UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A study on hot rolling of CP Titanium Koushik, Ray 1996

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1996-0169.pdf [ 9.07MB ]
JSON: 831-1.0078441.json
JSON-LD: 831-1.0078441-ld.json
RDF/XML (Pretty): 831-1.0078441-rdf.xml
RDF/JSON: 831-1.0078441-rdf.json
Turtle: 831-1.0078441-turtle.txt
N-Triples: 831-1.0078441-rdf-ntriples.txt
Original Record: 831-1.0078441-source.json
Full Text

Full Text

A STUDY ON HOT ROLLING OF CP TITANIUM BY KOUSHIK RAY B.Tech.(Hons.), Indian Institute of Technology, Kharagpur, India A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTERS OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES (DEPARTMENT OF METALS AND MATERIALS ENGINEERING) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA MARCH, 1996. ©Koushik Ray, 1996. In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. The University of British Columbia Vancouver, Canada Department of DE-6 (2/88) ABSTRACT A major source of concern in the direct breakdown rolling of titanium is the occurrence of surface cracks when the as-cast vacuum-arc or electron-beam melted ingots are hot-rolled for breakdown. Traditionally, this has been done by open-die forging, but recent developments in direct hot-rolling have revealed this problem. A possible reason why cracks are formed is that titanium forms a surface layer of alpha phase containing oxygen dissolved in solid solution through the diffusion of oxygen during the air soaking of the ingots prior to hot rolling. This layer cracks visibly during the rolling process. Available literature indicates the presence of a brittle surface layer, but its characterization has not been done. This work characterizes the solid solution case and measures the strain to failure of the material. Also, microstructural examination of the compression specimens showed cracks between Widmanstatten patterns which are formed when air-annealed samples are cooled. These results are helpful in understanding the reason for the formation of the cracks during direct rolling and could be used to remedy some of the problems. Constitutive equations for the hot working of CP.Titanium in the alpha and the beta phase have been developed by conducting axisymmetric compression tests on cylindrical specimens using a Gleeble 1500®* in the temperature range 750°C - 950°C for strain rates appropriate for direct hot-rolling. The above results are compared with the theoretical predictions of Ashby based on which Ashby had constructed deformation maps for CP. Titanium. The softening characteristics and the strain rate sensitivities of the two phases are noted to be different. Further, a simple model of the breakdown rolling using DEFORM®**, a finite element software package, has been studied to identify the criticality of some processing parameters. * This is a registered product name for a thermomechanical simulator marketed by Dynamic Systems Inc., Troy, New York. ** This is a registered product name for a software designed by Scientific Forming Technologies Corporation, Columbus, Ohio ii TABLE OF CONTENTS Page Abstract , ii List of Tables ix List of Illustrations x Nomenclature xiv Acknowledgments xvi CHAPTER ONE - INTRODUCTION 1 CHAPTER TWO - LITERATURE REVIEW ...4 1. BASIC TITANIUM METALLURGY* 4 1.1. PHASES OF TITANIUM - THE p o a TRANSFORMATION 4 1.1.1. Temperature of transformation 4 1.1.2. Rate of transformation 4 1.1.3. Nature of transformation 4 1.2. THE TITANIUM-OXYGEN PHASE DIAGRAM 5 1.3. EFFECT OF OXYGEN ON PHASE TRANSFORMATION 6 1.3.1. Effect of oxygen on transformation temperature 1.3.2. Effect of oxygen on rate of transformation 6 1.3.3. Effect of oxygen on nature of transformation 6 1.4. THE OXIDATION OF TITANIUM 7 1.4.1. The Oxidation Kinetics of Titanium 7 1.4.2. Change of Oxygen Kinetics with Temperature 8 1.4.3. Influence of Scale Thickness on Oxidation Kinetics 11 1.4.4. Mathematical Model 11 iii 2. ROLLING OF CP. TITANIUM 14 2.1. ROLLING IN DIFFERENT ATMOSPHERES 14 2.1.1. Vacuum 14 2.1.2. Inert atmosphere 15 2.2. RECRYSTALLIZATION AND RECOVERY DURING ROLLING 17 2.2.1. Recovery and recrystallization temperatures 17 3. INGOT CONVERSION - FORGING ...20 3.1. STAGES IN INGOT CONVERSION 20 3.1.1. Lubrication 21 3.1.2. Removal of contamination and defects 21 3.2. FORCE CONSIDERATIONS 22 3.2.1. Effect of Strain Rate 22 Strain Rates used for forging 23 3.2.2. Effect of temperature 23 Temperatures for forging 24 Effect of die temperature 26 3.2.3. Contamination during processing 26 Hydrogen 26 Oxygen and Nitrogen 26 4. THE CONSTITUTIVE EQUATION... 26 4.1. TYPES OF CONSTITUTIVE EQUATIONS 27 4.2. AVAILABLE EXPERIMENTAL DATA ON CP TITANIUM 28 5. DEFORMATION-MECHANISM MAPS ,. 29 5.1. ASSUMPTIONS IN CONSTRUCTING DEFORMATION MAPS] 30 5.2. DESCRIPTION OF DEFORMATION MAPS 31 5.3. POWER-LAW BREAKDOWN REGIME 31 iv 5.4. DEFORMATION MAP FOR CP TITANIUM 33 5.4.1. Limitations of the material constants 34 Reference 35 CHAPTER THREE - SCOPE AND OBJECTIVES 3 8 CHAPTER FOUR - SIMULATION OF THE ROLLING PROCESS 39 4.1. OBJECTIVE OF THE MODEL 39 4.2. DESCRIPTION OF DEFORM 39 4.3. ASSUMPTIONS 40 4.3.1. Simplifications in scale 40 4.3.2. Simplifications in the processing variables 40 4.3.3. A Critique on the Assumptions 41 4.4. ASSESSMENT OF THE VALUES OF PROCESS VARIABLES 41 4.4.1. Effective strain 43 4.4.2. Effective Strain Rate 44 4.4.3. Temperature ; 45 4.4.4. Stresses 45 4.5. EFFECT OF VARYING PROCESS VARIABLES ON THE ROLLING PROCESS 48 4.5.1. Heat transfer coefficient (Original run value = 40 W/m7K) 48 < 4.5.2. Effect of varying the rolling temperature 51 4.5.3. The effect of varying the interface shear friction 52 4.5.4. Effect of degree of reduction 54 4.5.5. Effect of varying the roll temperature.. 55 4.6. DISCUSSION AND SUMMARY 56 Reference ". 57 v CHAPTER FIVE - EXPERIMENTAL WORK 5.1. CP TITANIUM 58 5.2. SAMPLES SUBJECTED TO OXYGEN ANNEALING 58 5.2.1. Annealing of the samples 58 5.2.2. Surface oxide 58 5.2.3. Oxygen diffusion thickness 59 Microstructural examination..., 59 Hardness measurements 60 WDX analysis 60 5.3. SAMPLES SUBJECTED TO VACUUM-ANNEALING 60 5.4. COMPRESSION TESTING 61 5.4.1. Compression testing for the as received (non-annealed) samples ..61 Compression Testing using the Gleeble® 1500 Thermomechanical Simulator.. 61 Processing schedule 63 5.4.2. Compression Testing for the annealed samples 64 Processing Schedule 65 5.4.3. Deformation of the Vacuum annealed sample 65 5.5. EXAMINATION OF DEFORMED SAMPLES ,. 65 References .• -65 CHAPTER SIX - RESULTS AND DISCUSSION.. 66 6.1. OXYGEN DIFFUSION RESULTS 66 6.1.1. The oxide scale 66 6.1.2. The alpha-stabilized surface layer 66 6.1.3. The Widmanstatten pattern formation 66 6.1.4. Measurements to determine oxygen concentration 69 Constructing the oxygen profile 71 vi Constructing the oxygen profile 71 6.2. DISCUSSION ON RESULTS OF OXYGEN DIFFUSION 73 6.2.1. The oxide scale 73 Dependence on Temperature 73 6.2.2. The Oxygen stabilized Alpha case 74 6.2.3. Growth of the alpha phase 75 6.2.4. The Widmanstatten pattern development and formation 76 6.3. RESULTS OF COMPRESSION TESTS ON ANNEALED SAMPLES .... 77 6.3.1. Visual Inspection. 77 6.3.2. Microstructural Examination 78 6.3.3. Strain to Fracture 80 Comparison of Stress - Strain Curves 80 6.3.4. Effect of grain size on flow stress 82 6.4. DISCUSSION OF THE COMPRESSION TEST RESULTS 83 6.4.1. Microstructural examination 83 6.4.2. Stress-Strain Curve Analysis 85 6.4.3. Effect on grain size on flow stresses: 86 6.5. GENERAL DISCUSSION 87 6.5.1. Presence of interstitial oxygen 87 6.5.2. Phase transformation effects during breakdown rolling 87 6.5.3. Surface chilling effects during breakdown rolling 87 6.5.4. Grain size :....88 Effect on Oxygen Diffusion 88 Effect on susceptibility to Cracking 88 6.6. MICROSTRUCTURAL EVOLUTION OF THE SPECIMENS 89 6.7. FLOW STRESS STUDIES FOR CP TITANIUM 91 vii 6.7.1. The Alpha Regime 92 Comparison with Ashby's theoretical predictions 93 Constitutive Equations in the Alpha phase 94 6.7.2. The Beta regime 95 Comparison with Ashby's predicted values 96 Constitutive Equations in the Beta phase 97 6.8 DISCUSSION OF FLOW STRESS RESULTS 98 6.8.1. Alpha phase 98 Comparison with Ashby's predicted results 98 Constitutive Equations in the Alpha phase 99 Exponential Law 99 Power Law 99 6.8.2. Beta phase - Flow stress results 101 Comparison with Ashby's predictions 101 Constitutive Equations in the B eta phase 102 Reference '. 103 CHAPTER 7 - SUMMARY AND CONCLUSIONS ...105 7.1.SUMMARY AND CONCLUSIONS 105 7.2. RECOMMENDATIONS 106 7.3. SCOPE FOR FURTHER WORK 107 APPENDIX A 108 APPENDIX B -109 viii LIST OF TABLES Page Table 2.1 - Forging temperatures for CP titanium 25 Table 4.1 - The parameters incorporated in the model. 42 Table 5.1 - Chemical composition of the as-received sample 58 Table 5.2 - Composition of the Kroll's reagent used as etchant 60 Table 5.3 - Deformation conditions for the uncontaminated samples 64 Table 5.4 - Deformation conditions for the oxygen contaminated samples 65 Table 6.1 - Comparison of the oxide scale thickness after different times of soaking anneal treatments 66 Table 6.2 - Composition of oxide scale by weight analyzed by WDX measurements 67 Table 6.3 - List of strains to fracture as deduced from the flow curve comparisons between annealed and plain samples 80 Table 6.4 - Q, n and A values of exponential fit in the beta region at various strains 97 Table 6.5 - A,B and C values for the equation of the form Power Law Equation in the beta region 97 Table 6.6 - n values (for the power law fit) obtained at different temperatures 100 Table A . l - Q, n and A values for different values of a for the sinhyperbolic fit in the alpha region 108 ix LIST OF ILLUSTRATIONS Figure Page Figure 2.1 Phase Diagram of the Ti-0 system 5 Figure 2.2 Comparisons of the microstructure of unalloyed titanium and titanium containing dissolved oxygen on quenching from P field 7 Figure 2.3 Kinetics of oxidation of titanium 8 Figure 2.4 Duration of the parabolic kinetics as a function of temperature 9 Figure 2.5 Thickness of the oxide layer and gain of mass of the samples at the transition of the kinetics from parabolic to linear as a function of temperature 9 Figure 2.6 Adherence and colour of oxide as a function of its thickness and temperature 10 Figure 2.7 Schematic representation of the oxygen concentrations at the interfaces 12 Figure 2.8 Comparison of the diffusion coefficients of oxygen in alpha titanium presented by different workers 13 Figure 2.9 Comparison of properties of titanium hot-rolled in different atmospheres 15 Figure 2.10 Comparison of gas pick-up with temperature between argon and air 16 Figure 2.11 Schematic representation of recovery and recrystallization of titanium in the alpha phase 18 Figure 2.12 Change in structure with hot rolling reduction and temperature 19 Figure 2.13 Effect of forging temperature on required pressure for three Ti-alloys compared to 4340 alloy steel 24 Figure 2.14 Deformation Map of CP Ti with a grain size of 100 microns, also showing the data used to construct it 33 x Figure 4.0 Schematic diagram showing the finite-element mesh and the positions that were examined 41 Figure 4.1 Effective strain vs. Time at positions 1, 2, 3 and 4. 43 Figure 4.2 Effective strain rate vs. time at positions 1,2,3 and 4 44 Figure 4.3 Temperature vs. time at positions 1, 2, 3 and 4 45 Figure 4.4 The x-component of the stress vs. time at positions 1, 2, 3 ,4 47 Figure 4.5 The y-component of the stress vs. time at positions 1, 2, 3,4 47 Figure 4.6 The z-component of the stress vs. time at positions 1, 2, 3, 4 48 Figure 4.7 The shear component of stress vs. time at positions 1, 2, 3, 4 48 Figure 4.8 Effect of heat-transfer coefficient on material chill depth 49 Figure 4.9 Effect of heat-transfer coefficient on strain 50 Figure 4.10 Effect of rolling temperature on material chill 51 Figure 4.11 Comparison of the compressive rolling forces required at different rolling temperatures 52 Figure 4.12 Effect of degree of reduction on the chill of the material element at the surface and mid- centre 53 Figure 4.13 Effect of degree of reduction on the strain in the material element at the surface and mid-centre 54 Figure 4.14 Effect of varying the temperature of the rolls on the chilling of the material element at the surface and the mid-centre 54 Figure 4.15 Effect of decreasing the friction on the temperature decrease experienced at the surface and mid-centre element 55 Figure 5.1 Schematic diagram of the holding chamber of the Gleeble 1500® 62 Figure 6.1 Picture comparing the physical appearance of three samples - from left - soaked at 1100°C for 24 hours, at 750°C for 24 hours, untreated sample 67 xi Figure 6.2(a) Micrograph of a portion of the cross section of the CP Ti rod, showing the alpha case formed after annealing at 1100°C for 24 hours (Magnification 200X) 68 Figure 6.2(b) Micrograph of the cross section of the sample vacuum after annealing at 1100°C for 24 hours (Magnification 100X) 68 Figure 6.3(a) Vicker's microhardness indentation showing decreasing hardness with distance from the edge, and also the sudden fall in hardness at the interface suggesting a different phase.(200X) 70 Figure 6.3(b) Plot of microhardness vs. depth as measured in Figure 6.3 70 Figure 6.4 Plot of microhardness vs. depth as measured in Figure 6.3 70 Figure 6.5 Plot of oxygen concentrations by WDX at equivalent depths 72 Figure 6.6 Equilibrium Diagram of Ti-0 in the region of our interest (not to scale) 72 Figure 6.7 Comparison between deformed samples - the left one was oxygen contaminated previously by annealing at 1100°C, the right one was not. The temperature was 850°C, strain rate was 10 s"1 deformed to a final strain of 1. 11 Figure 6.8(a) Micrograph of oxygen contaminated sample deformed at 850°C at 10s"1 to a final strain of 1. Cracks arelimited to the outer boundary. Inside is the deformed ductile phase, which does not show any evidence of recrystallization. (100X) 78 Figure 6.8 (b) Micrograph of uncontaminated sample deformed at 850°C at 10 s"1 to a strain of 1. There is no evidence of surface cracking and the material has undergone recrystallization. (400X) 79 Figure 6.8 (c) Micrograph of vacuum annealed sample deformed at 850°C at 10 s"1 to a strain of 1. There is no evidence of surface cracking and the material which previously had large grains has undergon some recrystallization (200X) 79 Figure 6.9(a) Flow curve of oxygen contaminated sample compared against control sample tested in alpha phase region 81 Figure 6.9(b) Flow curve of oxygen contaminated sample compared against control sample tested in beta phase region 81 Figure 6.10 Photograph of sample deformed at 850°C at 10s"1 after annealing at 750°C instead of 1100°C. Sample shows no sign of cracking 82 xii Figure6.11(a) Micrograph showing Widmanstatten pattern formation (200X) 84 Figure 6.11(b) Micrograph showing air-annealed sample after deformation. The cracks have a regular arrangement, and different orientations for different grains (200X) 84 Figure 6.12 Microstructure of the as-received samples showing equiaxed grains (200X) 90 Figure 6.13 Microstructure of as-received sample deformed in the beta phase at 900 °C at a strain rate of 10s"1 to a strain of 1. (200X) 90 Figure 6.14 Flow stress curves for alpha titanium from axisymmetric compression tests 92 Figure 6.15 Comparision between the flow stress values predicted Ashby's data and those experimentally obtained, in the alpha phase 93 Figure 6.16 Flow stress curves form the Gleeble compression tests done in the beta region 95 Figure 6.17 Comparision between the stress values predicted by Ashby and those experimentally obtained at strain = 0.2 by compression test experiments in the beta phase 95 Figure 6.18 Comparison of the experimental and predicted values using the power law constitutive equation in the alpha regime. 101 Figure B.I Flow curves showing dynamic recrystallization at various temperatures for a strain rate of 0.01 s"1 109 xiii NOMENCLATURE a fitting parameter in the sinhyperbolic equation a hep low temperature phase of titanium a' fitting parameter a ' martensitic phase P bec high temperature phase of titanium 8 strain Zj strain to fracture S strain rate p temperature corrected shear modulus u 0 shear modulus at absolute zero temperature a flow stress A, A', A" constants A2 Dorn constant b Burger's vector C constant D diameter D diffusion coefficient D, diffusion coefficient in the alpha phase D2 diffusion coefficient across the oxide layer Deff lattice diffusion co-efficient at that temperature d grain diameter dj initial grain diameter e engineering strain xiv G elastic shear modulus H height Hv Vicker's microhardness K Arhennius Rate Law constant L0 initial length L, final length m interface shear coefficient of friction m mass unit in milligrams n power-law exponent or stress exponent n, n', n" related to strain-rate sensitivity Q the activation energy of deformation R gas constant initial radius R, final radius s surface area in cm2 T temperature in Kelvin T melting point t time z Zener-Hollomman parameter XV ACKNOWLEDGMENT I would like to thank Dr. Mitchell and Dr. Hawbolt for their, able guidance and direction throughout the course of my reading. I deeply appreciate their support during this entire period. I would also like to thank Dr. Poole for his involvement and interest, and for his ready help whenever requested. There have been a few people without whom accomplishing most of the work would have been daunting. I would like to acknowledge A l Schmalz for being there, and Mary Wells for walking me through a quirksome DEFORM software. Binh Chau and Xiande Chen unfailingly met my imposed urgencies for performing the Gleeble tests. Carl Ng, Ross McLeod and Serge Milaire would oblige - promptly - every request for experimental samples. I would also like to remember the old guards, most of whom have since graduated, who introduced me to the ways and works of the Department on my arrival. I am indebted to our graduate secretary, Joan Kitchen, for her help and patience in my dealings with the officialese. I would also like to thank our cosmopolitan department - the students, the staff, the faculty - for making my sojourn a thoroughly enjoyable international experience. Finally, I would like to thank my parents and my little sister for all that good cheer and support from across the seas. xvi CHAPTER ONE INTRODUCTION This work examines breakdown rolling of titanium ingots into slabs and billets. The rolling process is intended to substitute the forging process which presently achieves the same objective , but at significantly lower rates of productivity. Titanium is cast into ingots using Vacuum Arc Remelting (VAR) and Electron Beam Melting (EBM) techniques, because the Ti/Ti0 2 equilibrium has a very low oxygen potential, -18 8 with equilibrium partial pressure of oxygen 10" atm at 1000°C compared to 10 atm [1] of that of the Fe/FeO system. The VAR ingots are cylindrical and have large diameters (about 1000 mm) for economic. reasons. The EBM furnace can produce slabs which are usually thicker than 600 mm. The ingots have a rough surface layer, often with.cracks, due to non-uniform thermal contraction and due to the presence of solidified relatively volatile impurities on the surface. This surface layer must be removed by milling before the primary breakdown of the ingot. This milling step is expensive - the only redeeming factor is that the milling scrap can be recycled back to the remelting furnaces. Thin titanium pipes and tubes have corrosion resistance applications in heat exchangers and desalination plants. To make them, fabrication processes such as extrusion and sheet rolling are necessary. However, the ingots must be broken down into suitable shapes and micro structures before these processes can be carried out. Presently, the method used for the breakdown of these ingots is forging. The surfaee-machined ingots are usually open-die forged at very low strain rates. This process has worked successfully and is well characterized but introduces difficulty in temperature control. Other titanium alloys (such as Ti-6-4 and Ti-6242) also undergo primary breakdown by forging. To date, most of the applications of titanium has been as common alloys primarily in defence and aerospace, where the cost of production has been of lesser consequence and 1 concern. As a result, reliable, robust and reproducible processes formed a standard for processing the materials, even if they were not the possibly most economic ones. However, recent global economic and political changes necessitate that the titanium industry looks at other opportunities for growth. That is not a problem in itself because titanium is favoured for many applications, and is the sole possibility in many corrosion-resistant structures. The problem is that it cannot lend serious competition to steel because of its high cost, which can be up to 10 times that of stainless steel. Breakdown rolling of titanium is crucial because it increases the rate of production of slabs for the titanium manufacturers and thus lowers the cost. Also, since the strain-rate is higher in rolling as compared to forging, the microstructural refinement is better for the subsequent secondary working processes. The breakdown rolling process for titanium is still to be accepted as conventional practice, and these ingots are typically rolled in mills normally used for steels to convert them to billets and slabs. Before rolling, these ingots are soaked in steel soaking pits at temperatures between 1000 - 1100°C for long times* since it is desired that the ingots are rolled at a temperature in the ductile beta phase, and are transformed to the alpha phase during the course of the rolling process for the best combination of strength, microstructural properties, and processing ease. The soaking temperatures and heat-treating regimes vary from company to company, and in most cases the knowledge is proprietary. Problems can arise because cracks appear on the surface of the ingots during the initial passes. * Nominally, 1 hour per inch of thickness of the material - a 625 mm thick slab or a 1000 mm diameter ingot is considered. However, varying experimental heat-schedules are being tried by industries for better results 2 It is believed that one of the principal causes of these cracks is the mechanical failure of the solid-solution oxygen-stabilized alpha layer on the surface of the ingots. This layer is formed during the soaking anneal before rolling, because titanium readily forms solid solutions with oxygen. However, it must be emphasized that the titanium industry is yet to attain the size of the steel industry, and there could be other process reasons for cracks to occur; these might be over-heating and lack of pass control. Problems also arise from the use of equipment and personnel unfamiliar with titanium processing instead of working the ingots in dedicated facilities. Reference: 1. Y.K.Rao, Stoichiometry and Thermodynamics of Metallurgical Processes, Cambridge University Press, 1985; pp 375 3 CHAPTER TWO LITERATURE REVIEW 1. BASIC TITANIUM METALLURGY* 1.1. PHASES OF TITANIUM - THE p»a TRANSFORMATION 1.1.1. Temperature of transformation Titanium has two elemental crystal structures - the high temperature bcc phase (P) and the low temperature hep phase (a).The transformation temperature is 882.5°C (0.59 T m ) for the pure element. Addition of alloying elements stabilize the alpha or the beta phase making it possible to have alloys having alpha, alpha-beta and beta structures. Aluminium, oxygen and nitrogen stabilize the alpha phase; hydrogen, vanadium, molybdenum, iron, chromium and manganese stabilize the beta phase. 1.1.2. Rate of transformation The rate of the a=>p phase transformation is very rapid. Iodide-pure Ti has been shown to have completed the a=>P transformation in 200 micro-seconds under pulse-heating rates [2]. At lower heating rates, the transformation takes no more than 60 milliseconds [14]. For the completion of the P=>ct transformation, the corresponding supercooling prior to transformation is never more than 150°C, at quench rates. Thus, it is impossible to suppress the beta to alpha transformation [1]. 1.1.3. Nature of transformation Rapid quenching from the high temperature p phase does not prevent the p=oct transformation, which suggests diffusionless shear transformation of the martensitic type. The crystal structure of the martensite is hexagonal close-packed and thus similar to that of the * The heading format is different for this chapter. 5 distinct topics are dealt under 5 major headings, with sections and subsections for each of them. 4 equilibrium alpha phase normally obtained on air-cooling. The hardness change on formation of a' is not as obvious as that obtained in hardened steel because of the large beta grains [6]. The alpha phase formed by slow cooling has the same orientation relationship and habit planes as the acknowledgedly martensitically transformed alpha, which led McQuillan [1] to suggest that the transformation consists of athermal nucleation of the alpha phase followed by diffusion driven growth. Random nucleation is ruled out, since out of the many orientational relationships possible between the beta and the transformed alpha, only one single orientation results. 1.2. THE TITANIUM-OXYGEN PHASE DIAGRAM Commercially pure (CP) titanium invariably has some oxygen alloyed with it, which raises the aop transformation temperature and causes interstitial strengthening. The increase in strength has been well characterized and can be used as a measure of the interstitial content in the CP titanium. The Ti -O phase diagram is shown in Figure 2.1. Oxygen alloying is an integral part of the formulation of "commercially pure" titanium, as are small additions (>0.5 wt.%) of Fe and/or Cu. Atomic Percent Oxygen 0 10 20 30 40 M 60 70 2200 -i 1 • : i , i , 1 L . 0 S 10 15 20 25 M 33 40 45 Ti Weight Percent Oxygen . Figure 2.1 Phase Diagram of the Ti-0 system [ 9] 5 1.3. EFFECT OF OXYGEN ON PHASE TRANSFORMATION 1.3.1. Effect of oxygen on transformation temperature Increased oxygen favours the phase transformation to the alpha phase. When pure Ti is cooled from the beta region, the transformation, which normally begins at about 850°C, (the lower temperature being attributed to beta stabilizing effects of Fe and Si), starts at approximately 970°C, when the oxygen content is raised to approximately 0.75% [3]. 1.3.2. Effect of oxygen on rate of transformation When the solid solution oxygen content in titanium is increased, the rate of alpha nucleation increases, as observed by Bumps, Kessler and Hansen [1]. The rate of transformation and the subsequent growth of the alpha phase also increases, similar in effect to that of lowering the temperature [3]. In a study on Ti-Mo alloys done by DeLazaro and Rostoker and reviewed by McQuillan [1], oxygen accelerated the rate of transformation, shortening both the initiation and completion times of the P=>a transformation. 1.3.3. Effect of oxygen on nature of transformation The micro structure obtained on cooling titanium containing dissolved oxygen is different from that obtained on cooling unalloyed titanium, as shown in Figure 2.2. On quenching from the (3 phase field, needle-like martensitic structures are observed. High purity titanium subjected to a similar treatment show large alpha grains having extremely serrated boundaries. When the cooling rate is lower (10°C/min), large alpha plates are produced [3]. Presence of oxygen encourages the form, regularity and definition of needle-shaped alpha morphology, or lamellar alpha plates, depending on the cooling rate [1]. 6 Figure 2.2. Comparison of the microstructures of high-purity titanium (left) and titanium having 1% oxygen (right) after quenching from the P phase field [1] 1.4. THE OXIDATION OF TITANIUM 1.4.1. The Oxidation Kinetics of Titanium [4] Oxidation of titanium, as defined by gain of mass of oxygen vs. time curves, are initially parabolic, and then linear.The kinetics of the parabolic part can be represented by ( m / S ) 2 = K t (2.1) where m is in milligrams S is in cm2 and K is the Arhennius Rate Law constant K = 1.59xl08exp(-26,935/T) (2.2) Subsequently, this changes to a more rapid reaction governed by a linear rate law , of a form ( m / S ^ K t (2.3) where K = 1.59 xl0 8exp(-22,529/T) 7 It is suggested that the parabolic stage.represents the dissolution of oxygen in the hep alpha for the formation of the oxide layer. This is followed by the linear stage. Disintegration, or removal of the formed oxide layer causes the kinetics to revert to the parabolic state, which leads to the inference that the reaction changes to linear kinetics only after the effect of the oxide layer is removed, or ceases to become effective. In such a case, then, the linear kinetics represent further dissolution of oxygen in the solid metal by diffusion through the oxide layer when it is no more effective as a barrier [4]. 1.4.2. Change of Oxygen Kinetics with Temperature From Figure 2.3 it is seen that the gain of mass is higher for higher temperatures. At higher temperatures, the parabolic oxidation kinetics is of a shorter duration, as seen in Figure 2.4, which suggests a shorter duration of oxide layer formation. Also, it is to be noted that for the lower temperatures, most of the oxygen gain goes towards the development of the oxide scale, as seen in Figure 2.5. Thus, at higher temperatures, most of the oxygen is used for solid solution formation, whereas for lower temperatures, it is used for oxide scale formation. W% I 1 • 1 1 ' 1 "1 " I • ' '' W1 10* to 1 10* tt* Time (min) Figure 2.3 Kinetics of oxidation of titanium [4] 8 104 E E 10J 10 r • • —r-TTTj I fs 700 750 800 850 875 Temperature (°C) Figure 2.4 Duration of the parabolic kinetics as a function of temperature [4] w 1 1 t 1 1 1 « oxide layer weiaht • \ \ i l \t • \ \ \ \ \ \ ^ ^ 1 -N \ X 1 - fl 700 750 800 850 875 Temperature (°C) Figure 2.5 Thickness of the oxide layer and gain of mass of the samples at the transition of the kinetics from parabolic to linear as a function of temperature [4] 9 The oxide layer is gray and very adherent in the initial stages, up to about 20 microns thickness. Above this thickness the oxide layer, saturated with oxygen, consists uniquely of rutile and turns white, friable and less adherent. The initial changes are shown in Figure 2.6. The surface oxide lattice is similar in structure to the original unoxidized metal's surface structure (between 700°C and 875°C). The fractographic study of the oxide layer shows evidence of stratification. It is noted that the change of kinetics occurs after the development of several stratas, and is not dependent on the formation of the first stratum. The strata which forms first remains at the surface and confirms the hypothesis that oxidation takes place at the metal-oxide interface. Also, the fact that the parabolic kinetics continue after the first oxide stratum is formed indicates that the oxygen diffuses through the rutile layer progressively working towards saturation. ISO err ? 100 § I 50 T T Brownish compact oxide fragile and non adherent Brownish friable oxide and somewhat adherent f— Grey oxide and very adherent —j J I I I 700 750 900 050 «7S Temperature ( ° c ) 1 Figure 2.6 Adherence and colour as a function of oxide thickness and temperature [4] 10 1.4.3. Influence of Scale Thickness on Oxidation Kinetics While different researchers have maintained that the role of the oxide layer ceases after its saturation to rutile, Dechamps et al. [4] observed that when samples were replaced in the heating atmosphere after removal of scale, the rate of oxidation accelerated at all temperatures, suggesting that the scale plays some role as a protective coating. However, the differential thermal expansion of the scale and the oxygen-rich titanium underneath, coupled with the lack of plasticity of the surface layer, causes the scale to lose its cohesion. When the oxide layer becomes several strata thick, decohesion spreads across the ensemble. This results in a change of kinetics to linear and is accompanied by an increase in the speed of oxidation. Simultaneously, • the oxide becomes white and powdery. The oxide layer may regenerate to form a protective coating which would eventually rupture again. If an equivalent mean thickness of the oxide layer is to be considered as a barrier to diffusion its thickness should be approximately 3 microns [4]. This calculation takes into consideration all kinetic processes which are occurring. 1.4.4. Mathematical Model With the above assumption, the initial parabolic oxidation kinetics is diffusion controlled in a semi-infinite environment with a moving interface. For further simplification, four assumptions were made : 1) Oxygen was the only diffusing species - the diffusion of titanium was neglected. This was evident from comparison of the activation energy for the diffusion process with that required for the diffusion of titanium across the oxide layer - the values were found to be very different. 2) The diffusion coefficient of oxygen across the oxide layer was independent of the oxygen concentration. It must be mentioned, however, that only the alpha phase was examined. The problem would have been much more complicated if the temperature was higher and both the alpha and the beta phases were involved. 3) The kinetics of oxidation was known. 4) Steady state equilibrium existed at the the interface and was not the rate limiting step. 11 This stated, the problem can be solved if the concentration of oxygen at various interfaces responsible for the inward flow of oxygen were known. Three interfaces were under examination: (i) the metal-oxide interface: Several researchers have agreed on a value of 25 at.% oxygen in titanium at this metal-oxide interface [4]. (ii) the atmosphere-oxide interface: The problem was more difficult for this interface, since actually two different species with varying mechanisms of diffusion were involved. The mathematical model of Danckwert [4] was utilized, and it was assumed that the oxygen was in a condensed state at the interface of the oxide-atmosphere, with an equivalent concentration of 1.4g/cc at the oxide-atmosphere interface. This value is not much different than the concentration of oxygen in rutile, 1.5g/cc. (iii) On the basis of thermodynamic studies by Kubaschewski and Dench [4], it was hypothesized that the oxygen concentration at the metal-oxide interface was very low, less than 1% of that at the oxide-atmosphere interface. The interface conditions are elucidated in Figure 2.7. Oxide Metal F i g u r e 2.7 Schematic representation of the oxygen concentration at the interfaces [4] 12 With these assumptions, a value of the coefficient of diffusion of oxygen in the alpha titanium was obtained by Dechamps and Lehr [4] as D2 =.408exp(-47,040/7?r) cm 2s _ 1 ( 2 4 ) This result was in agreement with previous works, which are presented in Figure 2.8. noo looo 900 eoo 700 T 1 1 1 1 1 I 1 1 1-j(V ,«* <) Figure 2.8 Diffusion coefficients in the alpha phase calculated by different workers Similarly, they calculated the coefficient of diffusion of oxygen across the oxide layer to be D,-0142exp(-45,080//?r) cmV 1 (2.5) 2 -1 With these equations, the diffusion coefficients were calculated to be 0.003 cm s and 2-1 0.00013 cm s for Dj and D 2 respectively at 875°C. These values have been calculated from the linear portion of the oxidation kinetic curve, where the rate limiting process for oxidation is the kinetics of diffusion of oxygen in the Ti and across the strata [4]. This information is useful in estimating the role of the oxide layer as a barrier to diffusion and for predicting the thickness of the solid solution oxygen layer with time at temperature. It is pointed out that the diffusion estimation becomes complicated at higher temperatures in the beta phase field where oxygen diffusion takes place in three different materials - in the oxide layer, in the oxygen stabilized alpha phase, and in the beta phase. 13 2. ROLLING OF C P . TITANIUM The lack of literature on breakdown rolling of titanium or its alloys is a consequence of the fact that the breakdown of ingots has been done traditionally by forging processes. The major application for titanium rolling has been for making strip, sheet or plate products [20]. Sheet rolling procedures were established by the U.S. Department of Defense Sheet Rolling Program operative from 1955 till the early 60's. In that work, seven important alloys of titanium were studied for aerospace applications, and hot rolling schedules and procedures were established for them. The Titanium Metallurgical Laboratory, now known as the Defense Metals Information Center, also published several compilations of hot working detail for titanium alloys. As a result, considerable technical know-how has been generated for processing titanium alloys. However, this information could not be exploited for this study since work on CP Ti has been minimal compared to the work on the titanium alloys, since there was no application of CP titanium in defence. Also, most of the available data on CP titanium is on sheet-rolling [20]. 2.1. ROLLING IN DIFFERENT ATMOSPHERES Hot rolling of titanium requires that the metal be heated at elevated temperatures, above 900°C. However, at these temperatures, it is susceptible to gas pick-up. Since titanium is one of the most reactive metals [38], and shows a tendency to embrittle when worked in air [12,13,33,34,37], attempts were made, mostly in Russia during the mid 60's, to carry out the rolling of titanium (and other such reactive metals such as niobium and tantalum) in protective atmospheres such as a vacuum or inert gas. 2.1.1. Vacuum Only a vacuum better than 10~2 Pa (10"5 mm Hg) ensures protection against gas diffusion into the titanium - confirmed by mechanical tests and chemical analysis of the gas content in the samples as shown in Figure 2.9 [37].Samples were held at different temperatures, and apart from the increase in weight at increasing temperatures, the rate of oxygen diffusion increased sharply at temperatures above 900°C, coinciding with the phase transformation to the more open bcc structure (beta phase) [37]. 14 Figure 2.9. Comparison of properties of titanium hot rolled in different atmospheres [37] The low output capacity of vacuum rolling mills can be attributed to the difficulties in maintaining a high vacuum in large spaces, the tendency of the metal to adhere to the rolling surfaces during vacuum rolling, and the difficulty in manipulating the equipment under vacuum conditions [35]. 2.1.2. Inert atmosphere Figure 2.10 shows that argon atmospheres, apart from the vacuum atmospheres, are effective against oxygen pick-up, the efficiency of one over the other depending on the vacuum pressure [40]. 15 Temperature (C) Figure 2.10. Comparison of gas pick-up with temperature between argon and air [37] Inert atmosphere chambers, such as one made by Universal Cyclops Co. (U.S.A), was 2360 m-3 in capacity filled with argon of 99.9% purity and had an operator in a pressurized suit within the chamber. The Soviets also had a similar chamber - "Atmosfera" of capacity 44 m ,^ -and also had argon as the inert medium [35]. The amount of gas pick-up was studied for the reactive metals - titanium, vanadium, niobium - after ingots had been heated to 1200°C and rolled. It was found that for Ti, the gas saturation was negligible but increased at higher temperature. The slight increase in weight of the Ti specimens indicated a certain oxidation (or nitridation) of the metal. When compared with the results of another researcher, Pavlov [36], it was evident that the weight increment could be attributed to the use of commercial purity of argon as compared to using high purity argon. The only advantage of having a thin gas-saturated surface layer on the metal is that it prevents the metal from sticking to the working tool. This is important for metal processing which involves high-temperature plastic working [35]. The summary of the authors was that employing an inert atmosphere to hot-work metals of high chemical reactivity prevented scale formation and cracking [35,37]. However, maintaining such an atmosphere for daily commercial practice remains costly and certainly impractical for the large CP titanium ingots weighing over 10 tons. 16 2.2. RECRYSTALLIZATION AND RECOVERY DURING HOT ROLLING Apart from the surface layer embrittlement associated with the hot rolling of titanium in air, suitable refining conditions for the ingot structure are required. Studies have been done recently on the softening mechanisms of titanium during hot rolling, the results of which are summarized below. 2.2.1. Recovery and recrystallization temperatures The beta phase has a very low deformation activation energy, Q, [7,12] signifying rapid diffusion rates in the beta phase. Hence, the beta phase is fairly resistant to recrystallization, undergoing static recovery instead at the forming temperatures. Any recrystallization beginning at the grain boundaries or at shear bands proceeds slowly, because the interior of the grain has an already recovered structure. At 900°C, the beta grains are elongated on rolling, the recrystallized grains form the fringe at the beta grain boundaries, and complete recrystallization cannot occur even at increased reductions [39]. Recrystallization can only be enhanced by high strains and high temperatures after deformation. High strain rates allow the rapid accumulation of dislocations necessary for formation of the critical nuclei required for dynamic recrystallization (DRX) and higher temperatures bring about accelerated formation and growth of the nuclei. The recrystallization nose in the alpha region exists at 800°C, or slightly above. Hence, at these hot-rolling temperatures, the hardness does not change significantly with increasing reduction. However, whether the recrystallization is dynamic or post-dynamic, both in the beta and the alpha phase, is not defined [39]. The recrystallized grain size, d, in the alpha phase can be expressed by the formula similar to the behaviour of austenitic and ferritic stainless steels during hot rolling, as given by the equation [39] d = 1 0-0.102 8-0.017 z-0.012d i0.397 ( 2 6 ) 17 where Z is the Zener Hollomman parameter calculated with a Q value of 185 KJ/mol s is the strain dj is the initial grain size Dependence of the final grain diameter on strain suggests that recrystallization is static. Recrystallization often occurs at the sites of lamellar deformation twins at higher strain rates of 10 s"1 or more'[3 9]. The deformation twins in the hep alpha phase are lamellar above 700°C, and lenticular below 700°C [39]. Here recovery dominates and reduces the strain hardening [12]; recrystallization does not occur until significant strains are imparted. As the reduction increases, the concentration of lamellar deformation twins increases, slip deformation continues, but a recrystallized structure appears only at 40% reduction or higher. A good reason why recovery occurs is the fact that the c/a ratio of the h.c.p. phase is 1.59, which allows both basal and prismatic slip. Figure 2.11 is a schematic representation of the above softening processes. Recrystallization initiates in twins and twin intersections. With decrease in temperature, deformation continues by slip and lenticular twin formation, but recrystallization occurs.only locally, as shown in Figure 2.11 and Figure 2.12. Initial state Hot rolling Restoration process $-*a transformation or recrystallization or ^ _ Deformation twins Static recrystallization Recovery Figure 2.11. Schematic Diagram of the softening processes [39] I I E Reduction 0% 15% 43% 66% 1,173 K 1 . S i ^ • ^ ^ ^ ^ ^ ^ ^ m 1.073 K ' ^ ^ ^ ^ ^ m oo ' i i ' . \ '" s*v ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ 100/itn Figure 2.12 Change in structure with hot rolling temperature and reduction [39] At even lower rolling temperatures (700°C - 600°C), the hardness increases with increasing reduction and a uniform recrystallized structure cannot be obtained. Concurrently, the microstructure exhibits a recovered structure characterized by lenticular deformation twins, as shown in Figures 2.11 and 2.12 [39]. Below 600°C recrystallization is not observed [39]. Recovery and recrystallization do not occur as readily in structures with coarse grains since deformation twins and slip bands are not uniformly introduced[39]. The specimens made from the ingot have a coarser structure and tend to exhibit recrystallization only at higher temperatures compared to the slab material [39]. However, it is very important to note that these rolling practices are employed only after the initial breakdown of the ingot by forging. When direct rolling of the as-cast ingot has been attempted, cracks have developed. Hence, the standard method used for break-down of ingots until recently has been forging. It is thus considered necessary to examine the forging practice since it gives a background to the understanding of the hot-working characteristics of CP Titanium. This information should help to establish guidelines for direct rolling. Moreover, many of the established practices for forging could be necessary requisites for designing a direct rolling procedure. 19 3. INGOT CONVERSION - FORGING Titanium ingot conversion to slab is done by the primary titanium metal producer or by the forger. The ingot working procedures, the macro-structural, microstructural and mechanical property requirements are usually based upon the specific forging involved, the forging equipment, and cost considerations. The final forging structure and mechanical properties are subject to specifications by the forger, or are negotiated between the forger and the metal supplier. In addition, certain general, military or other governmental specifications, such as AMS 2380, impose specific requirements on the manufacture of the titanium (alloy) ingot, the fabrication process to make the stock, the macro and microstructural requirements for the forging stock, and the necessary validation tests. Primary fabrication converts the cast ingot into a mill product which can then be used for secondary manufacture of parts and structures. For this initial breakdown operation, opert-die forging is used; closed-die forging is later used to obtain complex shapes having close dimensional control [20]. 3.1. STAGES IN INGOT CONVERSION The surface or "skin" of the ingots to be fabricated is first machined off using large lathes. This "scalping" removes the volatile materials condensing to the side of the mould, which otherwise would cause defects in the mill products. After a protective coating1 is applied to reduce surface contamination, and also to provide lubrication, the ingot is heated and press-forged into billets. Following the press-forging, any surface defects are removed by machining or grinding - a process called "conditioning" - since any exposed cracks or defects in Ti will not heal during subsequent forming. If considerable conditioning is required, the billet is ground while still hot (315 - 590° C) or reheated before grinding [20]. This machining process is frequently omitted for reasons of yield in the case of small ingots with high surface quality. 1 The sample to be forged is cut from the forging stock, and preheated after coating with a ceramic to retard oxidation [42]. 20 Forging stocks are frequently reheated for forging, sometimes with ceramic coatings to reduce oxidation, as it is believed that if too low temperatures occur, then surface cracks or chill-cracks will be formed [12]. After each stage of deformation, the product is again conditioned, i.e., the bloom or billet is ground at dull red heat [20,42] to remove surface defects. The bloom may require reheating to about 700°C halfway through the operation. Hot grinding is followed by annealing at 700°C for two hours, then sand blasting and inspection for remaining defects; the latter are subsequently removed by machining. These billets or blooms may be further deformed on a blooming mill or two-high reversing mill, and fabricated into sheet bars, which are then extensively surface-conditioned. Finally, they may be hot-rolled, and then finished by cold-rolling in continuous bands to 0.125 - 0.250 " sheets [42]. 3.1.1. Lubrication The lubrication serves to increase die-life since the adherent oxide-scale on titanium is very abrasive and severely erodes dies [20,42]. The use of a glass-type lubricant serves a two-fold advantage - it reduces oxidation and provides lubrication [34,42]. Protective coating products such as Markal CRT, Du Pont J-400, and Crucible's No. 50 have been reported to reduce surface contamination, and provide lubrication during forging [34]. 3.1.2. Removal of contamination and defects The heavy rutile scale (Ti02) obtained due to oxidation is removed first by descaling in a molten salt-bath of sodium hydroxide, nitrate, or chloride. Beneath this scale is the hard, oxygen enriched layer which can be .005 to .025" thick, depending on soaking times. It is difficult to remove this material by machining. Pickling forgings in hydrofluoric acid (2 to 3% aqueous solution) removes the layer at the rate of .001 inch/min. To minimize hydrogen pick-up, 30% HN0 3 is also added, which however, decreases the removal rate by 50% [34]. Stringent ultrasonic inspection is frequently performed on Ti alloy forgings as a critical part of the overall quality assurance. Ultra Sonic Inspection (USI) of the billets is often preferred to the USI of the final forged shape because of the more regular geometry of the billet. Therefore, 21 Ti alloy billet or bar stock is typically ground or machined to remove all defects and to prepare the surface for ultrasonic inspection. 3.2. FORCE CONSIDERATIONS In open-die forging, the primary force is compressive and more localized compared to closed-die forging, where the forces are exerted on all surfaces of the workpiece. Also, the open-die requires a relatively small tool force because of reduced frictional contact [20]. For cylindrical ingots, two-sided lateral forging is preferred to axial forging because the initial cast structure has pronounced microstructural and chemical inhomogeneity. To obtain a homogenizing effect, more extensive deformation is required; hence, deformation must occur with a change of axis [30]. 3.2.1. Effect of Strain Rate For high strain rates of 10/s, typical of mechanical or rapid strain rate forging equipment, and similar to strain rates obtained on direct rolling, the flow stress of CP Ti is about 20% higher than that of 4340 alloy steel. Figure 2.13 compares the forging stresses of some Ti alloys with that of 4340 alloy steel. Also, CP Ti has flow stresses very similar to the alpha beta alloys for the given strain rate (10 s"1). Compared to steel or many aluminium alloys, titanium and its alloys are much more strain-rate sensitive. For a decrease of strain rate from 10 s"1 to 0.001 s"1 the flow stress can be reduced up to 10 times. It is obvious that it is advantageous to forge titanium (alloys) at low strain rates in order to reduce the resistance to deformation. However, at these slow strain rates, if the conditions of forging are non-isothermal, then the increase in flow stress due to a decrease in temperature far outweighs the low strain rate advantage. Thus forging processes have to be isothermal, which can require die-heating equipment. Else, the workpiece has to be reheated. 22 (0 CO o Ti-C.P. at 800 "CIWO-F) • Ti-6AI-4Vat900'C(1650*F) a Ti-13V-11Cr-3AI at 900 *C (1650 *F) A Ti-6AI-6V-2Sn at 900 *C (1650 *F) o Ti-8AI-1Mo-1Vat 955 *C (1750 "F) • Ti-10V-2Fe-3AI at 815 "C (1500 'F) ? 4340 at 1205 "C (2200 °F) 27/s Figure 2.13 Flow stress of commonly forged Ti alloys at 10 s"1 strain rate compared with 4340 alloy steel forged at 27 s'1 strain rate [43] 3.2.1.l.Strain Rates used for forging Ram speeds usually range from 0.03 - 0.1 m/sec, and for mechanical presses, are ordinarily higher at the start of forging and decrease towards the end [20]. Hammers and forging machines typically have strain rates above the equivalent of 6 m/sec [20] and are used for secondary forging purposes. 3.2.2. Effect of temperature Deformation pressures are very sensitive to temperature, more so than in low-alloy steels, with the alpha alloys being most temperature sensitive [12,20]. Next are the alpha-beta alloys, and finally the ductile beta alloys. Figure 2.14 compares the forging pressures of different materials at different temperatures which shows that the forging pressure required for titanium and its alloys is higher. 23 for forging The forging temperature range is between the upper temperature which is limited by excessive oxidation or grain growth, and the lower temperature which is limited by increasing flow stresses, lower ductility, and lack of recrystallization [7,34], and in the case of open-die forging, the desire to avoid excessive cracking and/or other surface quality problems. Figure 2.13 Effect of forging temperature on required pressure for 3 Ti alloys compared to 4340 alloy steel [43] 600 400 Q. S3 200 1 N 1 / 1 'Ti-13 1 /-l1Cr- 3AI — Ti-•8AI-1K o-W Ti-6A •4V ^ V 340 st ( 80 60 cn cn CB C H20 700 800 900 1000 1100 1200 1300 Temperature, °C Initial breakdown is done in the beta phase because the bcc structure is more ductile and forging pressure requirements are less [20]. Final forging is done at temperatures below the beta transus too for several reasons. Firstly, since the dynamic recoverability of the beta phase is high, it forms a stable sub-structure not amenable to recrystallization at moderate strains, leading to grain growth which cannot be corrected [7,12,20]. Secondly, forging above the beta transus cause beta embrittlement [20]. Also, billets are intended for further forging, rolling or extrusion and go through a grain refinement process. This technique utilizes titanium's characteristic to recrystallize when it is heated above the beta transus. However, grain boundary acicular alpha forms when most alloys are cooled from above the beta transus temperature [20,34]. Working in the alpha-beta region during continuous cooling through the beta transus may be needed to 24 eliminate the formation of the grain boundary alpha [34]. Modern processes use substantial amounts of working below the beta transus to produce billets with refined equiaxed structures [20,34]. If there is untransformed beta after the deformation is complete, it turns to acicular alpha [20]. Ideally, 50% of the forging reduction should be made below the beta transus to ensure optimum properties in the forged piece [20]. This reduction is carried out high in the alpha region to allow greater reduction and improved grain refinement. Lower temperatures in alpha region are not desired because pf higher forming pressures and dangers of surface rupturing [12,34,20]. The ingot breakdown temperature at the beginning of forging is 955 - 980°C for CP Ti Grades 1 - 4 . The beta transus temperature for the grades is raised from 885°C depending on the alpha stabilizing impurities present and varies between 900 and 950°C. The recommended temperature limit for forging is CP Ti is between 815 °C and 900°C. The lower temperature is the absolute lower limit at which the forging may be completed. Table 2.1. Forging temperatures for CP titanium [34] Process Temperature (°C) Breakdown of ingot (start) 955 - 980 Breakdown of ingot (forging and finish) 815-900 Severe forming (usually closed-die) 500 - 700 Hence, stock is preheated to 700°C and then heated to the forging temperature and forged as quickly as possible. Preheating avoids thermal-stress cracking as a result of too rapid heating, while reducing the time the metal is at the forging temperature thereby minimizes surface contamination and grain growth. Titanium alpha alloys have narrow forging limits, particularly below the beta transus temperature. Accordingly, small reduction steps and frequent reheating must be incorporated in the forging schedule[12,34]. As an approximate rule, forging temperatures should be about Pt r - 28°C for alpha-beta forging, P t r + 42°C for beta forging introducing the obvious requirement for accurate knowledge of Tp(r and accurate process temperature control. 25 of die temperature Die heating reduces surface chilling, which might cause excessive cracking. Open-die forging equipment is usually preheated to between 150 - 260°C. 3.2.3. Contamination during processing Carbon, nitrogen, hydrogen and oxygen can impair the properties and quality of the forging. Contamination by interstitials raises the beta transus and the strength level substantially [20]. Further, the oxygen and nitrogen form a stabilized and hardened alpha layer at the surface which is prone to cracking [12]. Hydrogen is absorbed at forging temperatures, diffuses inwards, raising the hydrogen content of the entire forging, which can lead to hydrogen embrittlement [20]. To avoid this, forgings are performed in a slightly oxidizing atmosphere when oil or gas fired furnaces are used [20,42]. and Nitrogen At temperatures above 540°C in air, oxygen reacts to form a surface scale on titanium. The scale consists almost entirely of rutile (Ti02), and a hard alpha-rich sub-surface layer [20]. Generally, when heating in air, 2 hr is the longest time permitted during secondary forging, which reduces to 20 min at 870°C [34] for subsequent finishing operations. These time/temperature boundaries are cumulative and include the total time that the metal is at the temperature for all forming operations. These limits can be somewhat relaxed when scale-inhibiting coatings are applied to prevent oxidation [34]. For open-die breakdown forging, the demands for surface quality are not as stringent. 4. THE CONSTITUTIVE EQUATION It was important to develop a constitutive equation for the processing of CP titanium in the parameter domain encountered during its direct rolling. For quantitative analysis of any deformation process using a technique such as the finite-element method [23], or even for macroscopic nominal estimations of processing variables, it is necessary to have a continuous 26 function relating the processing variables strain, strain-rate, temperature and stress in the domain where these processing conditions are experienced [23]. The conditions of strain, strain-rate and temperature in a hot-working process vary with position and time in a sample. Experimental techniques such as tension, compression, torsion and indentation tests are used to obtain stress values at various combinations of temperature, strain and strain-rate, and an equation relating all the process variables is fitted to the experimental data. This equation can be an empirical or semi-empirical relationship, and may attempt to signify the microstructural processes that are associated with hot working [8]. Since these processes are many, and some have predominance over the others at particular processing regimes, it is difficult to have a unique relationship for all parameter regimes [8]. Attempts have been made by several [23] to characterize this relationship in the strain hardening part of the curve, or in the steady state portion for different materials. Some authors, such as Dadras and Thomas [18] also attempted three distinct equations for three separate processing regimes for Ti-6242. 4.1. T Y P E S O F C O N S T I T U T I V E E Q U A T I O N S The three most commonly used Arhennius-type constitutive equations for predicting steady state flow stresses are given below : S = A exp(n CT) exp(-Q / RT) Exponential Equation-—(2.7) 8 = A"CT n exp(-Q / RT) Power Equation (2.8) 1 n 8 = A sinh((XCT) exp(-Q / RT) —Sinhyperbolic Equation-—(2.9) where A, A', A" and a are constants n, n\ n" are related to strain-rate sensitivity Q is the activation energy of deformation R is the gas constant 8 is the strain rate T is the absolute temperature 27 The constants A, a, n and Q must be experimentally determined The Q term in the equation is supposed to represent the activation energy of the process when steady-state flow stress conditions prevail. The choice of the particular equation depends on its ability to fit the data for the required domain. For example, the power law is applicable at low stresses and the exponential law at high stresses in the case of alpha iron. Only the sinhyperbolic model is valid for the entire stress range [7]. Since the above equations are applicable for steady-state conditions, they are not a function of strain. However, there are certain deformation conditions where the flow stress is a function of strain. To predict the stress values at a given value of strain, then, the constants A, n and Q for the above equations have to be modified according to the level of strain. Rao and Hawbolt [23] developed an equation of the form A i Y = - ^ - + C i (2.10) s 1 where Aj, Bj and C/ had to be calculated separately for each of the constants A, n and Q (represented by F in each case). This equation ensured that a single equation could be used at all strains for all hot-working processing conditions. 4.2. AVAILABLE EXPERIMENTAL DATA ON CP TITANIUM For the alpha phase, the data reported by Santhanam and Reed-hill [47] was obtained for strain-rate change tests in the temperature range -196°C to 700°C for very low strain rates - 2.7 X 10"4 s"lto 2.7 X 10"5 s"l. Experimental data on CP titanium has also been generated by Doner and Conrad [5] in the temperature range 375 - 875°C, at strain rates of 3.3 X 10"5 s"* to 1.7 X 10~2 s~l employing both constant strain rate and strain rate cycling tests. A value of n varying between 4.3 and 4.5 has been reported for the power law exponent in the dislocation creep domain (very low strain rates) in the temperature range from 0.4 - 0.6 T m , and an activation energy between 185 - 242 KJ/mole [12,13]. The flow curves were described to have the classic shape for dynamic recovery similar to Al and alpha-Fe. McQueen and Bourell[7,12] 28 have reported that the stress-strain relation in the hep a phase follows the sinhyperbolic law and the Arhennius temperature relationship for low strain rates - 10"V to 10"1 s"1. However, the validity of these laws are still to be tested for higher strain rates. Deformation tests have also been carried out in the P phase region, for temperatures between 900 and 1000°C at strain rates 5.5 X 10"2 s'1 to 5.5 X 10"5 s"1 by Oikawa et al. [10], who reported a stress-exponent n of 4.1. From the works of several authors summarized by McQueen and Bourell [7,12], the value of the power-law exponent was 4.0 in the beta phase region, and the activation energy, Q, was.quite low as compared to the other bec metals; in the range of 137 - 171 KJ/mole. An activation energy between 185 - 242 KJ/mole was also reported for the alpha phase. The only experimental work of the deformation of pure beta titanium in the range of temperatures and strain rates applicable for direct rolling was done by Buhler and Wagener [11]; they performed tests in the temperature range 900°C to 1300°C for strain rates between 0.25s"1 to 64s"1. However, neither the deformation characteristics of beta titanium, nor the stress exponent or activation energy are well established, because of lack of similar experiments and for reasons outlined in the subsequent section. 5. DEFORMATION-MECHANISM MAPS There is a requirement for predicting the material response to an intended manufacturing process. There are two approaches for doing so - the materials modelling approach and the deterministic approach [23]. Processing maps developed by Gegel et al. [23] utilize the Dynamic Materials Modelling approach. These maps represent the instantaneous dissipation characteristics of the material. No apriori knowledge or evaluation of the atomistic mechanisms occurring are required[23]. However, constructing such a map requires experimental data obtained over at least 5 decades of strain rate and a temperature range of 200°C. Since our interest is in strain rates in the range of breakdown rolling, data points have to be obtained from literature for cases spanning the regime of interest to construct an appropriate Processing map. However, on examining the literature it is apparent that different compositions of Ti have been used by different workers. Since the flow 29 stress in Ti is very sensitive to the oxygen content[l 1], and since the efficiencies plotted in these dissipation maps are so sensitive to the stress values, care has to be taken in collecting appropriate published data. In this study, the deformation mechanisms prevalent for processing conditions similar to breakdown rolling have been examined. In deformation-mechanism maps, plastic flow is determined by the kinetics of processes occurring on an atomic scale. These atomistic processes make certain deformation mechanisms dominant at a given processing combination of strain, strain rate and temperature. These mechanisms are [22]: a) Elastic collapse at ideal shear strength b) Low temperature plasticity by dislocation glide c) Low temperature plasticity by twinning d) Power law creep by dislocation glide or a combination of glide and climb e) Diffusional flow Ashby maps [22] are constructed from experimental data, fitted to rate equations which describe these deformation mechanisms. 5.1. ASSUMPTIONS IN CONSTRUCTING DEFORMATION MAPS [22] The change of the microstructural variables with change in deformation conditions is not well understood. Hence, this construction is limited by the assumption of constant structure throughout the process, i.e., Si=SQ (2.11) where Sj=Starting or initial structure $0= final structure dSj/dt = 0 . (2.12) This assumes that grain size, dislocation motion, recovery, grain boundary sliding remain constant throughout the process. Fracture is suppressed, if necessary, by applying a sufficiently large hydrostatic force. The state variables are defined at any stated condition. 30 e =/(State Variables, Microstructural Variables, Temperature, Stress) (2.13) Thus the above equation, combined with equations (2.12) and (2.13), reduces to s = /(Temperature, Stress) (2.14) 5.2. DESCRIPTION OF DEFORMATION MAPS The typical deformation-mechanism map is a diagram with normalized stress(o7p) and homologous temperature (T/Tm) as the two axes, where CT = flow stress u. = bulk shear modulus T = Temperature (in Kelvin) = Melting point of the material (in Kelvin) The map is divided into different regions demarcating the predominant deformation mechanism operative at specific combinations of temperature and stress. Constant strain-rate contours are marked as well; the latter is not additional information, but is obtained from equation (2.11). The maps could also be developed with strain rate and homologous temperature as its two axes, or with normalized stress and strain/However, only the stress-temperature map is useful in fitting constant strain-rate data. Mohamed and Langdon [15] have constructed these maps with structure parameters such as grain size as one of the variables. However, Ashby and Frost [22] believing that the microstructural variables cannot be controlled as external variables, respect equations (2.11) and (2.12), and assume constant structure. 5.3. POWER-LAW BREAKDOWN REGIME The stress levels appropriate to breakdown of Ti are above 10'^p [22], where the power-law breaks down. In this region the plastic flow mechanism is supposed to be a transition from climb control to glide control. An empirical description of it proposed by Jonas, Sellars and Tegart [17] has the form of the exponential equation stated earlier. The sinhyperbolic equation is an alternative more encompassing equation which can accommodate data both in the power-law and breakdown regime [22] with* the help of an added parameter a, where a is a constant indicative of their relative strengths in a particular material. 31 Since it is logical to anticipate a change in stress due to the temperature dependence of the shear modulus of the material, a better experimental fit has been found with a temperature corrected, shear modulus dependent, sinhyperbolic equation of the form [22] e = A[sinh(—)] n exp ( -Q /RT) (2.15) where ct'=p'p0 To correlate this equation with the rate equation for power-law creep developed by Ashby and co-workers, they propose an equation of the form s = A 2 [ s i n h ( ^ ) ] n ' ( ^ £ ^ ) (2.16) Z Lt kT where A2'=A2/a'n where A2 is the Dorn constant n is the power law exponent Z)eff is the lattice diffusion co-efficient at that temperature p is the temperature corrected shear modulus b is the length of the Burger's Vector k is the Boltzmann's constant n' is the power-law exponent a ' is a fitting parameter dependent on experimental values obtained and suggested by Ashby to be 300 for a-Ti This equation is nearly identical in form to the sinhyperbolic equation (apart from having different constants) except that it apparently has a temperature term in the denominator. The strength of this equation is that it is identical to the power-law rate equation proposed by Ashby and Frost [22] for low stresses. The limitation is that it uses the fitting parameter, a', to serve two purposes [22] - to determine the stress level of the power-law breakdown, and as a constant which determines the magnitude of dependence of the strain on stress, whereas the sinhyperbolic equation has two separate constants - a and A - for these two different conditions. 32 5.4. DEFORMATION MAP FOR CP TITANIUM Deformation maps were constructed by Ashby and co-workers [18] using their equations to demarcate between different deformation regimes on the normalized stress vs. temperature space, and using the experimental data of various researchers to place the boundaries of these regimes on the graph. The deformation map for CP Ti is given in Figure 2.14. Figure 2.14. A Deformation mechanism map for CP Ti with a grain-size of 100 microns, showing the data [22] 200 I 0 TEMPERATURE, (°C) 200 COO 600 800 KX30 — 1 1 , 1 1 t CO CO LU CO cr < LU X CO 10' Q LU CO _J < cr 10 o fqfc~ DRAG CONTROLLED—IQ^H PLASTICITY ^ CONTROLLED rPLASTICITYW 1200 K00 — I L _ 1600 TITANIUM d =100Lim DIFFUSIONAL FLOW) BUM LER ANO WAGENER0965) Grao*1, BUHUER AND, WAG£NER(1965) Grodrf | MALAKONOAIAH (I960) 195-7 200>jm\ DONER ANO C^o^ /.«--.. »  v RAO.RAO ANO GRIEST ET AL 10 0 (BOUNOARYl KLATTICE) —I^slQ 1 ^ -——KT3 ^-iffi V. \ \ - ^ c • ^FUSIONAL FLOW \ ( LATTICE) 0-2 Q.i 0.6 0.8 HOMOLOGOUS TEMPERATURE, Vj 1.0 M 33 5.4.1. Limitations of the material constants It is useful to know the limitations of the state variables used in the equation for the power-law breakdown regime to improve the understanding of the theoretically calculated values. The lattice diffusion coefficient for cc-Ti obtained by Malkandaiah [22], has an anomalously low activation energy for creep, Qcteep, which cannot be reconciled with experimental results. His test samples had an oxygen content of 0.3 wt.% and a grain-size of 200-400 microns and were tested in the difftisional flow regime. Similar values of Screep w e r e found by Conrad and Doner [5]. Also, the a value for power-law breakdown in Figure 2.2 is taken from the experimental results of Conrad and Doner [5]. Buhler and Wagener state that the low temperature plasticity of CP Ti is a very strong function of the interstitial content and could vary by a factor of 3.5 depending on its interstitial content. It is noted that the data used in the formulation of these maps have varied oxygen compositions. The temperature dependence of the shear modulus in the beta phase is assumed, since the shear modulus is known at only one temperature in the beta phase. Furthermore, the flow stress data in the power-law regime is very sparse - with only Buhler and Wagener's [11] experimental data shown on the map. The a value is assumed to be 10^, and the power law exponent calculated by fitting the above data with the assumed a value. The diffusion coefficient is taken to be 152 KJ/mole, as reported by de Reca and Libanati [16]. 34 References 1. McQuillan M.K., Metallurgical Reviews, (8), 1963, pp 41-104 2. Cezairliyan A, Miiller A.P., Journal of Research of the National Bureau of Standards, Vol. 83, No.2, March-April 1978 3. Dubrov V.A., Fiz. Metal, metallovedenie, 24, No. 2, 1967, pp 316 - 320 4. Dechamps M , Lehr P., Journal of the Less-Common Metals, 56 (1977) 193 - 207 5. Doner M . and Conrad H., Metall. Trans. A, vol 4, 1973, pp 2809-2817 6. Polmear I.J, 2nd Ed., 1989, Metallurgy of the Light Metals, Edward Arnold Publication (Division of Hodder and Staughton Ltd.), pp 218 - 220 7. McQueen H.J. and Bourell D.L., Journals of Materials Shaping Technology, Vol.5, No.l , 1987, pp. 53 - 57 8. Rao K.P. and Hawbolt E.B., Journal of Mater, and Tech., vol. 114, Jan. 1992, pp 116 - 123 9. "Alloy Phase Diagrams"- ASM Metals Handbook vol.3, 10th Edition, pp 2.324 10. Oikawa H., Nishimura K., Cui M.X., Scripta Metallurgica, vol.19, pp.825-828, 1985 11. Buhler H and Wagener, Bander Bleche Rohre, vol.6, pp 625 - 677, 1965 12. McQueen H.J. and Bourell D.L., Journal of Materials Shaping Technology, Vol.5, No.3, 1988, pp 163 -189 13. McQueen H.J. and Bourell D.L., Review of Hot Workability of Metals and Alloys, Journal of Metals, Sept. 1987, pp 28-35 14. Parker, R., Trans. Metall. Soc. (AIME) 223, 1545-1549, 1965/3 15. Mohamed F.A. and Langdon T.G., Metall. Trans. 5, 2339, 1974] 16. de RecaN.E., Libanati C M . , Acta Metall., 16, pp 1297,1968 17. Jonas J.J, Sellars C M . , Tegart W.J.M.M., Metall. Rev., 14, 1, 1969 18. Dadras P., Thomas J.R., Metall. Trans. A, vol 12A, Nov 1981, pp 1867-1876. 20. Gerds A.F. et al., Deformation processing of Titanium and its alloys - NTIS 634 -141 21. Asby M.F., Acta Metallurgica, vol 20, July 1972, pp 887-897 35 22. Frost H.J., Ashby M.F.: Deformation-Mechanism Maps - The Plasticity and Creep of Metals and Ceramics,, Oxfordshire, New York, Pergamon Press, 1982. 23. Gegel H.L., Malas J.C., Doraivelu S.M., Shende V.A.: Metals Handbook, 9th Edition, Vol. 14, pp 417-438, Materials Park, Ohio ASM. 24. Raj R., Metall. Trans.A, 1981, vol. 12A, pp 1089-97 25. Prasad Y.V.R.K., Gegel H.L., Malas J.C., Doraivelu S.M. et al, Metall. Trans. A, vol. 15A, 1984, pp 1883-92 . 26. Chakravartty J.K., Prasad Y.V.R.K., Asundi M.K., Metall. Trans. A, vol. 22A, 1991, pp 829-36 27. Venugopal S., Mannan S.L., Prasad Y.V.R.K., Materials Science and Technology, vol. 9, October 1993, pp 899-906 28. Srinivasan N. , Prasad Y.V.R.K., Mat. Sci. & Tech., Nov 92, 8(3), pp 206-213 29. Dieter G.E.- Mechanical Metallurgy - 3rd Ed., McGraw-Hill Book Company, 1976, pp531-583 30. Kopylov V.N., Metallovedenie i termicheskaia metallov (Metal Science and Heat Treatment), vol. 33, Sept-October 1991, pp 703 - 707 31. Gegel H.L. - "Optimizing Hot workability and controlling microstructures in difficult of process high strength and high temperature materials" - Dept. of the Air Force Washington DC, Pat. Appl - 698 728, AD 586, Feb 1985 32. Prasad Y.V.R.K, Gegel H.L., Doraivelu S.M. et al, Metall. Trans. A, vol 15A, Oct. 1984, ppl883 33. Grant N.J., et al., Investigation of fracture in connection with hot deformation processing of Metals, M.I.T. contract, pp 2, NTIS AD 648 451 Dec 1966 34. Donachie M. , Titanium - A Technical Guide - ASM International, 1988 pp 37-56 35. Borisov A.Y, Shekalov V.I., Gertsyk M.A., Aleksandrov A.A - Soviet journal of Non-ferrous metals", May 67, vol.8 (5), pp 99-101 36. Pavlov I.M., et al., Tsvetnye Metally, 1962, No. 7; No.5,1963 36 37. Gurevich,Y.b., Bashchenko A.P., Orzhekhovskii V.L.- Some specific features of hot rolling refractory metals in a vacuum and in an atmosphere of inert gas - Tsvetnye Metally, 1965, No. 7; 76-81 38. Rao Y.K., "Stoichiometry and Thermodynamics of Metallurgical Processes", Cambridge University Press 1985, pp 375 39. Hayashi M. , Yoshimura H., Ishii M. , Harada H., "Recrystallisation behaviour of commercially pure titanium during hot rolling", Nippon Steel Techical Report No. 62, July 1994 40. Lillie C.R., McPherson D.J., " Development of Titanium Base Alloys of High Strength and Toughness" - Armour Research Foundation, Chicago, Illinois, NTIS AD-952342, March 1955 42. Hart R.V. - "Research and Development in Titanium Processing and conversion to mill products" NTIS AD - A952259, September 1956 43. Kuhlman G.W., Forging of Titanium Alloys, Metals Handbook, 9th edition, Volume 14, Forming and Forging, pp 267-283 45. Layner D.I., Bay A.S., Slesareva Y.N., Dnepro M.J. - "Some features of oxidation of Titanium", Metalloved., 20, No.6, 864-867, 1965 46. Kofstad P., Anderson P., Krudtaa O., Journal of Less Common Metals, 3, No.3, 89 (1961) 47. Santhanam A.T. and Reed-Hill R.E., Metall. Trans. A, vol. 2, 1971, pp 2619 37 CHAPTER THREE SCOPE AND OBJECTIVES The principal objective is to determine possible reasons for the failure of CP titanium when it is subjected to deformation conditions similar to those encountered during industrial breakdown rolling. The samples are annealed and deformed under conditions similar to the current industrial rolling practice. The samples are then to be examined for possible reasons of failure. Since it is known that titanium forms an oxygen stabilized alpha case when treated under such conditions, the alpha case is to be characterized in terms of its formation, composition and mechanical properties. The overall aim is to determine whether or not the presence of this oxygen contaminated surface layer is responsible for the cracking of the ingots during direct rolling, and whether there could be other reasons for failure to occur. A secondary but necessary objective of this study is to develop a constitutive equation for the hot working of commercially pure titanium. The parameter regime for direct rolling is predicted from a simple model which would also identify the critical rolling parameters. The results of the deformation experiments are to be compared against the flow-stresses predicted by deformation maps to test for the latter's general applicability for the given hot-working process. 38 C H A P T E R FOUR SIMULATION O F T H E B R E A K D O W N R O L L I N G PROCESS 4.1. OBJECTIVE OF THE MODEL This model describes the stress, strain and temperature state of the material on rolling. Since the processing parameters for breakdown rolling of Ti have not been definitively established, a prefatory study of the influence of the processing parameters on the variables of the rolling process is called for. In particular, it is necessary to identify those parameters which critically influence the success of the process. However, it is recognized that laboratory simulations are useful but cannot replace industrial rolling tests. 4.2. DESCRIPTION OF DEFORM DEFORM®* is a commercial 2-D finite-element method based software package which can analyze the deformation mechanics of plastically deforming objects. DEFORM® uses the flow formulation approach (i.e., elasticity is neglected), and an updated Lagrangian procedure -the finite element mesh is embedded in the material and moves with it as it is being deformed. It has the capability of modelling large-scale deformations, and calculating the values of the processing variables over the whole range of deformation. 4.3. ASSUMPTIONS The simplifying assumptions that have been made for modelling the breakdown rolling of Ti can be classified into two distinct types - simplifications in the process or simplifications in the processing variables. * DEFORM® is a registered trademark of UES Company Ltd., Columbus, Ohio, USA. 39 4.3.1. Simplifications in scale Firstly, it is assumed that the ingot has an uniform temperature of 900°C at the point of entry of the first rolling pass. Any differences in temperature that may exist between the surface and centre of the ingot are neglected. Secondly, the first reduction in the breakdown rolling is taken to be 20%, although this may change from one schedule to another, and from one company to another. The third assumption is that the width of the workpiece is much greater than its thickness, and hence the workpiece experiences plane-strain during rolling. Finally, symmetry conditions have been assumed along the central plane of the workpiece, and only the upper half has been modelled. 4.3.2. Simplifications in the processing variables The temperature of the roll is assumed to be constant at 50° C, and the heat affected zone in the roll extends 10mm into its depth. It is assumed that the roll is perfectly rigid, and the workpiece is plastic. The emissivity of the surface of the workpiece is set at 0.85 and the interface-shear coefficient of friction between the rolls and the metal is set at 0.85. Also, the heat transfer coefficient between the workpiece and the roll is kept at a constant 40 Wm" K" and the roll-speed at 15 r.p.m.. The fraction of the energy of deformation which is converted to heat in the workpiece is assumed to be 95%. Finally, the effect of phase transformation on the enthalpy of the material is taken to be negligible compared to the chilling effect of the rolls and the heat of deformation. 40 4.3.3. A Critique on the Assumptions Firstly, many of the assumptions were made because of lack of relevant industrial data, rather than attempting to simplify the problem. Secondly, certain assumptions, such as limiting the thickness of the heat-affected layer of the roll to 10 mm, were made as a trade-off between computing expense and marginal information gain, as it is of not much interest to determine the temperature profile in the roll. Finally, the rigorous nature of the DEFORM software package imposes certain restrictions on the modelling. To minimize computing time, only those variables known to significantly affect the rolling process have been examined. The assumptions are summarized in Table 4.1. 4.4. ASSESSMENT OF THE VALUES OF THE PROCESS VARIABLES Analysis of the process variables during deformation in the roll gap were done at four points in the upper half of the slab on the same plane vertical to the rolling direction, as shown in diagram 4.0. X-axis Figure 4.0. Diagram showing the finite-element mesh and the 4 positions examined 41 Table 4.1. Parameters incorporated in the model. Those with asterisks have been varied. Material property data Rationale Reference Die (Roll) temperature* Assumed to be 50° C at steady state from Industrial experience and steel data Workpiece temperature* Temperature drop from 1100 C in 2 minutes Estimates by workshop personnel [2],[3] Workpiece Strain* Reduction of 20% for the standard model [2] Material property data Rationale Reference Workpiece Material CP Grade IITitanium Thermal Expansion Function of temperature [4] Thermal conductivity Function of temperature [4] Specific Heat Function of temperature [4] Emissivity 0.85 [5] Density Constant = 4510 kg/m3 the difference due to expansion is minimal - increases computation time tremendously [4] Rolls Material Gray cast iron with chilled outer surface - the rolls used in steel mills where CP Ti is also rolled [4] Thermal conductivity Constant (at 50°C) [4] Specific Heat Constant (at 50°C) [4] Emissivity Constant (at 50°C) [4] Density Constant (at 50°C) [4] Die characteristics : Roll speed* 10-15rpm [2] Interface Data : Friction Factor* Expect high value because unlubricated, and remnants of scale Heat transfer coefficient* Value for stainless steel, which has the most processing parameter similarities with Ti. 42 4.4.1. Effective strain As shown in Figure 4.1, the effective strain is approximately 0.5 near the surface due to redundant work as compared to the overall strain of about 0.23. Not only is the strain higher, the strain gradient is very steep; the decrease in strain in the first 10 mm depth is approximately the same as the difference between positions (3) and (4). The points under study physically enter the roll-gap at 0.12 s and leave the roll-gap at 0.50 s which explains the flattening of the curves after that time, denoting no further change in strain. In addition, the rise in strain is sharper closer to the surface as compared to points towards the centre. 0.600 2 0.500 H 0.400 e I CO OJOO 3 4 0.100 0.000 0.000 0.600 Time Figure 4.1. Effective strain vs time at positions 1, 2, 3 and 4. 43 4.4.2. Effective Strain Rate The overall strain rate of the process at the positions 1, 2, 3 and 4 is shown in Figure 4.2. The figure shows that the strain rate at the surface points reaches a peak of 8 s"', compared to that at positions (3) and (4), which reaches a peak of 1 s"^ . Also, the increased strain rate, though it lasts for approximately 0.1 s, is responsible for the increased strain which is shown in Figure 4.1. This means that the material experiences a sudden high strain rate which causes the strain to reach the maximum strain attained in the process. 10.00 - I 9.00 -8.00 -7.00 -«J 6.00 -rt CC £ 5 0 0 " 3= W 4.00 -3.00 -2.00 -1.00 -0.00 " f 0.000 0.100 0200 0.300 0.400 0500 Time Figure 4.2. Effective Strain Rate vs time at positions 1, 2, 3 and 4. The strain rate rise for the surface elements(l,2) is at the point of entry as compared to the inner elements (3,4). 44 4.4.3. Temperature The effect of roll chilling is to decrease the temperature by about 300°C for the element closest to the surface - position (1), but only about 40°C for position (2). This shows that the severity of the chill is confined to a very small depth. Thus, the depth to which the temperature of the material is quenched below the alpha transus is of the order of 10 mm. The result is consistent with the fact that titanium has poor thermal conductivity, about half that of steel. 823.0 -I 89Z5 860.0 H 827.5 H 795.0 H Q . E 762.5 H CD 730.0 H 697.5 H 665.0 H 632.5 H 600.0 0.000 I 0.100 0200 1 0.300 Time —I— 0.400 —I— 0.500 1 0.600 Figure 4.3. Temperature vs time for positions 1,2,3 and 4, showing the chilling effect of the surface elements on entering the roll. 45 4.4.4. Stresses The modelled rolling stresses, as shown in Figures 4.4 and 4.5, indicate that the maximum stress is compressive in the through-thickness (Y-direction), and the surface stresses are greater than the stresses obtained deeper within the workpiece. The stresses in the x-direction can be compressive or tensile at the surface points -positions (1) and (2), depending on their position relative to the neutral point; the compressive stress peak occurs at the neutral point. Figures 4.4 and 4.5 show that the centre of the workpiece, does not feel these subtleties, and experiences a general tensile stress throughout the rolling process, in response to the general compressive stress of the y-component. The chosen plane strain condition gives us the z-component as shown in Figures 4.6 and 4.7. The shear stress due to rolling has a peak at 42.5 MPa, and is most acutely felt at the surface points(l,2). However, in all cases the changes in the variation of stress during the rolling cycle are much more abrupt for the surface points. The study of stress is complicated because it involves effects due to chilling of the material. This causes a rise in the flow stress, apart from increased stress effects due to redundant work done near the surface, and stress effects due to frictional forces. 46 84.J -j sa 3 -) Figure 4.4. The x-component of the stress vs time for positions 1,2,3 and 4. The x-component is along the direction of the rolling. 2 0.000 0.100 0-2O0 OJOO 0.400 OJOO Time Figure 4.5. The y-component of the stress vs time, for positions 1,2,3 and 4, this being the compressive force that causes the rolling reduction. 47 Z STRESS H O -1 Figure 4.7. The shear component of the stress vs time for positions 1, 2, 3 and 4. 48 4.5. EFFECT OF VARYING PROCESS VARIABLES ON THE ROLLING PROCESS 4.5.1. Heat transfer coefficient (Original run value = 40 W/nr/IQ The heat transfer coefficient was varied from 30 to 60 W/m2/K. The variation of the degree of chill, as shown in Figure 4.8, is less than 75°C even when the heat-transfer coefficient is doubled from 30 to 60 W/m /K. It must be mentioned, however, that the heat transfer coefficient can be a complex function depending on the roll-pressure and the interfacial conditions, factors which have not been considered in this model. Also, Figure 4.4 shows that the lowest temperature has a defined minima in each case before the effects of thermal recovery from the inner material become noticed. This also suggests that for this particular case (with a draft of 100mm and roll speed of 15 r.p.m), the extent of maximum chilling is more dependent on the thermal diffusivity of the material than the time of contact with the rolls. The results in Figure 4.9 show that lower heat-transfer co-efficients allow higher strains at the surface elements due to easier plastic flow, but the effect is relatively small. 20 25 30 35 40 45 50 55 60 65 70 Heat Transfer co-efficient (W/m'/K) Figure 4.8. Effect of heat transfer coefficient on material chill depth. 49 ure 4.9. Effect of heat transfer co-efficient on strain. 50 4.5.2. Effect of varying the rolling temperature (Original run = 900°O ' As the rolling temperature changes from 900 to 950°C, slightly less roll-chilling is observed, as shown in Figure 4.10. A possible reason may be that at higher temperatures less compressive forces are required, less factional forces are produced, and hence roll-gripping becomes difficult. Secondly, the degree of chill is almost identical to the original case, suggesting that the thermal recovery limits the degree of chilling. Thirdly, the compressive (Y-component) deforming forces required are significantly lower at the higher temperature, by about 20%, as shown in Figure 4.11. 300 Mid-centre Surface element Figure 4.10. Effect of rolling temperature on material chill. 51 Temp=900 C • 1 5 Temp=950 C a. S - 5 5 - 1 1 5 - 1 3 5 Figure 4.11. Comparison of the compressive rolling forces required at different rolling temperatures 4.5.3. The effect of varying the interface shear friction (Original run: Interface shear co-efficient m = 0.85) On decreasing the interface shear coefficient of friction to m = 0.7 from m = 0.85, most chilling of the surface is decreased by 50°C. This is due to the increased time of contact with the rolls because of the reduced friction in gripping the material by the rolls. However, if the heat-transfer coefficient was set to be a function of roll-pressure, which is a very realistic assumption, then the chilling effect would tend to decrease with decreasing 52 friction, and not increase as observed in Figure 4.12. The x and y components of the stresses (tensile and compressive stresses respectively) decrease only marginally - the peak compressive y-component of stress falls from 125 MPa to 120 MPa as the friction coefficient falls from 0.85 to 0.7, subject to the limits of accuracy in plotting the DEFORM software. For the given reduction of friction to m < 0.6, the rolls are unable to grip the material. 350 I Friction m=.85 iFriction m=.7 SURFACE ELEMENT MID-CENTRE ELEMENT Figure 4.12. Effect of decreasing the interface shear friction coefficient from 0.85 to 0.7 on the temperature decrease experienced at the surface and mid-centre element. 53 4.5.4. Effect of degree of reduction (1000 mm reduced to 800 mm) When the draft is halved, the strain disparities between the surface and the centre elements still exist, as shown in Figure 4.13. The surface elements still strain considerably more than the nominal strain. In fact, the reduction in strain and in redundant deformation is not very apparent from the strains of the surface elements. The strain-rate is halved, in keeping with the reduced strain and its maxima limited to the entry point for the surface elements, whereas the strain rate is more or less constant for the central regions. The chill is significantly less near the surface for a reduced draft, as shown in Figure 4.14, and can be attributed to the direct thermal response obtained at the surface. Figure 4.13. Effect of degree of reduction on the chill of the material element at the surface and mid-centre. Figure 4.14. Effect of degree of reduction on Strain in the material elements at the surface and the mid-centre. 300 Mid-centre element Surface element Surface element Mid-centre element 54 4.5.5. Effect of varying the roll temperature (Original run 50gC) The effect of the temperature profile through the workpiece in the roll-gap is compared for cases where the roll temperatures were kept at 100°C and 200°C. The results are shown in Figure 4.15. The effect of roll-heating did not create a significant rise in the temperature profiles. The minimum surface temperature increased by approximately 60°C as the roll temperature increased from 50°C to 200°C , but the inner temperatures were unaffected. No significant effects in the strain rates or the stress components at the centre or surface were associated with increasing the roll temperature from 50°C to 200°C. 900 850 800 ^T^^^r^----^^^^ MID-CENTRE ELEMENT ' \ \ ^ ^ " ^ " H ^ - ^ _ _Roll_Temp = 200C \ \ Roll Temp = bUC \ \ \ \ \ \ , \ \ . \ \ . \ \ . \ \". \ V N \ \ \ \ • N \ s\ • \ s.*. \ - j ^ SURFACE ELEMENT Roll Temp = 200C Roll Temp = 100C Uoll lei up - LiUL.' « 750 o CL E t-700 650 600 0.1 0.2 0.3 Time (s) 0.4 0.5 0.6 Figure 4.15. Effect of varying the temperature of the rolls from 50 to 100 to 200°C on the chilling of the material at the surface and mid-centre. 55 4.6. DISCUSSION AND SUMMARY For the hot roll processing parameters, the nominal strain is fixed by the reduction. The variation of strain, then can only be in the readjustment of strains experienced by the individual elements. The modelling results of Figure 4.1 showed that the strain is about twice as large in the surface elements as compared to the inner elements due to the inhomogeneous nature of the deformation. For this reason, rolling processes should be designed in keeping with the ductile limit of the surface elements, rather than the nominal strain of the process. The strain rate reaches a maximum for the surface elements at the point of entry, and at a much later stage for the inner elements coinciding with the friction-hill point of maximum stress as shown in Figure 4.2. Hence, different strain rate vs. time responses are obtained for the surface and internal elements. The depth of the surface chill, i.e, the depth to which the material is immediately quenched below the alpha phase transus, is of the order of 10 mm. Surface chilling is a very strong function of the time of contact with the rolls. Just after entry, it is estimated that the surface loses 100°C for every 0.1 sec of contact with the roll. It is therefore very important to lower the time of contact with the rolls; this is more important than the alternative of roll-heating. Heating the rolls well above 200°C would be equivalent to extracting the material from the roll-gap 0.05 seconds earlier. Increasing reduction implies decreasing the total time of contact with the roll, and hence reducing chilling for the same roll-speed. Reduced friction might delay the gripping by the rolls, and hence result in greater chilling. The surface chilling is arrested when the material exits the rolls or when the thermal recovery affects the surface, whichever is earlier. The latter case is characterized by the arrest ,of the surface chill followed by an increasing temperature profile. Obviously, the magnitude of the heat transfer coefficient also significantly affects the degree of chilling, but its effect seems to become less above 50 W/m /K, as shown in Figure 4.8. 56 The redundant deformation criterion applied normally to rolling situations does not define the conditions at the point of entry where the forces are maximum for the surface elements. The Figures 4.4, 4.5, 4.6 and 4.7 all show a maximum for the surface elements within 0.2 seconds. (The delay is the time taken for the surface element to arrive at the roll-gap). After this initial anomaly between the surface and central elements, no further major disparities between their stresses are observed. Of the force components examined, the compressive y-component is not destructive, and the tensile x-component does not increase beyond a seemingly harmless 32 MPa. The shear stresses are also small, but they build up abruptly in the surface elements at the point of entry. Theoretically, the most effective way to reduce the rolling pressures, apart from increasing rolling temperatures, is to reduce friction. However, the decline in stresses for a 15% reduction in friction is only a 5 MPa drop in compressive load. It should be emphasized that the hot rolling simulation has been done with a few simplifying assumptions. The effect of phase transformation during rolling could be significant enough to alter some of the observed trends. However, as stated above, the purpose of this study is to obtain a window for the processing variables during rolling and to assess how they may affect the rolling process. This has been accomplished with the DEFORM® software. References 1. William Hosford and Robert M. Caddell, "Metal Forming - Mechanics and Metallurgy", Prentice Hall Inc., 1983, pp 116-13. 2. Private Communication from Toho Titanium Company, Ltd. 3. Private Communication from Titanium Hearth Technologies Ltd. 4. Deform® Users' Manual version 4.0 - Material Database, Appendix B, Scientific Forming Technologies Corporation, Columbus, Ohio. 5. Robert Siegel and J.R. Howell, Thermal Radiation Heat Transfer, 3rd Ed., Hemisphere Publishing Corporation, 1982 57 CHAPTER FIVE EXPERIMENTAL WORK 5.1. CP TITANIUM The starting samples for all the deformation experiments were cylindrical samples 10 mm in diameter and 15 mm in height machined from an extruded Grade II CP Titanium rod with the composition given in Table 5.1. Table 5.1. Composition of the as-received CP Titanium ELEMENT WT. % Oxygen 0.11 Nitrogen 0.01 Hydrogen 0.0005 Carbon 0.01 Iron 0.06 Titanium Balance Several samples to be deformed in compression and analyzed for cracking were annealed in air. The other samples were deformed in the as-received condition. The latter set of samples served as controls specimens to compare with the former ones, as well as to provide hot-working data for developing the constitutive response of the CP Ti. 5.2. SAMPLES SUBJECTED TO OXYGEN ANNEALING 5.2.1. Annealing of the samples Cylindrical samples were annealed for 24 hours at 1100°C in air in a vertical tubular furnace with Super Kanthal as the heating element. The samples were then air-cooled. 58 The choice of the time and the temperature of heating was made to simulate the industrial practice*. The aim was to develop a oxygen-rich surface layer on the specimen, and to test its susceptibility to cracking during deformation. 5.2.2. Surface oxide On completion of the anneal, the samples were coated with a friable oxide layer. This oxide layer flaked off or was removed by abrasion and examined for thickness, colour and composition to estimate the extent of the surface layer's role as a barrier against diffusion of oxygen reaching the underlying base metal. 5.2.3. Oxygen diffusion thickness Tests were performed to measure the concentration profile of the oxygen in solid solution as a function of the depth from the surface. A combination of methods was used: (1) visual inspection using an optical microscope and SEM, (2) indirect measurement using Vicker's microhardness measurements, and (3) direct oxygen concentration analysis by Wavelength Dispersive Spectrometry (WDX) analysis. Each method had its own merits and weaknesses, the combinations complementing each other. examination The annealed samples were cold mounted and ground on silicon-carbide paper of successively finer grits, then fine-polished with 5 and 1 micron diamond using a light polishing pressure to minimize cold working [1]. The samples were then etched with a reagent recommended for near-alpha and alpha alloys (composition given in Table 5.2) and viewed using the optical microscope. The time required for etching varied between 1 and 2 minutes. Further etching and polishing details are available in reference [1]. * Titanium hot-rolling schedules are still in a stage of development, and hence they change frequently. There is no unique soaking schedule at present - schedules vary from 6 hrs to 18 hrs soaking at 1100°C, often combined with soakings at other temperatures. 59 Table 5.2. Composition of etchant [1] Composition Volume % concentrated HF 2 concentrated HNO3 2 Distilled water balance 5,2.3.2.Hardness measurements After polishing, Vickers micro-hardness measurements were made as a function of depth from the surface using a Micromet® Hardness Indenter. The load used for the indentation was 100 grams. Measurements were made every 50 microns from the surface until a uniform hardness was attained. Since the hardness values exhibited considerable variation, a number of measurements were taken at each distance from the surface, and the average value was reported. analysis The oxygen concentration measurements at positions similar to where microhardness measurements were made, were obtained using the WDX analysis. The oxide scale was also analysed for oxygen using WDX to establish the nominal chemical composition of the oxide. It was noted that the WDX results for light elements are not very reliable. These results were therefore used to support the composition estimates made with the help of the other methods. 5.3. SAMPLES SUBJECTED TO VACUUM-ANNEALING An as-received sample was annealed in vacuum for 24 hours at 1100°C. The sample was then air cooled, and the microstructure of its radial cross-section was examined. 60 5.4. COMPRESSION TESTING The axisyrnmetric compression test is one standard test used for establishing the stress-strain response of a material at high temperatures. Without introducing necking problems as are obtained in the tensile test or material reorientation such as that occurring in the torsion test, a large strain can be achieved by compression testing. In hot tensile testing, the low rate of strain hardening induces necking at e ~ 0.1, which raises the subsequent strain rate at the neck by 10 to 20 times. The disadvantage of axisyrnmetric compression testing is barrelling which occurs due to lateral frictional restraint on the end faces. The resulting inhomogeneous deformation causes incorrect stress values after a strain of approximately 0.7 has been reached [2,3,4]. The surface friction can be reduced using graphite at the anvil/sample interface, or using cylindrical samples with small diameter to height (D/H) ratios. However, with increasingly slender specimens the chances of compressive buckling during compression increase for D/H> 0.5 [A]. 5.4.1. Compression testing for the as received (non-annealed) samples Compression Testing using the Gleeble® 1500 Thermomechanical Simulator The Gleeble® 1500 (manufactured by Duffers Scientific Incorporated, Troy, New York) is a computer controlled servo-hydraulic thermomechanical simulator. The specimen is placed in the Gleeble and held between deformation anvils, with tantalum foils at either end to prevent welding of the specimen to the anvil during deformation. The specimen is resistance-heated by a thermocouple feedback controlled alternating current. The equipment is capable of deforming a sample at a prescribed programmed strain-rate up to a specified strain at a particular temperature in a controlled surrounding environment. Figure 5.1 is a schematic diagram of the specimen support in the Gleeble test chamber. 61 Ta foil to stop welding to anvil Thermocouples Crosswise Strain Measurement (LVDT) Supports connected to load cells Anvil Anvil supports Figure 5.1. Schematic diagram of the holding chamber of the Gleeble. A thermal gradient exists along the specimen axis due to heat-transfer through the Inconel 718 anvils to the water-cooled anvil supports. Obviously, the thermal gradient is smaller for shorter specimens. Resistance heating produces isothermal diametrical test planes with a maximum temperature at the mid span of the specimen [3]. The sample temperature was measured using a chromel-alumel thermocouple of 0.6 mm diameter wire inserted into a drilled hole to the centre of the test sample. The thermocouple was held in place by deforming the entrance hole. This temperature was measured to determine the mean temperature at which the deformation took place, since it was thought that deformation heating could be substantial and could play a significant role in the softening of the material. In addition, a chromel-alumel thermocouple was spot-welded at the circumferential surface of the.mid span of the test speciment to measure the surface temperature; the measurement of both the surface and centre temperature provided a measure 62 of the radial thermal gradient during testing. The surface thermocouple also served to control the sample termperature by a feedback heating system and maintained the desired temperature during the thermal cycle and subsequent deformation. The test chamber inside the Gleeble was initially evacuated using a mechanical pump to 1 Pa and then backfilled with high-purity argon (99.99%). The complete thermal cycle could be programmed into the Gleeble control system. A clip-on quartz rod cross-wise Linear Variable Differential Transformer (LVDT) was mounted across the diameter of the test sample at the mid-span of the specimen to measure the diametrical strain (C-Strain) during testing. A similar LVDT was also mounted to measure the length change during hot deformation, but was not as reliable a measure as the cross-wise strain since it also measures the deflection of the tooling [3]. The compression tests were performed using constant ram velocity as it is easier to program compared to variable ram velocity, the mean strain rate at each temperature being given. For most tests the strain rate varied by +10% for a strain up to 0.7. The C-strain device measures the maximum cross-section and hence, the flow stress value calculated from it is less than the average flow stress calculated from the L-strain measurements which are based oh the constancy of volume. The C-strain response is used to calculate the true stress at the temperature controlled mid span plane and is very reliable when barrelling is minimum, which is considered negligible up to a strain of 0.7. schedule In the test chamber containing an argon atmosphere, a heating rate of 10°C/sec was obtained by resistive current control to heat the specimens to a holding temperature; the sample was then held at the temperature for 2 minutes to ensure uniform temperature conditions throughout the specimen. The temperature control during the isothermal holding period was excellent, maintaining the sample temperature within ±2°C of the desired temperature. A prescribed strain rate was applied to the specimen till it attained a true strain of 1.0. During this deformation, the surface and the centre temperature (in °C), the cross-wise and length-wise strain, the stroke-rate (in cm/s), and the load (in kg) were recorded at a frequency 63 varying between 5 and 500 Hz, depending on the stage of the processing schedule and the strain rate employed. The annealed samples were deformed at different combinations of strain rate and temperature as summarized in Table 5.3. Table 5.3. Deformation conditions for the non-annealed samples. * denotes the combinations for which the tests were performed. Temp Strain Rate ( s"1) ( ° Q 0.01 1 5 10 750 (a) * * * 800 (a) * * 850 (a) * * * 900 (p) * * * 950 (P) * * * * 5.4.2. Compression Testing for the annealed samples The oxide on the air annealed samples was flaked off and the sample polished at both ends. It was noted that the radius of the cylindrical samples decreased by about 0.25 mm due to material loss as oxide scale. For these samples, a platinium - platinium rhodium (Pt/Pt-10%Rh), Type S thermocouple was intrinsically spot welded to the oxygen contaminated specimen surface since the chromel-alumel thermocouples did not weld easily to the sample surface. The thermocouple wires were 0.25 mm in diameter. No central thermcouple was used during the test of the air-annealed (oxygen contaminated) specimens in the argon environment. 64 Schedule The processing schedule employed for the air-annealed (oxygen contaminated) material is summarized in Table 5.4. Table 5.4. Deformation conditions for the annealed samples Temp ( O Q Strain Rate ( s"1) 0.005 0.01 0.1 1 10 750 (a) * * * 800 (a) * * * 850(a) * * * 950 (P) * * * 5.4.3. Deformation of the Vacuum annealed sample The vacuum annealed sample was deformed at 850°C and a strain rate of 10s"1 to a strain of 1, in a manner similar to the other samples. It was then allowed to air cool from the deformation temperature. 5.5. EXAMINATION OF DEFORMED SAMPLES The deformed samples were cut axially, mounted, ground and polished for microstructural examination. References 1. Metallographic Technique for Titanium and Titanium Alloys - ASM Handbook, Vol.2, pp. 140-141 2. Dieter G.E., Mechanical Metallurgy , McGraw Hill Book Company, 2nd Ed., pp. 544 - 45 3. Rao K.P. and Hawbolt E.B., Journal of Engineering Materials and Technology, Vol.114, January 1992, pp. 116-123 4. Dieter G.E., Metals Handbook, 9th Edition, ASM International, pp. 373-386 65 CHAPTER SIX RESULTS AND DISCUSSION 6.1. OXYGEN DIFFUSION RESULTS 6.1.1. The oxide scale Cylindrical samples were air-annealed (oxygen-contaminated) for certain temperature-time combinations to study the growth of the oxide layer, and the alpha phase underneath. Table 6.1 lists the temperature-time combinations that were analyzed, the thickness of the oxide layer and the alpha case, and the visual appearance of the oxide. The first two cases are shown in Figure 6.1. This picture compares the visual appearance of three samples - one oxygen-contaminated at 1100°C for 24 hours, another at 750°C for 24 hours, compared with the as-received (oxygen uncontaminated) sample. There was no difference in the visual appearance between the samples of the second, third and fourth cases of Table 6.1. The scale of the sample that was oxygen-contaminated at 1100°C for 24 hours was analyzed using WDX with the results given in Table 6.2. It is understood that there are problems associated with the estimation of such light elements as oxygen and nitrogen. Table 6.1. Comparison of the physical characteristics of the oxide scale and its thickness after different annealing treatments. Temp (°C) Time (hr) Thickness(um) Scale/a-phase Comments 750 . 24 Very adherent, not removable 1100 24 700 / 275 Scaly and friable; yellowish exterior, white interior 1100 48 950/400 as above 1100 60 1000/500 as above 66 Figure 6.1. Picture comparing the physical appearance of three samples - from left to right - soaked at 1100°C for 24 hours, soaked at 750°C for 24 hours, untreated sample Table 6.2. Composition of oxide scale by weight analyzed by WDX measurement. Element Weight % Oxygen 20.82 Nitrogen 1.78 Titanium Balance 6.1.2. The alpha-stabilized surface layer Under the rutile surface layer, oxygen is dissolved in solid solution in the titanium. Microscopic examination of the air-annealed (oxygen contaminated) specimen reveals the presence of a distinct boundary, suggesting an interface between two different phases, as seen in the Figure 6.2. The alpha case also has cracks originating at the surface and travelling radially inwards. 6.1.3. The Widmanstatten pattern formation The morphology of the microstructure obtained just below the alpha case was of the Widmanstatten type as shown in Figure 6.2. The orientation of the platelets varied from one grain to another. 67 Figure 6.2(a). Micrograph of a portion of the cross-section of the CP Ti rod, after annealing at 1100°C for 24 hours, showing the oxygen-stabilized alpha case and Widmanstatten pattern (Magnification 200X) Figure 6.2(b). Micrograph of a portion of the cross-section of the CP Ti rod, after annealing in vacuum at 1100°C for 24 hours (Magnification 100X) 68 6.1.4. Measurements to determine oxygen concentration To characterize the oxygen profile of the oxygen solid-solution surface layer, Vicker's microhardness .indentation measurements were made, since the microhardness is a well-characterized function of the oxygen content. Figure 6.3 shows the variation of the microhardness indentation size with distance from the surface, and Figure 6.4 shows the corresponding microhardness values. Microhardness measurements had to be made since the maximum variation of oxygen took place within a millimetre of the surface. The hardness values showed reasonable reproducibility at similar depths from the surface, having variations ± 100 H v at high values of hardness (700 H v ) , and ±30 H v when the hardness values were lower. The increase in hardness of CP Ti with dissolved oxygen has already been characterized, as reported in [4]. A 0.1 wt.% oxygen causes the hardness to increase by 40 points on the Vicker's scale, translating to a yield strength increase of about 120 MPa [4]. However, the available literature did not contain hardness data for oxygen concentrations above 1% by weight. The maximum oxygen concentrations from the current study were certainly above 7%, ruling out the extrapolation of the hardness data. Thus, WDX measurements were made to determine the oxygen profile independently. It was recognized that such an analysis had its inherent limitations for determining the concentration of light elements such as oxygen. The oxygen profile was also examined in the light of the expected values obtained from the equilibrium phase diagram and diffusion calculations, and checked for mutual consistency. The WDX measurement was calibrated using the known oxygen/hardness relationship at a point where the oxygen concentration was known corresponding to the hardness value. Having established this internal standard, the oxygen concentration at the points close to the surface (where the microhardness extrapolation had failed) could be determined from the WDX measurements. 69 Figure 6.3(a). Vicker's microhardness indentations showing decreasing hardness with distance from the outer surface, and also the sudden decrease at the visible boundary (Magnification 200X) 800 700 | • 8 600 | c a 500-§ 400 | 3oo } s i i 200 4-100 0 Observed phase 100 200 300 Oepth (In microns) 400 500 Figure 6.3(b). Plot showing the Vicker's microhardness values with distance from the outer surface, corresponding to the indentations shown in Figure 6.3. 70 the oxygen profile All hardness measurements in the oxygen stabilized alpha case were greater than 500 H v . The microhardness at the interface was 500 H v , beyond which it dropped abruptly to values of about 250 H v . If the microhardness vs. oxygen content data available in the literature was extrapolated, 500 H v would correspond to an oxygen weight % between 1.8 and 2.8%, depending how strongly the function continued to behave beyond 1% oxygen. This is in very close agreement with the equilibrium oxygen concentration at the a/a-f3 interface, which predicts 3%, and is shown in the schematic phase diagram Figure 6.6, which represents the region of our present interest in the Ti-0 phase diagram already shown in Figure 2.1. Referring to the WDX measurement shown in Figure 6.5, the phase boundary value of 3% measured by the WDX is calibrated against the actual 3% suggested by the phase diagram, and that obtained by the extrapolated hardness in Figure 6.4. The oxygen concentration profile was established using the WDX measurements once the calibration was completed. It was noted that if the value of 3% was calibrated against an estimated value of 2.5%, all of the WDX measured oxygen values would be higher by a factor of 1.2. 71 o c TJ a c o k_ o 2 100 H : H 200 300 Depth (In microns) 400 500 Figure 6.4 Plot of Vicker's microhardness vs. depth. _ 8 Z 7 CD o a o u a a —*—• • Observed phase boundary ^ — 1 1 — • 1 • 1 100 200 300 Dapth (In microns) 400 500 ?ure 6.5 Plot of oxygen concentration by WDX measurements at equivalent depths 1100°C Ot+H. 882°C CPTi Temp. GradeU 1100 ppm 1% 2% 3% weight % oxygen F igure 6.6 Equilibrium diagram of Ti -0 in the region of our interest (not to scale). 72 6.2. DISCUSSION ON RESULTS OF OXYGEN DIFFUSION 6.2.1. The oxide scale The objective of studying the oxide scale was to ascertain whether it played a role as a barrier to oxidation, and if so, whether it was more effective at certain temperatures. Dependence on Temperature For low temperatures of air soaking, the oxide formed on the surface was very adherent and gray in colour. The material did not appear to have lost its metallic lustre entirely. At higher temperatures, the oxide became more scaly and friable. For the samples air-annealed (oxygen contaminated) at 1100°C, the oxide scale sometimes broke during handling. Previous studies of oxide scales have mentioned that the scales grow in strata, and the resulting oxide layer tends to break at higher temperatures because of internal thermal expansion mismatch. Furthermore, the oxide scale presents an effective barrier 3 microns thick at 875°C [1]. At higher temperatures, it is expected that the diffusion coefficient of oxygen through the scale would rise due to the temperature increase. One would also expect more defects and breakages at the higher temperatures [1]. Consequently, it could be surmised that the diffusion barrier due to the scale would be less effective at higher temperatures. If one were to assume that no oxide barrier exists, and that the only role it plays is to maintain a constant oxygen potential at the surface of the material, then the oxygen stabilized alpha case should grow in a manner dictated by the diffusion of oxygen through the alpha phase. In such a case, the difference in the calculated thickness of the alpha phase layer after 24, 48 and 60 hours using a parabolic growth rate would be nearly the same as experimentally measured and listed in Table 6.1. (The calculated thicknesses are about 25 - 30% more, which is a very good agreement under the circumstances). Thus, it can be said that the oxide layer barrier is not an effective barrier at 1100°C and does not change significantly with annealing times at 1100°C. 73 The visible colour change of the Ti scale provided an indication of the chemical composition development in the oxide scale. Stoichiometric rutile (TiC^) is powdery in texture and white in colour, and does not appear to form at alpha phase temperatures, even when exposed for long periods. Consequently, samples air-annealed (oxygen contaminated) at 750°C were grayish in colour retaining some metallic lustre. The change of colour from gray to yellowish at higher temperatures is attributed to the formation of small amounts of Titanium Nitride (TiN), which is brown in colour, amidst the white rutile. The sample appears yellow because of the sparseness of TiN. The WDX analysis of the oxide confirms the abundance of rutile and the sparseness of TiN. This reinforces the argument that oxygen - not nitrogen - is the main hardening agent. Not only is the diffusion co-efficient of nitrogen in titanium at least 10 times lower than that of oxygen at all temperatures [2], the surface potential of nitrogen on the alpha titanium is very low due to its sparseness in the scale. Thus, the oxide layer is beneficial insofar that it may serve as an adherent coating during heating prior to rolling when the temperatures are still below the alpha transus. Higher temperatures, i.e., above the a=>P transition temperature, causes the stratification of the oxide layer and breakdown of adherence. This causes further thickening of both the oxide scale and the development of the oxygen stabilized alpha case. Not only do higher temperatures of soaking enhance these phenomena, but also they allow increased difftisivity of oxygen which adds to the growth. 6.2.2. The Oxygen stabilized Alpha case The radial symmetry of the sample which were air-annealed (oxygen contaminated) producing an oxygen-stabilized alpha case and the increase of its thickness with increasing high temperature exposure to oxygen annealing time confirms that the thickness of the case is related to the oxygen content. Since the samples were air-annealed at 1100°C for long times, and then air-cooled, the Ti-0 equilibrium phase diagram (Figure 2.1) was used to interpret the resulting microstructure. A schematic diagram of 74 the area of interest on the phase diagram is shown in Figure 6.6. The surrounding outer phase, as in Figure 6.2, was identified as the oxygen stabilized alpha phase containing at least 3% oxygen in solid solution at 1100°C. The distinct boundary denotes the a/cc-P interface at 1100°C, which was associated with the change of oxygen concentration below the required value to have stabilized alpha at that temperature, viz. 3%. The adjacent material consisted of a small amount of mixed alpha-beta phases, and the adjacent high temperature beta phase stable at that temperature. 6.2.3. Growth of the alpha phase Since the boundaries of each of the phase regions change with oxygen and hence time, it is a multiple moving boundary which could be solved using finite difference method. Simplifications could be made - for instance, since the diffusion coefficient of oxygen in the beta phase is 100 times that of the alpha phase, the diffusion in the mixed alpha and beta region should be a result of the beta phase transport properties, if it is a continuous phase. But to attempt such a solution would have required the knowledge of the transport properties of oxygen through the scale and the equivalent oxygen content at the phase interfaces, since this determines the inflow of oxygen into the material. These properties of the scale are not known at the temperatures of interest. In fact, the scale consists of a mixture of non-stoichiometric and stoichiometric Ti-0 compounds, and the oxygen content of the aggregate would be quite complicated. The only data on oxygen transport and equivalent oxygen content is available for 875°C, as shown in Figure 2.7 in Chapter 2; that data too is obtained by indirect calculation from the study of growth of the alpha phase at that temperature [1]. It is not possible, however, to model the growth of the alpha phase because of the paucity of information about the nature of the diffusive barrier and growth of the oxide scale. Since the diffusion coefficient of oxygen in the alpha close-packed hexagonal phase is about one one-hundredth of that of the more open body-centred cubic beta phase at that temperature, the diffusion of oxygen is most likely controlled by the diffusion in 75 the beta phase beyond the alpha boundary interface into the bulk of the material, even though there exists a region of mixed alpha and beta phases. It is assumed that the beta phase is a continuous matrix. The growth of the alpha-phase, thus, becomes a semi-infinite two-phase diffusion problem with a moving boundary. This situation has been studied for an equivalent case of diffusion of carbon into steel (which is made up of the structurally open B.C.C. phase) through a F.C.C. stabilized carburized case which grows with time. Drawing parallels from such a study, as well as from results of simple calculations using Fick's Law to predict the order of the thickness of the alpha layer, it can be said that the growth of the solid solution alpha case is controlled by the rate of diffusion of oxygen through the alpha phase, the oxygen having a very low diffusion coefficient. It is for this reason, and because of the continuous growth of the oxide scale which grows at the expense of the alpha case beneath it, that even annealing under severe temperatures and for prolonged time periods creates a depth of the alpha case which is unlikely to exceed 1 mm. 6.2.4. The Widmanstatten pattern development and formation On air-cooling, asjn our case, the p phase transforms to a-Widmanstatten plates. The morphology of the alpha phase obtained on cooling can vary from serrated to basket-weave depending on the oxygen alloying content [3]. As discussed in Sec.1.3.3 in Chapter 2, the presence of oxygen in the sample encourages the formation and definition of the Widmanstatten plates. The obtained microstructure, as shown in Figure 6.2, was similar to that cited by McQuillan, as shown in Figure 2.2 (b). The Widmanstatten patterns were not observed in the vacuum-annealed samples when it was similarly air-cooled. The occurrence of the Widmanstatten patterns is thus attributed to the oxygen alloying in the sample. 76 6.3. RESULTS OF COMPRESSION TESTS ON ANNEALED SAMPLES To ascertain the reason for the occurrence of surface cracking at the processing conditions found in direct rolling, oxygen contaminated samples were deformed by the Gleeble at strain rates and temperatures comparable to those experienced during rolling; the samples were then examined for cracks and failures. As a basis for comparison, samples without the oxygen-annealing treatment, and a sample annealed under similar conditions in vacuum were subjected to the same test conditions. 6.3.1. Visual Inspection Surface cracks were readily visible on the oxygen contaminated samples, the severity of the cracks varying with the processing conditions. The as-received (oxygen uncontaminated) or the vacuum annealed sample showed no evidence of any kind of surface cracking. Figure 6.7 compares the typical visual appearance of an oxygen contaminated and an uncontaminated sample after the same deformation treatment at 850°C at a strain rate of 10 s"1 to a strain of 1 . Figure 6.7. Comparison between deformed samples - the left one was previously air-annealed at 1100°C , the right one was not. The deforming temperature was 850°C, and the samples were deformed to a strain of 1 at a strain rate of 10 s" . 77 6.3.2. Microstructural Examination The oxygen-contaminated samples, the untreated sample and the vacuum annealed sample were cut across the axis at the mid-span of the specimen and polished and etched to reveal the microstructure. The plain sample and the vacuum annealed samples showed no signs of cracks, either at the surface or in the interior, as shown in Fig. 6.8(b) and Fig. 6.8(c) respectively. There were two distinct types of cracks in the air-annealed (oxygen contaminated) samples, as illustrated in Figure 6.8(a). Firstly, the alpha case was totally cracked, as marked by the white arrow. Secondly, a regular pattern of cracks appeared along the existing Widmanstatten plates (marked by the black arrow), whose orientations varied from one grain to another. Deformation twins were also noticed, as marked by the dotted arrow, concurrent with the findings from a similar study [12]. Presence of the latter, and other deformation structures, if any, can only be confirmed by further studies. Figure 6.8(a). Micrograph of sample deformed at 850°C at 10s"1 to a strain of 1 after annealing. Cracks are usually limited to the outer boundary. Inside are the cracked Widmanstatten platelets and the twin bands. (100X) 78 Figure 6.8(b) Micrograph of as-received (uncontaminated) sample deformed at 850°C at 10 s"1 to a strain of 1. There is no evidence of surface cracking and the material appears to have undergone some recrystallization. (400X) strain of 1. There is no evidence of surface cracking and the material which previously had large grains appears to have undergone recrystallization.(200X) 79 6.3.3. Strain to Fracture Comparison of Stress - Strain Curves Samples of uncontaminated CP titanium when deformed in compression using the Gleeble at strain-rates comparable to that of direct rolling, showed steady-state flow stresses in the alpha region as shown in Figure 6.9(a), and some strain-hardening when tested in the beta region as shown in Figure 6.9(b). The magnitude of the stress varied as a function of the strain rate, increasing strain rate producing higher stresses. The flow stress curve of the vacuum annealed sample tested in the alpha region of 850°C at a strain rate of 10s"1 was identical to the flow stress curve of the oxygen uncontaminated sample tested under the same conditions. The stress-strain curve of the air-annealed (oxygen contaminated) specimen showed a steadily rising stress up to a peak point after which the stress reduced with increasing strain, as seen in Figures 6.9(a) and (b). Although this behaviour could also be indicative of dynamic recrystallization, no evidence of this is seen in the microstructure of the sample, as seen in Figure 6.8(a). Moreover, neither of the control specimes - the as-received (oxygen uncontaminated) sample or the vacuum annealed sample - showed a similar drooping curve that suggests dynamic recrystallization. Thus, the peak stress was evidence of failure resulting in a loss of load bearing capacity of the material, and was associated with a strain value (sf ~ 0.2) at those strain rates, as shown in Table 6.3. Table 6.3 Strain to failure as deduced from the flow curve comparisons between oxygen contaminated and plain samples. (.'?' means that the value was not judged reliable due to loss of c-strain gauge during test). Temp(°C) Strain Rate (s-1; > 0.01 1.0 10.0 750 Sf=0.12 ef=o.2 sf=0-3 850 S f 0 7 8f=0.12 6 f - ? 950 e=o.i2 Sf=0.13 Sf=0.12 80 Figure 6.9(a) Flow curve of the air-annealed (oxygen contaminated sample) compared against as-received (uncontaminated) sample tested in alpha phase region. The flow stress curve of the vacuum annealed sample was identical that of the oxygen uncontaminated sample. 120 Figure 6.9(b) Flow curve of the air-annealed (oxygen contaminated) sample compared' against as-received (uncontaminated) sample tested in beta phase region. 81 By visual observation, it was found that the most severe surface cracking of the air-annealed (oxygen contaminated) sample occurred at 800°C at 10s"1 .When an air-annealed (oxygen contaminated) sample which had been soaked at 750°C instead of 1100°C was subjected to the same deformation conditions, the sample did not show any surface cracks as seen in the Figure 6.10 below. The flow stress curve of the sample was identical to that of the flow stress curve of the as-received (oxygen uncontaminated) sample, which did not show a peak stress. 6.3.4. Effect of grain size on flow stress When tested at 850°C at 10 s"1 , the flow stress curves for the as-received material having grain size -100 micron, and that for the vacuum annealed sample having grain size about 1 mm, were identical. Figure 6.10. Photograph of sample deformed at 850°C at 10s"1 after annealing at 750°C instead of 1100°C. Sample shows no sign of cracking. 82 6.4. DISCUSSION OF THE COMPRESSION TEST RESULTS 6.4.1. Microstructural examination In the deformed air-annealed (oxygen contaminated) samples, there were two types of cracks. Firstly, there were cracks in the alpha case. They tended to terminate on reaching the matrix underneath, as shown in Figure 6.8(a). Most of these cracks already existed even before the samples were deformed. During deformation, the existing cracks became more severe, and new cracks formed. The second type of cracks, i.e., the cracks between the Widmanstatten plates, were formed during deformation. The reasons that support the fact that the cracking takes place between the Widmanstatten patterns are: i) The microstructure as shown in Figure 6.8(a) shows Widmanstatten platelets separating from each other. ii) The spacing between the successive cracks and adjacent plates are similar, as are their pattern, as compared in the micrographs in Figure 6.11 (a) and (b). iii) The orientation of the cracks are different in different grains, resembling the orientation pattern of Widmanstatten structures, as evident in Figures 6.11. Results from previous works [5,6] suggest that twinning is an operative deformation mechanism. However, it appears that the material is unable to accommodate slip. Failure between the Widmanstatten plates, which have pronounced directional properties, most likely occurs due to the pile-up of slip dislocations at these plates. This is also consistent with the fact that the alpha phase is hexagonal close-packed, and has limited operative slip sytems. 83 F i g u r e 6.11 (a) Micrograph showing Widmanstatten pattern formation (200X) F i g u r e 6.11 (b) Micrograph showing air-annealed sample after deformation. The cracks have a regular arrangement, and different orientations for different grains (200X) 84 The parent material showed wide-lens shaped structures and serrated structures -similar to the lamellar and lenticular twins, which are invariably known to occur when titanium is deformed under such conditions of temperature, strain rate and strain. Their presence has been reported in a similar study [5], and independently observed by others [6]. At this point, there have not been any reports indicating that their presence causes microstructural instability in the material resulting in failure. They appear to be a principal deformation mechanism operative during deformation at hot-rolling conditions. When the strain or strain rate increases, they are known to offer nucleation sites for static recrystallization [5]. The uncontaminated sample, or the vacuum annealed sample did not have Widmanstatten structure in its microstructure. Nor did they have an oxygen stabilized alpha case. After deforming the uncontaminated samples, the matrix was uncracked, recrystallized, and had very fine equiaxed grains, as seen in Figure 6.8(b). The recrystallization must have occurred after the deformation, since CP Ti is known to recrystallize statically. Further, evidence of dynamic recrystallization was not noticeable from the shape of their flow curves. The vacuum annealed samples had some recrystallized grains, though portions of of pre-existing large grains remained. The microstructure is shown in Figure 6.8(c). 6.4.2. Stress-Strain Curve Analysis The clustering of the strain to failure around 0.2 suggests that the failure of the material occurs when the strain reaches a critical value. The surface alpha case was subjected to the circumferential strain as the cylinders were compressed, which aggravates the conditions for failure at the periphery. It is noted that the circumferential strain is about half that of the axial strain if the constancy of 85 volume is assumed at small strains, since 7rRQ L 0 = 7rRj . L j implies that the circumferential strain ln(27lRj / 27lR 0) = -1 / 2 ln(Lj / L Q ). The failure of the load bearing capacity of the material appears to be due to the combination of the total failure of the alpha case, and the failure occurring between the Widmanstatten plates in the base metal underneath the alpha case. However, the existence of radial cracks even before deformation occurred must have considerably lowered the strength contribution of the alpha case. Further correlations between the processing conditions and the strain to fracture were not apparent on the basis of metallographic examination of the samples or comparison of their flow curves. The examination of the flow-stress curves for the air-annealed (oxygen contaminated) samples in Figure 6.9(a) shows that for 750°C, the peak stress for a strain rate of Is - 1 is actually higher than for the strain rate of 10s"1 . A probable reason could be that at the higher strain rate the material failed earlier. For the lower strain rate, the peak in the curve is indicative of the strain to fracture of the material. The difference in the flow stress values between the air-annealed (oxygen contaminated) samples and the uncontaminated ones, as evident in Figure 6.9(a) cannot only be due to the hard alpha case, as this would require impracticably high yield stress for the alpha case, it being only 0.5 mm in thickness. This difference must also be a consequence of the dissolved oxygen in the base metal. It has been shown that an increase of 0.1 weight % oxygen can cause the yield stress to rise by about 120 MPa at room temperature. Thus, a higher yield stress would be expected also at higher temperatures. 6.4.3. Effect of grain size on flow stresses Studies of deformation mechanisms [18] indicate that the climb mechanism becomes the rate-controlling step for higher stress levels (when o7G> 10~3). For even 86 higher stress levels in the power-law breakdown regime relevant to the present research, the rate-controlling mechanism is not clear, but the grain size dependence gradually diminishes, and the grain size becomes unimportant. This is consistent with our observations, where an increased grain size did not alter the shape of the flow stress curve. 6.5. GENERAL DISCUSSION In the light of the above results, certain key factors which are expected to influence the formation of cracks during breakdown rolling of C P titanium are discussed. 6.5.1. Presence of interstitial oxygen Presence of dissolved oxygen enhances Widmanstatten pattern formation during cooling from the beta region, as illustrated in Figure 2.2. These plates subsequently fail during deformation. Furthermore, the dissolved oxygen helps to form a stable alpha case which also fractures during deformation. 6.5.2. Phase transformation effects during breakdown rolling It is to be examined whether the p=>ct transformation would aggravate or alleviate the surface cracking problem that occurs during breakdown rolling. The factors which point towards aggravation are more than those which do not. Firstly, the (5=>a transformation at cooling rates such as due to roll-chilling is of the martensitic type, which causes surface stresses [3,14]. Secondly, the transformation to the alpha phase is accompanied by a reduction in volume, which results in tensile forces at the surface. 6.5.3. Surface chilling effects during breakdown rolling The depth to which the temperature is below the alpha transus, as obtained from the simulation results of breakdown rolling is about 10 mm (1 cm), of which the alpha case is less than 1 mm. Due to oxygen stabilization effects, the alpha transus temperature would increase considerably. Thus, the estimate of 1 cm is conservative - the depth upto which the material immediately transforms to the alpha phase on entering the roll is certainly more than 1 cm. 87 The cooling rate is enough to ensure martensitic transformation near the surface of the material; this will result in needle-shaped to Widmanstatten structures (depending on the depth from the surface), and accompanying surface tensile stresses. These Widmanstatten patterns subsequently disintegrate during deformation. Further, surface chilling causes thermal contraction of the surface layer in comparison to the relatively hot metal underneath, giving rise to surface tensile forces which would aggravate the tendency to rupture at the surface. 6.5.4. Grain size After the CP Ti ingot is cast, the surface layer of the ingot is removed to eliminate surface imperfections and volatile impurities sticking at the surface. Depending on the depth of the layer that is machined off, the fresh surface of the material can have a range of grain sizes. If the machining totally eliminates the as-cast chilled layer, the surface can have large grains (approximately 10 mm diameter). Given that our experiments were performed with material having a much smaller grain size, the effect of an increased grain size needs to be examined - its effect on the oxygen diffusion rate, and consequently, its susceptibility to cracking. Effect on Oxygen Diffusion When samples having grain sizes about 10-20 mm in diameter were annealed, it was found that the thickness of the alpha case was approximately half that of the annealed as-received samples having a grain size of 50 microns. It appears, thus, that the grain size does not have a significant effect on the rate of oxygen diffusion. Effect on susceptibility to Cracking Samples with large grain sizes (1-2 mm), also containing dissolved oxygen had pronounced Widmanstatten patterns, which subsequently failed during deformation. However, the Widmanstatten pattern was noticed when the material had dissolved oxygen as well - the effect of the two factors cannot be separated from the results of our experiments. It is known from past work, however that both large grain sizes and oxygen content encourage formation of lamellar alpha structures [14, 19]. 88 MICROSTRUCTURAL EVOLUTION OF THE SPECIMENS The as-received sample had equiaxed grains with an average grain size of about 100 microns, as shown in Figure 6.12. This served as the starting material for all subsequent procedures. For tests with the oxygen uncontaminated samples, these specimens were heated up to the test temperature at 10°C s"1 and compressed in the Gleeble, and then allowed to air-cool. For samples deformed in the beta phase, the material transformed to the alpha phase during air-cooling. The resulting microstructure consisted of small grains with sharp edges of an average size of 30-40 micron. The boundaries of the previously existing beta phase could be discerned, even though the imparted strain was sufficient to cause copious nucleation of the alpha phase during cooling. Figure 6.13 shows the microstructure of an uncontaminated sample which was deformed in the beta phase and then allowed to air-cool. For test specimens deformed at elevated temperatures in the alpha phase and then air-cooled, the material recrystallized to form fine grains, though some directionality could be observed. The resulting microstructure consisted of small grains with an average size of around 20 microns as already shown in Figure 6.8(b). 89 Figure 6.12 Microstructure of the as-received sample showing equiaxed grains. (200X) 90 The microstructure of the as-received sample which was annealed in air at 1100°C for 24 hours and then air cooled has already been shown in Figure 6.2 and discussed in Section 6.1, where the formation of an oxide scale, an alpha case, and Widmanstatten pattern was noted. The resulting microstructure on deforming the sample has been shown in Figure 6.8(a) and discussed already in Section 6.4.1. The microstructure of the as-received sample which was annealed in vacuum at 1100°C for 24 hours and then air cooled has already been shown in Figure 6.2(b). There is no formation of an oxide scale, alpha case, or Widmanstatten structure. The resulting microstructure on deforming the sample has been shown in Figure 6.8(c), and discussed in Section 6.4.1. 6.7. FLOW STRESS STUDIES FOR CP TITANIUM Axisyrnmetric compression . tests were performed using the Gleeble thermomechanical simulator at strain rates 1 - 20 s"1 comparable to those realized during breakdown rolling of CP Ti in the temperature range 750°C to 950°C. The temperatures were chosen so that deformation occurred both in the alpha and the beta regimes. These tests performed on the as-received material served as control samples against which the oxygen contaminated samples were previously compared. The specimens were deformed to a true strain of 1.0 and the flow stress curves were analyzed to a strain of 0.6 without applying any correction for friction*. * Rao and Hawbolt [8] quote various researchers who have suggested that for a specimen of 10 mm in diameter and 12 mm in height, the effect of friction is negligible up to a strain of -0.7. 91 6.7.1. The Alpha Regime For processing regime at temperatures between 750°C and 850°C, flow stress curves achieve a steady state response beyond a strain of approximately 0.1 when strain rate is 1 s"1 and beyond a strain of 0.25 when the strain rate is 10 s"1 , as shown in Figure 6.14. The temperatures stated in the figure are the temperatures measured at the surface of the specimen at the mid-plane location of the strain measuring device. The maximum temperature difference between the outer and the inner thermocouple varied between 30 and 5 0°C, mostly due to the thermal gradient. 200 100 0.1 0.2 0.3 0.4 C-strain O.S 0.6 0.7 0.8 Figure 6.14. Axisyrnmetric compression testing flow stress curves for alpha titanium. The flow stress curve for vacuum annealed sample at 850°C at a strain rate of 10s is identical to that of the uncontreated sample tested at the same conditions 92 with Ashby's theoretical predictions The flow stress results from the compression testing fell beyond the power-law description of constitutive equations, in the power-law breakdown regime. The equation proposed by Ashby [8] to model the experimental data in this regime is: (6.1) e = A I ^ [ s i n h ( a a ) ] n kT L I All the values necessary for the calculation of the flow stresses from Ashby's equation have been tabulated by Ashby and Frost [8] from the works of various researchers. In Figure 6.15, the flow stress values obtained in the current study are compared with the results obtained using the data published by Ashby and Frost [8]. Figure 6.15. Comparison between the flow stress values predicted by Equation 6.1 using Ashby and Frost's data [8] and those experimentally obtained by compression test measurements. 93 Equations in the Alpha phase A constitutive equation has been developed to predict the steady-state flow stress using the experimental data obtained in the alpha region. Three different equations were examined - the power-law, exponential and sinhyperbolic - to obtain the best fit. The value of aa > 1.2 obtained in this study suggests that creep power-law does not apply. However, this equation has been included in the analysis since the n value can be directly related to the strain-rate sensitivity. With this end in view, the best-fit power-law equation was: The Q and n values shown in Equation (6.2) are larger than expected. However, when the 850°C experimental points were removed, since these data points were suspected to have been obtained for structure which was softer (because of reasons discussed later), the resulting activation energy, Q, was similar to that reported for lattice controlled diffusion of 242 KJ/mole [8], as is shown in Equation (6.3): The data for this regime should ideally be represented by the exponential law - a simplification of the more encompassing sinhyperbolic law. The only problem with the exponential fit is that the n value does not make physical sense as it does in the former case. The best-fit exponential constitutive equation for describing the same data points with the 850°C points removed was: e = 8.65X10 3 Xr j 5 - 9 Xexp(-38300 / 8.312T) (6.2) s =.0437XG 5- 4 exp(-249800/ 8.312T) (6.3) 8 = 4 . 2 X 1 0 6 . e x p r j 0 1 exp(-216520 / 8.312T) (6.4) 94 The results of the sinhyperbolic fit are summarized in Appendix I. The 0 values obtained by the sinhyperbolic law are comparable to those obtained for the power law and the exponential law, but small values of a must be used to obtain Q-242 KJ/mole. 6.7.2. The Beta regime For formulating the constitutive equation in the beta regime, axisymmetric compression tests were performed at strain rates of Is"1 and 10s'1 at temperatures between 900°C and 975°C. The results are shown in Figure 6.16. Figure 6.16. Flow stress curves from the Gleeble compression tests done in the beta region The flow curves obtained in the beta region were different from those obtained in the alpha region in the following aspects. Firstly, the flow stresses were much lower in the beta phase, less than 100 MPa for all tests. Secondly, the softening mechanisms were not as dominant, resulting in all flow curves showing strain hardening with increasing strain. The only exception to this was the flow curve obtained at 900°C at Is"1 , which approached a nearly steady-state behaviour. 95 with Ashby's predicted values As in the case of the alpha region, the Ashby equation was used to predict the flow stresses in the beta region. The equation pertaining to the power-law breakdown was applicable here as well [8]. It should be emphasized that the standard constitutive equations are based on the attainment of a steady-state flow stress and thus independent of strain; that is, there is an underlying assumption of constant structure. However, flow curves in our case increased with increasing strain. The Ashby data was compared against the experimental stress values obtained at a strain of 0.2, as shown in Figure 6.17. Although reasonable agreement is seen, the experimental data did not agree with the calculated predictions using Ashby's equation (6.1) at other strain levels. 60 50 45 20/S 5/S • U 3 10/S • 10/s • 20/s OAshby 1/S • 1/S 900 950 Temperature (C) Figure 6.17. Comparison between experimental flow stresses obtained at a strain of 0.2 and Ashby's predictions 96 Equations in the Beta phase The exponential law has been used to describe the beta phase flow stress results. The values of Q, n, and A obtained at different strains are shown in Table 6.4. Table 6.4. Exponential equation Q, n and A values obtained for describing hot deformation in the beta phase region at different strains Strain Q n A 0.1 1187,816 ;478 7.52E41 0.2 612,741 .196 7.42E21 0.3 434,069 .141 1.96E15 0.4 . 407,912 .120 2.2E14 0.5 304,719 .112 1.83E10 0.6 216,518 .101 4.22E06 The effect of strain on each of the parameters can be described using an equation of the form (QoxnoxA)= 4r + C (6.5) with the constants A,B and C given in Table 6.5. Table 6.5. A, B and C values for Equation (6.5). Q n Log A A 330,394 .005 27.82 B .614 1.82 0.377 c -216,208 .08 -26.53 97 6.8. DISCUSSION OF FLOW STRESS RESULTS 6.8.1. Alpha phase with Ashby's predicted results It should be noted that the observed change in experimental flow stress with change in temperature and strain rate was much greater than that predicted by Ashby, i.e., the Ashby values show less strain rate sensitivity than the experimental results. The experimental -results were similar to Ashby's predicted results, but consistently larger in magnitude. This trend disappears when the test temperatures are very close to the beta transus. One reason could be that the nucleation of the softer beta phase has already started and hence the test is not within the alpha processing regime. It could also be due to another mechanism operative during deformation, such as dynamic recrystallization. The Ashby comparison has been used to "filter" the data, to classify it in distinct regimes, for then the anomalous data points show markedly different variations compared to the trend of the rest of the data points. The only experimental hot deformation data available in the alpha regime is that reported by Buhler and Wagener [3] and Malakandaiah [8]. However, since the oxygen content in their Ti samples were different from that in our experiments, and since the flow stresses are markedly dependent on the oxygen content [4], their data could not be used as a basis of comparison. 98 Equations in the Alpha phase .Exponential Law The activation energy obtained using the exponential equation activation energy for deformation obtained in this work is almost identical to the Q values reported for lattice diffusion, 242 KJ/mole; this is expected in the power-law breakdown regime where recovery is the predominant softening mechanism [12,13]. Law Assuming that n is a constant in the strain rates examined, the activation energies at a given strain rate can be written as: Q = R.n d l " ( ° > dln(l /T) (6.6) Using the steady-state flow stress data obtained in the alpha region, it is calculated that at low strain rate, Q is found to be about 220 KJ/mol/K, whereas, at high strain rates, it is about 265 KJ/mol/K. The value of Q for the entire strain rate range is thus a weighted average between these two extremes. The stress exponent in the power law, which corresponds to a strain-rate sensitivity factor of 0.185, is in the general range of the strain-rate sensitivity values reported for metals and alloys, which normally lie between 0.16 - 0.22. This is lower than the 0.23 value which has been quoted by Ashby for tests at lower strain rates, but similar to Donrad and Coner's power law exponent of 0.2 [11]. Ashby's value is generically quoted for the entire alpha regime spreading over several decades of strain rate, whereas the value obtained in the present study is more restricted to the breakdown rolling regime. When the power-law exponent is determined from the experimental slope of the logarithms of stress vs. strain, the material is found to have a low power-law exponent (high strain-rate sensitivity) at lower temperatures. This is also as expected, since at 99 higher temperatures the Q value is higher, and so n has to increase with increase in Q to keep the stress value constant. Table 6.6. n values (in the power-law fit) obtained at different temperatures. Temperature (C) n m 775 , 5.45 .19 810 5.88 .17 865* 7.31 .14 The abnormally high Q value which was obtained when the 850°C points was included in the analysis was a consequence of the mixed alpha and beta phase being tested, or another softening mechanism present during deformation. The equation had to accommodate the stress values of the mixed phase which were markedly different from the alpha phase stress values, and hence interpreted it as an extreme overall temperature sensitivity of the material. When the constitutive equation obtained with the tests performed in the single phase region is plotted, the experimental points of 850°C clearly stand out, as shown in Figure 6.18. The purpose of the constitutive equation, if it is to provide an accurate prediction of stress, is better served by including only the results obtained without the test points at 850°C, as deduced with the Ashby equations. An inexplicably high Q value obtained in constructing a constitutive equation suggests more than one operative mechanism for deformation. * Used only in the first power-law constitutive equation, equation (6.2) 100 s 600 500 • 400 300 A i 200 j i 100 0 Test temperature 850 C • Stress(Measured) j Stress (predicted) i 11 11.5 12 12.5 13 13.5 Los(Z-H) Figure 6.18. Comparison of the experimental and predicted values using the power law constitutive equation (6.4), in the alpha regime. The points obtained at 850°C stand out from the general trend. It should be emphasized that relatively few tests have been performed. The value of the limited data lies in the fact that they give an estimation of the activation energies and strain rate sensitivities. 6.8.2. Beta phase - Flow stress results 6.8.2.l.Comparison with Ashby's predictions The data points obtained at the lower strain rates obtained in the beta phase show good correspondence to the Ashby predictions both at 900°C and 950°C, as shown in Figure 6.17. However, the strain rate sensitivity is higher than that predicted by Ashby's data and Equation 6.1 for points around 900°C. The general trend is that the experimental values are higher than the predicted values. However, the predicted value at 900°C and Is"1 is higher than the experimental value. The possible reason for this is the presence of the alpha phase, which also results in higher strain rate sensitivities (and hence Q values). 101 It should be noted that the difference between the experimental and observed values at different strains is more at the lower temperature (900°C). At the higher temperatures, the Ashby's predicted values are very close to the measured values. The value of n in Ashby's equation is based on very few data points (Buhler and Wagener's experimental data only). Also, the stress coefficient a' has been set to an arbitrary value of 103, to agree with these data points. The overall agreement of the prediction is thus somewhat fortuitous, since the predicted equations depended on many parameters whose values are not conclusively established, as discussed in Chapter 2, Section 5.4.1. Equations in the Beta phase It can be seen in Table 6.4 that the Q and n values systematically decrease with increasing strain, tending towards steady state at higher strains. It is known that increasing strain increases the stored deformation energy, and hence decreases the activation energy. The Q value tends to the activation energy of the rate controlling step at higher strains. Mathematically, the low strain rate sensitivity at low strains (i.e. the relatively small difference between stress values at different strain rates at low strain) cause large values of n at low strain levels. Accordingly, the Q has to be large to counteract this effect and keep the L.H.S. of the Equation 2.8 constant. In other words, the lack of marked difference in the stress values at different strain rates measured at low strains is responsible for the high Q and n values. The softening mechanisms actually diminish above 900°C as a result of the different phase formed. The crystal structure, its lattice diffusion coefficient, and grain-size are known to affect the deformation mechanism. Since it is known that the bcc crystal structure permits more slip-systems to operate than does the low temperature hep structure, it is expected that the beta phase should exhibit greater - not lesser - softening. It is possible, then, that the reduced dislocation motion depends on the grain size and the 102 lattice diffusion coefficient. Beta grains in Ti have been known to be notoriously resistant to softening by recrystallization, new grain formation only taking place at the grain boundaries. This suggests that a small grain size would allow more softening. Since the test conditions did not permit enough time for the growth of the beta phase grains, the grain size of the beta phase was small. Thus, a relatively low lattice diffusion coefficient in the beta phase is thought to be responsible for the lack of softening. References 1. Dechamps M , Lehr. P, Journal of Less Common Metals, 56 (1977), pp. 193 - 207 2. Smithell's Metals Reference Book, Ed. Brands E.A., Brook G.B., 7th Ed., Butterworht and Heinemann, pp. 13-93. •3. Metallurgy of the Light Metals - Polmear I.J, 2nd Ed., 1989, Edward Arnold Publication (Division of Hodder and Staughton Ltd.), pp. 218 - 220 4. Donachie M. , Titanium - A Technical Guide , ASM International, 1988, pp. 162 5. Hayashi M. , Yoshimura H., Ishii M. , Harada H., Nippon Steel Techical Report No. 62, July 1994 6. Private communication from Axel Johnson Ltd., Morgantown, PA, USA 7. Krasnikov N.E. and Skryabin N.P. , Soviet Journal of Non-ferrous metals, May 67, 8, (5), pp. 99-102 8. Frost H.J., Ashby M.F. , Deformation-Mechanism Maps - The Plasticity and Creep of Metals and Ceramics, Pergamon Press, Oxfordshire, New York, 1982 9. Buhler H., Wagener H.W.,Bander Blache Rohre, vol 6, pp. 677, 1965 10. Malakondaiah G., PhD Thesis, Banaras Hindu University, Varanasi, India, 1980 11. Doner, M. And Conrad H., Met. Trans. Vol 4, pp. 2809, 1973 103 12. McQueen H.J. and Bourell D.L., Journal of Materials Shaping Technology, Vol.5, No.3, 1988, pp. 163-189 13. McQueen H.J. and Bourell D.L., Review of Hot Workability of Metals and Alloys, Journal of Metals, Sept. 1987, pp. 28-35 14. McQuillan M.K., Met. Reviews (8), pp. 41-48, 1963 15. Dubrov V.A., Fiz. Metal. Metaloved., 24., No.2, pp. 316-320, 1967 17. Chakravartty J.K., Prasad Y.V.R.K., and Asundi M.K., Met. Trans. A, vol. 22A, April 1991, pp. 829-836 18. Mohamed F.A, Langdon T.G, Met. Trans., 5, pp. 2339-47, 1974 19. Porter D.A, Easterling K. E., Phase Transformations in Metals and Alloys, Van Nostrand Reinhold Co., New York, 1981, pp. 320-321 104 CHAPTER SEVEN SUMMARY AND CONCLUSIONS 7.1. SUMMARY AND CONCLUSIONS Oxygen is the main interstitial contaminant. The oxide scale formed is not tenacious, and does not serve as a protective coating to the material at 1100°C. The oxygen forms a solid solution with titanium resulting in an alpha case which is stable at elevated temperatures. The growth of the alpha case is pronounced at 1100°C. The more open bcc structure appears to promote such solid solution formation. At temperatures up to 900°C, the growth of the alpha case is not significant. Under the alpha case, pronounced Widmanstatten patterns were observed. The presence of oxygen accentuated the formation of the Widmanstatten pattern. The alpha case or Widmanstatten patterns were not observed in samples that were air annealed at 750°C, or those annealed in vacuum. The samples which were oxygen contaminated at 1100°C and then deformed under compression showed a peak in the stress-strain curves, and the subsequent fall in the stress was attributed to their failure, which occurred at low ductility values. Radial cracks were already formed in the stabilized alpha case during air cooling from the elevated temperatures. The cause of the failure during compression was attributed to total fracture of the alpha case, and failure occurring between the Widmanstatten plates in the parent material underneath the alpha case. The as-received oxygen-free samples, air-annealed at 750°C, and the samples which were annealed in vacuum, showed no signs of cracking when subjected to similar deformation treatments. Due to oxygen contamination, there was also a marked increase in the flow stress of the air-annealed samples before failure occurred. 105 The as-received oxygen pure samples showed steady-state recovery characteristics in the alpha regime. For tests performed with the outer thermocouple control temperature at 850°C, the flow stress values were markedly lower, which indicated that the nucleation of the beta phase had already started inside the specimen. There was a strain hardening response of the material in the beta regime. The stress-strain curve at 900°C , however, gave a near steady-state response, similar to the alpha regime tests. Constitutive equations were developed for the alpha and beta phase, and the activation energies obtained were consistent with previous reports. When compared with the data and equations proposed by Ashby, the experimental data from tests in both the phases was found to be more strain rate sensitive than the predicted values. Further, deviations from predicted values were greater when results of tests done near the phase transition temperature were compared. The deviation from Ashby's predictions was shown to be indicative of the presence of mixed phases. The results of modelling of the rolling process indicated that the material at the surface of the ingot was also most susceptible to higher strains (double the strain at the centre) and strain rates (eight times the strain rate at the centre), and this happened most intensely at the point of entry of the rolls. The chilled layer (the depth to which the temperature is below the alpha transus temperature) was limited to 10 mm from the surface, and the extent of chill was determined by the time of contact with the rolls. The degree of reduction and the friction between the material and the rolls were also important in determining the extent of the chill. 7.2. RECOMMENDATIONS The alpha case and Widmanstatten pattern formation should be prevented to eliminate cracking. It does not seem necessary to heat the ingots for extended periods at 1100°C - the heating should be performed at temperatures below the alpha transus, up to a maximum upper limit of 900°C. Suitable coatings must be applied before such heating. The coatings should be adherent to the base metal. They should not disintegrate at such 106 temperatures, and should not have a harmful reaction with titanium. The thickness of the coating should be sufficient for it to act as a diffusion barrier against oxygen. The need for rigorous temperature control is emphasized, also suggesting the need for accurate temperature measurement during the processing operations. 7.3. SCOPE FOR FURTHER WORK The constitutive equations developed in this work needs to be backed by more experimental data generated from a robust and thorough experimental schedule, both in the alpha and beta regions. A comparison between different grades of CP Ti could help to understand the variation of the deformation response as a function of the interstitial content. In this regard, the deformation maps constructed by Ashby might be updated to predict the deformation response of CP . titanium with present-day oxygen contents. It is possible that the problems encountered during the breakdown rolling CP Ti . are not entirely limited to the problems addressed in this study. Titanium has its unique characteristics such as an uncommon ingot-casting route, strong "memory" of its processing history, textural effects, complex operative deformation mechanisms, phase transformation occurring during deformation, many of which might contribute to problems during its processing. More interaction with the titanium rolling industry and a first-hand appreciation and knowledge of their difficulties is necessary to remedy the problems that might occur due to such causes. Data from industrial trials could also serve to strengthen the finite-element model that has been developed in this study. 107 APPENDIX A SINHYPERBOLIC FIT IN T H E A L P H A REGION The following table lists the values of n, Q and A for different values of a for the sinhyperbolic equation of the form e = A sinh(aa)nexp(-Q/RT).. (A.l) Table A . l . Q, n, A values for different values of alpha for the sinhyperbolic equation for steady-state flow stress results in the alpha region. Alpha value Q (J/moI/K) n A .0001 235493 5.47 4.34 E 19 .0002 235295 5.45 9.23 E 17 .0005 235039 5.39 5.32 E 15 .001 234275 5.17 8.53 E 13 .002 213798 4.44 1.38 E 11 .004 218661 3.09 9.36 E 09 .007 210772 1.93 1.02 E 09 . .01 208233 1.34 4.78 E 08 .02 ' 207930 0.67 2.88 E 08 .05 207845 0.27 2.16 E 08 108 APPENDIX B DYNAMIC R E C R Y S T A L L I Z A T I O N IN T H E A L P H A AND B E T A PHASES For low strain rate tests at 0.01s"1 , both in the beta and alpha region, a peak stress was noted, as shown in Figure B.l. A peak stress followed by a decreasing stress is usually interpreted as dynamic recrystallization (DRX) taking place during deformation. The DRX is more pronounced in the alpha phase. 180 160 140 / 750C - Alpha region 120 100 80 / 850C - Primarily alpha region . *' J^ > |»' •n^.i^^j^y/r^ 60 40 950C - Beta region 20 0 i; 0.1 0.2 0.3 C - Strain 0.4 0.5 i 0.6 0.7 Figure B. 1. Flow curves showing dynamic recrystallization at different temperatures for a strain rate of 0.01 s'1. 109 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items