Skip to main navigation Skip to main content
  • E-Submission

JKSPE : Journal of the Korean Society for Precision Engineering

OPEN ACCESS
ABOUT
BROWSE ARTICLES
EDITORIAL POLICIES
FOR CONTRIBUTORS
REGULAR

심층 신경망을 이용한 패드 표면 거칠기 기반 CMP 재료 제거율 예측

Prediction of CMP Material Removal Rate based on Pad Surface Roughness Using Deep Neural Network

Journal of the Korean Society for Precision Engineering 2023;40(1):21-29.
Published online: January 1, 2023

1 부산대학교 대학원 기계공학부

1 School of Mechanical Engineering, Graduate School, Pusan National University

#E-mail: hdjeong@pusan.ac.kr, TEL: +82-51-510-3210
• Received: October 4, 2022   • Revised: November 1, 2022   • Accepted: November 1, 2022

Copyright © The Korean Society for Precision Engineering

This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

  • 12 Views
  • 3 Download
  • 2 Crossref
  • 3 Scopus
prev next

Citations

Citations to this article as recorded by  Crossref logo
  • Precision Engineering and Intelligent Technologies for Predictable CMP
    Somin Shin, Hyun Jun Ryu, Sanha Kim, Haedo Jeong, Hyunseop Lee
    International Journal of Precision Engineering and Manufacturing.2025; 26(9): 2121.     CrossRef
  • Prediction of Normalized Material Removal Rate Profile Based on Deep Neural Network in Five-Zone Carrier Head CMP System
    Yonsang Cho, Myeongjun Kim, Munyoung Hong, Joocheol Han, Hong Jin Kim, Hyunki Kim, Hyunseop Lee
    International Journal of Precision Engineering and Manufacturing-Green Technology.2025; 12(3): 869.     CrossRef

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

Include:

Prediction of CMP Material Removal Rate based on Pad Surface Roughness Using Deep Neural Network
J. Korean Soc. Precis. Eng.. 2023;40(1):21-29.   Published online January 1, 2023
Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:
Include:
Prediction of CMP Material Removal Rate based on Pad Surface Roughness Using Deep Neural Network
J. Korean Soc. Precis. Eng.. 2023;40(1):21-29.   Published online January 1, 2023
Close

Figure

  • 0
  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
Prediction of CMP Material Removal Rate based on Pad Surface Roughness Using Deep Neural Network
Image Image Image Image Image Image Image Image Image Image Image
Fig. 1 Schematic of CMP experiment set up
Fig. 2 Three-dimensional confocal image of pad surface deformation over time
Fig. 3 Comparison of CMP results in parameters over time by pressure
Fig. 4 Schematic of pad measurement process
Fig. 5 Schematic of RCA measurement system
Fig. 6 RCA image processing and analysis
Fig. 7 Actual parametric data scatterplots for deep learning-based modeling
Fig. 8 Deep Neural Network structure diagram
Fig. 9 Result of R2 according to optimized DNN training
Fig. 10 Validation of DNN with test data set
Fig. 11 Evaluation loss function according to epoch of optimized DNN
Prediction of CMP Material Removal Rate based on Pad Surface Roughness Using Deep Neural Network

Break-in conditions

Polishing machine POLI-500
(G&P Technology)
Wafer 200 mm Oxide blanket
Pressure Head [g/cm²] 210
Retainer ring [g/cm²]
Velocity Head [RPM] 87
Platen [RPM] 93
Slurry/DIW flow rate [ml/min] 150
Polishing time with DIW [min] 20
Polishing time with slurry [min] 10
Conditioning Ex-situ
Conditioning Sweep cyc [cyc/min] 9
Rotation speed [RPM] 101
Down force [kgf] 4
Time [min] 5

Experimental conditions

Polishing machine POLI-500
(G&P Technology)
Wafer 200 mm Oxide blanket
Pressure Head [g/cm²] 140, 210, 280, 350
Velocity Retainer ring [g/cm²] 210, 280, 350, 420
Head [RPM] 87
Platen [RPM] 93
Slurry flow rate [ml/min] 150
Polishing time [min] 1
Conditioning Ex-situ
Conditioning Sweep cycle [cyc/min] 9
Rotation speed [RPM] 101
Down force [kgf] 4
Time [sec] 10, 30, 60, 180, 300

Range of experiment result value for data construction

Data feature Value
Minimum Maximum
Pressure [g/cm²] 140 350
Reduced peak height [μm] 1.643 6.308
Real contact area [%] 0.065 1.834
Material removal rate [Å/min] 694 3356

Number of partitioned data set

Data set Value
Train 132
Validation 44
Test 44

Range of hyperparameter tuning

Hyper parameter Value
Minimum Maximum Step size
Hidden layer 0 10 1
Number of neurons 1 150 1
Batch size 2 64 2
Learning rate 3e-4 3e-1 1e-18
Epochs 1,000 10,000 500
Drop out 0.1 0.9 0.1

Result of random search

Hyper parameter Value
Hidden layer 6
Number of neurons 57
Batch size 8
Learning rate 0.000319
Epochs 5,500
Drop out 0.1

Test data with a prediction error rate above MAPE

No. Pressure
[g/cm²]
RCA
[%]
Rpk
[μm]
Real MRR
[Å/min]
Predicted MRR
[Å/min]
Prediction error rate
[%]
1 210 0.4766235 4.543 1,624 1,785.6068 9.951158
2 280 1.1841390 3.180 1,731 1,602.0144 7.451508
3 140 0.1846151 5.175 1,226 1,038.1718 15.320412
4 210 0.4943034 4.586 1,793 1,657.3381 7.566194
5 210 0.4660441 2.709 1,191 1,375.7162 15.509336
6 280 0.8840373 3.013 1,866 1,666.5496 10.688662
7 280 0.6425659 3.729 2,107 2,308.8945 9.582085
8 140 0.1851733 3.928 864 936.416 8.381483
9 280 1.1594971 3.056 1,340 1,571.9052 17.306355
10 140 0.3186116 4.454 843 951.67255 12.891168
11 210 0.3302002 3.946 1,396 1,759.6122 26.046718
12 210 0.4329015 3.048 1,624 1,473.16 9.288175
13 350 1.5340251 2.086 1,821 2,112.963 16.033108
14 350 1.4431559 2.282 1,877 2,339.1433 24.621380
Table 1 Break-in conditions
Table 2 Experimental conditions
Table 3 Range of experiment result value for data construction
Table 4 Number of partitioned data set
Table 5 Range of hyperparameter tuning
Table 6 Result of random search
Table 7 Test data with a prediction error rate above MAPE