Laser Micro-processing of Silicon using Nanosecond Pulse Shaped Fibre Laser at 1 µm Wavelength Institute for Manufacturing Department of Engineering Clare Hall This dissertation is submitted for the degree of Doctor of Philosophy by: Kun Li December, 2011 _____________________________________________________________________ ii Declaration This dissertation is the result of my own work and includes nothing which is the outcome of work done in collaboration except where specifically indicated in the text. Furthermore, the dissertation is not substantially the same as I have submitted or will be submitting for a degree or diploma or other qualification at any other University. Signed: Date: This thesis contains 48,479 words and 122 figures, which do not exceed the word limit for the respective Degree Committee. _____________________________________________________________________ iii Abstract Processing of Si in the semiconductor and solar cell industry has been dominated by the Diode Pumped Solid State (DPSS) Ultraviolet (UV) laser. Recent advances in laser source technology have produced fibre lasers with Master Oscillator Power Amplifier (MOPA) architectures that offer high repetition rates, high operational efficiencies, and pulse modulation controls exceeding those of typical Q-switched DPSS lasers. The aim of this research is to investigate 1 µm fibre laser machining of Si with a view to identifying the influential laser parameters for optimum processing of high quality, high efficiency micro drilling and surface texturing applications. A secondary aim is to develop a greater understanding of the laser material interactions and material removal mechanism when using fast rise-time nanosecond laser pulse envelopes. The IR fibre laser was able to perform percussion drilling and single pulse machining on the polished Si over a range of intensities up to 1.22 GW/cm2. With the optimum parameters, the micro-sized holes generated by the IR laser have a well defined edge, no heavy recast and no cracks. With a pulse shape of fast rise time (<7.5 ns for a 10-90% rise in signal), a high front peak power zone (approaching 14 kW) and an energetic long tail (40-180 ns), the absorption coefficient of Si at IR wavelength increased dramatically with time and temperature due to the fact that the liquid Si has a metal like absorption behavior. As a result, Si was quickly melted and the rest of pulse energy was able to remove the liquid Si effectively. The machining process left a limited amount of resolidified melt droplets and vapor condensates, which could be washed off ultrasonically. The drilling process was energy efficient when melt expulsion dominated the machining mechanism (0.08-0.2 mJ pulse energy depending on the pulse durations). The low energy pulse (~0.2 mJ) can achieve similar depth as the high energy pulse (~0.7 mJ), so high repetition rates of 100 kHz can be used to instead of 25 kHz, resulted in high processing speed. _____________________________________________________________________ iv In addition, by comparing the single pulse machining with the state of the art UV laser, the IR fibre laser machined deeper features and better surface finish in the pulse energy region of >0.07 mJ. With the pulse shaping capability, the material properties can be varied and the wavelength factor can be minimized. The results suggest that applications like microvia drilling can now be carried out with the more flexible and low cost IR fibre laser. The increased repetition rates of fibre laser can increase production speed to satisfy the needs of drilling ~10 thousands holes per second, required by the modern semiconductor and solar cell production. The shortened optical penetration length of 1 µm wavelength laser on Si with increasing temperature and sufficient thermal diffusion length resulted from the asymmetrical fibre laser pulse and the dynamic properties of Si produced a thick liquid layer. A one- dimensional heat conduction model based on the surface heating source predicted that this superheated liquid layer was able to stay above 4706 K (0.905 times the thermal critical temperature 5200 K of Si) for longer than 70 ns to induce explosive boiling. This proposed material removal mechanism was also confirmed by the shadowgraph images, showing particulates ejection lasting up to ten microseconds after the laser pulse. The estimated hole depth based on the explosive boiling alone were different from the measured ones at varying peak power densities (<1.22 GW/cm2) but fixed pulse duration (200 ns), since Si was removed by a mixture of mechanisms. With varying pulse durations (40-200 ns) but fixed peak power density (~0.63 GW/cm2), the estimated depth based on the explosive boiling was in close agreement with the measured ones (6% difference on average). The SEM images at this power density showed a micron- /submicron-sized debris field, which was also observed with the explosive boiling in the past. Although the improved quality of Si machining was demonstrated with the 1 µm MOPA based fibre laser, the setup of this system was only applicable to surface texturing, blind holes and through holes of less than 100 µm in depth. Further research is required to demonstrate the capability of more energetic pulse with higher peak power and large pulse duration range to explore more machining options. _____________________________________________________________________ v Acknowledgements Firstly, I would like to thank my supervisor, Dr. Bill O’Neill for his guidance and support throughout the whole research. His ideas and thoughts inspire me from time to time. The entire learning experience will be so valuable and rewarding in my future career development. In addition, Dr. Jack Gabzdyl from SPI Lasers helped me understand the importance of my research from an industrial point of view. He also supplied different types of material for my research. I would like to thank the colleagues from the Centre for Industrial Photonics; Dr. Martin Sparkes for his help with the thesis proof reading, knowledgeable discussion and laser safety measures, Dr. Ali Khan for his expertise in high speed imaging techniques and thermal modeling, Mr. Arthur Nightingale, Mr. Stuart Fordham and Mr. Simon Ford for their excellent engineering works and helps in the laboratory. The encouragement and motivated discussion from the rest of the team are highly appreciated; Dr. Rocco Lupoi, Dr. Andrew Cockburn, Mr. Krste Pangovski, Mr. Sina Habibi and Dr. Chris Forman. I would also thank my sponsors Engineering and Physical Sciences Research Council (EPSRC) and SPI Lasers for their financial support. I’m very grateful of my family who have supported me and brought me happiness throughout the difficult times. Kun Li _____________________________________________________________________ vi Table of Contents Declaration .......................................................................................................................... ii Abstract .............................................................................................................................. iii Chapter 1 Introduction ........................................................................................................ 1 1.1 Background ............................................................................................................... 1 1.2 Research aim and research questions ........................................................................ 2 1.3 Research methodology .............................................................................................. 3 1.3.1 Phase 1 ................................................................................................................ 4 1.3.2 Phase 2 ................................................................................................................ 4 1.3.3 Phase 3 ................................................................................................................ 5 1.4 Research scope .......................................................................................................... 5 1.4.1 Chapter 2 – Literature Review ............................................................................ 6 1.4.2 Chapter 3 – Experimental systems, measurement tools and methodology ........ 6 1.4.3 Chapter 4 – Fibre laser machining characterization of silicon ........................... 7 1.4.4 Chapter 5 – Performance comparison between IR and UV lasers ..................... 7 1.4.5 Chapter 6 – Theoretical analysis ........................................................................ 8 1.4.6 Chapter 7 – Conclusion ...................................................................................... 8 Chapter 2 Literature Review ............................................................................................... 9 2.1 Introduction ............................................................................................................... 9 2.2 Silicon........................................................................................................................ 9 2.2.1 Basic properties and applications ....................................................................... 9 2.2.2 Energy transfer in Si ......................................................................................... 14 2.2.2.1 Optical coupling ......................................................................................... 15 2.2.2.2 Thermal coupling ....................................................................................... 21 2.3 Laser micromachining ............................................................................................. 24 2.3.1 Laser parameters ............................................................................................... 26 2.3.1.1 Wavelength ................................................................................................ 26 2.3.1.2 Pulse energy and pulse shape ..................................................................... 29 2.3.1.3 Beam size ................................................................................................... 33 _____________________________________________________________________ vii 2.3.2 Machining mechanism ...................................................................................... 35 2.3.2.1 Melt expulsion ........................................................................................... 36 2.3.2.2 Vaporization ............................................................................................... 36 2.3.3 Machining processes ......................................................................................... 44 2.3.3.1 Solid state laser .......................................................................................... 47 2.3.3.2 Ultrafast laser ............................................................................................. 52 2.3.3.3 Hybrid processing of Si ............................................................................. 56 2.3.4 Machined depth ................................................................................................ 57 2.3.5 Fiber laser ......................................................................................................... 58 2.3.5.1 Single mode fibre ....................................................................................... 60 2.3.5.2 Advantages ................................................................................................. 61 2.4 Analytical models .................................................................................................... 62 2.4.1 Thermal conduction model ............................................................................... 63 2.5 Summary ................................................................................................................. 74 Chapter 3 Experimental systems, measurement tools and methodology .......................... 76 3.1 Laser systems .......................................................................................................... 76 3.1.1 Fibre laser ......................................................................................................... 76 3.1.2 Q-switched UV laser ........................................................................................ 78 3.2 Scanning systems .................................................................................................... 80 3.3 Machining systems .................................................................................................. 82 3.4 Laser beam diagnostics ........................................................................................... 85 3.4.1 Spatial beam profile measurements .................................................................. 85 3.4.2 Temporal pulse profile measurements .............................................................. 86 3.4.3 Average power measurements .......................................................................... 88 3.5 Sample management and focus test ........................................................................ 92 3.6 Sample Analysis ...................................................................................................... 93 3.6.1 White light interferometer ................................................................................ 93 3.6.1.1 Analysis Options ........................................................................................ 94 3.6.2 Scanning electron microscopy (SEM) .............................................................. 96 3.6.2.1 Operational practice ................................................................................... 96 _____________________________________________________________________ viii 3.6.3 High speed imaging .......................................................................................... 97 3.6.3.1 Triggering .................................................................................................. 97 3.6.3.2 Set up ......................................................................................................... 97 3.6.4 Shadowgraph imaging ...................................................................................... 98 Chapter 4 Fibre laser machining characterization of silicon .......................................... 101 4.1 Introduction ........................................................................................................... 101 4.2 Percussion drilling ................................................................................................. 101 4.2.1 PRF analysis ................................................................................................... 102 4.2.2 Peak power analysis........................................................................................ 106 4.2.3 Pulse shape analysis........................................................................................ 110 4.2.4 Cooling cycle analysis .................................................................................... 112 4.2.5 Discussion ....................................................................................................... 114 4.2.5.1 Thermal coupling ..................................................................................... 114 4.3 Single pulse interactions........................................................................................ 115 4.3.1 Constant pulse duration .................................................................................. 115 4.3.2 Constant pulse peak power ............................................................................. 120 4.3.3 Constant pulse energy ..................................................................................... 123 4.3.4 Discussion ....................................................................................................... 126 4.3.4.1 Machined quality and efficiency .............................................................. 126 4.3.4.2 Thermal diffusion..................................................................................... 127 4.4 Summary ............................................................................................................... 128 Chapter 5 Performance comparison between IR and UV lasers ..................................... 130 5.1 Introduction ........................................................................................................... 130 5.2 Performance comparison between IR and UV lasers ............................................ 130 5.2.1 Single pulse machined depth and efficiency .................................................. 131 5.2.2 Single pulse machined mechanism and finish quality .................................... 138 5.2.3 Discussion ....................................................................................................... 139 5.3 Comparison between fibre lasers of different M2 ................................................. 140 5.3.1 Single pulse machined depth and efficiency .................................................. 141 5.3.2 Single pulse machined mechanism and finish quality .................................... 143 _____________________________________________________________________ ix 5.3.3 Discussion ....................................................................................................... 146 5.4 Summary ............................................................................................................... 148 Chapter 6 Theoretical analysis ........................................................................................ 149 6.1 Introduction ........................................................................................................... 149 6.2 Fourier heat conduction ......................................................................................... 150 6.3 Volume heating source .......................................................................................... 153 6.4 Surface heating source .......................................................................................... 157 6.5 Explosive boiling................................................................................................... 162 6.5.1 Constant pulse duration .................................................................................. 163 6.5.2 Evidence of explosive boiling – Shadowgraph .............................................. 167 6.5.3 Constant peak power density .......................................................................... 169 6.5.4 UV laser .......................................................................................................... 173 6.6 Plasma plume expansion ....................................................................................... 176 6.7 Summary ............................................................................................................... 180 Chapter 7 Conclusion ...................................................................................................... 182 7.1 Fibre laser machining characterization of Si ......................................................... 182 7.2 Performance comparison between IR and UV lasers ............................................ 183 7.3 Theoretical analysis ............................................................................................... 184 7.4 Suggestion for future work .................................................................................... 187 Appendix A Solar cell market trend, production volume, and module technologies ..... 188 Appendix B List of basic laser parameters ..................................................................... 191 Appendix C Si properties ................................................................................................ 193 References ....................................................................................................................... 194 _____________________________________________________________________ x Nomenclature 𝐴 = absorbed radiation /absorptivity (%) 𝑅 = reflected radiation /reflectivity (%) 𝑇� = transmitted radiation (%) 𝐷 = heat diffusivity (cm2/s) 𝐹 = laser fluence (J/cm2) ℎ = Planck’s constant (eV·s) 𝑘 = Boltzmann constant (J·K-1) 𝑙� = thermal diffusion length (µm) 𝑙� = optical penetration depth (µm) 𝐿� = latent heat of melting (J/g) 𝐿� = latent heat of vaporization (J/g) 𝑛 = complex refractive index 𝑛� = the imaginary part of the refractive index 𝑛� = laser induced plasma ion density (per cm 3) 𝑄 = the average charge of the plasma (C) 𝑇 = temperature (K) 𝑇� = initial temperature of the material (K) 𝑇� = surface temperature (K) 𝑇� = melting temperature of Si (K) 𝑇� = boiling temperature of Si (K) 𝑇�� = thermodynamic critical temperature of Si (K) 𝑇� = plasma temperature (K) 𝛼 = optical absorption coefficient (cm-1) 𝛼� = absorption coefficient of the plasma (cm -1) 𝜅 = thermal conductivity (J/ cm·s·K) 𝜆 = the wavelength (µm) 𝜌 = density of solid (g/cm3) _____________________________________________________________________ xi 𝐶� = heat capacity of solid (J/g·K) 𝜏 = full duration of the asymmetrical pulses of the fibre laser (ns) 𝜏� = diffusion time (ns) 𝜔 = angular frequency (rad/s) 𝜁 = time span of the thermal diffusion (ns) 𝑣 = frequency of the laser light (s-1) 𝑧 = distance into the Si substrate measured from the surface with a direction perpendicular to the surface (µm) 𝑍 = machined depth into the Si substrate measured from the surface with a direction perpendicular to the surface (µm) _____________________________________________________________________ xii List of Figures Figure 1.1: Research methodology used in this thesis. ....................................................... 4 Figure 2.1: Si nanowire: (A) cross-sectional SEM image of vertically grown Si nanowires off of a Si(111) substrate, scale bar is 1 µm. (B) TEM image of SI nanowire surrounded in a conformal SiO2 coating, scale bar is 75 nm. (C) high resolution TEM image of a Si nanowire with an inner diameter that has been reduced to ~4.5 nm, scale bar is 4 nm (Goldberger, Hochbaum et al. 2006). ............................................................................... 11 Figure 2.2(a): Schematic and (b) SEM image of 19.8% efficient multicrystalline silicon solar cell with “honeycomb” surface texturing (Zhao, Wang et al. 1998). (c) The “inverted pyramid” texture on the top surface (Zhao, Wang et al. 1995). (d) A series of thin silicon strips (Weber, Blakers et al. 2004). ................................................................ 13 Figure 2.3: SEM image of the rear surface after laser processing and KOH etching for RISE-EWT solar cell (Harder, Hermann et al. 2009). ...................................................... 14 Figure 2.4: Optical penetration depth of mono-crystalline Si for various wavelength, measured at room temperature (Bauerle 2000). ................................................................ 16 Figure 2.5: Optical absorption coefficient of Si measured at wavelength of 1064 and 1152 nm as a function of temperature (Jellison and Lowndes 1982); (Weakliem and Redfield 1979); (Macfarlane, McLean et al. 1958). ......................................................... 17 Figure 2.6: Optical reflectivity mono-crystalline Si for various wavelength, measured at room temperature (Bauerle 2000). .................................................................................... 19 Figure 2.7: Time dependent transmission of a 1.152μm laser beam through Si during and after pulsed Ruby irradiation, where El is the pulsed Ruby laser energy density (Lowndes, Jellison et al. 1982). ........................................................................................ 20 _____________________________________________________________________ xiii Figure 2.8: Time dependent transmission T and probe reflection R of a 1.152 μm probe laser incident on Si during and after exposure to a pulsed laser excitation, 1 J/cm2, 8 ns pulse width and 455 nm wavelength (Aydinli, Lo et al. 1981)......................................... 21 Figure 2.9: Thermal diffusivity of Si from 300 to 1400 K (Shanks, Maycock et al. 1963). ........................................................................................................................................... 22 Figure 2.10: The effect of pulse width damage in Si using 355 nm wavelength DPSS laser, a greater amount of resolidified material resulted with a longer pulse width (Otani, Herbst et al. 2004). ............................................................................................................ 23 Figure 2.11: Commercially available laser systems versus optical penetration length and thermal diffusion length for mono crystalline Si at a temperature of 300K. With respect to the pulse width and wavelength various commercially available laser systems are drawn in the graph (Schoonderbeek, Kling et al. 2007). ............................................................. 24 Figure 2.12: Laser fired contacts technology simplifies the process sequence to implement the point contact pattern of passivated rear silicon solar cells. ....................... 25 Figure 2.13: Average etch depth as a function of intensity for 0.25 mm thick copper when operating the laser at 355, 532 and 1064 nm. ................................................................... 27 Figure 2.14: Two lines etched at different wavelengths. Horizontal line: 𝜆 =1047 nm (Nd:YLF, first harmonic); F=18 J/cm2. Vertical line: 𝜆 = 262 nm (Nd:YLF, fourth harmonic); F=3 J/cm2. Even the residues in the infrared laser etched groove could be completely removed by the subsequent ultraviolet irradiation. All lines, horizontal lines first, were etched at the scan speed 5 mm/s (Yavas and Takai 1998). ............................. 28 Figure 2.15: From left to right, the basic principle of stealth dicing, SEM observation of SD-processed MEMS overall chip, the enlarged section of edge periphery (Hamamatsu Photonics 2005) ................................................................................................................ 29 Figure 2.16: Advantages associated with working with ultrashort laser pulses (Clark- MXR), chapter 5. .............................................................................................................. 31 _____________________________________________________________________ xiv Figure 2.17: highlights the numerous physical phenomena that are present when machining with a long nanosecond laser pulse (Clark-MXR), Chapter 3. ....................... 32 Figure 2.18: Basic set-up to focus a Gaussian beam (Otani, Herbst et al. 2004). ............ 34 Figure 2.19: Laser-induced surface melting and vaporization (Bäuerle 2011). ............... 37 Figure 2.20: Sequence of images obtained by laser shadowgraphy for the laser irradiance of 8.7×1010 W/cm2 showing shockwave and particulate emission (Yoo, Jeong et al. 2000). ................................................................................................................................ 41 Figure 2.21: Plasma images obtained by laser shadowgraphy for femtosecond (left) and nanosecond (right) laser ablation of Si at 10 ns after the laser shot. The shock front (S), ionization front (I), contact front (C) and air breakdown plasma (A) are labelled in the figures (Zeng, Mao et al. 2005). ....................................................................................... 42 Figure 2.22: Electron number density as a function of time for femtosecond and nanosecond laser ablation at a distance of 0.6 mm above the target surface (Zeng, Mao et al. 2005) ............................................................................................................................ 43 Figure 2.23(a): SEM microphotographs of Si irradiated at 1.7 J/cm2 with 500 laser pulses at 45° view. (b) Magnified view of (a). (Sánchez, Morenza et al. 1996) ......................... 45 Figure 2.24: Etch crater on a Si target after being irradiated with 300 pulses at constant angles of incidence (40°) at (a) 6 J/cm2 and (b) 3 J/cm2 (Riet, Nillesen et al. 1993). ...... 46 Figure 2.25: SEM image of ion-polished cross-section 10 µm below the upper side of the wafer through a hole drilled in silicon with the copper vapour laser (Luft, Franz et al. 1996). ................................................................................................................................ 47 Figure 2.26: Influence of the focal position on the shape of the cutting edge and on the depth of cut (Dauer, Ehlert et al. 1999). ........................................................................... 48 Figure 2.27: Laser cut in silicon wafer using 1064 nm wavelength. ................................ 49 _____________________________________________________________________ xv Figure 2.28(a): Laser trepanned holes in silicon using high intensity (15 GW/cm2) at 355 nm (Karnakis, Rutterford et al. 2005). (b): Deep through hole with straight sidewalls drilled by two-cycle trepanning, the entrance is 50 µm and exit is 40 µm (Tan 2006). (c): SEM image of silicon machined by frequency tripled Nd: YAG laser (355 nm) followed by cleaning in the KOH solution (Chen and Darling 2005). (d) Profiles and SEM images (×1.5 k) of ablated a-Si under different irradiation conditions at 248 nm (Molpeceres, Lauzurica et al. 2007). ...................................................................................................... 51 Figure 2.29(a-d): Femtosecond-pulse laser processing of a 0.3 mm thick silicon target with 250 fs, 0.5 mJ, and F=2.5 J/cm2 second harmonic radiation (λ=309 nm) and different number of pulses: (a) 10, (b) 100, (c) 5×103, (d) 104 pulses (Chichkov, Momma et al. 1996). ................................................................................................................................ 53 Figure 2.30(a) & (b): Examples of silicon structuring with fs laser pulses by trepanning with a spot focus. (c): Kerf in thin silicon, cut by static exposure to fs laser pulses shaped to a line focus (Bärsch, Körber et al. 2003). ..................................................................... 53 Figure 2.31: Ablation depth vs. pulse number for femtosecond and nanosecond laser ablation of Si in air. The insets show the fs and ns crater profile after five laser pulses, notice that two axis are of different unit (Zeng, Mao et al. 2005). ................................... 54 Figure 2.32(a): Trench micromachined by laser beam with an energy of 50 nJ/pulse and (b): an energy of 200 nJ/pulse. High power machining resulting in surface contamination due to ejection of melt materials and thus, a large heat affected zone (Ngoi, Venkatakrishnan et al. 2001). ........................................................................................... 55 Figure 2.33: SEM image of cone inside dish laser etched into Si using chlorine ambient of 100 Torr. The dish has an upper radius of 80 µm and lower radius of 35 µm with radius of curvature of 82 µm and spacer layers extending 8 µm inward. The clipped cone has upper radius of 4 pm and lower radius of 16 µm. The laser was swept in a circular scan in each plane (Bloomstein and Ehrlich 1992). .......................................................... 57 Figure 2.34: Schematic Diagram of fibre laser (ORC 2009) ............................................ 60 _____________________________________________________________________ xvi Figure 2.35: Double-clad fibres convert the output of a high-power semiconductor pump laser to a diffraction-limited output beam in simple, end-pumped fibre laser cavities. Inset shows refractive index profile typical of a cladding-pumped fibre (Richardson, Minelly et al. 1997). ........................................................................................................................... 61 Figure 2.36: The shape of the temperature pulse at the carbon surface when the laser beam is absorbed. The temperature is calculated using the differential equation for one- dimensional heat flow and assuming a laser pulse as shown. Both the temperature pulse and the laser pulse are normalized to their peak values (Ready 1963). ............................ 64 Figure 2.37: Temperature versus distance into the solid at different normalized times. .. 66 Figure 2.38: Normalized temperature rise N(0, 0, W) at the surface (z = 0) and beam center (r = 0) for a Gaussian laser beam shape plotted versus W. The upper (solid) curve corresponds to the horizontal scale shown. The lower (dashed) curve is the upper curve replotted with the horizontal scale expanded by a factor of 5 to facilitate reading points in the small W regions. ......................................................................................................... 67 Figure 2.39: Normalised temperature rise as a functioin of R and Z for (a) W=0.8 and (b) W=12. .......................................................................................................................... 68 Figure 2.40: Schematic of nanosecond-pulse laser ablation. ............................................ 69 Figure 2.41: Distribution of the laser energy density between different channels for laser ablation of two materials as a function of laser fluence. E0 is incident laser pulse energy, E/E0 is the proportion of the how much other energy is with respect to E0; Et are the losses to the target due to heat conduction; ER is the reflected energy; Eab is the energy absorbed in the plume; EL is the part corresponding to latent heat of vaporization; Eth is the thermal energy of the vaporised particles (Bulgakova, Bulgakov et al. 2004). .......... 71 Figure 2.42: (a) Various laser pulse shapes, (b) Temperature profiles in the target at the time of evaporation for the pulse shapes shown in (a) (Franklin and Thareja 2004) ....... 72 _____________________________________________________________________ xvii Figure 2.43: Temperature profiles inside the ablated plume, at the end of laser pulse, at various intensities (Gaussian pulse s=0.2) (Franklin and Thareja 2004) .......................... 73 Figure 3.1: Internal structure of Yb doped fibre laser ...................................................... 76 Figure 3.2 (a): Laser pulse characteristics at varying pulse repetition rates of the SPI fibre laser (SPI Lasers 2008). (b): Pulse shapes of an IPG Q-switched fibre laser. .................. 77 Figure 3.3: Average power out of laser head (Coherent 2007). ........................................ 79 Figure 3.4: Images of the scan head and schematic data flow (Nutfield 2011). ............... 80 Figure 3.5: The experimental set up includes a laptop, a control pipeline unit, an auto-Z- axis enclosure, a sample holder, a fibre laser, a scanner and a vacuum pipe. .................. 81 Figure 3.6: Fibre laser line scanning on Si with 25 kHz repetition rate, 1.8 m/s and 10 overscans. .......................................................................................................................... 82 Figure 3.7: The machining system set up for the UV laser. ............................................. 83 Figure 3.8: The scanning system set up for the UV laser (left) and the control unit (right). ........................................................................................................................................... 84 Figure 3.9: Schematics of beam shape monitoring arrangement. ..................................... 85 Figure 3.10: Spatial intensity distribution of the raw laser beam, 25 kHz to 500 kHz pulse repetition rate, 20 W output power. .................................................................................. 86 Figure 3.11: Fibre laser pulse characteristics at varying pulse repetition rates measured after scanner lens (left), stated in the system manual (right). ........................................... 87 Figure 3.12: Temporal pulse profile of the DPSS UV laser at 60 kHz up to 10 W average power................................................................................................................................. 88 Figure 3.13: Average power available at the material substrate at 25 kHz, WFM 0, with ×2, ×3, ×3.3 and ×4 beam expansion.............................................................................. 89 _____________________________________________________________________ xviii Figure 3.14: Typical average power variation with PRFs for the selected waveform. ..... 90 Figure 3.15: Measured average power as a function of frequencies of the UV laser after optics. ................................................................................................................................ 91 Figure 3.16: Measured average power as a function of time of the UV laser at 60 kHz after optics. ........................................................................................................................ 91 Figure 3.17: Etch depth of machined lines in the focus test of the fibre laser scanning system. .............................................................................................................................. 93 Figure 3.18(a): Wyko NT3300 system. (b): Interference fringes at best focus (Veeco 1999). ................................................................................................................................ 94 Figure 3.19(a): Example of Wyko measurement of single pulse ablation on Si with SPI’s fibre laser at 25 kHz and 11 W. (b): Sub-region of Figure 3.19(a) (top right hole). ........ 95 Figure 3.20: High speed imaging setup with a light source, an iris to clean the illumination beam, a collimating lens, a sample stage with photodiode next to it, a lens and a telescope to magnify and form clear image on the CCD sensor. ............................ 98 Figure 3.21: Schematic setup of the shadowgraph imaging system. ................................ 99 Figure 3.22: Measurement of the Nd:YAG laser pulse width. ....................................... 100 Figure 4.1: Laser pulse characteristics at varying pulse repetition rates. ....................... 102 Figure 4.2: percussion drilled holes produced using beam parameters as detailed in Table 4.1.................................................................................................................................... 104 Figure 4.3: Ave. depth per pulse and total depth vs. frequencies in the drilling mode. . 105 Figure 4.4: Depth per pulse vs. peak power density in the drilling mode. ..................... 106 Figure 4.5: SEM images of percussion drilled holes with peak power density of 98 MW/cm2, (a) before ultrasonic cleaning, (b) after ultrasonic cleaning, (c) cross section. ......................................................................................................................................... 107 _____________________________________________________________________ xix Figure 4.6: SEM images of percussion drilled holes with peak power density of 270 MW/cm2, (a) before ultrasonic cleaning, (b) after ultrasonic cleaning, (c) cross section. ............................................................................................................................ 108 Figure 4.7: SEM images of percussion drilled holes with peak power density of 1,200 MW/cm2, (a) before ultrasonic cleaning, (b) after ultrasonic cleaning, (c) cross section. ............................................................................................................................ 108 Figure 4.8: SEM images of percussion drilled blind holes by 1064 nm Yb doped fibre laser with 10 pulses, 25 kHz PRF, 200 ns pulse duration and 20 W average power, (a) Secondary electron detection, (b) Inlens detection to look at the hole bottom. ........ 109 Figure 4.9: Pulse profiles of various pulse shapes at 125 kHz with 50, 75, 200 ns pulse. ......................................................................................................................................... 110 Figure 4.10: SEM image of percussion drilled holes with 50 ns pulse (a) before, (b) after ultrasonic cleaning, and (c) the cross section.................................................................. 110 Figure 4.11: SEM image of percussion drilled holes with 75 ns pulse (a) before, (b) after ultrasonic cleaning, and (c) the cross section.................................................................. 111 Figure 4.12: SEM image of percussion drilled holes with 200 ns pulse (a) before, (b) after ultrasonic cleaning, and (c) the cross section.................................................................. 111 Figure 4.13: Similar pulse profiles (200 ns pulse duration, ~870 W peak power, 83.4 MW/cm2 peak power density) at 25, 50, 75, 100 and 125 kHz for cooling cycle analysis. ........................................................................................................................... 112 Figure 4.14: Single pulse ablation and percussion drilling on Si with similar pulse profiles of different PRFs. ............................................................................................................ 113 Figure 4.15: Pulse shapes at 200 ns pulse duration with varying pulse energy/peak power. ......................................................................................................................................... 116 _____________________________________________________________________ xx Figure 4.16: Single pulse machined depth using a 200 ns pulse with variation in fluence. ......................................................................................................................................... 117 Figure 4.17: Single pulse ablation depth as a function of fluence at varying pulse durations from 50 to 200 ns. ........................................................................................... 119 Figure 4.18: Machining efficiency as a function of fluence at varying pulse durations from 50 to 200 ns. ........................................................................................................... 120 Figure 4.19: Pulse shapes of the same pulse peak power with pulse duration between 40 ns and 200 ns. .................................................................................................................. 121 Figure 4.20: Single pulse ablation depth as a function of fluence (bottom x-axis) and pulse duration (top x-axis) with pulses of constant peak power. .................................... 122 Figure 4.21: Pulse shapes of the same pulse energy (0.28 mJ/ 8.49 J/cm2) with pulse duration between 75 ns and 200 ns and peak power between 1.8 kW and 0.5 kW. ....... 123 Figure 4.22: Single pulse machined depth as a function of pulse duration at varying fluence values.................................................................................................................. 124 Figure 4.23: Average diameter and SEM images of the selected curves in Figure 4.22, with respect to pulse duration at varying fluence values. ............................................... 125 Figure 5.1: Temporal pulse profile of the DPSS UV laser at 60 kHz with varying average power up to 10W. ............................................................................................................ 131 Figure 5.2: Single pulse machining depth as a function of pulse energy for the UV laser at 58 ns. ............................................................................................................................... 132 Figure 5.3: Single pulse machining depth as a function of pulse energy for the fibre laser at 65, 200 and 220 ns. ..................................................................................................... 133 Figure 5.4(a): Temporal pulse profile of M2=1.2 fibre laser with 0.04 mJ pulse energy, 208 W peak power. (b): The temperature evolution of Si surface at the centre of laser irradiation with the laser pulse shown in (a). .................................................................. 134 _____________________________________________________________________ xxi Figure 5.5(a): Temporal pulse profile of the UV laser with 0.04 mJ pulse energy, 1740 W peak power. (b): The temperature evolution of Si surface at the centre of laser irradiation with the laser pulse shown in (a)..................................................................................... 135 Figure 5.6: Effects of pseudo Gaussian spatial energy distribution on the machined diameters. ........................................................................................................................ 136 Figure 5.7: Single pulse machining energy efficiency as a function of pulse energy for both the UV laser (58 ns) and IR fibre laser (200 ns). .................................................... 137 Figure 5.8(a): SEM image of single pulse machined hole (6.4 µm deep) by the M2 = 1.2 fibre laser, 0.171 mJ pulse energy. (b): SEM image of single pulse machined hole (4.6 µm deep) by the UV laser, 0.164 mJ pulse energy. ................................................. 139 Figure 5.9(a): Single pulse machining on Si with fibre lasers of different M2, depth plotted against pulse energy with a 200 ns pulse. (b): depth plotted against fluence. .... 141 Figure 5.10(a): Temporal pulse profile of two fibre lasers with 0.1 mJ pulse energy, ~605 W peak power. (b): Si surface temperature distribution of two fibre lasers with time. ......................................................................................................................................... 142 Figure 5.11: Single pulse machining efficiency with respect to pulse energy for both lasers with 200 ns pulse. ................................................................................................. 143 Figure 5.12: The M2 = 1.2 fibre laser single pulse machining of Si as a function of fluence with a 200 ns pulse and inset SEM images. ....................................................... 144 Figure 5.13(a): SEM image of single pulse machined hole by the M2 = 1.2 fibre laser. (b): by the M2 = 2 fibre laser. ................................................................................................ 145 Figure 6.1: Surface temperature distribution of Si under the 1064 nm fibre laser irradiation with a peak power density of 1.22 GW/cm2.................................................. 154 Figure 6.2: Thermal and optical properties of Si as a function of temperature. ............. 155 _____________________________________________________________________ xxii Figure 6.3: Surface temperature distribution of Si irradiated by laser pulses in Figure 4.15. The volume heating source was considered before the Si surface was melted. .... 156 Figure 6.4(a): Temperature distribution of Si by a surface heating source with respect to time at varying distances inside the Si substrate. (b): Close view of the surface temperature rise. .............................................................................................................. 160 Figure 6.5: Zoomed-in view of Figure 6.4(a), close to the beginning of the laser pulse. 161 Figure 6.6: Temperature distribution of Si by a surface heating source as a function of distance inside the Si substrate at different elapsed times. ............................................. 162 Figure 6.7: Pulse shapes at 200 ns pulse duration with varying pulse energy/peak power. ......................................................................................................................................... 164 Figure 6.8(a): Temperature distribution of the 1.22 GW/cm2 laser pulse, Figure 6.7, irradiation of Si as a function of time at varying distances beneath the Si surface. (b): Determination of the depth which is above 0.905𝑇𝑡𝑐 (4706 K) for at least 70 ns. . 165 Figure 6.9: The comparison of theoretical predictions based on the explosive boiling and measured depths. ............................................................................................................. 166 Figure 6.10: Sequence of mass ejection images obtained by laser shadowgraphy for the laser irradiance of 0.26 GW/cm2. .................................................................................... 167 Figure 6.11: Sequence of mass ejection images obtained by laser shadowgraphy for the laser irradiance of 1.22 GW/cm2. .................................................................................... 168 Figure 6.12: Pulse shapes of the same pulse peak power with pulse duration between 40 ns and 200 ns. .................................................................................................................. 169 Figure 6.13: The comparison of theoretical predictions and measured depths. .............. 170 Figure 6.14: Temperature distribution of 40 ns laser pulse. ........................................... 171 _____________________________________________________________________ xxiii Figure 6.15: Temporal pulse profile of the DPSS UV laser at 60 kHz with varying peak power densities................................................................................................................ 173 Figure 6.16: Temperature profile of Si under irradiation of the UV laser at 2.82 GW/cm2. ......................................................................................................................................... 174 Figure 6.17: The comparison between the theoretical predictions based on the explosive boiling and the measured depth of the UV laser machined holes on Si. ........................ 175 Figure 6.18: The overlaps of the ultrafast camera shutter signal and the fibre laser pulse. ......................................................................................................................................... 177 Figure 6.19: Plasma evolution of a 200 ns fibre laser pulse (1.22 GW/cm2) on Si sample. ......................................................................................................................................... 178 Figure 6.20: The fibre laser single pulse machined depth with respect to the peak power density. ............................................................................................................................ 180 _____________________________________________________________________ xxiv List of Tables Table 2.1: Basic Si properties. .......................................................................................... 10 Table 2.2: Laser parameters and influence on material processing (Ehrlich and Tsao 1989) ................................................................................................................................. 26 Table 2.3: Laser operation modes and typical pulse width. .............................................. 30 Table 2.4: Laser types with Characteristics and Applications for Si micromachining. .... 44 Table 3.1: Incident power density for a range of WFM settings of variable input bean diameters with a 125 mm focal length objective lens. ...................................................... 78 Table 3.2: System comparison of Yb fibre laser and DPSS UV laser. ............................. 79 Table 3.3: Technology modes of the Wyko NT3300 system. .......................................... 95 Table 4.1: Laser parameters for different PRFs when total energy of 3.54 mJ was kept constant, the total process time was 0.2 ms for each hole. ............................................. 103 Table 5.1: Two fibre laser parameters used to generate features in Figure 5.13 and the machined hole depth. ...................................................................................................... 145 Table 5.2: Laser parameters used to in literature and this study for explosive material removal mechanism. ....................................................................................................... 147 Table 6.1: Assumptions for the heat conduction model. ................................................ 150 Table 6.2: Si optical and thermal properties used in heat conduction Equations 6.7 and 6.10, at room temperature 𝑇0 =300 K. .................................................................... 153 Table 6.3: Si optical and thermal properties at 1 µm wavelength as a function of temperature (<1683 K). ................................................................................................... 155 Table 6.4: The onset time of melting and the percentage of pulse energy used to melt Si. ......................................................................................................................................... 157 _____________________________________________________________________ xxv Table 6.5: Optical and thermal properties of liquid Si used in Equations 6.15 and 6.16 at 1 µm wavelength. ............................................................................................................ 159 Table 6.6: Laser parameters and percentage energy used to melt Si. ............................. 170 Table 6.7: UV Laser parameters and the percentage energy used to melt the Si surface. ......................................................................................................................................... 173 Table 6.8: Optical and thermal properties of Si used in Equations 6.15 & 6.16 at 355 nm wavelength. ..................................................................................................................... 174 _____________________________________________________________________ 1 Chapter 1 Introduction 1.1 Background The micro-drilling of silicon components for use in the Micro-Electro-Mechanical Systems (MEMS) fabrication and microelectronics industry is well-established and continues to grow. One of the legacies of the electronic era is that Si has become the most intensely studied and developed material in manufacturing, and Si drilling processes are highly mature (Greuters 2003). Standard tools for processing Si, such as reactive ion etching (RIE) and diamond-saw cutting, have shown their limitations (Otani, Herbst et al. 2004). This makes Si an important material in the development of flexible and efficient laser materials processing solutions. The photovoltaic (PV) industry has experienced enormous growth over last few years (details in appendix A), and it is forecasted to grow to $100 billion in 2013 (Lux Research, NY). As the most important material, crystalline-silicon has a 77% market share of world production of solar cells (NanoMarkets 2009). In modern PV cell production, each wafer requires 100 via holes with a diameter of 50-100µm to increase the cell efficiency. A machine with a throughput of 30-40 wafers per second is considered state-of-the-art (Stolberg and Friedel 2010). Wet etching technology has a process time for a 200µm thick wafer in the range of 1 hour, additional process steps are required, and disposal of the byproducts are an environmental problem. Laser technology can satisfy the processing time with additional advantages such as reproducibility, energy selectivity, no tool wear-out and no mechanical contact. The industry standard state of the art for Si micro-drilling and micro-machining is the Q- switched Diode Pumped Solid State (DPSS) UV laser (Otani, Herbst et al. 2004); (Karnakis, Rutterford et al. 2005). A wavelength less than 532nm has advantages of _____________________________________________________________________ 2 higher absorption coefficients, shorter optical penetration depths, less plasma absorption (Breitling, Schittenhelm et al. 1999) and smaller focal spot diameters. The laser works in a low nanosecond regime which gives very little thermal defects. One could say that the laser industry has already approached the best Si machining output with DPSS UV laser systems using rods or discs as the gain medium. In order to achieve higher levels of power rating, beam quality, and wall plug efficiency, manufacturers are forced to deal with increasingly complex designs. Stable resonators, pump energy delivery and thermal management offer considerable challenges during scale-up. Recent advances in laser source technology have produced the fibre laser with the Master Oscillator Power Amplifier (MOPA) architecture that offers high repetition rates, high operational efficiencies, and pulse modulation controls exceeding those of typical Q- switched DPSS lasers. The advances in design, performance, cost reduction, and brightness for the Ytterbium (Yb) doped fiber lasers have opened up the possibilities of redefining the processing options and improving processing solutions of Si at a wavelength of 1064 nm. 1.2 Research aim and research questions This work seeks to explore the potential of 1 µm fiber laser machining of Si substrates by utilizing the advanced fiber laser specifications and exploring the influence these new parameters settings can have on the laser matter interaction. Of particular interest is the pulse modulation control (pulse shaping) capability, which provides extremely fast rise times for variable peak powers and tunable pulse durations. This work also seeks to understand physical phenomena related to the laser matter interactions. This thesis had the following aim: Research aim: To characterize 1 µm fibre laser machining of Si substrates with a view to identifying the most influential and optimum set of laser parameters for high quality, high efficiency micro drilling and surface texturing applications. A secondary _____________________________________________________________________ 3 aim is to develop a greater understanding of the laser material interactions and material removal mechanism when using fast rise-time nanosecond laser pulse envelopes. The research aim was translated into the research question shown below: Research question: Can a 1 µm fiber laser be used to machine Si with comparable or increased quality and energy efficiency compared to existing Q-switched DPSS UV lasers? For clarify this question was further split into three smaller research questions on the basis of the major tasks to be accomplished: Research Question 1: Which fibre laser parameters have the greatest influence on Si machining and what are their optimum settings? Research Question 2: How does the IR fibre laser compare with the DPSS UV laser on single pulse machining of Si? Research Question 3: How does theoretical analysis and imaging tools help understand the IR fibre laser Si interaction? 1.3 Research methodology The proposed methodology to solve research questions was mapped and is presented in Figure 1.1. There are strong connections between one phase and another, and between all phases and the research aim and objectives. _____________________________________________________________________ 4 Figure 1.1: Research methodology used in this thesis. 1.3.1 Phase 1 The first phase involved identifying and reviewing relevant literature regarding the laser machining of Si in terms of the machining parameters, machining mechanism, machining processes, and material properties. The current state of the art was studied with a focus on the machining quality. The information was used to design the base-line experimental configurations, develop the analysis procedures and review the research aim and objectives, hence affecting Phases 2 and 3. 1.3.2 Phase 2 For research question 1 of phase 2, a series of single pulse interaction and percussion drilling experiments were performed on 650 µm thick Si wafer to identify the influence of laser parameters on the quality and energy efficiency of Si drilling. Changes of the Phase 3 Theoretical analysis & high speed imaging Research questions 3 Phase 2 Process characterisation & performance comparison (Research questions 1 & 2) Research aim & objectives Phase 1 Literature review & Experimental setup _____________________________________________________________________ 5 mass removal mechanism are investigated by varying laser pulse energy, pulse duration, and pulse peak power. Hole depths were measured with a white light interferometer to obtain etch depth per pulse /unit energy. Surface morphology was studied using a Scanning Electron Microscope (SEM) to identify factors such as debris size, level of recast and the formation of micro cracks. The objective for research question 2 of phase 2, seeks to examine the differences in the machined depth and the machining quality of Si produced by the IR fibre laser and the industry leading UV laser. Both lasers work in the nanosecond regime, the difference in wavelengths (1064 nm fibre laser and 355 nm UV laser) gives UV laser advantages in Si processing due to a much higher absorption coefficient, but the pulse shaping capability of the fiber laser opens new processing opportunities. The UV laser experiments were carried out with single pulse interaction at varying pulse energy but a fixed pulse duration for comparison. 1.3.3 Phase 3 To answer research question 3, fast visible imaging and shadowgraph techniques were employed to examine the interaction zone in a greater detail, and a thermal model was applied to analyze the energy coupled into the material. The high speed imaging techniques can provide detailed information such as the speed of vaporised particles and the duration for material removal at varying pulse shapes. This information can then be used to validate theoretical analysis, which provides a more quantitative explanation of the laser material interaction phenomena. The results from the analytical model can lead to future developments for optimizing laser Si materials processing. 1.4 Research scope The material on which all the experiments were carried out in this report was single crystalline Si <111>. Si is the most used material in semiconductor and solar cell _____________________________________________________________________ 6 industry, and its properties and processing options have been studied extensively in the past. Laser machined depths were measured in this study. The relationship between the machined depth and the laser irradiance is the key parameter in characterizing how well the laser energy couples into a solid. The fibre laser used in this study is randomly polarized, and all the samples have been irradiated at normal incidence to minimize the effects polarization has on the angle of incidence at the centre of the irradiation. The processing conditions were standard temperature and pressure, and complied with typical industrial processing environment and conditions. Thesis synopsis 1.4.1 Chapter 2 – Literature Review The literature review chapter starts with an examination of the properties of Si and its applications. The thermodynamic properties of Si are discussed at room temperature and at elevated temperature, the variations have great effects on the optical and thermal energy coupling of the laser source. The chapter also provides relevant information on silicon machining with various types of lasers, laser parameters and their effects on processing, and material removal mechanisms. The information obtained here helped to inform and develop the experimental procedures and analysis. Thermal models are reviewed, which were used widely in the prediction of the temperature distribution and understanding of the laser matter interaction. At the end of this chapter, a summary is presented to highlight the unsolved problems and further define the investigation areas for this research. 1.4.2 Chapter 3 – Experimental systems, measurement tools and methodology Equipment and techniques used in this thesis are described. The list begins with lasers and beam delivering systems, and their specifications. Prior to the machining processes, _____________________________________________________________________ 7 laser properties are studied with the power meter, beam profiler and photodiode, to measure the average output power steadiness, to examine its spatial energy distribution and inspect its temporal profile at various parameters. Then detailed sample management and machining procedures are explained so that the work can be repeated. Measurement tools and processes are discussed with an aim to extract the depth information and examine the finishing quality of drilled holes. High speed imaging and shadowgraphy methods are illustrated to study the material removal mechanism and detect plasma generation /expansion. 1.4.3 Chapter 4 – Fibre laser machining characterization of silicon This chapter aims to answer research questions 1. The laser matter interactions with the pulse tuneable 1064 nm wavelength Yb fibre laser on Si for micro-via applications are characterized with various laser parameters. The parameters are number of pulses, pulse energy, pulse duration, pulse peak power and pulse repetition frequency (PRF). These parameters affect the machining mechanism, quality, debris formed, machining depth and energy efficiency. Effective laser parameters are examined collectively through percussion drilling. Then each laser parameter is analyzed independently through single pulse interaction experiments. Of particular interest is the laser temporal pulse profile effect on drilling performance of Si. The drilling depth is measured for each parameter set and processed quality is compared for a wide range of interaction conditions. The optimum laser parameters for Si drilling at varying peak power densities and varying pulse durations are determined. 1.4.4 Chapter 5 – Performance comparison between IR and UV lasers This chapter aims to answer research questions 2. Wavelength is the main difference between the IR fibre laser and DPSS UV laser, it is also the most important laser parameter that determines how much energy is absorbed by the material and how the energy is delivered. Single pulse machining of silicon using a non process optimised Q- swithced DPSS laser at 355 nm is compared with that achieved using a MOPA based _____________________________________________________________________ 8 fibre laser (M2 = 1.2) at 1064 nm. The comparison includes the machined depth, machining mechanism, energy efficiency and hole quality, and the surface temperature is calculated to identify the differences caused by the wavelength. The difference in performance of two fibre lasers (M2 = 2 and M2 = 1.2) are also compared. 1.4.5 Chapter 6 – Theoretical analysis This chapter aims to answer research question 3. The asymmetrical pulse shape of the fibre laser was thought to alter the way energy is delivered to Si, and cause different interaction mechanism from the normal 1 µm laser volumetric heating of Si. The temperature and phase dependent Si properties could also assist the heating process by changing the material response to the incoming laser beam. A thermal model is used here to predict the temperature change within the target, determine the dominant material removal mechanism that Si went through during the IR irradiation and estimate the machined depth. The validity of the model with empirical findings is discussed. Imaging tools, like laser shadowgraphy and visible imaging, were used to capture images of the ejected mass from the interaction zone, approximate ejected particle size and speed, and infer the time scale for the onset of mass ejection. The results are used to validate the material removal mechanism. 1.4.6 Chapter 7 – Conclusion The research aim and objectives are reviewed again, the main findings from Chapters 4 to 6 are summarized with respect to the research questions. The overall conclusion is drawn and suggestions for future work are given. _____________________________________________________________________ 9 Chapter 2 Literature Review 2.1 Introduction This chapter aims to summarise the relevant literature on laser material interactions. The material side includes details on the silicon’s basic properties, applications and energy transfer from a laser source. The optical and thermal energy coupling which are two forms of energy transfer are discussed for the elevated temperature and material’s transition from solid to liquid Si. The laser side starts with its advantages over other machining tools and laser’s classifications in terms of pulse widths. A list of basic laser parameters is defined in Appendix B, the most influential parameters are discussed alongside with their effects on micromachining. A range of mass removal mechanisms are talked about based on the laser heating of material. Silicon was processed with different types of lasers in the past, the results and side effects are demonstrated, with emphasis on the state of the art DPSS UV laser. The fiber laser, which is used in this study, and its features and advantages are introduced. The thermal model and its derivative forms were used to predict temperature and other important information like melt depth and ablation threshold. The variations of laser parameters and material properties were revealed in the calculated results, which is useful in understanding the complex interaction phenomena. 2.2 Silicon 2.2.1 Basic properties and applications Single crystalline Si was considered an excellent substrate choice for many mechanical sensor applications because of its intrinsic mechanical stability, Table 2.1, and the _____________________________________________________________________ 10 feasibility of integrating sensing and electronics on the same substrate. Compared to other substrate materials like ceramic, plastic and glass, Si is good for both metallisation and machinability. Single-crystal Si has higher thermal conductivity than steel and aluminum and about 100 times larger than glass and this makes Si accepted in devices as an efficient heat sink (Sakuma, Yonezu et al. 1978); (Madou 2002). Si has low thermal expansion coefficient and is stress free over a wide range of temperatures (same as Pyrex® glass), both of these properties make Si highly desirable as the support for sensors. One application uses the bonding between Pyrex® glass and Si, and produces no drift which stems from packaging and little stress. Table 2.1: Basic Si properties. Property (Solid Si) Values Units Reference Atomic weight 28.09 Energy Gap 1.12 eV (Bäuerle 2011) Hardness 7 Moh (Basu 2011) Young’s Modulus (different directions) 130-185 GPa (Ioffe 2011-5-27) Refractive index 3.4 (Basu 2011) Melting point (𝑇�) 1683 K (Yoo, Jeong et al. 2000); (Bäuerle 2011) Boiling point (𝑇�) 3514 K (Yoo, Jeong et al. 2000) Critical temperature (𝑇�) 5200 K (Lu, Mao et al. 2008) Liquid density (𝜌�) 2.533 g/cm 3 (Bauerle, Bermann et al. 2004) Latent heat of melting (𝐿�) 1800 J/g (Bäuerle 2011) Latent heat of vaporization (𝐿�) 12,800 J/g (Bäuerle 2011) Absorption coefficient (𝛼) 50 cm-1 (Bauerle 2000); (Weakliem and Redfield 1979) Optical penetration depth (𝑙�) 0.02 cm Reflectivity (𝑅) @ 1.152 µm 31.9 % (Jellison, Lowndes et al. 1986) Specific heat (𝐶�) 0.745 J/g·K (Desai 1986); (Hull 1999) Density (𝜌) 2.32 g/cm3 (Unamuno and Fogarassy 1989); (Lukeš, šášik et al. 1992) Thermal conductivity (𝜅) 1.397 J/ cm·s·K (Yoo, Jeong et al. 2000); (Unamuno and Fogarassy 1989) Thermal diffusivity (𝐷) 0.852 cm2/s (Yamamoto, Abe et al. 1991); (Shanks, Maycock et al. 1963) At room temperature, Si can only be deformed elastically, as under an applied load, no dislocation lines can move. Together with an extremely high yield strength, Si is a _____________________________________________________________________ 11 material superior to any metals in most MEMS applications. New generations of MEMS integrating a number of thick layers (>5 µm) made of different materials such as glass, Si, silicon carbide (SiC) and piezoceramics (PZT) could benefit from laser micromachining techniques (Desbiens and Masson 2007). Si exists in different forms, and they are used for various applications. The vast majority of electronic components are based on semi-conducting mono-crystalline Si (Boyd and Wilson 1983); (Madou 2002); (Huang and Lieber 2004). In the modern micro- and nanoelectonics industry, silicon nanowires, Figure 2.1, has been integrated into surrounding gate field effect transistors to miniaturize the transistor while still maintaining control over power consumption (Goldberger, Hochbaum et al. 2006). Figure 2.1: Si nanowire: (A) cross-sectional SEM image of vertically grown Si nanowires off of a Si(111) substrate, scale bar is 1 µm. (B) TEM image of SI nanowire surrounded in a conformal SiO2 coating, scale bar is 75 nm. (C) high resolution TEM image of a Si nanowire with an inner diameter that has been reduced to ~4.5 nm, scale bar is 4 nm (Goldberger, Hochbaum et al. 2006). _____________________________________________________________________ 12 Micromechanical components like acceleration sensors in car safety systems (Butefisch, Schoft et al. 2000); (Howe, Boser et al. 1996) and micro-fluidic circuits are made of Si (Takao, Ishida et al. 2002). X-ray mirrors also use Si as a substrate (Thielsch, Gatto et al. 2002). Due to the extreme flatness and well established coating procedures, Si, with or without passivating layers, is often the preferred substrate, especially for thin films (Smith, Carey et al. 1997); (Madou and Florkey 2000). The booming solar cell industry uses single crystal, polycrystalline and amorphous Si as wafers (Zhao, Wang et al. 1998); (Shah, Torres et al. 1999); (Weber, Blakers et al. 2004). Crystalline silicon wafers have dominated the PV market for two decades. The increase in solar cell efficiency and reduction in their cost are essential for crystalline-silicon solar cells to be competitive in the longer term. Reducing the wafer thickness is an effective method because one third of the Si solar cell cost is the material itself (Schoonderbeek, Stute et al. 2006); (Moalem, Schoonderbeek et al. 2007). Various solar cell surface structures and novel concepts have been developed to increase the cell efficiency. Figure 2.2 showed some examples of surface structures; the “honeycomb”, “inverted pyramid” surface textures and thin silicon strips on the surface of Si solar cell. _____________________________________________________________________ 13 (a) (b) (c) (d) Figure 2.2(a): Schematic and (b) SEM image of 19.8% efficient multicrystalline silicon solar cell with “honeycomb” surface texturing (Zhao, Wang et al. 1998). (c) The “inverted pyramid” texture on the top surface (Zhao, Wang et al. 1995). (d) A series of thin silicon strips (Weber, Blakers et al. 2004). Novel back contact solar cell concepts such as Emitter Wrap Through (EWT) (Gee, Schubert et al. 1993), Rear Interdigitated Single Evaporation EWT (RISE-EWT) (Harder, Hermann et al. 2009) and Metal Wrap Through (MWT) (Clement, Menkoe et al. 2010) require micro-vias drilled into the Si wafer. The vias are used to guide the contacts (metal filling) or the emitter to the backside, about 3,000 to 4,000 vias of about 100 µm in diameter must be drilled per second for the modern solar cell production. An example of the rear side for RISE-EWT solar cell is shown in Figure 2.3. _____________________________________________________________________ 14 Figure 2.3: SEM image of the rear surface after laser processing and KOH etching for RISE-EWT solar cell (Harder, Hermann et al. 2009). Another example, the passivated emitter and rear cell (PERC) concept deposits a dielectric passivation layer onto the solar cell’s rear side and subsequently makes electrical contact to the bulk silicon by a point pattern (Schneiderlöchner, Preu et al. 2002). A Q-switched Nd-YAG laser emitting a beam in TEM00-mode with 1064 nm wavelength was used to form the contact pattern with a movable X-Y table underneath the fixed laser beam axes. 2.2.2 Energy transfer in Si Laser light energy, in order to cause permanent damage on materials, must first be transferred optically or thermally or both. Si under laser irradiation undergoes phase transformation depending on the energy transferred, as a result the material properties are affected by the elevated temperature and materials’ form. _____________________________________________________________________ 15 2.2.2.1 Optical coupling The response of semiconductors under intense laser illumination and therefore high carrier density has been the subject of much research since the advent of the high power laser. The dynamic nature of optical energy coupling was proved by the observation of the transient absorption due to laser excitation of free carriers (Blinov, Bobrova et al. 1968), and enhanced reflectivity induced by intense laser pulses (Shvarev, Baum et al. 1975); (Lampert, Koebel et al. 1981); (Jellison and Lowndes 1987). When the laser radiation hits the target, its radiation can be divided into three parts; an absorbed, reflected and transmitted part as described by A + R + T = 1. The absorbed part A of the laser radiation penetrates the sample and heats the material, and the resulted heat can be transported into the bulk material by diffusion. The reflected part R of the laser radiation is given by the Fresnel expression; R = |(n– 1)/(n + 1)|�, where n is the complex refractive index. In some cases, the target is transparent dependent on the laser wavelength, and a part T of the laser radiation is transmitted through the sample (Schoonderbeek, Stute et al. 2006). When the target is not transparent or the target is thick enough that T is negligible, the absorptivity A = 1 − R (%) determines the amount of beam power engaged into the material (Allmen and Blatter 1995). However, the performances of laser micromachining processes are mainly dependent on the absorption coefficient (α) which is a strong function of the medium, wavelength and temperature. An enormous amount of work has been dedicated to investigating laser absorption mechanisms under various circumstances (Allmen and Blatter 1995); (Aydinli, Lo et al. 1981); (Jellison and Lowndes 1982). 2.2.2.1.1 Absorption The optical absorption coefficient α (cm-1), which characterises the absorption of light as it passes through an absorbing material, is determined by the equation α = 4πn2/λ, where n2 is the imaginary part of the refractive index (n = n1 + i n2), and λ is the wavelength of incident laser (Allmen and Blatter 1995). Instead of absorption coefficient α, the optical _____________________________________________________________________ 16 penetration depth lα = α-1 is often used in laser/material interaction, it is defined as the length in which the power density decreases to a fraction of 1/e with respect to its original value on the sample surface. Si is partially transparent for many laser wavelengths. As shown in Figure 2.4 that laser radiation of wavelength 1 µm transmits through a 200 µm thickness, while laser radiation of wavelength 355 nm is absorbed for the greater part in a surface layer of 10 nm thickness. As a result, the absorption of Si at longer wavelengths (>532 nm) is thought to be too low to achieve accurate micromachining, and short wavelengths (≤532 nm) are the preferred choice. Figure 2.4: Optical penetration depth of mono-crystalline Si for various wavelength, measured at room temperature (Bauerle 2000). According to Figure 2.4, the optical penetration length of Si decreases with decreasing laser wavelength only steadily near 1064 nm (hω=1.17 eV), but rather rapidly near 355 nm (hω=3.5 eV). Two different kinds of interband transitions are suggested in materials like Si (Allmen and Blatter 1995). The first kind is direct transitions, which permit transitions of electrons from valence band to conduction band with absorbed photon energy. The second kind is indirect transitions, which requires a change in both energy and momentum, and the momentum factor must be provided by phonons (Polleux and Rumelhard 2000). Silicon is an indirect bandgap material; it allows only phonon- _____________________________________________________________________ 17 assisted transitions. The probability of such transitions depends on the phonon occupancy and is relatively small and temperature dependent. The data in Figure 2.4, the optical penetration length against wavelength, are room temperature values and they do not relate to the conditions offered during laser interaction. At elevated temperature from 300 K to 1140 K, the absorption coefficient was measured to increase significantly, from just below 10 cm-1 to over 2 ×103 cm-1 at 1064 nm wavelength as shown in Figure 2.5 (Jellison and Lowndes 1982); (Weakliem and Redfield 1979); (Macfarlane, McLean et al. 1958). Figure 2.5: Optical absorption coefficient of Si measured at wavelength of 1064 and 1152 nm as a function of temperature (Jellison and Lowndes 1982); (Weakliem and Redfield 1979); (Macfarlane, McLean et al. 1958). _____________________________________________________________________ 18 The increased absorption coefficient corresponds to an optical penetration length of the order of 0.1 µm, which was also reported by (Blinov, Bobrova et al. 1968) with a decrease in the transmission. The temperature increase causes an increase in the phonon population, which then results more phonon-electron energy exchanges. Instead of oscillation and reradiation, the electrons are more likely to interact with the structure, and result in an increase in the absorption (Steen 1998). When Si becomes melted, the absorption coefficient also increases strongly since there are many more valence electrons, and molten Si adopts a metallic behavior (Sokolowski-Tinten, Bialkowski et al. 1998). In this case, as with metals, the optical penetration length of liquid Si decreases abruptly to 14.3 nm which allows the absorption of the laser radiation to take place very near to the surface (Muller, De Unamuno et al. 1996); (Shirk and Molian 1998). 2.2.2.1.2 Reflection and Transmission At wavelengths shorter than 400 nm (blue and UV), Si reflects over 60 % of the incident light as shown in Figure 2.6. The optical quality of Si surface gives very low insertion loss even after multiple reflections. This high reflectivity allows MEMS devices to have a significant commercial breakthrough, the micromachined mirrors for optical switching in both fibre optic communications and data storage applications (Wu 1997). The reflection of Si to 1 µm laser light at room temperature is 30 %, but it tends to increase when Si is excited by high energy laser irradiation or becomes melted, and the transmission decreases. _____________________________________________________________________ 19 Figure 2.6: Optical reflectivity mono-crystalline Si for various wavelength, measured at room temperature (Bauerle 2000). The time dependent optical properties of Si during nanosecond irradiation from a ruby laser were studied (Lowndes, Jellison et al. 1982). In Figure 2.7, the transmission of the 1152 nm line reduced to zero and stayed at zero for a time proportionate to the laser fluence El when El is above 0.8 J/cm2 (λ=694 nm, FWHM 14 ns). As the fluence was further increased, zero transmission lasted longer, and the interaction with 1µm wavelength laser occurred on the Si surface. The transmission then recovered to its initial value over several hundred nanoseconds. _____________________________________________________________________ 20 Figure 2.7: Time dependent transmission of a 1.152μm laser beam through Si during and after pulsed Ruby irradiation, where El is the pulsed Ruby laser energy density (Lowndes, Jellison et al. 1982). This low transmission behaviour is also accompanied by high reflectivity (close to 75 %) as shown in Figure 2.8, and is attributed to a thin layer of molten Si. The high reflectivity phase is characterised by high carrier density, high carrier temperature, low vibrational temperature and a self confinement of the plasma region of the order of 0.07 µm (Aydinli, Lo et al. 1981). All of the measurements revealed that the time resolved transmission recovered slowly with a time constant of ~500 ns even after the molten layer has solidified. One explanation for long tail behaviour is that there is enhanced absorption in the near surface region of the sample. _____________________________________________________________________ 21 Figure 2.8: Time dependent transmission T and probe reflection R of a 1.152 μm probe laser incident on Si during and after exposure to a pulsed laser excitation, 1 J/cm2, 8 ns pulse width and 455 nm wavelength (Aydinli, Lo et al. 1981). Given the highly dynamic response of Si under intense illumination and temperature variations, with the subsequent increase in absorption and reflection, and reduction in transmission when in a metallic state, the question is then; is it therefore possible to control the laser-Si interaction conditions at IR wavelengths by carefully selecting the basic laser parameters, and greatly improving the efficiency and quality of laser Si processing at infrared wavelengths? 2.2.2.2 Thermal coupling Most laser research has shown that processing Si with nanosecond lasers was only possible with thermal effects, as Si becomes molten, laser light is concentrated on the surface layer, and energy can only be thermally coupled into the material. Heat diffusivity (D) of Si, which is the thermal conductivity divided by the volumetric heat capacity, determines how efficient the thermal coupling process is. Laser pulse width governs the diffusion time (τl), and thermal diffusion length can then be calculated approximately through equation: lT ≈ 2(Dτl)1/2 (Schoonderbeek, Kling et al. 2007). For _____________________________________________________________________ 22 single crystalline Si (𝐷 = 0.852 cm2/s at T=300 K), the thermal diffusion length can be as deep as 8.26 µm at τl=200 ns, or it can be as shallow as 0.2 µm at τl=0.1 ns. Similar to the optical properties of Si, temperature affects the value of the thermal diffusivity. In Figure 2.9, thermal diffusivity of Si was measured to decrease exponentially with temperature from 300 to 1400 K (melting point). The thermal diffusion length is then reduced for the same pulse width, so high peak power and laser beam intensity which results in an increase in temperature can be used to reduce thermal diffusion length. Figure 2.9: Thermal diffusivity of Si from 300 to 1400 K (Shanks, Maycock et al. 1963). The other efficient way of reducing the thermal diffusion length is to use short pulses. The process can then produce less thermal defects, such as heat affected zone (HAZ )and recast layer as shown in Figure 2.10. In many cases, thermal defects caused by the laser beam can also be reduced by increasing the absorption strength via the laser wavelength selection or material doping. The spread of the damaged zone is related to the degree of localisation of the absorbed laser light energy and thereby to the thermal diffusion length _____________________________________________________________________ 23 and the optical penetration depth (Allmen and Blatter 1995). Reducing the thermal diffusion length on the other hand can lead to less material being removed, and the excessive heat is not used to remove more material, instead it is often used to generate plasma. Figure 2.10: The effect of pulse width damage in Si using 355 nm wavelength DPSS laser, a greater amount of resolidified material resulted with a longer pulse width (Otani, Herbst et al. 2004). By combining data on optical penetration length and thermal diffusion length, Figure 2.11 was constructed (Schoonderbeek, Kling et al. 2007). _____________________________________________________________________ 24 Figure 2.11: Commercially available laser systems versus optical penetration length and thermal diffusion length for mono crystalline Si at a temperature of 300K. With respect to the pulse width and wavelength various commercially available laser systems are drawn in the graph (Schoonderbeek, Kling et al. 2007). Figure 2.11 reveals the theoretical laser parameters of commonly used lasers for different micromachining features (Holes, Grooves and Texture), as shown in red lines. These parameter ranges are determined according to the allowable thermal damage of the Si. Fiber lasers, which became very popular in the last few years, are situated next to Q-Nd:YAG laser, the non-Gaussian pulse shape tunability enhances the pulse duration from 10-8 to 10-7 s. 2.3 Laser micromachining The requirement for the production of micron and sub-micron features at high-speed and with low-unit cost is essential to the manufacturing of high-tech devices. Laser micromachining is no exception, which received huge attention from industrial and scientific users (Allmen and Blatter 1995); (Duley 1996); (Kautek, Roas et al. 1990); (Shannon, Mao et al. 1995). Texture _____________________________________________________________________ 25 Laser machining as a competitive technology outperforms other processing techniques such as photolithographic process, mechanical drilling, reactive ion etching, and electron- beam machining in different ways. It has direct write capability (does not require a mask), requires less processing steps, Figure 2.12, and it is capable of both subtractive and additive machining. Photons have no charge, and they do not repulse each other like electrons. Short pulses of coherent light replaced electrons from the electron beam vacuum technology as the cutting tool by vaporizing the substrate (subtractive process) (Madou 2002). The laser is a tool for rapid prototyping, mold fabrication, and site- specific manufacturing. Lasers have now been used for machining micro-channel in micro-fluidic applications (Chen and Darling 2008); (Cheng, Wei et al. 2004); (Qi, Chen et al. 2009), micro-lenses (Lee and Wu 2007), micro thin film thermocouples (Choi and Li 2007), and semiconductor materials such as Gallium nitride (Gu, Howard et al. 2006). (1) Photolithographic process (2) Laser fired contacts process Figure 2.12: Laser fired contacts technology simplifies the process sequence to implement the point contact pattern of passivated rear silicon solar cells. Laser machining processes can be distinguished in different ways. In terms of material machining, it can be classified into pyrolitic (thermal) and photolytic (non-thermal) processes. In pyrolitic processes the laser energy is absorbed by heating the material, _____________________________________________________________________ 26 resulting in a temperature rise, melting or evaporation of the material. In photolytic processes, the energy is engaged in photon induced chemical reactions. The mass removal by laser ablation in this study is a thermal process. 2.3.1 Laser parameters Laser parameters determine how materials react with light, a range of influencing parameters and their effects on materials processing are summarized in Table 2.2. For micromachining purposes, the wavelength, beam size, pulse energy, pulse width, and peak power are the most important parameters to control. Table 2.2: Laser parameters and influence on material processing (Ehrlich and Tsao 1989) Laser parameters Influence on material processing Wavelength (µm) Optical absorption, reflection, transmission, resolution, and photochemical effects Average power (W) /Pulse energy (mJ) Temperature (steady state), process throughput Pulse width (s) Interaction time, transient processes Peak power (W) Peak temperature, damage/induced stress, nonlinear effects Beam size (mm) Focal spot size, depth of focus, intensity 2.3.1.1 Wavelength In the thermal process, wavelength is one parameter which determines how much laser energy is coupled into the workpiece. The lower wavelength possesses higher photon energy, and most materials absorb UV wavelength better than visible and IR wavelength. As an example of copper micromachining with different wavelengths, a much higher average etch depth per pulse was achieved at lower wavelength (Tunna, Kearns et al. 2001), Figure 2.13. Absorption of some materials can increase as wavelength moves from 1 to 10 µm, such as glass and Perspex. For Si3N4 and hard metals like tungsten carbide (WC) there is no significant change over the whole spectrum (Meijer, Du et al. 2002). In _____________________________________________________________________ 27 the non-thermal process, direct photochemical bond breaking plays an important role during the organic materials exposure to UV radiation. At this wavelength, photons have enough energy (3 to 10 eV), and one- or two-photon absorption by materials can cause direct bond cleavage in the solid, leading to rapid decomposition of the material into highly volatile molecular fragments (Madou 2002). Figure 2.13: Average etch depth as a function of intensity for 0.25 mm thick copper when operating the laser at 355, 532 and 1064 nm. Wavelength selection largely depends on applications, for example, microwelding needs large spot size and long penetration depth for the formation of a key hole throughout the thin sheets, i.e. long wavelength. In contrast, precision ablation requires small spot size and short penetration depth for the rapid and complete material removal, i.e. short wavelength. Laser tools operating at 1047 nm with typically a few tens of Watts of power have been used widely for the fine line scribing of Indium Tin Oxide (ITO) on glass substrates for thin film solar panel production. However, the IR irradiation results in the formation of a ripple-like morphology in the laser etched groove, ultraviolet laser irradiation yields a ripple-free smooth etch groove even at higher scan speeds, Figure 2.14. Stronger heating _____________________________________________________________________ 28 of the ITO–glass interface is considered to be the key for the improved pattern morphology in the ultraviolet region (Yavas and Takai 1998). Figure 2.14: Two lines etched at different wavelengths. Horizontal line: 𝜆 =1047 nm (Nd:YLF, first harmonic); F=18 J/cm2. Vertical line: 𝜆 = 262 nm (Nd:YLF, fourth harmonic); F=3 J/cm2. Even the residues in the infrared laser etched groove could be completely removed by the subsequent ultraviolet irradiation. All lines, horizontal lines first, were etched at the scan speed 5 mm/s (Yavas and Takai 1998). In recent years, (Hamamatsu Photonics 2005) from Japan have showed an interesting laser stealth dicing (SD) technique on a Si wafer, using a 1 µm wavelength nanosecond laser, which offers high quality, high speed, superior breakage strength, low kerf loss, a completely dry process and low running cost. The basic principle of stealth dicing is shown in Figure 2.15, a laser beam at a wavelength capable of transmitting through a Si wafer is condensed by an objective lens and focused onto a point inside Si wafer. It selectively forms a mechanical damage layer (stress layer) in a localized point near the light focus area so that Si is cut from the “inside”. _____________________________________________________________________ 29 Figure 2.15: From left to right, the basic principle of stealth dicing, SEM observation of SD-processed MEMS overall chip, the enlarged section of edge periphery (Hamamatsu Photonics 2005) One application of stealth dicing to MEMS devices is the MEMS chip sensor device. Figure 2.15 (left) is the SD process performed on the MEMS chip sensor device. Figure 2.15 (middle) shows the SEM images of overall MEMS device with a chip size of 2×2 mm and a thickness of 300 µm. A hollow with a depth of several µm is created in the center of the chip, and the actual device is formed on the opposite side of the chip. In this type of sample, the ultra-thin sections in the center of the chip are easily damaged by external impacts from the blade and cleaning water during dicing by the conventional blade dicing method. With SD process, an extremely sharp edge was obtained with no chipping or cracks occurred, Figure 2.15 (right). 2.3.1.2 Pulse energy and pulse shape Material damage is generally proportional to the pulse energy used, but the machining mechanism is also determined by the way the energy is delivered, i.e. the pulse shape. The laser pulse width and peak power are the key parameters which describe the pulse shape. For gain switched or asymmetrical pulse shapes, there is no effective way to describe the pulse through FWHM measurements as most of the energy is not contained in the FWHM region. The pulse shape of the fibre laser in this study has a high front peak followed by a long energy tail. The pulse duration describes the full width of the pulse, _____________________________________________________________________ 30 and the peak power is determined through integration of the energy contained in the pulse. 2.3.1.2.1 Pulse width Lasers are divided into three major operation modes based on the pulse width, Table 2.3, each operation mode has its own machining characteristics and applications. Table 2.3: Laser operation modes and typical pulse width. Operation modes Ultrashort pulse Short pulse Long pulse/ continuous wave (CW) Typical pulse width Pico- and femto-second pulses Nano-second pulses Micro-second pulses and longer In the ultrashort pulses range, extremely accurate micromachined parts are promised with no melt zone, no microcracks, no recast layer and no damage caused to adjacent structures as illustrated in Figure 2.16. _____________________________________________________________________ 31 Figure 2.16: Advantages associated with working with ultrashort laser pulses (Clark- MXR), chapter 5. In the short pulses range, (long nanosecond), the most elemental feature of laser/material interaction is that the heat deposited by the laser in the material diffuses away during the pulse width. This is good for laser welding, but for most micromachining, HAZ is unfavorable and limits the quality of the machining (Singh, Alberts et al. 2008). The disadvantages of long pulse width, long heat diffusion time are summarized, and illustrated in Figure 2.17. _____________________________________________________________________ 32 Figure 2.17: highlights the numerous physical phenomena that are present when machining with a long nanosecond laser pulse (Clark-MXR), Chapter 3. The long heat diffusion time causes the diffused heat away from the focal spot, and energy is not coupled efficiently to the machined area. With enough pulse energy for material ablation, the heat diffusion is causing larger feature sizes than the minimum spot size (resolution). The long heat diffusion time and laser generated shock wave can cause mechanical stress during laser heating or cooling cycles, and generate micro cracks in the adjacent areas. Nearby devices in integrated circuits can then be damaged and multilayer materials delaminated. A large amount of molten materials are generated by long pulse widths, some form a recast layer on sidewall of drilled holes, and some are left as surface debris. In the long pulse or CW range, lasers can be used to melt materials for micro welding (additive process), which the material is constantly in the liquid state. _____________________________________________________________________ 33 2.3.1.2.2 Peak power Peak power decides the peak temperature reached at the interaction zone at constant pulse width, which then decides the machining mechanism. High peak power is normally used to generate vaporization dominated ablation on metals and semiconductor materials, which leaves a thin recast layer and little heat affected zones (Ion 2005). High instantaneous peak power is also a factor which induces stress and causes cracks if the temperature is increasing or decreasing too quickly. 2.3.1.3 Beam size Laser beam size describes the raw beam diameter of the laser, it has influences on the minimum spot size, laser beam intensity and depth of focus. 2.3.1.3.1 Spot size The smaller spot size provides for a smaller machined feature size, for example, the demand to reduce exhaust emissions of automotive engines requires that fuel injectors be drilled with hole diameters below 100 µm. Moreover, reduced spot size increases fluence and power density of the same laser pulse energy. The focused spot size of a Gaussian beam in the focal plane is determined by properties of the laser beam and focusing lens as in Figure 2.18. The Rayleigh criterion suggested theoretically the smallest spot is about half the wavelength of the light used (Spencer 1982). The formula for calculating the minimum laser spot size is d = 4M�λ � �� . (2.1) _____________________________________________________________________ 34 Figure 2.18: Basic set-up to focus a Gaussian beam (Otani, Herbst et al. 2004). In Equation 2.1, λ is the wavelength of the radiation, M� is the beam quality factor (beam propagation ratio), D is the diameter of the beam at the focusing lens and f is the focal length of the lens, Figure 2.18. Using a beam expander with large magnification before the focusing lens or a focusing lens with shorter focal length are effective ways of getting smaller spot size for the same laser system. Another method is to use a mask, which contains feature sizes smaller than laser spot size. 2.3.1.3.2 Beam intensity Energetic optical pulses (µJ–mJ) with short pulse width (ns) and high focussed intensities (MW/cm2–GW/cm2) have proved capable of removing a wide range of materials from hard materials such as tungsten carbide and silicon carbide for tool technology to soft materials such as polymers for medical uses. Even ceramics, glass and diamond can be micro-machined with laser technologies (frequency tripled Nd:Vanadate lasers) with accuracy of less than 10 µm (Gillner, Holtkamp et al. 2005). High laser intensity is also necessary to overcome processing limitations that commonly arise from the high aspect ratio microvia drilling. 2.3.1.3.3 Depth of focus The depth of focus (DOF) of a laser beam depends on the modal structure of the beam, a general expression for the DOF of a general laser field does not exist. The lowest-order Gaussian mode, TEM00 is used for considering the DOF. One frequently used definition _____________________________________________________________________ 35 of the DOF is the distance over which the focal spot size changes ±5 %, which yields the following expression (Charschan 1977), Equation 2.2. DOF = ±0.32∙ ���� λ (2.2) where 𝑤� is the radius of the beam waist of the TEM00 beam. For a laser beam with a known M�, the DOF can be calculated from Equation 2.3. DOF = ±0.08∙λ∙ ����� ��∙λ � � (2.3) DOF is affected by the minimum laser spot size. When a beam expander is used before the focusing lens to reduce the spot size, the depth of focus will be smaller according to the formula. So both spot size and depth of focus need to be adjusted according to the application. In Si wafer surface patterning, a smaller spot size is preferred for high laser intensity. Because the wafer is very flat and smooth, a large depth of focus is not necessary. For laser cleaning, a large depth of focus is better since the surface to be cleaned is normally rough (debris of different sizes), and a larger spot size can increase the cleaning speed. For machining a corrugated topology or drilling into a relatively thick sample, an adjustable z-axis can be applied to continuously track the focal spot. 2.3.2 Machining mechanism When electromagnetic radiation strikes a surface, the force induced by the electric field is too small to vibrate an atomic nucleus, and the interaction occurs mostly between photons and electrons. The inverse Bremsstrahlung effect happens as photons are being absorbed by electrons, the Bremsstrahlung effect is the emission of photons from excited electrons (Steen 1998). As electrons vibrate, some reradiate in all directions, and others make collisions both among themselves and with lattice phonons (quantized microscopic vibrations in solid media). In good conductors, the mean free time between collisions for electrons is of the order of 10–14 to 10–12 seconds (Ready 1965). _____________________________________________________________________ 36 The lattice phonons then cause the structure to vibrate, and this vibration is transmitted through the structure as heat by the normal diffusion type processes. The diffused heat may result in melting and vaporization of the sample, when the molecular bonding is stretched and loosened due to strong molecular vibrations. The vapour with bound electrons can still absorb the radiation slightly, but when the gas gets sufficiently hot and electrons are shaken free the gas is then said to be plasma (Steen 1998). For practical purposes, laser irradiation can be regarded as a source of heat energy which is released at or near the surface of the substrate (Ready 1982). Thermal machining mechanisms include melting/ melt expulsion, vaporization and plasma effects, in most cases more than one phenomenon occur at the same time. Laser intensities, interaction time and materials used determine which mechanism is dominating the light matter interaction. The rate of material removal is determined by both evaporation and liquid- phase expulsion. 2.3.2.1 Melt expulsion Laser drilling frequently requires melt expulsion of the material. As laser energy is increased, mass removal in the form of micron-sized droplets can result from a hydrodynamic instability in the molten liquid layer (Zhang, Chu et al. 1997); (Brailovsky, Gaponov et al. 1995). The evaporating particles can exert a recoil pressure on the laser melted pool, and molten mass can be splashed away from the irradiated zone (Allmen and Blatter 1995); (Lee and Jeong 2004). Although melt expulsion is caused by the vaporization pressure, the whole process is still predominantly caused by melt expulsion. Melt expulsion is an energy efficient mechanism, and for limited laser beam intensities, it permits higher processing speed. The disadvantages are the generation of recast layers and wider extension of HAZ, which are not suitable for high precision micromachining. 2.3.2.2 Vaporization When the laser light intensity is high, enough pulse energy is transferred to the material which increases the matter temperature beyond vaporization threshold and overcomes the _____________________________________________________________________ 37 latent heat of fusion. Significant material vaporization is induced, and a cloud of vapour particles are formed. Figure 2.19: Laser-induced surface melting and vaporization (Bäuerle 2011). Thermalisation of species leaving the surface is intervened via collisions within a few mean free paths, typically within some microns from the surface. This region is called the Knudsen layer (Bäuerle 2011). The species leaving the surface generates a recoil pressure onto the substrate. In the presence of a molten surface layer and focused laser beam irradiation, the recoil pressure expels the liquid partially and the expelled liquids droplets resolidify again on the surface of material substrate as shown in Figure 2.19. The evaporation dominated mechanism has the advantages of smooth material edges, thinner recast layer, small extension of the oxidized surface and little HAZ, also loosely attached vapours can be washed off easily in an ultrasonic bath. Vaporization requires higher laser intensities and it is slower compare with melt expulsion dominated mechanism, and the generation of vapours can cause problems to the drilling process. In the range between material vaporization and ionisation of vapour, which is between 104 and 106 W/cm2 for CO2 laser radiation and metal substrates, interactions of the laser light with the vapour can often be ignored because vapour plume is not dense enough. To be more precise, this intensity range does have a loss due to absorption and scattering _____________________________________________________________________ 38 within the plume. There are slightly local intensity changes due to distortions of the beam shape by the hot vapour plume. 2.3.2.2.1 Plasma formation Plasma is caused by the interaction between the incident laser beam and the vapour plume generated from material substrate. With elevated laser intensities, the vapour becomes substantially ionized, and laser light becomes increasingly absorbed by photo- excitation and photo-ionisation of species within the plume, and this results in the formation of the plasma. Clearly there are no distinct division between plasma generation and vaporization. With plasma formation, the coupling between the laser light and the material becomes strongly nonlinear and material removal is affected. When laser light intensities are just above plasma formation intensity, the plasma is confined to a region near the surface. This intensity regime is employed in most types of laser micromachining because the effect on material removal is minimal. When the laser intensity is beyond the plasma formation intensity, the plasma plume expands in volume and its forward direction becomes more pronounced. When the intensity is even further increased, the plasma decouples from the substrate and propagates towards the incident beam. Finally when the intensity reaches some critical value, the laser light is essentially absorbed within the plasma and does not reach the substrate, this is the range of plasma shielding. The coupling of the plasma to the substrate becomes so weak that energy transfer is interrupted as well as the laser induced material vaporization (Bauerle 2000). The absorption of laser light by a plasma primarily occurs by an inverse bremsstrahlung process, which involves the absorption of a photon by a free electron (Singh and Narayan 1990). Based on the Saha equation, the absorption coefficient 𝛼� of the plasma can be expressed as 𝛼� = 3.69 × 10� � ��������.���� �1 − 𝑒�� �������, (2.4) _____________________________________________________________________ 39 where 𝑄, 𝑛�, and 𝑇�, are, respectively, the average charge, ion density, and temperature of the plasma, and ℎ, 𝑘, and 𝑣 are the Planck constant, Boltzmann constant, and frequency of the laser light, respectively. The laser energy is highly absorbed if (𝛼�𝑋) is large, where 𝑋 is the dimension perpendicular to the target of the expanding plasma. Equation 2.4 shows that the absorption coefficient of the plasma is proportional to 𝑛� . Thus, the plasma absorbs the incident laser radiation only at distances very close to the target where the densities of the charged particles are very high. The absorption coefficient of the plasma is inversely proportional to 𝑣, hence proportional to 𝜆, the laser wavelength, it means more laser energy is absorbed by the plasma at the longer wavelengths. In this equation, it is assumed that the plasma frequency is smaller than the frequency of the laser wavelength, otherwise, all the radiation would be reflected by the plasma. The term �1 − 𝑒�� ������� represents the losses due to stimulated emission. If we consider the excimer laser wavelength (𝜆 = 308 nm), the exponential term becomes unity for 𝑇� ≪40,000 K and can be approximated by �� ��� for 𝑇� ≫40,000 K. This absorption term shows a 𝑇���.� dependence for low temperatures ( 𝑇� ≪ 40000 K for 𝜆 =308 nm, 𝑇� ≪10,000 K for 𝜆 =1.06 µm) and 𝑇���.� for high temperatures. The plasma generated during the excimer laser (193 nm and 248 nm) ablation of Si was estimated to have pressures of several hundred bar (Shinn, Steigerwald et al. 1986). The energy density in the plasma and the pressure waves propagating to the surface are large enough to rapidly expel material from the liquefied zone. This offers efficient machining with increased laser intensity, since molten Si is expelled rather than vaporised. These shock waves are called laser supported combustion waves (LSC waves). At still higher laser intensities, there is further increase in the plasma density and temperature, and the laser-supported absorption wave propagates with supersonic velocity (Bäuerle 2011). Laser absorption occurs almost entirely in a small region near _____________________________________________________________________ 40 the crown of the plasma causing explosive heating of this region, initiating laser supported detonation waves (LSD waves), and less efficiency in the coupling of the laser energy to the substrate surface (Schittenhelm, Callies et al. 1998); (Chang and Warner 1996). High temperature plasma resulting from this form of laser heating has reported lifetimes that keeps the ablation process going through electron conductive heating (Boley and Early 1994). As an example, Figure 2.20 includes the time-resolved shadowgraph images of Nd:YAG laser (266 nm wavelength and 3 ns pulse width) machining of Si target. The shock wave was expanding within 50 ns, then ejection of large particulates from Si surface happened 400 ns later and lasted up to several tens of microseconds. Explosive boiling within a deep superheated layer was suggested to be a dominant mechanism at laser irradiance greater than 22 GW/cm2 (Yoo, Jeong et al. 2000); (Yoo, Jeong et al. 2000); (Russo, Mao et al. 1999). The bulge (additional shock wave) at the centre of the shock front, as seen in Figure 2.20 at 5 ns delay, was resulted by the plasma absorbing the latter part of the laser beam. _____________________________________________________________________ 41 Figure 2.20: Sequence of images obtained by laser shadowgraphy for the laser irradiance of 8.7×1010 W/cm2 showing shockwave and particulate emission (Yoo, Jeong et al. 2000). Plasma is often generated with ultrashort pulses, energy is deposited in the material at such a fast rate, and the material is forced into a plasma state. The material goes from a solid to a gas phase without first going through a melt phase. This plasma then expands away from the material as a highly energetic gas, taking almost all the energy away with it. The use of ultrafast lasers has the same result as the absorption of single high-energy photons, electrons of a molecule are hit by multi-photons simultaneously and solid material are removed as highly ionized plasma (Meijer, Du et al. 2002). Wavelength becomes less important in this case, and ultrafast lasers normally operate in the IR part of the spectrum with relatively simple optics. _____________________________________________________________________ 42 Plasma generated with femtosecond (Ti:sapphire, 100 fs, 800 nm) and nanosecond (Nd:YAG, 3 ns, 266 nm) laser on Si was compared (Zeng, Mao et al. 2005), and shadowgraphs of the shock waves generated by femtosecond and nanosecond lasers at 10 ns after a laser pulse were obtained in Figure 2.21. Figure 2.21: Plasma images obtained by laser shadowgraphy for femtosecond (left) and nanosecond (right) laser ablation of Si at 10 ns after the laser shot. The shock front (S), ionization front (I), contact front (C) and air breakdown plasma (A) are labelled in the figures (Zeng, Mao et al. 2005). There was a thin layer of shocked air between the shock front and the ionization front, and there was an ionized air between ionization front and contact front. In femtosecond laser ablation, air breakdown was induced because of the much higher laser irradiance of 112 TW/cm2, compared with the irradiance of 3.7 GW/cm2 for the nanosecond laser. The shock wave generated by the nanosecond laser expands spherically, which indicates the plasma expands in the perpendicular and lateral direction at similar velocities. The perpendicular expansion velocities were obtained by differentiating the point explosion model, in which the expansion distance between the surface of the target and the position of a blast wave depends on the energy converted into the plasma state. At 5 ns the perpendicular expansion velocities were 13.6×105 cm/s for fs- plasma and 7.9×105 cm/s for ns- plasma, the fs-plasma kinetic energy is therefore higher than ns-plasma at the early times. _____________________________________________________________________ 43 The electron number density values as a function of time for fs- and ns- plasma was determined, Figure 2.22, one can see there is an intersection at 20 ns. The reason could be explained by the way that plasma expands in Figure 2.21. The fs- plasma exploded in the direction of laser light, and electrons and particles escaped, while the ns- plasma was confined to the surface region, and electron number density accumulated and was in excess of fs laser at a distance of 0.6 mm above the target surface after 20 ns. Figure 2.22: Electron number density as a function of time for femtosecond and nanosecond laser ablation at a distance of 0.6 mm above the target surface (Zeng, Mao et al. 2005) Plasma generation can sometimes be used to promote drilling rates. Machining of transparent material (fused silica windows, Corning 7980) from the rear side was demonstrated to increase the processing rate by a factor of two with an IR and UV nanosecond pulsed laser (Salleo, Sands et al. 2000). Plasma created at the drilling front was propagating into the bulk material and the drilling rates were in excess of 100 µm per pulse. _____________________________________________________________________ 44 2.3.3 Machining processes Various lasers have been used to machine Si, most of them are working at the UV wavelength because of the high absorption coefficient. Table 2.4 summarises the wavelength, power, pulse width and some applications of some popular lasers for Si micromachining. Table 2.4: Laser types with Characteristics and Applications for Si micromachining. Wavelength Pulse width Applications Nanosecond lasers Excimer laser 157, 193, 248, 308 nm 10 ns - 100 ns Micro-machining, laser chemistry, medicine UV lithography, annealing of LCD displays Copper vapor laser 511, 578 – 611 nm (255, 289 nm frequency doubled) 6 ns - 120 ns Drilling, cutting, micro milling precision machining development, writing Fiber Bragg Gratings Solid state laser 1064, 532, 355, 266 nm 10 - 100 ns Materials processing (marking and micro structuring), medicine Fiber laser (Solid state) 1064 nm 20 – 400 ns Marking, micromachining Pico- & Femto-second lasers Diode pumped solid state laser or fibre laser 1064, 532, 355 nm 10 ps Micromachining, micro drilling, thin film scribing Ti-Sappire 775 nm 100 fs Micromachining The processing capability of any laser is strongly dependent on the spot size and power density. Industrial laser developers have made steady progress in delivering higher quality beams with ever increasing power levels. Excimer lasers offer applications that need high average output powers (up to several 100 W), large pulse energies (10 mJ to 1 J) at pulse durations around 30 ns by a discrete UV wavelength between 157 and 308 nm (Meijer, Du et al. 2002). The wavelength range of an Excimer laser exhibits high absorption on Si, but the large beam size and flat-top beam profile can only enable ablation of less than 1µm per shot with fluence up to 80 J/cm2, and the process time is long due to low frequencies and masks are often used _____________________________________________________________________ 45 (Shinn, Steigerwald et al. 1986); (Desbiens and Masson 2007). Surface modification of Si (100) was realised with an excimer laser (193 nm, 23 ns), in which whisker like structures grew after irradiation of several hundred low fluence laser pulses as shown in Figure 2.23 (Sánchez, Morenza et al. 1996). These columns were not formed by preferential etching, they grew due to a hydrodynamic process, and their height (20-30 µm) is between one and two orders of magnitude higher than the depth of the crater. They are thought to be microscopic local areas (impurities, different chemical composition) which are more resistant to ablation, thus ablated more slowly than the surrounding material. (a) (b) Figure 2.23(a): SEM microphotographs of Si irradiated at 1.7 J/cm2 with 500 laser pulses at 45° view. (b) Magnified view of (a). (Sánchez, Morenza et al. 1996) (Riet, Nillesen et al. 1993) noticed that the Si surface roughness decreases when higher laser fluence pulses were used from a Lambda Physik EMG 201 MSC excimer laser (308 nm wavelength, 28 ns pulse length). The centre of the etch crater remains relatively flat at the higher fluence, the boundary of the crater is rough where the fluence is lower, Figure 2.24(a). The whole crater is rough at the lower fluence, Figure 2.24(b). _____________________________________________________________________ 46 Figure 2.24: Etch crater on a Si target after being irradiated with 300 pulses at constant angles of incidence (40°) at (a) 6 J/cm2 and (b) 3 J/cm2 (Riet, Nillesen et al. 1993). Copper vapour laser (511/578 nm) was used to drill Si (001) but micro-cracking has limited its use, Figure 2.25. Holes approximately 120 µm in depth with a 25 µm diameter were drilled with copper vapour laser of 50 ns pulse width, 0.1 mJ pulse energy, 6.5 kHz pulse frequency, 15 J/cm2 fluence, 3 x 108 W/cm2 intensity and 50 pulses in air. _____________________________________________________________________ 47 Figure 2.25: SEM image of ion-polished cross-section 10 µm below the upper side of the wafer through a hole drilled in silicon with the copper vapour laser (Luft, Franz et al. 1996). The formation of radial and peripheral cracks was caused by the sudden rise in yield stress with rapid falling temperature (Luft, Franz et al. 1996). Multi-pulse drilling deposited a large amount of molten material on the sidewall as hole became deeper, the recast layer and local heat diffusion into the adjoining material lead to thermal distortion and mechanical distortion, and a larger HAZ was established on the sidewall than at the bottom of the hole (Körner, Mayerhofer et al. 1996). 2.3.3.1 Solid state laser In the late nineties, the introduction of diode pumped solid state (DPSS) laser systems heralded a new wave of processing capability with high quality beams and high efficiency systems. The DPSS lasers are, compared to excimer lasers, more compact, nearly maintenance free, and offer better beam quality which enables direct processing and a wider range of wavelength, pulse durations and pulse frequency. _____________________________________________________________________ 48 Processing Si at 1064 nm often shows signs of micro cracking, microcavities (Craciun and Craciun 1999), heavy recast layers, and surface damage. These effects are largely due to the low absorption coefficient of Si at 1 µm wavelength. The optical penetration length of Si at 1 µm is 200 µm, compared with 0.01 µm at 355 nm (Bauerle 1996); (Palik 1991). At 1 µm, the laser penetrates a greater distance and heating becomes much less effective with limited pulse energies. (Dauer, Ehlert et al. 1999) investigated the influence of the position of the focal plane on micro-cutting of Si using a Q-switched Nd:YAG laser with optional frequency doubling. A 450 µm thick Si wafer (N 100) was cut at 1064 nm wavelength at a cutting speed of 2.36 mm/s at various focal positions shown in Figure 2.26. The laser was working at 3 kHz pulse frequency and 4.58 W average power, the focal length of the lens is 68 mm. Figure 2.26: Influence of the focal position on the shape of the cutting edge and on the depth of cut (Dauer, Ehlert et al. 1999). The optimum focal position was determined to be 0.5-1 mm below the top surface of the _____________________________________________________________________ 49 wafer by varying the focal position between 1.5 mm above and 2.5 mm below the top surface of the wafer. Parallel cutting edges were achieved at optimum focal position, but a large amount of the solidified molten silicon remains in the cutting kerf shown in Figure 2.27. Figure 2.27: Laser cut in silicon wafer using 1064 nm wavelength. The debris consisted of solidified Si melt and SiO2, and they reduce mechanical strength of the processed substrate. The oxide can be removed by wet chemical etching in 40 % HF solution. The solidified melt could only be removed in a wet etching process in 30 % KOH at 343 K for 10 minutes, and the mechanical strength of Si was then recovered to its original value. The frequency doubling option (532 nm) could not deliver sufficient average power, and the wafer could not be cut through. The DPSS laser working at UV wavelength is considered as the state of art for micromachining Si. The focused laser beam with short focal length and a Gaussian intensity spatial profile enables high laser intensity, together with an optical trepanning device, DPSS UV can produce high quality, high aspect ratio holes (Karnakis, Rutterford et al. 2005), Figure 2.28(a). (Karnakis 2006) also reported an unusual single pulse ablation depth of up to 60 µm for laser intensities exceeding ~11.5 GW/cm2, which was achieved on monocrystalline Si with a nanosecond Nd:YAG laser at 355 nm. Percussion drilled through holes on Si wafers lead to deformations in hole shape, which were caused by high recoil pressure and expanding vapour and plasma. Drilling with two cycles of _____________________________________________________________________ 50 burst pulse train was demonstrated to achieve smooth but tapered sidewalls (Tan 2006), Figure 2.28(b). DPSS UV laser drilling and micromachining of Si often leaves a certain amount of solidified molten Si in the ablation area, which can be cleaned and etched using a 22% KOH solution at 348 K for 4 min (Chen and Darling 2005), Figure 2.28(c). DPSS UV lasers are also used for selective ablation of amorphous Si on glass for thin film photovoltaic applications (Molpeceres, Lauzurica et al. 2007), Figure 2.28(d). _____________________________________________________________________ 51 (a) (b) (c) (d) Figure 2.28(a): Laser trepanned holes in silicon using high intensity (15 GW/cm2) at 355 nm (Karnakis, Rutterford et al. 2005). (b): Deep through hole with straight sidewalls drilled by two-cycle trepanning, the entrance is 50 µm and exit is 40 µm (Tan 2006). (c): SEM image of silicon machined by frequency tripled Nd: YAG laser (355 nm) followed by cleaning in the KOH solution (Chen and Darling 2005). (d) Profiles and SEM images (×1.5 k) of ablated a-Si under different irradiation conditions at 248 nm (Molpeceres, Lauzurica et al. 2007). _____________________________________________________________________ 52 The thin disk is a power scalable concept of the diode-pumped high-power solid-state laser, which was demonstrated in the 1993 by the group of Adolf Giesen at the University of Stuttgart, Germany. The main difference to conventional rod lasers or slab lasers is the geometry of the gain medium. The laser crystal is a thin disk, where the thickness is considerably smaller than the laser beam diameter, so that the generated heat is extracted predominantly through one end face, i.e., in the longitudinal rather than in the transverse direction (Giesen, Hügel et al. 1994). The cooled end face has a dielectric coating which reflects both the laser radiation and the pump radiation. Furthermore, lasers with high power (up to 100 W, fundamental mode) are achieved with a higher electrical efficiency than that of all other commercially available solid state lasers with similar power (Giesen and Speiser 2007). 2.3.3.2 Ultrafast laser Ultrafast (pico- and femto-second) lasers deliver an astonishing amount of peak power of several GigaWatts with only a fraction of mJ pulse energy, and the laser intensity easily reaches hundreds of terawatts per square centimeter at the work spot itself. For femtosecond lasers, the duration of the laser pulse is shorter that the heat relaxation time (Madou 2002), which is in the order of several picoseconds (Meijer, Du et al. 2002). The hydrodynamic motion of the matter under laser irradiation can be ignored (Liu, Du et al. 1997). Ultrafast lasers are used for high precision micromachining. A number of limiting factors, such as heat affected zone (HAZ), micro cracks and recast layer, are not found after machining with the appropriate parameters, Figure 2.29 & Figure 2.30. Ultrafast lasers require less fluence to initialize ablation and they generate laser-induced optical breakdown through avalanche ionization and multiphoton ionization, which is useful in the ablation of transparent materials (Liu, Du et al. 1997). _____________________________________________________________________ 53 Figure 2.29(a-d): Femtosecond-pulse laser processing of a 0.3 mm thick silicon target with 250 fs, 0.5 mJ, and F=2.5 J/cm2 second harmonic radiation (λ=309 nm) and different number of pulses: (a) 10, (b) 100, (c) 5×103, (d) 104 pulses (Chichkov, Momma et al. 1996). (a) (b) (c) Figure 2.30(a) & (b): Examples of silicon structuring with fs laser pulses by trepanning with a spot focus. (c): Kerf in thin silicon, cut by static exposure to fs laser pulses shaped to a line focus (Bärsch, Körber et al. 2003). _____________________________________________________________________ 54 Femtosecond laser machining is more efficient, because the plasma is generated after the laser pulse and there is no energy loss through plasma. The crater depths were compared between femtosecond and nanosecond laser machining of Si, and plotted with respect to the pulse number, Figure 2.31. The machining efficiency of the 266 nm laser was 0.63 µm per pulse for ns ablation and 1.4 µm per pulse for fs ablation. The femtosecond laser produced less re-solidified molten Si around the crater, inset in Figure 2.31. Figure 2.31: Ablation depth vs. pulse number for femtosecond and nanosecond laser ablation of Si in air. The insets show the fs and ns crater profile after five laser pulses, notice that two axis are of different unit (Zeng, Mao et al. 2005). In addition to material ablation applications, ultrafast lasers are important in metrology and shadowgraphy. The first isolated soft-X-ray attosecond (10-18 s) pulses have been used to trace electronic dynamics (Hentschel, Kienberger et al. 2001), which the sources of attosecond pulses are the high-order harmonics of femtosecond lasers radiation (Paul, Toma et al. 2001). Femtosecond laser was used as the probe beam to look at the plasma expansion, particle generation and shock wave of laser ablation of Si in laser shadowgraphy (Zeng, Mao et al. 2004); (Liu, Mao et al. 2007). Ultrafast lasers have low throughputs and high cost of ownership; they are more often used in research labs rather than production lines. The low machining speed is caused by _____________________________________________________________________ 55 the low pulse energy, because majority of ultrafast lasers work at megahertz and the average power is limited. The creation of plasma caused by high peak power absorbs and screens away the next incoming laser pulse at high repetition rates, laser-plasma interaction can change dramatically both the laser beam intensity distribution on the solid surface and the laser plume expansion features. Furthermore, when effects of femtosecond laser was analyzed at the energy above damage thresholds in the line scanning of Si, although cleaner surface with little contamination was observed at lower energy levels, redeposition from the melt droplets and vapour condensation were found at higher fluences (Ngoi, Venkatakrishnan et al. 2001), Figure 2.32. (a) (b) Figure 2.32(a): Trench micromachined by laser beam with an energy of 50 nJ/pulse and (b): an energy of 200 nJ/pulse. High power machining resulting in surface contamination due to ejection of melt materials and thus, a large heat affected zone (Ngoi, Venkatakrishnan et al. 2001). _____________________________________________________________________ 56 2.3.3.3 Hybrid processing of Si Si has been laser processed with different techniques to remove debris. Si machining with an excimer laser (FWHM = 25 ns, λ = 248 nm) was explored in air and under water, the machining under water prevents debris from re-depositing on the finished surface, and it had the same effect as covering Si surface with an adhesive tape (Choo, Ogawa et al. 2004). The ablated surface is smooth after machining in air, while machining under water shows the resolidified dendritic structure. The laser ablation rate of Si was found to vary with the thickness of the water layer above the substrates, the most highly enhanced thickness was 1.1 mm (Zhu, Lu et al. 2001). The acoustic wave generated during laser ablation was found strongest at a 1.1 mm thickness, which indicated the plasma generated in the water confinement regime induced the strongest pressure, and resulted in highest ablation rate. Laser induced chemical etching is a common process, and with the assist of chemicals, the droplets produced during the process can form gas and be carried away from the work area. Such methods can also increase the removal rates and leave a cleaner machined surface afterwards. For machining Si, chlorine is often used as an etchant (Müllenborn, Dirac et al. 1996), because the Si vapour and debris will react with Cl2 to form SiCl4 gas. At the MIT Lincoln Laboratories, laser scanning at the speed of 7500 µm/s is used to remove Si layers in a Cl2 etchant (Bloomstein and Ehrlich 1992). When each layer of 1 µm is removed, the focusing objective is lowered 1µm and refocus, and a new pattern is etched. The multilevel-type structures were produced with this technique such as the cone inside dish in Figure 2.33. In addition, Br2, HCl, XeF2, KOH, etc., can also be used as the etchant in the laser photochemical etching of Si (Madou 2002). _____________________________________________________________________ 57 Figure 2.33: SEM image of cone inside dish laser etched into Si using chlorine ambient of 100 Torr. The dish has an upper radius of 80 µm and lower radius of 35 µm with radius of curvature of 82 µm and spacer layers extending 8 µm inward. The clipped cone has upper radius of 4 pm and lower radius of 16 µm. The laser was swept in a circular scan in each plane (Bloomstein and Ehrlich 1992). However, Cl2 is toxic and require a complete circulation system if used in industry, and it increases the cost of machining, and it is not applicable. Another method is using wet chemical etching after laser structuring to remove the Si damaged by the laser process. In this case, the thermal diffusion length and the optical penetration length of chosen laser system, i.e., the thickness of the thermally damaged layer, should be lower than the allowed wet chemical etching removing layer thickness. It is particularly critical in laser texturing, because laser processed patterns should never be distorted by etching process. 2.3.4 Machined depth A relationship between light absorption and the single pulse etch depth can be established by making some assumptions. If etching is viewed as a two-step process in which the initial light absorption is followed by material ablation, then the Beer-Lambert law can be utilized: _____________________________________________________________________ 58 𝐹� = 𝐹��� ∙ 𝑒���. (2.5) In Equation 2.5, 𝐹� is the attenuated energy fluence at depth z into the material, 𝐹��� is the incident pulse fluence, and 𝛼 is the absorption coefficient at the given wavelength. Assume that a material threshold fluence must be met or surpassed in order for ablation to occur. If it is now required that 𝐹� equals this threshold value 𝐹����� then 𝐹��� must be increased over 𝐹����� by a factor of 𝑒��� so that single pulse etching occurs to the desired depth z. The factor 𝑒��� compensates for the exponentially decreasing value of 𝐹 as light penetrates into the material. A rearranged form of Equation 2.5 is obtained: 𝐹��� = 𝐹����� ∙ 𝑒���. (2.6) The single pulse etch depth 𝑍 is then 𝑍 = 𝛼��ln ( ���� ������ ). (2.7) Equation 2.7 has been utilized in one form or another by several authors (Jellinek and Srinivasan 1984; Brannon, Lankard et al. 1985). 2.3.5 Fiber laser In the past the fiber laser operated in the telecommunications industry, providing high quality single mode low power systems with a mean time between failures of some 25 years, enabling the “bury and forget” approach to communication installations. With the contraction of the telecommunications markets in the late 1990’s, fiber laser producers shifted their focus on meeting the needs of industrial manufacturing, military, medical, and aerospace sectors. The shift to alternative markets requires higher power laser developments with an emphasis on operational and performance targets that must be reached to take a share of the multi-billion dollar laser materials processing markets (Gapontsev, Platonov et al. 2003); (Norman, Zervas et al. 2004). Fiber lasers have been around since Snitzer’s proposal and subsequent demonstration in 1961 (Canning 2006). The practical values of fiber laser were not demonstrated until mid _____________________________________________________________________ 59 1980s with improved fabrication methods at University of Southampton in the UK, but their work has been limited mostly to laboratory. With the maturation of fiber grating and double-clad technology, fiber lasers have yielded real commercial opportunities for material processing market. Modern fiber lasers use doped fibre as gain medium, it is a glass fibre doped with rare earth ions such as Erbium (Er3+), Neodymium (Nd3+), Ytterbium (Yb3+), Thulium (Tm3+), or Praseodymium (Pr3+) (RP Photonics web). Optical fibres on the other hand are a kind of waveguide, which is a spatially inhomogeneous structure for guiding light. The fibres are usually made of some kind of glass, which can be as long as hundreds of kilometres, and are fairly flexible compared to other waveguides. The most commonly used glass is silica, it is so widely used because of its outstanding properties, in particular its potential for extremely low propagation losses and its amazingly high mechanical strength against pulling and even bending. Two common types of fibre lasers are the Erbium-doped and Ytterbium-doped Fibre Laser. The rare earth atoms are mixed into the core of the fibre when it is made. The Er and Yb dopants provide energy levels that can absorb photons with a wavelength of 980nm from a low cost diode laser pump source, and then decay to a meta-stable state and generate a potentially very high power beam at 1550nm and 1060nm respectively. Within the doped fibre, the Er and Yb atoms form laser medium. Light, in order to be guided along an optical fibre, must be confined to a central core by reflection from the cladding that surrounds it (Cregan, Mangan et al. 1999), Figure 2.34. To create the lasing cavity, Bragg Gratings are added. A Bragg Grating is a section of glass where the refractive index has been changed, and the grating acts like a very efficient mirror. Each time the light goes across a boundary between one refractive index and another, a portion is reflected back. _____________________________________________________________________ 60 Figure 2.34: Schematic Diagram of fibre laser (ORC 2009) The diode pump laser source can be stacked to pump a single fibre laser with large amounts of power. The diode laser beam is focused into the large cladding around the core shown in Figure 2.34, and the fibre is clad with an outer sheath to contain the pump laser beam. The pump beam reflects inside the fibre, which a part is absorbed every time the beam crosses the core. 2.3.5.1 Single mode fibre Transverse electromagnetic mode (TEM) distributions are controlled by the waveguiding characteristics of the active core. The core of a fibre is the region in which the light is guided, usually it is a region of slightly increased refractive index. Single mode fibres usually have a relatively small core, with a diameter of only a few microns as in Figure 2.35. This fibre core is surrounded by an inner cladding layer heavily multimode at the pump wavelength. This inner cladding layer with slightly lower refractive index promotes rapid and efficient pump light transfer into the core. The outer cladding layer which has the lowest refractive index is made of polymer or glass. It is around the inner cladding, and confines the pump light within the outer core. _____________________________________________________________________ 61 Figure 2.35: Double-clad fibres convert the output of a high-power semiconductor pump laser to a diffraction-limited output beam in simple, end-pumped fibre laser cavities. Inset shows refractive index profile typical of a cladding-pumped fibre (Richardson, Minelly et al. 1997). The dual-clad fibre design has become standard because its outer core can efficiently collect pump light, while the smaller inner core concentrates laser power in a smaller volume for a higher quality output beam. The outer core is often non-circularly symmetric so that the pump power can be absorbed better. In Figure 2.35, the inner cladding layer is adapted in shape and dimensions to allow efficient end-coupling of the output from high-power semiconductor diode stripes and arrays (Richardson, Minelly et al. 1997). 2.3.5.2 Advantages Fibre lasers are unique in offering instant operation without the need of regular tuning and realignment due to their robust all fibre, all guided architecture. The inherently large surface to volume ratio facilitates heat removal and minimizes the external cooling requirements. There are no thermal lensing effects, which in turn results in excellent beam pointing stability. The beam quality can be engineered by proper fibre core design to match the application requirements. Fibre lasers are usually pumped by combining a _____________________________________________________________________ 62 number of extremely robust, telecommunication grade, single emitter broad stripe multimode pump diodes. This results in laser systems with long lifetime, maintenance free operation. Fibre lasers show very small and compact footprint that facilitates system integration (Richardson, Minelly et al. 1997). Fibre lasers also offer high operating efficiencies, maximum levels are of the order of 25%, a factor 2 better than current DPSS systems (O'Neill and Li 2009). The cost of a fibre laser is 8-10 times less than the DPSS system. There are a wide range of fiber lasers with a choice of power levels and output quality that cater for a host of industrial applications. Commercial products are provided by IPG with output powers <50 kW, and SPI Lasers with output powers <400 W. Other manufacturers operating in the low to medium power regimes are IMRA, Spectra Physics Inc, Synchronous Inc, Optigain Inc, and GSI. The single mode systems, of high power levels (<5 kW), are increasingly applied in materials processing and medical applications. It is clear that single mode fibre lasers have stolen a march on other low power laser sources, with higher power systems already challenging the existing CO2 and Nd:YAG industrial systems. 2.4 Analytical models Laser material interactions depend on a number of laser parameters and material optical and thermal properties. Based on these factors, numerous models were developed to predict temperature distributions, energy coupling and machined depth. In order to understand how the fibre laser interacts with Si in greater detail, a model that describes the interaction is essential. Laser irradiation in nanosecond regime is normally considered as a heat source. There are three distinct methods of heat flowing from the hotter parts to the cooler. The first one is conduction, in which heat passes through the substance of the body itself. The second one is convection, in which heat is transferred by relative motion of portions of the heated body. The third one is radiation, in which heat transfers directly between distant portions _____________________________________________________________________ 63 of the body by electromagnetic radiation (Carslaw and Jaeger 1959). In liquids and gases, convection and radiation are crucially important, but in solids, conduction is the one considered. Thermal conductivity (κ) is an important property, it indicates material’s ability to conduct heat, and it depends upon temperature. When the range of temperature is limited, this change in thermal conductivity may be neglected. Thermal conductivity also depends on the material phase, it is high in solids and low in liquids. By the definition of conductivity, a solid is homogeneous or isotropic, such that when a point within is heated, the heat spreads out equally well in all directions. As opposed to crystalline and anisotropic solids, in which certain directions are more favourable for the conduction of heat than others and the conditions of conduction vary from point to point (Carslaw and Jaeger 1959). For the ease of calculations, Si is normally assumed to be homogeneous. 2.4.1 Thermal conduction model (Ready 1963) was the first to use the differential equation for one-dimensional heat flow, Equation 2.4, a closed form analytical solution in a semi-infinite solid was developed to calculate the temperature rise (Carslaw and Jaeger 1959), Equation 2.9. 𝜕2𝑇(𝑧,𝑡) 𝜕𝑧2 − 1𝐷 𝜕𝑇𝜕𝑡 = − 𝐼(𝑡)𝑒−𝛼𝑧𝜅 (2.8) 𝑇(𝑧, 𝑡) = ��(�) � √𝐷𝑡 ∗ 𝑖𝑒𝑟𝑓𝑐 � � �√�� � − �(�)���� �� + �(�) ��� 𝑒�� ����� ∗ 𝑒𝑟𝑓𝑐 �α√𝐷𝑡 − � �√�� � + �(�) ��� 𝑒�� ����� ∗ 𝑒𝑟𝑓𝑐 �α√𝐷𝑡 + � �√�� � (2.9) 𝑇 is the temperature, 𝐷 the thermal diffusivity, 𝜅 the thermal conductivity, 𝛼 the absorption coefficient, and 𝐼 the heat production per unit area per unit time; 𝑒𝑟𝑓𝑐 and 𝑖𝑒𝑟𝑓𝑐 denote the complementary error function and its integral. In the following study, (Ready 1963) plotted the temperature distribution at the surface of the carbon block in air, Figure 2.36, a laser beam with a 30 MW peak power was focused _____________________________________________________________________ 64 on an area of about 3x10-3 cm2. He also observed with high speed framing camera (10 ns frame time) that the emission of vaporised material does not begin until after the surface temperature pulse has declined from its peak value, Figure 2.36. Figure 2.36: The shape of the temperature pulse at the carbon surface when the laser beam is absorbed. The temperature is calculated using the differential equation for one- dimensional heat flow and assuming a laser pulse as shown. Both the temperature pulse and the laser pulse are normalized to their peak values (Ready 1963). (Ready 1965) predicted that a longer, lower pulse shape can ablate a deeper hole than a shorter, higher pulse shape with the same pulse energy. In the case of a giant laser pulse (nanoseconds pulse width, 2 MW peak power, and 109 W/cm2 peak power density) produced by a Q-switched laser, the formation of superheated state was caused by the recoil pressure of vaporised material. The maximum depth at which metals can be heated above the critical point was estimated. Different laser parameters including peak power density, total energy density, pulse width, and the time from start of pulse to peak power, were used in the calculations, but the knowledge of high temperature thermal properties of the metals were limited and were not included in the calculations. Besides that, the model only dealt with the heating cycle (not the cooling cycle), it neglected the liquid phase, as a metal’s latent heat of melting is much smaller than its latent heat of _____________________________________________________________________ 65 vaporization, it neglected re-radiation of energy from the surface, and it considered that the vapour plume did not interface greatly with energy in the laser pulse. (Dabby and Paek 1972) introduced subsurface heating caused by a Q-switched laser pulse (pulse width of less than 100 ns, intensity of 3.25x108 W/cm2), and with temperature higher below the material surface than that at the surface, explosive material removal occurred. The analytical solution was based on the front-surface energy balance and heat conduction model, Equation 2.10, Equation 2.11, which allowed for transmission of laser energy past the front surface. 𝜌𝐿�?̇? = 𝐷 ���� ⃒z=Z’ (2.10) 𝜌𝐶� �� �� = 𝛼𝐼𝑒��(���) + 𝐷 ��� ��� , 𝑧 > 𝑍 (2.11) 𝐼 is the laser beam intensity in W/cm2, 𝐷 the thermal conductivity in J/cm·s K, 𝛼 the absorption coefficient in cm-1, 𝜌 the density of the solid in g/cm3, 𝐶� the specific heat in J/g K, 𝐿� the latent heat of vaporization in J/g, 𝑍 the depth to which material has been removed in cm, and ?̇? the velocity of the front surface in cm/s. The energy balance requires that the energy given to the vaporised material equals the energy conducted from the solid. _____________________________________________________________________ 66 Figure 2.37: Temperature versus distance into the solid at different normalized times. The model predicted material removal velocity by vaporization (moving boundary) and the temperature profile in the metal solids. As a result, the temperature inside the material was higher than the surface temperature, Figure 2.37, this was because that energy at the surface region was removed by the front surface vaporization. If the sufficient heat generation within the solid could cause vaporization beneath the surface, tremendous pressure arising from vaporised material would explosively remove the intervening material. This mechanism is more efficient than conventional drilling by vaporization, and material removal would continue after the end of laser pulse. The model assumed that laser beam intensity was sufficient to cause surface vaporization, the vaporised gas was transparent to incident laser energy, no heat was lost due to re-radiation, optical absorption coefficient and material thermal properties did not depend on laser intensity and temperature, and radial heat conduction and liquid phase was ignored. (Lax 1977) developed a closed-form solution for the maximum temperature rise induced by a laser beam of the spatial Gaussian distribution absorbed in a solid. The temperature rise was written 𝑇 = 𝑇���𝑁(𝑅,𝑍,𝑊), where 𝑅 = 𝑟/𝜔, 𝑍 = 𝑧/𝜔, 𝑊 = 𝛼/𝜔, 𝑇��� is the _____________________________________________________________________ 67 maximum temperature rise when the heat is absorbed in an infinite small layer at the surface, 𝑁 is the normalized temperature rise value ( � ���� ), 𝜔 is the beam waist, 𝑟 the radial distance from the centre of the beam, 𝑧 the depth beneath the sample surface, 𝛼 the absorption coefficient. Figure 2.38: Normalized temperature rise N(0, 0, W) at the surface (z = 0) and beam center (r = 0) for a Gaussian laser beam shape plotted versus W. The upper (solid) curve corresponds to the horizontal scale shown. The lower (dashed) curve is the upper curve replotted with the horizontal scale expanded by a factor of 5 to facilitate reading points in the small W regions. _____________________________________________________________________ 68 (a) (b) Figure 2.39: Normalised temperature rise as a functioin of R and Z for (a) W=0.8 and (b) W=12. The normalized surface temperature N(0,0,W) in the beam centre was increasing with the absorption coefficient of beam in the solid, Figure 2.38. The temperature rise beneath the surface decreased with the increasing distance from the surface, and the radial distance from the beam centre, Figure 2.39. If the heat source was confined to an infinitesimal layer near the surface, i.e. the laser beam attenuation length approaches zero, the maximum temperature rise would depend on the integrated beam power, thermal conductivity, and the beam radius. The model assumed the semi-infinite medium and neglected heat flow from the single solid surface to the air. (Chichkov, Momma et al. 1996) presented the one-dimensional, two-temperature diffusion model developed by (Kaganov, Lifshitz et al. 1957) and (Anisimov, Kapeliovich et al. 1974) during the interaction of low intensity short laser pulses (0.2- 5000 ps) with metal targets, Equation 2.12-Equation 2.15. A thermal conductivity in the lattice subsystem (phonon component) is neglected. 𝐶� ��� �� = −��(�) �� − 𝛾(𝑇� − 𝑇�) + 𝑆, (2.12) 𝐶� ��� �� = 𝛾(𝑇� − 𝑇�), (2.13) _____________________________________________________________________ 69 𝑄(𝑧) = −𝑘� ����� , (2.14) 𝑆 = 𝐼(𝑡)𝐴𝛼𝑒���. (2.15) 𝑇� is the electron temperature, 𝑇� is the lattice temperature, 𝑧 is the direction perpendicular to the target surface, 𝑄(𝑧) is the heat flux, 𝑆 is the laser heating source term, 𝐼(𝑡) is the laser intensity, 𝐴 = 1 − 𝑅 and 𝛼 are the surface transmissivity and the material absorption coefficient, 𝐶� and 𝐶� are the heat capacities (per unit volume) of the electron and lattice subsystem, 𝛾 is the parameter characterizing the electron-lattice coupling, 𝑘� is the electron thermal conductivity. In the nanosecond regime, there is enough time for the coupling between electrons and lattice to reach equilibrium, i.e. 𝑇� and 𝑇� are equal and Equation 2.12-Equation 2.15 can be reduced to 𝐶� �� �� = �(������) �� + 𝐼�𝛼𝑒���, 𝐷 = 𝑘𝑜𝐶𝑖, (2.16) where 𝑘� is the conventional equilibrium thermal conductivity of a metal, 𝐷 is the heat diffusion coefficient. With nanosecond laser ablation, the thermal wave propagates into the target and a relatively large layer of melted material is created, which the evaporation occurs from the liquid metal, Figure 2.40. Figure 2.40: Schematic of nanosecond-pulse laser ablation. _____________________________________________________________________ 70 (Ren, Orlov et al. 2004) showed that the vapour phase front propagation rate is more relevant to the material removal rate when surface vaporization was dominating the ablation mechanism during UV nanosecond laser micromachining of Si at a lower irradiance range (0.5-4 GW/cm2). Good agreement between calculations and experimental verifications was found while 1-D equilibrium vaporization theory developed by (Anisimov 1968) was used with Hertz-Knudsen equation. It was also emphasised that the laser generated plasma plays a vital role in screening and re- distribution of laser energy during the ablation at medium high intensity regime (4-18 GW/cm2), and causing ablation rate to slow down and saturate with increasing intensity. Furthermore, previous experimental observations (Yoo, Jeong et al. 2000); (Yoo, Jeong et al. 2000); (Jacques and Gofstein 1991) suggested that the majority of mass ablation takes place a long time after the pulse; (Ren, Orlov et al. 2004) thought the hot plasma and plasma induced strong shock wave allowed uninterrupted heat exchange after the laser pulse and provided the increased ablation rates at high laser irradiance (20-150 GW/cm2). (Bulgakova, Bulgakov et al. 2004) focused on a new feature of the thermal model; the calculation of the solid-liquid interface. He took the experimental conditions of (Liu, Mao et al. 1999) (266 nm, 3 ns FWHM), which he modeled with the thermodynamic and optical properties taken from (Unamuno and Fogarassy 1989). He then explained that the contributions of the different channels of energy distribution depend strongly on many factors, for example, properties of the target material, and the loss due to heat conduction especially at low fluence. In Figure 2.41, the Si target behaves like metal (Nb in this case) under the considered conditions, it has high fraction of reflected energy and a small part of the useful inputs Eth + EL (thermal energy of the vaporised particles + energy less due to latent heat of vaporization). Plasma absorption here is the important channel for laser energy losses at nanosecond regimes of ablation before phase explosion occurs. _____________________________________________________________________ 71 Figure 2.41: Distribution of the laser energy density between different channels for laser ablation of two materials as a function of laser fluence. E0 is incident laser pulse energy, E/E0 is the proportion of the how much other energy is with respect to E0; Et are the losses to the target due to heat conduction; ER is the reflected energy; Eab is the energy absorbed in the plume; EL is the part corresponding to latent heat of vaporization; Eth is the thermal energy of the vaporised particles (Bulgakova, Bulgakov et al. 2004). In terms of the laser energy absorbed by the plasma, a few factors need to be determined, such as incident laser wavelength, electron density, plasma dimensions and plasma temperature. The latter three are strongly time dependent, and thus not much progress has been made in the estimation of the plasma absorption effects during excimer laser ablation (193, 248, 308 nm wavelengths) of materials. (Singh and Viatella 1994) developed the concept of the simulated plasma absorption coefficient αsimul, which is assumed to only depend on the evaporation thickness of the material that controls the concentration of species inside the plasma. They neglected the plasma thermal conduction effects, and the melting of the solid. (Franklin and Thareja 2004) calculated ablation threshold, ablation depth and ablation onset time for aluminium targets in ambient gas by numerically solving the one- dimensional heat conduction equation. They also included the temperature and phase dependence of the material properties, and marked differences in ablation characteristics with and without the temperature and phase dependence. For simulation of various parameters, a 1.06 µm wavelength laser of pulse width 8 ns (𝜏�) was used on aluminium _____________________________________________________________________ 72 target, the laser has a spot size of 200 µm with intensity range of 108-1010 W/cm2. The laser pulse is assumed to have temporal dependence of the form 𝑒𝑥𝑝[−(𝜏� − 𝑡)�/(𝑠𝜏�)�], various pulse shapes given by different values of 𝒔 are shown, Figure 2.42(a). The total energy in the pulse was kept fixed. The temperature distribution inside the target at the time of the evaporation for various pulse shapes were revealed, Figure 2.42(b). It suggested that the target surface was heated up quickly when laser pulse front rises sharply (s=0.1), resulting in smaller melt depth. On the other hand (s=1.0), the target surface was heated slowly, which resulted in further heat conduction inside the material and an increase in the melt depth (Franklin and Thareja 2004). The study suggested that the temperature rise of the material depends on the laser pulse shape and the conduction of heat within the target. (a) (b) Figure 2.42: (a) Various laser pulse shapes, (b) Temperature profiles in the target at the time of evaporation for the pulse shapes shown in (a) (Franklin and Thareja 2004) The laser generated plume expands into the surrounding ambient gas after the laser pulse ends. The temperature profiles in the ablated plume after the pulse ends (s=0.2) was shown, Figure 2.43. As the laser intensity increases the laser absorption in the plume decreases with the increase in vapour density, and the incoming laser pulse gets reflected _____________________________________________________________________ 73 (shielding). At 1010 W/cm2 a region of thickness 2.1 µm reflecting laser was seen inside the plume, Figure 2.43. Figure 2.43: Temperature profiles inside the ablated plume, at the end of laser pulse, at various intensities (Gaussian pulse s=0.2) (Franklin and Thareja 2004) The relative magnitude of absorption depth and thermal influence length (Mazhukin, Smurov et al. 1995) defines whether volumetric laser heating is important for laser-matter interaction. An analytic model based on volumetric heating has proven more accurate than a model based on surface heating (Li, Li et al. 2004). CO2 laser (λ=9.3 µm) drilling of polymeric dielectrics, which are used as multilayer electronic substrates, was investigated by (Zhang, Salama et al. 2006) with a one-dimensional transient analysis based on volumetric heating. The results showed that with a 20µs rectangular pulse and laser irradiance of above 1 kW/cm2, overheating existed inside the polymer. The temperature increased with time and laser irradiance, the high temperature inside the polymer caused explosive material removal and thermal stress above the yield stress. The depth, where maximum temperature occurred, increased first and then decreased due to reduced polymer thickness and reflection off the copper layer beneath the polymer. This depth also decreased with pulse width due to increased drilling speed and reduced polymer thickness caused by the increased laser irradiance. The vaporization mass rate was predicted larger with a spatially uniform beam rather than Gaussian beam. _____________________________________________________________________ 74 Nevertheless, the model used here did not consider temperature dependent thermo- physical properties and plasma formation. (Wu and Shin 2006) developed a thermal model for the nanosecond pulsed laser ablation, which included plasma generation and evaporation /recondensation. The simulated results were compared with the 248 nm Excimer laser (26 ns FWHM) ablation of nickel. Both the simulation and measurement results indicated that the transient transmissivity started to decrease at 15 ns after the beginning of the laser pulse. By the end of the laser pulse, only half of the laser energy could pass through the plasma. A lower surface temperature was obtained with plasma formation, which means a lower corresponding vapour saturation pressure. Therefore, a less strong evaporation process, less evaporation depth and less ablation depth were resulted with the plasma consideration. The model is valid when the phase explosion does not occur (no liquid ejection), which is below laser fluence of 5.0 J/cm2 for nickel (Xu 2002), and ablation depth was in the range of several tens of nanometers. 2.5 Summary At the commencement of this thesis it is clear that Si processing using a MOPA based Yb fibre laser has not been studied in any detail due to the recent arrival of this new technology. The literature review has been shown that IR wavelengths are not preferred in Si machining applications due to its low absorption coefficient, and large optical penetration depths. However, the optical and thermal properties of Si were found to change with temperature in a way that could significantly favour Si machining with a 1 µm laser source. Together with the fibre laser’s characteristics of being flexible, lost cost, maintenance free, and pulse tunable capability, the 1 µm laser processing of Si should be re-explored, and possibly redefined for future application potential. A range of laser parameters define the analysis options for characterizing Si laser machining. Laser pulse energy, pulse duration and pulse peak power are the key parameters besides wavelength. Different combination of laser parameters can result in _____________________________________________________________________ 75 various machining mechanisms from melt expulsion, surface vaporization to plasma formation or a mixture of them. Machining of Si in the vaporization dominated regime has advantages of smooth material edges, thinner recast layer, small extension of the oxidized surface and little HAZ. The machining processes with different types of lasers defined performance criteria. The excimer laser, copper vapor laser and 1 µm laser source showed machining defects on Si, such as micro cracks, heavy resolidified melt and recast layer. In comparison, the DPSS UV laser and ultrafast laser have demonstrated great machining quality on Si with little defects and little HAZ, but they are more expensive, require more delivery optics, are less stable and less flexible than the fibre laser. There is a significant research gap in that fibre laser characterization of Si machining has not been carried out previously. This study aims to fill this gap, whilst comparing the performance with that of the state of the art DPSS UV laser. Thermal models have been used extensively to provide a theoretical understanding, and predict key values like temperature distribution during laser material interaction. In this study, the fibre laser’s pulse shape effects on material’s response, temperature distribution and machining mechanism will also be analyzed theoretically with a thermal model, in order to provide insight into the development of the Yb laser machining characteristics. _____________________________________________________________________ 76 Chapter 3 Experimental systems, measurement tools and methodology 3.1 Laser systems The fibre laser and the Q-switched DPSS laser used in the study are described in the following section along with their pulse characteristics. 3.1.1 Fibre laser The fibre laser (Model G3 HS, SPI Lasers, UK) used in this work was a 20 W pulsed Yb doped laser based on Master Oscillator Power Amplifier (MOPA) architecture. A directly modulated seed laser (single emitter telecom based laser diode) pumped a fibre based Yb amplifier, Figure 3.1. This seed laser approach is enabled by a high gain and high saturation power characteristics of the active Yb fibre. Ytterbium is one of the most effective ions in a silica-based host; it offers an unusually broad absorption band that stretches from below 850 nm to above 1070 nm, a key feature for high efficiencies (Lou, Zhou et al. 2003). The system provides the following specifications: peak emission wavelength 1064 +/−10 nm; CW-500 kHz pulse repetition rate; maximum pulse energy 0.8 mJ; maximum peak power 14 kW; collimated output beam diameter 3.1 mm; and an integrated Faraday isolator. The beam intensity spatial profile is near Gaussian with an M2 of <2, and is stable across its operational frequency range at constant power. Figure 3.1: Internal structure of Yb doped fibre laser _____________________________________________________________________ 77 The output pulse-to-pulse energy stability is typically ≤5% with a beam divergence (full angle) of < 3 mrad (SPI Lasers 2008). The PulseTuneTM capability presets 25 output waveforms (WFMs) for optimised peak pulse power at specified pulse repetition frequencies (PRFs), the WFMs are shown in Figure 3.2(a) for a number of settings. Waveforms at high PRFs (>125 kHz) are near Gaussian in shape, but waveforms at low PRFs (<125 kHz) are distinctly non Gaussian with a gain switched leading edge followed by long lower intensity trailing edge. Pulse peak power and pulse duration vary with changing waveforms. At higher frequencies the MOPA system is able to modify the pulse envelope in order to produce optimised peak power output with short pulse duration. For example in Figure 3.2(a) and (b), the SPI fibre laser has peak power of 6.5 kW at 65 kHz compared to the Q-switched fibre laser with less than 2 kW at 60 kHz. (a) (b) Figure 3.2 (a): Laser pulse characteristics at varying pulse repetition rates of the SPI fibre laser (SPI Lasers 2008). (b): Pulse shapes of an IPG Q-switched fibre laser. The preset WFMs offer much greater control of pulse duration and peak power than the traditional Q-switched systems, Figure 3.2(b), which have flattened pulse profiles at higher pulse repetition rates. The Q-switched systems have long fixed pulse duration of 350 ns compared with 20-200 ns for the MOPA based systems. The SPI fibre laser provides high power densities at the focal point with a variable beam expander and a flat _____________________________________________________________________ 78 field focal lens. Table 3.1 summarises the incident beam parameters at focus for a variety of WFM outputs when using a 125 mm focal length flat field lens. Table 3.1: Incident power density for a range of WFM settings of variable input bean diameters with a 125 mm focal length objective lens. WaveForm Number (Peak power density MW/cm2 ) Input beam diameter (mm) Focal spot diameter (µm) 0 1 2 3 3.1 109 149.1 74.6 53.3 42.6 6.2 55 596.4 298.2 213.0 170.4 9.3 36 1341.9 671.0 479.3 383.4 With an average power of 20 W, a laser peak power density of more than 1.3 GW/cm2 is available with a spot size of 36.45 µm, at a pulse energy of 0.8 mJ, and a pulse repetition rate of 25 kHz. The scanner aperture is the factor which limits the input beam diameter to less than 10 mm. 3.1.2 Q-switched UV laser The UV laser (355nm wavelength) used for performance comparison was a Coherent AVIA 355-10 (Santa Clara, CA, USA). This solid state Q-switched UV laser is commonly used to process Si in the semiconductor industry because of its high absorption at 355 nm wavelength. The system produce different levels of average power at varying pulse repetition rates, Figure 3.3, where the maximum 10W average power is optimized between 50 and 60 kHz. _____________________________________________________________________ 79 Figure 3.3: Average power out of laser head (Coherent 2007). The fibre laser and UV laser system specifications are summarised in Table 3.2 for comparison. The UV laser has a pulse width (FWHM) of less than 35 ns and pulse duration of less than 60 ns for frequencies up to 60 kHz, compared with a range of 20- 200 ns pulse duration at 25-500 kHz for the fibre laser. The output pulse-to-pulse energy stability (RMS) is <5% up to 80 kHz, with a beam divergence (full angle) of < 0.3 mrad between 50 and 70 kHz. The beam circularity is greater than 85%, and the typical warm up time is less 15 minutes from standby mode and less than 40 minutes from cold start, while the fibre laser offers instant turn on capability. Table 3.2: System comparison of Yb fibre laser and DPSS UV laser. Laser parameters SPI fibre laser Coherent UV laser (Coherent 2007) Wavelength λ 1064+10 nm 354.7 nm Max. Ave. Power Pmax 20 W 10 W at 60 kHz Max. pulse energy Ep 0.8 mJ @ 25 kHz 0.167 mJ @ 60 kHz Pulse Rep. Frequency 1-500 kHz 1-300 kHz Peak power 14 kW 6.6 kW (measured) Waveforms 25 preset Pseudo Gaussian Beam diameter 3.1 mm 3.5 ± 10% mm Full pulse duration 20-200 ns <35 ns (FWHM) up to 60 kHz M2 <2 (Pseudo Gaussian Distribution) 1.3 Polarization Random Horizontal _____________________________________________________________________ 80 3.2 Scanning systems The laser beam was delivered to the workpiece through a Nutfield Technology Inc. XLR8 Galvo X-Y scan head, Figure 3.4. The scanner provides the synchronised actions of two galvanometer servo-controlled turning mirrors, which direct the laser beam to specific locations on a target material surface in both the X and Y directions. A flat field lens of focal length 125 mm was used which gave a maximum field size of 90×90 mm for the 1064nm wavelength. Scan patterns are generated within the Waverunner software (Nutfield Inc.) which also has full control of laser parameters. The minimum spot size at the focal plane was calculated to be 36 µm with the beam input diameter of 9.3 mm (×3 beam expansion). Figure 3.4: Images of the scan head and schematic data flow (Nutfield 2011). Figure 3.5 are pictures of the experimental set up (overall and zoomed-in view), the setup consists of a laptop, a control pipeline unit, and an enclosure with fume extraction. The control pipeline unit takes commands from the software and sends them out to both laser and scanner. The laser output isolator, scan head, sample holder, and a z-focus device are integrated within the enclosure. The z-stage is tailored for different sized samples. The fibre laser unit was placed on the shelf below the enclosure, the beam was delivered through the fibre to the isolator and scan head. The whole system is Class 1 laser safety standard in operation. A FumeCube (PUREX, Rotherham, UK) extraction was used to protect both the user and optics. _____________________________________________________________________ 81 Figure 3.5: The experimental set up includes a laptop, a control pipeline unit, an auto-Z- axis enclosure, a sample holder, a fibre laser, a scanner and a vacuum pipe. The fibre laser system driven by the scan head can be classified with two operational modes; drilling mode and scanning mode. Drilling mode refers to percussion drilling, in which there was a delay of one hundred microseconds approximately for the laser to reach its maximum power. Drilling becomes less effective in this mode. The drilling mode is used for drilling deep features, as it does not leave traces of acceleration and deceleration. Scanning mode refers to the situation where a line is scanned at high speed so adjacent spots are separated. In the scanning mode, the system accelerates to the preset scanning speed and reaches preset power after about one hundred microseconds, Figure 3.6. Measurements were taken at latter part of the line at the full power and full speed for single pulse interaction. Holes scanned at high PRF (>135 kHz) cannot be separated due to the limited spot size (37 µm) and scanning speed (5 m/s). Holes drilled with a varying number of pulses were achieved by scanning the same line the same number of times. _____________________________________________________________________ 82 The time between two consecutive pulses was not determined by the PRF value, instead it is the delay between two consecutive scans. Figure 3.6: Fibre laser line scanning on Si with 25 kHz repetition rate, 1.8 m/s and 10 overscans. 3.3 Machining systems The UV laser was integrated with both a four axis machining system and a scanning system. For the machining system, the laser beam was delivered by mirrors and a focusing lens (focal length = 40 mm) via an Aerotech precision table (ES13636). The theoretical minimum spot size was calculated to be 6.7 µm, which delivers more power density to the workpiece than the scanning system. Figure 3.7 illustrates the full integrated system, which was enclosed with interlocks to a Class 1 laser safety standard in operation. _____________________________________________________________________ 83 Figure 3.7: The machining system set up for the UV laser. Two reflective mirrors and the focusing lens were mounted on an optical breadboard which was fixed to the z-axis linear stage of the Aerotech precision system. The z-axis stage was secured to a granite bridge, mounted directly onto the xy stage at the base. All components are fixed to an optical table. The breadboard with focusing lens was free to move in z-direction, which was useful to focus on various sample z-positions. The sample was placed on the Aerotech table and was driven by two linear stages moving in the x and y directions. The linear air bearing stage had a movement resolution of 20 nm, and maximum speed of 300 mm/s. The fourth rotary (U) axis was available for machining cylindrical parts. Both the linear stages and the UV laser were connected to a NPAQ drive rack and driven by Automation 3200 (Aerotech) controller. G-code was used to program the axis movements and M codes used to control the state of the laser. _____________________________________________________________________ 84 Figure 3.8: The scanning system set up for the UV laser (left) and the control unit (right). For the scanning system, the scan head was mounted on the same optical breadboard for focusing, Figure 3.8. The mirrors were used to deliver laser beam to the scan head, and a UV scan head and focal lens (focal length=103.3 mm) were used for 355 nm wavelength. The theoretical minimum spot size was calculated to be 17 µm. The manually configured control unit in Figure 3.8 was used to control the scan head and laser through WaveRunner software (Nutfield Inc). The DPSS UV laser does not provide pulse shape tunability, and the controlled parameters are fewer compared with the fibre laser system, they were laser on /off, switching frequency, scanning direction and speed. The scanner can perform single pulse ablation experiments by scanning at a few thousands mm/s. _____________________________________________________________________ 85 3.4 Laser beam diagnostics Laser pulses were diagnosed in terms of spatial, temporal distribution and average power for full operational machining experiments. 3.4.1 Spatial beam profile measurements A LBA-FW-SCOR20 digital camera (Ophir-Spiricon Inc. USA) was used to examine the spatial power distribution at different power levels and at different repetition rates. Tools were arranged according to the schematics in Figure 3.9. A black anodised spinning plate which reduces specular reflection was used to image the beams. Two attenuators were used to reduce average power for the protection of the black screen and the camera sensor (Damage threshold: 50 W/cm2 or 1 J/cm2 with all filters installed for <100 ns pulse width). Figure 3.9: Schematics of beam shape monitoring arrangement. The aperture of camera was adjusted to a level just below sensor saturation, maximizing dynamic range. At 20 W, the laser beam spatial profiles are near Gaussian and independent of repetition rates, Figure 3.10. _____________________________________________________________________ 86 Figure 3.10: Spatial intensity distribution of the raw laser beam, 25 kHz to 500 kHz pulse repetition rate, 20 W output power. 3.4.2 Temporal pulse profile measurements The aim of the measurements was to verify temporal pulse profile characteristics at full operational conditions. The tools used included a Si Biased Detector (DET10A/M, 200- 1100 nm, 1 ns Rise Time, THORLABS), a digital oscilloscope (Lecroy WaveRunner 6050A, 500MHz analogue bandwidth, 5 GS/s single-shot sample rate). The photodiode was placed a few centimetres from the laser material interaction zone to collect reflected and scattered light, and the laser beam was defocused to avoid ablation. Stainless steel (SS) was used due to its high reflectivity at 1 µm wavelength (>60%) (Karlsson and Ribbing 1982). During the measurement process the SS was not melted, which limited the variations of the amount of the reflected or scattered light with temperature. The oscilloscope was connected to the photodiode with BNC cable and the signal averaged over a few tens of seconds. The measured data was then integrated with pulse energy to calculate the peak power, Figure 3.11. _____________________________________________________________________ 87 Figure 3.11: Fibre laser pulse characteristics at varying pulse repetition rates measured after scanner lens (left), stated in the system manual (right). The maximum peak power striking substrate material was 12.77 kW after the beam expander and scanner, it is lower than the specified value (14 kW) in the system manual. The peak power and pulse shape of the UV laser are not provided, but they were measured using the same photodiode and the measured pulse shape data plotted, Figure 3.12. Unlike the fibre laser pulse shapes, the UV pulses do not have a fast leading edge and long tail, they are pseudo-Gaussian shaped pulses with a 60 ns full time and less than 30 ns FWHM. _____________________________________________________________________ 88 Figure 3.12: Temporal pulse profile of the DPSS UV laser at 60 kHz up to 10 W average power. 3.4.3 Average power measurements The average power was measured with a low power thermal sensor (30A-BB-SH-18, replacing 30A-SH-V1, OPHIR), the sensor was placed after the focusing lens at a defocused position to measure the average power available to the material. The power sensor was connected to a laptop via a Smart Head to USB Interface (OPHIR), and the measured values were recorded by the laptop. A plot was generated with actual power against specified system power, Figure 3.13. The power readings were taken at 25 kHz, WFM 0, with ×2, ×3, ×3.3 and ×4 beam expansion, which corresponds to 6.2 mm, 9.3 mm, 10.2 mm and 12.4 mm beam diameters. _____________________________________________________________________ 89 Figure 3.13: Average power available at the material substrate at 25 kHz, WFM 0, with ×2, ×3, ×3.3 and ×4 beam expansion. The highest actual power incident on the substrate was less than 18 W after the lens. The power curve with ×4 beam expansion was lower than the other three curves, which suggests that the beam was larger than the optics in the scan head and should not be used due to undesirable effects at the focus. Average output power was also measured when waveform variations are used at increasing frequencies. Figure 3.14 shows that the 18 W maximum average power can be maintained at frequencies higher than PRF0 (a specified switching frequency with optimised waveform for highest possible peak power). _____________________________________________________________________ 90 Figure 3.14: Typical average power variation with PRFs for the selected waveform. The average output power of the UV laser after the focusing optics was also measured using the same power meter, and the results are plotted against the corresponding frequencies, Figure 3.15. The maximum average power on the substrate was 8.81 W at 50 kHz, 10.11 W without the lens. An initial overshoot was observed at each measurement from 30 kHz to 100 kHz, Figure 3.16. These overshoot values were also plotted as red data points in Figure 3.15. _____________________________________________________________________ 91 Figure 3.15: Measured average power as a function of frequencies of the UV laser after optics. Figure 3.16: Measured average power as a function of time of the UV laser at 60 kHz after optics. _____________________________________________________________________ 92 3.5 Sample management and focus test The materials used were polished single crystalline silicon wafer of 650 µm thickness with a <111> crystallographic orientation. The polished single crystalline silicon wafer was cut into small pieces (15×15 mm) by scribing with a diamond pen and cleaving. The samples were placed on top of an aluminum sample stage for machining. The sample stage was polished every 3 months to maintain flatness for maximum machining accuracy. The laser beam was focused on the top surface of each sample with a motorized Z-axis (10 µm resolution). A set of test runs were performed to find the best focus on the sample surface. A series of high fluence lines were scribed with laser at stand-offs from 163 mm to 164 mm for Si at 0.1 mm steps. The focal distance is measured downwards from the focusing lens. Once the best location was selected the experiment was repeated at 0.02 mm increments for fine focal position tuning. The scribed lines were measured with a white light interferometer (details in section 3.6.1) and results plotted, Figure 3.17. For this trial, the focal distance was calculated to be 163.30 mm by identifying the highest etch depth. This procedure was done every time the laser was switched on to ensure consistent results. All the samples were kept in plastic sample bags before and after processing, and they were handled with powder free lab gloves to avoid contamination. _____________________________________________________________________ 93 Figure 3.17: Etch depth of machined lines in the focus test of the fibre laser scanning system. 3.6 Sample Analysis The processed samples were examined using a range of imaging and measurement tools including white light interferometer and scanning electron microscope. Samples were imaged before and after cleaning in an ultrasonic bath (FS7652, Fisher Scientific) in ethanol and deionized water, they were then blown dry with oxygen free N2. Chemical etching was not used and all process interactions were carried out in ambient air at room temperature. 3.6.1 White light interferometer Wyko NT3300 system (Veeco Metrology Group, Arizona, USA) used in this study is shown in Figure 3.18(a), it consists of several key components which work together to provide surface information on the sample. The components are an interference microscope mounted on a motorized z-axis, a set of magnification objective lenses, a _____________________________________________________________________ 94 vibration-isolation table with a motorized x/y sample stage on the top, a computer with Wyko® Vision32TM software installed, and an operator station (Veeco 1999). (a) (b) Figure 3.18(a): Wyko NT3300 system. (b): Interference fringes at best focus (Veeco 1999). The basic working principle is that light from the illuminator passes through the Integrated Optics Assembly and is reflected down to the objective by a beamsplitter. Another beamsplitter separates the light into two beams, one beam is reflected from a super smooth reference mirror, and the other beam is reflected from a sample surface. Both beams reflect through the microscope objective to the detector array and form interference fringes when in focus, Figure 3.18(b). 3.6.1.1 Analysis Options The system has two technologies to measure both smooth and rough surfaces, Table 3.3. _____________________________________________________________________ 95 Table 3.3: Technology modes of the Wyko NT3300 system. Technologies Applications Vertical resolution Measurement range Phase-shifting interferometry (PSI) Smooth surfaces and small steps, such as mirrors, highly polished samples. 0.3 nm <160 nm Vertical scanning interferometry (VSI) Rough surfaces. 3 nm <2 mm VSI was used because machined dimensions are in the microscale regime. As an example, a machined Si sample was scanned with VSI, and a 2-D surface plot of four drilled holes generated, Figure 3.19(a). The colour bar on the right of the main image gives relief information. A sub-region was then selected, Figure 3.19(b) and the hole depth is shown to be 5.66 µm, the lowest value on the scale bar. Eight holes were measured for each laser parameter setting, and the average was taken to represent the hole depth at that setting. The black parts in the image were invalid pixels (system limitation), probably as a result of low intensity surface reflections from sharp peaks or valleys. (a) (b) Figure 3.19(a): Example of Wyko measurement of single pulse ablation on Si with SPI’s fibre laser at 25 kHz and 11 W. (b): Sub-region of Figure 3.19(a) (top right hole). _____________________________________________________________________ 96 3.6.2 Scanning electron microscopy (SEM) SEM works by detecting secondary electrons which are emitted from the surface due to excitation by the primary electron beam. Compared to an optical microscope, SEM provides improved resolution capacities, high depth of focus, and simplified interpretation of the images due to the 3D impression. The advantages allow visualisation of the machined hole shape, measurements of holes diameter, debris size, and other surface defects. 3.6.2.1 Operational practice Fine surface images can normally be obtained with low accelerating voltages. At high accelerating voltages the electron beam penetration and diffusion area become larger resulting in unnecessary signals, like backscattered electrons being generated from within the specimen. Image contrast is also reduced, fine surface structures are veiled. A charge-up problem happens in nonconductive specimens when the sample is directly illuminated by the electron beam. The electrons collect locally and prevent normal emission of secondary electrons. As a result, the sample experiences unusual phenomena like abnormal contrast, image deformation or shift. There are several ways to diminish this problem, lower the accelerating voltage and tilting the specimen to find a balanced point between the amount of incident electrons and the amount of backscattered electrons that leave the specimen. Another way is to coat the nonconductive sample with uniform conductive metal layer. For some parts of specimen which are difficult to coat, the conductive paint can be applied when fixing sample on the stub. Specimen contamination is caused by a long radiation exposure of the electron probe on the sample surface, and shows reduced image contrast and fidelity. This problem is alleviated by focusing near the point of interest, and then returning back at a later time. _____________________________________________________________________ 97 3.6.3 High speed imaging The Specialized Imaging (SIM 16, Herts, UK) camera was used to capture the vapour expansion of Si under fibre laser illumination. The camera takes 16 images at a time, and it works at up to 200 MHz. Each image can be individually set to a different exposure time form 5 ns to 10 ms in 5 ns steps, and the interframe time between each image can be independently set between 0 ns and 20 ms in 5 ns steps. External triggering is available from the rising or falling edge of a 2 to 50V signal. The camera is connected via Ethernet to a PC via dedicated software running under Win XP. 3.6.3.1 Triggering A photodiode was used as the external triggering source. It was placed next to the interaction zone and when fibre laser started pulsing, the scattered and reflected light was detected, amplified and sent to the camera. A delay time was set within the camera software so that the vapour expansion caused by the next laser pulse inline could be imaged. Due to the power ramp up nature of the fibre laser, the detected signal was not from the first pulse, but the fourth. 3.6.3.2 Set up The complete system is shown in Figure 3.20. The light source was a continuous Xenon light (HBO 100 Mciroscope Illuminating System, W/2, Carl Zeiss, Germany). The fibre laser beam was delivered from the scan head situated above the sample stage, percussion drilling was performed on Si. All the parts except the camera and the laser were on the optical bench for ease of alignment. _____________________________________________________________________ 98 Figure 3.20: High speed imaging setup with a light source, an iris to clean the illumination beam, a collimating lens, a sample stage with photodiode next to it, a lens and a telescope to magnify and form clear image on the CCD sensor. 3.6.4 Shadowgraph imaging The schematic setup of the shadowgraph imaging system is shown in Figure 3.21. The system included two lasers: a fibre laser for ablation and a Dual Cavity Nd:YAG laser (532 nm wavelength with 4 ns FWHM, Oxford lasers, Oxon, UK) for imaging. The pulse timing between the two lasers was controlled via a digital delay generator (Stanford DG535). The illumination laser was unstable at low pulse energy, so the pulse energy was set high intentionally and an optic attenuator was used to reduce the pulse energy to the usable level. A beam splitter was used to separate the illumination beam and feed one part to the photodiode. The other photodiode was placed next to interaction zone to collect the ablation laser signal. Signals from both photodiodes and the control signal _____________________________________________________________________ 99 from the ablation laser were monitored by the oscilloscope to ensure the correct timing. A green bandpass filter was placed in front of the Nikon D70 camera to remove other wavelengths generated duration ablation. A variable magnification telescope was used to increase the size of the interaction zone to fit into the CCD sensor. Figure 3.21: Schematic setup of the shadowgraph imaging system. During the imaging process, the room which the experiments took place was kept dark, and the camera shutter was left open for 1.6 second. The exposure time of the image was dependent on the pulse width of the illumination laser. Single pulse ablation was performed in air at atmospheric pressure, the fast scanning speed separated two neighbouring spots at the operating frequency. Each shadowgraph image was captured on ablation of a fresh piece of Si sample. The trigger signal of the G3 fibre laser set off the ablation laser and the illumination laser. The timing was set according to the trigger signal, ablation laser pulse signal, and the delay between the trigger signal and the illumination laser pulse signal. The illumination laser was pointing at the sixth pulse of ablation laser on Si along the scanned line. The pulse width of the imaging laser (Nd:YAG) shortens from more than 100 ns to less than 10 ns FWHM with increasing pulse energy. The particle speed leaving the interaction zone was estimated based on the 140 ns imaging laser pulse width, Figure _____________________________________________________________________ 100 3.22. The Nd:YAG laser pulse signal is not stable by external triggering, it fluctuates by a few hundred nanosecond. Figure 3.22: Measurement of the Nd:YAG laser pulse width. _____________________________________________________________________ 101 Chapter 4 Fibre laser machining characterization of silicon 4.1 Introduction This chapter aims to answer Research Question 1. Research Question 1: Which fiber laser parameters have the greatest influence on Si machining and what are their optimum settings? The laser matter interactions with a pulse tuneable 1064 nm wavelength Yb fibre laser on Si for micro-via applications were characterized with various laser parameters. The laser parameters were number of pulses, pulse energy, pulse duration, pulse peak power and pulse repetition frequency (PRF). These parameters affect the machining mechanism, quality, debris formed, machining depth and energy efficiency. Section 4.2 explores effective laser parameters by examining them collectively through percussion drilling. Section 4.3 investigates each laser parameter independently through single pulse interaction experiments. Of particular interest is the laser temporal pulse profile effect on drilling performance of Si. The drilling depth is measured for each parameter and processed quality is compared for a wide range of interaction conditions. Each section has discussion merged with results. The answers to the research questions are summarised in section 4.4. 4.2 Percussion drilling Observations of percussion drilled holes on Si are described in this section with various fibre laser parameters; pulse repetition frequency (PRF), peak power, pulse shape and cooling cycle. _____________________________________________________________________ 102 4.2.1 PRF analysis The fibre laser’s output waveforms (WFM) for a number of settings are shown in Figure 4.1. At low PRF, especially at 25 kHz (200 ns pulse duration), the asymmetrical pulse has a long tail which contains more than half of the pulse energy. Figure 4.1: Laser pulse characteristics at varying pulse repetition rates. Holes were percussion drilled at various PRFs. The equivalent total energy and drilling time were applied to each hole, but the pulse energy, pulse duration, peak power and inter pulse period were different. Table 4.1 summarises the laser parameters used when a total energy of 3.54 mJ was delivered to each hole at 17.7 W average power after the optics. _____________________________________________________________________ 103 Table 4.1: Laser parameters for different PRFs when total energy of 3.54 mJ was kept constant, the total process time was 0.2 ms for each hole. Repetition Rate (kHz) Pulse Energy @17.7W (mJ) Pulse Period (µs) Full pulse duration (ns) Peak Power (kW) Pulse number to achieve 3.54 mJ 25 0.708 40 200 12.77 5 65 0.272 15.4 75 7.03 13 125 0.142 8 50 6.41 25 250 0.071 4 30 3.85 50 500 0.035 2 21 2.88 100 The waveform produced with a 25 kHz PRF has the longest inter pulse period (40 µs) and the longest pulse duration (200 ns). It also delivers the highest pulse energy (0.708 mJ) and the highest peak power (12.77 kW). The SEM images of the percussion drilled holes with laser parameters listed in Table 4.1 are shown in Figure 4.2 before and after ultrasonic cleaning. _____________________________________________________________________ 104 a(1) 25kHz before cleaning a(2) 25kHz after cleaning b(1) 65kHz before cleaning b(2) 65kHz after cleaning c(1) 125kHz before cleaning c(2) 125kHz after cleaning d(1) 250kHz before cleaning d(2) 250kHz after cleaning e(1) 500kHz before cleaning e(2) 500kHz after cleaning Figure 4.2: percussion drilled holes produced using beam parameters as detailed in Table 4.1. The SEM images in Figure 4.2 show variations in hole depth, hole diameter and finishing quality of the percussion drilled holes at various PRFs. Holes drilled at low PRFs (≤ 125 kHz) produced wide and shallow holes, with excellent quality (very little melt resolidification and recast layer). The remaining debris consists of sub-micron/ nano- _____________________________________________________________________ 105 sized particles, which washed off ultrasonically. At high PRFs (> 125 kHz), the drilled holes were narrow, deep, with resolidified melt deposited near the rim, Figure 4.2d(1) and e(1). The debris consists of micro-sized particles and tens of microns of resolidification. The ultrasonic wash damaged the hole rim and compromised the sidewall quality. The machined hole depth was measured with a white light interferometer, the total depth and depth per pulse plotted in Figure 4.3 for PRFs from 25 kHz to 500 kHz. The process time (0.2 ms) and total energy (3.54 mJ) per hole were kept constant. Figure 4.3: Ave. depth per pulse and total depth vs. frequencies in the drilling mode. The depth per pulse curve (dark blue) in Figure 4.3 suggests a transition at around 100 kHz. The value was above 1 µm at PRFs below 100 kHz, and the debris size suggests the vaporization dominated removal mechanism. The value was below 1 µm at PRFs above 100 kHz, and the debris size suggests the melt expulsion removal mechanism. _____________________________________________________________________ 106 The total drilled depth increases with PRFs despite the decreasing pulse energy. The inter pulse period becomes much shorter, 2 µs at 500 kHz compared to 40 µs at 25 kHz. The residual heat is likely to accumulate and contribute to the subsequent laser pulse heating at high PRFs (Tan and Venkatakrishnan 2006); (Tan and Venkatakrishnan 2007). 4.2.2 Peak power analysis In percussion drilling, promising results were found at low PRFs (< 125 kHz) where the pulse energy is high (> 0.14 mJ) and cooling time is long (> 8 µs). Percussion drilling with 10 pulses at 25 kHz with a 200 ns pulse duration were performed with varying peak power densities, depth per pulse was plotted in Figure 4.4 with inset SEM images. Figure 4.4: Depth per pulse vs. peak power density in the drilling mode. The machining quality has been shown to improve (sub-micron sized debris, no raised rim, sharp sidewall) when the process is governed by vaporization, and peak power densities in excess of 160 MW/cm2. The high peak power and fast rising edge produced _____________________________________________________________________ 107 results that have not been achieved using the 1 µm laser source before. The detailed machining mechanism and true single pulse interactions are included in section 4.3. Figure 4.5 – Figure 4.7 are SEM images of percussion drilled holes before/after ultrasonic cleaning and the cross sections of them. The laser parameters used are 25 kHz PRF, 10 pulses and 200 ns pulse duration with varying peak power density (98, 270 and 1,200 MW/cm2). These cross-sections were made with a diamond cutting pen and some edges were broken due to the brittleness of silicon. Figure 4.5: SEM images of percussion drilled holes with peak power density of 98 MW/cm2, (a) before ultrasonic cleaning, (b) after ultrasonic cleaning, (c) cross section. (a) (b) (c) _____________________________________________________________________ 108 Figure 4.6: SEM images of percussion drilled holes with peak power density of 270 MW/cm2, (a) before ultrasonic cleaning, (b) after ultrasonic cleaning, (c) cross section. Figure 4.7: SEM images of percussion drilled holes with peak power density of 1,200 MW/cm2, (a) before ultrasonic cleaning, (b) after ultrasonic cleaning, (c) cross section. 20µm I I (a) (b) (c) (c) (a) (b) _____________________________________________________________________ 109 The heavy melt re-solidification and smooth surface morphology shown in Figure 4.5 suggests that the melt ejection is the dominant mass removal mechanism. From Figure 4.5 – Figure 4.7, there is a significant reduction in the amount of the melt ejection. Although the sub-micron sized debris on the substrate surface suggest vaporization as the dominant mass removal mechanism, Figure 4.6 and Figure 4.7, there were traces of melt flows on the surface of the cavity walls. Figure 4.5(b) shows some signs of recast layer and damage at the hole rim due to strong bonding between re-solidified melt ejection and surface substrate. Figure 4.6 and Figure 4.7 show very little recast layer at the hole rim, sharp sidewalls, no cracks and no HAZ into the bulk material. By using laser pulses at 25 kHz PRF, 200 ns pulse duration and 1,200 MW/cm2 peak power density, blind micro vias were percussion drilled on single crystalline silicon, Figure 4.8. With a depth of 46.4 µm and standard deviation of 0.27 µm, the laser demonstrated great repeatability and consistency. (a) (b) Figure 4.8: SEM images of percussion drilled blind holes by 1064 nm Yb doped fibre laser with 10 pulses, 25 kHz PRF, 200 ns pulse duration and 20 W average power, (a) Secondary electron detection, (b) Inlens detection to look at the hole bottom. _____________________________________________________________________ 110 4.2.3 Pulse shape analysis Holes were percussion drilled at various pulse shapes. The equivalent PRF, pulse energy, total energy, drilling time and number of pulses were applied to each hole, but the pulse duration and peak power were different, Figure 4.9. Figure 4.9: Pulse profiles of various pulse shapes at 125 kHz with 50, 75, 200 ns pulse. Figure 4.10-Figure 4.12 illustrate the differences for percussion drilled holes with pulses of 50 ns, 75 ns and 200 ns in duration, 3.54 mJ total energy, 25 pulses, 125 kHz PRF. Figure 4.10: SEM image of percussion drilled holes with 50 ns pulse (a) before, (b) after ultrasonic cleaning, and (c) the cross section. (a) (b) (c) _____________________________________________________________________ 111 Figure 4.11: SEM image of percussion drilled holes with 75 ns pulse (a) before, (b) after ultrasonic cleaning, and (c) the cross section. Figure 4.12: SEM image of percussion drilled holes with 200 ns pulse (a) before, (b) after ultrasonic cleaning, and (c) the cross section. The total depth drilled with a 200 ns pulse is about four times deeper than that machined with 50 ns pulse. The lower peak power at 200 ns results in incomplete evaporation leaving expelled molten droplets on the surface. The average diameter of holes drilled with 200 ns pulse is smaller than those machined with 50 and 75 ns. The hole quality in Figure 4.12 was compromised with reduced edge finish, and micro-cracks were found at the side wall of the hole drilled with 200ns pulse. So for a given pulse energy above the ablation threshold, short pulse width i.e. high peak power would drill holes with vaporization dominated ablation mechanism. There would be limited resolidified droplets for maximum hole quality. A long pulse width on the other hand would produce narrower, deeper holes due to enhanced thermal diffusion, fluid formation and fluid ejection. The number of pulses required to drill through Si of fixed thickness would be less for longer pulses, thus delivering faster drilling rate. (a) (b) (c) (c) (b) (a) (a) (b) (c) _____________________________________________________________________ 112 4.2.4 Cooling cycle analysis The cooling cycle effect on percussion drilling was investigated while the pulse duration, pulse peak power, pulse energy and total energy are maintained, Figure 4.13. The hole depth was plotted, Figure 4.14, with inset SEM images. The number of pulses (15) was selected to maximise the inter pulse period effect, but not generate much recast. Single pulse ablation was also carried out and plotted in the same figure, Figure 4.14. Figure 4.13: Similar pulse profiles (200 ns pulse duration, ~870 W peak power, 83.4 MW/cm2 peak power density) at 25, 50, 75, 100 and 125 kHz for cooling cycle analysis. _____________________________________________________________________ 113 Figure 4.14: Single pulse ablation and percussion drilling on Si with similar pulse profiles of different PRFs. The depth of single pulse ablation shows slight increase from 3.9 to 4.5 µm due to the minor variations in pulse profile, Figure 4.13, and the increased PRF. Despite the maximum feasible scanning speed (4500 mm/s) used at 125 kHz, the distance between adjacent spots were getting closer with increasing PRF. The SEM image at 125 kHz shows ~12% overlap, the next pulse was deposited on a preheated and disturbed surface instead of an original polished substrate surface. The percussion drilled holes had increased depth (37.6 to 45.6 µm) and decreased diameter (32.8 to 19.3 µm) with the decreasing cooling cycle (from 40 µs to 8 µs). The holes at with 8 µs cooling cycle was badly deformed, the hole rim was no longer circular with a narrow neck, similar findings were reported with a Q-switched DPSS UV laser percussion drilling on Si (Tan and Venkatakrishnan 2006). _____________________________________________________________________ 114 The increased depth was due to the reduced cooling period, the increased duty cycle (pulse duration divided by pulse period) enhances temperature rise and shortens the time to the melting temperature in the consecutive pulses (Shuja, Yilbas et al. 2007). At high PRFs more material was melted and vaporised, the molten material was re-solidified and formed a recast layer on the rim and sidewall. A higher rate of stress development in the workpiece was also a result of the increased PRFs (Shuja, Arif et al. 2002). 4.2.5 Discussion 4.2.5.1 Thermal coupling The surface temperature of Si rises as a result of photonic energy absorption during a laser pulse. The temperature reaches T at the end of a pulse, the relationship defined by one-dimensional heat conduction (Tan and Venkatakrishnan 2006), Equation 4.1. T − T� = 0.783 �� �√������� (4.1) F is the laser fluence, κ is the thermal conductivity, ρ is the density of solid, and C� is the heat capacity of solid. τ normally represents the FWHM of the Gaussian pulse shape, in this study it is defined as the full duration of the asymmetrical pulses of the fibre laser. T� is the initial temperature of the material, room temperature. After an individual laser pulse, the surface temperature T reduces slowly until it reaches room temperature T0. The one-dimensional homogeneous thermal diffusion equation can be used to determine the time (ζ) span of the thermal diffusion, which is assumed to be the dominant mechanism for the heat loss. For the multi-pulse percussion drilling, the initial temperature (T0) will be higher when the next pulse arrives as there has not been sufficient time for the heat to completely dissipate (t < ζ). After n laser pulses for a pulse separation of t, the initial temperature at a time just before the next pulse can be expressed in Equation 4.2 (Ball, Feurer et al. 1996). T�(n) = T� + nδT (4.2) _____________________________________________________________________ 115 δT is the temperature rise due to undissipated heat at the end of a time delay t. The actual surface temperature T(n) after n successive pulses can be written as in Equation 4.3. T(n) = T�(n) + T (4.3) High repetition rates will cause overheating through the increased surface temperature of the ablation front. For percussion drilling, overheating deteriorates the geometry of the hole, resulting in a barrel shaped hole (Tan and Venkatakrishnan 2006). Increasing the cooling time between consecutive pulses lowers the degree of heat coupling. When the cooling time is long (t ≥ ζ), the machining quality of a multi-pulse percussion drilled hole will be as good as that of a single pulse machined hole. 4.3 Single pulse interactions MOPA lasers offer much considerable flexibility in energy profiling and can tune the pulse shape to meet the needs of the process. Single pulse interactions isolated the effects of pulse shape and pulse energy from the influences resulted from multiple pulsing, such as inter pulse period and multiple reflections. 4.3.1 Constant pulse duration When the pulse duration is kept constant, the peak power can be varied with pulse energy, this is achieved by changing the average power for a given PRF. Figure 4.15 reveals the change in pulse peak power at the leading edge with varying average power at 25 kHz PRF and 200 ns pulse duration. _____________________________________________________________________ 116 Figure 4.15: Pulse shapes at 200 ns pulse duration with varying pulse energy/peak power. Single pulse machining experiments were conducted on Si to investigate the effects of pulse energy in terms of peak power variation with 200 ns pulse, Figure 4.15. The maximum average power delivered to the substrate was 17.7 W, a maximum fluence of 67.86 J/cm2. The machined hole depths were measured with a white light interferometer, and the results plotted against fluence, Figure 4.16. The inset SEM images show the holes produced with a particular fluence. From these images, the average debris size can be seen to decrease with increasing fluence, demonstrating the transition from melt expulsion dominated machining to vaporization dominated machining. Similar observations were also found with Si machining using a Q-switched DPSS laser as the intensity was increased from 108 W/cm2 to 109 W/cm2 (Lim, Ki et al. 2006). _____________________________________________________________________ 117 Figure 4.16: Single pulse machined depth using a 200 ns pulse with variation in fluence. At low fluence, the machined depth is below 1 µm, and the process is dominated by surface vaporization verified by the fine vapour particles seen in the SEM image at 7.32 J/cm2. The surface deformation is caused by weak absorption on Si at the 1 µm wavelength and the low peak power (0.45 kW). As the fluence approaches 10 J/cm2, volumetric material removal in the form of micron-sized droplets is formed near the rim of the cavities. The molten silicon is pushed in the radial direction by vapour pressure and the hole rim is pushed above the surrounding wafer due to the hydrodynamic instability of the liquid layer (Christensen and Tillack 2003); (Willis and Xu 2000). Between 10 and 20 J/cm2, the evaporating particles exert increased recoil pressure on the Si melt pool, and molten mass is completed ejected from the irradiated zone (Allmen 1976); (Allmen and Blatter 1995); (Lee and Jeong 2004). An SEM image at 14.34 J/cm2 reveals a mixture of condensed vapour particles and re-solidified micro droplets (Chan and Mazumder 1987); (Semak and Matsunawa 1997). _____________________________________________________________________ 118 The machining depth curve begins to saturate at around 20 J/cm2. This occurs at the transition from the melt expulsion dominated mechanism to vaporization dominated mechanism. As the vapour density increases with increased laser peak power, the vapour becomes substantially ionized, and laser radiation becomes increasingly absorbed by the photo-excitation and photo-ionisation of species within the plume. When the plasma becomes dense, energy losses through inverse Bremsstrahlung become significant (Mao and Russo 1996); (Shinn, Steigerwald et al. 1986). With further increase in the plasma density and temperature, laser absorption occurs almost entirely in a small region near the crown of the plasma resulting in explosive heating of this region, and reduced coupling of the laser energy to the surface (Shinn, Steigerwald et al. 1986); (Chang and Warner 1996). High temperature plasma resulting from this form of laser heating has reported lifetimes that keep the ablation process going through electron conductive heating (Boley and Early 1994); (Russo, Mao et al. 1999). The experiments were repeated with pulse duration from 50 to 180 ns, Figure 4.17. Drill depth for varying pulse durations show a similar response to 200 ns, increasing rapidly after the damage threshold then reaching a saturation level which is lower for short pulses and higher for long pulses. The damage threshold is lower for shorter pulses in Figure 4.17, a phenomenon also observed with epoxy paint (Brygo, Dutouquet et al. 2006) and confirmed theoretically, numerically and experimentally on Si (Wang, Shen et al. 2010). The possible explanation proposed for this reduction in damage threshold is that the required threshold peak power can be achieved with lower fluence at a shorter pulse, which also results in a lower heat diffusion length and more efficient heating. _____________________________________________________________________ 119 Figure 4.17: Single pulse ablation depth as a function of fluence at varying pulse durations from 50 to 200 ns. The ratio of machined depth and pulse energy provides one-dimensional machining efficiency data (µm/mJ), which is plotted in Figure 4.18 against fluence for pulse durations between 50 ns and 200 ns. The highest machining efficiencies occur at fluences which are the saturation transition points for the depth curves of Figure 4.17. They are the points at which vaporization starts to dominate the machining mechanism rather than melt expulsion. _____________________________________________________________________ 120 Figure 4.18: Machining efficiency as a function of fluence at varying pulse durations from 50 to 200 ns. 4.3.2 Constant pulse peak power When the pulse peak power is kept constant, the pulse duration and the energy contained in the tail varies, Figure 4.19. The machining pulses, Figure 4.19, have equivalent peak power and rise time and provide the equivalent heating rate. _____________________________________________________________________ 121 Figure 4.19: Pulse shapes of the same pulse peak power with pulse duration between 40 ns and 200 ns. The depth of machined holes was plotted against fluence and pulse duration with inset SEM images, in Figure 4.20. The ablation is dominated by vaporization with limited melt resolidification, the machining mechanism is the same at all fluences as shown in the images. The increase in the fluence through increased pulse duration enriched further vapourisation, the hole diameters were seen to increase from 40 µm to 48 µm with finer debris. _____________________________________________________________________ 122 Figure 4.20: Single pulse ablation depth as a function of fluence (bottom x-axis) and pulse duration (top x-axis) with pulses of constant peak power. In the vaporization dominated regime and same fluence range of 18-52 J/cm2, the ablation depth increased by 5 µm when increasing pulse duration from 40 ns to 200 ns, compared to 0.8 µm when increasing peak power from 115 MW/cm2 to 630 MW/cm2 of Figure 4.16. The variation in ablation depth suggests that the pulse duration has much larger effect on ablation depth than pulse peak power within the same machining mechanism. The hole depth grows at a constant rate with pulse duration, and the drilling velocity (30 m/s) was taken as the slope of the linear part of the depth-vs-pulse duration curve, Figure 4.20. Similar observation was found on 1 µm wavelength Nd:YAG laser drilling of copper with a rectangular pulse shape, where the drilling velocity was approaching an asymptotic value of 22 m/s with increased intensity (Allmen 1976). The drilling depth in the (Allmen 1976) paper reached 300 µm at the pulse duration of 15 µs, it suggests there _____________________________________________________________________ 123 is still potential for the Si ablation depth to grow with pulse duration, which is limited to 200 ns by the fibre laser. 4.3.3 Constant pulse energy As described in the sections 4.3.1 and 4.3.2, pulse duration mostly influences the ablation depth at fixed peak power through a vaporization dominated mechanism, and pulse peak power influences the machining mechanism at fixed pulse duration. Both variations involve changes in pulse energy, but when the pulse energy is kept constant increasing pulse duration causes the peak power to decrease and vice versa. Figure 4.21 shows the pulse shapes available when peak power is reduced from 1.8 kW to 0.5 kW while maintaining constant pulse energy. Figure 4.21: Pulse shapes of the same pulse energy (0.28 mJ/ 8.49 J/cm2) with pulse duration between 75 ns and 200 ns and peak power between 1.8 kW and 0.5 kW. The longer pulse durations were predicted to ablate a deeper hole than the shorter, higher peak power pulses (Ready 1965). Physically, this means the heat has a better chance to flow into the interior of the material. This prediction was experimentally confirmed on _____________________________________________________________________ 124 laser ablation of Al targets (Breitling, Ruf et al. 2004), where the maximum melt depth and ablation depth were found to increase with pulse duration. Experiments were conducted on Si to investigate the influence of pulse shape at constant energy. The machined depths were plotted against pulse duration at various fluence values from 5.9 to 26.1 J/cm2, Figure 4.22. The hole diameters of the four fluence values were plotted against pulse duration with inset of several SEM images, Figure 4.23. Figure 4.22: Single pulse machined depth as a function of pulse duration at varying fluence values. _____________________________________________________________________ 125 Figure 4.23: Average diameter and SEM images of the selected curves in Figure 4.22, with respect to pulse duration at varying fluence values. The high fluence (>12 J/cm2) exhibit linear slopes, and the images (20.47 J/cm2) demonstrate the vaporization dominated ablation. Both observations suggest that the reduction of peak power is attained above the vaporization threshold, so the increase in pulse duration can still lead to a steady increase in ablation depth. Similar findings were reported by (Hendow, Shakir et al. 2010) with poly-crystalline silicon scribed by a 1 µm MOPA fibre laser. At medium fluence (8 – 12 J/cm2), the measured depth Figure 4.22 vary by less than 1 µm, which suggests the decrease of peak power is causing the transition in dominant machining mechanism from vaporization to melt expulsion, again seen in the images (8.49 J/cm2), Figure 4.23. The increase in pulse duration causes a slight increase in drill depth when material removal is vaporization dominated and a slight decrease when melt expulsion dominated. At low fluence (<8 J/cm2), the machined _____________________________________________________________________ 126 depth decreased with the increase in pulse duration as melt expulsion and surface vaporization are the dominant mechanisms. 4.3.4 Discussion 4.3.4.1 Machined quality and efficiency Si micro drilling at 1µm wavelength using gain switched pulses from Master Oscillator Power Amplifier based Yb fibre lasers presented an improved finish quality (O'Neill and Li 2009). Optimum pulse profiles were shown to contain an initial high peak power element with a fast rise time, less than 7.5 ns for a 10-90% rise in signal. Pulses such as those caused the surface layer of Si to heat rapidly leading to subsequent melting and vaporization regimes. The rapid melting of Si during the leading edge of the pulse leads to modification of the refractive index coefficient. The increased number of conduction electrons within liquid forces silicon to behave like a metal (Sokolowski-Tinten, Bialkowski et al. 1998). The transmission of 1 µm wavelength light is then reduced to almost to zero and remains low for several hundred nanoseconds (Lowndes, Jellison et al. 1982). This low transmission zone is accompanied by high reflectivity (Aydinli, Lo et al. 1981) and strongly increased absorption coefficient (Weakliem and Redfield 1979), which is characterised by high carrier density, and high carrier temperature. Within the several hundred nanoseconds timescale, vaporization dominated machining and plasma formation were enriched with the long energy tail of the laser pulse near the surface region, resulting in slower cooling rates and the minimisation of thermal defects such as micro-cracks usually formed during resolidification. Melt expulsion is an energy efficient mechanism, and for limited laser beam intensities it permits higher processing speeds at high repetition rates. The disadvantages are the generation of resolidified material, recast layers and wider extension of HAZ which are not suitable for high precision micromachining. The vaporization dominated mechanism has the advantage of smooth material edges, thinner recast layer, small extension of the oxidized surface and little HAZ; also loosely attached vapour condensates can be _____________________________________________________________________ 127 removed easily with ultrasonic cleaning. Vaporization requires higher laser intensities, is less energy efficient compared to melt expulsion. The energy needed to vaporise a unit volume of Si is nearly eight times that needed for melting (27.8x103 J/g compared with 3.6x103 J/g) (ICKnowledge 2003). The generation of a high density vapour plume can also lead to energy loss through vapour heating and plasma absorption via photoionization during the drilling process (Chang and Warner 1996). In the case of fixed pulse energy, the users can select the pulse shape which satisfies their needs, they can use higher, shorter pulse for maximum quality or they can use longer, lower pulse for maximum hole depth. There was a threshold (1 kW) on how low the peak power can go in order to have deep feature at longer pulse, this threshold was the minimum value that the 1 µm fibre laser was able to change Si optical properties. 4.3.4.2 Thermal diffusion The linear relationship between the ablation depth and pulse duration in Figure 4.20 is caused by the pulse duration extension. The energy contained in the tail of the long pulse promotes further vaporization and the ejection of molten material through sustained recoil pressures. The depth of the drilled hole is dependent on thermal diffusion length of the laser pulse, which is determined by heat diffusivity 𝐷 and the diffusion time 𝜏� shown in Equation 4.4. 𝑙� ≈ 2�𝐷𝜏� (4.4) For a 200 ns laser pulse working at 1 µm wavelength on Si at 300 K, the thermal diffusion length (𝑙�) is 8.26 µm based on Equation 4.4, which is of the same order as the ablation depth 6.05 µm. Shorter pulses limit the thermal diffusion depth resulting in reduced ablation depth. The thermal diffusivity decreases from 0.8 to 0.14 cm2/s with elevated temperature from 300 to 1400 K (Shanks, Maycock et al. 1963), which results in a decrease in thermal diffusion length from 8.26 to 3.35 µm according to Equation 4.4. There is an anomaly when Si is melted, the thermal diffusivity is measured to be 0.228 cm2/s (4.27 µm), which is twice as much as the value when Si is still in the crystal state at _____________________________________________________________________ 128 the melting point (Yamamoto, Abe et al. 1991). Under the vaporization dominated regime, the liquid region caused by the thermal diffusion is likely to be vaporised and expelled due to recoil pressure. There is evidence of resolidified melt expulsion around the hole rim, which was found with a 40 ns pulse, Figure 4.20. However, there is a mixture of liquid droplets and vapor ejected from the target with longer pulses (>100 ns), Figure 4.20, which a number of researcher claimed to be phase explosion (Lu, Mao et al. 2002); (Yoo, Jeong et al. 2000). More discussions on the phase explosion are in Chapter 6. 4.4 Summary The data obtained in this chapter allows answers to Research Question 1 to be provided. Which fibre laser parameters have the greatest influence on Si machining and what are their optimum settings? Percussion drilling and single pulse interactions were performed on Si using a 1 µm MOPA based fibre laser, great quality holes with sharp sidewalls and little defects (microcracks, recast layer or HAZ) were drilled which were not reported before. Various laser parameters were characterised and it was found that the pulse shape (peak power and pulse duration) affect Si machining processes the most. The pulse energy determines the amplitude of peak power at constant pulse duration and vice versa. The optimum setting of this fibre laser for quality purpose is to use the highest possible pulse peak power at pulse durations greater than 50 ns. With pulse durations less than 50 ns, the pulse profiles lose the energy tail and the machined results show considerable amount of resolidified melt splashes. Peak power was found to strongly influence the material removal mechanism. High peak powers (>1 kW) gave vaporization dominated ablation, left a limited resolidified molten layer and clean hole formation. For energy efficiency purpose, Figure 4.18 summarized the most energy efficient points for various pulse durations. The maximum drilling efficiency of 31.86 µm per mJ was _____________________________________________________________________ 129 achieved with a laser pulse of 0.14 mJ energy (13.3 J/cm2 fluence), 200 ns pulse duration and 0.97 kW peak power. The most energy efficient fluence value was at the transition point between melt expulsion and vaporization dominated mechanisms. The recoil pressure at this point was high enough to cause effective melt expulsion, while the vapour density at this point was low enough not to affect the incoming laser beam. _____________________________________________________________________ 130 Chapter 5 Performance comparison between IR and UV lasers 5.1 Introduction This chapter aims to answer Research Question 2. How does the IR fibre laser compare with the DPSS UV laser on single pulse machining of Si? The main difference between the IR fibre laser and the DPSS UV laser is the wavelength, it is the most important laser parameter which determines how much energy is absorbed by the material and how the energy is delivered. In the case of Si, the wafer is thick enough that no visible light will transmit through the substrate, so the value of reflection allows the amount of energy available to the material to be calculated. The absorption coefficient determines how far light travels inside the material, hence the heating efficiency can be determined. In addition, the wavelength determines the machining resolution, i.e., the minimum spot size which has an influence on the peak power density. A MOPA based fibre laser with an M2 = 1.2 was used, closer to the mode of typical DPSS UV laser used in industrial Si micro-machining. Single pulse machining of silicon using a non process optimised Q-swithced DPSS laser at 355 nm is compared with that achieved using a MOPA based fibre laser (M2 = 1.2) at 1064 nm. The comparison includes the machined depth, machining efficiency and hole quality, with surface temperature calculations to identify the differences caused by the wavelength. The difference in performance of two fibre lasers (M2 = 2 and M2 = 1.2) are also compared. 5.2 Performance comparison between IR and UV lasers The M2 = 1.2 fibre laser was selected because it has similar M2 values and spot sizes, to the DPSS UV laser. _____________________________________________________________________ 131 5.2.1 Single pulse machined depth and efficiency The UV laser provides a maximum, 10W, average power at 60 kHz, giving a maximum pulse energy of 0.17 mJ. The UV laser is integrated with a scan head and lens with a focal length of 103 mm. The minimum spot size is calculated to be 17 µm, with a maximum fluence of 69.5 J/cm2. The pulse shapes of the UV laser were measured with a photodiode and results plotted in Figure 5.1. Unlike the fibre laser pulse shapes, the UV pulses do not have a fast rise time and long plateau. The UV pulses are pseudo-Gaussian shaped with full durations 35-58 ns, FWHM 20-30 ns, and 12-14 ns for a 10-90% rise in signal. The second burst of energy seen at around 70 ns were generated in the sensor circuitry and do not represent the actual beam. The pulse parameters in Figure 5.1 were used during the comparison experiments. Figure 5.1: Temporal pulse profile of the DPSS UV laser at 60 kHz with varying average power up to 10W. _____________________________________________________________________ 132 The single pulse machined depth was plotted with respect to pulse energy for both the UV laser and IR fibre laser, Figure 5.2 and Figure 5.3 (the inset images are of the same scale). The fibre laser machining data were obtained for three pulse durations (65 ns, 200 ns and 220 ns), Figure 5.3. Figure 5.2: Single pulse machining depth as a function of pulse energy for the UV laser at 58 ns. _____________________________________________________________________ 133 Figure 5.3: Single pulse machining depth as a function of pulse energy for the fibre laser at 65, 200 and 220 ns. When the pulses of similar pulse duration are compared, the UV laser (58 ns) outperforms the fibre laser (65 ns) with lower machining threshold and higher machined depth in the same pulse energy range (< 0.16 mJ). The longer pulses of the fibre laser were, 200 and 220 ns, and the results compared in three distinct regions: Low pulse energy (≤ 0.04 mJ). There was little material removal due to poor energy coupling between the IR wavelength laser beam and Si. The surface temperature 𝑇�� at the centre of the irradiated zone for a volume source can be calculated from Equation 5.1, which was derived from the heat conduction differential equation (Metev and Veiko 1998). Details on how the equation is derived are included in Section 6.2. 𝑇� = ��(���)������ + 𝑇� (5.1) _____________________________________________________________________ 134 𝐼� is the laser irradiance at the surface, 𝑅 is the surface reflectivity, 𝜌 is the density of solid Si, 𝐶� is the specific heat of solid Si, 𝑙� = �� is the optical penetration depth and 𝑇� is the sample’s initial temperature (300 K). With a pulse energy of 0.04 mJ, the long optical penetration depth (200 µm) and flat pulse profile only produced a low peak power of 208 W, Figure 5.4(a). The surface temperature of Si was raised to 1300 K by the end of the pulse, Figure 5.4(b). The melting temperature of 1683 K was not reached during the laser pulse even with the reduced optical penetration depth due to elevated temperature. (a) (b) Figure 5.4(a): Temporal pulse profile of M2=1.2 fibre laser with 0.04 mJ pulse energy, 208 W peak power. (b): The temperature evolution of Si surface at the centre of laser irradiation with the laser pulse shown in (a). For the UV laser, the optical absorption depth was only 10 nm for solid Si (Bauerle 2000) and 7.7 nm for liquid Si (Jellison, Lowndes et al. 1986). The 0.04 mJ laser pulse produced a peak power of 1740 W and was able to raise the surface temperature to the melting point within 6 ns from the beginning of the laser pulse, Figure 5.5. The equation for a surface heating source was used, Equation 5.2 (Metev and Veiko 1998), and it was based on the heat conduction differential equation, Section 6.2. _____________________________________________________________________ 135 𝑇� = ���(���)√���√� + 𝑇� (5.2) 𝜅 is the thermal conductivity and 𝐷 is the thermal diffusivity. (a) (b) Figure 5.5(a): Temporal pulse profile of the UV laser with 0.04 mJ pulse energy, 1740 W peak power. (b): The temperature evolution of Si surface at the centre of laser irradiation with the laser pulse shown in (a). Intermediate pulse energy (0.04 – 0.1 mJ). There was a considerable increase of the IR laser machined depth, Figure 5.3. As the laser heating becomes more effective with the increasing peak power, Si was melted and the molten layer was expelled by the vaporization and plasma pressure (Tunna, Kearns et al. 2001). There was little increase in the UV laser machined depth in this region, the machined hole starts to widen instead of deepen, Figure 5.2. The 0.04 mJ pulse energy corresponds to the power density of 0.65 GW/cm2, it is beyond the plasma shielding threshold of 0.2-0.3 GW/cm2 (~0.015 mJ) (Russo 1995). The later part of laser pulse energy was absorbed, reflected and scattered from the dense plasma before it reaches the solid sample as a result. The M2=1.3 UV laser has a pseudo Gaussian spatial distribution, the energy curves were plotted with the normalized unit, Figure 5.6. A value of 0.3 was assumed to be the melting threshold, the energy points above the threshold will melt/machine the Si surface, and the points below the threshold will only heat the Si substrate up to the melting temperature. Gaussian curves represent varying pulse energies, and they result in different machining diameters. _____________________________________________________________________ 136 Figure 5.6: Effects of pseudo Gaussian spatial energy distribution on the machined diameters. High pulse energy (> 0.1 mJ). The IR fibre laser soon reached the plasma shielding threshold (0.2 GW/cm2, 0.15 mJ), and the amount of laser energy reaching the surface was limited. The machined depth for both lasers reached a plateau; 6.8 µm/pulse was achieved with the fibre laser compared with a 4.5 µm/pulse with the UV laser with pulse energies close to 0.17 mJ. The single pulse machining energy efficiency was plotted for the UV laser and fibre laser with inset images, Figure 5.7. _____________________________________________________________________ 137 Figure 5.7: Single pulse machining energy efficiency as a function of pulse energy for both the UV laser (58 ns) and IR fibre laser (200 ns). The highest machining efficiency is greater than 400 µm/mJ for the UV laser, which is six times higher than the fibre laser (~60 µm/mJ), Figure 5.11(a), and it occurred with a much lower pulse energy of 0.003 mJ compared with 0.085 mJ. With the high absorption coefficient, the 0.003 mJ UV laser pulse can reach the melting temperature during the laser pulse. The smaller spot size which resulted in higher power density was also in favor of the UV laser. The theoretical melting threshold based on the fact that the surface temperature was raised beyond melting temperature is calculated to be 0.0003 mJ pulse energy. With pulse energies higher than 0.07 mJ, the machining efficiency was higher for the fibre laser. The UV laser reached a plasma shielding threshold with a pulse energy of 0.015 mJ, while the IR laser did not reach such a threshold until 0.15 mJ. _____________________________________________________________________ 138 5.2.2 Single pulse machined mechanism and finish quality The UV laser machined holes with high energy efficiencies greater than 200 µm/mJ were only 80 % circular, Figure 5.7. They are also inconsistent, because the specification for the beam circularity was only >85% for a new laser, with an unspecified power range. Circularity was seen to improve with increased power. Circular holes were formed with the fibre laser, which had a radially uniform beam. The UV produced a smaller hole size (~20 µm with pulse energies < 0.07 mJ), but the inconsistent system performance, which resulted in significant error bars, is not suitable for industrial applications such as surface texturing. The hole size of both lasers was comparable with pulse energies higher than 0.07 mJ. The holes produced with high pulse energies of the UV laser reveal a large amount of resolidified melt beyond the rim. Holes produced by the IR fibre laser show a mixture of micron and sub-micron sized debris, with streaks of melt ejection close to the rim, Figure 5.8. Comparing the performances of two lasers, the hole machined with the IR fibre laser show clearer defined edges and less resolidifed melt splashes. The intensity of the UV laser is high enough to compensate the power loss associated with the full liquid expulsion, but there is no energy enriched long tail to further vaporise the molten material. _____________________________________________________________________ 139 (a) (b) Figure 5.8(a): SEM image of single pulse machined hole (6.4 µm deep) by the M2 = 1.2 fibre laser, 0.171 mJ pulse energy. (b): SEM image of single pulse machined hole (4.6 µm deep) by the UV laser, 0.164 mJ pulse energy. 5.2.3 Discussion An interpretation for physical effects taking place during IR and UV wavelength processing must take into consideration the material properties of silicon. The optical properties of a material such as optical absorptivity and reflectivity strongly depend on the nature of the material and the wavelength of the incident radiation (Gray 1972). The UV laser was able to achieve higher power density with a smaller spot size of the same pulse energy compared with the IR fibre laser. The energy was strongly absorbed in an optical penetration depth of 10 nm, which resulted in Si becoming molten within a few nanoseconds after the beginning of the laser pulse based on the thermal conduction equation. The UV laser managed to cause melting/melt expulsion and achieve much higher energy efficiency than the IR laser with low pulse energies (< 0.07 mJ), Figure 5.7. _____________________________________________________________________ 140 The IR laser did not manage to achieve a high heating rate because the peak power was low (<410 W) with low pulse energies (< 0.07 mJ). The IR laser was considered as volume source with solid Si, the energy was spread out over an optical penetration depth of 200 µm. As the peak power increased with pulse energy, the surface temperature increased faster and the Si optical properties were modified through the elevated surface temperature (Jellison and Lowndes 1982); (Weakliem and Redfield 1979); (Macfarlane, McLean et al. 1958). Si was melted faster with increased peak power. The molten Si, which has metal like optical properties such as high reflectivity and low transmissivity, was then exposed to the increasing remaining part of the fibre laser’s long pulse (200 ns). The rest of pulse energy was then considered to be a surface source as the optical penetration depth decreased to 14 nm. The long thermal diffusion length resulted in a thick liquid layer, which was removed through vaporization and melt expulsion. When the liquid layer is being heated rapidly to the thermodynamic critical temperature (5200 K), explosive boiling is likely to occur (Lu, Mao et al. 2002). In case of the explosive boiling, there was a delay of several hundred nanoseconds for the droplets to leave the silicon surface which could last up to several tens of microseconds (Yoo, Jeong et al. 2000). A shadowgraph technique has been used to investigate the material removal mechanism induced by the IR fibre laser and more details of explosive boiling are documented, Chapter 6. 5.3 Comparison between fibre lasers of different M2 Single pulse experiments were conducted on Si with a 200 ns pulse at varying peak powers using two fibre lasers; one with M2 = 1.2, the other M2 = 2.0. Smaller spot sizes and better focus performance were expected from the M2 = 1.2, which have effects on machining processes such as depth, efficiency, mechanism and finished quality. The importance of these effects are discussed. _____________________________________________________________________ 141 5.3.1 Single pulse machined depth and efficiency The single pulse machined depth with respect to pulse energy and fluence was determined for both lasers, Figure 5.9. The maximum pulse energy was 0.55 mJ for the M2 = 1.2 laser compared to 0.71 mJ for the M2 = 2 laser due to different operating frequencies, Figure 5.9(a). The theoretical spot size of 25 µm was calculated from the M2 = 1.2, which resulted in a higher corresponding fluence of 109.0 J/cm2, Figure 5.9(a). (a) (b) Figure 5.9(a): Single pulse machining on Si with fibre lasers of different M2, depth plotted against pulse energy with a 200 ns pulse. (b): depth plotted against fluence. The curves in Figure 5.9 are of similar shape; a sharp increase after the damage threshold followed by saturation after the transition. The M2 = 1.2 laser had depth advantage over the M2 = 2 laser; around 3 µm with pulse energies less than 0.15 mJ and around 1 µm with pulse energies higher than 0.15 mJ, Figure 5.9(a). For a given pulse energy (0.1 mJ) the heating becomes more efficient when concentrated on a smaller spot. The resulting higher fluence and peak power density can cause faster rise of the surface temperature based on the thermal conduction model. The solution for the volume source (𝑇�� = ���(���)√���√� + 𝑇�) was used here to calculate the temperature distribution at the centre of the irradiated surface based on the temporal pulse profiles of two lasers, Figure 5.10(a). The results showed that the onset of melting phase started 100 ns earlier for the M2=1.2 laser, Figure 5.10(b). From the melting point the optical _____________________________________________________________________ 142 energy was coupled into the molten Si much more effectively, because the optical penetration depth of 1 µm wavelength decreases dramatically from solid Si to liquid Si (from 200 µm to 14.3 nm). With the temporal pulse profiles similar to each other, the M2 = 1.2 has more energy left in the pulse to remove the molten Si. (a) (b) Figure 5.10(a): Temporal pulse profile of two fibre lasers with 0.1 mJ pulse energy, ~605 W peak power. (b): Si surface temperature distribution of two fibre lasers with time. In the fluence plot, Figure 5.9(b), both lasers achieved a similar machined depth at fluences lower than 20 J/cm2. Above 20 J/cm2 the M2=1.2 laser outperformed by less than 10 %, which was a result of the more Guassian like spatial distribution since the factor of spot size was ruled out with the fluence comparison. With fluences greater than 20 J/cm2, there was no significant change in the machined depth for both fibre lasers which suggest a physical limitation in the drilling mechanism. The change in the laser material interaction was replaced by the interaction between laser induced plasma plume and the irradiated surface. The additional energy is being absorbed or scattered by the dense plasma, resulting in the constant material removal mechanism (Ihlemann, Wolff et al. 1992). During the irradiation of a flat surface, the plume can expand freely and heating of the surface might be limited. _____________________________________________________________________ 143 Figure 5.11: Single pulse machining efficiency with respect to pulse energy for both lasers with 200 ns pulse. Considering the energy efficiency of the process, Figure 5.11, both curves exhibit similar shapes; a sharp increase followed by steady decrease with the pulse energy. The highest machining efficiency for both lasers occurs at the transition points for the depth curves in Figure 5.9(a). The M2 = 1.2 laser achieved higher machining efficiencies at all pulse energies, this is particularly evident with low pulse energies, e.g. 60 µm/mJ compared with 30 µm/mJ, Figure 5.11. With a smaller spot size, the peak power density was 0.12 GW/cm2 for the M2 = 1.2 laser, twice the value compared with the M2 = 2 laser with the same 0.1 mJ pulse energy. The M2 = 1.2 laser has advantages on Si which are particularly noticeable with low pulse energies (<0.15 mJ). With the limited average power, the associated higher repetition frequencies can be used to increase production rates. 5.3.2 Single pulse machined mechanism and finish quality The M2 = 1.2 laser machined holes were examined, Figure 5.12, and the surface debris used to determine the machining mechanism. _____________________________________________________________________ 144 Figure 5.12: The M2 = 1.2 fibre laser single pulse machining of Si as a function of fluence with a 200 ns pulse and inset SEM images. As with the M2 = 2 laser, the machining mechanism of the M2 = 1.2 laser progressed from surface vaporization at low fluence (< 10 J/cm2), to a melt expulsion dominated regime between 10 J/cm2 and 25 J/cm2, finally reaching a vaporization dominated regime at fluences above 25 J/cm2. Peak power density was approaching 0.2 GW/cm2 at 25 J/cm2 transition, it is the onset of plasma shielding and less energy is coupled to the work piece (Russo 1995). Hole quality was also comparable between the two lasers, but between 25 J/cm2 and 50 J/cm2, the holes machined by the M2 = 1.2 laser show clear sign of splashing while the M2 = 2 laser left an extensive particulate field. Enlarged images of finished holes are presented, Figure 5.13, the laser parameters and machined depth are tabulated, Table 5.1. _____________________________________________________________________ 145 (a) (b) Figure 5.13(a): SEM image of single pulse machined hole by the M2 = 1.2 fibre laser. (b): by the M2 = 2 fibre laser. Table 5.1: Two fibre laser parameters used to generate features in Figure 5.13 and the machined hole depth. Laser M2 = 1.2 M2 = 2 Pulse energy (mJ) 0.2 0.4 Fluence (J/cm2) 39.4 37.6 Peak power density (GW/cm2) 0.3 0.38 Machined hole depth (µm) 6.4 6.0 The particle field around the hole rim, generated by the M2=2 laser, consisted of fine molten droplets and condensation of the Si vapour, Figure 5.13(b). It extends for hundreds of microns radially away from the hole. The observation is consistent with the vaporization dominant regime associated with high peak powers (Jackson and O'Neill 2003). While the machining mechanism of the M2 = 1.2 laser tended more towards the melt expulsion dominated regime, Figure 5.13(a). The bottom of the hole machined by the M2 = 1.2 laser was smooth, because the peak power was consistently in the centre of the drilled hole, and the laser generated vapour _____________________________________________________________________ 146 pressure forced the molten material out of the hole with a high degree of symmetry, Figure 5.13(a). The machined depth was slightly higher as a result, with molten layer not falling back to the interaction zone. The bottom of the hole machined by the M2 = 2 laser contains non uniform resolidified melt. It is thought this is a consequence of “hot spots” from a non uniform laser energy distribution, which resulted in the low vapour pressures from the reduced intensities in the centre. 5.3.3 Discussion The micron- /submicron-sized particulate field in Figure 5.13(b) was thought to be caused by an explosive material removal mechanism, called phase explosion/explosive boiling. It is still under debate what the conditions are to induce phase explosion. (Yoo, Jeong et al. 2000) observed a dramatic increase of the depth and volume for a threshold irradiance of above 22 GW/cm2 for phase explosion, drilling Si with a single pulse from a Nd:YAG laser. (Craciun and Craciun 1999) observed the formation of micron-sized, round-shaped cavities as signs of subsurface volume boiling with a Nd:YAG laser of 400 MW/cm2 laser irradiance. This study observed a clean hole formation and fine particulate field with a long asymmetrical pulse of 160 MW/cm2 peak power density threshold. The peak power density of an asymmetrical pulse is different to that from the laser irradiance of a Gaussian pulse, the Gaussian pulse is missing the long energy tail. The complete laser parameters for these three studies are in Table 5.2. _____________________________________________________________________ 147 Table 5.2: Laser parameters used to in literature and this study for explosive material removal mechanism. Lasers Nd:YAG (Yoo, Jeong et al. 2000) Nd:YAG (Craciun and Craciun 1999) Fibre (this study) Wavelength (nm) 266 1064 1064 Pulse width (ns) 3 (FWHM) 15 (FWHM) 200 (full duration) Pulse energy (mJ) 0.64 N/A 0.23 Fluence (J/cm2) 66.5 6 22 Laser irradiance/ peak power density (GW/cm2) 22 0.4 0.16 However, the most important thing is the energy available to the material for the heat generation and conduction after all the losses are considered. The plasma shielding work by (Mao and Russo 1996) showed that only 20% of the laser beam energy was transmitted through the plasma to the sample surface at 22 GW/cm2, so the threshold irradiance on the sample was 4.4 GW/cm2 (0.13 mJ pulse energy). A slightly different threshold irradiance of 3.1 GW/cm2, pulse energy of 0.09 mJ was calculated based on the thermal conduction model of a one-dimensional semi-infinite solid with a constant surface heat flux (Yoo, Jeong et al. 2000). Two values were close to each other, the 0.04 mJ energy difference could be used to cover other losses besides plasma shielding. The other losses during the machining process include heat loss due to the surface vaporization, energy loss due to the latent heat of melting and vaporization, and they are not considered in Russo’s model. The 0.23 mJ pulse energy threshold for the IR fibre laser would have little plasma shielding due to a much lower peak power density (Russo 1995). The calculated 0.09 mJ from Figure 5.10 was used to raise the Si surface temperature to the melting point, so the rest of the energy could be considered as a surface source with the metal-like liquid Si. The pulse energy value was not available in the (Craciun and Craciun 1999) paper, but the IR fibre lasers in our study have higher fluences and peak power densities, Table 5.2. In addition, the pulse profile has an effect on how energy is conducted inside the _____________________________________________________________________ 148 material. A thermal conduction model was used to investigate the temperature distribution with the IR fibre laser pulse to identify the possibility of phase explosion, Chapter 6. 5.4 Summary The data obtained in this chapter allows answers to Research Question 2 to be provided. How does the IR fibre laser compare with the DPSS UV laser on single pulse machining of Si? The single pulse machined depth, efficiency and hole quality were compared between the M2 = 1.2 fibre laser and the M2 = 1.3 DPSS UV laser. With lower pulse energies (<0.04 mJ), the UV laser achieved higher machined depth, energy efficiency and lower machining threshold. This is due to the high absorption coefficient (106 cm-1) on solid Si of UV wavelength, compared with 50 cm-1 at IR wavelengths. As pulse energy increases, the fibre laser started to melt Si earlier in the pulse, and more energy was left to machine liquid Si, which has absorption coefficient of 7×105 cm-1 at IR. With pulse energies of approaching 0.17 mJ, the maximum machined depth of the fibre laser was 2.3 µm higher than the DPSS laser (6.8 µm/pulse compared to 4.5 µm/pulse). The UV laser machined holes have a large amount of resolidified melt beyond the hole rim, while the fibre laser machined holes leave a mixture of vapour condensates and limited re-solidified droplets, demonstrating the effective drilling of Si using these fibre laser sources at longer pulse duration. The M2 = 1.2 fibre laser was also compared with the M2 = 2 fibre laser with a fixed 200 ns pulse duration. The better focusibility and smaller spot size allowed efficient material removal to happen with a lower pulse energy; a 5 µm hole depth was machined with less than 0.1 mJ pulse energy for the M2 = 1.2 laser while the same result was reached with twice the energy from the M2 = 2 laser. _____________________________________________________________________ 149 Chapter 6 Theoretical analysis 6.1 Introduction This chapter aims to answer Research Question 3. How does theoretical analysis and imaging tools help understand the IR fibre laser Si interaction? The IR fibre laser achieved great results at Si micro-drilling, which is unusual at the 1 µm wavelength. The interaction mechanism is thought to be different from the normal 1 µm laser volumetric heating of Si, since the asymmetrical pulse shape alters the way energy is delivered to the Si. When the laser irradiates a solid target, the temperature at the target surface rises and the target goes through melting/melt expulsion, vaporization or even explosive boiling. During the irradiation process a range of laser parameters and material properties could affect the energy coupling, the temperature rise and the subsequently machining mechanism. Some of material properties are temperature and phase dependent, which change the material response to the incoming laser beam. A one-dimensional thermal model is used here to predict the temperature change on the surface and within the target, which helps to determine the dominant material removal mechanism that Si went through during the IR irradiation and estimate the machined depth. Some imaging tools, like laser shadowgraphy and visible imaging, were used to examine the interaction zone in greater details. The time scale for the onset of mass ejection can be estimated, the size and speed of the ejected particle can be approximated. The obtained information can be used to validate the material removal mechanism. _____________________________________________________________________ 150 6.2 Fourier heat conduction With the nanosecond pulsed laser interaction, the temperature equilibrium between electrons and phonons of the irradiated target is reached. The time dependent temperature distribution in the target can be calculated with the classic Fourier heat conduction equation (Ready 1965); (Bogaerts, Chen et al. 2003); (D. Bäuerle, M. Geiger et al. 2004); (Hendow and Shakir 2010), Equation 6.1. 𝜌𝐶� ��(�,�) �� = � �� �𝜅 ��(�,�) �� � + 𝛼𝐼(𝑧, 𝑡) (6.1) where 𝑇 represents the temperature inside the Si target, 𝑧 is the distance into the Si substrate from the surface with a direction perpendicular to the surface, 𝑡 is time, 𝜅 is the thermal conductivity, 𝜌 is the density of solid Si, 𝐶� is the specific heat of solid Si, 𝛼 is the optical absorption coefficient. Some initial assumptions are tabulated in Table 6.1. Table 6.1: Assumptions for the heat conduction model. Assumptions Justifications Semi-infinite medium (Si) The sample thickness (650 µm) > the optical penetration depth (200 µm @ 1 µm wavelength), thermal penetration depth (8.3 µm) and spot size (36.5 µm) No lateral heat conduction (1-D) The radius of laser spot size (18.2 µm) > the thermal penetration depth (8.3 µm) Ambient conditions Constant pressure (1 atm), initial room temperature (300 K) No plasma screening Low peak power density (<1.22 GW/cm2) No surface heat convection No re-radiation at the surface The first term of the right-hand side in Equation 6.1 represents the heat conduction, whereas the second term refers to the energy source; 𝐼(𝑧, 𝑡) stands for the laser irradiance as a function of time and position in the target. Assuming there is no surface vaporization, the energy term of a volumetric heating source can be written according to the Lambert’s law of radiation absorption: _____________________________________________________________________ 151 𝐼(𝑧, 𝑡) = (1 − 𝑅)𝐼�𝑒��� (6.2) 𝐼� is the incident laser irradiance at the surface, and 𝑅 is the surface reflectivity. The appropriate initial and boundary conditions are: 𝑇(∞, 𝑡) = 𝑇(𝑧, 0) = 𝑇�, (6.3) 𝜅 ��(�,�) �� = 0 (6.4) where 𝑇� is the sample’s initial temperature (300 K) across the target. All the material optical and thermal constants in Equation 6.1-6.4 are assumed to be temperature independent, the surface irradiation is assumed to be constant and uniform. The solution of the boundary problem has the following form: 𝑇(𝑧, 𝑡) = 𝑇� + ��(�)� √𝐷𝑡 · 𝑖𝑒𝑟𝑓𝑐 � ��√��� − �(�)������ + �(�)��� 𝑒������� · 𝑒𝑟𝑓𝑐 �α√𝐷𝑡 − ��√��� + �(�) ��� 𝑒�� ����� · 𝑒𝑟𝑓𝑐 �α√𝐷𝑡 + � �√�� � (6.5) where 𝑒𝑟𝑓𝑐(𝑢) = � √� ∫ 𝑒�� � 𝑑𝜁 � � and 𝑖𝑒𝑟𝑓𝑐(𝑢) = ∫ 𝑒���𝑑𝜁�� = ����√� − 𝑢 · 𝑒𝑟𝑓𝑐(𝑢) (6.6) are the complementary error function and its integral. 𝐷 = � ��� is the thermal diffusivity. Depending on the relationship between the thermal diffusion length 𝑙� = 2√𝐷𝜏 (𝜏 is the full pulse duration) and the optical penetration length 𝑙� = ��, the expression for surface temperature (𝑇�) calculations at the centre of the irradiated zone can be simplified (Metev and Veiko 1998); (D. Bäuerle, M. Geiger et al. 2004): 𝑇�(𝑡) = ��(���)������ + 𝑇� for 𝑙� ≫ 𝑙� (volume heating source) (6.7) 𝑇�(𝑡) = ���(���)√���√� + 𝑇� for 𝑙� ≪ 𝑙� (surface heating source) (6.8) _____________________________________________________________________ 152 Equations 6.7-6.8 are applied to the cases where the heating source is constant and there is no phase change. The time needed for Si to reach its melting temperature (𝑇� =1683 K) can also be estimated. However, the fibre laser has a non-square temporal pulse shape, i.e. 𝐼(𝑡) is a prescribed function of time, so heat conduction Equations 6.1-6.8 cannot be applied directly to calculate the temperature distribution. In this case, a general solution for the temperature was derived, Equation 6.9 (Carslaw and Jaeger 1959); (Ready 1965); (Ready 1971). 𝑇(𝑧, 𝑡) = ∫ ∫ �(�) �� � �� ���(��,���) ��� 𝑑𝑧�𝑑𝜏 � � � � (6.9) 𝑇�(𝑧, 𝑡) is the calculated temperature for the case of a square pulse of power density 𝐼�. When one attempts to use this formula for functions 𝐼(𝑡), analytic solutions are no longer available (Ready 1965). For an asymmetrical pulse shape used in this study, the pulse was temporally divided into many small time intervals 𝑡 = 𝑡� + 𝑡� + 𝑡� … + 𝑡� = 𝑛𝑡� and 𝑡� = 𝑡� = 𝑡� = ⋯ = 𝑡� where 𝑛 is the number of time intervals at time (𝑡). A square pulse was considered within each time interval (𝑡� =0.2 ns). The pulse shape was then measured and the peak power density at each time interval calculated (𝐼�, 𝐼�, 𝐼�, 𝐼�, 𝐼� … 𝐼�). In the case of a volume heating source, Equation 6.7 becomes 𝑇�(𝑡) = ��(���)�������� + 𝑇� = (���)����� (𝐼�𝑡� + 𝐼�𝑡� + 𝐼�𝑡� + ⋯𝐼�𝑡�) + 𝑇� = (���) ����� 𝑡�(𝐼� + 𝐼� + 𝐼� … + 𝐼�) + 𝑇�, (6.10) similarly in the case of a surface heating source, Equation 6.8 becomes 𝑇�(𝑡) = ���(���)√���√� + 𝑇� = �(���)√��√� 𝐼��𝑛𝑡� + 𝑇� = �(���)√� �√� �𝐼� �𝑡� + 𝐼��𝑡� + 𝐼��𝑡� … + 𝐼��𝑡� + 𝑇� _____________________________________________________________________ 153 = �(���)√���� �√� �𝐼� � + 𝐼�� + 𝐼�� … + 𝐼�� + 𝑇�, (6.11) and 𝐼� = ������������…����� . (6.12) 6.3 Volume heating source When a 1064 nm wavelength laser irradiates Si, the optical penetration depth (𝑙� = 200 µm) is much longer than the thermal diffusion length (𝑙� = 8.26 µm), and the laser beam is normally considered as a volume heating source. In this case, the forward heat conduction can be neglected. The surface temperature evolution of Si irradiated by the fibre laser with the highest peak power density (1.22 GW/cm2) was calculated according to Equation 6.10 with the material properties in Table 6.2. The results are plotted in Figure 6.1. Table 6.2: Si optical and thermal properties used in heat conduction Equations 6.7 and 6.10, at room temperature 𝑇� =300 K. Property (Solid Si) Values Units References Absorption coefficient (𝛼) 50 cm-1 (Bauerle 2000); (Weakliem and Redfield 1979) Optical penetration depth (𝑙�) 0.02 cm Reflectivity (𝑅) @ 1.152 µm 31.9 % (Bauerle 2000); (Jellison, Lowndes et al. 1986) Specific heat (𝐶�) 0.745 J/g·K (Desai 1986) Density (𝜌) 2.32 g/cm3 (Unamuno and Fogarassy 1989); (Lukeš, šášik et al. 1992) Thermal conductivity (𝜅) 1.397 J/ cm·s·K (Yoo, Jeong et al. 2000); (Unamuno and Fogarassy 1989) Thermal diffusivity (𝐷) 0.81 cm2/s (Yamamoto, Abe et al. 1991) _____________________________________________________________________ 154 Figure 6.1: Surface temperature distribution of Si under the 1064 nm fibre laser irradiation with a peak power density of 1.22 GW/cm2. The Si surface did not reach the melting temperature by the end of the laser pulse, which is not true in the experiments. In a real case scenario, the material properties largely depend on temperature and the phase of the material. For example, the absorption coefficient at 1 µm wavelength can be written as a function of temperature, α ≈1.13𝑇-289, approximated from the temperature curve between 300 and 1140 K in Figure 2.5 (Jellison and Lowndes 1982); (Weakliem and Redfield 1979); (Macfarlane, McLean et al. 1958). The optical penetration depth of the molten Si is reduced to 14.3 nm (Muller, De Unamuno et al. 1996). The other properties of Si can also be written as a function of temperature, Table 6.3, and they are plotted in Figure 6.2. _____________________________________________________________________ 155 Table 6.3: Si optical and thermal properties at 1 µm wavelength as a function of temperature (<1683 K). Property (Solid Si) Values Units References Absorption coefficient (𝛼) 1.13𝑇 − 289 cm-1 (Jellison and Lowndes 1982); (Weakliem and Redfield 1979); (Macfarlane, McLean et al. 1958) Optical penetration depth (𝑙�) 11.13𝑇 − 289 cm Reflectivity (𝑅) @ 1.152 µm 31.2 + 2.2 × 10��𝑇 % (Jellison and Lowndes 1982) Specific heat (𝐶�) 0.694 × 𝑒�.���×����� J/g·K (Yoo, Jeong et al. 2000); (Unamuno and Fogarassy 1989) Density (𝜌) 2.32 g/cm3 Thermal conductivity (𝜅) 1521𝑇��.��� for 𝑇 ≤1200 K 8.98𝑇��.��� for 1200 K< 𝑇 < 𝑇� J/ cm·s·K Thermal diffusivity (𝐷) 𝐷 = 𝜅 𝜌𝐶� cm 2/s Figure 6.2: Thermal and optical properties of Si as a function of temperature. The absorption coefficient of Si at 1 µm wavelength increased from 101 to 103 cm-1, which could make laser heating of material much more effective. The heating is further enhanced when Si becomes molten. The decrease in the thermal conductivity and diffusivity could also slow the energy loss due to heat conduction. With time being divided into many small intervals, the material properties are considered to be constant within each interval. The surface temperature evolutions based on _____________________________________________________________________ 156 temperature dependent material properties are calculated and plotted, Figure 6.3, for a number of pulse profiles with a fixed pulse duration (200 ns) but varying peak power densities (<1.22 GW/cm2) in Figure 4.15. With a high peak power density (1.22 GW/cm2) and fast rise (<7.5 ns for a 10-90% rise in signal) laser pulse, the surface temperature of Si increases much faster and Si surface is melted in only ~10 ns. When laser pulses with lower peak power densities are used, the Si surface is melted in longer times. Figure 6.3: Surface temperature distribution of Si irradiated by laser pulses in Figure 4.15. The volume heating source was considered before the Si surface was melted. The energy used to melt the Si surface and the energy left in the pulse are calculated for laser parameters in Figure 6.3, Table 6.4. _____________________________________________________________________ 157 Table 6.4: The onset time of melting and the percentage of pulse energy used to melt Si. Laser pulse energy (mJ) Fluence (J/cm2) Peak power density (GW/cm2) Onset time of melting (ns) Pulse energy used to melt Si (mJ) Percentage of pulse energy used to melt Si (%) Energy left as a surface heating source (mJ) 0.71 67.86 1.22 10.4 0.0935 13.17% 0.62 0.56 53.22 0.63 19.0 0.0926 16.54% 0.47 0.39 37.61 0.38 29.6 0.0915 23.47% 0.30 0.31 29.52 0.27 41.4 0.0909 29.31% 0.22 0.23 22.08 0.16 66.4 0.0906 39.39% 0.14 0.15 14.34 0.098 114.0 0.0903 60.22% 0.06 0.11 10.93 0.069 160.2 0.0903 82.05% 0.02 0.08 7.32 0.043 N/A N/A N/A N/A The energy used to melt the Si surface is ~0.09 mJ, independent of the laser pulse energy and the peak power density. The energy left in the pulse which acts as a surface heating source decreases with the decreasing pulse energy. 6.4 Surface heating source Laser irradiation is regarded as a surface heating source which is released at or near the surface of the work piece when the optical penetration length is much smaller than the thermal diffusion length (𝑙� ≪ 𝑙�), such as the UV laser irradiation of Si or the IR laser irradiation of metals. The IR laser irradiation of Si can also be considered as a surface heating source when Si is melted, as 𝑙� = 14.3 ns for liquid Si compared with 𝑙� = 8.26 µm. The long range atomic order of a crystalline material is lost when heated beyond the melting point, and there are considerable changes in material properties such as absorption coefficient, thermal conductivity and heat capacity (Muller, De Unamuno et al. 1996); (Bauerle, Bermann et al. 2004). The reflectance of molten silicon measured at high temperature is in good agreement with a metallic behavior (Lampert, Koebel et al. 1981). With these metallic absorption properties, the temperature distribution of Si can be calculated using the analytical solution given by (Carslaw and Jaeger 1959), _____________________________________________________________________ 158 𝑇(𝑧, 𝑡) = 𝐼�(1 − 𝑅) �√��� 𝑖𝑒𝑟𝑓𝑐( ��√��). (6.13) When a laser pulse ends, the material will cool for 𝑡 ≥ 𝜏 according to the relationship 𝑇(𝑧, 𝑡) = ���(���) � �√𝐷𝑡 ∗ 𝑖𝑒𝑟𝑓𝑐 � � �√�� � − �𝐷(𝑡 − 𝜏) ∗ 𝑖𝑒𝑟𝑓𝑐 � � ���(���)��. (6.14) When 𝐼(𝑡) is a prescribed function of time, Equations 6.13 and 6.14 become 𝑇(𝑧, 𝑡) = ������������…���� � (1 − 𝑅) �√�� � 𝑖𝑒𝑟𝑓𝑐( � �√�� ) and (6.15) 𝑇(𝑧, 𝑡) = ������������…���� � �(���) � × �√𝐷𝑡 ∗ 𝑖𝑒𝑟𝑓𝑐 � � �√�� � − �𝐷(𝑡 − 𝜏) ∗ 𝑖𝑒𝑟𝑓𝑐 � � ���(���)�� (6.16) The optical and thermal properties of liquid Si used in Equations 6.15 and 6.16 are tabulated in Table 6.5. The thermal conductivity of liquid Si takes the average of its values between the melting and boiling points because of the linear relationship with temperature, Figure 6.2. In addition, no information of material properties is available when temperature is beyond the boiling point, so the properties of liquid Si are always used. _____________________________________________________________________ 159 Table 6.5: Optical and thermal properties of liquid Si used in Equations 6.15 and 6.16 at 1 µm wavelength. Property (Liquid Si) Values Units References Absorption coefficient (𝛼) 699300 cm-1 (Muller, De Unamuno et al. 1996) Optical penetration depth (𝑙�) 14.3×10-7 cm Reflectivity (𝑅) 72.0 % (Jellison, Lowndes et al. 1986); (Muller, De Unamuno et al. 1996) Specific heat (𝐶�) 1.05 J/g·K (Muller, De Unamuno et al. 1996); (Wood and Jellison Jr 1984) Density (𝜌) 2.52 g/cm3 Thermal conductivity (𝜅) 0.766 J/ cm·s·K (Yoo, Jeong et al. 2000); (Muller, De Unamuno et al. 1996) Thermal diffusivity (𝐷) 0.289 cm2/s The calculated temperature distributions for single pulse machining are shown with respect to time and the distance into Si, Figure 6.4-Figure 6.6. The laser pulse used here has an average power of 18 W at 25 kHz repetition rate with a pulse duration of 200 ns and a peak power of 12.77 kW. The percentage of energy used to melt the surface is deducted from the source term, 𝐼(𝑧, 𝑡) = (1 − 𝑅)(1 − 𝐸����)𝐼�𝑒���, (6.17) where 𝐸���� is the energy loss, 13.17% for a 12.77 kW peak power pulse according to Table 6.4. _____________________________________________________________________ 160 (a) (b) Figure 6.4(a): Temperature distribution of Si by a surface heating source with respect to time at varying distances inside the Si substrate. (b): Close view of the surface temperature rise. With a surface heating source, the surface temperature of Si shows a dramatic increase at the beginning of the laser pulse (0-20 ns), Figure 6.4(b), illustrating the impact of the varying silicon thermal dynamic properties with time and temperature under the fast rise high peak pulse. The long energy tail of the fibre laser pulse kept raising the surface temperature to greater than 3.5×104 K till the end of the pulse. The Si surface researched the boiling temperature (3514 K) in ~ 2.7 ns and the thermal critical temperature (5200 K) in ~3.3 ns, Figure 6.5. _____________________________________________________________________ 161 Figure 6.5: Zoomed-in view of Figure 6.4(a), close to the beginning of the laser pulse. Thousands of Kelvin were achieved at and beneath the Si surface, which resulted in a liquid layer. The liquid layer became thicker with time, which was further heated to the thermal critical temperature at hundreds of even a thousand nanoseconds after the beginning of the laser pulse, Figure 6.6. _____________________________________________________________________ 162 Figure 6.6: Temperature distribution of Si by a surface heating source as a function of distance inside the Si substrate at different elapsed times. 6.5 Explosive boiling The theory of explosive boiling may be considered from either a thermodynamic or kinetic viewpoint (Lu, Mao et al. 2002). According to the thermodynamic theory, when the temperature of a material exceeds a limitation of about 0.8Ttc, Ttc = 5200 K is the thermodynamic critical temperature of silicon (Lu, Mao et al. 2008), the liquid is heated rapidly (≤ ~1000 ns) and the system becomes metastable (Miotello and Kelly 1995). When the target material reaches ~0.9Ttc, the liquid experiences large density fluctuations and homogeneous bubble nucleation occurs. The superheated liquid at this point makes a rapid transformation into a mixture of liquid droplets and vapor which can facilitate explosive boiling (Miotello and Kelly 1995). _____________________________________________________________________ 163 The kinetic mechanism models the vapor bubble formation and growth rate at a given temperature. These bubbles only grow spontaneously if they reach a critical radius 𝑟�, they are likely to collapse when their radii are less than 𝑟�, and it takes the bubble a time of 𝜏� to reach 𝑟�. The values of 𝑟� and 𝜏� were calculated to be 0.6 µm and 70 ns with the approximation of the superheated liquid pressure using recoil pressure and a power-law relation of surface tension (Lu, Mao et al. 2002). The IR fibre laser pulse is 200 ns long; therefore the bubble has enough time reach the critical radius during the laser pulse. The thermal penetration depth of liquid Si is calculated to be 4.3 µm (over 200 ns), the thermal diffusivity, D, of liquid Si is 0.228 cm2/s (Yamamoto, Abe et al. 1991). The critical diameter of the bubble is then 𝑑� = 2𝑟� = 1.2 µm, which is smaller than the thermal penetration depth and the bubble can grow to its critical diameter during the laser pulse. (Martynyuk 1974) argued that the rate of homogeneous nucleation, 𝐼� , is numerically significant (i.e., 𝐼� ≥ 1 ) only near 𝑇�� . For example, the value of 𝐼� equals 1 nucleus/cm3 s at 𝑇 = 0.874𝑇�� and equals 1026 nuclei/cm3 s at 𝑇 = 0.905𝑇�� . The number of homogeneous nuclei which would be generated during the laser pulse is 𝐼�𝑉 ∗𝜏 , where 𝑉∗ is the heated volume during the laser pulse ( 𝑉∗ = 𝑙�𝜋 ��� ≈ 4.5 × 10�� cm3, 𝑑 =36.45 µm is the minimum spot size of the laser pulse). If 𝐼� is taken to be 1026 nuclei/cm3 s, 𝑙� is 4.3 µm and 𝜏 is 200 ns, the homogeneous nuclei generated during the laser pulse will be 9 × 10��. During the laser pulse, a huge number of bubbles are generated near the target surface and they have enough time to grow to a critical size, therefore, explosive boiling will occur under these conditions. 6.5.1 Constant pulse duration The thermal conduction solutions, Equations 6.15-6.16, were used to evaluate the depth to which the temperature was raised beyond 0.905𝑇�� (4706 K) for at least 70 ns. The laser parameters used are summarized in Table 6.4, and represented again in Figure 6.7. _____________________________________________________________________ 164 Figure 6.7: Pulse shapes at 200 ns pulse duration with varying pulse energy/peak power. The laser pulses have a fixed pulse duration (200 ns), and varying pulse peak power densities up to 1.22 GW/cm2, Figure 6.7. The temperature distribution of the 1.22 GW/cm2 pulse was re-plotted with respect to time, Figure 6.13(a). The green curve stayed above 4706 K for more than 70 ns, Figure 6.13(b), and that value (8.4 µm) is estimated to be the machined depth resulted from the explosive boiling. _____________________________________________________________________ 165 Figure 6.8(a): Temperature distribution of the 1.22 GW/cm2 laser pulse, Figure 6.7, irradiation of Si as a function of time at varying distances beneath the Si surface. (b): Determination of the depth which is above 0.905𝑇�� (4706 K) for at least 70 ns. The depth values of other pulses in Figure 6.7 were predicted in the same way as in Figure 6.13 and results plotted in comparison with the measured values, Figure 6.14. An energy loss of ~0.09 mJ was taken as a percentage from the source term for all pulses, Table 6.4. This energy was used to melt the Si surface so the optical and thermal properties of liquid Si were used and it was independent of other laser parameters as shown in Table 6.4. >70 ns _____________________________________________________________________ 166 Figure 6.9: The comparison of theoretical predictions based on the explosive boiling and measured depths. With the varying peak power densities, the material removal mechanism kept changing. The predicted results based on the explosive boiling was not in good agreement with the measured values, Figure 6.9, it suggests that the explosive boiling only contributes partly to the machining mechanism. At 1.22 GW/cm2, the predicted depth was higher than the measured value, the excess energy which caused the difference in depth could have been absorbed by the dense plasma. The high speed images of plasma expansion are shown later in Section 6.6, but the plasma absorption is not considered in the thermal model as the peak power density of the fibre laser is limited to 1.22 GW/cm2. At 0.63 GW/cm2, the plasma absorption was thought to be less effective, and the predicted value was close to the measured value. Between 0.63 and 0.16 GW/cm2, the vaporization started to become the main machining mechanism with part of the material removed by explosive boiling. At power densities lower than 0.16 GW/cm2, the explosive _____________________________________________________________________ 167 boiling no longer existed, the machining mechanism was dominated by melt expulsion and vaporization. 6.5.2 Evidence of explosive boiling – Shadowgraph Time sequenced images of ejected mass at selected laser irradiances were measured using laser shadowgraphy, covering a region of ~150 µm in front of the target surface. Each image was taken on a fresh piece of Si wafer and the bottom edge of the images is the Si surface. Images of the ejected mass were obtained at 0.26 GW/cm2 and 1.22 GW/cm2 (200 ns pulse duration), the time of the frame was indicated below each image, Figure 6.10 and Figure 6.11. The white spot at the centre of the lower part of the image represents the interaction zone; the green bandpass filter was removed to allow more information to be captured by the camera. 5 ns 50 ns 300 ns 1.67 µs 2.47 µs 5.67 µs Figure 6.10: Sequence of mass ejection images obtained by laser shadowgraphy for the laser irradiance of 0.26 GW/cm2. Figure 6.10 shows a sequence of violent mass ejection within and after the laser pulse. The speed of particles leaving the surface was approximately 500 m/s within the 200 ns laser pulse. It then slowed down after the laser pulse to ~200 m/s at 300 ns, ~50 m/s at _____________________________________________________________________ 168 1.67 µs and ~10 m/s at 2.47 µs and later. The mass ejection continued after the laser pulse and lasted until 6-8 µs after the laser pulse. The size of particles increased from sub-micron vapors to micron-sized liquid droplet. A precise determination of the ejected particulate size is difficult due to the presence of diffraction rings around the particles in the images. Larger particles have fewer diffraction rings and a sharper boundary; the size of the largest particle is estimated to be approximately 10 µm. Figure 6.11 shows similar behavior as shown in Figure 6.10, except that the mass ejection lasted slightly longer until about 10 µs after the laser pulse. With the same pulse duration, the higher laser irradiance means more pulse energy was available to heat the Si surface to a higher temperature, Figure 6.3. The heat conducted further inside Si and the temperature reached the critical temperature and stayed above it long enough to induce explosive boiling, which resulted in deeper features and a longer mass ejection process. 5 ns 100 ns 200 ns 1.5 µs 5.5 µs 9.5 µs Figure 6.11: Sequence of mass ejection images obtained by laser shadowgraphy for the laser irradiance of 1.22 GW/cm2. Figure 6.10 and Figure 6.11 do not show the formation of shock wave, and so the point explosion model cannot be used. Previous literature suggests that shockwaves are normally detected with the UV laser ablation of Si at pulse widths of less than 10 ns. The _____________________________________________________________________ 169 long pulse duration of machining laser used here may prevent the generation of the shock wave, also the long pulse width of the imaging laser made it difficult to detect the shock wave. Still a mixture of vapors and liquid droplets were observed during and after the laser pulse, which is evidence of explosive boiling. In addition a smooth sidewall and bottom surface of the hole were left after the fibre laser machining, they are different from the observations in the case of volume boiling or subsurface heating (Craciun, Craciun et al. 1998); (Craciun and Craciun 1999). 6.5.3 Constant peak power density The laser pulses of constant peak power density of ~0.63 GW/cm2 and varying pulse durations from 40 ns to 200 ns, Figure 6.12, were also tested with the thermal conduction model for the explosive boiling. Figure 6.12: Pulse shapes of the same pulse peak power with pulse duration between 40 ns and 200 ns. _____________________________________________________________________ 170 The machined depth values of laser pulses in Figure 6.12 were predicted in the same way as in Figure 6.8 and results plotted in comparison with the measured values, Figure 6.13. An energy of 0.092 mJ which was used to melt the Si surface, was deducted as a percentage from the source term of all pulses, Table 6.6. Table 6.6: Laser parameters and percentage energy used to melt Si. Pulse duration (ns) Peak power (kW) Pulse energy (mJ) Energy used to melt Si (mJ) Percentage of the energy used to melt Si as in pulse energy 40 6.53 0.189 0.092 48.677% 60 6.53 0.264 0.092 34.848% 80 6.56 0.319 0.092 28.840% 100 6.55 0.359 0.092 25.627% 120 6.55 0.400 0.092 23.000% 140 6.55 0.431 0.092 21.346% 160 6.54 0.469 0.092 19.616% 180 6.52 0.511 0.092 18.004% 200 6.56 0.540 0.092 17.037% Figure 6.13: The comparison of theoretical predictions and measured depths. _____________________________________________________________________ 171 The theoretical predictions are very close to the actual measured values, demonstrating that the explosive boiling is the dominant material removal mechanism at this peak power density. The zero value at 40 ns resulted from the fact that the temperature did not stay above 0.905𝑇�� (4706 K) for more than 70 ns, Figure 6.14, and there was no explosive boiling with a low pulse energy. The image at 40 ns in Figure 6.13 showed very little mixture of vapors and droplets, instead it has large amount of micro-sized debris re- solidified in the radial direction, suggesting a melt expulsion by the recoil pressure. Figure 6.14: Temperature distribution of 40 ns laser pulse. The energy loss due to the laser generated vapor plasma and surface vaporization were not included. Nevertheless, these losses play an important role in determining the laser energy available for heat conduction at high power densities. The predicted machined depth due to the explosive boiling was higher when less energy losses were considered, an example shown in Figure 6.9 at 1.22 GW/cm2. The explosive boiling observed in this study differs from that of the short nanosecond (3 ns) UV (266 nm) laser (Lu, Mao et al. 2008). In case of the UV laser, the thermal <70 ns _____________________________________________________________________ 172 penetration depth is too short for the bubbles to grow to its critical radius and the homogeneous nuclei generated during the laser pulse are too few to cause significant changes. Instead, the superheated liquid layer will propagate into the target with thermal diffusion after the laser pulse. During this process new bubbles will appear and will grow spontaneously when the thickness of liquid layer becomes greater than the critical diameter. The superheated layer will eventually leave the target in the form of mixed vapors and droplets (Lu, Mao et al. 2008). The machined depth is likely to be determined by the temperature within the superheated layer. The increasing laser irradiance resulted in the increase of the superheated layer temperature, and was likely to cause the further diffusion to the bulk Si. In case of the fibre laser, the pulse duration was long enough to cause sufficient heat diffusion depth for the bubble to grow within the laser pulse, and the explosive boiling can started before the end of laser pulse. With the same peak power density, Table 6.6, the longer laser pulse kept heating Si for a longer time. As a result, the temperature further inside Si reached the critical temperature and stayed above it long enough to induce explosive boiling, which resulted in deeper features. The higher the laser irradiance, the higher the plasma density, and the more laser energy is absorbed in the plasma (Lu, Mao et al. 2008); (Liu, Mao et al. 1999). Russo’s work used a UV laser with fixed pulse duration (3 ns FWHM) but increasing power density (1 − 100 GW/cm2) (Yoo, Jeong et al. 2000); (Yoo, Jeong et al. 2000); (Liu, Mao et al. 2007). The fibre laser in this study has low peak power density (<1.22 GW/cm2) but long pulse duration (<200 ns). With the same pulse energy, the fibre laser has lower laser irradiance, lower plasma absorption and more energy left for heat conduction. In addition, the fibre laser has faster rise time than a general Gaussian temporal pulse shape, and Si can be melted quickly with only 0.09 mJ of pulse energy. The rest of pulse energy can act on the melted Si, which is similar to a UV laser Si interaction. _____________________________________________________________________ 173 6.5.4 UV laser The UV (355 nm) laser pulses of varying peak power densities (< 2.82 GW/cm2) with pulse width < 35 ns FWHM, Figure 6.15 & Table 6.7, were tested with the thermal conduction model for explosive boiling. Figure 6.15: Temporal pulse profile of the DPSS UV laser at 60 kHz with varying peak power densities. Table 6.7: UV Laser parameters and the percentage energy used to melt the Si surface. Average power (W) Pulse duration (ns) Peak power (kW) Pulse energy (mJ) Energy used to melt Si (mJ) Percentage of the energy used to melt Si as in pulse energy (%) 9.85 51.5 6.66 0.164 0.000210 0.1285% 8.20 57.2 5.19 0.136 0.000185 0.1363% 5.4 59.2 3.32 0.090 0.000273 0.3038% 4.32 60.2 2.76 0.072 0.000320 0.4451% 3.31 53.4 2.22 0.055 0.000321 0.5841% 2.21 46 1.60 0.037 0.000308 0.8325% 1.56 47 1.15 0.026 0.000348 1.3375% 1.06 44.8 0.79 0.018 0.000366 2.0312% 0.63 45.8 0.45 0.011 0.000404 3.6740% _____________________________________________________________________ 174 Under the high UV laser irradiation of 2.82 GW/cm2, the Si surface was melted in only 2 ns, Figure 6.16, calculated using the surface heating model. Therefore the reflectivity, specific heat and density of liquid Si were used in the subsequent calculations. The thermal conductivity and diffusivity were taken as the average of their values between the melting temperature and thermal critical temperature, Table 6.8. Figure 6.16: Temperature profile of Si under irradiation of the UV laser at 2.82 GW/cm2. Table 6.8: Optical and thermal properties of Si used in Equations 6.15 & 6.16 at 355 nm wavelength. Property Values Units References Absorption coefficient (𝛼) 106 cm-1 (Bauerle 1996); (Palik 1991) Optical penetration depth (𝑙�) 10×10-7 cm Reflectivity (𝑅) @ 355 nm 70 % (Jellison, Lowndes et al. 1986) Specific heat (𝐶�) 1.05 J/g·K (Muller, De Unamuno et al. 1996); (Wood and Jellison Jr 1984) Density (𝜌) 2.52 g/cm3 Thermal conductivity (𝜅) 1.0104 J/ cm·s·K (Yoo, Jeong et al. 2000) Thermal diffusivity (𝐷) 0.382 cm2/s Thermal diffusion length of liquid Si is calculated to be 1.79 µm based on the pulse width of less than 35 ns FWHM, it is greater than the diameter of the bubbles (1.2 µm), but during the laser pulse, the bubbles do not have sufficient time (70 ns) to grow to their _____________________________________________________________________ 175 critical radius. Therefore, the explosive boiling is unlikely to occur during the laser pulse. A superheated liquid region is likely to be formed near the surface, which the high temperature would eventually penetrate into the target by conduction. The bubbles then have time to reach the critical radius after the laser pulse. The depth resulted from the explosive boiling using laser pulses in Figure 6.15 were predicted in the same way as in Figure 6.8 and results plotted in comparison with the measured values, Figure 6.17. An fraction of the pulse energy which was used to melt the Si surface, was deducted as a percentage from the source term of all pulses, Table 6.7. Figure 6.17: The comparison between the theoretical predictions based on the explosive boiling and the measured depth of the UV laser machined holes on Si. The predicted data show a great difference from the measured data when the explosive boiling alone was considered as the material removal mechanism. The measured depth was caused by a mixture of removal mechanisms; melt expulsion, vaporization, explosive _____________________________________________________________________ 176 boiling and plasma effects. At laser irradiances higher than ~1 GW/cm2, plasma was likely to be generated and its density and temperature were growing with the laser irradiance (Russo 1995). A lateral pressure difference between the plasma above the sample surface and the surrounding ambient may result in the motion of the molten Si, and cause the liquid splashes from the interaction zone, Figure 6.17. In addition, phenomena like plasma screening and scattering were preventing part of the laser energy from reaching the substrate. The predicted depth of the explosive boiling did not take into account the plasma effects, therefore they are greater than the measured value. Especially the one at 2.82 GW/cm2, the predicted depth is more than twice as much as the measured value. At laser irradiances between 0.5 and 1 GW/cm2, the vaporization and melt expulsion were likely to be the dominant machining mechanisms, only part of the material was removed by the explosive boiling. At laser irradiances lower than 0.5 GW/cm2, the thermal critical temperature was not maintained for longer than 70 ns and that no phase explosion would occur, the material was totally removed by the melt expulsion and vaporization. 6.6 Plasma plume expansion The 200 ns fibre laser pulse rises rapidly to the peak value of 12.77 kW in about 10 ns, and falls off slowly, stops at 200 ns. During the laser pulse, 15 high speed imaging frames were captured starting 10 ns after the beginning of the laser pulse, Figure 6.18. Each frame lasts only 10 ns long and there is a 5 ns delay between the two adjacent frames. _____________________________________________________________________ 177 Figure 6.18: The overlaps of the ultrafast camera shutter signal and the fibre laser pulse. During the process of fibre laser percussion drilling, plasma was initiated after 10 ns of laser irradiation, and it grew and expanded with time, Figure 6.19. The plasma was self illuminant as it released photons in the visible wavelength. The high density Si atoms and ions travelled several hundred micrometers from the sample surface at the speed of ~2 km/s. They collided with each other as they expanded into the atmosphere. The ionization potential of Si is about 8.3 eV (Hildenbrand 1969); (Fuke, Tsukamoto et al. 1993), the photon energy at 1 µm wavelength of the IR fibre laser is 1.17 eV; at least eight photons are needed for multiphoton ionization of a single Si atom. The chances of Si atoms getting ionized are little with a low photon concentration, i.e. low laser power density, hence the plasma brightness and density was low at the later part of the pulse. During the nanosecond laser pulse, the gas environment influence is not significant, as collisions of gas species with the ejected atoms and ions are negligible (Russo 1995). _____________________________________________________________________ 178 10 ns 25 ns 40 ns 55 ns 70 ns 85 ns 100 ns 115 ns 130 ns 145 ns 160 ns 175 ns 190 ns 205 ns 220 ns Figure 6.19: Plasma evolution of a 200 ns fibre laser pulse (1.22 GW/cm2) on Si sample. _____________________________________________________________________ 179 The initiation and evolution inside the low threshold breakdown plasma of a laser supported detonation wave caused the detachment of the laser radiation absorption front from the target surface and the subsequent decrease of the energy coupling coefficient (Ursu, Apostol et al. 1985). This so called plasma shielding happens when the laser induced surface plasma becomes optically dense at high power density (0.2 – 0.3 GW/cm2), the amount of laser energy coupled to the target then reaches a plateau. The density and temperature of the plasma increase with laser power density (Russo 1995). A number of researchers measured the temporal profiles and energy of the reflected or transmitted pulse, and found that the laser pulse duration was truncated at higher power densities (Tokarev, Marine et al. 1994); (Wolff-Rottke, Ihlemann et al. 1995); (Ihlemann, Scholl et al. 1995); (Mao and Russo 1996). The roll-off power density for the nanosecond laser was found to be almost independent of the target materials, whose melting temperature ranges from 533 K to 3273 K (Shannon, Mao et al. 1995). At 0.2 – 0.3 GW/cm2, the amount of mass removed directly by the laser ceased to increase with increasing laser energy, which was found during the IR fibre laser machining of Si, Figure 6.20. The heat exchange between the laser induced plasma and sample surface, however, can further vaporize material at a slower rate. _____________________________________________________________________ 180 Figure 6.20: The fibre laser single pulse machined depth with respect to the peak power density. 6.7 Summary A one-dimensional thermal conduction model was developed to examine the effects of the temperature dependent material properties and the fibre laser’s asymmetrical pulse shape. Based on the volume heating source, which is typical for 1 µm laser machining of Si, the temperature of Si surface was predicted to reach the melting point with a pulse energy of 0.09 mJ. The value is independent of the laser parameters. With a higher peak power density, the melting phase was achieved earlier, Figure 6.3. Once Si is melted, its thermodynamic properties change dramatically, and the 1 µm laser source was considered as a surface heating source. With the adapted analytical solutions, the temperature distribution as a function of time and distance inside Si was obtained. It was shown that the temperature inside Si was able to stay above 0.905𝑇�� for longer than _____________________________________________________________________ 181 70 ns. Together with the sufficient liquid thickness, the explosive boiling was thought to occur and becomes the dominant material removal mechanism. The estimated hole depths due to the explosive boiling alone were different from the measured ones at varying peak power densities (<1.22 GW/cm2) but fixed pulse duration (200 ns), since Si was removed by a mixture of mechanisms. Nevertheless a close agreement was found (6% difference on average) with varying pulse durations (40-200 ns) but fixed peak power density (~0.63 GW/cm2) between the estimated data and measured ones. The fibre laser’s long duration pulse and the DPSS UV laser’s high irradiance pulse both resulted in the explosive boiling, but the long pulse duration is more effective. Firstly, the long pulse duration loses less energy through plasma due to the limited peak power density (Russo 1995). Secondly, the long pulse duration increases the thermal diffusion length, and the bubbles can grow to the critical diameter during the laser pulse. The explosive boiling was proved by the shadowgraph and SEM images. The shadowgraph images showed mass ejection during and after the laser pulse and the SEM images revealed debris in the form of a mixture of vapors and liquid droplets. A dense plasma was also observed during the laser processing of Si with a 1.22 GW/cm2 pulse, which a number of researcher proved that a part of laser energy was lost through the plasma. With less energy available after the energy losses due to the dense plasma, surface reflection and surface melting, there is still enough energy to induce explosive boiling with the IR fibre laser on Si. _____________________________________________________________________ 182 Chapter 7 Conclusion The aim of this thesis is to characterize fibre laser machining of Si substrates with a view to identifying the pulse shape influences and optimum set of laser parameters for high quality, high efficiency micro drilling and surface texturing applications. Concurrently to develop a greater understanding of the laser material interactions and material removal mechanism, in light of the unique pulse shapes offered by the fibre laser. The IR fibre laser system developed in this work was able to perform percussion drilling and single pulse machining on the Si <111> surface over a wide range of processing options. With the optimum parameters, the IR laser generated micro-sized holes on Si with a well defined edge, no heavy recast, and no micro-cracks, which has not been reported before at a wavelength of 1 µm. The temporal laser pulse shape was then tuned to gain an understanding of the pulse parameter effects; Si was machined with fixed pulse durations, fixed peak powers and fixed pulse energies separately, Chapter 4. In Chapter 5, the comparison between the IR fibre laser and the state of the art DPSS UV laser was investigated. In Chapter 6, a thermal conduction model was used to examine the explosive boiling theory behind the proposal. Some key findings from each chapter are given in the sections below. 7.1 Fibre laser machining characterization of Si • The peak power and pulse duration (temporal pulse shape) affect machining processes the most during the 1 µm laser irradiation of Si. The pulse energy determines the amplitude of peak power at constant pulse duration and vice versa. • Peak power was found to strongly influence the material removal mechanism with a fixed pulse duration. High peak powers (>1 kW) gave vaporization dominated _____________________________________________________________________ 183 ablation, left a limited resolidified molten layer and delivered clean hole formation. • Pulse duration was found to strongly influence the machined depth. Longer pulse durations generated deeper holes with constant peak power (>1 kW). • With the fixed pulse energy, it is more energy efficient to use longer pulses when the peak power is maintained above the vaporization threshold. • For quality purpose, the optimum setting is to use the highest possible pulse peak power at pulse durations greater than 50 ns. • With pulse durations less than 50 ns, the pulse profiles lose the energy tail and the machined results show considerable amount of resolidified melt splashes. • For energy efficiency purpose, the highest efficiencies for varying pulse durations (50-200 ns) occur at the transition point between melt expulsion and vaporization dominated mechanisms. The recoil pressure at this point was high enough to cause effective melt expulsion, while the vapour density at this point was low enough not to affect the incoming laser beam. • The maximum energy efficiency of 31.86 µm per mJ was achieved with a laser pulse of 0.14 mJ energy (13.3 J/cm2 fluence), 200 ns pulse duration and 0.97 kW peak power. 7.2 Performance comparison between IR and UV lasers • With lower pulse energies (<0.04 mJ), the UV laser achieved higher machined depth, energy efficiency and lower machining threshold due to the high absorption coefficient (106 cm-1) on solid Si, compared with 50 cm-1 of IR wavelength. • As pulse energy increases, the fibre laser started to melt Si earlier in the pulse, and more energy was left to machine liquid Si, which has absorption coefficient of 7×105 cm-1 at IR. _____________________________________________________________________ 184 • With pulse energies of approaching 0.17 mJ, the maximum machined depth of the fibre laser was 2.3 µm higher than the DPSS laser (6.8 µm/pulse compared to 4.5 µm/pulse). • The UV laser machined holes have a large amount of resolidified melt beyond the hole rim, while the fibre laser machined holes leave a mixture of vapour condensates and limited re-solidified droplets, demonstrating the effective drilling of Si using these fibre laser sources at longer pulse duration. • The M2 = 1.2 fibre laser was also compared with the M2 = 2 fibre laser with a fixed 200 ns pulse duration. The better focusibility and smaller spot size allowed efficient material removal to happen with a lower pulse energy; a 5 µm hole was machined with less than 0.1 mJ pulse energy for the M2 = 1.2 laser while twice the energy was required for the M2 = 2 laser. 7.3 Theoretical analysis • The thermal conduction model was developed to include the temperature dependent material properties and the fibre laser’s asymmetrical pulse shape. • Based on the volume heating source, the temperature of Si surface was predicted to reach the melting point with a pulse energy of 0.09 mJ. The value is independent of the laser parameters. With higher peak power density, the melting phase was achieved earlier. • Based on the surface heating source, once Si is melted, and the adapted analytical solution, the temperature distribution showed that the temperature inside the Si was able to stay above 0.905𝑇�� for longer than 70 ns. Together with sufficient liquid thickness, explosive boiling is thought to occur and become the dominant material removal mechanism. • The estimated hole depth due to the explosive boiling alone were different from the measured data at varying peak power densities (<1.22 GW/cm2) but fixed pulse duration (200 ns), since Si was removed by a mixture of mechanisms. Nevertheless a close agreement was found (6% difference on average) with _____________________________________________________________________ 185 varying pulse durations (40-200 ns) but fixed peak power density (~0.63 GW/cm2). • The longer pulse duration is more effective in generating explosive boiling than the higher laser irradiance. Firstly, the longer pulse duration will lose less energy through plasma due to the fixed peak power density. Secondly, the longer pulse duration increases the thermal diffusion length, and the bubble can grow to its critical diameter during the laser pulse. • The explosive boiling was proved by the shadowgraph images, which showed mass ejection during and after the laser pulse, and the debris in the SEM images, which was in the form of a mixture of vapors and liquid droplets. • A dense plasma was also observed during the laser processing of Si with a 1.22 GW/cm2 pulse, which a number of researcher proved that a part of laser energy was lost through the plasma. With less energy available after the energy losses due to surface reflection, surface melting and plasma absorption, there is still enough energy to induce explosive boiling with the IR fibre laser. Overall it has been shown that the IR fibre laser machining of Si is an extremely complex process that, as expected, is a function of both laser parameters and material thermodynamic properties. Laser parameters have significant effects on how energy is delivered, whereas material properties have more effects on the energy conduction and temperature rise which in turn changes material properties. The dramatic increase in hole quality with this 1 µm MOPA based fibre has not been reported before, it redefines the processing options for Si, and redefines the landscape for large volume Si fabrication at low cost. With a fast rising pulse shape, a high peak front and an energetic long tail, Si is quickly melted and the rest of pulse energy was able to remove liquid Si effectively, leaving limited resolidified melt droplets and vapor condensates. The explanation of the melt expulsion and vaporization mechanism, and energy cost of the melt and vapour in terms of the machining efficiency have not been discussed for _____________________________________________________________________ 186 these laser parameters before. Melt expulsion dominated machining mechanism with little vaporization resulted in an energy efficient machining regime. With the pulse shaping capability, the low energy pulse can be used to do the same job as the high energy pulse. As a result, higher repetition rates can be applied in the production line, and the production speed is increased to satisfy the needs of modern semiconductor and solar cell production. In addition, the performance comparison of Si machining with 355 nm DPSS laser of Gaussian pulse shapes with 1064 nm fibre laser of non-Gaussian pulse shapes is insightful. In the high pulse energy region, the IR fibre laser machined deeper features with better surface finish than the UV laser. The debris left by the IR laser was in the form of thin streaks of resolidfied melt splashes and vapor particles, compared with thick resolidified melt splashes by the UV laser. The observation demonstrated that with the optimum pulse profile, the wavelength factor can be minimized. The comparison also suggests that some applications (microvia drilling) of the DPSS UV laser can be replaced with the more flexible, low cost IR fibre laser. The proposed explosive boiling material removal mechanism was confirmed by the thermal conduction model and shadowgraph images. The predicted results were in close agreement with the measured ones. The pulse shape of long pulse duration and moderate peak power was suggested to lose less energy through plasma screening effects. Therefore with the limited pulse energy of the fibre laser, the explosive material removal can still be induced which produced increased machining depth and a micron- /submicron-sized debris field. The majority of findings in this PhD come from direct experimental observations, the explanations from existing literature have been used to interpret them. The empirical findings are used as the basis for a numerical investigation of the temperature distributions. Key areas of study for experimentally improving the process are identified, some suggestions for which are given below. _____________________________________________________________________ 187 7.4 Suggestion for future work The focus of further work should be on characterizing more temporal pulse profiles. The pulse profiles in this study were limited due to the design of the laser; the pulse duration was limited to 20-200 ns and the pulse peak power was limited to 14 kW. The pulse peak power and pulse energy decrease when exploring shorter pulses and the shorter pulses are only available with high repetitions rates (>125 kHz). The high repetitions rates could not machine single pulse with the current scanning system, because the scanning speed is limited to 5 m/s and it could not separate the two neighbouring pulses. The single shot with the current system cannot produce full power until the fifth pulse. Using a more powerful laser with more flexibility would enable greater laser parameter range to be tuned independently, for example, keeping high peak power while reducing pulse duration. A threshold value of pulse duration can then be found with the minimum interference with the plasma, so more pulse energy is used to machine Si. The use of an ultrafast imaging laser with a more stable and reliable triggering method would improve the shadowgraph images. The high speed particles would show less streaking, the smaller particles would be better captured and observed, and timing and sequence of each event would be more accurate. It can also reveal more information in the interaction zone, such as shock wave with high peak power pulses or plasma thickness. _____________________________________________________________________ 188 Appendix A Solar cell market trend, production volume, and module technologies Figure A. 1 represents a significant proportion of worldwide installed PV capacity (the utility electricity network refers to as the grid). There was strong growth between 2008 and 2009 despite difficult economic conditions, with Germany and Italy in dominant position. Figure A. 1: Cumulative installed grid-connected and off-grid PV power in the reporting countries (IEA web, 2010) _____________________________________________________________________ 189 Figure A. 2 reveals the PV cell production in selected countries between 2007 and 2009, China, Germany, Japan and Taiwan were among the biggest producers. The global PV cell production volumes exceeded 11.5GW in 2009. Figure A. 2: PV cell production in selected countries between 2007 and 2009 IEA web (2010) _____________________________________________________________________ 190 Figure A. 3 is the current performance and price of different PV module technologies, percentage share of 2008 market. Crystalline silicon modules represent 85-90% of the global annual market, which is subdivided into single crystalline and multi-crystalline. Figure A. 3: Current performance and price of different PV module technologies, percentage share of 2008 market (Technology Roadmap, Solar photovoltaic energy, IEA- PVPS web) _____________________________________________________________________ 191 Appendix B List of basic laser parameters Wavelength (µm) The wavelength of a light wave is the spatial period of the wave; the distance between subsequent maxima or minima. Pulse energy (mJ) Average power and frequencies of a pulsed laser determines how much pulse energy is available to machine a sample material; 𝑝𝑢𝑙𝑠𝑒 𝑒𝑛𝑒𝑟𝑔𝑦 = ������� ����� ����� ���������� ����. Laser repetition rate (Hz) Number of laser pulses per second, as determined by the laser specifications. Laser pulse width (ns) Most laser pulses are Gaussian shaped or pseudo-Gaussian shaped, which the pulse width is measured at full width half maximum, and peak power is calculated as pulse energy divided by pulse width. Fluence (J/cm2) Fluence equals energy of the laser beam divided by the exposed area. Even if a Gaussian beam is used for this work, uniform fluence over the beam area was assumed (Desbiens and Masson 2007). Spot size (µm) Exposed area of the Gaussian laser beam; dependent on the fluence at the sample surface. For a fluence lower than the threshold fluence (Fth) of a material, the ablated diameter can be less than the exposed area and for a fluence higher than Fth, it can be larger (Desbiens _____________________________________________________________________ 192 and Masson 2007). A constant exposed spot size is assumed for fluence calculation; minimum spot size. Laser beam intensity (W/cm2) Laser beam intensity is the optical power per unit area. High laser beam intensity can be achieved with high pulse energy, short pulse width, small spot size, and properly and efficiently delivered beam to the ablation region. The average laser intensity of a CW laser is given by the average power divided by the focal spot area. For a pulsed laser, the average power is replaced by the pulse peak power. _____________________________________________________________________ 193 Appendix C Si properties Figure C. 1: Properties of silicon at 300K (ICKnowledge 2003) _____________________________________________________________________ 194 References Allmen, M. v. (1976). "Laser drilling velocity in metals." Journal of Applied Physics 47(12): 5460-5463. Allmen, M. v. and A. Blatter (1995). Laser-Beam Interactions with Materials: Physical Principles and Applications, Springer. Anisimov, S. I. (1968). "Vaporization of metal absorbing laser radiation." SOVIET PHYSICS JETP-USSR 27(1): 182-&. Anisimov, S. I., B. L. Kapeliovich, et al. (1974). "Electron emission from metal surfaces exposed to ultrashort laser pulses " Soviet Physics - JETP 39: 375-377. Aydinli, A., H. W. Lo, et al. (1981). "Induced Absorption in Silicon under Intense Laser Excitation: Evidence for a Self-Confined Plasma." Physical Review Letters 46(25): 1640. Ball, Z., T. Feurer, et al. (1996). "Thermal and mechanical coupling between successive pulses in KrF-excimer-laser ablation of polyimide." Applied Physics A Materials Science & Processing 62(3): 203-211 Bärsch, N., K. Körber, et al. (2003). "Ablation and cutting of planar silicon devices using femtosecond laser pulses." Applied Physics A: Materials Science & Processing 77(2): 237-242. Basu, S. (2011). CRYSTALLINE SILICON – PROPERTIES AND USES. Croatia, InTech. Bauerle, D. (1996). Laser Processing and Chemistry. Berlin, Springer. Bauerle, D. (2000). Laser Processing and Chemistry. Berlin, Springer. Bäuerle, D. (2011). Laser Processing and Chemistry. Berlin Heidelberg, Springer. Bauerle, D., H. W. Bermann, et al. (2004). Laser Applications. Berlin, Springer-Verlag. Blinov, L. M., E. A. Bobrova, et al. (1968). Soviet Physics Solid State 9: 2537. Bloomstein, T. M. and D. J. Ehrlich (1992). "Stereo laser micromachining of silicon." Applied Physics Letters 61(6): 708 Bogaerts, A., Z. Chen, et al. (2003). "Laser ablation for analytical sampling: what can we learn from modeling?" Spectrochimica Acta Part B: Atomic Spectroscopy 58(11): 1867-1893 _____________________________________________________________________ 195 Boley, C. D. and J. T. Early (1994). Computational model of drilling with high radiance pulsed lasers. Proceedings of the Laser Material Processing Conference, ICALEO 94, Orlando, FL. Boyd, I. W. and J. I. B. Wilson (1983). "Laser processing of silicon." Nature 303(5917): 481-486. Brailovsky, A. B., S. V. Gaponov, et al. (1995). "Mechanisms of melt droplets and solid- particle ejection from a target surface by pulsed laser action." Applied Physics A: Materials Science & Processing 61(1): 81-86. Brannon, J. H., J. R. Lankard, et al. (1985). "Excimer laser etching of polyimide." Journal of Applied Physics 58(5): 2036-2043 Breitling, D., A. Ruf, et al. (2004). Fundamental aspects in machining of metals with short and ultrashort laser pulses. Photon Processing in Microelectronics and Photonics III. P. R. Herman, J. Fieret, A. Piqueet al, SPIE. 5339: 49-63. Breitling, D., H. Schittenhelm, et al. (1999). "Shadowgraphic and interferometric investigations on Nd:YAG laser-induced vapor/plasma plumes for different processing wavelengths." Applied Physics A: Materials Science & Processing 69(0): S505-S508. Brygo, F. o., C. Dutouquet, et al. (2006). "Laser fluence, repetition rate and pulse duration effects on paint ablation." Applied Surface Science 252(6): 2131-2138. Bulgakova, N. M., A. V. Bulgakov, et al. (2004). "Energy balance of pulsed laser ablation: thermal model revised." Applied Physics A: Materials Science & Processing 79(4): 1323-1326. Butefisch, S., A. Schoft, et al. (2000). "Three-axes monolithic silicon low-g accelerometer." Microelectromechanical Systems, Journal of 9(4): 551-556. Canning, J. (2006). "Fibre lasers and related technologies." Optics and Lasers in Engineering 44(7): 647-676 Carslaw, H. S. and J. C. Jaeger (1959). Conduction of Heat in Solids. Oxford, Clarendon. Chan, C. L. and J. Mazumder (1987). "One-dimensional steady-state model for damage by vaporization and liquid expulsion due to laser-material interaction." Journal of Applied Physics 62(11): 4579-4586. Chang, J. J. and B. E. Warner (1996). "Laser-plasma interaction during visible-laser ablation of methods." Applied Physics Letters 69(4): 473-475. Charschan, S. S. (1977). Guide for Material Processing by Lasers. Toledo, OH, Laser Institute of America. _____________________________________________________________________ 196 Chen, T.-C. and R. B. Darling (2005). "Parametric studies on pulsed near ultraviolet frequency tripled Nd:YAG laser micromachining of sapphire and silicon." Journal of Materials Processing Technology 169(2): 214-218 Chen, T.-C. and R. B. Darling (2008). "Laser micromachining of the materials using in microfluidics by high precision pulsed near and mid-ultraviolet Nd:YAG lasers." Journal of Materials Processing Technology 198(1-3): 248-253 Cheng, J.-Y., C.-W. Wei, et al. (2004). "Direct-write laser micromachining and universal surface modification of PMMA for device development." Sensors and Actuators B: Chemical 99(1): 186-196 Chichkov, B. N., C. Momma, et al. (1996). "Femtosecond, picosecond and nanosecond laser ablation of solids." Applied Physics A: Materials Science & Processing 63(2): 109-115. Choi, H. and X. Li (2007). "Fabrication and application of micro thin film thermocouples for transient temperature measurement in nanosecond pulsed laser micromachining of nickel." Sensors and Actuators A: Physical 136(1): 118-124. Choo, K. L., Y. Ogawa, et al. (2004). "Micromachining of silicon by short-pulse laser ablation in air and under water." Materials Science and Engineering A 372(1-2): 145-162. Christensen, B. and M. S. Tillack (2003). Survey of mechanisms for liquid droplet ejection from surfaces exposed to rapid pulsed heating. San Diego, Fusion Division, Center for Energy Research, University of California. Clark-MXR Clark-MXR Micromachining Handbook. Clement, F., M. Menkoe, et al. (2010). "High throughput via-metallization technique for multi-crystalline metal wrap through (MWT) silicon solar cells exceeding 16% efficiency." Solar Energy Materials and Solar Cells 94(1): 51-56 Coherent (2007). Coherent AVIA 355-10 Solid-State Q-Switched Ultraviolet Lasers Data Sheet, Coherent, http://www.coherent.com/downloads/AVIA_355- 10_DS_final.PDF. Craciun, V. and D. Craciun (1999). "Evidence for volume boiling during laser ablation of single crystalline targets." Applied Surface Science 138-139: 218-223. Craciun, V., D. Craciun, et al. (1998). "Subsurface boiling during pulsed laser ablation of Ge." Physical Review B 58(11): 6787 Cregan, R. F., B. J. Mangan, et al. (1999). "Single-Mode Photonic Band Gap Guidance of Light in Air." Science 285(5433): 1537 -1539 _____________________________________________________________________ 197 D. Bäuerle, H. W. B., F. Dausinger, K. Dörschel, A. Gebhardt,, M. G. M. Geiger, H. Haferkamp, C. Hertzler, H. Hügel, O. Minet,, et al. (2004). Group VIII: Advanced Materials and Technologies Volume 1 Laser Physics and Applications Subvolume C: Laser Applications. Germany, Springer-Verlag. Dabby, F. and U.-C. Paek (1972). "High-intensity laser-induced vaporization and explosion of solid material." Quantum Electronics, IEEE Journal of 8(2): 106- 111. Dauer, S., A. Ehlert, et al. (1999). "Rapid prototyping of micromechanical devices using a Q-switched Nd:YAG laser with optional frequency doubling." Sensors and Actuators A: Physical 76(1-3): 381-385 Desai, P. D. (1986). "Thermodynamic Properties of Iron and Silicon." Journal of Physical and Chemical Reference Data 15: 967 Desbiens, J.-P. and P. Masson (2007). "ArF excimer laser micromachining of Pyrex, SiC and PZT for rapid prototyping of MEMS components." Sensors and Actuators A: Physical 136(2): 554-563. Duley, W. W. (1996). UV Lasers: Effects and Applications in Materials Science. Cambridge, Cambridge University Press Ehrlich, D. J. and J. Y. Tsao (1989). Laser microfabrication: thin film processes and lithography, Academic Press Franklin, S. R. and R. K. Thareja (2004). "Simplified model to account for dependence of ablation parameters on temperature and phase of the ablated material." Applied Surface Science 222(1-4): 293-306. Fuke, K., K. Tsukamoto, et al. (1993). "Photoionization of small silicon clusters: ionization potentials for Si2 to Si40." Zeitschrift Für Physik D Atoms, Molecules and Clusters 26(S1): 204-206 Gapontsev, V. P., N. S. Platonov, et al. (2003). 400 W low-noise single-mode CW ytterbium fiber laser with an integrated fiber delivery. Lasers and Electro-Optics, 2003. CLEO '03. Conference on: 3 pp. Gee, J. M., W. K. Schubert, et al. (1993). Emitter wrap-through solar cell. Conference Record of the Twenty Third IEEE Photovoltaic Specialists Conference, 1993, IEEE: 265-270 Giesen, A., H. Hügel, et al. (1994). "Scalable concept for diode-pumped high-power solid-state lasers." Applied Physics B: Lasers and Optics 58(5): 365-372 _____________________________________________________________________ 198 Giesen, A. and J. Speiser (2007). "Fifteen Years of Work on Thin-Disk Lasers: Results and Scaling Laws." Selected Topics in Quantum Electronics, IEEE Journal of 13(3): 598-609. Gillner, A., J. Holtkamp, et al. (2005). "Laser applications in microtechnology." Journal of Materials Processing Technology 167(2-3): 494-498 Goldberger, J., A. I. Hochbaum, et al. (2006). "Silicon Vertically Integrated Nanowire Field Effect Transistors." Nano Lett. 6(5): 973-977. Gray, D. E. (1972). American Institute of Physics Handbook. New York, American Institute of Physics. Greuters, J. (2003). UV laser micromachining of silicon, indium phosphide and lithium niobate for telecommunications applications. Proceedings of SPIE. Galway, Ireland: 479-486. Gu, E., H. Howard, et al. (2006). "Microfabrication in free-standing gallium nitride using UV laser micromachining." Applied Surface Science 252(13): 4897-4901 Hamamatsu Photonics (2005). Stealth Dicing Technology and Applications. Harder, N.-P., S. Hermann, et al. (2009). "Laser-processed high-efficiency silicon RISE- EWT solar cells and characterisation." Physica Status Solidi C 6(3): 736-743. Hendow, S. T. and S. A. Shakir (2010). "Structuring materials with nanosecond laser pulses." Opt. Express 18(10): 10188-10199. Hendow, S. T., S. A. Shakir, et al. (2010). MOPA fiber laser with controlled pulse width and peak power for optimizing micromachining applications. Photonics West 2010, Laser Applications in Microelectronic and Optoelectronic Manufacturing XV, San Francisco, California, USA, SPIE. Hentschel, M., R. Kienberger, et al. (2001). "Attosecond metrology." Nature 414(6863): 509-513 Hildenbrand, D. L. (1969). "Dissociation Energy and Ionization Potential of Silicon Monoxide." The Journal of Chemical Physics 51(2): 807 Howe, R. T., B. E. Boser, et al. (1996). "Polysilicon integrated microsystems: technologies and applications." Sensors and Actuators A: Physical 56(1-2): 167- 177. Huang, Y. and C. M. Lieber (2004). "Integrated nanoscale electronics and optoelectronics: Exploring nanoscale science and technology through semiconductor nanowires." Pure and Applied Chemistry 76(12): 2051-2068. _____________________________________________________________________ 199 Hull, R. (1999). Properties of crystalline silicon. Exeter, IET ICKnowledge (2003). Properties of silicon at 300K, ICKnowledge.com. Ihlemann, J., A. Scholl, et al. (1995). "Nanosecond and femtosecond excimer-laser ablation of oxide ceramics." Applied Physics A Materials Science & Processing 60(4): 411-417. Ihlemann, J., B. Wolff, et al. (1992). "Nanosecond and femtosecond excimer laser ablation of fused silica." Applied Physics A: Materials Science & Processing 54(4): 363-368. Ioffe (2011-5-27). Physical properties of Silicon, Ioffe Institute Database. Ion, J. C. (2005). Laser processing of engineering materials: principles, procedure and industrial application. Oxford, Butterworth-Heinemann Jackson, M. J. and W. O'Neill (2003). "Laser micro-drilling of tool steel using Nd:YAG lasers." Journal of Materials Processing Technology 142(2): 517-525 Jacques, S. L. and G. Gofstein (1991). Laser-flash photographic studies of Er:YAG laser ablation of water. Proceedings of SPIE, Berlin, Germany. Jellinek, H. H. G. and R. Srinivasan (1984). "Theory of etching of polymers by far- ultraviolet high-intensity pulsed laser- and long-term irradiation." J. Phys. Chem. 88(14): 3048-3051. Jellison, G. E. and D. H. Lowndes (1982). "Optical absorption coefficient of silicon at 1.152µ at elevated temperatures." Applied Physics Letters 41(7): 594-596. Jellison, G. E., D. H. Lowndes, et al. (1986). "Time-resolved reflectivity measurements on silicon and germanium using a pulsed excimer KrF laser heating beam." Physical Review B 34(4): 2407. Jellison, G. E. J. and D. H. Lowndes (1987). "Measurements of the optical properties of liquid silicon and germanium using nanosecond time-resolved ellipsometry." Applied Physics Letters 51(5): 352-354. Kaganov, M. I., I. M. Lifshitz, et al. (1957). Soviet Physics - JETP 4: 173. Karlsson, B. r. and C. G. Ribbing (1982). "Optical constants and spectral selectivity of stainless steel and its oxides." Journal of Applied Physics 53(9): 6340-6346 Karnakis, D., G. Rutterford, et al. (2005). High Power DPSS Laser Micromachining of Silicon and Stainless Steel for Device Singulation. Miami Florida, USA. ICALEO 2005. _____________________________________________________________________ 200 Karnakis, D. M. (2006). "High power single-shot laser ablation of silicon with nanosecond 355nm." Applied Surface Science 252(22): 7823-7825 Kautek, W., B. Roas, et al. (1990). "Formation of Y---Ba---Cu---oxide thin films by pulsed laser deposition: A comparative study in the UV, visible and IR range." Thin Solid Films 191(2): 317-334. Körner, C., R. Mayerhofer, et al. (1996). "Physical and material aspects in using visible laser pulses of nanosecond duration for ablation." Applied Physics A: Materials Science & Processing 63(2): 123-131. Lampert, M. O., J. M. Koebel, et al. (1981). "Temperature dependence of the reflectance of solid and liquid silicon." Journal of Applied Physics 52(8): 4975-4976. Lax, M. (1977). "Temperature rise induced by a laser beam." Journal of Applied Physics 48(9): 3919-3924. Lee, D. J. and S. H. Jeong (2004). "Analysis of recoil force during Nd:YAG laser ablation of silicon." Applied Physics A: Materials Science & Processing 79(4): 1341-1344. Lee, Y.-C. and C.-Y. Wu (2007). "Excimer laser micromachining of aspheric microlenses with precise surface profile control and optimal focusing capability." Optics and Lasers in Engineering 45(1): 116-125 Li, J. F., L. Li, et al. (2004). "Comparison of volumetric and surface heating sources in the modeling of laser melting of ceramic materials." International Journal of Heat and Mass Transfer 47(6-7): 1159-1174 Lim, D. J., H. Ki, et al. (2006). "Mass removal modes in the laser ablation of silicon by a Q-switched diode-pumped solid-state laser (DPSSL)." Journal of Physics D: Applied Physics 39(12): 2624-2635 Liu, C. Y., X. L. Mao, et al. (2007). "Time Resolved Shadowgraph Images of Silicon during Laser Ablation: Shockwaves and Particle Generation." Journal of Physics: Conference Series 59: 338-342 Liu, H. C., X. L. Mao, et al. (1999). "Nonlinear changes in plasma and crater properties during laser ablation of Si." Applied Physics Letters 75(9): 1216-1218 Liu, X., D. Du, et al. (1997). "Laser ablation and micromachining with ultrashort laser pulses." Quantum Electronics, IEEE Journal of 33(10): 1706-1716. Lou, Q., J. Zhou, et al. (2003). Wavelength characteristics of Yb3+ -doped double cladding fiber laser. Advances in Fiber Devices, SPIE, San Jose, CA, USA. _____________________________________________________________________ 201 Lowndes, D. H., G. E. Jellison, et al. (1982). "Time-resolved optical studies of silicon during nanosecond pulsed-laser irradiation." Physical Review B 26(12): 6747 Lu, Q., S. S. Mao, et al. (2002). "Delayed phase explosion during high-power nanosecond laser ablation of silicon." Applied Physics Letters 80(17): 3072 -3074. Lu, Q., S. S. Mao, et al. (2008). "Theory analysis of wavelength dependence of laser- induced phase explosion of silicon." Journal of Applied Physics 104(8): 083301. Luft, A., U. Franz, et al. (1996). "A study of thermal and mechanical effects on materials induced by pulsed laser drilling." Applied Physics A: Materials Science & Processing 63(2): 93-101 Lukeš, I., R. šášik, et al. (1992). "Study of excimer laser induced melting and solidification of Si by time-resolved reflectivity measurements." Applied Physics A Solids and Surfaces 54: 327-333 Macfarlane, G. G., T. P. McLean, et al. (1958). "Fine Structure in the Absorption-Edge Spectrum of Si." Physical Review 111(5): 1245 Madou, M. and J. Florkey (2000). "From Batch to Continuous Manufacturing of Microbiomedical Devices." Chem. Rev. 100(7): 2679-2692 Madou, M. J. (2002). Fundamentals of Microfabrication, the Science of Miniaturization, CRC Press. Mao, X. and R. E. Russo (1996). "Observation of plasma shielding by measuring transmitted and reflected laser pulse temporal profiles." Applied Physics A: Materials Science & Processing 64(1): 1-6. Martynyuk, M. M. (1974). "Vaporization and boiling of liquid metal in an exploding wire." Sov Phys Tech Phys 19(6): 793-797. Mazhukin, V. I., I. Smurov, et al. (1995). "Overheated metastable states in pulsed laser action on ceramics." Journal of Applied Physics 78(2): 1259 Meijer, J., K. Du, et al. (2002). "Laser Machining by short and ultrashort pulses, state of the art and new opportunities in the age of the photons." CIRP Annals - Manufacturing Technology 51(2): 531-550 Metev, S. M. and V. P. Veiko (1998). Laser-Assisted Microtechnology. Berlin, Springer- Verlag. Miotello, A. and R. Kelly (1995). "Critical assessment of thermal models for laser sputtering at high fluences." Applied Physics Letters 67(24): 3535-3537 _____________________________________________________________________ 202 Moalem, A., A. Schoonderbeek, et al. (2007). Laser micro welding in silicon solar module manufacturing. Proceedings of LPM2007 – 8th International Symposium on Laser Precision Microfabrication, University of Vienna, Austria. Molpeceres, C., S. Lauzurica, et al. (2007). "Advanced 3D micromachining techniques using UV laser sources." Microelectronic Engineering 84(5-8): 1337-1340 Müllenborn, M., H. Dirac, et al. (1996). "Fast three-dimensional laser micromachining of silicon for microsystems." Sensors and Actuators A: Physical 52(1-3): 121-125 Muller, O., S. De Unamuno, et al. (1996). "Surface Morphology and Structural Modifications of Si Targets Submitted to Intense Pulsed Laser Irradiations." physica status solidi (a) 158(2): 385-396 NanoMarkets, L. (2009). Materials for Thin-Film Silicon Photovoltaic. Ngoi, B. K. A., K. Venkatakrishnan, et al. (2001). "Effect of energy above laser-induced damage thresholds in the micromachining of silicon by femtosecond pulse laser." Optics and Lasers in Engineering 35(6): 361-369. Norman, S., M. N. Zervas, et al. (2004). Latest development of high-power fiber lasers in SPI. Fiber Lasers: Technology, Systems, and Applications, San Jose, Ca, USA, SPIE. Nutfield (2011). Nutfield XLR8 Two Axis Scan Head, Nutfield Technology, http://www.nutfieldtech.com/pscanheads/xlr8_15.asp. O'Neill, W. and K. Li (2009). "High-Quality Micromachining of Silicon at 1064 nm Using a High-Brightness MOPA-Based 20-W Yb Fiber Laser." IEEE Journal of Selected Topics In Quantum Electronics 15(2): 462-470. ORC (2009). University of Southampton. Otani, T., L. Herbst, et al. (2004). "Microdrilling and micromachining with diode- pumped solid-state lasers." Applied Physics A: Materials Science & Processing 79(4): 1335-1339. Palik, E. D. (1991). Handbook of Optical Constants of Solids II. London, Academic Press. Paul, P. M., E. S. Toma, et al. (2001). "Observation of a Train of Attosecond Pulses from High Harmonic Generation." Science 292(5522): 1689 -1692 Polleux, J. L. and C. Rumelhard (2000). Optical absorption coefficient determination and physical modelling of strained SiGe/Si photodetectors. High Performance Electron Devices for Microwave and Optoelectronic Applications, 2000 8th IEEE International Symposium on: 167-172. _____________________________________________________________________ 203 Qi, H., T. Chen, et al. (2009). "Micromachining of microchannel on the polycarbonate substrate with CO2 laser direct-writing ablation." Optics and Lasers in Engineering 47(5): 594-598 Ready, J. F. (1963). "abstract only." J. Opt. Soc. Am. 53: 514. Ready, J. F. (1963). "Development of plume of material vaporized by giant-pulse laser." Applied Physics Letters 3(1): 11-13. Ready, J. F. (1965). "Effects Due to Absorption of Laser Radiation." Journal of Applied Physics 36(2): 462-468. Ready, J. F. (1971). Effects of High-Power Laser Radiation. New York, Academic Press, pp 73. Ready, J. F. (1982). "Material processing: An overview." Proceedings of the IEEE 70(6): 533-544. Ren, J., S. S. Orlov, et al. (2004). Nanosecond laser silicon micromachining. San Jose, Ca, USA, SPIE. 5339: 382-393 Richardson, D., J. Minelly, et al. (1997). "Fiber laser systems shine brightly." Laser Focus World 33(9): 87-96. Riet, E. v. d., C. J. C. M. Nillesen, et al. (1993). "Reduction of droplet emission and target roughening in laser ablation and deposition of metals." Journal of Applied Physics 74(3): 2008. Russo, R. E. (1995). "Laser Ablation." Applied Spectroscopy 49(9): 14A-28A Russo, R. E., X. L. Mao, et al. (1999). "Time-resolved plasma diagnostics and mass removal during single-pulse laser ablation." Applied Physics A: Materials Science & Processing 69(0): S887-S894. Sakuma, I., H. Yonezu, et al. (1978). Semiconductor laser device equipped with a silicon heat sink. United States Patent, Nippon Electric Co., Ltd. (Tokyo, JA). Salleo, A., T. Sands, et al. (2000). "Machining of transparent materials using an IR and UV nanosecond pulsed laser." Applied Physics A-Materials Science & Processing 71(6): 601-608. Sánchez, F., J. L. Morenza, et al. (1996). "Whiskerlike structure growth on silicon exposed to ArF excimer laser irradiation." Applied Physics Letters 69(5): 620. Schittenhelm, H., G. Callies, et al. (1998). "Two-wavelength interferometry on excimer laser induced vapour/plasma plumes during the laser pulse." Applied Surface Science 127-129: 922-927 _____________________________________________________________________ 204 Schneiderlöchner, E., R. Preu, et al. (2002). "Laser-fired rear contacts for crystalline silicon solar cells." Progress in Photovoltaics: Research and Applications 10(1): 29-34 Schoonderbeek, A., R. Kling, et al. (2007). Efficiency increase of silicon solar cells by laser processing. Proc ALAC 2007, Boston, Ma. Schoonderbeek, A., U. Stute, et al. (2006). Laser technology for cost reduction in silicon solar cell production. ICALEO 2006, Laser Microprocessing Conference, Scottsdale, Ariz., USA. Semak, V. and A. Matsunawa (1997). "The role of recoil pressure in energy balance during laser materials processing." Journal of Physics D: Applied Physics 30(18): 2541-2552. Shah, A., P. Torres, et al. (1999). "Photovoltaic Technology: The Case for Thin-Film Solar Cells." Science 285(5428): 692-698. Shanks, H. R., P. D. Maycock, et al. (1963). "Thermal Conductivity of Silicon from 300 to 1400°K." Physical Review 130(5): 1743. Shannon, M. A., X. L. Mao, et al. (1995). "Laser ablation mass removal versus incident power density during solid sampling for inductively coupled plasma atomic emission spectroscopy." Analytical Chemistry 67(24): 4522-4529. Shinn, G. B., F. Steigerwald, et al. (1986). "Excimer laser photoablation of silicon." Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures 4(6): 1273-1277. Shirk, M. D. and P. A. Molian (1998). "A review of ultrashort pulsed laser ablation of materials." Journal of Laser Applications 10(1): 18-28. Shuja, S., B. Yilbas, et al. (2007). "Laser repetitive pulse heating influence of pulse duty on temperature rise." Heat and Mass Transfer 43(9): 949-955. Shuja, S. Z., A. F. M. Arif, et al. (2002). "Laser Repetitive Pulse Heating of Steel Surface: A Material Response to Thermal Loading." Journal of Manufacturing Science and Engineering 124(3): 595-604. Shvarev, K. M., B. A. Baum, et al. (1975). "Optical properties of liquid silicon." Sov. Phys.—Solid State 16: 2111-2. Singh, R., M. J. Alberts, et al. (2008). "Characterization and prediction of the heat- affected zone in a laser-assisted mechanical micromachining process." International Journal of Machine Tools and Manufacture 48(9): 994-1004 _____________________________________________________________________ 205 Singh, R. K. and J. Narayan (1990). "Pulsed-laser evaporation technique for deposition of thin films: Physics and theoretical model." Physical Review B 41(13): 8843-8859 Singh, R. K. and J. Viatella (1994). "Estimation of plasma absorption effects during pulsed laser ablation of high-critical-temperature superconductors." Journal of Applied Physics 75(2): 1204-1206 Smith, P. M., P. G. Carey, et al. (1997). "Excimer laser crystallization and doping of silicon films on plastic substrates." Applied Physics Letters 70: 342. Sokolowski-Tinten, K., J. Bialkowski, et al. (1998). "Transient States of Matter during Short Pulse Laser Ablation." Physical Review Letters 81(1): 224. Spencer, M. (1982). Fundamentals of light microscopy. Cambridge, UK, Cambridge Univeristy Press SPI Lasers (2008). Beam Delivery Options and Output Beam Specification. SM- S00090_A 2008-08-05. Southampton, UK, SPI Lasers. Steen, W. M. (1998). Laser material processing. London, UK, Springer Stolberg, K. and S. Friedel (2010). High-speed Laser Drilling of EWT/MWT Vias for Backcontact Solar Cells, http://www.renewableenergyworld.com/rea/news/article/2010/08/high-speed- laser-drilling-of-ewtmwt-vias-for-backcontact-solar-cells. Takao, H., M. Ishida, et al. (2002). "A pneumatically actuated full in-channel microvalve with MOSFET-like function in fluid channel networks." Journal of Microelectromechanical Systems 11(5): 421-426. Tan, B. (2006). "Deep micro hole drilling in a silicon substrate using multi-bursts of nanosecond UV laser pulses." Journal of Micromechanics and Microengineering 16(1): 109-112. Tan, B. and K. Venkatakrishnan (2006). "Thermal coupling in multishot laser microvia drilling for interconnection application." Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures 24(1): 211-215. Tan, B. and K. Venkatakrishnan (2007). "Nd-YAG laser microvia drilling for interconnection application." Journal of Micromechanics and Microengineering 17(8): 1511-1517. Thielsch, R., A. Gatto, et al. (2002). "A comparative study of the UV optical and structural properties of SiO2, Al2O3, and HfO2 single layers deposited by reactive evaporation, ion-assisted deposition and plasma ion-assisted deposition." Thin Solid Films 410(1-2): 86-93. _____________________________________________________________________ 206 Tokarev, V., W. Marine, et al. (1994). "Time-resolved measurements of plume shielding during ArF laser ablation of silicon." Thin Solid Films 241(1-2): 129-133. Tunna, L., A. Kearns, et al. (2001). "Micromachining of copper using Nd:YAG laser radiation at 1064, 532, and 355 nm wavelengths." Optics & Laser Technology 33(3): 135-143 Unamuno, S. d. and E. Fogarassy (1989). "A thermal description of the melting of c- and a-silicon under pulsed excimer lasers." Applied Surface Science 36(1-4): 1-11 Ursu, I., I. Apostol, et al. (1985). "Energy plasma-target coupling in the case a periodical structure is formed within the irradiation spot as an effect of powerful laser irradiation." Optics Communications 55(6): 409-412. Veeco (1999). WYKO Surface Profilers Technical Reference Manual. USA, Veeco Metrology Group. Wang, X., Z. H. Shen, et al. (2010). "Laser-induced damage threshold of silicon in millisecond, nanosecond, and picosecond regimes." Journal of Applied Physics 108(3): 033103. Weakliem, H. A. and D. Redfield (1979). "Temperature dependence of the optical properties of silicon." J. Appl. Phys. 50(3): 1491-1493. Weber, K. J., A. W. Blakers, et al. (2004). "A novel low-cost, high-efficiency micromachined silicon solar cell." IEEE Electron Device Letters 25(1): 37- 39. Willis, D. A. and X. Xu (2000). "Transport Phenomena and Droplet Formation During Pulsed Laser Interaction With Thin Films." Journal of Heat Transfer 122(4): 763- 770. Wolff-Rottke, B., J. Ihlemann, et al. (1995). "Influence of the laser-spot diameter on photo-ablation rates." Applied Physics A Materials Science & Processing 60(1): 13-17. Wood, R. F. and G. E. Jellison Jr (1984). Melting Model of Pulsed Laser Processing. Semiconductors and Semimetals. New York, Elsevier. 23: 165 Wu, B. and Y. C. Shin (2006). "Modeling of nanosecond laser ablation with vapor plasma formation." Journal of Applied Physics 99(8): 084310-8 Wu, M. C. (1997). "Micromachining for optical and optoelectronic systems." Proceedings of the IEEE 85(11): 1833-1856. Xu, X. (2002). "Phase explosion and its time lag in nanosecond laser ablation." Applied Surface Science 197-198: 61-66 _____________________________________________________________________ 207 Yamamoto, K., T. Abe, et al. (1991). "Thermal Diffusivity of Crystalline and Liquid Silicon and an Anomaly at Melting." Japanese Journal of Applied Physics 30(Part 1, No. 10): 2423-2426. Yavas, O. and M. Takai (1998). "High-speed maskless laser patterning of indium tin oxide thin films." Applied Physics Letters 73(18): 2558-2560. Yoo, J. H., S. H. Jeong, et al. (2000). "Explosive change in crater properties during high power nanosecond laser ablation of silicon." Journal of Applied Physics 88(3): 1638-1649 Yoo, J. H., S. H. Jeong, et al. (2000). "Evidence for phase-explosion and generation of large particles during high power nanosecond laser ablation of silicon." Applied Physics Letters 76(6): 783-785 Zeng, X., X. L. Mao, et al. (2005). "Experimental investigation of ablation efficiency and plasma expansion during femtosecond and nanosecond laser ablation of silicon." Applied Physics A: Materials Science & Processing 80(2): 237-241 Zeng, X. J., X. Mao, et al. (2004). Ultraviolet femtosecond and nanosecond laser ablation of silicon: ablation efficiency and laser-induced plasma expansion. Proceedings of SPIE, Taos, NM, USA. Zhang, C., I. A. Salama, et al. (2006). "One-dimensional transient analysis of volumetric heating for laser drilling." Journal of Applied Physics 99(11): 113530 Zhang, X., S. S. Chu, et al. (1997). "Excimer laser ablation of thin gold films on a quartz crystal microbalance at various argon background pressures." Applied Physics A: Materials Science & Processing 64(6): 545-552 Zhao, J., A. Wang, et al. (1995). "Twenty-four percent efficient silicon solar cells with double layer antireflection coatings and reduced resistance loss." Applied Physics Letters 66: 3636 Zhao, J., A. Wang, et al. (1998). "19.8% efficient “honeycomb” textured multicrystalline and 24.4% monocrystalline silicon solar cells." Applied Physics Letters 73: 1991 Zhu, S., Y. F. Lu, et al. (2001). "Laser ablation of solid substrates in a water-confined environment." Applied Physics Letters 79(9): 1396