Pattern-enhanced spatial correlation of test structures to die level responses
Enhancement of less dominant patterns for parametric wafer measurements. Dominant patterns are removed from the parametric pattern thereby revealing a less dominant pattern. The less dominant patterns can be used to identify root causes for yield loss that are not visible in the original parametric measurements.
1. A processor-based modeling method, comprising:
generating a first wafer map having a plurality of values corresponding to a multiplicity of parametric measurements obtained from a semiconductor wafer in a semiconductor manufacturing operation;
identifying a dominant pattern of the plurality of values on the first wafer map;
identifying the dominant pattern as a known spatial variation caused by a specific physical feature of the semiconductor wafer;
removing the dominant pattern from the first wafer map thereby generating a second wafer map and revealing a less dominant pattern of the plurality of values on the second wafer map;
training a machine learning model to determine the root cause on the basis of the less dominant pattern; and
deploying the machine learning model in a semiconductor equipment operating environment.
2. The method of claim 1 , further comprising:
identifying a plurality of dominant patterns in the plurality of values on the first wafer map; and
removing the plurality of dominant patterns from the first wafer map thereby generating the second wafer map.
3. The method of claim 1 , further comprising:
computing a statistical median or a statistical mean of the plurality of values on the first wafer map;
identifying a common spatial pattern of the plurality of values;
subtracting the statistical median or the statistical mean of the plurality of values and subtracting the common spatial pattern of the plurality of values from the plurality of values to thereby reveal the less dominant pattern.
4. The method of claim 3 , further comprising:
subtracting the statistical median or the statistical mean of the plurality of values, the common spatial pattern of the plurality of values, and the spatial variation caused by a specific physical feature from the plurality of values to thereby reveal the less dominant pattern.
5. The method of claim 1 , further comprising:
determining a correlation between the less dominant pattern and a bin yield pattern for the first semiconductor wafer.
6. A method, comprising:
obtaining a multiplicity of parametric measurements from a test structure formed on a semiconductor wafer in a semiconductor manufacturing operation;
plotting a plurality of values corresponding to the multiplicity of parametric measurements to generate a parametric wafer map image;
identifying at least one dominant pattern on the parametric wafer map image;
removing the at least one dominant pattern from the parametric wafer map image thus forming an enhanced wafer map image;
identifying, after the at least one dominant pattern is removed, a less dominant pattern of on the enhanced wafer map image;
training a machine learning model to determine the root cause on the basis of the less dominant pattern; and
deploying the machine learning model in a semiconductor equipment operating environment.
7. The method of claim 6 , further comprising:
identifying an additional dominant pattern in the multiplicity of parametric measurements on the parametric wafer map image; and
removing the at least one dominant pattern and the additional dominant pattern from the parametric wafer map image to form the enhanced wafer map image.
8. The method of claim 7 , further comprising:
computing a statistical median or a statistical mean of the parametric wafer map image;
identifying a common spatial pattern in the multiplicity of parametric measurements on the parametric wafer map image; and
subtracting the statistical median or the statistical mean and the common spatial pattern from each of the plurality of values on the parametric wafer map image.
9. The method of claim 6 , further comprising:
computing a statistical median or a statistical mean of the parametric wafer map image; and
subtracting the statistical median or the statistical mean from each of the plurality of values on the parametric wafer map image.
10. The method of claim 6 , further comprising:
identifying a spatial variation caused by a specific physical feature;
subtracting the statistical median or the statistical mean, the common spatial pattern and the spatial variation from each of the plurality of values on the parametric wafer map image.