Coursehero >> Texas >> Texas >> JAMIESONAT 042

Course Hero has millions of student submitted documents similar to the one below including study guides, homework solutions, papers, exam answer keys and textbook solutions.

...by Copyright Andrew Thomas Jamieson 2004 The Dissertation Committee for Andrew Thomas Jamieson Certifies that this is the approved version of the following dissertation: Top Surface Imaging for Sub-100nm Lithography Committee: C. Grant Willson, Supervisor Roger T. Bonnecaze Benny D. Freeman Gyeong S. Hwang Scott A. MacDonald Top Surface Imaging for Sub-100nm Lithography by Andrew Thomas Jamieson, B.S. Ch.E, B.S. Ch., M.S.E. Dissertation Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy The University of Texas at Austin May, 2004 Acknowledgements I would like to thank my friends, family, coworkers and financial supporters for their help. Without them, I would not be where I am today. The amount of assistance that I received throughout this doctoral program (and life in general) was enormous, and I will be forever grateful. Specifically, I d like to thank the Semiconductor Research Corporation for its generous funding in the form of a graduate research fellowship. I would also like to thank the UT College of Engineering for its Thrust fellowship. Numerous individuals in industry were instrumental in my research. Foremost among them was Scott MacDonald from IBM (now Hitachi). He was a pleasure to work with during his sabbatical at UT. Jeff Byers from KLA-Tencor is a great person and teacher- he taught me an enormous amount about lithography. Adam Pawloski from AMD was great to work with on the base quencher project. I would also like to thank Chris Mack and Roger Bonnecaze for their insightful discussions. Mike Sheehan from DuPont Electronic Materials was generous to allow us to have samples of semiconductor grade polymers. Ralph Dammel of AZ Clariant was helpful in theoretical discussions and in material donations. The collaboration with ETEC was great, and I enjoyed working with the people there. I was fortunate to have the opportunity to run many experiments at International SEMATECH. In the course of these experiments, I received help from countless people. Danny Miller in particular was a pleasure to work withiv he is a great combination of hard work, good ethics and a great personality. Will Conley provided great assistance and guidance. The crew of Vicky Graffenburg, Georgia Rich, Mike Rodriguez, and Shashi Patel are helpful, hard working people that were a great to spend hours gowned up with. David Stark, Paul Zimmerman, Stefan Hein, and Jeff Muete taught us a lot. In the ATDF, Larry Looger and Jordan Owens were generous for allowing to us to use their steppers and tracks. Chari, Dennis Hanslik, and Arnie Ford were very helpful in all they did. The Willson research group was a tremendous place to work for four and a half years; Dr. Willson has created an amazing research group. I am indebted to him for his guidance, friendship, and everything he has made possible. Kathleen Sparks is a great person, both professionally and personally. There have been so many graduate students I ve had the pleasure to work with that it will be hard to not leave anyone out. Without the help of Mark Somervell, my graduate program would ve been longer and more frustrating. His teaching and guidance were excellent. I was fortunate to collaborate with Matt Pinnow. He is a first-rate scientist and person. Without his assistance, this dissertation would be a fraction of what it is. There are a number of other chemists that I worked with closely, including: Charles Chambers, Brian Osborn, Shintaro Yamada, Takashi Chiba, Raymond Hung, Les Carpenter, and HV Tran; a heartfelt thank you to them all. In chemical engineering, I had the pleasure of working with a number of great people. Pavlos was about the best lab-mate a person could have. He is a great person to talk to at any time. Tim Michaelson was an excellent collaborator on many projects, and he played a significant role in much of the work presented v here. Steve Johnson, Matt Schmid, Matt Colburn, Sean Burns, Jason Meiring, EK Kim, Mike Dickey, Michael Stewart, Heather Johnoson, and everyone else in engineering were a pleasure to work side by side for my time there. I was fortunate enough to work with a number of excellent undergraduate students. Zach Hogan played a huge role in my research. He is a top-notch person, both personally and intellectually. Terry Farmer was great to work with. Never before has such a wealth of chemical engineering knowledge, and UT football obsession been combined in one mind. I was also lucky enough to work for shorter times with Ryan Deschner, Rashid Hameed and Cyrus Tabery. I have no doubt that they will all succeed in anything they put their minds to. Finally, and most importantly, I d like to thank my family and close friends. Without their amazing support, I would not be where I am today. I cannot express how grateful I am to my parents for all that they have done for me. Peter, Emily and Mike: thank you for always being there. Thank you Monika for your encouragement, patience and caring. Thank you all again, -ATJ vi Top Surface Imaging for Sub-100nm Lithography Publication No._____________ Andrew Thomas Jamieson, Ph.D. The University of Texas at Austin, 2004 Supervisor: C. Grant Willson Advances in semiconductor microlithography have resulted in reduced transistor dimensions and consequent improvements in chip performance and cost. In the microlithographic process a photoactive material called a photoresist is uniformly spun cast on a substrate and selectively exposed to radiation, causing a chemical change in the exposed areas. The polymer is subsequently removed in either the exposed or unexposed regions, typically using an aqueous base developer. An alternative to the traditional lithographic method is a process called Top Surface Imaging (TSI). TSI has a number of advantages over traditional base-development techniques but is highly susceptible to line edge deformities commonly referred to as line edge roughness (LER). In TSI, the top surface of the photoresist is exposed to radiation, resulting in the generation of reactive sites. A gas phase, silicon-containing compound called a silylation agent reacts with vii these sites, causing selective incorporation of silicon. The silicon then acts as an etch mask in an anisotropic oxygen etch process. As the industry continues to improve resolution by shifting to shorter wavelengths, TSI s lax transparency requirements provide it with a distinct advantage over traditional lithographic techniques. In this work, TSI was evaluated for use with three industrially relevant radiation sources, including 157nm light, low-voltage electron beams, and extreme ultraviolet light. An investigation into the origins of low frequency LER in TSI systems showed it to be the result of surface tension induced capillary instabilities (also known as a Rayleigh instability). The polymer contribution to LER was investigated by the synthesis of numerous high-Tg TSI polymers. Although many of these polymers are capable of producing high-resolution features, they all suffer from significant levels of LER. The kinetics of the gas phase reaction of the TSI polymer poly(hydroxystyrene) and the silylation agent dimethylaminodimethylsilane was investigated using variable angle spectroscopic ellipsometry, and was found to be front propagated and reaction limited. It was found that the addition of base quenchers to chemically amplified photoresists greatly minimizes LER, and a theoretical approach to understand this effect is presented. Overall, TSI performs well, but the nature of LER in TSI systems is still not fully understood. viii Table of Contents List of Tables......................................................................................................... xv List of Figures ...................................................................................................... xvi Chapter 1. Introduction to Microlithography and Thin layer Imaging .................. 1 Background .................................................................................................... 1 Device scaling and economics .............................................................. 1 Microlithographic Process..................................................................... 2 Photoresist Chemistry ........................................................................... 4 Thin Layer Imaging............................................................................... 7 2. Top Surface Imaging by Vapor-Phase Silylation.................................... 13 Introduction to Top Surface Imaging .................................................. 13 1. Application of TSI to New Exposure Sources. .............................. 22 Goals / Motivation:..................................................................... 22 Accomplishments: ...................................................................... 23 2. Investigation of Mechanisms of Line Edge Roughness Formation in TSI ........................................................................ 24 Goals / Motivation:..................................................................... 24 Accomplishments: ...................................................................... 25 3. Development of New TSI Polymer Systems................................. 26 Goals / Motivation:..................................................................... 26 Accomplishments: ...................................................................... 27 4. Fundamental Investigations of the Silylation Process.................... 27 Goals / Motivation:..................................................................... 27 Accomplishments: ...................................................................... 29 Chapter 2: Top Surface Imaging at 157 nm .......................................................... 31 Abstract: ....................................................................................................... 31 Introduction: ................................................................................................. 32 ix Experimental and Materials ......................................................................... 35 Materials.............................................................................................. 35 Experimental Apparatus...................................................................... 37 Results and Discussion................................................................................. 39 Imaging Experiments .......................................................................... 39 Top Surface Imaging of tBOC Styrene ...................................... 39 Top Surface Imaging of PFAS ................................................... 42 Top Surface Imaging of Poly(NBHFA) ..................................... 44 Contrast Curves ................................................................................... 48 Generation of contrast curves..................................................... 48 Discussion of tBOC Styrene contrast curve............................... 49 Discussion of contrast curves for PFAS..................................... 51 Discussion of contrast curves for poly (NBHFA) ...................... 53 Conclusions .................................................................................................. 54 Chapter 3: Low-voltage Electron Beam Lithography Resist Processes: Top Surface Imaging and Hydrogen Silsesquioxane Bilayer.............................. 56 Abstract: ....................................................................................................... 56 Introduction: ................................................................................................. 57 a. Motivation for low-kV electron beam exposure tools.................... 57 b. Hydrogen Silsesquioxane ............................................................... 59 c. Top Surface Imaging ...................................................................... 61 Experimental and materials:......................................................................... 62 a. Exposure Conditions and Metrology.............................................. 62 b. Processing Conditions and Materials ............................................. 62 Results and Discussion:................................................................................ 64 a. HSQ Single Layer Imaging ............................................................ 64 b. HSQ Bilayer Imaging..................................................................... 66 c. HSQ Proximity Effects................................................................... 68 d. Exposure Latitude ......................................................................... 71 x e. Effect of Base Quenchers on Top Surface Imaging Resolution and Line Edge Roughness .......................................................... 74 f. Low voltage electron beam exposure using TSI............................. 78 Conclusions .................................................................................................. 80 Chapter 4: Surface Energy Induced Low Frequency Line Edge Roughness in Top Surface Imaging .................................................................................... 83 Abstract: ....................................................................................................... 83 Introduction .................................................................................................. 83 Experimental Section ................................................................................... 89 Results .......................................................................................................... 90 Theoretical Analysis..................................................................................... 93 Case A: Fixed Footprint Analysis ...................................................... 95 Case B: Variable Footprint Analysis.................................................. 98 Discussion .................................................................................................. 100 Conclusions ................................................................................................ 102 Chapter 5: Synthesis and Evaluation of Polymers for Top Surface Imaging Applications. .............................................................................................. 104 Background: ............................................................................................... 104 Section A: Polymers examined in this thesis. ........................................... 105 Introduction to TSI polymers: ........................................................... 105 Phenolics ........................................................................................... 106 Fluorinated Norbornenes................................................................... 107 Motivations for each polymer: .......................................................... 112 Section B: Typical Polymer evaluation..................................................... 116 Experimental ..................................................................................... 117 Synthesis of High MW MAST................................................. 117 N-(p-Hydroxyphenyl)maleamic Acid (1)................................. 119 N-(p-Acetoxyphenyl)maleimide (2)......................................... 119 Poly(p-(acetoxystyrene-alt-N-(p-(acetoxyphenyl)maleimide) (3) .................................................................................... 119 xi Poly(p-(hydroxystyrene)-alt-N-(p-hydroxyphenyl)maleimide) (4) .................................................................................... 120 Poly (p-(tert-butoxycarbonyloxy) styrene-alt-N-(p-(tertbutoxycarbonyloxy) phenyl) maleimide) (tBOCMAST) (5)............................................................... 120 Synthesis of Low Molecular Weight MAST..................................... 121 Low molecular weight poly(p-(acetoxystyrene-alt-N-(p(acetoxyphenyl)maleimide) (3) ....................................... 121 Low Molecular Weight Poly(p-(hydroxystyrene)-alt-N-(phydroxyphenyl)maleimide) (4) ....................................... 122 Low Molecular Weight Poly(p-(tert-butoxycarbonyloxy) styrene-alt-N-(p-(tert-butoxycarbonyloxy) phenyl) maleimide) (t-BOCMAST) (5)........................................ 122 Evaluation of Photochemistry and Silylation.................................... 122 Ellipsometry of MAST Polymer ....................................................... 125 Tg Measurement by Ellipsometry...................................................... 126 Tg Sample Preparation and Analysis........................................ 127 Results ............................................................................................... 128 Discussion of Tg measurements........................................................ 130 Exposure and Processing for High Molecular Weight Polymer ....... 131 SEM Images and Results for Low MW MAST ................................ 133 Discussion and Conclusion on Maleimide Polymer .................................. 135 Chapter 6: The Kinetics of poly(hydroxystyrene) Silylation by Dimethylaminodimethylsilane. .................................................................. 137 Abstract: ..................................................................................................... 137 Background: ............................................................................................... 138 Experimental ..................................................................................... 142 Ellipsometry and Silylation Chamber ............................................... 142 Materials and Processing............................................................................ 145 Discussion of Results ................................................................................. 150 Conclusions ................................................................................................ 152 xii Chapter 7: The Effects of Base Additives on Contrast in Acid-Catalyzed Chemically Amplified Photoresists............................................................ 155 Abstract ...................................................................................................... 155 Background: ............................................................................................... 155 Theory ........................................................................................................ 160 Graphical Model................................................................................ 162 Simple Kinetics Model...................................................................... 165 Experimental .............................................................................................. 168 Materials............................................................................................ 168 Processing.......................................................................................... 170 Discussion of Results ................................................................................. 175 Conclusions ................................................................................................ 176 Chapter 8: Summary & Future Work................................................................. 178 Summary .................................................................................................... 178 Discussion / Conclusions ........................................................................... 179 Future work: ............................................................................................... 181 Background ....................................................................................... 181 Future Work Topics .......................................................................... 183 1. Fundamental investigations of line edge roughness & the deprotection profile: ........................................................ 183 2. Basic simulations of the deprotection process .................... 184 3. Non-chemically Amplified TSI schemes. ........................... 185 Appendix A: Silylation Chamber Construction ................................................. 189 Background ................................................................................................ 189 General Design Considerations .................................................................. 190 Specifics ..................................................................................................... 194 Chamber ............................................................................................ 194 Valve Structure.................................................................................. 195 Thermal Calculations ................................................................................. 197 xiii General Heating Background ............................................................ 197 Chamber temperature uniformity ...................................................... 198 Sample Stage Temperature Uniformity............................................. 200 Agent Thermal Expansion Issues ...................................................... 200 Automation................................................................................................. 203 Control Box ................................................................................................ 205 LabView Program ...................................................................................... 206 Safety.......................................................................................................... 208 Typical Operation....................................................................................... 210 Agent Loading................................................................................... 210 Sample loading and measurement..................................................... 210 Hotblock v.II Construction......................................................................... 211 Background ................................................................................................ 211 Construction and Schematic.............................................................. 212 Bibliography........................................................................................................ 216 Vita .......223 xiv List of Tables Table 3.1: Proximity effect induced iso-nested bias: simulated vs. experimental. .................................................................................... 71 Table 3.2: Experimental exposure latitude for various imaging conditions vs. theoretically predicted values........................................................... 74 Table 5.1: Summary of polymer development & evaluation ............................ 112 Table 5.2: Tg of Polymers................................................................................... 130 Table A.1: Heating Configurations ................................................................... 197 Table A.2: SCXI output connections ................................................................. 205 Table A.3: SCXI input Connections.................................................................. 206 Table A.4: PID parameters for the various heating regions. ............................. 207 xv List of Figures Figure 1.1: The microlithographic process (ion implantation is just an example of many possible processing steps that could occur using the resist as a mask). ................................................................. 3 Figure 1.2: Schematic of thin layer imaging technique. ....................................... 7 Figure 1.3: Schematic of a positive tone bilayer imaging process...................... 10 Figure 1.4: Schmatic of Infineon s CARL Process............................................. 12 Figure 1.5: Process Schematic for Negative Tone TSI System ........................... 15 Figure 1.6: High resolution, high aspect ratio features printed with a TSI scheme employing t-BOC styrene with binary illumination at 193nm............................................................................................... 17 Figure 1.7: Top-down scanning electron micrographs of line-edge roughness in t-BOC stryene, a) low levels of LER (but still unacceptable), largely at footing of lines. ................................................................ 17 Figure 1.8: PFAS and DMADMDS. ................................................................... 19 Figure 1.9: Smooth Imaging performed with PFAS. .......................................... 19 Figure 2.1: Process Schematic for Negative Tone TSI System ........................... 34 Figure 2.2: The three polymer systems used in this study. ................................. 37 Figure 2.3: Absorbance (a) t-BOC styrene, (b) PFAS, and (c) poly(NBHFA as a function of wavelength.............................................................. 40 Figure. 2.4: Top down and Tilted SEM of tBOC Styrene imaged with a binary mask at 157 nm. .................................................................... 41 xvi Figure 2.5: Top down and Tilted SEMs of PFAS imaged at 157 nm with a binary mask. All features are 270 nm tall ....................................... 43 Figure 2.6: Infrared Spectra of poly (NBHFA) (a) after coating, (b) after exposure and PEB and (c) after silylation with DMADMS............. 45 Figure 2.7: Top down and Tilted SEMs of poly (NBHFA) imaged at 157 nm with a phase-shifted mask. All features 450 nm tall. ...................... 47 Figure 2.8: Contrast curves for tBOC Styrene ................................................... 49 Figure 2.9: Contrast curves for PFAS ................................................................. 51 Figure 2.10: Contrast curves for PFNA .............................................................. 53 Figure 3.1: Chemical structure of HSQ............................................................... 60 Figure 3.2: HSQ images generated at 1kV. a) 33nm thick resist exposed at 44 C / cm2; note the inadequate penetration of the 1kV electrons into the resist, resulting in delamination. b) 24nm thick resist exposed at 40 C / cm2. Image shows 50, 40 and 30nm lines on a 1:1 line:space pitch........................................................................... 65 Figure 3.3: 60nm and 50nm nested features. ....................................................... 67 Figure 3.4: Line-width vs. Dose for nominal 70nm isolated and nested lines exposed at 1,2 and 3 kV. .................................................................. 68 Figure 3.5: Energy deposition profiles of 1,2 and 3 kV electrons in HSQ. Nested lines are on the left; isolated are on the right (Units on energy contours are J / cm3 for a dose of 30 C / cm2). ................... 69 Figure 3.6: Energy deposition profiles at 1, 2, and 3kV (dotted lines show feature edge). .................................................................................... 73 xvii Figure 3.7: Effect of variable TOA and TPS-Nf concentration on LER. Nested 200nm lines exposed with 248nm illumination using a) 2wt% TPS-Nf / 10mol% TOA, b) 6wt% TPS-Nf / 30mol% TOA, c) 6wt% TSP- Nf / 50mol% TOA. ..................................... 77 Figure 3.8: High resolution imaging using TSI at 1kV. a) 60 nm nested features in ~120 nm of resist, b) 50 nm 1:1.5 features in ~120nm of resist c) 40 nm 1:1.5 features in ~50nm of resist. ..................... 79 Figure 4.1. 120 nanometer lines and spaces generated with top surface imaging, showing significant low frequency LER 6. ....................... 85 Figure 4.2. Process flow for a typical top surface imaging process. A resist film is coated, selectively exposed to light, silylated and Oxygen etched, forming high-resolution features. ........................................ 86 Figure 4.3. Scanning electron micrograph of a bank of lines prior to etch demonstrating the formation of truncated cylindrical crosssections. ............................................................................................ 87 Figure 4.4. Two perturbations on a truncated cylinder: (a) a perturbation on the width of the cylinder footprint, with a constant contact angle, and (b) a perturbation on the contact angle of the cylinder, , on a fixed footprint................................................................................ 89 Figure 4.5. (a) A tilted cross-sectional micrograph showing smooth truncated cylinders created with TSI and (b) a tilted cross-sectional micrograph showing truncated cylinders at a larger level of swelling that appear to have undergone a capillary instability. ....... 91 xviii Figure 4.6. a) A SEM of a line space pattern created by TSI. This pattern was held at high temperature for a prolonged period of time, allowing the sample to achieve a more stable configuration. b) A photograph of water in a macroscopic fluoropolymer trench showing a similar stable configuration............................................. 92 Figure 4.7. The geometry of the cross section of the line of fluid. ..................... 94 Figure 4.8. Surface area change as a function of normalized disturbance wavelength for several initial contact angles ................................... 97 Figure 4.9. Critical Wavelength for the onset of instability for contact angle disturbances...................................................................................... 98 Figure 4.10. Surface area change as a function of normalized disturbance wavelength for several contact angles.............................................. 99 Figure 4.11. Critical wavelength for the onset of instability for variable footprint disturbances ..................................................................... 100 Figure 5.1: Phenolic TSI polymers that were synthesized and evaluated......... 109 Figure 5.2: Fluorinated Norbornene polymers that were synthesized and evaluated......................................................................................... 110 Figure 5.3: Mark Somervell s polymer that were synthesized and evaluated. . 111 Figure 5.5. IR of t-BOC MAST Deprotection and Silylation ............................ 124 Figure 5.6. Two Layer Fit of Silylation Ellipsometry Data ............................... 125 Figure 5.7. Tg Determination of Silylated PHOST at Vacuum .......................... 129 Figure 5.8. SEM of Features Formed with High Molecular Weight MAST...... 132 Figure 5.9. SEM of Features Formed with Low Molecular Weight MAST ...... 134 xix Figure 6.1: The gas phase silylation of poly (hydroxystyrene) with dimethylaminodimethylsilane. ....................................................... 138 Figure 6.2: Schematic diagram of controlled environment silylation chamber for use with Woolham M2000 Ellipsometer. ................................. 144 Figure 6.3: Schematic of the gas manifold for environmental control ............. 144 Figure 6.4: Silylation of poly(hydroxystyrene), as measured by ellipsometry & Infrared Spectroscopy. ............................................................... 146 Figure 6.5: Silylation of poly(hydroxystyrene), (A) as measured by ellipsometry using a two-layer fit & by staining techniques & (B&C) typical SEMs of stained PHS, 3 minute & 8 minute silylations respectively. .................................................................. 147 Figure 6.6: Reaction Rate as a function of temperature and pressure as measure ellipsometrically............................................................... 148 Figure 6.7: Sorption and desorption curve of dimethylaminodimethylsilane in silylated poly (hydroxystyrene) at 60oC..................................... 149 Figure 6.8: Henry s law coefficients as a function of temperature. The points represent experimental data, and the curve represents the simple Flory-Huggins model. .................................................................... 150 Figure 6.9: A semi-log Arrhenius plot of silylation rate divided by the concentration of silyation agent vs. the inverse temperature. ........ 152 Figure 7.1: Arbitrary Aerial Image. .................................................................. 163 Figure 7.2: Chemical Contrast curves for two hypothetical resists. ................. 164 Figure 7.3: Materials used in these experiments ............................................... 169 xx Figure 7.4: Blocking fraction as a function of dose. Solid lines represent theory, triangles denote TOA, squares denote TBAH, and triangles triethanolamine. ............................................................... 172 Figure 7.5: Chemical Contrast as a function of dose. Solid lines represent theory, triangles denote TOA, squares denote TBAH, and triangles triethanolamine. ............................................................... 173 Figure 7.6: Top down and cross-section SEMs of 180nm nested lines with both the zero base (a & b), and 20mol% TOA (c & d) resists. Note the improved sidewall angles and lower line edge roughness in the 20mol% base case. .............................................. 174 Figure A.1: Schematic of ellipsometry sample chamber .................................. 192 Figure A.2: Schematic of Valve Stucture ......................................................... 195 Figure A.3: Temperature distribution simulation of the Stainless Steel sample chamber.............................................................................. 199 Figure A.4: Screen shot of the control software user interface......................... 207 Figure A.5: Schematic of the hotblock v.II........................................................ 213 Figure A.6: Temperature heating profiles in the hotblock as a function of time for five locations............................................................................. 215 xxi Chapter 1. Introduction to Microlithography and Thin layer Imaging BACKGROUND Device scaling and economics Semiconductor manufacturing is a highly competitive industry in which device cost and speed are the major economic driving forces. The competitiveness of the industry was described by Gordon Moore from Intel, when he predicted that the number of transistors per chip would double approximately every year and half 1. This relation has become known as Moore s law . Any company that cannot keep pace will fall behind and not be competitive. Moore s law has held true for nearly 40 years, and shows no signs of ending in the near future. The average cost per chip has remained nearly constant over that time period, so the cost per transistor has plummeted. This amazing accomplishment is based on several factors. First, the size of the silicon wafers, which chips are made from has gradually increased from 1 to 12 , while the cost per wafer for most processing steps hasn t changed dramatically. Second, the size of the average chip has increased, increasing the area on which transistors can be placed (although this has stabilized in recent times). Finally, and most importantly, the size of transistors has shrunk dramatically. Printing smaller features on a wafer (and thus smaller transistors) has enabled more transistors to be placed on a given wafer area, and thus more transistors are produced per processing step. Smaller transistors are faster and cheaper to make. Chips with small transistors are both 1 faster and have more transistors. Intel s current Pentium 4 chip has features that are 130nm, possess 42 million transistors per chip and operates at speeds of over 2 gigahertz. This amazing miniaturization has been the result of the patterning process called microlithography. Microlithographic Process Microlithography is a process that selectively places patterns of material (polymer) on a surface. The polymer acts as mask for subsequent semiconductor processing steps, protecting the underlying region. For instance, during ion implantation, the polymer shields the underlying silicon from dopants that are accelerated at the wafer surface, resulting in selective doping of certain regions of the silicon. The microlithographic process is illustrated schematically in figure 1.1. First, a solution of a polymer is evenly spin coated onto a wafer surface. The solution is typically made up of a polymer, a moderately volatile solvent, a photo active compound and other additives. After coating, the wafer is baked (called a post-apply bake, or PAB), causing the solvent to evaporate, and resulting in a uniform, thin film of polymer on the wafer. The wafer is then selectively exposed to radiation. The radiation interacts with the photo-sensitive coating, resulting in a chemical change within the exposed region. The film is then baked again (called the post-exposure bake, or PEB), resulting in limited diffusion of the exposed photoactive compound (this is necessary to help remove undesired optical effects called standing waves, and in the case of some resists, to allow a catalytic reaction to occur within the exposed region). The chemical change 2 within the exposed areas allows the material to be selectively removed (usually by dissolution into an aqueous base solution). This occurs by either removing the exposed areas (called positive tone) or by removing the un-exposed areas (called negative tone). Resist Substrate h Mask (Positive Tone) (Negative Tone) Expose Coat & PAB PEB & Develop Ion Implant Strip Resist Figure 1.1: The microlithographic process (ion implantation is just an example of many possible processing steps that could occur using the resist as a mask). The ultimate goal of microlithography is to produce features as small as possible. In optical exposure, the feature size is limited by Rayleigh s equation: k (1.1) Minimum feature size = 1 NA where k1 is a proportionality constant, is wavelength of light, and NA is the numerical aperture of the lens. This equation holds true for a pattern of lines and 3 spaces of equal width. The semiconductor industry has decreased this minimum feature size aggressively by changing all three parameters. k1 has been decreased by improved resist performance, and moving to alternate exposure techniques (such as off-axis illumination and phase-shifting), but it is quickly reaching the theoretical limit of 0.5 for partially coherent light. The numerical aperture of lenses has been increasing, with current state-of-the-art tools having an NA of 0.8. This too is approaching the theoretical maximum of 1 for tools (NA = n sin , where n is the index of the medium, 1 in this case for air, and is the diverging angle from the normal of the sample to the edge of the lens, with 90o being the maximum). The exposure wavelength has been continuously decreased from the Hg arc lamp bands of 436nm and 365nm, down to excimer laser wavelengths of 248nm, 193nm and 157nm. Current state-of-the-art production tools are just now beginning to use 193nm light, and research is underway to develop either a viable 157nm process or a 193nm process with water in between the photoresist and the bottom lense element (increases NA by increasing the index of refraction). Photoresist Chemistry Typically, the subsequent processing steps require the polymer to have certain material properties. The process with the most stringent requirements is plasma etching. In order for a polymer to adequately act as an etch mask, it must have a relatively low etch rate in plasma environments, which is accomplished by using polymers that contain cyclic structures and often have relatively high 4 degrees of unsaturation. In addition, the resist coating must be of sufficient thickness that it will not be entirely consumed during the etch processes. In traditional resist processes, the light renders the exposed resist either soluble or insoluble in aqueous base developer. The photoresist must be relatively transparent at the exposure wavelength in order to have the solubility switch occur throughout the thickness of the film. This requirement has grown progressively more challenging as imaging wavelengths have decreased. In fact, most polymeric materials are extremely absorbing in the UV, and significant research has been performed to create materials that will function at 248nm, 193nm and 157nm. One can avoid this problem by going to thinner and thinner films; however, the plasma etch process requirements typically place a minimum value upon thickness. Microlithography with wavelengths of 365nm or longer was performed using diazonapthoquinone (DNQ)/ novolac based resists. Novolac is a polymer that is inherently soluble in aqueous base developer. DNQ is a photoactive compound that when unexposed, inhibits the dissolution of novolac. Upon exposure, the DNQ undergoes a chemical reaction to form a compound that enhances the solubility of the novolac. Thus, the resist system is positive tone. It takes at least one photon to chemically alter one DNQ molecule. The reader is referred to Ralph Dammel s book2 for a thorough discussion of DNQ / Novolac resists. Unfortunately, novolac is not sufficiently transparent for use at 248nm, so the industry was forced to change to a new polymer resin. Additionally, for 5 various reasons (including throughput), it was decided that the new photoresist should be more sensitive to light than the DNQ / Novolac system. It was quickly found out that poly(hydroxystyrene) (PHS) was transparent at 248nm, and soluble in base developer. Unfortunately, DNQ does not inhibit the dissolution of PHS, and no other dissolution inhibitor could be found. Instead, Ito and Willson proposed a different mechanism for introducing a solubility switch to this system3. Called chemical amplification , the process consisted of placing an acid labile protecting group on the acidic hydroxyl group, and incorporating a photoacid generator into the resin solution. Upon exposure, acid is generated in the film, and during the post-exposure bake the acid catalytically removes the protection group. Thus, within the exposed areas, soluble PHS is generated. Reaction A illustrated this scheme using t-butoxycarbonyl (t-BOC) on PHS. Chemically amplified resists are predominant in the semiconductor industry for imaging at wavelengths of 248nm and shorter. + O O O H+(catalyst), heat CO2 OH Reaction A 6 Thin Layer Imaging Thin layer imaging (TLI) techniques are alternatives to traditional single layer processes and have long been considered for industrial implementation4. Fundamentally, a TLI technique is one in which the imaging is performed in a very thin section at the top of the resist, and then the image is transferred anisotropically down through the resist, usually utilizing reactive ion etch processes. Schematically, the process is shown in figure 1.2. Imaging Layer Coat & PAB Substrate h Mask Expose Hardmask Generation Oxygen Etch Figure 1.2: Schematic of thin layer imaging technique. 7 There are many advantages to thin layer imaging processes. First, the transparency requirements are dramatically lower. One only needs to have good energy deposition profiles in the top ~150nm of the resist in order to generate tall features. The height of the final profile is dictated by the transfer layer (the layer which is etched through), rather than the imaging depth. Second, thin layer imaging techniques do not suffer from feature collapse due to capillary forces, as occur in traditional wet-developed resist systems. In single layer systems, during the drying step after development, water between lines can pull features down if the pitch (distance between the centers of sequential lines) is too close. Finally, since imaging must be performed in a thinner film, these systems should perform with better depth-of-focus (errors in focus of the stepper, or topography can cause features to print poorly; a better depth-of-focus allows for better process control). Unfortunately, these thin film imaging techniques also suffer from added complexity. At the bare minimum, they require a plasma etch step, and often include additional coating steps, or gas phase reaction steps. As such, they have been slow to be accepted into practice. However, as lithography processes grow more and more challenging, people have grown more accepting of these techniques as alternatives, especially if no single layer process exists that will work for a given application. Two thin film imaging techniques are used in industry today: bilayer processes and the CARL process. A bilayer technique is shown schematically in figure 1.3. It involves the coating of two layers. The bottom layer is typically a non-functional polymer (often novolac) that contains only carbon, hydrogen and oxygen. The layer is 8 usually baked at a high temperature after coating, resulting in cross-linking. The top layer is a thin (~150nm), functional resist that typically contains at least 12% silicon in addition to carbon, hydrogen and oxygen. The stack is exposed, and the top layer undergoes a solubility switch. The material is developed in aqueous base, resulting in silicon containing features atop a uniform transfer layer. The stack is then etched in a reactive ion etcher containing an environment of mainly oxygen gas. The silicon containing resist quickly oxidizes to silicon dioxide, which is inert to oxygen plasma, while the regions that do not possess the top resist are quickly etched away, forming water and carbon dioxide. One can tune the etch process to perform the etch anisotropically, resulting in vertical features. This technique is shown schematically in figure 1.3. 9 Imaging Layer Transfer Layer Substrate h Mask Coat & PAB Expose PEB & Develop Oxygen Etch Figure 1.3: Schematic of a positive tone bilayer imaging process. Bilayer processes have all the aforementioned advantages of thin layer imaging techniques. Typically, they are used in industrial processes which require exceptionally high aspect ratio features patterned over topography. The major drawbacks to the techniques are the addition of an extra coating step, and a plasma etch step, both of which are not required with traditional resists (although most resists require both the spin-casting of an anti-reflection coating and a trim etch step after development, so bilayer resists don t add too much complexity). A further drawback is the added potential for contamination of the bottom lens 10 element of an exposure system with silicon. During exposure, small amounts of resist can degrade and outgas into the volume above the resist. These materials can contaminate the lens, degrading optical performance, which necessitates cleaning of the lens. Unfortunately, contamination by silicon onto the fused silica lens element is nearly impossible to clean. This results in costly replacement of the bottom lens element. As such, bilayer processes are considered moderately controversial in many circles. The second thin layer imaging process that is practiced in industry is called Chemical Amplification of Resist Lines (or CARL). Originally developed by Infineon, this process is similar to the bilayer process, except for the fact that the silicon is incorporated into the top resist in a liquid silylation step after base development. The process is shown schematically in figure 1.4. 11 Imaging Layer Transfer Layer Substrate h Mask Expose Coat & PAB PEB & Develop Liquid Silylation Oxygen Etch Figure 1.4: Schmatic of Infineon s CARL Process. Clearly, CARL has significantly more process complexity than a traditional resist. It incorporates one extra coating step, one extra etch step, and a liquid silylation step. It possesses all the benefits of the bilayer process, with two additional advantages. First, the top layer doesn t contain nearly as much silicon as is present in bilayer, thereby decreasing the chances of uncorrectable lens contamination. Second, and more important, is a result of the liquid silylation step. During the silylation step, the resist swells in size, both vertically and horizontally. This allows lithographers to print a space at the Rayleigh limit, and then shrink the space further. So, they are able to print so-called trenches beyond the Rayleigh limit of their tool. Alternatively, they can over-expose the 12 resist, making the pre-silylated images smaller than they desire, and then swell the images up to the size they want. The advantage of this over-exposure is that typically the so-called the iso-focal dose - the dose at which focus errors minimally affect feature size - occurs at an over-exposed dose. Therefore, they are able to gain significant process latitude by using this thin layer imaging process. 2. TOP SURFACE IMAGING BY VAPOR-PHASE SILYLATION Introduction to Top Surface Imaging Top surface imaging (TSI) is a thin film imaging technique with a fairly simple process flow. As opposed to the bilayer and CARL processes discussed earlier, TSI only employs a coating step, an exposure step, a gas phase reaction step and an oxygen etch step. The fundamental idea behind TSI is that a single resist film is coated on a substrate and exposed to light or electrons in such a way that only the top of the resist is chemically altered (either by working at a wavelength at which the resist absorbs, or at an electron energy where the penetration depth is shallow). The resist is then put in the presence of a silylation agent, which causes silicon to be incorporated into either the exposed or unexposed region. The film is then etched in an oxygen plasma, where the incorporated silicon acts as a hard mask, protecting that region. In the TSI system studied here, silylation contrast is created by a photoinduced switch in the chemical reactivity of the polymer. A polymer that is 13 unreactive to vapor phase silylation is rendered reactive through a chemically amplified mechanism. Fig. 1.5 shows how this process works using poly(4-tbutoxycarbonyloxystyrene), also referred to as t-BOC styrene5. In this system the polymer has two states, reactive and unreactive. Since the polymer is always in one of these two states, the silylation is digital. The other feature is subtle but important. Exposure and heat causes deprotection (which changes the polymer s reactivity to silylation) with concurrent loss of volatile products. As a result of this volatilization, the polymer film shrinks in the regions of exposure. The silylation reaction, however, changes the mass and volume of the polymer substantially and causes it to swell. The net change in thickness is small in the case where dimethylaminodimethylsilane, a silylation agent, is used, which results in very little image deformation throughout the process. 14 E x p o se PAG n h H+ n Bake O C O 2 C (C H 3 ) 3 n H+ OH n R 3 S iN R ' 2 OH O S iR 3 + + CO2 + S ily la te N R '2 H E tc h silyla ted u n silyla ted S iO 2 CO 2 + H2O Figure 1.5: Process Schematic for Negative Tone TSI System 15 Top surface imaging using t-BOC styrene has produced high resolution, high aspect ratio images when exposed with 248nm and 193nm light6. Figure 1.6 shows images patterned at 193nm. As can be seen, the images have vertical sidewalls, and are patterned near the resolution limit of the exposure tool. Unfortunately, upon closer inspection, the images suffer from significant levels of line edge roughness (LER), as is shown in top-down scanning electron micrograph in figure 1.7. 16 0.15 m 3.8 mJ/cm2 0.11 m 6.2 mJ/cm2 90 nm 6.2 mJ/cm2 Figure 1.6: High resolution, high aspect ratio features printed with a TSI scheme employing t-BOC styrene with binary illumination at 193nm. Figure 1.7: Top-down scanning electron micrographs of line-edge roughness in t-BOC stryene, a) low levels of LER (but still unacceptable), largely at footing of lines. 17 The fundamental nature of line edge roughness in TSI systems is still poorly understood, and is subject of paramount interest. Fortunately, a TSI Originally system has been developed which possesses nearly no LER6. discovered by Mark Somervell, the system employs a co-polymer of tertbutyloxycarbonyl protected 3-bicyclo[2.2.1]hept-5-ene-2-yl)-1,1,1-trifluoro-2- (trifluoromethyl)propan-2-ol (tBOC-NBHFA) and sulfur dioxide. The polymer is shown in figure 1.8 and is henceforward referred to as PFAS. The norbornene ring is employed for etch resistance in post-lithography processing. The hexafluoroalcohol unit was used due its pKa of 12, which behaves similarly to poly(hydroxystyrene) in reactions with silylation agents. The sulfone linkage was incorporated largely out of necessity in the polymerization process, as techniques had not been developed to homopolymerize NBHFA (this has subsequently changed, and will be discussed later). Due to the increased mass of the repeat unit as compared to poly(hydroxystyrene), a new silylation agent was necessary in order to increase the silicon content of the fully silyated polymer above 12% (wt) (the threshold level of silicon above which the features will not etch in an oxygen plasma environment). David Wheeler of Sandia National Laboratories generated a synthetic route to create dimethylaminodimethyldisilane (DMADMDS), also shown in figure 1.8. This polymer system is capable of printing with very smooth images, as is shown in figure 1.9. 18 O S O n H Me Me Si Si NMe2 H H O CF3 CF3 O O Figure 1.8: PFAS and DMADMDS. Figure 1.9: Smooth Imaging performed with PFAS. 19 The fundamental material property differences between these two polymers are many. However, if one could determine the critical parameters that enable the second system to print as cleanly as it does, one might be able to apply these lessons to the design of future TSI systems. PFAS suffers from additional problems which would make it difficult to use in industry, such as cleavage of the backbone which releases sulfur-dioxide. Additionally, the challenging synthetic route to DMADMDS greatly limits its availability, and thus the potential for process optimization. As such, generation of new polymer systems would be extremely beneficial. It was believed that two material properties were largely responsible for the smooth imaging behavior of PFAS. The first property is the high glass transition temperature of the polymer, especially in its silylated state. The Tg of silylated PFAS is around 120oC (the Tg of silylated poly(hydroxystyrene) is ~70oC). This Tg is substantially higher than the silylation process temperature of 90oC. The structures that are created in lithographic processes are thermodynamically unstable, and processing above the Tg allows material flow, usually with poor results. The fundamental mechanisms of this material flow are not well understood in relation to these systems. The second material property that is dramatically different in these two systems is the opacity of the polymers. PHS is extremely absorbing at 193nm, and as such, the feature edge is poorly defined. The random nature of the deprotection reaction and the etch process could stochastically degrade this poor feature edge, resulting in roughness. PFAS on the other hand is fairly transparent, and the resulting feature edge is more 20 clearly defined. Figure 1.10 shows energy deposition profiles in the two resists at 193nm. A B Figure 1.10: Energy deposition profiles in a) PHS and b) PFAS (red corresponds to high relative intensity, while blue represents low relative intensity). Note the clear edge definition in the more transparent polymer on the right. 3. RESEARCH OBJECTIVES & ACCOMPLISHMENTS: The goals of this dissertation were originally stated in my preliminary document. They were broken into four distinct area of research, including: 1. Application of TSI to new exposure sources. 2. Investigation of mechanisms of line edge roughness in TSI. 3. Development of new TSI polymer systems. 4. Fundamental investigations of silylation reaction. All of these areas were studied thoroughly, and a brief synopsis of the goals and accomplishments of each area are shown below. 21 1. Application of TSI to New Exposure Sources. Goals / Motivation: Resists and microlithography processes at current commercial exposure wavelengths are well developed and optimized. Resist processes for future exposure sources are largely undetermined. Ideally, industry would prefer to work with a traditional single layer process, but that is often very difficult. Thus, it is extremely important to study alternate techniques for use in these next generation lithography (NGL) systems. From a researcher s perspective, these NGL exposure sources are capable of printing much higher resolution features than are currently available at today s wavelengths of 248nm and 193nm. Working with higher resolution features and different energy deposition profiles in general allows investigation of fundamental mechanisms of resist behavior that cannot be seen when working with larger features. Therefore, the first area of research that I planned to investigate was the application of TSI to alternate exposure sources. There are many exposure sources that have been proposed for future industrial use. Three exposure sources that are particularly interesting to investigate are 157nm light, low-voltage electron beams (1-3kV) and extreme ultra violet (EUV, or 10-14nm). 157nm light is slotted to be the next major wavelength that will be used in industry, and is currently slated to go into production at the 70nm feature node, which will occur sometime around 2006. International SEMATECH in Austin has a 157nm stepper that we have access to, and TSI applications can be investigated thoroughly with this exposure source. Low voltage electrons have been suggested 22 for use in mask making applications due to their amenability for use in microcolumn arrays (multiple exposure sources simultaneously exposing a single sample). A major challenge associated with application of this technique lays in finding a suitable resist system. Low voltage electrons have very shallow penetration depths (~25nm for 1kV electrons), and as such, a thin film imaging technique is absolutely necessary to produce adequate etch resistance for later processing. Top surface imaging is particularly well suited for this application, as even bilayer applications may not work here since spin coating films of less than 70nm is difficult due to thin film instabilities. Additionally, electron beam exposure often has very high resolution, so probing fundamental mechanisms of TSI is particularly possible. EUV is currently slated for introduction to industrial practice after 157nm lithography, and as such, is a wavelength whose resist processes haven t been well investigated. Accomplishments: In the course of this research, TSI was applied to nearly every next generation lithography (NGL) exposure source that is presently being considered by industry. These include 157nm light, 13nm light (aka Extreme Ultraviolet (EUV) and low voltage electron beam exposures. TSI performed fairly well in all of these studies, but in the end still suffered from LER (although much headway was made in the understanding of this area). The work at 157nm is presented in chapter 2. Chapter 3 is a summary of work performed in collaboration with ETEC systems in Hayward, CA on the application of TSI to low-voltage electron 23 beam exposure sources. Along with TSI, studies based on another promising thin-film imaging technique, HSQ bilayer, are presented. Preliminary studies on the utility of TSI for use as a EUV resist were performed, but is not presented here since it did not provide much new knowledge. Overall, TSI performed adequately in these new applications, but suffered from a somewhat larger magnitude of LER than is desired. The main benefit of these experiments was the increased knowledge and data that resulted. Although it is often tempting to do so, creating theories on limited amounts of data is a risky proposition. These imaging experiments not only examined the application of the TSI process to new conditions, but created a large body of data from which to work. 2. Investigation of Mechanisms of Line Edge Roughness Formation in TSI Goals / Motivation: Line edge roughness is the largest problem in TSI, and the origins of the phenomenon are not fully understood. This problem caused Intel, Texas Instruments, Bell Labs, and other companies to halt research into TSI. Until the paper of Somervell, et al6 it was thought that TSI and line edge roughness were not separable. The beginning of my thesis was a unique time in research of this area since examples of materials that print roughly and materials that print smoothly were available. The critical material properties related to line edge roughness proposed in the previous section seem reasonable, but a detailed analysis of the effects of aerial image, and low glass transition temperature had 24 not been performed when those theories were made. In fact, when this work began, there were many questions that remained about the nature of line edge roughness that the previous theories did not entirely encompass. For instance, although silylation of PFAS above its Tg at 130oC does produce rough images, silylation of PHS below its Tg of 70oC (say at 50oC) does not produce smooth images. It has been postulated that this results from silylation agent sorption, and hence plasticization, of the resist, but this had not been carefully studied. It was my aim to investigate line edge roughness in TSI systems from a mechanistic standpoint. The goal was to acquire an understanding that could aid in the design of polymers for TSI. Accomplishments: LER in chemically amplified systems in general is an extremely complicated topic that is still poorly understood in the lithography community today. In the author s opinion, LER may become the major obstacle to continuation of Moore's law sometime in the not too distant future, barring a revolution in photoresist design. Since LER in chemically amplified systems in general isn t understood fully, a complete understanding of LER in chemically amplified TSI systems is still limited. Previous work by Somervell indicated that alteration of the polymer structure was capable of fixing LER in TSI in and of itself. Although true for that particular case, duplication of that performance in other polymer systems was elusive. Nearly 20 new TSI polymers were This work is synthesized and none performed nearly as well as PFAS. 25 documented in chapter 5. The hypothesis that the use of high Tg polymers could fix LER was tested and proven incorrect. High Tg does however have the benefit of kinetically quenching phenomena associated with polymer mobility. Chapter 4 shows how mobility can result in alteration of structures by using a Rayleigh Instability analysis. In essence, this chapter shows theoretically that some of the structures that are creacted lithographically are thermodynamically unstable, and need to be kinetically quenched. Finally, the low-kV TSI experiments revealed the amazing effects that base quenchers have on line edge roughness in TSI systems. In fact, base quenchers have dramatic, but poorly understood effects on all chemically amplified systems, base-developed included. Certain aspects of their The work was performance-improving behavior are studied in chapter 7. performed on a base-developed system due to its larger relevance, but applies equally to TSI systems as well. 3. Development of New TSI Polymer Systems. Goals / Motivation: In the end, a working polymer system that uses readily available materials to produce high quality images is the goal of any research program in top surface imaging. As is often the case, sometimes the best approach to solve a problem is to simply go out and experiment. The fundamental understanding that we had at the beginning of our experiments (aim for high Tg materials that are moderately 26 transparent) we hoped would meet the requirements adequately to make a workable TSI process. A major portion of the Willson research group was devoted to the development of single layer resists for application at 157nm. Close collaboration with the skilled chemists working on that project had a good chance of bearing fruit in this area. As such, it was my plan to synthesize, characterize and test new polymers for TSI applications. At the very least, having more data points when it comes to systems that work, and systems that don t, would increase our understanding of the fundamental processes that contribute to line edge roughness. Accomplishments: Numerous novel TSI imaging systems were developed and evaluated, and none of them matched the Line edge roughness present in previous work with PFAS. Chapter 5 summarizes all the polymers that were developed in the course of this dissertation. Overall, it appears as though alteration of polymer structure in nearly all cases doesn t dramatically improve imaging performance. 4. Fundamental Investigations of the Silylation Process Goals / Motivation: A chemically amplified top surface imaging (TSI) scheme consists of three main processing steps. First, one coats a resist, bakes it, exposes it to radiation and then bakes it again, resulting in a film patterned with a latent image 27 of reactive sites. Second, one exposes the resist to a gaseous silylation agent, which incorporates silicon into the reactive sites through covalent bond formation. Finally, one performs an oxygen reactive ion etch on the resist, anisotropically removing resist from the regions that lack silicon. Most TSI systems suffer far greater levels of line edge roughness than wet-developed processes. In order to understand the cause of this problem, one must investigate the process of which we have the least understanding. The first processing step has been studied in great depth by countless researchers around the world. There are still many fundamental questions associated with that step, but overall it has been well investigated. Numerous people studied the third step as well, and although it is not completely understood, the process space in this area has been explored thoroughly. It is likely that these two steps are not the cause of the dramatic line edge roughness found in TSI. The second step however, has not been fully probed, and many fundamental questions remain. The standard techniques that have been used to investigate thin film processes in resists, such as infrared spectroscopy, interferometry, and quartz crystal microbalance are a challenge to apply to in-situ studies of the silylation reaction. In order to get infrared spectroscopy to probe the reaction, one must create a cell with a very limited pathlength in order to bring the film signal out from the background of the gas phase. Interferometry lacks the ability to compensate for changes of index within a film which would likely occur during reaction and swelling. Finally, a quartz crystal microbalance lacks the ability to distinguish between changes in mass and changes in modulus when dealing with 28 sorption in polymer films, and thus lacks the ability to generate quantitative data with any accuracy. Therefore, it was proposed that a technique to investigate the fundamental mechanisms associated with the silylation reaction be developed. The construction of a silylation cell that will work in a real-time Woolham ellipsometer was the planned research route. Accomplishments: A silylation chamber that allows for in-situ ellipsometric monitoring was constructed and automated. The silylation kinetics of poly(hydroxystyrene) were investigated thoroughly and are presented in chapter 6. The reaction was found to proceed by a front-propagated, reaction-limited mechanism. The silylation agent is very soluble in the reacted film, and nearly insoluble in the unreacted film. The solubility was measured by monitoring the unidirectional expansion and contraction of a thin film upon exposure to the silylation agent. 29 References: (1) (2) (3) (4) (5) (6) Moore, G. E. Electronics 1965, 36. Dammel, R. R. Spie Press 1993, TT 11. Ito, H.; Willson, C. G. Polymers in Electronics 1984, ACS Symposium Series 242, 11-23. Thompson, L. F., Willson, C. Grant, Bowden, M. J. Introduction to Microlithography; ACS, 1983. Ito, H.; MacDonald, S. A.; Miller, D. A.; Willson, C. G. US Patent 1985, 4,552,83. Somervell, M. H.; Byers, J.; Willson, C. G. Abstracts of Papers of the American Chemical Society 1999, 218, 29-PMSE. 30 Chapter 2: Top Surface Imaging at 157 nm ABSTRACT: Top surface imaging (TSI) has had an interesting history. This process showed great promise in the late 1980 s and several attempts were made to introduce it to full-scale manufacturing. Unfortunately, defect density problems limited the process and it fell from favor. TSI emerged again as an important part of the EUV and 193 nm strategies in the early stages of those programs because it offered a solution to the high opacity of common resist materials at both wavelengths. A flurry of research in both areas identified new, transparent polymers for single-layer resists, and the seemingly insurmountable problem of line edge roughness in TSI resists, which caused the process to be dropped from both programs. Recent developments in TSI have demonstrated the ability to print high-resolution, high-aspect ratio images at 193 nm with less line edge roughness than typical single layer resist systems. This has largely been due to the development of a polymer specifically tailored for that end use. The 157 nm program has much in common with the early stages of the 193 nm program. The optical density of even 193 nm resist materials at 157 nm is far too high to allow their use in single layer applications. The less stringent optical density requirements of TSI make it a potentially viable imaging scheme for use at 157 nm. Various TSI materials, including the traditional t-BOC styrene, as well as novel aliphatic cyclic polymers bearing bis-trifluoromethylcarbinol substituents, 31 have been investigated for use at 157 nm, and smooth high-resolution images have been generated. INTRODUCTION: The semiconductor industry continues to drive toward production of devices with smaller and smaller critical dimensions. This quest has generated many difficult challenges. When the industry moved to 248 nm exposure light, the challenge was to find a material that was appropriately transparent and functioned with low energy doses. Chemical amplification with poly(hydroxystyrene) based systems overcame these obstacles. At 193 nm, the primary challenge lay in optical transparency, since phenolic materials absorbed strongly. Fortunately, it was found that alicyclic materials such as norbornene and adamantane provided both transparency and adequate etch resistance. At the exposure wavelength of 157 nm, the industry faces the same optical density challenge, but with a much more limited set of transparent functional groups from which to work. At 157 nm, nearly everything absorbs light, including polyethylene, which is made up of only C-C and C-H bonds. Preliminary work by Kunz1 and others2 indicates that polymer systems based on siloxanes and fluoro-carbons are the most transparent platforms for resist design, and functional, etch resistant photoresists have now been formulated with 32 absorbances of around 1 m-1 3. However, even with polymer systems that are this transparent, in order to achieve appropriate sidewall angles, resist coatings will have to be fairly thin, usually less than 200 nm. This thickness is not adequate for many subsequent etch processes, so in case polymers with lower absorbance are not found for single layer systems, many people are looking into alternative imaging schemes including inorganic BARC hardmasks 4,5, bi-layer processes6 and top surface imaging7. Further motivating the push towards alternative imaging schemes is the problem of feature collapse in base-developed systems. Studies by researchers at University of Wisconsin and International SEMATECH show that the critical aspect ratio of collapse scales roughly linearly with pitch, and as such, 70 nm nested lines will likely be limited to aspect ratios of less than 2 in wet developed systems8. This chapter focuses on the application of top surface imaging to 157 nm lithography. The TSI systems in this paper have silylation contrast created by a photo-induced switch in the chemical reactivity of the polymer. A polymer that is unreactive to vapor phase silylation is rendered reactive through a chemically amplified mechanism. Fig. 2.1 shows how this process can be implemented in tBOC styrene.9,10 This example serves to display some important features of the process design. In this system the polymer has two states, reactive and unreactive. Since the polymer is always in one of these two states, the silylation is digital. The other feature is subtle but important. Exposure and heat causes deprotection (which changes the polymer s reactivity to silylation) with concurrent loss of volatile products. As a result of this volatilization, the polymer film shrinks in the 33 regions of exposure. The silylation reaction, however, changes the mass and volume of the polymer substantially and causes it to swell. Silylation of poly(hydroxystyrene) with dimethylaminodimethyldisilane (DMADMDS) is shown in Reaction Scheme A. The mass loss due to deprotection and the mass gain due to silylation are balanced to make the net change in the resist film thickness zero.11 This effect is desirable because thickness changes in the film can lead to unwanted image distortion that drastically diminishes the quality of imaging. This system may therefore be described as a digital, zero-volume change, TSI process. Expose PAG n h H+ n Bake OCO2C(CH3)3 n H + + OH n CO2 + Silylate OH R3SiNR'2 OSiR3 + NR'2H Etch silylated unsilylated SiO 2 CO2 + H2 O Figure 2.1: Process Schematic for Negative Tone TSI System 34 n n H + H3C CH3H N Si Si H H CH3 O n + OH H3C + H Si CH3 H Si CH3 H NH(CH3)2 O O O 220 g / monomer Scheme A 120 g / monomer 208 g / monomer EXPERIMENTAL AND MATERIALS Materials tBOC-Styrene was prepared by polymerization of 4-t- butyloxycarbonyloxystyrene monomer donated by Triquest chemical company. A typical polymerization procedure is given below. t-BOC-styrene monomer, (20g, 0.091 mol), was dissolved in dry THF (40mL) in a round bottom flask. The solution was heated to 60oC and then 0.2g of AIBN (0.2g) was added. The resulting solution was stirred overnight, cooled to room temperature, and diluted with an additional 20ml of THF. The THF solution was added to 1L of rapidly stirred methanol whereupon a white precipitate formed. The polymer was isolated by filtration, dried in vacuo at room temperature for 24 hours, redissolved in THF, and then reprecipitated. The resulting white polymer powder was 35 obtained in 85% yield after the two precipitations and typically had a number averaged molecular weight of 25,000 with a Mw / Mn of 1.8. PFAS was prepared by the free radical polymerization of SO2 with tBOCNBHFA. A typical polymerization procedure is given below. Sulfur dioxide (25mL) was condensed in a round bottom flask at 41oC (dry ice / acetonitrile). A solution of tBOC-NBHFA (5g, 13.3 mmol) in 22.5 ml dry THF was added. A solution of t-butylhydroperoxide (23 l, 0.31 mmol) was added in 2 ml dry THF, and the reaction vessel was stirred for 4 hours. In order to quench the reaction, hydroquinone (70mg, 0.64 mmol) was added, and the flask was allowed to warm to room temperature after the polymerization, evolving SO2 gas. The polymer solution was mixed with 75ml of ethyl acetate and washed with 5% NaHCO3 in water until neutral. The organic phase was then washed with water (3x) and with brine (2x), and dried with MgSO4. The ethyl acetate was removed by rotoevaporation, and the polymer was redissolved into a minimum volume of THF. The polymer was precipitated into hexane, filtered, and dried in vacuo at room temperature overnight (3.0g, 51%). Poly (t-BOC-NBHFA) was synthesized by t-BOC protection of poly(NBHFA) with di-tert-butyl dicarbonate. Poly-(NBHFA) was prepared by the Pd2+ catalyzed addition polymerization in dichloromethane at room temperature, as described in Hung et al 4. Poly-(NBHFA) (3.13 g, 11.4 mmol) and di-tertbutyl dicarbonate (5.4 g, 22.8 mmol) were dissolved in 40 ml THF. The solution was stirred at room temperature for 5 minutes. 4-dimethylaminopydridine (0.21 g, 36 1.14 mmol) was added to the solution. The resulting mixture was stirred overnight at room temperature. The resulting polymer was precipitated into methanol (500 ml), filtered, and dried in vacuo at 50 C to give a white powder (3.2 g). Dimethylaminodimethylsilane (DMADMS) was purchased from Silar Laboratories, and 1,2-dimethylaminodimethyldisilane (DMADMDS) was donated to us by Dr. David Wheeler from Sandia National Laboratories. DMADMDS was used in imaging and contrast curve experiments, while DMADMS was used in the FTIR experiments with poly (NBHFA) for proof of silylation O S O n n CF3 O CF3 O O . t-BOC Styrene O CF3 O O CF3 O O O PFAS poly (NBHFA) Figure 2.2: The three polymer systems used in this study. Experimental Apparatus Wafers were coated and baked using an FSI Polaris 2000 track. All 157 nm exposures were done using an Exitech micro-stepper at International 37 SEMATECH (NA = 0.6). For binary mask exposures, = 0.7, for alternating aperture phase-shifted exposures, = 0.3. Silylation was performed using a Genesis 200C at 90oC with DMADMDS for varying times and pressures that are described for each polymer system. Etching was performed on a LAM 9400 SE with the following settings: top power = 260 watts, bottom power 75 watts, pressure = ~3.2 mTorr O2, flowrate 60sccm O2, and bottom chuck T = -25oC. Endpoint detection for imaging was done using Endpoint Plus with 30% overetch. Thickness was measured using a Prometrix interferometer using cauchy coefficients obtained for each polymer system using a Woolham variable angle spectroscoptic ellipsometer. SEM work was done using a Joel tilt SEM and a Hitachi 4500. Absorbance data was taken with a Woolham VUV VASE ellipsometer. IR spectra were taken on-wafer using a Nicolet Magna FTIR 550. For the contrast curve experiments, pad exposures were performed at 248 nm using an Ultratech XLS stepper. 38 RESULTS AND DISCUSSION Imaging Experiments Top Surface Imaging of tBOC Styrene Previous work7 with tBOC Styrene at 193 nm led to the proposal that line edge roughness was largely due to two factors. The first factor is the high opacity of tBOC Styrene at 193 nm. The high opacity results in poor feature edge definition, and a stochastic etch mask degradation process along the edge of the feature during the etch step. The second factor is the low Tg of the silylated form of the polymer of 70oC. The silylation reaction is run at a temperature of 90oC, and as such, during the silylation step, the polymer is mobile, and the image can blur as a result of flow, phase separation or some other mechanism. For this work at 157 nm, the transparency of the resist is significantly better than at 193 nm (as can be seen in Fig. 2.3, although it is still relatively high at ~6 um-1), which is expected to improve imaging performance. This is a necessary, but not sufficient characteristic. Based on this theory, the low Tg of the silylated polymer should still result in roughness. 39 Absorbance of t- BOCPS, PFASO2 and PNFA 16 14 12 (a) Abs. ( m-1 ) 10 8 6 4 2 0 155 (b) 160 (c) 165 170 175 180 185 190 195 200 Wavelength (nm) Figure 2.3: Absorbance (a) t-BOC styrene, (b) PFAS, and (c) poly(NBHFA as a function of wavelength 40 tBOC Styrene was formulated with 4% DPI-Nf (wt PAG / wt Poly) and 5% trioctylamine (moles base / moles PAG) in PGMEA. PAB was 100oC for 60s, PEB 100o for 60s. The resist was silylated at 90oC for 60s at 30 torr DMADMDS. The resist was imaged at 157 nm using a binary mask and the resulting images are shown in Fig. 2.4. 120 nm 1:1 s, 400 nm tall 2.7 mJ / cm2 90 nm 1:1.5 s, 400 nm tall 3.3 mJ / cm2 80 nm 1:2 s, 300 nm tall 3.3 mJ / cm2 Figure. 2.4: Top down and Tilted SEM of tBOC Styrene imaged with a binary mask at 157 nm. 41 This resist prints with rough sidewalls, but the aspect ratio is extremely high, and the resolution is near the limit of the tool with a binary mask. Top Surface Imaging of PFAS Previous work7 with PFAS at 193 nm showed the ability to print highresolution, high-aspect ratio images with extremely low Line Edge Roughness (LER). The extremely low LER of this system was attributed to two advantageous material properties: low opacity and a high Tg of the silylated polymer (~105oC). At 157 nm the opacity of the resist is higher, as can be seen in Fig. 2.3, and as such, the effects of the wavelength change are unclear. Fortunately though, the glass transition temperature is above the temperature of the silylation step. PFAS was formulated with 4% TPS-Nf (wt PAG / wt Poly) and 10% trioctylamine (moles base / moles PAG). PAB and PEB were both run at 90oC for 60s. The resist was silylated at 90oC for 300s at 40 torr DMADMDS. The resist was imaged at 157 nm using a binary mask and the resulting images are shown in Fig. 2.5. 42 120 nm 1:1 s 100 nm 1:2 s 90 nm 1:3 s Figure 2.5: Top down and Tilted SEMs of PFAS imaged at 157 nm with a binary mask. All features are 270 nm tall 43 The imaging quality is excellent and the edges are relatively smooth. The images show some low-frequency LER, but the high-frequency roughness that is characteristic of most TSI systems is not present. The dose required to print these images is between 80- 110 mJ / cm2. The process hasn t been optimized, (in fact these images were taken from the second wafer we processed at high doses), and it is likely that with optimization of the bake temperatures and etch conditions, the low-frequency roughness can be overcome. Overall, these images show that printing smooth lines with TSI at 157 nm is possible. Top Surface Imaging of Poly(NBHFA) Poly (NBHFA) is a material that was generated as a byproduct of our 157 nm single-layer resist program. It has promising material properties in that it has high silicon content after silylation (15.5% wt), is more transparent than PFAS, (see Fig. 2.3), and has a high Tg in its silylated state (nearly all homopolymers of norbornenes have high Tgs). In order to insure that poly (NBHFA) deprotected and silylated properly, we first studied the silylation of the material with IR. Poly (NBHFA) was formulated with 6% TPS-Nf (wt PAG / wt Poly) and 10% trioctylamine (moles base / moles PAG). The resist was coated, baked, flood exposed at 248 nm, and silylated. PAB and PEB were both run for 60s at 130oC. Silylation was at 90oC 44 for 120s with 40torr DMADMS. The infrared spectrum of poly (NBHFA) was collected after the PAB, after the PEB, and after silylation. The resulting spectra are shown in Fig. 2.6. The complete disappearance of the carbonyl peak at ~1770 cm-1 coupled with the appearance of the OH peak at ~3500 cm-1 after exposure indicates complete deprotection, and the appearance of the Si-H peak at ~2100 cm-1 in the silylated spectra accompanied by the complete disappearance of the OH peak at ~3500 cm-1 indicates complete silylation. Analogous spectra of tBOC Styrene and PFAS have been published previously7,12 ). (A) T % 100 90 80 70 60 (B) T % 90 80 70 90 (C) T % 80 70 60 4000 3500 3000 2500 2000 1500 1000 500 W avenumbers (c m-1) Figure 2.6: Infrared Spectra of poly (NBHFA) (a) after coating, (b) after exposure and PEB and (c) after silylation with DMADMS. 45 The resist was imaged at 157 nm using an alternating aperture phase-shifted mask and the resulting images are shown in Fig. 2.7. Process conditions were the same as described for the IR experiment, the only exception being the use of DMADMDS instead of DMADMS. Unfortunately, the images printed with rough sidewalls. Imaging the resist at 193 nm, where the polymer is essentially completely transparent did not help. These results indicate that a high Tg, and transparency may be necessary, but not sufficient, material properties for smooth imaging. This material has some, but not all of the attributes required for 157 nm TSI applications. This issue will be discussed in the ensuing chapters of the dissertation. 46 110 nm 1:1 s 90 nm 1:1.5 s 70 nm 1:2 s Figure 2.7: Top down and Tilted SEMs of poly (NBHFA) imaged at 157 nm with a phase-shifted mask. All features 450 nm tall. 47 Contrast Curves Generation of contrast curves In order to gain a better understanding of the fundamental behavior of these TSI systems, contrast curves were generated through each step of the process by recording film thickness as a function of dose. It was verified that thickness changes correlate nearly exactly with changes in the chemical composition of the film by IR, by either monitoring the growth of the Si-H peak (at ~2100 cm-1 )or by monitoring the loss of the carbonyl peak, (the thickness changes were not the result of density changes). Thickness was measured with an interferometer in each field of a 10 x 10 array after each of the following steps: PAB, PEB, silylation, 10s etch, 50s etch, and 110s etch. Flood exposures were performed at 248 nm where all three polymer systems are transparent. These curves are presented as Figs. 2.8, 2.9 and 2.10. Using an ellipsometer, it was found that the Cauchy coefficients varied little through deprotection and silylation, and as such, Cauchy coefficients for the protected compounds were used. All process conditions (formulation, PAB, PEB, silylation, etc) were the same as those used for imaging, with the exception of the tBOC Styrene, where 2% TPS-Nf was used for contrast curves instead of 4% DPI-Nf. 48 t -BOCPS Contrast Curves 4500.00 4000.00 3500.00 3000.00 2500.00 2000.00 1500.00 1000.00 500.00 0.00 0.00 Thickness (A) Pre exposure Post exposure Post silylation 10s Etch 50s Etch 110s Etch 2.00 4.00 6.00 8.00 2 10.00 12.00 14.00 Dose (mJ/cm ) Figure 2.8: Contrast curves for tBOC Styrene Discussion of tBOC Styrene contrast curve These curves show a number of interesting aspects of this system. The chemical amplification step of this resist generates excellent contrast, as can be seen by the sharp thickness loss after exposure greater than ~1 mJ / cm2. The silylation of this resist also enhances the contrast of the system. Until the resist is ~50% deprotected, the silylation agent does not permeate into the film (as can be seen by the lack of thickness increase upon silylation at doses of less than ~3 mJ / cm2). However, above 50% deprotection, the silylation agent readily permeates and reacts with the film. This is the only system that we have seen that exhibits 49 this silylation contrast behavior. As is usually the case, tBOC Styrene shows great contrast in its final etch step. excellent. The overall contrast in this system is The very high silicon content of the PHOST silylated with DMDSDMA (27 wt %) results in an extremely low etch rate of the silylated resist. The blanket etch rate of the silylated resist after the silicon dioxide etch mask has been formed is ~ 0.2 nm / s. The blanket etch rate of the protected polymer is ~13 nm / s. This excellent etch contrast enables the TSI system to function at wavelengths where the resist is nearly opaque since the light (and hence deprotection and silicon incorporation) does not need to go very deep into the film for an adequate etch barrier to be formed. 50 PFASO2 Contrast Curves 4000.00 3500.00 3000.00 Thickness (A) 2500.00 2000.00 1500.00 1000.00 500.00 0.00 0.00 Pre exposure Post exposure Post silylation 10s Etch 50s Etch 110s Etch 5.00 10.00 15.00 20.00 25.00 Dose (mJ/cm2) Figure 2.9: Contrast curves for PFAS Discussion of contrast curves for PFAS The contrast curves of the PFAS system reveal a number of interesting aspects. First, the thickness curve after exposure and PEB shows that the deprotection step of this polymer is extremely slow. Typically, one would expect to see a much sharper transition, and this result is very puzzling. This curve was replicated by looking at the carbonyl peak in the IR across all the fields and by synthesizing a second batch of polymer that was more rigorously neutralized after 51 polymerization than that described by Ito et al13. This poor deprotection contrast might be related to the relatively low PEB temperature of 90oC, but since the polymer begins to decompose at higher temperatures, it is not appropriate to run it at higher temperatures. The post-silylation curve of this polymer indicates that no silylation contrast is present in the film, since the film fills back up to the initial thickness. Fortunately, the high etch contrast of system results in acceptable overall contrast in the system. The blanket etch rate of the silylated resist after the silicon dioxide etch mask has been formed of 0.4 nm/s is slightly higher in this system than in tBOC Styrene. The blanket etch rate of the protected resist is 17 nm / s. 52 PNFA Contrast Curves 5500 5000 4500 4000 Thickness (A) 3500 3000 2500 2000 1500 1000 500 0 0 2 4 2 Pre-exposure Post-Exposure Post Silylation 10s Etch 110s Etch 6 8 10 Dose (mJ / cm ) Figure 2.10: Contrast curves for PFNA Discussion of contrast curves for poly (NBHFA) The deprotection step in poly (NBHFA) generates sharp etch contrast at around 4 mJ / cm2. The silylation step fills the image in to its original thickness, not enhancing overall system contrast. Finally the etch step enhances contrast, resulting in a sharp overall contrast at around 4 mJ / cm2. The slow deprotection step that we saw in the PFAS has vanished by removing the SO2 from the polymer backbone, and by raising the PEB temperature. The etch rate of this silylated polymer is still relatively high compared to tBOC Styrene, but the 53 increased transparency and better contrast of the resist drops the imaging dose at 157 nm to a more reasonable level of ~15mJ / cm2 for a binary mask. CONCLUSIONS TSI is a process that has the potential to overcome many of the challenges facing the lithographic community at 157 nm. It is capable of printing smooth, high aspect ratio, high resolution images that do not suffer feature collapse problems, while using polymers that are substantially more absorbing than those used in single layer applications. Until recently, TSI has had significant line edge roughness problems, but this might be overcome if the correct material properties are present in the polymeric resin (although these material properties are not yet fully understood). This paper demonstrates that TSI is capable of printing smooth images at 157 nm, and with proper development, could be a viable imaging alternative to single layer resists. 54 6. REFERENCES (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) R. R. Kunz, T. M. Bloomstein, D. E. Hardy, R. B. Goodman, D. K. Downs, J.E. Curin, Proc. SPIE 3678, 13 (1999) C. Brodsky, et al, J. Vac. Sci. Technol. B, 18 (6), 3396-3401 Nov/Dec (2000) R. J. Hung, et al, Proc. SPIE 4345 (2001) A. P. Mahorowala, et al. SPIE 4343 (2001) J. Cobb, W. Conley, F. Huang, T. Lii, S. Usmani, S. Hector, W. Wu, Proc. SPIE 4345, (2001). R. Sooriyakumanaran, D. Fenzel-Alexander, N. Fender, G.M. Wallraff, R.D. Allen, Proc. SPIE 4345, (2001). M. H. Somervell, D. S. Fryer, B. Osborn, K. Patterson, C. G. Willson, J. Vac. Sci. Technol. B 18(5), 2551-2559 Sep/Oct (2000) W. D. Domke, V. L. Graffenberg, S. Patel, G. K. Rich, H. B. Cao, P. F. Nealey, Proc. SPIE 3999, 313 (2000) S. A. MacDonald, H. Schlosser, N. J. Clecak, C. G. Willson, J. M. J. Fr chet, Chem. Mater. 4, 1364 (1992). H. Ito, S. A. MacDonald, R.D. Miller, C. G. Willson, U.S. Patent 4,552,83, (1985). S. V. Postnikov, M. H. Somervell, C. L. Henderson, C. G. Willson, S. Katz, J. Byers, A. Qin, Q. Lin, Proc. SPIE 3333, 997 (1998). S. A. MacDonald, H. Schlosser, H. Ito, N. J. Clecak, C. G. Willson, Chem. Mater. 3, 435-442 (1991) H. Ito, N. Seehof, R. Sato, T. Nakayama, M.Ueda, Micro- and Nanopatterning Polymers, ACS Symposium Series 706, 210-212 55 Chapter 3: Low-voltage Electron Beam Lithography Resist Processes: Top Surface Imaging and Hydrogen Silsesquioxane Bilayer. ABSTRACT: A Hydrogen Silsesquioxane (HSQ) bilayer process and a Top Surface Imaging (TSI) process have been investigated for application as low-voltage electron beam resist systems. Namatsu, van Delft, and others have reported printing exceptionally small features using high voltage electron beam exposure of HSQ at high exposure doses (~2000 C/cm2 at 100kV). The shallow penetration depth of low voltage electrons results in greatly reduced dose requirements, and smooth, high resolution images were generated at 1kV with an exposure dose of less than 60 C/cm2. HSQ s high silicon content enabled it to be used in a bilayer form utilizing reactive ion etching with an oxygen plasma, thus generating high aspect ratio images. TSI has been studied in the past by numerous researchers at low-voltages using various TSI schemes. Here we investigate the use of a chemically amplified TSI resist process based on t-BOC Styrene. The effect of base quencher loading in the resist formulation on line edge roughness and resolution was investigated, and found to have dramatic influence. High resolution (40nm), high aspect ratio images were printed that exhibited only moderate levels of line edge roughness. Further, proximity effects and exposure latitude at 1, 2 and 3 kV were examined and compared to simulation. 56 INTRODUCTION: a. Motivation for low-kV electron beam exposure tools As feature sizes shrink, the demands on mask writing become increasingly difficult to meet. In particular, the popularity of optical proximity correction (OPC) has put tremendous pressure on the mask industry to accelerate their plan for shrinking of minimum feature sizes. Prior to OPC, the resolution of mask features was dictated by the reduction ratio of the lens systems, so the minimum mask feature size was approximately four times that of the chip generation. However, the writing of OPC features demands far higher resolution. Optical mask making tools are reaching their optical resolution limits, and electron beam systems are quickly becoming the tools of choice for the most demanding masks. Traditionally, the resolution of electron beam systems has been improved by increasing the accelerating voltage. Higher energy electron beams have better resolution due to enhanced beam stiffness and the improved energy deposition profile within the resist. There are two aspects of electron behavior in a resist that dictate the energy deposition profile. The first, known as forward scattering, occurs when the electrons are scattered as they travel through the resist. The second, backscattering, is the result of electrons that scatter off the substrate and expose the resist as they exit the material. Increasing the exposure energy reduces the range of forward scattering but also increases the range of the backscattered electrons. Fortunately, a process of dose modulation called proximity correction can largely compensate for the backscattered component, but it is a 57 computationally intensive and time-consuming process. A way of circumventing this process would be very valuable. In addition, the capture cross-section of materials is inversely proportional to the accelerating potential, which results in much lower sensitivity of resists. One way to solve the problems of low resist sensitivity, and proximity correction in electron beam systems is to lower the accelerating voltage dramatically (to 1-3keV) to the point where the forward scattering and backscattering ranges are smaller than the feature size 1-5 . The low penetration depth of electrons within a solid at these voltages makes proximity corrections unnecessary, 6, 7 and results in very high sensitivity. In addition, the low voltages within the column make it particularly amenable for use in micro-column form, allowing the design of multiple beam writing systems8-10. Charge induced pattern distortion is a potential issue associated with low-voltage electron beam lithography11-13 , although many ways exist to circumvent this issue by imaging near the cross-over energy, or by altering the cross-over energy of a resist film through the use of additives. A major difficulty associated with low voltage imaging is the development of a suitable resist process. The low penetration depth of the electrons necessitates the use of a thin film imaging technique to generate features with sufficient etch resistance for subsequent etch transfer steps. In addition, the resist must have sufficiently high resolution to meet the demands of future imaging processes, while maintaining extremely low levels of line edge roughness. Many researchers have investigated this topic14-19. Top Surface Imaging and a bi-layer 58 resist scheme employing a thin, (~20nm thick) hydrogen silsesquioxane imaging layer atop a novolac transfer layer are promising candidates for this application. b. Hydrogen Silsesquioxane Hydrogen Silsesquioxane (HSQ) is a material that is typically used for spin-on-glass applications. It has been commercialized by Dow-Corning under the trade name FOx (standing for Flowable Oxide). FOx solutions are available with varying concentrations of HSQ dissolved in methylisobutylketone (MIBK). The chemical structure of HSQ shown in figure 3.1. As can be seen, it is basically a cross-linked silicon dioxide cage structure, with reactive Si-H bonds on the corners. The material cures at high temperature, releasing H2, and forming a film of porous silicon dioxide20, 21. The Si-H bonds are very reactive, and as such, the resist is sensitive to contamination. The material should not be stored in glass- it should only be stored in poly-ethylene or fluorocarbon bottles at low temperatures. 59 H O Si Figure 3.1: Chemical structure of HSQ. The substantial chemical structure change that occurs during curing of HSQ led researchers to investigate its utility as a resist. Its very high silicon content made it even more appealing because of its potential as a top layer in a bilayer process. In 1998, Namatsu published the surprising result22 that the unexposed resist dissolves in traditional base developer, and demonstrated the functionality of HSQ as a resist. Namatsu, Maile, van Delft, and others have printed very high-resolution features in HSQ with high-voltage e-beam exposure (50-100kV)23-25. Unfortunately, the resist requires unusually high dose in order to function at these voltages (up to >1000 C/ cm2 at 50 kV25). This is likely due to the large number of cross-linking events required to render the resist insoluble coupled with the lack of chemical amplification. The dose can be decreased 60 through the use of a photo-base generator, as was demonstrated by Harkness et al26, although this study used organic developers, such as MIBK, n-octane or toluene. c. Top Surface Imaging Top surface imaging (TSI) is a promising technique for use as a low kV resist process. The fundamental idea behind TSI is that a single resist film is coated on a substrate, and exposed to light or electrons in such a way that only the top of the resist is chemically altered (either by working at a wavelength at which the resist absorbs, or at an electron energy where the penetration depth is shallow). The resist is then put in the presence of a silylation agent, which causes silicon to be incorporated into either the exposed or unexposed region. The film is then etched in an oxygen plasma, where the incorporated silicon acts as a hard mask, protecting that region. The traditional problem associated with TSI is line edge roughness (LER). LER in TSI systems is quite a complicated and involved subject. Recent research has shown that LER in TSI systems can be overcome 27. These experiments strive to determine if TSI has the appropriate resolution capabilities for electron beam applications, and to investigate the factors that influence that resolution. As such, we focus on a traditional TSI system that is easier to work with than those that have minimal LER for the sake or efficiency. Thus, the LER of these images isn t indicative of the absolute limits of TSI. Numerous researchers have investigated the application of TSI to low-voltage electron beam lithography 14, 17, 19. 61 EXPERIMENTAL AND MATERIALS: a. Exposure Conditions and Metrology Imaging was performed on a Raith 150 electron beam exposure tool, which is capable of high resolution at voltages between 500 V and 30kV. The imaging was performed in a vector mode (the beam is electronically positioned at each location and then un-blanked for a single exposure for a designated duration and current). Typical imaging conditions were a step-size of 5 or 10 nm, and a current of 7-10pA. During exposure the aperture was 7.5 um and minimum blanker time was 100ns. Beam focus, aperture alignment and astigmatism were performed on 40nm and 100nm gold nanoparticles that were placed on the wafer. Beam spot sizes at 1, 2 and 3kV were measured to be below 10nm. 248 nm imaging was performed at International SEMATECH on a SVGL Micrascan III using a binary mask (NA = 0.6, = 0.6). Top-down metrology was performed on the Raith 150, or on a Joel Tilt SEM. Critical dimensions were measured using a signal-averaged linescan coupled with a threshold protocol (typically, 8 lines were scanned and averaged for each data point). Cross-section SEMs were made using a Hitachi 4500 and a Hitachi S-4700. b. Processing Conditions and Materials A solution of FOx-14 was purchased from Dow-Corning, and diluted with semiconductor grade MIBK in order to facilitate spin coating at a thickness on the 62 order of 20- 30nm (~13 wt.% FOx-14). The thickness was measured with a Tencor Alpha-step profilometer that was calibrated at the beginning of the study. The diluted FOx was contained in fluoro-polymer or poly-ethylene bottles to avoid degradation, or else diluted in glass immediately prior to coating. The post apply bake was 5 min. at 120 C. Development was performed in two steps: 1min. in Shipley LDD-26W, and 10sec. in 1:9 LDD-26W:H20, followed by rinse and dry as described by Van Delft in Ref. 25 . The bottom layer used in bilayer imaging was the bottom resist used for Infineon s 248nm CARL process. It was coated at varying thicknesses (between 400 and 100nm), and baked for 1min at 140 C, followed by 5min at 250 C. t-BOC styrene was synthesized as described in chapter 2. The post apply bake was 1 min. at 100 C. The post exposure bake was 1min. at 90 C. Silylation was performed on Genesis Microstar 250 using 30 torr of Dimethylaminodimethylsilane (purchased from Silar Laboratories) at 90 C for 1 min. The etch process was performed on a PlasmaTherm 760, with a chuck temperature of -40 C, operated at a top and bottom power of 100W and 800W, respectively. The pressure was 5 mTorr, with a flowrate of 40/10 sccm O2/He respectively. The etch duration was calculated based on measured etch rates plus 30% over etch. Etch work for the optical exposures was performed at International SEMATECH using a Lam 9400 SE operated with a top and bottom power of 260W and 75W respectively. The pressure was 2.5mTorr with an oxygen flowrate of 60sccm and a chuck temperature of -25 C. 63 RESULTS AND DISCUSSION: a. HSQ Single Layer Imaging Initially, the HSQ resist was imaged as a single layer in order to determine imaging performance without the added complication of the etch transfer step. The resist was coated at a thickness of 33nm, and exposed between 40 and 90 C / cm2 at 1kV. The imaging results are shown in figure 3.2a. The resist suffered from significant levels of adhesion failure and delamination. This is in accordance with the simulated penetration depth of the electrons within the resist. The electrons do not penetrate the full thickness of the film. They only insolubilize the top of the resist, and the developer is able to dissolve out the underlying underexposed resist. The resist was then coated at a thickness of 24nm, and exposed at 1kV. A resulting image is shown in figure 3.2b. The resist performed well, with very low line edge roughness, and fairly high resolution, down to 30nm for nested features. (The low contrast of the SEM results from the fact that it is of ~20nm features of basically SiO2 on a wafer surface of native oxide, and therefore we have very little topographical or elemental contrast.) 64 A B Figure 3.2: HSQ images generated at 1kV. a) 33nm thick resist exposed at 44 C / cm2; note the inadequate penetration of the 1kV electrons into the resist, resulting in delamination. b) 24nm thick resist exposed at 40 C / cm2. Image shows 50, 40 and 30nm lines on a 1:1 line:space pitch. 65 b. HSQ Bilayer Imaging Once a satisfactory process was developed for the HSQ in single layer experiments, the work on the bilayer stack was undertaken. The etch rates of the top and bottom layers were measured, and found to be <2 /s and ~180 /s, respectively. A layer of 20nm of HSQ was coated on 180nm of image transfer layer, and exposed at 1, 2, and 3kV. The stack was developed, and etched for 13s. The resulting images are shown in figure 3.3. The resist printed exceptionally smoothly, and produced very high aspect ratio features. Figure 3.3b shows feature collapse that occurs at this tight pitch and high aspect ratio. Resist feature collapse in wet-developed systems is attributed to capillary forces. In this case, since the high aspect ratio features dry-developed in an etcher, the cause isn t immediately obvious. It is likely the result of condensation on the wafer that was still cold when removed from the etcher, or possibly intermolecular forces. The undercutting is the result of an un-optimized etch process, which could be fixed. 66 A B Figure 3.3: 60nm and 50nm nested features. 67 c. HSQ Proximity Effects In order to investigate proximity effects in low-voltage electron beam systems, isolated and nested 70nm features were patterned under the same bilayer conditions that were described in Section 3b. The line-width was measured as a function of dose, and the results are presented in Fig. 3.4. Data were compared with simulated energy distribution profiles, which are shown in figures 3.5 a,b, and c. The thickness of the imaging layer is substantially thinner than the electron penetration depth at 2 and 3 kV. It is therefore likely that this condition results in a lower level of proximity effect, and a higher dose-to-size compared to a thicker film. Feature Size vs. Dose 100 90 80 Feature Size (nm) 70 60 50 40 30 20 10 0 20 40 60 Dose ( C/cm ) 2 3kV 1:2 3kV 1:1 2kV 1:2 2kV 1:1 1kV 1:2 1kV 1:1 80 100 Figure 3.4: Line-width vs. Dose for nominal 70nm isolated and nested lines exposed at 1,2 and 3 kV. 68 70nm 70nm lines exposed at 1kV 1kV lines exposed at 0 0 70nm70nm linesexposed2kV 2kV lines exposed at at 0 0 10 10 10 10 y-position (nm) y-position (nm) Depth into resist (nm) 20 20 30 30 40 40 1000 2000 3000 4000 5000 6000 7000 Depth into resist (nm) 20 20 30 30 40 40 1000 2000 3000 4000 0 0 200 200 X-position (nm) x-position (nm) A 400 400 600 600 800 800 1000 1000 0 0 200 200 X- position (nm) x-position (nm) B 400 400 600 600 800 800 1000 1000 70 nm lines exposed at 3kV 70nm lines exposed at 3kV 0 0 10 10 y-position (nm) Depth into resist (nm) 20 20 30 30 40 40 C 0 0 500 1000 1500 2000 2500 3000 800 800 1000 1000 0 200 200 X- position (nm) x-position (nm) 400 400 600 600 Figure 3.5: Energy deposition profiles of 1,2 and 3 kV electrons in HSQ. Nested lines are on the left; isolated are on the right (Units on energy contours are J / cm3 for a dose of 30 C / cm2). 69 Resist exposure simulations were done using the Monte Carlo method of K. Kim28 which has been shown to give results that agree with experiments at low electron energy, (<3keV). The method calculates the point spread function for the structure and materials used. Once the point-spread function has been calculated it is convolved with the spot shape to determine the instantaneous dose of the electron beam. Summing the spots that would be exposed then creates the pattern. In this case, a two-dimensional profile of lines and spaces was created. The input that produced figure 3.5 assumed the spot to have a Gaussian profile with a FWHM of 10nm and the spot was stepped with a 5 nm grid spacing to create the 70nm lines. Figure 3.5 shows the energy deposition contours for 70nm lines and spaces at 1, 2, 3 keV. The low penetration depth and short scattering length of low voltage electrons has been shown to minimize proximity effects. The simulations in figure 3.5a, b, and c predict proximity effects to appear at 70nm feature width starting somewhere around 2kV, as evidenced the onset of wider energy deposition contours of the electrons within the nested features. If a simple threshold model for development is assumed (meaning that below a certain critical deposition energy, the resist will dissolve completely, and above that energy, the resist will not dissolve at all), it is possible to predict the iso-nested bias from the critical deposition energy at a given depth within the resist (preferably the widest point in the energy deposition profile). The predicted isonested bias is compared with the experimental data in Figure 3.3. Table 3.1 below shows these results. 70 Table 3.1: Proximity effect induced iso-nested bias: simulated vs. experimental. 1kV 2kV 3kV Predicted Iso-Nested Bias (threshold Experimental Iso-Nested Bias model) (nm) (nm) 0 0 3 7 8 11 The model predicts the experimental trends, but underestimates the magnitude of the phenomenon. For 70nm features, proximity effects are not present at 1kV, begin to appear at 2kV and increase at 3kV. It is likely that the discrepancy in magnitude results from the threshold development rate assumption. d. Exposure Latitude A comparison can be made between predicted and experimental exposure latitude in many ways. At one end of the spectrum, a threshold model can be evoked, as was done in the previous section. At the opposite end of the spectrum is the complete model of the entire lithographic process. A rigorous model is not available, but by using the log-slope method , one can compare theoretical and experimental exposure latitude, and gain some insight into resist performance. The log-slope method uses the normalized image-log-slope, or NLS (defined in Eq. 3.1 below) evaluated at the nominal feature edge as the metric for the imaging conditions. 71 NLS = w d ln I dx (3.1) In the equation above, w is the nominal line-width, I is the intensity, and x is the horizontal position. In general, the exposure latitude (EL) for a given imaging system can be related to the NLS by the relationship in Eqn. 3.2. EL = ( NLS ) (3.2) In the above equation, and are adjustable parameters that vary from resist to resist and are process dependent. The parameter is the minimum NLS for which a resist is able to resolve a pattern, and is the percent change in exposure latitude per unit change in NLS. Typically, the EL is defined as the exposure latitude for which the developed line-width is within 10% of the target line-width. For an ideal resist, is 10, and =0, and the performance of a resist can be measured by how close it is to these values. A typical resist has values of = 8.1, and =1.1. The reader is referred to ref the image-log-slope method. In optical systems, where this technique has been traditionally applied, the NLS is calculated from the aerial image at the point that is at the air-resist interface and nominal feature edge. At the air resist interface, the intensity of the aerial image is at its highest (optical absorption in the film has not occurred yet), and the NLS is at its highest. In order to apply the relationship to the low-voltage 29 for a thorough discussion of 72 electron beam situation, the natural place to evaluate the NLS is still at the nominal feature edge, but at the depth where the energy deposition profile is at its widest. The imaging behavior of the system is dictated by what is occurring at this point, since the energy deposition is drastically higher than at the air-resist interface. The intensity at a fixed depth was extracted from the simulations, and the NLS was calculated. For 1kV, the depth used was 14nm; for 2 and 3 kV, the bottom of the resist film, or 20nm, was used. A typical fixed-depth energy deposition profile for nested 70nm lines at 1,2, and 3kV is shown in figure 3.6. Figure 3.6: Energy deposition profiles at 1, 2, and 3kV (dotted lines show feature edge). The experimental EL data is shown in Table 3.2, along with the predicted EL using a threshold model and the NLS model. Overall, the threshold model 73 performed poorly, but the NLS model using typical resist parameters for and did a fairly good job. Table 3.2: Experimental exposure latitude for various imaging conditions vs. theoretically predicted values 1kV (nested 30nm) 1kV (isolated 70nm) 1kV (nested 70nm) 2kV (isolated 70nm) 2kV (nested 70nm) 3kV (nested 70nm) NLS Exp. EL (%) 2.4 0 6.3 44 6.3 44 4.4 22 4.0 13 4.4 22 Typical Resist EL (%) Threshold Model EL (%) =8.1, =1.1 50 11 65 42 65 42 43 27 39 23 42 27 e. Effect of Base Quenchers on Top Surface Imaging Resolution and Line Edge Roughness In formulating a traditional chemically amplified resist, one typically adds four components: polymer, solvent, photo-acid generator, and a quenching base. The polymer contains polar phenolic or carboxylic acid groups that are protected with acid labile protecting groups. The deprotection of these groups enables the solubility switch (and in this TSI system, generates a reactive OH site). Within the exposed areas, the photo-acid generator produces a strong acid which catalytically deprotects numerous sites in the vicinity of where it was formed. The acid is thought to diffuse until the acid is either quenched by a basic 74 site, or until it is trapped in a low diffusion area of the polymer. From a simplistic perspective, the quenching base decreases the number of deprotection incidents, thus pushing the required imaging dose to higher levels. Practically, many bases perform this action in a surprising way. Rather than changing the slope of a deprotection vs. dose curve, the base causes the initiation of the deprotection reaction to move to a higher dose. However, the slope itself doesn t change dramatically. The fundamental reason for this is likely the fast rate of the acidbase reaction relative to the deprotection reaction. The effect of this onset shift is extremely beneficial to resist performance since it serves to increase the contrast of the resist. The critical parameter for contrast in resists is the relative change in deprotection per relative change in dose, not relative change of deprotection per absolute change in dose. In mathematical terms, we care about I/ X (dX/dI), (equivalent to d lnX / d lnI), and not d lnX / dI, where X is the fraction of sites deprotected, and I is the dose. This increase in contrast greatly improves imaging performance of resists; it increases the resolution capabilities of a resist by allowing it to work with a lower image-log-slope (although these bases also may affect other parameters such as iso-nested bias). The influence of base quenchers on photoresists is further discussed in chapter 7. In order to investigate the affects of additive base on the imaging performance of this TSI system, three formulations were tested at 248nm. Using a 248nm stepper at International SEMATECH, we were able to investigate LER and resolution more accurately than in the ebeam system due to the higher reproducibility of the 248nm aerial image as compared to that of the ebeam tool. 75 Three formulations of t-BOC Styrene were generated containing: 2wt% TPS-Nf / 10mol% TOA (mol% measured as base: PAG), 6wt% TPS-Nf / 30mol% TOA, and 6wt% TSP- Nf / 50mol% TOA. So, the ratio of overall base loadings was 1:9:15. The PAG loading was increased as well in order to keep the dose reasonable. These resists were then exposed at 248 nm to determine the effect on LER. The resulting 200nm images are shown in Figure 3.7. As can readily be seen, line edge roughness decreased as base loading increased. In fact, the high base loading resist was capable of resolving down to 160nm, far lower than the other two resists. Unfortunately, the high base resist required a dose of 115 mJ / cm2, far higher than would be practical for application. However, as we shall see, at 1kV the shallow deposition profile decreases the required dose by so much that we can still work at a reasonable level compared to tradition electron lithography tools. 76 A B C Figure 3.7: Effect of variable TOA and TPS-Nf concentration on LER. Nested 200nm lines exposed with 248nm illumination using a) 2wt% TPSNf / 10mol% TOA, b) 6wt% TPS-Nf / 30mol% TOA, c) 6wt% TSP- Nf / 50mol% TOA. 77 f. Low voltage electron beam exposure using TSI. The high base formulation discussed in the previous section was exposed at 1kV. The resulting images are show in Figure 3.8. Two different thicknesses of resist were imaged in order to limit feature collapse at the smaller features. As can readily be seen, very small features were patterned, with only moderate levels of line edge roughness. Dose to size for all of these features was less than 6 C / cm2. 78 A B C Figure 3.8: High resolution imaging using TSI at 1kV. a) 60 nm nested features in ~120 nm of resist, b) 50 nm 1:1.5 features in ~120nm of resist c) 40 nm 1:1.5 features in ~50nm of resist. 79 CONCLUSIONS Low voltage electron beam imaging was investigated using an HSQbilayer resist platform and a TSI resist platform. High resolution, high aspect ratio patterns were generated that had extremely low line edge roughness using HSQ. At 70nm, 1kV imaging showed no evidence of proximity effects, where 2 and 3kV did show moderate levels. The results of Monte Carlo simulations were using the method of Kim were compared qualitatively with experimental results. Top surface imaging generated high resolution, high aspect ratio images with only moderate levels of LER, using a resist formulation containing high levels of quencher base. 80 References: (1) Pease, R. F. W. The IEEE 9th Annual Symposium on Electron, Ion, and Laser Beam Technology, Edited by RFW Pease (San Francisco, CA) 1967, 176. Yau, Y. W.; Pease, R. F. W.; Iranmanesh, A. A.; Polasko, K. J. Journal of Vacuum Science & Technology 1981, 19, 1048-1052. McCord, M. A.; Newman, T. H. Journal of Vacuum Science & Technology B 1992, 10, 3083-3087. Peterson, P. A.; Radzimski, Z. J.; Schwalm, S. A.; Russell, P. E. J. Vac. Sci. Technol., B 1992, 10, 3088-3093. Lee, Y. H.; Browning, R.; Maluf, N.; Owen, G.; Pease, R. F. W. J. Vac. Sci. Technol., B 1992, 10, 3094-3098. Polasko, K. J.; Yau, Y. W.; Pease, R. F. W. Proc. SPIE-Int. Soc. Opt. Eng. 1982, 333, 76-82. Stark, T. J.; Edenfeld, K. M.; Griffis, D. P.; Radzimski, Z. J.; Russell, P. E. J. Vac. Sci. Technol., B 1993, 11, 2367-2372. Chang, T. H. P.; Kern, D. P.; Muray, L. P. J. Vac. Sci. Technol., B 1992, 10, 2743-2748. Kratschmer, E.; Kim, E. S.; Thomson, M. G. R.; Lee, K. Y.; Rishton, S. A.; Yu, M. L.; Chang, T. H. P. J. Vac. Sci. Technol., B 1995, 13, 24982503. Kratschmer, E.; Kim, H. S.; Thomson, M. G. R.; Lee, K. Y.; Rishton, S. A.; Yu, M. L.; Chang, T. H. P. Journal of Vacuum Science & Technology B 1994, 12, 3503-3507. Kudryashov, V. A.; Krasnov, V. V.; Prewett, P. D.; Hall, T. J. Microelectron. Eng. 1997, 35, 165-168. Liu, W.; Ingino, J.; Pease, R. F. J. Vac. Sci. Technol., B 1995, 13, 19791983. Satyalakshmi, K. M.; Olkhovets, A.; Metzler, M. G.; Harnett, C. K.; Tanenbaum, D. M.; Craighead, H. G. Journal of Vacuum Science & Technology, B: Microelectronics and Nanometer Structures 2000, 18, 3122-3125. Boettcher, M.; Bauch, L.; Stolberg, I. J. Vac. Sci. Technol., B 1994, 12, 3473-3477. 81 (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) Lo, C. W.; Rooks, M. J.; Lo, W. K.; Isaacson, M.; Craighead, H. G. J. Vac. Sci. Technol., B 1995, 13, 812-820. Tanenbaum, D. M.; Lo, C. W.; Isaacson, M.; Craighead, H. G.; Rooks, M. J.; Lee, K. Y.; Huang, W. S.; Chang, T. H. P. J. Vac. Sci. Technol., B 1996, 14, 3829-3833. Whelan, C. S.; Tanenbaum, D. M.; La Tulipe, D. C.; Isaacson, M.; Craighead, H. G. J. Vac. Sci. Technol., B 1997, 15, 2555-2560. Schock, K. D.; Prins, F. E.; Strahle, S.; Kern, D. P. J. Vac. Sci. Technol., B 1997, 15, 2323-2326. Lee, K. Y.; Hsu, Y.; Le, P.; Tan, Z. C. H.; Chang, T. H. P.; Elian, K. Journal of Vacuum Science & Technology, B: Microelectronics and Nanometer Structures 2000, 18, 3408-3413. Siew, Y. K.; Sarkar, G.; Hu, X.; Hui, J.; See, A.; Chua, C. T. J. Electrochem. Soc. 2000, 147, 335-339. Belot, V.; Corriu, R.; Leclercq, D.; Mutin, P. H.; Vioux, A. Chem. Mater. 1991, 3, 127-131. Namatsu, H.; Takahashi, Y.; Yamazaki, K.; Yamaguchi, T.; Nagase, M.; Kurihara, K. J. Vac. Sci. Technol., B 1998, 16, 69-76. Namatsu, H.; Yamaguchi, T.; Nagase, M.; Yamazaki, K.; Kurihara, K. Microelectron. Eng. 1998, 41/42, 331-334. Maile, B. E.; Henschel, W.; Kurz, H.; Rienks, B.; Polman, R.; Kaars, P. Japanese Journal of Applied Physics, Part 1: Regular Papers, Short Notes & Review Papers 2000, 39, 6836-6842. van Delft, F. C. M. J. M.; Weterings, J. P.; van Langen-Suurling, A. K.; Romijn, H. Journal of Vacuum Science & Technology, B: Microelectronics and Nanometer Structures 2000, 18, 3419-3423. Harkness, B. R.; Takeuchi, K.; Tachikawa, M. Macromolecules 1998, 31, 4798-4805. Somervell, M. H.; Byers, J.; Willson, C. G. Abstracts of Papers of the American Chemical Society 1999, 218, 29-PMSE. Lee, Y.; Lee, W.; Chun, K.; Kim, H. Journal of Vacuum Science & Technology B 1999, 17, 2903-2906. Mack, C. A. Opt. Eng. (Bellingham, Wash.) 1993, 32, 3350-3362. (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28) (29) 82 Chapter 4: Surface Energy Induced Low Frequency Line Edge Roughness in Top Surface Imaging ABSTRACT: Top Surface Imaging (TSI) is a lithographic technique that possesses many advantages over traditional processes. However, TSI suffers from the In this chapter, we poorly understood phenomenon of line edge roughness. propose that low frequency roughness in TSI systems is the result of the surface energy minimization of a low Tg polymer in a thermodynamically unstable geometry. We model the lithographic system as a truncated cylinder on a surface, and present a capillary instability analysis of two degradation pathways. Modeling demonstrates that in nearly all circumstances the geometries created in TSI are thermodynamically unfavorable. INTRODUCTION Surface energy induced instabilities occur in a variety of systems. The classic example of this phenomenon is the break-up of a liquid jet into small droplets; however, any system with the ability to flow is capable of minimizing its free surface energy and is therefore susceptible to this so-called Rayleigh or capillary instability. Interfacial instabilities have also been used to explain instabilities in numerous systems including, jets of liquid crystalline polymers1, liquid metal in welding applications2 and MEMs based atomizers3 to name a few. One area where these phenomena are beginning to appear more regularly is nanostructures. Fundamentally, small-scale features (often created by 83 lithography) have very large surface area to volume ratios. If the material can flow, the features have a tremendous driving force to rearrange themselves to minimize their surface free energy. This driving force dramatically increases as feature sizes decrease since the surface tension based Laplace pressure scales inversely with the radius of curvature of the feature. A good example of this phenomena was presented by Tanaka et al, who investigated degradations associated with the annealing of nanofabricated silicon structures, and attributed them to a Rayleigh instability4. In the semiconductor industry people strive to create very small features with sharp, uniform edges. Roughness in this edge is typically referred to as line edge roughness (LER). Top surface imaging (TSI) is an alternative lithographic process that has not been accepted into semiconductor manufacturing for various reasons, but one of the dominant problems associated with TSI is a large amount of LER. There have been many proposed explanations for LER in TSI systems5, 6 , but the problem has yet to be fully understood. Traditionally, line edge High roughness has been characterized as either low or high frequency. frequency LER is a very complicated phenomenon that has been studied extensively in both traditional wet-developed systems as well as in TSI systems7-9. Low frequency roughness is less well understood, but it is still a significant issue. Figure 4.1 shows a 120nm line and space pattern with a significant amount of low frequency LER. This paper seeks to explain this low frequency LER in terms of a capillary instability of the features that have been created by TSI. 84 Figure 4.1. 120 nanometer lines and spaces generated with top surface imaging, showing significant low frequency LER 6. A TSI process flow is shown schematically in figure 4.2. A single layer resist film is coated on a substrate and exposed to light or electrons, which chemically alters the film. If the resist is highly absorbing to the light, only the top portion of the resist is affected. Silicon is then selectively incorporated into the top portion of the film, often by exposure to a gas phase silylation agent. This process leads to a resist film containing silicon in the exposed regions and none in the unexposed regions (or vice versa, depending on the chemistry). The material is then etched in an oxygen plasma, where the incorporated silicon acts as a hard mask, providing etch resistance to the exposed regions. The fundamental advantage of TSI is that it works even when the photoresist is very absorbing to the wavelength of light to which it is exposed, making it easier to print using short wavelengths of light (thus enabling printing of smaller features, as dictated by the Rayleigh criterion). 85 Expose with light Silylate O2 Etch Figure 4.2. Process flow for a typical top surface imaging process. A resist film is coated, selectively exposed to light, silylated and Oxygen etched, forming high-resolution features. As illustrated in Figure 4.2, the resist film often swells during the silylation step, creating a truncated cylinder of silylated polymer. Depending on the chemistry of the TSI system in use, this level of swelling can vary greatly. Figure 4.3 shows a cross sectional scanning electron micrograph of a group of lines that exhibit the swelling phenomenon prior to etch (the white caps are the silylated polymer). 86 Figure 4.3. Scanning electron micrograph of a bank of lines prior to etch demonstrating the formation of truncated cylindrical cross-sections. In order for a capillary instability to occur, the material must be sufficiently mobile to rearrange its surface. In the case of TSI, during the silylation step, the glass transition temperature (Tg) of the polymer greatly decreases, thereby lowering the polymer s viscosity5. Furthermore, an even lower Tg can be achieved during the etch process because oxidation of the silylated polymer can result in low Tg siloxane intermediates10. As the glass transition temperature of the silylated regions drops below the processing temperatures, the low Tg features gain the ability to flow; and the features thereby become susceptible to capillary instabilities. This chapter seeks to explain low frequency line edge roughness in top surface imaging systems as a capillary instability of low Tg, silicon containing regions in the resist film. The system is modeled as a truncated, liquid cylinder bounded by a solid resist film below. Its stability to axial disturbances is tested by 87 applying a low amplitude sinusoidal disturbance to the cylinder and examining the change in energy of the system. Previous researchers have investigated similar geometries. Grinfeld11 investigated a fixed and a variable footprint case with disturbances that allowed the radius of curvature to vary at a given cross-section, which does not closely match the observed phenomena in our small systems. Gratzke2 investigated a fixed radius of curvature geometry that is similar to our system, but placed a disturbance the radius of curvature while allowing the width and the contact angle to float. Neither of these disturbances adequately described our observations. In this work, two cases are examined. In the first case, the cylinder is assumed to have a fixed footprint on the resist film below. In other words, the bottom portion of the truncated cylinder is assumed to be in contact with a fixed portion of the underlying substrate. In the second case, the footprint of the cylinder is allowed to vary but the cylinder maintains a fixed contact angle with respect to the substrate. Figure 4.4 shows an example of each type of perturbation considered. 88 a. b. (Z) Figure 4.4. Two perturbations on a truncated cylinder: (a) a perturbation on the width of the cylinder footprint, with a constant contact angle, and (b) a perturbation on the contact angle of the cylinder, , on a fixed footprint. EXPERIMENTAL SECTION A traditional binary chemically amplified top surface imaging scheme based on the work of Somervell5 was employed to generate the truncated cylinder topographies discussed earlier. TBOC styrene was synthesized and formulated in a propyleneglycol methylether acetate solution containing 2% (by weight) of the photo acid generator triphenylsulfonium nonaflate. The solution was spin cast on an anti-reflection coated silicon substrate using an FSI Polaris 2000 and post application baked for 1 min. at 100 C. The wafer was exposed at 248 nm on a SVGL Micrascan III using binary illumination (NA = 0.6, = 0.6). It was then post exposure baked for 1 minute at 90 C. Silylation was performed on a Genesis 89 Microstar 250 using 30 torr of dimethylaminodimethylsilane (purchased from Silar Laboratories) at 90 C for 1 min (or for 10 min in the long-time case). Scanning Electron Micrographs were produced on a Joel Tilt SEM, and a Hitachi S 4500. All processing and exposures were performed at International SEMATECH on 8 inch silicon wafers. RESULTS Two line space patterns were prepared by TSI. One was swollen by a small amount, whereas the other was swollen greatly. Figure 4.5 shows the results of these experiments. Figure 4.5a shows the truncated cylinder geometry at low levels of swelling. The lines appear smooth and uniform. Figure 4.5b shows lines that were allowed to swell. These lines are much less smooth and appear to have undergone an instability-based degradation. The results of this experiment tend to indicate that higher degrees of swelling lead to more susceptibility to capillary instability. 90 (a). (b). Figure 4.5. (a) A tilted cross-sectional micrograph showing smooth truncated cylinders created with TSI and (b) a tilted cross-sectional micrograph showing truncated cylinders at a larger level of swelling that appear to have undergone a capillary instability. In order to demonstrate that this phenomenon was induced by capillary instabilities, an attempt was made to generate the results on a bulk scale. Liquid was placed in a trench with fluoropolymer surroundings. In almost all cases, the liquid immediately formed a single bulge at one place along the line. It became apparent that this was the stable equilibrium state. A TSI sample was held at high 91 temperature for a long period of time to determine if it achieved the same geometry. The results are shown in Figures 4.6a and 4.6b. (a) (b) Figure 4.6. a) A SEM of a line space pattern created by TSI. This pattern was held at high temperature for a prolonged period of time, allowing the sample to achieve a more stable configuration. b) A photograph of water in a macroscopic fluoropolymer trench showing a similar stable configuration. 92 THEORETICAL ANALYSIS Capillary instabilities are driven by the minimization of surface free energy in systems that are capable of flow. This phenomenon can be investigated by examining the energy of a system when subjected to a small perturbation. If the energy of the perturbed system is less than the energy of the unperturbed system, the perturbation will grow and therefore, the system is unstable and will tend to evolve to a lower surface energy state. The type of perturbation typically used for this type of analysis is a low amplitude sinusoidal variation in a variable that describes the geometry of the system. Because all perturbations can be broken down into a summation of sinusoids via Fourier transforms, this type of perturbation is generally applicable to any arbitrary disturbance or perturbation. In this specific case, the system of interest is modeled initially as cylindrical cap of liquid on a surface. The liquid is assumed to be infinitely long in the axial, or z, direction and is assumed to have a cylindrical cross-section. All changes in the energy of the system under consideration are changes in surface energy, which are directly proportional to changes in surface area. Thus, by comparing the surface area of the original truncated cylinder to the surface area of the perturbed cylinder, including any interactions between the liquid and the surface it is resting upon, one can determine the stability of the original system. It should be noted that the geometry we are investigating is not, per se, the exact geometry that always occurs in these systems, but it is representative of that which occurs. The most convenient coordinate system for performing these calculations is cylindrical coordinates as shown in Figure 4.7. As can be seen, the origin is 93 located centrally on the surface upon which the liquid rests. The variable w represents the half width of the cylinder and represents the contact angle of the liquid. The variables r, , and z are standard cylindrical coordinates. r ( ) w Figure 4.7. The geometry of the cross section of the line of fluid. The top surface of the truncated cylinder is given by 1 2 w 2 2 r ( ) = (1 cos cos ) sin cos . sin (4.1) Given this coordinate system, one can compute the surface area of this shape by applying: S= Surface r r z d dz , (4.2) 94 where r and rz are tangent vectors given by x y z , , x y z , , z z z r = (4.3) rz = (4.4) and x, y and z equal r cos , r sin , and z, respectively. As stated earlier, a fixed footprint model and a variable footprint model were examined. The fixed footprint model would apply to a TSI system that could swell but for which the swollen material would remain in the silylated region. The variable footprint model would govern a system in which the swollen cylinder was able to move into the surrounding region. It is possible that energetic interactions or degree of swelling could govern which model is applicable. Case A: Fixed Footprint Analysis In this case, the cylinder maintains a constant footprint; hence, the interactions with the substrate are the same for both the perturbed and unperturbed cases and therefore can be neglected. As a result, any changes in energy in this system will be proportional to the change in surface area of the disturbed cylindrical cap. Mathematically, one could impose this disturbance in any number of forms, but in this case, the disturbance was modeled as a sinusoidal disturbance to the contact angle, 95 ( z ) = o 1 + A sin 2 z L (4.5) where A represents the amplitude and L represents the wavelength of the sinusoidal perturbation. For this disturbance, the total volume of a length L of the undisturbed cylinder must be conserved. To satisfy this constraint, o was determined based on conservation of volume between the initial and final state. The volume of the perturbed cylinder can be easily calculated by integrating the cross sectional area of the cylinder along the z axis. Setting this integral equal to the volume of the perturbed integral yields L i ( z) 1 1 = V = w2 2 Lw 2 2 sin ( z ) tan ( z ) dz , sin i tan i 0 (4.6) where i is the undisturbed contact angle. Equation 4.6 was solved numerically to give the correct value for o. Once this parameter was determined, the surface area was computed by combining equations 4.1 through 4.4 and numerically integrating. This process was repeated for a range of L/w ratios and at a variety of contact angles. The amplitude of the disturbance was fixed at one percent. The normalized energy change due to the perturbation is plotted against L/w for various contact angles in Figure 4.8. 96 Energy Changes for Various Contact Angles: Fixed Footprint Model 8 6 4 E/Eo * 10 30 Degrees 50 Degrees 70 Degrees 90 Degrees 110 Degrees 130 Degrees 150 Degrees 1 10 L/w 100 4 2 0 -2 -4 Figure 4.8. Surface area change as a function of normalized disturbance wavelength for several initial contact angles The point on the graph where the change in energy equals zero corresponds to the normalized wavelength at the onset of instability. This wavelength is plotted against contact angle in Figure 4.9. The system is unstable to disturbance wavelengths longer than the critical wavelength. 97 Onset of Instability for Fixed Footprint Case 100 80 60 L/w 40 20 0 90 100 110 120 130 140 150 160 170 180 Contact Angle Figure 4.9. Critical Wavelength for the onset of instability for contact angle disturbances Case B: Variable Footprint Analysis In the case of the variable footprint disturbance, a similar perturbation, 2 z w = wo 1 + A sin , L (4.7) was applied to the cylinder. The parameter wo was determined by conservation of volume as before. In this case, an additional complexity arises from the interactions of the cylinder with the surface. The change in energy for this case can be described as 98 E = S + (2wi L 2wo L ) cos , (4.8) where E is the change in energy of the system, is the surface energy of the liquid, wi is the initial drop half width, wo is the average perturbed drop half width, and S is the change in surface area of the cylinder. This equation is derived from an energy balance around the system and a force ba...

Find millions of documents here - Study Guides, Homework Solutions, Papers, Exam Answer Keys and more. Course Hero has millions of course related materials that will enable you to learn better, faster and get an A in all your courses.
Below is a small sample set of documents: