Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Quality evaluation
Reexamination Certificate
2001-12-11
2004-09-14
Pert, Evan (Department: 2829)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Quality evaluation
C438S014000
Reexamination Certificate
active
06792366
ABSTRACT:
CROSS-REFERENCES TO RELATED APPLICATIONS
NOT APPLICABLE
BACKGROUND OF THE INVENTION
The present invention relates generally to semiconductor manufacturing, and more specifically to inspection of semiconductor wafers for defects.
In the manufacturing of semiconductor wafers, inspection of the wafer is executed after each manufacturing process, the idea being to improve yields by detecting process changes that might negatively affect yield. Process variations manifest themselves as defects in and on the wafer. An initial inspection of each wafer is made to detect that there are defects at all. A variety of inspection tools are available for making such inspections. The kinds of defects that can be detected by these inspection techniques include, for example, pattern short-circuits, pattern disconnections, foreign materials, depressions, scratches, separation of conductor, unevenness, etc. The initial inspection can only identify that defects exist, the number of defects, their locations on the wafer, etc., but generally cannot identify the kinds of defects. Consequently, the defects are subjected to a detailed review subsequent to the initial inspection using technologies including conventional optical microscopes, scanning electron microscopes (SEMs), and the like in order to determine the specific kinds of defects (e.g., shorts, disconnections, and so on). This additional detailed information facilitates an understanding of the causes of the defects, to detect that the process is changing, how the process is changing, and how to adjust the processes accordingly to avoid such defects.
Ideally, a detailed inspection of all defects detected on all semiconductor wafers coming off the production line is made in order to provide as complete an understanding as possible of the manufacturing process. However, such a brute force approach is not feasible or practical due to the large numbers of wafers that are produced and the large numbers of defects per wafer that can occur. Consequently, in practice, the review task is limited to a small population of defects selected from among all of the detected defects. The smaller population of defects are then subjected to further detailed review to gain an understanding of the manufacturing process and to detect process variations, albeit a less accurate understanding.
A conventional technique for selecting defects for inspection is shown in
FIG. 11. A
semiconductor wafer
1101
is initially inspected to detect the existence of defects (including location of the defect on the wafer), which are indicated by the diamond shapes and identified generally by reference numeral
2
. The defects (which can number in the thousands) are collected and stored as defect data
1112
, for example in a database. An area of interest
1122
on the semiconductor wafer representation
1103
serves as a selection criterion for obtaining a small set of defects from among all of the defects. The area of interest is input by a user as a specified region
1114
. A defect selection process
1116
makes a selection of defects from the defect data in accordance with the specified area of interest, selecting all of the defects within the user-specified area. The selected defect data
1118
are then produced as an output to the user in a suitable format. As can be seen in the figure, semiconductor wafer
1101
′ illustrates the selected defects indicated by the filled diamond shapes and identified generally by reference numeral
3
. The selected defects are then subjected to closer more detailed examination to determine the kind of defect, distribution patterns of the different kinds of defects, etc. to gain a further understanding of the condition of the manufacturing process.
Another conventional technique for selecting defects is shown in
FIG. 12. A
semiconductor wafer
1201
is initially inspected to detect the existence of defects
2
. The defects are collected as defect data
1212
. The defects for further detailed study are selected purely at random. The number of defects selected is based on a selection criterion known as a selection ratio
1214
which is input to a selection process
1216
. The selection ratio is simply a percentage value of the total defects; for example a selection ratio of 0.3 would refer to 30% of the total detected defects. The defects are randomly selected until the ratio of the number of selected defects to the total number of detected defects is equal to the selection ratio; this constitutes the selected defect data
1218
. The selected defects are then presented to the user in a suitable format.
FIG. 12
shows a typical random pattern of selected defects
3
′ on semiconductor wafer
1201
′. A subsequent detailed examination of the selected defects is then made to identify process variations as discussed above in connection with FIG.
11
.
A selection criterion based on a particular region of the semiconductor (such as in
FIG. 11
) is likely to be inadequate for process variations in which the process error is location-specific. Wafer defects will tend to have a non-uniform distribution pattern. A selection of defects based on a user-specified region of the semiconductor wafer can result in a low number of selected defects if the defects are clustered in a region of the semiconductor wafer other than the region specified by the user. Consequently, the subsequent detailed inspection of the sampled defects will not provide an accurate indication of the process error.
A selection criterion based purely on a random sampling of the entire wafer as explained in connection with
FIG. 12
can mask out the effect of process errors which result in location-specific defects. When defects for further detailed study are selected purely at random, the non-uniformity of distribution of the defects is not reflected in the selected defects. Consequently, it is difficult to ascertain certain information such as the distribution of the different kinds of defects on the wafer.
Some process errors result in a distribution of defects on the semiconductor wafer where the defects are grouped in clusters having varying densities of defects; for example, some defects may be located in a region of densely distributed defects (hereinafter referred to as the “densely distributed region”) and other defects may be located in a region of sparsely distributed defects (hereinafter referred to as the “sparsely distributed region”) have different causes and different classified-by-kind defect percentages. In such cases, the selection of defects for further detailed inspection may be conducted separately for clusters of defects of different densities. In the method of selecting a particular cluster, an operation of dividing the defects into those of the densely distributed region(s) and those of the sparsely distributed region requires the decision making of an operator. Due to the subjective nature of this approach, it is likely that the defined clusters will vary from operator to operator. On the other hand, if the selection is made randomly, the majority of the selected defects will tend to be those defects located in the more densely distributed region(s). The defects in the sparsely distributed region are not likely to be adequately sampled. Therefore, it is difficult to gain an understanding of process errors when defects are clustered in this way.
There is a need to provide an inspection semiconductor wafers which improve on shortcomings of conventional semiconductor wafer inspection methods and systems.
SUMMARY OF THE INVENTION
A semiconductor wafer inspection and review method and system comprise detecting defects in a semiconductor wafer and storing information relating to the detected defects in an appropriate data store. Along with the detected defects, additional information relating to clusters of defects on the semiconductor wafer is produced and stored. The defects are sampled based on statistical criteria to produce sampled defects. Subsequent detailed inspection and analysis of the sampled defects produce information relating to the su
Hosoya Naoki
Obara Kenji
Shibuya Hisae
Takagi Yuuji
Pert Evan
Townsend and Townsend / and Crew LLP
LandOfFree
Method and apparatus for inspecting defects in a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for inspecting defects in a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for inspecting defects in a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3207941