矽加工中的表徵（簡體書）

Although it is much more a function of etch parameter variations than ofpolysiliconmaterials properties, another problem which frequently arises is that of variations inthe cross-sectional profile of the etched potysilicon lines. The standard procedure formonitoring this problem is to prepare samples destructively by cleaving etched wafersperpendicular to the poly lines, coating the samples, and then evaluating them in crosssection in an SEM. The time and wasted material in this process can be significantlyreduced by using LS, which has been shown to produce signatures that are extremelysensitive to dimensions and angles in periodic fields of etched structures. This technique requires only the momentary reflection of a light beam from the area of interestto produce a signature which can be compared to that of a properly etched area.
2.5 Subsequent Processing
During most of the subsequent processing steps that polysilicon layers may see, theonly process input which produces any change is the time-temperature product whichresults in grain growth and dopant diffusion effects. The two exceptions are thedeposition, anneal, and oxidation steps involved in the formation of polycide (metalsilicide on polysilicon) structures and the dielectric encapsulation of the polysilicon.
Polycides
The polycide formation process will be illustrated with the example of a tungsten-polycide system. Various other metal-silicide systems are used, but many of thematerials-related problems are similar. The silicide film is formed either by the deposition of a pure metal film and an anneal to form the silicide layer or by direct deposition of the metal silicide onto the polysilicon. In the tungsten-polycide case, it isusually done through the CVD ofWSi2.2. (The excess silicon will be depleted duringa subsequent oxidation step.) Two important materials issues in this step are that thepolysilicon surface must be clean so that the silicide layer will adhere to it and thatthe deposition system must have a low background pressure so that tungsten oxidesare not formed during deposition. The glaze discussed previously, which is formedduring some types of doping, must be removed, as must any other contaminantspicked up after polysilicon deposition. For the level of surface contamination controlrequired here, AES is typically the best tool to check cleanliness. Volume contamina-tion in the deposited silicide, principally oxygen, picked up from poor depositionsystem base pressures may also be found by AES, although post-deposition profilingby dynamic SIMS may be required to see the very low levels of oxygen which canbegin to form tungsten oxides. Monitoring the Si/W ratio, which can be of concern,is best done by RBS, since differential sputtering effects between the tungsten andthe silicon make all sputter-profiling techniques problematic.