The Pennsylvania State University. The Graduate School. College of Engineering PARAMETER OPTIMIZATION OF LASER-DOPED SELECTIVE EMITTERS FOR

Transcription

1 The Pennsylvania State University The Graduate School College of Engineering PARAMETER OPTIMIZATION OF LASER-DOPED SELECTIVE EMITTERS FOR APPLICATIONS IN THE SILICON PHOTOVOLTAIC INDUSTRY A Thesis in Engineering Science by Holly E. Heinrichs 2012 Holly E. Heinrichs Submitted in Partial Fulfillment of the Requirements for the Degree of Master of Science December 2012 i

2 The thesis of Holly E. Heinrichs was reviewed and approved* by the following: S. Ashok Professor of Engineering Science Thesis Co-Advisor Edward W. Reutzel Head of Laser System Engineering and Integration Applied Research Lab Thesis Co-Advisor Suzanne Mohney Professor of Materials Science and Engineering Judith Todd P.B. Breneman Department Head Chair Head of the Department of Engineering Science and Mechanics *Signatures are on file in the Graduate School ii

3 ABSTRACT The advancement of commercial silicon solar cells has plateaued over the past decade; however, in recent years with the addition of selective emitters, overall cell efficiency has increased by % absolute. A selective emitter increases efficiency by heavily doping the region directly below the front-side metal contacts, providing a potential energy barrier for the minority charge carriers and thus lowering recombination at the metal-semiconductor interface. Laser doping provides a very attractive method to create the selective emitters due to its ability to locally heat and dope the surface of the cell without additional patterning steps. This work investigates the influence of laser processing parameters and dopant concentration in the creation of selective emitters. Q-switched lasers of 1064, 532, and 355 nm wavelengths are used at a range of pulse energies on p-type FZ silicon with a pre-deposited thin n+ dopant film with varying thickness. In addition to the pulsed lasers, a 1070 nm continuous wave laser is used with varying power and pulse duration. The results are then characterized electrically through current-voltage measurements, and diode characteristics are extrapolated from the curves. The characteristics are compared via 3-D graphs and the ideal processing conditions and dopant concentrations for each laser are determined. Electrically, the best LDSEs were created with the 1064-nm 4-ns Q-switched laser on dopant structures with 60 to 80-nm n+a-si:h with pulse energies of 80-µJ and above for the tophat beam profile and of 200-µJ and greater when using the Gaussian beam profile. The poorest LDSEs were created with the 355-nm 30-ns Q-switched laser. These diodes exhibited high series resistance and saturation current which are both undesirable for selective emitters. iii

4 TABLE OF CONTENTS LIST OF FIGURES... vi LIST OF TABLES... xi LIST OF EQUATIONS... xii LIST OF SYMBOLS...xiii ACKNOWLEDGEMENTS... xvi CHAPTER 1: INTRODUCTION... 1 CHAPTER 2: LITERATURE REVIEW... 3 Background of Laser Science... 3 Introduction to Lasers... 3 Necessary Conditions for Laser Operation... 3 Continuous Wave vs. Pulsed Operation... 7 Gaussian vs. Top hat Beam Manipulation... 8 Neodymium: Yttrium Aluminum Garnate (Nd:YAG) Lasers... 9 Neodymium: Yttrium Vanadate (Nd:YVO 4 ) Lasers Fiber Lasers Laser Material Interaction Laser Assisted Diffusion Introduction to Photovoltaics History of Photovoltaic Power Generation Solar Cell Device Physics Solar Cell Characterization Silicon Solar Cells Laser Processing in the Photovoltaic Industry Laser Fired Contacts Laser Doping Laser Doped Selective Emitters Laser-Induced Damage CHAPTER 3: DESIGN OF EXPERIMENTS: LASER DOPED SELECTIVE EMITTERS Goals and Objectives iv

5 Laser Doped Selective Emitter Creation Laser Doped Selective Emitter Results CHAPTER 4: LASER DOPED SELECTIVE EMITTER DISCUSSION CHAPTER 5: CONCLUSIONS AND FUTURE WORK REFERENCES APPENDIX A: Extended Literature Review History of the Laser Metallization Techniques used for Photovoltaics Screen Printing Lithography Light-induced Electro-plating Laser Sintering Chemical Plating of Nickel Metal Aerosol Jet Printing Pad Printing Metallization Pattern Design v

6 LIST OF FIGURES Figure 1: (a) Population distribution under thermodynamic equilibrium and Boltzmann distribution and (b) population distribution when population inversion is occurring... 5 Figure 2: Energy-level arrangements and pumping and decay processes for three- and four-level lasers. 6 Figure 3: (a) A Gaussian intensity distribution and (b) A top hat intensity distribution profile... 9 Figure 4: Absorption coefficient of silicon vs. wavelength Figure 5: Example of a laser pulse which caused ablation of the material Figure 6: The energy band diagram for a p-n junction Figure 7: Diagram of a p-n junction which depicts the electric field and the depletion width: Figure 8: Air Mass schematic [10] Figure 9: Spectral irradiance vs. wavelength [10] Figure 10: Typical solar cell current density vs. voltage characteristics Figure 11: Universally accepted equivalent electrical circuit of a solar cell Figure 12: The effect of series resistance on the J-V characteristics of a solar cell [10] Figure 13: Ln(I) vs. V for a diode with series resistance Figure 14: The effect of shunt resistance on the J-V characteristics of a solar cell [10] Figure 15: Ohmic metal-contact to p-type semiconductor Figure 16: Non-ohmic metal contact to p-type semiconductor Figure 17: Selective ohmic contact to anode which utilizes a material with a new bandgap Figure 18: Selective ohmic contact at the cathode which utilizes effective forces Figure 19: Current-voltage characteristics of a two-region diode [24] Figure 20: Passivation and dopant structures used in LDSE experiments Figure 21: SIMS profile for the deposited 60 nm n+a-si:h doped thin film Figure 22: Experimental optic schematic for lasers Figure 23: Parameter schematic for Q-switched lasers Figure 24: Parameter schematic for IPG Figure 25: 10 x 10 metallization mask layout for the laser doped selective emitters Figure 26: Schematic of sample with lithography layers after the sample has been subjected to ultraviolet radiation and developed in CD Figure 27: Portion of sample post metallization and annealing Figure 28: SEM images of the 355-nm, 30-ns, pulsed laser at various energies under 1000x magnification Figure 29: SEM images for the 532-nm, 4-ns, pulsed laser at various energies under 1000x magnification Figure 30: SEM images of the Gaussian 1064-nm, 4-ns, pulsed laser at various energies under 1000x magnification Figure 31: SEM images of the 1070-nm continuous wave laser at highest powers and longest pulse durations at 1000x magnification vi

7 Figure 32: SEM images of the 1070-nm continuous wave laser at highest powers and shortest pulse durations at 1000x magnification Figure 33: SEM images of the 1070-nm continuous wave laser at lowest powers and longest pulse durations at 1000x magnification Figure 34: SEM images of the 1070-nm continuous wave laser at lowest powers and shortest pulse durations at 1000x magnification Figure 35: Comparison of a CW vs. pulsed laser spot taken in an SEM under 1000x magnification Figure 36: Current-voltage characteristics for the 355-nm, 30-ns pulsed laser on the 10/20/150 stack.. 80 Figure 37: Current-voltage characteristics of the 355-nm, 30-ns pulsed laser on the 10/40/150 stack Figure 38: Current-voltage characteristics of the 355-nm, 30-ns pulsed laser on 10/60/150 stack Figure 39: Current-voltage characteristics of the 355-nm, 30-ns pulsed laser on the 10/80/150 stack Figure 40: Current-voltage characteristics of the 355-nm, 30-ns pulsed laser on the 10/120/150 stack.. 84 Figure 41: 3D graph comparing the ideality factor for the various stacks and pulse energies of the 355- nm, 30-ns pulsed laser Figure 42: 3D graph comparing the decades of linearity for the various stacks and pulse energies of the 355-nm, 30-ns pulsed laser Figure 43: 3D graph comparing the series resistance for the various stacks and pulse energies of the 355- nm, 30-ns pulsed laser Figure 44: 3D graph comparing the saturation current for the various stacks and pulse energies of the 355-nm, 30-ns pulsed laser Figure 45: 3D graph comparing the shunt resistance for the various stacks and pulse energies of the 355- nm, 30-ns pulsed laser Figure 46: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/20/150 stack Figure 47: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/40/150 stack Figure 48: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/60/150 stack Figure 49: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/80/150 stack Figure 50: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/120/150 stack Figure 51: 3D graph comparing the ideality factor for the various stacks and pulse energies of the 532- nm, 4-ns pulsed laser Figure 52: 3D graph comparing the decades of linearity for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser Figure 53: 3D graph comparing series resistance for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser Figure 54: 3D graph comparing the saturation current for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser Figure 55: 3D graph comparing shunt resistance for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser Figure 56: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/20/150 stack Figure 57: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/40/150 stack vii

8 Figure 58: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/60/150 stack Figure 59: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/80/150 stack Figure 60: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/120/150 stack Figure 61: 3D graph comparing the ideality factor for the various stacks and pulse energies of the nm, 4-ns pulsed Gaussian laser Figure 62: 3D graph comparing the decades of linearity for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed Gaussian laser Figure 63: 3D graph comparing series resistance for the various stacks and pulse energies of the nm, 4-ns pulsed Gaussian laser Figure 64: 3D graph comparing the saturation current for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed Gaussian laser Figure 65: 3D graph comparing shunt resistance for the various stacks and pulse energies of the nm, 4-ns pulsed Gaussian laser Figure 66: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/20/150 stack Figure 67: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/40/150 stack Figure 68: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/60/150 stack Figure 69: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/80/150 stack Figure 70: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/120/150 stack Figure 71: 3D graph comparing the ideality factor for the various stacks and pulse energies of the nm, 4-ns pulsed top hat laser Figure 72: 3D graph comparing the decades of linearity for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed top hat laser Figure 73: 3D graph comparing series resistance for the various stacks and pulse energies of the nm, 4-ns pulsed top hat laser Figure 74: 3D graph comparing the saturation current for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed top hat laser Figure 75: 3D graph comparing shunt resistance for the various stacks and pulse energies of the nm, 4-ns pulsed top hat laser Figure 76: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/20/150 stack Figure 77: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/40/150 stack viii

9 Figure 78: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/60/150 stack Figure 79: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/80/150 stack Figure 80: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/120/150 stack Figure 81: Current-voltage characteristics of the 10/20/150 stack with 1070-nm CW laser with highest power and longest pulse duration Figure 82: Current-voltage characteristics of the 10/20/150 stack with 1070-nm CW laser with highest power and shortest pulse duration Figure 83: Current-voltage characteristics of the 10/20/150 stack with 1070-nm CW laser with lowest power and longest pulse duration Figure 84: Current-voltage characteristics of the 10/20/150 stack with 1070-nm CW laser with lowest power and shortest pulse duration Figure 85: Ideality factor for the 10/20/150 stack with the 1070-nm CW laser Figure 86: Decades of linearity for the 10/20/150 stack with the 1070-nm CW laser Figure 87: Series resistance for the 10/20/150 stack with the 1070-nm CW laser Figure 88: Saturation current for the 10/20/150 stack with the 1070-nm CW laser Figure 89: Shunt resistance for the 10/20/150 stack with the 1070-nm CW laser Figure 90: Current-voltage characteristics of the 10/40/150 stack with 1070-nm CW laser with highest power and longest pulse duration Figure 91 Current-voltage characteristics of the 10/40/150 stack with 1070-nm CW laser with highest power and shortest pulse duration Figure 92 Current-voltage characteristics of the 10/40/150 stack with 1070-nm CW laser with lowest power and longest pulse duration Figure 93 Current-voltage characteristics of the 10/40/150 stack with 1070-nm CW laser with lowest power and shortest pulse duration Figure 94: Ideality factor for the 10/40/150 stack with the 1070-nm CW laser Figure 95: Decades of linearity for the 10/40/150 stack with the 1070-nm CW laser Figure 96: Series resistance for the 10/40/150 stack with the 1070-nm CW laser Figure 97: Saturation current for the 10/40/150 stack with the 1070-nm CW laser Figure 98: Shunt resistance for the 10/40/150 stack with the 1070-nm CW laser Figure 99: Current-voltage characteristics of the 10/80/150 stack with 1070-nm CW laser with highest power and longest pulse duration Figure 100: Current-voltage characteristics of the 10/80/150 stack with 1070-nm CW laser with highest power and shortest pulse duration Figure 101: Current-voltage characteristics of the 10/80/150 stack with 1070-nm CW laser with lowest power and longest pulse duration Figure 102: Current-voltage characteristics of the 10/80/150 stack with 1070-nm CW laser with lowest power and shortest pulse duration Figure 103: Ideality factor for the 10/80/150 stack with the 1070-nm CW laser ix

10 Figure 104: Decades of linearity for the 10/80/150 stack with the 1070-nm CW laser Figure 105: Series resistance for the 10/80/150 stack with the 1070-nm CW laser Figure 106: Saturation current for the 10/80/150 stack with the 1070-nm CW laser Figure 107: Shunt resistance for the 10/80/150 stack with the 1070-nm CW laser Figure 108: Current-voltage characteristics of the 10/120/150 stack with 1070-nm CW laser with highest power and longest pulse duration Figure 109: Current-voltage characteristics of the 10/120/150 stack with 1070-nm CW laser with highest power and shortest pulse duration Figure 110: Current-voltage characteristics of the 10/120/150 stack with 1070-nm CW laser with lowest power and longest pulse duration Figure 111: Current-voltage characteristics of the 10/120/150 stack with 1070-nm CW laser with lowest power and shortest pulse duration Figure 112: Ideality factor for the 10/120/150 stack with the 1070-nm CW laser Figure 113: Decades of linearity for the 10/120/150 stack with the 1070-nm CW laser Figure 114: Series resistance for the 10/120/150 stack with the 1070-nm CW laser Figure 115: Saturation current for the 10/120/150 stack with the 1070-nm CW laser Figure 116: Shunt resistance for the 10/120/150 stack with the 1070-nm CW laser Figure 117: 3D graph comparing the ideality factor for the various stacks and average powers of the 1070-nm, CW laser at 200-µs pulse duration Figure 118: 3D graph comparing the decades of linearity for the various stacks and average powers of the 1070-nm, CW laser at 200-µs pulse duration Figure 119: 3D graph comparing series resistance for the various stacks and average powers of the nm, CW laser at 200-µs pulse duration Figure 120: 3D graph comparing the saturation current for the various stacks and average powers of the 1070-nm, CW laser at 200-µs pulse duration Figure 121: 3D graph comparing shunt resistance for the various stacks and average powers of the nm, CW laser at 200-µs pulse duration Figure 122: 3D graph comparing the ideality factor for the various stacks and pulse durations of the 1070-nm, CW laser at W average power Figure 123: 3D graph comparing the decades of linearity for the various stacks and pulse durations of the 1070-nm, CW laser at W average power Figure 124: 3D graph comparing series resistance for the various stacks and pulse durations of the nm, CW laser at W average power Figure 125: 3D graph comparing the saturation current for the various stacks and pulse durations of the 1070-nm, CW laser at W average power Figure 126: 3D graph comparing shunt resistance for the various stacks and pulse durations of the nm, CW laser at W average power Figure 127: SEM photograph of a sample which has a laser sintered seed layer which is thickened by light-induced silver plating [26] Figure 128: Schematic of a metal Aerosol jet printing head [26] x

11 LIST OF TABLES Table 1: Laser parameters and spot dimensions for the 355-nm, 30-ns pulsed laser Table 2: Laser parameters and spot dimensions for the 532-nm, 4-ns pulsed laser Table 3: Laser parameters and spot dimensions for the 1064-nm, 4-ns pulsed laser with Gaussian and top hat beam profiles Table 4: Laser parameters and spot dimensions for the 1070-nm, CW laser xi

12 LIST OF EQUATIONS Equation 1 Wavelength calculation based on energy levels... 4 Equation 2: Intensity within gain material... 5 Equation 3: Gain calculation equation... 5 Equation 4: Fick's first law for steady-state diffusion (in one dimension) Equation 5: Air mass calculation for solar spectrum Equation 6: Spectral conversion efficiency equation Equation 7: The Beer-Lambert Law Equation 8: Dark current density for an ideal diode Equation 9: The illumination current density for an ideal solar cell Equation 10: Maximum short circuit current density for a solar cell Equation 11: Open-circuit voltage equation Equation 13: Open-circuit voltage approximation equation Equation 14: The fundamental thermodynamic limit to open-circuit voltage as discovered by Keiss and Rehwals [22] Equation 15: The electrical output power of a solar cell Equation 16: The theoretical maximum electrical output power of a solar cell Equation 17: The fill factor equation Equation 18: Solar cell efficiency at the maximum power point Equation 19: An accurate method for determining series resistance within a diode Equation 20: Current for a two-region diode under photon illumination Equation 21: Ideality factor for a diode Equation 22: Current equation for the universal electrical circuit representation of a solar cell xii

13 LIST OF SYMBOLS = Absorption coefficient = Angle of incident sunlight measured from a perpendicular line to the Earth s surface = Energy conversion efficiency of a solar cell = Spectral conversion efficiency = Wavelength = Photons absorbed by a material per unit time = Work function of a metal = Work function of a n-type semiconductor = Work function of a p-type semiconductor = True capacitance = Speed of light in vacuum = 10 8 m/s = Measured capacitance = Bandgap energy of a semiconductor = Quantized energy value at the nth energy level = Measurement frequency = Fill Factor = Conductance ( ) = gain coefficient = Statistical weight of the lower energy level = Statistical weight of the upper energy level = Planck constant = J-s = Current (with regards to electrical characterization) = Intensity of the beam (with regards to laser fundamentals) = The saturation current of a solar cell diode xiii

14 = Initial intensity of the beam (with regards to laser fundamentals) = The saturation current of the quasi-neutral region of a diode = The saturation current of the space charge region of a diode = The current due to incoming photons of a diode under illumination ( ) = The intensity of impingent light a distance into the material = The dark current density = The current density under illumination = The current density maximum power point = The reverse saturation current = The photo-generated current density = constant = Boltzman constant = The ideality factor of a solar cell diode = The ideality factor of the quasi-neutral region of a diode = The ideality factor of the space charge region of a diode = Total number of atoms per unit volume in the lower energy level = Total number of atoms per unit volume in the upper energy level = The electrical power output per area of a solar cell = The theoretical maximum electrical power per area of a solar cell = The power of the sun at a given wavelength = Charge = The parallel resistance in a solar cell. It stands for both recombination and shunt resistances. = Series resistance = The slope of the linear segment of a diode curve = Temperature xiv

15 = Measured voltage across entire diode = Built-in potential voltage = Diode voltage = The voltage maximum power point = Open circuit voltage = The theoretical thermodynamic maximum open-circuit voltage xv

16 ACKNOWLEDGEMENTS Firstly I would like to thank my advisor Dr. Ashok for his continual wisdom, guidance and patience while sharing his knowledge with me. I would like to thank Dr. Reutzel for his thought provoking questions and inspiring me to learn why things behave as they do. To Dr. Mohney, for her insight into material science and for giving me access to the resources to complete the project. Also, to the BP Solar team for their research guidance and financial support of the project as well as the laser processing division at the Applied Research Lab for the hard work they put into the project. I would also like to thank Dr. Wolfe for his generosity with his laboratory equipment. Finally, I would like to thank my friends and family who have stood by me through this experience. Your continuous support was invaluable and has helped me through both the good and the bad times. Philippians 4:13. xvi

17 CHAPTER 1: INTRODUCTION In the past decade, much research has been focused on creating and improving alternative energy sources and decreasing dependence on fossil fuels. Within the photovoltaic industry, silicon solar cells have received the bulk of research because of their relatively low production costs and ease of manufacturability. Advancements in both the quality of silicon along with device improvements such as laser fired contacts have allowed solar energy to become more competitive in the global energy market. Selective emitters have been one such advancement which has helped propel the efficiency gain in the solar industry. In a typical cell design, a lightly-doped shallow emitter covers the entire cell which allows for charge separation while maintaining maximal carrier lifetimes. In a cell with a selective emitter, the blanket shallow emitter is still present, but there are also select regions under the metal contact fingers which shield minority charge carriers from entering the contact. This is achieved via some type of potential energy barrier which could be a product of material design or through a concentrated highly doped area. By preventing minority carriers from entering the metal contacts and reducing parasitic resistance, recombination rates are greatly reduced thereby increasing overall cell efficiency Laser doped selective emitters in particular prevent recombination by utilizing effective forces and creating an area of higher dopant concentration directly below the metal contact. Lasers are an ideal processing tool for this job due to their highly coherent beam and their ability to concentrate large amounts of energy on an extremely small area without physically contacting the cell. The objective of this thesis is to investigate the effect of varying laser processing parameters when creating selective emitters. Additionally, the dopant concentration within the laser doped selective emitter is altered by changing the dopant film thickness. This allows for parameter 1

18 optimization from not only the laser processing side, but the material science side as well. The results are compared qualitatively by examining the current-voltage characteristics of the individual diodes and extrapolating key values. 2

19 CHAPTER 2: LITERATURE REVIEW Background of Laser Science Introduction to Lasers A laser is a device which amplifies light to produce a high intensity and directional beam consisting of a single wavelength or frequency. The term laser is actually an acronym for Light Amplification by Stimulated Emission of Radiation. [1] Lasers have found applications in almost every modern day industry including electronics, welding, engineering, military and the medical industry to name a few. Their major claim-to-fame is their ability to concentrate coherent electromagnetic radiation in a single direction. When comparing lasers to everyday white light, lasers have much narrower frequency spread or bandwidth, much higher intensity and a greater degree of collimation. Lasers can be found in a variety of types and sizes. They have powers which can range from 10-9 to W and frequencies ranging from to Hz. [1] Some lasers behave as continuous wave (CW) sources whereas others require Q-switching and emit pulses of energy. These pulses can vary with time reaching as short as 5 x seconds. [1] Because there is such a spectrum of laser properties, certain lasers are much more applicable to certain jobs than others. One major advantage of using lasers in many applications is their ability to transfer energy in a non-contact manner. Necessary Conditions for Laser Operation As previously stated, the goal of a laser is to amplify previously generated light signals; however, there are a few essential design elements required for this to happen. One of these is that the laser must consist of some kind of gain or amplifying medium. This can be found in a variety of forms including a solid, liquid, gas, or plasma. [1] This medium should be between a set of mirrors which are designed to feed the light back into the amplifier for continued growth of the beam. These mirrors are 3

20 typically located in the resonator cavity and one of these mirrors must be a partially transparent output coupler which allows for controlled emission of light from the cavity. [1] Finally, a laser must consist of a pumping device which will excite electrons from lower energy levels to higher energy levels. This device is typically a flashlamp or another laser. Now that the essential laser mechanical parts are in place, the laser also has two theoretical conditions needed for operation: population inversion and light amplification. Population inversion occurs when the number of electrons in the excited state is larger than that of the lower energy state. According to the Boltzmann principle, while at thermal equilibrium there will be a larger number of atoms in a lower energy than the number in a high energy state. [1] However, when population inversion occurs, the reverse is true. This leads to a higher stimulated emission where a photon is emitted as the electrons return to the lower energy level. [1] The wavelength of these photons is dependent on the difference in energy between the levels and is found using the following equation: Equation 1 Wavelength calculation based on energy levels Where is the wavelength, is Plank s constant, is the speed of light and and are the corresponding energy levels of the electron states. The following figure demonstrates a Boltzmann distribution of energy levels in part (a) and a distribution undergoing population inversion in (b). 4

21 Figure 1: (a) Population distribution under thermodynamic equilibrium and Boltzmann distribution and (b) population distribution when population inversion is occurring In order to create this population inversion, one must look at the fundamental equations governing laser growth. The following equation describes the intensity of the beam as it travels through a gain material: Equation 2: Intensity within gain material Where is the intensity of the beam at a distance into the gain material and is the initial intensity of the beam as it enters the gain material and ( ) is referred to as the gain coefficient and is equal to: Equation 3: Gain calculation equation 5

22 Here, is a constant, and represent the total number of atoms per unit volume in the upper and lower levels of the transitions and and are the statistical weights of the upper and lower energy levels. As can be seen from these two equations, in order for the intensity of the beam to grow within the gain material, the sign of the exponent must be positive, and therefore ( ) must be a positive value. The only way for this to hold true is if and this is the necessary term for population inversion. Typically, this scenario is created using a pump which excites atoms to a higher level and the corresponding decay processes will produce the desired light output. This is traditionally done using either three or four level lasers. A schematic of this is shown below: Figure 2: Energy-level arrangements and pumping and decay processes for three- and four-level lasers 6

23 In traditional solid-state lasers, a pump excites atoms up to a third energy level. Here the atoms experience rapid decay to the second level but this is the location where they begin to build up. This buildup is what creates the population inversion and in return, the laser beam emits a wavelength related to the energy difference between the second and first levels. Gas lasers traditionally exhibit a similar three-level process but the laser wavelength is typically between the third and second levels while the atoms experience rapid decay between the second and first level. Four-level lasers, such as the Nd:YAG, add another step in this excitation process. Here, a pump excites atoms up to a fourth level, where they experience rapid decay. The population inversion occurs between the third and second levels (and correspondingly this is how the wavelength is determined) and again, rapid decay occurs between the second and first. Now that population inversion has been explained, it is time to return back to the other necessary theoretical condition for laser growth: light amplification. Light amplification is a term which implies that the quantity of stimulated emission due to the light passing through the gain medium is greater than the quantity that is simultaneously being absorbed by the said medium. What this is essentially saying is that as the beam makes a pass through the gain material, it must be able to gain more energy from the pumping device than what it loses through its interaction with the gain material. Continuous Wave vs. Pulsed Operation Lasers can be classified as operating in either continuous or pulsed mode. The distinction depends on if the power output is continuous over time or if its output takes the form of pulses of light. However, this distinction essentially comes down to the type of pumping device used inside the laser. 7

24 For continuous wave lasers, they are traditionally pumped by high pressure krypton or tungsten-iodine lamps which are applied continuously. [2] As a result, there is a continuous supply of energy to the gain medium and thus a continuous beam is emitted. Conversely, pulsed lasers tend to be pumped by medium pressure xenon or krypton flash lamps. [2] As can be interpreted from their name, the flash lamps provide an intense flash to supply the energy to excite the gain material. In return, the beam itself also emits a short pulse. One key characteristic of pulsed lasers is they can exhibit extremely high powers due to the short pulse durations. [1] Gaussian vs. Top hat Beam Manipulation Laser beams can come in many different transverse modes but the most common has a Gaussian distribution. With a Gaussian shape, the beam profile has a much higher intensity at the center and it decays as it moves farther away from the center. Unfortunately, this type of distribution is not always desired for certain applications because it allows for uneven energy distribution throughout the spot size. This allows material interaction to be different at the center of the spot compared to that at the edges. For example, ablation could occur at the center whereas melting occurs at the edges of the spot. To combat this problem, one common solution is to manipulate the beam to exhibit a top hat distribution instead. A top hat shape applies a uniform intensity to a well-defined spot and enables sharp and accurate transitions to the material undergoing processing. This has found considerable advantages particularly in the micro-machining industry. [3] There are currently a variety of lenses and π-shaping devices to transform a Gaussian beam into a top hat beam, each are wavelength dependent. [1] 8

25 Figure 3: (a) A Gaussian intensity distribution and (b) A top hat intensity distribution profile Neodymium: Yttrium Aluminum Garnate (Nd:YAG) Lasers A Neodymium: Yttrium Aluminum Garnate laser, more commonly referred to as Nd:YAG, is a four-level solid-state laser. With this laser, Y 5 Al 5 O 12 serves as the host crystal for the neodymium dopant. The resultant crystal can produce a strong emission at a wavelength equal to 1064 nm. Neodymium is a rare earth metallic element which is commonly used in laser fabrication because it has multiple energy bands between 1.56 ev and 2.32 ev which absorbs light from the visible to the infrared regions. [1] A typical Nd:YAG laser setup is relatively simple and consists of the crystal rod surrounded either by flash lamps of noble gases, such as xenon or krypton, or by diode lasers. There is generally a total reflector located at one end of the rod, and a partial reflector on the other end which allows for controlled output of the beam. The rod is usually cooled by flowing water. As previously mentioned, the Nd:YAG laser uses a four-level energy system when exciting the electrons. In YAG, the laser transition is homogeneously broadened by thermally activated lattice vibrations. Putting this into practice, visible to infrared light from the flash lamps generally excite the electrons in the rod where they then experience immediate thermalization and drop non-radiatively to an energy level of 1.38 ev. At this energy level, the rod is in a metastable (equilibrium) state. Metastable 9

26 states are long lived, but not infinitely stable. In this case, it means that electron energy can no longer be lost due to thermallization and spontaneous radiative transitions are highly unlikely. [4] Additionally, because these atoms are in this metastable state, they are able to retain this position for longer periods of time, allowing for significant population inversion to occur. The result of this is that upper level lifetime for the excited atoms in a Nd:YAG crystal is around 230 µs which is relative long compared to many other laser materials. When the excited ion finally does emit radiation, the photon wavelength is 1064 nm and corresponds to the energy jump between the 1.38 ev level and the 0.22 ev level. Upon reaching this level, the ion immediately thermalizes again to the ground state of the system. Neodymium-doped lasers are often operated in frequency-doubled or frequency-tripled modes which produce 532 nm and 355 nm wavelengths, respectively. This frequency manipulation takes the output beam of 1064 nm and manipulates it in a process of nonlinear frequency conversion to achieve the desired wavelength. A Q-switching device is designed to either permit or prevent optical oscillations. It can either be a mechanical switch or an electro-optical device. To utilize the theory of Q-switching, the switch must first be closed to allow for continuous pumping of atoms into the metastable upper laser state. After sufficient energy has been built up, the switch is opened and emits an extremely high energy, short duration pulse, typically less than 10 ns, while draining the upper state. Q-switching allows for large energies to be released in a short time resulting in extremely high power secretion. Because of this, many applications of Q-switched lasers involving material ablation. Neodymium: Yttrium Vanadate (Nd:YVO 4 ) Lasers Neodymium-doped yttrium vanadate is a laser that has frequently been compared to the Nd:YAG laser. There are many similarities including that they both operate at nearly the same wavelength. Nd:YVO 4 naturally operates at a nm wavelength although it typically undergoes 10

27 frequency-doubling (532 nm) or frequency-tripling (355 nm) output. Compared to the Nd:YAG though, it has a few advantages. The first is that it has a 5-times-larger stimulated emission cross section which leads to a higher gain. Additionally, it has a 4-times-stronger absorption cross section which allows for a 6-times-wider pump absorption band. [1] The resultant of these features is that it can be effectively pumped with crystals of only a few millimeters in length making it attractive for use in small diodepumped lasers. The major applications for the 355 nm laser are in stereolithography and in the printed circuit board manufacturing industry. [1] Fiber Lasers Fiber lasers are quite different from the solid-state lasers. They utilize a glass fiber doped with rare earth metals such as erbium, ytterbium, neodymium, dysprosium, praseodymium, and thulium. [1] The entire fiber then acts as both resonator and delivery system. Typically fiber lasers are pumped by semiconductor laser diodes or by other fiber lasers. They have many advantages associated with them, but perhaps the most obvious is that it allows for energy to be easily delivered to a movable element because the beam is already coupled into a flexible fiber. Additionally, fiber lasers tend to have higher output than many other lasers (kw power-levels) due to their long active region and high surface area to volume ratio. Lastly, fiber lasers tend to have extremely high beam quality. This has been found to be linked to the fiber s wave-guiding properties which eliminate thermal distortion of the optical path. Laser Material Interaction Laser-material interaction depends on several different factors. Some of them include wavelength, power or energy, energy distribution and whether the laser has continuous or pulsed operation. One of the primary areas of interest though in determining how a material interacts with the beam is the absorption coefficient of the material. This absorption coefficient, commonly referred to as α, determines the amount of energy which is absorbed within each interval of distance the 11

28 electromagnetic beam permeates into the specific type of material. [1] The reciprocal of the absorption coefficient is the absorption depth which is the average depth over which the absorption of the beam occurs. The absorption of a certain type of material depends not only on the wavelength it is exposed to, but also to the temperature of the material. The following figure shows how the absorption coefficient of silicon is dependent on wavelength. Figure 4: Absorption coefficient of silicon vs. wavelength When energy is absorbed by a material, it gives rise to vibrations, or phonons, within the material s lattice. [4] Depending on the amount of energy supplied, several different outcomes can occur. If the phonons are sufficiently strong, they can cause the lattice to break from its uniform structure and as a result, a liquid state can form. When dealing with lasers, this is commonly referred to as a melt pool. If even more energy is absorbed by the material, yet another phase transformation can take place through vaporization. Here, atomic-size particles obtain enough energy to leave the surface of the material in a gaseous form. [5] However, melting and vaporization are not the only laser-material interactions that can take place. Returning back to vaporization, if a particle is able to vaporize, there must be a vapor pressure 12

29 gradient between the surface of the material and the ambient air. As the particle vaporizes away from the surface, there is a resulting recoil force that becomes present. This recoil force is directed downward into the melt pool and it can become the dominant driving mechanism of melt motion within the pool. [1] [4] If this recoil force is larger than the surface tension of the melt pool, ablation occurs. Ablation is the explosive ejection of molten material from the melt pool. This ejection is uncontrolled and creates nanometer to micrometer size particles which spray out of the pool. Ablation can be a desired result depending upon the application. Typically pulsed lasers operating between a few tens of nanoseconds are used for ablation removal processes. [6] Figure 5: Example of a laser pulse which caused ablation of the material Laser Assisted Diffusion When dealing with laser material interaction and melt pools, one important factor to take into consideration is diffusion, particularly when dealing with heterogeneous structures. Diffusion is the phenomenon of material transport by atomic motion. [5] Although self-diffusion occurs (atoms of one species moving around in that species), interdiffusion is much more commonly studied because it 13

30 involves the transmission of different molecules within each other. The driving force behind diffusion is a concentration gradient and at steady-state diffusion the flux is traditionally described by Fick s first law. Equation 4: Fick's first law for steady-state diffusion (in one dimension) In this law, is the diffusion flux, is the diffusion coefficient, and is the change in concentration over the change in distance. [5] [7] As can be seen by Fick s law, at steady state only two factors influence the rate of diffusion, the concentration gradient and the diffusion coefficient. When studying diffusion in relationship to lasers, the primary parameter to consider is the diffusion coefficient. The diffusion coefficient is determined experimentally and it varies depending on the two substances which are interacting with each other, the phases of the substances, and the temperatures of the substances. This variation can be quite dramatic. For example, most solids tend to have a diffusion coefficient on the order of m 2 /s or smaller whereas many liquid materials have a different coefficient on the order of 10 9 m 2 /s. [7] With regards to laser-material interaction, if the material is heterogeneous, as the beam creates a melt pool it is able to greatly speed up the diffusion of the two materials into each other by changing the diffusion coefficient between them. Not only will it change it several orders of magnitude because of the phase change, but diffusion coefficients generally tend to increase as temperature increases as well; so as the beam provides thermal energy to the material, it also quickens the diffusion. [7] 14

31 In addition to increasing the rate of diffusion of two heterogeneous materials, another interesting phenomenon can occur when using lasers to aid in diffusion; the concentration of the doped substance can actually reach a higher concentration than found at steady state. This is due to the rapid heating and then cooling of the area under the laser beam and comes from the diffusion not reaching steady state thermal equilibrium before the melt pool recrystallizes, thus resulting in super-saturated phases. This can be desired in some cases and tends to be more dramatic when dealing with ultra-short pulse durations. One negative aspect of using lasers to speed up diffusion processes is that they generally have a tiny beam and therefore must be scanned over a large area, and it can be rather time consuming and energy costly. 15

32 Introduction to Photovoltaics Photovoltaics refers to the direct conversion of electromagnetic radiation contained in sunlight into electrical energy. Though this definition includes the entire electromagnetic spectrum, the term generally refers to wavelengths near or in the optical regime. At first, the focus of photovoltaics was used exclusively in non-terrestrial (space) applications, but due to an increased demand in renewable energies and inevitable depletion of non-renewable energy supplies, the photovoltaic energy source has become quite prevalent in today s world. [8] Unfortunately, the photovoltaic solar cell industry is plagued by high manufacturing costs and low conversion efficiencies. Only a few decades ago solar cells were dismissed as a non-viable technology because the energy and costs required to manufacture the cell were greater than the energy supplied by the technology over its lifetime. [9] Most current research is focused on improving the conversion efficiency, lowering production costs, and increasing module lifetimes. In the past few years with the aid of government programs, particularly in Germany, Japan and Spain, the photovoltaic industry has experienced large growth. Between 2005 and 2008, worldwide production volume increased between 40-70% every year. [10] Although the solar cell industry itself is growing quite rapidly, it is only slightly impacting the overall global energy consumption. According to the 2011 Global Energy Statistical Yearbook, the total global energy consumed was about 1.52 x 10 8 GW; out of this, only 67 GW were by the means of solar energy. History of Photovoltaic Power Generation In 1839, Edmond Becquerel discovered that when two different brass plates were immersed in a liquid and one of the plates was illuminated with sunlight, a continuous current was produced. [11] What he really discovered was the first copper-cuprous oxide thin film solar cell; however, it was not until over a century later, in 1954, that the first intentional solar cell was created at Bell Labs by Chapin 16

33 et al. [8] The cell was a simple p-n junction diode made of a single silicon crystal with 6% efficiency. [8] Over the following years, researchers were quickly able to bring this efficiency up to about 15%; however, the cost of a cell was hundreds of dollars per Watt. [11] Solar Cell Device Physics Basic Photovoltaic Principles Photovoltaic energy conversion is defined as the direct production of electrical energy in the form of current and voltage from electromagnetic energy. [12] Every photovoltaic device can be broken down into four major steps. The first is that a light absorption process must occur which causes a transition in the absorber material from a ground state to an excited state. This excited state can be known as an exciton and is essentially a semi-bound electron and hole pair which is consequently not free for movement. Second, the excited state must be separated into a free negative- and a free positive-charge carrier pair. In this case, the negative charge carrier must be brought up to the conduction energy band while the positive-carrier is brought down to the valance energy band. Thirdly, there must be some transport mechanism which allows for separation of the carriers. This can be done through either the presence of an electric field, the presence of an effective force field, or through diffusion. [12] When the charge carriers are being separated, the negatively charged electrons will move in one direction to a contact called the cathode while the positively charged holes will move in the opposite charge to the anode contact. Once at their respective contact points, the electrons will travel through some type of external path or electrical circuit until they reach their electronic load where they use their energy. In the fourth and final step of any photovoltaic process, the electrons continue through the circuit and arrive at the anode where they combine with an arriving positive-charge carrier, thereby returning the absorber to the ground state. 17

34 As previously stated, charge carriers can be separated through either the presence of an electric field, the presence of an effective force field, or through diffusion; however, not all of these separation methods are equal. From basic theory as well as computer model analysis, it has been found that built-in electric fields and built-in effective force fields are the principal sources of photovoltaic action. [12] Diffusion by itself does not give rise to much photovoltaic action simply because it does not make one direction inherently different from the other; only the relatively minor difference in diffusion constants of electrons and holes yields a second-order PV effected called Dember potential. However, when diffusion is combined with another form of separation, it can be quite instrumental in moving carriers to their respective metal contact locations. Basic Semiconductor Terminology and the p-n Junction Diode Semiconductor materials have a bandgap energy which is the difference in energy levels between the lowest energy band of the conduction band and the highest energy level of the valance band. The energy gap is fixed for a given material and is only slightly dependent on temperature. For silicon, this value is 1.12 ev at 25 C. Semiconductors can also be broken down into two general types: direct and indirect bandgap. These terms refer to the interaction between the incoming light and the semiconductor itself. If the semiconductor is direct gap, absorption can occur with just a photon whereas indirect bandgaps require both a photon and a phonon to bring an atom to an excited state. A phonon is a quantized unit of energy related to lattice vibrations. Because of this additional interaction with a phonon, indirect bandgap materials have a significantly reduced absorption probability compared to direct bandgap materials. Silicon is an indirect bandgap material therefore when using silicon solar cells, the materials must be relatively thick to ensure that the probability of a phonon interaction due to the increased lattice size. 18

35 The p-n junction is the simplest case of a diode and is an essential element of countless electronic devices including solar cells. The n-type region corresponds to the region doped with donor impurities which leads to an increased density of electrons, whereas the p-type region is doped with acceptor atoms leading to an increased density of holes. In solar cells this is typically done thermally in a furnace exploiting diffusion properties by introducing a gas containing the opposing doping atoms. When these two types of semiconductors come into contact, a space charge region occurs where acceptors and donors diffuse into each other in an effort to maintain an equal Fermi energy level throughout the cell. Electrons diffuse from the n-type to the p-type region whereas holes diffuse from p- to n-. A potential energy barrier or internal electric field then arises to produce the compensating electron and hole drift currents needed to maintain thermal equilibrium. The potential barrier or builtin voltage is the difference between the work function of the p-type semiconductor and the work function of the n-type semiconductor. This difference is always smaller than and for silicon solar cells, it is generally around 1 V. Figure 6: The energy band diagram for a p-n junction 19

36 Figure 7: Diagram of a p-n junction which depicts the electric field and the depletion width: The internal electric field is strictly limited to the depletion layer in the p-n diode and it is this field which allows free charge carrier atoms to drift and separates holes to one side of the cells while electrons go to the opposite. Because it is governed by the built-in voltage, the higher the built-in voltage, the stronger the electric field thereby increasing separation qualities. Solar Spectrum and Absorption Earth is constantly being bombarded by a supply of energy from the sun in the form of photons; the issue becomes utilizing the maximum amount of this solar energy possible. In space, the sun can be modeled as a blackbody emitting radiation at 6000 K. This provides a fairly standard solar spectrum. By the time this output from the sun has reached the Earth s surface; however, it has been scattered and absorbed in various ways which depend on variables such as latitude, time of day and atmospheric conditions. [12] To factor in all of these variables, scientists have developed a standard known as the air mass (AM) value. This value accounts for the solar spectrum modified by Rayleigh scattering, scattering by 20

37 aerosols and absorption by constituent gases of the atmosphere. Typically UV wavelengths are absorbed by the ozone while infrared absorption occurs by water vapor. [10] Air Mass 1 is defined for normal incidence of the sun at the zenith, while for other inclinations, the AM coefficient is found by simple geometry from the way the sunlight hits the earth s surface, with the following equation: Equation 5: Air mass calculation for solar spectrum where is the angle between the solar ray and a line perpendicular to the Earth s surface. Thus, AM1 corresponds to the solar spectrum when the sun is directly overhead. AM0 corresponds to the solar spectrum just outside the atmosphere and is used for space solar applications. The solar spectrum is shown below at various AMU values and the corresponding incidence of light is also shown: Figure 8: Air Mass schematic [10] 21

38 Figure 9: Spectral irradiance vs. wavelength [10] Setting up a standard solar spectrum was necessary when comparing solar cells in both research and industrial settings. Therefore, scientists have chosen AM1.5 as the standard value for testing because it corresponds to a reasonable average value of air mass through which most incoming sunlight passes. It is known that if a photon has greater incoming energy than the energy bandgap of a semiconductor, then the photon can be absorbed and if the photon energy is significantly higher than the bandgap energy it will lose the excess energy above the bandgap due to thermalization. Taking this into consideration, one can determine that there is indeed an optimal energy conversion relationship between semiconductor bandgap and the energy that can be extracted when only considering bandgap and the solar spectrum energy. It is compared through spectral conversion efficiency. 22

39 Equation 6: Spectral conversion efficiency equation Here, is the spectral conversion efficiency, is the number of photons per unit area absorbed per unit time, is the energy bandgap of the material and is the solar power density (W/cm 2 ). Using this relationship, we can determine the range of bandgaps that will yield the maximal spectral conversion efficiency. What is discovered is that the maximal values for are close to 50% and are obtained for bandgaps in the range of 0.8 ev to 1.5 ev. [10] Therefore semiconductors such as silicon which fall in the middle of this spectrum are ideal. In fact, silicon s spectral conversion efficiency is around 48%. [10] One should note though, that this limit does not apply as even the theoretical upper limit to the overall conversion efficiency because there are other loses throughout the cell, particularly in the separation of the charge carriers. These values are assuming only simple p-n junction solar cell. Optical absorption in a material is characterized by the Beer-Lambert law: Equation 7: The Beer-Lambert Law where ( ) is the intensity at a distance into the material, is the initial intensity and is the socalled absorption coefficient. At, ( ), thus implying 63% of the light has been absorbed by the sample. Hence serves as a crude measure of the sample thickness needed for substantial absorption, and is called the absorption length. In practice, one would desire near-100% absorption and this would require a thickness approximately 3-5 times the absorption length. The absorption coefficient α is a function of wavelength λ, so absorption is not only nonuniform, it will also be strongly dependent on the wavelength of light under consideration. 23

40 Another factor to take into consideration when dealing with solar cells is reflection at the surface. Silicon has a relatively high refractive index ( 3.5) and so reflects 30-35% of incident light. [13] To combat this, anti-reflection coatings are placed on the front side of the cells. Several different types of coatings can be developed though one of the most common is SiN x. [14] Some make use of pyramidal texture which exploits multiple internal reflections. This allows the incoming light to bend and come into contact with the solar cell multiple times thereby increasing the chances of absorption. [15] [16] Another possible approach of reducing reflection consists of depositing inhomogeneous dielectric whose index decreases from the bulk to the ambient. With this, theoretically the reflection would be null; however, it is found that these materials generally have high absorption coefficients themselves. This means that the dielectric will absorb most of the photons before it has time to reach the surface of the solar cell. [17] Multi-layer antireflective coatings have also been adopted and shown extremely promising results in laboratory scale. They utilize different materials with different refractive indexes to reduce the total reflectance over the entire solar spectral range. The major downfall of these types of coatings is that they add to production costs. [18] Current and Voltage Characteristics Put simply, a solar cell is always just a diode. It is designed to separate charge carriers in as efficient method as possible and the manner in which it leaves is known as current. When no light is applied, the cell is said to be in the dark condition and the current density flowing through the device is given by the following equation: Equation 8: Dark current density for an ideal diode 24

41 Here, is the dark current density, is the reverse saturation current density, is the magnitude of electron charge, is the applied voltage, is the Boltzmann constant and is the temperature. However, as soon as light with enough energy hits the solar cell, a photo-generated current density,, will be superimposed on the dark current density. The above equation then becomes: Equation 9: The illumination current density for an ideal solar cell where is the current density under illumination. There are a few important things to note, the first is that this is the equation of an ideal solar cell, therefore recombination, resistances, and other loses are not taken into account. Also, there is now a negative sign in front of the dark current density. This is because the photo-generated current and the diode forward current flow in the opposite direction. Finally, this equation assumes that both the photo-generated current and the dark current of the diode can simply be additively superimposed, without any additional terms. [19] [20] The short circuit current density corresponds to the case where the voltage across the cell is zero or zero load resistance; hence, from the equation above,. Essentially, the electrical current delivered by the cell directly depends on the quantity of solar photons absorbed by the semiconductor. Therefore, if one were to find a theoretical maximum of it would be as follows: Equation 10: Maximum short circuit current density for a solar cell where is the number of electron-hole pairs generated per unit time. This value depends on several factors including crystal quality and the bandgap of the semiconductor. It can be seen easily that as the 25

42 bandgap energy decreases, the short circuit current density will increase. This is because more photons will have enough energy to create free electron-hole pairs. For crystalline silicon, the maximum short current density is approximately 42.5mA/cm 2. [10] The short current density is only slightly affected by any temperature changes. Because the bandgap of a semiconductor usually decreases slightly as temperature increases, a larger number of photons can be absorbed by the semiconductor. However, this effect is rather small and therefore the increase of is almost negligible. In contrast to, the open-circuit voltage,, is the other extreme when the load resistance is infinity or open circuit. This value is found by setting cell current density to zero and is given by the following expression: Equation 11: Open-circuit voltage equation Since, this equation is approximated to: Equation 12: Open-circuit voltage approximation equation As seen from this equation depends on ; i.e. the illumination level of the solar cell as well as the reverse saturation current density. Quantitatively, if the incoming light is decreased by a factor of 10, the will decrease by roughly 60 mv. Temperature is another factor that affects and can be shown in two ways. First, in the pre-factor of, but secondly by, which has a very strong temperature dependence. As a net result, it can be shown that increasing temperature will decrease. 26

43 The ideality factor of the cell will also have influence on the open-circuit voltage, though it is more complex because of simultaneous changes in. Just as has a maximal value (set by absorption), so does. This fundamental thermodynamic limit was first given by Schockley and Queisser and later by Keiss and Rehwals. [21] [22] The latter derived the following expression: Equation 13: The fundamental thermodynamic limit to open-circuit voltage as discovered by Keiss and Rehwals [22] Here, is Plank s constant and is the speed of light. Since this is the fundamental maximum, it has assumed an ideal diode, with. For crystalline silicon, this fundamental maximum value is around 680 mv. [10] Power, Fill Factor, and Efficiency The electrical output power per area produced by a solar cell, is found by multiplying the current density by the operating voltage, or Equation 14: The electrical output power of a solar cell Taking into consideration that both the current density and the voltage have maximum values, one can therefore estimate the upper bound of output power as: Equation 15: The theoretical maximum electrical output power of a solar cell 27

44 However, this maximum value can never be achieved because a cell cannot simultaneously be at short circuit and open voltage. Instead the maximum power is achieved slightly below and for an optimal value of load resistance. Therefore, a term known as fill factor ( ), has been developed to show how close the actual output power relates to the product: Equation 16: The fill factor equation Because the fill factor equation attempts to achieve the highest possible power ratio, there exists a maximum power point for both the current density,, and the voltage,. The higher the fill factor, the more the current-voltage curve approaches the ideal case. Non-idealities such as series resistance and shunt conductance cause further reductions in FF. Figure 10: Typical solar cell current density vs. voltage characteristics In practice, the load resistance is found either by using electronic circuits called maximum power point trackers, or in a much more rudimentary manner, it is often done by selecting a fixed load 28

45 that more or less keeps the solar module near maximum power output for usual illumination conditions. Fill factor values of between 0.7 and 0.85 are achieved for a solar cell of good quality. For crystalline silicon, the theoretical limit for fill factor is approximately [10] The efficiency of a cell is determined by taking its output power divided by the total power into the cell. When operating at its maximum power point, we find that Equation 17: Solar cell efficiency at the maximum power point where is the energy conversion efficiency of the solar cell. Cell temperature has a strong influence on efficiency. As previously stated, increasing temperature lowers the open-circuit voltage which in turn lowers power. It has been found that when operating a solar cell at 80 C compared to standard test conditions at 25 C, the efficiency of the cell can be about 20 to 30% lower. [10] For silicon solar cells, a red shift in the spectrum of incoming light will lead to an increase in efficiency whereas a blue shift will lead to a decrease. Red shifts occur near dusk and dawn whereas blue shifts can occur near large bodies of water surfaces. Taking into account all of the limitations of silicon, the maximal achievable efficiency under AM1.5 lies somewhere in the range of 26.5% to 29%. At this point, there have been laboratory scale solar cells which have achieved efficiencies as high as 25% which indicates that crystalline silicon solar cells are becoming a very mature technology. [23] 29

46 Resistance Values in Solar Cells The ideal representation of a solar cell is simply a diode, but in practice there are additional losses due to series resistance and shunt resistances. The universally accepted equivalent electrical circuit of a solar cell device is shown below with the addition of series resistance,, and the shunt resistance denoted Figure 11: Universally accepted equivalent electrical circuit of a solar cell As seen, the series resistance lies in series with the voltage of the cell whereas the shunt resistance lies parallel to the diode providing an alternative path for current to travel through. Series resistance degrades device performance and is always present in devices. Resistance values as low as a few ohms-cm 2 are detrimental for a solar cell. It can come from many different sources within the cell but the major contributors come from the semiconductor resistivity, the contact resistance and on geometric factors. [24] Because the series resistance is in series with the voltage developed by the cell, it causes a drop in the total available voltage. The effect of this drop is most prevalent at high light intensities and is modeled below: 30

47 Figure 12: The effect of series resistance on the J-V characteristics of a solar cell [10] As seen, the fill factor also dramatically decreases as series resistance increases. It has been found that the fill factor of a solar cell decreases by about 2.5% for each 0.1 Ω increase in series resistance. [25] Studies have also shown that series resistance is dependent upon the operating current of the cell. They have shown that the fill factor can be lowered even more as the current increases. [25] There are several ways to measure the series resistance within a cell; some involve dynamic approaches, some under dark current, some under illumination. One such technique is to measure the deviation of the experimental log dark-current-voltage curve from the extrapolated straight line. [24] This is depicted below: 31

48 Figure 13: Ln(I) vs. V for a diode with series resistance A more accurate series resistance is obtained from the below expression: Equation 18: An accurate method for determining series resistance within a diode Here, one can see that taking a plot of vs., the equation should give a straight line with the intercept equal to and the slope being equal to the series resistance. The other type of resistance in a solar cell is known as shunt resistance and is a conductive path in parallel with the diode of the solar cell which allows for a leakage of current through it. Since current leaves through this path, it does not flow into the load and hence does not contribute to the power output of a cell. In the universal equivalent electrical circuit of a solar cell, the shunt resistance, shown as, not only designates actual ohmic shunts through the material, but it is also meant to symbolize 32

49 recombination losses within the solar cell. Physically, this is not a satisfactory way of representing recombination because recombination is dependent upon the photocurrent,, however if one assumes that the light intensity does not vary dramatically, then holding constant is a perfectly acceptable approximation. To examine the effect that the shunt resistance has on a solar cell, one can use the following figure to model the J-V characteristics. Figure 14: The effect of shunt resistance on the J-V characteristics of a solar cell [10] As can be seen from the figure, when there is an infinite value of the shunt resistance, the output power of the cell most closely replicates the ideal rectangle. Conversely, as the shunt resistance decreases, the J-V curve becomes more linear which decreases the fill factor of the cell and consequently the output power. The shunt resistance of a solar cell is a parameter which strongly depends on the fabrication process and structure of the cell. Therefore, even within the same material there exists a large 33

50 fluctuation in values. As a general rule, higher quality silicon achieves the desired high shunt resistance values because it has fewer conductive paths through it. It is also known that the effect of the shunt resistance increases at lower intensities of light. Transport Processes at Metal-Contact Interfaces Metals contacts are typically used on solar cells to collect the generated charge carriers; however, the transport phenomenon that occurs at these interfaces is dramatically different from that in the bulk of the material and can have a large impact on overall performance depending upon the cell design. The classical process of transport is through thermionic emission, where the carriers transfer from an allowed state in one material to an allowed state in another with no change in total energy. [12] These ohmic contacts are ideal for solar cells because they provide a practical way to transport charge carriers without encountering unwanted potential barriers. Metal Contacts Not all of the metal-surface transport processes are beneficial to cell efficiency. Interface recombination tends to play a large role in decreasing overall cell efficiency. When the semiconductor, such as silicon, comes to an abrupt end at the interface, defects and localized states occur. These states trap charge carriers and essentially work as recombination-ladders for the photogenerated carriers in the cell. To help combat this recombination, a special type of ohmic contact the selective ohmic contact, discussed later is desired at this interface. First considering the normal ohmic contact, it is realized on p-type Si by depositing a metal with a large work function. This allows for banding bending to occur which promotes majority charge mobility while hindering minority charge carrier. The band diagram of that situation is shown below: 34

51 Figure 15: Ohmic metal-contact to p-type semiconductor As can be seen from this structure, if the metal has a work function that is greater than the work function of the p-type material,, a dipole develops which encourages the majority charge carriers (the holes) to move towards the anode. If the opposite scenario is true, the dipole is oriented in the opposite direction. This creates a potential energy barrier for the majority charge carriers and can decrease cell efficiency. This situation is shown below: Figure 16: Non-ohmic metal contact to p-type semiconductor Using this same logic, on the n-type side of the material a low work function metal is desired so that it too, creates a dipole oriented to help the majority charge carriers to migrate. In this case, 35

52 . Obviously though the metal made into contact cannot solely depend on the work function. Other factors need to be taken into consideration as well such as Fermi level pinning, chemical reactions, cost, diffusion, or damage may necessitate using another approaches. If suitable metals are not available, an alternative approach is to develop a potential barrier for the majority carriers, but one makes it ultrathin with heavy doping. Carriers flow by tunneling thus producing an ohmic contact. [12] In silicon solar cells, aluminum is typically used as the p-type contact material and a silveraluminum paste is used for the n-type contact. The work function of silver is 4.73 ev while aluminum is 4.08 ev. [26] Cost is a major factor to consider when selecting practical metal contacts for industrial purposes along with manufacturability. Additionally, laser fired contacts are generally used for the back contact for p-type wafers which uses back surface field effects. Selective Ohmic Contacts Selective ohmic contacts are of particular importance in the photovoltaic industry. They are defined as a contact which passes only one type of charge carrier with ideally no voltage drop. [12] This helps with photocarrier collection since the photogenerated minority carriers are now deflected away from the ohmic contact towards the photovoltaic junction. These selective ohmic contacts can be made with a semiconductor heterointerface or by altering doping concentration. For the former, one must utilize a material with a larger bandgap than the original semiconductor. This new material is placed between the semiconductor and the metal contact. Then, if proper doping levels occur in the new material and the bandgap lines up appropriately, it can form a potential barrier for one of the charge carriers (the photogenerated minority carriers, in a solar cell) to prevent it from leaving at that contact point. This is shown schematically below: 36

53 Figure 17: Selective ohmic contact to anode which utilizes a material with a new bandgap As shown from this figure, minority electrons which are trying to exit the anode will encounter a large potential barrier blockage which will prevent them from exiting the contact or recombining at the edge. This technique needs to have another material added onto the photovoltaic device which increases the cost of production. However, there are alternative methods to create a selective ohmic contact that do not require the addition of other materials. These methods utilize effective forces and involve altering the dopant concentration of the semiconductor at location of contact with the metal. For example, to increase cell efficiency it is desirable to prevent holes from entering the cathode. To do this, increasing the number of acceptors within the semiconductor, one can create an effective force which causes a potential barrier to be encountered for the holes. This is shown schematically below: 37

54 Figure 18: Selective ohmic contact at the cathode which utilizes effective forces There are several advantages to this approach, the major being that only one material is needed so there is less total cost to cell production. Another is that while it creates an effective force to prevent hole exit, it simultaneously creates an effective force which encourages electrons to leave. However, one disadvantage to using effective forces compared to using a larger bandgap material is that effective forces do not repel minority charge carriers as effectively. [12] Another disadvantage is that if the depletion width is too thin, tunneling may occur which would be an undesired result, though this is more important when dealing with electrons than holes. Solar Cell Characterization A p-n junction solar cell exhibits diode behavior by design which translates to rectifying currentvoltage characteristics. Traditionally, these characteristics can be broken down into two distinct portions: a quasi-neutral region (QNR) and a space charge region (SCR). These two regions are shown below on a log current vs. voltage plot. 38

55 Figure 19: Current-voltage characteristics of a two-region diode [24] As can be seen, the SCR dominates the current at low voltages whereas the QNR is more prevalent at higher voltages. The bending off of the curve at high voltages indicates that series resistance is becoming the dominant factor controlling the current. The above plot is taken under dark current conditions; however, taking into account illumination current is of much more practical use for applications. Adding this term into the diode allows for the following equation to represent the current flowing through the diode at any given voltage. Equation 19: Current for a two-region diode under photon illumination In this equation, is the current traveling through the diode, is the voltage, is the photon current from illumination, and are the saturation currents of the QNR and SCR respectively, 39

56 and are the ideality factors of the QNR and SCR, is the series resistance, is the shunt resistance, is the Boltzmann constant, and is the temperature of the diode. Frequently, the saturation current, ideality factor, and leakage current of a cell are used to quantitatively determine the quality of a diode. To determine the saturation current and ideality factor for the quasi-neutral and space charge regions, a linear equation should be fit to the distinct sections on the log current vs. voltage plot. This is shown on the previous figure depicting a diode. For each region the slope can be used to calculate the ideality factor by the following equation: Equation 20: Ideality factor for a diode In this equation, refers to the value of the slope for the linear portion. Traditionally, the ideality factor of the QNR should be equal to 1 while the ideality of the SCR is approximately equal to 2. The y-intercept of the linear line for each segment corresponds to the saturation current. Low saturation currents indicate a high quality diode. The final term traditionally used to determine the quality of the diode is the leakage current. Again, this is desired to be as low as possible. For solar cell applications however, this term has limited importance because the cell is only designed to work in one direction, therefore leakage current does not come into play. For solar cells, the current equation for a diode is simplified to follow the universal electrical circuit equivalent of a solar cell. Under this simplification, the SCR is eliminated from the current equation because it does not have a dominating influence in the diode qualities, and in most cases, it is not easily identified on a graph as it blends in to the QNR. The equation becomes as follows: 40

57 Equation 21: Current equation for the universal electrical circuit representation of a solar cell This equation also assumes that both the saturation current and the ideality factor are constant over the entire current-voltage curve though in reality they may vary at the extreme edges of the diode. Silicon Solar Cells As previously stated, silicon is an indirect bandgap semiconductor with an energy gap of 1.12 ev which corresponds to a wavelength of 1100 nm. Crystalline silicon technology has, to this point, dominated the photovoltaic market and will probably continue to do so in the near future. For example, in 2008, 88% of the cell and module production was based on crystalline silicon. [10] With such an increase in demand for silicon the past few decades, at first the industry was limited by the availability of purified poly-silicon. This limited supply of device quality silicon has triggered interest in many other thin-film technologies though they are still somewhat new technologies. In silicon solar cells, boron is the normal dopant for p-type semiconductors whereas phosphorous is common for n-type semiconductors. Generally, the p-n junction is created through thermal diffusion. The cell processing step is extremely crucial to the cell as it has a large correlation to efficiency of the module which directly corresponds to cost. Typical industrial cell efficiencies are around 15-16% for multi-crystalline silicon cells and around 16-17% for mono-silicon solar cells. Upon adding these to a module, these efficiencies are lowered to around 13-15%. [10] The major hindrance to the growth of silicon photovoltaic industry is cost. At this point, solar cells are not the most competitive sources of electrical energy. Government incentive programs have helped to increase the deployment of photovoltaics in several areas and there is still a great potential 41

58 for cost reduction. [9] Several studies show that crystalline silicon technology has potential to reach production costs of $1.20-$1.50/Watt within the next five years. [27] 42

59 Laser Processing in the Photovoltaic Industry To date, laser processing has been used in a large variety of applications within the solar cell industry. Lasers have been used as a cutting method to remove the edges of the completed cell, they have been used to scribe into cells, and they have even been used as a drilling tool for some solar cell designs such as the emitter wrap-through cell. They can also be used to sinter contact metal onto the cell as well. Some of the most important applications of lasers are through the creation of laser fired contacts, using lasers to dope materials, and using lasers to create selective emitters. These three applications are described in detail below along with information regarding laser-induced damage that may be a disadvantage to their intended processes. Laser Fired Contacts Laser fired contacts hold much potential in today s industry to increase solar cell efficiencies while simultaneously simplifying the manufacturing process. They serve as a way to obtain rear side electrical contact while concurrently creating a back surface field or selective ohmic contact. Traditional methods of creating the back metal contact involve depositing a thin aluminum layer and then annealing at high temperatures to allow the aluminum to diffuse into the silicon which creates a back surface field which repels minority carriers by utilizing a potential barrier. [4] As the industry is moving towards thinner, cost saving wafers, this traditional approach is reaching its limitations. The thin wafers are much more susceptible to warping during the annealing stage which leads to wafer breaking in subsequent processing steps. Additionally, a lower percentage of photons are absorbed because the thickness of the wafer is smaller. To combat this, dielectric layers are sometimes desired which allow for reflectance of the photons back into the silicon wafer. Additionally, such a layer would provide passivation to the substrate thereby removing many dangling bonds at the cut surface of the wafer and dramatically decreasing recombination. 43

60 Laser fired contacts provide a way to eliminate both of the major problems associated with thin wafers. In this design, a dielectric layer is sandwiched between the back surface of the silicon and an aluminum layer. Then, laser pulses penetrate through the aluminum and dielectric layer to come into contact with the back of the silicon, creating selective back electrodes. [4] To combat wafer warping, laser fired contacts only selectively heat tiny individual areas thereby eliminate bulk annealing. Additionally, the laser pulse can create an aluminum-silicon alloy which then generates the desired back surface field which repels minority charge carriers thereby eliminating any need for an annealing step. [4] The dielectric layer is the other key step in the benefits of laser fired contacts. In this design, there can still be a dielectric layer which promotes optical reflection and decreases carrier recombination at the back of the cell while still providing electrical conductivity through the aluminum. Cells with laser fired contacts have exceeded 20% efficiency and have the capability of reducing several processing steps in the manufacturing line. [4] Laser Doping Laser doping has been used to make a multitude of electronic devices, from light emitting diodes to solar cells to silicon carbide power devices. In the microelectronics industry, it is being applied to create ultra-shallow metal-oxide-semiconductor field-effect transistor source and drain where other techniques such as ion implantation are reaching their limits. [28] It is also used to create complementary metal-oxide-semiconductor and microelectromechanical systems devices leading to smaller chip sizes and lowering the cost of mass production. Laser doping has also been used to form a three-dimensional structure of silicon in GaAs in the electronic industry. [29] 44

61 The core principle behind laser doping lies in the fact that the diffusion coefficient in a liquid is several orders of magnitudes higher than that in the solid state. The laser pulse selectively melts a small area allowing for selective diffusion to take place. The concentration, depth and to some extent the distribution of the dopant may be controlled by altering the laser energy and precursor dopant source. [30] It is also important to qualitatively compare laser doping to other doping techniques. Thermal diffusion is a very common doping technique where dopants are introduced at the surface of the wafer in a high temperature environment thus greatly enhancing diffusion. This technique allows for precise control of the diffusion depth but it is always limited by the solid solubility limits. Unfortunately, the high temperatures may also lead to impurity contamination, sublimation etching of the surface, degradation of minority carrier lifetimes in the substrate, and deterioration of the lattice structure of the substrate. [31] In the creation of solar cell, the thermal doping is one of the most costly steps in the entire solar cell production process. Ion implantation is another common technique where doped ions are bombarded onto the surface of the substrate and imbed themselves within the material. Using this technique, the dopant concentrations can actually exceed the solubility limit, but it comes with a price. Because the atoms crash and imbed themselves within the substrate, severe lattice damage occurs which is detrimental to many devices including solar cells. To combat this, it is common to thermally anneal the wafer after ion implantation to help remove the defects. Lasers can be used in this stage to selectively melt the surface and allow for it to recrystallize epitaxially. [32] During this annealing process, the implanted dopant atoms are redistributed allowing for a junction to be formed. 45

62 A final alternative doping technique is in situ epilayer doping which allows for doping to occur during the wafer creation phase. However, this is also a limited approach because it cannot exceed the solubility limits. In addition, it has been known to reduce wafer yield by altering the planarity of the substrate. [33] Additionally, it is extremely difficult to produce different dopant profiles for different devices on the same wafer using this approach. [34] Out of the aforementioned techniques for doping, thermal doping is the most popular used for creating a p-n junction in solar cells; however, laser doping may provide an elegant and cost-effective alternative to this technique. Already cells have been created with a full area shallow emitter by using exclusively laser doping. Eisele et al. reported 18.7% efficiency and a defect free p-n junction by using a pre-deposited phosphorus precursor layer in ambient atmosphere. [35] A 65 ns pulse, 532 nm Nd:YAG laser was used to melt the surface to a depth of about 1 µm. [35] An open circuit voltage of 377 mv and a short-circuit density of 34.0 ma/cm 2 were found from the resulting cell, leading to the conclusion that laser doping did not result in significant lattice defects. [35] Esturo-Breton et al. used a similar 532 nm Nd:YVO 4 laser with frequency of 100 khz and power less than 1 W to yield a 14.2% efficient cell. [36] This efficiency is lower because there were no laser fired contacts on this cell. Esturo-Breton used transmission electron microscopy to reveal that the crystalline structure of the silicon recrystallized free of dislocations and with the same orientation as the underlying layers. [36] Ogane et al. used an ultraviolet wavelength instead of the conventional 532 nm wavelength to selectively dope the emitter of silicon solar cells at room temperature in ambient conditions. [37] Despite the fact that the solar cells had comparable short circuit and open circuit characteristics to those 46

63 cells created via thermal diffusion, the group discovered that these cells exhibited high series resistance which deteriorated the fill factor. [37] It is also possible to dope shallow emitters using a narrow phosphoric acid jet. What happens in this scenario is the injected beam propagates through the acidic jet and grooves the surface of the silicon solar cell. [38] Subsequently, the laser beam provides energy to melt the surface allowing for the phosphorous to penetrate through the surface creating a diode. [38] This approach is rather costly and at this stage, not applicable to creating the full area shallow emitter in a solar cell. Hameiri investigated creating junctions from dopants contained in the silicon nitride antireflective coating layer. He discovered that a wide variety of laser energies could be used to create a diode. Diodes were not only created when the energy was low enough not to completely melt the pyramidal structure on the top surface, but they were also created when the pulse energy was high enough to create ablation. [14] Cell performance worsened when ablation occurred, though there was a small range with minimal ablation that did not significantly degrade the cell quality. [14] The results showed that high laser pulse energies lowered the open circuit voltage and the fill factor. [14] Sheet resistivity was also reduced with increasing energy up to a threshold energy level. Laser Doped Selective Emitters Laser doped selective emitters (LDSEs) utilize lasers to create highly doped areas directly under the metal contact fingers, thereby providing ohmic contact and reducing recombination. In a conventional cell, a trade-off between conductivity and carrier lifetime is a necessary choice. Shallower doping profiles allow for longer lifetimes, but result in decreased conductivity. [39] LDSEs provide a way to utilize the best of both situations and allow for long carrier lifetimes throughout the absorber bulk, while having high conductivity directly below the metal contact fingers where a majority of the current must flow. LDSEs can also decrease the saturation current which results in an increased open circuit 47

64 voltage. [40] Studies have shown that when properly utilizing LDSEs, cell efficiency can be increased by % absolute. [41] [42] There are currently several problems that are hindering LDSEs from being incorporated into industrial manufacturing. The main hindrance is the difficulty of aligning the metal contact fingers to the LDSE with the currently used, screen printing methods. One of the proposed solutions is to make the metal contact fingers perpendicular to the LDSE lines, resulting in a 0.3% absolute efficiency gain. [43] This eliminates the need to line up the metal contacts while still obtaining a portion of the beneficial efficiency gain. Additional problems include a decrease in demand due to the development of new, better conducting, Ag screen-printing pastes. [44] Economic studies have shown though, that by combining commercial lasers and stat-of-the-art metallization processes, LDSEs would be competitive in the market if two machines are working in parallel in the manufacturing line. [45] LDSEs have been created using both pulsed and continuous wave lasers ranging from UV to IR wavelengths. Wavelengths in the green visual spectrum range have proven to be the most promising due to the 1 µm penetration depth which is comparable to the depth of the shallow emitter. [39] IR lasers have a much larger penetration depth which leads to increased bulk heating and thus requiring power levels of about four times larger than the visible spectrum lasers. [46] The governing parameters for the creation of LDSEs are the power and fluence used, though the pulse length is also an important consideration. The pulse length is typically between 50 ns and 300 ns where the shorter pulses correspond to using lower laser power levels due to less diffusion losses in the wafer. [46] Conversely, longer pulses seem to reduce surface and lattice damage. However, with longer pulses undesirable evaporation of the Si occurs. [39] Additionally, longer pulses have been shown to allow for a much larger power processing width for lasers and a decrease of sheet resistance. [45] The 48

65 pulse duration has a minor influence on the total amount of active dopants and it has little to no influence on the emitter depth. [45] The concentration profile of the selective emitter is primarily controlled by the type of beam. Gaussian beams have a much higher concentration in the center whereas top hat beams have a uniform distribution over the entire area. In addition, the texture of the wafer has a large influence on the creation of the LDSE. [47] This is due to different coupling conditions for the laser radiation and the differences in heat conduction due to the texture. [47] An attractive method for the creation of LDSEs uses a phosphorus doped spin-on glass (PSG) film. This method requires only one additional processing step as it uses the left-over PSG from the thermal diffusion step as the doping source. After laser processing, the PSG layer is etched away and the rest of the processing steps are the same. The only additional problem is aligning the metallization pattern with the LDSE. Kohler et al. used the PSG method with a 532 nm, 65 ns pulsed top hat focused laser beam to form the selective emitter. [42] They observed a 0.5% absolute efficiency gain with a maximum efficiency of 18.1%. [42] The increase in efficiency is believed to result from reduced Auger recombination and better surface passivation. [42] Compared to the control cell, in the LDSE cell, both the short circuit and the open voltage increased while the fill factor slightly decreased. The decrease in fill factor is believed to be due to the non-optimized metal grid design. Kohler also calculated that with the PSG method once higher power visible spectrum lasers are available it would only require an additional 1.3 seconds to process a 6 inch wafer. [42] Besi-Vetrella et al. used the concept of spin-on phosphorous dopants to make LDSEs with a 532 Q-switched Nd:YAG laser with fluence close to 2.4 J/cm 2. [48] They observed no change in shunt 49

66 resistance, saturation current or diode ideality factor indicating that in this fluence regime no damage occurred to the cell. [48] What is of particular interest is that they were able to devise a system to combine screen printing and LDSEs. They used reference markers, a digitizing camera, and customized software that acquires both image and marker positioning to achieve screen printing with an error of less than 10 µm. [48] Another promising method of creating selective emitters involves using the antireflection coating (ARC) as a mask for the metallization. In this method, the ARC is placed on top of the PSG layer, prior to the creation of selective emitters. Then, a laser patterns the cell with fluences high enough to ablate the anti-reflection coating, while simultaneously doping the underlying area with phosphorus. The cell is then metallized using light-induced electroplating. The major advantage of this process is that it allows for a self-aligned metallization scheme. Additionally, it allows for the formation of lines as small as 20 µm which is significantly below the tolerances for screen printing. [46] As a result, fingers can be spaced more closely together while still decreasing shadow losses, thereby increasing short-circuit densities. However, adhesion between the metal and the silicon has been an issue for cells processed by this method. Researchers at the University of New South Wales used this self-aligning method with a 532 nm, Gaussian, Q-switched Nd:YVO 4 laser to create the selective emitters. [49] They used SiN x deposited by plasma-enhanced chemical vapor deposition and used nickel and copper light-induced plating to form the front contact grid. An efficiency of 18.7% was observed though there was a slight decrease in shortcircuit current density which can be explained by increased shading due to the variation in line width. [49] 50

67 As previously mentioned, with this self-aligning metallization pattern for creating selective emitters, adhesion between the metal and the semiconductor has been identified as a possible drawback. This problem can be avoided by increasing the roughness of the surface or creating holes in the laser-doped areas to act as anchor points for the metal. Again, the University of New South Wales group is pursuing this research using a 532 nm, Nd:YVO 4, Q-switched laser. [32] Using high enough powers, both the dielectric antireflective layer and a portion of the underlying silicon were ablated. This left for a rougher surface while still creating the desired p-n junction. The results suggested that intermediate levels of ablation can be used as a possible solution for the poor adhesion without significant loss of efficiency. [32] With regards to laser fluence, it was found that the selective emitter could be created with powers low enough to not remove the dielectric, as well as with powers which were high enough to cause significant ablation. With the highest powers though, shunts were observed; therefore, there appears to exist a middle regime which is desirable for creating the desired ablation, while minimizing shunts. [32] Lasers in the ultraviolet range have also been used to create LDSEs. Poulain et al. used a frequency tripled Nd:YAG laser with 355 nm wavelength and a pulse duration of 10 ns to create the selective emitters. [50] Because of the very strong absorption of silicon at this wavelength, localized thermal effects caused decomposition of the SiN x layer and activation of the dopants present in the emitter. [50] Beam modification has also been explored for the creation of LDSEs decreased overall processing times for cells. Line shaping and beam splitting in particular have been investigated. In one such research project, Jager et al. used a top hat line-shaped beam with dimensions of 15 µm x 200 µm with a 515 nm wavelength and 300 ns pulsed laser. [51] They used PSG to create the LDSE and observed no degradation of the material even at very high fluences. [51] Gaussian beams are undesirable when 51

68 using line-shaping devices because of the non-uniform doping profile. With the addition of line-shaping devices, less processing time would be needed to create selective emitters on the wafers. In addition to line shaping, research is also being done with a beam splitter capable of producing up to 70 top hat profiles. [51] Laser-Induced Damage Quite a bit of uncertainty exists with regard to how much damage occurs to a solar cell undergoing laser processing. The principal effect of laser-induced defect formation in solar cells is lowered carrier lifetime which decreases cell efficiency. Mitigating the formation and influence of these defects is essential when incorporating laser processing into the photovoltaic industry. The first full-area laser doped solar cells had considerably lower efficiencies, indicataing increased recombination from laser-induced defects. Mooney et al. observed electrically active (apparently point) defects with a Q-switched ruby laser process at depths up to 10 µm using deep level transient spectroscopy. [52] However, further research using a transmission electron microscopy did not reveal any defect clusters, dislocations, or stacking faults. [52] This leads to the conclusion that a defectfree epitaxial layer is grown from the melt. Further research showed that only crystals on a (1 0 0) orientated crystal grew defect free, whereas there was a high defect density on substrates with a (1 1 1) orientation. [52] These observations were independent of the type of laser used. Despite the fact that TEM images revealed no defects, a reduction in electrical performance of molten areas has still been observed and is most likely due to point defects. In solar cells, this translates to a reduction of the open-circuit voltage and slightly lower fill factors. One possible source for these point defects can be due to the rapid cooling the melt pool experiences. [53] Another possibility is an increase in contaminants. If the cell undergoes laser 52

69 processing in ambient air, contaminants from the air are more likely to penetrate into the material during the melting phase. This is confirmed by SIMS measurements which indicate a large incorporation of oxygen in these samples. [52] Processing in an inert gas can help combat the contamination and therefore increase electrical performance of the cell. Laser pulse energy has an influence on the number of defects generated. There appears to be two energy thresholds that affect the bulk lifetime of the carriers. The higher threshold corresponds to visible changes in the surface morphology of the wafer whereas at the lower threshold defects result without any visible changes to the wafer. [53] Arora and Dawar theorize that the root causes behind the surface damage produced from a laser pulse are thermally-induced stress and the generation of long-lived traps. [52] Because the thermal expansion coefficient of SiO 2 is one order of magnitude lower than that of silicon, and because silicon is particularly weak when subjected to tensile stress, the researchers concluded that with the presence of a thin oxide layer defects were caused at the surface by the lattice mismatch during the cooling phase. [52] In general, ablation is an undesired effect on solar cells because it causes splattering and decreases carrier lifetimes; however, some solar cell designs, such as the buried contact solar cell require ablation. With this design, a laser ablates the solar cell creating a groove in which metal is deposited to form the electrodes. Generally an infrared laser is used for this process due to the high absorption coefficient. In ablation, the vaporizing silicon ejects a volume of material which leaves surface defects. If left untreated, the defects can propagate through the material resulting in an oval-shaped edge dislocation formed deep into the cell. [54] This propagation process can be quickly expedited through thermal 53

70 cycling. Research has shown that the surface damage can be removed with a NaOH etch immediately after laser ablation. As etch time increases, surface damage decreases the density of defects. [54] For example, even after thermal cycling, Yang etches have revealed that there are no deep surface defects in the material caused from laser ablation if a 20 minute NaOH etch is implemented on the wafer immediately after laser processing. [54] Abbott et al. concluded that laser-induced defects do not cause shunting of the junction; however, the saturation current of the space-charge region was increased compared to samples without laser processing. [54] From this they concluded that there were laser-induced defects in the depletion region of the cell leading to a reduction in quantum efficiency. This is further confirmed by the decrease in blue-response for the cells undergoing laser processing indicating an increase in recombination at the front surface of the emitter. [54] 54

71 CHAPTER 3: DESIGN OF EXPERIMENTS: LASER DOPED SELECTIVE EMITTERS Goals and Objectives The goal of the laser doped selective emitter experiments is to determine the optimal laser processing conditions that could potentially be implemented in an industrial setting. The parameters varied were the laser wavelength, pulse duration, pulse energy and the thickness of the dopant layer. The selective emitters were then characterized electrically through current-voltage measurements. The ideal emitter would exhibit low series resistance, low saturation current, high shunt resistance, and an ideality factor close to unity. Additionally, surface morphologies would demonstrate limited ablation, indicating minimal damage to the crystal lattice structure. Laser Doped Selective Emitter Creation Wafer Creation The laser doped selective emitters (LDSEs) were all fabricated on 100 mm diameter, B-doped p- type FZ (Float Zone) silicon wafers of <100> orientation. The wafers were polished on the front side and etched on the back side with a thickness ranging between 255 and 305-µm and resistivity between 1 and 5 Ω-cm. Five different wafer stacks were made in order to investigate the influence of dopant layer thickness on the creation of LDSEs. The stacks contained the dopant layer (n-type a-si:h) was prepared by deposition in a Plasma Enhanced Chemical Vapor Deposition (PECVD) reactor and had the following profiles: 55

72 Figure 20: Passivation and dopant structures used in LDSE experiments Prior to deposition, the wafers were cleaned in an HF bath to remove any impurities and oxide layers that may have developed on the wafer over time. To do this, the wafer was placed in a 1:1 ratio of 49% HF and distilled water for 45-s and then rinsed in deionized water for an additional 45-s. At this point the surface of the wafer demonstrated hydrophobic behaviors characteristic of the absence of any SiO x layer. The wafers were then immediately placed under vacuum in a PECVD reactor located at the Nanofabrication facility, Materials Research Institute, Penn State University. The quick transfer ensured minimal impurity contamination and native oxide growth before the depositions. 56

73 The 10-nm a-si:h layer, was deposited at 120 C, with a pressure of 2 Torr, and power of 25 W, with SiH 4 and H 2 flowing at a rate of 33 sccm and 300 sccm respectively for 28 s. The doped amorphous silicon layer was deposited at 200 C, 2 Torr and 50 W with PH 3, SiH 4, and H 2 with flow rates of 15 sccm, 30 sccm, and 285 sccm, respectively. It was assumed that the deposition rate was linear and therefore to deposit 20 nm 120-nm, the deposition time ranged over s. The SiO x capping layer was also deposited in the PECVD at 250 C, 3.5 Torr, and 100 W power level. This deposition lasted 47.4 s to deposit 150-nm of SiO x with N 2 O, SiH 4, and N 2 flowing at rates of 1800 sccm, 18 sccm, and 1800 sccm respectively. To analyze the doping concentration profile, a sample was sent for Secondary-Ion Mass Spectroscopy (SIMS). The sample sent for analysis had the top oxide removed, leaving the underlying 60- nm n+a-si:h and 10-nm a-si:h on the wafer. The results are as follows: 57

74 Figure 21: SIMS profile for the deposited 60 nm n+a-si:h doped thin film The results show that the film has 1 x phosphorous atoms per cm 3 indicating that the layer is indeed n-type. It is noted that the phosphorous atoms tail off upon entering the sample. This is most likely due to an artifact present from the high H and P concentrations in the a-si layers and is assumed not to affect the sample in any way. Laser Firing Set-up and Processing Three different types of lasers were used to create the LDSEs. A Coherent AVIA-355 frequencytripled Nd:YVO 4 Q-switched laser with a 30-ns pulse was used for the 355-nm UV wavelength. A Quantel Brilliant Nd:YAG Q-switched laser with a 4-ns pulse was operated at both the fundamental (1064-nm) 58

75 and the frequency doubled (532-nm) wavelengths. Finally, an IPG Photons continuous wave fiber laser with a 1070-nm wavelength was used. This laser was used in conjunction with gates to create various pulse durations within the nanosecond regime. Both the Quantel and the IPG lasers used a train of optics to manipulate the beam and focus it along the axis. The Coherent laser used a GSI Lumonics 3-axis scanner which turned the beam perpendicular to its originating direction. The optic schematics for all three lasers are shown below: 59

76 Figure 22: Experimental optic schematic for lasers A three-axis Aerotech vacuum stage held the samples perpendicular to the incoming beam and the same stage was used for every laser setup. Focus spot size for each beam was determined via observations of burn marks and processed spots. The IPG is assumed to be 20 µm in diameter, the Quantel has a diameter of 40 µm when operated at 1064 nm, and 20 µm when operated at 532 nm; the AVIA has a 30 µm diameter. These values are used for the fluence calculations. Parameter Selections To test the various parameters, each stack was subjected to a variety of laser parameters. These parameters were tested on a 13 mm x 13 mm square sample with varying parameters along each axis on a 10 spot x 10 spot array. For the Q-switched lasers, the only parameter which was varied was the pulse energy. This allowed for duplicates to be made of each sample to ensure repeatable results. The AVIA laser had pulse range of µj/pulse, the Quantel at 1064-nm wavelength had a range of µj/pulse, and at the 532-nm wavelength, varied in the range of µj/pulse. 60

77 Figure 23: Parameter schematic for Q-switched lasers For the continuous wave laser, both the pulse duration ( µs) and the output power (4-125 W) were varied along the axis thereby creating a 2- dimensional array. 61

78 Figure 24: Parameter schematic for IPG Metallization Following laser processing, the LDSEs were metallized. This metallization ensured that good electrical contact was made to the diode and was especially important in the case diodes corresponding to the lowest fluences because of the pulse size. Because electrical measurements were desired for a single diode at a time, lithography was used to selectively metallize the top of the samples. The samples were first cleaned by a 30-s bath in acetone, followed by a 30-s bath in isopropanol, followed by a 30-s bath in distilled water. The samples were then dried in N 2 gas. Two layers were deposited via lithography. The bottom layer, the lift-off resist layer, was used to prevent the metal from coming into contact with itself along the edges and interfere with the 62

79 development stage. Because of this, the lift-off resist layer was designed to be at least 25% thicker than the thickness of the deposited metal. The second layer, #3012 photoresist blanket coating was used for the development process. The lift-off resist layer was spun on in a dynamic process at 3000 revolutions per minute for 45 s and then baked on a 180 C hotplate for 5 mins. Then, the second layer of the #3012 photoresist was spun on in another dynamic process of 4000 revolutions per minute for 45 s and then baked at 95 C for 60 s The samples were then placed under a photolithographic mask and developed under 400-nm UV wavelength. The LDSE samples were processed with the layout of 10 rows by 10 columns. Correspondingly, the mask used is shown below: Figure 25: 10 x 10 metallization mask layout for the laser doped selective emitters 63

80 In this design, there are 100 different metallization pads each with an area of 0.5 mm x 0.5 mm. The distance from the center to center of each pad is 1 mm. The sample was then placed in hard contact with the mask and developed under the ultraviolet radiation for 8 s. Upon finishing the mask process, the samples were then developed to allow for the lithography films to open up to subsequent metallization. To do this, the samples were placed in tetramethylammonium hydroxide (CD-26) for 90 s followed by a distilled water rinse for 30 s. Figure 26: Schematic of sample with lithography layers after the sample has been subjected to ultraviolet radiation and developed in CD-26 The samples were then cleaned using 30-s baths of acetone, isopropanol, and deionized water respectively. Then, the samples were etched with a 50% diluted concentration of 1:10 BOE for 3-5 s. This process ensured that the surface was clean for deposition, but still had some remaining oxide on the sample. Previous studies investigating the rate of oxide removal had found that after 6 s there would be no oxide remaining on the sample. This is an undesirable result because the oxide layer prevents shunting through the silicon and is essential when trying to measure the electrical characteristics of the diode. 64

81 The samples were coated in nickel in a Denton sputtering machine. The nickel was deposited at 5 mtorr at a deposition rate of approximately 1.7 Å/s with a total thickness of 100 nm. Following deposition, the samples were placed in n-methyl pyrrolidinone (Remover PG) to remove the photoresist and reveal the final metallized pattern. This was again followed by 30-s baths in isopropanol and deionized water. To aid in the electrical contact between the nickel and the silicon, the samples were annealed using a rapid thermal anneal at 450 C for 15 min. using argon as the ambient gas. The samples then had their electrical characteristics measured using a Karl Suss four-probe station and a Keithly 4200 semiconductor characterization system. The samples were placed on a conductive copper tape surface with one copper contact probe touching the copper tape (providing back contact), while the other probe was placed on the individual metal pad on the surface of the wafer. The voltage was varied from -1 to 1 volts and the current was measured accordingly. Figure 27: Portion of sample post metallization and annealing 65

82 Laser Doped Selective Emitter Results Scanning Electron Microscope Micrographs To qualitatively compare the extent of Si surface damage under the various laser fluences, all of the processed samples were viewed under a Scanning Electron Microscope (SEM). The results show that even under the smallest fluences, there were visible changes in surface morphology to the wafer by the laser as evident by the distinct pulse shape. The surface morphology also showed that there was a dramatic difference between the various lasers. For the wavelengths where there was a much higher absorption coefficient, more energy was absorbed closer to the surface in a short amount of time leading to ablation at the surface. Ablation is extremely undesirable for solar cells because it causes large amounts of lattice damage thereby leading to recombination centers. When comparing pulsed laser SEM images to images from the continuous wave laser, laser pulses from Q-switched lasers showed more ablation, whereas pulses from the continuous wave laser were much more uniform in their distribution. This is especially evident when comparing similar energy transfers and can be directly related to the fact that the pulsed lasers deliver energy in a much shorter amount of time, thereby increasing the power levels. The wafer stack-up used for the SEM images was 10-nm a-si:h/ 20-nm n+a-si:h/ 150-nm SiO x. The SiO x capping dielectric is essentially transparent to the wavelengths used in this study, so absorption in this layer is to be negligible. set of energies: Shown below are the SEM photographs for the 355-nm, 30-ns pulse duration laser for a given 66

83 Figure 28: SEM images of the 355-nm, 30-ns, pulsed laser at various energies under 1000x magnification 67

84 There is ablation present at fluences greater than 2 J/cm 2 which increases as the fluence continues to increase. The beam propagates the deepest in the center as evident by the tiny circular hole in the middle of the pulse. There is a large amount of spider-web created ablation evident at the higher pulses showing that the wafer experiences extremely rapid cooling. From qualitatively viewing the SEMs for the use of the solar cell industry, fluences with large ablation should be avoided; therefore fluences should be kept below 15 J/cm 2. shown below: The pulse sizes of the 355-nm, 30-ns laser were measured via the SEM photographs and are WAVELENGTH PULSE ENERGY (µj/pulse) DIAMETER AREA (µm 2 ) (nm) (µm) Table 1: Laser parameters and spot dimensions for the 355-nm, 30-ns pulsed laser The following are the SEM images for the 532-nm 30-ns pulse duration pulsed laser: 68

85 Figure 29: SEM images for the 532-nm, 4-ns, pulsed laser at various energies under 1000x magnification 69

86 At the 532-nm wavelength, the penetration depth of the LDSEs is about 1 µm, which is highly desirable for the creation of solar cells because this also corresponds to the average depth of the shallow emitter. From the SEM images it can be seen that the laser fluence creates a shallow pulse which does not appear to cause much ablation. The shallow profile is most likely due to the short pulse duration allowing for minimal beam exposure time to create a melt pool. The 532-nm, 4-ns pulsed laser shows dramatically less ablation compared to the 355-nm, 30-ns laser which indicates less lattice damage. Additionally, the spot size appears to be much larger than for the 355-nm laser. The pulse sizes of the 532-nm were measured via the SEM images and the following table summarizes the results: WAVELENGTH PULSE ENERGY DIAMETER (µm) AREA (µm 2 ) (nm) (µj/pulse) Table 2: Laser parameters and spot dimensions for the 532-nm, 4-ns pulsed laser The following are the SEM images for the 1064-nm, 4-ns pulsed laser with Gaussian profile: 70

87 Figure 30: SEM images of the Gaussian 1064-nm, 4-ns, pulsed laser at various energies under 1000x magnification 71

88 At the 1064-nm wavelength, the IR wavelength penetrates further into the wafer allowing for a deeper melt pool. As with the 532-nm wavelength, minimal ablation occurs even at the highest fluences. The SEM images are of a Gaussian beam profile which is apparent by noticing the small dot observed in the center of the spot in most images. The 1064-nm wavelength also used a π-shaper to obtain a top hat beam profile which yielded smaller, more uniform spots. The following table summarizes the parameters used: WAVELENGTH (nm) PULSE ENERGY (µj/pulse) GAUSSIAN DIAMETER (µm) GAUSSIAN AREA (µm 2 ) TOP HAT DIAMETER (µm) TOP HAT AREA (µm 2 ) Table 3: Laser parameters and spot dimensions for the 1064-nm, 4-ns pulsed laser with Gaussian and top hat beam profiles The following images are of the continuous 1070-nm laser: 72

89 Figure 31: SEM images of the 1070-nm continuous wave laser at highest powers and longest pulse durations at 1000x magnification Figure 32: SEM images of the 1070-nm continuous wave laser at highest powers and shortest pulse durations at 1000x magnification 73

90 Figure 33: SEM images of the 1070-nm continuous wave laser at lowest powers and longest pulse durations at 1000x magnification Figure 34: SEM images of the 1070-nm continuous wave laser at lowest powers and shortest pulse durations at 1000x magnification 74

91 The CW laser spot is dramatically different from that of the pulsed laser. In the CW laser, the spot is much more uniform and exhibited signs of a melt pool whereas a pulsed spot is more prone to display ablation and material ejection from the melt pool. The longer pulses experienced a longer melt time resulting in a very uniform melt pool. Conversely, in the ultra-short pulse ranges, below 100 µs, the melt pool solidifies rapidly before equilibrium is reached, resulting in pits and cracks along the outer circumferential ring of the spot. This pitting and non-uniformity are seen more readily at lower power levels. At the highest powers, it is only observed at the lowest pulse durations. From a strictly qualitative point of view, CW lasers generally create less lattice damage because they are less prone to ablation leading to less damage creation. However, for some LDSE applications such as self-aligning electroless metallization, ablation is desirable for adhesion purposes, in which case CW lasers would not be the ideal choice. WAVELENGTH (nm) Average Power (W) Pulse Duration (µs) DIAMETER (µm) AREA (µm 2 )

92

93 Table 4: Laser parameters and spot dimensions for the 1070-nm, CW laser 77

94 Another interesting factor to note is the dramatic difference between the amount of energy delivered in the CW laser vs. the pulsed laser. The CW laser delivered thousands of times more energy per unit area than the pulsed laser, yet the peak power of the CW laser was several orders of magnitude lower than that of the pulsed lasers. This is due to the short pulse duration for the Q-switched laser leading to powers in the multiple-digit kilowatt range. The following image shows the dramatic difference between the different types of lasers. On the left, is the CW pulse with the smallest power and shortest pulse, W and 40-µs, resulting in approximately 3000 µj being delivered to the surface of the wafer. On the right, is the 1064-nm wavelength at the highest pulse energy, µJ/pulse. This is smaller by more than an order of magnitude. Figure 35: Comparison of a CW vs. pulsed laser spot taken in an SEM under 1000x magnification 78

95 Electrical Characterization To quantitatively compare the differences between LDSEs, current-voltage measurements were made for each diode. Most of the emitters exhibited diode-like behavior with the exception of some of the lowest power and shortest pulse duration parameters, particularly on the thicker dopant layer stacks. The desirable LDSE will exhibit low series resistance, a low saturation current, an ideality factor close to unity, a high number of decades of linearity, and a high shunt resistance. These values were obtained by interpolating the diode curve of each individual LDSE. Each laser shows dramatically different results and trends indicating that parameter optimization for LDSE creation is laser dependent. 355-nm 30-ns pulsed laser The following are the current voltage characteristics of the 355-nm wavelength for the various stacks and pulse energies: 79

96 Figure 36: Current-voltage characteristics for the 355-nm, 30-ns pulsed laser on the 10/20/150 stack 80

97 Figure 37: Current-voltage characteristics of the 355-nm, 30-ns pulsed laser on the 10/40/150 stack 81

98 Figure 38: Current-voltage characteristics of the 355-nm, 30-ns pulsed laser on 10/60/150 stack 82

99 Figure 39: Current-voltage characteristics of the 355-nm, 30-ns pulsed laser on the 10/80/150 stack 83

100 Figure 40: Current-voltage characteristics of the 355-nm, 30-ns pulsed laser on the 10/120/150 stack 84

101 From these curves, the important characteristics were extrapolated and are organized below to determine trends: Figure 41: 3D graph comparing the ideality factor for the various stacks and pulse energies of the 355-nm, 30-ns pulsed laser In terms of ideality factor and the 355-nm wavelength, the ideality factor was near unity for most of the energy levels with a lower dopant concentration. When the dopant concentration got particularly high and on the 10/120/150 stack, there was a dramatic increase in ideality factor. Because of this, the highest dopant stack would not be recommended for LDSE creation. Additionally, the ideality factor tends to increase at pulse energies lower than 20 µj/pulse. It is believed that this is due to insufficient energy to create a large melt pool for the LDSE. 85

102 Figure 42: 3D graph comparing the decades of linearity for the various stacks and pulse energies of the 355-nm, 30-ns pulsed laser As far as decades of linearity are concerned, the 355-nm wavelength did not have remarkably good results, especially at the lower dopant concentrations. In general, stacks with 60 or 80-nm of n+a- Si:H had the largest number of linear decades, however they are still lower than expected. 86

103 Figure 43: 3D graph comparing the series resistance for the various stacks and pulse energies of the 355-nm, 30-ns pulsed laser The unusually high series resistance is most likely the cause for the low decades of linearity. As seen in the 3-D graph above, at the lower pulse energies the series resistance dramatically increases. This is particularly evident when examining the thicker dopant stacks. Because of this, it is further assumed that the lowest energies did not have sufficient penetration into the wafer to create a reliable LDSE. Since series resistance is detrimental to a solar cell, this is one of the leading parameters to examine. 87

104 Figure 44: 3D graph comparing the saturation current for the various stacks and pulse energies of the 355-nm, 30-ns pulsed laser The saturation current also displays apparent trends at the 355-nm wavelength. At the highest powers and the lowest dopant concentrations, the saturation current dramatically increases. This is believed to be caused by increased laser damage due to more ablation of the material. Figure 45: 3D graph comparing the shunt resistance for the various stacks and pulse energies of the 355-nm, 30-ns pulsed laser 88

105 With regards to shunt resistance, there does not appear to be a major trend in the data. This is apparent throughout all of the samples with the different lasers; therefore, it will not be factored into discussions. Taking all of the factors into consideration, at the 355-nm wavelength, it is particularly important to avoid the lowest pulse energies, particularly at the thick dopant stacks because there is insufficient energy to properly melt the region. Furthermore, to maximize the decades of linearity and to minimize the saturation current, one should stick to higher concentration levels. Taking all of these things into consideration, the most reliable LDSEs should be created with a dopant layer thickness between 60 and 120 nm, and pulse energy within the range of 50 to 200 µj. It should be noted that in general, the 355-nm wavelength created poor LDSEs which can draw its root causes from the high absorption coefficient causing high levels of ablation and minimal depth penetration. 532-nm 4-ns pulsed laser The following are the current voltage characteristics of the 532-nm, 4-ns laser for the various stacks and pulse energies: 89

106 Figure 46: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/20/150 stack 90

107 Figure 47: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/40/150 stack 91

108 Figure 48: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/60/150 stack 92

109 Figure 49: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/80/150 stack 93

110 Figure 50: Current-voltage characteristics of the 532-nm, 4-ns pulsed laser on the 10/120/150 stack 94

111 From these curves, the following data was extrapolated: Figure 51: 3D graph comparing the ideality factor for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser Figure 52: 3D graph comparing the decades of linearity for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser 95

112 For the 532-nm wavelength, the ideality factor was relatively close to unity for the stacks with 60-nm n+a-si:h and lower. Once the number of dopants atoms increased more than that in the 10/60/150 stack, the ideality factor increased dramatically. Additionally, the lowest pulse energies had higher ideality factors indicating that at these levels, there was not enough energy to melt a sufficient spot size to create the LDSE. The decades of linearity experienced the opposite trends of the ideality factor. The lowest energies and highest concentrations had the most linear decades while the higher energies and lower concentrations had the best ideality factors, yet only experienced a few decades of linearity. Figure 53: 3D graph comparing series resistance for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser The 532-nm wavelength yielded very low series resistance which is ideal. At the very lowest pulse energies (8.65 J/pulse) and the very high pulse energies (205.1 µj/pulse and higher), the series resistance increased significantly. This is believed to be caused by insufficient melting and diffusion in the first case and high amounts of ablation in the latter. 96

113 Figure 54: 3D graph comparing the saturation current for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser Figure 55: 3D graph comparing shunt resistance for the various stacks and pulse energies of the 532-nm, 4-ns pulsed laser The saturation current for the 532-nm wavelength shows the same general trends as the decades of linearity. It shows that at the highest pulse energies and lowest concentrations, there is an increase in the saturation current, but relatively constant throughout the rest of the parameters. 97

114 Taking everything into consideration, the highest and lowest pulse energies should be avoided along with the highest and lowest concentration levels. Therefore, for ideal selective emitter creation, pulse energies should stay within the range of 30 to 150 µj/pulse and near the dopant concentration of the 10/60/150 stack. This region experiences ideality close to unity, several decades of linearity, low series resistance and a low saturation current nm 4-ns pulsed Gaussian laser The following are the current voltage characteristics of the 1064-nm, 4-ns laser for the various stacks and pulse energies with a Gaussian beam. Following the I-V curves, there are three-dimensional graphs depicting the extrapolated information from the curves. 98

115 Figure 56: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/20/150 stack 99

116 Figure 57: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/40/150 stack 100

117 Figure 58: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/60/150 stack 101

118 Figure 59: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/80/150 stack 102

119 Figure 60: Current-voltage characteristics of the 1064-nm, 4-ns pulsed Gaussian laser on the 10/120/150 stack 103

120 Figure 61: 3D graph comparing the ideality factor for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed Gaussian laser Figure 62: 3D graph comparing the decades of linearity for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed Gaussian laser The 1064 nm Gaussian beam produiced a higher ideality factor than typically desired for almost all processing parameters. Only the highest pulse energies experienced ideality close to unity and the lowest concentrations achieved the most ideal results. 104

121 Despite the higher ideality factor, the 1064-nm wavelength showed extremely promising results in terms of decades of linearity for these diodes. All processing parameters and concentrations had several linear decades with the exception of the 120-nm n+a-si:h stack, indicating that at the highest dopant concentrations, poorer diodes are created. Figure 63: 3D graph comparing series resistance for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed Gaussian laser The series resistance for many of the diodes was relatively low (< 200 Ω), although lower results would be desired. It should be noted that while an individual LDSE series resistance is somewhat large, once they are placed in parallel with other LDSEs the total diagnostic cell s series resistance becomes much lower. The lowest two concentration stacks experienced the highest series resistance; therefore this processing zone should be avoided when operating at similar laser processing parameters. 105

122 Figure 64: 3D graph comparing the saturation current for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed Gaussian laser Figure 65: 3D graph comparing shunt resistance for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed Gaussian laser The 1064-nm wavelength showed that at the highest concentration stack, the saturation current grew quite dramatically. This is also followed by the low number of decades of linearity. 106

123 Taking everything into consideration, concentrations similar to the 10/60/150 and 10/80/150 should be used at higher pulse energies of at least 200 µj/pulse. These diodes had the best combination of ideality factor, many linear decades, low series resistance, and low saturation current. It is believed that the higher energies were able to create a larger, more uniform melt pool which to create the diode. The 1064-nm wavelength required higher energies to achieve this compared to the visible and UV wavelengths due to the deeper penetration of the beam nm 4-ns pulsed top hat laser The following are the current voltage characteristics of the 1064-nm, 4-ns laser for the various stacks and pulse energies with a top hat beam profile. Following the I-V curves, there are 3-D graphs depicting the extrapolated information from the curves. 107

124 Figure 66: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/20/150 stack 108

125 Figure 67: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/40/150 stack 109

126 Figure 68: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/60/150 stack 110

127 Figure 69: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/80/150 stack 111

128 Figure 70: Current-voltage characteristics of the 1064-nm, 4-ns pulsed top hat laser on the 10/120/150 stack 112

129 Figure 71: 3D graph comparing the ideality factor for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed top hat laser Figure 72: 3D graph comparing the decades of linearity for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed top hat laser Similar to the Gaussian beam, the top hat profile of the 1064-nm, 4-ns pulsed laser produced higher ideality factors at lower pulse energies. This trend indicates that at the lowest pulse energies, 113

130 there was not sufficient energy to properly create the selective emitter. This trend was more noticeable as the dopant concentration became larger. The decades of linearity showed that the 10/80/150 stack showed the most promise while the lowest concentration stack demonstrated a rather poor linear region. Figure 73: 3D graph comparing series resistance for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed top hat laser The lowest two stacks, 10/20/150 and 10/40/150 showed extremely high series resistance. However, once the concentration was sufficiently high, there was minimal series resistance. This trend was uniform throughout the entire stack and therefore did not appear to have any correlation to the pulse energy. 114

131 Figure 74: 3D graph comparing the saturation current for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed top hat laser Figure 75: 3D graph comparing shunt resistance for the various stacks and pulse energies of the 1064-nm, 4-ns pulsed top hat laser Saturation current was relatively uniform throughout the processing parameters with the exception of the highest pulse energies at the lowest stack regions. This is also apparent by the high series resistance and the low number of linear decades. 115

132 Overall, the most promising results for the 1064-nm, 4-ns pulsed top hat profile laser beam were in the concentration stack of 10/60/150 and 10/80/150 at higher pulse energies of approximately 80 µj/pulse and above. There is most likely an upper limit to this energy level but it was not reached in the parameters of this experiment nm 4-ns pulsed Gaussian vs. top hat comparison To compare the Gaussian and top hat beam profile results for the LDSE, the I-V characteristics were plotted on the same graphs to gain visible trends between the two. The following are plots with the top hat I-V curve in red and the Gaussian I-V curve in blue: 116

133 Figure 76: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/20/150 stack 117

134 Figure 77: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/40/150 stack 118

135 Figure 78: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/60/150 stack 119

136 Figure 79: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/80/150 stack 120

137 Figure 80: Current-voltage characteristic comparison of the 1064-nm, 4-ns pulsed Gaussian (blue) vs. top hat (red) laser on the 10/120/150 stack 121

138 The most apparent trend the I-V curves shows is that the top hat beam profile tends to have higher saturation currents and higher reverse current values compared to the Gaussian beam. This trend becomes less apparent at the higher concentration levels. However, the top hat beam tends to have an ideality factor closer to unity compared to that of the Gaussian profile. The top hat profile shows more distinct trends than the Gaussian beam. The trends on the 3-D graphs have distinct regions whereas the trends for the Gaussian beam tend to blend into each other more. This is probably due to the well-defined nature of the top hat beam shape with distinct beam edges. Conversely, the Gaussian beam tapers gradually at the edges and allow for a more gradual impact on the material. The overall recommendations for both the Gaussian and top hat profiles are quite similar. Both work best with concentrations within the range of 10/60/150 to 10/80/150 and higher pulse energies. The only difference is that the Gaussian beam profile has a much smaller range of optimal pulse energies (200 µj/pulse and greater), whereas the top hat beam profile can create optimal LDSEs at pulse energies of approximately 80-µJ/pulse and above. It is believed that this is due to the particular depth and area penetration that is needed for an ideal LDSE. Since the top hat profile can concentrate uniform amounts of energy over a wider area than the Gaussian beam, the desired depth and size penetration is achieved at lower energies nm continuous wave laser The continuous wave laser was also utilized to create selective emitters. Since both the power and pulse duration were varied, the results from each stack are shown through I-V curves and via 3-D graphs. Following this, the medium pulse duration (200-µs) and the medium average power ( W) were selected to compare the different stacks to each other. 122

139 Figure 81: Current-voltage characteristics of the 10/20/150 stack with nm CW laser with highest power and longest pulse duration Figure 82: Current-voltage characteristics of the 10/20/150 stack with nm CW laser with highest power and shortest pulse duration 123

140 Figure 83: Current-voltage characteristics of the 10/20/150 stack with nm CW laser with lowest power and longest pulse duration Figure 84: Current-voltage characteristics of the 10/20/150 stack with nm CW laser with lowest power and shortest pulse duration 124

141 Figure 85: Ideality factor for the 10/20/150 stack with the 1070-nm CW laser Figure 86: Decades of linearity for the 10/20/150 stack with the 1070-nm CW laser 125

142 Figure 87: Series resistance for the 10/20/150 stack with the 1070-nm CW laser Figure 88: Saturation current for the 10/20/150 stack with the 1070-nm CW laser 126

143 Figure 89: Shunt resistance for the 10/20/150 stack with the 1070-nm CW laser At the lowest concentration stack, 10/20/150, there were not very many noticeable trends in the current-voltage data. The ideality factor appeared to increase with high pulse power and short pulse duration. In general, the ideality factor was relatively high for every laser parameter varied. The decades of linearity were also quite large throughout the parameter range as well. Series resistance was much higher than desired and it was the highest at the lowest average power values. To obtain the lowest saturation current, the lowest pulse energies and the shortest pulse durations were used, but these were also the regions of high series resistance. 127

144 Figure 90: Current-voltage characteristics of the 10/40/150 stack with nm CW laser with highest power and longest pulse duration Figure 91 Current-voltage characteristics of the 10/40/150 stack with nm CW laser with highest power and shortest pulse duration 128

145 Figure 92 Current-voltage characteristics of the 10/40/150 stack with nm CW laser with lowest power and longest pulse duration Figure 93 Current-voltage characteristics of the 10/40/150 stack with nm CW laser with lowest power and shortest pulse duration 129

146 Figure 94: Ideality factor for the 10/40/150 stack with the 1070-nm CW laser Figure 95: Decades of linearity for the 10/40/150 stack with the 1070-nm CW laser 130

147 Figure 96: Series resistance for the 10/40/150 stack with the 1070-nm CW laser Figure 97: Saturation current for the 10/40/150 stack with the 1070-nm CW laser 131

148 Figure 98: Shunt resistance for the 10/40/150 stack with the 1070-nm CW laser For the stacks with slightly higher concentration, there were more noticeable trends discovered in many of the current-voltage parameters. Ideality factor was again rather high throughout all processing parameters, but decreased significantly at the highest pulse powers and longest pulse durations. There were in general 2-3 decades of linearity with almost every laser parameter. The series resistance was significantly lower throughout the sample. Regions with 120-µs pulse durations and greater with average powers greater than 50 W showed extremely low series resistance. The series resistance was extremely high for the lowest average powers. There was also a dramatic increase at the shortest pulse durations less than 100 µs. The saturation current increased at the highest average powers and longest pulse durations and showed similar trends as that of the ideality factor. Shunt resistance was in general much higher than with the 10/20/150 stack. 132

149 Figure 99: Current-voltage characteristics of the 10/80/150 stack with nm CW laser with highest power and longest pulse duration Figure 100: Current-voltage characteristics of the 10/80/150 stack with nm CW laser with highest power and shortest pulse duration 133

150 Figure 101: Current-voltage characteristics of the 10/80/150 stack with nm CW laser with lowest power and longest pulse duration Figure 102: Current-voltage characteristics of the 10/80/150 stack with nm CW laser with lowest power and shortest pulse duration 134

151 Figure 103: Ideality factor for the 10/80/150 stack with the 1070-nm CW laser Figure 104: Decades of linearity for the 10/80/150 stack with the 1070-nm CW laser 135

152 Figure 105: Series resistance for the 10/80/150 stack with the 1070-nm CW laser Figure 106: Saturation current for the 10/80/150 stack with the 1070-nm CW laser 136

153 Figure 107: Shunt resistance for the 10/80/150 stack with the 1070-nm CW laser At the 10/80/159 stack, the ideality factor was close to unity for a majority of the samples, indicating a much higher quality of diodes created. Pulse durations lower than 120 µs showed high ideality factors. This is most likely due to inadequate time for the melt pool to allow for diffusion. This can also be seen through the decades of linearity. The lowest pulse durations experience high decades of linearity, but the diodes are more indicative of a space charge region rather than the wanted LDSE. For this stack the series resistance was relative low with again the exception of the lowest powers and shortest pulses. It is believed that this is again due to insufficient energy to create the melt pool and inadequate time to allow diffusion to take place. Converse to many of the other results, the saturation current is much higher for the highest powers and longest pulse durations despite the more ideal curves. This could be due to lattice damage. 137

154 Figure 108: Current-voltage characteristics of the 10/120/150 stack with 1070-nm CW laser with highest power and longest pulse duration Figure 109: Current-voltage characteristics of the 10/120/150 stack with 1070-nm CW laser with highest power and shortest pulse duration 138

155 Figure 110: Current-voltage characteristics of the 10/120/150 stack with 1070-nm CW laser with lowest power and longest pulse duration Figure 111: Current-voltage characteristics of the 10/120/150 stack with 1070-nm CW laser with lowest power and shortest pulse duration 139

156 Figure 112: Ideality factor for the 10/120/150 stack with the 1070-nm CW laser Figure 113: Decades of linearity for the 10/120/150 stack with the 1070-nm CW laser 140

157 Figure 114: Series resistance for the 10/120/150 stack with the 1070-nm CW laser Figure 115: Saturation current for the 10/120/150 stack with the 1070-nm CW laser 141

158 Figure 116: Shunt resistance for the 10/120/150 stack with the 1070-nm CW laser With the highest concentration stack, the diode quality was rather poor. Ideality factors were close to unity only at the highest powers and longest pulse durations. It is believed this is because there was enough energy to penetrate the thicker dopant stack and create a junction only at high power levels. At the lowest powers and shortest pulse durations, no p-n junction formation occurs. In addition, the current-voltage curves of the remaining diodes experienced extremely high reverse current, which indicates poor quality. The 10/120/150 stack had very poor decades of linearity although the series resistance was rather low with the exception of the highest powers with pulse durations of less than 120 µs. The saturation current was much higher than many of the other stacks and was particularly high for the higher average powers. This may be due to lattice damage. The following compare the diode quality at the same pulse duration of 200 µs: 142

159 Figure 117: 3D graph comparing the ideality factor for the various stacks and average powers of the 1070-nm, CW laser at 200-µs pulse duration Figure 118: 3D graph comparing the decades of linearity for the various stacks and average powers of the 1070-nm, CW laser at 200-µs pulse duration 143

160 When comparing trends across the different stacks, the thicker dopant structures had much better ideality factors than that of the thinner structures. This trend seemed unaffected by the average power used. Additionally, the decades of linearity also followed a similar trend. The lowest concentration stacks had significantly higher decades of linearity compared to the higher ones. Figure 119: 3D graph comparing series resistance for the various stacks and average powers of the 1070-nm, CW laser at 200-µs pulse duration 144

161 Figure 120: 3D graph comparing the saturation current for the various stacks and average powers of the 1070-nm, CW laser at 200-µs pulse duration. Figure 121: 3D graph comparing shunt resistance for the various stacks and average powers of the 1070-nm, CW laser at 200-µs pulse duration 145

162 The series resistance was lowest at the thicker dopant stacks with the highest powers. At lower powers, insufficient diffusion takes place. The saturation current showed opposite trends though, and jumped quite significantly at the highest dopant concentrations. This is most likely due to the low decades of linearity. Comparing the same pulse power is also important when determine the ideal processing parameters. The following are the data from the 1070-nm CW laser at constant average power of W. Figure 122: 3D graph comparing the ideality factor for the various stacks and pulse durations of the 1070-nm, CW laser at W average power 146

163 Figure 123: 3D graph comparing the decades of linearity for the various stacks and pulse durations of the 1070-nm, CW laser at W average power Across the various stacks, the pulse duration did not appear to have a dramatic effect on the ideality factor or the decades of linearity. The major contribution appeared to come from the concentration used. The one exception is that there appears to be a critical pulse length of greater than 120 µs required to create good diodes. There was a dramatic change in decades of linearity and ideality factor for all stacks when below this critical pulse length. It is believed that at < 120 µs, there is insufficient time for the dopant atoms to properly diffuse; the result is formation of poor quality junctions. Therefore, when determining the ideal processing parameters for an LDSE, pulse durations should be kept greater than 120 µs. 147

164 Figure 124: 3D graph comparing series resistance for the various stacks and pulse durations of the 1070-nm, CW laser at W average power Figure 125: 3D graph comparing the saturation current for the various stacks and pulse durations of the 1070-nm, CW laser at W average power 148

165 Figure 126: 3D graph comparing shunt resistance for the various stacks and pulse durations of the 1070-nm, CW laser at W average power In terms of series resistance, saturation current, and shunt resistance, the pulse duration did not appear to alter these parameters. There was an obvious correlation between these functions and the stack used, but not pulse duration. Taking all of the above data into consideration, for the 1070-nm CW laser, the best overall diodes occurred with the 10/40/150 and the 10/80/150 stacks. These stacks exhibited low series resistance and low saturation current. The 10/40/150 stack had many decades of linearity, but showed poor ideality factors. Conversely, the 10/80/150 stack displayed good ideality factors yet few linear decades. Because of this, it is believed that the medium stack of 10/60/150 would best suit the 1070-nm wavelength because it would be optimal for achieving both ideality factors close to unity and multiple decades of linearity. 149

166 For the 10/40/150 stack, pulse duration should be larger than 120 µs, but below 240 µs. At longer pulse lengths, the saturation current increases significantly. Additionally, the average power should be kept greater than 55 W to allow for proper melting and diffusion of the material. For the 10/80/150 stack, again the pulse duration should be kept greater than 120 µs, but for this stack should be less than 320 µs before electrical degradation results. It is believed that this increased size window correlates to the increased thickness level of the dopant material. Again, the average power should be kept greater than 55 W for diffusion purposes. For ideal results at the 10/60/150 stack, a compromise between these two processing parameters should be considered while maintaining average powers greater than 55 W and pulse durations of greater than 120 µs. 150

167 CHAPTER 4: LASER DOPED SELECTIVE EMITTER DISCUSSION Laser Doped Selective Emitter Comparisons While each different type of laser has its own parameter optimization, a comparison must be made between the lasers to determine which might be best for the creation of LDSEs. To do this, each electrical characterization parameter will be compared across the lasers. In terms of ideality factor, most of the lasers did not have the most ideal results and the ideality factor appears to increase with increasing wavelength. The 355-nm laser showed the most promise of any lasers and was between 1.0 and 1.2 for almost all parameters. Next the 532-nm wavelength had a slightly higher ideality factor of about 1.2, for all areas except the lowest pulses and the highest concentrations. Next was the 1064-nm wavelength. The top hat beam had decent ideality within the range of 1.2 to 1.4 for most values whereas the Gaussian beam was slightly higher and between 1.3 and 1.4. Finally the 1070-nm CW laser had extremely high ideality factors between 1.6 and 1.8 for many of the conditions. Equally as important as ideality factor is the number of decades of linearity. Here, the 355-nm laser exhibited the poorest linear region and in general was between 1 and 2 decades. The 532-nm laser was slightly better and was between 2 and 3 whereas the IR lasers did significantly better. The 1064-nm Gaussian beam had the best out of all lasers with almost all results coming between 3 and 4 decades. The top hat beam had equally good results, just a slightly larger window of 2.5 decades to 4. The nm CW laser had difficult results to compare. At the lower thickness stacks, linearity extended only to 1-2 decades, whereas for the thicker dopant layers, this range was increased to 3-4 decades for most processing parameters. Overall, the IR wavelengths exhibited the largest decades of linearity whereas 151

168 the UV wavelength had the poorest. The variation in absorption coefficient may be the cause of the results for both ideality and decades of linearity being dependent on wavelength. Across the board, the series resistance was higher than desired for the lasers used in this study, though some were worse than others. The 355-nm laser was the worst in terms of series resistance but still had a majority of the values below 150 Ω with a minimum value of 42 Ω. The Quantel results all had series resistance below 100 Ω for most conditions. Within these, the minimum values for the 532-nm, the 1064-nm Gaussian beam, and the 1064-nm top hat beam were 15, 46, and 58 Ω respectively. The 1070-nm CW laser showed slightly lower average series resistance than the above-mentioned with its minimum of around 36 Ω. As a comparison of the lasers in terms of series resistance, almost all of them display fairly similar results with the exception of the 355-nm wavelength laser which is significantly higher. Saturation current is one of the most important factors to consider when comparing the various lasers. This term makes it fairly easy to tell which laser adequately dopes the LDSE and which lasers do not. In the lasers chosen for this study, the IR lasers showed the lowest saturation current, with the 1070-nm CW laser being the lowest with a minimum value of 36 pa. Coming up similarly close were the 1064-nm Gaussian and top hat with minimum values of 560 and 460 pa respectively. However, it is important to note that these results are higher than laser doping literature values which have obtained saturation currents in the single digit pico-range. The green laser exhibited almost a decade higher minimum value of saturation current of around 3350 pa. Similarly, the UV wavelength showed values only as low as 6180 pa. These high values for non-ir wavelengths indicate that these lasers should be avoided when optimizing LDSE parameters. 152

169 After a thorough electrical comparison, it appears as if the 355-nm, 30-ns pulsed laser should be avoided due to its high series resistance, low decades of linearity, and extremely high saturation current. The 532-nm 4-ns pulsed laser shows promising characteristics in many aspects, yet also has the undesirable high saturation current. Longer pulse lengths or a CW green laser would help with this electrical characteristic. The IR lasers showed overwhelmingly the best all-around electrical characteristics. The 1070-nm CW laser has high potential to optimize settings between the dopant stacks, power and pulse duration. It also had the lowest saturation current of all the lasers of 36 pa. The major downfall of the CW laser was the high ideality factors. The best diode characteristics came from the 1064-nm, 4-ns pulsed laser with both the Gaussian and top hat beam profiles. Using this laser showed minimal series resistance, low saturation current, high decades of linearity and decent ideality factors within the range of 1.2 to 1.4 were found. The top hat beam profile had the slight edge on the Gaussian profile due to the slightly lower ideality factors and saturation current. Discussion of Error Within the design implementation, there were several sources of error which most likely occurred during experimentation. These errors can be seen through some inconsistencies in the electrical results. Starting with the laser setup, care was taken to maintain consistent mounting of different samples, but slight variations are likely. These could include small changes in the beam output power and energy along with minor changes in the beam focus. Additionally, the samples were processed in ambient air so impurities could have been incorporated into the LDSEs and these impurities could have varied depending on the day and the laser system being used. 153

170 Both the 532-nm and the 1064-nm Quantel beams displayed poor quality multi-mode beams. To reduce this, processing was performed close to the aperture where the beam would exhibit a more Gaussian energy profile. Additionally, the π-shaper may not have made a perfectly top hat beam and there may have been minimal amounts of energy lost within the device as well. The measurement methods for the I-V characteristics may also have contributed to error. The metallization helped to improve consistency between the samples, but there was still some variance which could have occurred. The largest contributor to the variance probably stemmed from the size of the device matrix from which the electrical measurements taken with each parameter. For better trends, more samples should be collected from the various laser parameters to ensure better statistical conclusions. 154

171 CHAPTER 5: CONCLUSIONS AND FUTURE WORK Conclusions Laser doped selective emitters were prepared and junction formation was studied via SEM images and electrical characterization. Fabrication of the LDSEs was performed on silicon wafers deposited with five different dopant structures, each with varying dopant thickness levels thereby altering the doping concentration within the LDSE. A total of four different lasers were used to create the LDSEs along with a π-shaper for the 1064 nm wavelength. For the Q-switched lasers, pulse energy was varied as a processing parameter. For the CW laser, both average power and pulse duration were varied. The resulting diodes were metallized and then characterized via current-voltage measurements. From the diode curves, electrical parameters such as ideality factor, decades of linearity, series resistance, saturation current and shunt resistance were extrapolated. The aforementioned parameters were then compared by 3-D graphs to determine the optimal laser processing parameters. The ideal parameters were found for each individual laser as well as overall optimization for the various types of lasers. The results showed LDSE creation for almost all of the processing parameters used. The best diodes were created with the 1064-nm, 4-ns pulsed laser with both a Gaussian beam and a top hat beam. Within these, the dopant stacks of 10/60/150 and 10/80/150 produced the best results with energies of 80 µj/pulse and 200 µj/pulse or greater for the Gaussian and top hat modes, respectively. The 355-nm, 30-ns pulsed laser produced the poorest overall results. LDSEs with this wavelength exhibited high series resistance, high saturation current and poor decades of linearity. Additionally, because of the high absorption coefficient, diodes of this wavelength created undesirable ablation creating lattice damage and decreasing carrier lifetime. 155

172 Future work Future work of the laser doped parameter experiment will entail using a larger variety of lasers to compare selective emitter electrical results. In particular, green lasers with longer pulse duration would be of interest because it would maintain the ideal absorption depth, yet allow a lower peak power during operation. In particular, a CW green laser would be of interest because of the lower power levels thereby creating less ablation. Lifetime measurements should be taken to observe quantitatively the damage caused by laser processing. Modeling simulations of the pulses would also be of interest and could show material interactions in the time domain. Additionally, modeling the theoretical doping concentration of the p-n junction compared to the experimental results would be of interest to help diagnose areas of improvement. A fully designed experimental diagnostic cell with both shallow and selective emitter would aid in the comparisons for the parameter optimization. Finally, a full economical study should be done to determine if the increase in cell efficiency is worth the additional processing cost. 156

173 REFERENCES [1] W. T. Silfvast, Laser Fundamentals, New York: Cambridge University Press, [2] E. Kannatey-Asibu Jr., Principles of Laser Materials Processing, Hoboken: John Wiley & Sons, Inc., [3] O. Homburg, F. Toennissen, H. Ganser, T. Mitra and V. Lissotschenko, "Refractive Gauss-to-Tophat Beam Shapers Improve Structure Quality and Speed in Micro-machining," in Conference on Quantum Electronics and Laser Science, Homburg, [4] B. L. DeCesar, Investigation of Process Parameter Optimization of Laser-Fired Back Contact Silicon Solar Cells, University Park: The Pennsylvania State University, [5] W. D. Callister Jr., Materials Science and Engineering and Introduction, York: John Wiley & Sons, Inc., [6] P. E. Millon and E. Fogarassy, "Laser Cleaning: State of the Art," Recent Advances in Laser Processing of Materials, pp , [7] C. J. Geankoplis, Transport Processes and Separation Process Principles, Upper Saddle River: Pearson Education, Inc., [8] O. B. Cruz and L. D. Olavarrieta, "A Bird's Eye View of Materials and Manufacturing Processes for Photovoltaic Cells," Circuits and Devices Magazine, IEEE, vol. 20, no. 4, pp. 5-12, [9] T. Jackson and M. Oliver, "The Viability of Solar Photovoltaics," Energy Policy, vol. 28, pp , [10] A. Shah, Thin-Film Silicon Solar Cells, Italy: EPFL Press, [11] L. Fraas and L. Partain, Solar Cells and their Applications, Hoboken: John Wiley & Sons, Inc., [12] S. J. Fonash, Solar Cell Device Physics, Burlington: Elsevier Inc., [13] L. Zighed and A. Mahdjoub, "Comparative Study Between Various Antireflective Coatings For Solar Cell Applications," in IOP Conference Series: Materials Science and Engineering, [14] Z. Hameiri, L. Mai, N. Borojevic, S. Javid, B. Tjahjono, S. Wang, A. Sproul and S. Wenham, "The Influence of Silicon Nitride Layer Parameters on the Implied Voc on CZ Silicon Wafers After 157

174 Annealing," IEEE, pp , [15] S. C. Baker-Finch and K. R. Mcintosh, "Reflection of Normally Incident Light from Silicon Solar Cells with Pyramidal Texture," Progress in Photovoltaics: Research and Applications, vol. 19, pp , [16] M. Z. Pakhuruddin, K. Ibrahim, M. K. M. Ali and A. A. Aziz, "Surface Morphology and Optical Reflection of Thermally Evaporated Thin Film Al-Doped Silicon on Plastic Substrates for Solar Cell Applications," in International Conference on Enabling Science and Nanotechnology, Malaysia, [17] M. Lipinski, P. Panek, Z. Swiatek, E. Beltowska and R. Ciach, "The Influence of Surface Modifications on Monocrystalline Silicon Solar Cells," Solar Energy Materials & Solar Cells, vol. 72, p. 271, [18] Ouellette M. F., R. Lang, K. L. Yan, R. W. Betram, R. S. Owles and D. Vincent, "Comparative Study Between Various Antireflective Coatings for Solar Cell Applications," Journal of Vacuum Sciences and Technology, vol. 9, [19] M. A. Green, Solar Cells, Operating Principles, Technology and Systems Applications, Englewood Cliffs, N. J.: Prentice Hall, [20] H. J. Moller, Seimconductors for Solar Cells, Boston: Artech Houxe, [21] W. Schockley and H. J. Queisser, "Detalied Balance Limit of Efficiency of p-n Junction Solar Cells," Journal of Applied Physics, vol. 32, p. 510, [22] H. Kiess and W. Rehwald, "On the Ultimate Efficiency of Solar Cells," Solar Energy Materials and Solar Cells, vol. 38, pp , [23] M. A. Green, K. Emery, Y. Hishikawa and W. Warta, "Solar Efficiency Tables (Version 35)," Progressin Photovoltaics: Researchand Applications, vol. 18, pp , [24] D. K. Schroder, Semiconductor Material and Device Characterization, New York: John Wiley & Sons, Inc., [25] M. Dadu, A. Kapoor and K. N. Tripathi, "Effect of Operating Current Depdent Series Resistance on the Fill Factor of a Solar Cell," Solar Energy Materials & Solar Cells, vol. 71, pp , [26] S. W. Glunz, A. Mette, M. Aleman, P. L. Richter, A. Filipovic and G. Willeke, "New Concepts for the Front Side Metallization of Silicon Solar Cells," in 21st European Photovoltaic Solar Energy Conference and Exhibition, Dresden, [27] C. del Canizo, G. del Coso and W. C. Sinke, "Crystalline Silicon Solar Module Technology: Towards 158

175 the 1 Euro per Watt-peak goal," Progress in Photovoltaics: Research and Applications, vol. 17, pp , [28] T. Sarnet, G. Kerrien, N. Yaakoubi and A. Bosseboeuf, "Laser Doping for Microelectronics and Microtechnology," in SPIE, Bellingham, WA, [29] K. Toyoda and K. Sugioka, "Microfabrication of Semiconductors by Means of Excimer Laser Doping," Microelectronic Engineering, vol. 20, pp , [30] K. Ibbs and M. Lloyd, "Laser Doping: Bipolar Structures in SIlicon," Optics and Laser Technology, vol. 30, pp , [31] S. Bet, N. Quick and A. Kar, "Effect of Laser Field and Thermal Stress on Diffusion in Laser Doping of SiC," Acta Materialia, vol. 55, pp , [32] Z. Hameiri, L. Mai, T. Puzzer and S. R. Wenham, "Influence of Laser Power on the Properties of Laser Doped Solar Cells," Solar Energy Materials & Solar Cells, vol. 95, pp , [33] S. Bet, N. Quick and A. Kar, "Laser Doping of Chromium as a Double Acceptor in Silicon Carbide with Reduced Crystalline Damage and Nearly All Dopants in Activated State," Acta Materialia, vol. 56, pp , [34] Z. Tian, N. R. Quick and A. Kar, "Laser-enhanced DIffusion of Nitrogen and Aluminum Dopants in Silicon Carbide," Acta Materialia, vol. 54, pp , [35] S. J. Eisele, T. C. Roder, J. R. Kohler and J. H. Werner, "18.9% Efficient Full Area Laser Doped Silicon Solar Cell," Applied Physics Letters, vol. 95, [36] A. Esturo-Breton, M. Ametowobia, J. R. Kohler and J. H. Werner, "Laser Doping for Crystalline Silicon Solar Cell Emitters," in 20th European Photovoltaic Solar Energy Conference, Barcelona, [37] A. Ogane, K. Hirata, K. Horiuchi, Y. Nishihara, Y. Yakahashi, A. Kitiyanan and T. Fuyuki, "Laser-Doping Technique Using Ultraviolet Laser for Shallow Doping in Crystalline Silicon Solar Cell Fabrication," Japanese Journal of Applied Physics, vol. 48, [38] K. Lee, S. Bae, Y. Kim, M. Jeon, J. Lee, J. Hong, Y.-G. Joh and J. Yu, "Laser Doping of Phosphorous in Grooved Silicone Surface," in SPIE, [39] D. Trusheim, M. Schultz-Ruhtenberg, T. Baier, S. Krantz, D. Bauer and J. Das, "Investigation of the Influence of Pulse Duration in Laser Processes for Solar Cells," Physics Procedia, vol. 12, pp ,

176 [40] B. Paviet-Salomon, S. Gall, R. Monna, S. Manuel and A. Slaoui, "Experimental and Analytical Study of Saturation Current Density of Laser-Doped Phosphorus Emitters for Silicon Solar Cells," Solar Energy Materials & Solar Cells, vol. 95, pp , [41] T. Roder, P. Grabitz, S. Eisele, C. Wagner, J. R. Kohler and J. H. Werner, "0.4% Absolute Efficiency Gain of Industrial Solar Cells by Laser Doped Selective Emitter," in 34th Photovoltaic Specialist Conference, Philadelphia, [42] J. R. Kohler, P. Grabitz, S. J. Eisele, T. C. Roder and J. H. Werner, "Laser Doped Selective Emitters Yield 0.5% Efficiency Gain," in 24th European Photovoltaic Solar Energy Conference, Hamburg, [43] L. Zhu, J. Gong, J. Huang, P. She, M. Zeng, L. Li, M. Dai and Q. Wan, "Improving the Efficiency of Crystalline Silicon Solar Cells by an Intersected Selective Laser Doping," Solar Energy Materials & Solar Cells, vol. 95, pp , [44] C.-H. Lin, S.-P. Hus, J.-J. Liou, C.-P. Chuang, W.-H. Lu and W.-L. Chang, "Characterization of Selective- Emitter Solar Cells Consists of Laser Opened Window and Subsequently Screen-Printed Electrodes," IEEE, vol. 10, pp , [45] M. Schulz-Ruhtenberg, A. Haeberle, R. Russell, J. L. Hernandez and S. Krantz, "Influence of the Pulse Width for Visible Pulsed Laser Doping for Crystalline Solar Cells Using Phosphosilicate Glass," Journal fo Laser Micro/Nanoengineering, vol. 6, no. 1, pp , [46] P. Oesterlin and U. Jager, "High Throughput Laser Doping for Selective Emitter Crystalline Si Solar Cells," in 18th IEEE Conference on Advanced Thermal Processing of Semiconductors, [47] S. Germershausen, L. Bartholomaus, U. Seidel, N. Hanisch, A. Schieferdecker, K. H. Kusters, M. Kittler, M. Ametowobla, F. Einsele and G. Dallmann, "Investigation of Emitter Homogeneity on Laser Doped Emitters," Energy Procedia, vol. 8, pp , [48] U. Besi-Vetrella, L. Pirozzi and E. Salza, "Large Area, Screen Printed Silicon Solar Cells with Selective Emitter made by Laser Overdoping and RTA Spin-on Glasses," IEEE, vol. 97, pp , [49] Z. Hameiri, L. Mai, A. Sproul and S. R. Wenham, "18.7% Efficient Laser-Doped Solar Cell on p-type Czochralski Silicon," Applied Physics Letters, vol. 97, [50] G. Poulain, C. Boulord, D. Blanc, A. Kaminski, M. Gauthier, C. Dubois, B. Semmache and M. Lemiti, "Direct Laser Printing for High Efficiency Silicon Solar Cells Fabrication," Applied Surface Science, pp , [51] U. Jager, P. Oesterlin, A. Kimmerle and R. Preu, "Beam Shaping - The Key to High Throughput 160

177 Selective Emitter Laser Processing with a Single Laser System," IEEE, vol. 10, pp , [52] Z. Hameiri, T. Puzzer, L. Mai, A. B. Sproul and S. R. Wenham, "Laser Induced Defects in Laser Doped Solar Cells," Progress in Photovoltaics: Research and Applications, vol. 19, pp , [53] K. Ohmer, Y. Weng, J. R. Kohler, H. P. Strunk and J. H. Werner, "Defect Formation in Silicon During Laser Doping," IEEE Journal of Photovoltaics, [54] M. Abbott, P. Cousins, F. Chen and J. Cotter, "Laser-Induced Defects in Crystalline SIlicon Solar Cells," IEEE, pp , [55] J. Hecht, Beam: The Race to Make the Laser, New York: Oxford University Press, [56] J. Hecht, Laser Pioneers, San Diego: Academic Press, Inc., [57] N. Taylor, Laser, New York: Simon & Schuster, [58] B. H. King, M. J. O'Reilly and S. M. Barnes, "Characterizing Aerosol Jet Multi-Nozzel Processing Parameters for Non-Contact Front Side Metallization of Silicon Solar Cells," IEEE, [59] J. H. Lee, Y. H. Lee, J. Y. Ahn and J.-W. Jeong, "Analysis of Series Resistance of Crystalline Silicon Solar Cell with Two-Layer Front Metallization Based on Light-Induced Plating," Solar Energy Materials & Solar Cells, vol. 95, pp , [60] V. A. Chaudhari and C. S. Solanki, "A Novel Two Step Metallization of Ni/Cu for Low Concentrator c- Si Solar Cells," Solar Energy Materials and Solar Cells, vol. 94, no. 12, pp , [61] S. Kontermann, A. Wolf, D. Reinwand, A. Grohe, D. Biro and R. Preu, "Optimizing Annealing Steps for Crystalline SIlicon Solar Cells with Screen Printed Front Side Metallization and an Oxide- Passivated Rear Surface with Local Contacts," Progress in Photovoltaics: Research and Applications, vol. 17, no. 8, pp , [62] B. e. al., "Photoelectrochemical Plating of Silver". United States Patent , 10 Jan [63] M. Aleman, N. Bay, M. Fabritius and S. W. Glunz, "Characterization of Electroless Nickel Plating on Silicon Solar Cells for the Front Side Metallization," in 20th European Photovoltaic Solar Energy Conference and Exhibition, Milano. [64] A. Mette, P. L. Richter, M. Horteis and S. W. Glunz, "Metal Aerosol Jet Printing for Solar Cell Metallization," Progress in Photovoltaics: Research and Applications, vol. 15, pp , [65] D. M. Hulijic, S. Thormann, R. Preu, R. Ludemann and G. Willeke, "Pad Printed Front Contacts for c- Si Solar Cells - A Technological and Economical Evaluation," in 29th IEEE Photovoltaic Specialists 161

178 Conference, New Orleans, [66] A. R. Burges, "How to Design Optimal Metallization Patterns for Solar Cells," Progress in Photovoltaics: Research and Applications, vol. 7, pp , [67] H. B. Serreze, "Optimizing Solar Cell Performance by Simultaneously Consideration of the Grid Pattern Design and Interconnect Configuration," in IEEE Photovoltaic Specialists Conference, [68] P. Morvillo, E. Bobeico, F. Formisano and F. Roca, "Influence of Metal Grid Patterns on the Performance of Silicon SOlar Cells at Different Illumination Levels," Materials Science and Engineering B, pp , [69] A. Meisel, M. Burrows, M. Abbott, F. Lemmi and H. Antoniadis, "Impact of Metal Contact Misalignment in Silicon Ink Selective Emitter Solar Cells," IEEE, vol. 10, pp , [70] B. L. Hedrick, Characterization of Laser Fired Contacts, Laser Fired Emitters, and Fixed-Charge Passivation for Improved Silicon Solar Cells, University Park: The Pennsylvania State University,

179 APPENDIX A: Extended Literature Review History of the Laser Although laser technology is around a half-a-century old, its roots can be traced as far back as World War I, when Albert Einstein first proposed his theory of light emission. [55] More directly however, the maser tends to be thought of as the first practical starting place for laser technology. Created by Charles Townes at Columbia University in 1953, the maser utilized beams of microwave emission to help aide in communication applications. [56] The first maser was produced using ammonia vapor which produced gain at a wavelength of 1.25 cm. This wavelength was comparable to the dimensions of the device. [1] Interest soon turned to extending this maser concept to the optical and infrared wavelengths; however problems arose because the same principle ideas could not be utilized since these wavelengths were several orders of magnitudes smaller than the 1.25 cm box. In 1958, Charles Townes and Arthur Schawlow proposed a solution to this problem. They developed the concept of an optical amplifier which was surrounded by an optical mirror resonate cavity to allow for growth of the beam. [1] This paper eventually earned them a Nobel Prize. Another great mind was also behind the solution: Gordon Gould. Gould was the first person to ever use the word laser and was inspired to build his version of the optical laser in 1958 however did not file for a patent until [57] As a result, his patent was refused, however he spent almost 30 years fighting for it claiming that Townes original design wouldn t work. Eventually the courts ruled in Gould s favor but due to the delay, his technology had already been exploited by others. [57] Just two years after Townes and Schawlow s paper, Theodore Maiman produced the first laser using a ruby crystal as the amplifier and a flashlamp as the energy source. [56] The rod-shaped ruby crystal was surrounded by a helical flashlamp and the optical cavity was formed by coating the flattened 163

180 ends of the ruby rod with a highly reflective material. [1] As a result, an intense red beam was observed to emerge from the rod when the flashlamp was fired. After the first laser was created, there was a time referred to as The Great Laser Boom. [56] Within a matter of years, dozens of new types of lasers were created, each utilizing a slightly different property. In the first five years, the gas laser was created using a mixture of helium and neon gases, the neodymium laser, the semiconductor laser, the carbon dioxide laser and the ion laser which used mercury vapor. In subsequent years, more laser types were continued to be discovered. [1] Metallization Techniques used for Photovoltaics The front metallization of a solar cell has the capability to either make-or-break the cell efficiency. There is a naturally occurring increase in recombination at these contacts and there is both contact resistance and series resistance that decrease cell efficiency as well. Conventional industrial methods use screen printing to create these contact lines however it is becoming evident that screen printing has limitations which may be hindering the photovoltaic industry. Because of this, other metallization methods have been developed; many of which utilize a two-stage metallization process with a seed layer followed by light induced plating. Screen Printing Screen printing is a relatively simple and cost effective process for the metallization of solar cells and it is the most common technique in use today for silicon solar cells. [58] During the process, a screen is lowered onto the cell along with a paste. A squeegee drags across the screen forcing the metal paste through the holes in the screen print mask. This mask is then removed leaving behind a thick layer of wet metal paste chis is dried in an oven to remove any organic solvents. After the oven, it is then placed 164

181 in furnace of much higher temperature to allow the metal to come into good electrical contact with the silicon. This method is ideal for industry because of its simplicity and the speed at which metal can be deposited. One disadvantage of screen printing is that it has a low aspect ratio (a ratio of height to width of the printed grid line). In order to achieve the desired conductance, lines need to have a larger width which leads to an increase in shadow loss. Another negative feature is that the silver paste has conductivity which is 2-3 times lower than that of pure silver. [59] This can be contributed to the presence of SiO 2 in the silver paste. [60] It has been found that screen printing methods tend to have higher series resistance than many other metallization methods. Investigations with a scanning electron microscope have shown that there are gaps and defects at the silicon-silver interface which partially separates the silver crystals from the silicon. [61] It is believed that this contact resistance is a major cause of the higher series resistance. The series resistance can be noticed very clearly when attempting to operate the cell at concentration levels of more than one sun as there is a large drop in cell efficiency. [60] Studies have shown that through proper annealing conditions post metallization, the number and size of these defects can be reduced thereby lowering the series resistance. [61] As the industry is moving toward thinner wafers with higher efficiencies, screen printing is reaching its limits and it is opening the doors for promising new metallization alternatives. Lithography Though lithography is not appropriate for large scale industrial production, this approach for metallization has achieved the highest solar cell efficiencies to date. In this approach, the solar cell is coated with a thin layer of photo-resist while spinning the substrate at high speeds. The cell is then 165

182 placed under a photomask displaying the desired front metallization pattern. It is then exposed to UV light and the photons cause the exposed film to become soluble and dissolve when placed in a developing solution leaving behind a thin mask on the cell. From here various thin-film metallization techniques can be used such as sputtering or e-beam to deposit the metal. The remaining layer of photo-resist is then removed leaving behind a neat and uniform front metallization pattern. Light-induced Electro-plating Light-induced electro-plating is of particular interest for the front side metallization of solar cells. This is the second part of a two-step metallization process where a metal, typically silver, is deposited photo-electrochemically to thicken an already existent thin seed layer. The results in aspect ratios close to 1:2 (height: width). [26] The overall goal of this two-step process is to reduce shadow loss while keeping series resistance as low as possible. The seed layer is designed to adhere well to the cell well offering low contact resistance whereas the light-induced layer is designed to decrease the series resistance. In particular, silver is used in many of these applications because it is relatively chemically stable and has high thermal and electrical conductivity characteristics. [62] 166

183 Figure 127: SEM photograph of a sample which has a laser sintered seed layer which is thickened by light-induced silver plating [26] This process has a much higher deposition rate than electro-less plating methods which is better for industrial purposes. In addition, the process can be used in junction with screen printing. It has been found to increase efficiencies of narrow screen-printed lines by %. [26] The increase in efficiency is due to a reduction in both shadow losses and in series resistance. Studies have shown that the series resistance could decrease by 0.5 Ω cm 2 on crystalline silicon solar cells with grid widths of 80 µm. [59] Laser Sintering Laser sintering is a metallization process where a powder of metal particles is deposited on the surface of the cell. A scanning laser then sinters and melts the particles to form the contact lines of the cell. The excess powder is then removed leaving a very fine high contact line. This line must then be thickened by light induce electroplating to enhance the conduction characteristics. Though this process involves melting and diffusion of both the metal and the underlying silicon, it has been shown that there are not sever damages in the emitter which would increase recombination. [26] However, this approach 167

184 is not used widely in industry because of the high costs and the additional processing steps. Furthermore, research has shown that efficiencies are not as high as conventional metallization methods. [26] Chemical Plating of Nickel Chemical plating of nickel is a very attractive process for creating the metal contact grid to the front of the solar cell. The process involves placing the cell in an alkaline nickel sulfate bath with sodium hypophosphite acting as a reducing agent. [63] To get the desired metallization pattern, laser ablation techniques remove the antireflection coating, typically SiN x. The ARC is chemically inert to the nickel bath and therefore plating does not occur on it. In cases where SiO2 is the antireflection coating, laser ablation methods are not applicable. Here, a photo-resist can be used to mask a chemical etching step thereby opening the antireflection layer in the desired manner. [26] The final nickel layer acts as a good metal-contact layer for the cell, but typically the cell is placed in a light induced silver plating bath to increase the height of the metal contact, thereby increasing conductivity. Cells have been created at a laboratory scale to have efficiencies as high as 18.9% when combined with laser fired contacts. [63] [26] One advantage of chemically plating nickel is that there no mechanical pressure required, thereby lowering the waste. Additionally, there is no need for a vacuum or for high temperature firing steps because nickel offers a low resistivity contact to the silicon. Finally, when combined with laser ablation removal of the antireflection coating, it can provide a thin, and very detailed, front grid without needing costly photolithography. Silver light induced plating is not the only step that can increase the conductivity of the nickel lines. Research has shown that annealing the nickel after plating it, followed by electroplating copper to the cell, dramatically lowers the series resistance present in the metal contact. When comparing this 168

185 nickel and copper process to conventional screen printing, the series resistance dropped by 50% in this new process. [60] This translates to higher fill factors and higher efficiencies for the module. Metal Aerosol Jet Printing Metal ink jet printing is a non-contact printing approach where diluted inks are directed in a stream through a nozzle tip and land on the substrate. Unfortunately this concept has many difficulties with it. One struggle with metal jet printing is that it uses unconventional pastes. With standard pastes, the particles are quite large (5 10 µm) which creates a severe problem of plugging the nozzle. [26] In general, the nozzle should be at least seven times larger than the particle size, and this results in rather large printed grid fingers which are undesirable for solar cells. These line widths were no better than those achieved by screen printing, making this an unattractive industrial process. [58] To combat these problems, an Aerosol printing head was developed by Optomec to aide in the jet printing process. [58] In this design, the printing head is wrapped up in a ring-shaped gas flow which allows for the gas to flow and avoid contact with the nozzle tip, thereby eliminating clogging. [26] The process begins with the atomization of ink producing 1 µm in diameter droplets. [58] These droplets are entrained in a gas stream and delivered to the print head where an annular flow of clean gas is introduced to the stream. This gas eliminates clogging of the nozzle and helps to converge the atomized droplets before they exit the nozzle tip. The resultant stream impinges and adheres to the surface of the substrate allowing for line widths to be considerably smaller than the diameter tip due to the highly focused particles exiting the system. A schematic of an Aerosol jet system is shown below: 169

186 Figure 128: Schematic of a metal Aerosol jet printing head [26] Metal Aerosol jet printing has been found to give fine, continuous line widths of 14 µm with aspect ratios between 1:2 and 1:3. [64] The system is highly sensitive to changes in the velocity of the focusing gas and the temperature of the tip. It was discovered that varying these parameters, the size of the finger widths can vary from µm. [58] Because the focused Aerosol particles exit the tip at a high velocity, approximately 50 m/s, the substrate does not have to be directly below the tip. [58] This makes the process relatively flexible as the working distance can vary by 3 5 mm allowing for nonplanar objects to be printed without increasing the size of the finger width. [58] Fortunately, despite the high exiting velocity, the printing process itself is relatively gentle and does not cause much substrate damage or splattering from excess particles. [58] The process is not limited to just the fingers. By changing the nozzle print head, thicker busbars (~1 mm) could also be printed in a single pass in approximately 3 seconds per cell. [58] The Aerosol jet process is combined with light induced chemical plating to create fingers with desirable aspect ratios and low series resistance. Efficiencies in the range of % were achieved 170