LVTS Dose&Focus recognition by image
0 likes430 views
The document discusses using image processing and morphology operations to better characterize features for focus exposure matrices compared to traditional CD SEM measurements. It shows that parameters like orientation, area, and elongation extracted from binary images provide better separation of dose and focus conditions than CD alone. A minimal mean distance classifier is proposed to recognize imaging sets based on these multi-dimensional morphology measurements. Further steps mentioned include collecting more focus exposure data and recipe optimization to fully validate the approach on production images.
![FE after morphology operation
FE
0.0
20.0
40.0
60.0
80.0
100.0
120.0
140.0
160.0
180.0
200.0
Focus
steps
Focus
CD's
Dose0
Dose1
Dose2
Dose3
Dose4
Dose5
Dose6
1. By replacing CD measurement by particle Area*sign[orientation]
measurement we getting strong separation between negative and
positive focus as well as better dose separation..
FE per Via Area combined Orientation
-6000
-4000
-2000
0
2000
4000
6000
1 2 3 4 5 6 7 8
Focus
Area*sign[Orientation]
D0
D1
D2
D3
D4
D5](https://image.slidesharecdn.com/lvtsprojectsdfrecognitionbyimage-141217001554-conversion-gate02/85/LVTS-Dose-Focus-recognition-by-image-9-320.jpg)
![Results
Focus0Dose0 Focus2Dose0 Focus1Dose5 Focus5Dose3 Focus5Dose0 Focus4Dose0
Focus0Dose1 Focus2Dose1 Focus0Dose0 Focus3Dose1 Focus3Dose0 Focus5Dose0
Focus0Dose3 Focus0Dose2 Focus2Dose1 Focus3Dose0 Focus5Dose1 Focus5Dose4
Focus0Dose2 Focus0Dose1 Focus2Dose3 Focus3Dose4 Focus3Dose3 Focus4Dose1
Focus0Dose4 Focus1Dose3 Focus0Dose5 Focus3Dose0 Focus4Dose1 Focus5Dose4
Focus3Dose5 Focus1Dose5 Focus0Dose2 Focus3Dose2 Focus4Dose2 Focus4Dose3
Focus0Dose0 Focus1Dose0 Focus3Dose1 Focus3Dose0 Focus4Dose0 Focus4Dose0
Focus0Dose1 Focus2Dose2 Focus1Dose0 Focus3Dose1 Focus3Dose0 Focus5Dose0
Focus1Dose5 Focus2Dose4 Focus2Dose1 Focus3Dose0 Focus5Dose1 Focus4Dose3
Focus0Dose2 Focus2Dose4 Focus2Dose3 Focus3Dose1 Focus3Dose3 Focus4Dose4
Focus0Dose4 Focus1Dose3 Focus2Dose5 Focus4Dose2 Focus4Dose4 Focus5Dose4
Focus0Dose5 Focus1Dose5 Focus1Dose4 Focus3Dose4 Focus5Dose3 Focus5Dose5
Focus0Dose0 Focus1Dose0 Focus3Dose1 Focus3Dose0 Focus4Dose0 Focus4Dose0
Focus0Dose1 Focus2Dose1 Focus2Dose1 Focus3Dose1 Focus3Dose0 Focus5Dose0
Focus0Dose3 Focus0Dose1 Focus2Dose1 Focus3Dose0 Focus5Dose1 Focus4Dose1
Focus0Dose2 Focus1Dose2 Focus2Dose3 Focus3Dose1 Focus3Dose3 Focus4Dose4
Focus0Dose4 Focus1Dose3 Focus2Dose5 Focus5Dose3 Focus4Dose4 Focus5Dose4
Focus0Dose5 Focus1Dose5 Focus1Dose4 Focus3Dose4 Focus5Dose3 Focus4Dose5
Sign [Orientation]*Area*Elongation
Sign [Orientation]*Area; Area; Area*Elongation
Sign [Orientation]*Area*Elongation; Area
For all cases high
probability to find
Dose/Focus value
within one step of
dose or/and focus
error](https://image.slidesharecdn.com/lvtsprojectsdfrecognitionbyimage-141217001554-conversion-gate02/85/LVTS-Dose-Focus-recognition-by-image-14-320.jpg)
LVTS Dose&Focus recognition by image
- 1. Is it possible to make the life of Litho Engineer easier or how to trace DOSE and FOCUS by image Vladislav Kaplan, July 2008
- 2. Agenda •Classical FE’s ▫Purpose ▫Mathematical presentation ▫Algorithms and models •Way of Control –CD (CDSEM) ▫Methodology of measurements •Is it another way…? ▫Methodology of Image Processing ▫Dose/Focus correlated parameters ▫Classification ▫SW presentation •Things to be concerned with…and next steps ▫Litho tool impact ▫Metrology tool impact ▫Layer/Mask impact
- 3. Classical FE FE 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.0 Focus steps Focus CD's Dose0 Dose1 Dose2 Dose3 Dose4 Dose5 Dose6 Classical FE’s • Choose most sensitive feature. (Could be more than one) • Run FE. • Define best dose and focus based on allowable process window. • Control it with CDSEM.
- 4. Classical FEFE 140.0 145.0 150.0 155.0 160.0 165.0 170.0 175.0 180.0 185.0 190.0 Focus steps Focus CD's Dose0 Dose1 Dose2 Dose3 Dose4 Dose5 Dose6 Wide process window. • PW – 160-180 nm • Dose2Focus5 optimal. • Still in PW up to +/-2 steps for focus and up to 1 step for dose fluctuation. Tight process window. • PW – 170-180 nm • Dose2Focus5 still optimal. • Still in PW up to +/-1 steps for focus and no allowable dose fluctuation. CDSEM. • For CD measure 175nm dose fluctuation could from Dose0 to Dose2 including and +-2 step of focus, so CDSEM could not be a ”definitive tracer” of litho tool dose and focus stability.
- 5. CDSEM measurement Grayscale level approach for CD measurement •Build radial vector. •Apply LPF alongside the vector. •Find cross over threshold location. •Calculate distance from center to location. •Repeat it N times with different angles. •Find average. CDSEM pros and cons. •Excellent for predefined feature measurement. •Good for edge location definition. •Bad for form of feature characterization. •90 percent of pixel information in image lost. Morphology •Discipline in Image processing specialized on extraction and characterization of binary images. Could be very helpful for form description and recognition.
- 6. Examples of Morphology measurement •Threshold image –get binary one. (0 –black or 1 –red) •Fill holes •Reduce particle of particular area or shape –irrelevant for decision making •Calculate morphology parameters for each remaining particle. •Parameters could be: •Elongation, Orientation, Type factor, Waddler disk diameter, Heywood circularity factor, Compactness factor, Different moments of inertia etc.. Conclusion In order to increase amount of information used for processing need to add morphology operations and particle analysis.
- 7. FE after morphology operation 1. By replacing CD measurement by particle morphology area measurement be getting better separation per dose. FE 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.0 Focus steps Focus CD's Dose0 Dose1 Dose2 Dose3 Dose4 Dose5 Dose6 FE per Via area 0 1000 2000 3000 4000 5000 1 2 3 4 5 6 7 8 Focus Area D0 D1 D2 D3 D4 D5
- 8. FE after morphology operation FE 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.0 Focus steps Focus CD's Dose0 Dose1 Dose2 Dose3 Dose4 Dose5 Dose6 1. By replacing CD measurement by particle orientation measurement we getting strong separation between negative and positive focus. FE per Via Orientation 0 50 100 150 200 1 2 3 4 5 6 7 8 Focus Degrees of rotation D0 D1 D2 D3 D4 D5
- 9. FE after morphology operation FE 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.0 Focus steps Focus CD's Dose0 Dose1 Dose2 Dose3 Dose4 Dose5 Dose6 1. By replacing CD measurement by particle Area*sign[orientation] measurement we getting strong separation between negative and positive focus as well as better dose separation.. FE per Via Area combined Orientation -6000 -4000 -2000 0 2000 4000 6000 1 2 3 4 5 6 7 8 Focus Area*sign[Orientation] D0 D1 D2 D3 D4 D5
- 10. FE after morphology operation FE 0.0 20.0 40.0 60.0 80.0 100.0 120.0 140.0 160.0 180.0 200.0 Focus steps Focus CD's Dose0 Dose1 Dose2 Dose3 Dose4 Dose5 Dose6 1. By replacing CD measurement by particle Elongation measurement we getting visible separation of dose. FE by Elongation 1 1.1 1.2 1.3 1.4 1.5 1 2 3 4 5 6 Dose Elongation Focus0 Focus1 Focus2 Focus3 Focus4 Focus5 Focus6 Focus7
- 11. What we are trying to characterize 1.Via changing orientation from <90 degree to > than 90 degree as result of focus changes 2.Via changes its area as result of focus change 3.Via changes its elongation as result of focus change 4.For every other dose this process has different rate.
- 12. Engine for classification 1.Two obvious ways for recognition of Focus/Dose imaging sets are fuzzy classifier and Minimal Mean Distance classifier. 2.Advantages for Fuzzy classifier – •Strait forward –just define the required range of values per Dose/Focus imaging set. •Simple –no need for complicated mathematical operation, for example normalizations and decompositions. 3.Disadvantages for Fuzzy classifier – •Build as is –SW need to be recompiled in order to correct/change any additional parameter or information. •Highly Non-linear, rigid structure –problematic tool for research and definition for new parameters We choose MMDC –for the research purposes and relative simplicity in reconfiguration.
- 13. Simple MMDC classifier SW demonstration -Cool
- 14. Results Focus0Dose0 Focus2Dose0 Focus1Dose5 Focus5Dose3 Focus5Dose0 Focus4Dose0 Focus0Dose1 Focus2Dose1 Focus0Dose0 Focus3Dose1 Focus3Dose0 Focus5Dose0 Focus0Dose3 Focus0Dose2 Focus2Dose1 Focus3Dose0 Focus5Dose1 Focus5Dose4 Focus0Dose2 Focus0Dose1 Focus2Dose3 Focus3Dose4 Focus3Dose3 Focus4Dose1 Focus0Dose4 Focus1Dose3 Focus0Dose5 Focus3Dose0 Focus4Dose1 Focus5Dose4 Focus3Dose5 Focus1Dose5 Focus0Dose2 Focus3Dose2 Focus4Dose2 Focus4Dose3 Focus0Dose0 Focus1Dose0 Focus3Dose1 Focus3Dose0 Focus4Dose0 Focus4Dose0 Focus0Dose1 Focus2Dose2 Focus1Dose0 Focus3Dose1 Focus3Dose0 Focus5Dose0 Focus1Dose5 Focus2Dose4 Focus2Dose1 Focus3Dose0 Focus5Dose1 Focus4Dose3 Focus0Dose2 Focus2Dose4 Focus2Dose3 Focus3Dose1 Focus3Dose3 Focus4Dose4 Focus0Dose4 Focus1Dose3 Focus2Dose5 Focus4Dose2 Focus4Dose4 Focus5Dose4 Focus0Dose5 Focus1Dose5 Focus1Dose4 Focus3Dose4 Focus5Dose3 Focus5Dose5 Focus0Dose0 Focus1Dose0 Focus3Dose1 Focus3Dose0 Focus4Dose0 Focus4Dose0 Focus0Dose1 Focus2Dose1 Focus2Dose1 Focus3Dose1 Focus3Dose0 Focus5Dose0 Focus0Dose3 Focus0Dose1 Focus2Dose1 Focus3Dose0 Focus5Dose1 Focus4Dose1 Focus0Dose2 Focus1Dose2 Focus2Dose3 Focus3Dose1 Focus3Dose3 Focus4Dose4 Focus0Dose4 Focus1Dose3 Focus2Dose5 Focus5Dose3 Focus4Dose4 Focus5Dose4 Focus0Dose5 Focus1Dose5 Focus1Dose4 Focus3Dose4 Focus5Dose3 Focus4Dose5 Sign [Orientation]*Area*Elongation Sign [Orientation]*Area; Area; Area*Elongation Sign [Orientation]*Area*Elongation; Area For all cases high probability to find Dose/Focus value within one step of dose or/and focus error
- 15. Next steps Need significant amount of FE data in order to: •Characterize stability of measurement factors. •Create real MMDC –not just on one set of FE’s •Create robust normalization of parameters in SW –Fisher decomposition? Need significant amount of real CDSEM images from production recipes: •Full verification process Need CDSEM recipe optimization –close 0 nm shifts in array •Characterize stability of measurement factors. •Full verification process Conclusion Similar process of focus/dose characterization possible to perform on any closed shape feature, not just via. In that case amount of information that we could extract with the help of morphology operations are greater significantly.