UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Inverse segregation in Al-Cu alloys Prabhakar, Balasubramaniam 1973

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-UBC_1973_A7 P73_2.pdf [ 8.12MB ]
Metadata
JSON: 831-1.0078741.json
JSON-LD: 831-1.0078741-ld.json
RDF/XML (Pretty): 831-1.0078741-rdf.xml
RDF/JSON: 831-1.0078741-rdf.json
Turtle: 831-1.0078741-turtle.txt
N-Triples: 831-1.0078741-rdf-ntriples.txt
Original Record: 831-1.0078741-source.json
Full Text
831-1.0078741-fulltext.txt
Citation
831-1.0078741.ris

Full Text

INVERSE SEGREGATION IN Al-Cu ALLOYS BY ML^UBRAMANIAM PRABHAKAR B.Sc. , Osmania University, India, 1967 B.E. (Metallurgy), Indian Institute of Science, India, 1970 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in the Department of METALLURGY We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA A p r i l , 1973 In present ing th is thes is in p a r t i a l fu l f i lment o f the requirements fo r an advanced degree at the Un ivers i ty of B r i t i s h Columbia, I agree that the L ibrary sha l l make i t f r ee ly ava i l ab le for reference and study. I fur ther agree that permission for extensive copying of th is thes is for s c h o l a r l y purposes may be granted by the Head of my Department or by h is representa t ives . It is understood that copying or pub l i ca t ion of th is thes is f o r f i n a n c i a l gain sha l l not be allowed without my wr i t ten permission. Department of Metallurgy  The Univers i ty of B r i t i s h Columbia Vancouver 8, Canada Date 12-bft June 1973 ABSTRACT Inverse segregation has been studied in unidirectionally cast ingots as a function of alloy composition and casting con-ditions. C h i l l face segregation and the extent of segregation adjacent to the c h i l l face has been examined in the Al-Cu and the Al-Ag systems, using radioactive copper (Cu 6 4) and radio-active s i l v e r (Ag 1 1 0) to measure the segregation. The results indicate that the observed segregation is markedly dependent on the alloy composition, melt superheat, thermal conductivity of the c h i l l , temperature gradient in the casting, and other casting variables. The results obtained are compared to those reported i n the literature for the Al-Cu alloys, and to the values pre-dicted theoretically. i i i TABLE OF CONTENTS Page 1. INTRODUCTION 1 1.1 Inverse segregation: Reported experimental results and theories 2 1.1.1 Summary of the different theories of inverse segregation -3 1.1.2 Quantitative predictions of inverse segregation 5 1.1.2.1 Theoretical model for c h i l l face segregation • • • 5 1.1.2.2 Comparison with experimental results 12 1.1.2.3 Extensions of the theories to predict composition variations within the ingot, away from the c h i l l face ,. ... 15 1.2 Exudations 24 1.3 Aim of the present work 25 2. EXPERIMENTAL PROCEDURES 27 2.1 Materials 27 2.2 Alloy preparation 27 2.3 Casting 29 2. 4 Metallography and autoradiography 32 2.5 Measurement of the variations of the solute concentrations 33 2.5.1 Radioactive tracer method 33 2.5.2 Chemical analysis 37 2.6 Microprobe analysis , 38 2.7 Temperature measurements 38 i v Page 2.8 Modification of casting procedures 39 2.8.1 Degassing 39 2.8.2 Grain refining 40 2.8.3 Exudations HO 2.8.4 Thermal conductivity of the c h i l l s 40 3. OBSERVATIONS 42 3.1 Tabulation of experiments 42 3.2 Homogeneity of the melt 42 3.3 Cast structure 48 3.4 Measurement of the i n i t i a l melt composition 51 3.5 Distribution of Cu6k along the ingot length 54 3.5.1 Observation of macrosegregati.cn near the c h i l l 54 3.5.2 Exudations '. 59 3.5.3 Effect of superheat 60 3.5.4 Effect of changing the rate of heat removal 65 3.5.5 Effect of changing composition 65 3.5.6 Special treatments 71 3.6 Saturation effects in the s c i n t i l l a t i o n counter 79 3.7 Temperature distributions in the casting during sol i d i f i c a t i o n 83 3.8 Chemical analysis of ingot samples 95 3.9 Inverse segregation in Al-Ag ingots 99 4. DISCUSSION 103 4.1 Scatter of the results 103 4.2 C h i l l face segregation 105 V Page 4.3 Effect of superheat and thermal conductivity of the c h i l l 108 4.4 Exudations on the c h i l l face l l 4 4.5 General mechanism of inverse segregation 117 4.6 Effect of cast structure on segregation 121 4.7 Effect of the gas content of the melt 123 4.8 Comparison of theories 124 4.9 Turbulence during casting 124 5. SUMMARY AND CONCLUSIONS 126 6. SUGGESTIONS FOR FUTURE WORK 128 APPENDIX A 129 A.l S t a t i s t i c a l consideration of radioactivity measurements 129 A. 1.1 Confidence limits and errors 131 A. 1.2 Errors in the counting rate determinations 132 REFERENCES 136 v i LIST OF FIGURES Figure Page 1(a) Schematic representation of the soli d - l i q u i d region near the c h i l l face. (b) Schematic l i q u i d concentration distribution i n this region 7 2 Constitution diagram and variation of the specific volumes of sol i d and liquid for Al-Cu alloys , ,.. 13 3 Theoretical predictions and experimental results for maximum c h i l l face segregation vs. alloy composition for the Al-Cu system8 13 4 Constitutional - specific volume diagram for the Bi-Sb system 14 5 Theoretical curve and experimental results for the maximum c h i l l face segregation vs. alloy compositions for Bi-Sb a l l o y s 9 14 6 Constitutional - specific volume diagram for the Al-Zn system 16 7 Theoretical curve and experimental results for the c h i l l face segregation vs. alloy composition for Al-Zn a l l o y s 1 2 16 8 Constitutional - specific volume relationships for the Al-Si system 17 9 Theoretical curve and experimental results for the c h i l l face segregation vs. alloy composition for Al-Si a l l o y s 1 3 17 10(a) Representation of crystal growth and accompanying concen-tration changes in an unidirectionally s o l i d i f i e d ingot (dotted lines indicate constantly increasing length of the solid-liquid region t i l l the ingot top i s reached) 19 (b) Liquid mass distribution curves i n the model ingot for representative times during s o l i d i f i c a t i o n 6 ' 19 11 Theoretical and experimental segregation curves for an Al-10% Cu ingot 20 v i i Figure Page 12 Theoretical and experimental segregation curves for an Al-15% Cu ingot 9 20 13 Comparison of the theoretical predictions of the maximum c h i l l face segregation in Al-Cu alloys 23 14 Schematic representation of the casting arrangement for casting procedure (a) 30 15 Schematic representation of the casting arrangement for casting procedure (b) 31 16 Gamma ray spectrum for Cu 6 t f obtained on the Picker Nuclear s c i n t i l l a t i o n counter 36 17 Autoradiograph and corresponding etched surface of an Al-Cu ingot (ingot 94) 46 18 Autoradiograph and corresponding etched surface of an Al-Cu ingot (ingot 84) V 47 19 Macrostructure perpendicular to the c h i l l face of Al-Cu. ingots cast (a) in the furnace arrangement (b) in fibrefrax molds (c) in the vacuum induction furnace 50 20 Normalized activity of samples taken at different times from the same melt 52 Normalized activity vs. distance from the c h i l l face for Al-Cu alloys as indicated below: 0 21 Al-10% Cu, fibrefrax mold, copper c h i l l , 41 C superheat 55 22 A l-5% Cu, fibrefrax mold, copper c h i l l , 54°C superheat 57 23 Al-10% Cu, fibrefrax mold, copper c h i l l , 91°C superheat 58 24 Al-15% Cu furnace mold, copper c h i l l , 40°C superheat 61 25 Al-15% Cu,. furnace mold, copper c h i l l , 80°C superheat 62 26 Al-15% Cu, furnace mold, copper c h i l l , 115°C superheat 63 27 Al-15% Cu, furnace mold, copper c h i l l , 160°C superheat 64 28 Al-10% Cu, fibrefrax mold, copper c h i l l , 93°C superheat '66 v i i i Figure Page 29 Al-10% Cu, fibrefrax mold, copper c h i l l , 65?C superheat 67 30 Al-10% Cu, fibrefrax mold, s:.-ekee& c h i l l , 91 °C superheat 68 31 Al-10% Cu, fibrefrax mold, a&teeb c h i l l , 67°C superheat. 69 32 Al-5% Cu, fibrefrax mold, stainless steel c h i l l , 107°C superheat 72 33 Al-15% Cu, fibrefrax mold, stainless steei c h i l l , 100°C superheat 73 34 Al-20% Cu, fibrefrax mold, stainless steel c h i l l , 103°C superheat 74 35 Al-25% Cu, fibrefrax mold, stainless steel c h i l l , 103°C superheat 75 36 Al-10% Cu, fibrefrax mold, copper c h i l l , 84 C superheat 77 37 Al-10%, Cu, fibrefrax mold, stainless steel c h i l l , 91°C superheat, 78 38 Al-10% Cu, furnace mold, copper c h i l l , 1(5 C superheat, high counting rate 80 o 39 Al-10% Cu, furnace mold, copper c h i l l , 10 C superheat, intermediate counting rate 81 40 Al-10% Cu, furnace mold, copper c h i l l , 10°C superheat, low counting rate 82 41 Cooling curves for an Al-10% Cu ingot under the casting conditions indicated • 84 42 Temperature distributions i n a solidifying Al-10% Cu ingot • cast under the conditions indicated 85 43 Movement of eil^ ectwC-j isotherms for Al-10% Cu ingots cast under the conditions indicated..... 87 44 • Movement of e^tectce, isotherms for Al-10% Cu ingots cast under the conditions indicated 88 45 Microstructures of sections par a l l e l tp the c h i l l face for Al-10% Cu ingots 92 (a) cast in the furnace (procedure (a) ) 10OX (b) cast in fibrefrax molds 10OX i x 46 Dendrite arm spacing vs. superheat for Al-10% Cu ingots cast under the conditions indicated 93 47 Lu. vs. superheat for Al-10% Cu ingots cast under the conditions indicated 94 48 Normalized concentrations from chemical analysis super-imposed on the normalized activity distribution 96 49 Normalized concentrations from chemical analysis super-imposed on the normalized activity distribution 97 50 Normalized concentrations from chemical analysis super-imposed on the normalized activity distribution 98 51 Normalized activity vs. distance from the. c h i l l face for an Al-10% Ag ingot 100 52 Normalized activity vs. distance from the c h i l l face for an Al-15% Ag ingot 101 53 Theoretical predictions and experimentally obtained c h i l l face segregations 106 54 Theoretical predictions .and experimentally obtained c h i l l face segregations 107 55 C h i l l face-segregation vs. superheat for Al-5% Cu ingots cast under the conditions indicated 109 56 C h i l l face segregation vs. superheat for Al-10% Cu ingots cast under the . conditions indicated I l l 57 C h i l l face segregation vs. superheat for Al-15% Cu ingots cast under the conditions indicated.. , 113 58 Microstructure of the base of an Al-10% Cu alloy ingot showing exudations (after Youdelis 6) 110X 115 59 Microstructure of the base of an Al-10% Cu ingot, per-pendicular to the c h i l l face from the present investigation, i n which exudations were induced. 100X . 115 60 Microstructure of the base of an. Al-10% Cu ingot, per-pendicular to the chill, face, cast normally., showing absence'of exudations. 100X 116 61 L vs. superheat for Al-10% Cu alloys for the casting conditions indicated 119 X Figure Page 62 Comparison of the Poisson and Gaussian distributions 130 63 The error of counting determinations A. Probable error B. Standard deviation C. Nine tenths error D. Ninety-five hundredths error E. Ninety-nine hundredths error 130 64 Activity vs. weight of the Al-Cu sample containing Cu 6 4 134 x i LIST OF TABLES Table Page I Data and results for Al-5% Cu, Al-20% Cu, A l - 2 5 % Cu ingots 43 II Data and results for Al-10% Cu ingots 44 III Data and results for Al-15% Cu ingots 45 IV Temperature measurements for Al-Cu alloys 90 V Errors used to define confidence i n t e r v a l s 2 0 133 ACKNOWL£rX_ENT The author wishes to express his sincere gratitude to Dr. F. Weinberg, for his advice and assistance during the course of this investigation. Thanks are also extended to members of the faculty, and fellow graduate students for helpful discussions. The assistance of the departmental technical staff, i n particular, the help extended by Mr. J. Brezden during much of the experimental work, i s greatly appreciated. Financial aid from the U.B.C. Research Committee Grants and a Killam Predoctoral Fellowship i s gratefully acknowledged. 1-1. INTRODUCTION Cast materials are used extensively, in a wide variety of appli-cations. With the increasing sophistication of modem technology, greater demands are being made on the quality and r e l i a b i l i t y of cast products. In addition, competitive manufacturing processes and new materials have resulted in efforts being made to increase the e f f i c -iency of the casting process. Accordingly, the production, structure and properties of castings, in both ferrous and non-ferrous materials, warrants continuing investigation. The fundamental nature of the so l i d i f i c a t i o n of alloys, in which solid of one composition forms from li q u i d of another composition, makes i t impossible to cast homogeneous materials with uniform struc-ture and properties. Inhomogeneities can occur both on a macroscopic and microscopic scale i n the s o l i d i f i e d metal. On a macroscopic scale, macrosegrega-tion can occur i n three different forms, which may or may not occur . together.' The three forms commonly observed can be clas s i f i e d as follows: 1. Normal segregation: Solute i s segregated in a manner pre-dicted by the phase diagram, i n which the alloy composition increases with increasing distance from the c h i l l face, for a system in which the segregation coefficient i s less than unity. 2. Gravity segregation: Solute segregates as a result of den-sity differences i n the constituents of the casting during 2. s o l i d i f i c a t i o n . This process i s most pronounced in slowly s o l i d i f i e d large castings. 3. Inverse segregation: In this case, solute content decreases with increasing distance from the c h i l l face, for a system with a segregation coefficient less than unity. It i s most pro-nounced in a small region near the c h i l l face. The present investigation i s primarily concerned with the last category, namely inverse segregation. 1.1 Inverse segregation: Reported experimental results and theories Inverse segregation has been studied extensively, both theoret-i c a l l y and experimentally for many years, and many theories have been proposed to account for the apparently anomalous segregation behaviour. It i s also of considerable interest, commercially, particularly i n non-ferrous materials, since inverse segregation can be seriously detri-mental to the surface quality of the cast product. Pell-Walpole 1 has reviewed in considerable detail the experimental evidence and theories presented prior to 1949, primarily in cast bronzes and other non-ferrous alloys. ' A short summary of the theories w i l l be presented below. More recently, Vosskuhler 2 has collected the results of further work, and has given a c r i t i c a l review of these results and the newer theories. He has also added experimental evidence which he obtained on a number of alloy systems with different types of equilibrium.phase diagrams. 3. Since the review of Vosskuhler, more experimental results have been reported. In addition, theoretical models for macro-segregation, of which inverse segregation is a part, have been extended and refined, primarily by Flemings and his co-workers.3'4'5 1.1.1 Summary of the different theories of inverse segregation These theories have been discussed in detail elsewhere1'2'6; as a result, only brief summaries will be presented here. The theories f a l l broadly into two groups. In the first group, i t is assumed that there is a solute enrichment of the liquid next to the mold surface before solidification occurs, and that the composition stays enriched .during the solidification process. In the second group, i t is-assumed that solidification proceeds as predicted by the equilibrium diagram. Inverse segregation results from interv-dendritic fluid flow during solidification. The second, and more recent group, is now considered to correctly describe the major processes occurring during inverse segregation. The theories are briefly summarized as follows: (a) The theory of mobile equilibrium: Le Chatelier's principle is assumed to be operative during solidification. The hypothesis is then put forward that, in order to retard solidification, the com-ponent cf the alloy which would tend to lower the melting*point segregates toward the mold surface. (b) Benedicks' theory: The Ludwig-Soret effect, which involves the appearance of a concentration gradient under a temperature grad-ient, is the mechanism leading to inverse segregation at the ch i l l 4 . face. (c) Undercooling of the alloy: The l i q u i d underccols next to the mold face and therefore deposits so l i d richer i n the solute element in this region. (d) Theory of crystal migration: Depending on the density differences, there i s a migration of primary crystals away from a suddenly chil l e d mold wall. (e) Gas evolution theory: This theory was developed to explain the segregation in cast bronzes. The gas present in the cooling alloy becomes concentrated in the last region to s o l i d i f y , and the evolution of the gas on cooling tends to force the enriched li q u i d into the interdendritic regions towards the outside of the casting, i.e. , towards the mold face. (f) Crystallization pressure theory: Crystallization pressure, or the pressure exerted on the li q u i d by the growing dendrites, i s the driving force for transport of the enriched residual liquid towards the mold surface of the casting. (g) Shell contraction pressure theory: A crystalline shell i s formed surrounding the molten metal. On cooling, this shell con-tracts , and the resultant pressure set up i n the liquid causes the shell to rupture and the liquid to flow to the outside. (h) Interdendritic flow theory: This theory was f i r s t pro-posed in 1921, and, in a modified form, i s now the accepted one. The sol i d i f i c a t i o n shrinkage of the primary dendrites i s followed by interdendritic backflow of the enriched residual liquid. The 5. segregation depends on the growth rate and the direction of the primary dendrites, solute diffusion i n the solid and the liq u i d , and the temperature dependence of the specific volume of the solid and liquid. 1.1.2 Quantitative predictions of inverse segregation Scheil 7 was the f i r s t to derive an analytical expression to define the amount of segregation that could occur at the c h i l l face of a binary alloy during unidirectional s o l i d i f i c a t i o n , as a func-tion of the composition of the alloy. The predictions made by the theory involve the knowledge of the specific volumes of the solid and the liq u i d , as a function of temperature near the melting point, and the equilibrium phase diagram of the constituents of the binary alloy. The equations given by Scheil have been modified, and the theory has been extended by Kirkaldy and Youdelis 6' 8 to predict the composition distribution along the longitudinal direction (the direc-tion of heat removal) of the ingot away from the c h i l l face. 1.1.2.1 Theoretical model for c h i l l face segregation The theoretical model for inverse segregation 9 makes the follow-ing assumptions: • (a) Surface exudations are absent on the c h i l l face. (b) There i s negligible undercooling at the solid-liquid inter-face, i . e . , at the interface, equilibrium conditions exist. 6. (c) There i s no restriction on the backflow of li q u i d through the interdendritic regions. (d) Gas levels i n the ingot are low and no porosity exists. (e) No significant solute diffusion occurs in the solid state, i.e. , conplete coring i s assumed to be present i n the solid-i f i e d metal. (f) There is no diffusion or l i q u i d mixing in the li q u i d in a direction pa r a l l e l to the growth direction. The concentration gradients in the li q u i d along this direction are assumed small. There i s complete mixing in the liquid perpendicular to the growth direction. (g) The distribution of the liquid and solid mass throughout the solid-liquid region can be estimated from temperature measure-ments during s o l i d i f i c a t i o n and the equilibrium phase diagram. The analysis uses a "control volume", a representative section of the system, and evaluates the mass and concentration balances r for material entering and leaving this volume. With assumed or measured boundary conditions, the whole system i s then defined, and the processes taking place during s o l i d i f i c a t i o n are characterized. The control volume selected i s shown in Figure 1 ( a ) , position A, adjacent t o t h e c h i l l . This volume i s part of the solid-liquid region, drawn schematically i n the Figure. The solute concentra-tion i n .the liquid (x-x) as a function of the distance from the c h i l l i s shown in Figure Kb). Fig. 1(a) (b) Schematic representation of the solid-liquid region near the c h i l l face. Schematic liquid concentration distribution in this region. 8 . Denoting so l i d by the subscript s and liquid by the subscript X, the mean composition C of the volume element after s o l i d i f i c a t i o n i s given by r m i C l + m C " s s + m s (1) where a =/ v -r V ( -v v „ (2) v g represents the specific volume of the so l i d , and the specific volume of the li q u i d , and m^m represent the masses of the liq u i d and s o l i d , respectively. The segregation AC i s given by AC = C - Co (3) where Co i s the mean composition of the ingot. For the binary eutectic alloy system, the solute concentration C i s given by C v ™ £ E f * E + m s E C s E. m a + msE (M where m^ and are the liquid and solid masses respectively i n the representative volume at the eutectic temperature, c ^ i s the mean solute content in the liquid at the eutectic, and the mean solute 9. content i n the cored crystals. For the volume element at the c h i l l face, there i s flow only into the element, and the volume V of the element i s given by V =v m +v„ m„ s s -SL SL (5) On so l i d i f i c a t i o n and on cooling, there i s shrinkage i n the volume element, and i f this i s completely f i l l e d by backflow of liq u i d , then dV = 0 = v„dm„ + v dm + m„dv„ + m dv l SL s s I SL s s (6) If v i s considered to be constant, this gives S 5 6 or dmc = dm = SL -dm -bdm rr V s + m £ d v £ v. v„ dm (7: ( 8 ) where ' b = (9) Considering mass balances i n the volume element after some so l i d i f i c a t i o n , and assuming complete coring, d(m„C„) = -C dm + I SL - s s C dm (v - v ) - Cm id S S . J6 J6 J6 (10) 10. where the right hand side of the equation i s composed of terms considering the solute mass i n the s o l i d i f i e d region, the solute mass gained due to volume shrinkage in the elements, and the solute mass gained due to concentration change of the liquid. The solute mass added due to temperature change i s considered to be negligible. from equation (9), and equation (10) reduces to drn^  = bdC £ = _ b dC £ ( 12) using, CH ' °s = 1-K = A • (13) since both A and b vary with composition, stepwise integration becomes necessary, between the i and i+l"1"^1 step. m m . . f i + l . / dms = - 1 / dm^  (ik) m b m. s. £. 1 x gives m = ms> + 1 (m^ - m ) l+l x b. x l+l v ^ 1 x 11. The solute mass in the cored primary crystals i s given by d (m C ) = C dm s s s s (18) The right.hand side of the equation reduces to C dm =/ m„ dC„ - m„dC s s ( a a A and i+1 • d(m C ) s s 1 - A . 1 A. -b. A. T[ W * ] dC, (19) gives m C m C 1+1 1+1 1 1 + X1 - A i l m„.ci'. 1 1 n A.-b. c l . 1 1 ( i + l \ A. - 1 ( 2 0 ) Stepwise integration i s carried out u n t i l m E C E and m £ E are reached. The values obtained are substituted into equation (4) and (3) to give the maximum c h i l l face segregation. 1.1.2.2 Comparison with experimental results For the Al-Cu system, the temperature dependence of the specific volumes of sol i d and l i q u i d , and the equilibrium phase diagram are given i n Figure 2. The specific volume data was obtained by Sauerwald 1 0. Using this data, Scheil calculated values of the maximum segregation at the c h i l l .face for unidirectionally s o l i d i f i e d alloys with compositions varying between 0 and 33% Cu, obtaining the solid line shown i n Figure 3. The experimental results of the c h i l l face segregation as determined by Scheil, Adams11 and Kirkaldy and Youdelis 8 are also presented in the Figure. Comparing the experimental values with the theoretical curve, i t is apparent that there i s excellent agreement between the experimental results and the theoretical curve for a l l three investigations over the entire composition range examined and for both slow and fast cool i n the results of Kirkaldy and Youdelis. Similar measurements and calculations of inverse segrega-tion for the Bi-Sb, Al-Zn and Al-Si alloy systems were made. The specific volume and phase diagram data for the Bi-Sb system is given in Figure 4 and the corresponding theoretical curve and 13. 5401 1 i I I I I I 0 5 10 15 20 25 30 33 C O N C E N T R A T I O N , w t - % Cu Fig. 2 Constitution diagram, and variation of the specific volumes of so l i d and li q u i d for Al-Cu alloys. Fig. 3 Theoretical predictions and experimental results for maximum c h i l l face segregation vs. alloy composition for the Al-Cu system 8. SPECIFIC V O L U M E , cmVqm 016 0_I4 0-12 (HO 4 Constitutional - s p e c i f i c volume diagram f o r the Bi-Sb system. — Theoretical • Experimental Sb 10 20 30 40 50 60 70 80 90 Bi C O N C E N T R A T I O N , w t - % Bi 5 Theoretical curve and experimental res u l t s f o r the maximum c h i l l 1 face segregation vs. a l l o y composition f o r Bi-Sb a l l o y s 9 . experimental points for the c h i l l face segregation as reported by Youdelis 9 are given i n Figure 5. This system i s of particular interest since pure Bismuth expands on freezing. No volume shrinkage occurs and therefore no inverse segregation should occur for alloys high in Bismuth concentration. As shown i n Figure 5, this i s what i s observed. Again, the f i t between theory and experiment i s excellent. The specific volumes of the s o l i d and liquid and the equi-librium phase diagrams for the Al-Zn system are shown i n Figure 6, and the corresponding theoretical curve for c h i l l face segre-gation, with experimental r e s u l t s 1 2 i s shown i n Figure 7. The experimental scatter (error bars) reported i s much larger i n this system, but using average values, there i s s t i l l excellent agreement between theory and experiment. The results for the Al-Si system have been reported recently by Ba l l and Youdelis 1 3 using the data of Figure 8. In this case, Figure 9, there i s a large discrepancy between theory and experiment, the measured concentrations at the c h i l l face being far below the calculated values. 1.1.2.3 Extensions of the theories to predict composition variations within the ingot, away from the c h i l l face Kirkaldy and Youdelis have extended the Scheil analysis to calculate tlie variation in the solute concentration with distance 16. Fig. 7 Theoretical curve and experimental results for the ch i l l face segregation vs. alloy composition for Al-Zn alloys 1 2. 17. C o n s t i t u t i o n a l - s p e c i f i c volume r e l a t i o n s h i p s f o r the A l - S i system. 0*24 0-20 • Theoretical Maximum Experimental 0-16 to 0*12 - / s SEGREGATION, 0-08 0*04 / • + • 1 * • • 0 I 1 I I 1 l \ 1 2 4 6 8 10 12 14 CONCENTRATION , W t / % Si - 0 « 0 4 • T h e o r e t i c a l curve and experimental r e s u l t s f o r the c h i l l face segregation vs. a l l o y composition f o r A l - S i a l l o y s 1 3 . from the c h i l l f a c e 6 ' 8 ' 9 . In this case, the control volume element i s shifted away from the c h i l l face to position B as shown in Figure 1. Solute transfer now occurs both into and out of the volume element, in order to feed the contraction to the l e f t of the volume element. The amount of the l i q u i d (mass distribution of liquid) in the sol i d - l i q u i d region i s assumed to increase linearly with distance in the freezing direction from the completely so l i d interface. The length of the solid-liquid region i s assumed to start from zero at the c h i l l face, increases to a steady state value, and then decreases as s o l i d i f i c a t i o n reaches the f i n a l part of the casting to freeze. The change in the length of the solid-liquid zone i s shown schematically in Figure 10(a), and the corresponding slope of the liquid mass distributions i n the casting at various stages of s o l i d i f i c a t i o n i s shown in Figure 10(b). The theoretical predictions of the solute segregation in an Al-10% Cu alloy, using the data in Figure 2, i s shown as the sol i d line i n Figure 11. Trie corresponding experimental values .are plotted and i t i s apparent that there i s excellent f i t between the experimental values and the theoretical curve over the entire casting. -One point at .about 0.2" i s a l i t t l e low; the remaining points coincide with dr are very close to the theoretical curve. The corresponding curve: and the experimental points for an Al-15% Cu alloy i s shown in Figure 12, again showing excellent agreement 19. Fig. 10(a) Representation of crystal growth and accompanying concentration changes i n an unidirectionally s o l i d i f i e d ingot. (Dotted lines indicate constantly increasing length of the solid-liquid region t i l l the ingot top i s reached). (b) Liquid mass distribution curves in the model ingot for representative times during s o l i d i f i c a t i o n 6 . 20. 10 4 9 4 9 2 90 88 86 Chill face Inqot top Theory — o - - Experiment i i I I I Fig. 11 0 0 2 04 06 0-8 10 12 14 16 18 20 DISTANCE , in Theoretical and experimental segregation curves for an Al-10% Cu ingot. C h i l l foe* Inqot top 07 04 0-6 0-8 10 I i 14 16 IB D I S T A N C E , In Fig. 12 Theoretical and experimental segregation curves for an Al-15% Cu ingot. 9 21. between theory and experiment. This excellent agreement in a l l of the results reported between theory and experiment (with the exception of the Al-Si system) i s , on closer examination, somewhat surprising. The extensive results of inverse segregation measurements summar--zed by Pell-Walpole1 where no attempt was made to compare experiments with theory, shows a very wide scatter of results. The theory of inverse segregation does not take into account the cast structure, the thermal conditions during solidification, exudations on the surface, dissolved gas and porosity, superheat, nucleation, or other variables in the casting process. In add-ition, the accuracy of the measured average concentrations seem remarkably good considering that a l l the castings have micro-segregation, and sampling to give highly reproducible results is difficult. In Figure 12, the difference between the average composition of the casting and the composition remote from the ch i l l face is 0.2%, in an alloy of 14.75% Cu composition. This difference involves establishing the average composition of the casting, which is .difficult, as well as locally measuring the composition accurately. The values of the solute composition at the c h i l l face are based on one point near the face, which is suspect due to the presence' of the exudations on the c h i l l face, or on extrapola-tion of two 'or three points showing scatter. Some of the assumptions of the model, in particular, that there is complete mixing in the liquid perpendicular to the freezing direction, and no mixing parallel to the freezing dir-ection, seems to be only a fir s t approximation. Flemings et a l . 3 ' 4 ' 5 have adopted the volume element analysis to calculate the macrosegregation in an entire casting, with or without c h i l l faces being present, and with solidification pro-ceeding in three dimensions. The assumptions made are essentially the same as those used by Scheil. The general solutions as derived by Flemings et al . , are stated to reduce to those derived by Scheil for inverse segregation at the c h i l l face, using suitable boundary conditions related to casting directionally against a c h i l l face. The important contribution of the general treatment of Flemings et al. is the generality of their solutions which con-sider normal segregation, density effects, and inverse segrega-tion concurrently. They give numerical examples of their analysis for the Al-4.5% Cu alloy and compare their calculations with their experimental measurements for unidirectionally solidified castings 5' 2 3 , in general, obtaining exce'llent agreement. The maximum segregation at the c h i l l face was calculated from the general solutions, as a part of this investigation, to test the applicability of the general theory to inverse segre-gation in Al-Cu alloys. The calculated segregation obtained is Fig. 13 Comparison of the theoretical predictions of the maximum c h i l l face segregation in Al-Cu alloys. 24. shown in Figure 13 and compared to the values obtained from the Scheil analysis. It is seen that the two calculations give markedly different results, the Flemings analysis giving much larger values than the Scheil analysis. The basic reason for this difference is that simplifying assumptions are introduced in the Flemings analysis in order to reduce the complex math-ematical expressions to simpler analytical solutions in order for the calculations to be made. Specifically, a constant partition ratio (k = 0.172) is used as well as constant values of the solid and liquid densities, yielding a constant value for the solidification shrinkage (3 = 0.055). The assumption of constant liquid density is not justified in the Al-Cu system since i t is this variation that is the major factor leading to the observed inverse segregation. It can be observed from the Figure that the two theories predict maximum.chill face segrega-tion at different compositions, i.e. , the maxima of the two curves correspond to different compositions. 1.2 Exudations Exudations on the surface of the ingot are not considered part of the theoretical model of inverse segregation. However, the presence of exudations is generally observed on the c h i l l face and is very important in the casting industry. This is because exudations are normally eutectic material, and are often hard, brittle, and abrasive. As a result, the surface layer of ingots with exudations often has to be removed before further fabrication can be carried out. The formation of exudations is normally due to the partially solidified ingot drawing away from the mold wall due to solidification shrinkage, which causes a decrease in the rate of heat removal from the surface of the ingot. As a result, the solid shell of the ingot may reheat, due to the superheat normally present in the liquid portions of the casting. This in turn causes the flow of en-riched residual liquid through the partially remelted shell to the surface of the casting through interdendritic pipes. In addition to the formations of exudations on the surface, there may be a depletion of solute in the regions next to the c h i l l face. Studies of this phenomenon have been carried out and models proposed for the quantitative prediction of this effect 1 4 1.3 Aim of the present work This investigation was undertaken to experimentally' investi gate the amount and extent of inverse segregation at the 'chill face and in the c h i l l zone. By using radioactive tracer tech-niques , i t was hoped to obtain a sufficient number of points in the region df the c h i l l face to be able to extrapolate the values with, confidence and compare measured c h i l l face values with extrapolated values. The effect of a l l the major casting variables would be con-sidered experimentally to determine their effect on inverse segregation. This would include the cast structure, the rate of heat removal from the c h i l l face, the superheat, the tempera-ture gradient in the freezing direction in the liquid during solidification, the gas content of the liquid metal, as well as the alloy composition. Both c h i l l face and the c h i l l zone segregation would be examined. Initial experiments were carried out on the Al-Ag system using Ag 1 1 0 as the tracer. No significant inverse segregation was detected, contrary to what had been anticipated. However, a quantitative estimate of the c h i l l face segregation could not be predicted from theory for this system since values of the specific volumes of the solid and the liquid as a function of composition and temperature could not be obtained from the literature. Accordingly, i t was decided to revert to the Al-Cu system on which the bulk of the previous observations reported in literature had been made. Specific volume data is available for this system; this system is one of the best systems for the examination of inverse segregation; and preliminary measurements demonstrated that Cu 6 4 could be successfully by used as tracer material, even though i t has a short half l i f e (12.8 hours). The short half l i f e of the copper permitted normal chemical analysis to be carried out on the same samples that were used to.determine the activity variations along the ingot length. This was done by allowing the radiation in the samples to decay to a negligible level prior to analysis. 27. 2. EXPERIMENTAL PROCEDURE 2.1 Materials Aluminium: Superpure Aluntinium (99.995%) obtained from Alcan was used in the preparation of a l l the ingots. Copper: Copper used in the preparation of the Aluminium-Copper ingots was of 99.999% purity and was obtained from the American Smelting and Refining Co., New Jersey. Silver: High purity silver, obtained from Cominco Ltd., was used in the Aluminium-Silver ingots. Radioactive Copper: Cu 6 4 (half'life 12.8 hours, g+y). Specimens were irradiated in the form of sheet, 0.025 cm. thick, weighing 0.5 grams each. Irradiation was carried out by the International Nuclear and Chemical Corporation, Buffalo, New York. Radioactive Silver: Ag 1 1 0 (half l i f e 253 days, 8+y). Silver wire of 0.1 cm. diameter was irradiated by International Nuclear and Chemical Corporation, Buffalo. Specimens were cut from the wire when required. 2. 2 Alloy preparation The alloys were melted in a graphite crucible (under an argon atmosphere), by induction, using an Inductotherm power supply. Weighed amounts of the pure constituents were added to the crucible to give an alloy of the desired composition. Melting occurred in about five min-utes, using a power input of 10 Kw. After melting, the cover of the crucible was removed and the melt was vigorously stirred using a 28. graphite plunger with a perforated graphite disc at its end, the diam-eter of the disc being slightly smaller than the inside diameter of the crucible. Prior to casting, the melt temperature was measured with a bare chromel-alumel thermocouple and recorded on a Honeywell chart recorder. The pouring temperature was determined by extra-polating the falling temperature of the melt, just prior to pouring, to the start of pouring. Specific superheats at pouring were obtained by adjusting the time of pouring to the melt temperature and the cooling rate prior to pouring. In general, the melt size was three hundred grams. For the Al-Cu alloys, approximately 0.5 grams of Cu 6 4 having an activity of about 8 mC at the time of the test was added to each melt, giving a concen-tration of radioactive copper of about 1600 ppm. After the addition, the liquid was vigorously stirred, a sample was removed for the measure-ment of the melt composition, and the alloy cast as soon as the desired pouring temperature was attained. A number of procedures were tried to obtain suitable samples for the measurement of the melt composition. In i n i t i a l tests, liquid was drawn up into 0.32 cm l.D. Vycor tubing. This method was found to be unsatisfactory, since reproducible results were not obtained, as des-cribed in section 3.3. Subsequently, the method adopted was to scoop some of the well mixed melt into a graphite cup and pour the liquid into a copper c h i l l mold, giving cylindrical samples 2 cm. in diam-eter and 2.5 cm. long. These samples proved to be satisfactory for subsequent sectioning and counting, and gave reproducible values for the melt composition. 2.3 Casting Two different casting procedures were used. In one case, (a), the melt was heated i n a furnace and s o l i d i f i e d i n situ by water cooling a c h i l l adjacent to the melt. In the second case, (b), the melt was poured into a preheated fibrefrax mold against a water cooled c h i l l . Details of these procedures are given below. The furnace used i n procedure (a) i s shown in Figure 14. The liquid metal was cast into a graphite sleeve with either a copper or stainless steel c h i l l in the bottom of the sleeve. The graphite sleeve and the c h i l l were surrounded by a graphite core, heated by two 750 watt elements, the graphite core serving to provide a uniform temperature i n the centre of the furnace prior to casting. A chromel-alumel thermocouple inserted in the core was used to monitor the tem-perature of the furnace. With power supplied independently to each half of the fumace through two variacs, the temperature of the furnace could be brought up to 700°C in about 0.5 hours. The copper and stainless steel c h i l l s were used in order to obtain different cooling rates. The surface of the c h i l l was polished and then coated with Aquadag (a colloidal suspension o f graphite in a volatile medium), to prevent the liquid metal from alloying with the top surface of the c h i l l on casting. A stream of argon was passed through the furnace prior to casting to reduce oxidation during cast-ing, and also to prevent the oxidation of the Aquadag coating on the c h i l l . This gas stream, by displacing the a i r i n the furnace, also served to minimize the burning away of the graphite sleeve. insulated heating -elements graphite sleeve furnace support 13 cm 30. water inlet 1 /l|*3-2cm-> v. •thermocouple graphite -core 30 cm. insulating "ring 6 mm I.D. J water outlet 25cm I.D. Fig. 14 Schematic representation of the casting arrangement for gasting procedure (a). 31. 32 mm. l.D fibrefrax tube 35 cm. water inlet 6mm. LD. / / / / / / / / / / / / / / L A \ \ A i 4 46 mm. l.D •**Vycor tube LcWH s s V J3 cm. water outlet Fig. 15 Schematic representation procedure (b). of the casting arrangement for casting To cast, the liquid melt was poured into the preheated furnace and the temperature of the liquid monitored. When the required super-heat was reached, the power to the furnace was turned off and the water flow to the c h i l l was turned on, causing the liquid metal to solidify progressively from the c h i l l . In the second casting procedure, as illustrated in Figure 15, a preheated fibrefrax tube was inserted into a 46 mm. I.D. Vycor tube, the metal solidifying progressively from the water cooled metal c h i l l . The tube was inclined slightly to the vertical to obtain an even flow of liquid onto the c h i l l during the i n i t i a l stages of pouring. Pour-ing into a vertical tube with a cold c h i l l was found to cause local, irregular i n i t i a l solidification. The fibrefrax tubing was normally preheated to between 800°C and 900°C in a resistance furnace. The water flow to the c h i l l was turned on just prior to casting, to elim-inate condensation on the ch i l l . Both copper and stainless steel chills, coated with Aquadag, were used, as in'procedure (a). The ingots obtained were about 3 cm. in diameter and about 12.5 cm. in length. Shrinkage pipes were absent in a l l the ingots. 2.4 Metallography and Autoradiography The macrostructure of the castings was determined by sectioning the casting longitudinally, mechanically polishing the exposed sur-faces, and then etching in a modified Kellers reagent consisting of HF,HN03, HCl and water in the proportion 2:3:5:25, followed immediately by immersion in concentrated HN03 to remove 'the layer formed on the 33. surface due t o the etching. Ingots containing C u 6 4 were sectioned and autoradiographed t o determine whether the radioactive material was d i s t r i b u t e d reasonably uniformly throughout the ingot, i . e . , t o e s t a b l i s h that the l i q u i d was w e l l mixed before casting. A f t e r the material adjacent t o the c h i l l face was removed f o r a n a l y s i s , transverse and l o n g i t u d i n a l sections were cut from the ingot. These were mechanically polished, and then placed on Kodak Contrast Process Ortho f i l m , backed by a glass p l a t e . The exposure time required f o r the autoradiographs v a r i e d from 15 to. 50 hours depending on the a c t i v i t y o f the t r a c e r i n the section. The darkening o f any region of the f i l m was taken as d i r e c t l y r e l a t e d t o the concentration of C u 6 4 adjacent to the f i l m i n that region. Transverse sections, 1 cm. from the c h i l l face, were taken from the ingots used f o r the temperature measurements (section 2.7), t o obtain the average dendrite spacings. These samples, a f t e r being mechanically polished f o r metallographic examination, were etched i n the K e l l e r s reagent, d i l u t e d with water i n the proportion 1:9. The washing of the sample with concentrated HNO3 was dispensed with. Den-d r i t e spacings were measured from enlarged photographs of the polished and etched sections. 2.5 Measurement of the v a r i a t i o n of the solute concentration 2.5.1 Radioactive Tracer method The cast ingot was placed i n a lathe and layers 5O0j*thick p a r a l l e l to the c h i l l face were progressively machined from the ingot, the cuttings from each layer being caught in a box enclosing the cutting tool and the ingot. Before these layers were cut, the outside sur-face of the ingot in the longitudinal direction was removed, so that the layers in contact with the mold would not affect the analysis. Cuttings from each layer, to make approximately one gram of material, were packed in 3 dram stoppered vials. The weight of the cuttings was determined from the measured weight of each vial before and after the cuttings were introduced into them. The activity of the cuttings in each vial was measured with a Picker Nuclear Twinscaler II automatic scintillation counter. One hundred samples at a time were loaded into the automatic changer device of the instrument and the activity of each sample was counted in turn over a 0.2 minute interval by two detectors simultaneously.-The output of the scalers was automatically printed on a teletype, and a punched tape output was also generated on a papertape punch machine coupled to the teletype. The printout information included the sample number (as loaded on the machine), the counts per sample for each scaler, and the counting time. The average background count, as determined by manually measuring the background level a number of times at the start of counting, was set on the machine and was auto-matically subtracted from each reading. The symmetrical design of the sample tray ensured a constant background level at the detectors, independent of the position of the samples. 35. The counter1 was calibrated with a Cesium 1 3 7 standard source prior to counting. In counting the samples containing Cu 6 4, the energy range of the gamma activity detected by the scalers was adjusted to be between 400 KeV and 1 MeV, since the activity peak for Cu6l+ is about 500 KeV. (Figure 16). For Ag 1 1 0 the energy range over which measure-ments were made was 50 KeV to 1 MeV. All the samples from the ingots were counted at least three times each, to ensure that counting or machine errors would be detected. In addition, by having repeat counts made over a two day period, satur-ation effects associated with high counting rates could be detected. For the samples cut from the c h i l l cast buttons (cast to obtain the i n i t i a l liquid composition), counts were made at least ten times over at least four half lives, and decay curves were drawn from this infor-mation. The half l i f e obtained from these curves, when compared with the published value for Cu6lf, further established the stability and reliability of.the counting system. The information on the punched tapes from the counters was read and stored on magnetic tape at the U.B.C. Computing Centre.. The weight of each sample was recorded on Fortran Coding sheets, and added to the tape information. A correction was made to the activity measure-ment for each 'sample to allow for decay from the time of the start of-counting.for each .set of 100 samples. Since the amount of data handled per set of experiments was quite large (about. 2 ,000 readings), the data was processed with the IBM 360/67 computer at the U.B.C. Computing Centre. .After applying,the corrections for the decay of each specimen, the•activities were normalized for the weight of the sample, these in turn were normalized by dividing by the melt composition determined from the activity of the cast buttons and the results were obtained both in printed form, as well as in plotted form, as a function of the distance from the c h i l l . The plots were produced on the Calcomp Plotter at the U.B.C. Com-puting Centre. To check on the performance of the Scintillation Counting equipment, a large number of observations (350) were made on a long lived isotope ( T i n 1 1 3 ) , and the data put through a Goodness of Fit Test. No significant difference was detected between the expected and the observed distributions of the data, i.e. , the data corres-ponded to the "normal" distribution. 2.5.2 Chemical Analysis To establish that the variations in composition detected by activity measurements do correspond to variations in the solute con-centrations , chemical analyses were carried out on some of the samples used for the activity measurements. The method employed was the electrodeposition of copper from a solution of the alloy sample onto a platinum cathode. This was done by first dissolving the cuttings in a mixture of H 2 S O 4 , H N O 3 and water (in the proportion 2:1:2), heating to dryness, adding an excess of 1:1 H2SCv, and then redissolving the paste in boiling water. The platinum cathode was placed in the solution so obtained and a current 38. of approximately 1 amp. was passed through the solution for about 1.5 hours, by which time effectively a l l the copper was removed from solution. The amount of copper so deposited on the cathode was then determined from the increase of weight of the cathode, and the alloy composition then determined from this weight and the i n i t i a l weight of the alloy cuttings used in making up the solution. 2.6 Microprobe Analysis In an attempt to measure directly the variations in the solute concentrations as a function of the distance from the c h i l l face, scans of ingot sections were made on a JEOL Model JXA-3A Electron Probe Microanalyzer. It was found that the short range variations in the composition due to microsegregation completely masked any macro-segregation present. Continuous scans were also made with a large spot size of 50 microns to follow the variations qualitatively, but this did not yield acceptable results either. 2.7 Temperature measurements Five thermocouples were positioned along the ingot length at dist-ances of 0., 1.2,. 2.5, 5 and 11.5 cm., respectively, from the c h i l l face. The chromel-alumel thermocouples were made of 20 gauge wire, sheathed in glass wool. The sheathed thermocouples were coated with a layer of Sairset alumina cement, and then thoroughly dried, in order to prevent electrical contact between the thermocouples and the cast ingot during solidification. The thermocouples were connected through an ice-water reference cold junction to a Hewlett Packard Digital Voltmeter (Model 5055 A), and the emf. generated by each thermocouple recorded on the printout machine connected to the voltmeter. A reading was recorded for each thermocouple every 1.25 seconds. Other measurements were made with a single pen chart recorder, switching from one thermocouple to another with a multipoint switch, at varying time intervals. However, this procedure was not found to be satis-factory primarily due to the large time intervals between successive readings on any one thermocouple. The measured values of the emf., taken at known intervals of time from the start of pouring, enabled temperature profiles to be established as a function of time, at five positions along the ingot axis, corresponding to the five thermocouple positions. 2. 8 Modification of Casting procedures 2.8.1 Degassing Degassing was carried out in some of the ingots by adding Foseco 190 Degaser to the melt prior to casting. Following the instructions enclosed with the degaser, the degaser was introduced 'into a bell-shaped perforated graphite plunger, which in turn was introduced into the melt and agitated. The degassing was carried out when the melt was cooling to the desired pouring temperature. However, the degaser had an unpredictable effect on the structure of the cast ingot, yield-ing both equiaxed and columnar structures at different times. 40. Efforts were also made to reduce gas dissolution in the cast alloys by melting and casting in a Vacuum Induction furnace- This , procedure was found to be unsatisfactory because the casting process tended to be cumbersome, and because the rate of heat removal from the bottom of the mold was too slow. In addition, the ingots pro-duced contained large interior cracks. 2.8.2 Grain Refining Fine grained equiaxed structures were obtained by adding about 0.5 grams of 5% TiB2 in aluminium Master alloy to 300 gram melts just prior to casting. 2.8.3 Exudations To examine the effect of a change in thermal contact between the casting and the c h i l l during solidification, some castings were moved away from the c h i l l to a distance of about 5 mm., 15 seconds after pouring. This caused a reheating of the solid portion of the casting, and consequently, formation of exudations on the ingot surface in i t i a l l y adjacent to the ch i l l . 2.8.4 Thermal conductivity of the chills Two types of chills were used, one copper (thermal conductivity .9 cal./sec. sq.cm. C at 100 C) and one of stainless steel (thermal conductivity .04 cal./sec. sq.cm. °C at 100°C). In both cases, the polished top surfaces were coated with a thin layer of colloidal graphite, which did not affect the- thermal characteristics of the ch i l l . A large flow of tap water was passed through the c h i l l , the temperature and volume of water being considered as effectively constant through-out a l l the tests. 3. OBSERVATIONS 3.1 Tabulation of Experiments An appreciable number of ingots were cast under a variety of casting conditions. Five different alloy compositions were investigated, and a series of factors associated with the casting conditions were considered. This included superheat, thermal con-ductivity of the c h i l l , cooling rate, cast structure, gas content, separation of the cast metal from the c h i l l during solidification, and others. The castings to which Cu 6 4 was added to the melt, are listed in Tables I ,II,III, along with the casting conditions. Individual castings will be referred to by the heat number listed. Most of the castings were of the Al-10% Cu alloy, since this alloy should exhibit the maximum inverse segregation at the c h i l l face, as re-ported in previous investigations 8. The values of the superheat listed in the Tables were determined from the difference between the pouring temperature and the equilibrium liquidus temperature for that particular alloy. 3.2 Homogeneity of the melt In order to determine whether the castings containing the radioactive tracer were macroscopically homogeneous, i.e., whether the tracer was thoroughly mixed in the melt, autoradiographs were TABLE I DATA AND RESULTS FOR Al-Cu INGOTS COMPOSITION CASTING NUMBER CHILL HOLD SUPERHEAT °c • MELT ACTIVITY SAMPLING SPECIAL TREATMENT INGOT STRUCTURE NORMALIZED INGOT COMPOSITION Co XTRAPOLATED CHILL FACE COMPOSITION % Cu CHILL ZONE L (x500) c JX FIGURE # COMMENTS C- COLUMNAR E= EQUIAXED C+E= MOSTLY COLUMNAR j Al-5% Cu 130 Copper Fumaoe 10 °c C h i l l K 1.15 5.26 18 93 CoDper Fibrefrax 39°C C h i l l - C+E 1.00 5.28 9 •34 Copper Fibre f rax 5M°C C h i l l Exudations C+E 1.28 15.6 22 •30 Stainless Fibrafrax 107°C C h i l l - r 1.03 32 Al-20% Co 71 72 Stainless Stainless Fumaoe Furnace 180°C 180°C - -C Excessive scatter due to snail sample sizes 82 Stainless Fibrefrax 103°C C h i l l - C .99 - - 34 126 Copper Fibrefrax 30°C C h i l l — C+E 1.19 20.32 12 Al-25% Cu 83 Stainless Fibrefrax 103°C C h i l l _ C 1.01 35 127 Copper Fibrefrax V 30°C. C h i l l - C+E 1.22 - -TABLE II DATA AND RESULTS FOR Al-Cu INGOTS Al-10% Cu C A S T I N G ' ' WJMBi'R . CHIL  MOLD SUPERHEAT °C KELT ACTIVITY SAMPLING SPECIAL TREATMENT INGOT STRUCTURE NORMALIZED INGOT COMPOSITION Co EXTRAPOLATED CHILL FACE COMPOSITION * Cu CHILL ZONE L <xbO0> c ^ FIGURE » COMMENTS C= COLUMNAR E= EQUAXED C+E= MOSTLY COLUMNAR •12 i Copper Furnace 10°C Chill - E 1.04 10.2 5 18.39, 40 9G Copper Furnace 70°C Chill - C 1.04 10.42 8 7b 7G Stair.les; Stainless Furnace Furnace 160°C IBO^C - - C+E C+E - - : Scatter too high due to snail sample sizes. 93 85 Copper Copper Fibrefrax Fibrefrax 94°C 91°C Chill Chill slow water flow exudations C C+E .87 1.24 10. Ml 25.8 19 23 44 Copper Fibrefrax ' 93°C Tube - C+E 0.99 - - 28 •97 Ccpper Fibrefrax 84°C Chill structure changed C+E 1.00 - 36 45 Copper Fibrefrax 65°C Tube - C+E 1.14 10.05 15 29 179 Copper Fibrefrax 50°C Chill - C+E 1.12 10.16 15 95 Copper Fibrefrax 41°C degassed C+E 1.00 10.5 9 21 1?5 Copper Fibrefrax 20°C Chill - C+E 1.21 10.15 19 46 Stainles: Fibrefrax 91°C Tube - C 1.04 10.2 8 30 84 Stair.les; Fibrefrax 91°C Chill TiB 2 E 1.01 10.27 E 37 47 Stainless Fibrefrax 67°C Tube C+E 1.06 10.37 18 31 TABLE III DATA AMD RESULTS FOR Al-Cu INGOTS Al-15% Cu CASTING NUMBLK CHIli, MOLT1 SUPERHEAT °C MELT ACTIVITY SAMPLING SPECIAL TREATMENT t INGOT TRUCTURE NORMALIZED INGOT COMPOSITION Co EXTRAPOLATED CHILL FACE COMPOSITION *Cu CHILL ZONE Lc(x5jjg> FIGURE # COMMENTS C= COLUMNAR £= E0UIAXED C+E= MOSTLY COLUMNAR 21 Cooper Furnace 160 °C Tube C 0.74 15.05 6 27 •o Copper Furnace 115 °C Tube snail ingot C 1.31 - - • 26 18 Copper Furnace 80 °C Tube - C+E .96 15.42 12 25 19 Copper Furnace 40 °C Tube - C+E .96 15.42 7 24 124 Copper Furnace 10 °C C h i l l - E 1.28 15.41 20 73 74 Stainless Stain les : Furnace ; Furnace 1S0°C 150 °C - - C+E C+E - - Excessive scatter due to sna i l sairple sizes. 128 Copper Fibrefrax 20 °C C h i l l T iB 2 E 1.23 15.37 ' 1C 81 Copper Fibrefrax 100 °c C h i l l - C 1.00 15.52 10 33 46. Fig. 17 Autoradiograph and corresponding etched surface of an Al-Cu ing< (ingot 94). 47. Fig. 18 Autoradiograph and corresponding etched surface of an Al-Cu ingot (ingot 84). made of the parts of the ingot remaining after the cuttings re-quired to measure the composition variation were removed from the ingot. The autoradiographs established that the liquid was mixed sufficiently to uniformly disperse the tracer addition, and also gave an indication of the structure of the casting. Examples of two such autoradiographs and the corresponding etched surface of columnar and equiaxed ingots are shown in Figures 17 and 18. The Figures show that there is no significant macro-segregation of the tracer in the ingot, and, that the cast struct-ure on the etched surface is delineated in the autoradiograph. Since the lower part of the casting, adjacent to the c h i l l was not autoradiographed, inhomogeneities may be considered to be present in this region. However, in view of the uniform distributions shown in the autoradiographs, the probability of such an occurrence is considered to be small. The relative position of the auto-, radiographs with respect to the whole ingot is outlined in the Figures. 3.3 Cast structure Most of the castings that were used for the determination of macrosegregation were columnar, for at least half of the overall ingot length, starting at the c h i l l face. In view of the fact that only the first 2 cms. of the ingot was significant for the present analysis, no attempt was made to make the whole of the ingot columnar. It was noticed that the use of higher superheats yielded ingots with longer columnar zones. The castings in which the grain refiners were added, had a fine-grained equiaxed crystal structure uniformly throughout the casting. The grain size of one such casting, of Al-10% Cu alloy was about 1 mm. diameter. It was generally observed that exudations or sweat beads were not present on the c h i l l face in nearly a l l of the ingots examined. It will be shown later that exudations are present on the c h i l l face of the ingot, as deduced from composition analysis, but that the thickness of the exuded layer must be very small. In general, the c h i l l face of the ingots was smooth and flat. Microporosity was observed to be present in most of the ingots examined, as detected optically. The microporosity is believed to be due to the presence of hydrogen in the molten aluminium. The hydrogen was introduced into the melt during the relatively short period of time that the melt was exposed to air during casting. In general, ingots cast in the preheated furnace using tech-nique (a), (Section 2.3), showed a longer columnar zone than the ones cast in the fibrefrax tubing, employing procedure (b). The columnar grains had a smaller diameter when the ingots were cast in the fibrefrax tube. Macrostructures of three ingots, one cast using casting technique (a), one using (b), and the third cast in the vacuum induction furnace are shown in Figure 19. 50. Fig. 19 Macrostructure perpendicular to the c h i l l face of Al-Cu ingots cast (a) in the furnace arrangement (b) in fibrefrax molds (c) in the vacuum induction furnace. 51. 3. 4 Measurement of the i n i t i a l melt composition Tests were carried out to determine the change of the mean composition of the melt with time, due to the loss of either the aluminium, or copper from the liquid alloy into the oxide layer covering the melt. Cu6i+ tracer was added to the melt and was vigorously stirred in. Samples were taken at 5 minute intervals from the melt, over a duration of 20 minutes, and cast into copper molds. For the duration of the test, the temperature of the molten metal in the crucible was kept at 700°C, by adjusting the power to the induction coil. The c h i l l cast samples were then machined into small cuttings, which were packed into vials, and the activ-ity of these samples determined on the scintillation counter. It was found that there was a small loss of Al from the melt with time, but that this loss was too small to make any significant difference to the measurement of the melt composition. The rate of increase of the copper composition in the melt is shown in Figure 20. The activities have been normalized with respect to the average activity of 10 samples obtained from the ingot cast from the remaining melt. Since the normal time between the re-moval of the sample from the melt, and casting was usually less than 2 minutes, i t was considered that this loss was insignificant in the present investigation. For each ingot, direct measurements of the activity and there-fore Cu6I+ concentrations in the melt were made prior to casting. The sampling procedure used in each test (tube or c h i l l casting) _ | J I I 1 Q 1 » L 2 4 6 8 O 12 14 16 18 TIME (min.) Fig. 20 Normalized activity of samples taken at different times from the same melt. 53. is listed in Tables I, II, and III. The melt activity measured was used to normalize the activity measurements of cuttings taken to establish the segregation in the casting. If segregation is confined to the c h i l l zone of the casting, then the normalized concentration of Cu 6 4 away from the segregated zone should be unity. These values (normalized ingot composition C q ) are listed in Tables I, II, and III. An examination of the values obtained from the c h i l l casting technique shows that 90% of the samples tested had a value of C q obtained by normalizing,, to within 5% of unity. This does not include samples 123 to 130 which had high activities giving slightly erroneous results due to saturation effects during counting. For the melt sampling using a quartz tube, approximately 25% of the values of C q were within 5% of unity, 50% within a range of 5 to 15% of unity, and the rest greater than 15% from unity. These results indicate that the tube sampling technique is not as satisfactory as the c h i l l casting technique to determine the melt activity. The accuracy and reproducibility of C q indicates re-producible measurements of concentration in the cuttings taken from ingots away from the c h i l l zone. In the case of ingot 129, samples for the mean composition of the ingot were obtained using the method of Kirkaldy and Youdelis. The ingot was split longitudinally, and samples were machined from one half of the ingot for the determination of the segregation of copper with increasing distance from the c h i l l face. The other half of the ingot was clamped on a vise and the cut surface milled over the entire length of the ingot, the cuttings were thoroughly mixed, and their activity measured to determine the average ingot com-position. It was found that the activities measured from standard samples taken from the cuttings did not give consistent results. This method for obtaining C was therefore not considered satis-& o factory in the present tests. 3. 5 Distribution of Cu6"4 along the ingot length 3. 5.1 Observation of macrosegregabion near the c h i l l Figure 21 illustrates the main features found in the solute distribution near the c h i l l face of the ingots. (a) The first point, adjacent to the c h i l l , has an abnorm-ally high composition in comparison to the adjacent points and the remainder of the curve. (b) There is a gradual decrease in composition with dist-ance from the c h i l l face. (c) After the gradual decrease, the composition remains constant over the rest of the ingot examined. .This compo-sition is taken as the mean composition of the ingot. (d) In some cases, a small solute depleted zone, just before .• the increase in composition (b) was observed. The concentration indicated by the first point (feature (a)) in Figure" 21, is 10.7% Cu, which is much lower than the eutectic i n MEAN COMPOSITION RL-1QJ CU CAST NUMBER 95 MOLD : SUPERHEAT (°CJ 41 CHILL DEGRSED FIBREFRAX COPPER >-2 C J CC UJ—1 CO CL DC — i n cn 1 0 . 0 ~~1 1 1 1 1 5 . 0 1 0 . 0 1 5 . 0 2 0 . 0 2 5 . 0 DISTANCE FROM CHILL FACE I X .05 CM 3 3 0 . 0 Fig. 2 1 Normalized activity vs. distance from the c h i l l face for the composition and casting conditions indicated. composition of 33% Cu for Al-Cu alloys. In other ingots of Al-10% Cu, the i n i t i a l composition measured varied in the range 10.0 to 11.5% Cu. Such a high concentration could only result from the formation of a thin layer of eutectic exudation on the ch i l l face. The thickness of the eutectic to produce a compo-sition of 11.5% Cu in the first point was calculated to be 3 5 microns which would not normally be observed visually. As a result of the presence of the eutectic layer, the first point was neglected in the analysis of the composition distribution. The c h i l l face concentrations reported here were obtained by extrapolating the composition distribution curves back to the ch i l l face, provided that at least three points were present in the decreasing concentration region (feature (b)). The composition given by the horizontal portion of the normalized activity distributions (normalized with respect to the melt sample activity), was taken as the mean composition of the ingot in a l l cases. A l l the measured compositions were then normalized so as to set the horizontal portion to the value 1. In many of the following activity distributions, multiple points have been represented for the activity of the same sample. These indicate repeat measurements taken on the same sample at different times. The activities reported in Figures not bearing such multiple points are averages of such repeat measurements. Al l activities reported are corrected for decay. 2 s * * » * MEAN COMPOSITION AL- 5; CD CAST NUMBER 94 MOLD : FIBREFRAX SUPERHEAT C O 54 CHILL : COPPER EXUDATIONS — i i 1 r 1 1 1 5-0 10.0 15,0 20.0 25.0 30 0 35 0 DISTANCE FROM CHILL FACE ( X .05 CM J Normalized activity vs. distance from the c h i l l face for the composition and casting conditions indicated. MEAN COMPOSITION RL-IO; cu CD -4- "~~ CAST NUMBER 85 MOLD : FIBREFRAX SUPERHEAT (°C) 91 CHILL : EXUDATIONS COPPER St rn ~ X >-ea t— • -X • X 5 >, 4 > RCT. .ISED 2.2 .ISED 2.2 X CL z; Q I C * * « I * * * * 0 f .0 5.0 DISTANCE i 10.0 FROM i f r 15.0 20.0 25.0 CHILL FACE ( X .05 CM ) i i 30.0 35.0 Fig. 23 Normalized activity vs. distance from the c h i l l face for the composition and casting conditions indicated. 3.5. .2 Exudations The distribution of Cu5it near the c h i l l resulting from the removal of the ingot from the c h i l l face 15 seconds after casting is shown in Figures 22 and 23. A marked increase in the concen-tration of CuSi+ at the c h i l l face is observed in both curves, as compared with Figure 21, with the concentration adjacent to the ch i l l approaching the eutectic composition in Figure 23. The effect of removing the ingot from the c h i l l has clearly caused local remelting to occur and the liquid near the eutectic compo-sition to be exuded past the i n i t i a l cast surface through the remelted interdendritic liquid pipes. Since the amount of eutectic liquid present in the Al-5% Cu alloy due to non-equilibrium solidification is small as compared to the Al-10% Cu alloy, the composition of the exuded material adjacent to the c h i l l does not reach the eutectic composition. Comparing Figures 22 and 23, the distance through which there is a rise in composition differs, the distance in Figure 2 3 being appreciably larger. This is attributed to either a larger separation between the c h i l l and the ingot, or more local remelt-ing and subsequent flow of liquid to form exudations in Figure 23. The value of the horizontal portion of the curve in both cases is high (1.25). This is probably due to the flow of the enriched interdendritic fluid towards the c h i l l face of the casting during the formation of the exudations. 60. 3.5. 3 Effect of superheat The composition distributions obtained for ingots of Al-15% Cu, cast against a copper c h i l l in the preheated furnace chamber (procedure (a)), is shown in Figures 24 to 27. Here, the variable being changed is the melt superheat, which varies from 40°C for Figure 24 to 160°C for Figure 27. A l l four Figures show an i n i t i a l high concentration at the ch i l l face, with the concentration decreasing away from the c h i l l , tending towards a constant value. There is a wide scatter in the points, particularly in Figures 24 and 25, making an extra-polation of the concentration values near the c h i l l , back to the chi l l face, ambiguous. Comparing the compositions adjacent to the c h i l l , the results show a decrease in the concentrations with increasing superheat. They also show the region adjacent to the c h i l l exhibiting higher solute concentrations is larger for low superheats. A quantitative comparison of maximum c h i l l face composition and superheat can be made using the first point adjacent to the c h i l l , or by extra-polating the points in the region adjacent to the c h i l l , back to the c h i l l face. Both procedures should give the same results. However, using the first point gives much higher values for con-centration at the c h i l l face than the values obtained from the extrapolating procedures. The large values are attributed to the exudation of eutectic material at the c h i l l face, separated slightly CM MEAN COMPOSITION CAST NUMBER 19 SUPERHEAT (°C3 4 0 fll-15; CU MOLD : FURNACE CHILL : COPPER 03 CC f M U J -CO cc 2: Qcn -J; ±_ + * + + + • ^ + • + + + cn o ~ 1 — 5.0 I 1 1 1 1 10.0 15.0 20.0 25.0 30.0 DISTANCE FROM CHILL FACE ( X .05 CM ) o . o 35.0 40.0 45.0 Fig. 24 Normalized activity v-S- distance from the c h i l l face for the composition and casting conditions indicated. CM O UJ-* CO MERN COMPOSITION CRST NUMBER 18 SUPERHERT (° CJ gO fll-15/ CU MOLD : FURNACE CHILL : COPPER 5.0 10.0 15.0 2 0 . 0 2 5 . 0 3 0 . 0 DISTANCE FROM CHILL FACE ( X .05 CM 3 3 5 . 0 4 0 . 0 4 5 . 0 CD Fig. 25 Normalized activity vs. distance from the c h i l l face for the composition and casting conditions indicated. MEAN C0MP05J TJON flL-15; CU ^_ CAST NUMBER 20 MOLD : FURNACE SUPERHEAT (°C3 115 CHILL : COPPER QO -rn CJ CC ft + UJ-* + . + C O ± -L 4 * 4 + ' + ' + ' — ' + * + ' 4 + 4 + t ^ + , , + t , l _ f CX | + + * + + + IZ Cilia CM 1 I I 1 1 1 1 1 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 4.0 DISTANCE FROM CHILL FACE ( X .05 CM ) Fig. 26 Normalized activity vs. distance from the c h i l l face for the composition and casting conditions indicated. 03 A MEAN COMPOSITION AL-152 CU CAST NUMBER 21 MOLD : FURNACE SUPERHEAT (°C)160 CHILL : COPPER »* •<& A £ ^ A tl h & a A A A A ft A A • & A A A A A A * A A 03 " L D cr L iJo ' CO CC 51 CCCD a I 1 1 1 1 1 0 . 0 1 5 . 0 2 0 . 0 2 5 . 0 3 0 . 0 DISTANCE FROM CHILL FACE ( X .05 CM ) 0 . 0 5 . 0 3 5 . 0 4 0 . 0 4 B . 0 Fig. 27 Normalized activity vs. distance from the c h i l l face for the composition and casting conditions indicated. from the c h i l l due to volume shrinkage on cooling. Accordingly, i t is considered that the first point (and possibly the second) cannot be directly related to the c h i l l face concentrations due to inverse segregation. Extrapolation of the decreasing con-centration region is only approximate due to the scatter of the points, the relatively few points in this region, and the non-linearity of the region being extrapolated. A similar decrease in the c h i l l face concentration with superheat was observed in Al-10% Cu alloys. Ingot 20 (Figure'26) was a short ingot (Table I), cast to examine the effect of ingot length on segregation in the c h i l l region. No significant difference was observed using the short ingot. However, the comparison was made at a high superheat where the segregation at the c h i l l is small. No comparisons were made at low superheats. 3.5.4 Effect of changing the rate of heat removal The rate of heat removal was changed by using either stain-less steel or copper chills, the former giving a lower rate of heat removal from the ingot during casting. Typical results for Al-10% Cu alloys cast in Fibrefrax molds, using the two chills, are shown in Figures 28 to 31. Figures 28 and 30 are curves of normalized concentration versus distance from the c h i l l for ingots cast on copper and stainless steel chills respectively, at the i n CM MEAN COMPOSITION AL-102 CU CAST NUMBER 44 MOLD : FIBREFRAX SUPERHEAT (°C) 93 CHILL : COPPER CD >-2 o cc L U — • C O CC + + -t- -t- T + + + + * + + + + ' + + + + + + + + + + + LO CD a 0 . 0 5 . 0 1 0 , 0 1 5 . 0 2 0 . 0 2 5 . 0 3 0 . 0 3 5 . 0 4 0 . 0 DISTANCE FROM CHILL FACE ( X .05 CM ) 45.0 Fig. 28 Normalized activity vs. distance from the ch i l l face for the composition and casting conditions indicated. MEAN COMPOSITION RL-IOJ CU CAST NUMBER 45 MOLD : SUPERHEAT (°C) 65 CHILL FIBREFRAX COPPER 5.0 1 1 - 1 1 r 10.0 15.0 20.0 25.0 30.0 DISTANCE FROM CHILL FACE ( X .05 CM )" 35.0 40.0 46.0 Fig. 29 Normalized activity vs. distance from the ch i l l face for the composition and casting conditions indicated. MERN COMPOSITION fiL-lC* CU CR?T NUMBER 4 6 MOLD : FISREPRRX S'JPERHERT (°C) 91 CHILL : S. S T E E L -\— ° DISTANCE C H I L t W F f\° .05 zfi 3 5 0 * ed activity vs .distance from the c h i l l face for the composition and casting conditions indicated. in t'M MEAN COMPOSITION CAST NUMBER . SUPERHEAT l°C) R L - I O ; CU 47 MOLD t 67 CHILL FIBREFRAX S. STEEL a. i 10.0 15.0 DISTANCE FROM 20.0 CHILL FACE 25.0 { X 30.0 05 CM ) 25.J 40.0 45.0 F i g . 31. Normalized a c t i v i t y vs. distance from the c h i l l face f o r the composition and casting c o n d i t i o n s i n d i c a t e d . same superheat (92°C). Ignoring the first point adjacent to the c h i l l , the extrapolated concentration at the ch i l l face is higher for the stainless steel c h i l l (Figure 30), and the distance of the higher concentration adjacent to the c h i l l is larger. (The fourth point on Figure 30 is abnormally low for unknown reasons and is neglected). The concentration change at the c h i l l zone at a lower super-heat (66°C) for the copper and stainless steel chills is shown in Figures 29 and 31. Lowering the superheat with a copper c h i l l increases the amount and extent of the segregation at the ch i l l zone, (as was expected from the observations in the previous section), as seen by comparing Figures 28 and 29. When the c h i l l is changed to stainless steel, there is a further corresponding increase (Figures 29 and 31). Accordingly, reducing the thermal conductivity of the c h i l l increases the amount and extent of segregation in the c h i l l region for the same alloy cast at the same superheat. 3. 5. 5 Effect of changing composition The variations of the normalized activity distributions obtained for ingots of Al-5% Cu, Al-15% Cu, Al-20% Cu and Al-25% Cu respectively, cast on stainless steel chills in fibrefrax molds, at effectively constant superheat of 100°C, are illustrated in Figures 32 to 35. For the case of Al-10% Cu alloy, Figure 30 will be used for comparison, although this alloy was cast at a 71. slightly lower superheat than the others. Considering the curves shown in the Figures mentioned above, only Figures 30 and 32 show high concentrations in the fi r s t layer adjacent to the ch i l l . These points are associated with exuda-tions, and will be neglected. At low (5%, Figure 32) and high (25% \, Figure 35) copper concentrations, no significant segre-gation is observed in .the c h i l l region. For the Al-20% Cu alloy (Figure 34), the first point indicates the possibility of segre-gation immediately adjacent to the c h i l l , but this may partly result from exudation on the surface. Both the Al-10% Cu alloy (Figure 30) and the Al-15% Cu alloy (Figure 33) exhibit appre-ciable segregation at the c h i l l face, which extends into the ingot away from the c h i l l face. A quantitative comparison of the segregation will be presented later when the present results are compared to the published observations and theoretical pre-dictions . 3.5.6 Special treatments The variations in the solute distributions along the ingot axis for alloys of Al-10% Cu, which have been subjected to special treatments as outlined in Section 2-. 8, are presented in Figures 21, 36, and 37. The effect of adding TiB 2 to the melt of Al-10% Cu alloy before casting, on the segregation in the c h i l l region is shown in Figure 37. In this case, the ingot had a fine, equiaxed in CM cn o cr r-Q ° . UJ—•' CO i n MEAN COMPOSITION CAST NUMBER SUPERHEAT l°C) 107 AL-5J CU 80 MOLD CHILL FIBREFRAX S. STEEL X x 3 S ik i * 1 * 1 TT <9 ^ 4 i-U 0 . 0 ~~i 1 1 1 1— 5.0 10.0 15.0 20.0 25.0 DISTANCE FROM CHILL FACE ( X .05 CM ) 30.0 35.0 Fig. 32 Normalized activity AS. distance from the c h i l l face for the composition and casting conditions indicated. m cn C J CC Q ° . u 1 - 4 ' CO CX Q O in aj MEAN COMPOSITION AL-15JT CU CAST NUMBER .81 MOLD : SUPERHEAT (°C) 100 CHILL FIBREFRAX S. STEEL u » g 1 * i 11 i 1 8 M a t s * 1 M l * ! 1 0.0 —1 1 1 5.0 10.0 15.0 DISTANCE FROM CHILL FACE 20.0 25.0 ( X .05 CM ) 30.0 35.0 - J CO Fig. 33 Normalized activity vs. distance from the c h i l l face for the composition and casting conditions indicated. i n cn en CJ (X • ° 4 LU—• CO C t l — 0 ° 2 i n cn 0.0 MEAN COMPOSITION A L - 2 0 * CU CAST NUMBER 82 MOLD : SUPERHEAT (°C) 103 CHILL i % t t t ' t x i * M * » « « » I * g , , » i 1 » » * s * $ t » I 1 1 ! 1 5.0 1D.0 15.0 20.0 25.0 DISTANCE FROM CHILL FACE ( X .05 CM ) 30.0 35.0 FIBREFRAX S . STEEL Fig. 3 4 Normalized activity v s .distance from the ch i l l face for the composition and casting conditions indicated. MEAN COMPOSITION AL-25* CU CAST NUMBER 83 MOLD s FIBREFRAX SUPERHEAT (°C) 103 CHILL : S. STEEL CD o cr r-• ° J LU-C O cr z: i n * * * 1 i $ i t * 18 » * * i * » « * 8 i 1 4 * A M | » i a i | | 1 1 1 1 1 1 1 0.0 5.0 10.0 15.0 2 0 . 0 2 5 . 0 3 0 . 0 3 5 . 0 DISTANCE FROM CHILL FACE ( X .05 CM ) C n Fig. 35 • Normalized activity vs .distance from the c h i l l face for the composition and casting conditions indicated. grain size. Comparison of Figure 37 with Figure 30, the equiv-alent melt without the addition of the grain refiner, shows that the two curves are essentially equivalent, with the exception of the point for the section at .25 cm. at the c h i l l in Figure 37, which is low. This drop in the concentration in this region may be attributed to the drop in the solute concentration resulting from the feeding of the interdendritic liquid from this region towards the ch i l l face during solidification. The addition of the TiB 2 to the melt, therefore, does not significantly change the extent of the segregation at the ch i l l . To determine whether changing the structure during solidif-ication would affect the segregation, a casting was made in which the columnar growth was disturbed at a 'distance of 2.5 cm. from the c h i l l . The resulting composition distribution curve is shown in Figure 36. Comparing this curve to the equivalent curve for a normally cast ingot, shown in Figure 28, i t is observed that the two curves are essentially similar. Accordingly, i t is seen that changing the cast structure near the c h i l l during the solidifica-tion does not influence the segregation near the c h i l l zone. The segregation in the c h i l l zone for the Al-10% Cu alloy which was degassed is shown in Figure 21. Segregation is clearly evident, and more extensive than in normal casting. This would suggest that the presence of fine pores in the casting reduces the extent of segregation in the c h i l l zone. i n C M > - ! ! 2 (_) cr L U -cr 2 : in cn a ' MEAN COMPOSITION RL-10; CU CAST NUMBER .97 MOLD : FIBRFRAX SUPERHEAT l°C) 84 CHILL : COPPER STRUCTURE CHANGE * t * $ * ^ 4 ? A S * + y M * A A i ^ — * * 1 * i » * * * 1 1 * * * s < a » * 1 1 »^ 1 1 1 1 1 1 1 0 . 0 5 . 0 1 0 . 0 1 5 . 0 2 0 . 0 2 5 . 0 3 0 . 0 3 5 . 0 DISTANCE FROM CHILL FACE ( X .05 CM ) Fig. 36 Normalized activity vs .distance from the c h i l l face for the composition and casting conditions indicated. i n MEAN COMPOSITION CAST NUMBER SUPERHEAT l°C) 91 AL-IOJ CU 84 MOLD : FIBREFRAX CHILL : S. STEEL FINE GRAINED cn i o CX r~~ Q Q - . L U — ' C Q X 4 CX z: i ****** * 4 ft I - r 4 * * J * J * i n 0 . 0 I 1 1 1 5 . 0 1 0 . 0 1 5 . 0 2 0 . 0 DISTANCE FROM CHILL FACE ( X 0 5 — i — 2 5 . 0 CM 3 —I— 3 0 . 0 3 5 . 0 oo Fig. 37 Normalized activity vs..distance from the c h i l l face for the composition and casting conditions indicated 3.6 Saturation effects in the scintillation counter Saturation effects in the counter, caused by high activity of the radioactive samples, was investigated by counting samples, init i a l l y possessing a high activity, and then again, later, when the samples had decayed to lower activity levels. The results of these tests are shown in Figures 38, 39, and 40, the only var-iable being the counting time for each sample. The i n i t i a l rate of 700,000 counts in 6 seconds gave the curve for the normalized activity versus distance from the c h i l l shown in Figure 38. At lower counting rates of 500,000 counts in 12 seconds and 500,000 counts in 60 seconds for the same samples, the curves shown in Figures 39 and 40 were obtained. The decay time between Figure 22 and Figure 24 is equivalent to four half lives of the isotope. It is noticed from these Figures that the high counting rate employed in the first case has led to the saturation of the detectors, thus giving erroneous values for the activities measured at different times. In the other two cases, employing intermed-iate and low counting rates, i t is seen that the normalized activ-ity distributions are nearly similar, no significant improvements in the measurements resulting from the use of the much lower counting rates employed in the third case. Use of this low count-ing rate, would, in addition, lead to a greater error in the activity measurement, since the sample would decay during the MEAN COMPOSITION RL-io; cu CAST NUMBER 123 MOLD : FURNACE SUPERHEAT (°C) 10 CHILL : COPPER CM X + HIGH COUNTING RATE A 1— " — 1—4 "* > 1 ) 1— CJ cc CM 111— CO •—4 X 4 X X A + * 4 A * + A A 4 A A X * v * M + V + 4 A +  A + A A + A A A X X X + 4 * 1 A X 4 A X X X +  + * X ! * * * " + I X • * * * A 4 X A 4 A + A A A A X 4 X X A 4 • A A NORMAL 1.06 o - — 4 0 .0 1 DISTANCE 1 10.0 FROM CHILL 1 15.0 FACE ( 1 1 20.0 25.0 X .05 CM ) 1 30.0 Fig. 38 Normalized activity vs. distance from the c h i l l face for the composition and casting conditions indicated. MEAN COMPOSITION CAST NUMBER 123 SUPERHEAT ( C) 10 A L - I O ; CU MOLD : FURNACE CHILL : COPPER MEDIUM COUNTING RATE CJ CX CO CX * * * * * * • * t » , » * « T T T 0.0 5.0 10.0 15.0 DISTANCE FROM CHILL FACE I X 20.0 . 05 CM 25.0 30.0 Fig. 39 Normalized activity vs. distance from the ch i l l face for the composition and casting conditions indicated. oo C J CC rv U J — CO cc GCLlD MEAN COMPOSITION CAST NUMBER 123 SUPERHEAT (°C) 10 A A A A A A A 4 A * 4 A * A A * A T T T 1 0.0 5.0 10.0 15.0 20.0 25.0 DISTANCE FROM CHILL FACE ( X . 05 CM ) R L - I O ; CU MOLD s FURNACE CHILL : COPPER LOW COUNTING RATE 3 0 . 0 oo Fig. 40 Normalized activity vs.distance from the ch i l l face for the composition and casting conditions indicated. counting period, which is a factor that is not corrected for in the calculation of the normalized activities. 3. 7 Temperature distributions in the casting during solidification Temperature measurements were not made on the castings in which Cu 6 4 was added in order to relate the freezing conditions directly to the measured segregation. The temperature measure-ments required the insertion of a series of thermocouples in the casting which would have seriously interfered with the mach-ining of the casting to obtain suitable sample cuttings. As an alternative, castings without Cu 6 4 were made under apparently identical conditions with thermocouples, and the results compared to the corresponding castings containing Cu61+. An example of the variation of temperature with time at five different positions in the casting are shown in Figure 41 for an Al-10% Cu alloy. The curves rise rapidly i n i t i a l l y , indicating fast and uniform response of the thermocouples and the recording equipment to rapid thermal changes. From the curves of Figure 41, the temper-ature distribution across the casting can be plotted as a function of the distance from the c h i l l at progressively increasing times. The results are shown in Figure 42. At any given time the solid-liquid zone in the casting can be considered to be the portion of the casting between the equilibrium liquidus and eutectic temper-atures for this alloy. Accordingly, i t is evident in Figure 42 ° - \ 1 1 1 1 1 1 1 1 1 1 O.J 25.0 50.0 75.0 100.0 125.0 150.0 175.0 200 0 225.0 250.C : TI.MF IN SFCflNDS Fig. 41 Cooling curves for an Al-10% Cu ingot under the casting conditions indicated. oo that the s o l i d - l i q u i d zone increases i n length with increasing time a f t e r casting, varying from a length of .95 cm. 5 seconds a f t e r casting, to 7.5 cm., 60 seconds a f t e r casting. No reheating (which would r e s u l t from separation of the casting from the c h i l l ) of the ingot at the c h i l l face was observed. The advance of the s o l i d - l i q u i d interface during u n i d i r e c t -i o n a l s o l i d i f i c a t i o n may be assumed to r e l a t e to time following the r e l a t i o n s h i p 1 5 , x t A 2 ( t ) ^ where x i s the distance from the c h i l l and t i s the elapsed time, A 2 being a constant. To determine i f t h i s i s applicable i n the present case, the eutectic temperatures obtained from the cooling curves was plotted as a function of t 2 f o r four ingots cast on . a copper c h i l l , giving the r e s u l t s shown i n Figure 43. Corres-ponding results f o r ingots cast on a stainless s t e e l c h i l l are given i n Figure 44. A l l the curves are l i n e a r i n the i n i t i a l part of the ingot showing that the assumed relationship i s applic-able i n t h i s region and i s consistent with the d i r e c t i o n a l s o l i d -llquidus i f i c a t i o n of the casting. The/curves do not pass through the o r i g i n , because of the superheat of the melt and d i f f i c u l t i e s i n establishing the precise time at which the s o l i d i f i c a t i o n commenced. The deviation of the curves from l i n e a r i t y i n the l a t e r stages of s o l i d i f i c a t i o n i s a r e s u l t of non-directional s o l i d i f i c a t i o n and EUTECTIC ISOTHERMS A l - 1 0 % Cu alloys /TIME (sec^) Fig. 43 Movement of e u t e c t i c isotherms f o r Al-10% Cu i n g o t s c a s t under the c o n d i t i o n s i n d i c a t e d . 00 Al-10% Cu alloys EUTECTIC ISOTHERMS Fig. 141+ Movement of eutectic isotherms for Al-10% Cu ingots cast under the conditions indicated. 0 3 C O heat losses from the top surface of the liquid in the mold. Also, the relationship would hold only i f the temperature of the c h i l l face is constant, and i t is seen from cooling curves that this is not the case for these ingots. Seven ingots were cast containing thermocouples under a variety of casting conditions, as listed in Table IV. From the temperature measurements the gradient of the temperature in the solid-liquid region G was determined (this is when the eutectic temperature is just reached by the c h i l l face) , and the length of the solid-liquid zone L^, as listed in the Table. Using the linear portion of the x = Ai + A 2(t) J s curves, where A^  is the intercept of the curve with the abcissa, the values of Aj and A2 were determined as shown in Table IV. The cooling rates and the freezing rates obtained from the temperature data are also listed. It is seen from Figure 43 that a l l the curves for the alloys cast in the Fibrefrax molds are quite close, the only significant difference being the displacement of the curves from one another due to the different values of A 1 # Due to the difficulties in establishing the exact time of the start of solidification, these values could be slightly in error. However, i t would be expected that for the ingots with larger superheats, the values of A± would increase. Though this is the case for ingots 14-6 and 147, the TABLE IV TEHPL'RATURT MT^IREMENTS Al-10% Cu AI.LOYS - CAST NUMBER CHILL MOLD SUPERHEAT °C s cC/cm. cm. (sec. j'"2 (sec. )** FREEZING RATE AT l.?5 CM. cm./sec. AVERAGE DENDRITE ARM SPACING, v COOLING RATE -LOCAL SOLIDIFICATION TIME'AT 1.25 CM. (sec.) COMMENTS 72-146 Copper Fibrefrax 70°C 74 1.08 2.6 .25 .66 26 15 72-147 Copper Fibre frax 30°C 32 2.5 2.1 .25 .84 28.7 14 73-5 Copper Fibrefrax 15°C 70 1.77 3.0 .25 .625 30.8 20 72-143 Stainless Fibrefrax 65°C 42 1.9 3.0 .36 .417 31.7 25- separated from c h i l l 73-1 Stainless Fibrefrax 90°C 140 .57 1.7 .41 .069 150 73-4 Stainless Fibrefrax 50°C 64 1.2S 2.4 .25 .464 28.2 22 73-6 i ; Copper Furnace 50°C 20 • 3.9 .5.3 .25 .26 39.4 45' 91. value of Ai f o r ingot 73-5 i s much l a r g e r than the value obtained f o r the abovementioned ingots. The same e f f e c t of decreasing A i with increasing superheat i s found to e x i s t i n Figure 44 f o r the ingots cast on the s t a i n l e s s s t e e l c h i l l s . The ingot 148 separated from the c h i l l on s o l i d i f i c a t i o n , and therefore shows a r e s u l t d i f f e r e n t than expected. . I t has been shown that the s o l i d i f i c a t i o n r a t e i s r e l a t e d to the dendrite arm spacing i n the c a s t i n g 1 6 . To determine how the dendrite spacing v a r i e s with c o o l i n g r a t e i n the present case, transverse sections o f the castings used i n the temperature measurements were made at 1.25 cms., from the c h i l l and the dendrite arm spacing measured on the po l i s h e d and etched surfaces. Examples of the d e n d r i t i c structures o f two ingots cast on copper c h i l l s are shown i n Figure 45: (b) f i b r e f r a x mold, and (a) furnace cast. The d e n d r i t i c structure o f (h) i s appreciably f i n e r than that of (a) , consistent with the higher f r e e z i n g rates and steeper gradients i n the f i r s t casting. Values o f the average dendrite arm spacings f o r the seven castings are l i s t e d i n Table IV. The dendrite arm spacings are observed to vary appreciably with casting conditions. The dependence o f the spacing on the superheat i s shown i n Figure 46, f o r ingots cast on the copper and s t a i n l e s s s t e e l c h i l l s . The dendrite arm spacing i s observed to pr o g r e s s i v e l y decrease with superheat, with l i t t l e d i fferences between the copper and the stainless, s t e e l c h i l l s . Much l a r g e r spacings were observed f o r the -92. a Fig. 45 Microstructures of sections p a r a l l e l to the c h i l l face f o r Al-10% Cu ingots (a) cast i n the furnace (procedure (a) ) 10OX (b) cast i n f i b r e f r a x molds 100X 93. O 381 c s , » 361 e O 34) 2 cr < 32 LU cE 301 Q Z LU O UJ 281 < cr LU £ 261 chill mold copper fibrefrax 0 copper furnace 0 s.steel fibrefrax \ fct A. \ r\ separated ^ from chill _L J __L 20 40 60 80 SUPERHEAT (°C) Fig. 46 Dendrite arm spacing vs. superheat for Al-10% Cu ingots, cast under the conditions indicated. 4r-AI-10% Cu alloys chill mold o copper furnace • copper fibrefrax A s steel fibrefrax i 2 1 20 40 60 SUPERHEAT ( °C ) 80 Fig. 47 vs. Superheat for Al-10% Cu ingots cast under the conditions indicated. C O - P furnace cast ingot at about 50 superheat, as compared to casting in fibrefrax molds. An appreciably larger dendrite arm spacing was observed when the ingot separated from the c h i l l during solidification. The variation in the length of the solid-liquid region with superheat is shown in Figure 47. Increas-ing the superheat progressively reduces the value of for both the copper and stainless steel chills, the values for the two chills being essentially the same. The results for ingot 73-5 may be erroneous due to difficulties encountered during the measurement of the temperature distributions. 3.8 Chemical analysis of ingot samples The purpose of conducting chemical analysis on a series of samples which had been counted was to establish the correlation between the changes in the measured activity with changes in the copper concentration. It is assumed that the radioactive copper atom is the same as a normal copper atom in this case and the radioactive atoms are uniformly distributed throughout the alloy. The results of chemical analyses of samples taken from three castings are shown in Figures 48, 49, and 50, superimposed on the normalized activity measurements of the same samples. To plot the chemical analysis results in the same way, the copper concentration values obtained were normalized by copper concentra-tion obtained from the same melt samples used for the activity measurements. Only one analysis was done for any one ingot sample rsi cn cr LU—' CO x + A * , i ft 4 X A MEAN COMPOSITION AL-5JT CU CAST NUMBER 80 MOLD t FIBREFRAX SUPERHEAT (°C) 107 CHILL : S. STEEL * Results from chemical analysis + x o.o T T —1 1 1 5.0 10.0 15.0 20.0 25.0 DISTANCE FROM CHILL FACE ( X .05 CM J 30.0 35.0 Fig. 48 Normalized concentrations from chemical analysis superimposed on the normalized activity distribution. MEAN C O M P O S I T I O N A L - 1 5 J C U s . _ C A S T NUMBER 81 MOLD : F I B R E F R A X S U P E R H E A T ( ° C ) 100 C H I L L : S. S T E E L • Results from chemical analysis >-2 CJ CL r-CO CX a:— in cn ' ^ l l i j H * n i j i « , t t i i f 1 i f » f U * ! 1 1 1 1 1 1 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 D I S T A N C E FROM C H I L L F A C E ( X .05 CM ) Fig. 49 Normalized concentrations from chemical analysis superimposed on the normalized activity distribution. i n f M cn >-2 cx Q ° . CO CX i n cn ! MEAN COMPOSITION flL-2G7 CU CAST NUMBER 82 MOLD : FIBREFRAX SUPERHEAT (*C) 103 CHILL : S. STEEL • Results from chemical analysis I I 1 [ 1 1 1 1 ° - ° - - 5 . 0 1 0 . 0 1 5 . 0 2 0 . 0 2 5 . 0 3 0 0 3 5 0 DISTANCE FROM CHILL FACE ( X .05 CM) CO oo Fig. 50 Normalized concentrations from chemical analysis superimposed on the normalized activity distribution. since only one gram of material was available per sample. As such, the magnitude of errors in the measurement could not be established. From Figures 48, 49, and 50, i t is apparent that there is no significant difference between the normalized concentrations obtained either from the activity measurements, or by chemical analysis. As such the correspondence of the normalized concen-trations was taken to be one-to-one and i t was concluded that the activity of the sample containing the radioactive copper is pro-portional to the concentration of copper in the sample, in a l l further tests. 3.9 Inverse segregation in Al-Ag alloys The variation of activity with increasing distance from the c h i l l face for Al-10% Ag and Al-15% Ag ingots cast on copper chills in fibrefrax molds at superheats of 50°C are shown in Figures 51 and 52. In Figure 51, i t is seen that there is an i n i t i a l point with very high activity, which would be caused by the presence of exudation on the c h i l l face of the casting. There is a solute depleted region immediately adjacent to this point, and the compo-sition increases to a steady value towards the interior of the ingot. There is no region clearly showing an increase in compo-sition near the c h i l l face. In view of the criteria used in measuring the c h i l l face segregation for the Al-Cu alloy ingots, i t is seen that there is no c h i l l face segregation evident in 100. •22 •19 •16 113 > HO o < Q 107 UJ CO rr o z •04 01-•98-AI - IO%Ag alloy chill • copper mold : fibrefrax superheat^ 60°C o o " o o o -0 n O O O O O o 0 1 1 51 4 8 12 16 20 D I S T A N C E F R O M CHILL (x 05 cms) Normalized activity vs. distance from the c h i l l for an Al-10% Ag ing 101 . •13 > o < Q LU CO o •10 07 o •04 o AI- l5%Ag alloy chill :copper mold fibrefrax superheat160°C o o Q-£)___JO_ _ £L_Q_ .--_-o_ 9_0 I oil— o o o •98 D I S T A N C E F R O M CHII 1 1 16 20 (x 05 cms) Fig. 52 Normalized activity vs. distance from the c h i l l for an Al-15% Ag ingot. 102. this ingot. In Figure 52 for the Al-15% Ag alloy, i t is observed that the composition is essentially constant, with a slight increase towards the c h i l l face. This increase cannot be definitely attributed to the presence of c h i l l face segregation in the ingot in view of the scatter of the measurements. As is the case with other castings, the first point adjacent to the c h i l l shows a high activity due to the presence of exudations. 103. 4. DISCUSSION In the Al-Cu system, the present results show that: 1. Maximum segregation occurs at the c h i l l face of cast alloys. 2. The amount of segregation at the c h i l l face is small. 3. The extent of the segregation away from the c h i l l face is small. 4. The c h i l l face composition is markedly dependent on: a) alloy composition b) melt temperature on pouring (superheat) c) temperature gradient in the cast material The present results will be considered in detail, as they relate to the observations and theoretical predictions reported in the literature. 4.1 Scatter of the results Kirkaldy and Youdelis8 (Figure 3) report close agreement between their experimental observations and the theoretical values for maximum c h i l l face segregation, with l i t t l e experimental scatter. To obtain this f i t , extrapolations were made on the basis of only one or two experimental points and the results complicated by the presence of exudations having a high alloy concentration at the c h i l l face. In the results summarized by Pell-Walpole1, 104. appreciable scatter of results are reported in inverse segrega-tion measurements as a function of alloy composition and other casting variables. In the present results, variations of the activity and con-centration measurements for neighbouring specimens remote from the c h i l l zone were observed. This scatter is directly attribu-ted to microsegregation in the ingot which is always present. The sample size for counting ( 1 gram) was chosen to be sufficiently large to minimize the effect of microsegregation and s t i l l enable enough samples to be obtained from the casting to measure the segregation. Scatter is s t i l l present in the measurements, the amount varying for different castings depending on the casting conditions. In a l l cases, the f u l l set of samples for a given casting were counted at least three times, and the results averaged. In the observations, the multiple sets of data have been plotted in several figures, to show the scatter between successive count-ing sequences. This scatter is due to the stability of the count-ing system and the statistical fluctuations of the activity of a given sample, the loss of activity due to the decay between the counting sequences having been corrected for. It can be seen that there is much less scatter of repeat counts of a given sample, than scatter between neighbouring samples due to microsegregation, in most cases. A horizontal line of best f i t is drawn through the average data of normalized activity as a function of distance from the 105. c h i l l , and the extrapolation of the curve in the ch i l l zone to the c h i l l face is based on at least three points above the mean composition. It is considered that the present extrapolation is more reliable than that previously reported in the literature, based on more points being used, which f i t a reasonably smooth curve, and rejection of the value of concentration of the mater-i a l adjacent to the c h i l l face. 4.2 Chill face segregation The c h i l l face segregation measured in the present investi-gation, based on the extrapolation of the segregation in the ch i l l zone, is plotted in Figures 53 and 54, as well as the theoretical curve derived from the Scheil analysis. The points in Figure 53 are for ingots directionally cast in the preheated furnace as in casting procedure (a). The points in Figure 54 are for ingots directionally cast in the fibrefrax molds (procedure (b) ). The type of c h i l l used, the superheats, and the cast numbers are indicated in the curves. Considering the experimental results plotted in Figures 53 and 54, i t is evident that there is a very large scatter in the ch i l l face segregation observed for both casting procedures used. In some of the castings, no inverse segregation was detected. This is particularly anomalous in the light of the excellent agreement between theory and experiment reported by Kirkaldy and Youdelis. This difference may be accounted for in part, by the A l - C u a l l o y s C O M P O S I T I O N ( % C u ) F i g . 53 T h e o r e t i c a l predictions and experimentally obtained c h i l l face segregations. AI~Cu alloys 8 i ( 9 o ° o mold 1 fibrefrax C O M P O S I T I O N ( % C u ) Fig. 54 Theoretical predictions and experimentally obtained c h i l l face segregations. o <1 108. difference in t-he casting superheats employed, and the c h i l l used. It can be observed from Figure 54, that the c h i l l face segregation is markedly dependent on both the casting temperature and the c h i l l used. Decreasing the pouring temperature or re-ducing the thermal conductivity of the c h i l l increases the ch i l l face segregation. These observations are in accord with those described by Pell-Walpole1, in which the extent of segregation observed in various alloys increases with a decrease in the rate of cooling. However, in discussing the inverse segregation in Al alloys reported by Woronoff, he notes that increasing the mold conductivity increases the amount of inverse segregation observed, which is clearly in conflict with the present results. The superheat used by Kirkaldy and Youdelis for a l l their ingots was 40°C, and the c h i l l material was polished steel coated with a thin layer of alundum paste. Accordingly, these conditions give results closest to the stainless steel c h i l l and lower super-heats of the present experiments, where the results tend to approach the theoretical predictions. 4.3 Effect of superheat and the thermal conductivity of the c h i l l To show the dependence of the segregation on the superheat more clearly, the results of Figure 53 and 54 have been replotted. The c h i l l face segregation of three alloys Al-5% Cu, Al-10% Cu, and Al-15% Cu are shown in Figures 55, 56, and 57 respectively, as a function of the superheat. The c h i l l face segregation x3 5 -AI-5%Cu alloys chill mold o copper furnace copper fibrefrax A s.steel fibrefrax m from Kirkaldy 8 Youdelis THEORETICAL <*2 LU rr CD L U CO •I-o 130 93 80 40 60 80 SUPERHEAT ( ° C ) 100 F i g . 55 C h i l l face segregation vs. superheat f o r Al-5% Cu ingots cast under the conditions indicated. i — 1 o CO measured by Kirkaldy and Youdelis are also indicated in the Figures, and the values of segregation as calculated from the Scheil theory are shown as solid horizontal lines. In Figure 55, only three points are shown corresponding to the measurements of the present investigation. A l l of these points are seen to lie below the expected value of the segrega-tion, with no segregation being measured in the case of ingot 80. It can be observed that lowering the melt superheat results in an increase in the segregation measured. In Figure 56, a l l the points l i e below the theoretical value. Consider the data from the ingots cast on the copper c h i l l in the fibrefrax molds (filled circles), ignoring the values ob-tained from ingots 98 and 125. The c h i l l face composition is seen to decrease progressively with increasing superheat, with the segregation reaching zero past a casting superheat of 80°C. Ingot 125 shows a smaller segregation, possibly due to non-directional solidification. In case of ingot 98, the water flow to the c h i l l block was markedly reduced, significantly changing the casting conditions. This situation is anologous to reducing the thermal conductivity of the c h i l l . The c h i l l face segregation with the stainless steel c h i l l also decreases with increasing superheat, but is significantly higher than the segregations obtained for the copper c h i l l , for a given superheat. Ingot 96, cast on the copper ch i l l in the furnace arrangement, shows a high ch i l l face segregation, close to the .theoretically expected value. 3 o 6 •5 5-4 o ^ r— LU or e> LU CO Al- 10% Cu alloys o 123 '125 20 95 chill mold o • copper furnace © copper fibrefrax A s.steel fibrefrax a from Kirkaldy 8 Youdelis THEORETICAL O -A47 96 slow water flow 98 _ TiB2 84 A A 4 6 ^129 45 4 0 6 0 SUPERHEAT ( ° C ) .97 .44 80 Fig. 56 Chill face segregation vs. superheat for Al-10% Cu ingots cast under the conditions indicated. However, the ingot 123 cast at low superheat shows a much lower value of ch i l l face segregation, though i t would be expected that the slower rate of heat removal would lead to segregation higher than those obtained from casting, either on the copper ch i l l or the stainless steel c h i l l in the fibrefrax molds. This low value may be due to extensive nucleation in the bulk of the liquid resulting from the low casting superheat. In Figure 57, the emphasis is on Al-15% Cu alloys cast in the furnace arrangement, using a copper c h i l l . Here i t is seen that lowering the superheat causes an increase in the amount of the c h i l l face segregation. Changing the size of the ingot cast (using a smaller melt size for ingot 21) does not significantly affect the ch i l l face segregation obtained. This demonstrates that the amount of segregation only depends on the freezing con-ditions in the first parts of the ingot to solidify. Ingot 81, cast in the fibrefrax mold with a stainless steel c h i l l shows a very large c h i l l face concentration, much larger than the theoretically expected maximum, even though the superheat is high. The reason for this is not known. Ingot 128, which had a fine grained structure due to the addition of a grain refiner, shows a chi l l face segregation close to the theoretically expected value. The presence of inverse segregation in ingots possessing an equiaxed grain structure is in accord with the findings reported by Adams11. 3 ' 5 o ^ 4 2 -3h 3-2 LU rr CD LU CO •i AI-l5%Cu alloys a 124 > 128 T i B ? ifer c h i l l m o l d 0 c o p p e r f u r n a c e © c o p p e r f i b r e f r a x A s . s t e e l f i b r e f r a x • f r o m K i r k a l d y & Y o u d e l i s A 81 T H E O R E T I C A L O19 0 | 8 20 O _ L 21 4 0 6 0 8 0 100 S U P E R H E A T (°C) 120 140 160 Fig. 57 Chill face segregation vs superheat for Al-15% Cu ingots cast under the conditions indicated. 114. 4.4 Exudations on the c h i l l face It was generally found that the c h i l l faces of ingots cast in the present work were free of massive exudations or sweat beads, the faces appearing smooth. Kirkaldy and Youdelis re-ported that exudations were generally observed on c h i l l faces of their ingots. A section of one of their ingots, perpendicu-lar to the c h i l l face, showing exudations, is reproduced in Figure 58. The structure of the c h i l l face of ingot 85 of this investigation, in which exudations were produced by moving the ingot away from the c h i l l face, is shown in Figure 59. The po-sition of the c h i l l face of the ingot in this Figure, before separation of the ingot from the c h i l l , is marked x x. The thickness of the exudation was large, due to the imposed large separation; only the part immediately below the c h i l l face is shown in the Figure. At the i n i t i a l c h i l l face, some remelting of the primary dendrites has occurred, and long interdendritic channels can be observed. For a normal casting, a section through the middle of the ingot, perpendicular to the c h i l l face is shown in Figure 60, the c h i l l face being marked Y Y. No evidence of remelting or exudations is observed and the interface is flat. In the tempera-ture measurements, no evidence of reheating was detected, indi-cating no separation between the ingot and the c h i l l face. Since there was no indication of exudations, the corrections applied by Kirkaldy and Youdelis to their data, which are uncertain, were 115. Fig. 59 Microstructure of the base of an Al-10% Cu ingot perpendicular to the c h i l l face from the present investigation, in which exudations were induced. 10OX Fig. 60 Microstructure of the base of an Al-10% Cu ingot, perpendicular to the c h i l l face, cast normally, showing absence of exudations. 10OX t-not applied. No such corrections have been mentioned in the determinations of the c h i l l face compositions carried out by earlier workers (except Adams), who determined the extent of inverse segregation in various alloy systems. In view of the fact that many of them did observe sweat beads on the c h i l l face of the ingots they examined, i t is possible that their results are in error. The c h i l l face segregation they measured could be a combination of the actual segregation and the exuded eutectic. The theory of the mechanism of inverse segregation assumes that no exudations are present. 4.5 General mechanism for inverse segregation As described previously, lowering the melt superheat causes the following effects: a) increase of the c h i l l face composition b) increase of the extent of segregation into the ingot away from the c h i l l face (increase of the c h i l l zone L ) c) increase of the solid-liquid zone length L^. d) increase of the dendrite arm spacing The effects b), c), and d) are a l l interrelated and affect the magnitude of a). Since inverse segregation is due to inter-dendritic feeding to compensate for solidification shrinkage and specific volume changes with temperature, the feeding must take place in the solid-liquid zone, and the segregation near the c h i l l face will depend on the length L^. The length of the c h i l l zone L will also be greater with larger L^. The results of the present work generally indicate that there is no increase in the solute concentration past L , contrary to the theoretical and experimental results reported by Adams, and Kirkaldy and Youdelis. It is possible that in the present investigation, there is an overall small increase in the solute concentration over the bulk of the casting, smaller than the scatter limits of the data. The concentration increase reported by Kirkaldy and Youdelis beyond is very small, and :may not be evident in the present results. Flemings et al. have predicted that for parallel liquidus and eutectic isotherms, moving at a constant rate in a direction perpendicular to the c h i l l face, the composition in the region of the ingot at a distance greater than from the c h i l l , is equal to the mean composition of the ingot. In the Flemings et al. analysis, i t was assumed that L , the c h i l l zone length, equals L^, the mushy zone length. In the present results, i t was found that L c does not equal L^, being smaller for a l l the cases examined. The values of L for the ingots examined are c 6 i listed in Tables I, II, and III, and plotted in Figure 61 as a function of superheat for the Al-10% Cu alloys. In Figure 61, i t is evident that L c decreases rapidly with increasing superheat, and for a given superheat, L c is larger for the stainless steel c h i l l than for the copper ch i l l . The dotted lines in the Figure have been included to show the trend of the data and should not be considered as representative of the functional dependence of L c 18 14 A 47 129 45 0 98 slow water flow Al-10% Cu alloys chill mold O copper furnace 9 copper fibrefrax A s.steel fibrefrax to E i n IOh p X 6r-O 123 95 .96 A 4 6 A 84 2 h I 97 4 4 , J—m L _ a 1 20" 4 0 6 0 , 80 SUPERHEAT ( ° C ) 100 Fig. 61 L vs. & c superheat for Al-10% Cu alloys for the casting conditions indicated. i—• CO 120. on superheat. Comparing the values of L c in Figure 61 with the values of (Figure 47), at a given superheat, i t is noted that L c is smaller than L^. The difference between L q and can be attributed to incomplete back feeding of liquid through the interdendritic channels, characterized by a "Tortuity Factor". If the flow of liquid down the channels is restricted during solidification, small voids should be produced due to inadequate feeding of liquid. Voids, in the form of microporosity were observed inter-dendritically in the castings examined, as described previously, The presence of voids would not affect the values of the normal-ized activities obtained after sectioning the ingot. This is because the voids were quite small, and, as compared with the bulk of cuttings taken for the analysis, the total volume of the pores is negligible. The presence of voids is only signifi-cant when density measurements are used to determine the varia-tions in composition of the ingot. The extent of interdendritic flow associated with inverse segregation will be related to the size and distributions of the liquid channels present. This, in turn, is related to the dendrite arm spacing, and therefore, to the superheat on casting. At higher superheats, the dendrite spacing in the c h i l l zone is small, and the liquid channels are therefore small leading to retardation of backflow and low c h i l l face concentrations. Lowering the super-heat increases the relative channel size and therefore results in 121. high c h i l l face concentrations. Stainless steel chills produce higher solute concentrations at the c h i l l face than copper chills for similar conditions of casting. The present results show a small decrease in for stainless steel, compared to copper, and an increase in both the L c and the dendrite arm spacing. The increased dendrite arm spacing results in backflow occurring more readily with the stainless steel c h i l l , in agreement with the larger L c obtained. The increased backflow would result in an increase in the concentra-tion at the c h i l l face, as observed. In the ingots cast at very low superheats, l i t t l e c h i l l face segregation was observed, contrary to the general trend of increasing segregation with decreasing superheat. This is attribu-ted to a change in the manner of solidification. Under normal casting conditions, progressive solidification occurs away from the c h i l l face. At very low superheats, i t is likely that ex-tensive nucleation occurs well ahead of the general solidifica-tion front. Under these conditions, backflow of enriched liquid would be markedly reduced, giving low c h i l l face segregation. 4 . 6 Effect of the cast structure on segregation Fricke 1 8has reported, on the basis of decanting experiments on Al-Cu alloy castings, that inverse segregation results from a columnar to equiaxed transition during solidification. This conclusion has been refuted by Edwards and Spittle 1 9 on the basis of inverse segregation detected in small, fully columnar ingots produced by electroslag melting. The present results show that inverse segregation is obtained in a columnar structure, in agreement with Edwards and Spittle, and previously reported results. It is s t i l l possible that inverse segregation may occur when the structure transition occurs. To test this hypothesis, in one test (97), the columnar structure was changed to equiaxed during solidification by adding Foseco degassing compound when the interface was 2.5 cm. from the c h i l l . It was found that changing the structure had no effect on the distribution of copper along the ingot, i.e. , the distribution was the same as that observed in a fully columnar casting (44). Castings with a fine equiaxed grain structure were made using TiB 2 grain refiners. The segregation obtained in this case (84) is very similar to casting (46) produced under similar con-ditions . In the case of the fine grained ingots, inverse segre-gation occurs by liquid feeding through the channels between partially solidified grains. Solidification of the fine grained material is considered to occur progressively, roughly similar to columnar growth, with nuclei forming just ahead of the advancing interface. 123. 4.7 Effect of the gas content of the melt One of the ingots (95) was degassed with the use of Foseco degaser. The c h i l l face composition for this ingot showed very good agreement with the predicted value from Scheil's theory. It is possible that the presence of small amounts of gas at the last stages of solidification would block or hinder the passage of liquid through the interdendritic channels. This would cause a lower value of c h i l l face composition than would be expected i f flow were not hindered. This was not the case in the test on ingot 95. Pell-Walpole reports instances in which the extent of segregation has been increased by the presence of large volumes of gas in the melt, the gas being released in the last regions of the casting to solidify. However, the presence of gas in the melt had always led to ingots with exudations, and without correcting for the exudations in the value of the c h i l l face segregations determined, these conclusions could be in error. Roth 1 7 has cast Al-Cu ingots with large gas content in some ingots, and found that the amount of segregation at the c h i l l face and the extent of the c h i l l zone was reduced by the presence of gas. In view of this, the effect of the presence of gas in the ingot on the final segregation pattern is unclear. 124. 4.8 Comparison of theories The theory proposed by Flemings et al. is not equivalent to the Scheil theory as claimed, primarily due to the assumption of constant values for the segregation coefficient and specific volumes of solid and liquid. Their demonstration of equivalence was only for a small region of composition variation, and they selected appropriate constant values for that region to give good f i t . To treat the quantities mentioned above as variables, makes solutions of the equations derived by Flemings et al. d i f f i -cult. The excellent agreement between theory and experiment re-ported by Kirkaldy and Youdelis was not obtained in the present results. Their good agreement may be partly due to their low casting superheat which results in high c h i l l face concentrations. However, they apply a correction for the exudations observed in their castings, and this may lead to erroneous results. They also give no indication of the scatter associated with repeat observations or scatter resulting from small changes in the casting variables. 4.9 Turbulence during casting In the present investigation, the casting of the ingot was a dynamic process, in which the molten metal was poured onto a cold c h i l l as in casting procedure (b). The prediction of the theories, however, hold for the condition that the melt is quiescent 1 2 5 . when the freezing of the ingot is initiated. Thus an additional variable is introduced into the process, which cannot be ex-pli c i t l y defined. Roth17has reported that stirring the melt vigorously ahead of the advancing solid-liquid interface causes a reduction in the segregation. In the present investigation, the liquid was poured smoothly and quickly into the mold against the c h i l l in the same manner for a l l the castings produced, using fibrefrax molds. The turbulence on pouring was likely small and died away quickly. For the furnace cooled castings, the melt was quiescent before solidification was initiated. It is considered that the effect of turbulence in the present investigation on the observed segregation is small and can be neglected. It is thus seen that, fundamentally, Scheil 1s theory appears satisfactory for determining the c h i l l face segregation, but only under a relatively narrow range of solidification conditions. Changing the solidification variables affects the magnitude of the segregation detected, and in order to be more general, the theory will have to incorporate the effect of these variables. 5. SUMMARY AND CONCLUSIONS (1) Inverse segregation results from interdendritic backflow of enriched liquid towards the c h i l l face of the ingot. (2) The segregation is markedly dependent on casting variables such as casting superheat, gas content in the ingot, tempera-ture gradients in the ingot during solidification, the thermal conductivity of the c h i l l and the cast structure. (3) The c h i l l face segregation increases with decreasing super-heat. At very low values of superheat the segregation process changes to give a reduced segregation at the c h i l l face. (4) Decreasing the thermal conductivity of the c h i l l in the range of superheat from 100°C to 30°C increases the segregation at the c h i l l face. (5) Reducing the gas content of the ingot enhances the ch i l l face segregation in the ingot. (6) Inverse segregation is present in completely equiaxed ingots that are cast unidirectionally. (7) The theory of macrosegregation as put forward by Flemings et a l . , does not become equal to the Scheil solutions for the Al-Cu system, except for a narrow range of compositions, eve-though the authors claim equivalence. This is because of the simplifying assumptions used to reduce the complex mathematical expressions to yield analytical solutions. (8) The magnitude of the c h i l l face segregation has been found to differ from the values reported in the literature and the quantitative predictions of the theory of inverse segregation.. ( 9 ) No significant inverse segregation was observed in ingots of Al-Ag cast unidirectionally. 6. SUGGESTIONS FOR FUTURE WORK Using the same radioisotope techniques, the segregation for unidirectionally cast ingots cast at low superheats can be investigated. The effect of the mold and c h i l l conduct-ivity can be studied in further detail, in order to reduce the amount of the c h i l l face segregation, as well as to provide means of reducing the amount of exudations on the c h i l l face. The variation of the composition throughout the length of the ingot perpendicular to the c h i l l face can also be studied, and the results compared to the theoretically pre-dicted composition distributions, under different cooling conditions. Numerical integration of the theoretical equa-tions will have to be carried out. The effect of radial temperature gradients on the segregation can also be investigated. 1 2 9 . APPENDIX A A.l Statistical consideration of radioactivity measurements The activity emanated by radioactive materials exhibits random fluctuations in intensity due to the particulate, random nature of the emission. In addition, the intensity decays ex-ponentially, the rate defined by the half l i f e of the isotope. As a result, probability laws are applied in analyzing the count-ing measurements to assess accidental or "indeterminate" errors. The systematic or "determinate" errors are minimized by applying known correction or calibration procedures to the data, and by careful planning of the experiments. The basic frequency distribution governing random events such as radioactive decay is the binomial distribution, from which the simpler and more familiar Poisson and Gaussian (or Normal) distri-butions are derived. Since the Gaussian distribution has been found to be relatively easy to handle mathematically, i t is used frequently, and is almost as accurate as the Poisson distribution over a wide range. The Poisson and Normal distributions are represented mathematically as P n n: and G n 1 130. Fig. 63 The error of counting determinations. A. Probable error B. Standard deviation C. Nine tenths error D. Ninety-five hundredths error E. Ninety-nine hundredths error respectively, where y is the actual or true mean of the quantity being measured and where n is the value of any one observation. The difference between the Poisson and the Gaussian distributions is shown in Figure 62. To test for non-statistical behaviour of the counter, 350 measurements of activity of a sample containing Sn 1 1 3, (which is a long lived isotope, with a half l i f e of 118 days) were made. The activity of the sample used in this test was comparable to the activities of the samples containing Cu6t+, so as to provide a test under conditions to those used in the measurements of the variation of the activity along the ingot length for the Al-Cu ingots. The results of the experiment showed no significant difference from the expected normal distribution. The activity measurements were processed for the "Goodness of Fit" tests with the help of the "Freq" library program at the U.B.C. Computing Centre. A.1.1 Confidence limits and errors The confidence of data may be represented in many ways, the most common being those'listed below:20 (a) The one sigma level is defined by the standard deviation, and there is a 31.73% chance that this level will be exceeded. (b) The nine tenths error is so named because there are nine chances out of ten that the error in a specific determination will be 132. less than this value. (c) The ninety-nine hundredths error signifies a confidence that errors of 99% of the determinations will be less than this value. This degree of confidence is highly significant. Errors used to define confidence levels are listed in Table V. A.1.2 Errors in the counting rate determinations From Figure 63 2 1, i t is seen that in the region where counting was performed on the ingot samples (about 250,000 counts), the error in the counting rate is seen to be less than 1% of the total number of counts. Thus, i t was concluded that the activity measurements made are of high accuracy. The background counting rate, due to naturally occurring radioactive materials present in and around the counter, as well as due to cosmic rays, is an additional source of error encountered. It has been assumed that there was no loss of the average activity during the actual counting operation on each sample. To minimize the error arising from this assumption, the counting times were chosen to be small. For a counting time of At, and half l i f e t± , the correction for such an error would be 0.34% of the total for At / t = 1%, and 0.69% for At / t = 2%.22 T A B L E V Errors used to define confidence intervals 2 0 probability of observing error larger than / y where y = mean value Name of error T D Probable Standard deviation Nine tenths Ninety-five hundredths Two sigma Ninety-nine hundredths Three sigma 0.6745 1.0000 1.6449 1.9600 2.0000 2.5758 3.0000 0.5000 0.3173 0.1000 0.0500 0.0455 0.0100 0.0027 Fig. 64 Activity vs. weight of the Al-Cu sample containing Cu 6 4. 135. It was attempted to present samples of similar geometry to the counter, as changes in the geometry of the sample will affect the measurements. Samples of different weights, cut from c h i l l cast buttons prepared from a well stirred melt with Cu6k tracer dissolved in i t , were used to check on the combined effect of geometry and absorption of radiation in the specimen. Chill casting small buttons served to reduce segregation effects in the samples. The variation of the activity obtained with respect to weight of the sample is shown in Figure 64. Muminium is a weak absorber for gamma rays. * It has an absorption coefficient of .08 cm-1 for 0.5 MeV gamma rays as com-pared to 1.127 cm 1 for lead. As such, no appreciable loss of activity due to absorption in the aluminium is expected. In case of larger sample sizes, i.e. , larger volumes, the absorption effect may play a larger part. For larger sample weights, i t is observed from Figure 64 that the variation of activity with weight changes from the linear relation-ship found with lower sample sizes. In the region of around 1 gram,(most samples counted were in this region), the variation is linear, and the normalization of the activity with respect to weight in this region is justified. 136. REFERENCES 1. Pell-Walpole, W.T., and Hanson, D. , Chill Cast Tin Bronzes, Edward Arnold and Co., London, 1951, p. 211. 2. Vosskuhler, von H., Z. Metallkunde, 1965, 56, 719. 3. Flemings, M.C. , and Nereo, G.E., Trans. Met. Soc. AIME, 1967, 239, 1449. 4. Flemings, M.C, Mehrabian, R. , and Nereo, G.E. , Trans. Met. Soc. AIME , 1968, 242, 41. 5. Fleniings, M.C, and Nereo, G.E. , Trans. Met. Soc. AIME, 1968, 242, 50. 6. Youdelis, W.V. , Ph.D. Thesis, McGill University, 1958. 7. Scheil, E., Metallforschung, 1942, 20, 69. 8. Kirkaldy, J.S. , and Youdelis, W.V., Trans. Met. Soc. AIME, 1958, 58, 212. 9. Youdelis, W.V., The Solidification of Metals (Brighton Confer-ence ), The Iron and Steel Institute, London, p. 112. 10. Sauerwald, F. , Metallwirtschaft, 1943, 22_, 543. 11. Adams, D.E. , Journal, Institute of Metals, 1948, 75_, 809. 12. Youdelis, W.V., and Colton, D.R. , Trans. Met. Soc. AIME, 1960, 218, 628. 13. Ball, L.S., and Youdelis, W.V. , Journal, Institute of Metals, 1971, 99_, 29. 14. Kaempffer, F.L., M.A.Sc. Thesis, University of British Columbia, 1970. 15. Szekeley, J. , and Themelis, N.J. , Rate Phenomena in Process  Metallurgy, Wiley Interscience, 1971, p. 307. 16. Brody, H.D., and Flemings, M.C., Trans. Met. Soc. AIME, 1966, 236, 615. 137. 17. Roth, von W., Z. Metallkunde, 1965, 56, 713. 18. Fricke, W.G., Trans. Met. Soc. AIME, 1969, 245, 1126. 19. Edwards, K.P., and Spittle, J.A. , Metallurgical Transactions, 1972 , 3_, 1004. 20. Chase, G.D., and Rabinowitz, J.L., Principles of Radioisotope  Methodology, Burgess Publishing Co., 1963. 21. Kohl, J., Zentner, R.D., and Lukens, H.R., Radioisotope  Applications Engineering, Van Nostrand and Co., 1961 22. Leymonie, C., Radioactive Tracers in Physical Metallurgy, Chapman and Hall, London, 1963. 23. Mehrabian, R. , Keane, M.A. , and Flemings, M.C., Metallurgical Transactions, 1970, 1, 3238. 

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}]}"
                            data-media="{[{embed.selectedMedia}]}"
                            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:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0078741/manifest

Comment

Related Items