THE USE OF GRAPHENE AS A SOLID STATE DIFFUSION BARRIER By WAYNE KENNETH MORROW A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DO CTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2014
Â© 2014 Wayne Kenneth Morrow
To Celicia, Danelle, Adalae and Nicholas, the ones who truly sacrificed
4 ACKNOWLEDGMENTS First and foremost I have to acknowledge D r. Steve Pearton who had more overstated in a multitude of actions and he displayed significant patience in a distance student achieving his goals. A special acknowledgement must be paid to Dr. Yves Chabal for allowing me access into the UT Dallas LSM lab, where the majority of this research took place and for many useful discussions. I would also like to thank Dr. Brent Gila and the staff at NRF, specifically Al Ogden and Da vid Hays. These three provided insight, direction and common sense both technically and experimentally. Also a special thanks to Shawn Brown and Jeff Veyan who assisted greatly in the design and construction of the rapid thermal graphene growth system. To all the above people I am eternally grateful.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 MOTIVATION ................................ ................................ ................................ ......... 14 1.1 Graphene: A New Two Dimensional Material ................................ ................... 14 1.2 An Impermeable Single Atomic Layer Membrane ................................ ............. 16 1.3 Statement Of Th esis ................................ ................................ ......................... 17 2 TECHNICAL BACKGROUND ................................ ................................ ................. 19 2.1 Properties Of Graphene ................................ ................................ .................... 19 2.1.1 Carbon Bonding ................................ ................................ ...................... 19 2.1.2 Graphene Crystal Structure ................................ ................................ ..... 19 2.1.3 Graphene Electrical Properties ................................ ................................ 20 2.1.4 Elastic Modulus Of Graphene ................................ ................................ .. 23 2.1.5 Thermal Properties Of Graphene ................................ ............................ 24 s Chemical Resistance ................................ ........................... 25 2.1.7 Defects In Graphene ................................ ................................ ............... 28 220.127.116.11 Stone Wales defect ................................ ................................ ........ 29 18.104.22.168 Vacancies ................................ ................................ ...................... 29 22.214.171.124 Line defects ................................ ................................ .................... 30 2.2 Graphene Growth Techniques ................................ ................................ .......... 30 2.2.1 Survey Of Techniques ................................ ................................ ............. 30 2.2.2 Chemical Vapor Deposition Of Graphene ................................ ............... 31 2.2.3 Growing Gr aphene On Copper Substrates ................................ .............. 33 2.2.4 Modern Methods Of Graphene Growth ................................ ................... 35 2.3 Metrology Of Graphene ................................ ................................ .................... 37 2.3.1 Raman Spectroscopy Of Graphene ................................ ......................... 37 2.3.2 Strain In Graphene ................................ ................................ .................. 40 2.3.3 Doping Effects And Graphene ................................ ................................ . 41 2.3.4 Temperature Effects And Graphene ................................ ........................ 42 2.4 X Ray Photoelectron Spectroscopy ................................ ................................ .. 42 2.5 The Four Point Probe ................................ ................................ ........................ 43 2.5.1 Sheet Resistance Determination By The Four Point Probe Method ........ 43
6 2.5.2 Diffusion Constants By Four Point Probe Method ................................ ... 46 2.6 Diffusion Barriers ................................ ................................ .............................. 49 2.6.1 Carbon Diffusion Barriers ................................ ................................ ........ 50 2.6.2 Self Assembled Monolayers As Carbon Based Diffusion Barriers ........... 51 2.7 Rapid Thermal Processing ................................ ................................ ................ 52 2.7.1 The History Of Rapid Thermal Processing ................................ .............. 53 2.7.2 Infrared Halogen Lamp Based Rapid Thermal Processing ...................... 54 2.7.3 Rapid Thermal Processing Chamber Materials ................................ ....... 56 2.7.4 Susceptor Based Rapid Thermal Processing ................................ .......... 57 3 DESIGN AN D FABRICATION OF A RAPID THERMAL GRAPHENE GROWTH SYSTEM AND SYNTHESIS TECHNIQUES ................................ ........................... 69 3.1 Introduction ................................ ................................ ................................ ....... 69 3.2 Design Consideration s Of A Graphene Growth System ................................ ... 69 3.2.1 A Unique Dual Chamber Design ................................ ............................. 70 3.2.2 Materials Science In Chamber Designs ................................ ................... 74 3.3 Graphene Synthesis In A Rapid Thermal Graphene Growth System ............... 77 3.4 An Attempt Of Graphene Synthesis Without Transfer ................................ ....... 79 3.5 Concluding Remarks ................................ ................................ ......................... 80 4 TESTING OF GRAPHENE AS A SOLID STATE DIFFUSION BARRIER ............... 88 4.1 Introduction ................................ ................................ ................................ ....... 88 4.2 Physisorption And Chemisorption: Bonding With Graphene ............................. 88 4.3 A Comparison Of Commercial Grade G raphene ................................ .............. 89 4.4 Graphene Barrier Testing Of Chemisorbed Metals ................................ ........... 90 4.5 Graphene Barrier Testing Of Physisorbed Metals ................................ ............ 93 4.6 Diffusion Barrier Testing For NiAu Contacts On GaN ................................ ....... 94 4.6.1 Challenges In GaN Processing ................................ ............................... 94 4.6.2 Graphene as a prevention to Ni Au Intermixing ................................ ....... 96 5 CONCLUSIONS AND PROSPECTIVE ................................ ................................ . 111 LIST OF REFERENC ES ................................ ................................ ............................. 113 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 123
7 LIST OF TABLES Table page 2 1 Raman Tendencies of Gr aphene. Recreated from reference 122 ......................... 68 2 2 Batch Furnace/RTP Comparisons ................................ ................................ ...... 68
8 LIST OF FIGURES Figure page 1 1 Graphene crystal structure created in CrystalMaker TM . ................................ ...... 18 2 1 An illustration of the different carbon carbon hybridization structures showing the bond angles of sp1, s p2 and sp3. ................................ ................................ . 59 2 2 The hexagonal lattice of graphene and its Brillouin zone. ................................ .. 59 2 3 The two dimensional band structure of a sing le atomic layer of graphite as modeled by Painter and Ellis ................................ ................................ .............. 60 2 4 An illustration of the linear dispersion of graphene as compared to a single electron dispersion and a quadratic dispersion. ................................ .................. 60 2 5 Illustration of AFM technique to measure deflection of graphene. ...................... 61 2 6 Examples of graphene reconstruction around defects ................................ ........ 61 2 7 A representation of the gas flow dynamics in growing graphene on copper ....... 62 2 8 The evolution of Raman shifts ................................ ................................ ............ 62 2 9 The Nicolet Alamega Ramn Spectrometer used in this research at UT Dallas NSERL lab . ................................ ................................ ................................ ......... 63 2 10 The Ulvac Phi 2300 XPS used in this resea r ch at the UT Dallas NSERL la b ..... 63 2 11 The Alessi Four Point Probe used in the research in the UT Dallas NSERL cleanroom ................................ ................................ ................................ ........... 64 2 12 The use of the four point probe method to determin e diffusion constants at the temperature of barrier breakdown ................................ ................................ 65 2 13 Selective formation o f SAM barriers in a Cu dual damascene integration scheme ................................ ................................ ................................ ............... 65 2 14 tral Distribution ................................ ................................ ............. 66 2 15 An assembly of the quartz halogen lamps used in this research as provided by Ushio Incorporated ................................ ................................ ........................ 66 2 16 A comparison of the Reflectance vs. Wavelength for different metals in the infrared region of the spectrum ................................ ................................ ........... 67 2 17 The optical transmission of 214 grade quartz as compared to quartz with a high percentage of hydroxyl groups ................................ ................................ .... 67
9 3 1 A picture of the clamshell approach to the graphene growth system during fab rication . ................................ ................................ ................................ .......... 81 3 2 The design of the wall of water approach of water cooling in the Rapid Thermal Graphene Growth System ................................ ................................ .... 81 3 3 The completed design and build of the Rapid Thermal Graphene Growth System used in this research with panels removed ................................ ............ 82 3 4 T he final design of the graphene growth system as a mockup designed in Solidworks ................................ ................................ ................................ .......... 83 3 5 The graphene growth chamber and copper conductors. ................................ .... 84 3 6 The graphene growth chamber showing the location of the failed o ring u nderneath the quartz window . ................................ ................................ ........... 85 3 7 Raman Spectroscopy of Graphene on Copper grown in the rapid thermal graphene growth system measured at multiple locations . ................................ .. 85 3 8 The process run of successfully growing graphene on copper ........................... 86 3 9 Growing graphene directly on SiO 2 ................................ ................................ .... 86 3 10 Raman spectroscopy of an attempt at growing graphene on SiO 2 . .................... 87 4 1 Bonding energies of various metals with gr aphene ................................ .......... 10 0 4 2 The band structures of graphene with physisorbe d and chemisorbed metals .. 100 4 3 Microscope analysis of Graphene on SiO 2 provided by the Graphene Supermarket . ................................ ................................ ................................ .... 101 4 4 Raman analysis of five samples of graphene on SiO 2 provided by the Graphene Supermarket ................................ ................................ .................... 101 4 5 Raman analysis of five samples of graphene on SiO 2 provided by ACS Materials ................................ ................................ ................................ ........... 102 4 6 Titanium films on single layer and double layer graphene annealed at 600 o C and 800 o C . ................................ ................................ ................................ ........ 102 4 7 Sheet resistance results at increasing annealing temperatures for 1000Ã… Ti on single layer and double layer graphene . ................................ ...................... 103 4 8 STEM pictures of titanium on graphene and titanium on silicon. Note both pictures are at the same magnification . ................................ ............................ 103
10 4 9 Magnified STEM images of titanium on double layer graphene and titanium on silicon . ................................ ................................ ................................ ......... 104 4 10 Copper delamination on graphene after a 825 o C anneal at 5X an d 50X . ......... 104 4 11 XPS data on a control sample of Cu on SiO2 and Cu/graphene on SiO2 after a copper wet etch . ................................ ................................ ............................ 105 4 12 Sheet resistance measurements of Ni/Au on SiO 2 and Ni/Au on p GaN . ......... 105 4 13 An image of graphene on nickel after transfer . ................................ ................. 106 4 14 An analysis from the substrate backside demonstrating the evolution of Ni Au intermixing at different processing temperatures on p GaN substrates ............ 106 4 15 Microscope 10X image of particles after 600 o C anneal in areas where graphene is not present . ................................ ................................ ................... 107 4 16 An SEM image of the contamination in areas in which graphene is not present . ................................ ................................ ................................ ............ 107 4 17 XPS spectrum of anneals of Ni/Au on p GaN . ................................ .................. 108 4 18 STEM image of the complete intermixing of Ni Au . ................................ .......... 109 4 19 STEM image of N i/Au on graphene after 600 o C anneal . ................................ .. 109 4 20 Magnified STEM image of Ni/Au on p GaN after 600 o C anneal that shows no sign of the Ni layer indicating the Ni Au intermixing process has occurred. ..... 110 4 21 Magnified STEM image of Ni/Au on graphene/ p GaN after 600 o C anneal that shows clearly the Ni layer is present and intact. ................................ ............... 110
11 LIST OF ABBREVIATIONS 2D Two Dimensional 3D Three Dimensional . AFM Atomic Force Microscopy BOE Buffered Oxide Etch CTE Coefficient of Thermal Expansion DFT Density Functional Theory DVM Discrete Variatio nal Method h BN Hexagonal Boron Nitride HOPG Highly Ordered Pyrolytic Graphite NEMS Nonoelectromechanical Systems p GaN P type doped Gallium Nitride RTP Rapid Thermal Processsing SEM Scanning Electron Microscopy STEM Scanning Tunneling Electron Micr oscopy VdW Van der Waals forces XPS X Ray Photoelectron Spectroscopy
12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosop hy THE USE OF GRAPHENE AS A SOLID STATE DIFFUSION BARRIER By Wayne Kenneth Morrow August 2014 Chair: Stephen J. Pearton Major: Materials Science and Engineering Conventional thin film diffusion barriers consist of three dimensional bulk films with hi gh chemical and thermal stabilities. The purpose of the barrier material is to prevent intermixing or penetration from the two materials that encase the barrier material. Adhesion to both top and bottom materials is critical to the success of the barrier. The aim of this thesis is to test the effectiveness of a single atomic layer of graphene as a solid state diffusion barrier for common semiconductor based conductors. Two classes of metals were studied, physisorbed metals, which form a weak bond with graph ene and chemisorbed metals, which are strongly bonded to graphene. E beam deposited metal films on silicon were annealed with and without graphene to determine if graphene will prevent metal silicide layers from forming. Thin layers of metal films on silic a were studied with a graphene barrier to determine if the graphene would prevent diffusion into the underlying substrate at various temperatures. In addition, a graphene diffusion barrier was used to study the inter diffusion of a sandwich structure of ni ckel and gold on Mg doped p GaN wafers. In order to synthesize graphene, a custom vacuum based Rapid Thermal Processing (RTP) system was built. The RTP system was utilized to explore the growth
13 of graphene with various chemistries of methane and hydrogen a t different vacuum levels and temperatures. Analytical techniques such as X ray Photoelectron Spectroscopy, Raman Spectroscopy, Scanning Electron Microscopy and the Four Point Probe technique were used to determine the effectiveness of the graphene diffus ion barrier. For titanium, a chemisorbed metal, the graphene layer is sacrificial and prevents or inhibits silicide formation. In the case of copper a physisorbed metal on graphene, surface diffusion of clusters on graphene is the failure mechanism and no t diffusion through the graphene barrier. For nickel and gold sandwich structures, graphene effectively prevented an inter diffusion reaction from occurring .
14 CHAPTER 1 MOTIVATION 1.1 Graphene: A N ew Two Dimensional Material Historically the word graphene comes from the Greek word gra phein , which means to write. In the 1800s the name graphite was given to the bulk material used in pencils by the German chemist Wagner. The most important historical application of graphite was in the molds to make cannon bal ls. Because of its strategic importance, the British monarchy imposed an embargo on graphite during the Napoleonic wars. Other uses of graphite include crucibles due to its refractory nature, lubrication because graphite planes can slide against each other and electrodes for motors due to its high conductivity and ease of machining. In addition , carbon fibers used for composites were achieved in 1980s providing structural strength for commercial items such as tennis rackets and airplane wings. Graphene cons ists of a single two dimensional layer of carbon atoms in a honeycomb pattern. Until as recently as ten years ago, any free standing 2 D film was believed to not exist until graphene discovery by the team of Novoselov and Geim of Manchester University, who used adhesive tape to extract a single layer of carbon from an ordinary piece of HOPG graphite. From a thermodynamic standpoint, past theories predicted a two dimensional single layer of any solid substance is unstable as a stand alone material. Experimen tal observations demonstrated the melting temperature of thin films rapidly decreases with decreasing thickness and the films will degrade into smaller more thermodynamically stable particles, decompose or form islands if not free standing. 1 The refore, the 2 D structure will try to reduce its surface area by transitioning into a 3 D structure. Although graphene has been known to exist as single layers within
15 graphite held together by van der Waals forces, the isolation and verification of graphene in 2004 r esulted in the scientific excitement we are currently realizing. Since 2002 there have been ~25,000 academic papers published on graphene and ~8400 patent applications (2013) have been submitted. 2 The honeycomb lattice of graphene is a giant aromatic mole cule which consists of two carbon atoms per unit cell and with two triangular sublattices or a rhombus as the unit cell. 3 The 2 D structure of graphene consists of hexagons with a carbon atom at each corner. The bond length between the next neighbor carbon atoms is 1.42 Ã… and the lattice constant is 2.46 Ã… . 4 The monolayer thickness of graphene is 3.4 Ã… . 5 Large area graphene is a semimetal with a zero band gap and cone shaped valence and conduction bands touching at the K points in the Brillouin zone. 5 ,6 Due to the zero bandgap, electron conduction in graphene behaves like a two dimensional gas. Electrons in graphene act as massless fermions with a large mean free path of several millimeters. 3 The scotch tape experiment of Novoselov and Geim demonstrated a 2 D material can exist in nature. The revolution in the science community from this simple process is unmatched in patents, papers published and potential applications. Indeed , graphene has been researched for virtually any imaginable scientific field of stu dy including transparent conductors, 7,8 nanoelectromechanical systems (NEMS), 9,10 high speed electronics, 11 supercapacitors, 12 batteries, 13,14 fuel cells, 15 19 solar cells, 20,21 biosensors 22 27 and strengthening composite materials. 28 Although the impressi ve properties of graphene allow for distant commercial applications, the fact remains that defect free,
16 low temperature and easy to use films will be the limiting factor for graphene to achieve its unsurpassed commercial potential. 1.2 An Impermeable Singl e Atomic Layer Membrane Several case studies in the literature demonstrate the use of graphene as an impermeable membrane to all gases except hydrogen. Studies of lateral diffusion of various metals have also been reported. However, there is little publish ed work on diffusion through the graphene basal plane especially in a solid state application. The lack of effusion for gases and liquids through graphene is easily understood by analyzing the molecular structure and bonding. Each carbon atom in graphene i s bonded to three other carbon atoms with strong covalent bonds. The opening in the et al. used DFT to find the penetration barrier for single crystal graphene requires an extremely high energy of >11.7 eV for Helium atoms. 29 For solid state diffusion barriers, although different mechanisms are at work, notably temperature and chemical reaction s, similar arguments can be made concerning the penetration barrier energies for graphene. A seminal paper by Nicolet in 1970 gave a blueprint of materials properties when used in barrier applications 30 . In this paper, Nicolet described solid state diffusi on barriers generally require several key properties to be effective. These properties include thermodynamic stability between layers, strong adhesion, low contact resistance and resistant to mechanical and thermal stresses. In addition, for semiconductor applications, the barrier should be highly thermally and electrically conducting . Additionally, Nicolet described d iffusion barriers in general should either be amorphous or have large grain sizes in order to reduce the densities of fast diffusion
17 paths a long grain boundaries and defects. Although graphene theoretically has physical properties that exceed many materials for diffusion barriers, it is a two surface material with no bulk. The interfacial properties of graphene in contact with a diffusing spe cies as well as the interfacial contact with the host substrate are the key parameter s in determining whether pristine graphene can prevent diffusion across its basal plan . 1.3 Statement O f Thesis The purpose of this research is to advance our understandin interface to chemisorbed and physisorbed metals and to test the effectiveness of graphene as a solid state diffusion barrier. Graphene growth and transfer techniques with a custom built Rapid Thermal Processing System are also reported. The principal experimental results are summarized below: Challenges and growth of single layer graphene on copper by chemical vapor deposition in a Rapid Thermal Processing System. Single layer and double layer graphene as a successful diffusion barrier to pr event titanium silicide formation at temperatures up to 900 o C. Single and double layer graphene failure mechanisms as a diffusion barrier for gold and copper. Graphene as a successful diffusion barrier to prevent the intermixing of a gold and nickel sandwi ch structure on Mg doped p GaN substrates.
18 Figure 1 1. Graphene crystal s tructure created in CrystalMaker TM which shows the hexagonal lattice structure. The unit cell and directions are indicated in blue.
19 CHAPTER 2 TECHNICAL BACKGROUND 2.1 Properties O f Graphene 2.1.1 Carbon Bonding If the electronegativity difference between two atoms is small, there is no potential for either atom give up electrons to become an ion. Atoms will attempt to acquire a noble gas configuration by sharing electrons with neigh boring atoms. Delocalization of electrons in carbon leads to hybridization. The ground state of an isolated carbon atom is 1s 2 2s 2 2p 2 . However, by promoting an electron from the 2s level f hybrid orbitals, 1s 2 2s 2 2p x 2p y 2p z . This hybridization of atomic orbitals allows covalently bonded sp 2 , sp 3 as well as unhybridized 2p orbitals. The directional bon ding of a two atom sp hybrid is a linear molecule with 180 degree separation. The sp 2 hybrid orbitals with three atoms form the backbone to the planar graphene hexagon and the bond directionality is separated by 120 degrees. sp 3 hybrids form a tetragonal, diamond, hybrid structure with all directional angles separated by 109.5 degrees. 2.1.2 Graphene Crystal Structure The crystal structure of graphene can be described as a planar structure of hexagons with carbon atoms at each of the vertices. There are tw o carbon atoms per unit cell of graphene. The graphene unit cell vectors a 1 and a 2 are expressed in Cartesian coordinates as: 3 (2 1)
20 Where Ã…, and is the lattice constant of graphene. The unit vectors in reciprocal space are given by 3 : (2 2) a. The two points at the corner of (2 3) These are the Dir ac points for graphene in which the conduction and valence bands converge. 2 configuration and the fourth electron forms the 2p z orbital which is perpendicular to the graphene plane and forms coval carbon atom provides graphene with its superior electrical performance. The C C bond length is 0.142 nm a nd the vdw radius of carbon of 0.11 nm 10 . The opening in the center 10 . 2.1.3 Graphene Electrical Properties Graphene is a semi metal with a zero bandgap and linear dispersions leading to cones in two dimensional reciprocal space 1 . The accurate modeling of the graphene band structure was accomplished in 1970 by G.S. Painter and D.E. Ellis with an advanced modeli ng technique developed at the University of Florida, working under Professor J.C. Slater, in the Quantum Theory Project. 31 Painter and Ellis developed a method for computing energy bands in solids called the Discrete Variational Method
21 (DVM). Their algorit hms were extremely flexible and were independent of any particular form of basis set or approximations for the crystalline potential. Using their code and an early form of what was later to be called the Local Density Approximation, they computed the first accurate energy bands for graphite and diamond. 32 Painter and Ellis utilized their Variational Approach applied to a single atomic layer of graphite and locations of the Brilloui n Zone. Most matter wavefunctions have quadratic dispersions as part of the Schrodinger equation: 33 (2 4) Where E c,v are the conduction band and valence band edges and m c,v are the effective masses of el ectrons in the conduction band and holes in the valence band respectively. In contrast to the Schrodinger equation, the dispersion for the Dirac equation is: 33 (2 5) Where c is the speed of light and m is the relativistic mass. For m=0 , equation (2 5) becomes: (2 6) For graphene the Fermi Velocity is 10 6 m/s or 0.003c and equation (2.6) becomes: (2 7) This analogy led to the suggestion that electrons and holes ac t like Dirac fermions in graphene with zero mass and zero gap. 3 In other words, charge carriers are
22 massless relativistic fermions with the point of intersection between conduction and valence bands labeled as Dirac points and the dispersion cones labeled as Dirac cones. 3 The electron mobility within the graphene 2D structure is theoretically predicted to be > 2x10 5 cm 2 /VÂ·s with a mean free path of several microns. 1 Indeed, this result has been verified in a study by Bolotin et al . of Columbia University 34 . In this research graphene was transferred onto a 200nm SiO 2 /Si wafer, followed by etching the underlying oxide layer in BOE. Gold contacts were deposited and the mobility of the suspended graphene was measured to be 200,000 cm 2 /VÂ·s at electron densities of ~2Ã—10 11 cm 2 . In most cases, however, graphene is not useful in a suspended state. In transferring graphene to a host substrate, the interaction between graphene and surface phonons limits mobility significantly less than the suspended mobility. For graphene on SiO 2 , which is the most common substrate for graphene, carrier mobility is limited by scattering from charged surface states and impurities, 35,36 substrate surface roughness 37,38 and SiO 2 surface optical phonons. 39 Hexagonal boron nitride (h BN) is an attracti ve dielectric substrate for use in improved graphene based devices. h BN has a large (5.97 eV) bandgap 40 and a small (1.7%) lattice mismatch with graphite. 41 In addition, h BN is relatively inert, atomically planar and is expected to be free of dangling bo nds or surface charge traps. Dean et al. from Columbia performed carrier mobility experiments on exfoliated graphene on exfoliated h BN. 42 The experimental results show when compared to graphene on SiO 2 , the carrier mobility of graphene on h BN is 3X large r with values of 60,000 cm 2 /VÂ·s.
23 2.1.4 Elastic Modulus O f Graphene The elastic modulus is a measure of bulk bond strength in 3D materials. Generally the higher the melting point of a material, the higher the intrinsic modulus. For 3D materials there will always be surface and bulk defects which limit the tensile strength to values well below the estimated intrinsic strength. For 2D materials, ultrahigh strength measurements can be obtained because the material defects are limited to in plane and not bulk d efects. Lattice defects in low dimensional materials, theoretically, should have a larger impact than 3D materials since there is no mechanism to support or relax the defect. In fact, for a single one dimensional atomic chain, a single vacancy would reduce the strength of the material to zero. Multiple studies of modeling of the strength of graphene have been conducted, including an equilibrium continuum approach 43 , atomistic modeling 44 , quantum molecular dynamics 45 and a combination of the multiple approac hes 46 TPaÂ±0.4. The difference between the various approaches was based on the estimated thickness of the single atomic layer of graphene. In further studies, Lee et al . of Columbia University, u sed nano indentation of freely suspended graphene films in an atomic force microscope (AFM) to verify that graphene isolated through mechanical exfoliation is the strongest known material in its defect free pristine state. 47 The measurements showed pristin e graphene can achieve its intrinsic strength before rupturing occurs. In another study the Lee group analyzed samples with CVD graphene grain sizes of 1 5Âµm and 50 200Âµm. 48 The results show a slight decrease of at most 15% in graphene with a higher densit y of grain boundaries than samples with a larger grain size. For both sets of samples, the defective graphene approached the strength
24 highest of any measured material. 49 Unlike other materials, the modulus does not decrease with a high density of high angle grain boundaries. 50 Graphene has a unique ability to covalently re bond around defects and grain boundaries. In metals the elastic modulus decreases with increasing temperatu re. But in graphite and graphene an anomaly exists. In the temperature range from 77 150 o C, the elastic modulus decreases as expected, however from 150 2000 o C, the elastic modulus increases followed by another decrease at even higher temperatures. 2.1.5 T hermal Properties O f Graphene In general, when using a material as a diffusion barrier, it is imperative to look at the bonding between layers and thermal expansion coefficients in order to assess the interlayer stresses. For graphene, the interaction with an underlying substrate or film is by weak van der Waals forces, except in the case of chemisorption, and thus bond breakage or strain failures from a lattice mismatch mechanism are minimized. In analyzing the Coefficient of Thermal Expansion (CTE) compon ents, graphene is an odd material in which the thermal expansion coefficient is positive at low temperatures and transitions to a negative CTE at higher temperatures. 51 There is much speculation at what temperature and the amount of expansion or contractio n of a graphene film upon heating or cooling. The difficulty of ascertaining an exact plot of CTE vs Temperature is due to the investigational difficulties in testing a single atomic layer. Free standing graphene is extremely difficult to measure. Tests with graphene transferred onto a host substrate are biased due to the CTE of the underlying substrate. Suspended graphene tests are pinned in two or more locations. Very small interactions of graphene with a substrate material can greatly bias the value o f CTE, while strong interactions (chemisorbed) will
25 force graphene to have a positive CTE at all temperatures. Zakharchenco et al. modeled the CTE of graphene by Monte Carlo simulations.51 It was discovered that the negative CTE of graphene was consistent to 900 K, at which stage the thermal function to investigate the TCE of graphene and carbon nanotubes. 52 At close to absolute zero, the graphene carbon atoms expand close to a distance of 1% from their equilibrium positions. For simulations, the CTE of graphene at 300K is 6x10 6K 1. Bao et al. measured the coefficient of thermal expansion of single layer graphene by suspending graphene across small trenches etched in SiO2 53. By measuring the deflection before and after heating to 700 K with an SEM, it was determined the CTE of graphene is 7 x 10 6 K 1, which is much larger than measured CTE of graphite ( 1 x 10 6 K 1). Yoon et al. used Raman Spectroscopy to measure the g raphene CTE in the temperature ranges of 200 400K. 54 This research measured the Raman G band during heating and used the shift of the Raman G band of single layer graphene on SiO2. The SiO2 substrate positive CTE was subtracted from the G band shift to ob tain the CTE of graphene. It was determined the CTE of graphene is 8.0 x 10 6K 1. For diffusion barriers, the chemical resistance of the film must be sufficient to withstand further processing steps. A barrier that can be etched, degraded, oxidized or reduced by downstream process steps will fail inadvertently without actually failing from a diffusion process. 30 Graphene, like its 3D counterpart graphite, is chemically resistant to a variety of solvents, acids, bases an d gases. In CVD graphene, growth on copper or nickel substrates and subsequent transfer process, the main etchant for substrate removal is ammonium persulfate (NH 4 ) 2 S 2 O 8 , iron nitrate Fe(NO ) or iron chloride
26 FeCl 3. 55,56 These chemistries obviously do not detrimentally affect the graphene. In addition the graphene film is subjected to acetone treatments, for hours to days in length, to remove PMMA or PMDS support structures. A study of comm on semiconductor based chemistries and their effects on the conductivity of graphene by Chen et al. 55 These experiments demonstrated graphene is chemically resistant to acetone, isopropanol, diluted hydrochloric acid, diluted ammonium hydroxide and UV Ozon e. However graphene is etched in low density oxygen plasmas. Additionally, a study by Liang et al ., 56 used standard RCA cleaning methods to remove contaminants after copper substrate etching. The solutions for the RCA cleaning process consisted of 20:1:1 H 2 O/H 2 O 2 /NH 4 OH and 20:1:1 H 2 O/H 2 O 2 /HCL. For electrochemical etching of the copper substrate, Wang et al . demonstrated that graphene is not affected by a electrolytic solution of K 2 S 6 O 8 . 57 For other types of graphene growth mediums, graphene growth on nickel substrates require a nitric acid etch process to remove the nickel underlying film during the transfer process, generally this is 4:1 H 2 O:HNO 3 58 or FeCl3. 59 For graphene growth on ruthenium substrates the support structure is etched in a solution of ceric ammonium nitrate and acetic acid. 60 For gold based substrates, the etchant for removal of the growth substrate is potassium iodide. For gaseous base chemistries, graphene can be etched at high temperatures in reducing and oxidizing environments. Zhang et al. 61 annealed graphene samples in hydrogen at 800 o C. The results were an anisotropic process that selectively etched graphene grain boundaries. Xie et al . 62 used a hydrogen based plasma at 300 o C to remove defective edges to create graphene nano ribbons. F or oxidizing chemistries,
27 Cui et al. 63 utilized graphene grown on Ru for gas detector applications. The graphene was stable in a 100% oxygen environment up to 600K. At temperatures above this value, graphene would start degrading at defect sites followed b y bulk removal. In another study, Gong et al . 64 used various gases, consisting of O 2 , forming gas, NO 2 , CO 2 and C 2 H 4 , in order to remove PMMA residues from the surface of graphene. This research showed that by monitoring the graphene Raman peaks in situ, o xygen partially etches graphene at temperatures exceeding 500 o C, NO 2 complete etches graphene at 500 o C, while CO 2 is effective in removing PMMA residues at 500 o C while leaving the graphene surface intact. For the reducing gases, both forming gas and C 2 H 4 a re not adequate in removing PMMA residues, yet they are not detrimental chemistries to the graphene surface. For long term stability of graphene, results may not be as positive. Schriver et al. 65 performed a long term study of graphene as an oxidation bar rier to copper corrosion. Graphene on copper samples were exposed to air up to 250 o C for seventeen hours and to room temperature for 18 months . Their results show over long time scales, O 2 and H 2 O infiltrate defects in the graphene and oxidize or corrode the surface of the underlying copper substrate. Graphene catalyzed corrosion proceeds nonuniformly, which leads to cracks in the resulting oxide and further corrosion. Conducting graphene coatings can further promote corrosion by facilitating electrochemic al reactions both across the surface and through the bulk of the substrate, whereas typical native oxides passivate the surface and terminate the electron transfer needed for continued corrosion. In another study by Zhou et al . 66 graphene on copper samples were stored in a plastic Petri dish at room temperature for up to 6 months in
28 the dark. Graphene coated copper foil developed a much thicker oxide film in places than the uncoated bare copper samples. In addition, when transferred to a SiO 2 substrate, the graphene in areas of copper oxidation were cracked and defected, while graphene in areas without copper oxidation remained pristine. Although graphene is compatible with the previous listed chemicals, it is not a comprehensive list. In general graphene is immune to degradation from most acids, bases and solvents but it can be etched in aggressive plasma environments and high temperature oxidizing environments. Long term exposure to air and humidity even at room temperatures will degrade graphene on copper substrates by stresses from copper oxidation. 2.1.7 Defects I n Graphene For any diffusion barrier, the efficiency is proportional to the number of structural defects in the film. 30 For a single atomic layer of graphene, the importance of defects is compou nded since a 3 D structure is non existent for prevention of foreign atom diffusion. Graphene is unique in that unlike any other material the surface reconstruction can host lattice defects with minimum energy costs. 1 The number of carbon hybridization bon ding that allow different numbers of nearest neighbors leads to energetic stable configurations without dangling bonds but with non hexagonal bonding. The sp 2 carbon atoms can reconstruct into various polygons without under coordinated atoms. 67 Graphene h as similar defects as a three dimensional crystal, however the defects are limited to two dimensions. Similar to the 3 D crystals, defect migration has important consequences on the properties of graphene. Defect migration as in 3 D
29 crystals is governed by an activation barrier which is defect dependent and increases exponentially with temperature. 67 126.96.36.199 Stone Wales d efect A Stone Wales defect is generated by a pure reconstruction of the graphene lattice with no atoms added or removed. The reconstructio n consists of a switching to pentagons, hexagons and heptagons. Two hexagons are transformed into two pentagons and two heptagons by rotation of the in plane C C bonds by 90 degrees. This defect has a formation energy of approximately 5 eV. 67 The high form ation energy of the Stone Wales defect reflects a very small equilibrium concentration in the graphene film. 188.8.131.52 Vacancies Single vacancies in graphene result in a nonlinear atomic arrangement. Local atoms rearrange to lower the free energy of the latt ice. The rearrangement leads to the filling of two of three dangling bonds towards the missing atoms. The formation energy of a single vacancy is approximately 7.5 eV and the migration of graphene vacancies is 1.3 eV which is low enough to allow migration at lower temperatures. 67 Double vacancies in graphene are created by the combination of two single vacancies or by two missing carbon atoms. The resulting reconstruction of carbon bonds results in two pentagons and one octagon and no dangling bonds, instea d of the pristine four hexagon model. The energy of formation of a double vacancy is 8 eV which is close to the formation energy of a single vacancy 1 . However the energy per missing atom is 4 eV which is much lower than for a single vacancy. The migration energy of a double vacancy is 7 eV which essentially eliminates movement until very high
30 temperatures are reached. It is more energetically favorable for the double vacancy to reconstruct rather than migrate. 184.108.40.206 Line d efects Line defects in graphene g enerally impact the performance properties more than any other defect. Line defects, which include grain boundaries and dislocation lines, are formed due to separate domain growths of different crystal orientation. As with 3 D crystals, the density of the line defects affects the physical and electrical performance more than any other defect. Although similar physically, the line defects in graphene differ from 3 D line defects because of the 2 D nature of the film. Conventionally, the definition of a dislo cation line in a 3 D crystal requires a dislocation line and a burgers vector which is not applicable in a 2 D space and obviously screw dislocations do not exist in 2 D space. 2.2 Graphene Growth Techniques 2.2.1 Survey O f Techniques For graphene, there a re four methods to obtain single atomic layer films of graphene. These include mechanical exfoliation, which in its simplicity, uses a layer of scotch tape to peel individual atomic layers from a piece of Highly Ordered Pyrolytic Graphite (HOPG). The graph ene can be attached to a host substrate simply by placing the tape/graphene on the substrate and re peeling. Although the discovery of graphene and thus the Nobel prize was awarded based on this method, it is not consistent or efficient. The second method of forming graphene is to heat a single crystal substrate of SiC to temperatures up to 1500 o C in ultrahigh vacuum. Carbon will segregate to the surface of the SiC substrate as silicon layers evaporate leaving a single atomic layer of graphene. Although hig h quality graphene can be achieved with this method, as of now,
31 there is no method of transferring the graphene to another substrate. The third method of obtaining graphene is to chemically derive graphene from graphite oxide. In this method, graphite is t reated with a strong oxidizer consisting of a mixture of sulfuric acid (H 2 SO 4 ), sodium nitrate (Na NO 3 ), and potassium permanganate (KMnO 4 ). This process is called intercalation, indivi dual graphene sheets are peeled away. Once in solution the graphene oxide is reduced back to graphene in a mixture of hydrazine hydrate. 68 The fourth method, which is the most common and the method we will concentrate, is growing graphene in a carbon conta ining chemistry on catalytic metal substrates. Graphene on metal surfaces has been observed at least 40 years ago. The formation of graphene was first observed during heating of Pt and Ru single crystal samples and their surface reconstructions. 69 71 LEED measurements showed the surface reconstruction was due to the presence of an aromatic carbon molecule of one, two or three layers. 72 With the discovery of single layer graphene and the subsequent research boom currently happening today, there has been a re newed interest of carbon on catalytic metal surfaces. 2 .2.2 Chemical Vapor Deposition O f Graphene Graphene can be grown on catalytic metal surfaces by two separate mechanisms, segregation of bulk dissolved carbon to the surface and by surface decompositi on of a gaseous hydrocarbon molecule. Note, the vernacular gaseous decomposition and chemical vapor deposition (CVD) is interchangeable with respect to graphene growth in this study. The segregation of bulk dissolved carbon to the surface method requires a substantial solid solubility of carbon in the metal substrate. Metals that have been used for graphene growth with the segregation mechanism include
32 Co 73,74 , Rh 75,76 , Ni 77 81 and Ru 82 85 for surface decomposition Pd 86 , Pt 87 89 , Ir 90 93 and Cu 94 102 have be en used. In a 2009 study, the UT Austin Ruoff group verified the precipitation mechanism for graphene grown on Ni and the surface decomposition method on Cu. 103 By alternating equal amounts of 12 C and 13 C as the precursor gas and analyzing the separation o f 12 C and 13 C Raman modes, it was found that equal amounts of 12 C and 13 C segregated to the surface of Ni regardless of the precursor flow sequence. For the Cu tests, a clear distinction in graphene domains while flowing 13 C or 12 C was evident across the g rowth plane. Recent results of graphene growth on relatively inexpensive polycrystalline Ni and Cu substrates have triggered interest in optimizing CVD conditions for large area deposition and transfer for these two metals. Analyzing the other catalytic metals it is apparent that some are exotic and expensive. Since the graphene host substrate is sacrificial, etching is much more difficult with exotic metals as compared to Ni and Cu. The fundamental limitation of utilizing Ni as the catalyst is that singl e and few layered graphene is obtained over few to tens of microns regions and not homogeneously over the entire substrate. The lack of control over the number of layers is attributed to the fact that the segregation of carbon from the metal upon cooling o ccurs rapidly within the Ni grains and at the grain boundaries. In contrast to Ni, exceptional results in terms of uniform deposition of high quality single layered graphene over large areas have been achieved on polycrystalline copper foils. For graphene on copper, over 95% of the copper surface is covered by single layered graphene and the reaction is self limiting. The remaining graphene is 2 3 atomic layers, which is independent of growth time o r heating and cooling rates. The growth on
33 copper is simple and somewhat inexpensive, making high quality graphene over large area readily manageable. Furthermore, thin copper foils can be easily etched with standard chemistries and transfer onto desired substrates can be easily achieved. These features as well as the fact that copper is inexpensive, make it an appealing process for the deposition of graphene. For these reasons we will concentrate on the graphene growth on copper for this study. 2.2.3 Gro wing Graphene On Copper S ubstrates In growing graphene on copper, a thin copper substrate (25 microns) is heated in vacuum to >1000 o C in a small partial pressure of hydrogen for removal of surface oxi des, followed by the graphene growth in a mixture of hydrogen and a volatile hydrocarbon such as methane, ethylene 104 or acetylene. 105 Generally, prior to graphene growth the copper substrate is etched in nitric acid to remove copper oxides. The importanc e of hydrogen in the copper substrate pretreatment and the graphene growth process is one of the critical controlling factors in creating high quality graphene. In the pre treatment step, a flow of hydrogen at elevated temperatures (>1000 o C, 30 min) reduce s and removes the CuO and Cu 2 O native oxide on the copper substrate. 106 The copper substrate will not act as a catalyst for graphene growth unless the copper oxide layer is removed. In addition, annealing in hydrogen also increases the copper grain size an d eliminates surface defects. Hydrogen in the graphene growth stage has been studied by the Ruoff group at UT Austin. 107 In their experiments a 200 nm amorphous carbon film was sputtered on copper substrates followed by annealing at the standard graphene g rowth temperature of 1035 o C without H 2 flow and with diluted H 2 and pure H 2 . The Ruoff group discovered that no graphene was detected on the samples unless a partial pressure of hydrogen
34 was present. The conclusions of their study demonstrated that graphen e growth is through a reaction with hydrocarbons rather than simply carbon species. For graphene growth on copper the reaction can be modeled similar to a general CVD process including dependencies on residence times and boundary layers. The kinetics of lo w pressure graphene CVD are 1) carbon species diffuse through a boundary layer and reach the surface. 2) At the surface there is catalytic decomposition of the carbon atoms to form CxHy, 3) carbon atoms nucleate 4) carbon atoms through surface diffusion co mbine to form the graphene lattice and graphene islands begin to grow 5) inactive species such as hydrogen are desorbed from the surface, form molecular hydrogen and 6) diffuse away from the surface and through the boundary layer and are carried or pumped away. 108,109 Like most CVD processes, in graphene CVD there are two limiting processes; mass transport and surface reaction. The first order equations for the two processes are: 110 F 1 =h g (C g C s ) (2 7) F S = K s C s (2 8) Where F 1 is the flux of active specie s through the boundary layer and F 2 is the flux of active species consumed at the surface, h g is the mass transport coefficient and K s is the surface reaction rate constant. These fluxes are in series and the slower of the two processes is the rate limitin g step during graphene synthesis. At steady state F 1 =F 2 =F T and combining equations F T =[K S *h g /(K s +h g )]/C g (2 9) In mass transport limited processes the flux of active species through the boundary layer is the rate limiting step (h g <
35 processes (h g >> K s ) , the consumption of active species at the surface is rate limiting, all other sub processes including transport of active species to the surface, adsorption, desorption and transport away from the surface are faster than the surface reaction rate. 2.2.4 Modern Methods O f Graphene Growth Although the graphene growth process has been well studied and well understood in terms of the kinetics and growth mechanisms, there are two recent studies that stand out in terms of growth of defect free graphene. The first study was conducted by the Ruoff group at UT Austin and published in late 2013 111 . In this study, oxygen was intentionally added to the graphene growth chemistry. This approach was counterintuitive of past studies which ind icate oxygen etches graphene at high temperatures. The Ruoff group discovered that oxygen on the Cu surface not only decreases the graphene nucleation density but also fosters ultra large domain sizes. Electrical measurements from the Ruoff study indicate the graphene quality is comparable to mechanically exfoliated single crystal graphene, in spite of being grown in the presence of oxygen. The carrier mobility measured for different samples ranged from 40,000 to 65,000 cm 2 V 1 s 1 at 1.7 K and from 15,000 to 30,000 cm 2 V 1 s 1 at room temperature. The mechanism of the large domain sizes is based on suppressing the nucleation density of the graphene. It is well known during nucleation and growth processes, larger nucleation de nsities lead to smaller grain sizes and vice versa. In addition it was observed that the time for complete coverage of the copper host substrate by graphene was reduced by half. In general, graphene growth is an edge attachment limited growth. In contrast, graphene domains grown with oxygen became dendritic which is typical of diffusion (mass transport) limited growth. The morphology
36 change indicates that, with the introduction of O, carbon attachment at domain edges is no longer rate limiting, and domain g rowth is instead governed by carbon diffusion. For the Ruoff experiments, adding oxygen into the growth chemistries oxidizes high energy nucleation sites on the copper growth substrate which in turn reduces the density of nucleation sites. This reduction in the density of high energy sites reduces the nucleation density and has led to grain sizes of 1.0 centimeters. At the time of publishing, the grain size of the Ruoff graphene was close to an order of magnitude larger than all other previous experiments. The second study which is perceived as a major breakthrough in the growth of graphene is based on Lee et al. from SKKU in the use of Germanium as the host substrate. 112 In most previous work to date graphene was grown on metals that act as a catalyst to g fundamental theories of grain growth. The first approach involves growing a single grain to as large of size as possible from a single nucleation site. The second approach involves the epitaxial growth of graphene on a single crystal substrate. If multiple nucleations of the graphene seeds occur but with perfect rotational alignment, the uni directionally aligned seeds can grow and coalesce into a uniform single crystal layer without grain boundar y defects, even if the nucleation density is high. The second catalytic activity, which ca n lower the energy barriers for the catalytic decomposition of carbon precursor. 2) The extremely low solubility of carbon in Ge even at its melting temperature (<10 8 atoms/cm 3 ). 3) The well defined and anisotropic atomic arrangement
37 of the single crystal Ge surface. 4) The availability of a large area single crystal surface via epitaxial Ge growth on Si wafers. 5) The small difference in thermal expansion coefficients between Ge and graphene, which suppresses intrinsic wrinkle formation. The process for gr owing graphene on Ge begins with the surface preparation of a hetero epitaxial Ge on Silicon wafer. The wafer is cleaned in 10% HF to remove a native oxide and to create a hydrogen terminated surface. This is followed by re growing an additional Ge layer w as grown in GeH 4 . In the same chamber, without breaking vacuum, graphene was grown in a standard mixture of CH 4 and H 2 . The graphene was transferred by depositing a thin layer of gold followed by lifting the graphene from Germanium with sticky tape followe d by etching of the gold overlayer. over a 50mm substrate, which is quite impressive. Essentially, the holy grail of graphene growth has been discovered. The room temperatur e mobility values given by Lee were 10,620 cm 2 /VÂ· s which is roughly half of the Ruoff study. If indeed mobility is a function of defect densities, it should be expected that the single crystal graphene should have a much higher mobility. Closer examinatio n of both papers reveals the Ruoff mobility measurements were based on graphene on a hexagonal BN substrate and the Lee experiments were based on graphene on an SiO 2 substrate. 2.3 Metrology O f Graphene 2.3.1 Raman Spectroscopy O f Graphene Although there a re many techniques for characterization of carbon based materials such as Infrared Spectroscopy, Photoluminescence, Optical Absorption etc., Raman Spectroscopy is the easiest, fastest and provides the most information. Raman Spectroscopy is a metrology tec hnique that provides information about molecular
38 vibrations, the technique involves shining a monochromatic laser at a sample and measuring the scattered light. The majority of the scattered light is due to elastic Rayleigh scattering which is at the same frequency as the incident beam; a very small amount of light from inelastic scattering is shifted in frequency from interactions with the incident laser and the vibrational energy levels of the sample. Rayleigh scattering is considered in most systems to b e at the 0 cm 1 position. The Raman Effect is based on molecular deformations in an electromagnetic field. The laser electromagnetic wave interacts with the sample and induces a dipole 113 Because of periodici ty within the sample, molecules start vibrating with a characteristic frequency m . The vibrating molecules can emit light with three different frequencies. In Raleigh scattering, a molecule absorbs a photon with frequency o , the excited molecule returns back to the original vibrational state and emits a photon with the same frequency o . 113 If a photon with frequency o is absorbed by a Raman active molecule, part of the photons energy is transferred to the Raman active mode with frequency m and the res ulting frequency of light is reduced to o m . This is called Stokes frequency. If Raman active molecule absorbs a photon with frequency o and is already in a vibrational state the excess energy of excited Raman active is released and the molecule retur ns to the basic vibrational state and the resulting frequency of scattered light is o + m . This Raman frequency is called the Anti Stokes frequency. 113 For graphene, Raman Spectroscopy is a rapid, nondestructive technique that is also an invaluable tool to determine the quality of films and carrier densities prior to device fabrication or electrical measurements. Raman modes are sensitive to physical
39 distortions and are also used to distinguish between single layer, bilayer and multilayers of graphene, 113 1 17 strain, 118,119 doping, 119,120 stacking sequences, 117 and defect density. 121 Graphene has two atoms per unit cell and therefore six separate phonon branches 113 which include out of plane transverse acoustic (oTA), in plane transverse acoustic (iTA), lon gitudinal acoustic (LA), out of plane transverse optic (oTo) (not Raman active but IR active), in plane transverse optic (iTO) and longitudinal optic (LO). The G band (for graphite) is present in the Raman spectra of all sp 2 carbon systems. It is related t o the in plane C C bond stretching and is centered at (1582 cm 1 ). 114 The G band is a first order Raman scattering process involving iTO and LO phonon modes at the BZ center. Due to the small mass of carbon atoms and the strong C C covalent bonding, the G band has a relatively high Raman peak. 1 which has the largest intensity, is a second order process which involves two in plane transverse optic phonons. The large intensity as compared to the other bands is due to a triple resonance process, where all steps in the scattering process are resonant 114 . The 2D band exhibits strong frequency dependence and this dispersive behavior causes a strong sensitivity to any changes in the electronic or phonon structure. The third peak in a graphene Raman spectrum is the D band (for defect) at 1350 cm 1 . This peak is also a second order process except one phonon and one defect participates in the Raman scattering instead of the two phonon scattering in the 2D band. A lattice defect is required to observe the 2D band because the double resonance with a single phonon requires a defect to absorb the extra
40 momentum of the scattering process. 113 active at a frequency of 1626 cm 1 . 114 When a nalyzing Raman spectra, we can learn valuable information for graphene by studying the band shifts and relative intensities of the three peaks. To determine the number of graphene layers, the G band intensity is studied. The intensity of the G peak increas es in a linear relation with the number of graphene layers. As more sp 2 bonded carbon atoms are being detected, the intensity of the G peak increases. In addition a red shift occurs with increasing layers. However, band shifts of the G band also occur with doping and strain and thus a red shift in not a valid indicator of the number of graphene layers. However, the G band intensity is independent of doping and strain and is the key source of determining the number of graphene layers. 116 In determining the n umber of graphene layers, the 2D peak is also studied. 117 In single layer graphene, the 2D peak is sharp and distinct at 2700cm 1 and is approximately 4X the intensity of the G band. 119 In graphene layers >1, the 2D band splits into four Lorentzian bands d ue to interactions between planes. 117 The overall 2D band will broaden and shift to higher or lower frequencies depending on the AB stacking of the graphene layers. 117 For layers > 5, the 2D band Raman signal will transform into two components reflecting a transformation into graphite. 2.3.2 Strain I n Graphene Strain levels in graphene are an important issue especially when the graphene is placed on any substrate subjected to temperatures. When subjected to strain, the C C bonds stretch and distort the hex agonal lattice. In graphene strain can induce a slight bandgap from the distortion. In general tensile strain induces phonon softening (red shifts) and compression causes phonon hardening (blue shifts) in both the G and 2D
41 peaks. In addition the G peak spl its into two components at higher strain levels due to symmetry distortion of the crystal lattice. 119 Uniaxial strain obviously also reduces the hexagonal symmetry of the system and softens the frequency( Red Shift)of the optical phonon branches. 118 To qu antify strain in graphene a higher shift in the 2D mode is utilized to distinguish strain since the G band frequency is dependent on doping. 2.3.3 Doping Effects A nd Graphene The G band Raman spectra of graphene not only can be used to determine the number of graphene layers but can also be used to determine doping levels by analyzing the band shift and the FWHM. It is imperative to note that unintentional doping of graphene due to interactions with the substrate (usually SiO 2 ) usually occurs unless the gr aphene is free standing. As a reference, the intrinsic carrier density of graphene is < 5x10 11 cm 2 . Interactions with the substrate can unintentionally dope the graphene to levels an order of magnitude above intrinsic levels. The maximum intentional dopi ng level of graphene is approximately 5x10 13 cm 2 . 120 In analyzing the G band shift of graphene, Caridad et al . showed a doping influence of 9x10 11 cm 2 per cm 1 shift increase in the G band Raman signal. 121 For the G band a shift to higher frequencies occ urs for both electron and hole doping. The 2D peak instead of the G peak in graphene is utilized to study doping levels. The 2D peak is more sensitive to doping and the type of doping. For electrons, the 2D peak decreases as n type doping increases, while the 2D peak increases with p type doping. The shifts are due to stiffening or softening of the phonon modes with excess charges leading to an expansion of the crystal lattice or excess holes lead to a contraction of the crystal lattice. 116 The 2D band se nsitivity to n type doping is 2.6 x10 12 cm 2 per cm 1 shift and 2.0x10 12 cm 2 per cm 1 shift for p type. 120
42 2.3.4 Temperature Effects A nd Graphene The Raman signal of graphene with increasing temperature was studied by Calizo et al ., in the temperature range of 113 373K. 122 A downshift of the G band was observed with increasing temperature and is based on the elongation of the C C bond due to thermal expansion and coupling of phonon modes. Calizo discovered a linear relationship of the G and 2D band freq uency and temperature exists in the range studied. However for the 2D band the frequency shift is 4X that of the G band shift. 2.4 X Ray Photoelectron Spectroscopy X Ray Photoelectron Spectroscopy (XPS) was developed in the 1960s by K. Siegbahn at the Uni versity of Uppsala who won the Noble Prize in 1981. Surface analysis by XPS is accomplished by irradiating a sample under high vacuum with monochromatic x ray types can be utilized on a particular sample if discrepancies exist or can be used as a quality check to e nsure rays are use throughout this research. The x ray photons have limited penetration depth in a solid and usually the limit is 1 10Âµm. The photons interact with the atomic electron cloud causing electrons to be emitted. The emitted electrons have kinetic energies given by: (2 10) Where h is the energy of the photon, BE is the binding energy of the atomic orbital and is the spectrometer work function. Each ele ment has a unique set of binding energies. XPS can be used to identify and determine the concentration of elements on the surface. Variations in the binding energies occur due to chemical
43 potentials and polarizability of compounds. These chemical shifts ca n be used to identify the chemical bonding of a sample. The probability of a photon interacting with a materials orbital cloud is small compared to an ejected electron interacting with the orbital cloud. Therefore the photoelectrons that are detected are limited to tens of angstroms from the surface. The electrons leaving the surface are detected by an electron spectrometer which is based on the electron KE. The analyzer utilized an energy pass filter is which only electrons of a certain energy will reach the detector. A variable electric field is used to scan for different electrons energies which give the corresponding elemental information. The spectrum is obtained as a plot of the number of detected electrons per energy interval versus the electron kin etic energy. The XPS system used in this research is an Ulvac Phi 2300 XPS system at UT Dallas. The UT Dallas XPS has Al (E =1486.6 eV) and Mg (E =1253.6 eV) dual monochromatic sources. 2.5 The Four Point Probe 2.5.1 Sheet Resistance Determination By T he Four Point Probe Method The four point probe technique was originally developed by Wenner in 1916 to 123 1954 Valdes adopted the technique for semiconductor wafer res istivity measurements by measuring the resistivity of Germanium. 124 The sheet resistance ( Rs ) of a thin film or diffused layer is the resistance exhibited in a region of equal length and width which has a thickness of t. 125 The value of Rs is expressed in units of ohms/square and is related to the resistivity of the thin film or diffused layer. In order to determine Rs , a current I is forced between two outer
44 probes and the voltage drop V between the two inner probes is measured. To prevent erroneous readi ngs due to thermoelectric heating and cooling, the measurement is performed first with current in the forward direction followed by a reverse current measurement. The two voltage measurements are averaged. The use of four probes has important advantages o ver measuring resistivity with a two probe approach. With the two probe method, each probe serves as a current and voltage source. The total resistance between the probes is given by: 125 (2 11) Where R c is the contact resistance at each probe, R sp is the spreading resistance under each probe and R s is the resistance of the measured film. For a four probe method, the two current carrying probes still have current and spreading resistance but the two voltage prob es do not. The voltage is measured with a high impedance voltmeter or potentiometer which draws very little or no current. The two parasitic resistances R c and R sp are negligible because the voltage drop is very small due to the very small current that flo ws through the voltage probes. The potential at a particular distance from a current carrying electrode is given by: 125 (2 12) Where V is the voltage potential, is the resistivity of the material, I is current and r is the distance where V is measured. For four probes of equal spacing ( s) , in a linear arrangement the resistivity can be simply calculated from : (2 13)
45 The above equation assumes a semi infinite material which is not the case for most semiconductor measu rements. Generally probe spacings are on the order of 635 equation must be modified for finite geometries by adding a correction factor F . (2 14) T he correction factor in four point probe resistivity measurements corrects for edge, thickness and probe placement effects. The correction factors have been calculated by various techniques, including the method of images, complex variable theory, and meth od of Corbino sources among others and can be combined into a product of three separate correction factors: 125 (2 15) Where F 1 corrects for sample thickness, F 2 corrects for lateral sample dimensions and F 3 corrects for probe placement relative to the sample edge. For very thin samples and a non conducting bottom wafer boundary and as long as the film thickness ( t ) is less than or equal to one half the probe spacing ( s ) the correction factors simplify and the resistance can be determined with the following equation: 125 (2 16) The sheet resistance in ohms/square is then: (2 17 ) The four point probe used in this research is a manual load Alessi Four point probe which uses a C4S probe head with 1mm spacing between stationary tungsten carbide probe tips. The current generation and voltage measurement is with a Quadtech
46 LR2000 milli ohmmeter. The constant current ranges of the Quadtech system are between 1mA to 1 amp with measurement accuracy of 0.05%. 2.5.2 Diffusion Constants B y Four Point Probe Method The four point probe measurement in this research is used to determine the diffusion constants, activation energies and diffusion barrier effectiveness. Typically other analytical techniques such as XRD, RBS, Auger, SIMS or cross sectio nal analysis are used to determine penetration through a diffusion barrier with SEM or TEM analysis. Literature searches show the first correlation of sheet resistivity measurements with barrier failures was originally conceived by Karen Holloway of IBM Re search Labs in 1990 126 . Holloway conducted in situ four point probe sheet resistance measurements while annealing 100nmCu/50nmTa/Si stacks from 25 o C to 750 o C in 3 o C/min ramp rates. At 660 o C an abrupt rise in the sheet resistance was observed indicating a m ore resistive structure was formed. At higher temperatures above 660 o C the measured resistance decreases with temperature in a manner which is consistent with the intrinsic semiconducting behavior of the silicon substrate. The sheet resistance measurements were continued while cooling to room temperature. It was observed the resistance increase was irreversible. Holloway analyzed the failed samples with RBS, Auger, TEMs and SEMs to determine if the abrupt rise in sheet resistance was due to the tantalum bar rier failure or due to a reaction between the Tantalum layer and silicon which will form TaSi 2 . The analysis revealed at annealing temperatures of 750 o C there was a Tantalum silicide formation, however there was little to no copper left on the sample surfa ces. TEMs showed copper penetration past the Tantalum barrier, a 120nm TaSi 2 layer which had formed and approximately 1
47 H. Ono et al of Kawasaki Steel LSI Research Center in 1993 showed a correlation between sheet resistance and XRD measurements while investigating Cu/M/Si multilayers with M=Cr , Ti, Nb, Mo, Ta and W 127 . The thickness of the M barrier metal was fixed at 60nm while the Copper overlayer was set at 500nm. All structures were annealed in H 2 at 200 800 o C. The testing showed a direct correlation of the normalized intensity change of Cu (111) XRD diffraction peaks with the resistivity change of all samples measured. RBS measurements confirmed the Cu layer was reduced in thickness as annealing temperatures increased and as the barrier layer failed. The effectiveness of the M barrier mater ials is demonstrated by only slight drops an almost non existent Cu (111) diffraction peak. The correlation with sheet resistivity shows almost no change in the W, Ta s amples and a drastically large increase for the failed barrier materials. The correlation of sheet resistivity measurements with barrier failures and diffusion coefficients of the barrier was published in 2003 by Kuo et al of the National Taiwan Universit y of Science and Technology. 128 In this research Kuo investigated the effectiveness of 60 nm (Ti,Zr)N barriers to copper diffusion at elevated temperatures of 500 850 o C. This study compared XRD measurements with four point probe sheet resistivity measurem ents. In the XRD analysis Kuo analyzed samples for copper (111) and (200) diffraction peaks before and after annealing. As expected as the annealing temperature increased to a point where the kT thermal energy neared or exceeded the activation energy of me tal diffusion through the barrier and the (Ti,Zr)N diffusion barrier ceased to prevent copper penetration. Kuo observed the copper (111) and (200)
48 diffraction peaks decreased with time until a Cu 3 Si peak emerged. In conjunction with the XRD analysis four p oint probe sheet resistivity measurements were also conducted on all samples. Kuo measured the threshold time to barrier failure by measuring sheet resistance of the copper at various time intervals at a constant temperature. The resulting data showed a dr astic increase in copper resistivity as the (Ti,Zr)N barrier increase in sheet resistivity occurred at a time of three minutes before the XRD measurement could detect a silicide formation. Thus, a more accurate indication of barrier failure was ascertained. The theory of why the sheet resistivity of the copper thin film was a better predictor of barrier failure was due to the fact that XRD analysis needs a large volume of compound, formed by reaction between copper and silicon, to be able to detect the barrier failure. Thus, the FPP analysis for the calculation of Cu diffusion coefficient is more precise and sensitive than that of XRD. The correlation to diffusion constant s was made by utilizing the Einstein equation for diffusion in thin films: L=2(Dt) 1/2 (2 18 ) The average diffusion length of atoms L can be approximated by 2(Dt) 1/2 . where D is the diffusion coefficient of Cu in (Ti,Zr)N and t is the diffusion time. The D value of Cu in the barrier layer at elevated temperatures can be obtained by substituting the barrier thickness of 6nm for L . By plotting the threshold times of the onset of increasing sheet resistivity of copper with respect to temperature, the variable s of the diffusion Arrhenius equation can be calculated. The Arrhenius plots of the D values as a function of annealing temperature in the range of 500 850Â°C are shown in Figure 2 11 . The
49 temperature dependence of the diffusion coefficient is expressed by t he Arrhenius relationship of D= D 0 exp( Q/kT) (2 19) Where D 0 is the pre exponential factor and is the slope of the D vs 1/T plot, Q is the activation energy for diffusion, k is the Boltzmann constant, and T is absolute temperature. 2.6 Diffusion Barrier s Diffusion Barriers in microelectronics have been studied for many years. For aluminum based metallization schemes, Ti and TiN are commonly used as a diffusion barrier to junction spiking with the Titanium used as an adhesion promoter. In addition Ti/TiN has been used in Tungsten plug processes to reduce contact resistance, improve adhesion and protect silicon from fluorine in WF 6 CVD chemistries. Diffusion barrier properties have been well studied as well for multilevel copper/low k metallization schemes of current semiconductor devices where copper is a severe contaminant. Copper not only is a fast diffusant in silicon and SiO 2 , but it forms low temperature silicides and midgap traps, recombination centers and reduces minority carrier lifetimes. 129 Many i nvestigations have been carried out using various metals and compounds to prevent copper from diffusion into silicon or SiO 2 . These single element diffusion barriers include Ta, 126,130 132 W 127 Ni 133 Cr 134 Ti 135 Nb 136 and Mo 136 . Refractory materials for dif fusion barriers have gained acceptance due to their high temperature stability and low chemical reactivity. Examples of these compounds include TaC 137 WN x 138 HfN x 138 TaSi 130 TaN 131 TiN 139 and TiON 133 . The current state of the art copper diffusion barrier is a Ta/TaN bi layer stack. The tantalum layer is used for adhesion while TaN is the actual diffusion barrier. The
50 advantage of Ta is it is a refractive metal thermodynamically stable at high temperatures and is completely immiscible in Cu up to the meltin g point. In addition the adhesion strength of Ta/TaN to copper is high at 7.5 J m 2 . 140 As semiconductor devices continue to shrink and 3 D interconnects and through silicon vias are incorporated and the copper/liner thickness decreases, the use of Tantalu m based diffusion barriers may not be practical. Tantalum forms a columnar structure as deposited. As the barrier thickness decreases, the Ta/TaN bi layer will not provide adequate protection from copper diffusion. The liner thickness currently used in the latest generation of semiconductor devices is 3.3nm. 141 With the advent of Atomic Layer Deposition (ALD) new low dimensionality diffusion barriers are being investigated. The advantage of ALD over physical vapor deposition or e beam deposition is the depo sited films are formed atomic layer by atomic layer. Complete conformality of films can be obtained as well as coating aspect ratios of 1000:1 or better. In addition chemical compositions of the diffusion barriers can be tailored to adjust materials proper ties by adjusting the precursor chemistry. ALD has been used to create new exotic compounds of low dimensional films(<3nm): MnO x 142 Ta x N y C z 143 Co W 144 V N 145 and Mn Ru 146 as examples. 2.6.1 Carbon Diffusion Barriers The use of carbon based materials for diffusion barriers is not a recent development. In 1989, Chang et al . of IBM studied amorphous carbon as a diffusion barrier to prevent a intermixing between aluminum and Pd 2 Si and aluminum and PtSi 147,148 . In both experiments, a 1000Ã… amorphous carbon la yer was used as a barrier. The experimental results showed no intermixing of the aluminum and the silicide layers up to 500 o C for annealing times of 30 minutes. In other study, Chang explored
51 the use of carbon as a barrier to the outdiffusion of copper 149 151 . In these studies, a 1000Ã… carbon film was sputtered on copper and annealed at 750 o C for six hours. The results showed no copper out diffusion through the carbon layer as analyzed by auger analysis. However, an interesting note in these studies is when the researchers deposited a thin gold layer on top of the carbon with the stacking sequence of Au/C/Cu/SiO 2 there was a significant interaction of copper with the top gold layer. Apparently, the carbon barrier failed in these experiments, however the pres ence of copper within the carbon layer on experiments without the gold overlayer was not detected by the Auger surface analysis. 2.6 .2 Self Assembled Monolayers A s Carbon Based Diffusion Barriers In exploring a low dimensional material like graphene as a d iffusion barrier, there is precedence in carbon based materials of low order that prevent diffusion. Self assembled monolayers (SAM) have been explored as diffusion barriers to copper in the last decade 152 155 , this time as a wet chemistry approach. Self A ssembled Monolayers with various functional chains were investigated to prevent copper from penetrating into a silicon or SiO 2 dielectric layer. The thought process was low dimensional (<5nm) films could be deposited by vapor or wet chemistry methods. Thes e methods would provide the conformality needed for metal steps and via structures. The underlying rationale is to immobilize Cu through strong local interfacial bonding with short chained molecular functional groups and create a barrier between the Cu and SiO 2 or Si. The molecules of the SAM are composed of three parts: the head group, the alkyl chain, and the terminal functional group 152 . Each part can be tailored in order to achieve specific layer surface chemistry, structure, and thickness. In particula r, when selecting the correct SAM molecule for barrier applications, several structural aspects must be taken into account.
52 For instance, the selectivity and adhesion strength between the SAM head group and the substrate impact the film packing density and thermal stability. 152 The SAM chain length influences the film packing density and order. 153 The choice of the SAM terminal group affects the adhesion between the SAM layer and the Cu over layer. Gandhi et al . has demonstrated that SAM films can be used no t only as inhibitors to prevent Cu diffusion in the underlying SiO 2 layer, but also to enhance Cu/SiO 2 interfacial adhesion. 154,155 Examples of SAMs currently being investigated for diffusion barriers include: 3 aminopropyltrimethoxysilane derived self ass embled monolayer , (NH2SAM), 156 alkylsiloxane (SAM) 157 , 3 mercaptopropyltrimethoxysilane (MPTMS), 158,159 2 (diphenylphosphino)ethyltriethoxy silane, 160 3 [2 (trimethoxysilyl) ethyl pyridine 161 . Indeed these very thin films have garnered much success in adhe sion promotion and diffusion barrier properties for copper. In several studies by Krishnamoorthy, 161 and Ganesan et al ., 162 organosilanes(SAMs) increase the Cu diffusion induced device failure time by a factor of four and Cu SiO 2 interfacial adhesion by a factor of three when compared to interfaces without a barrier layer. In addition a 1.7nm 2 (diphenylphosphino)ethyltriethoxy silane SAM is equivalent in performance to 20nm of a Ta film as measured by bias stress testing. All of these experiments were con ducted with a barrier thickness of less than 2nm. 2.7 Rapid Thermal Processing Although there is no formal definition for Rapid Thermal Processing, in the science community the term implies a fast heating of a semiconductor substrate by electromagnetic ra diation or a beam of electrons. In comparison to a general batch or tube furnace, a Rapid Thermal Processor has shorter processing times, fast heating and cooling rates, the substrate is thermally isolated from the processing chamber and
53 is cold walled wit h precise ambient control. There are many subsets of Rapid Thermal Processing which includes Rapid Thermal Annealing (RTA) Rapid Thermal Oxidation (RTO), Rapid Thermal Nitridation (RTN), Rapid Thermal Silicidation (RTS) and Rapid Thermal Chemical Vapor Dep osition(RTCVD). All of these processes are inclusive in the general Rapid Thermal Processing definition. RTP is based on the energy transfer between a radiant heat source and a substrate or object. At elevated temperatures, radiation energy transfer domin ates over convection and conduction heat transfer. For RTP, which utilizes optical energy, the chamber wall is not in thermal equilibrium with the substrate, short process times(minutes) are possible. In contrast to batch furnaces, which convection and co nduction heat flow dominate, the process times for these systems are on the scale of hours. 2 .7.1 The History O f Rapid Thermal Processing The emergence of Rapid Thermal Processing (RTP) for semiconductors can be traced back to Fairfield and Schuttke at IBM in the 1960s 163 , in which a pulsed ruby laser(at 0.6943Âµm) was used to activate boron and phosphorus doped silicon for p n junctions. In 1968 W.K. Mammel of the Western Electric Company, filed Patent halogen lamp based system with six 2KW halogen lamps and wafer rotation 164 . The purpose of this system was phosphorus doping from the gas phase. An interesting concept of the Mammel patent was wafer levitation from precise gas distribution. upon the infrared halogen lamp concept, which would become the industry standard even today. However, very similar halogen lamp based systems had been develope d
54 for non semiconductor applications. In 1957, N.K. Hiester at Caltech demonstrated a solar furnace using a parabolic reflector and a linear array of halogen lamps and had the capability of heating at rates of 1000 o C/sec. 165 In 1961, F.J. White at the Nava l Research Labs demonstrated a halogen based system for the conversion of thermal energy to mechanical energy to produce aerodynamic heating. This system is curiously similar to the RTP systems being sold today. 166 2.7.2 Infrared Halogen Lamp B ased Rapid T hermal Processing In infrared halogen lamp based RTP systems, radiative heating of the substrate is the dominant energy transfer mechanism over conductive and convection heating. For cooling, radiative energy transfer from substrate to the chamber walls is dominant as well. Because the substrate is not in thermal equilibrium with the chamber walls, rapid cool down rates can be achieved. Highly reflective chamber walls with gold or nickel coatings, allow the majority of the halogen lamp energy to be imparted on the substrate with low energy losses to the chamber itself. Infrared halogen lamps consist of a high melting point, tungsten filament (Melting Point=3422 o C) surrounded by a halogen gas. The tungsten filament is activated through resistive heating. The halogen gas, usually consisting of one or more components of the halogen groups (I, Br, F, Cl) is used to prevent the tungsten filament from evaporating and degrading the lifetime of the lamp. As the tungsten filament evaporates, the gaseous tungsten atom s react with the halogen gas to form a tungsten halide gas. When the tungsten halide particle approaches the filament, the halide is cracked and released and the tungsten atom is re deposited on the filament which greatly increases the lifetime of the lamp . The tungsten evaporation, reaction with the halide gas and tungsten deposition is called the Halogen Cycle. 167
55 The fundamental radiative output of the halogen lamps can be determined by d emits all body per unit surface area at a given temperature in Wm 2 K 4 is: 168 (2 20) Where is the emissivity and is the ratio of the emitted radiation from a real surface to that of a perfect black body. A is the emission area, k is the Boltzmann constant, c is the speed of light, h T is temperature. The emissivity is unity for a perfect blackbody. For the tungsten filaments, the emissivity is equal to 0.35. 168 zero, which is not realistic. Fo r non becomes: (2 21) Solving this equation with a T o value of room temperature or 300K yields a temperature T f of 2570K. The spectral irradiance or the distribution of the radiant radiative power per unit area, per unit wavelength: 169 (2 22) The maximum intensity of the Planck radi displacement law. 169 (2 23)
56 For infrared halogen lamps the emission wavelength is from 0.5Âµm(visible) to 3.0Âµm (intermediate IR) with a peak wavelength close to 1.0 Âµm as describe d 2.7.3 Rapid Thermal Processing Chamber Materials When infrared radiation from the halogen lamps impinges on a substrate, nsmit, absorb or reflect the radiation. Various materials have different optical absorption characteristics. The loss in energy of the impinging electromagnetic wave on a substrate can be through scattering, polarization of the molecules, lattice vibratio ns or excitation of electrons from the valence band to the conduction band as well as other mechanisms such as luminescence. 168 For rapid thermal processing, the chamber materials must be highly reflective in order to deliver the maximum energy to the sub strate and minimize absorption to the chamber walls. Generally the chamber walls are made of aluminum or stainless steel and plated with a highly reflective metal. The reflection coefficients of different metals are shown in Figure 2 16 . Although Au has t he highest reflectance in the infrared region 170 , there is a concern with Au contamination, especially if silicon wafers are annealed. Au is a fast diffusant in silicon and also occupies both an interstitial and substitutional sites. In order to reduce the potential contamination from a gold coated chamber, the wafer is isolated by either quartz windows or a quartz process chamber. The quartz must be transparent to the halogen lamp radiation in order to reduce absorption and heating of the quartz. The quart z is synthesized from purified silica sand with a beneficiation process. Semiconductor grade quartz has the properties of very high optical transparency over a wide spectral range, chemical inertness, high
57 temperature deformation resistance, ultra low ther mal expansion coefficient, excellent thermal shock resistance and extremely high purity. The optical properties of the quartz window or chamber obviously must be transparent in the photon frequencies generated by the halogen lamps. The most common quartz i n the semiconductor industry is the 214 grade developed by the General Electric Corporation and now produced by Momentive Performance Materials Inc. 171 The 214 grade quartz has very low levels of hydroxyl (OH ) groups, which are typical of common quartz. T he low levels of OH groups allow the quartz to have a high transmission at wavelengths ~2.71Âµm, which is the absorption band of water and is within the spectrum of the halogen lamp output. 2.7.4 Susceptor Based Rapid Thermal Processing A susceptor is a b asic substrate holder that is fabricated out of graphite. The susceptor generally consists of a base with a cavity that holds a substrate with a lid that covers the assembly. The graphite susceptor is coated with a layer of SiC of approximately 25Âµm and is used to prevent carbon contamination on the substrate and to reduce the porosity of the graphite. The graphite to SiC conversion process is carried out in a temperature range of 1400 1600 o C in a silane atmosphere. The emissivity of graphite approaches a p erfect blackbody with measurements between 0.8 0.95. In semiconductor processing, a susceptor is a must for rapid thermal annealing of compound semiconductors. Generally for most compound semiconductors, the infrared halogen lamp photon energy is less than the bandgap of the substrate and is thus transparent. By utilizing the susceptor, which absorbs mostly all wavelengths, the halogen lamp radiation heats the susceptor, which in turn radiates in mostly all wavelengths and heats the compound semiconductor s ubstrate. The substrate is heated isothermally and thus without thermal gradients. An additional advantage of
58 susceptor based rapid thermal processing in compound semiconductors, is to reduce the Group V component due to outgassing from the substrate. 172 G enerally the Group V atoms (As, P, N, Sb) has a higher vapor pressure than the Group III component. During high temperature anneals, the Group V component will exit the semiconductor surface and leave behind non stoichiometric defects. A method of preventi ng the Group V component from outgassing, is to create an overpressure of the Group V element within the susceptor. An overpressure reduces the evaporation rate of the Group V element. The overpressure can be accomplished by either placing pellets of the G roup V element within the susceptor cavity or by charging the susceptor with the Group V element. Charging the susceptor consists of running multiple processes with expendable wafers similar to the actual device based wafer and sacrificially allowing the G roup V element to outgas and coat the inside of the susceptor. 172 Over time a saturation of the Group V element develops and a vapor pressure equilibrium of the Group V component is established. In later sections we will describe the methodology of using a charged susceptor in the processing of graphene.
59 Figure 2 1. An illustration of the different carbon carbon hybridization structures showing the bond angles of sp1, sp2 and sp3. Note, sp1 and sp2 structures are in a single (x,y) plane, while the sp3 structure is in three dimensions. Figure 2 2. The hexagonal lattice of graphene and its Brillouin zone. A) The lattice structure of graphene indicating the unit cell vectors a1 and a2 in Cartesian Coordinates. B) The corresponding Brillouin Zone illust rating the Dirac points
60 Figure 2 3. The two dimensional band structure of a single atomic layer of graphite as modeled by Painter and Ellis. [reproduced with permission from ref. 33 ] Figure 2 4. An illustration of the linear dispersion of graphene as compared to a single electron dispersion and a quadratic dispersion. A) The linear dispersion of graphene. B) The quadratic dispersion of a free electron. C) The quadratic dispersion of a direct band semiconductor.
61 Figure 2 5. Left: Illustration of AFM technique to measure deflection of graphene. A) A representation of the nanoindentation technique. B) Graphene covered SiO 2 structures for nanoindentation tests. [reproduced with permission fro m ref. 47 ] Figure 2 6. Examples of graphene reconstruction around defects. A) Stone Wales Defect. B) A single vacancy within graphene. C) A graphene double vacancy [reproduced with permission from ref. 67 ]
62 Figure 2 7. A representation of the gas flow dynamics in growing graphene on copper. A) The reaction mechanisms showing hydrocarbon decomposition. B) A diagram that details the graphene growth reaction limitations. The process is either surface reaction limited or mass transport limited. [reproduced with permission from ref. 109 ] Figure 2 8. The evolution of Raman shifts for single(red line), double layer graphene (blue line) and few layer graphene(green line) transferred onto SiO2 [reproduced with permission from ref. 114 ]
63 Figure 2 9. The Nicol et Alamega Ramn Spectrometer used in this research at UT Dallas NSERL lab. The raman has dual wavelength capability and is adaptable for raman measurements in different atmospheres. Photograph courtesy of the author. Figure 2 10. The Ulvac Phi 2300 XPS used in this research at the UT Dallas NSERL lab. Photograph courtesy of the author.
64 Figure 2 11. The Alessi Four Point Probe used in the research in the UT Dallas NSERL cleanroom. Photograph courtesy of the author.
65 Figure 2 12. The use of the fou r point probe method to determine diffusion constants at the temperature of barrier breakdown. A) A plot of annealing time to breakdown at 750 o C. B) A plot of barrier breakdown at a constant time at different temperatures. The slope of the line represents Do the pre exponential factor. Note the change in slope is representative of change from grain boundary diffusion to bulk diffusion through the barrier. [reproduced with permission from ref. 126 ] Figure 2 13. Selective formation of SAM barriers in a Cu dual damascene integration scheme. The SAM B acts as a copper diffusion barrier while promoting adhesion.[reproduced with permission from ref. 156 ]
66 Figure 2 167 ] Figure 2 15. An assembly of the quartz halogen lamps used in this research as provided by Ushio Incorporated. [reproduced with permission from ref. 167 ]
67 Figure 2 16. A comparison of the Reflectance vs. Wavelength for different metals in the infrared region of the spect rum. Gold has the highest reflectivity at the wavelength of halogen lamps. [Reprinted with permission from reference 170 ] Figure 2 17. The optical transmission of 214 grade quartz as compared to quartz with a high percentage of hydroxyl groups. Note the drop off in transmission at the water absorption wavelength of 2.71Âµm. [Reprinted with permission from reference 171 ]
68 Table 2 1 . Raman Tendencies of Graphene. Recreated from reference 122 G band Frequency FWHM Intensity Number of layers Red shif t with increased layers Constant Linear increasing with increased layers Doping concentration Blue shift with both electron and hole doping Narrow Constant Temperature dependence Red shift with increasing temperature Not sensitive Constant Uniaxial strain Blue shift with compressed strain Not sensitive Constant Biaxial strain Blue shift with compressed strain Not sensitive Constant 2D band Frequency FWHM Intensity I2D/IG Number of layers Splitting to some components w ith increased layers Larger for bilayer than single layer Not sensitive Decreasing with increased layers Doping concentration Blue shift with hole doping and red shift with electron doping Not sensitive, becomes narrow Decreasing with increased electr on concentration Decreasing with increased electron concentration Temperature dependence Red shift with increasing temperature Not sensitive Constant Uniaxial strain Blue shift with compressed strain Not sensitive Constant Biaxial strain Blue shift with compressed strain Not sensitive Constant Table 2 2 . Batch Furnace/RTP Comparisons Furnace RTP Batch Single Substrate Hot Wall Cold Wall Long Process Times Short Process Times Small dT/dt Large dT/dt High Cycle Time Low Cycle Time Temp Measurement Environment Temp Measurement Substrate
69 CHAPTER 3 DESIGN AND FABRICATION OF A RAPID THERMAL GRAPHENE GROWTH SYSTEM AND SYNTHESIS TECHNIQUES 3.1 Introduction In this chapter we discuss the design and build of a haloge n lamp based rapid thermal system and the advantages and limitations in reference to graphene growth. A through explanation of the challenges of building a custom designed rapid thermal graphene growth system will be discussed. Graphene process optimizati on and experimental results will be presented that demonstrate the successful growth of single layer graphene on copper foils as well as an attempt at growing graphene directly on SiO 2 . 3.2 Design Considerations Of A Graphene Growth System The design of a rapid thermal graphene growth system is not trivial. After all, the majority of all graphene growth has been synthesized in tube based systems with inductive coils at the heating source. As detailed in section 2.5, the advantages of a halogen lamp based ra pid thermal system as compared to a tube furnace approach is fast heating and cooling rates, the ability to alter temperature uniformity in situ, instantaneous gas chemistry changes and low process switching times. However, rapid thermal processing systems are not noted for long process times at high temperatures. Initially, when considering the project of the graphene growth system and after thorough literature searches, several important features must be incorporated in the design. It was imperative that a high vacuum based on turbo pumping was a necessity. There are several references that detail the graphene growth quality is directly dependent on the cleanliness of the chamber and the minimization of oxygen contaminants. The second important feature is chamber cooling. Unlike conventional
70 RTA systems, which operate in time scales of seconds to minutes, the graphene growth chamber would be operated for time scales of hours. The chamber cooling design would have to warrant enough cooling capacity to operat e for long process times, without overheating or damaging chamber components. With these two criteria in mind, we decided on a clamshell type chamber in which two separate components would mesh to form the single chamber. The first component is the chamber lid, in this component are housed the halogen lamps, gold reflectors, quartz window and high vacuum electrical feedthroughs. In the bottom section, which we will designate as the process chamber, is the gas inlets, the thermocouple inlets and pumping port s. It is also the location in which the process substrate rests and the graphene growth occurs. 3.2.1 A Unique Dual Chamber Design window which isolates the halogen lamps, in our case, from the process chemistries. The quartz window itself proved to be a source of great deal of work and in later sections will prove to be the source of chamber failures. In Section 2.5 it was described in detail the reasons for separating gold reflec tors from the process substrate. It was also a safety imperative to isolate the halogen lamps and high voltage power delivery from process gases. Since the graphene growth is in chemistries of 100% H 2 and methane, any chance of a spark or arc may have been disastrous. The quartz window accomplishes both tasks of isolating gold reflectors from the substrate and also isolating hazardous gases from high voltage. Although the quartz used in the graphene growth system is >90% transparent in the infrared frequen cies, the photon absorption increases as a function of thickness. This is due to small defects within the glass that act as scattering centers. In addition
71 the ambient in between the halogen lamps and the quartz window also absorbs photon energy. Therefor e, the approach to maximize the optical throughput is by using a thin quartz window and having the top chamber in a low vacuum environment. The quartz thickness chosen was a standard 3mm. By calculations of the flexure strength of 214 quartz, the 3mm thick ness would allow a 10 Torr differential between the vacuum levels of the top chamber and the bottom process chamber. High low interlock switches were wired into vacuum gauges such that if a pressure differential between the top and bottom chambers were >5 Torr, a solenoid valve connected to both chambers would open. This would allow equalization of pressures between both top and bottom chambers. For vacuum integrity, both chambers were machined from a single block out of 6061 T6 aluminum. There were no wel ds within the system. The graphene growth system utilizes 21, 1200 Watt halogen lamps for a total power output of 25.2KW. The lamp connectors were initially machined out of oxygen free copper. The halogen lamps would screw into the copper blocks for power distribution. Four high voltage, 90 amp vacuum feedthroughs, supplied by Lesker tied the copper blocks inside the top vacuum chamber to connector wires outside the chamber. A MKS 722A pressure transducer was used to measure the vacuum level inside the top chamber. The vacuum readout was with an Instrutech B RAX controller. The B RAX controller has relays that can be manually programmed for high low settings and activate once a programmed vacuum level is reached. For high reflectivity, the top chamber was e lectroless nickel plated with a thickness of 2.5Âµm. The nickel coating is an adhesion layer between gold and aluminum. The nickel coating is followed by a 1.27Âµm plating of 99% pure gold.
72 The bottom or process chamber is the location where the graphene gro wth process occurs. This chamber has a port for a Pfeiffer HiPace 80 turbo pump as well as feedthroughs for multiple thermocouples as well as the process gas lines. Gases enter the chamber from the left hand side and are pumped out the right hand side eith er through the turbo pump or an Edwards 8 oil based pump with a cold trap. The turbo pump is backed by an Adixem diaphragm based, dry vacuum pump model AMD 4C. The ultimate base pressure of the bottom vacuum chamber is measured by a cold cathode gauge conn ected into a MKS 943 cold cathode vacuum controller. During processing, the chamber pressure was measured by a baratron transducer connected to a MKS PDR 2000 dual capacitance manometer controller. The PDR 2000 also has high low relays which activate once a certain vacuum level is reached. The bottom chamber utilized double o rings with slots that allowed differentially pumping. The differential pumping essentially allows the chamber to have an interface to a reduced pressure of 10mTorr rather than an inte rface to atmospheric pressure. The differential pumping was through the Adixem turbo backing pump. The unique two chamber design of the graphene growth system provided a challenge in order to obtain a correct pumping sequence. As stated earlier, if there w as a 10 Torr differential between either the top or bottom chambers, the quartz window was susceptible to breaking. This was solved by using the turbo backing pump to evacuate both chambers simultaneously from atmosphere to a rough vacuum. The pumping sequ ence utilized is as follows. The common valve for both chambers was opened which allowed pumping to vacuum levels of 10mTorr, this was followed by opening the turbo pump gate valve. The turbo would be engaged and spin up to 1500 Hz to reach a base pressure of <10 6
73 Torr. The turbo pump gate valve was closed and hydrogen was flowed to reach a pressure of 10mTorr. The process pump valve was opened and gas flows were adjusted to reach the desired pressure levels. To vent the system, all valves were closed and the common valve to the top and bottom chamber was opened. Nitrogen was flowed continuously until atmospheric pressure was obtained. Temperature measurements and control of the graphene growth system were with an Omega XL Type K thermocouple with a stainl ess steel sheath. The accuracy of the thermocouple is within 2 o C from 500 1300 o C with a maximum temperature capability of 1335 o C. The thermocouple was vacuum sealed with a custom modified Swagelok ultra torr fitting. The gas delivery system was a custom b uilt stainless steel manifold connected to four separate MKS M100B mass flow controllers calibrated for H 2 , CH 4 , N 2 and O 2 . Isolation valves were fast response time, Swagelok ALD 6 diaphragm valves. The process gases used in the graphene growth system were Ultra High purity H 2 and CH 4 supplied by Airgas with 99.999% purity. The nitrogen, which is used for venting both chambers, was the UTD NSERL house nitrogen, also supplied by Airgas. One of the critical design issues for the rapid thermal graphene growth system is the chamber cooling. Since both top and bottom chambers would be expected to remain cool for temperatures > 1000 o C for hours, a unique approach was necessary. The most other words, we would try to obtain the maximum amount of surface area of both top and bottom chambers exposed to cooling water. Both top and bottom chambers were milled with large cooling passages approximately one inch in width. The water was
74 dir ected from inlet to outlet ports by small tracks approximately 5mm in width. The wall of water approach allowed the top and bottom chambers to have approximately 96% of the total chamber area exposed to cooling water. Initially a Neslab HX300 chiller was used for chamber cooling. The Neslab HX300 has an operating range of 5 35 o C, a cooling capacity of 5.0 KW and a flow rate of 3.3 gallons/min. During initial tests, prior to gold plating of the top chamber, the HX300 chiller would overheat after 60 minutes during process temperatures of 1035 o C. It was determined that a larger flow rate was necessary in order to cool the chamber consistently for longer process runs. A Neslab Thermoflex 10,000, which has 10KW cooling capacity and a flow rate of 10 gallons/min. , was subsequently connected. This chiller provided the necessary cooling capacity necessary for unlimited process times at temperatures of 1035 o C and above. During the ramp up stage of the process, in which the maximum amount of energy is supplied by the halogen lamps, the water temperature would rise from a setpoint of 19 o C to approximately 25 o C. Once at the oven setpoint temperature, the water temperature would decrease to 21 o C and stay at that level indefinitely. 3.2.2 Materials Science I n Chamber Desi gns The graphene growth optimization was hindered by an unforeseen issue with the halogen lamp, copper connector blocks. The original thought of using a highly conductive copper connectors was valid, however what we did not anticipate is a copper reaction at moderate vacuum levels and high temperatures. A symptom that continuously occurred while trying to grow graphene, was the 90 amp protection fuses on one or two phases of the three phase power distribution to the halogen lamps kept blowing. The blown fus es were not consistently on one particular phase. While we believed there was a short that was present in the wiring circuitry, no known problem
75 could be found. The problem would not occur with both chambers at atmospheric pressure. After more troubleshoot ing, it was realized that on the chamber walls there was burn marks next the copper connectors. We suspected that an arc from the copper conductor blocks to the grounded chamber walls was the source of blown circuit breakers. However, intuitively, since th e dielectric constant of air at atmospheric pressure and of a vacuum are approximately the same, it was puzzling why the arc would occur under a moderate vacuum, but not at atmospheric pressure. Ultimately, after further troubleshooting, it was determined that the copper connectors were receiving reflective lamp radiation and heating. The combined effect of an increased vapor pressure from being in vacuum combined with an elevated temperature resulted in a conductive pathway for arcing to occur. In trying t o ascertain the thermal energy and thus how hot the copper connectors would have to be to create a pathway, we used the Clausius Clapeyon equation which relates vapor pressure to temperature. For copper at temperatures of 727 o C, 857 o C and 900 o C, the vapor pressure of copper is 10 8 , 10 6 and 10 4 respectively. It is apparent that at 900 o C and higher, the copper vapor is significant, especially since the top chamber is usually in the low mTorr range. The next step to alleviate the arcing issue, was to scrap the two chamber approach. The top chamber would remain at atmospheric pressure, while the process chamber would still be capable of high vacuum. In order to accomplish this goal, we would need to recalculate the quartz thickness in order to withstand a pr essure differential of atmosphere to high vacuum(14.7 psi). The calculations resulted in a quartz thickness of 12.5mm. After installing the thicker quartz, tests were run to
76 determine if the graphene growth system would reach the desired temperatures of >1 000 o C, which the system was indeed capable. However, the extra thickness of quartz resulted in slower maximum ramp rates (50 o C/sec vs 120 o C/Sec) and a much slower ramp down rate. The additional quartz thickness acted as a secondary heat source which limite d the cool down rates from 600 o C to 200 o C to <1 o C/sec. After much hesitation of having to redesign the graphene growth system in order to solve the copper evaporation issue, we moved forward with trying to develop a process for growing graphite. This was short lived as we were faced with an additional problem, once again with the copper connectors. Although, we had resolved the arcing problem by having the copper connectors and the top chamber at atmospheric pressure, we were confronted with a problem of t he copper connectors oxidizing and developing a low conductivity black scale on the surface. Once again, the copper connectors were at a high temperature resulting from reflected lamp radiation. In an air atmosphere, the copper surface would essentially co rrode. During each process run, the amperage being delivered to the lamps would stochastically increase with process time. We attempted to etch the surface in ammonium persulfate (Transcene copper etch) but this did not treat the underlying issue, as the s caling continued. Again, we had to make a decision to redesign a component of the graphene growth system. We decided to replace the copper blocks with machined, 316 stainless steel. The 316 stainless has a ~22% higher melt temperature than copper (1400 o C v s 1083 o C) and a better high temperature oxidation resistance. However, the conductivity of 316 stainless is an order of magnitude less than copper (6.21x10 6 vs 5.98x10 7 S/m), which apparently did not cause any issues. Ultimately, by transitioning to stainl ess conductors and by
77 eliminating the top chamber vacuum, we were able to proceed with graphene growth, without any other failures (on this component). With the graphene growth process optimization back on track, we were able to successfully test several p rocess conditions that enabled us to create single layer graphene on copper. However, with every step forward, it seems there is a step backwards. In this instance, we did not foresee another component failing. After each graphene growth process, it was no ticed that the chamber base pressure would increase. In addition, we noticed that we would have one successful growth of graphene followed by a second process where we see little to no graphene on the copper surface. Once again, we would have to troublesho ot the reasons for this inconsistency. It was found that after each process run, which is detailed later, the Viton o rings that seal the quartz window were substantially degrading. By switching to a thicker quartz window, which is 4X the thickness of the original quartz window, we found the quartz was significantly hotter than previously. Even though the o rings are not in a direct path of the lamp based radiation, the quartz itself was heating to a point where the temperature was above the glass transitio n temperature of the Viton o rings. At this point, we tried to implement a gas cooling loop which would blow CDA over the quartz window. This technique did help, however, ultimately we could not supply enough cooling gas to prevent the o ring degradation. In our research, we are able to grow graphene successfully, however with each process run the chamber needs disassemble and the o rings have to be replaced. 3.3 Graphene Synthesis In A Rapid Thermal Graphene Growth System The initial process conditions fo r graphene growth on copper were based upon a recipe published by the UT Austin, Ruoff group. 173 The sequence of general process
78 steps is 1) Ramp to 1035 o C in 100% H 2 at 2 sccm. 2) Hold at 1035 o C for 20 minutes in 2 sccm 100%H 2 to reduce surface copper oxi des. 3) Change chamber chemistry to 1 sccm CH 4 /2sccm H 2 for 150 seconds for graphene nucleation. 4) Grow graphene to full coverage in 35 sccm CH 4 /2sccm H 2 for 180 seconds. We repeated this process in the rapid thermal graphene growth system. Our samples co (Alfa Aeser 99.8%) that were precleaned in 10% HNO 3 for three minutes. After several DI water rinsed, the samples were blow dried in N 2 and placed in the graphene growth chamber. The system was pumped to 10 6 Torr and back f lowed with 10sccm H 2 . The chamber pressure at this flow rate was 360mTorr. The sample was heated at 5 o C/sec to 1035 o C. For reduction of copper oxides, the sample was held for 20 minutes. Methane at a rate of 5 sccm was introduced to nucleate graphene for two minutes. This was followed by an increase in methane flow to 35 sccm for 30 minutes. The pressure was stabilized at 0.5 Torr. The lamps were turned off and the sample was allowed to cool as quickly as possible. Initial results showed an incomplete surf ace coverage of graphene with very small islands of graphene a few millimeters in diameter. It was determined the nucleation process was successful, but the growth process was incomplete. The recipe was modified to keep the same condition except increase t he growth time to 60 minutes from 30 minutes. This experiment resulted in an increase in substrate coverage to 50% of the total. Since the graphene growth process is self limiting, it was decided in order to maximize the process times to ensure graphene gr owth would be complete. We altered the methane gas flows to 50 sccm at a pressure of 625 mTorr and increased the growth times to one hour. In these tests, we achieved complete copper coverage of graphene.
79 3.4 An Attempt Of Graphene Synthesis W ithout Trans fer Once graphene was consistently grown on copper, we shifted our attention to growing graphene directly on SiO 2 . In a paper by Kim et al., 174 it was demonstrated that an SiO 2 wafer in close proximity to a copper substrate, in an atmosphere of hydrocarbons would result in graphene growth directly on the SiO 2 . The theory behind this mechanism is at high temperature and low pressures, a copper vapor would form and act as a gaseous catalyst to graphene growth. In the Kim experiments, the distance from the copp er substrate to the SiO 2 was 50Âµm. The process conditions for these experiments was a growth temperature of 1000 o C in flows of 30 sccm of CH 4 and 20 sccm of H 2 at a total pressure of 5 Torr, for 30 minutes. Kim verified, through Raman analysis, the graphene on SiO 2 was single layer. It was also verified, through XPS, that copper was not incorporated in the graphene film and acted simply as a growth catalyst. Our method of transferless , graphene growth on SiO 2 , was to use a susceptor based approach. A 100 mm graphite susceptor, supplied by GMSI, Phoenix, Arizona was sputter coated with copper in a Lesker Multi source RF and DC sputter system at NRF. The deposited copper thickness was 50 00Ã…. The copper coated graphite susceptor should provide a copper over pressure within the susceptor that may act as the catalyst for graphene growth. The susceptor was loaded into the graphene growth system, followed by placing a clean 100mm, 300nm SiO 2 /S i wafer on top of the susceptor which acted as the lid and growth substrate. The susceptor has eight gas slots that allow pumping of the inside of the susceptor and allows gas flow through the susceptor. The standard graphene growth recipe, described earli er, was used to grow graphene. The results showed a distinct discoloration of the SiO 2 substrate, which we believed to be graphene. However, these results were not verified by Raman. The
80 Raman data showed no G,D or 2D peaks. We believe the deposit was carb on, however it was not structurally graphene. The close proximity of the Kim experiments may explain why graphene growth was successful, as compared to our results. The distance from our SiO 2 sample to the copper substrate is 0.8mm, while the Kim samples w ere only at a distance of 50Âµm. 3.5 Concluding Remarks In this section we demonstrated the reasoning and challenges of a rapid thermal graphene growth system. We were able to grow successfully single layer graphene on copper substrates. Although valuable data on design aspects has been generated, looking ahead, a complete chamber redesign is necessary to continue with optimizing the growth process. The new design will have all electronics outside of the heating zone. In addition, shielding from the halogen lamps of the o rings as well as extended CDA cooling will prevent the degradation problems we have experienced. We expect to continue the concept of growing graphene on dielectric substrates without using a transfer process.
81 Figure 3 1. A picture of the clamshell approach to the graphene growth system during fabrication. A). The top chamber houses the halogen lamps and high voltage resting on a quartz wafer holder. Photograph courte sy of the author. Figure 3 2. The design of the wall of water approach of water cooling in the Rapid Thermal Graphene Growth System. Note, the top cover is removed in this drawing. . Photograph courtesy of the author.
82 Figure 3 3. The completed design and build of the Rapid Thermal Graphene Growth System used in this research with panels removed. . Photograph courtesy of the author.
83 Figure 3 4. The final design of the graphene growth system as a mockup designed in Solidworks
84 Figure 3 5. The grap hene growth chamber and copper conductors. A,B. The location of arcing to the grounded chamber walls. C),D) Corroded copper conductors after a 1035 o C graphene growth process. Photograph courtesy of the author.
85 Figure 3 6 . The graphene growth chamber sho wing the location of the failed o ring underneath the quartz window. Photograph courtesy of the author. Figure 3 7. Raman Spectroscopy of Graphene on Copper grown in the rapid thermal graphene growth system measured at multiple locations. Note the large 2D to G peak ratio indicating single layer graphene.
86 Figure 3 8. The process run of successfully growing graphene on copper. Note the dips in temperature (red line) and lamp power (green line)indicate a change in gas flows. Figure 3 9. Growing graphen e directly on SiO 2 . A) The copper coated graphite susceptor used in growing graphene in a copper overpressure. B) An illustration of the copper catalytic mechanism of growing graphene on SiO 2 . [reproduced with permission from ref. 174 ]
87 Figure 3 10 . Raman spectroscopy of an attempt at growing graphene on SiO 2 . One small peak appears at approximately 2325 cm 1 which is not related to any graphene peaks. Inset, the SiO2 with what appears to be a carbon deposition..
88 CHAPTER 4 TESTING OF GRAPHENE AS A SOLID STA TE DIFFUSION BARRIER 4.1 Introduction Solid state diffusion barriers generally require several key properties to be effective. These properties include thermodynamic stability between layers, strong adhesion, low contact resistance and resistance to mechan ical and thermal stresses. In addition, for semiconductor applications, the barrier should be highly thermally and electrically conducting . Diffusion barriers in general should either be amorphous or have large grain sizes in order to reduce the densitie s of fast diffusion paths along grain boundaries and defects. The interfacial properties of graphene in contact with a diffusing species as well as the interfacial contact with the bulk substrate are the key parameter in determining whether a single layer of graphene can prevent diffusion across its interface. In this section we will discuss the properties of bonding of different metals to graphene and the effects on the properties of graphene. We will validate the choices of obtaining high quality graphen e with an analysis of three commercial vendors. Furthermore, we present the successful application of graphene as a solid state diffusion barrier. 4.2 Physisorption And Chemisorption: Bonding W ith Graphene Many studies of electrical contacts to graphene de monstrate weak or strong bonding to the graphene surface. Previous investigations of graphene contacts with different metallic systems show a distinct separation between metallic groups that strongly bond to graphene. The transition metals that have a dist inct solid solubility of carbon, thus are capable of forming carbides through covalency. These are preferable
89 candidates for graphene in the use as a diffusion barrier. Alkaline metals which form weak ionic bonds to graphene, cause little distortion to the graphene lattice, however the cohesiveness of these metals reduce their viability as barriers. 175 Previous studies of common contact metals on graphene, distinguish whether the metal/graphene interface is through physisorption or chemisorption. 176 179 . Th e metals that physisorb with very weak bond formation (0.03 0.05 eV) include Ag, Al, Cu, Cd, Ir, Pt, and Au while the chemisorb group (0.09 0.4 eV) includes Ni, Co, Ru, Pd, and Ti. For physisorption metals, the adatoms tend to either p type dope graphene for Au and Pt or n type dope graphene for Al, Ag, Cu. 176 . However in all physisorbed metals, the graphene electronic band structure with a linear crossing at the Dirac point remains intact. The chemisorption process includes breaking of the C C bonds of gr aphene and forming hybridized states in the graphene band structure, leading to the disappearance of Dirac points when carbidization occurs. 176,178 4.3 A Comparison O f Commercial Grade Graphene Prior to experimenting with graphene as a diffusion barrier we evaluated three commercial vendors for the quality of their graphene. These three vendors include the Graphene Supermarket, Graphene Square Inc. and ACS Materials. The criteria for determining the highest quality graphene and the highest success in our use of graphene as a barrier material, we performed a microscope analysis as well as Raman Spectroscopy. The sample we obtained from Graphene Square was CVD graphene on copper. Prior to performing any transfer of the graphene, it was immediately clear that the sample provided had copper oxidation. In performing Raman analysis, we had to search for areas that had a graphene signature. We ruled out this vendor immediately. The next set of samples was from the Graphene Supermarket. These samples had
90 single lay er graphene transferred onto a 300nm SiO 2 /Si substrate. A microscope analysis showed incomplete copper etching in many places prior to the transfer. Additionally, several of the samples had large particulates on top or underneath the graphene. Also the sam ples appeared to have scratches within the graphene. Raman analysis showed the standard characteristic G band at ~1580 cm 1 , the 2D band peak at ~2700 cm 1 and a relatively large D band or defect peak at 1350cm 1 . The large defect peak was disconcerting, y et the underetched copper was the determining factor in not using this vendor. Afterall, copper is one of the metals that was proposed to be studied in the diffusion barrier experiments described later. The third vendor that was evaluated and ultimately ch osen was ACS Materials. The ACS samples were transferred single layer graphene on silicon and 300nm SiO 2 /Si. The provided samples were clean, flat and had a minimum amount of defects except at the graphene edges. Raman analysis showed the standard G band, 2D band and a small D band peak. The grain size of these samples was approximately 10Âµm in diameter. 4.4 Graphene Barrier Testing O f Chemisorbed Metals In this study we investigate the high temperature properties of graphene as a barrier against a chemiso rbed atomic diffusion. In order to ascertain if graphene is a stable diffusion barrier for chemisorbed metals, we chose titanium as the metal. reaction mechanisms with Si. The TiSi 2 formation energy is ~2 eV, and the silicide formation rate is proportional to (time) 1/2 180 . Sili con is the diffusing species to the growing TiSi 2 interface. Single layer and double layer graphene samples on Si and SiO 2 were obtained from ACS Materials, Medford, MA 02155. The samples were 10mm square with either
91 300nm thermal SiO 2 or (100) silicon as the substrate. An approximately 8mm square graphene layer was transferred such that the peripheral areas were uncovered by graphene. The double layer graphene samples were transferred twice and obviously did not have the AB stacking sequence of graphite. All samples were analyzed by a Nicolet Almega XR Raman spectrometer measured at 514nm for verification of single layer and double layer graphene. All samples showed the characteristic G band at ~1580 cm 1 , the 2D band peak at ~2700 cm 1 and a relatively sm all D band or defect peak at 1350 cm 1 , as shown in Figure 4 5 . The ratio of the G /2D peak indicates single layer graphene or double layer graphene. Optical inspection also verified the graphene quality for wrinkles, rips or tears and contamination. A PV D Products e beam deposition system with a base pressure of 10 7 Torr was used for metal depositions. This was followed by annealing in Nitrogen in a SSI Solaris 150 rapid thermal processing system in 100 degree increments from 200 900 o C for 10 minutes. An Alessi four point probe was utilized to determine sheet resistance after annealing. A Dektak 150 profilometer was used to determine step heights after etch and STEMs were taken to verify graphene barrier performance. In order to obtain a baseline for shee t resistivity, we annealed control samples of 100nm Ti/Si at 100 degree increments from 200 o C 900 o C for 10 minutes, which is well beyond the usual thermal annealing time for Ti silicide formation. The sheet resistance of a 100nm Ti/Si stack was measured at each incremental step. As predicted in the literature 190 and as shown in Figure 4 6, the sheet resistance increased from initial values of 6 8 ohms sq to a value of 16 ohms sq at 400 o C, which signifies the formation of the higher resistance metastable C49 base centered orthorhombic structure of TiSi 2 .
92 After a further increase in temperature, the sheet resistance dropped significantly to a value of ~0.8 ohms sq which demonstrates the low resistivity C54 face centered orthorhombic lattice structure of TiSi 2 has been formed. The C49 C54 transition begins at ~650 o C 191 . For graphene barrier testing, we repeated the annealing steps starting at 600 o C and measured sheet resistivity of the 100nm Ti/graphene/Si stack. An evaluation of the color change showed a distin ct discoloring of the Ti/graphene/Si samples on the periphery where graphene was not present. These peripheral non graphene locations turned to a brownish color, indicating a reaction with the titanium layer and the silicon substrate similar to the Ti/Si s amples. The areas with graphene showed no such discoloration. Sheet resistance measurements of the SLG and DLG samples showed that at 600 o C the sheet resistance was approximately 16 19 ohms sq, while the control Ti/Si sheet resistance dropped to the lower resistance value of 3.8 ohms sq. At 700 o C, the sheet resistance of the SLG samples decreased to 0.9 ohms sq, which is equivalent to the Ti/Si control samples. Further increase in temperature showed the same low sheet resistance values as the control sample s. For the DLG samples, the sheet resistance continued to decrease at increasing temperatures up to 900 o C. However these samples never reached the low resistivity of the Ti/Si control samples. An analysis of the data indicates the DLG samples have approxim ately the same trend as the Ti/Si control samples, except this value is shifted by a few hundred degrees. In order to determine the amount of titanium that was consumed in the annealing process, we etched the unreacted titanium in a solution of 5:1:1 H 2 O:H 2 O 2 :NH 4 OH, which has a high selectivity of Ti to TiSi, for 12 minutes and measured the step height across areas with and without graphene. For anneals at 600 o C, both the Ti/SLG/Si and Ti/DLG/Si showed
93 a 100nm step, which indicates the full thickness of tit anium was unreacted. The control Ti/Si sample showed a zero step height which indicates 100% of the titanium had formed a silicide. Similar results were obtained for Ti/DLG/Si at 700 o C however for Ti/SLG/Si the step height was not measurable, indicating th at the full 100nm had reacted. For TiSi 2 the thickness of the resulting silicide per nm thickness of deposited metal is ~2.5 nm. At 800 o C, measurements show roughly 60% of the titanium in the Ti/DLG/Si has been reacted. Cross sectional TEMs were taken to verify the four point probe and profilometer data on the 600 o C annealed samples. See Figure 4 8. It is clear that the titanium on double layer graphene samples showed a sharp interface after 10 minutes anneals. The titanium on silicon sample distinctly sho ws the silicide growth front with most if not all the titanium having been silicided. C. Gong et al. showed through angle resolved XPS, that titanium spontaneously forms a carbide with graphene during deposition at room temperature. 192 The silicide thickn ess in Figure 4 8 shows an approximate 2X the thickness of the unreacted titanium on graphene indicating the Ti/graphene reaction layer(TiC), formed during deposition, prevents the diffusion of silicon to the titanium growth interface, thereby stopping the silicide reaction at lower temperatures and inhibiting the silicide formation at higher temperatures. 4.5 Graphene Barrier Testing O f Physisorbed Metals In order to determine the effectiveness of graphene as a barrier to copper diffusion, single layer a nd double layer graphene samples on Si and SiO 2 were obtained from ACS Materials, Medford, MA 02155. The substrate size and quality of the ACS graphene was discussed earlier. 100nm of copper was e beam deposited on transferred single layer and double layer graphene on 300nm SiO 2 . For control samples, 100nm of
94 copper was deposited on SiO 2 without graphene. The samples were annealed in N 2 for temperatures from 600 825 o C. At 600 o C, it is apparent the copper has agglomerated on the graphene areas of all samples . At 825 o C, the copper is non existent on the graphene regions. This result was not unexpected since copper is physisorbed on graphene and carbon has an extremely low solid solubility in copper. This is consistent with previous work that demonstrated coppe r on highly ordered pyrolytic graphite, diffuses in clusters laterally rather than individual atoms 193 . The activation energy for cluster movement is less than the energy required to separate an individual atom from the cluster. Nevertheless, in order to d etermine if graphene prevented copper from diffusion into the underlying SiO 2 layer, the copper was etched in persulfuric acid (Transcene Copper Etch) for 10 minutes. XPS analysis was utilized to determine if copper had penetrated the graphene barrier for samples annealed at 825 o C. The results for the Cu2p3 binding energy in Figure 4 11, show a slight Cu peak on control samples of Cu/SiO 2 . The Cu/G/SiO 2 sample showed no trace of copper after etching. Raman data was utilized to determine if the graphene had survived the annealing after copper etching. For samples up to 825 o C, the graphene was still present, albeit in a defective state. As shown in figure 4 10, graphene is still present but appears to have holes in the film. 4.6 Diffusion Barrier Testing For NiAu Contacts O n GaN 4.6.1 Challenges I n GaN Processing The III V nitride semiconductors have wide bandgaps and therefore high breakdown voltages, good thermal conductivity and chemical stability. GaN, the most common nitride semiconductor, is an ideal mat erial for high temperature, high frequency and high power devices. 194 In addition GaN is used in optoelectronic
95 applications for light emitting diodes (LED) and laser diodes. Despite the desirable materials properties of GaN, there are still challenges tha t remain to be solved, specifically p type doping of GaN and stable Ohmic contacts. The difficulty of doping GaN is due to the fact that the only effective p type dopant is magnesium. 195 In order to achieve reasonable p type conductivity, Mg doping levels have to be relatively large, on the order of 10 19 cm 3 . 196 This is due to the fact that only one percent of all magnesium atoms are ionized. The high doping levels leads to crystal quality degradation through stacking faults, inclusions and dislocation f ormations. 197 The second obstacle for GaN processing is stable, low resistivity Ohmic contacts. For low hole concentration materials like P GaN, high work function metals such as Ni 198 , Pd 199 , Cr 200 and Pt 201 with an overlayer of Au are required for low re sistivity contacts. Ni Au contacts tend to be transparent on GaN which is advantageous for optoelectronic applications. 202 204 However, these contacts are characterized by a high specific contact resistance (10 3 2 ). Low resistance Ni Au contacts were discovered by Ho in 1999. 205 By annealing the Ni Au contacts in an oxygen environment, a low resistance contact on the order of 10 4 2 was obtained. Since Au contacts have been well studied. It has been recognized that during annealing inter diffusion between the Au layer and the Ni layer takes place with or without an oxygen environment. 206 211 . In general materials, the inter diffusion between metals would not be an issue, however for Ni Au structures on p GaN, a separation of layers occurs. Essentially, the Ni and Au layers systematically trade locations. The Au layer becomes the interface to p GaN and Ni becomes a high resistance incoherent top layer once oxidized. Nickel is necessary in the Ni Au bilayer
96 since Au depositions directly on GaN cause a non uniform incongruent layer. The Ni acts as a glue layer. The inter diffusion mechanism, described by Ponce, 206 is due to two effects. Ni Au has a miscibility gap in the phase diagram for all compositions below 810 o C. The second reason is due to the electronegativity difference of Au, Ni and Ga which are 2.4, 1.8 and 1.6 respectively. The largest difference in electronegativity is between Au and Ga, permitting Au to preferentially attach to Ga. 4.6.2 Graphene a s a p revention t o Ni Au Intermixing In this section we report on the use of graphene as a preventer of the Ni Au inter diffusion mechanism. Previously, we reported the lack of adhesion between gold and graphene. However, in some instances, graphene sheathed wi th different metals will actually increase the bonding energy of both interfaces. Gong et al. showed an Au/graphene/Ag stack increased the binding energy by 26%. 212 A Pt/graphene/Cu stack had a binding energy increase of 60% for both films. For this testin g, the starting substrate was 50mm Mg doped GaN on sapphire. The GaN epitaxial thickness is 1.0 Âµm and the Mg doping is 5x10 18 cm 3 . Prior to depositions, the GaN substrate was cleaned with acetone followed by IPA and a DI water rinse. 20nm of Ni followed by 200nm of Au were deposited with the PVD products e beam system in the NRF cleanroom. The films were deposited without breaking vacuum. The samples were annealed in a Solaris 150 rapid thermal processing system at temperatures from 500 700 o C for one min ute in an N 2 atmosphere. Sheet resistance measurements were performed with an Alessi four point probe, on GaN samples and also on Ni Au on SiO 2 , to determine if a change in sheet resistance would indicate whether the Ni Au layers mechanism occurs with a di electic substrate. XPS was
97 performed to look for Ni on the surface samples. A Ni peak would imply that the Ni Au layers had swapped positions. Samples with graphene were prepared following the same technique as mentioned above. The p GaN on sapphire was cl eaned in acetone and IPA and rinsed in DI water. A 20nm Ni layer initial layer was e beam deposited. The graphene used in these experiments was Trivial Transfer Graphene TM obtained from ACS Materials. The Trivial Transfer graphene is a simple easy to use P MMA/graphene stack on water soluble tape. In order to transfer this graphene to a host substrate, the sample is immersed in DI water for a few seconds. The water soluble tape releases the PMMA/graphene and is left floating on the water surface. With tweez ers, the graphene is maneuvered to the center of the Ni coated p GaN sample. The sample is carefully blow dried with N2, ensuring no water is under the graphene. The graphene is then air baked at 200 o C for five minutes on a hot plate. In order to remove th e PMMA, the sample was placed in an acetone bath for one hour. This was followed by a DI rinse and N 2 blow dry. In order to ensure the PMMA was completely removed the sample was baked in the PVD products e beam system at 250 o C at <10 6 Torr for three hours and allowed to cool. Without breaking vacuum, 200nm of Au was e beam deposited. Post anneal analysis of Ni/Au on SiO 2, visually show a coherent , planar film which appear to be fully intact. For Ni/Au on p GaN, the films appear less bright and are dull in appearance. The benefit of using a sapphire, transparent substrate in our experiments is backside microscope analysis of the frontside films obtainable. It is clear in visually analyzing the p GaN from a top down approach that the surface appears more gr ay in color as annealing temperatures are increased. It is also clear from visually analyzing
98 the wafer backside that the bottom metal layer is becoming more bright and yellow in color. Obviously, this is indicating the Ni Au intermixing is occurring. We u sed XPS analysis to verify these results. For samples annealed at 600 o C and above, a Ni2p3 peak appears at 852 eV on all samples without graphene. For a 600 o C anneal with graphene XPS did not show any trace of Ni on the substrate surfaces. Four point prob e results are virtually the same for anneals up to 700 o C for both Ni/Au on p GaN and Ni/Au on SiO 2 . Since there is minimum reaction with both substrates and the fact the Au layer is 10X the thickness of the Ni layer, we can conclude that four point probe m easurements are not an accurate indicator of the intermixing process. Top down SEM analysis show surface defects that appear to be large and round. Interestingly, these defects are not in the areas where graphene is present. At first glance, it would appe ar the defects were an incongruent Ni film on the top surface. However, XPS analysis on a 500 o C annealed sample showed no presence of Nickel. The defects are most likely residues from the graphene transfer process. STEM (Evans Analytical Group) analysis wa s performed on samples with and without graphene annealed at 600 o C. The analysis shows for areas without graphene, a continuous layer of Ni and Au is the interface to p GaN. There is no differentiation between Ni and Au films. This implies the Ni Au inter mixing process has occurred and verifies the XPS data. However, there is an abundance of pinholes at the initial Ni/Au interface which are likely related to the residues discussed earlier. For samples with graphene, top down SEM analysis demonstrates a sm ooth, residue free film. The STEM analysis clearly shows a distinct Ni layer as the interface to
99 p GaN. The analysis also shows a discoloration on the top interface of the Ni film. This could be a reaction between the Ni and graphene that forms NiC. The th icknes s of this second film is 10 15Ã… . In a similar study by Kim et al . 213 . a Au/graphene/Ni stack on silicon was annealed for one minute at 600 o C without delamination of the gold. In conclusion, through XPS and STEM analysis, we have demonstrated that gr aphene prevents an interdiffusion between Ni and Au. The Ni and Au transformation is detectable by XPS at 600 o C and confirmed by STEM images on samples without graphene .
100 Figure 4 1. Bonding energies of various metals with graphene. A more negative bind ing energy represents a stronger bond to graphene. [reproduced with permission from references 176,178,179 ] Figure 4 2. The band structures of graphene with physisorbed and chemisorbed metals. A) The band structure of pristine graphene. B). The band st ructure of graphene with a copper interface that demonstrates the graphene bands at the Dirac point are unaffected. C). The band structure of graphene and nickel which shows the destruction of the graphene band structure. [reproduced with permission from r eferences 178 ]
1 01 Figure 4 3. Microscope analysis of Graphene on SiO 2 provided by the Graphene Supermarket. The green particles are unetched copper residues and the discolored area are tears in the graphene film. Photograph courtesy of the author. Figu re 4 4. Raman analysis of five samples of graphene on SiO 2 provided by the Graphene Supermarket. Note the large defect peaks at 1400cm 1 .
102 Figure 4 5. Raman analysis of five samples of graphene on SiO 2 provided by ACS Materials. The defect peak at 1379 cm 1 is minimal on the majority of measurement locations Figure 4 6. Titanium films on single layer and double layer graphene annealed at 600 o C and 800 o C. A) The 600 o C annealed samples show a bright titanium film where the graphene is located. The periphe ral areas show a brownish discoloration indicative of a titanium silicide reaction. B) 800 o C anneals show the single layer graphene is similar color of the peripheral Ti on Si areas indicating the graphene barrier has failed. For the DLG sample the graphen e areas still appear bright and unreacted. Photograph courtesy of the author.
103 Figure 4 7. Sheet resistance results at increasing annealing temperatures for 1000Ã… Ti on single layer and double layer graphene. The control is simply Ti on silicon. Note the delay in forming a low resistance titanium silicide up to 600 o C for single layer graphene and 900 o C for double layer graphene. Figure 4 8. STEM pictures of titanium on graphene and titanium on silicon. Note both pictures are at the same magnification. A) The titanium on double layer graphene shows a clear distinct titanium film and unreacted silicon surface. B) Titanium on silicon sample shows a clear interaction with silicon. The thickness of the titanium film is roughly 2.X the original thickness, whi ch indicates the entirety of the titanium film has been reacted to form a silicide.
104 Figure 4 9. Magnified STEM images of titanium on double layer graphene and titanium on silicon. A) The gap in between the titanium film and the silicon surface is approx imately 40Ã… which could be attributed to a TiC layer. B) A clear growth from of a titanium silicide formation without the graphene diffusion barrier. Figure 4 10. Copper delamination on graphene after a 825 o C anneal at 5X and 50X.A) The graphene areas a re almost clear of copper while the SiO 2 outer areas still have a copper coating. B) Close up picture graphene still intact albeit with defects. Photograph courtesy of the author.
105 Figure 4 11. XPS data on a control sample of Cu on SiO2 and Cu/graphene on SiO2 after a copper wet etch. A) A clear copper peak is seen on samples without graphene. B) No copper peak is seen on samples with graphene even though the copper delaminated. Figure 4 12. Sheet resistance measurements of Ni/Au on SiO 2 and Ni/Au on p GaN. The data shows no difference in sheet resistance with temperature, indicating the four point probe method is not a valid indicator of the Ni Au intermixing.
106 Figure 4 13. An image of graphene on nickel after transfer. Photograph courtesy of the autho r. Figure 4 14. An analysis from the substrate backside demonstrating the evolution of Ni Au intermixing at different processing temperatures on p GaN substrates. Photograph courtesy of the author.
107 Figure 4 15. Microscope 10X image of particles after 600 o C anneal in areas where graphene is not present. Note the clear demarcation of the particle free graphene areas. Photograph courtesy of the author. Figure 4 16. An SEM image of the contamination in areas in which graphene is not present. Note the cl ear particle free surface of the Ni/Au on graphene.
108 Figure 4 17. XPS spectrum of anneals of Ni/Au on p GaN. A) 600 o C which shows a Ni peak at 852 eV. B) 650 o C anneal. C) 700 o C anneal. D) 600 o C anneal of Ni/Au on single layer graphene. There is no detecti on of the nickel peak at 852eV.
109 Figure 4 18. STEM image of the complete intermixing of Ni Au. Note the contamination at the interface to p GaN. Note, there are no distinct Au or Ni films. Figure 4 19. STEM image of Ni/Au on graphene after 600 o C anneal . The white layer is a contrast reversal and is the graphene layer.
110 Figure 4 20. Magnified STEM image of Ni/Au on p GaN after 600 o C anneal that shows no sign of the Ni layer indicating the Ni Au intermixing process has occurred. Figure 4 21. Magnified STEM image of Ni/Au on graphene/ p GaN after 600 o C anneal that shows clearly the Ni layer is present and intact.
111 CHAPTER 5 CONCLUSIONS AND PROSPECTIVE The results described in this thesis can be summarized as follows: we demonstrated the reasoning and cha llenges of a rapid thermal graphene growth system. We were able to grow successfully single layer graphene on copper substrates. Although valuable data on design aspects has been generated, looking ahead, a complete chamber redesign is necessary to continu e with optimizing the graphene growth process. The new design will have all electronics outside of the heating zone. In addition, shielding from the halogen lamps of the o rings as well as extended CDA cooling will prevent the degradation problems we have experienced. We expect to continue the concept of growing graphene on dielectric substrates without using a transfer process. For graphene diffusion barrier testing, it was shown through the formation of a carbide layer with metals that are chemisorbed on graphene, an effective barrier can be achieved. As for physisorbed metals, the barrier failure mechanism is adhesion at the metal and graphene interface. In the case of Au Ni on p GaN, graphene effectively blocked the intermixing between the two metals, h owever a closer inspection is needed to determine if graphene will be beneficial in this application. Through the testing of three different interfaces to graphene, it is clear that diffusion through the graphene barrier was not the cause of failures. Th state diffusion barrier is positive. We expect, through continued research, that adhesion promoters or surface treatments that allow bonding to graphene, yet leave the hexagonal lattice untouched, will be
112 still a relatively new material and much is still to be learned about the phenomenal properties of graphene.
113 LIST OF REFERENCES 1) A.K. Geim and K.S Novoselov, Nat. Mater. , 6 , ( 2007 ) 2) Graphene, The Worldwide Patent Landscape in 2013, The Intellectual Property Office, Concept House, Newport, NP10 QQQ, www.ipo.gov.uk/informatics 3) A. Geim, Science , 324 (2009) 4) F. Schwi erz, Nature Nanotechnology , 5 (2010) 5) M. Dragoman et al ., IEEE Micro. , 1527 3342 (2010) 6) T.N.Narayanan, Physics Alumni News , Cochin University, 3 , ( 2010 ) 7) A. Reina et al ., Nano Lett. 9 , 30 (2009) 8) K.S.Kim et al ., Nature 457 , 706 (2009) 9) J. T. Robinson et al., Nano Lett. 8 , 3441 (2008) 10) J.S. Bunch et al., Science 315 , 490 (2007) 11) J. Hass, W. de Heer and E. Conrad E. H., J. Phys Condens. Mat . 20 , 27 (2008) 12) M.D. Stoller et al ., Nano Lett., 8 , 3498. (2008) 13) E. Yoo et al ., Nano Lett., 8 , 2277 (2008) 14) D.H. Wang et al ,, ACS Nano 3 , 907 (2009) 15) B. Seger, and P. Kamat, J. Phys. Chem. C 113 , 7990 (2009) 16) R. Kou et al ., Electrochem. Commun . 11 , 954 (2009) 17) E. Yoo et al ., Nano Lett. 9 , 2255. (2009) 18) Y.C. Si and E.T. Samulski, Chem. Mater. , 20 , 6792 (2008) 19) Y.M. Li, L.H. Tang and J.H. Li, Electrochem. Commun. 11 ,846 (2009) 20) X. Wang et al ., Angew. Chem. Int. Ed. 47, 2990 (2008) 21) J.B. Wu et al ., Appl. Phys. Lett . 92 , 263302 (2008) 22) Z. Liu et al ., J. Am. Chem. Soc. 130 , 10876 (2008)
114 23) H. Chen et al ., Adv. Mater. 20 , 3557 (2008) 24) C.S. Shan et al ., Anal. Chem. 81 , 2378 (2009) 25) Z. Wang et al., J. Phys. Chem. C 113 , 14071 (2009) 26) C.H. Lu et al ., Angew. Chem. Int. Ed.t . 48 , 4785 (2009) 27) Y. Wang et al ., Anal. Chem. 81 , 9710 (2009) 28) S. Vadukumpully et al ., Carbon 49 ,1 (2011) 29) O. Leenaerts, B. Partoens and F. Peeters , Appl. Phys. Lett. 93 , 193107 (2008) 30) M.A. Nicolet , Thin Solid Films, 52, 415 443 (1978) 31) A. Kaloyeros, and E. Eisenbraun , Annu. Rev. Mater. Sci . 30 , 363 385 (2000) 32) G.S. Painter and D.E. Ellis, Int. J. Quantum Chem . (1969) 33) G.S. Painter and D.E. Ellis, Phys.Rev. B 1 , 12 (1970) 34) K.I. Bolotin, et al., Solid State Commun. 146 , 9 (2008) 35) K. Nomura and A. MacDonald, Phys. Rev. Lett. 98 , 076 602 (2007) 36) E. Hwang, S. Adam, S. and S. Das Sarma, Phys. Rev. Lett. 98 , 186806 (2007) 37) M. Ishigami et al., Nano Lett. 7 , 1643 1648 (2007) 38) S.V. Morozov et al ., Phys. Rev. Lett. 100 , 016602 (2008) 39) S. Fratini, S. and F. Guinea, Phys. Rev. B 77 , 195415 (2008) 40) K. Watanabe, T. Taniguchi and H. Kanda, Nature Mater. 3 , 404 409 (2004) 41) G. Giovannetti, et al ., Phys. Rev. B 76, 073103 (2007). 42) C.R. Dean et al ., Nature Nanotechnology , 5 (2010) 43) A. Sakhaee Pour, Solid State Commun. 149 (2009) 44) G. Van Lier et al. Chem. Phys. Lett. 326 (2000) 45) Y. Gao and P. Hao, Physica E 41 (2009)
115 46) E. Cadelano et al ., Phys. Rev. Lett. 102 , 235502 (2009) 47) C. Lee et al., Science 321 , 385 (2008); 48) G. Lee et al., Science 340 , 107 3 (2013) 49) C. Lee et al., Science , 321 (2008) 50) R. Grantab, V. Shenoy and R. Ruoff , Science , 330 (2010) 51) K.V. Zakharchenko et al . Phys. Rev. Lett. 102 , 046808 (2009) 52) J.W. Jiang et al . Phys. REV. B 80 , 205429 (2009) 53) W. Bao et al . Nat ure Nanotechnology , 4 , (2009) 54) D.Yoon ,Y. Son and H. Cheong , Nano Lett. , 11 , 8 (2011) 55) W. Chen et al ., J. Vac. Sci. Tech. B 30 , 040602 (2012) 56) X. Liang et al . ACSNano, 5 , 11 (2011) 57) Y. Wang et al . ACSNano 5 ,12 (2011) 58) L. De Arco, IEEE Trans. Nanotech . 8 , 2, (2009) 59) Y. Lee et al . Nano Lett. 10 , 490 493 ( 2010) 60) P. Sutter, P. Albrecht and E. Sutter, Appl. Phys. Lett . 97 , 213101 (2010) 61) Y. Zhang et al ., ACSNano 6 , 1 (2012) 62) L. Xie, L. Jiao and H. Dai, J. Am. Chem. Soc . 9, (2010 ) 63) Y. Cui et al . J. Phys. Chem. C 113 (2009) 64) C. Gong et al ., J. Phys. Chem. C , 117 (2013) 65) M. Schriver et al ., ACSNano 7 , 7 (2013) 66) F. Zhou et al ., ACSNano 7 , 8 (2013) 67) F. Banhart, J. Kotakoski and A. ACS Nano 5 ,1 (2011) 68) M. Allen et al . Chem. Rev . 110 (2010)
116 69) S. Hagstrom, H. Lyon and G. Somorjai, Phys. Rev. Lett. 15 , 491 (1965). 70) H.B. Lyon and G.A. Somorjai, J. Chem. Phys. 46 , 2539 (1967) 71) A.E. Morgan and G.A. Somorjai, Surf. Sci . 12 , 405 (1968) 72) J.W. May, Surf. Sci. 17 , 267 (1969) 73) D. Eom et al., Nano Lett. , , 9 , 8 ( 2009) 74) H. Ago et al., ACS Nano , 4 , 12 (2010) 75) E.V. , A.V. and N.R. Gall, Phys. Solid State + 53 ,5 (2011) 76) S. Roth , J. Osterwalder and T. Greber , Surf. Sci . 605 , 9 10 (2011) 77) G. Nandamuri, S. Roumimov and R. Solanki, Nanotechnology 21 ,145604 (2010) 78) H. Chen et al., Carbon 48 , 12, (2010) 79) L. Baraton et al., Europhys. Lett. 96 , 4 (2011) 80) H. Park et al., Carbon 48 , 4, (2010) 81) Z. Juang, Carbon 48 , 11, (2010) 82) Y. Cui, J. Phys. Chem. C 113 (2009) 83) Loginova, E. et al ., New J. Phys . 11 , 063046 (2009) 84) K.F. McCarty et al ., Carbon , 47, 1806 (2009) 85) S. Marchini, S. Gunther and J. Wintterlin, J. Phys. Rev. B 76 , 075429 (2007) 86) S. Kwon et al., Nano Lett. 9 , 12 (2009) 87) T.A. Land et al., Chem. Phys. 97 , 6774 (1992) 88) H. Ueta et al ., Surf. Sci. 560 , 183 (2004) 89) D.E. Starr et al ., Surf. Sci. 600 , 2688 (2006) 90) I.V. Makarenko et al., Appl. Surf. Sci. 252 , 1221 (2005) et al ., Phys. Rev. Lett. 97 , 215501 (2006)
117 92) J. Coraux et al ., Nano Lett. 8 , 565 (2008) et al ., New J. Phys. 10 , 043033 (2008) 94) Wu, W. et al ., Adv. Mater. 23 , (2011) 95) Z. Luo et al . ACS Nano , 5 , (2011) 96) Q.K. Yu et al ., Nat. Mater. 10 (2011) 97) L.B. Gao et al ., Nat. Commun ., 3 , 699 (2012) 98) Z. Yan et al ., ACS Nano , 6 (2012) 99) Y. Wu et al ., Chem. Mater. 23 (2011) 100) H. Wang et al ., J. Am. Chem. Soc. 134 (2012) 101) X.S. Li et al ., J. Am. Chem. Soc. , 133 (2011) 102) X.S. Li et al ., Nano Lett . 10 (2010) 103) X. Li, Nano Lett. 9 , (2009) 104) J.D. Wood et al ., Nano Lett. 11 (2011) 105) G. Nandamuri , S. Roumimov S. and R. Solanki, Nanotechnology , 21 , 14 (2010) 106) C. Mattevi, H. Kim and M. Chhowalla, J. Mater. Chem. 21 , 10 (2011) 107) H. Ji et al ., ACS Nano 5 , 9 ( 2011 ) 108) Z. Li et al ., Nano Letters 10 (2010) 109) S. Bhaviripudi et al ., Nano Letters 10 (2010) 110) Silicon Processing for the VLSI Era , Vol 1. S. Wolf & R. Tauber, Lattice Press (1986) 111) T. Hao et al ., Science 8 , 34 2 (2013) 112) H. Lee et al ., Science 344 , 286 (2014); 113) Raman Spectroscopy in Graphene Related Systems , Ado Jorio, Riichiro Saito, Gene Dresselhaus, Mildred Dresselhaus Wiley VCG (2011)
118 114) L. Malard et al ., Phys. Rep . 473 (2009) 115) M. Dresselhaus et al ., Phys. Rep . 409 (2004) 116) R. Saito et al ., Adv. Phys., 60 , 3 (2011) 117) J. Park, et al ., Carbon 47 (2009) 118) M. Mohr, J. Maultzsch and C. Thomsen, Phys. Rev. B 82 201409 (2010) 119) O. Frank et al ., ACS Nano 5 , 3 (2011) 120) A. Das et a l ., Nature Nanotechnology , 3 (2008) 121) J. Caridad et al ., J. Appl. Phys . 108 , 084321 (2010) 122) I. Calizo et al ., Nano Letters 7 , 9 (2007) 123) F. Wenner, Journal of the Franklin Institute 180.3 (1915) 124) L.B. Valdes, Proceedings of the IRE , 42.2 ( 1954) 125) Semiconductor Material and Device Fabrication , Second Edition, D. Schroder, , John Wiley & Sons. (1988) 126) K. Holloway and P. Fryer, Appl. Phys. Lett. 57 , 1736 (1990) 127) H. Ono T. Nakano and T. Ohta, Appl. Phys. Lett. 64 , 1511 (1994) 128 ) Y. Kuo et al ., Electrochem. Solid St. 7 , 3 ( 2004) 129) H. Yan et al ., 11th Electronics Packaging Technology Conference (2009) 130) E. Kolawa et al ., J. Appl. Phys . 70 , 1369 (1991) 131) K. Holloway et al ., J. Appl. Phys . 71 , 5433 (1992) 132) P. Catan ia, J. Doyle and J. Cuomo, J. Vat. Sci. Technol. A 10 , 3318 (1992) 133) J. Olowolafe et al., Appl. Phys. Lett. 58 , 469 (1991) 134) P. Bai et al ., Appl. Phys. Lett. 60 , 1824 (1992) 135) C. Apblett et al ., J. Appl. Phys . 71 , 4925 (1992) 136) J. Imahori et al ., Thin Solid Films , 301 , Issues (1997)
119 137) J. Kelsey et al , J. Vac. Sci. Technol . B 17 , 1101 (1999) 138) S. Rawal et al ., J. Appl. Phys. 100 , 063532. (2006) 139) M.B. Chamberlain, Thin Solid Films 91 , 155 (1982) 140) A. Cara A et al ., Funct. Mat er., 20 (2010) 141) L. Henderson, J. Ekert, Thin Solid Films , 517 (2009) 142) Matsumoto, K et al ., IEEE International, Interconnect Technology Conference (IITC) , (2013) 143) H. Wojcik et al ., Microelectronic Engineering , 110 , (2013) 144) H. Shimizu, et al ., IEEE International, Interconnect Technology Conference and 2011 Materials for Advanced Metallization (IITC/MAM), (2011) 145) G. Rampelberg et al ., 13th International Conference on Atomic Layer Deposition (ALD 2013) 146) A. McCoy et al ., Appl. Phys. Lett. 101 , 231603 (2012) 147) C. Chang, Appl. Phys. Lett. 54 , 1656 (1989) 148) C. Chang, J. Appl. Phys. , 66 , 2363 (1989) 149) C. Chang et al ., Appl. Phys. Lett. 54 , 2545 (1989) 150) C. Chan, J. Appl. Phys ., 66 , 1163 (1989) 151) C. Chang, J. Appl. Phy s. , 67 , 6184 (1990) 152) C. M. Whelan et al ., Mater. Res. Soc. Symp. Proc. 812 , F2.2.1 (2004) 153) K. Limura, Y. Nakajima and T.Kato, Thin Solid Films 379 , 230 (2000) 154) D. D. Gandhi et al ., Appl. Phys. Lett. 90 , 163507 (2007) 155) D. D. Gandhi et al ., Nature 447 , 299 (2007) 156) A. Caro et al ., Adv. Funct. Mater. , 20 , (2010) 157) Z.Kong, et al ., Appl. Surf. Sci. 279 (2013) 158) X.Liu et al ., J. Electrochem. Soc . 153 , 3 (2006)
120 159) N. Semaltianos, Surface & Coatings Technology, 201 (2007) 160) N. Mikami et al ., Appl. Phys. Lett ., 83 , 25 (2003) 161) Krishnamoorthy et al ., Appl. Phys. Lett . 78 , 17 (2001) 162) P. Ganesan et al ., Appl. Phys. Lett , 85 , 4 (2004) 163) J. M. Fairfield and G. H. Schwuttke, Solid State Electron. 11 , (1968) 164) United St Mammel 165) N.K. Hiester et al ., Journal of the American Rocket Society , 27 (1957) 166) B. Lojek, RTP . Vol. 99 . 1999 167) Ushio, Halogen Lamp Technical Notes , Ushio, Inc 168) Principles o f Electronic Materials and Devices , S.O. Kasap, 3rd Edition, McGraw Hill, (2007) 169) Rapid Thermal Processing Science & Technology , R.B. Fair, Academic Press (1993) 170) Momentive Performance Materials Inc., Technical Bulletin 171) A. Katz and S.J. Pea rton, J. Vac. Sci. Technol. B8 (6) (1990) 172) S.J. Pearton and R. Caruseo, J. Appl. Phys . 66 ,663 (1989) 173) X. Li et al ., Nano letters 10 .11 (2010) 174) H. Kim et al ., ACS Nano 7 , 8, (2013) 175) Q. M. Ramasse et al., ACS Nano , 6 , 5, (2012) 176) Q. Ra n et al., Appl. Phys. Lett. 94 , 103511 (2009) 177) Q. J. Wang and J. G. Che, Phys.Rev. Lett. 103 , 066802 (2009) 178) C. Gong et al., J. Appl. Phys. 108 , 123711 (2010) 179) G. Giovannetti et al., Phys.Rev. Lett. 101 , 066802 (2009) 180) L.S. Hung et al . , J. Appl. Phys. 54 (9) (1983)
121 190) T. Brat, et al ., J. Electrochem. Soc. 133 , 7 (1986) 191) Z. Ma, L.H. Allen and D.D. Allman J. Appl. Phys. 77, 9 (1995) 192) C. Gong et al., ACS Nano 8 , 1 (2014) 193) D.Q. Yang and E. Sacher, J. Appl. Phys . 90 , 9 (2 001) 194) J. Smalc Koziorowska et al ., Solid State Elect. 54 (2010) 195) R. Piotrzkowski et al ., Physica B 308 (2001) 196) E. Litwin Staszewska et al., J. Appl. Phys. 89 :7960 (2001) 197) Z. Liliental Weber et al ., Phys Stat Sol B 228 ,345 (2001) 198) T. Mori et al ., Appl. Phys. Lett. 69 , 3537 (1996) 199) L. Wang et al ., Appl. Phys. Lett. 68 , 1267 (1996) 200) M. Lee et al ., Appl. Phys. Lett. 91 , 182106 (2007) 201) J. Jang et al ., Appl. Phys. Lett. 74 , (1999) 202) J. Jang et al ., J. Appl. Phys. 88 , 5 490 (2000) 203) J. Sheu et al ., Appl. Phys. Lett. 74 , 2340 (1999) 204) A. Motayed et al ., J. Appl. Phys . 92 , 5218 (2002) 205) J. Ho et al ., Appl. Phys. Lett. 74 , 1275 (1999) 206) H. Omiya and F. Ponce et al ., Appl. Phys. Lett. 85 , 6143 (2004) 207) H .Omiya. et al ., AIP Conference Proceedings 772, 421 (2005) 208) P.J. Hartlieb et al . J. Appl. Phys ., 91 , (2002) 209) C.Hu. et al. , Mat. Sci.and Eng.B Solid 128 (2006) 210) Z. Chen et al ., Mat. Sci.and Eng.B Solid 100 (2003) 211) C. Hu et al ., J. Cryst. Growth 298 (2007) 212) C. Gong et al ., ACSNano 6 (2012)
122 213) H.Y. Kim et al ., J. Vac. Sci. & Technol. B 30 , 030602 (2012)
123 BIOGRAPHICAL SKETCH Wayne Morrow was born in Lakeland, Florida in 1963. He spent the majority of his childhood in Lake City, Flor ida. After a brief two year family hiatus in Wyoming and Colorado, he returned to Florida and spent his teen years in Lakeland. In May 1987, he graduated from the Department of Materials Science and Engineering at the University of Florida and was the firs t class to grad uate with the new specialty of electronics m aterials. Later, after pursuing a career at Motorola in Phoenix, Arizona, Wayne returned to academia and in 1996 earned a master ' s degree in materials s cience at Arizona State University, Tempe, Ar izona , with a thesis based on sp 3 diamond like carbon for field emission devices. In 1999, he returned to the University of Florida and enrolled in a newly created EMBA program in the Warrington College of Business and was the first class to graduate with an MBA in 2001, where classes were completely off campus. With both a business and engineering background in hand, Wayne started his own business in 2004 named Surface Science Integration in order to provide quality, low cost semiconductor processing equi pment to Universities and Government Labs. He continues to manage Surface Science Integration to this date. In 2010, after discussions with Dr. Pearton, who probably had more confidence in Wayne ever had himself , he enrolled once again in the Department of Materials Science and Engineering at the University of Florida to pursue a PhD. Once again, as with his m aster s, the topic was carbon based materials, this time sp 2 graphene. He spent the next four years in pursuit of industrial base d applications of graphene. Wayne plans to continue building semiconductor equipment tailored to enhancing the properties and uses of graphene.