Forward Selection and Best Subset Regression for Semiconductor Yield Enhancement

Forward Selection and Best Subset Regression for Semiconductor Yield Enhancement
George Chang, Jonathan Bass, Bala Pinnangudi*, Sreeram Venkataraman*
ON Semiconductor, Phoenix, Arizona 85008
*
Arizona State University, Tempe, Arizona 85281
Key words: Yield Enhancement, Forward Selection, Best Subset Regression
Abstract
Quick yield enhancement using limited data is essential for semiconductor industry. This is extremely critical
to catch up with today’s short product life cycle and global competition. There are several commercial available
packages for data mining and yield analysis. However, most of tools focus on identifying critical parameters from
huge data base. If number of parameters is much larger than number of sample, new method of data analysis is
necessary. In this paper, we have demonstrated using forward selection and best subset regression method to identify
critical parameters from huge suspected process parameters with limited number of wafer lots. Follow up design of
experiment lot confirmed critical parameter selection was a success and provided optimal process condition. Yield
was improved shortly after new technology was introduced to production. Critical data interpretation technique during
analysis will be discussed also.
Analysis examples:
Forward selection. Alpha-to-Enter: 0.05
Response is Yield on 20 predictors, with N = 176
Step
Constant
DNIT11-SPACER (DMP01)
T-Value
P-Value
1
12.0298
2
11.4356
3
13.5473
4
1.4726
5
0.4013
-0.02328
-12.98
0.000
-0.02202
-16.44
0.000
-0.02359
-24.19
0.000
-0.02346
-27.63
0.000
-0.02373
-28.78
0.000
-4.27
-11.87
0.000
-5.90
-20.35
0.000
-4.96
-17.54
0.000
-5.15
-18.48
0.000
-0.0263
-12.64
0.000
-0.0198
-9.87
0.000
-0.0181
-9.04
0.000
0.00286
7.48
0.000
0.00321
8.39
0.000
PCONTACT (PIV01) X Overlay
T-Value
P-Value
EPOLY12 (EPO09)
T-Value
P-Value
DWETOX14 (DME01)
T-Value
P-Value
Ptrench11(PCD01)
T-Value
P-Value
-0.73
-3.58
0.000
Best Subsets Regression:
Response is Yield-UIS
Vars
1
1
2
2
3
3
4
4
5
5
R-Sq
49.2
28.3
72.0
69.4
85.5
82.8
89.1
86.1
89.8
89.3
R-Sq(adj)
48.9
27.8
71.7
69.0
85.2
82.5
88.8
85.8
89.5
88.9
Mallows
C-p
708.0
1070.6
315.0
360.0
83.4
129.2
23.4
74.9
12.1
21.9
S
0.21827
0.25936
0.16250
0.16988
0.11733
0.12759
0.10213
0.11519
0.098775
0.10150
P
t
r
e
n
c
h
1
1
(
P
C
D
0
1
)
D
N
I
T
1
1
S
P
A
C
E
R
(
D
M
P
0
1
)
X
X
X
X
X
X
X
X X
X
P
C
O
N
T
A
C
T
D
W
E
T
O
X
1
4
E
P
O
L
Y
1
2
(
D
M
E
0
1
)
(
E
P
O
0
9
)
(
P
I
V
0
1
)
F
M
E
T
A
L
1
T
X I
O
v
e
r
l
a
y
(
F
R
S
0
1
)
X
X
X
X
X
X X
X
X X
X X
X
X
X
X X
X
X X
6
90.2
89.9
7.0
0.097055
X X X X X X