Title: Moir fringe method for the measurement of distortions of hotembossed polymeric substrates Hayden Tay
1Moiré fringe method for the measurement of
distortions of hot-embossed polymeric
substratesHayden Taylor 3, Xu Zhiguang 1, 2, 4,
Li Shiguang 1, 2, Kamal Youcef-Toumi 4, Yoon
Soon Fatt 5, and Duane Boning 32 July 20081
Singapore-MIT Alliance2 Singapore Institute of
Manufacturing Technology3 Microsystems
Technology Laboratories, MIT4 Department of
Mechanical Engineering, MIT5 School of
Electrical and Electronic Engineering, NTU
2Outline
- Why do distortions of hot-embossed parts need to
be measured? - Why use the moiré principle to do so?
- Advantages of our method
- Image processing procedures used
- Outlook
3Why measure distortions of hot-embossed parts?
- Hot embossing of thermoplastic polymers an
attractive manufacturing process for inexpensive
microfluidic devices - Two main types of dimensional faults with the
output - Out-of-plane distortions (from demolding while
soft) - In-plane distortions (from post-demolding thermal
contraction) - Focus here on in-plane distortions
- Harder to eliminate from the process
- Present challenges when registering multiple
layers
4Why use the moiré principle?
- Has the potential to be effective for determining
small distortions over large areas - Has already been demonstrated for quantifying
distortions of micropatterned layers of the
elastomer PDMS 1 - Patterned areas 1 cm2 were measured
- Microscopic angular positioning of samples was
required - Processing of moiré patterns required extensive
user input - Our aim is to demonstrate a simpler, more
automated procedure that works over larger
substrate areas. - 1 John A. Rogers, et al. Quantifying
distortions in soft lithography, J. Vac. Sci.
Technol. B, 1998, 16(1) 88-97 - polydimethylsiloxane
5Advantages of our method
- Simplicity the only apparatus required is a
flatbed scanner and a printed, square reference
grid. The part is embossed using a stamp with the
same reference pattern. - Samples may be positioned by hand within an
allowable angular range of a few degrees - Scans are made at a variety of part-reference
angles, and precise orientations are determined
from the captured images
Scanners cover
Reference grid(printed acetate)
Embossed part
Scanners platen
Light from scanner
6Information obtained from one scan
where
10 mm
Schematic of scanned image
Actual scanned image
dM, and are specific to each image
captured
7Fitting global part shrinkage using multiple scans
One part multiple scans
dM
10 mm
- For this part, global average post-demolding
contraction of 0.4 is extracted - Consistent with known material properties
- Could increase confidence in estimate by taking
more scans
8Processing of scanned images
- Strength of the tonal contrast of moire fringes
varies among (and within) parts depends on the
relief and shape of the embossed topography. - Our image-processing strategy depends on the
strength of contrast
Harder casesweak contrast
Easier casesstrong contrast
9Strong fringe contrast
Steps for determining the positions of
intersections of moiré fringes for simple cases
with strong image contrast
- Signal filtering
- Image rotation
- Image sharpness
- Fine rotation, continuous region separation, and
centroid point judgment
10Weak fringe contrast
Steps for determining the positions of
intersections of moiré fringes for hard cases
with weak image contrast one additional step
- Signal filtering
- Image rotation
- Image sharpness
- Correlation calculation between the typical
extracted cross and our image - Image sharpness, fine rotation, continuous region
separation, and centroid point judgment
11Processing steps in detail (weak contrast)
Original image (19442629 pixels)
12Processing steps in detail (weak contrast)
Image after signal filtering (step 1)
13Processing steps in detail (weak contrast)
Image after rotation (step 2)
Fringe orientations identified
14Processing steps in detail (weak contrast)
Image after sharpness filter (step 3)
15Processing steps in detail (weak contrast)
Cross points searching correlation calculation
between one extracted cross and our image (step 4)
16Processing steps in detail (weak contrast)
Image sharpness, fine rotation, continuous region
separation, and centroid point judgment, to get
the positions of the fringe crosses (step 5)
17Processing steps in detail (weak contrast)
Obtaining the best fitted grid by rotating,
translating and shrinking
Fringe cross
??
Fitted grid
extracted dM
18Outlook
- A promising method is described for detecting
post-demolding distortions in hot-embossed
substrates - The principle is demonstrated of using multiple
scanned images to extract global-average part
distortion - Demonstrated with a 50 mm square sample
- A high level of automation has so far been
achieved in image interpretation - Identification of fringe locations is mostly
automatic for low-contrast images,
identification of one fringe location is
user-assisted - Identification of sample and reference grid edges
is user-assisted - Have yet to confirm whether local deviations of
the fringes from their fitted grid positions
correspond to real part distortions.
19Acknowledgements
- Funding the Singapore-MIT Alliance
- Use of the Microsystems Technology Laboratories
at MIT - Matthew Dirckx
- Brian Anthony