steps are arranged around a central wafer handling mechanism that is fed through a single load lock, which lends itself to yield improvement thus reducing overall costs. Such configurations also lower the area of clean room floor space required for equipment. A process that isolates the wafer from the factory environment throughout the manufacturing process will alleviate the need for such a large area of expensive clean room facilities as well as the need to isolate workers from the factory environment, all of which are expensive overhead items. A way of achieving this is to more fully integrate the manufacturing process.
Achieving integrated processing requires elimination of the numerous inspections that are now carried out to ensure that only "good" wafers are carried forward in the fabrication process. These inspections are now necessary to minimize the cost of processing unacceptable wafers and assure the proper functioning of the various process steps. Elimination of inspections can be achieved if sufficient in situ monitors can be inserted into the process steps and the output from these monitors can be fed into a real-time process control network. This represents a major advance from the way that most process steps are now carried out. As noted in Chapter 6, most process tools monitor and control the input variables to the machine (e.g., current, voltage, gas flow). At the present time, except for a few cases, such as measurement of ion current in ion implantation and end-point detection in plasma etching (by optical techniques), no direct monitoring is performed for what is happening at the wafer.
Development of real-time process monitors requires, for each process step, establishing the critical parameters to be measured, and then identifying the appropriate noninvasive sensors that must be developed; work on such sensors is important to this technology. To use the output from these sensors for real-time process control, process simulation models must be developed and combined with expert system artificial intelligence technology to develop "smart" real-time process controls.
Another problem to be overcome is maximizing the throughput of an integrated process. Different process steps require different operation times and may be used more than once in the fabrication of a given circuit. Using current technology, a typical fabrication facility for a 16 M-bit dynamic RAM (DRAM) circuit production of 3 million units per month will consist of 50 lithography stations, 35 ion implanters, 60 etchers, 20 sputtering units, etc. Such a multiplicity of tools also allows for redundancy, which permits continued production when a single tool breaks down or is inoperative for any reason. Proper architecture of an integrated system must accommodate these situations. A point to be made here is that each process must have sufficient uptime to ensure long continuous runs of the entire system. Otherwise the amount of redundancy that must be built into the integrated process will increase the cost, the size of real estate occupied by the system, and the complexity of the control processor. These factors could render total integration unachievable. Another feature needed for achieving integrated systems is to have the various vendors' machines capable of being joined to each other or to some common third body. In addition, each machine's microprocessor should be capable of talking to the system computer.
To achieve fully integrated semiconductor processing, any liquid steps, such as spin-on photoresist or dips to clean up etching or deposition steps, most probably will have to be replaced by dry processes. The time to develop new resists and application systems will undoubtedly make this the last operation to be integrated.
Implementation of the above advances will require close cooperation among the process engineers, equipment vendors, and process designers. It is most likely that additional wafer area will be required for process monitoring. Also, equipment will have to be redesigned to accommodate line-of-sight to such areas. These modifications will occur only if the overall economics are favorable. The example of integrated processing of integrated circuits has been discussed in some detail because the same generic problems apply to integrated processing of other materials. Determining what properties to measure, and developing the sensor systems and process simulation models, will still be required, as will development of "smart" control algorithms. One significant difference between silicon and many other materials relates to the size and shape of the material being processed. Whereas the small size of silicon wafers allows development of single-process modules connected serially or in clusters, this may not be the case for other materials. As was shown in Chapter 6, several applications involve large sheets or continuous ribbons of material that must be manufactured by in-line processes.
A more interesting case of integrated beam processing is the use of various energy beams to shape and harden a gear. Here, measurement of significant variables is easier because the workpiece is not in a protected environment. Processes for other materials, requiring as many steps as found in silicon device fabrication, have yet to be developed, but, as application of beam technologies increases, these materials will follow the lead of silicon processing. Nevertheless, it is recognized that modular integrated processes offer the opportunity to achieve process upgrades by changing modules rather than building whole new factories. In addition, for small fabrication runs, software-controlled flexibility would allow fewer tools to do a broader range of processes.
Semiconductor International. December 1989. Industry News. p. 22-23.