Citation
Simulation, Part Path Correction, and Automated Process Parameter Selection for Ultrashort Pulsed Laser Micromachining of Sapphire

Material Information

Title:
Simulation, Part Path Correction, and Automated Process Parameter Selection for Ultrashort Pulsed Laser Micromachining of Sapphire
Creator:
Blood, Daniel A
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (198 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Mechanical Engineering
Mechanical and Aerospace Engineering
Committee Chair:
SHEPLAK,MARK
Committee Co-Chair:
SCHUELLER,JOHN KENNETH
Committee Members:
GREENSLET,HITOMI
ARNOLD,DAVID P
Graduation Date:
8/9/2014

Subjects

Subjects / Keywords:
Acceleration ( jstor )
Fluence ( jstor )
Ionization ( jstor )
Laser beams ( jstor )
Laser power ( jstor )
Lasers ( jstor )
Machining ( jstor )
Photons ( jstor )
Sapphire ( jstor )
Simulations ( jstor )
Mechanical and Aerospace Engineering -- Dissertations, Academic -- UF
al2o3 -- laser -- micromachining -- sapphire -- ultrashort
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Mechanical Engineering thesis, Ph.D.

Notes

Abstract:
This dissertation describes an ultrashort pulsed laser material removal simulator with x-y stage acceleration profile consideration and part path compensation. Ultrashort pulsed lasers offer the advantage of single step processing of various materials with high repeatability. Over the past 30 years the laser repetition rate and power output have increased, and although this increases the material removal rate, it also introduces new challenges. The acceleration rates of the x-y stages on a laser micromachining setup are finite, but this has been neglected. In the past the acceleration rate has been negligible due to low repetition rates; however, for high repetition rates the acceleration and deceleration regions introduce local variations in the material removal. A novel method is presented that accounts for the stage dynamics to produce a simulated cut. In addition to the simulator, a technique for modifying the part path to reduce non-uniformity in the material removal is discussed. The laser operator has access to a variety of process parameters that ultimately affect the cost and quality of the machined component. Choosing the correct combination of these parameters requires knowledge of the machining process, and the wrong combination can result in a feature that is unsatisfactory and/or overly expensive. The modification of these parameters, and a correction of the part path allows for a more uniform depth of cut and higher feature quality. This dissertation contains three main contributions. The first contribution is to quantify the relationship between ultrashort pulsed laser machining parameters and the ablation depth of sapphire. The second is to produce a pulsed laser micromachining simulator that includes not only the laser-material interaction, but also the nuances of controlling the position of the laser beam on the workpiece. The final contribution is to produce a part path correction program with an automated process parameter routine. This program will simplify the process parameter selection and reduce depth irregularities in the machined geometry. Ultrashort pulsed lasers are a relatively new laser type; consequently, there is a plethora of aspects that may be added into future iterations of the simulator, automated parameter selection, and part path correction software. These aspects include, but are not limited to: sidewall angle compensation, thermal diffusion modeling, and modeling of additional workpiece materials. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2014.
Local:
Adviser: SHEPLAK,MARK.
Local:
Co-adviser: SCHUELLER,JOHN KENNETH.
Statement of Responsibility:
by Daniel A Blood.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Blood, Daniel A. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
969976492 ( OCLC )
Classification:
LD1780 2014 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

1 SIMULATION, PART PATH CORRECTION, AND AUTOMATED PROCESS PARAMETER SELECTION FOR ULTRASHORT PULSED LASER MICROMACHINING OF SAPPHIRE By DANIEL A. BLOOD A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 4

PAGE 2

2 © 201 4 D aniel A. B lood

PAGE 3

3 To my wife , Kristen J. Blood

PAGE 4

4 ACKNOWLEDGMENTS I thank my family for their unwavering support during my entire education. I would also like to thank Dr. Scott Duncan for the encouragement and lessons taught at Valparaiso University . I thank my lab mates for their help through these past years, and appr eciate all of the information they have passed down. Last but not least, I would like to thank my advisor s , Dr. Sheplak and Dr. Schmitz, for their guidance and patience throughout my time at The University of Florida.

PAGE 5

5 TABLE OF CONTENTS Page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURE S ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................... 15 CHAPTER 1 INTRODUCTION AND MOTIVATION ................................ ................................ ..... 17 1.1 Introduction ................................ ................................ ................................ ....... 17 1.2 Motivation ................................ ................................ ................................ ......... 17 1.2.1 Advantages of a Simulator ................................ ................................ ...... 23 1.2.2 Advantages of Automation/Correction Routines ................................ ...... 25 1.2.3 Complexity of Temporal Processes ................................ ......................... 28 1.3 Research Contributions ................................ ................................ .................... 29 1.4 Dissertation Organization ................................ ................................ .................. 30 2 LASER MICROMACHINING BACKGROUND ................................ ........................ 31 2.1 History and Physics of Lasers ................................ ................................ ........... 31 2.2 Types of Lasers ................................ ................................ ................................ 34 2.2.1 Gas Lasers ................................ ................................ .............................. 34 2.2.2 Dye Lasers ................................ ................................ .............................. 37 2.2.3 Semiconductor Lasers ................................ ................................ ............. 38 2. 2.4 Solid State Lasers ................................ ................................ ................... 39 2.2.5 Laser Pulse Generation ................................ ................................ ........... 41 2.2.5.1 Q Switching ................................ ................................ .................... 42 2.2.5.2 Mode Locking ................................ ................................ ................. 43 2.2.5.3 Cavity Dumping ................................ ................................ .............. 45 2.2.5.4 Gain Switching ................................ ................................ ............... 47 2.2.6 Coherent Talisker Laser Specifications ................................ ................... 48 2.3 Laser Material Removal Mechanisms ................................ ............................... 48 2.3.1 Types of Ablation ................................ ................................ ..................... 49 2.3.1.1 Pyrolitic Ablation ................................ ................................ ............ 51 2.3.1.2 Photolitic Ablation ................................ ................................ ........... 54 2.3.2 Weak and Strong Ablation Regions ................................ ......................... 56 2.4 Machining Characteristics for Different Pulse Durations ................................ ... 63 2.4.1 Continuous Wavelength (CW), Millisecond, and Micro second Pulses ..... 64 2.4.2 Nanosecond ................................ ................................ ............................ 65 2.4.3 Picosecond ................................ ................................ .............................. 65 2.4.4 Femtos econd ................................ ................................ ........................... 66

PAGE 6

6 2.5 Existing Material Removal Models ................................ ................................ .... 66 2.5.1 Molecular Dynamics ................................ ................................ ................ 66 2.5.2 Empirical Models Based on Statistical Data ................................ ............ 69 2.5.3 Summation of Individual Pulses ................................ .............................. 72 2.6 Chapter Summary ................................ ................................ ............................. 76 3 LASER MACHINING SIMULATION ................................ ................................ ........ 78 3.1 New Ablation Simulator ................................ ................................ ..................... 78 3.2 Ablation Profile ................................ ................................ ................................ .. 78 3.3 Acce leration Profile ................................ ................................ ........................... 82 3.4 Acceleration Simulation ................................ ................................ .................... 86 3.4.1 Acceleration Profiles ................................ ................................ ................ 86 3.4.1.1 Constant Acceleration ................................ ................................ .... 89 3.4.1.2 Constant Jerk ................................ ................................ ................. 91 3.4.1.3 Sinusoidal Acceleration ................................ ................................ .. 94 3.4.2 Excessive Overlap Regions ................................ ................................ ..... 96 3.4.3 Command Delays ................................ ................................ .................... 97 3.5 Proc ess Flow ................................ ................................ ................................ .... 97 3.6 Chapter Summary ................................ ................................ ........................... 102 4 LASER MACHINING PART PATH CORRECTION AND AUTOMATED PROCESS PARAMETER SELECTION ................................ ................................ 104 4.1 Introduc tion ................................ ................................ ................................ ..... 104 4.2 Excessive Pulse Overlap ................................ ................................ ................ 105 4.3 Ablation Depth ................................ ................................ ................................ 106 4.3.1 Part Path Correction Steps ................................ ................................ .... 109 4.3. 2 Added Passes Geometry ................................ ................................ ....... 111 4.3. 3 Strategies for Automating the Process Parameter Selection ................. 113 4.4 Process Flow ................................ ................................ ................................ .. 114 4.5 Chapter Summary ................................ ................................ ........................... 118 5 OXFORD J 355PS CHARACTERIZATION ................................ .......................... 121 5.1 Introduction ................................ ................................ ................................ ..... 121 5.2 Linear X Y and Galvo Stage Characterization ................................ ................ 123 5.2.2 Motion Controller and Laser Interface Delays ................................ ....... 129 5.2.2.1 Linear X Y Stage Delays ................................ .............................. 130 5.2.2.2 Galvo Undershoot Phenomenon ................................ .................. 132 5.3 Beam Profile ................................ ................................ ................................ ... 135 5.3.1 Focal Beam Diameter Analysis Technique ................................ ............ 136 5.3.2 Beam Shape Quality ................................ ................................ .............. 138 5. 4 Attenuator Characterization ................................ ................................ ............ 140 5.5 Lasing Command Delay ................................ ................................ .................. 142 5.6 Chapter Summary ................................ ................................ ........................... 144

PAGE 7

7 6 ULTRASHORT PULSED LASER MICROMACHINING OF SILICON ................... 145 6.1 Introduction to Silicon ................................ ................................ ...................... 145 6.2 Laser Processing Parameters ................................ ................................ ......... 145 6.2.1 Pulse Delay ................................ ................................ ........................... 146 6.2.2 Pulse Overlap ................................ ................................ ........................ 147 6.2.3 Av erage Pulse Fluence ................................ ................................ ......... 147 6.2.4 Machining Environment ................................ ................................ ......... 148 6.2.5 Number of Passes ................................ ................................ ................. 148 6.3 Laser Silicon Interaction ................................ ................................ ................. 148 6.3.1 Empirical Data ................................ ................................ ....................... 150 6.3.2 Gas Flow ................................ ................................ ............................... 151 6.4 Chapter Summary ................................ ................................ ........................... 153 7 ULTRASHORT PULSED LASER MICROMACHINING OF SAPPHIRE ............... 155 7.1 Introduction to Sapphire ................................ ................................ .................. 15 5 7.2 Laser Sapphire Interaction ................................ ................................ .............. 156 7.3 Traditional Laser Material Removal Model ................................ ...................... 157 7.4 Empirical Sapphire Model ................................ ................................ ............... 157 7.4.1 Gas Flow ................................ ................................ ............................... 158 7.4.2 Gentle Strong Ablation ................................ ................................ .......... 162 7.4. 3 Model Verification ................................ ................................ .................. 166 7.4. 4 Sidewall Angle ................................ ................................ ....................... 168 7.4. 5 Surface Roughness ................................ ................................ ............... 171 7.5 Parameter Modification, Optimization Routine, and Part Path Modification .... 172 7.5.1 Parameter Modification Strategies ................................ ........................ 172 7.5.2 Optimization Strategies ................................ ................................ ......... 174 7.5.3 Parameter Modification Routine Verification ................................ .......... 178 7.6 Added Passes ................................ ................................ ................................ . 185 7.7 Chapter Summary ................................ ................................ ........................... 189 8 CONCLUSION ................................ ................................ ................................ ...... 190 LIST OF REFERE NCES ................................ ................................ ............................. 193 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 198

PAGE 8

8 LIST OF TABLES Table P age 1 1 General guidelines for pulse durations to be categorized as long, short, or ultrashort pulses ................................ ................................ ................................ . 20 1 2 Processing parameters and the resulting effect on the machining process ........ 21 1 3 Stock material cost comparison between traditional (1100 Aluminum and 304 Stainless Steel) and exotic materials (C Plane Sapphire, Silica Aerogel, and Non Porous Alumina) ................................ ................................ ......................... 23 1 4 Comparison of CNC mill processing parameters and the analogous laser processing parameters ................................ ................................ ....................... 26 2 1 Pros and cons for different laser types ................................ ............................... 34 2 2 Output wavelength for different combinations of dyes and solvents . .................. 38 2 3 Rough estimate of the maximum pulse durations for photolitic m aterial removal to dominate ................................ ................................ ........................... 51 2 4 A comparison of the energy required to remove material for different pulse durations. ................................ ................................ ................................ ............ 56 2 5 Respective pulse duration, pulse energy, peak intensity, and wavelength for the h oles drilled in Figure 2 22 . ................................ ................................ ......... 64 2 6 Depth of cut comparison bet ween gentle phase ablation and strong phase ablation. ................................ ................................ ................................ .............. 70 3 1 The resulting number of pulses impacting a workpiece for different repetition rates. ................................ ................................ ................................ .................. 85 3 2 The acceleration rates of each region in a jerk system ................................ ....... 92 3 3 Regional times for a jerk type acceleration profile that does not reach the maximum acceleration. ................................ ................................ ....................... 93 3 4 Regional times for a jerk type acceleration profile that reaches the maximum acceleration. ................................ ................................ ................................ ....... 94 3 5 Time required for each region for a sinusoidal acceleration profile ..................... 95 4 1 Explanation of the different cut types availa ble in the modification routine ....... 108

PAGE 9

9 4 2 Cut results for the three simulations, one without any additional processing, one with automated parameter selection, and lastly one with modified parameters, and part path corrected. ................................ ............................... 116 4 3 Input parameters for the simulation modification comparison. ......................... 118 5 1 Oxford J 355PS m anufacturer specifications ................................ .................... 122 5 2 Aerotech PRO115 linear stage and controller specifications ............................ 125 5 3 Listing of the various command combinations to determine the delay introduced by each command type. ................................ ................................ .. 130 5 4 Extrapolated command delays for specific operations. ................................ ..... 131 5 5 Comparison of measured and predicted cut t imes for X Y linear stage moves 131 7 1 Sapphire material properties and UF laser pr o perties ................................ ...... 157 7 2 The range of values used for each of the three main parameters of interest for the sapphire characterization. ................................ ................................ ..... 162 7 3 Comparison of the average difference in predi cted versus measured cut depths ................................ ................................ ................................ ............... 167 7 4 Processing paramete rs for Figure 7 12 through 7 14 ................................ ....... 169 7 5 Explanation of the method of se lecting the modified parameters ..................... 173 7 6 Explanation of the e ffect of the tolerance selection on the depth const raint ..... 174 7 7 List of the average difference in cut depth between the desired and measured, and the average time to complete the cut. ................................ ...... 178 7 8 Maximum depth of cut for Figures 6 16, 6 17, and 6 18. ................................ .. 185 7 9 Comparison of Desired depth of cut to the maximum and average cut depth. . 185

PAGE 10

10 LIST OF FIGURES Figure P age 1 1 Diagram of the different laser types and applications ................................ ......... 18 1 2 Visualization of the difference between CW, and pulsed lasers. ........................ 19 1 3 Visualization of the difference machining parameters ................................ ........ 22 1 4 Demonstration of the three main processing parameters ................................ ... 26 1 5 Automated part path generation by a CAM program (Surfcam ® ) ........................ 27 2 1 Diagram of a three energy level laser ................................ ................................ . 33 2 2 Schematic of an externally pumped CO 2 gas lase r ................................ ............. 35 2 3 Schematic of the three different modes of vibration of a CO 2 molecule .............. 36 2 4 Schematic of an externally pumped dye laser ................................ .................... 38 2 5 Schematic of a semiconductor laser ................................ ................................ ... 39 2 6 Visualization of a solid state laser. Diodes can also be used as the excitation source in place of flashlamps ................................ ................................ ............. 40 2 7 Gains and losses as a function of time for an actively Q switched laser ............ 43 2 8 Optical power and losses as a function of time i n an actively mode locked laser ................................ ................................ ................................ .................... 44 2 9 Optical power, losses, and gain as a fun ction of time in a passively mode locked laser ................................ ................................ ................................ ........ 44 2 10 Schematics for two types of cavity dumping systems ................................ ......... 46 2 11 Output power, and pump power as a function of time in a gain switching laser . 47 2 12 ............................... 49 2 13 Visualization of the pyrolitic material removal process ................................ ....... 52 2 14 Visualization of the HAZ in laser machining ................................ ....................... 53 2 15 Thermal finite element simulation temperature distribution results ..................... 54 2 16 Material removal by ionization ................................ ................................ ............ 55

PAGE 11

11 2 17 Demonstration of single photon ionization ................................ .......................... 57 2 18 Demonstration of 2 multi photon excitation methods ................................ .......... 58 2 19 A comparison between multi photon ionization and impact ionization ................ 60 2 20 Free electon densities and the respective contributions from multi photon ionization and impaction ionization ................................ ................................ ..... 61 2 21 Strong and gentle ablation traditional models ................................ ..................... 62 2 22 Recreation of holes drilled in steel foil with different pulse duration lasers ......... 63 2 23 Illustration of the LJ 2D model with the photons bombarding individual atoms and electrons transferring their energy into the lattice via phonons .................... 67 2 24 Multishot MD study simulating the effect of multiple pulses on silicon ................ 68 2 25 Ablation profile for different materials ................................ ................................ . 72 2 26 A comparison of the predicted and measured depth profile for a two level cut in sodalime ................................ ................................ ................................ ......... 73 2 27 Simulated versus measured depth profiles of D2 steel ................................ ....... 75 3 1 Gaussian beam intensity profiles ................................ ................................ ........ 80 3 2 Comparison of acceleration profiles ................................ ................................ ... 83 3 3 Visualization of the pulse area overlap ................................ ............................... 84 3 4 Demonstrations of acceleration region effects on the ablation profile of silicon .. 85 3 5 Typical acceleration profiles and velocity profiles as a function of time .............. 87 3 6 The above graphs show four different acceleration profiles for a constant jerk scenario ................................ ................................ ................................ .............. 88 3 7 The acceleration profile of a constant jerk system with a maximum acceleration rate ................................ ................................ ................................ . 91 3 8 Demonstration of the two profiles of a s inusoidal acceleration ........................... 95 3 9 Visible cracks in a sapphire machined component that are likely cause by centripetal acceleration problems ................................ ................................ ....... 96 3 10 Operation flowchart for the UF ultrashort pulsed laser ablation simulator .......... 98 3 11 Screenshot of the UF picosecond puls ed laser ablation simulator GUI .............. 99

PAGE 12

12 3 12 2D depth of cut gr aph for the 'GATOR' cut program ................................ ......... 100 3 13 3D depth of cut p lot for the 'GATOR' cut program ................................ ............ 100 3 14 Plot of the velocity as a function of posit ion for the 'GATOR' cut program ........ 101 3 15 Calculated machining time as a function of commanded velocity for the 'GA TOR' cut program ................................ ................................ ....................... 101 3 16 Simulation resu lts for the 'GATOR' cut program ................................ ............... 102 4 1 direction motion ................................ ................................ ................................ 107 4 2 Visualization of the different cut types ................................ .............................. 109 4 3 Demonstration of the path addition process ................................ ..................... 110 4 4 of the cut profile ................................ ................................ ................................ 111 4 5 Visualization of a linear added passes routine ................................ .................. 112 4 6 Demonstration of a feature machined in R plane sapphire with added linear passes ................................ ................................ ................................ .............. 112 4 7 Visualization of a circular added passes routine ................................ ............... 113 4 8 A si mulated cut in R Plane sapphire without any parameter or path modification ................................ ................................ ................................ ...... 1 15 4 9 Simulation results with the ne w modified machining parameters ..................... 116 5 1 Oxford J 355PS laser micromachining station installed at the University of Florida ................................ ................................ ................................ .............. 122 5 2 Location of the three PRO115 stages (X, Y, and Z) and the galvo ................... 123 5 3 Visualization of the path of a laser beam through a galvo ................................ 124 5 4 Position and velocity measurements of the PRO115 linear stage .................... 126 5 5 Sample silicon cut measurement to determine the acceleration profile and rate of the galvo ................................ ................................ ................................ 127 5 6 Measured data with SWLI resolution error bars included ................................ . 128 5 7 Measured and constant acceleration profile least squares error fit of the silicon pulse location data ................................ ................................ ................. 128

PAGE 13

13 5 8 Comparison of the computed galvo velocity a nd the LSE theoretical velocity .. 129 5 9 2D plot of data from a SWLI measurement of overlapping rectangles machine in R plane sapphire ................................ ................................ ............ 133 5 10 2D plot of data from a SWLI measurement of overlapping rectangles machine in R plane sapphire with a delay added into the cut program ............ 134 5 11 Visualization of how a scanning slit profilometer operates ............................... 135 5 12 Ophir S ................................ ......... 136 5 13 Measured and least squares fit theoretical beam waist at various distances fro m the focal plane in the Z axis ................................ ................................ ...... 138 5 14 Demonstration of peak saturation at z heights close to the fo cal plane for both X and Y axes ................................ ................................ ............................ 139 5 15 Gaussian intensity profile at a distance of 13.35 mm from the focal plane ....... 139 5 16 Visualization of how a laser beam variable attenuator operates ....................... 141 5 17 Comparison of commanded and measured attenuation values ........................ 142 5 18 Visualization of the beam on/beam off lasing time ratio experim ent ................. 143 6 1 The process of forming plasma over the workpiece surface ............................. 146 6 2 Zygo Newview 7200 SWLI setup at the University of Florida ........................... 149 6 3 Fluence depth of cut relationship for silicon ................................ ..................... 150 6 4 Intensity depth of cut relationship for silicon determined by single pulse ablation depth ................................ ................................ ................................ ... 151 6 5 The effect of gas flow on a machined silicon feature ................................ ........ 152 6 6 Demonstration of a typical nozzle workpiece setup for blowing different gases across the workpiece surface ................................ ................................ . 153 7 1 Experimental setup for flowing different gases over a sapphire wafer .............. 159 7 2 Gas flow machining experiment r esults for a 5 µm pulse spacing .................... 160 7 3 Gas flow machining experiment res ults for a 2.5 µm pulse spacing ................. 160 7 4 Gas flow machining experiment resu lts for a 1.25 µm pulse spacing ............... 161

PAGE 14

14 7 5 Surface roughness as a function of depth for all of the various pulse frequenc y fluence gas flow combinations ................................ ......................... 162 7 6 Probability distribution shapes ................................ ................................ .......... 163 7 7 Average volume removed per pulse for 5 passes ................................ ............. 165 7 8 Average volume removed per pulse for 10 passes ................................ ........... 165 7 9 Average volume removed per pulse for 25 passes ................................ ........... 166 7 10 Average difference between the theoretical depth of cut and measured depth of cut for the data points used in the Monte Carlo characterizatio n routine ...... 167 7 11 Experiments on sapphire and silicon have shown maximum attainable sidewall angles ................................ ................................ ................................ . 169 7 12 Sidewall angle tests in R plane sapphire with a fixed fluence, number of passes, and pulse frequency b ut a varying pulse area overlap ........................ 169 7 13 Sidewall angle tests in R plane sapphire with a fixed feedrate, number of passes, and pulse frequency but a varying fluence ................................ .......... 170 7 14 Sidewall angle tests in R plane sapphire with a fixed fluence, frequency, and pulse spacing but a varying number of pas ses ................................ ................. 170 7 15 Surface roughness as a function of cut depth for a v ariety of processing parameters ................................ ................................ ................................ ....... 172 7 16 Plot of the absolute difference between the simulated cut depth of a given combination of pulse spacing/passes and the desired depth of cut .................. 177 7 17 Cut depths for the different parameter modification routines ............................ 179 7 18 Cut depths for the different parameter modification routines ............................ 180 7 19 SWLI measurements for a 20 µ m deep cut in R plane sapphire ...................... 182 7 20 SWLI measurements for a 25 µ m deep cut in R plane sapphire ...................... 183 7 21 SWLI measurements for a 30 µm deep cut in R plane sapphire ...................... 184 7 22 .............. 186 7 23 .............. 187 7 24 ......... 188

PAGE 15

15 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy SIMULATION, PART PATH CORRECTION, AND AUTOMATED PROCESS PARAMETER SELECTION FOR ULTRA SHORT PULSED LASER MICROMACHINING OF SAPPHIRE By Daniel A. Blood Chair: Mark Sheplak Major: Mechanical Engineering This dissertation describes a n ultrashort pulsed lase r material removal simulator with X Y stage acceleration profile consideration and part path compensation. Ultrashort pulsed lasers offer the advantage of single step processing of various materials with high repeatability. Over the past 30 years the laser repetition rate and power output have increased , and although this increases the material removal rate, it also introduces new challenges. The acceleration rates of the X Y stages on a laser micromachining setup are finite , but this has been neglected . In the past th e acceleration rate has been negligible due to low repetition rates ; however, for high repetition rate s the acceleration and deceleration regions introduce local variations in the material removal. A novel method is presented that accounts for the stage dynamics to produce a more robust simulated cut. In addition to the simulator , a technique for modifying the part path to reduce non uniformity in the material removal is discussed. The laser operator has access to a variety of process parameters that ultimately affect the cost and quality of the machined component . Choosing the correct combination of these parameters requires knowledge of the machining process , and the wrong combination can result in a feature that is

PAGE 16

16 unsatisfactory and/or overly expensive. The modification of these parameters, and a correction of the part path allows for a more uniform depth of cut and higher feature quality. This dissertation contains three main contributions. The first contributio n is to quantify the relationship between ultrashort pulsed laser machining parameters and the ablation depth of sapphire . The second is to produce a pulsed laser micro machining simulator that includes not only the laser material interaction, but also the nuances of controlling the position of the laser beam on the workpiece. The final contribution is to produce a part path correction program with an automated process parameter routine. This program simplifies the process parameter selection and reduce s dep th irregularities in the machined geometry. Ultrashort pulsed lasers are a relatively new laser type; consequently, there is a plethora of aspects that may be added into future iterations of the simulator , automated parameter selection routine , and part pa th correction software. These aspects include, but are not limited to: sidewall angle compensation, thermal diffusion modeling, and modeling of additional workpiece materials.

PAGE 17

17 CHAPTER 1 INTRODUCTION AND MOTIVATION 1.1 Introduction Subtractive manufacturing has been, and remains, one of the most widely utilized manufacturing methods in industry. Since the advent of lasers in 1960, laser machining has shown promise in adding another tool for subtractive manufacturing. While continuous wavelength (CW) and long pulse lasers allowed for machining of metals and pla stics, ceramic machining has not been an appropriate application for lasers until the advent of ultrashort pulsed lasers. The introductio n of these new types of lasers inherently brings with it new challenges in selecting the correct processing parameters, and part paths. A new set of tools is needed to streamline the parameter selection process, and reduce the complexity of implementing these new lasers into industry. This chapter begins with the motivation for a laser machining simulator that accounts for stage dynamics as well as other processing parameters. It then gives an overview of the complexity of laser machining, and why a correction/automation routine aid s in the manufacturing process. R esearch contributions a re then reviewed, and finally a selection of additional research topics related to this dissertation are discussed . 1.2 Motivation L.A.S.E.R, or light amplification through stimulated emission of radiatio n, is a highly coherent beam of light that has gained acceptance in a plethora of applications ever since it s development in 1960 [1] . Even though the technology is barely fifty years old, it has been widely adapted in both the industrial and private sector [2, 3] . Whether it is a laser pointer, DVD player, range finder, surgical device, or one of many other items used daily, they all exhibit the same basic characteristics; high monochromat icity

PAGE 18

18 and high temporal coherence [4] . These two qualities have driven the development of lasers with higher power output, different wavelengths, and most importantly (for this research) shorter pulse durations. Figure 1 1. Diagram of the different laser types and applications. The grey region bandgap cutting and drilling of metals. Deep penetration welding is located in Region [4] .

PAGE 19

19 Laser machining h as gained popularity in the past thirty years due to its broad machining capabilit ies [3] . These capabilities include being able to machine materials that are metallic as well as non metallic, soft to hard, hydrophobic to hydro philic, highly conductive to t hermally sensitive, and ductile to brittle . It also includes the ability to machine large and small components depending on the setup of the laser machin ing station . Figure 1 1 provides a visual of the different applications a s function of the interaction time and the laser light intensity. Explanations of the difference between ionization types is explored in sections 2.3.2. Although there are many different characteristics of lasers, one of the most important for machining sa pphire is the pulse duration . Sapphire is susceptible to thermal shock, and short pulsed lasers greatly reduce the amount of thermal energy imparted into the workpiece. Figure 1 2 provides a visualization of pulsed lasers versus continuous wavelength (CW) lasers. In the pulsed laser category there are three sub classifications: long pulsed, short pulsed , and ultrashort pulsed. Figure 1 2 . Visualizatio n of the difference between CW, and pulsed lasers.

PAGE 20

20 Table 1 1 provides a guideline for the typical pulsed duration of the three different pulse classifications. Note that the classification of a laser as ultrashort is a function of the workpiece material. The cause for this material dependen ce is explained fully in section 2.3 . Table 1 1. General guidelines for pulse durations to be categorized as long, short, or ultrashort pulses [5] . Pulse Length Category Pulse Duration Long Millisecond to Microsecond Short Nanosecond Ultrashort Material Pulse Length Metals 1 picosecond Ceramics 10 picoseconds Plastics 1 n anosecond Millisecond , microsecond, and nanosecond pulsed lasers (of the long and short category) have the advantage of high material removal and reliability, but they are limited on the materials they can process and also the quality of the cut [4] . This is in part due to the high thermal load that these pulsed lasers imp art on the machined components. Imparting a high thermal load into the workpiece can produce a heat affected zone (HAZ), initiate cracking due to thermal strain, and/or change the geometry of the machined component. Lasers that are in the low p icosecond to femtosecond pulse duration are better suited for materials that are thermally sensitive because they impart minimal thermal load on the material surrounding the cut [4] . With the broad range of exotic materials being utilized in modern systems, ultrashort pulsed lasers are likely to become an essential component in the manufacturing process; however, the complexity of the new materials , increased capabilities of lasers, and complex features requires a new toolbox to fully utilize these machines.

PAGE 21

21 Two components of laser machining that are essential to the implementation of ultrashort pulsed lasers into industry are modeling the laser material interaction ( either with firs t princip le s or empirically ) , and understanding the dynamics of the laser as it traverses over the workpiece surface . For the sake of simplicity, this dissertation refer s to any piece of equipment that manipulates the position of the laser beam relative to the workpiece as X Y stages and galvanometer scanners (aka galvo(s)) . Table 1 2. Processing parameters and the resulting effect on the machining process. Processing Parameter Input Parameters Possible Effects on Mac hining Process Pulse power Attenuation Excessive ablation per pulse, pulses too weak to surpass the ablation threshold, sidewall angle, and/or imparting excessive thermal energy into workpiece Pulse area overlap Feedrate , pulse frequency High surface roughness, imparting excessive thermal energy in one location, and/or increasing or decreasing ablation rate due to surface defects initiated by previous pulse(s) Pulse delay Pulse frequency Increased thermal energy over a shorter time period, and/or inc reasing or decreasing ablation rate due to surface defects initiated by previous pulse(s) Fluence/intensity profile Attenuation, f ocal depth Size and shape of ablation profile, peak intensity, and/or sidewall angle Laser material interaction requires investigation of the ablation profile, especially the combined effect that the laser parameters have on that profile. The main parameters of interest , shown in Table 1 2 , are the pulse power, pulse area overlap, pulse de lay, pulse duration, and pulse power. The pulse delay is the time delay between consecutive pulses, pulse area overlap is the area overlap between consecutive pulses, pulse duration is the time between the rise and fall of a laser pulse, and lastly the pul se power

PAGE 22

22 is a measure of the photonic energy per pulse. Visualizations of the input parameters affecting these processing parameters is shown in Figure 1 3 , and more in depth explanations of these variables are found in the following chapters. A B C Figure 1 3 . Visualization of the difference machining parameters . A ) F eedrate of the laser as it scans across the workpiece , B ) pulse frequency, and C ) fluence/intensity profile with various levels of attenuation .

PAGE 23

23 New lasers with higher power output and increased repetition rates have increased the complexity of modeling these parameters. Further investigation of their interactions is necessary to produce a reliable laser material ablation model. The i nteraction is discussed in more detail in Chapter 2. Faster repetition rates of lasers also introduce another important component: stage dynamics. While the stage dynamics are essentially negligible with low repetition rates, the improved rates of current lasers magnify the effect of the acceleration and deceleration regions. This introduces non linearity in the cut depth that can cause failure of the machined component , especially for depth sensitive parts . Inclusion of the stage dynamics in a s imulation is essential for accurate modeling of the entire process . 1.2.1 Advantages of a S imulator A disadvanta ge of machining exotic materials (e.g. sapphire, diamond, and Aerogel) is that the stock materials are typically expensive. Trial and error might be an acceptable method for producing accurate parts with inexpensive materials, but when the price of a test run is increased several magnitudes , as seen in Table 1 3 , it becomes a non viable solution. Tabl e 1 3 . Stock material cost comparison between traditional (1100 Aluminum and 304 Stainless Steel) and exotic materials (C Plane Sapphire, Silica Aerogel, and Non Porous Alumina) [6, 7, 8, 9, 10] . Material / mm 3 C Plane Sapphire 98.68 Silica Aerogel 0.94 Non Porous Alumina 0.13 304 Stainless Steel 0.08 1100 Aluminum 0.01 Another factor that makes simulator s more attractive manufacturing. Industrial lasers consume anywhere from several kilowatts to several

PAGE 24

24 hundred kilowatts while lasing [11, 12] . The move to minimize energy consumption reduces both the operati ng costs and the . New tools are necessary to allow the end user to minimize both trial and error test ing , and to reduce overall machining time. Additionally, with the high cost of ultrashort pulsed laser micromachining stati ons the machine time required for trials becomes quite costly ; especially in a production setting. A new method of determining the effect of parameters on the cut part is essential in reducing ultrashort machining costs and to reduce energy consumption . A simulator is one tool that may fulfill this need. Simulators allow the user to run part programs , with varying machining parameters, to predict the machined geometry of the finished component , and the processing time/cost for each respective set of parame ter s . It has the advantage of both verifying that the part path is correct, and allowing the user to tweak the parameters such that the desired component is produced. There are a variety of commercially available simulation programs for millisecond and nan osecond pulsed lasers [13, 14] , but there are few options for ultrashort lasers. One of the reasons for this is the high level of complexity in modeling the laser material interaction [15] , but it is also due to this technology being relatively new compared to longer pulsed lasers [4] . An other advantage of simulators is the ability to determine areas where excessive pulsing is occurring. Excessive pulsing occur s when the laser lingers in one location for an extended period of time or travels slower than desired , and causes a higher number of pulses to occur in that region. The excessive pulsing results in an increased amount of material removed at that location, and ultimately changes the depth profile of the workpiece. The extra material removed reduce s the mechanical

PAGE 25

25 strength of the finished component by introducing stress risers, and alter s the final geometry of the workpiece surface. Excessive pulsing also in crease s the amount of thermal energy imparted into the workpiece. Each pulse contains a certain amount of photonic energy, and when a pulse impacts the workpiece a portion of the photonic energy may be absorbed as thermal energy. When a large number of pul ses is absorbed in a small region over a short amount of time the relatively large amount of thermal energy can damage the workpiece because there is not sufficient time or material surrounding the region to dissipate the energy effectively. Characteristi cs of excessive pulsing causing an increase in the thermal load are discussed further in section 3.3. The advantages of utilizing a simulator are numerous. The potential to increase machining quality while simultaneously decreas ing operational costs and ca rbon output is attractive for industrial and research applications [16] ; however, the required knowledge to model this complex pr ocess is immense , as seen Table 1 3 . An accurate simulator require s the modeler to analyze the entire process, not just the laser material interaction. 1.2.2 Advantages of Automation/Correction Routines There is no known software currently available on the market that enables automated processing parameter selection for ultrashort pulsed lasers ; h ow ever, the idea of an automated process parameter software is not novel in traditional manufacturing. A variety of programs are available for CNC mills that perform these automated tasks [17, 18] . These programs are abl e to simplify p rocess parameter selection and part path optimization because it has been studied extensively for material removal by a defined cutting edge (e.g. CNC mills and lathes) [19, 20] . These CNC mill

PAGE 26

26 software packages are similar to the UF laser automated processing parameter program because there are analogous parameters of interests , shown in Table 1 4, between these two program types. Visualizations of the CNC mill parameters listed in Table 1 4 are provided in Figure 1 4 . Table 1 4. Comparison of CNC mill processing parameters and the analogous laser processing parameters. CNC Mill Processing Parameter Analogous Laser Processing Parameter Rotational Velocity (rpm) Pulse Repetition Rate (Hz) Tool Chip Load (µm) Pulse Area Overlap (%) Cut Depth (µm) Pulse Power (W) Endmill Diameter Focal Spot Size Endmill Material Laser Wavelength/Laser Pulse Duration A B Figure 1 4 . Demonstration of the three main processing parameters . A ) T ool chip load, and tool rotational velocity. B ) Endmill cut depth . Currently there are several methods for determining the optimal combination of processing parameters for CNC mills . First, t here are readily available tables and equations that relate suggested endmill tooth chip load and rotational velocity for a given tool and workpiece material combination. Several programs are also available that

PAGE 27

27 analyze the stiffness of the mill's spindle to determine the optimal combination of these parameters for a specific milling machine. Lastly, there are computer aided manufacturing (CAM) programs that generate part paths, as seen in Figure 1 5 , for a given geometry and suggests processing parameters based on the selected workpiece and tool. The availability of this information reduces the complexity of the machining process and allows for decreased part costs due to improved processing parameters. Figure 1 5 . Automated part path generation by a C AM program (Surfcam ® ) . The grey area is the workpiece surface, and the green lines reveal the cut path, aka the part path, that was generated by the CAM program. Similar to traditional machining machines, a laser operator has access to a variety of process parameters that ultimately affect the cost and quality of the machined component. Choosing the most appropriate combination of these parameters is difficult without knowledge of the machining processes , and a non optimal combination can result in a featur e that is unsatisfactory and/or overly expensive. Although a simulator allows the user to see the effect parameters have on the machined component, picking

PAGE 28

28 the optimal set of parameters for a desired type of cut is not a straight forward task . A large numb er of parameter combinations require additional time for simulations, and then the data would need to be analyzed to determine the most appropriate combination. An automated process parameter routine simplifies the parameter selection process, and decreases the amount of time required to determine the most appropriate set of parameters for each component. 1.2.3 Complexity of Temporal Processes One difficult aspect of laser machining is that it is a te mporal based process and not a position based process. Milling or turning operations are position based because a defined cutting edge removes material wherever that tool is located [21] . As long as there is no significant defl ection in either the cutting edge or component , it is straight forward to determine what material will be removed by analyzing the part path. If the part paths overlap in one location it has virtually no impact on the finished component because the defined cutting edge has already removed the material in that location. Therefore, modeling the final depth of cut is a relatively simple process in non deflection position based processes. Lasers machining is a temporal based process because material removal i s controlled by the laser energy workpiece molecular interaction, and the amount of energy imparted onto the workpiece by the laser is a function of time. Temporal based operations are more complex because one must analyze not only the part path but also t he position as a function of time. The laser beam removes a certain amount of material with every pulse , regardless of its location. If the laser is slower in one region, or overlaps another laser path, the final depth of cut is affected in that region. Th is can

PAGE 29

29 result in non uniform depths of cut, a characteristic that is highly undesired in many machining operations , or damage to the workpiece . Part path correction routines can reduce the effect of these depth irregularities by modifying both the machining parameters and also adding additional passes to the part path in specific locations . Reducing the pulse frequency and the commanded velocity helps minimize the effect of acceleration irregularities, but reducing the rates also increases the time to machine a component. A correction routine can analyze the part path to determine the best tradeoff between the required time to machine the component and the resulting depth irregularity. The correction routine can also append additional passes into the cut program. The process of adding in additional passes includes analyzing the simulated cut geometry and determining where and how many additional passes are necessary. This is a n a ttractive feature because it compensates for acceleration regions and imp erfect part paths. 1.3 Research Contributions The first contribution of this dissertation is the quantification of the relationship between machining parameters and ablation depth. T he effect s of pulse delay , pulse power , and pulse overlap on the ablation depth of sapphire is considered and an empirical relationship is presented. The computed relationship allow s for computation of the ablation profile for any setup, regardless of the combination of parameters selected. The second contribution is the pulse d laser micro machining simulator . The custom Matlab ® program read s g code files and produces a 3D simulated cut. It include s compensation for stage dynamics , which enable s mitigation of excessive pulse overlap trouble zones . The program is capable of different acceleration profiles

PAGE 30

30 (constant acceleration, constant jerk, and sinusoidal profile). Lastly, the program estimate s the amount of power consumed , cost, and time to cut a component . The final contribution is the part path correction p rogram with an automated process parameter routine . Th e Matlab ® routine is integrated into the simulator. The routine analyzes the simulator data and correct s the part path for the acceleration profile of a given laser machining station. Corrections can ei ther change machining parameters for the whole program, divide the feature into multiple cuts with separate parameters, or change parameters on the fly. 1.4 Dissertation Organization This chapter established the need for an ultrashort pulsed laser simulato r and automation/correction routines for machining of materials in an industrial setting . The complexity of the machining process was explored, along with an explanation of the contributions of this dissertation. In Chapter 2 the methods for producing lase r and pulsed lasers is explored, as well as the characteristics of pulsed lasers, and a review of previous simulation models. Chapter 3 explores the methodology behind the new simulator with acceleration compensation. Then , Chapter 4 delves into methods for correcting the part path and automating process parameter selection of the laser cut programs . J 355ps micromachining station . Next, Chapter 6 discusses sapphire mate rial properties and explains the results of the simulation/modification routines. Lastly, Chapter 7 concludes the dissertation with a summary of the research contribut ions and a discussion of possible future work.

PAGE 31

31 CHAPTER 2 L ASER MICROMACHINING B ACKGROUN D This chapter covers the background of laser micromachining. First, the history of the laser is discussed. Then, the different types of lasers are explored, along with the different methods for producing laser pulses. Next, a comparison of the different l aser types is discussed, and finally a review of previous simulator models is performed. 2.1 History and Physics of Lasers Albert Einstein first theorized the driving physics of a laser in a 1917 paper titled Zur Quantentheorie der Strahlung (On the Q uantum Theory of Radiation ) [22] . p aper stated that photon radiation could stimulate additional photon radiation when it collides with an atom [23, 24] . The requirement was that the energy of the stimulated radiation is approximately equal to the energy lost when the atom relaxes from an upper energy level to a lower energy level [23] . In the late 1920s Rudolph Ladenburg conf and negative absorption [1] . Although much effort went into stimulated emission, it was not until April 1954 that Charles Townes and Jim Gordon produced th e first microwave a mplifier; a maser (microwave amplification by stimulated emission of radiation) [1] . state. An amplifier would excite ammonia ato ms in the ground state to the excited state through absorption. At this point the atoms in the excited state seek to drop back to the ground state, and this is accomplished by either spontaneous emission or stimulated emission. In spontaneous emission a se t of atoms spontaneously decays to the ground state. This produces non coherent emission because there is no fixed phase between the decaying groups of atoms; consequently, this type of emission is undesired in lasers

PAGE 32

32 mulated emission, which occurs when an incident photon passes by the excited atom and causes it to relax. When the atom relaxes, it emits a photon of the same frequency , polarization, direction as the passing photon. The repeated absorption emission cycle produces a highly coherent beam, and is the driving force of the maser/laser [1, 23] . population inversion, which rendered it incapable of continuous operation [1, 23] . Equation 2 1 shows the equilibrium eq uation for population inver sion (2 1) In the above equation and are the number of atoms in each state, and are the respective energy levels of each state, is the absolute thermodynamic temperature, and was unable to repopulate the excited state faster than the rate of atoms relaxing, and consequently, it was unable to maintain the rate of radiation emission [1, 23] . In 1955 Nikolay Basov and Aleksandr Prokhorov, and Nicolaas Bloembergen independently theorized that a multi level (two or more excited levels) system would allow a maser to maintain the population inversion. This multi level system, or quantum oscillator as it is better known, uses a third state in betwee n the excited and ground states. Figure 2 1 demonstrates this concept with a three energy level system. Atoms transition between three levels, with a high ly excited level (E 2 ) , a metastable level (E 1 ) , and a ground level (E 0 ) . The first step in producing e mission is to excite atoms up to the high ly excited level (E 2 ). Then, the atoms spontaneously decay do wn to the metastable

PAGE 33

33 level. This decay can produce spontaneous emission, but the change in energy is typically not sufficiently high enough for that to occur; consequently, the energy is dissipated as vibrations in the lattice. The last step is for the atoms to relax down to the ground state and to emit a photon. Note that due to the losses from the spontaneous decay, excitation energy must be greater tha n the energy required for laser emission (E 2 >E 1 E 0 ) [25] . In 1956 Bloemberger and Henry Scovil created the first solid state multi level maser [1, 23] . Figure 2 1. Diagram of a three energy level laser , after [25] . After the maser, the race was on to produce the first laser. Townes and Gordon Gould both came up with preliminary ideas for the laser in 1957, but Gould is generally contributed with the invention of the laser acronym. Although Townes would eventually pu blish a more detailed laser concept along with his brother in law in 1958, Theodore

PAGE 34

34 synthetic ruby with silver coated ends to make the ruby a resonator. A flashlamp s upplied the required excitation energy. Even though Maiman's laser was the first of its kind, it was incapable of continuous operation. It was not until 1960 that Ali Javan, William Bennett, and Donald Herriott produced a HeNe laser that operated continuou sly. In 1961 the first commercial lasers became available, and ever since then laser technology has been growing tremendously [1, 23] . 2.2 Types of Lasers The number of applications for lasers is immense, and the need to customize lasers for specific applications has driven the development of a wide variety of laser types [3] . Lasers are categorized into four main groups: gas, dye, semiconductor, and solid state [3] . Some of the most important differences in these laser types are power output, pulsing capabilities , beam wavelength, cost of operation, and reliability. A sampling of the most common types is discussed, along with their method of operation, applicat ions, and characteristics. Table 2 1. Pros and cons for different laser types [26] . Laser Type Pros Cons Gas High power, reliable Large cavities due to low gas densities Dye Higher excited molecule density Dyes and solvents are typically poisonous and/or carcinogenic Semiconductor No moving components Low laser power output Solid state High energy output without moving parts Laser material degradation and the need for exotic materials 2.2.1 Gas Lasers The first laser type of interest is the gas laser. Gas lasers have been around since 1960, with the advent of the HeNe laser [27] . Typically, the excitation energy in these lasers is supplied by electrical current discharge in to a gas. The main advantages

PAGE 35

35 of gas lasers is the gas acts as a homogeneous laser medium, is easily transported for cooling and replenishment, and the gases are relatively inexpensive. The biggest disadvantage is the low density of the gas that requires a greater volume to satisfy the required number of atoms to maintain the population inversion. Consequently, these laser types are generally large. Several lasers in this category are the argon ion, carbon dioxide, carbon monoxide, excimer, krypton, nitroge n, and xenon ion. Two of the more common gas lasers are discussed below [24, 25, 26] . Figure 2 2. Schematic of an externally pumped CO 2 gas laser , after [2 4] . The first gas laser of interest is the carbon dioxide (CO 2 ) laser. The CO 2 laser was developed in 1964 by Kumar Patel [27] . A schematic of an externally pumped CO 2 laser is shown in Figure 2 2 , and this schematic is representative of other externally pumped gas lasers as well . The laser functions by CO 2 molecules emitting photons via stimulation, then the pump removes the relaxed molecules, electrical discharge (AC current for this schematic) e xcites the relaxed molecules, and the process is repeated again [24, 25, 26] .

PAGE 36

36 A B C Figure 2 3. Schematic of the three d ifferent modes of vibration of a CO 2 molecule . A) Asymmetric stretch mode, B) bending mode, and C) symmetric stretch mode , after [26] . This laser utilizes the three modes of vibration for the CO2 molecule , as shown above in Figure 2 3 . Emission occurs when the molecule transitions from the highest energy level, asymmetric stretching , to either symmetric stretching that produces 10.6 µ m radiation or bending that produces 9.6 µ m radiation. The power output of these la sers can r un into the hundreds of kilowatts , which makes them attractive for welding, laser machining, and surface modification. It is also one of the more efficient (ratio of output photonic energy to input electrical energy) lasers, with energy efficiencies as hig h as twenty seven percent [25] . Additionally, it is possible to configure CO 2 laser s to operate as either a CW or pulsed laser. The ability to run in pulsed mode increases the usefulness of the laser as a machining tool, but th e CO 2 -

PAGE 37

37 nanosecond pulses limits the materials applicable to this laser. Overall, the CO 2 laser has gained wide acceptance in industry and is known for its low cost of operation and high reliability [24, 25, 26] . The next gas laser examined is the excimer laser. The excimer laser was first developed by Nikolai Basov, V. A. Danilychev, and Yu. M. Popov in 1970 [27] . These lasers are typically made up of a noble gas (Ar, Kr, Xe) and a reactive gas (Cl, F). In the ground state, the formation of a diatomic molecule of these two gases is unstable. However, in the excited state a dimer molecule is formed that is stable. This laser has the advantage of short wavel engths (157 nm to 351 nm) that result in high photon energies. They also have the ability to operate in nanosecond scale pulsed operation. The typical applications for an excimer laser are micro patterning, surfa ce modification, and thin film deposition. T his type of laser is typically used for micro patterning and thin film deposition because the high photonic energies allow for direct photodecomposition of target material s. Additionally, t he short pulse length m inimizes damage to the surrounding material . Lastly, the highly multimodal output results in poor spatial coherence that reduces interference effects that would otherwise reduce the quality of the micro patterning [4] . 2.2.2 Dy e Lasers The next type of laser is the dye laser. The first dye laser, a chloro aluminum phtalocyanine dye laser, was developed in 1966 by Peter P. Sorokin and John R. Lankard [27] . A dye laser utilizes a liquid, consisting of an organic dye and a solvent, to absorb and radiate (see Figure 2 4) . A key aspect of dye laser s is the ability to tune the output wavelength by altering the dye solvent combination. Table 2 2 shows the output of several different dye solvent combinations .

PAGE 38

38 Figure 2 4. Schematic of an externally pumped dye laser [24] . Table 2 2. Output wavelength for different combinations of dyes and solvents [25] . Dye Solvent(s) Output Wavelength Acridine red EtOH 600 630 nm Rhodamine 6G EtOH, MeOH, H 2 0, DMSO, Polymethyl methacrylate 570 610 nm Rhodamine B EtOH, MeOH, Polymethyl methacrylate 605 635 nm Na fluorescein EtOH, H 2 0 530 560 nm 7 Hydroxy coumarin H 2 0 (pH~9) 450 470 nm 2.2.3 Semiconductor Lasers The third laser type is semiconductor lasers. Nick Holonyak Jr. was the first to develop a semiconductor laser, a GaAsP type, in 1962 [27] . Semiconductor lasers are made up of two types of semiconductors; p type and n type (see Figure 2 5) . When a positive potential is applied to the p type region, the potential barrier is reduced and electrons can diffuse from the n type to p type regions. Additionally, h oles diffuse from

PAGE 39

39 the p type to n type region. During this diffusion both electrons and holes act as charge carriers and undergo recombination. The recombination is the driving factor in radiation emission. Semiconductor wavelengths range between 630 680 n m for AlGaInP lasers and 1.15 1.65 um for InGaAsP lasers. The key characteristic of semiconductor lasers is the wide divergence angle of the laser [24, 25, 26] . Figure 2 5. Schema tic of a semiconductor laser, after [24] . 2.2. 4 Solid State Lasers The last major type of laser is the solid state variety. Although the semiconductor lasers might at first appear to fall into this category the two types utilize different methods of producing photons. Semiconductor lasers produce emission by the recombination of charge carriers, and solid state lasers produce e mission by the decay of an atom from an excited state to a lower energy state. The use of energy states to

PAGE 40

40 produce emission has several advantages that include higher output power and output wavelength t u nability [26] . The firs t ever laser was a solid state laser, the ruby laser, developed in 1960 by Maiman [27] . The basics of these lasers are that a material is pumped with a light source , such as the one in Figure 2 6 . The energy is absorbed and at oms are excited. Depending on the material, t here is anywhere from a single excited state to 4+ excited states. These laser types are used primarily in laser machining [24, 25, 26] . Figure 2 6 . Visualizatio n of a solid state laser. Diodes can also be used as the excitation source in place of flashlamps , after [24] . Although each solid state laser is slightly different, the Nd:YAG laser is the most widespread so it is a typical example of how they operate. The Nd:YAG laser consists of a host material, Y 3 Al 5 O 12 , and a impurity ion, . The laser is composed of four pumping energy levels (four possible levels that excited atoms are pumped into) and four final levels (four levels atoms can relax to after emission). There are several possible wavelengths (266 nm, 355 nm, 532 nm, and 1064 n m) due to the four

PAGE 41

41 relaxation levels, but the most common and strongest wavelength for a Nd:YAG laser is 1064 n m. This laser type has become popu lar because of the output's ability to be continuous or pulsed. The pulsing mechanism ranges from a Kr or Xe arc lamp for continuous operation to flashlamps for pulsed operation. Typical applications of Nd:YAG lasers are laser machining, and surface modifi cation [24, 25, 26] . The above mentioned lasers are just a sampling of the ones available on the market . I t is likely that in the future even more laser type s with a wider range of capabilities will become available . A variety of articles with more detailed explanations and reviews of different laser types and applications is available from the Journal of Laser Applications . 2.2.5 Laser Pulse Generation Many lasers have the capability to be op erated in either a continuous wavelength mode (CW), or a pulsed mode [24] . Pulsed lasers are advantageous to laser machining continuously bombard t he workpiece surface with photons, but pulsed lasers send packets of photons that only interact with the surface for the duration of the pulse. The short interaction time decreases the amount of energy that diffuses into the surrounding workpiece [4] . This phenomenon decreases the depth of the HAZ in the cut region , and can reduce or eliminate cracking due to thermal shock [5] . The first laser invented was a pulsed laser. This was not due to the des ire for a pulsed mode, but rather due to an inefficient pumping design [1] . Current methods of producing pulsed lasers allow for different pulse durations, pulse repetition rates, and peak power. Present day pulsed lasers typic ally use one or two of the following four methods to produce pulses: Q switching, mode locking, cavity dumping, and gain

PAGE 42

42 switching [28] . The characteristics of each of these techniques is explored and the typical applications reviewed. 2.2.5.1 Q Switching The quality factor of a resonator cavity is denoted by Q , and it correlates to the amount of loss occurring in the resonator. Lasers operated by Q switching function by modifying the Q factor of the resonator cav ity. This is accomplished by either modifying resonator cavity. There are two main methods for switching the Q factor; actively or passively. Figure 2 7 demonstrate s the gain energy, resonator losses, and output power of the laser before, during, and after pulse generation for an actively Q switched laser . At first the losses in the resonator are set to a high level that inhibits lasing. Then, the gain medium is pump ed up until spontaneous emission or lasing occurs . Next, the Q factor is rapidly increased so that stimulated emission occurs quickly. The peak power of the pulse builds until the energy left in the gain medium equals that of the resonator losses [25, 28] . The pulse duration for a Q switched laser depends on the rate of energy depletion in the gain medium. Consequently, this type of laser is not suitable for ultrashort pulse generation, but it is capable of nanosecond pulse generation. Typical repetition rat es are in the range of 1 to 100 kHz, and peak power in the kilojoules range [25, 28] .

PAGE 43

43 Figure 2 7. Gains and losses as a function of time for an actively Q switched laser, after [28] . 2.2 .5.2 Mode Locking The next method of pulse generation is mode locking. This method utilizes either a passive or active element in the resonator to produce an ultrashort pulse that circulates in the resonator. In active mode locking either the round trip ph ase change or the resonator losses are modulated periodically. T he modulation is performed by either an acousto optic (AOM), electro optic (EOM), Mach Zehnder integrated optic, or a semiconductor electroabsorption modulator [28] . Figure 2 8 demonstrates the way an actively modulated mode locked laser operates. The modulator cycles between a high amount of loss and minimal loss. When a pulse synchronizes with this cycle, the leading and trailing edges of the pulse are attenuated. This produces a shortened pulse that typically is in the picosecond duration [25, 28] . 0 0.05 0.1 0.15 0.2 0.25 0.3 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 Magnitude Time ( s) Gain Total Losses Output Power

PAGE 44

44 Figure 2 8. Optical power and losses as a function of time in an actively mode locked laser, after [28] . Figure 2 9. Optical power, losses, and gain as a function of time in a passively mode locked laser, after [28] . 0 0.2 0.4 0.6 0.8 1 1.2 0 0.5 1 1.5 2 2.5 3 Magnitude Time ( s) Losses Output Power 0 0.2 0.4 0.6 0.8 1 1.2 0 0.5 1 1.5 2 2.5 3 Magnitude Time ( s) Losses Output Power Gain

PAGE 45

45 Passive mode locking utilizes a saturable absorber as the modulation mechanism. The modulator absorbs a certain amount of energy before it becomes saturated. Figure 2 9 demonstrates the losses associated with the absorber. When a random photon hits the modulator it is simply absorbed because the losses are greater than the amplification, but when a pulse with sufficient energy impacts the absorber it is reflected. When a pulse impacts the absorber the leading edge of the pulse is absorbed, and if the absorber recovers quickly enough it can also ab sorb the trailing edge. The continual attenuation of the leading and trailing edge of the pulse creates ultrashort pulses, typically in the femtosecond range [25, 28] . 2.2.5.3 Cavity Dumping Cavity dumping is another me thod of producing laser pulses. In this method both of the mirrors in the resonator are highly reflective, instead of one being partially transmissive. Cavity dumping is paired with either Q switching or mode locking to produce the high energy pulses. The selection of which method s are paired depends on the desired length of the pulse. The incorporation of Q switching and mode locking with cavity dumping is explored below [25, 28] . Figure 2 1 0 (a) illustrates the setup for a Q switching cavity dumping laser. The laser first acts similarly to a standalone Q switched laser, building up energy in the gain medium. After the medium is sufficiently built up, then the Q factor is suddenly raised and the laser power b uilds up quickly. Once the power reaches a certain level or number of resonator cycles, an AOM is activated and pulses are extracted. This method typically produces nanosecond pulses, and the pulse length is limited by the response time of the AOM [28] .

PAGE 46

46 A B Figure 2 10. Schematics for two types of cavity dumping systems . A) Q switched cavity dumping, and B) mode locked cavity dumping , after [28 ] . It is possible to generate u ltrashort pulses with cavity dumping when mode locking is utilized. This system is similar to that of mode locked lasers, but without a partially transmissive mirror emitting the pulses. Figure 2 1 0 (b) illustrates that i nstead of the partially transmissive mirror , a fast photodiode monitors the pulses and when they have achieved the maximum energy level an EOM and a polarizer extract the pulses.

PAGE 47

47 This method allows for a much higher level of energy than a standalone mode l ocked laser, typically an order of magnitude higher [28] . 2.2.5.4 Gain Switching The last pulse generation method is gain switching. This method uses a simple technique of generating pulses; it pulses the pump power. Figure 2 1 1 demonstrates that t he pump power is first either turned off or in a lower emission state . T hen , the medium is quickly pumped to a high energy level and shut off. Shortly after, a high energy pulse develops and continues to build until the medium energy i s drained. The generated pulse can have a duration shorter than that of the pump pulse. This allows for the generation of nanosecond or picosecond pulses, depending on the laser type [25, 28] . Figure 2 11. Output powe r, and pump power as a function of time in a gain switching laser, after [28] . 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.2 0.4 0.6 0.8 1 Magnitude Time ( s) Pump Power Output Power

PAGE 48

48 2.2.6 Coherent Talisker Laser Specifications The above section explored a variety of laser types ; each type having its respective pros and cons. One desired property for machining sapphire is the ability of the laser type to produce ultrashort pulses. In the early 1980s the typical choice was dye lasers, but these laser types have limited power and repetition capabilities. Ti:Sapphire lasers are capab le of producing femtosecond duration pulses, but the power output is typically only in the 0.3 1.0 W range [28] . Nd:YAG lasers have higher output power, but are typically limited to picosecond range pulse duration [29] . Research at The University of Florida (UF) is geared towards producing industrially scalable machining processes; therefore, a Nd:YAG laser is utilized to allow for higher MRR. The Coherent Talisker laser at UF utilizes a Coherent Ptarmigan diode seed laser that produces low energy 5 picosecond pulses at a rate of 40 MHz. The pulses are fed to a Nd:YAG crystal via a fiber optic cable. The high energy pulses are generated by a mode locked cavity dumping resonator system, an d the repetition rate is further selected using an AOM. Lastly, a pair of Lithium T riborate (LBO) crystals allow for 2nd and 3rd harmonic generation [30] . Although 1064 nm and 532 nm outputs are available, the UF system utilize s the third harmonic of 355 nm because it is more easily absorbed by transparent materials than the first two harmonics [4] . The laser is rated at a maximum of 4 watts average power, 200 kHz max repetition rate, and a nominal 1 0 15 picosecond pulse duration [30] . 2.3 Laser Material Removal Mechanisms Laser material removal is a function of several parameters, including but not limited to: pulse duration, pu lse fluence, and pulse area overlap . Pulse d uration is

PAGE 49

49 defined as the time elapsed between the initial moment the laser beam amplitude reaches a certain threshold , and the moment when the amplitude drops below that value again (see Figure 2 1 2 ) [28] . The pulse duration is a highly important aspect of laser machining because it has a significant effect on the type of ablation and consequently on the amount of thermal energy imparted into the surrounding material. Typically, the shorter the pulse duration, the less thermal energy is imparted, but with shorter pulses also comes the drawback of a lower average power output from the laser [5] . Figure 2 set that determines where the pulse begins and ends , after [31] . 2.3.1 Types of A blation Laser machining involves two modes of material removal; pyrolitic and photolitic. Pyrolitic relies on localized heating, while photolitic utilize s direct ionization (The detailed physics of these material removal mechanisms is discussed in section s 2.3.1.1 and 2.3.1.2). What separates the two modes is the difference in the optical penetration depth of the laser energy, and the thermal diffusivity of the material. T he initial intensity of the incoming beam , before it strikes the surface of the workpiece is

PAGE 50

50 , (2 2) and the intensity, at a depth into th e workpiece is (2 3) In the above equations is the pulse energy, is the pulse duration, is the focal beam radius, and is the optical absorpti vity of the material [5, 32] . From Eq uation 2 3 we can derive the optical penetration depth, where most of the laser energy has been dissipated (~86%) [5] , (2 4 ) To determine if the mode of material removal is pyrolitic or photolitic , one compares the optical penetration depth of the laser to the thermal diffusion depth , (2 5 ) w here is the thermal diffusivity, and is the diffusion time (generally considered to be the pulse duration) [5, 32] . The thermal diffusion depth is therefore a measure of the distance over which temperature changes occur during the pulse's duration. To minimize the quantity of thermal energy diffusing into the surrounding material , it is necessary for photons to penetrate fu rther into the material than the depth of thermal diffusion [5] . If the optical penetration depth is equal to or greater than the thermal diffusion depth ( ) , then p hotolitic material removal dominate s ; however, if the

PAGE 51

51 thermal diffusion depth is greater ( ) , then p yrolitic is the main mode for material removal [5] . Combining the photolitic requirement , Equation 2 4, Equation 2 5, and substituting for in Equation 2 5, the p ulse duration for a process to be dominated by photolitic ablation is . (2 6) The term ultrashort refers to lasers that have sufficiently short pulses that the process is considered photolitic dominant. It is important to note that to ensure a machining operation is actually photolitic dominated, one would need to analyze the materi al and the respective wavelength of the laser system. This is a result of materials being more or less transparent to certain wavelengths; this affects the absorbed photonic energy and optical penetration depth [5] . Table 2 3 s hows roughly typical maximum pulse length values for photoliti c material removal to dominate . Table 2 3. Rough estimate of the m aximum pulse durations for photolitic material removal to dominate [5] . Material Pulse Length Metals 1 picosecond Ceramics 10 picoseconds Plastics 1 nanosecond 2.3. 1.1 Pyrolitic Ablation Pyrolitic ablation is the heating of a material to produce melting or vaporization at the surface. P yrolitic ablation dominates f or CW , microsecond, and nanosecond pulsed lasers. Figure 2 1 3 shows a diagram of the pyrolitic process that involves three steps . First, photons bombard the surface and cause a heated zone. Next, as photon

PAGE 52

52 bombardment continues a part of the heated zone liquefi es . Lastly, the liquid vaporizes causing mass ejection of the material [33] . Figure 2 13. Visualization of the p yr olitic material removal process, after [33] . There are several advantages to pyrolitic material removal. First, the lasers that generate pyrolitic ablation are longer pulse or CW. These types of lasers are generally regarded as more reliable than their ultrashort counterparts and typically cost less [28] . Also, these laser types usually have a much higher average power output that leads to an increased material removal rate [5] . There are also disadvantages of pyrolitic material removal. The h igh thermal load imparted on the workpiece creates a large HAZ in metals , as seen in Figure 2 14, that can alter material properties [4] . In plastics this high thermal load can melt or burn the workpiece, and in ceramics the th ermal shock can lead to cracking [3] . In addition to the thermal damage , liquefied materials can reform on the surface of the workpiece and require additional steps to remove the redeposited material [34 ] . Lastly, it is difficult to

PAGE 53

53 produce narrow features because the excessive thermal energy is not easily dissipated to the surrounding material. A B Figure 2 14. Visualization of the HAZ in laser machining. A) P yrolitic dominated material removal processes produces a large HAS, and B) photolitic dominated process has a reduced HA Z . To demonstrate the difficulty of machining narrow features with a high thermal load, a thermal finite element simulation is performed in Solidworks Simulation ® on a silicon wafer (Solidworks assumes a thermal conductivity of 124 W /mK for silicon) . A spot size of 10 µm produces a simulation with temperature distributions that more closely match the UF laser system, but that int ensity level results in material vaporization. A simulation with that level of complexity is beyond the scope of this example , so a larger spot size is deemed sufficient to convey the high thermal load effect. A continuous 4 watt thermal load is placed on a 1 mm diameter circle in the middle of the wafer , and the edges of the wafer are constrained at 50 degrees Celsius . Then, the simulation is performed on the two geometries shown in Figure 2 1 5 . The figure illustrates the difference between a trench cut in to a large sheet versus one on a

PAGE 54

54 2 mm wide by 8 mm long member. The only direction to dissipate the thermal energy away from the laser spot in the latter case is through the thin member that has a high er thermal resistance ; c onsequently, unintentional melting , warping, or buckling of the member is a possibility (especially with higher power lasers) . A B Figure 2 1 5 . Thermal finite element simulation temperature distribution results . A) Simulati on with a whole wafer, and B) simulation with a thin member. The simulated temperature difference for the thin member is ~5X greater than the whole wafer. 2.3.1.2 Photolitic Ablation Photolitic ablation is the removal of material due to the ionization of the workpiece atoms/molecules. Ionization occurs when an electron in the valence band of an atom or molecule absorbs sufficient energy to move the electron into the conduction band. Lasers accomplish this by bombarding the surface of the workpiece with photons that , when absorbed by the atoms/molecules , result in mass ejection of the ionized atoms/molecules (see Figure 2 16) [5] . When photolitic ablation dominates as the main mode of material removal, the laser is considered to have ultrashort pulses. Picosecond

PAGE 55

55 and femtosecond lasers generally have properties that result in a mostly photolitic ablation process; therefore, picosecond and femtosecond lasers are typically classified ultrashort. A B Figure 2 1 6 . Material removal by ionization. A) Material is bombarded by photons resulting in electrons being stripped from the atoms, and B ) highly ionized plasma is ejected from the surface , after [33] . Photolitic ablation allows for a higher degree of precision in the final machined component when compared to pyrolitic ablation . The flexibility of machining virtually any material is unsurpassed by longer pulses [33] . Also, the use of ionization rather than thermal heating produces virtually no HAZ [33] . This is important in machining ceramic materials, especially those that are sensitive to thermal shock , because the HAZ can initiate cracks and possibly lead to failure of the component . Lastly, the ablation efficiency ( the ratio of material removed for a given amount of photonic energy) of photolitic ablation is higher than that of pyrolitic [35] . Energy is only needed to ionize the workpiece and not to produce a phase change. This decreases the required average

PAGE 56

56 power in order to remove the same amount of material. An example of the efficiency difference is shown in Table 2 4. Th e femtosecond laser removes approximately 50% 75% more material for the same pulse fluence, and the thermal diffusion depth of the femtosecond laser is [5] . Table 2 4 . A co mparison of the energy required to remove material for different pulse durations. The machined material was diamond, and the wavelength for both pulse durations was 248 nm [5] . Pulse Length Pulse Fluence Removal Rate Per Pulse Removal Rate Per Fluence Unit (J/cm 2 ) (nm 3 ) (nm 3 cm 2 /J) 20 ns 2 10 5 20 ns 4 20 5 20 ns 6 30 5 500 fs 2 15 7.5 500 fs 4 35 8.8 500 fs 6 50 8.3 There are also disadvantages associated with photolitic material removal. As stated above, the lasers that produce picosecond and femtosecond pulses are generally more expensive ; leading to increased operating expenses [5] . The technology behind ultrashor t pulsed lasers is still relatively untested ; consequently, the reliability of these lasers is less than that of CW, long, and short laser systems [36] . Lastly, while the ultrashort pulses commonly have higher peak power, the average power output is less than that of CW or long pulsed lasers [5] . So although the photolitic method is more efficient, the smaller power output leads to a lower material removal rate that increase s operating costs. 2.3.2 Weak and Strong Ablation Regions Several diff erent types of ionization can occur as a result of ultrashort pulsed laser bombardment . The three main methods of accomplishing ionization are single -

PAGE 57

57 photon, multi photon, and avalanche [4] . The type of ionization is dependent on not only the laser intensity but also the laser wavelength, number of pulses , the pulse frequency , the ionization threshold of the workpiece, and the workpiece absorptivity . The three ionization types are generally grouped into one of two categories : ge ntle and strong [37] . Gentle ablation is generally single and multi photon ionization, and avalanche ionization is classified as strong ablation. Gentle ablation is considered to have a more consistent material removal rate from one pulse to the next, but the rate is lower than that of strong ablation [37] . The reason for these different ablation characteristics is revealed by analyzing th e ionization types . These ionization types are discussed below. A B Figure 2 1 7 . Demonstration of single photon ionization. A) A single photon bombards an atom , and B) the impact results in the loss of an electron, a fter [4] . Single photon ionization occurs when a laser bombards the surface of a workpiece with photons of sufficient energy that a single photon can ionize workpiece atom s /molecule s (see Figure 2 17) . This is only possible if the photonic energy of a

PAGE 58

58 single photon is greater than or equal to the ionization energy, . The energy of a photon, , is defined as [25] , , (2 7 ) where is the speed of light, and is the photon wavelength. The ionization energy is dependent on the type of element and is generally extrapolated through empirical methods. Single photon ionization is rarely found in micromachining because the maximum wavel ength of the photon is lower than the wavelength that most industrial lasers operate. Although single photon absorption is of minimal relevance to laser micromachining, it is of importance in laser pulsing so it is an essential aspect to understand [4] . Figure 2 1 8 . Demonstration of 2 multi photon excitation methods, after [4] .

PAGE 59

59 If the energy of a single photon is not sufficiently high to ionize an atom/molecule there is another way to cause ionization; multi photon absorption. Multi photon ionization occurs when two or more photons impact an electron and the combined energy is sufficiently high to cause ionization . The two ways two photon absorption occur , as shown in Figure 2 1 8 , are through coherent absorption and sequential absorption. In sequential absorption a single photon is absorbed and results in the energy level of a valence electron to be raised to a higher level. Then, before the valence electron dissipates causes the valence electron to become unbound . The second method, coherent absorption, occurs when two photons are absorbed simultaneously. Coherent absorption is typically only possible if the l aser has a relatively high intensity because the density of photons in a particular region must be sufficiently high that two photons interact with a single atom/molecule at the same time [4, 24] . When lasers reach a s ufficiently high intensity, multi photon ionization can lead to avalanche ionization, as seen in Figure 2 19. The rate of electron excitation is related to the laser intensity. Excited electrons dissipate energy via phonon generation, but if the rate of ex citation is greater than the rate of phonon generation, then electrons become so highly charged that they can interact with other molecules and cause impact ionization. The transition from multi photon ionization dominating to avalanche ionization dominati ng is not immediate because a sufficient number of highly charged electrons must first be generated. This transition can last anywhere from several femtoseconds to several picoseconds depending on the material, pulse duration, photon wavelength, and intens ity. Figure 2 20 shows two graphs of different pulse durations

PAGE 60

60 and their respective contributions of free electrons due to multi photon and avalanche ionization. The total density of free electrons is denoted by , the density contributed by multi photon ionization is mp , and the density contributed by impact ionization is imp . The trend is that as the pulse duration increases, the fraction of free electrons generated by impact ionization increases [4, 24] . A B Figure 2 1 9 . A comparison between multi photon ionization and impact ionization. A) A n electron in the valence band is shifted into the conduction b and by a strong electric field in multi photon ionization . B) In impact ionization a n electron with a high kinetic energy transfers some of the energy to another electron in the valence band; resulting in that electron to also sh ift it into the conduction band, aft er [38] .

PAGE 61

61 A B Figure 2 20 . Free electon densities and the respective contributions from multi photon ionization and impaction ionization . A) Shows the contribution for a 50 fs pulse, and B) shows the contribution for a 200 fs pulse as a function of time after a 500 nm wavelength laser pulse with an electron field of 150 MV/cm impacts SiO 2 , after [38] . 0 0.2 0.4 0.6 0.8 1 1.2 0 25 50 75 100 125 150 mp, imp (10 21 /cm 3 ) t (fs) Multi-Photon Impact Free-Electron 0 1 2 3 4 5 6 7 0 50 100 150 200 250 300 , mp, imp (10 21 /cm 3 ) t (fs) Multi-Photon Impact Free-Electron

PAGE 62

62 A B Figure 2 2 1 . Strong and gentle ablation traditional models. A) The effect of pulse number on the transition, and B) the effect of pulse power on the material removal rate , after [37] . Figure 2 2 1 demonstrates a generally accepted trend for the transition from a gentle, multi photon dominant, ablation to a strong, avalanche ionization dominant, ablation . Predicting the transition from gentle to strong a blation is not a simple process, because each material reacts slightly different to photon excitation. Avalanche ionization is generated by either body defects, such as Frenkel defects that are generated after a series of pulses, or by a high intensity pul se that causes a high order multi photon ionization [4, 24] . Lower energy pulses may never transition to strong ablation because they are unable to create the minimum number of body defects for avalanche ionization to occur. In contrast, high energy pulses may experience strong ablation on the first pulse because they generate high order multi photon ablation. While simulations may eventually reduce the complexity of this topic, current methods for determining the tran sition require empirical data. Methods for determining the transition are discussed in Chapter 6 [4, 24] .

PAGE 63

63 2.4 Machining Characteristics for Different Pulse Durations The method of pulse generation and the theoretical p hysical interaction resulting from the pulse interaction is discussed in the previous sections , but it is important to review experimental results as well to grasp a better understand of machining parameters . A B C Figure 2 2 2 . Recreation of holes drilled in steel foil with different pulse duration lasers . A) Shows a representation of a nanosecond laser hole, B) a representation of a picosecond laser hole, and C) a representation of a fe mtosecond laser hole , after [39] . The duration of a laser pulse has an effect on feature quality. Figure 2 2 2 demonstrates the drilling of a 100 m diameter hole in steel foil using

PAGE 64

64 t hree different lasers (see Tabl e 2 5 for laser properties) . T he nanosecond laser , 2 2 2 ( A ), generates melting and resolidifcation of the melted material around the opening of the hole [39] . This may be an acceptable result in certain applications, but if the surface finish is important then a nanosecond laser would be a poor tool for this feature. The picosecond hole , 2 2 2 ( B ) , shows a large amount of debris inside of the hole, but less redeposit of material at the opening. It is possible that the material in t he hole is redep osit that further processing can remove , but the additional step(s) add to the overall processing time and cost . The last hole , 2 2 2 ( C ), drilled with a femtosecond laser, shows minimal melting and redeposit. Table 2 5. Respective pulse dur ation, pulse energy, peak intensity, and wavelength for the holes drilled in Figure 2 2 2 [39] . Hole Pulse Duration Pulse Energy Peak Pulse Intensity Wavelength A 3.3 ns 1 mJ 4.2 J/cm 2 780 nm B 80 ps 3.7 J/cm 2 780 nm C 200 fs 0.5 J/cm 2 780 nm There are more aspects to lasers than just fea ture quality though. Below is a summary of characteristics for each respective pulse duration. While these characteristics are not concrete for every laser in the category, a basic knowledge of the characteristics is necessary to understand why the simulation , parameter selection, and part path of a picosecond laser is important. 2.4.1 Continuous Wavelength (CW) , Millisecond, and Microsecond Pulses CW , millisecond, and m icrosecond lasers exhibit similar machining characteristics. All three rely heavily on thermal material removal, and generally have the highest average power [5] . Millisecond and m icrosecond lasers are considered long pulse and ha ve some thermal advantages over CW, but all three produce a large

PAGE 65

65 amount of damage to the surface and surrounding material [4] . They are, however, typically the cheapest of the four listed here, with the highest material rem oval rate. They are also considered the most reliable because of their simplicity and the relative maturity of the systems. 2.4.2 Nanosecond Nanosecond lasers also utilize thermal material removal, but the shorter pulses help minimize thermal energy dissi pation . They still produce damage on the surface, with some materials exhibiting worse finishes than others ( e . g . nanosecond machining of plastics results in minimal surface damage ) [4] . This technology is also relatively more mature than the ultrashort systems and it requires fewer components to produce the pulse; consequently, the reliability is high [28] . The cost of these systems is dependent on the type and power output, but is generally less th an ultrashort systems. 2.4.3 Picosecond Picosecond lasers dominantly utilize photolitic ablation for material removal , but pyrolitic is still present, especially in metals [5] . Higher power picosecond systems with higher repetition rates are still in their infancy , and the trustworthiness is uncertain; however, t he reliability of the technology increases every year as manufacturers are able to optimize components in the laser [40, 41] . Picosecond systems have lower power than nanosecond or CW/ millisecond/ microsecond, but the photolitic process is more efficient [35] . Nevertheless, these systems have lower material removal rates than their longer pulse counte rparts. The price of these systems varies according to the capabilities, but in general the increased complexity results in a more expensive system .

PAGE 66

66 2.4.4 Femtosecond Femtosecond lase rs are almost purely photolitic. The photolitic material removal produce s virtually no damage to the surrounding material [42] . The downfall of these lasers is the low average power output and the relatively low repetition rates [5] . Because of this , the material removal rate of these lasers is quite low. T he power and repetition rates may increase with time; however , current systems are only suitable for a handful of applications due to the high system costs and low machining rate. The cost and reliability of these systems is comparable to picosecond systems. 2.5 Existing Material Removal Models Although ultrashort pulsed lasers are a relatively new laser type , several laser machining models are currently available in journals like Applied Surface Science and Applied Phy s ics A . Three main approaches to this simulation have been recognized as : molecular dynamics, empirical models based on statistical data , and a summing strategy to account for the effect of individual pulses . 2.5.1 Molecular Dynamics Molecular dynamics (MD) is the study of the interaction between laser pulses and the workpiece at the atomic level [43] . One advantage of MD is that it offers a detailed view into the laser material interaction. Also, e xtremely short pul ses cause the ionization process to occur over a short period of time, which renders real time data capturing impracticable [35] . Additionally , the small geometric scale of micromachining presents problems because ablated mater ial and the plasma plume can shield the interaction ; t herefore, theoretical modeling can reveal information that would otherwise be unattainable [43] .

PAGE 67

67 Another advantage of MD is that the process does not require a real laser source. It is possible to test v arious theoretical lasers to determine their theoretical ablation characteristics. This is a valuable tool for laser manufacturers because it reveals the parameters that have the greatest effect on the laser material interaction. MD also does not require a real workpiece, which allows for theoretical testing of machining parameters without wasting potentially expensive test pieces. [15] Figure 2 23. Illustration of the LJ 2D model with the photons bombarding individual atoms and electrons transferring their energy into the lattice via phonons, after [43] . The complex laser material interaction requires numerous variables to be accounted for in the model; consequently, there are several approaches to modeling the interaction . Lewis et al. identifies two different models: a two dimensional model based off of the Lennard Jones (LJ) potential, and a three dimensional model based off of the

PAGE 68

68 Stillinger Weber (SW) potential [43] . While the SW model is more realistic, it also requires a significant increase in computational power. This drove Lewis et al. to utilize the LJ model, and an illustration of this model is shown in Figure 2 2 3 . Lewis et al. represents individual atoms as red circles, photons as yellow circles/lines, free electrons as black dots, and phonons as the green line. Simulations li ke th is are valuable in determining the time for ablation plumes to dissipate, and aiding in selection of machining parameters. Figure 2 24 . Multishot MD study simulating the effect of multiple pulses on silicon. The results of studies are shown for A) 1 ps pulse and 1 shot, B) 1 ps pulse and 10 shots, C) 10 fs pulse and 1 shot, and D) 10 fs pulse and 14 shots, after [15] . One of the difficulties with MD is modeling the effect of multiple pulses [15] . It is difficult to model the effect of previous pulses because ablated atoms may adhere to the surface of the workpiece. The redeposited material can shield the target atoms and cause non linear material removal . Herrmann et al. develops a 2D model that can

PAGE 69

69 analyze the effect of multiple shots by reanalyzing the position of each atom for each time step [15] . Figure 2 2 4 shows the difference between a single pulse and ten pulses for both picosec ond and femtosecond lasers. The model is successfully able to simulate redeposited material at the surface, but the usefulness of these simulations as a quantitative tool is uncertain due to the randomness inherent in redeposit. While MD is a powerful tool to determine the physics behind material removal, it is not suitable for machining simulations. The ability to accurately model every aspect of the process is a daunting task . Komashko et al. notes this saying [35] , " It is intrinsically difficult to estimate theoretically the amount of material removal resulting from laser energy absorption . " Lewis et al. also states [43] , " T he chain of events that ultimately leads to ablation is extremely comp lex . " Possibly the largest drawback of MD is the required computational investment. Herrmann et al. states their simulations took 3 days to complete, which is impractical in a manufacturing setting [15] . More powerful computer s inevitably reduce this time, but the trend of models becom ing more complex over time increases the computational load even further. 2.5.2 Empirical Models Based on Statistical Data In contrast to MD, statistical simulators model the pulse d laser micromachining process via empirical data. One goal of simulators is to determine the depth of cut for a given set of parameters, and through experimentation these values are determined within a certain range of certainty. There are two types of sta tistical analysis identified : single pulse modeling and modeling of the processing parameter s . Single pulse modeling aims to determine the effect of single or multiple pulses on the surface

PAGE 70

70 morphology in one location [29] . Para meter modeling looks at the combined effect of various machining parameters to determine the resulting depth of cut per pass for a given combination [16] . Ashkenasi et al. investigated the effect of various parameters on a sing le pulse location in sapphire [29] . Their work focused on the difference between the gentle and strong ablation phases due to the processing parameters. The work revealed that there is a significant difference between the abla tion rate of the gentle and strong phase s . Ashkenasi et al. deduced that for single shot strong ablation , the intensity must be above a certain threshold , which is in agreement with theoretical models. Ashkenasi et al. also investigated the effect of multiple pulses on the ablation rate [29] . Table 2 5 shows that after a certain number of incubation pulses the ablation phase changes from gentle to strong. While single pulses require 6 photon excitation to achieve strong abl ation in sapphire , multiple shots allow for long lived Frenkel defects to accumulate and produce avalanche ionization. This is the reason that in Table 2 6 the higher intensity pulses require fewer shots to reach the strong phase ; a higher rate of body def ects occurs for the higher intensity pulses. Table 2 6. Depth of cut comparison between gentle phase ablation and strong phase ablation. The test specimen was sapphire, wavelength 790 nm, and pulse duration 2. 3 ps [29] . Intensity Gentle Phase Ablation Rate # of Shots to Reach Strong Phase Strong Phase Ablation Rate (J/cm 2 ) (nm/shot) (nm/shot) 3.2 <10 25 350 6.4 61 10 270 12.8 71 5 480 Orazi et al. also detailed a statistics based simulation [16] . It is based on a regression model approach, where empirical coefficients are identified to describe the

PAGE 71

7 1 material removal rate. The model assumes the dominant parameters effecting material removal to be laser current, pulse frequency, and laser scanning speed. The material removal rate (MRR) is related to the processing parameters by (2 8 ) where is the laser current (related to intensity), is the pulse frequency, is the scanning speed, and through are the regression coefficients [16] . The empirical coefficients are established by machining multiple 5 mm squares and determining the material removal rat e. Twenty seven different combinations of machining parameters were tested three times each. Following the testing , an ANOVA, or analysis of variance, test was performed to determine the coefficients. The main advantage of statistical simulations is the m inimal computation al power required. A user can quickly input the processing parameters into Orazi et al.'s software and get a prediction of the depth of cut, or use Ashkenasi et al.'s data to determine the hole depth. The quick turnaround time of this met hod makes it a useful tool. With proper regression model selection it can produce accurate simulations of 2D features. The main drawback of statistical simulation is the limitation to 2D features [16] . It is difficult to predi ct the final cut geometry of 3D features because the focal spot size can change as the depth of the surface changes. Also, complex cut paths can make simulating the final geometry extremely challenging. Lastly, the empirical data is only applicable to one particular laser to machine one particular material.

PAGE 72

72 2.5.3 Summation of Individual Pulses The summation method is an extension of statistical simulation. Instead of simply modeling the ablation depth of a single pulse, it utilizes a finite element approach to sum the effects of the ablation profile at each pulse location [44, 45] . The summation method eliminates many of the shortcomings of basic statistical simulation by taking into account the position of each pulse , b ut it also introduces some complexity into the modeling process because the laser ablation profile is affected by processing parameters. A B C Figure 2 2 5 . Ablation profile for different materials. A) B orosilicate glass , B) sodalime glass , and C) silicon. The respective beam power for each material was 383 mW, 357 mW, and 53.2 mW, after [45] . Hayden et al. described an etch profile based simulation for borosilicate, sodalime, and silicon [45] . The model assumes a consistent ablation profile, one that -5 -4 -3 -2 -1 0 0 50 100 150 200 Depth (µm) Distance (microns) -5 -4 -3 -2 -1 0 0 50 100 150 200 Depth ( µm ) Distance (microns) -4 -3 -2 -1 0 1 2 0 100 200 Depth ( µm ) Distance (microns)

PAGE 73

73 accumulates over multiple passes. The measured ablation profile of each material is shown in Figure 2 2 5 . To ensure the ablation phase is not changing with the num ber of pulses, Hayden et al. ran several tests with various intensities to determine if the ablation depth per pass remained linear. For their respective testing p arameters Hayden et al. found the depth per pass remained linear. Once an ablation profile was established, each pulse's ablation profile was superimposed onto a 1 µ m X 1 µ m X Y grid. The ablation resolution was set to 0.1 µ m. Figure 2 2 6 shows the results of the simulation and measured profile . The average cut depths were predicted to be 15.1 µ m and 30.2 µ m, which is comparable to the m easured of 13.2 µ m and 28.6 µ m. Figure 2 2 6 . A comparison of the predicted and measured depth profile for a two level cut in sodalime , after [45] . Davis et al., also presented a beam profile based simulation methodology , but their technique utilized a Gaussian function to represent the ablation profile of each pulse [44] . The intensity profile of many lasers is of a Gaussian shape, and Davis et al.'s -60 -50 -40 -30 -20 -10 0 10 0 100 200 300 400 500 600 700 Etch Depth ( µm ) Distance ( µm ) Measured Profile Simulated Profile

PAGE 74

74 m odel assumes a linear correlation between intensity and ablation depth [25] . Therefore, the ablation profile is (2 9 ) where A is the maximum single pulse ablation depth, and are the coordinates of the respective pulse, and and are empirical values controlling the spread of the ablation profile [44] . A finite mesh with spacing 100 times smaller than the beam diameter was created with a matrix slightly larger than the estimated feature size. Then, each location in the mesh is analyzed to determine the respective depth of cut. The cut depth is defined by (2 10 ) A comparison of the simulated versus measured ablation profile is shown in Figure 2 2 7 . The profiles in Figure 2 2 7 ( A ) appear to agree with some degree of accuracy, but the trench profiles in Figure 2 2 7 ( B ) vary greatly. Davis et al. attributes this to red eposit, but it might also be the plasma plume laser interaction as well. Methods for reducing these factors are discussed in the next chapter.

PAGE 75

75 A B Figure 2 2 7 . Simulated versus measured depth profiles of D2 steel . A) A s ingle pulse , and B ) machined squares with v arious pulse area overlap . The laser wavelength was 532 nm, average power 0.10 W, and pulse frequency of 800 Hz, after [44] . -1.8 -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0 20 40 60 80 100 Ablation Depth (µm) Horizontal Distance (µm) Gaussian Model Experimental Data -40 -35 -30 -25 -20 -15 -10 -5 0 0 10 20 30 40 50 Depth (µm) Y Profile (µm) 50% Overlap Predicted 50% Overlap Measured 87.5% Overlap Predicted 87.5% Overlap Measured 95% Overlap Predicted 95% Overlap Measured

PAGE 76

76 2.6 Chapter Summary Although lasers have only been around since the 1960s, there are numerous methods for stimulating radiation with different lasing mediums and different excitation sources. Four different types of lasers have been discussed, but there are numerous additiona l methods for generating a laser beam. Each laser type has its own respective advantages and disadvantages, and the properties of each laser type allows users to be highly selective in the wavelength, pulse energy, beam quality, pumping efficiency, and pul se duration. The ability to generate different length pulses with different pulse generation techniques has increased the number of possible applications for these laser types. In the field ultrashort pulsed machining, the advent of picosecond and femtosec ond lasers has increased the capabilities and quality of machining ceramics, plastics, and metals. The two competing methods of laser removal by laser machining, pyrolitic and photolitic, utilize different material removal mechanisms. The thermal cycle of heating boiling vaporization for pyrolitic allows for higher power longer pulses, but it also results in a larger HAZ. Photolitic material removal, with its reliance on ionization, has the capacity to greatly reduce HAZ, but the process is not as well unde rstood as that of pyrolitic. The attempt at using molecular dynamics and empirical models based on statistical data has provided initial insight into the ionization process, but a new tool is necessary to allow for reduced computational loads and a broader range of processing parameters. The summation modeling method is a promising simulation tool . The ability to account for different beam shapes and more complex part paths may increase the accuracy of ultrashort pulsed laser simulators , but current models run into limitations

PAGE 77

77 w hen accounting for redeposit and the acceleration profile of X Y stages. A refinement of this simulation technique is explored in the following chapters.

PAGE 78

78 CHAPTER 3 LASER MACHINING SIMULATION This chapter covers the development of an empirical model for ultrashort pulsed laser machining . The model accounts for various beam shapes, pulse area overlaps , pulse delays, stage acceleration profiles, and each cut program's respective part path. The first topic discussed is the method for modeling the ablation profile. Next, the effect of acceleration and deceleration regions are discussed al ong with equations for modeling the laser position as a function of time . Then , the effects of excessive pulse overlap due to these acce leration regions are explained. Finally, a walkthrough of the simulation software is performed. 3.1 New Ablation Simulator This work considers the roles of power, attenuation , pulse width , work material, and beam shape in determining the ablation profile. A statistics based approach is applied to relate the material removal rate to intensity. Once the intensity ablation depth relationship is determined, it is combined with the beam profile to produce an ablation profile. 3.2 Ablation Profile While there ar e many process parameters in ultrashort pulsed laser machining, the simulator focus es on only a few key parameters. Laser beam shape is the first parameter of interest and is also one of the most complicated. The final beam profile at the surface of the pa rt is affected by both the laser source and the transmitting/focusing optics [46] . For the sake of simplicity, the simulator initially assumes an ideal Gaussian beam profile because the Coherent Talisker Ultra 355 4 laser system at the University of Florida (UF) theoretically

PAGE 79

79 is listed as TEM 00 with a beam quality factor , M 2 , rated as <1.3. The beam quality factor is defined as , ( 3 1) w here is the half angle beam divergence, is the focal beam radius, and is the laser wavelength [28] . Measuring t he actual beam profile of the UF Talisker laser system increases the accuracy of the simulation; however, because of the high rating of the Talisker laser the initial beam shape is approximated by an ideal Gaussian profile . In an ideal Gaussian beam, the intensity and beam radius vary according to the position in the beam shape. The equations for these two parameters are (3 2 ) and (3 3 ) In Equations 3 2 and 3 3 , I is the intensity, is the peak intensity, is the radius at which the intensity is reduced to , is the effective beam radius, r is the radial distance from the focal point, is the vertical distance from the focal point, and is the Rayleigh length [46] . The Rayleigh length is the vertical distance from the focal point where the cross sectional area is doubled. It is given by (3 4 ) where is the wavelength of the laser [46] .

PAGE 80

80 Figure 3 1 (a) demonstrates the dependence on the beam intensity with the radial distance from the beam center, . Figure 3 1 (b) shows that, as the vertical distance, , from the ideal focal point is increased, the effective beam radius relative to the focus radius, , increases as well. A B Figure 3 1. Gaussian beam intensity profiles. A) Effect of radial position on intensity , and B ) the effect of verti cal position on the beam radius, after [25] . The second para meter of interest is the peak intensity of the laser beam. An attenuator can reduce t he peak and average intensity of each pulse. This enables the user to tune the amount of energy imparted on the surface and, ultimately, the material removal rate. The rel ationship between the peak intensity and the attenuation is , (3 5 ) where att is the commanded percentage of laser power (100% at full power) allowed through the attenuator, and is the attenuated peak intensity. By combining Equations 3 2 through 3 5 , the intensity profile for an ideal Gaussian beam is

PAGE 81

81 (3 6 ) Typically is not measured directly, but is calculated by measuring the average power output. The conversion from measured power to the pulse power is (3 7 ) and the peak intensity is (3 8 ) In Equations 3 7 and 3 8 , is the measured average power, is the pulse length in seconds, and is the pulse frequency of the laser, in Hz, during the power measurement. is measured by lasing 100% power at pulse frequency over a power meter. Typically, the lase r is turned on for a set amount of time before the measurement is taken to ensure that no transient effects, due to the power meter's response time, affect the measurement. Combining Equations 3 6 , 3 7 , and 3 8 yields the final intensity equation (3 9 ) The last step is to determine the relationship between intensity and the depth of cut. Chapter 6 discusses the characterization of sapphire and includes details on additional processing parameters that affect the ablation profile.

PAGE 82

82 There are several advantages of this method of ablation profiling compared to the previous work by Orazi et al., Ashkenasi et al, and Davis et al. First, this method allows for the modelin g of beam shapes with various transverse modes and compensation of imperfect Gaussian beam profiles. Relating the depth of cut to intensity allows for the computation of any ablation profile, as long as the intensity profile is known (theoretically or expe rimentally ) . Details about the UF laser system beam shape are discussed further in Chapter 5. The second advantage of the proposed method is that the theoretical ablation profile is known at any vertical location in the beam. Previous methods produced profiles that were only valid for the vertical beam position for which they were calibrated, but the new method allows for simulation at any focal point. Lastly, this method is valid for multiple power settings . This gives the flexibility to use the simul ation at any attenuation and with any laser of similar properties. The greatest disadvantage of this method is the volume of data that must be collected to produce a statistically significant model. Every material is machined numerous times with a variety of peak intensities to determine the intensity ablation depth relationship. I f a parameter other than intensity is deemed to have a significant effect, then that parameter is tested as well. 3.3 Acceleration Profile As discussed in the previous section, t here are a number of existing techniques for modeling ultrashort pulsed laser material removal ; however, most efforts have focused only on the laser material interaction [15, 16, 29, 35, 44, 45] . The laser material interaction is important, but it is the entire laser micromachining system performance that dictates the final part quality. In addition to those issues that have already been

PAGE 83

83 identified, the stage dynamics also play an important role because they affect where the pulses interact with the workpiece. Figure 3 2 demonstrates two classical dynamic velocity profile assumptions for the stages. Figure 3 2 (a) shows the typical assumption that the stage instantly reaches the commanded velocity of 50 mm/s, travels 1 mm, and then instantly stops. Figure 3 2 (b) shows the velocity profile for a stage with an acceleration rate of 9.8 m/s 2 . These two scenarios yield different position s as a function of time. For position dependent material removal operations, such as turning, this minimally affects the finished part, but for a temporal based material removal operation, such as laser micromachining, this is an important consideration. Figure 3 2 . Comparison of acceleration profiles. A) A n infinite acceleration profile , and B) a constant acceleration profile for commanded velocity of 50 mm/s. The acceleration profile has an effect on pulse area overlap, and this in turn can affect the MRR. Davis et al. discussed the importance of pulse area overlap in laser machining. Pulse area overlap is a measure of the amount of area two sequential pulses o verlap. A visualization of this parameter is shown in Figure 3 3 and t he equation for calculating the percent overlap is

PAGE 84

84 (3 10 ) w here is the commanded feedr ate , in mm/s, of the laser beam as it scans over the workpiece surface, is the pulse frequency in Hz, and is the beam radius on the workpiece surface in mm. The above equation shows that, for a constant velocity/frequency ratio, the pulse area overlap is a constant. Consequently, when repetition rates were lower (prior to improvements in laser capabilities), the commanded velocity was also typically low. This is the primary rea soning for this author's hypothesis that the transition from sub kHz to >100 kHz repetition rates result s in the velocity profile playing a more significant role in the machining process; increased pulse repetition rates result in increased commanded veloc it ies for a given pulse overlap . Figure 3 3 . Visualization of the pulse area overlap. Certain combinations of pulse frequency and feedrate result in pulse overlap. The effect of the increased pulse repetition rate on the mach ining process is demonstrated in Table 3 1 . It shows that for a low repetition rate and low commanded velocity the infinite acceleration assumption results in the same number of pulses as the finite acceleration. However, as the repetition rate and frequen cy are increased, the finite acceleration pulse count becomes much greater than the assumed infinite

PAGE 85

85 acceleration pulse count. This results in a discrepancy between the expected depth of cut and the realized depth of cut. Table 3 1 . The resulting number of pulses impacting a workpiece for different repetition rates. Commanded velocity (mm/s) and repetition rate (kHz) = 0.1 f = 0. 1 = 5 f = 5 = 50 f = 50 = 200 f = 200 Move l ength (mm) 1 1 1 1 Time to complete (infinite acceleration) 10 0.2 0 0.02 0.005 Full pulses (infinite acceleration) 1000 1000 1000 1000 Time to complete (a = 9.8 m/s 2 ) 10 0.20 0.0 3 0.02 Full pulses (a = 9.8 m/s 2 ) 1000 100 0 1 300 4 000 Figure 3 4 . Demonstrations of acceleration region effects on the ablation profile of silicon . Start stop locations (circled in red) become evident in from the secondary left right moves, after [45] . Examples of the stage acceleration effects are visible in several of Hayden et al.'s test cuts [45] . Figure 3 4 shows the profile of a multi level cut in silicon where r aster scanning in two directions was performed (left right and into out of the figure). R a ster scanning left and right prod uced deeper trenches at the start stop locations because of the acceleration/deceleration effects. -75 -65 -55 -45 -35 -25 -15 -5 5 0 100 200 300 400 500 600 700 Etch Depth (µm) Distance (µm) Measured Profile Simulated Profile

PAGE 86

86 Orazi et al., Ashkenasi et al, and Davis et al.'s models provide insight into the laser machining process [16, 29 , 44] . However, with higher repetition rate lasers the need to model the stage dynamics becomes more evident, and increased accuracy of models are realized with the inclusion of this aspect. 3.4 Acceleration Simulation There are several advantages to including stage dynamics into a simulation. From a simulation aspect, the ability to calculate the depth of cut in regions affected by acceleration and deceleration is vital. It also allows for identification of locations of excessive pulse area overlap where excessive thermal energy is imparted. For ceramics and thermally sensitive materials , this is necessary to minimize thermal shock in the workpiece due to the machining process [3] . Additional ly, it allows for later compensation of these excessive ablation regions to produce a more uniform ablation depth . Lastly, from an administrative and planning perspective it helps determine the actual time required to machine a component. This is helpful i n not only planning of the manufacturing process , but also in pricing out machining operations. 3.4.1 Acceleration Profiles One of the most difficult aspects of modeling the acceleration/deceleration regions is determining the acceleration profile because d ifferent types of stages and stage controllers produce different profiles. Some controller stage combinations are designed to produce acceleration profiles similar to those in classical dynamics (e.g. the UF linear X Y stages) , while others may exhibit a mass spring damper type acceleration profile (e.g. the UF galvanometer scanner) . Additionally , when one account s for the maximum torque/velocity of the stage motors it increases the complexity even further.

PAGE 87

87 Figure 3 5. Typical acceleration profiles and velocity profiles as a function of time. A) A constant acceleration system, B) constant jerk system, and C) sinusoidal acceleration system. For this research the profile is assumed to be either constant acceleration, const ant jerk (with max acceleration), or sinusoidal acceleration. Figure 3 5 shows the

PAGE 88

88 acceleration and velocity profiles for these acceleration types. These profiles are modeled because the Aerotech Pro115 X Y linear stages are controllable in the Cimita software [47] . This allow s for comprehensive testing of each profile type to establish the validity of the models. The acceleration profile and rate of the galvo is not c ontrollable, but it is assumed that the dynamics of the galvo can be approximated by one of the above three profiles. Chapter 5 goes into more detail about fitting the galvo profile to one of the classical profiles. Figure 3 6 . The above graphs show four different acceleration profiles for a constant jerk scenario. A) Represents a profile that never reaches the max acceleration rate or commanded velocity, B) reaches the commanded velocity, C) reaches the max acceleration rate, and D) reaches both the commanded velocity and max acceleration rate.

PAGE 89

89 Calculating the time required to machine a feature is not a trivial task. While the equations of motion for these acceleration types are not incredibly complex, accounting for the comman ded velocity , move length , and max acceleration is challenging . When command delays are included, as discussed in Chapter 5, it becomes even more complex. An example of this is demonstrated in the constant jerk type that can have four different acce leratio n profiles , such as the profiles shown in Figure 3 6 . A method for determining which profile is applicable to each move length, acceleration rate, and commanded velocity combination is discussed below. 3.4.1.1 Constant Acceleration A constant acceleration profile is mathematically the simplest of the three to characterize. The acceleration, velocity, and position as a function of time are (3 1 1 ) (3 1 2 ) and (3 1 3 ) where is the acceleration rate of the stage, the original velocity, and the original position [48] . Due to a constant acceleration rate, the only variable that needs to be analyzed is whether or not the commanded velocity is reached before the middle of the move. This is accomplished by comparing the distance to reach the commande d velocity with the move length (3 1 4 ) (3 1 5 )

PAGE 90

90 and (3 1 6 ) The commanded velocity is not reached i f Equation 3 1 6 is valid . C omputation of the move time is simplified by only analyzing the first half of the move. This is an acceptable simplification because the acceleration and velocity plots are symmetric about (3 1 7 ) In the case of the stage never reaching the commanded velocity , this is a rather simple calculation because the stage is always accelerating. The move time can then be calculated by substituting and into E quation 3 1 3 and solving for (3 1 8 ) and (3 1 9 ) The assumption that is valid for the UF simulator because the stage comes to a stop before the start of the next move. Also, assuming incremental movements instead of absolute movements simplifies the calculation by eliminating and . For the case where the commanded velocity is reached, there be comes two move segments: one where and one where . The time to complete the first segment is computed by plugging in into Eq uation 3 1 2 , (3 20 ) and (3 2 1 )

PAGE 91

91 Next, the time to complete the rest of the move is determined by calculating the distance left to travel after the acceleration segments, (3 2 2 ) and solving for the constant velocity time, (3 2 3 ) Finally, the time to complete the move is given by, (3 2 4 ) 3.4.1.2 Constant Jerk The constant jerk scenario is a more complex computation. Instead of just analyzing the commanded velocity, the max acceleration of the system must also be accounted for in the analysis . Figure 3 7 shows seven region s in the acceleration profile, of that only regions one, three, five, and seven are required. Time Figure 3 7 . The acceleration profile of a constant jerk system with a maximum acceleration rate. There are seven possible zone, but only zones 1, 3, 5, and 7 are found in every constant jerk situation. Acceleration

PAGE 92

92 The equations of motion for a jerk system without a max acceleration rate is (3 2 5 ) (3 2 6 ) and (3 2 7 ) w here is the jerk rate [48] . The above acceleration rate is valid for region one, but the other regions are governed by different rates, as shown in Table 3 2 . Table 3 2 . The acceleration rates of each region in a jerk system. represents a system where both the max acceleration and commanded velocity are reached. reaches the max acceleration, reaches the max acceleration, and does not reach either. Region 1 2 3 4 5 6 7 The process of determining what regions are active in a particular move requires analysis of the time before the max acceleration, commanded velocity, and commanded position are reached . The first step in this analysis is determining if the max acceleratio n rate is reached. If either of the conditions in E quations 3 2 8 or 3 2 9 is not met, then the max acceleration rate is not realized , (3 2 8 ) and (3 2 9 ) If it is not realized, then the next step is to determine if the feedrate is reached. Equation 3 30 gives the distance required to reach the feedrate. If the feed distance is

PAGE 93

93 less than the move length, then the commanded feedrate is reached. The times for each region are listed in Table 3 3 , (3 30 ) Table 3 3 . Regional times for a jerk type acceleration profile that does not reach the maximum acceleration. Region Feedrate Reached Feedrate Not Reached 0 0 0 For the case of reaching the max imum acceleration, time is used to determine the max feedrate. The time to compete the move is (3 3 1 ) and the time to reach the commanded feedrate is (3 3 2 ) The commanded feedrate is not reached i f . Times to complete each region are listed in Table 3 4 . Finally, the total time to complete the move is given by, (3 33)

PAGE 94

94 Table 3 4 . Regional times for a jerk type acceleration profile that reaches the maximum acceleration. Region Feedrate Reached Feedrate Not Reached 0 3.4.1.3 Sinusoidal Acceleration A sinusoidal acceleration profile is relatively simple compared to the jerk profile. Figure 3 8 shows that there are only three possible regions in this profile. One caveat with this profile type is that the UF system assumes the commanded maximum accelera tion rate is the mean rate of the sinusoidal profile. This has been accounted for in the following equations. For other systems without this assumption, a correction to the max acceleration rate would need to be performed. The equations of motion for this system are [48] (3 3 4 ) (3 3 5 ) and (3 3 6 )

PAGE 95

95 Time Time A B Figure 3 8 . Demonstration of the two profiles of a sinusoidal acceleration. A ) A situation where the commanded velocity is reached , and B) when the commanded velocity is not reached. For this acceleration type, the two variables compared to see which regions are active are the time to reach the commanded velocity, (3 3 7 ) and the time to reach the commanded position, (3 3 8 ) The regional move time equations are listed bel o w in Table 3 5 , and t he time to complete the move is (3 3 9 ) Table 3 5 . Time required for each region for a sinusoidal acceleration profile. Region (Commanded feedrate not reached) (Commanded feedrate reached) 0 Velocity Velocity

PAGE 96

96 3.4.2 Excessive Overlap Regions Excessive pulsing can occur in not only the start/stop locations, but also in small radius moves . This problem was discovered during early attempts at machining R plane sapphire H bar structures. Figure 3 9 shows the cracks in several places on one of these early attempts. It was later realized that the high laser velocity, coupled with small feature radii , required an excessively large force to be exerted by the stages to maintain the feedrate . This centripetal acceleration force is given by (3 40 ) where is the laser velocity, and is the feature radius [48] . For a scanning speed of 2 , or roughly 80 times greater than earth's gravitational acceleration. This is an exceedingly high value ; therefore, this aspect has been included into the simulation. Figure 3 9 . Visible cracks in a sapphire machined component that are likely cause by centripetal acceleration problems. The picture on the right shows the half H bar structure with several visible cracks outlined. The blown up views on the left show a greater detail image of the cracks that appear to propagate from radius sections.

PAGE 97

97 There are two possible results from a move with insufficient centripetal acceleration. Either the feedrate is reduced or the radius increased, which causes geometric irregularities . Examination of the encoder signals from the X Y stages on the UF system revealed that the controller simply limited the maximum feedrate . Therefore in the UF simulator, the maximum feedrate is calculated for each circular move and the commanded feedrate reduce if it exceeds the capabilities of the stage. 3.4.3 Command Delays Command delays are the last identified situation where excessive pulsing can occur. Beam on and beam off delays were discovered during single pulse ablation profile characterization . During the tests the beam was commanded to turn on, lase long enough for a single pulse to impact the workpiece, and then commanded to turn off. Initial tests revealed that multiple pulses were impacting the surface, even when the delay between beam comman ds was reduced to zero. It was determined that during the beam on and beam off commands there is a slight delay between the command being sent to the laser and the laser responding. Because of this, the laser controller incorporate s a slight delay to ensur e the beam is in the correct state before it attempts the next command. This delay cause s lasing while the stages are stationary ; resulting in excessive pulses at beam on or beam off command locations. These two delays and several others are explored furth er in Chapter 5. 3.5 Process Flow T he ultrashort pulsed laser micro machining simulator is a complex program that includes all of the above mentioned properties and nuances of laser machines . The simulator runs in Matlab ® , and a graphic al user interface, or GUI , allows the user to load the cut program, input process parameters, input laser station settings, run the

PAGE 98

98 simulation, and inspect the simulation results all from one screen . The process flow for the simulator is outlined in Figure 3 1 0 , and a screenshot of the GUI is shown in Figure 3 1 1 . Figure 3 1 0 . Operation flowchart for the UF ultrashort pulsed laser ablation simulator.

PAGE 99

99 Figure 3 1 1 . Screenshot of the UF picosecond pulsed laser ablation simulator GUI. Once th e simulation program routine is complete, the results table and currently selected graph are updated. A demonstration of the outputs are shown in Figure s 3 1 2 through 3 1 5 . The wor d 'GATORS' was tested. Figure 3 1 2 shows the 2D depth of cut graph. Several red spots are visible on the graph that indicate excessive ablation has occurred in those regions. Figure 3 16 shows the simulated results for the nine parameters .

PAGE 100

100 Figure 3 1 2 . 2D depth of cut graph for the 'GATOR' cut program. Figure 3 13. 3D depth of cut plot for the 'GATOR' cut program.

PAGE 101

101 Figure 3 14. Plot of the velocity as a function of position for the 'GATOR' cut program. (c) Figure 3 15. Calculated machining time as a function of commanded velocity for the 'GATOR' cut program.

PAGE 102

102 Infinite Acceleration Acceleration Compensated Time To Machine (s) 8.65E+00 2.64E+01 Laser On Time (s) 5.94E+00 8.19E+00 Energy Consumed (kWh) 8.44E 03 1.97E 02 Cost of Cut ($) 1.20E 01 3.67E 01 Pulse Area Overlap (%) 85.9 M ax Realizable Feed 46 Max Depth of Cut (um) 9.41E+01 Avg Depth of Cut (um) 2.37E+01 Min Depth of Cut (um) 1.36E+01 Figure 3 16. Simulation results for the 'GATOR' cut program. 3.6 Chapter Summary The details for a comprehensive ultrashort pulsed laser machining simulator have been discussed in this chapter. An ideal TEM 00 Gaussian beam profile is currently being utilized due to the widespread use of this pr ofile; however, the program also allows for other beam shapes to be used in the profile. This capability aid s in increasing the accuracy of the simulations. Several different acceleration profiles from classical dynamics have been discussed, along with t heir respective governing equations. It was shown that these profiles can take on several different shapes depending on the commanded velocity and move length, as well as the jerk/acceleration properties of the respective stages. Three different scenarios h ave been identified that can cause excessive pulsing on the workpiece. The acceleration/deceleration regions of a move are inevitable, but with the proper parameters selection, most specifically the pulse frequency and

PAGE 103

103 commanded feedrate, the effect of t hese regions is minimized . The second scenario is small radius circular moves where the necessary centripetal acceleration may be greater than the capabilities of the stages or stage controllers. Once again, the possibility of incorrect geometries or exces sive pulsing due to these small circular moves are mitigated by the proper commanded feedrate selection. The last scenario is not affected by the feedrate, but rather the response time of the laser system to turn on or turn off the laser. The delay caused by these beam commands result s in excessive ablation at the beam on/beam off locations ; therefore, the location of the on/off locations is an important aspect . Next , the UF ultrashort pulsed machining simulation program was discussed. The process flow of p reparing and running a simulation was discussed, along with a demonstration of the graphical user interface of the program. Finally, the various outputs of the program were revealed. In the following chapter the part path correction and automated process parameter selection routine is discussed. There is a comparison of non corrected part paths to corrected part paths, along with a demonstration of the UF correction program.

PAGE 104

104 CHAPTER 4 LASER MACHINING PART PATH CORRECTION AND AUTOMATED PROCESS PARAMETER SELECTION In this chapter the strategies and steps for part path correction and automated process parameter selection for ultrashort pulsed laser operation are discussed. First, the two recognized causes of excessive pulse overlap are explored. Next, the e ffects of excessive pulse overlap are examined. Then, the proposed part path correction steps and strategies are reviewed, along with simulations of these correction steps . Following that , three methods for automating the process parameter selection are discussed. Finally, the process flow for the correction/automation program is reviewed . 4.1 Introduction One of the most difficult challenges in utilizing ultrashort pulsed laser s is selecti ng the best combination of machining parameters. Machine operator s may not always select the most favorable input parameters or strategies, which can cause excessive ablation and/ or excessive thermal load . Additionally, if the operator is too conservative with their respective selection the time to machine may be unnecessarily long . This inevitably increase s the cost of the final product, and hamper the acceptance of this technology into the manufacturing field. Incorporating a correction/automation feature into the simulation software alleviate s many of the current problems by providing a qua ntitat ive solution to parameter selection. Although a simulator is the first step in qua ntit atively selecting parameters, it still requires a vast amount of experimenta tion to discover a successful combination of inputs . An automated process parameter selection routine exam ine s key aspects of the machining process and quickly determine s the best set of parameters. The two main aspects examined are excessive pulse area overlap, and excessive ablation . These

PAGE 105

105 aspects are i mportant because they can damage the workpiece through thermal shock and/or failure of the part by geometric inaccuracies . The various user defined inputs ( e . g . pulse frequency, pulse power, feedrate, an d number of passes) are selected to minimize these effects. Also, the location and number of additional passes to account for excessive ablation is determined from the same analysis. 4.2 Excessive Pulse Overlap Excessive pulse overlap is the result of two scenarios. The first scenario , acceleration zones, stems from the combination of low acceleration rates and high feedrates . Because the acceleration rate of an X Y stage or galvo is limited, the best alternative is to modify the feedrate. Selection of a pr oper feedrate is a two step process. First, the part path is analyzed for radial moves, and the smallest radius , , is used to calculate a maximum feedrate, . (4 1) Next, a maximum acceleration length to reach the feedrate , , is specified, and the corresponding maximum feedrate , , is calculated. This calculation varies depending on the acceleration profile of the stage. Equation 4 2 is used with constant acceleration systems, and Eq. 4 3 for sinusoi dal systems , , (4 2) and . (4 3) The max feedrate for the jerk system is more complicated because it is depend ent on whether the maximum acceleration rate is reached. If Eq uation 4 4 is

PAGE 106

106 valid, then the max acceleration is not reached and Eq uation 4 5 is used to calculate . Otherwise Eq uation 4 6 is used to calculate . , (4 4) , (4 5) and . (4 6) Lastly, the lesser of and is selected as the feedrate. In the second scenario, the operator can select a pulse frequency and feedrate combination that results in excessive overlap at every point in the cut. Modifyin g the pulse frequency , feedrate , or beam radius eliminates the excessive overlap ; however, it requires an understanding of each material's respective optimal pulse overlap to make the correction . While both the feedrate and beam radius could be used to alt er the overlap, the frequency is selected instead. This is due to the optimal feedrate being defined in the previous scenario , and an assumption that the beam radius is constant due to the tendency to focus at the focal point . Unfortunately , there is no explicit solution to solve for in the pulse area overlap equation (Equation 3 9) ; however, an iterative solution is possible given a range of valid feedrates. 4.3 Ablation Depth Control of the f inal cut depth of a feature is an important capability . W ithout a simulator it is difficult to identify regions with undesired cut depth s . This is due to four factors: stage kinetics causing excessive ablation, possible overlap in the cut path, uncertainty in the pulse ablation depth, and uncertaint y in the number of necessary passes . These four factors can combine to cause a complex part floor profile , and a part

PAGE 107

107 path correction tool is necessary to analyze and compensate for all four factors. Excessive overlap is minimized through the parameter mod ifications listed above, but for most scenarios there is still some type of a non linear ablation at the beginning and end of the move , as demonstrated in Figure 4 1 . A part path correction scheme is integrated into the UF simulator to account for these re gions of undesired cut depths. A B C D Figure 4 direction motion. A) Represents a simulated cut with a pulse frequency of 1 kHz, B ) 5 kHz, C) 50 kHz, and D) 200 kHz. The stage is considered to have a constant acceleration profile with an acceleration 9.8 m/s 2 . The diameter of p ulse area overlap is 86% , and average measured power of 4 wat ts .

PAGE 108

108 The first step in the ablation depth control process is to select a pulse power. This value is dependent on not only the material being machined, but also the desired quality of the feature. H igher power results in higher uncertainty in the ablation p rocess due to an increase in avalanche ionization. Additionally, the greater pulse power typically results in a greater amount of redeposit due to a higher MRR and a larger amount of thermal absorption into the workpiece . Quality selection by the user is i ncluded to allow the user to adjust the allowable power due to time (higher power) or quality (near ablation threshold) constraints. The quality selection also affects the feedrate and pulse frequency. This gives the operator a greater flexibility in machi ning parts for different applications. Table 4 1 . Explanation of the different cut types available in the modification routine. Cut Type Explanation of Cut Features Through Through cutting is used to produce a hole through the entire substrate with the desired feature profile. Through cuts disregard the variations in the cut floor and instead focus on removing enough material to produce a through hole pattern that most closel y matches the feature profile. The desired depth of cut is considered a minimum depth constraint. Face Face milling is used to thin down a workpiece for further processing. Face milling disregards the acceleration regions at the beginning and end of the c ut moves because those regions are typically beyond the wafer edges. This feature attempts to minimize the difference between the desired and machined depth on the actual substrate. The desired depth of cut is considered an average depth constraint. Pocke t Pocket milling is used to produce a feature in the workpiece that is at a given depth. This strategy analyzes the depth due to the acceleration regions. The desired depth of cut is considered a maximum constraint. This type of cut typically requires addi tional passes to fill in the non acceleration regions to the desired depth.

PAGE 109

109 Once the pulse power is selected , the operator selects a depth of cut and decides if the cut type is a through, face, or pocket feature. Selection of the cut type affects the interpretation of the desired cut depth value as either a maximum, minimum, or average value. Table 4 1 provides explanations of the parameter modification strategy for each cut typ e , and Figure 4 2 provides a visual . It is important to note that the simul ation/modification program analyzes the material type and the cut feature to determine the proper parameters for each cut type. Consequently, a custom set of parameters is produced for each part path material laser power output combination. Figure 4 2 . Visualization of the different cut types. A) Original unmodified wafer, B ) Demonstration of a through cut with 4 holes , C ) demonstration of a face cut in the center of the wafer , and D ) demonstration of a pocket cut with 4 circles . 4.3.1 Part Path Correction Steps The correction process involves two steps. First , the unaltered part path is simulated with the newly selected pulse power, feedrate, and pulse frequency. Next, the number of passes are modified until the desired cut dep th constraint is achieved. Once

PAGE 110

110 the baseline cut profile is established, additional cuts are added to the primary cut path. Figure 4 3 demonstrates the process of adding passes via a simulated line cut . As more passes are added , the depth of cut becomes more uniform. Further depth uniformity is realized with a smaller spacing between the additional passes, as seen in the simulation results in Figure 4 4 . A B C D Figure 4 3 . Demonstration of the path ad dition process. A) Depth of cut profile for the primary pass, B) primary pass and 1 secondary pass, C) primary pass and 2 secondary passes, and D) primary pass and 6 secondary passes. The pulse spacing is 8 µ m. The n on uniform depth profile is slowly impro ved through the addition of more passes. The machining parameters used are the same ones to those used in Figure 4 1.

PAGE 111

111 Figure 4 4 of the cut profile. The machining parameters are th e same as the ones to those used in Figure 4 1. 4.3. 2 Added Passes Geometry Figure 4 5 demonstrates the linear moves involved in adding passes. This technique was originally selected because the part path of the added passes remains on the original cut p ath. Following the original part path is an attractive option because it minimizes the effect of one added pass on another cut region; effectively insulating other cut regions from the added move. However, the strategy was found to be ineffective because o f the galvo undershoot issue that is discusses in section 5.1.2.2. The added passes are so short that the galvo effectively disregards them. Figure 4 6 shows a 3D plot from a scanning white light interferometry (SWLI) measurement of one cut in R plane sap phire that utilized the straight line added passes, and it is evident that the cut depth is not consistent across the entire feature floor. The SWLI used in this research is a Zygo Newview 7200 with a 20X optic . The settings for this measurement are a 1X zoom tube , scan mode, 3 repeated measurements are used to determine the average measured depth at each data point in the image, and a scanning depth of 65 µm .

PAGE 112

112 Figure 4 5 . Visualization of a linear added passes r outine. Additional moves are added in zone 1, 3, 6, and 7. The other sections represent regions that do not require additional passes (cut depth criteria is obtained). Figure 4 6 . Demonstration of a feature machined in R plane sapphire with added linear passes. Data from a SWLI measurement is used in producing the 3D plot. The cut floor is not uniform due to the galvo skipping over the short added passes (5 µm added passes spacing ). The desired depth of cut was 50 µm. The fluence for this measurement was 2.35 J/cm 2 , feedrate of 37.5 mm/s, and pulse frequency of 50 kHz.

PAGE 113

113 A solution to the short move skipping is to increase the added passes spacing and added passes moves. The problem with this is the roughness of the feature floor increases and the precision of the added passes decreases , as seen in Figure 4 3 and Figure 4 4 . A workaround to this dilemma is to utilize a non linear added passes move, as demonstrated in Figure 4 7 . The short back and forth moves are replaced by a single circular move. This has several advantages, including the ability to compensate for non perpendicular sidewalls, and the flexibility to add a single circular move in single increments, instead of increments of two for the linear added passes. Figure 4 7 . Visualization of a circular added passes routine. Additional moves are added in zone 1, 3, 6, and 7. The other sections represent regions that do not require additional passes (cut depth criteria is obtained). 4.3. 3 Strategies for Automating the Process Parameter Selection The simulation/modification software is capable of modifying a given feature in several different ways to incorporate the a bove routines, and t hree automation strategies have been identified. The first strategy corrects the e n tire program and assigns the feedrate, pulse power, and pulse frequency according to the most

PAGE 114

114 troublesome areas. This is the simplest method to implement ; however, it also results in the longest cut time because the conservative parameters affect the entire cut. The next technique is to reduce the feedrate and power in select regions. Areas of small radii and short moves utilize relatively low power and f eedrates as compared to the rest of the program. Unfortunately, the pulse frequency is not a parameter that the UF laser system is capable of modifying on the fly ; t herefore, the pulse overlap is increased in these select regions. Reducing the pulse power can minimize the effect of the increased overlap, but a larger thermal load may still be realized due to the tighter spacing of the pulses. This method reduces the cut time of the part, but may introduce unacceptable thermal loads. The last strategy is to split the feature into two separate cut programs. This allows for the reduction of the pulse frequency in the modified regions, thereby producing a relatively constant pulse overlap. The realized gains of this method over the first two methods depends on t he time it takes the user to switch pulse frequencies; therefore, small feature cuts are likely not an appropriate application for this automation strategy. The first and third strategies are the two methods integrated into the UF simulator program. The second strategy is not utilized because there is a delay associated with altering the beam attenuator. This delay of only a couple milliseconds is high enough to cause unacceptable excessive lasing; therefore, the second strategy is not as effective at improving the feature quality. 4.4 Process Flow The techniques and processes to determine the most appropriate set of parameters for each respective cut are laid out in the previous sections. To demonstrate

PAGE 115

115 the capabilities of the modification routines , a 2 X 3 pocket is virtually machined in R plane sapphire . The desired depth of cut is 3 0 µ m , and a cons tant jerk rate of 1.35 e11 m/s 3 is assumed for this simulation . The simulated processing parameters are 1.93 J/cm 2 pulse fluence, 50 kHz pulse repetition rate, and 50 mm/s commanded feedrate. A tolerance of +/ 10% of the depth is selected, and the modification type mixed Figure 4 8. A simulated cut in R Plane sapphire without any parameter or path modification. Figure 4 8 show s the first attempt at machining the square without the aid of the modification routine . The results of the simulation, listed in Table 4 2 reveal that the start/stop and b eam on/beam off locations produce trenches that extend farther into the work piece than the main cut area (27.0 µm compared to 8.93 µm) . Additionally, the average depth is less than a third of the desired depth. A number of additional trials would be neces sary to find the optimal number of passes if a correction/automation

PAGE 116

116 routine was not utilized . The machining parameters for these simulations are listed in Table 4 3 Table 4 2. Cut results for the three simulations, one without any additional processing, o ne with automated parameter selection, and lastly one with modified parameters, and part path corrected. Modification Type Time to Machine Laser On Time Maximum Depth Average Depth Minimum Depth (s) (s) (µm) (µm) (µm) None 2.03 1.96 27 .0 8.93 6.94 Process Parameters 5.34 5.21 38 .0 29.5 17.8 Process Parameters and Corrected Part Path 7.5 7.23 36.8 31.2 17.0 Figure 4 9. Simulation results with the new modified machining parameters. Next, in Figure 4 9 the input parameters are modified and the required number of passes calculated. Table 4 2 shows the new parameters with a reduction of the feedrate and an increase in the pulse power. The reduced feedrate increases the pulse area

PAGE 117

117 overla p, but it reduces the difference in depth at the start/stop locations relative to the average cut depth. The increase in pulse power results in a fluence that is above the ablation threshold for a single pass. This reduces the amount of thermal energy that is imparted into the workpiece, and increases the consistency in the depth of cut from one pass to the next. A full automated parameter selection with part path correction is shown in Figure 4 10. The maximum depth of cut is reduced compared to the parame ter only modification, but Figure 4 8 shows a rougher cut finish when compared to 4 8 C. This is due to the relatively small acceleration zones; smaller zones results in a reduction in the start/stop excessive lasing. Therefore, added passes are likely to be more useful on complex geometries because the number of passes required to reach a given depth may vary in different cut sections. A demonstration of the effectiveness of added passes is explored in Chapter 6.

PAGE 118

118 Figure 4 10 . Simulation results for a cu t with modified cut parameters and part paths. Table 4 3. Input parameters for the simulation modification comparison. Modification Type Main Passes Feedrate Pulse Frequency Pulse Fluence (mm/s) (kHz) ( J/cm 2 ) None 5 50 50 1.9 Process Parameters 10 37.5 50 2.3 Process Parameters and Corrected Part Path 21 87.5 50 2.3 4.5 Chapter Summary Excessive pulse overlap can cause damage due to thermal shock and/or excessive material removed. Two scenarios have been identified that may produce excessive pulse overlap; acceleration zones and incorrect process parameter selection. Equations were presented to calculate the maximum feed rate for small radius moves and also for an acceleration length . Then, the difference in cut profiles for various pulse

PAGE 119

119 frequencies was analyzed, and it was found that higher pulse frequencies led to reduced cut floor uniformity. After it was realized that the cut floor uniformity is reduced with higher pulse frequencies, a strategy for including additional passes t o account for the non uniform depth of cut was ex plored. The analysis showed that as the number of additional passes increased and the spacing between the passes decreased, the uniformity of the cut floor increased. Then , three strategies for modifying the processing parameters was discussed. Finally, a comparison of results for simulations of non modified, parameter modified, and parameter modified plus additional passes cuts was performed. The results show that the processing param eter modification was c apable of altering the simulated cut depth to the desired depth . Additionally, the additional passes improved the uniformity of the cut floor. Marked improvements is realized with the addition of an automation/correction routine. It enables quick and efficient determination of machining parameters, and reduces the time investment for planning a cut operation. The shortcoming of this technique is the necessary addition al information about the workpiece, specifically the thermal properties and thermal shock characteristics; however, these values only need to be established once for each material. Once a library of acceptable acceleration lengths is established, the routi nes require nothing further than the laser machine station's properties. The automated processing parameter selection and part path correction routines show the potential to increase the viability of the ultrashort pulsed laser as an industrial tool.

PAGE 120

120 The f ollowing chapter discuss es the characterization process for the Oxford J 355PS laser micromachining station. It include s details on characterizing the X Y linear and galvo stages, quantifying command delays, determining the beam profile, characterizing the attenuator, and quantifying the lasing time for beam on and beam off commands.

PAGE 121

121 CHAPTER 5 OXFORD J 355PS CHARACTERIZATION In this chapter , the strategies and steps for characterizing the Oxford J 355PS laser micromachining station are discussed. First, the acceleration profile and rate of the X Y linear stage and galvo are explored. Next, the tool and techniques for determining the beam diam eter are discussed. Then, the characteristics of a diffractive attenuator are explored along with a calibration technique. Finally, the amount of time the system lases during beam on and beam off commands is quantified. 5.1 Introduction As discussed in Cha pter 2, c haracterization of laser machining stations is quite similar to the process of characterizing CNC milling machines. Each brand and model of CNC mill has a unique combination of motion controllers, motors, ball screws, spindle geometry , spindle bea rings , etc . Operators must account for all of these parameters while characterizing the CNC mill. Although some components have minimal effects on the optimal milling parameters, others such as the spindle and spindle bearings have a significant role in determining the optimal parameters. This is true with laser micromachining stations as well; some parameters have significant roles in determining optimal parameters and the effect of others is negligible. The characterization of laser machining stations r equires the operator to consider the focal beam radius , pulse fluence, pulse repetition rate, stage dynamics, motion controllers, and attenuator. All of these parameters are incorporated into the simulator software. In the previous chapters the methods for simulating and modifying ultrashort pulsed lasers are discussed, and in this chapter these methods are applied to the UF laser system. It is important to note that the simulation software is designed such that

PAGE 122

122 virtually every ultrashort pulsed laser machi ning station can be characterized and the software modified for that particular machining station. The UF system is characterized in this chapter to demonstrate the techniques and required experiments to characterize a machining station. The University of Florida operates an Oxford J 355PS laser micromachining station, shown in Figure 5 1, that utilizes a Coherent Talisker Ultra laser head to produce picosecond pulses . The specifications of the laser head were discussed in section 2.2, and a review of the s pecifications are listed below in Table 5 1. Figure 5 1. Oxford J 355PS laser micromachining station installed at the University of Florida. Daniel Blood. 2010. Gainesville. Table 5 1. Oxford J 355PS manufacturer specifications [47] . Nominal Pulse Duration Wavelength Theoretical Beam Diameter Max Rated Output Beam Attenuator Pulse Repetition Rate 10 15 ps 355 nm 9 µm 4 W 0 100% up to 200 kHz

PAGE 123

123 5. 2 Linear X Y and Galvo Stage Characterization The Oxford laser micromachining station has two methods of controlling the X Y location of the laser beam : linear servo driven ballscrew Aerotech PRO115 stages and an Aerotech galvo [49] . Figure 5 2 shows the location of each of these stages. The need for two separate control systems arises from the pros and cons of each system. The linear stages have the advantage of a relatively long travel range, 200 mm, and 0.5 µm resolution [47] . The limitation of these stages is their relativ ely high inertia when compared to the servo power rating. These stages are limited to 50 mm/s feedrates in part because the acceleration distance is excessively long for higher speeds. Additionally, these stages are not advantageous for high speed moves be cause the workpiece is accelerated along with the stage. This requires the workpiece to be fixtured to the stage, and can cause deflections in the workpiece due to the stage acceleration/deceleration. Due to all of these reasons the linear stages are prima rily used to transition the workpiece between galvo cuts. Figure 5 2. Location of the three PRO115 stages (X, Y, and Z) and the galvo . Note that the galvo is enclosed in an aluminum housing, and the arrow simply points at the l ocation of the housing. Daniel Blood. 2012. Gainesville.

PAGE 124

124 Galvo s are unlike the linear stages because they utilize two mirrors to move the laser beam instead of the actual workpiece. The UF galvo setup, as seen in Figure 5 3, has a lower inertia to motor p ower ratio that allows for much faster acceleration rates. Additionally, a small mirror angle change results in a large linear movement of the laser beam on the workpiece. Lastly, the galvo has a listed resolution of 0.3 µm, an improvement over the linear stages [47] . Figure 5 3. Visualization of the path of a laser beam through a galvo . Small changes in the angle of the mirrors alters the position of the laser on the workpiece . Daniel Blood. 2014. Gainesville. 5. 2 .1 Acceleration Rate Measurement The technique for characterizing the acceleration profile and rate of the two stage , A3200 , allows the operator to record the encoder signals from the servo motors of the linea r X Y stages . The encoder signals

PAGE 125

125 are then analyzed by the A3200 software, and displacement and velocity data is generated. The displacement , , is given by , ( 5 1) and the velocity , , is given by . ( 5 2 ) In the above equations is the number of encoder pulses recorded, is the ball screw pitch, is the number of pulses per revolution of the encoder, is the multiplication factor of the MXH multiplier, is the controller multiplication factor, and is the elapsed time between the initial and final measured pulse [50] . The displacement resolution, , is given by . ( 5 3 ) Table 5 2, and the resulting displacement resolution is 0.5 µm. Table 5 2. Aerotech PRO115 linear stage and controller specifications [49] . Ball Screw Pitch Encoder Pulses Per Revolution MXH Multiplication Factor Controller Multiplication Factor 5 mm/rev 2500 1 4 One advantage of using the linear stage is the ability to set the acceleration rate within a given range, and the acceleration profile type in the Cimita software. This has the advantage of allowing the operator to minimize the acceleration forces exerted on

PAGE 126

126 the workpiece, and it also aid s in verifying the acceleration profiles. Figure 5 4 demonstrates one sample data set of the position and velocity for the PRO115 stages with a constant acceleration profile. The figure shows a series of 10 mm move s with 10 , 20, 30, and 40 mm/s commanded feedrates . Note that the last move in Figure 5 4 does not reach the commanded feedrate due to the move being too short for the commanded feedrate. Figure 5 4. Position and velocity measurements of the PRO115 linear stage. Four feedrates, 10, 20, 30, and 40 mm/s were ordered for move lengths of 10 mm. A constant acceleration profile with an acceleration rate of 200 mm/s 2 is shown. Calculating the galvo acceleration rate is not as simple as the linear stages because the motio n controller is either incapable of capturing the galvo encoder signals, or the galvo does no t contain a feedback mechanism. It is then compounded by the fact that the acceleration rate is significantly higher than the linear stages ; consequently, the 0 5 10 15 20 25 30 35 40 0 0.5 1 1.5 2 2.5 3 3.5 4 Position (mm) and Velocity (mm/s) Time (s) Position Velocity

PAGE 127

127 acce leration and deceleration regions are more difficult to capture. Additionally, Cimita does not have the capability to control the acceleration rate of the galvo . A variety of methods were investigated to determine acceleration profile and rate. One method is to utilize either a single or an array of photodiode position sensors, such as the OSI FIL C4DG sensor [51] . These type of sensors were deemed unusable because the ir response time is too slow, the position resolution too large, the lack of linearity in the response at different points on the surface, and the potential that the high intensity pulses may damage the sensor. Figure 5 5. Sample silicon cut measurement to determine the acceleration profile and rate of the galvo. The oval spots mark locations where the laser beam impacted the silicon wafer. The implemented technique analyze s the position of individual pulses on a silico n wafer as the laser beam decelerate s and accelerate s . Figure 5 5 shows one silicon wafer where the laser decelerates horizontally at the top, performs a vertical move in the middle to mark the stop location, and then accelerates horizontally at the bottom . To increase the resolution of the measurement this process is repeated 30 times for 50

PAGE 128

128 kHz, 66 kHz, and 100 kHz respectively. The machine d sections are measured using the SWLI with a 2X zoom tube, a scanning depth of 10 µm, and 3 averaged measurements. T he measurements result in over 1200 data points for two 50 µm move s . Figure 5 6 shows the resulting position velocity plot with all 1200 data points and their respective 95% uncertainty bars due to X Y SWLI measurement resolution , which is 276 nm for the 20X lens and 2X zoom tube combination. Figure 5 6. Measured data with SWLI resolution error bars included. Figure 5 7 . Measured and constant acceleration profile least squares error fit of the silicon pulse location data. An acceleration rate of 3920 m/s 2 is show n. The next step in characterizing the galvo acceleration profile and rate is to analyze the velocity position curve and determine if a constant acceleration, or a

PAGE 129

129 constant jerk profile fit is the best fit. A l east squares error (LSE) method is utilized for each profile type to determine the most appropriate acceleration or jerk value. Figure 5 7 shows the comparison between the measured and theoretical velocity for a constant acceleration profile, and Figure 5 8 shows the constant jerk LSE fit. A visual inspection of the two figures leads to the conclusion that the jerk profile is a better fit for the galvo , and the root sum square (RSS) error comparison confirms this with a value of 8 7.1 mm/s for the constant acceleration profile and 38.9 mm/s for the jerk profile. The equation used for the RSS analysis is , ( 5 4 ) where is the number of data points available, is the respective velocity for each data point, and is the theoretical velocity at the location of the data point. Figure 5 8 . Comparison of the computed galvo velocity and the LSE theoretical velocity. The jerk va lue for this plot is 1.35 e11 mm/s 3 . 5. 2 . 2 Motion Controller and Laser Interface Delays T he next step in characterizing the stages is to compare the simulated cut time to the measured cut time. Initial comparison experiments revealed a considerable

PAGE 130

130 difference between the simulated and measured cut times , and it appeared that the difference was consistent ly independent of the cut parameters. This consistent difference revealed that there were delays due to the software controller stage interface, and that these values need to be accounted for in the simulator model. 5. 2 .2.1 Linear X Y Stage Dela ys A series of experiments were conducted to determine which commands contribute to the time difference, and to quantify the time delay. A list of the different command combinations is shown in Table 5 3 , along with the recorded time to complete each combi nation. The X Y linear stage was utilized for these experiments with a feedrate of 10 mm/s and a commanded acceleration rate of 5 m/s 2 . Table 5 3 . Listing of the various command combinations to determine the delay introduced by each command type. Number of Repeats Command #1 Number of Empty Internal Repeats Time (s) 60,000 180 1 60,000 480 2 60,000 660 60,000 ON 720 60,000 OFF 420 60,000 ON, and OFF 960 60,000 ON, ON, OFF, OFF 1740 10,000 2 1 mm moves, ON, and OFF 2170 1,500 2 1 mm moves 345 5 2 1 mm moves 300 327 Table 5 4 lists the extrapolated delay times for each command type. It is important to note that the zero and non zero length move delays are only applicable to the X Y linear stage, and that the galvo displayed no move delay. The X Y move delay is likely caused by the time required to analyze the encoder signal and to ensure that

PAGE 131

131 the stage s reach the commanded position. Table 5 5 shows the results of the verification tests with the move and acceleratio n/deceleration delays incorporated into the prediction time. The uncertainty in the measure time is +/ 1 second; consequently, the majority of the difference s are within the measurement uncertainty. Table 5 4 . Extrapolated command delays for specific oper ations. Command Delay (ms) Zero Length Move 1 Non Zero Length Move 6 .4 Initiate Repeat Loop Sequence 5 Repeat Loop Sequence Iteration 3 Toggle Beam On 9 Toggle Beam Off 4 Table 5 5 . Comparison of measured and predicted cut times for X Y linear stage moves . Move Length Acceleration Rate Velocity Number of Passes Measured Time Predicted Time Difference (mm) (m/s 2 ) (mm/s) (s) (s) 5 0.1 5 4 676 677 0.1% 5 0.1 10 4 388 389 0.2% 5 0.1 15 4 314 314 0.0% 5 0.1 20 4 292 293 0.2% 5 0.1 25 4 291 293 0.5% 1 0.5 5 25 865 869 0.4% 1 0.5 10 25 506 509 0.5% 1 0.5 15 25 417 415 0.4% 1 0.5 20 25 385 389 1.0% 1 0.5 25 25 386 389 0.7% 0.5 1 5 40 719 718 0.1% 0.5 1 10 40 425 430 1.2% 0.5 1 15 40 360 355 1.3% 0.5 1 20 40 335 334 0.3% 0.5 1 25 40 335 334 0.3% 0.1 5 5 100 453 451 0.4% 0.1 5 10 100 293 307 4.6% 0.1 5 15 100 261 270 3.3% 0.1 5 20 100 261 259 0.7% 0.1 5 25 100 261 259 0.7%

PAGE 132

132 5. 2 .2. 2 Galvo Undershoot Phenomenon One issue that has arisen from the galvo not utilizing a closed loop control system is undershoot. An example of this is shown in Figure 5 9 . A crossing double rectangle feature was cut in R plane sapphire with a 30 µm stepover distance. The pulse frequency was 50 kHz, feedrate 50 mm/s, pulse power 5%, and measured power of 2.13 W. Figure 5 9 ( A ) shows the overall cut design, with the rectangles intersecting in one quadrant of the machined area. Figure 5 9 ( B ) shows that the long cut moves undershoot the desired position by 2.8 µm , and that the desired rectangular end moves ar e rounded. This is due to the control system assuming that the time to complete a move, , is only , (5 5 ) where is the move distance and is the commanded feedrate. In actuality the time to complete a move is also a function of the galvo acceleration and deceleration. Therefore , the stage controller is ordering the galvo to start the next move before the current move is finished. This resul ts in rounded cut paths and incorrect beginning/end move locations.

PAGE 133

133 A B Figure 5 9 . 2D plot of data from a SWLI measurement of overlapping rectangles machine in R plane sapphire. A) Plot of the entire cut , and B) a zoomed in view of the undershoot present at the beginning and end of the long moves.

PAGE 134

134 A B Figure 5 10 . 2D plot of data from a SWLI measurement of overlapping rectangles machine in R plane sapphire with a delay added into the cut program . A) Plot of the entire cut, and B) a zoomed in view of the undershoot present at the beginning and end of the long moves. The figures show that a delay reduces the undershoot and results in more rectangular features at the small stepover positions.

PAGE 135

135 It is important to replicate the desired cut path as closely as possible in order to minimize geometric differences between the desired and machined component. The UF simulator acco unts for the incorrect move time by calculating the actual time for each move and comparing it to the time from Equation 5 5. Then, the simulator adds in a delay between each move to account for the discrepancy and to increase part path accuracy. Figure 5 10(A) shows the same cut path that was utilized in Figure 5 9(A), but this cut program included compensating delays. A comparison of Figure 5 9(B) and 5 10(B) shows that the delays successfully eliminated the undershoot and resulted in a more rectangular e nd feature. 5. 3 Beam Profile The next parameter of interest for the Oxford J 355PS system is the beam profile at the workpiece. Traditionally, Nd:YAG lasers have beam shapes that are highly Gaussian in shape, but imperfections in the laser cavity and focu sing optics can alter the beam shape ; therefore, it is necessary to measure the profile at the workpiece [25] . Figure 5 1 1 . Visualization of how a scanning slit profilometer operates, after [52] .

PAGE 136

136 To the beam measurement a n Ophir Sphiricon PH00016 NanoScan silicon scanning slit profilometer with a 3.5 mm entrance aperture, and 1.0 µm wide scanning slit is used to measure the beam profile at various Z locations. A visual ization of how a scanning slit profilometer operates is shown in Figure 5 1 1 , and a 3D rendering of one beam profile measurement s is shown in Figure 5 1 2 . Figure 5 1 2 . Ophir 3D rendering of the beam profile. Note the shape is extrapolated from two 2D beam profiles ( X axis and Y axis respectively). 5. 3 .1 Focal Beam Diameter Analysis Technique The standard operating procedure for measuring the beam shape is to position the lase of measurements; however, the Coherent Talisker laser head produces pulses that are so intense that even at max attenuation levels the pulses can still damage the rotating s ilicon drum . As a work around to this dilemma, the beam radius was measured at a series of locations in the Z axis where the max intensity is below the ablation threshold.

PAGE 137

137 is capable of computing the beam radius via two methods: 1/e 2 and 4 sigma. The 1/e 2 method, also known as the 13.5% method, analyzes the intensity profile to determine the maximum value. Then, the beam radius is determine from the distance between the two points at which the intensity drops to 13.5% of the maximum value. The downside to this technique is that it relies heavily on the maximum intensity value from the intensity profile measurement. T he high peak intensity of the Coherent laser likely causes the peak intensity measurement to be in accurate due to saturation of the sensor. One way to work around this dilemma is to utilize the 4 sigma approach. The 4 sigma approach analyzes the entire intensity profile and determines the respective X and Y beam radii, and , from , (5 6 ) and , (5 7 ) where and are the distances away from the peak intensity in the X and Y axes [28] . Note that this technique considers every point on the intensity curve ; thereby , reducing the significance of the peak intensity measurement. After the beam radius is recorded for each Z location, t hen the data is loaded into a custom Matlab ® program that use s a LSE approach to fit the data to Gaussian shape equations. Figure 5 1 3 shows a comparison between the measured beam diameter and the theoretical Gaussian beam diameter at different locations i n the Z axis .

PAGE 138

138 Figure 5 1 3 . Measured and least squares fit theoretical beam waist at various distances from the focal plane in the Z axis. 5. 3 . 2 Beam Shape Quality As stated in section 3.2, Coherent lists the beam quality, M 2 , of the Talisker Ultra 355P S as <1.3 [30] . However, as stated above the laser beam must travel through a series of optics and the galvo before it reaches the workpiece ; consequently, the beam shape is software includes a routine to calculate the beam quality. software measures the beam profile at various Z locations and computes the resulting M 2 for the laser wavelen gth. This routine resulted in a calculated beam quality of 0.25, but it is likely that this value is underestimated due to the certain saturation of the sensor at dist ances close to the focal plane. Figure 5 14 shows one measurement ~350 nm from the focal beam plane where the plot demonstrates a saturation effect on the peak region. 0 100 200 300 400 500 600 700 800 900 1000 -20 -10 0 10 20 Beam Diameter (µm) Distance from Focal Plane (mm) Measured Axis 1 Measure Axis 2 Calculated

PAGE 139

139 Figure 5 14. Demonstration of peak saturation at z heights close to the focal plane for both X and Y axes . A B Figure 5 15. Gaussian intensity profile at a distance of 13 .35 mm from the focal plane for A) theoretical profile, and B) measured X and Y profiles.

PAGE 140

140 While a high quality beam shape is necessary in certain applications, it is not always the case for laser micromachining . This is because the pulses typically overlap a sufficient amount that the actual beam shape has a minimal effect on the finished feature . However, to increase the accuracy of the simulator it is important to determine if it is acceptable to approximate the beam shape as Gaussian, or if another beam shape is more appropriate. Figure 5 1 5 shows a comparison of the measured intensity profile of the Coherent Talisker laser versus the theoretical Gaussian profile at a distance of 13.4 mm from the focal plane . The measured profile appears to have a flatter peak than the theoretical, but the overall beam shape is similar. 5. 4 Attenuator Characterization An important aspect of laser machining is determining the most appropriate peak intensity for each pulse. Selecting a peak intensity that is too high can im part excessive thermal energy and cause failure, while a small peak value may not exceed the ablation threshold. The Oxford system utilizes a diffractive attenuator to modify the peak intensity. The operator is able to select a value between 0 and 100% in 0.5% increments. A selection of 100% produces the maximum peak intensity, and 0% theoretically should not allow any photons to pass through. A visualization of how a variable diffractive attenuator operates is shown in Figure 5 1 6 .

PAGE 141

141 Figure 5 1 6 . Visualization of how a laser beam variable attenuator operates. Depending on the position of the attenuating disc, the fraction of light passing through the slot changes, after [53] . The actual attenuation o f the laser beam is an important factor in characterizing the laser material interaction; therefore, a verification of the attenuation capability is performed to ensure that the commanded attenuation level matched the realized attenuation. To test this par ameter the built in pyrometer is utilized in measuring the average power at the workpiece over the full range of attenuation commands. Figure 5 1 7 shows the results of two trials : one with an increasing attenuation over time (Trial 1) and one with a decrea sing attenuation (Trial 2) . The increasing and decreasing trials ensured that the experiment would not suffer from hysteresis errors that could arise from heating of the pyrometer. Visual inspection of Figure 5 17 reveals that the measured attenuation is less than the commanded attenuation. To account for this disparity a second order polynomial is fitted to the measured data and a relationship between the commanded and actual attenuation is

PAGE 142

142 , (5 8 ) where is the commanded attenuation in decimal form, and is the realized attenuation in decimal form. Figure 5 1 7 . Comparison of commanded and measured attenuation values. 5. 5 Lasing Command Delay In section 5.1.2 the beam on and beam off command delay times are extrapolated . Knowing the delay for the beam on / beam off commands is important for time estimation, but it does not reveal how long the laser is lasing for each command. To determine the fraction of time the system is lasing during the delay a two step approach is utilized . 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0% 20% 40% 60% 80% 100% Measured Normalized Power Commanded Attenuation Assumed Linear Relationship Ascending Measurement Decending Measurement

PAGE 143

143 First, the combined lasing time duri ng a beam on and beam off command set is determined . To calculate this time a square is machined by moving a certain distance in the X axis, performing a beam on beam off command set, then repeating this over the entire square . This creates a grid of beam on / beam off command craters that can then be measured and the average depths compared to previous depth data. The analysis revealed that the combined time of lasing for the beam on / beam off command is 6.5 ms +/ 0. 2 ms. Next, it is necessary to determine how much each delay contribute s to the overall lasing time. This is accomplished by produced two extra deep cut regions (one due to beam on command s , one due to beam off command s ). A visualization of the cut path is demonstrated in Figure 5 1 8 . Then , the average cut depths in these regions are measured, and a ratio of beam on delay lasing to beam off delay lasing is established : 1. 8 +/ 0. 5 . This results in a beam on lasing time of 4.2 ms +/ 1 .1 ms and a beam off lasing time of 2. 4 ms +/ 1.0 ms . Figure 5 1 8 . Visualization of the beam on / beam off lasing time ratio experiment. The laser starts at location A, moves to B, and turns the beam off . Then the laser moves over to C, turns the laser on, and moves back to B. Lastly, the laser moves to C, turns the laser on, moves over to D, and turns the beam off .

PAGE 144

144 5. 6 Chapter Summary In this chapter the techniques and experimental setups for characterizing the Oxford J 355PS laser micromachining station were discussed. Silicon pulse distance measurements in acceleration and deceleration regions revealed that the galvo is best modeled b y a constant jerk profile and that the acceleration rate is 1.35e11 m/s 3 . Next, the beam profile was measured at a variety of Z locations. The data was fitted to Gaussian beam shape equations, and through the use of a LSE analysis the resulting beam diame ter is with . Additionally, a comparison between the measured and theoretical Gaussian intensity profile was presented, and a difference in the spread of the intensity profile at the peak was noted. Then, the method of op eration for a diffractive attenuator was discussed, along with the function of this item. A calibration of the commanded versus realized attenuation was performed that revealed the need to incorporate the calibration into the simulation software. Lastly, the respective lasing times for the beam on and beam off commands was explored. The two step process involved first determining the combined lasing time of a beam on / beam off command set. Then , the ratio of lasing times was calculated by performing a uniqu e cut strategy. Finally , the respective lasing times were quantified.

PAGE 145

145 CHAPTER 6 ULTRASHORT PULSED LASER MICROMACHINING OF SILICON In this chapter the characterization of <100> P type silicon is discussed. First, an introduction to why silicon is characterized before the main material of interest, sapphire, is discussed . Next, the processing parameters of interest for ultrashort pulsed laser micromachining are explored . Then silicon characterization results are discussed. Finally, the application o f a gas flow to reduce ablation plume shading is examined. 6.1 Introduction to Silicon The material of primary interest for this research paper is sapphire, but it is difficult to run a large number of tests on sapphire due to the high cost of the wafers. In order to refine the characterization process, and to reduce the number of sapphire wafers needed, silicon wafers are first characterized. Silicon is selected as the initial characterization material because of its relatively low cost and widespread use in the micro electro mechanical systems (MEMS) community [54] . Although techniques exist for fabricating components out of silicon, the ability to produce features with single step processing in non cleanroom environments is a n attractive option; consequently, it was determined that the use of silicon as the initial characterization material will be useful in refining the characterization process and for future silicon laser micromachining research [3] . 6.2 Laser Processing Parameters The first step in characterizing the ablation process of silicon is to determine the main machining parameters of interest . As discussed in previous chapters, the main parameters of interest are the pulse delay, the pulse area overlap, average pulse fluence, machining environment, and the number of passes. These four parameters

PAGE 146

146 combine to affect the overall material removal r ate, and the effect of each of these parameters must be explored to fully understand the ablation process. 6.2.1 Pulse Delay One parameter that potentially has a strong effect on the MRR is the delay between pulses. There are two aspects to the pulse delay : energy lost to the ablated material plume, and the rate of electron relaxation. The amount of energy that actually reaches the cut material's surface is affected by the recently ablated material plume above the workpiece, as demonstrated in Figure 6 1. I f there is not sufficient time for the plume to either be removed or dissipated, then the plume can absorb some of the laser energy and form a plasma over the workpiece [4] . The plasma can further block laser energy from reachi ng the surface. This effect can lead to high uncertainties in the depth of cut from each respective laser pulse. A B C Figure 6 1. The process of forming plasma over the workpiece surface. A) Pulse #1 impacts the workpiece, B) ablated material is ejected, and C) pulse #2 interacts with the weakly ionized molecules to form a plasma and the pulse intensity is decreased.

PAGE 147

147 A shorter or longer pulse separation also has an effect on the relaxation of the electrons in the cut materi al's lattice. Avalanche ionization plays a strong role in ultrashort pulsed ablation [38] . During avalanche ionization the transparency of materials is modified, allowing previously transparent materials to become strongly abso rbent [4] . This absorptivity increase can have a significant effect on the ablation rate. The effect of the pulse delay needs to be examined in order to determine the relationship between pulse frequency and single pulse ablati on depth. 6.2.2 Pulse Overlap Pulse overlap is another parameter that can affect the MRR. Similar to pulse delay, the overlap can affect the distance the plume must dissipate before the subsequent pulse. The smaller the spacing between pulses, the further the ablated plume must dissipate in order to not affect the following pulse. Also, if the pulses are sufficiently apart, then free electrons from previous pulses may not interact with the new pulse. If this occurs, avalanche ionization may not be possible; therefore, multi photon ionization may dominate. 6.2. 3 Average Pulse Fluence The fluence of an individual pulse affects the type of ionization that is possible. Multi photon ionization is typically possible as long as the workpiece material is not overl y transparent to the wavelength of the laser; however, avalanche ionization can only occur if the fluence is above a certain threshold. One important characteristic of ultrashort pulsed lasers is the phenomenon of absorption saturation at a certain fluenc e level. At low levels, as the fluence increases the amount of energy absorbed and reflected rises, and the amount of energy transmitted through the workpiece decreases. When the energy reaches a high enough

PAGE 148

148 level the amount of energy absorbed plateaus, an d the amount of energy reflected continues to rise. This is due to the ablation plume shading the workpiece and becoming more reflective [4] . Although it is possible that the maximum fluence available from the laser is below the transition point, it is still an important phenomenon to consider. 6.2. 4 Machining Environment An alternative to selecting the pulse delay and pulse overlap that minim izes the ablation plume shading is to remove the plume from the cut region before it can affect the following pulse. This technique is discussed further later in the chapter. 6.2. 5 Number of Passes In an ideal world the ablation rate per pulse would remain constant regardless of the number of repeat passes, but in reality this is often not the case. As the number of passes increases the surface roughness typically increases as well. This has an effect on the ablation rate because it changes the absorptivity of the workpiece material; consequently, the ablation rate can increase with an increase in the number of passes. However, with greater depth features also comes a greater amount of redeposit due to the increased difficulty in removing the ablated materia l. It is necessary to consider these two trends during the modeling process. 6. 3 Laser Silicon Interaction As discussed in sections 6.2.1 through 6.2.5 the processing parameters can have a variety of effects on the MRR. The main interest of this initial ch aracterization is to refine the characterization process for sapphire; consequently, the scope of characterizing silicon is limited to only a few experiments. The experiments are all carried out on <100> P type silicon wafers with the UF laser micromachini ng station.

PAGE 149

149 After they are machined, the features are measured using the Zygo Newview 7200 SWLI , as shown in Figure 6 2, to determine the average depth of cut. The settings for measuremen ts are used to determine the average measured depth at each data point in the image, and the scanning depths are either 10 µm or 30 µm depending on the feature depth . Then the average depth per pulse, , is then calculated by , (6 1 ) where is the average depth of cut, is the stepover distance, is the commanded feedrate, is the pulse frequency, and is the number of passes. Figure 6 2. Zygo Newview 7200 SWLI setup at the University of Florida. Daniel Blood. 2014. Gainesville.

PAGE 150

150 6. 3 . 1 Empirical Data The first parameter tested is the average fluence. To separate the fluence variable from the other variables non overlapping single pulses are machined on a silicon wafer and the maximum depth per pul se is recorded. The results of this test are shown in Figure 6 3. The figure shows that for the range of fluences tested that the fluence depth of cut relationship is relatively linear, and that the uncertainty in the maximum depth of cut increases as the fluence increases. Figure 6 3 . Fluence depth of cut relationship for silicon. The data represents the maximum measured depth of cut for a single pulse. The error bars represent the 95% uncertainty due in the depth that was computed by repeating the single pulse measurement 10 times each. Next, 100 µm squares are machined in silicon with the fluence and number of passes kept constant at 2.2 J/cm 2 and 5 passes respectively, and the pulse area overlap and pulse repetition rate varied. Figure 6 4 shows t he results of those tests. It reveals that as the pulse area overlap is increased, and as the pulse repetition rate are increased that the average volume of material removed per pulse decreases. This is characteristic of ablation plume shading because the higher overlap results in a smaller 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 2 4 6 8 10 Single Pulse Max Depth of Cut (µm) Average Pulse Fluence (J/cm 2 )

PAGE 151

151 distance between pulses, and a higher repetition rate decreases the time the plume has to dissipate. Figure 6 4 . Intensity depth of cut relationship for silicon determined by single pulse ablation depth. Depths were measured using a Zygo Newview 7200 SWLI and analyzed performed with a custom Matlab ® code. 6. 3 . 2 Gas Flow In section 6.2.1 the effect of ablation plume shading on the MRR is discussed, and in Figure 6 4 the empirical data confirms that silicon is affected by the plume shading. In order to reduce the plume shading it was theorized that the flow of a gas over the workpiece surface can remove the ablated pa rticles before the subsequent pulse can interact with the plume. To test the gas flow theory several 100 µm square s were machined in a silicon wafer , each square machined with a different gas flowing over the surface . The results of these tests are shown in Figure 6 5 . The gases were blown through a 40 mm wide Lechler air nozzle at a flow rate of 40 L/min, and at a distance of 12.7 mm from the workpiece surface, as seen in Figure 6 6 . The data in Figure 6 5 (A) indicates that the presence of gas flow increa ses the material removal rate, with further increases realized 3 3.5 4 4.5 5 5.5 6 40 50 60 70 80 90 100 Average Volume Removed Per Pulse (µm 3 ) Pulse Area Overlap (%) 250 Hz 500 Hz 1 kHz 2.5 kHz 10 kHz 25 kHz 50 kHz 100 kHz

PAGE 152

152 using more inert gases. Additionally, 6 5 (B) shows a decrease in the surface roughness for a given depth of cut. A B Figure 6 5 . The effect of gas flow on a machined silicon feature. A) Depth of cut and, and B) surface roughness data for different gases flowing across a silicon wafer as it is being machined. One possible explanation for these two results is that the fl ow of gas minimizes redeposited material by blowing it out of the machining area. Another possibility is that the more inert gases retard plasma formation in the ablated material plume above the surface. The degree of ionization, , depends on the densi ty of ions, , and the density of neutral atoms, [4] , 0 5 10 15 20 25 0 5 10 15 20 Depth of Cut (µm) Number of Repeat Cut Passes Stagnant Air Air Nitrogen Nitrogen Argon Argon 0 100 200 300 400 500 600 700 800 900 1000 0 5 10 15 20 Ra (nm) Depth of Cut (µm) Stagnant Nitrogen Nitrogen ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )

PAGE 153

153 . (6 2 ) The gas flow has the effect of decreasing the density of ions and increasing the density of neutral atoms. This is important because even a partially ionized gas (~1%) can have characteristics of plasma. A B Figure 6 6 . Demonstration of a typical nozzle workpiece setup for blowing different gases across the workpiece surface. A) Visualization of the nozzle gas flow over the wafer, and B) two gas flow setups where the long cut moves are either parallel or perpendicular to the gas flow. Daniel Blood. 2012. Gainesville. 6. 4 Chapter Summary In this chapter the usefulness of an ultrashort pulsed laser micromachining station processing silicon was discussed, along with the motivation that silicon was tested before sapphire. The main processing parameters of interest were introduced, and the effect of each of these parameters on the finished feature was discussed. Next, the effect of pulse intensity on the MRR of silicon was explored, along with the effect of Gas Flow Gas Flow Parallel ( ) Perpendicular ( )

PAGE 154

154 pulse area overlap an d pulse frequency. Lastly, the ability of a gas flow to reduce the ablation plume shading effect was presented.

PAGE 155

155 CHAPTER 7 ULTRASHORT PULSED LASER MICROMACHINING OF SAPPHIRE In this chapter the technique for characterizing the ablation rate of R plane sa pphire is discussed. First , an introduction to sapphire as a sensor material is introduced. Then, an analysis is performed to determine if the UF laser system is sufficient to be considered ultrashort for sapphire. Next, the characterized processing parame ters are explored, along with the introduction of a Monte Carlo analysis to determine the gentle and strong ablation slopes and the gentle strong fluence transition. Then the strategies for modifying parameters is discussed, and three modification verifica tion data sets are analyzed. Finally, the effectiveness of the added passes routine is explored. 7 .1 Introduction to Sapphire As engineers seek to design more efficient gas turbines, they require a detailed understanding of fundamental thermal fluid phenom ena, as well as active control methods , in high temperature environments . The high temperature requirement is based on the increasing turbine inlet temperatures that have risen to 1500 C in combined cycle systems in order to improve turbine peak power and efficiency [55] . Silicon is a widely utilized material in sensor fabrication; however, the limited survivability of silicon based sensors in high temperature and harsh environments has also caused researchers to investigate other materials for high temperature MEMS based sensors . sensors promising, but it also renders most traditional MEMS manufacturing methods ertness does not allow for effective dry or wet

PAGE 156

156 etching; consequently, a more effective method of machining the material is necessary [56] . Ultrashort pulsed laser micromachining is one technique that is capable of machining sap phire with a high degree of precision, and with minimal material damage. However, to minimize material damage and increase the material removal rate the sapphire ablation process needs to be fully characterized. 7 .2 Laser Sapphire Interaction Modeling the laser material interaction is a complicated task, especially when different materials require different energies to move electrons from the valence band to the conduction band. The first step in analyzing sapphire is to ensure that the UF laser system sapp hire combination is considered ultrashort, as defined in Equation 2 6. To accomplish this task the thermal diffusivity and optical absorptivity must first be defined, , ( 7 1 ) and , ( 7 2 ) w here is the thermal conductivity, is the bulk density, is the specific heat, is the refractive index, is the extinction coefficient, and is the laser wavelength [57, 58] . The values of the above variables for sapphire are listed in Table 7 1. Note that the extinction coefficient varies by an order of 4 magnitude. This is due to the extinction as a function of pulse energy. The literature does not prov ide a measurement of the coefficient at the 355 nm wavelength so l inear interpolation may have a strong effect on the number. To analyze a worst case scenario of a shallow optical penetration depth the larger value in the range, 7.8e 4 is used.

PAGE 157

157 Table 7 1. Sapphire material properties and UF laser properties [59, 60, 61] . Thermal Conductivity Bulk Density Specific Heat Extinction Coefficient Laser Wavelength 30.3 W/mK 3.98 g/cm 3 0.184 cal/gK 7.8e 4 to 2.8e 8 3 55 nm Inserting the values in Table 7 1 into Equations 2 4, 2 5, 7 1 , and 7 2 results in a thermal diffusion depth of 24.4 nm, and an optical penetration depth of 72.3 µ m. T he thermal diffusion depth is less than the optical penetration depth ; consequently, this process is considered ultrashort. 7 . 3 Traditional Laser Material Removal Model There are several atomic scale ionization models available for ultrashort pulse material interaction. In section 2.3.2 the difference between multi photon and avalanche ionization MRR was discussed. Figure 2 20 showed a generally accepted model of an increasing number of pulses resulting in a greater MRR, and a higher pulse power also and and the strong results in a higher but less controllable MRR. Although previous experiments in silicon did not reveal a gentle to strong transition, optical absorpt ivity of silicon, ~10.6e5 1/cm, is greater than that of sapphire, ~280 1/cm; therefore, it is likely that there will be a transition when the fluence or overlap is sufficiently high to alter the amount of energy absorbed versus transmitted through the work piece [62] . 7 . 4 Empirical Sapphire Model The process of empirically modeling the laser pulse material interaction is two step. First the effect of gas flowing over the surface is explored. This determine s

PAGE 158

158 whether there is a ne ed for separate gentle strong characterizations of stagnant and gas flow, or if only one set of data is necessary. Next, 400 µm by 250 µm rectangles are machined in R plane sapphire with a variety of fluences, pulse area overlaps, and number of passes. Th e size of the rectangle is dictated by the measurement area of the Zygo Newview 7200 SWLI with a 20X optic and 1X zoom tube. The Newview SWLI is capable of stitching together several measurements into one image, but this capability is not being utilized be cause the stitching process can at times result in errors between the depths of the respective stitched images. Therefore , to ensure the reliability of the measurements the 400 µm by 250 µm size is the characterization geometry. Once the SWLI data is collected, a Matlab ® program analyze s the depth of cut for each set of parameters. Then the program computes the fluence at which the gentle to strong ablation transition occurs for a particular pulse area overlap and number of pulses. Next , four matrices (gentle slope, gentle y intercept, strong slope, and strong y intercept) are calculated with the respective number of passes on the X axis and the pulse area overlap on the Y axis. Once this is complete the matrices are integrated into the simulat or program, and verification of the predicted to actual cut depths are established. 7 . 4 . 1 Gas Flow The effect of gas flow over the workpiece is examined because p revious gas flow tests using silicon showed an improvement in the MRR. The test s revealed that the mor e inert the gas, the greater the reduction in ablation plume shading. With this in mind three machining conditions are tested: stagnant, air flow, and argon flow. The selected flowrate is 21 L/min, nozzle distance 15 mm, and the nozzle setup is sho w in

PAGE 159

159 Figure 7 1 . The nozzle is a loc line 1.5 inch swivel nozzle with 16 3/64 inch diameter holes. Figure 7 1. Experimental setup for flowing different gases over a sapphire wafer. Daniel Blood. 2014. Gainesville. In addition to testing the gas flow, the pulse repetition rate is also tested. This aspect is included for two reasons: 1. the longer delay between pulses gives the plume more time to dissipate, and 2. the dependence on pulse frequency is tested in order t o determine if the ablation rate is consistent among different pulse frequencies. The number of passes for all of these experiments is 10. Figures 7 2 through 7 4 shows the results of these experiments, and analysis of the figures shows that the flow of ga s has minimal effect on the ablation rate. The flow of argon in Figure 7 4 even appears to decrease the overall ablation rate. Due to these results, the following characterization of sapphire neglects the effect of gas flow over the

PAGE 160

160 workpiece as an operati ng parameter. However, a gas flow may still be utilized while machining deep trenches to aid in removing ablated material from the cut regions. Figure 7 2 . Gas flow machining experiment results for a 5 µm pulse spacing. Figure 7 3 . Gas flow machining ex periment results for a 2.5 µm pulse spacing.

PAGE 161

161 Figure 7 4 . Gas flow machining experiment results for a 1.25 µm pulse spacing . The effect of the pulse frequency on the ablation rate does appear to have an effect. Figure 7 2 and Figure 7 3 show a trend of decreasing material removal for the 50 kHz repetition rate data points above 10 J/cm 2 . I t is theorized that the cause of this trend is due to a slight ablation plume shading that is not correctable by the flow of gas . This trend is not app arent in Figure 7 4 . The majority of pulse spacing combinations are below 2 µm; therefore, the effect of the pulse repetition rate should have minimal effect on the ablation rate. The above results are likely due to the higher ionization potential of sapph ire relative to silicon. The amount of energy required to ionize the ablation plume is greater for sapphire than it is for silicon; therefore, to produce the same phenomenon in sapphire as in silicon either the amount of energy imparted into the plume is i ncreased and /or the distance between the ablated particles is reduced . Lastly, Figure 7 5 shows the surface roughness, Ra, as a function of cut depth. Unlike the silicon results, this plot shows a relatively consistent trend between the different parameter combinations.

PAGE 162

162 Figure 7 5 . Surface roughness as a function of depth for all of the various pulse frequency fluence gas flow combinations. 7 . 4 . 2 Gentle Strong Ablation The next step in characterizing sapphire is to machine the 400 µm by 250 µm rectangles with various combinations of parameters and determining the gentle strong transition points, and the fluence ablation depth relationships. The range of machining parameters tested is listed i n Table 7 2 . Note tha t the middle parameter in Table 7 2 , pulse spacing, is another method for describing the pulse area overlap. It is used in place of the area overlap because it is easier to visualize the spacing differences. Each set of parameters was repeated 10 times to produce a probability distribution for each set of parameters . 269 different combinations of parameters are tested. Table 7 2 . The range of values used for each of the three main parameters of interest for the sapphire characterization. Fluence (J/cm 2 ) Pul se Spacing (µm) Number of Passes 1. 2 21.5 10 0.5 1 50 Once the parameter combinations are machined and measured, then the average cut depth is calculated using a custom Matlab ® program. Then the data is analyzed to determine the gentle strong fluence transition points, and the respective

PAGE 163

163 gentle/strong ablation slopes. This analysis is complicated by the probability distribution of the depth of cut and the fluence for each data point. A B Figure 7 6 . Probability distribution shapes for A) the depth of cut, and B) the laser fluence. To account for the uncertainties a Monte Carlo routine is utilized with 100,000 iterations to calculate the mean transition fluence and ablation slope value s. The routine also produces a distribution curve for the transition fluence and ablation slope values

PAGE 164

164 that are important in modification routines . One caveat of this analysis is that t he Monte Carlo routine requires knowledge of the probability distributi on shapes for the fluence and depth uncertainties; consequently, the overall distribution shape s w ere computed. Figure 7 6 shows the probability distribution for both the depth of cut and also the fluence . A Matlab ® chi squared goodness of fit test (chi2go f) was performed on both distributions to determine if the null hypothesis that both data sets have a G aussian distribution shape is rejected . The hypothesis was not rejected with a 5% significance level. The results of the Monte Carlo analysis are shown in F igure s 7 7 through 7 9 . G entle ablation line fits are marked in blue, and strong ablation line fits marked in red. Several trends can be extrapolated from the analysis of the graphs. First, at 5 passes the gentle to strong ablation transition point s occur at lower fluence levels for pulse spacings below 2 µm, but the difference in the transition fluence level decreases as the number of passes increases. Secondly, the shorter pulse spacings result in higher removal rates. This is likely due to an incre ase in the utilization of short lived defects that allow for a greater removal effect from avalanche ionization [4] . Third , the ablation threshold is approximately 1 J/cm 2 , which is than ~3 J/cm 2 for a 3.7 picos econd laser [63] . laser photons; consequently it requires fewer photons to be absorbed to provide the necessary energy to free an electron. Lastly, once the gentle strong ablation transition occurs there is an apparent linear relationship between increasing fluence and

PAGE 165

165 increasing material removal. This aspect is vital in predicting removal rates for various fluence levels. Figure 7 7 . Average volume removed per pulse for 5 passes. Figure 7 8 . Average volume removed per pulse for 10 passes.

PAGE 166

1 66 Figure 7 9 . Average volume removed per pulse for 25 passes. 7 . 4 . 3 Model Verification Three steps are involved in verifying that the calculated fluence transition points and slopes accurately predict the cut depth of R plane sapphire. First, the predicted cut depth is compared to the 2690 machined data points . This comparison reveals whethe r the theoretical model accurately predicts the depth of cut by performing a goodness of fit analysis. The analysis reveals that t he cut depth of 96. 6 % of the squares fall within the +/ 2 sigma uncertainty range. Additionally, when you consider only the s quare s with greater than 2 passes the average error in the cut depth prediction is only 8.5 %. A histogram of the percent difference between the theoretical and measured cut depth is shown in Figure 7 1 0 .

PAGE 167

167 Figure 7 1 0 . Average difference between the theoretical depth of cut and measured depth of cut for the data points used in the Monte Carlo characterization routine. The second step is to compare the theoretical model to data that was not used in the Monte Carlo simulat ion. Table 7 3 shows the parameter combination and depth comparison of a data set not used in the characterization analysis. The relatively high error of 19.2% is typical for low fluence cuts because the value is near the ablation threshold. Table 7 3. Co mparison of the average difference in predicted versus measured cut depths. Fluence (J/cm 2 ) Pulse Spacing (µm) Number of Passes Pulse Repetition Rate Average Error Percent Within +/ 2 Sigma Uncertainty Range 1.5 1.25,1.5 1 25 100 kHz 19.2% 87.5% The last comparison is performed in section 7 .4.2 when the results for the modification routine are discussed. The primary goal of the simulator is the ability to

PAGE 168

168 predict the depth of cut for a given set of features, and the modification results provide a metric to determine the effectiveness of the simulator. 7 . 4 . 4 Sidewall Angle A lesser understood phenomenon affecting ablation modeling is the sidewall angle. Testing has shown that as the aspect ratio of a feature is increased, the rate of material ablati on decreases. Figure 7 1 1 shows a typical depth profile of a laser machined trench. Several experiments were performed on 110 µm thick R plane sapphire wafers to test the effects of pulse area overlap, fluence, and number of passes . 100 µm wide by 500 µm long rectangles were machined in the sapphire, and the width of opening on the top of the wafer is compared to the width of the opening on the bottom. The equation for calculating the sidewall angle is , ( 7 3 ) where is the sidewall a ngle, is the width of the opening on the top of the wafer, is the width of the opening on the bottom of the wafer, and is the wafer thickness. The parameters used in the experiments are listed in Table 7 4 , and the results are shown in Figure s 7 1 2 through 7 1 4 .

PAGE 169

169 Figure 7 1 1 . Experiments on sapphire and silicon have shown maximum attainable Table 7 4 . Processing parameters for Figure 7 12 through 7 14 . Figure Fluence Feedrate Pulse Frequency Passes (J/cm 2 ) (mm/s) (kHz) 7 12 7.5 1000 7 13 100 100 1000 7 14 7.38 100 100 Figure 7 1 2 . Sidewall angle tests in R plane sapphire with a fixed fluence, number of passes, and pulse frequency but a varying pulse area overlap. 0 5 10 15 0% 20% 40% 60% 80% 100% Sidewall Angle (degrees) Pulse Area Overlap (%) Left Opening Right Opening

PAGE 170

170 Figure 7 1 3 . Sidewall angle tests in R plane sapphire with a fixed feedrate, number of passes, and pulse frequency but a varying fluence. Figure 7 1 4 . Sidewall angle tests in R plane sapphire with a fixed fluence, frequency, and pulse spacin g but a varying number of passes. Figure 7 1 2 shows that as long as the pulse area overlap is above approximately 75% that the sidewall angle is relatively constant. However, when the fluence is varied, as seen in Figure 7 1 3 , it results in a decreasing angle for an increasing fluence. The third graph, Figure 7 1 4 shows a decreasing angle for an increasing number of passes, but once the passes reach 500 the reduction becomes negligible. While the number of passes appears to ha ve an impact on the sidewall angle, for the current stage of the 0 5 10 15 0.0 5.0 10.0 15.0 20.0 25.0 Angle (degrees) Fluence (J/cm 2 ) Left Opening Right Opening 0 5 10 15 20 0 500 1000 1500 2000 2500 Sidewall Angle (degrees) Number of Passes Left Opening Right Opening

PAGE 171

171 simulator the sidewall angle prediction only rel ies on the fluence. Figure 7 1 3 shows two linear fit lines that are used to predict the sidewall angle for two fluence ranges. If the fluence i s less than or equal to 9.66 J/cm 2 then the sidewall angle is , ( 7 4 ) and if the fluence is greater than 9.66 J/cm 2 then the sidewall angle is . ( 7 5 ) In the above equations is the sidewall angle in degrees, and F is the fluence in J/cm 2 . While this aspect warrants further research to fully model the sidewall angle problem, it is out of the scope of this dissertation . There are plans to implement additional tilt stages to hel p reduce or eliminate the sidewall problem, and at that time this phenomenon can undergo a more rigorous examination. 7 . 4 . 5 Surface Roughness During analysis of the gentle strong data an interesting relationship was discovered. Figure 7 1 5 shows the surface roughness (RA) as a function of the cut depth for all of the gentle strong data points. In previous silicon machining experiments there a strong relationship between machining parameters and the surface roughness was noted, but the Figur e below does not follow that trend. T he data does not display a definitive surface roughness cut depth relationship, but it is sufficiently close to a polynomial that a quantitative relationship between the two variables is given in Equation 7 6 . It is imp ortant to note that this variable was not previously expected to be characterized, and that the relationship in Equation 7 6 is only a rough estimate.

PAGE 172

172 Figure 7 1 5 . Surface roughness as a function of cut depth for a variety of processing parameters. ( 7 6 ) In the above equation is the surface roughness in µm, and is the cut depth in µm. 7 . 5 Parameter Modification , Optimization Routine, and Part Path Modification The next step in simplifying the operation of ultrashort pulsed lasers for micromachining of sapphire is to implement parameter modification, and part path modification routines. The wide range of pulse frequencies (200 kHz to single digit Hz), feedrates ( 1000 mm/s to <1 mm/s), attenuation levels (100% to 0%), and almost infinite number of passes makes selecting the most appropriate set of parameters very challenging. The following routines simplify the selection process. 7 . 5 . 1 Parameter Modification Strate gies In previous iterations of the simulator a variety of input variables w ere available to the operator; however, the large number of variables were found to be overly

PAGE 173

173 confusing and required a greater understanding of the program than most users would likely contain. With this in mind the input selection has been narrowed down to two variables: the cut type (through cut, face mill, or pocket mill) and the depth tolerance (speed, mixed, or tolerance). A review of each cut type is available in Table 4 1 , and the method of parameter selection is explained in Table 7 5 . Table 7 5 . Explanation of the method of selecting the modified parameters. Through Cut Speed 200 kHz repetition rate and the pulse area ov erlap is selected such that the time to reach the desired depth is minimized. Mixed 2 sigma removal rate matrices are analyzed. Tolerance Face Mill Speed 50 kHz repetition rate and the pulse area overlap is selected such that the time to reach the desired depth is minimized. Mixed 50 kHz repetition rate and the pulse are a overlap is selected to minimize the desired simulated cut difference. Tolerance , mean, and 2 sigma matrices are analyzed and the largest of th e three differences is minimized . Pocket Mill Speed 50 kHz repetition rate and the pulse area overlap is selected to most closely match the acceleration region cut depths to the desired depth. The beam on and beam off delays are not included in the simulation. Mixed on and beam off delays are included. Tolerance 2 sigma matrices are considered in the cut depth. The last component of the modification strategy is the cut tolerance. The user is able to indicate the desired tolerance level for each feature. The tolerance has different effects on different cut type depth constraints , and these constraints are listed in Table 7 6.

PAGE 174

174 Table 7 6 . Explanation of the effect of the tolerance selection on the depth constraint. Cut Type Depth Constraint(s) Through Cut Face Mill Pocket Mill 7 . 5 . 2 Optimization Strategies Modification of parameters for through cuts or facing operations does not require repeated simulation of the cut feature to determine a set of parameters that satisfies the desired cut depth criterion. Through cuts and face milli ng parameter modification routines only analyze the depth of cut for constant velocity cutting conditions. Any excessive ablation due to beam on/beam off commands, acceleration regions, or cut path overlap either aids in producing a through feature, or for facing operations they are outside of the area of interest. However, for pocket milling those excessive ablation regions are of interest because they can exceed the desired depth of cut; therefore, it is necessary to simulate every parameter combination of interest to ensure that the excessive ablation zones do not exceed the desired depth. The problem with simulating every set of parameters is that it is computationally intense, and the simulation/modification routine can take hours or days to finish. Th is situation requires the use of an optimization routine to minimize the number of necessary simulations, while achieving the desired depth of cut.

PAGE 175

175 A general purpose optimization routine, like the built in Matlab ® function fmincon, attempts to find the con strained minimum of a nonlinear multivariable function. It is a gradient based optimizer that seeks out the minima by following the gradient/slope to the lowest point. Optimization routines like these work well when the system of interest is well defined b y a function, but they are less effective with highly non linear systems. The number of passes and attenuation level are both discrete values that result in non smooth mapping [64] . The inability to utilize discrete values decreases the accuracy of the optimization because approximating the system with non discrete values and then rounding to discrete values may shift the cut from a gentle to a strong ablation rate. Another optimizat ion routine is necessary to optimize the discrete variables. Genetic algorithms, such as Matlab ® ga function , do not analyze the gradient, but rather utilize a series of random initial points and tweak those points to find a solution. Each optimization i teration eliminates the less desirable ( higher fit) data points and creates new data points with slight random changes to the original parameters. This process is repeated until a solution is converged upon [65] . The function ga was implemented into the simulator routine and an optimal set of parameters was successfully determined; however, the highly non linear and non smooth acceleration zones, maximum sidewall angle, and part path overlap cause the system to run for 43 minutes before converging on the solution. The simulated feature was a relatively simple 100 µm x 200 µm rectangle, and with larger or more complex cut features the required optimization time is likely to increase exponentially; t herefore , a quicke r optimization routine is necessary.

PAGE 176

176 A solution to the above problems is to combine the rapid convergence characteristic of gradient based routines and the discrete variable handling properties of g enetic optimization algorithms. First, an initial point is selected based on the estimated number of necessary passes for a preselected pulse spacing. Next, the depth of cut differential is calculated for the init ial point and all of the neighbor points . The respective differentials are then analyzed to determine the lowest fit data points. The lowest fit point then becomes the new parent and the process is repeated. If the immediately surrounding data points are found to have higher fitness than the parent, then an additional set of simulations are run for parame ter combinations neighboring the children data points. This last step is to ensure that the optimization routine is not converging on a local minimum. An example of this technique is demonstrated in Figure 7 1 6 . Figure 7 1 6 (A) shows the complete 3D visualization of the depth differential for a fluence of 2.1 J/cm 2 , a feedrate of 50 mm/s, pulse frequency of 50 kHz, and a desired cut depth of 50 µm. An initial pulse spacing of 1 µm is selected in order to minimize the acceleration zone effects, and the optimization routine estimates the initial number of passes as 15 . It is important to note that this graph is not available at the beginning of the modification routine, but it is provided in this example to aid the read in understanding the optimization surface. Figure 7 1 6 ( B ) shows the iteration process as the processing parameters of 16 passes and 0.5 µm pulse spacing is converged upon by the optimization routine.

PAGE 177

177 A B Figure 7 1 6 . Plot of the absolute difference between the simulated cut depth of a given combination of pulse spacing/passes and the desired depth of cut. A) Shows a plot of all the entire parameter field, and B) demonstrates the technique for iterating to the most app ropriate set of processing parameters.

PAGE 178

178 7 . 5 . 3 Parameter Modification Routine Verification To verify the parameter modification routine three sets of cuts were simulated, modified, and then machined. The first two data sets, shown in Figure 7 17 and 7 18 are 100 µm x 300 µm through, face, and pocket cuts in R plane sapphire. Each cut parameter was repeated 5 times to produce the 95% uncertainty bars shown on the graphs. The cuts were performed on separate dates , and due to power fluctuations of the max laser output the computed parameters vary for each data set. The results of the cut tests are shown in Table 7 7. Each data set was measured using the SWLI with 20X scanning depths. Table 7 7 . List of the average difference in cut depth between the desired and measured, and the average time to complete the cut. Data Set # Cut Type Modification Strategy Average Difference Average Time (s) 1 Through Speed 5% 1.51 Mixed 11% 1.59 Tolerance 3% 5.13 Face Speed 1% 4.00 Mixed 5% 4.44 Tolerance 5.00 Pocket Speed 5% 4.00 Mixed 16% 2.69 Tolerance 14% 2.41 2 Through Speed 1% 1.14 Mixed 5% 1.25 Tolerance 34% 4.81 Face Speed 6% 2.71 Mixed 11% 3.69 Tolerance 10% 3.51 Pocket Speed 6% 3.08 Mixed 8% 3.15 Tolerance 9% 3.23

PAGE 179

179 A B C Figure 7 17 . Cut depths for the different parameter modification routines . A) M inimum depth of cut for through features, B ) average depth of c ut for facing operations, and C ) maximum depth of cut for pocket features. 0 10 20 30 40 50 60 10 15 20 25 30 50 Measured Depth ( µm) Desired Depth ( µm) Speed Mixed Tolerance 0 10 20 30 40 50 60 70 10 15 20 25 30 50 Measured Depth ( µm) Desired Depth ( µm) Speed Mixed Tolerance 0 10 20 30 40 50 60 70 10 15 20 25 30 50 Measured Depth ( µm) Desired Depth ( µm) Speed Mixed Tolerance

PAGE 180

180 A B C Figure 7 18 . Cut depths for the different parameter modification routines . A) M inimum depth of cut for through feature s , B ) average depth of cut for facing operation s , and C ) maximum depth of cut for pocket features. 0 10 20 30 40 50 60 5 10 20 30 40 Measured Depth ( µm) Desired Depth ( µm) Speed Mixed Tolerance 0 10 20 30 40 50 5 10 20 30 40 Measured Depth ( µm) Desired Depth ( µm) Speed Mixed Tolerance 0 10 20 30 40 50 5 10 20 30 40 Measured Depth ( µm) Desired Depth ( µm) Speed Mixed Tolerance

PAGE 181

181 Table 7 7 shows that the modification routine in data set #2 more closely matches the overall desired depth of cut. This is in part due to the integration of the depth tolerance value, which was set at 10% for th is verification test. The through cut feature achieves a full feature hole in all but the speed strategy, and it is only off by 1%. All of the face cuts fall within the tolerance range except for the mixed strategy, which is also 1% outside of the toleranc e range. Next , all of the pocket cuts are less than the maximum depth tolerance and within the acceptable range of 10%. Finally, t he table also shows that the speed cut consistently finished before the other two strategies, except for the speed pocket feature in data set #1. One problem that plagued the verification of the sapphire MRR model is the variation in the laser output powe r. An example of this is that during the verification cuts in Figure 7 17 the power dropped from an initial measured value of 2.36 watts to 2.18 watts at the end of the verification cuts. The power drop of ~8% affects the depth of cut and may account for the less than desired depths of cut. The third verification is the modification of parameters for the crossing double rectangles feature . A mixed modification strategy was used to machine 20, 25, and 30 µm deep pockets. Figure 7 19 , 7 2 0 , and 7 2 1 show the cut profile for the three respective depths o f cut, and Table 7 8 shows the maximum depth of cut for each commanded depth. The features were measured using the same SWLI settings as the previous experiment.

PAGE 182

182 A B Figure 7 19 . SWLI measurements for a 20 µ m deep cut in R plane sapphire. A ) A 3D rendering of the cut floor, and B ) shows a contour plot.

PAGE 183

183 A B Figure 7 2 0 . SWLI measurements for a 25 µ m deep cut in R plane sapphire. A ) A 3D rendering of the cut floor, and B ) shows a contour plot.

PAGE 184

184 A B Figure 7 2 1 . SWLI measurements for a 30 µm deep cut in R plane sapphire. A) A 3D rendering of the cut floor, and B ) shows a contour plot.

PAGE 185

185 Table 7 8 . Maximum depth of cut for Figures 6 16, 6 17, and 6 18. Commanded Depth ( µ m) Measured Maximum Depth ( µ m) 20 20.7 25 25.0 30 30.0 7 . 6 Added Passes Correcting the cut floor of the crossing double rectangles in Figures 7 19, 7 20, and 7 21 is of upmost importance if the cut floor depth needs to be consistent. Although this cut geometry has larger depth variations than most cuts, it is still useful in demonstrating the effectiveness of parameter selection and adding passes. The maximum depth of cut for a desired depth of 30 µm, 35 µm, and 40 µm is given in Table 7 9, along with the average depth. Figures 7 2 2 , 7 2 3 , and 7 24 show a comparison of the simulated and measured depth of cut profile for the 40 µm cuts. Table 7 9 . Comparison of Desired depth of cut to the maximum and average cut depth. Desired Depth (µm) Maximum Depth (µm) Average Depth (µm) Speed 30.0 27.0 18.6 Mixed 30.0 32.5 23.5 Tolerance 30.0 30.2 22.5 Speed 35.0 33.7 23.2 Mixed 35.0 31.7 23.6 Tolerance 35.0 38.8 27.8 Speed 40.0 37.1 26.4 Mixed 40.0 39.2 29.8 Tolerance 40.0 39.5 29.6

PAGE 186

186 A B Figure 7 2 2 . Added passes depth of cut profile for a speed modification strategy. A) Shows simulated cut, and B) measured cut.

PAGE 187

187 A B Figure 7 2 3 . Shows simulated cut, and B) measured cut.

PAGE 188

188 A B Figure 7 2 4 . Shows simulated cut, and B) measured cut.

PAGE 189

189 7 . 7 Chapter Summary In this chapter the characterization of the laser sapphire interaction was performed. F irst an an alytical analysis was performed to determine that the UF laser system sapphire combination is considered an ultrashort interaction. Secondly a Monte Carlo analysis was performed to determine the slope and intercepts for different pulse spacing and number o f passes combinations. It was determined that 96.6% of the data points used in the analysis fall within the +/ 2 sigma predicted depths of cut. To verify the capability of the parameter modification routine, two different verification sets were machined and their respective measured depths were compared to the desired depths. It was found that the resulting depth of cut was typically within 10% of the desired depth. Next, the parameter modification routine was tested on a double overlapping set of rectang capability to predict the necessary processing parameters to produce a certain cut depth even when the cut floor was not consistently deep. Lastly, the added passes routine was te sted on the double overlapping rectangles to determine if the routine could increase the uniformity of the cut floor. The routine was successful in modifying the processing parameters and adding passes to make the cut floor more uniform.

PAGE 190

190 CHAPTER 8 CONCL USION This dissertation focused on developing a computer aided manufacturing software to reduce the complexity of operating an ultrashort pulsed laser. Additionally, it focused on the characterization of a picosecond pulsed laser micromachining station. Ul trashort pulsed lasers are a promising technology because of their ability to machine a wide array of materials, but the lack of knowledge on how to properly operate the lasers has hindered their acceptance into industry. The hope is that t his research wil l allow the ultrashort pulsed laser to become more useful to the end user, and to reduce the cost of machining components by: reducing setup time, reducing processing time, reducing part failure due to incorrect parameters, and reduce geometric differences between the desired component and realized component. The primary contributions of this research are as follows : 1. A Matlab ® based simulation program is complete that allows the end user to input a part path, simulate the machined feature, and analyze a variety of graphical and numerical results. 2. An automated parameter selection routine and part path correction routine is avai lable in the simulator as well. These routines allow the end user to select the desired cut feature, cut type, and depth tolerance for each respective feature. The program is capable of modifying the cut parameters to achieve the desired cut depth, and if desired additional passes can be added to decrease cut floor irregularities.

PAGE 191

191 3. The relationship between an Oxford J 355PS laser micromachining station and R plane sapphire is established, along with the uncertainty of the relationship. Several additions to t he current simulation/modification program can be added in future iterations. Two of the most important additions to the machining process are further analysis of the physics behind the production of a non zero sidewall angle, and modeling of the thermal e nergy involved in cutting sapphire. The first addition requires an analysis into the driving physics behind the sidewall angle. It is important to determine if the beam profile is producing the sidewall angle, or if it is inherent with any beam shape. Steps to reduce the sidewall angle can i nclude, but are not limited to, installation of g oniometer s to compensate for the sidewall angle, and optimizing the beam shape to minimize the angle. Both of these steps require additional equipment to be installed for the techniques to be effective. Modi fication of the beam shape may be as simple as installing new focusing optics, but it may also require acoustic or electronic actuators to accomplish this task. The addition of goniometers changes the classification of the current laser setup from a 3 axis system to a 5 axis system. Although this method would almost certainly reduce or eliminate the sidewall angle, it also introduces a great deal of complexity in modeling and controlling the additional stages. A side benefit of goniometers is the ability to drill non perpendicular holes in a variety of materials. The second addition that would yield large dividends in simulating the ablation process is a high level model of both the laser material interaction, and the subsequent diffusion of thermal energy i nto the workpiece. This will aide in the machining of thin

PAGE 192

192 substrates, reduction of the HAZ, and minimize thermal strain during the machining process. Currently the fluence level is selected right above the ablation threshold, but it may be overly conserv ative. Features have been machined in 500 µm thick R plane wafers with fluence levels three times greater than the threshold without failure of the substrate. If these fluence levels are found to be acceptable in thinner substrates they can further reduce the time require to machine individual features, and reduce the overall power consumption required for each feature. Lastly, further research into annealing of machined feature in sapphire is necessary. Previous research has shown the capability of shrink ing or eliminating cracks and small features in sapphire wafers. If this can be realized with laser machined components then the surface roughness of the finished component may become inconsequential. Additionally, it may reduce or eliminate any thermal mo difications to the features, such as micro cracks. This post processing technique shows high potential for increasing the viability of ultrashort pulsed lasers as an industrial fabrication tool.

PAGE 193

193 LIST OF REFERENCES [1] C. H. Townes, How th e Laser Happened, New York: Oxford University Press, Inc., 1999. [2] G. Chryssolouris, Laser Machining Theory and Practice, New York: Springer Verlag, 1991. [3] N. Dahotre, Laser Machining of Advanced Materials, Leiden: CRC Press/Balkema, 2011. [4] D. Bauerle, Laser Processing and Chemistry, London: Springer, 2011. [5] J. Meijer, "Laser Machining by Short and Ultrashort Pulses, State of the Art and New Opportunities in the Age of the Photons," CIRP Annals Manufacturing Technology, vol. 51, no. 2, pp. 531 550, 2002. [6] University Wafer, "Sapphire," University Wafer, 2013. [Online]. Available: https://order.universitywafer.com/default.aspx?cat=Sapphire. [Accessed 10 December 2013]. [7] "Classic Silica Disc," Aerogel Technologies, LLC., 2 013. [Online]. Available: http://www.buyaerogel.com/product/classic silica disc/. [Accessed 10 December 2013]. [8] "NON Porous High Alumina Ceramic Sheet," McMaster Carr, 2013. [Online]. Available: http://www.mcmaster.com/#8462k21/=q4fy9x. [Accessed 10 D ecember 2013]. [9] "Multipurpose 304 Stainless Steel Foil," McMaster Carr, 2013. [Online]. Available: http://www.mcmaster.com/#3254k93/=q4fyqn. [Accessed 10 December 2013]. [10] "Ultra Corrosion Reistant 1100 Aluminum," McMaster Carr, 2013. [Online]. Available: http://www.mcmaster.com/#9060k16/=q4fz1v. [Accessed 10 December 2013]. [11] Epilog Lasers, "Epilog Zing Laser Systems," 2014. [Online]. Available: http://www.epiloglaser.com/products/zing laser series.htm. [Accessed 7 June 2014]. [12] IPG Ph otonics, "Industrial Fiber Lasers for Material Processing," 2012. [Online]. Available: http://www.ipgphotonics.com/Collateral/Documents/English US/HP_Brochure.pdf. [Accessed 7 June 2014].

PAGE 194

194 [13] "Laser Interface+," Universal Laser Systems, 2013. [Online]. Available: http://www.ulsinc.com/products/features/laser interface/. [Accessed 10 December 2013]. [14] "SigmaNEST for Laser Cutting Machines," Sigmanest, 2013. [Online]. Available: http://www.sigmanest.com/en us/machines/laser/. [Accessed 10 December 201 3]. [15] R. Herrmann, "Ultrashort Pulse Laser Ablation of Silicon: an MD Simulation Study," Applied Physics A, vol. 66, no. 1, pp. 35 42, 1998. [16] L. Orazi, "An Automated Procedure for Material Removal Rate Prediction in Laser Surface Morphology," I nternational Journal of Advanced Manufacturing Technology, vol. 46, pp. 163 171, 2010. [17] Vero Software, "Surfcam 2 Axis," [Online]. Available: http://www.surfcam.com/surfcam2axis. [Accessed 7 June 2014]. [18] CNC Software, Inc., "Mastercam Mill," [Online]. Available: http://www.mastercam.com/Products/Mill/Default.aspx. [Accessed 7 June 2014]. [19] G. S. Duncan, "Stability Lobe Uncertainty," in Proceedings of American Society for Precision Engineering Annual Meeting , 2005. [20] M. Traverso, R. Zapata, T. Schmitz and A. Abbas, "Optimal Experimentation for Selecting Stable Milling Parameters: A Bayesian Approach," in ASME 2009 International Manufacturing Science and Engineering Conference , West Lafayette, 2009. [21] J. Black, DeGarmo's Material s & Processes in Manufacturing, Hoboken: John Wiley & Sons, Inc., 2008. [22] A. Einstein, "Zur Quantentheorie der Strahlung," Physikalische Zeitschrift, vol. I, no. 18, pp. 121 128, 1917. [23] M. Bertolotti, The History of the Laser, London: Institut e of Physics Publishing, 1999. [24] P. W. Milonni, Lasers, New York: Wiley Interscience, 1988. [25] J. T. Verdeyen, Laser Electronics, Upper Saddle River: Prentice Hall, 1995. [26] N. B. Dahotre, Laser Fabrication and Machining of Materials, New York: Springer Science + Business Media, LLC, 2008.

PAGE 195

195 [27] M. Rose, "A History Of The Laser: A Trip Through The Light Fantastic," May 2010. [Online]. Available: http://www.photonics.com/Arti cle.aspx?AID=42279. [Accessed 20 November 2013]. [28] R. Paschotta, Encyclopedia of Laser Physics and Technology, Berlin: Wiley VCH, 2008. [29] D. Ashkenasi, "Laser Processing of Sapphire with Picosecond and Sub Picosecond Pulses," Applied Surface Sci ence, vol. 120, no. 1 2, pp. 6 80, 1997. [30] Coherent, Operator's Manual Talisker Ultra Industrial Picosecond Laser System, Santa Clara, 2010. [31] I. Ross, P. Matousek and M. Towrie, "The Prospects for Ultrashort Pulse Duration and Ultrahigh Intensity Using Optical Parametric Chirped Pulsed Amplifiers," Optics Communications, no. 144, pp. 125 133, 1997. [32] M. Lukac, "Ablation and Thermal Depths in VSP Er:YAG Laser Skin Resurfacing," Journal of the Laser and Health Academy, vol. 2010, no. 1, pp. 56 71, 2010. [33] B. H. Jared, "Application of Short Pulse Laser Systems for Micro Scale Processing," in SME MicroManufacturing , Minneapolis, 2013. [ 34] N. Rykalin, Laser Machining and Welding, Moscow: Mashinostroyenie Publishers, 1975. [35] A. Komashko, "Simulation of Material Removal Efficiency with Ultrashort Laser Pulses," Applied Physics A, vol. 69, no. 1, pp. S95 S98, 1999. [36] X. Liu, "l aser Ablation and Micromachining with Ultrashort laser Pulses," IEEE Journal of Quantum Electronics, vol. 33, no. 10, pp. 1706 1716, 1997. [37] I. V. Hertel, "On the Physics of Material Processing with Femtosecond Lasers," RIKEN Review, vol. 2, no. 32, pp. 23 30, 2001. [38] A. Kaiser, "Microscopic Processes in Dielectrics Under Irradiation by Subpicosecond Laser Pulses," Physical Review B, vol. 61, no. 17, pp. 437 450, 2000. [39] B. Chichkov, "Femtosecond, Picosecond, and Nanosecond Laser Ablation of Solids," Applied Physics A, vol. 63, no. 2, pp. 109 115, 1996.

PAGE 196

196 [40] A. Heidt, Z. Li and J. Sahu, "100 kW peak power picosecond thulium doped Fiber Amplifier System Seeded by a Gain Switched Diode Laser at 2 um," Optics Letter, vol. 38, no. 10, pp. 1615 1617, 2013. [41] C. Heese, A. Oehler, L. Gallmann and U. Keller, "Hi gh Energy Picosecond Nd:YVO4 Slab Amplifier for OPCPA Pumping," Applied Physics B, no. 103, pp. 5 8, 2011. [42] N. H. Rizvi, "Production of Novel 3D Microstructures Using Excimer Laser Mask Projection Techniques," in Proceedings of SPIE: Deisgn, Test, a nd Microfabrication of MEMS and MOEMS , Paris, 1999. [43] L. J. Lewis, "Laser Ablation with Short and Ultrashort Laser Pulses: Basic Mechanisms from Molecular Dynamics Simulations," Applied Surface Science, vol. 255, no. 10, pp. 5101 5106, 2009. [44] T. A. Davis, "Effect of Laser Pulse Overlap on Machined Depth," Transactions of NAMRI/SME, vol. 38, pp. 291 298, 2010. [45] C. Hayden, "A Simple Three Dimensional Computer Simulation Tool for Predicting Femtosecond Laser Micromachined Structures," Journal of Micromechanical and Microengineering, vol. 20, no. 2, pp. 1 11, 2010. [46] E. Hirleman, "Intensity Distribution Properties of a Gaussian Laser Beam Focus," Applied Optics, vol. 17, no. 21, pp. 3496 3499, 1978. [47] Oxford Lasers Ltd., J 35 5PS System Operation Manual, Oxford, 2010. [48] R. Hibbeler, Engineering Mechanics Dynamics, Upper Saddle River: Pearson Prentice Hall, 2007. [49] Aerotech, "PRO115 Series Stage User Manual," Pittsburg, 2010. [50] Aerotech, "Computing Resolution, " [Online]. Available: http://www.aerotech.com/media/247161/section%207_computing%20resolution.p df. [Accessed 19 07 2014]. [51] OSI Optioelectronics, "Tetra Lateral PSD," [Online]. Available: http://www.osioptoelectronics.com/standard products/silicon ph otodiodes/position sensing detectors/tetra lateral psds.aspx. [Accessed 18 May 2014]. [52] Thor Labs, Inc., "Scanning Slit Optical Beam Profilers," [Online]. Available: http://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=804. [Accessed 24 May 2014].

PAGE 197

197 [53] Del Mar Ventures, "Variable Attenuators for High Power Lasers," [Online]. Available: http://www.sciner.com/Opticsland/high power_laser_attenuation.htm. [Accessed 24 May 2014]. [54] V. Lindroos, M. Tilli and A. Lehto, Handbook of Silicon Based MEMS Materials And Technologies, Oxford: Elsevier, 2010. [55] L. Smith, H. Karim, S. Etemad and W. Pfefferle, The Gas Turbine Handbook, US Department of Energy National Energy Technology La boratory, 2006. [56] K. Williams, K. Gupta and M. Wasilik, "Etch Rates for Micromachining Processing Part II," Journal of Microelectromechanical Systems, vol. 12, no. 6, pp. 761 778, 2003. [57] T. Eisel, S. Feigl, J. Bremer, G. Burghart, T. Koettig a nd F. Haug, "Cooling of Electrically Insulated High Voltage Electrodes Down to 30 mK: Dynamic Measurements," in DKV Tagung Magdeburg , Geneva, 2010. [58] F. Wooten, Optical Properties of Solids, New York and London: Academic Press, 1972. [59] D. E.R., Sapphire: Material, Manufacturing, Applications, New York: Springer Science + Business Media, LLC, 2009. [60] S. M. Edlou, A. Smajkiewicz and G. A. Al Jumaily, "Optical Properties and Environmental Stability of Oxide Coatings Deposited by React ive Sputtering," Applied Optics, vol. 32, no. 28, pp. 5601 5605, 1993. [61] J. A. Harrington, B. L. Bobbs and M. Braunstein, "Ultraviolet Visible Absorption in Highly Transparent Solids by Laser Calorimetry and Wavelength Modulated Spectroscopy," Applie d Optics, vol. 17, no. 10, pp. 1541 1546, 1978. [62] H. Phillipp and E. Taft, "Optical Constants of Silicon in the Region 1 to 10 ev," Physical Review, vol. 120, no. 1, pp. 37 38, 1960. [63] I. V. Hertel, R. Stoian and D. Ashkenasi, "On the Physics o f Material Processing with Femtosecond Lasers," Riken Review, no. 32, pp. 23 30, 2001. [64] The Mathworks, Inc., "fmincon," 2014. [Online]. Available: http://www.mathworks.com/help/optim/ug/fmincon.html. [Accessed 16 June 2014]. [65] The Mathworks, In c., "How the Genetic Algorithm Works," 2014. [Online]. Available: http://www.mathworks.com/help/gads/how the genetic algorithm works.html. [Accessed 16 June 2014].

PAGE 198

198 BIOGRAPHICAL SKETCH Daniel Blood attended Valparaiso University (VU) in 2006, and he received his Bachelor of Science in 2010. He then moved to Florida in the summer of 2010 to attend the mechanical engineering graduate school at the University of Florida (UF). In 2012 he earned his Master of Science degree , and is continui ng to work towards his goal of a PhD. Daniel Blood has worked on many projects in his time at VU. H e was the president of the Valparaiso Independent Robotics Team, technical chair of the local chapter of Engineers Without Borders, and also headed the desig n and fabrication of a Braille bible press. His most notable project is the high temperature solar furnace at VU, where he headed the design and oversaw the fabrication of the solar concentrator. He is the recipient of several awards, including the Edgar J . Luecke Z* Award. Daniel's dissertation, Simulation, Part Path Correction, a nd Automated Process Parameter Selection f or Ultrashort Pulsed Laser Micro m achining o f Sapphire , was supervised by Dr. Sheplak of the University of Florida and Dr. Schmitz of the University of North Carolina Charlotte.