Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Experimental study and modeling of IF steel oxidation process in air Ren, Xuzhan 2000

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

Item Metadata

Download

Media
831-ubc_2000-0142.pdf [ 8.42MB ]
Metadata
JSON: 831-1.0078705.json
JSON-LD: 831-1.0078705-ld.json
RDF/XML (Pretty): 831-1.0078705-rdf.xml
RDF/JSON: 831-1.0078705-rdf.json
Turtle: 831-1.0078705-turtle.txt
N-Triples: 831-1.0078705-rdf-ntriples.txt
Original Record: 831-1.0078705-source.json
Full Text
831-1.0078705-fulltext.txt
Citation
831-1.0078705.ris

Full Text

EXPERIMENTAL STUDY AND MODELING OF IF STEEL OXIDATION PROCESS IN AIR by XUZHAN REN B.A.Sc, Northwest Polytechnic University, 1987 M.A.Sc.,Northwest Polytechnic University, 1990 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE(M.A.Sc) in THE FACULTY OF GRADUATE STUDIES (Department of Metals and Materials Engineering) We accept this thesis as conforming to the rebuked standard THE UNIVERSITY OF BRITISH COLUMBIA February, 2000 © Xuzhan Ren, 2000 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. Department of Mftf Jc twlj l ^ ^ ' r y ^ E^ivAoSfT / \Op The University of British Columbia Vancouver, Canada Date 2 2 , 0 0 0 ( 9 DE-6 (2/88) ABSTRACT This thesis examines the oxidation of Interstitial-free steels in air by means of an experimental investigation and process modeling. The ultimate objective is to develop a computer model which could be employed to predict scale growth under industrial conditions. A purely theoretical approach to developing a model for industrial use is not appropriate because there are too many factors involved in the oxidation process and it would be difficult to make realistic predictions. The experiments were conducted under conditions as close as possible to industrial practice. With IF steels, two different scale growth mechanisms were found. Based on those different mechanisms, an innovative model for simulating heat transfer and predicting scale growth has been formulated and implemented on a personal computer. The model gives the best possible results in comparison to previous studies of predicting scale growth for the whole process and has the potential of being applied in the steel industry. For oxidation experiments, samples bigger than those used in the past were employed to reduce geometrical effects. Samples that are too small tend to give oxidation data that deviates greatly from the industrial situation due to geometric effects. Below 950 °C, adherent and dense scale is usually formed while above 950 °C, detached, laminated and porous scale accounts for most of the scale formed. The gap formed due to detachment reduces the scaling rate. So after gap formation, the scaling rates at 1000 °C and 1050 °C are lower than that at 950 °C. However the porosity and fissures of the scale increase the scaling rate. Temperature plays an even bigger role in oxidation. At temperatures higher than 1050 °C, the scaling rates are higher than that at 950 °C in spite of the detachment of the scale formed. These three factors compete with each other. The primary isothermal oxidation data is very different from the secondary isothermal oxidation data. This difference is very important in developing a practical model. The microstructural aspects of the scale formed were also comprehensively examined. For modeling, the basic idea is to use a whole set of isothermal oxidation data to simulate the oxidation process for any given temperature history. However if a sample goes through two different oxidation mechanisms, the primary isothermal oxidation data cannot be used for the secondary mechanism. With other details taken into account, the scale growth model gives results in good agreement with the experimental measurements. Table of Contents ABSTRACT ii TABLE OF CONTENTS iv LIST OF FIGURES vi ACKNOWLEDGEMENTS xi CHAPTER 1 INTRODUCTION 1 CHAPTER 2 LITERATURE REVIEW 5 2.1 Process and Kinetics of Oxidation of Steel 5 2.1.1 General Oxidation Process 5 2.1.2 Transport Mechanisms 10 2.1.3 Properties of Oxides of Iron 12 2.1.4 Kinetics of Oxidation 17 2.1.5 Secondary Aspects during Oxidation Process 24 2.1.6 Effects of Alloying Elements 28 2.2 Morphology and Microstructure of Scale on Steels 33 2.3 Modeling of Oxide Growth Process 38 CHAPTER 3 SCOPE AND OBJECTIVES 42 CHAPTER 4 EXPERIMENTAL INVESTIGATION 44 4.1 Experimental Setup and Procedures 44 4.2 Growth Curves 50 4.3 Microstructures of Scale of IF steel 58 4.4 Other Process Characteristics 82 4.5 Discussion: Scale Types and Growth Mechanisms 83 CHAPTER 5 MODELING OF OXIDATION PROCESS 89 5.1 Heat Transfer Model 89 5.2 Scale Growth Model 95 5.2.1 Theoretical Model of Scale Growth 95 5.2.2 Implementation Techniques 99 5.3 Program Structure 102 5.4 Simulated Results and Verification 105 CHAPTER 6 CONCLUSIONS AND FUTURE WORK 110 BIBLIOGRAPHY 113 APPENDIX A Explanation of Data Input File 116 - V -List of Figures 1.1 Amount of oxidation of carbon, low-alloy and stainless steels in 100 hours in air 3 2.1 Typical fractograph of adherent scale of IF steel. 900 °C, 24h 6 2.2 Oxidation mechanism of iron to form a three-layered scale above 570 °C showing diffusion steps and interfacial reactions 7 2.3 Post-oxidation SEM micrographs showing morphologies of oxide nuclei developed on a specimen, heated to 1050 °C for 20 min followed by exposure to oxygen for 3 sec, lOOOx 9 2.4 Iron-oxygen phase diagram showing oxygen contents of three oxides 13 2.5 Approximate percentages of wustite, magnetite and haematite on iron oxidized in oxygen 16 2.6 Typical weight gain curves during oxidation of low carbon steel in a gas mixture containing 6% O2 and 94% N2 20 2.7 "Specpure" iron oxidized at 950 °C 22 2.8a Adherent scale formed on pure iron at 950 °C in oxyegn 25 2.8b Nonadherent scale formed on pure iron at 950 °C in oxygen 25 2.9 Precipitates of S1O2 forming internal oxidation 31 2.10 Scale formed on mild steel in air at 1000 °C after 250 min 33 2.1 la Mild steel sheet scaled at 950°C in dry oxygen 35 2.1 lb Oxide scale on iron wire after 14 hours at 805 °C in pure oxygen 35 2.1 lc "Specpure" iron sheet scaled at 950°C in 88% oxygen, 12% steam 36 2.11 d BISRA pure iron sheet scaled at 950°C in 88% oxygen, 12% steam 36 2.12 Model for formation of microchannels in oxide scales by preferential outward diffusion of metal ions along grain boundaries 38 4.1 Experimental setup for IF steel oxidation 45 4.2 Three weight gain curves of IF steel oxidation at 950 °C 53 4.3 Average primary weight gain curves of IF steel oxidation below 950 °C 54 4.4 Three weight gain curves of IF steel oxidation at 1050 °C 54 4.5 Average primary weight gain curves of IF steel oxidation above 950 °C 55 4.6 Average primary weight gain curves of IF steel oxidation at various temperatures 55 4.6a Average primary weight gain curves of IF steel oxidation at various temperatures within first five minutes .56 4.7 Secondary weight gain curves of IF steel oxidation(850 °C) 56 4.8 Secondary weight gain curves of IF steel oxidation(950 °C 57 4.9 Secondary weight gain curves of IF steel oxidation(l 100 °C) 57 4.10 Fractograph of broad face scale and narrow face scale. 900 °C, 24 hours 63 4.1 la Outer surface of adherent scale. 900 °C, 24 hours 63 4.1 lb Outer surface of adherent scale. 900 °C, 24 hours 63 4.1 lc Fractograph of adherent scale. 900 °C, 24 hours 63 4.1 Id Fractograph of adherent scale. 900 °C, 24 hours 64 4.1 le Under surface of adherent scale showing macro voids. 900 °C, 24 hours 64 4.1 If Under surface of adherent scale. 900 °C, 24 hours 64 4.1 lg Contact area on under surface of adherent scale. 900 °C, 24 hours. 64 4.1 lh Noncontact area of steel surface at scale/steel interface. 900 °C, 24 hours 65 4.12a Outer surface of adherent scale. 950 °C, 24 hours 65 4.12b Fractograph of adherent scale showing concave scale/steel interface. 950 °C, 24 hours 65 4.12c Under surface of adherent scale showing the contact areas and. noncontact areas 950 °C, 24 hours 66 4.12d Contact area on under surface of adherent scale. 950 °C, 24 hours. 66 4.12e Noncontact area on under surface of adherent scale. 950 °C, 24 hours 66 4.12f Outer surface of steel showing the contact areas and noncontact areas corresponding to adherent scale in Fig. 4.3c. 950 °C, 24 hours 66 4.12g Noncontact area on steel surface under adherent scale. 950 °C, 24 hours. 67 4.12h Contact area on steel surface under adherent scale showing the same orientation of some grain surfaces. 950 °C, 24 hours 67 vii 4.13a Cross-section of adherent scale showing decreasing precipitation of magnitite in wustite layer from outer surface to scale/steel interface, i.e. from (b) to (f). 1200 °C, 24 hours 67 4.13b Outmost part of cross-section of adherent scale. 1200 °C, 24 hours 67 4.13c An area of cross-section of adherent scale. 1200 °C, 24 hours 68 4.13d An area of cross-section of adherent scale. 1200 °C, 24 hours 68 4.13e An area of cross-section of adherent scale. 1200 °C, 24 hours 68 4.13f An area of cross-section of adherent scale. 1200 °C, 24 hours 68 4.14a Cross-section of nonadherent scale and their structures. 1200 °C, 24 hours 69 4.14b An area of cross-section of nonadherent scale. 1200 °C, 24 hours 69 4.14c An area of cross-section of nonadherent scale. 1200 °C, 24 hours 69 4.14d An area of cross-section of nonadherent scale. 1200 °C, 24 hours 69 4.14e An area of cross-section of nonadherent scale. 1200 °C, 24 hours... 70 4.14f An area of cross-section of nonadherent scale. 1200 °C, 24 hours 70 4.14g An area of cross-section of nonadherent scale. 1200 °C, 24 hours 70 4.15a Outer surface of first layer of nonadherent scale. 1200 °C, 24 hours 70 4.15b Outer surface of first layer of nonadherent scale. 1200 °C, 24 hours 71 4.15c Outer surface of first layer of nonadherent scale. 1200 °C, 24 hours 71 4.15d Fractograph of first layer of nonadherent scale, showing different, layers 1200 °C, 24 hours 71 4.15e Under surface of first layer of nonadherent scale. 1200 °C, 24 hours 71 4.15f Magnified under-surface of first layer of nonadherent scale, showing porosity 1200 °C, 24 hours 72 4.16a Outer surface of growing second layer of nonadherent scale and steel 1200 °C, 24 hours 72 4.16b Fractograph of second layer of nonadherent scale showing irregular and porous nature. Oxidized 1200 °C, 24 hours 72 4.16c Fractograph of second layer of nonadherent scale. Oxidized 1200 °C, 24 hours 72 4.16d Under surface of second layer of nonadherent scale. 1200 °C, 24 hours 73 4.16e Under surface of second layer of nonadherent scale. 1200 °C, 24 hours. 73 4.17a Outer surface of first layer of nonadherent scale. 1200 °C, 19 hours 73 4.17b Outer surface of first layer of nonadherent scale. 1200 °C, 19 hours 73 4.17c Fractograph of first layer of nonadherent scale. 1200 °C, 19 hours 74 4.17d Under surface of first layer of nonadherent scale. 1200 °C, 19 hours 74 4.17e Under surface of first layer of nonadherent scale. 1200 °C, 19 hours 74 4.18a Outer surface of second layer of nonadherent scale. 1200 °C, 19 hours 74 4.18b Outer surface and fractograph of second layer of nonadherent scale showing wustite grains. 1200 °C, 19 hours 75 4.18c Outer surface and fractograph of second layer of nonadherent scale 1200 °C, 19 hours 75 4.18d Under surface of second layer of nonadherent scale. 1200 °C, 19 hours 75 4.18e Under surface of second layer of nonadherent scale. 1200 °C, 19 hours. 75 4.19a Outer surface of first layer of nonadherent scale. 1100 °C, 24 hours 76 4.19b Outer surface of first layer of nonadherent scale. 1100 °C, 24 hours 76 4.19c Fractograph of first layer of nonadherent scale. 1100 °C, 24 hours 76 4.19d Under surface of first layer of nonadherent scale. 1100 °C, 24 hours 76 4.19e Under surface of first layer of nonadherent scale. 1100 °C, 24 hours 77 4.20a Fractograph and outer surface of second layer of nonadherent scale 1100 °C, 24 hours 77 4.20b Outer surface of second layer of nonadherent scale. 1100 °C, 24 hours 77 4.20c Under surface of second layer of nonadherent scale. 1100 °C, 24 hours. 77 4.20d Under surface of second layer of nonadherent scale. 1100 °C, 24 hours 78 4.21 Cross section of second layer of nonadherent scale showing pores (1.2 microns in diameter). 1050 °C, 26 hours 78 4.22a Outer surface of second layer of nonadherent scale showing growth process of new oxide grains. 1050 °C, 26 hours 78 4.22b Outer surface of second layer of nonadherent scale showing growth process of new oxide grains. 1050 °C, 26 hours 78 4.23a Outer surface of first layer of nonadherent scale. 1000 °C, 24 hours.. 79 ix 4.23b Outer surface of first layer of nonadherent scale. 1000 °C, 24 hours 79 4.23c Fractograph of first layer of nonadherent scale. 1000 ? C , 24 hours 79 4.23d Fractograph of adherent scale. 950 °C, 24 hours 79 4.23e Under surface of first layer of nonadherent scale. 1000 °C, 24 hours 80 4.23f Under surface of first layer of nonadherent scale. 1000 °C, 24 hours. 80 4.24a Fractograph of first layer of nonadherent scale. 1000 °C, 24 hours 80 4.24b Fractograph of first layer of nonadherent scale. 1000 °C, 24 hours 80 4.24c Fractograph of first layer of nonadherent scale. 1000 °C, 24 hours 81 4.24d Fractograph of first layer of nonadherent scale. 1000 °C, 24 hours 81 4.25 Cross section of first layer of nonadherent scale showing two distinct layers 1200 °C, 24 hours 81 5.1 Differential element for for heat transfer and scale growth model 90 5.2 Definition of intervals and time steps 94 5.3 Ramping the loads between the user-defined values Fo and F n s 94 5.4 Weight gain calculation model 97 5.5 Overall structure of program 104 5.6 Simulated and measured weight gain curves of IF steel oxidation 107 5.7 Simulated and measured weight gain curves of IF steel oxidation 107 5.8 Simulated and measured weight gain curves of IF steel oxidation 108 5.9 Simulated and measured weight gain curves of IF steel oxidation 109 ACKNOWLEDGEMENTS I would like to express my most sincere thanks to my supervisor, Dr. Indira V. Samarasekera for all her supervision, support and help during my study. She provided me the best study and research environment, gave me the quickest responses when I need help and was always so nice to her students. My greatest appreciation goes to my beloved son, Jiesi Ren, who so far knows nothing about science but always inspires me to advance forward, and my wife, Yanjun Wu, who took full workload during my study to take care of my family and gave my study the greatest possible support. I am very grateful to Wenman Patrick for setting up the experiment, Mary Jansepar for her prompt help all the time, and Mary Mager for helping take these SEM pictures. I gratefully acknowledge the assistance in various kinds from Joan Kitchen, Glenn Smith, Ross and Carl. Finally, my sincere thanks should also go to these faculty members who helped me and from whom I learned something. x i CHAPTER 1 INTRODUCTION The earliest serious study of the oxidation kinetics of a pure metal is the work of Langmuir 1 J on the reaction of tungsten lamp filaments with low pressure oxygen gas in the 1910s. Important progress was made by Wagner's reformulation in 1951 of his theory of parabolic growth rate, based on the independent migration of ions and electrons in local equilibrium within the scale. The migration is driven by the free energy decrease accompanying the oxide formation from solid metal and gaseous oxygen. Since that time very little noteworthy work has been carried out without Wagner's guidance and later work has focused on the interdependence of structural, morphological, thermodynamic, kinetic aspects and mechanical properties of scale. Before the 1970's, most research on oxidation had concentrated on oxidation kinetics and scale morphology. Afterwards, increasing efforts were directed at the mechanical properties of oxide scale. Very important future work is to combine, through computer simulation, the kinetic, microstructural data and mechanical properties to simulate and predict the process Ch. 1 Introduction 2 of scale growth and its failure. The simulated results could be used for steel processing control and improvement of descaling techniques and equipment. Most characterizations of the chemical and physical nature of oxide scale continue to be dependent upon observations made on scale following cooling to ambient temperature and deduction of the probable situation prevailing at the oxidation temperature. The ideal approach would be real time observations of the changes in scale physical topography, chemical composition and stress etc. both throughout oxidation and on subsequent cooling to ambient temperature. However, up to date the employment of such in-situ techniques has been limited primarily due to experimental difficulties and lack of suitable equipment.[2] For most practical conditions, steels are thermodynamically unstable with respect to ambient gases or liquids and tend to react with them, although at room or low temperature this instability may be of little or no practical consequence due to the low rate of reaction. However at elevated temperature the oxidation reaction is accelerated. Temperature is one of the most important parameters affecting rates of oxidation. Higher temperature accelerates the kinetics of oxidation in an exponentially increasing manner.P] For plain carbon steel, the amount of oxidation in air is negligible below 535 <>C. Above this temperature, the rate of oxidation of carbon steel increases rapidly. This can be seen from Fig. 1.1 which shows the oxidation levels of various steels in air. Since the extensive scaling of iron at high temperature and in the presence of an oxygen rich atmosphere involves a number of competing processes, two samples, subjected to conditions as closely identical as practically feasible, in general will not exhibit exactly identical quantitative behavior.[4] The study and modeling of oxidation process is a quite difficult task Ch. I Introduction 3 to perform. Therefore a number of simplifying assumptions are absolutely necessary to obtain meaningful results which could in turn deepen insights into the problem. TEMPERATURE, T 9 0 0 1000 1100 1200 1300 1100 1500 1600 I7Q0 I B 0 0 5 0 0 5 5 0 ' 6 0 0 6 5 0 7 0 0 750 BOO 6 5 0 9 0 0 9 5 0 TEMPERATURE, 'C Fig. 1.1 Amount of oxidation of carbon, low-alloy and stainless steels in 100 hours in air^ Most oxidation studies related to iron and steel are aimed at improving the resistance of materials against degradation in aggressive environments at high temperatures which entirely depends upon the formation of protective oxide scale and maintenance of its integrity throughout service life. However, the oxidation studies of steel during its processing are almost entirely related to the reduction and removal of oxide scale. Steel processing is an important and large industry. Thus knowledge of the precise mechanisms involved both in the development and the failure of scale is of crucial importance. Ch. 1 Introduction 4 During steel processing operations such as continuous casting and rolling, oxidation is an extensive and inevitable problem which usually is so severe that capital consuming operations such as descaling and pickling are necessary. The water pressure used for descaling could be up to 400 bars which is a real challenge for equipment. The scale formed during steel processing is generally undesirable for following reasons: (1) the weight loss just in reheating operations varies from 0.75 to 3.25% and the overall scale loss in the mass production of steel may be up to 4 percent^; (2) a reduction in overall efficiency of the heating process; (3) huge amount of capital investment related with additional operations such as descaling and pickling; (4) adverse effect on the product quality. In order to achieve a regular, good surface of the hot strip, it is absolutely necessary to remove the scale before each reduction by spraying with high pressure water. In particular, the scale must be completely removed before the strip enters the finishing train. Only partial removal results in the appearance of "red scale";[6] (5) increased roll wear due to rolled-in residual scale. So the study of oxide scale behavior and its removal are of great practical significance for steel processing operations. Significant efforts have been made to remove scale in steel processing by mechanical or chemical processes. CHAPTER 2 LITERATURE REVIEW This chapter reviews the related current literature on oxidation. Enough details have been presented in order to provide a complete view of oxidation. 2.1 Process and Kinetics of Oxidation of Steel At elevated temperature, the rate of oxide formation on steel is dependent on (1) chemistry and physical characteristics of steel; (2) temperature, composition, rate of flow and pressure of the oxidizing atmosphere;[7] and (3) the length of time that the steel is exposed to oxidizing conditions. In the study of oxidation, the aspects of particular interest are usually rate of oxidation, structure of scale, transport mechanisms, configuration of the scale/steel interface, changes in composition of the various phases, stresses and scale failure. 2.1.1 General Oxidation Process 5 Ch.2 Literature Review 6 As shown in Fig.2.1, above 570°C the oxidation of iron usually forms three layers of oxide, i.e. wustite(FeO), magnetite(Fe304) and haematite(Fe203). The scale grows in thickness by a complex mechanism, in which oxygen is adsorbed at the outer surface of the hematite and is Fig.2.1 Typical fractograph of adherent scale of IF steel. 900 °C, 24h dissolved in the lattice as soon as an anion vacancy is available for it in the oxide crystal lattice. The oxidation mechanism of iron can be schematically shown in Fig.2.2 although the mechanism needs further clarification:t8] (a) At the iron-wustite interface, iron ionizes according to Fe = Fe2++2e~ (2.1) The iron ions and electrons migrate outwards through FeO layer over iron vacancies and electron holes respectively. Ch.2 Literature Review 7 FeO Fe 3* / Fe»Fe 3 *+2e" Fe ,0 4 Fe3* Fe3* ^ ^ 2 F e 3 * + 3 0 3 _ = Fe ; Fa"" * ne" + 4Fe20j = 3Fe 30 4 2Fe3* + 6e" • | 0 2 = Fe j0 3 -;o, +2e- = 0 3 ' 0, Fe3* + 2e" + F e 3 0 4 - 4FeO Fig.2.2 Oxidation mechanism of iron to form a three-layered scale above 570 °C showing diffusion steps and interfacial reactions[8] (b) At the wustite-magnetite interface, magnetite is reduced by iron ions and electrons according to Fe2+ +2e~ +Fe304 = 4FeO (2-2) Iron ions and electrons surplus to this reaction proceed outward through the magnetite layer over iron ion vacancies on the tetrahedral and octahedral sites and over electron holes respectively. (c) At the magnetite-haematite interface, magnetite is formed according to Fe"+ +ne~ +4Fe203 = 3Fe304 (2-3) The value of n is 2 or 3 for Fe 2 + or Fe 3 + ions respectively. (d) If iron ions are mobile in the haematite they will migrate through this phase over iron ion vacancies v™ together with electrons and new haematite will form at the haematite-gas interface according to 2Fe3+ +6e~ + | c 9 2 = Fe203 (2.4) Ch.2 Literature Review 8 At this interface also, oxygen ionizes according to -02 +2e~ =0 2 2 ,2- (2.5) If oxygen ions are mobile in the haematite layer, the iron ions and electrons, in excess of the amount required for reduction of haematite to magnetite, will react with oxygen ions diffusing inwards through the haematite layer over oxygen vacancies forming new haematite according to Lee and Rapp[ ] continuously observed the growth process of wustite grains at the initial stages of iron oxidation at 1050 °C and an oxygen partial pressure of 10"4 atm in a hot-stage environmental SEM. They found that at the initial stages of oxidation, discrete wustite nuclei developed epitaxially on the austenite substrate grains as shown in Fig.2.3. This epitaxial growth manifested a highly aligned arrangement of oxide nuclei. Once the small oxide nuclei grew laterally and impinged, very rapid oxide grain growth took place. Oxide grain coalescence by annihilation of grain boundary dislocations was believed to be the major mode of oxide grain growth. These coalescing oxide grains were the subgrains on the same metal substrate grain with only small misorientations. It was also observed that all the oxide nuclei formed on a given austenite grain coalesce to form a single wustite grain. The equiaxed grain grown from early elongated grains was almost the same size as that for the original austenite grain from which the oxide grains nucleated. Grain coalescence was more active during the early stages of oxide grain growth, i.e. from the moment when the elongated oxide grains impinged to the moment when they converted into equiaxed grains of size up to about 20 ^. Because of the large grain boundary area during the early stages of oxidation, the driving force for grain growth was the greatest. 2Fe}+ + 3 0 2 ~ = Fe203 (2.6) Ch.2 Literature Review 9 Fig.2.3 Post-oxidation SEM micrographs showing morphologies of oxide nuclei developed on a specimen, heated to 1050 °C for 20 min followed by exposure to oxygen for 3 sec, lOOOx [ 9 ] As the oxide scale grew thicker, grain coalescence did not continue because of the low mobility of dislocations and the lack of low-angle boundaries. In this case, nucleation of a new grain took place at an area near an original grain boundary. During the oxidation process, the following are very important aspects that deserve serious study. (1) Oxygen Adsorption Smith and Crane[10] investigated the oxygen adsorption on a freshly exposed SAE4340 steel surface at room temperature. The adsorption process could be illustrated with following steps: Ch.2 Literature Review 10 Step 1: adsorption of molecular oxygen 02 + S o S • o2 (2.7) Step 2: chemisorption with dissociation (o\ S--02+2Mo2 (2.8) Step 3: formation of the first oxide layer o M -M-M —> M(-M -O-M) (2.9) Where S the steel M the iron atom To form multilayer oxide, the location exchange between metal ions and oxygen ions should follow. After the first oxide layer is formed, the rate decreases considerably and the amount of oxygen that can penetrate into the first lattice layers depends upon the concentration of adsorbed oxygen molecules in the first activated state, which in turn is directly proportional to the oxygen pressure. 2.1.2 Transport Mechanisms Diffusion and the related phenomenon of conductivity in oxides are commonly associated with the migration of ions and electrons rather than electrically neutral atoms or molecules. Although the cations, anions and electrons all have some mobility in oxides, the mobility of one of the component particles far exceeds that of the others. In most oxides, ionic transport can be attributed almost entirely to the metallic ions either through the migration of cation Ch.2 Literature Review 11 vacancies or interstitial ions. [ U ] Over the bulk of the scale(wustite and magnetite), the oxygen anion lattice is fixed and iron ions diffuse outward. Although there is some disagreement, the following transport mechanism is generally accepted: [ 5 ' "~ 1 5 ] (a) Wustite grows entirely by diffusion of iron ions. (b) Magnetite grows largely by iron ion diffusion, but there seems to be an appreciable contribution from oxygen ion diffusion. (c) Haematite grows by oxygen ion diffusion with practically no iron participation. Due to the much greater mobility of defects in wustite, the wustite layer will usually be very thick compared with the magnetite and haematite layers above 700 °C. The diffusion coefficient of iron ions in iron oxides depends on temperature and oxide composition(number of vacant sites).[11] The rate of self-diffusion of iron in wustite increases continuously as the concentration of vacant iron sites increases. Furthermore, this dependence is essentially linear and the effect is noticeably greater at higher temperatures. The specific self-diffusion coefficient of iron in iron oxides has been measured.[11] In addition to being a sensitive measure of the deviation from stoichiometric composition, the vacancy concentration has fundamental significance with regard to the transport mechanism in wustite.[11] If only vacant cation sites exist, then only cations will be expected to diffuse. On this basis, wustite and magnetite should diffuse only iron and hematite should diffuse only oxygen.[I4J Each iron atom that enters the oxide is assumed to leave a vacancy in the metal. Also, before the oxide separates from the metal, the vacancies are assumed to be uniformly distributed through the metal. As the concentration of vacancies increases, they diffuse back to the iron Ch.2 Literature Review 12 surface under the separated oxide. Every crystal has an equilibrium number o f lattice vacancies at any temperature. The actual number depends not only upon the temperature but upon the energy required to form a vacancy. The case of oxidation is different from ordinary self diffusion in that there is a vacancy concentration gradient resulting from the generation of vacancies at the Fe/FeO interface. 1 1 4 3 The supply of metal to the oxide w i l l be greatest in the early part of oxidation. Oxide may freely grow to some critical thickness before it is necessary to dispose of the excess vacancies stored up in the metal. The diffusion coefficients of iron ions in the oxygen-rich phases, magnetite and hematite, are lower than in wustite. However, the diffusion rate of iron ions is usually uniform across the entire thickness of scale. Thus the iron ions depart from the wustite/magnetite interface at the same rate at which they arrive. The chemical equilibrium at the interface keeps the thickness of each layer in balance with the diffusion coefficient. A n y departure from this balance, say in a direction which causes iron to arrive more quickly from the wustite, would lead to a rise in iron concentration, converting magnetite to wustite, moving the wustite/magnetite interface outward and thus increasing thickness of the wustite layer. The same process applies to the magnetite layer. 2.1.3 Properties of Oxides of Iron To analyze the oxidation products of iron, an invaluable tool is the iron-oxygen phase diagram shown in Fig.2.4. From the phase diagram, it is noted that there are three oxides formed at different composition and temperatures, i.e. wustite Fe i . y O, magnetite F e 3 0 4 and hematite F e 2 0 3 Ch.2 Literature Review 13 Since the radius of the oxygen ion is approximately 1.4 ^ as compared to roughly 0.8 j \ for Fe^ and 0.7 ^ for Fe + + + , the three oxides are often regarded as relatively close packed networks of oxygen ions with the smaller iron ions arranged in the interstices. R A T I O 0 7 F e 10 11 12 1.3 1 i 15 1600 •WOO cj 1200 I al 1000 U J 600 " " 2 2 24 26 28 30 WEI 'GHT V . O X Y G E N Fig.2.4 Iron-oxygen phase diagram showing oxygen contents of three oxides[5] (a) Wustite As a rule of thumb, wustite is the innermost and most iron rich phase of the steel scale. The chemical composition of wustite is more accurately written as Fe^O where y defines the degree of deviation from the stoichiometric ratio. The value of y generally increases with distance away from the metal/scale interface, i.e. nonstoichiometry of wustite increases with the distance from the iron/scale interface. In wustite the oxygen ions and the metal ions occupy separate, interpenetrating cubic lattices in the NaCl-type structure. Ch.2 Literature Review 14 The wustite layer is by far the thicker and looser layer as diffusional transport in wustite is much faster than in magnetite and haematite. The density of wustite is about 5.87g/cm3. Due to its defect structure, the transport mechanism is essentially diffusion of iron cations via vacancies. Its structure becomes more and more non-stoichiometric as the temperature increases to the liquid region at about 1400 °C, ranging from Fe0 9 5 0 to Fe0 8 8 0 and does not seem to reach the stoichiometry of FeO. It is demonstrated by X-ray lattice parameter that an appreciable fraction of the lattice sites normally occupied by iron ions are vacant.01] Wustite is a p-type metal deficit semi-conductor and always has 5% to 16% iron sites vacant. With the non-stoichiometry of this extent, wustite has an extremely high mobility of cations and electrons. In order for the crystal to be electrically neutral, there must be electron holes which correspond to the formation of trivalent iron ions. Since the oxygen lattice is thought to be perfect, two Fe + + ions must be promoted to F e ^ ions for each vacant iron site in order to maintain electrical neutrality. Universal agreement has been reached on the transport mechanism in wustite. Ionic transport in wustite occurs essentially by the interchange of vacancies with normal lattice sites. Wustite is stable only above about 570 °C and it will not exist below 570 °C under equilibrium conditions. Although the normal transformation temperature is about 570 °C, thin films of FeO on iron exists at temperatures as low as 400 °C. This lower limit of existence is strongly dependent upon film thickness and shifts toward 570 °C as the film becomes thicker.[I2] During slow cooling of the hot steel, most of FeO slowly decomposes into a eutectoid mixture of a-iron and magnetite which is most predominant after cooling according to 4FeO-+ Fe304 +Fe- However the exact reaction can vary depending on the Ch.2 Literature Review 15 cooling rate and oxygen content of parent wustite. During rapid cooling, the reaction may take place only partially or not at all, such that wustite can exist at room temperature. In practice, air cooling of sheet, plate and bar products is fast enough to allow only partial decomposition of wustite.[16] (b) Magnetite Magnetite, nominally represented as Fe 30 4, is the intermediate phase of scale. It is also a p-type conductor and is the main equilibrium constituent of scale below 570°C. Magnetite may be considered as a slightly distorted close packed cubic structure of oxygen ions with metal ions in certain interstices. Among the sites available to the iron ions in the ideal unit cell of the spinel structure, eight are tetrahedral positions and sixteen octahedral positions. At the stoichiometric compositions, the eight tetrahedral sites are occupied by Fe + + + ions while the distribution of the remaining eight Fe + + + and Fe + + ions at the sixteen octahedral positions is random and fluctuating. So its crystal structure is cubic inverse spinel with Fe 2 + occupying octahedral sites and Fe3+tetrahedral sites. Magnetite has a density of 5 to 5.4g/cm3. It exists as a metal deficient oxide but at a much smaller level than wustite. (c) Hematite Hematite, Fe 20 3, is the outermost layer of the scale and has the highest oxygen content. It exists in two forms. a -Fe 2 0 3 is a slightly distorted, hexagonal close-packing of oxygen ions with the metal ions residing in the interstices, and y-Fe203 has a cubic structure.112] However, the y-Fe 20 3 is a metastable form. The only form stable at high temperature is a -Fe 2 0 3 . It is an n-type conductor. The density of hematite is about 5.24g/cm3. Corresponding to the properties of three oxides of iron, thicknesses of various scale layers show some regularity. With pure iron above 700 °C, the three phases of iron oxides consist Ch.2 Literature Review 16 of three well-defined, parallel-sided layers with wustite occupying about 95% of the scale layer, magnetite about 4%, hematite about 1% at equilibrium as shown in Fig.2.5. [ 5' 1 2' 1 7 ] This balance could be disturbed by the presence of carbon in the alloy and the scale layers are much less regular in thickness and are often highly porous. [ 1 7 ] It is seen that at lower temperatures, the magnetite layer grows at the expense of wustite. If the scale formed is overoxidized, it will contain higher proportions of magnetite and hematite than are present in the scale formed on pure iron. [17J 100 FeO 90 00 70 -60 -50 -40 -30 -20 -10 -FejO, 0 i , f _ 1 r - * ! ° J 600 700 800 900 1 000 1100 1 200 TEMPERATURE, °C Fig.2.5 Approximate percentages of wustite, magnetite and haematite on iron oxidized in oxygen[5] It has been well established and generally accepted that:^1'12'13,15'18'191 (1) over the range 200-570 °C, iron forms a two-layer oxide scale consisting of an inner magnetite layer and an outer hematite layer with the inner layer being slightly thicker. The overall oxidation rate of iron in oxygen is to a great extent determined by the growth rate of magnetite; (2) above 620 Ch.2 Literature Review 17 °C, iron forms wustite which has a more defective structure than magnetite and hematite. The scale is mainly wustite with thin outer layers of magnetite and haematite. The overall rate seems to be determined principally by the rate of wustite growth. The three oxides are formed simultaneously and the ratio of thicknesses of the three layers is claimed to be independent of time and temperature, i.e. the relative thicknesses of these layers remain unchanged during parabolic oxidation.[15] The growth of each of the three oxides in oxygen rich atmosphere was found to follow the parabolic rate law. (3) Between 570 and 620°C, as the temperature increases, the rate controlling layer changes from magnetite to wustite. The corresponding parabolic rate constant is a complex function of temperature. 2.1.4 Kinetics of Oxidation Two kinetic growth rate laws are often observed: parabolic and linear growth, (a) Parabolic Growth Rate Law Wagner's theoretical treatment of the oxidation process permits evaluation of the rate constant of oxidation from certain crystal properties. However, these values are usually not known or incomplete so that it is still necessary to actually run oxidation tests and measure the growth rates.tl4] Therefore the real value of Wagner's analysis lies in providing a complete mechanistic understanding of the process of high temperature oxidation under the conditions set out. In a general oxidizing atmosphere with enough oxygen potential, parabolic growth is followed.1'2'20' By analyzing the particle flux in chemical potential, the following parabolic rate equation applies[8] Ch.2 Literature Review 18 x2=kpt + C (2-10) K ^ l ^ d ^ (2-11) (2.12) RT 1 t'M RT where # parabolic rate constants C integration constant x weight gain t oxidation time D M diffusion coefficient for metal ions MM>M"M chemical potentials of metal ions at metal-scale and scale-gas interfaces respectively C M mole concentration of metal ions Z M charge of metal ion F Faraday constant R gas constant (b) Linear Growth Rate Law If the oxygen potential is low enough, the process and reaction on steel surface could be rate controlling. In this case, a linear rate law could be observed, which could be expressed as x = k,t + C (2-13) The rate of delivery of oxygen to the surface depends not only on the composition of the atmosphere but also on the flow rate. The oxidation rate increases progressively with gas flow rate until a critical flow rate is reached. Beyond this, the oxidation rate is parabolic and Ch.2 Literature Review 19 remains constant. Some results quoted by Sachs and Tuck [5] were such that the critical flow rate was 2.54 cm/sec for C 0 2 and air and 11.68 cm/sec for steam. However a critical flow rate of 4.20 cm/sec was also reported for oxidation of steel in air. p l ] In the course of actual oxidation, part of the scale/steel interface could become detached. Various areas of the sample surface oxidize in radically different manners.p4] For the detached part of scale, the kinetic rate cannot accurately be expressed analytically. The closest rate equation would probably be linear. For the adherent part of scale, the oxidation is by solid diffusion and the rate equation would be parabolic. Therefore, part of the sample is oxidizing according to the parabolic law and part in a pseudo-linear fashion. As a result, a combination of parabolic and linear growth is frequently encountered. Typical scale growth curves are as shown in Fig.2.5. The shape of the sample, the impurity level in the iron and the atmosphere in which the sample are raised to the oxidation temperature are important in determining the oxidation behavior of iron in some oxidizing atmospheres such as oxygen+steam.[22] Therefore, in oxidation study, the time and the surface for the beginning of oxidation should be precisely defined. Above 570 °C, the oxidation of iron becomes more complex owing to the wide variation of the composition limits of wustite phase field with temperature. The oxidation process is dependent on diffusion and in turn the diffusion rate varies with the concentration gradient in the growing phase. That is, for the same thickness of wustite formed at different temperatures, there will be an effect on the growth rate due to the fact that the composition limits of the phase will be different, in addition to the normal increase of the average diffusion coefficient with increase in temperature.1121 Ch.2 Literature Review 20 100 -r 90 -0 10 20 30 40 50 60 70 80 ' Oxidation Time (minutes) .'' Fig.2.6 Typical weight gain curves during oxidation of low carbon steel in a gas mixture containing 6% 0 2 and 94%N 2 [ 2 1 ] The products of combustion of oil or gas are mixtures of C0 2 , steam with some excess 0 2 or CO and contaminants such as S0 2. In atmospheres consisting predominantly of products of combustion, particularly if the combustion has been efficient, the oxygen potential is substantially lower and the supply of oxygen absorbed from the atmosphere is reduced. In this case, the rate of absorption of oxygen into the scale may become slower than the rate of arrival of iron by outward diffusion through the scale. This will affect both the kinetics of scale growth and scale structure. The rate of oxidation will be very substantially reduced, because the adsorption of oxygen from the atmosphere is now a much slower step in the reaction sequence and is in fact the rate controlling step. When a phase-boundary reaction such as the adsorption of oxygen at the outer surface is rate-controlling, the rate of scale growth is independent of scale thickness and the amount of scale formed is directly proportional to time. A linear oxidation rate has been found in some oxidizing atmospheres.[7,21,23] Ch.2 Literature Review 21 In parabolic oxidation where the rate-controlling step is the arrival of iron at the surface, the oxidation mechanism favors the formation of a scale with smooth interfaces and uniform thickness. Local irregularities of scale are gradually corrected. If the scale is thin, diffusion is faster than elsewhere and the scale grows more quickly until it has caught up with thicker scale elsewhere. However, in the earlier stages of oxidation when diffusion is faster than the surface reaction, this regulating mechanism does not operate. The scale of pure iron oxidized in C 0 2 at initial stages has a smooth metal/oxide interface but a very irregular outer scale surface.^ In principle, a changeover from a process controlled by the surface reaction to one controlled by diffusion must occur in all atmospheres. In air or oxygen, the surface reaction is so fast that the linear stage is not perceptible and the diffusion control has taken over by the time the reliable experimental reading can be taken so that oxidation is virtually parabolic throughout. In steam, above 900 °C the linear stage is also difficult to detect, but the gradual transition to parabolic oxidation is very evident. In carbon dioxide, the linear oxidation predominates. The duration of the linear reaction rate is shorter the higher the temperature. Weight-gain data[5'22] have been obtained for oxidation of pure iron at 950 °C in oxygen, air and steam as shown in Fig.2.7. It was found that the scale was thicker in oxygen than in air and within the early stages, say the first three hours, the oxidation in oxygen and air was more severe than in steam. But after a certain time, the oxidation in steam exceeded that in air or even in oxygen. The scaling rate in oxygen + steam was somewhat faster than that in oxygen alone, especially after 75 min oxidation. Evidently, steam or hydrogen within the scale prevented the loss of contact between scale and metal which must be occurring on the samples oxidized in oxygen. Possibly this loss of contact which may be small and local Ch.2 Literature Review 22 during the early stages is prevented by the increased plasticity of the oxide formed in a steam-bearing atmosphere. •t, min Fig.2.7 "Specpure" iron oxidized at 950 °C [ 2 2 ] Creep or plastic flow of the oxide can only occur by means of the movement of dislocations in the iron and oxygen particle lattice. An increased ability to flow in the presence of steam or hydrogen can only imply that the steam or hydrogen either increased the number of dislocations or the number of dislocation sources or sinks or the mobility of those already existing dislocations^221 The oxygen transport inwards, as suggested by Rahmel,[24] is through unstable micropores which are being continually formed and healed as the oxidation/reduction mechanism takes Ch.2 Literature Review 23 place. These invisible micropores could act as both sources and sinks for dislocation movement and thus enable the creep of the scale to occur and at the same time allow the gases to be transported through the scale to the scale/metal interface in these mobile micropores. So, from Rahmel, the effect of steam to prevent the decrease in oxidation rate when a gap forms at the scale/metal interface is due to two contributory causes: (a) the scale formed in steam is more plastic, creeps more readily and therefore does not form gaps so soon; (b) the gas reactions in the pores permit oxygen transport across the scale and thus counteract the disruption of the diffusion path. Strain can have an effect of accelerating oxidation.p5] One effect of applied strain is to enhance oxidant ingress. The observed grain boundary sliding and subsequent infilling with magnetite is one possible cause of enhanced oxidant ingress. The oxidation behavior of pure iron is significantly modified by the presence of quite small alloying additions. In general, the oxidation rate of steels is much slower than that of pure iron and it is not easy to establish which component is responsible.[5'26] In oxidation study of steels, some subsequent deceleration of oxidation can be attributed to porosity formation or detachment and the subsequent rapid increase in weight almost certainly is due to an oxygen leak.[4] Carbon is peculiar in forming a gaseous reaction product. The main effect of carbon on oxidation rate is to make it more erratic.[5] Gas pressure of C 0 2 and CO in scale may cause gross cracking so that the atmosphere gains access to the metal core and the rate of oxidation is increased. On the other hand, gaps and voids will be less likely to heal if they contain gases. Unless the gas contains sufficient carbon to facilitate oxygen transport, the earlier formation of gaps and the stabilization of gaps, may result in a slower oxidation rate. Ch.2 Literature Review 24 In the early stages of an oxidation process, a transient temperature rise of the sample of up to 40-50 °C can be observed. It results from the exothermity of the chemical reaction. The average value for the heat of formation of wustite at 298 K is -63.5 kcal per mole.p 7 ] 2.1.5 Secondary Aspects During Oxidation Process (1) Separation at Scale/Metal Interface In the process of oxidation of iron, the contact between scale and metal can only be maintained by "creep" of the scale. If the creep is prevented orinhibited(this is more likely to occur at lower temperatures), the contact is lost and a gap is formed at the scale/metal interface. Iron cannot bridge that gap and the absorption and diffusion of iron are disrupted. If the gap extended along the entire interface, the metal loss would stop completely. Nevertheless, the iron ions already in oxide would continue to diffuse out of the wustite phase until the concentration is uniform throughout the layer at a level corresponding to equilibrium with magnetite. Meanwhile, the slowing down of diffusion would lead to a growth of magnetite within the wustite layer. Fig.2.8a shows fully coherent scale consisting of three parallel layers. On cooling from oxidation temperature, some magnetites have been precipitated from the wustite, indicating the oxygen gradient across this layer. Fig.2.8b shows a scale grown under similar experimental conditions, in which a gap has formed. The scale is thinner, there is relatively more magnetite and the precipitation of magnetite on cooling has occurred throughout the wustite, indicating the uniform saturation with oxygen and deficiency in iron.[5] When Ch.2 Literature Review 25 cooling, saturated wustite precipitates magnetite particles but pure magnetite does not undergo any structural changes. Loss of contact over the entire interface stops the loss of metal instantaneously, but the absorption of oxygen continues although at a decelerated rate. In practice, gaps appear locally at the interface. They may be healed and others may appear elsewhere so that their effects on the oxidation kinetics and structure are not consistent and reproducible. The effect of local separation of scale from metal is to increase locally the amount of the higher oxides in the scale.[12] The fall in plasticity with lower temperatures presumably increases the probability of gap formation and the proportion of the interface area where contact is lost and thus may account for the increase in the proportion of magnetite at lower temperatures. Fig.2.8a Adherent scale formed on pure Fig.2.8b Nonadherent scale formed on iron at 950 °C in oxyegn.[5] pure iron at 950 °C in oxygen.[5] (2) Phase Transformation and Structures When the scale on the iron surface is heated to a higher temperature than the transformation temperature, the transformation of iron from a phase to y phase has certain effects on the Ch.2 Literature Review 26 oxidation J 1 At the transformation temperature, where the iron suffers a contraction, the film may become partially separated from the iron. If the film is plastic at the transformation temperature, then no rupture will occur. The effect of the partial separation of scale and iron is to reduce the area of contact, thus reducing the supply of iron ions to the scale. This will reduce the overall rate of oxidation and increase the percentage of higher oxides in the scale. The rates of growth of iron oxide layers are rather insensitive to changes in the substrate. There seems to be no great change in the rate of growth of wustite and magnetite when the substrate changes from a to y iron. [ 1 2 ] Baud et a l [ 1 9 ] investigated the effect of structures on the oxidation rate and morphology by oxidizing the austenite or a mixture of ferrite and cementite at the same temperature. For Fe-C alloy, the scale adherence depends not on the temperature, but on the structure of the metal. When the carbon is in solid solution in the metal in the y state, the scale is adherent. On the other hand, the scale formed on a matrix made up of ferrite and cementite is always detached. The detachment could be alternatively explained by the structure difference between wustite and cementite. Wustite is face-centered and cementite is orthorhombic. The structure difference produces sufficient incohesion between the two phases. Scale detachment in turn decreases the oxidation rate. Hence, at a tempearture near that of the transformation point A[ where two structures can exist(a+y or a+cementite), the oxidation rate depends not only on temperature, but also primarily on the structure and hence on the thermal history of the sample. (3) Decarburization In the reheating furnace, oxidation is accompanied by decarburization, i.e. oxidation in reheating furnaces removes carbon from the surface layer of the steel. The effect is a Ch.2 Literature Review 27 deterioration of some important mechanical properties of the final product. Carbon is oxidized more rapidly than iron.[5] One of the possible ways to restrict decarburization seems to bind carbon as carbides by introducing some carbide forming alloying elements into the alloy matrix. Steels rich in alloying element for carbides, e.g. TiC, VC, NbC, TaC or WC provide considerable resistance to oxidation therefore decreasing the oxidation rates.p6] 1.2 percent carbon alloys generally have the highest concentration of voids in the scale which is presumably due to rapid and higher carbon losses during oxidation. The oxidation of iron is controlled at the beginning by the rate of surface reaction. Carbon dioxide, either from the atmosphere or as a product of reactions in the scale, is the main decarburizing medium. Initially, the decarburization is linear and the rate of carbon loss increases with the partial pressure of carbon dioxide. As carbon is lost from the surface, it is replenished by diffusion. As the decarburized layer grows thicker, the rate of arrival of carbon at the surface slows down until it becomes slower than the surface reaction and determines the rate of decarburization so that the rate of decarburization becomes parabolic. In the course of decarburization, the products of all decarburization reactions are gaseous and the gas must escape or build up a pressure corresponding to equilibrium with the carbon activity of the steel surface. If the scale withstands this pressure, the surface reaction is suppressed and decarburization comes to a halt. If the scale cracks under the pressure, the gases escape and a sudden increment of decarburization occurs, accompanied by a sudden increment in oxidation as the atmospheres gain access to the metal surface. On iron, there is a delicate balance between the strength of the scale and the gas pressure. Ch.2 Literature Review 28 Above 900 °C, the scale formed is not highly adherent.1 9 ] Decarburization and oxidation occur simultaneously. The reaction rates are high and the oxidation rapidly consumes the decarburized metal. Between 730 and 900 °C, even in the presence of an oxidizing atmosphere, an enrichment of carbon at the scale/steel interface can be observed. The oxide layer formed in this temperature interval is adherent, compact and impermeable to gaseous oxygen. As the carbon can not diffuse through the scale, it remains in the metal which becomes carbon-enriched. 2.1.6 Effects of Alloying Elements The oxidation behavior of pure iron is significantly affected by even small alloy additions. In general, the oxidation rate of steels is less than that of pure iron. The effect of a given alloy element can depend partially on whether it is more or less noble than iron. More noble elements such as Cu and Ni can become concentrated at the metal/scale interface and less noble elements such as Si can give rise to internal oxidation in the scale. Impurities and alloy elements can (a) facilitate separation of the oxide scale layer from the steel substrate; (b) form new oxide phases within steel substrate(internal oxidation); (c) form new phases within the scale and (d) concentrate at the scale/steel interface or become occluded in the scale if they are more noble than iron. (1) Effect of Silicon Many elements present in commercial steels form separate phases in the scale. The most important of these is silicon. Although killed steels contain only about 0.25% silicon, this is sufficient to form pools and extended stringers of iron manganese silicate in the scale layer.p] Ch.2 Literature Review 29 In severely oxidizing atmospheres a layer of this phase covers the metallic core. While contact between scale and metal is maintained, this layer remains at the metal surface. When the scale is fissured and inward migration of oxygen occurs, some wustite grows beneath the silicate phase. In many cases successive layers of silicate and wustite are formed. In steels with higher silicon contents a film of silica may form at the interface and the rate of scaling then becomes very slow. In atmospheres of low oxygen potential, when oxidation is controlled by the surface reaction, the absorption of iron into wustite at the scale/metal interface is relatively slow and the scale advances fairly slowly toward the metal. This gives the iron under the scale an opportunity to dissolve some oxygen from the scale. The oxygen diffuses inward and reacts with silicon in the iron, forming particles of silica or fayalite inside steel. It may lead to surface defects in rolled products. (2) Effects of Copper and Nickel Copper and nickel are oxidized less readily than iron. At the scale/metal interface, iron enters the wustite lattice while the nickel or copper is rejected. So the surface layer of the metal core is enriched in nickel or copper. This sets up a concentration gradient in the metal core. If the diffusion rate of the more noble elements in the metal core is faster than the rate of oxidation, the alloying elements gradually increase their concentration in the metal core with no great build-up at the metal surface. This can happen with copper. However, nickel does not diffuse rapidly back into the core. The nickel rich surface layer is more resistant to oxidation and scale advances inward by preferential oxidation along grain boundaries. The overall increase in scale thickness is not significantly slower, but particles and filaments of nickel-rich metals remain entangled in the scale. The volume of entangled metal is greater in Ch.2 Literature Review 30 less severely oxidizing atmospheres. The presence of nickel entanglement in the scale makes it adhere during rolling, so that it is sometimes forced into the metal, giving rise to surface defects. The effect of nickel up to 4.50% on the scale is merely to increase metal entanglement. Copper is also sometimes found as discrete particles in the scale. Enrichment of copper in the surface layer may lead to the precipitation of molten metalic copper and this may cause intergranular attack, hot shortness and surface defects. (3) Effects of Chromium and Molybdenum In the normal range of Mo content below 1%, the oxidation rate is reduced by the formation of a Mo-spinel(Fe2Mo04) layer at the scale/metal interface. Many low-alloy steels contain only about 0.25%Mo and there is not enough spinel to form a continuous layer at the interface. The presence of precipitates(or inclusions) in ordinary steels leads to poor scale adherence.[19] Chromium forms a similar spinel and since content up to 5% is quite common, the spinel often forms a significant part of the scale. It inhibits outward diffusion to such an extent that the little iron that does reach the outer scale layer is readily oxidized to magnetite and hematite. At higher chromium content, a protective film of Cr 2 0 3 is formed. (4) Effects of Alloying Elements Resulting in Internal Oxidation There are two types of oxidation in which oxides are formed within an alloy: internal oxidation and intergranular oxidation.[28] The internal oxidation describes the process in which formation of oxides occurs at a reaction front which advances in a uniform manner into the alloy and is not apparently affected by grain boundaries or other microstructural features. In intergranular oxidation, oxides form along alloy grain boundaries to a greater Ch.2 Literature Review 3 1 depth than in the bulk grains. Heterogeneous precipitation of oxides in the grain boundaries within the internal oxidation zone is not in itself intergranular oxidation. Precipitation of internal oxides can have very significant effects on the mechanical properties of the substrate, particularly if the strengthening elements are removed by oxidation. As the concentration of the alloying element which forms the internal oxides are increased, there is eventually a transition to external oxidation. For iron, chromium, silicon, vanadium or aluminum could form internal oxidation. Internal oxidation usually occurs in the alloy beneath an external scale of the more noble metal oxide. Oxygen saturates the surface of the alloy, either directly from the atmosphere or by dissociation of the iron oxides, diffuses inwards through the internal oxidation zone, and reacts with the solute element, precipitating its oxide at the internal oxide/alloy interface which thus advances inwards as shown in Fig.2.9. Fig.2.9 Precipitates of Si0 2 forming internal oxidation[5] When internal oxide precipitates form in a metal lattice, there is an increase in volume since most oxides have a Pilling-Bedworth ratio greater than unity. Hence high compressive Ch.2 Literature Review 32 stresses are generated at the internal oxidation front and within the internal oxidation zones. The Pilling-Bedworth ratio(PBR) is the volume of oxide divided by the volume of the iron from which the oxide is obtained. But with oxidation of iron, the PBR is of no significant use since the iron ions diffuse outward, not inward. (5) Effect of Carbon The effect of carbon is to reduce the oxidation rate of pure Fe. [ 1 7 ] The oxidation rate can be expected to be reduced by approximately two orders of magnitude depending on the percentage of magnetite in the scale layer. In effect, the presence of carbon in the alloy lowers the activity of Fe at the inner surface of the scale to the point where magnetite is the stable phase. The decrease in oxidation rate of Fe-C alloys is clearly related to some modifications to the alloy-scale interface. The rejection of carbon at the alloy/scale interface caused poor contact between scale and alloy. The poor contact increases the proportion of magnetite in the scale.1271 Alternatively, it may be due to general loss of contact between scale and alloy, aided by the formation of CO and C 0 2 at the interface. If alloy structure could dissolve more carbon, there is less possibility to get a separated scale. (6) Effects of Sulfur, Phosphorus and Manganese Sulfur dioxide is extremely corrosive, but in the amounts normally present in industrial furnace its effect is marginal. Higher sulfur contents in steels can increase oxidation rate. Effects of sulfur segregation at the interface on scale adhesion can be explained by an influence on the interfacial bonding and thus on the fracture toughness of the interface. Phosphorus lowers the oxidation resistance of iron slightly.[5] Manganese can substitute for iron in wustite and magnetite and the effects attributed directly to manganese are thought to be slight. Ch.2 Literature Review 33 2.2 Morphology and Microstructure of Scale on Steels A typical structure of scale of mild steel in air is shown in Fig.2.10. Obviously, the scale is severely fissured and blistered and consists largely of magnetite and hematite.P] The fissuring of these scale has been so severe that the scale is largely converted to magnetite. Oxygen continues to be absorbed at the outer surface and the weight continues to increase, but diffusion is slowed down because the diffusion path is mainly through magnetite and even in wustite the concentration gradient is shallower. In Goursat and Smeltzer's study of the oxidation of pure iron at 800 °C with oxygen pressure range of 0.3-152 Torr, magnetite consists of small grains whereas hematite appears as even smaller grains and platelets.[29] The grains in the magnetite layer were approximately one-tenth the average size of the wustite grains. Fig.2.10 Scale formed on mild steel in air at 1000 °C after 250 min [ 5 ] Ch.2 Literature Review 34 Hematite grew preferentially from sites at the grain boundaries of the magnetite surface. It grew as islands and finally covered the entire surface. Hematite within the expanding islands grew by vertical and lateral enlargement of the localized growths. They appeared as whiskers or platelets. Al l of the oxides finally developed into polycrystalline layers. Dunnington et al [ 1 4 ] studied the oxidation of pure iron. The separated scale had a variety of relative thicknesses. No constant relationship among the thicknesses of oxide layers existed on the separated oxide. It was seen that different structures occurred at different locations on the same sample. In the investigation of Himmel et al [ 1 1 ] , the resulting oxides were always coarse polycrystalline with a well-developed columnar grain structure. For the most part, the columnar grains appeared to be oriented so that the axis of preferred growth was normal to the flat surface of the sample. The average oxide grain diameter was generally between 1 and 2 mm for the wustites and of the order of 0.2 to 0.5 mm for the magnetites. Nazrazadani and Raman [30] also found similar columnar structure. Tuck et a l [ 2 2 ] classified the scale structures found into three main types as follows: (1) The oxygen atmosphere type This scale structure is characterized by the loss of contact between scale and metal that occurs on one complete face in the case of sheet(Fig.2.11a) or one part of the circumference in the case of rod-shaped samples(Fig.2.11b). The loss of contact resulted in a greatly increased thickness of primary magnetite, and saturation of the wustite with oxygen. Ch.2 Literature Review 35 Fig.2.1 lb Oxide scale on iron wire after 14 hours at 805 °C in pure oxygens (2) The oxygen+steam atmosphere, continuous contact type This is shown in Fig.2.1 lc and is characterized by maintenance of contact between scale and metal on both faces, even at long exposure times. Therefore there is only a very thin layer of Ch.2 Literature Review 36 primary magnetite, no sign of oxygen saturation of wustite through all thickness, nor any large-scale pore formation. (3) The oxygen+steam atmosphere, porous type This type of structure also shows a thin primary magnetite layer, but its chief characteristic is the considerable pore formation(Fig.2.11d) and in the case of cylindrical samples, the uniform consumption of the metal around the circumference. As a rule, scaling of mild steel samples in an oxygen+steam atmosphere always leads to the pore-containing structure, regardless of the shape of the test piece or the nature of the preheating atmosphere. While in the case of sheet samples of the "Specpure" iron, the continuous contact and pore-free structure is always found. Fig. 2.1 lc "Specpure" iron sheet scaled at 950 °C in 88% oxygen, 12%steam[22] Fig. 2.1 Id BISRA pure iron sheet scaled at 950 °C in 88% oxygen, 12%steam[22] Ch.2 Literature Review 37 Oxide scale develops porosity and microchannels that permit inward transport of molecular species from the ambient gas even under conditions when there is no evidence of cracking of scale.[31] Such porosity and microchannels develop as a result of grain growth and of plastic deformation(grain-boundary sliding, diffusion creep etc.) under compressive stresses in the scale. The presence of small amounts of impurities enriched at grain boundaries in the scale may greatly affect deformation and mechanical and transport properties in scale. Numerous studies of high temperature corrosion of metals and alloys have shown that protective oxide scale may gradually become pervious to molecular species in the ambient gas.[31] This penetration of gaseous species through protective scale can be explained by scale cracking: the cracks allow an inward penetration of gaseous molecules into or through the scale. This is a valid explanation and mechanism for many oxidation processes. However, the gaseous penetration takes place also when the oxide is able to deform plastically and when there is no evidence of cracking of the scale. In order to explain this phenomena, Mrowec [ 3 2 ] proposed that microchannels might develop in growing scale. The proposed model is illustrated in Fig.2.12. It was assumed that the scale grew by outward diffusion of metal ions and as a result, voids eventually developed at the scale/metal interface. Grain boundary diffusion of the metal ions was assumed to be much faster than lattice diffusion and as a result, the grain boundary above the void opened up and gradually formed a mcirochannel. It remained open as long as the chemical potential of the oxygen was equal at the surface of the channel and in the neighboring lattice. Oxygen might penetrate the channel to the inner part of the scale. However, this model has difficulty in explaining how the microchannels can remain open. To overcome this, Kofstad further Ch.2 Literature Review 38 proposed that the development of porosity and microchannels was a result of plastic deformation and creep in the scale caused by the growth stresses in the scale.p l ] 0 2 ( g j O x i d e Rapid metal diffusion a long gra in boundary - 0 2 ( g ) C o n s t a n t po ten s i a l I n n e r o x i d e 2 5 7777777, 02<g) i Open ing of - grain boundary 0 2 ( g ) M i c r o c r o c k . i n w a r d O2 t r a n s p o r t Stabi l izat ion of micro-crock. inward O2 transport, outward metal transp. Fig. 2.12 Model for formation of microchannels in oxide scales by preferential outward diffusion of metal ions along grain boundaries^] 2.3 Modeling of Oxide Growth Process There is extensive literature which reported the measured oxide growth data or calculated the rate C Onstants. [ 7' 1 1~ l 4" 1 7' 1 8 , I 9 , 2 1' 2 2' 2 3' 2 6' 2 9' 3 4' 3 5 ] However due to the complexity of the problem, few papers tried to predict oxide growth under nonisothermal oxidation and no satisfactory results have been obtained. Yurek et a l [ 1 8 ] put forward a model dealing with the simultaneous formation of two layers of oxide on metal and Garnaud and Rapp [ 1 3 ] applied this model to the oxidation of iron when only wustite and magnetite formed. The ratio of oxide thicknesses could be expressed by^13] Ch.2 Literature Review 39 1 Fe304 FeO 3 ^FeQ^p.Fe30A ^ V ^Fe-p^p.FeO Fe3Ot FeO 3 VFepk p.Fe-fii ^ V ^Fe3Ot^p.FeO + 4' ^p.Fe3Ot ^p.FeO 1/2 (2.14) where ^ thickness of magnetite, the outer oxide ^ thickness of wustite, the inner oxide VF Q molar volume of magnetite v molar volume of wustite Y FeO k parabolic rate constant for the exclusive formation of magnetite on p.Fe3Ot wustite k parabolic rate constant for the exclusive formation of wustite on iron Kp.FeO r The rate constant for the growth of magnetite on wustite by the exclusive diffusion of cations was given by 1 1 2J S (2.15) where n the diffusion coefficient of Fe in magnetite Fe.Fe3Ot P° p> the oxygen pressure at the oxygen/magnetite and magnetite/wustite Ol ' ®2 interfaces Abuluwefa et a l [ 2 1 ] expressed parabolic rate law for wustite growth in oxygen-nitrogen mixture as follows: , _6PFeoK . • • _ v ^ "-p.FeO—v , R 2 ^ Fe1*^ FeOIFe&i yFelFeO) M FeO DFeU = 0.118e -124300/RT (2.16) (2.17) where p density of FeO Ch.2 Literature Review 40 j l / molecular weight of atomic oxygen n* iron self-diffusion coefficient in iron oxides v iron ion vacancy concentration at Fe/FeO interface SFe/FeO J And the linear rate constant was expressed as follows: kl=M0kMTC{CG0i-Cli) (2-18) where h mass transfer coefficient MTC Q molar concentration of oxygen in gas, G for bulk gas, * for sample surface Re, Sc Reynolds number and Schmit number D diffusion coefficient of oxygen in gas mixture Similar formulae for oxidation in carbon dioxide and steam were also formulated.1231 In his most recent study of oxidation,[7] Abuluwefa et al used a thought similar to the present work to predict the scale growth with non-isothermal oxidation. His model assumed that scale layers formed at lower temperatures had no effect on scaling rates at higher temperatures. This assumption may be correct for initial linear oxidation period which is almost invisible for oxidation in air. And even for linear oxidation, this model will give a large error if scale/steel separation develops. Parabolic oxidation rate law is so extensively followed. Therefore a model applicable to both linear and parabolic rules or any combination of them should be developed. Al l of the above methods attempted to use somewhat pure theoretical approaches to simulate the scale growth. Thus a lot of material properties have to be measured first and those Ch.2 Literature Review 41 measured data are applicable only to limited conditions. This process itself will introduce certain inaccuracy. It is very hard to obtain those properties, not mention their completeness. So a possibly feasible approach for developing a practical model is to study in detail one important steel grade under practical oxidizing conditions and establish a special model just for this steel grade. CHAPTER 3 SCOPE AND OBJECTIVES The oxidation of iron and steel has been the subject of extensive studies. However most previous research concentrated on oxidation kinetics, diffusion mechanisms, microstructure, and oxide properties. Due to complexity arising when oxidation is considered as a whole process, little work has been done on modeling oxidation with the aim of developing software usable in a steel plant. The goal of this investigation is to develop a model of the oxidation of IF steel for any thermal history based on the isothermal oxidation data. Practical methods have been adopted to extract meaningful results. To accomplish this, the following specific tasks have been carried out: 1. Sixty isothermal oxidation runs of IF steel were conducted to obtain weight gain-data. Macro-characteristics were recorded and about seven hundred SEM photographs were taken for the study of micro-structures. 2. The isothermal weight-gain data and oxide structures were analyzed systematically to provide a basis for modeling. 42 Ch. 3 Scope and Objectives 43 3. Temperature evolution of the scale/steel system has been modeled with FEM which serves as the basis for scale growth and strain-stress models. 4. A scale growth model was developed to predict the weight-gain data for oxidation in air with any thermal history. 5. The above models were implemented into a C++ code which consists of about 2,500 statements. CHAPTER 4 EXPERIMENTAL INVESTIGATION This chapter presents details of the experimental setup and procedures used and discusses the experimental data and observations. The resulting conclusions serve as the basis for modeling. 4.1 Experimental Setup and Procedures (1) Experimental Setup The experimental setup in the present study is shown schematically in Fig.4.1. It is composed of five major components: a vertical tube furnace, a mechanical jack, temperature recording system, weight gain recording system and atmosphere control system. A vertical tube furnace was used to heat the samples and provide isothermal temperature from 500 to 1200°C. The alumina tube in the furnace has an inner diameter of 56 cm which is large enough to accommodate large samples. The furnace profile was obtained before 44 Ch.4 Experimental Investigation 45 experiments. The furnace used has an isothermal zone of 80 mm around the middle section. A jack was used to raise and lower the sample inside the furnace. The sample temperature was monitored through a dummy sample which was placed beside the sample and was raised and lowered together with the sample. A type K thermocouple was inserted and sealed firmly onto the inner surface of the hole drilled in the dummy sample so that stable temperature reading could be recorded. This temperature was taken as the sample temperature which was used to decide whether the sample temperature has attained the desired value. Weight gain data aquisition computer Air/argon Dummy sample temperature acquisition Temperature . recorder furnace inicrobalance Dummy.. sample I ~1 Suspension wire sample jack Fig.4.1 Experimental setup for IF steel oxidation Ch. 4 Experimental Investigation 46 The Denver Instruments' M310 electronic microbalance, utilized to monitor the sample weight, has an accuracy of O.lmg. It provides a continuous recording of the weight gain of the sample during oxidation. A suspension wire of Ni80Cr20 with diameter of 1 mm was used to connect the sample to a hook at the bottom of the microbalance. For the temperature range used, the Ni80Cr20 suspension wire has good oxidation resistance. So, its influence on oxidation weight gain is negligible especially after it has undergone some period of oxidation. The atmosphere control system could introduce air or argon flow of various flow rates as required. Argon was used to minimize the amount of oxidation during heating up of the sample to the test temperature. The effect of argon flow on preventing oxidation was examined. It was found that within the initial first five minutes, the oxidation weight gain with argon protection is less than 1.50 percent of the oxidation weight gain with air at 1200 °C. While at 20 minutes, the oxidation with argon protection is about 5.0 percent of that with air. In this investigation, the desired temperature was attained within 3~4 minutes. So the effect of argon protection is large and it is absolutely necessary to use argon as a protective atmosphere so that the measurements are reproducible. Similar protective atmosphere were also used by other researchers. [ i U 4,2 I ,23] For continuous casting, the strand speed usually ranges between 2 to 6 m/min. Above a certain flow rate the oxidation rate of iron in air is constant and a value of 2.52 m/min was reported for this critical flow rate.f5-21] So an air flow of 3.1 m/min was selected for the isothermal oxidation experiments. The measured results could be applied to higher speed situations. If the air density at 1000 °C is considered, the actual flow speed is about 4 times the calculated value. Anyway it is above the critical flow speed. Ch.4 Experimental Investigation 47 The chemistry of IF steel used is as follows: 0.002%C, 0.106%Mn, 0.010%P, 0.008%S, 0.001%Cu, 0.003%Sn, 0.01%Ni, 0.019%Cr, 0.002%Mo, 0.059%Ti, 0.009%Nb, 0.033%A1 and 0.004%N. The dimensions of the samples employed are 27.70x37.50x3.55 mm, which were cut from a rolled strip. This sample size is larger than almost all of the previous samples used in oxidation study by others. [7,19,21,22,23,26,27,35] The cut samples were ground and polished successively with polishing paper of 120, 240 and 320 grit so as to remove possible residual stresses and obtain consistent surface condition. A hole of 2 mm in diameter was drilled in the sample to pass the suspension wire. With a hole of this size, the suspension wire could be easily removed from the sample after the oxidation experiment. In oxidation studies, two methods are usually used to determine the scale growth, the continuous weighing method and the microscopic examination method. The continuous weighing method results in an averaging over the entire surface. Oxidation kinetic studies by the continuous weighing is an averaging result which does not separate different reactions that may be occurring at different areas on the sample. The microscopic examination method may be made at a particular point so that if one section is reacting differently from another, this may be noted and measured. In this investigation, both methods were employed. (1) Experimental Procedures for Weight Gain Measurements Before the experiment was begun, the furnace was set to the desired isothermal temperature. The microbalance was raised high enough for convenient loading of the sample. The data acquisition software was set up for recording the weight-gain data and the temperature recorder was also adjusted to record the temperature evolution. This allowed identification of possible problems by checking the stability and correctness of all reading about the weight gain and the temperature. At the same time the sample was further polished with polishing Ch.4 Experimental Investigation 48 paper of 320 grit. The corners were polished into small fillets so as to reduce the geometric effects. Then the sample was rinsed with acetone and alcohol to remove the grease, dirt and residues associated with polishing. The sample was dried in air after alcohol cleaning. Just before the sample was loaded on to the hook, the furnace was flushed with argon twice and the argon flow rate was kept high enough about 10 m/min in this investigation so as to minimize oxidation before the experiment started. The sample was quickly loaded and lowered to the isothermal section of the furnace. Once the temperature of the sample reached the desired temperature, the argon flow was cut off and the atmospheric air was flushed in with a flow rate similar to continuous casting. This commenced the isothermal oxidation run. After the desired period of time, the sample was raised up either slowly or quickly to examine the effect of thermal stresses on cracking during or after cooling. After the sample was raised above the furnace, it was completely examined by the naked eye and as many as possible characteristics of the sample such as colors of sample surface, cracking, sample thickness and scale appearance, were observed and recorded. To obtain a complete set of oxidation data, isothermal experiments were carried out from 700 to 1200 °C with a temperature increment of 50 °C. Since 24 hours is sufficient to cover almost any steel process, the oxidation experiments usually lasted for about 24 hours. Air buoyancy is different at different temperatures. However it does not have to be considered since only the weight gain is the desired data and the isothermal oxidation is carried out at a constant temperature. (2) Experimental Procedures for Microstructural Examination For the microstructural observation, a Hitachi S-570 SEM microscope was used which provided satisfactory images for magnification from 25 to 2,000 times. The sample was Ch.4 Experimental Investigation 49 prepared according to the purposes of observation. SEM scanning was conducted on the outer surface, under-surface, fracture surface or polished cross-section of the scale or sample. To extract oxidation rules from observation, large quantities of images on various locations and at various angles were taken. For the observation of the outer surface or under-surface of a scale layer, the scale layer to be observed was detached from the sample by hands or tweezers. The detached scale has enough conductivity which makes special treatment of the scale unnecessary for observation. But attention was required to make it have a good contact with the SEM object platform so as to obtain clear images. SEM was adjusted to focus on the desired location and the image was selected and saved on disks. For the observation of a fracture surface of the scale, the fracture surface was prepared by making the scale fracture along the desired path. The fractured scale was then loaded into SEM. The fracture surface was observed as was done in observing the scale surface. For the observation of the cross-section of the sample, the whole sample was mounted in a metallurgical resin. Otherwise the scale would break and detach during polishing. Then the mounted sample was polished with the sand paper of increasing grit. Polishing cloth with 0.1 micron diamond solution was employed for polishing final observation surface. The cross-section of the sample was etched with 35 percent HC1 solution to show the microstructures of the sample. During the polishing, moisture was absorbed into the resin. The moisture must be completely removed by a drying oven since the SEM cannot tolerate any moisture due to its high vacuum environment. The resin is not conductive so it was necessary to coat a very thin layer of charcoal on the surface of the mounting material before SEM observation. After the Ch.4 Experimental Investigation 50 charcoal-coating, the mounted sample was loaded into SEM for the microstructural observation. 4.2 Growth Curves (1) Growth Curves The measured oxidation weight-gain data are schematically presented in Figs.4.2-4.9. At each isothermal temperature, more than three experimental runs were usually carried out to verify the reproducibility of the data and to obtain an average. It is seen that below 950 °C, the individual experimental runs show quite good data consistency and convergence(Fig.4.2) while above 950 °C, the data consistency is still good(Fig.4.4). Therefore the average weight-gain curves could be used to represent the weight-gain data at specific temperatures. Average weight-gain curves below 950 °C are shown in Fig.4.3. It is noted that the weight gain increases with temperature. And the oxidation rate(weight gain per unit time) is increased with higher isothermal oxidation temperature. The weight-gain curves generally follow a parabolic rate law pattern. The following parabolic rate constants give a quite good approximation to the actual weight gain curves: At 850°C, w = 0.00195f0 8 2 3 kg/m2 At 900°C, w = 0.0278?° 5 3 3 kg/m2 At950°C, w = 0.0789f0437 kg/m2 . With the average weight-gain curves above 950 °C which are shown in Fig.4.5, it was also observed that the weight gain increases with higher temperature. The rate constants for these weight gain courves above 950 °C are as follows: Ch.4 Experimental Investigation 51 for 0 ~ 500min for > 500 min for 0 ~ 500 min for > 500 min for 0 ~ 400 min for > 400 min for 0 ~ 200min for > 200 min for 0-190 min for > 190 min where w the linear rate constant of oxidation, kg/m2 t the time of oxidation, m m From the above regression results, it is seen that the oxidation at 900 °C and 950 °C essentially follows parabolic rate law since their exponential index is very close to 0.5. For oxidation above 950 °C, the weight gain curves follows an initial non-linear rate law(2~7 hours) which is not parabolic. And later they changed to a linear rate law. This has not been mentioned in the previous literature. The most frequently observed phenomenon is an initial linear stage plus a later parabolic stage. In general, the oxidation process above 950 °C is faster and less stable. Fig.4.6 combines the weight-gain curves below 950 °C and above 950 °C. It is found that after 28 minutes of oxidation, the weight gain at 1000 °C is less than that at 950 °C and after 400 minutes of oxidation, the weight gain at 1050 °C is less than that at 950 °C. This seems abnormal since the weight gain at higher temperature should be larger. This has not been mentioned in the previous publications. When the isothermal oxidation experiments were ^1000°C, w = 0.0921*"Jt" w = 0.698 + 0.000363* ^1050°C, w=0.159*0330 w = 0.959+ 0.000449* ,4*1100 °C, w = 0.260/°2 6 4 w = 0.902 + 0.000751/ ,4*1150 °C, w = 0.290*0250 w = 0.860 + 0.00128* ^*1200°C, w = 0.307*0293 w = 1.082 + 0.00198* Ch.4 Experimental Investigation 52 performed, Abuwefa et aK7] noticed that there was a decrease in the oxidation rate around 900 °C due to scale/steel detachment. But they did not systematically analyze this phenomenon as part of the whole oxidation temperature range from 0 to 1200 °C. A method for modeling this phenomenon has not been proposed. This phenomenon suggests that a different mechanism has probably occurred between 950 and 1000 °C in this case. Due to different oxidation mechanisms, it is not feasible to simulate the nonisothermal oxidation process by simply superimposing the isothermal oxidation data if the actual oxidation process involves two different mechanisms. One feasible solution is to obtain another set of oxidation data, i.e. the secondary oxidation data which is weight-gain data under one mechanism after a layer of oxide scale(primary scale) has formed under a different mechanism. The isothermal oxidation data obtained when only a single mechanism is operating is termed the primary isothermal oxidation data. The isothermal oxidation data obtained after some scale has formed under a different mechanism, is termed the secondary isothermal oxidation data. If the oxidation of the first five minutes is observed as in Fig.4.6a, it is seen that in general the weight gain is larger with higher temperature. However some weight gain curves cross each other. The weight gain curve for 1000 °C crosses the curves for both 950 °C and 1050 °C. And in a large part, the weight gain for 1150 °C is a slight less than that for 1100 °C. The present thesis concentrates on the long time and overall behavior. For more detailed study of very short time oxidation, some measurement and equipment such as sampling time and the speed of the jack used should be adjusted. Ch.4 Experimental Investigation 53 Fig.4.7 to Fig.4.9 show the secondary weight-gain data. The data with zero hours of oxidation under the primary growth mechanism are essentially the primary isothermal oxidation data. It is evident that the magnitude of the secondary oxidation rates is much lower than the primary isothermal rates. The secondary weight-gain curves are not strongly influenced by the temperature at which the primary scale is formed as long as the primary growth mechanism is the same(Fig.4.8). But the longer the sample is exposed to a different growth mechanism, the larger the difference between the resulting secondary growth curves. Therefore the thickness of the primary scale has a strong influence on the growth of the secondary scale. This supports the premise that if the nonisothermal oxidation process is simulated by simply superimposing the isothermal oxidation data, a large error would result, particularly when the actual oxidation process involves two different oxidation mechanisms. i ' ' i * i i i ' < • • > • ' • * • < I > < i i i i t i i i 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 time(min) Fig. 4.2 Three weight gain curves of IF steel oxidation at 950 °C Ch.4 Experimental Investigation 54 4.00 r 3.50 '-3.00 -D) 2.50 F time(min) Fig. 4.3 Average primary weight gain curves of IF steel oxidation below 950 °C 4 . 0 0 r 3 . 5 0 '-^ 3 - 0 0 : 1 E O) 2 . 5 0 -I 2 . 0 0 -1 . 0 0 0 . 5 0 0 . 0 0 ' ' ' ' ' i • i < i i • i i • i i i i , i , • • . . . . i I 0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0 1 1 0 0 time(min) Fig. 4.4 Three weight gain curves of IF steel oxidation at 1050 °C Ch.4 Experimental Investigation 55 4.00 0 100 200 300 400 500 600 700 800 900 1000: 1100 1200 1300 time(min) Fig. 4.5 Average primary weight gain curves of IF steel oxidation above 950 °C 4.00 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 time(min) Fig. 4.6 Average primary weight gain curves of IF steel oxidation at various temperatures Ch.4 Experimental Investigation 56 0.50 time(min) Fig. 4.6a Average primary weight gain curves of IF steel oxidation at various temperatures within first 5 minutes 4.0 r-3.5 -^ 3.0 -E D> 2.5 -0 100 200 300 400 500 600 . 700 800 900 1000 1100 t i m e ( m i n ) Fig.4.7 Secondary weight gain curves of IF steel oxidation(850 °C) Ch.4 Experimental Investigation 57 4 . 0 r -3 . 5 -3 . 0 -0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 : 8 0 0 9 0 0 1 0 0 0 time(min) Fig.4.9 Secondary weight gain curves of IF steel oxidation(l 100 °C) Ch.4 Experimental Investigation 58 4.3 Microstructures of Scale of IF Steel The microstructures of the scale developed on the surface of IF steel in this investigation are shown in Figs.4.10~4.25. It was found that below 950 °C, an adherent single layer of scale was formed(Fig.4.11c, 4.23d) while above 950 °C, nonadherent multilayers of scale was formed(Figs.4.14a, 4.18b, 4.19c). In the latter case, the scale layers from the outmost layer to the innermost layer are called the first layer scale, second layer scale and third layer scale and so on. The adherent scale has a relatively tight contact with the underlying steel surface while usually there is a gap between two successive layers of nonadherent scale. Though in most cases, nonadherent scale are somewhat porous, some do not look porous. Therefore "nonadherent" rather than "porous" is a more appropriate word to describe them. (1) Scale Formed Below 950 °C Below 950 °C, the adherent scale is formed. Fig.4.1 la-h shows the outer surface, fractograph and under-surface of the adherent scale and the surface of the steel base. The grains in Fig.4.1 la and Fig.4.1 lb are the hematite grains which form the outmost layer of the scale. There seems to be a different layer growing along the grain boundaries. This is more apparently manifested in Fig.4.19b. Probably this results from faster diffusion along the grain boundaries since there is a more rapid metal diffusion along grain boundariesJ31>33>36] Fig.4.1 lc and Fig.4.1 Id shows the characteristics of the fractograph of the adherent scale. The wustite layer in the adherent scale has a relatively smooth fracture surface indicating that the wustite grain boundaries are stronger so that fracture is through grains instead of along the grain boundaries. The hematite and magnetite layers in the adherent scale fracture along Ch.4 Experimental Investigation 59 the hematite and magnetite grain boundaries indicating that the grain boundaries are weak. Wustite usually accounts for more than 93 percent of the adherent scale. This was also reported in other literatureJ5-13'37] In some cases, some small areas ofwustite layer also show some fracture surfaces along wustite grain boundaries(Fig.4.12b), but most of the area does not. The strong adherent wustite still predominates. F i g A l l e , Fig.4.1 If and Fig.4.1 lg shows the under-surface of the adherent scale. It can be seen that even the adherent scale only has contact with the steel at some locations and the macro voids exist. Even within the region of contact, it is maintained on smaller needle-shaped areas or ball-shaped areas(Fig.4.12d). Consistent with this under-surface characteristics, Fig.4.1 lh and Fig.4.1 l i shows the corresponding steel surface characteristics. This was seldom described in the previous literature. Fig.4.12a to Fig.4.12h shows the corresponding set of pictures at 950 °C. In this series, the contact areas and noncontact areas are more clearly manifested. Within contact areas on steel surface(Fig. 4.12h), most steel grains have common faces which are parallel to the scale/steel interface. Within noncontact areas on steel surface(Fig. 4.12g), the steel grains don't have common parallel faces. For adherent scale, three layers are usually formed: an outermost hematite layer, a middle magnetite layer and an inner wustite layer. If the sample is cooled slowly enough, the wustite layer will not theoretically exist after cooling. In this case, the wustite will be converted to iron and magnetite. But in practical situations, such a condition does not usually occur. So after cooling, some magnetite grains precipitate within the wustite layer. From the iron-oxygen phase diagram, oxygen accounts for 23.1 percent to 25.3 percent in weight ofwustite. During the oxidation process, there exists a gradient of oxygen potential within the wustite Ch.4 Experimental Investigation 60 layer. So the extent of magnetite precipitation during cooling is different within wustite if the sample is not overoxidized. Those theoretical derivations are reflected exactly in Fig. 4.13a to Fig. 4.13f. After etching in about 35 percent HC1, the three layers on the cross section of scale are manifested as different shades(Fig. 4.13a). The three layers could also be seen as three distinct layers on the fractograph(Fig.4.11d) although sometimes it is not clearly discernable. This is in agreement with the previous studiesJ5-14'16-22'29] In this investigation, the adherent scale could form above 950 °C but it usually just accounts for less than 25 percent area of the total scale surface. (2) Scale Above 950 °C Above 950 °C, multiple layers of nonadherent scale are usually formed as shown in Fig.4.14a to Fig.4.14g. It is noted that the scale layers themselves are not regular and there are pores, voids and fissures within the layers. Fig.4.15a to Fig.4.15f shows the surface, fractograph and under-surface of the first layer of nonadherent scale formed at 1200 °C. There seems to be some special phenomena along the hematite grain boundaries(Fig.4.15c). What role this extra growth along hematite grain boundaries plays deserves further study. This first layer of nonadherent scale fractures along grain boundaries(Fig.4.15d) indicating that with nonadherent scale, the weakness is along the grain boundaries. Its under-surface is porous and fissured(Fig.4.15e, Fig.4.15f). The characteristics of the second layer of nonadherent scale are shown in Fig.4.16a to Fig.4.16e. The outer surface is highly irregular andporous(Fig.4.16a to Fig.4.16c). There are a lot of fissures along grain boundaries and holes on the under-surface(Fig.4.16d, Fig.4.16e). This may be due to preferential oxide dissociation along oxide grain boundaries.PI Fig.4.17a Ch.4 Experimental Investigation 61 to Fig.4.17e gives another set of photographs of the first layer of the nonadherent scale. The fracture characteristics along grain boundaries are clearly manifested(Fig.4.17c). The set of photographs in Fig.4.18a to Fig.4.18e shows slightly different characteristics of the second layer of nonadherent scale. This scale consists of beautiful columnar grains(Fig.4.18a to Fig.4.18c). A lot of fissures on the under-surface exist along the grain boundaries(Fig.4.18e). The nonadherent scale formed between 950 °C and 1200 °C also shows similar characteristics. Fig.4.19 and Fig.4.20 show the first layer and second layer of the nonadherent scale at 1100 °C. The outer surface of the first layer of the nonadherent scale has the extra growth along grain boundaries(Fig.4.19b). The fracture is along the grain boundaries(Fig.4.19c). And its under-surface is fissured and porous(Fig.4.19d, Fig.4.19e). The second layer of this nonadherent scale has a seemingly compact and strong outer surface(Fig.4.20a, Fig.4.20b). But its grain boundaries are still weak since its fracture is also along grain boundaries(Fig.4.20a). Furthermore its under-surface shows a lot of fissures and holes as well(Fig.4.20c, Fig.4.20d). Fig.4.21 shows the cross section of the second layer of the nonadherent scale with some pores formed at 1050 °C. The outer surface of the second layer of this nonadherent scale shows the growth process of new grains(Fig.4.22). Fig.4.23a to Fig.4.23f shows the typical characteristics of the first layer of the nonadherent scale at 1000 °C. It has almost the same characteristics as that of the scale formed above 950 °C. The big difference between the nonadherent scale formed above 950 °C and the adherent scale formed below 950 °C is evident from Fig.4.23c and Fig.4.23d although the temperature difference is only 50 °C. Details of this kind of the second layer of nonadherent scale are shown in Fig.4.31 which is taken from another area on the same sample shown in Fig.4.23c. Ch.4 Experimental Investigation 62 The nonadherent scale mainly consists of magnetite and/or hematite. This conclusion is based on the following observations: (a) There are usually two distinct layers with the nonadherent scale(Fig.4.25); (b) The precipitation of magnetite inside wustite has never been found in the nonadherent scale even if the sample was cooled quickly, (c) With the nonadherent scale, the oxygen potential is high because the oxidizing atmosphere penetrates the pores and fissures inside the nonadherent scale. The high oxygen potential is conducive to forming oxide scale with high oxygen content, i.e. magnetite and hematite. The separation of scale layers greatly reduces the iron supplyJ5] Therefore when a small amount of iron has diffused to the scale surface, it could not be replenished adequately from the steel core and the scale is converted to magnetite. W Ch.4 Experimental Investigation outer surface narrow face scale Fig. 4.10 Fractograph of broad face scale and narrow face scale. 900 °C, 24 hours Fig. 4.1 lb Outer surface of adherent scale. 900 °C, 24 hours 63 Fig. 4.1 la Outer surface of adherent scale. 900 °C, 24 hours 03E967 £6KV K5§! 0 ' ' ! Ikm Fig. 4.1 lc Fractograph of adherent scale. 900 °C, 24 hours Ch.4 Experimental Investigation Fig. 4.1 Id Fractograph of adherent scale. 900 °C, 24 hours Fig. 4.1 If Under surface of adherent scale. 900 °C, 24 hours 64 Fig. 4.1 le Under surface of adherent scale showing macro voids. 900 °C, 24 hours Fig. 4.1 lg Contact area on under surface of adherent scale. 900 °C, 24 hours Fig. 4.1 lh Noncontact area of steel surface at scale/steel interface. 900 °C, 24 hours Fig. 4.1 l i Contact area of steel surface at scale/ steel interface. 900 °C, 24 hours Ch.4 Experimental Investigation 66 Fig. 4.12c Under surface of adherent scale showing Fig. 4.12d Contact area on under surface of the contact areas and noncontact areas. 950 °C, 24 adherent scale. 950 °C, 24 hours hours Fig. 4.12e Noncontact area on under surface of Fig. 4.12f Outer surface of steel showing the adherent scale. 950 °C, 24 hours contact areas and noncontact areas corresponding to adherent scale in Fig. 4.3c. 950 °C, 24 hours Ch. 4 Experimental Investigation 67 Fig. 4.12g Noncontact area on steel surface under adherent scale. 950 °C, 24 hours Fig. 4.13a Cross-section of adherent scale showing decreasing precipitation of magnitite in wustite layer from outer surface to scale/steel interface i.e. from (b) to (f). 1200 °C, 24 hours Fig. 4.12h Contact area on steel surface under adherent scale showing the same orientation of some grain surfaces. 950 °C, 24 hours Fig. 4.13b Outmost part of cross-section of adherent scale. 1200 °C, 24 hours Ch.4 Experimental Investigation 68 Fig. 4.13c An area of cross-section of adherent Fig. 4.13d An area of cross-section of adherent scale. 1200 °C, 24 hours scale. 1200 °C, 24 hours Fig. 4.13e An area of cross-section of adherent Fig. 4.13f An area of cross-section of adherent scale. 1200 °C, 24 hours scale. 1200 °C, 24 hours Ch.4 Experimental Investigation 69 Fig. 4.14a Cross-section of nonadherent scale and Fig. 4.14b An area of cross-section of nonadherent their structures. 1200 °C, 24 hours scale. 1200 °C, 24 hours Fig. 4.14c An area of cross-section of nonadherent Fig. 4.14d An area of cross-section of nonadherent scale. 1200 °C, 24 hours scale. 1200 °C, 24 hours Ch.4 Experimental Investigation 70 Fig. 4.14e A n area of cross-section of nonadherent Fig. 4.14f A n area of cross-section of nonadherent scale. 1200 °C, 24 hours scale. 1200 °C, 24 hours Fig. 4.14g A n area of cross-section of nonadherent Fig. 4.15a Outer surface of first layer of scale. 1200 °C, 24 hours nonadherent scale. 1200 °C, 24 hours Ch.4 Experimental Investigation 63 £951 28KV WW " ' " l l i u s ) tt 3 £ 9 5 1 b 9 U Ffi Fig. 4.15b Outer surface of first layer of nonadherent scale. 1200 °C, 24 hours Fig. 4.15c Outer surface of first layer of nonadherent scale. 1200 °C, 24 hours Fig. 4.15d Fractograph of first layer of nonadherent scale, showing different layers. 1200 °C, 24 hours N H B H B H B I 0 3 £ 9 5 1 £0KV ';•< 17 .8 1.76 nn Fig. 4.15e Under surface of first layer of nonadherent scale. 1200 °C, 24 hours Ch.4 Experimental Investigation 72 Fig. 4.15f Magnified under-surface of first layer of Fig. 4.16a Outer surface of growing second layer nonadherent scale, showing porosity. 1200 ° C , 24 of nonadherent scale and steel. 1200 ° C , 24 hours hours Fig. 4.16b Fractograph of second layer of Fig. 4.16c Fractograph of second layer of nonadherent scale showing irregular and porous nonadherent scale. Oxidized 1200 ° C , 24 hours nature. Oxidized 1200 ° C , 24 hours Ch.4 Experimental Investigation 932951 28KV k\K 8* " i ' . ?lm Fig. 4.16d Under surface of second layer of nonadherent scale. 1200 °C, 24 hours Fig. 4.17a Outer surface of first layer of nonadherent scale. 1200 °C, 19 hours 73 9 3 2 9 5 1 28KV X l l l ! § § M Fig. 4.16e Under surface of second layer of nonadherent scale. 1200 °C, 24 hours 032951 28KV X l ! 00K* ' ' 3©um Fig. 4.17b Outer surface of first layer of nonadherent scale. 1200 °C, 19 hours Ch.4 Experimental Investigation 74 Fig. 4.17c Fractograph of first layer of nonadherent Fig. 4.17d Under surface of first layer of scale. 1200 °C, 19 hours nonadherent scale. 1200 °C, 19 hours Fig. 4.17e Under surface of first layer of Fig. 4.18a Outer surface of second layer of nonadherent scale. 1200 °C, 19 hours nonadherent scale. 1200°C, 19 hours Ch.4 Experimental Investigation @ 3 £ 9 5 1 £ 0 K V K30 ! 0 * ' i ! SOmrri Fig. 4.18b Outer surface and fractograph of second layer of nonadherent scale showing wustite grains. 1200 °C, 19 hours Fig. 4.18d Under surface of second layer of nonadherent scale. 1200 °C, 19 hours 75 Fig. 4.18c Outer surface and fractograph of second layer of nonadherent scale. 1200 °C, 19 hours Fig. 4.18e Under surface of second layer of nonadherent scale. 1200 °C, 19 hours Ch.4 Experimental Investigation 76 Fig. 4.19a Outer surface of first layer of nonadherent scale. 1100 °C, 24 hours Fig. 4.19c Fractograph of first layer of nonadherent scale. 1100 °C, 24 hours 032952 £8KV X' i ! 00K* ' ' 30urn Fig. 4.19b Outer surface of first layer of nonadherent scale. 1100 °C, 24 hours Fig. 4.19d Under surface of first layer of nonadherent scale. 1100 °C, 24 hours Ch.4 Experimental Investigation 832952 28KV X 5 8 8 & 8 U B Fig. 4.19e Under surface of first layer of nonadherent scale. 1100 °C, 24 hours 032952 £ 0 K V )[ 3 8 8' ' i 8 8 u I 77 Fig. 4.20a Fractograph and outer surface of second layer of nonadherent scale. 1100 °C, 24 hours 032952 28KV 'A\W ' ' ' '. 30mm Fig. 4.20b Outer surface of second layer of nonadherent scale. 1100 °C, 24 hours Fig. 4.20c Under surface of second layer of nonadherent scale. 1100 °C, 24 hours Ch.4 Experimental Investigation 78 • 0^ • •.. . . 032952 20KV X3 00 I00UM Fig. 4.20d Under surface of second layer of nonadherent scale. 1100 °C, 24 hours Fig. 4.21 Cross section of second layer of nonadherent scale showing pores(1.2 microns in diameter). 1050 °C, 26 hours Fig. 4.22a Outer surface of second layer of nonadherent scale showing growth process of new oxide grains. 1050 °C, 26 hours Fig. 4.22b Outer surface of second layer of nonadherent scale showing growth process of new oxide grains. 1050 °C, 26 hours Ch.4 Experimental Investigation 79 Fig. 4.23c Fractograph of first layer of nonadherent Fig. 4.23d Fractograph of adherent scale. 950 °C, scale. 1000 °C, 24 hours 24 hours Ch.4 Experimental Investigation 80 Fig. 4.23e Under surface of first layer of Fig. 4.23f Under surface of first layer of nonadherent scale. 1000 °C, 24 hours nonadherent scale. 1000 °C, 24 hours Fig. 4.24a Fractograph of first layer of Fig. 4.24b Fractograph of first layer of nonadherent scale. 1000 °C, 24 hours nonadherent scale. 1000 °C, 24 hours Ch.4 Experimental Investigation Fig. 4.25 Cross section of first layer of nonadherent scale showing two distinct layers. 1200 °C, 24 hours Ch.4 Experimental Investigation 82 4.4 Other Process Characteristics (1) Geometry Effect The shape and dimensions of the sample have a big effect on the scale formed. The sample in the present work was 27.70x37.50x3.55 mm which is larger than the samples employed in most of the previous studies. Fig.4.10 shows the adherent scale formed below 950 °C. It was found that the scale thickness on the broad face is double the scale thickness on the narrow face. It is evident that this is due to the geometric effects, i.e. the deformation of the scale is constrained by the geometry of the sample. The geometric effects are even larger around the corner of the sample. The nonadherent scale formed above 950 °C also showed a similar thickness distribution pattern. Therefore to simulate the actual situation in steel processing, it is very important to use large samples. If possible, samples with dimensions as close to actual products as possible should be employed as in some plant trials. Abuluwefa et al have conducted a lot of work in oxidation of steels[7'21'23] and almost all of their experiments used a sample of 18x8x8 mm. They observed long linear oxidation periods in their experiments. This may be, as they said, due to low oxygen potential in the furnace atmosphere. But too early and extensive separation of scale from steel due to geometric effects most probably is another important reason for the linear rate law they found. As in the current investigation, after certain time, the nonadherent scale above 950 °C gives a linear weight-gain curve. Small samples are easy to deal with, but they also have larger geometric effects. To reduce the geometric effects, the edges and corners were properly rounded in the current investigation. Ch.4 Experimental Investigation 83 (2) Scale Cracking During the experiments, cracking sounds were heard within 2~6 minutes after the sample was raised above the upper end of the furnace if the adherent scale below 950 °C was cooled rapidly by taking it out of the furnace. Extensive cracks were evident on the scale surface. On the other hand, if the scale was cooled very slowly, the audible cracking sounds could not be heard indicating that the thermal stresses had been relaxed during the slow cooling. J f the scale was adherent and very thick, the cracking was not found but cracking sounds could still be heard several minutes after raising up. With the thick nonadherent scale, cracks could never be found but cracking sounds could also be heard several minutes after the sample was taken out quickly. These observations were also reported by other workers, t38'39! (3) Scale Colors With oxidation of IF steel in air, it was found that the various kinds of scale have different colors. The adherent scale is gray and nonadherent scale is usually black with or without tiny white points on it. From appearance, both kinds of scale look smooth. 4.5 Discussion: Scale Types and G r o w t h Mechanisms Based on the previous observations of the growth curves and microstructures for IF steel oxidation, the growth curves(weight-gain curves) observed can be divided into two categories. Various influencing factors contribute to the differences. Those factors include the nature of the scale(adherent or nonadherent), temperature, sample geometry etc. The adherent scale follows the lattice diffusion rule, i.e. parabolic growth during most of the growth process. With the nonadherent scale, the gap between distinct layers is formed and Ch.4 Experimental Investigation 84 each layer itself is porous or fissured. The gap between distinct layers tends to slow the rate of oxidation but pores and fissures tend to increase the rate. With multiple layers of scale, the innermost layer tends to be adherent at first. It may later change to a detached nonadherent layer. Temperature plays an important role in all oxidation process. (a) Growth Mechanism Corresponding to Adherent Scale below 950 °C Below 950 °C, the scale formed is adherent and compact. The oxidation process follows a solid state diffusion rule, i.e. the parabolic rate law is followed. The oxidation rate becomes smaller with increasing scale thickness since ions have to travel longer. And the parabolic oxidation rate is increased with higher temperature.[5'11'12,14,19'22'27'291 Under this mechanism, the weight gain is larger at higher temperature(Fig.4.3). The microstructural examination, presented in the previous section, shows that the scale formed below 950 °C is essentially the same and it is compact and adherent. This suggests that the weight gain for nonisothermal conditions could be simulated with the weight-gain data for isothermal conditions as long as the oxidation occurs below 950 °C all the time. (b) Growth Mechanism Corresponding to Nonadherent Scale above 950 °C Above 950 °C, the scale formed is nonadherent and porous. The oxidation process follows a combination of an initial nonlinear stage and a subsequent linear stage(Fig.4.5). A combination of an initial linear stage and a subsequent parabolic stage were previously reported.15'21'231 But a combination of an initial nonlinear and subsequent linear growth has not been previously presented. This characteristics is probably due to the effects of gap, porosity and fissures associated with the scale formed above 950 °C. At the very beginning of oxidation of IF steel in air, the oxidation rate should also be linear since the surface reaction controls the whole process. However since the oxygen potential of Ch.4 Experimental Investigation 85 air is so large, this initial linear stage is not visible before the rate law becomes parabolic due to a change of oxidation to solid state diffusion. The scale growth then follows a parabolic rate law. After a certain thickness of scale has formed, it becomes porous and separated from the steel surface. Due to the microfissures and porosity of the scale, air penetrates this scale layer and the oxidation rate is mainly controlled by the oxidation at the innermost steel surface which is periodically refreshed. Although air penetrates the outer scale layers, its oxygen potential is greatly reduced when it reaches the inner scale layer. This also accounts for the linear rate law since low oxygen potential atmosphere tends to give linear oxidation. [ 5 ' 2 1 ' 2 3 ] Nevertheless the conversion of wustite to magnetite or magnetite to hematite also contributes slightly to the resulting weight gain. The microstructural examination of the scale formed above 950 °C indicates that the scale formed is of the same type and is multiple-layered, porous and fissured. The weakness along oxide grain boundaries indicates that porousness and microfissures, to great extent, are caused by preferential dissociation of oxide along grain boundaries. Both the microstructure and weight-gain data show consistency and convergence. Therefore it is reasonable to simulate the nonisothermal oxidation process with isothermal oxidation data above 950 °C since they follow the same mechanism. If the weight-gain curves above 950 °C and below 950 °C are combined together(Fig.4.6). It is found that after 28 minutes of oxidation the weight gain at 1000 °C is less than that at 950 °C and after 400 minutes of oxidation the weight gain at 1050 °C is less than that at 950 °C. , This seems abnormal since in principle the weight gain at higher temperature should be larger. This has not been reported by other workers. Most literature investigated just Ch.4 Experimental Investigation 86 segments of temperature range between 500 and 1200 °C so that it is hard to find this phenomena. Although Abuluwefa et al [ 2 1 ] noticed that at 900 °C the scale changes from adherent to separated scale, they did not give enough attention to it. They did not report the abnormal weight-gain curves. If the weight-gain curves and the microstructural observation are considered together, the reason for this abnormality is found. The gap between distinct layers of scale tends to decrease the oxidation rate. However the pores, fissures and other defects tend to increase the rate. Temperature increases the oxidation rate in an exponential manner.[5'8'21'36] Those three main factors determine the rate of whole oxidation process. At 1000 °C, the gap between different layers greatly reduces the total oxidation rate while the extent of porosity and fissures are not so much and temperature is not so high to balance the gap's effect. Therefore the weight gain at 1000 °C is less than that at 950 °C. Due to different oxidation mechanisms, it is not feasible to simulate the nonisothermal oxidation process by simply applying or superimposing the isothermal oxidation data if the actual oxidation process involves two different mechanisms. One possibly feasible solution is to obtain another set of oxidation data, i.e. the secondary isothermal oxidation data. The secondary isothermal oxidation weight gain is much smaller than the primary isothermal oxidation weight gain. The secondary weight-gain curves are not greatly influenced by the temperature of the primary growth mechanism as long as the oxidation remains within the same growth mechanism(Fig.4.8). And the longer the sample is exposed to the primary growth mechanism, the larger the difference between various secondary growth curves. Abuluwefa et al [ 7 ] observed the scale separation with low carbon steel at 900 °C. They attributed the phenomenon to the phase transformation of ferrite to austenite and the Ch. 4 Experimental Investigation 87 formation of more rigid magnetite layer. However in the current investigation, it was observed that the temperature of transition to separated scale is between 950 °C and 1000 °C which is confirmed by many experimental runs carried out. The steel phase transformation temperature cannot be raised to such a high temperature. The separation is hardly related with phase transformation. Three layers of scale, wustite, magnetite and hematite, account for about 95, 4 and 1 percent respectively between 700 °C and 1200 °C. [ 3 ] The more rigid magnetite and hematite layers do not have a sudden increase in the scale. So it is hard to attribute the sudden increased separation of scale to more magnetite and hematite. The earlier separation of scale in Abuluwefa's case is most probably related with the too small samples employed which were parallepipeds of 18x8x8 mm. With this kind of samples the edge effects should be very large. To reduce the geometric effects the sample surface should be flat and as large as possible. In the present experiments, the very thick adherent scale formed which should be rigid enough but still maintained contact with the steel surface. This thesis attributes the detachment of scale layers to the combined effects of the rate of scale growth, the microstructure of the scale and the scale creep deformation process. This understanding could satisfactorily explain why at high temperature the nonadherent scale is formed while at low temperature the adherent scale is formed. It was mentioned by some papers that at high temperature it was easy for scale to maintain contact with steel since the creep was larger at high temperature. With increasing temperature, the adherent scale formed grows faster and faster due to the faster rate of diffusion until 950 °C(Fig.4.3). In the steel oxidation process, iron ions diffuse outward leaving certain voids and porosity behind. If the scale growth process is not too fast, Ch.4 Experimental Investigation 88 these voids and porosity could be compensated by newly arrived iron ions. Therefore there is a balance between the formation of voids and arrival of iron ions. In the meantime, the creep of the scale has enough time to deform to maintain contact with the steel surface. When the temperature is high enough, the scale growth process is so fast that this balance cannot be maintained. The voids and porosity are extensively and dynamically formed. There is not enough time for the scale creep process to maintain contact with the steel surface. This permits the scale to easily detach from the steel. CHAPTER 5 MODELING OF THE OXIDATION PROCESS This chapter presents the models used to predict scale growth. The ultimate objective is to develop a simulation model to predict scale growth under industrial conditions. The basic idea is to use a whole set of isothermal oxidation data to simulate an oxidation process of any temperature evolution. To form a working package, a heat transfer model is programmed to provide the necessary temperature history for the scale growth model. The scale growth model serves as the foundation for additional models such as stress-strain model and descaling model. 5.1 Heat Transfer Model140'41'421 To simulate the growth of scale, an essential part is the thermal history of the sample upon which the scale growth model is based. For steel processing operations such as continuous casting and rolling of slabs, the heat transfer could be treated as an initial-boundary-value problem. Since the slab and strip usually have a large width/thickness ratio, it is reasonable to 89 Ch. 5 Modeling of Oxidation Process 90 treat the heat transfer as a one-dimensional initial-boundary-value problem. A combination of the finite-element method for the spatial part of the model and the finite difference method for the temporal part of the model was used to solve the problem. For a differential element of cross-section area A and thickness dx(Fig.5.1), the energy balance could be expressed as Fig.5.1 Differential element for for heat transfer and scale growth model rate of energy added - rate of energy lost = rate of int ernal energy increase qA + QAdx -(q + dq)A - pc — Adx dt (5.1) (5.2) where p(j? x ) mass density which may change with respect to T and x c(T, x) specific heat which may change with respect to T and x q(x,t) heat flux Q(x,t) heat source Ch. 5 Modeling of Oxidation Process 91 Using Fourier's law of heat conduction q(x^ ty = -k(J, x) d^( x ' ^ , equation (5.2) becomes dx .dT(x,t) d p(T,x)c(T,x)— — dt dx k(T,X)dT^ = Q(.x,t) (5-3) dx To develop a more general code for various heat transfer situations, the equation (5.3) could be generalized to the following form dT{x,t) ^dT{x,t) d M(T,x)— — dt dx a(T,x)-dx + P(T,x)T(x,t) = f(x,t) (5-4) Equation (5.4) is essentially a 2-D problem with an infinite size of time domain. One convenient way to solve it is to separate the variables x and t. The element trial solution could be written as r w ( x , 0 = £ f l y ( ^ ' ) ( x ) (5-5) where a unknown function values <f>(e)(x) element shape functions(Lagrange interpolation polynomial) For any given value of t, equation (5.5) becomes a standard element trial solution for boundary-value problem. The numerical values for a rt) may vary from one instant to the next. This way we transform the initial-boundary-value problem(mixed initial-value problem and boundary-value problem) into a pure initial-value problem which could be easily solved by time stepping technique. The basic equations could be found in the relevant books J4 0>4 1 - 4 2 1 Only the main points are presented here. Through the steps of writing the Galerkin residual equations for a typical element, integrating by parts and substituting the general form of element trial solution into the interior integrals in residual equations, we get the following element equations for a typical element Ch. 5 Modeling of Oxidation Process 92 dt i—\ dx ctx +fiiej^e)(x)6(T,x)^\x)dx a At) (e) = $f(x,t)$e)(x)dx- r . a ( T , x ) ^ ^ dx tie\x) i = 1,2,..., n (5.6) Above element equation may also be written in the usual matrix form: where [cr{^}+w e )ko}={F(o} w O) Cf = \^e\x)M(T,x)tf>f(x)dx (5.7) Kf = Kaf + K6f = l^i^a(T,x)-^^-dx+ frf(x)6(T,x)<f,f(x)dx M(ir\ («) F/'>(0 = F / : w ( 0 + Fr, ( e ) (0 = ' 'j f(x, t)<j>f(x)dx • dT(e)(x,t)^ -a(T, x) dx tie)(x) (5.8) After further developing specific expressions for the shape functions ^ <e>(x), substituting the shape functions into the element equations, we transform the integrals inEqs(5.8) into a form appropriate for numerical evaluation(the isoparametric element is used here). /=i <xM(T,%nl) i djjtfy d{ Ch.5 Modeling of Oxidation Process 93 In above model, the property terms are functions of temperature. But within a short time step, properties do not change so much that it is assumed that the properties are constant within a time step. Once having evaluated the element stiffness equations and load vector equations, we assemble them into a system equation for the problem in the following form In order to form Eqs(5.10), a time-stepping method was used. In the time-stepping methods the time axis is divided into time intervals each of which correspond to a loading period. Each interval is further divided into a succession of time steps ^ j _ L2,...* beginning at time t. Instead of seeking a solution for {a(t)} o v e r m e continuous domain of time, we look for an approximate solution consisting of discrete values for {a(t)} a t t n e e n ( ^ °f e a c n s teP> i - e -at time ^, | a | 2 at time ^ etc., starting from the known initial value at time t . After we get the temperature values at nodes by the time-stepping method, the heat flux for isoparametric elements could be calculated by J d^ The implementation of the above model needs to consider a lot of specific details. To simplify the data input but still give the program certain flexibility, the user defines intervals of time(Fig.5.2). Within each interval there may be an arbitrary number of time steps, all the Ch.5 Modeling of Oxidation Process 94 same size. Each interval is therefore defined by two numbers: the number of steps n and the size of steps A ? • At the beginning of each interval the user defines all loadsfboundary and interior) that are to be applied at the end of the interval, i.e. at the end of the last step in the interval(step n ). The loads at the ends of all the intermediate steps are calculated by ramping, i.e. linearly interpolating from the loads at the end of the previous interval(Fig.5.3). The ramping formula is as follows: Interval I Interval II (All steps are size A/,) (All steps are size Af n) I l . l I I I I i U _ l J 1 1 | i I i Fig.5.2 Definition of intervals and time steps (5.12) where p n the load applied at the end of the nth step. One interval Load applied to a DOF 2 3 n - 1 n Fig.5.3 Ramping the loads between the user-defined values F and F ns Ch. 5 Modeling of Oxidation Process 95 , Here p represents a natural or essential boundary load or an interior load. For a "sudden" load(step load), the entire load could be applied to a very short one step interval or a few step interval. 5.2 Scale Growth Model The scale growth model is formulated by the author based on a large number of experimental observations. Some details are quite complicated to present and are just left in the code itself. 5.2.1 Theoretical Model of Scale Growth The question whether the isothermal oxidation data could be used to simulate nonisothermal oxidation process has been asked quite a few times. But no papers have been found which have implemented the idea with the complete oxidation process. It is believed that the complexity arisen hindered the efforts. A practical way has to be taken. In the present study, a temperature evolution is divided into a series of segments, each of which corresponds to a time step and could be approximately thought of as an isothermal segment. If the oxidation has been continued within the same mechanism, the scale growth can be additive as if the previous accumulated scale has grown at this new isothermal temperature. Complexity arises if some scale of a different growth mechanism has grown. As shown in Fig.4.7 to Fig.4.9, the secondary weight-gain data are much different from the primary weight-gain data. Amount of difference mainly depends upon how long the Ch. 5 Modeling of Oxidation Process 96 oxidation of the previous mechanism has proceeded or how much the weight has been gained. The oxidation weight-gain curves with different periods of oxidation of the previous mechanism have been obtained from experiments. The secondary oxidation periods for which there is no corresponding secondary experimental weight-gain curve will be interpolated using available weight-gain curves. As specific examples, let's examine the various cases. (1) Oxidation Process under the Same Mechanism Fig.5.4 shows the model to calculate the weight gain under the same mechanism. In time interval I, temperature is constant at 750 °C and it has a corresponding isothermal weight-gain curve. Therefore the oxidation should exactly follow the isothermal oxidation data although the weight-gain data are calculated using the same method described later for the time period II. The time period II represents a general case. At is the constant time step in a time interval. Different time intervals could have different time steps. Let's consider time steps a, b and c in the interval II. W A Q , W F E 0 and W C 0 are the accumulated weight gain at beginning of each time step. W G , W F E and W C are the accumulated weight gain at the end of each time step. A W A L , A W B L and A w c i a r e m e s t e P weight gain corresponding to the lower isothermal oxidation curve of two isothermal oxidation curves enclosing the actual step temperature. A W ^ , A W B 2 and A W R 2 are the step weight gain corresponding to the higher isothermal oxidation curve. T , T b and T. are the average temperature within each time step. ta, t^  and t are time values at the end of each time step. T i s o l and T. s o 2 are temperatures of two isothermal oxidation curves encompassing the step average temperature such as 750 °C and 800 °C for T (760 °C). Fig.5.4 Weight gain calculation model Ch. 5 Modeling of Oxidation Process 98 After the time interval I, the accumulated weight gain at the beginning of time period II is W a Q . Since Ta(760 °C) is between 750 ° C and 800 ° C , it is necessary to find the points corresponding to W a 0 on the 750 ° C isothermal weight-gain curve and 800 °C isothermal weight-gain curve. After time step At, we get step weight-gain data A^a\ on 750 °C curve and A ^ A 2 on 800 ° C curve. The actual step weight gain A W 3 will be calculated as weighted sum of ^ W a l and A ^ A 2 by the following equation: AW = Ti*°2~Ta AW.+ T"~Tm AW, (5-13) O rp rp Cll rp rp 01 isol isol isol isol Wa=Wa0+AWa (5-14) At the beginning of the step b, the accumulated weight gain is W b 0 which is equal to W . Tb(780 ° C ) is still between 750 ° C and 800 °C. So we should also use 750 ° C curve and 800 ° C curve. Find the points on 750 °C curve and 800 °C curve corresponding to W b 0 and calculate A W m and AW b 2 . Use the equation (5.13) and (5.14) to calculate W b . With time step c, everything is the same except the encompassing isothermal temperatures are 800 °C and 850 °C. (2) Oxidation Process under Different Mechanisms If a scale layer of different mechanism has grown, the primary weight-gain data cannot be simply applied to predict scale growth for the secondary oxidation data. The amount of scale of the primary mechanism greatly influences the growth of, the scale of the secondary mechanism. Although nonadherent scale is penetrable to an oxidizing atmosphere, the oxidizing potential is greatly reduced compared to the original oxidizing atmosphere. A layer of adherent scale could form under nonadherent scale. After this new scale has grown to a certain thickness, it Ch. 5 Modeling of Oxidation Process 99 will also detach from the steel core and become porous, thus forming the second layer of the scale. If a layer of adherent scale has grown below 950 °C, the scale could become nonadherent later if the temperature is increased above 950 °C. A set of secondary weight-gain curves have been measured after different periods of scale growth, say 3 hours and 6 hours, under the primary oxidation mechanism(Figs4.7-4.9). If the actual period of scale growth under the primary mechanism, say 4 hours, is between those time periods, interpolation is used to obtain corresponding secondary weight-gain data for the scale growth of four hours under the primary mechanism. However after implementation of this scheme in the code, it has been found that better results could be obtained if the equivalent weight gain under the primary mechanism is used instead of the oxidation time under the primary oxidation mechanism. This scheme is also implemented in the code. The scale prediction model has not been used by previous researchers and was formulated for the current investigation. 5.2.2 Implementation Techniques for Scale Growth Model (1) Representation of Weight-gain Curves Convenient and effective representation of the weight-gain curves in the computer is an important issue. From experiments, the time and the corresponding weight-gain data were measured. Easy input format is time vs. weight-gain data points read directly from measured weight-gain curves. There are usually curved portions and straight line portions on the weight-gain curves. So using piecewise 2 n d order polynomials would eliminate the use of complicated curve fitting and still give good approximation of the actual curves. Ch. 5 Modeling of Oxidation Process 100 Let's represent time by t and weight gain by w . Given three points (f 0 , w0), (tx, w,) and (f2, w2) > » curve could be decided by where w = a 0 +axt + a2t2 (5.15a) w,,^^ w,f0f2 w2f0f, "o — " r "1" a, =-(t0 ~ h X'o ~ h ) ('i " 'o ) ( ' i ~ h ) (h ~ h )(h ~ tx) ^ 0 ( ^ 1 + ^ 2 ) ^ 1 ( ^ 0 + ^ 2 ) ^ 2 ( ^ 0 + ^ 1 ) (tv-tXh-h) (tx-t^){tx-t2) (t2-t0)(t2-tx) w0 wx w2 a2 = + 1 + - -(5.15b) (^0 - h X'o - h) d - '0 - h) (h - h )(t2 - tx) The oxidation weight-gain curves and property curves in the modeling are numerically represented by this piecewise 2 n d order polynomial. Once the three points are input, the program automatically calculates and stores those coefficients in appropriate arrays. In the current model, three segments could represent an oxidation weight-gain curve very well. (2) Relationship between Weight-gain Data and Scale Thickness For adherent scale, there is an approximate conversion relation between weight-gain data and scale thickness. Goursat and Smeltzeri29! converted 1 kg/m2 weight gain to 0.00072 meter thick FeO scale. But in the current investigation, scale usually consists of three layers. If an average oxygen content of scale is taken to be 26 percent and the average density 5500 kg/m2, 1 kg/m2 weight gain is equivalent to 0.00069 meter thick scale. However for the nonadherent scale, the scale thickness is hard to define. We just use equivalent scale thickness for modeling purpose. Weight-gain data is more meaningful in the case of nonadherent scale. Ch.5 Modeling of Oxidation Process 101 (3) The Thickness Change during Simulation During the scale growth process, the thickness of the scale layers and the steel layer are constantly changing. In the present model, the thickness values of the various layers are modified once the thickness change is equal to or greater than the element size. This model will not cause too big a discontinuity in the heat transfer model if the element size is small enough. (4) The Boundary Conditions for Heat Transfer At the surface of the sample, the following heat transfer boundary conditions were applied: q f = h(T -T ,) + e cr(T4 -T\) (5-16) "surf V env surf/ "ox \ env surf) where h the heat convectivity of air to the sample surface T the environment temperature env r T s u r f the sample surface temperature q s u r f the sample surface heat flux other symbols listed in the following paragraph (5) Some Parameters Used in the Model The following parameters are used in the model. The emissivity of scale surface: e = 0.35 Stefan Boltzmann constant: a = 5.6697 x 10~8 W/m2KA Specific heat of oxide scale: Q = 641 + 0.227/ JlkgK Heat conductivity of scale: £ = 10.5-7.8x10""7/ WlmK Specific heat of steel: Specific heat of steel: Ch. 5 Modeling of Oxidation Process 102 Temperature range of 300-600 K: £ p = 309 + 0.416T JlkgK Temperature range of 600-1473 K: Q = 356 +1.5257: JlkgK Heat conductivity of steel: Temperature range of 300-400 K: £ = 49.10 - 0.0387 WlmK Temperature range of 400-1473 K: £ - 43.35 _ 0.0447 WImK Density of scale: p = 7854 kg/m3 Density of steel: ^ = 5700 kg/m3 Heat conductivity of air to the sample surface: ^ = 50.40 W lm2°C When temperature change is large, air buoyancy is different. For a steel bar of thickness of 1.0 cm and for a temperature change of 600 °C, the buoyancy difference is about 0.001 kg/m2. This is negligible for the time period used in the present investigation. 5.3 Program Structure The overall program structure is shown in Fig.5.5. The function of each major module is as follows: « init_data: read in problem type data, initialize some values, set limits and integration points # meshdata: read in node data and element data and automatically generate node or mesh if necessary # propdata: read in and process physical properties data # grow_data: read in and process the measured oxide scale growth data Ch.5 Modeling of Oxidation Process 103 # load_data: read in and process the load data and BCs , form_sys: form the element matrix and assemble it into global matrix # apply_bc: apply natural BCs and essential BCs to system equations # cal_temp: solve the system equations for nodal temperature values , scale_grow: simulate the scale step growth according to step temperature change and adjust scale layer thickness if necessary « cal_flux: calculate the flux value at some points . print: tabulate the calculated results These are the major modules. Each of them may consist of several sub-modules. The module "scale_grow" is quite complicated which involves judging if oxidation mechanism has changed, accumulating weight gain, adjusting interface node position, changing element property etc. Ch. 5 Modeling of Oxidation Process 104 ( end ) Fig.5.5 Overall structure of program Ch.5 Modeling of Oxidation Process 105 The input to the program is explained in appendix A which is an actual input data file for the modeling. The modeling code is not attached in this paper due to its length. 5.4 Simulated Results and Verification Four verification experiments have been carried out and all of them are compared with model predictions. With the current model, there are only a limited number of changeable process variables due to the empirical nature of the model. On the other hand this is exactly why this model could possibly be applied to actual plant operations for IF steel while others could not. For the current model, the oxidizing atmosphere and steel grade are given, i.e. air and IF steel. The most important process variable is thermal history of the sample, i.e. the temperature and the corresponding time period in which the sample is exposed to air. The absolute deviation of an experimental weight-gain curve from other curves at the same temperature is somewhat constant during the whole oxidation period(Fig.4.2). Therefore it gives a bigger relative deviation in the initial period since the magnitude of the weight-gain value at that time is small. The accuracy referred to here is after some time of oxidation, say, one hour. (1) Simulated Single Mechanism Oxidation above 950 °C Fig.5.6 shows the measured weight-gain curve from verification experiment and the simulated weight-gain curve and heat evolution data under single oxidation mechanism(not crossing transition temperature of 950 °C). In the initial period, the measured data shows the abnormal fluctuations(Fig.5.6). However this usually does not occur. If the general average Ch. 5 Modeling of Oxidation Process 106 trend is checked, the model predicts the scale growth within fifteen percent accuracy. Considering the high scattering nature of oxidation process, this is a satisfactory result. During the initial isothermal oxidation period at 1125 °C, the modelgives weight-gain values that fall between these of 1100 °C and 1150 °C. Therefore the current model could be used to derive isothermal oxidation weight-gain data which has not been experimentally measured. To verify the convergence of the current model, the time step was reduced from 6 seconds to 2 seconds. The predicted results are also shown in Fig.5.6. But these curves with different time steps cannot be discernable since they are overlapped. It is seen that the convergence of the model is very good. During the simulation, the model calculated the temperature history and the heat flux. Fig.5.7 gives another example of oxidation under single oxidation mechanism. It also shows a good agreement between the experimental curve and the predicted curve. It is seen that when the temperature is increased the weight-gain shows an increased oxidation rate(Fig.5.7). Ch. 5 Modeling of Oxidation Process 107 : • simulated temperature ; ' -'---1 : ! : i measured weight gain -1 \ • predicted weight gain ~ 'ii y\\ -1200 1100 1000 ~ q , aT 900 is 800 S 700 600 100 200 300 time(min) 400 500 Fig. 5.6 Simulated and measured weight gain curves of IF steel oxidation - s imulatec t e m p e r a t i ire -'•_ y ---measured w e i g m gam — -/ P redicted we sight gain --1200 1100 1000 o "ST La 900 is a> a. E 800 £ 700 600 100 200 300 400 500 600 700 time(min) Fig.5.7 Simulated and measured weight gain curves of IF steel oxidation Ch.5 Modeling of Oxidation Process 108 (2) Simulated Single Mechanism Oxidation below 950 °C Fig.5.8 shows the results for single mechanism oxidation below 950 °C. It is shown that the simulation gives good results(within fifteen percent accuracy) during a very long time period. Due to the decrease in oxidation rate, the weight gain is much smaller during 820 °C oxidation period than that during the period at 925 °C. O) 4.00 3.50 3.00 2.50 « 2.00 a cn 1.50 3 5 1.00 0.50 0.00 ---simulated temperature --— measured weight c ain predicted weight gam : 1 -100 200 500 600 1200 1100 1000 5-900 is a a. E 800 £ 700 600 700 300 400 t ime(min) Fig.5.8 S i m u l a t e d a n d m e a s u r e d we ight ga in c u r v e s of I F steel o x i d a t i o n (3) Simulated Cross-mechanism Oxidation Fig.5.9 shows the measured experimental weight-gain curve and the predicted curve when the sample undergoes different oxidation mechanisms during oxidation. As expected for the model, the predicted curve satisfactorily follows the measured curve in the initial oxidation stage 1120 °C and then radically changes to the slow oxidation period at 920 °C. Even in this case, the predicted weight-gain curve also shows good agreement with the measured weight-Ch. 5 Modeling of Oxidation Process 109 gain curve. This is much better than Abuluwefa et al's modeH7] which would give very big disagreement in high oxygen atmosphere such as air even at about three hours. From the above verification, it is concluded that the current model gives results in good agreement with the experimental curves during very long time period in all four cases. The cross-mechanism oxidation is also well simulated. These results satisfactorily justify the proposed model of oxidation growth. 4.00 3.50 3.00 I. 2.50 "5 2.00 O) .S> 1.50 a> 5 1.00 0.50 0.00 -s i m u l a t e d t e m p e r a t u r e -: --m e a s u red w e igh t g a in pre< j i c t e d v v e i g h t gain • 1 --1200 1100 1000 ~ o 5 900 is Li 0) Q. E 800 ^ 700 600 0 100 200 300 400 500 600 700 800 900 1000 1100 time(min) Fig.5.9 Simulated and measured weight gain curves of IF steel oxidation CHAPTER 6 CONCLUSIONS AND FUTURE WORK The following is a summary of the findings and conclusions of this investigation of IF steel oxidation: (1) The model proposed for predicting the scale growth of IF steel is feasible and usable. The predicted weight gain data is in good agreement with the measured weight gain curve during long time periods in all verification experiments. (2) There are two different oxidation mechanisms for IF steel. One corresponds to the adherent, dense and compact scale formed below 950 °C. The repeated experimental runs under this mechanism show a good consistency with each other. Another mechanism corresponds to the nonadherent, detached, laminated and porous scale formed above 950 °C whose growth is influenced by many factors such as pores and fissures in the scale, temperature, oxygen potential and interface reaction. Repeated experimental runs under this mechanism also show an acceptable consistency with each other. 110 Ch. 6 Conclusions and Future Work 1 1 1 (3) The adherent scale below 950 °C usually shows a parabolic weight gain curve since it follows the lattice diffusion rule. The nonadherent scale growth above 950 °C shows an initial nonlinear growth period and a following linear growth period. (4) The growth of the scale is through the growth of oxide grains. In some cases, long columnar oxide grains are formed which are often manifested by the fractograph of nonadherent scale layers. (5) For the adherent scale, there is no apparent weak interface. The fracture of scale is usually through oxide grains. For the nonadherent scale, the weakness and fracture is along the grain boundaries. (6) For the adherent scale, the contact between the scale layer and the steel base is maintained on some island-shaped areas, not uniformly through the whole interface area. Even on these island-shaped contact areas, the contact is essentially maintained through smaller needle-shaped or ball-shaped areas. So the assumption that the contact is maintained along whole interface area will give an apparent discrepancy with the actual situation. (7) The transition temperature between the adherent scale growth mechanism and the nonadherent scale growth mechanism is not influenced by the phase transformation temperature from a-iron to y-iron but possibly by the oxide growth dynamics and creep process. (8) For the adherent scale, a decreasing number of magnetite particles are precipitated from outer to inner areas inside the wustite layer if the sample is cooled rapidly while for the nonadherent scale, precipitation is not observed. Therefore, the adherent scale mainly Ch.6 Conclusions and Future Work 112 consists of wustite and the nonadherent scale mainly consists of magnetite and/or hematite. (9) The sample geometry has a significant influence on the oxidation process. If the oxidation of strip-like material is to be studied, samples as large as possible should be employed. (10) The idea to simulate the oxidation process with a set of isothermal oxidation data is feasible but a lot of specific features have to be accommodated. This makes the simulation complicated. (11) The model could be applied as a reference to industrial operation with the same conditions, i.e. oxidation of IF steel in air. Two important aspects of the oxidation process deserve further study to make the present model a complete package: (1) The corresponding sets of oxidation data in combustion atmosphere and air/steam atmosphere should be obtained and integrated into the present model. There will be some corresponding significant modifications to the present model. But the basic approaches should be similar. (2) 3D Stress-strain model based on the scale growth model could be developed to simulate descaling process. 113 BIBLIOGRAPHY [I] C.E.Birchenall: "A brief history of study of oxidation of metals and alloys, high temperature corrosion", NACE 6, 3/1981 [2] M.J.Bennett: "Real time studies of scale development and failure", High Temperature Corrosion of Advanced Materials & Protective Coatings, Y.Saito et al, Elsevier Science Publishers B.V., 1992 [3] W.T. Lankford et al: "The making, shaping and treating of steel", 10th Edition, USS, 1985 [4] D.W.Juenker, R.A.Meussner & C.E.Birchenall: "Cavity formation in iron oxide", Corrosion—National Association of Corrosion Engineers, 14/1958 [5] K.Sachs & C.W.Tuck: "Surface oxidation of steel in industrial furnaces", Reheating for Hot Working, Proc. of Iron and Steel Institute, 1/1968 [6] Ludwig Hachtel: "Scale structure and scale defects on hot strip", Prakt. Metallogr., 32/7/1995 [7] H.T. Abuluwefa, R.I.L.Gutherie & F.Ajersch: Oxidation of Low Carbon Steel in Multicomponent Gases. Part II. Reaction Mechanisms during Reheating", Metallurgical & Materials Transactions A, 28A/8/1997 [8] N.Birks & G.H.Mier: "Introduction to high temperature oxidation of metals", Edward Arnold Ltd, 1993 [9] M.Lee & R.A.Rapp: "Coalescence of wustite grains during iron oxidation in a hot-stage environmental SEM", Oxidation of Metals, 27/3/4/1987 [10] T.Smith & L.W.Crane: "Adsorption of oxygen on steel", Oxidation of Metals, 10/2/1976 [II] L.Himmel, R.F.Mehl & C.E.Birchenall: "Self-diffusion of iron in iron oxides and the Wagner theory of oxidation", Transactions AIME 197, Journal of Metals, 6/1953 [12] M.H.Davies, M.T.Simnad & C.E.Birchenall: "On the mechanism and kinetics of the scaling of iron", Trans. AIME 191(1951); Journal of Metals, 10/1951 [13] G.Garnaud & R.A.Rapp: "Thickness of the oxide layers formed during the oxidation of iron", Oxidation of Metals, 11/4/1977 1 1 4 [14] B.W.Dunnington, F.H.Beck & M.G.Fontana: "The mechanism of scale formation on iron at high temperature", Corrosion—National Association of Corrosion Engineers, Vol.8, 1/1952 [15] J.Kucera & M.Ffajduga: "Long-time oxidation of iron at 1100 C in air", Oxidation of Metals, 37/5/6/1992 [16] J.W.Picklens: "The microstructural characteristics of oxide scale formed during bar processing", Republic Steel Corporation, 39/1983 [17] A.U.Malik & D.P.Whittle: "Oxidation of Fe-C alloys in the temperature range 600-850 °C", Oxidation of Metals, 16/5/6/1981 [18] G.J.Yurek, J.P.Hirth & R.A.Rapp: "Formation of Two-phase Layered Scales on Pure Metals", Oxidation of Metals, 8/1974 [19] J.Baud, A.Ferrier, J.Manenc & J.Benard: "The oxidation and decarburizing of Fe-C alloys in air and the influence of relative humidity", Oxidation of Metals, 9/1/1975 [20] J.Paidassi, Trans. AIME 191(1952); Journal of Metals, 5/1952 [21] H. Abuluwefa, R.I.I.Guthrie & F.Ajersch: "Effects of Oxygen Concentration on Oxidation of Low-carbon steel in Temperature Range 1000 to 1250 °C", Oxidation of Metals, 46/5/6/1996 [22] C.W.Tuck, M.Odgers & K.Sachs: "The oxidation of iron at 950 °C in oxygen/water vapour mixtures", Corrosion Science, 9/1969 [23] H.T.Abuluwefa, R.I.L.Gutherie & F.Ajersch: Oxidation of Low Carbon Steel in Multicomponent Gases. Part I. Reaction Mechanisms during Isothermal Oxidation", Metallurgical & Materials Transactions A, 28A/8/1997 [24] A.Rahmel & J.Tobolski: "Einfluss Von Wasserdampf Und Kohlendioxyd Auf Die Oxydation Von Eisen in Sauerstoff Bei Hohen Temperaturen", Corrosion Science, 5/1965 [25] P.C.Rowlands, M.I.Manning and J.Soo: "The effect of substrate creep on the structure of oxide scale formed on Fe-9%Cr steels", 1980 [26] A.U.Malik and D.P.Whittle: "Influence of carbon on the oxidation behavior of iron base alloys in the temperature range of 600 - 850 °C", Proceedings JIMIS-3/1983 [27] A.U.Malik and D.P.Whittle: "Oxidation of Fe-C alloys in the temperature range 600-850 °C", Oxidation of Metals, 16/1981 115 [28] F.H.Scott & G.C.Wood: "Internal oxidation", Materials Science & Technology, Dec. 1988 Vol.4 [29] A.G.Goursat & W.W.Smeltzer: "Kinetics and morphological development of the oxide scale on iron at high temperatures in oxygen at low pressure", Oxidation of Metals, 6/2/1973 [30] S.Nazrazadani & A.Raman: "Morphology of oxide phases in the scales formed on steels", Pract. Met., 28/1991 [31] P.Kofstad: "On the formation of porosity and microchannels in growing scales", Oxidation of Metals, 24/6/1985 [32] S.Mrowec: "On the Mechanism of High Temperature Oxidation of Metals and Alloys", Corrosion Science, 7/1967 [33] Per Kofstad: "Oxidation Mechanism for Pure Metals in Single Oxidant Gases", High Temperature Oxidation, R.A.Rapp, NACE-6, Houston, 1983 [34] M.Lee & R.A.Rapp: "Development of scale morphology during wustite growth on iron at high temperature", Oxidation of Metals, 30/1/2/1988 [35] D.Caplan, G.I.Sproule, R.J.Hussey & M.J.Graham: "Oxidation of Fe-C Alloys at 700 °C", Oxidation of Metals, 13/3/1979 [36] Per Kofstad: "High Temperature Corrosion", Elsevier Applied Science, 1988 [37] J.Tominaga et al: "Manufacture of Wire Rods with Good Descaling Property" Transaction ISIJ, 22/1982 [38] Y.Zhang & D.A.Shores: "Cracking and spalling of oxide scale from 304 stainless steel at high temperatures", Journal of Electrochem. Soc, 141/5/1994 [39] D.L.Deadmore & C.E.Lowell: "The effect of AT(oxidizing temperature minus cooling temperature) on oxide spallation", Oxidation of Metals, 11/2/1977 [40] S.S.Rao: "Finite Element Method in Engineering", Pergamon Press, 2 n d Edition, 1989 [41] Lewis Morgan & Zienkiewicz: "Numerical Methods in Heat Transfer", John Wiley & Sons, 1981 [42] D.S.Burnett: "Finite Element Analysis", Addison-Wesley Publishing Co., 1987 116 A P P E N D I X A (1) Format of Program Input File 1 2 3 node node 4 5 6 node 7 8 9 element element 10 11 12 13 14 15 16 17 element 18 19 20 21 22 23 24 25 property 26 27 alpha 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 beta 43 44 45 46 47 48 49 50 51 51 52 53 54 55 56 mu 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 alpha 73 74 75 76 77 78 79 80 beta 81 82 83 84 85 86 87 88 mu 89 90 91 92 93 94 95 96 grow_curves grow_curves 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 grow_curves 127 128 129 130 131 132 133 134 139 140 141 146 147 148 149 150 151 152 135 136 137 138 142 143 144 145 153 154 155 156 117 grow_curves 157 158 159 160 161 162 163 164 165 166 167 168 step step 169 170 171 172 interior_load interiorjoad 173 174 175 176 environment_temperature 177 178 step step 179 180 181 interiorjoad interiorjoad 182 183 184 185 environmenttemperature 186 187 188 (2) Explanation of Various Terms Term 1: any string less than four letters such as " — " Term 2: problem type. It could be "boundary_value" or "intial_boundary_value" Term 3: problem description(string less than 50 letters). It will be output to tabulate output file. Term 4: node number. Term 5: increment in node numbers between this node and node read on next line. Term 6: coordinate of this node Terms 7, 8,9: same as terms 4, 5, 6 Term 10: element number. Term 11: increment in node numbers between this element and element read on next line. Term 12: element type number. 11— 2-node linear element 12— 3-node quadratic element 13— 4-node cubic element 14— 5-node quartic element Term 13: number of Gauss points. N for n-point quadrature rule. Term 14: identification number(group ID) for group of physical properties associated with this element. It could be 1, 2, 3,... Term 15: node number of first node in this element. Term 16: node number of second node in this element. Term 17: node number of third node in this element. Term 26: number of property groups. Term 27: group ID of this physical property group. It could be 1, 2, 3,... Term 28: number of segments in this "alpha" property curve. It could be 1, 2, 3, ... • Term 29: order of this segment of property curve. It could be 1, 2, 3,... Term 30: temperature value for first property data point. Term 31: property value of "alpha" for first property data point. Term 32: temperature value for second property data point. Term 33: property value of "alpha" for second property data point. Term 34: temperature value for third property data point. Term 35: property value of "alpha" for third property data point. Terms 36, 37, 38 39 40,41,42: same as terms 29,30, 31, 32,33, 34, 35 except it is for the second portion of the same curve. Term 43: number of portions of "beta" property curve. Terms 44,45,46,47,48,49, 50: same as terms 29, 30, 31,32, 33, 34, 35 except it is for 118 the first portion of "beta" curve. Term 72: group ID of second physical group data. Term 97: number of primary oxidation weight gain curves. Term 98: first weight gain curve ID — 1. Term 99: number of portions of first weight gain curve. Term 100: isothermal temperature for first weight gain curve. Term 101: weight gain value under previous oxidation mechanisms. Term 102: ID of first portion of first weight gain curve — 1. Term 103: time value of first point of first portion on first weight gain curve. Term 104: weight gain value of first point of first portion on first weight gain curve. Term 105: time value of second point of first portion on first weight gain curve. Term 106: weight gain value of second point of first portion on first weight gain curve. Term 107: time value of third point of first portion on first weight gain curve. Term 108: weight gain value of third point of first portion on first weight gain curve. Terms 109 to 115: same as terms 102 to 108 except it's for second portion of first weight gain curve. Terms 116 to 126: same as terms 98 to 115 except it's for second weight gain curve. Terms 127 to 156: same as terms 97 to 126 except it's for secondary weight gain curves below 950 °C. Terms 157 to 168: same as terms 97 to 126 except it's for secondary weight gain curves above 950 °C. Term 169: number of steps in first time interval. Term 170: time step value for all steps in first time interval. Term 171:0 value for heat transfer model, 0 < 9 < 1. Term 172: initial temperature value at beginning of program run. Term 173: first element number acted on by interior load. Term 174: value of interior load. Term 175: last element number acted on by interior load. Term 176: increment in element numbers between these first and last elements. Term 177: environment temperature value at beginning of first time interval. Term 178: environment temperature value at end of first time interval. Term 179: number of steps in second time interval. Term 180: time step value for all steps in second time interval. Term 181:9 value for heat transfer model, 0 < 6 < 1. Term 182: first element number acted on by interior load. Term 183: value of interior load. Term 184: last element number acted on by interior load. Term 185: increment in element numbers between these first and last elements. Term 186: environment temperature value at beginning of second time interval. Term 187: environment temperature value at end of second time interval. Term 188: another set of "step" data for third time interval or "end" if there are just two time intervals. (3) An Example of the Program Input File — initial_boundary_value simulate_heat_transfer_and_predict_scale_growth(15 8) node node 1 1 0.0 node 101 1 0.0018 element element 1 2 12 2 1 1 2 3 element 50 2 12 2 1 99 100 101 property 2 1 alpha 2 1 10.85 60.5 126.85 56.7 326.85 48.0 2 326.85 48.0 726.85 30.0 1200.0 8.71 beta 1 1 0.0 0.0 600.0 0.0 1200.0 0.0 mu 2 1 10.85 3.41e+6 126.85 3.82e+6 326.85 4.39e+6 2 326.85 4.39e+6 726.85 9.18e+6 1200.0 1.48e+7 2 alpha 1 1 10.85 10.26 526.85 9.87 1250.00 9.31 beta 1 1 0.0 0.0 600.0 0.0 1200.0 0.0 mu 1 1 10.85 4.03e+6 526.85 4.66e+6 1250.00 5.56e+6 grow_curves growcurves 13 1 1 0 0.0 1 0.0 0.0 1000. 0.0 3000. 0.0 2 1 500 0.0 1 0.0 0.0 1000. 0.0 3000. 0.0 3 2 700 0.0 1 0.0 0.0 19.35 0.0129 29.35 0.0132 2 29.35 0.0132 697.41 0.0350 1300.16 0.0421 4 2 750 0.0 1 0.0 0.0 36.02 0.0186 56.02 0.0197 2 56.02 0.0197 684.08 0.0447 1409.51 0.0738 5 2 800 0.0 1 0.0 0.0 200.58 0.1106 248.90 0.1217 2 248.90 0.1217 739.50 0.2308 1264.78 0.3478 6 2 850 0.0 1 0.0 0.0 532.00 0.3318 935.10 0.5666 2 935.10 0.5666 1512.36 0.7754 2508.25 1.0493 7 3 900 0.0 1 0.0 0.0 57.30 0.1900 114.60 0.3127 2 114.60 0.3127 287.90 0.5899 387.90 0.6849 3 387.90 0.6849 793.20 0.9740 1309.20 1.2563 8 3 950 0.0 1 0.0 0.0 33.33 0.3266 73.33 0.5370 2 73.33 0.5370 154.65 0.7503 251.96 0.8920 3 251.96 0.8920 758.58 1.4415 1379.85 1.8786 9 3 1000 0.0 1 0.0 0.0 61.31 0.4213 73.31 0.4300 2 73.31 0.4300 150.65 0.5970 262.63 0.7095 3 262.63 0.7095 670.58 0.9636 1263.85 1.1978 10 3 1050 0.0 1 0.0 0.0 49.33 0.5261 106.66 0.7763 2 106.66 0.7763 202.66 0.9755 312.02 1.1127 3 312.02 1.1127 768.02 1.3440 1288.03 1.5885 11 3 1100 0.0 1 0.0 0.0 49.35 0.7010 69.35 0.7986 2 69.35 0.7986 148.02 0.9744 205.37 1.0296 3 205.37 1.0296 625.42 1.3255 1254.83 1.8465 12 3 1150 0.0 1 0.0 0.0 41.35 0.6201 94.68 0.9093 2 94.68 0.9093 150.68 1.0500 244.03 1.2116 3 244.03 1.2116 689.42 1.7837 1269.48 2.5007 13 3 1200 0.0 1 0.0 0.0 12.00 0.5372 28.00 0.8080 2 28.00 0.8080 43.80 0.9400 105.53 1.2106 3 105.53 1.2106 477.40 2.0748 1077.48 3.2582 grow_curves 6 1 1 0 0.0 1 0.0 0.0 1000. 0.0 3000. 0.0 2 1 500 0.0 1 0.0 0.0 1773.0 0.0 3653.0 0.0 3 2 850 0.0 1 0.0 0.0 532.00 0.3318 935.10 0.5666 2 935.10 0.5666 1512.36 0.7754 2508.25 1.0493 4 1 850 1.4028 An Example of the Program Input File(continued) 1 0.0 0.0 536.10 0.0646 1356.20 0.1359 5 3 950 0.0 1 0.0 0.0 33.33 0.3266 73.33 0.5370 2 73.33 0.5370 154.65 0.7503 251.96 0.8920 3 251.96 0.8920 758.58 1.4415 1379.85 1.8786 6 1 950 1.3233 1 0.0 0.0 540.10 0.1164 1352.20 0.1767 grow_curves 6 1 3 950 0.0 1 0.0 0.0 33.33 0.3266 73.33 0.5370 2 73.33 0.5370 154.65 0.7503 251.96 0.8920 3 251.96 0.8920 758.58 1.4415 1379.85 1.8786 2 1 950 1.3233 1 0.0 0.0 540.10 0.1164 1352.20 0.1767 3 3 1100 0.0 1 0.0 0.0 30.66 0.5955 69.35 0.7986 2 69.35 0.7986 148.02 0.9744 205.37 1.0296 3 205.37 1.0296 625.42 1.3255 1254.83 1.8465 4 3 1100 0.983 1 0.0 0.0 26.66 0.1100 65.35 0.1675 2 65.35 0.1675 134.68 0.2425 184.03 0.2780 3 184.03 0.2780 730.77 0.4753 1344.16 0.6550 5 3 1200 0.0 1 0.0 0.0 12.00 0.5372 28.00 0.8080 2 28.80 0.8080 69.35 1.0093 105.53 1.2106 3 105.53 1.2106 477.40 2.0748 1077.48 3.2582 6 3 1200 1.647 1 0.0 0.0 36.00 0.2480 50.70 0.2690 2 50.70 0.2690 101.40 0.3477 158.70 0.4318 3 158.70 0.4318 740.10 0.9157 1386.90 1.3641 step step 3000 2.0 1.0 930.0 interior_load interiorjoad 1 0.0 50 1 environment_temperature 930.0 930.0 step step 180 2.0 1.0 interior_load interiorjoad 1 0.0 50 1 environment_temperature 930.0 820.0 step step 38820 2.0 1.0 interiorjoad interiorjoad 1 0.0 50 1 environment_temperature 820.0 820.0 end 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0078705/manifest

Comment

Related Items