Dynamic inline yield analysis and prediction of a defect limited yield using inline inspection defects
In one embodiment, a method for predicting yield includes calculating a criticality factor (CF) for each of a plurality of defects detected in an inspection process step of a wafer, and determining a yield-loss contribution of the inspection process step to the final yield based on CFs of the plurality of defects and the yield model built for a relevant design. The yield-loss contribution of the inspection process step is then used to predict the final yield for the wafer.
1. A computer-implemented method comprising:
calculating, using a computer, a criticality factor (CF) for each of a plurality of defects detected in at least one inspection process step of a wafer, the wafer comprising a plurality of dies;
determining a non-defect yield component of a final yield for the wafer using a subset of the plurality of dies that have no visual defects;
determining a yield-loss contribution of the inspection process step to the final yield based on the CFs of the plurality of defects, a yield model built for a relevant design, and the non-defect yield component; and
predicting the final yield for the wafer using the yield-loss contribution of the inspection process step.
2. The computer-implemented method of claim 1 , further comprising determining an adjusted number of wafer starts based on the predicted final yield.
3. The computer-implemented method of claim 1 , further comprising building the yield model for the relevant design by:
distributing defects among a number of inspection steps; and
for each step:
classifying the defects according to similar CF values into one or more CF bins;
creating a second subset of dies from the subset of the plurality of dies, wherein the subset is a set of dies having inspection results in all inspection steps, wherein the second subset is a set of dies having defects only in a specific inspection step;
binning together the dies in the second subset according to a similar CF value to create one or more sub-subsets;
for each sub-subset, calculating a defect-specific fail probability (FP defects s,b ); and
plotting the defect-specific fail probability in a functional form, FP defects =f(CF), to create a correlation curve.
4. The computer-implemented method of claim 3 , wherein calculating the defect-specific fail probability (FP defects s,b ) for each sub-subset is calculated as:
F
P
defects
s
,
b
=
1
-
1
c
·
Y
DirtyDies
SubsetC
=
1
-
1
c
·
GoodDiesInSubsetC
AllDiesInSubsetC
.
5. The computer-implemented method of claim 4 , wherein c is a constant value representing the non-defect yield component for the subset of dies having no visual defects calculated as a ratio of passing dies having no visual defects to all dies having no visual defects.
6. The computer-implemented method of claim 1 , wherein calculating the CF comprises:
identifying an open area, wherein the open area is a total area of all locations inside a design layout area to cause an open failure, and identifying a short area, wherein the short area is a total area of all locations inside the design layout to cause a short failure; and
calculating a probability of the defect to cause an open failure and a probability of the defect to cause a short failure.
7. The computer-implemented method of claim 6 , wherein the probability of the open failure is a ratio between the open area and a design layout window defined by a location inaccuracy of an inspection tool.
8. The computer-implemented method of claim 6 , wherein the probability of the short failure is a ratio between the short area and the entire design layout area.
9. The computer-implemented method of claim 1 , wherein determining the yield-loss contribution of the process step to the final yield comprises:
for each defect on an inspected layer, extracting the defect-specific fail probability value (FP defects ) from a correlation curve corresponding to the step; and
computing the yield-loss contribution to final yield as:
Y
L
defects
s
=
1
-
∏
b
=
1
B
s
(
1
-
F
P
defects
s
,
b
×
N
s
,
b
N
)
N
b
.
10. The computer-implemented method of claim 1 , wherein predicting the final yield for a wafer that has completed all inspection process steps comprises:
using the defect-specific fail probability value (FP defects ) from the correlation curve and a defective die count per CF bin to compute the final yield as:
Y
final
=
c
·
∏
s
=
1
S
∏
b
=
1
B
s
(
1
-
F
P
defects
s
,
b
×
N
s
,
b
N
)
N
b
.
11. The computer-implemented method of claim 1 , wherein predicting the final yield for an inline wafer comprises:
using the defect-specific fail probability value (FP defects ) from the correlation curve and a defective die count per CF bin to compute the final yield as:
Y
final
=
c
·
∏
s
=
1
M
∏
b
=
1
B
s
(
1
-
F
P
defects
s
,
b
×
N
s
,
b
N
)
N
b
·
∏
s
=
M
+
1
S
(
1
-
〈
Y
L
defects
s
〉
)
where YL defects s is an average yield-loss contribution over a historical time span.
12. The computer-implemented method of claim 1 , wherein there is more than one defect per die, further comprising:
aggregating all of the defects in a die into a virtual defect; and
for each virtual defect, calculating an aggregated CF value as:
CF
die
=
1
-
∏
i
=
1
DefectsPerDie
(
1
-
CF
i
)
.
13. A computer-implemented system comprising:
a memory;
processor coupled to the memory;
a model constructor executed from the memory by the processor to build a yield model for a design; and
a yield analyzer, coupled to the model constructor, to calculate a criticality factor (CF) for each of a plurality of defects detected in at least one inspection process step of a wafer, the wafer comprising a plurality of dies, to determine a non-defect yield component of a final yield for the wafer using a subset of the plurality of dies that have no visual defects, to determine a yield-loss contribution of the inspection process step to the final yield based on the built yield model and the criticality factors calculated for individual defects detected at different process steps, and the non-defect yield component, and to predict the final yield for the wafer using the yield-loss contribution.
14. The computer-implemented system of claim 13 , wherein the yield analyzer comprises:
a CF calculator, to calculate the CFs of the individual defects; and
a yield predictor, to calculate the yield-loss contribution of an inspection process step to a final yield using the CFs.
15. The computer-implemented system of claim 14 , wherein the yield-loss predictor to calculate the yield-loss contribution of the inspection process step to the final yield comprises correlating the CFs calculated for the defects in a specific process step to an observed die yield provided by the yield model.
16. The computer-implemented system of claim 13 , further comprising:
a yield model database, coupled to the model constructor, to store the yield model;
a design data database, coupled to the yield analyzer, to store design data, wherein the design data includes at least one of design layouts and routing information for the design layouts; and
a defect reporter, coupled to the yield analyzer, to provide geometric characteristics of individual defects detected during different process steps.
17. The computer-implemented system of claim 13 , wherein the CFs are calculated based on at least one of: a size of the defect, a volume of the defect, and a layout of the design on which the defect is found.
18. The computer-implemented system of claim 13 , wherein to build a yield model for a design uses sort yield data and the subset comprising a set of dies having inspection results for all critical process steps.
19. A non-transitory computer readable storage medium, comprising executable instructions which when executed on a processing system cause the processing system to perform a method comprising:
calculating a criticality factor (CF) for each of a plurality of defects detected in an at least one inspection process step of a wafer, the wafer comprising a plurality of dies;
determining a non-defect yield component of a final yield for the wafer using a subset of the plurality of dies, wherein the subset of dies has no visual defects;
determining a yield-loss contribution of the inspection process step to the final yield based on CFs of the plurality of defects, a yield model built for a relevant design, and the non-defect yield component; and
predicting the final yield for the wafer using the yield-loss contribution of the inspection process step.