Algorithmic Techniques for Parametric Model Recovery
AJ Garcia Honors Thesis sample
1. Finishing of Individual Sapphire Wafers
University of Florida
Center for Manufacturing Innovation
A.J Garcia
November 24, 2014
2. Sapphire (Al203) Background
• Hexagonal structure
• 9 mohs scale hardness
• Chemically and biologically
inert
• Non-thrombogenic
• 2040 °C melting point
• Low thermal expansion coeff.
• Wide transmission range
• 0.18 µm – 5.5 µm
• Stretches from IR to UV
• Birefringent
• Electrical insulator
• Anisotropic
2
(http://www.cyberphysics.co.uk/topics/light/emspec
t.htm)
3. Applications
• Electronics
• Epitaxial growth of semiconductors
• Gallium nitride LED manufacturing
• Silicon-on-sapphire integrated circuits
• Radiation hardened devices
• Scratch resistant screens
• Corrosion resistant components
• Nozzles, crucibles
• Optical windows and lenses in extreme
environments
4. Motivation
• Applications demand
precision surfaces
• Electrical industry
• Uniform semiconductor
growth
• Precision form
requirements
• Optical industry
• Image distortion
• Incomplete transmission
5. Effect of magnet arrangements
Guide magnet arrangement investigation
Workpiece
10 mm x 10 mm x 1 mm rectangular
sapphire
Slurry
50-70 µm diamond abrasive mixed with
lubricant
Abrasive surface P120 grit abrasive paper
Guide magnet arrangement 1, 2, 3
Set guide magnet rotation
speed
350 rpm
Finishing time 20 min
0.85 0.9
1.75
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
1 2 3
Ave.ThicknessReductionµm/min
Guide magnet arrangement
• Aim to improve tool motion
• Eliminating sticking
• Arrangement 3 enabled ideal motion
• Motion produced greater thickness reduction across
surface
6. Surface roughness
• 6 measurements along diagonal
• 2.02 mm spacing
• Abrasive: 0-0.5 µm diamond abrasive slurry
• Average initial roughness: 937 nm Sa
d
0
y
x
10 mm
10 mm
Measurement locations
(equally spaced)
7. WOT roughness reduction: Results
• Average roughness after finishing: 700 nm Sa
• Partial finishing of surface
• Smooth plateaus amid rough valleys
• 0-0.5 µm abrasive does not penetrate valleys
0
200
400
600
800
1000
1200
2.02 4.04 6.06 8.08 10.1 12.12
SurfaceRoughnessRa[nm]
Diagonal Distance From Corner [mm]
Before After
9. Surface Roughness with Flooded
Basin
0
0.5
1
1.5
2
2.5
3
3.5
4
0 6.3 12.6 18.9 25.2 31.5
RoughnessSa[nm]
Radial distance from center [mm]
Before finishing
After finishing
10. Radial
Distance [mm] 0 6.3 12.6
Unpolished
surface
Polished
surface
Radial
Distance [mm] 18.9 25.2 31.5
Unpolished
surface
Polished
surface
Sa = 1.77 nm Sa = 3.24 nm Sa = 1.40 nm
Sa = 1.55 nm Sa = 1.68 nm Sa = 1.68 nm
Sa = 0.67 nm Sa = 0.77 nm Sa = 1.52 nm
Sa = 1.67 nm Sa = 1.35 nm Sa = 1.99 nm
11. Tool magnet sticking
• Evidence of excessive magnetic flux
• High magnetic force
• High normal reaction force
• High friction
• Exclusion of lubricant and diamond particles
N
S
N
S
Magnetic force
Guide
magnet
Tool magnet
Workpiece
Iron particle
Abrasive
particle
Normal
force
Jig
12. Magnetic field density
-10
10
30
50
70
90
110
130
0 5 10 15 20 25 30
Magneticfluxdensity[mT]
Radial distance from center, r [mm]
3mm
6mm
9mm
12mm
15mm
12.7 mm
r
Tool
magnet
Guide
magnet
• 3 mm jig height
• Steep drop in magnetic flux density
• Maximum of 121 mT
• 9 mm jig height
• 71 mT
• Consistent across tool magnet
Jig height
0
16 mm +
jig height
13. Surface roughness with 9 mm jig
height
• Average before: 6.0 nm Sa
• Average after: 0.9 nm Sa
0
1
2
3
4
5
6
7
8
9
10
0 6.3 12.6 18.9 25.2 31.5
SurfaceRoughnessSa[nm]
Radial Distance from Center of Polishing [mm]
Before Finishing
After finishing
14. Abrasive path simulation
• End goal:
• Develop method for predicting material removal
• Plan:
• Design mathematical model of ideal particle motion
• Observe correlation between surface changes and
number of particle passes
• Observe changes caused by parameter variation
• Introduce corrective terms
15. Simulation Plot
• Example parameters:
• R = 3, r = 1, h = 0.5, ω = 6.28, t = 1, res = 2
• xmin = -5, xmax = -4, ymin = -1, ymax = 1
16. Simulation inaccuracies
• Inaccuracies noticed when h = 0
𝑥 𝑡 = 𝑅 + 𝑟 cos(ω𝑡) − ℎ cos
𝑅 + 𝑟
𝑟
ω𝑡
𝑦 𝑡 = 𝑅 + 𝑟 sin(ω𝑡) − ℎ sin
𝑅 + 𝑟
𝑟
ω𝑡
𝑥 𝑡 = 𝑅 + 𝑟 cos(ω𝑡)
𝑦 𝑡 = 𝑅 + 𝑟 sin(ω𝑡)
• Becomes parametric
equations of a circle
• Simulation generates an
annulus
17. Simulation inaccuracies
• Generate circle using parametric circle equations
𝑥 𝑡 = 𝐴 cos(ω𝑡)
𝑦 𝑡 = 𝐴 sin(ω𝑡)
• A = 10, ω = 21 rad/s
• t = 0 s : 900 s
• Also generates an
annulus
• Thickness increases with
t
• res too low
18. High res circle
• Same parameters to generate circle
• res increased from 4 to 6
• Correctly represents a circle
• Lower res limit set to 6
Editor's Notes
Structural and functional component
High dielectric constant
90 % of blue LEDs