Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Quality evaluation
Reexamination Certificate
1999-03-25
2002-05-14
Bui, Bryan (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Quality evaluation
C702S081000, C438S021000, C700S117000
Reexamination Certificate
active
06389366
ABSTRACT:
TECHNICAL FIELD
This invention relates generally to process control methods for use in manufacturing and, in particular, to process control methods that efficiently identify sources of patterns resulting from processing effects.
BACKGROUND
Many products and in particular semiconductor devices require multiple discrete processing steps to manufacture the products. For example, several hundreds of steps may be required to produce an integrated circuit from raw semiconductor material. The starting substrate is usually a slice of single crystal silicon referred to as a wafer. Multiple circuits, as many as several hundred, are fabricated on a single wafer and singulated later. Wafers may go through processing steps one at a time or, maybe processed together in batches called “lots” or “runs.”
Data gathered during the course of wafer processing is used to diagnose yield problems and forms the basis of yield improvement efforts. Such data includes parametric electrical test data gathered on individual circuits and test structures fabricated on the wafers, as well as final yield test data which tests the suitability for use once wafer processing is completed.
It has been recognized that one of the sources of yield variation is the order in which wafers in a lot are processed at a given processing step or the physical location of the wafer in batch processing equipment. The practice of tracking wafer processing order at critical processing steps and correlating this processing order to device performance in order to improve yields has become known as “Wafer Position Tracking.” The benefits of this practice are described, for example, in the paper by Scher et al., IEEE Transactions of Components, Hybrids, and Manufacturing Technology, Vol. 13, No. 3, pp 484-489 (1990).
In wafer position tracking, typically, a scribed identifying code on each wafer is read and recorded by specialized equipment. In many implementations, the wafers are placed in random order before critical processing steps to ensure effects from different steps are not compounded and the order of the wafers is recorded before and after randomization. When final yield data or other test data is plotted as a function of wafer order at critical processing steps, the signature of the piece of processing equipment responsible for any wafer-to-wafer variations in output often can readily be identified, facilitating corrective action by process engineers.
An example of wafer position tracking plots, referred to as scatter plots, is shown in
FIG. 1
for three processing steps. In this example, the monotonic increase in yield with wafer sequence number in the plot for step A indicates a “warm-up” effect in step A, while no positional dependence is observed in the wafer sequence of steps B and C. Statistical methods, such as correlation statistics, run statistics, and control charts, can be used to screen the scatter plots for systematic dependence on wafer sequence. Specific implementations of wafer position tracking are described, for example, in U.S. Pat. No. 5,716, 856 to Lin et al. and U.S. Pat. No. 5,761,065 to Kittler et al., both commonly assigned with the present application and incorporated herein by reference.
As described above, wafer position tracking typically uses summary statistics of whole wafers to track wafer-to-wafer variation in yield where yield may be considered the percentage of individual dies on a wafer that pass a final quality control test. Wafer position tracking is less sensitive to wafer-to-wafer variations affecting only part of a wafer. In addition, some processing problems, instead of introducing wafer-to-wafer variation, may introduce patterns of yield variations as a function of physical position on a single wafer. Wafer position tracking will not reveal a yield variation that appears as the same pattern on each wafer.
To address variation within a single wafer, a sector version of wafer position tracking has been introduced. In sector tracking, the wafer surface is divided into sectors, for example, into the pattern of nine sectors illustrated in FIG.
2
. Then wafer position tracking is performed for each sector individually. However, sector wafer position tracking may not be an optimal process control solution. The division into sectors is arbitrary and may not reveal a problem yield pattern in any particular lot. Performing sector position tracking multiplies the amount of work performed in process control by the number of sectors, in this example by about an order of magnitude.
Furthermore, massive amounts of data are potentially available for process control. After processing is completed, multiple properties of individual circuits on each wafer can now be measured and recorded routinely. In some cases, data for individual circuits is recorded at intermediate processing steps, as well. Typical numbers of circuits per wafer are several hundreds of circuits aid typical batch sizes are between 12 and 50 wafers per batch, with 24 wafers per batch quite common.
What is needed is a method to recognize both wafer-to-wafer and within-wafer variations in output induced by processing effects. What is needed is a method of data reduction to take advantage of the massive amounts of data collected during wafer fabrication for process control of both wafer-to-wafer and within-wafer variations in yield.
SUMMARY
Methods are provided to identify within-wafer and wafer-to-wafer yield variations resulting from processing steps in a multi-step manufacturing process. The methods are implemented along with wafer position tracking for controlling the manufacturing process.
According to an embodiment of the present invention, wafer position tracking with rotation reveals a static pattern in yield data, or other site-specific processing data, present on each wafer independent of position in a batch. In wafer position tracking with rotation, after wafers are ordered and the order recorded, as in previous implementations of wafer position tracking, the wafers are systematically rotated according to their position in the batch before entering a processing step. When a visual representation of yield data for each wafer is displayed in the wafer sequence order of the processing step which is responsible for introducing the static pattern, the pattern will appear in a successively rotated position on successive wafers. A process operator can, thus, readily identify the process step responsible for introducing the static pattern and take corrective action.
According to another aspect of the present invention, data reduction methods are provided to provide a compact representation of site-specific data. For each wafer, measurements on a discrete grid of locations, such as the location of each circuit on a wafer are transformed into a continuous function, termed a wafer function. A distance matrix and an angular rotation matrix are calculated from the wafer functions. Each element of the rotation matrix gives the rotation angle by which one wafer would be rotated to have a pattern as similar as possible to the other wafer. Each element of the distance matrix is a measure of the difference between the patterns on the two wafers when one or both wafers are rotated to realize this most similar pattern.
From the distance and angular rotation matrices, multidimensional scaling methods are applied to determine distance factor scores and angle factor scores. The distance factor scores and angle factor scores are the coordinates of vectors of very low dimension and thus provide a compact representation of the site-specific data at many grid points. The distance factor scores track changes in pattern on the wafers. Wafers similar in pattern after rotation have similar distance factor scores. The angle factor scores track rotation of patterns on wafers. Wafers that have patterns that are similar with little or no rotation have angle factor scores that are close to each other. The angle factor scores are preferably used to reveal patterns in data obtained with systematic rotation as described above for wafer position tracking with r
Bui Bryan
Kwok Edward C.
Skjerven Morrill & MacPherson LLP
LandOfFree
Methods for identifying sources of patterns in processing... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Methods for identifying sources of patterns in processing..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Methods for identifying sources of patterns in processing... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2865068