Joint Symposium 2017 – eMDC & ISSM

ID t em p l at e - Jo in t 2 017

To improve the Measure Fail Rate On SEM through Six Sigma DMAIC methodology -
Chien-Hui Lu
Ricky Lu1, CC Chao2 , Yi Cheng
[email protected], [email protected]
[ Powerchip Technology Corporation ]
[ No.12, LI-HSIN RD. 1, HSINCHU SCIENCE PARK, HSINCHU, TAIWAN, R.O.C ]
Phone: +03 -5795000 Fax: +03-5792142
Keywords: CD-SEM , Six Sigma , DMAIC
proceed continuously the quality improvement with Six
ABSTRACTS Sigma Methodology for other defect improvement in the
future.
A Critical Dimension SEM (CD-SEM: Critical Dimension
Scanning Electron Microscope) is a dedicated system for INTRODUCTION
measuring the dimensions of the fine patterns formed on a
semiconductor wafer. CD-SEM is mainly used in the C D - SE M ( cr it i c al -d i me n s ion s c a nn in g e l e ct ron
manufacturing lines of electronic devices of microscope) is an essential tool for measuring the fine
semiconductors. The measurement process includes OM pattern dimensions formed in semiconductor processes
Alignment, SEM Alignment, Addressing, Measurement, that require to fabricate high-quality semiconductor
and Image Saved. From May 2015, we found CD-SEM devices. Hitachi has developed a diverse range of
measure fail rate ( measured by S-9380 SEM ) is too high elemental technologies to address the needs of ever
( its point reject rate is 46.13% ). High measure fail rate smaller dimensions(1), and many of these innovations
will reduce the equipment capacity utilization, and increase were incorporated in Hitachi’s conventional CD-SEM. The
manufacturing costs. So , we apply six-sigma DMAIC measurement process includes OM Alignment, SEM
methodology to improve CD SEM fail rate that is our Alignment, Addressing, Measurement, and Image Saved.
primary task , this is big Y. According to entitlement rules, From May 2015, we found CD-SEM measure fail rate
we make a target to improve CD SEM fail rate from ( measured by S-9380 SEM ) is too high ( its point reject
46.13% to 15.4%. Drilling down CD SEM fail ERR ( Error rate is 46.13% ). High measure fail rate will reduce the
code reject rate) and analyze the data, the most of equipment capacity utilization, and increase manufacturing
measurement errors are in addressing code (9007) costs. To reduce measure fail rate and to tackle this
=43.731%. We choice addressing code (9007) improvement problem, the Six Sigma Methodology of DMAIC (Define,
as our small y. Measure, Analyze, Improve, Control) is deployed on it.
To improve this problem, the Six Sigma approach of
DMAIC (Define, Measure, Analyze, Improve, Control) is SIX SIGMA METHODOLOGY
deployed on this project. There are several tools applied in
this process improvement, such as SIPOC, Detailed Map, The Six Sigma Way developed and promoted by Motorola

Cause & Effect Matrix analysis, ANOVA, and DOE. Initial is a philosophy for company-wide quality improvement.

investigation indicates several factors in addressing of The Six Sigma level of the performance means a product

Photo process contribute this measurement fail defect rate of 3.4 per million opportunities for defects. Six

phenomenon. By Cause & Effect Matrix, several factors Sigma approach is defined as “a disciplined method of

are clarified as most possible root causes, they are using extremely rigorous data gathering and statistical
X1:Magnification ࠊ X2:Algorithm ࠊ X3:Kind ࠊ and X4: analysis to pinpoint sources of defects and ways of

Method. After the statistical comparisons, X1: eliminating them. “. Six Sigma project starts with a

Magnification and X4: Method are significant & important business strategy or a customer requirement, and the most

factors. DOE is conducted to figure out the optimized widely used Six Sigma approach for key process

process control conditions and the results showed improvement is the DMAIC approach

significant difference from original conditions. (Define-Measure-Analyze-Improve-Control), which

With such implementation, the CE-SEM measurement fail integrates Tests of Statistical Significance, Correlation and

rate is improved from 46.13% to 1%. The total benefits Regression Analysis, Design of Experiments (DOE),

lead to 39 million cost saving! It is a good practice to Response Surface Methods (RSM), Statistical Process

- 1 -
Control (SPC), Control Charts, Failure Modes and Effect
Analysis (FMEA), and many other statistical tools. Six
Sigma approach can result in increased market penetration,
higher productivity, and lower overall costs of
manufacturing and services. The main subject of this study
is to apply Six Sigma approach to reduce measure fail rate
On SEM. Fig.3 The project output and goal

3. DEFINE PHASE 3. MEASURE PHASE

Define the Problem. What problem would you like to fix? Map Out the Current Process. How does the process

The Define Phase is the first phase of the Lean Six Sigma currently perform? Measurement is critical throughout the

improvement process. In this phase, the leaders of the life of the project and as the team focuses on data collection

project create a “Project Charter”, create a high-level view initially they have two focuses: determining the start point

of the process, and begin to understand the needs of the or baseline of the process and looking for clues to

customers of the process. This is a critical phase of Lean understand the root cause of the process. Since data

Six Sigma in which your teams define the outline of their collection takes time and effort it’s good to consider both at

efforts for themselves and the leadership (executives) of the start of the project.

your organization. By means of the process map and C&E matrix (as shown

From May 2015, we found CD-SEM measure fail rate in Fig.4), we obtained four possible factors. They are
X1:MagnificationˣX2:AlgorithmˣX3:Kindˣand X4: Method.
( measured by S-9380 SEM ) is too high and its point reject
rate is 46.13%. The problem is described in 5W1H way as The Gauge R&R (as shown in Fig.5 ) is carried out to

shown in Table.1. High measure fail rate will reduce the assess the measurement system.. The Gauge R&R result

equipment capacity utilization, and increase manufacturing showed that:

costs. Then, we use SIPOC to define project scope and 1. All of the points in the chart are outside the control

project goal to reduce Error code 9007 (Addressing fail ) limits and the variation is mainly due to part-to-part

reject rate from 43.731% to 13%. The SIPOC and the differences. It shows that this measurement system could

project output goal are shown in Fig.2 & Fig.3. distinguish the difference between the parts.
2. All ranges are within the R chart’s control limit, it is

Table.1 Problem Description in 5W1H way consistent.

3. The part-to-part variation is major source of the total
variation.
4. No interaction effect between three certified operators.
5. %Gauge R&R of system is 19%. The variation for
repeatability and reproducibility is less than 10%. hence,
the measurement system is accepted.

Fig.2 The SIPOC of the project.

Fig.4 The C&E matrix

- 2 -
 5. IMPROVE PHASE

Implement and Verify the Solution(s). How will you fix

the problem? Once the project teams are satisfied with their
data and determined that additional analysis will not add to
their understanding of the problem, it’s time to move on to
solution development. The team is most likely collecting
improvement ideas throughout the project, but a structured
Fig.5 The GR&R result improvement effort can lead to innovative and elegant
solutions.
4. ANALYZE PHASE According to 2 significant factors in A phase, we conduct
a two-level full factorial designs. Total 8 wafers are
Identify the Cause of the Problem. What does your data
measured for this experiment and there are 10 scores for
tell you? This phase is often intertwined with the Measure
each wafer. The Fig.8 & Fig. 9 shows that the key
Phase. As data is collected, the team may consist of
significant factors are X1 (code A ) and X2 (code B) and
different people who will collect different sets of data or
the interaction between X1 and X2 are found. The best
additional data. As the team reviews the data collected
setting for ( X1 , X2 ) are ( 26K , fast method ) , as shown in
during the Measure Phase, they may decide to adjust the
Fig. 10. The Confirmation experiment buy off the best
data collection plan to include additional information. This
setting and its result is shown in Fig. 11.
continues as the team analyzes both the data and the
With such implementation, the CE-SEM measurement
process in an effort to narrow down and verify the root
fail rate is improved from 46.13% to 1%. The total
causes of waste and defects.
benefits lead to 39 million cost saving! It is a good
Collecting inline defect data and using Box plot and
practice to proceed continuously the quality improvement
T-test , we obtained that X1 : Magnification ( P-value <
with Six Sigma Methodology for other defect improvement
0.05 as shown in Fig.6 ) and X2 : Method ( P value < 0.05
in the future
as shown in Fig.7 ) are two significant factors for y2.

.
Fig.6 2-sample T result for X1
Fig. 8 . the key factors are X1(code A) and X2 (code B)

Fig.7 2-sample T result for X2

Fig. 9 . the interaction plot between X1(code A) and X2

(code B)

- 3 -
 REFERENCES
1. J. Neter, W. Wasserman and M.H. Kutner, ”Applied
Linear Statistical Models ”, Second Edition. Irwin, Inc,
1985.
2. Coleman, D. E. and D. C. Montgomery, “A Systematic
Approach to Planning for a Designed Industrial
Experiment” (with discussion), Technometrics, Vol.
35, No. 1. ,1993.
3. Montgomery, D.C., “Introduction to Statistical Quality
Control”, Wiley, New York, 1991.
Fig. 10. The best setting for ( X1 , X2 ) are ( 26K , fast 4. Montgomery, D.C.,”Design and Analysis of
Experiments” ,Willy, New York, 2001.
method ). 5. Nair, V.N., “Taguchi’s Parameter Design: A Panel
Discussion”, Technometrics, 34, 127-161, 1992.
6. M. Harry and R. Schroeder, Six Sigma. The
Breakthrough Management Strategy Revolutionizing
the World’s Top Corporations. New York: Doubleday,
2000.
7. Chao-Ton Su Chia-Jen Chou Li-Fei Chen, ”
Application of Six Sigma Methodology to optimize the
performance of the Inert-Metal Dielectric Process “ ,
IEEE Trans. Semiconductor Manuf.,vol.22, No.2, pp.
2009 .
8. Forrest W. and Breyfogle III, “ Implementing Six
Sigma : Smarter Solution Using Statistical Methods“ ,
John Wiely & Sons, Inc, (1999)
9. J. E. Brady and T. T. Allen, “Six sigma literature: A
Fig. 11. The result of confirmation experiment. review and agenda for future research,” Quality and
6. CONTROL PHASE Reliability Eng. Int., vol. 22, no. 3, pp. 335-367, 2006.
10. K. Linderman, R. G. Schroeder, S. Zaheer, and A. S.
Choo, “Six sigma: A goal-theoretic perspective,” J.
Maintain the Solution(s). How do you sustain the newly Oper. Manage., vol. 21, no. 2, pp. 193-203, 2003.
achieved improvement? This phase is a mini version of 11. T. N. Goh and M. Xie, “Improving on the six sigma
process management. The team has been building a form of paradigm,” TQM Mag., vol. 16, no. 4, pp. 235-240,
2004.l
infrastructure throughout the life of the project, and during
the Control Phase they begin to document exactly how they
want to pass that structure on to the employees who work
within the process.
The control plan, and Failure Mode and Effect Analysis
(FMEA) table was develop and updated in this phase.

SUMMARY AND CONCLUSION

The DMAIC approach has employed to define the key

significant factors to reduce measure fail rate on CD-SEM.
The key significant factors are X1 :Magnification and
X2 :Method and the interaction between X1 and X2 are
found. The best setting for ( X1 , X2 ) are ( 26K , fast
method ). The Confirmation experiment buy off the best
setting and with such implementation, the CE-SEM
measurement fail rate is improved from 46.13% to 1%.
The total benefits lead to 39 million cost saving! It is a
good practice to proceed continuously the quality
improvement with Six Sigma Methodology for other defect
improvement in the future

- 4 -