Castability of HAYNES 282 alloy. Abstract

Transcription

1 Workshop Advanced Ultrasupercritical Coal-fired Power Plants, EVN Forum Maria Enzerdorf Vienna, Austria, September Castability of HAYNES 282 alloy Natalia Sobczak 1, Zenon Pirowski 1, Robert M. Purgert 2, Waldemar Uhl 1, Krzysztof Jaskowiec 1, Lukasz Boron 1, Stanisław Pysz 1, Jerzy J. Sobczak 1 1 Foundry Research Institute, Krakow, Poland 2 Energy Industries of Ohio, Cleveland, USA Abstract This paper is focused on casting trials by the Energy Industries of Ohio and performed in cooperation with the Foundry Research Institute in Krakow, Poland. Wrought HAYNES 282 alloy was examined from the point of view regarding its applicability for use in casting applications, particularly as a possible candidate alloy for Advanced UltraSuperCritical (A-USC) turbine components for next generation power plants. A systematic study of the alloy castability was done using spiral fluidity tests and stepped tests for characterization of the alloy s sensitivity to shrinkage cavity formation and casting constriction, accompanied with thermal analysis and detailed chemistry and structure characterization. A comparison between cast and wrought properties is provided by optical light and scanning electron microscopy observations coupled with computed tomography and quantitative image analysis that were done in order to characterize the structure of wrought and cast alloys as well as casting defects. INTRODUCTION Relatively new wrought γ strengthened age-hardenable nickel superalloy HAYNES 282, designed by Haynes International (USA) for improved high temperature properties, has a unique combination of creep strength, thermal stability, weldability, and fabricability not found in currently available commercial alloys [1]. Due to a good oxidation resistance this alloy could potentially be used in highperformance, high-temperature environments, e.g. as an attractive candidate material for high temperature/high pressure boiler and turbine components in Advanced Ultrasupercritical Power Plants (A-USC) [2-4]. Several programs are in place in the United States and Overseas to develop and gather the necessary data in order to determine the utility of the material for various applications in power generation [5]. This paper is focused on casting trials of commercial wrought HAYNES 282 alloy as part of a supplier development effort initiated by the State of Ohio in support of the U.S. DOE/Ohio Coal Development Office Advanced UltraSuperCritical Materials Program and performed by the Energy Industries of Ohio in cooperation with the Foundry Research Institute (FRI) in Poland.

2 The aim of this work is to study the applicability for use of commercial wrought HAYNES 282 in casting applications, particularly as a possible candidate alloy for A- USC large-scale turbine components for next generation power plants. The experiments conducted and outlined in this paper have been mainly directed to optimize the technology of melting HAYNES 282 nickel alloy in an open induction furnace, evaluation of its technological properties and the collection and analysis of this preliminary data to be used for next stages of research. EXPERIMENTAL PROCEDURE Materials Rods of industrially available 75x2000 mm HAYNES 282 nickel-based superalloy (H282) were used in this study (Fig. 1). The alloy chemical composition, given by producer, is shown in Table 1. For comparison, a synthetic alloy of an assumed similar chemical composition was prepared from technically pure metals at the FRI experimental research facility in Krakow, Poland. Fig. 1. As-received rods of 75x2000 mm H282 alloy Table 1. Chemical composition of H282 alloy given by producer Content of elements, wt.% Al C Co Cr Fe Mo Ti Ni bal. Melting procedure Both the commercial H282 alloy and synthetic alloy, with a chemical composition believed to be similar to the Haynes produced alloy, were melted in air under gravity conditions in a medium-frequency induction furnace (Radyne Inc., UK) using a home-made crucible of 40 kgs capacity and lined with Al 2 O 3. First, the synthetic alloy (marked as H0) was melted in order to pre-wash the crucible. Next, the same crucible was used for melting Haynes provided H282 alloy Three heats were done and marked as H1, H2 and H3, respectively. Figure 2 shows the subsequent steps of the melting procedure used. After cutting the H282 rods to the desired length and immediately after mechanically cleaning all surfaces, the material was loaded into the dried crucible (Fig. 2a), and a gaseous protection system was put into motion (Figs. 2b and 2c). 2

3 During heating and melting the alloy surface was covered with inert gas (argon) resulting in a satisfactory mirror-like quality liquid metal surface (Fig. 2d). In order to decrease oxygen content, the molten alloy was deoxidized with NiMg alloy. The molten alloy was then cast into sand molds coated with zirconia-based binder. Before pouring the molten metal, the molds were preheated in air at 200 C for 24 h. a) b) c) d) Fig. 2. Following stages of melting of H282 alloy: loading of charge (a), controlling of argon feed (b), gas cover (c) and the surface of molten metal under protective atmosphere (d) Chemical analysis A detailed comparative analysis of the chemical composition was done for both alloys by means of two independent methods: 1) X-ray fluorescence analysis (XRF) in selected places on mechanically cleaned surfaces by means of a NITON XL3t 900S GOLDD apparatus (Thermo Scientific, USA) used in practice as an express non-destructive and low-cost test for chemistry characterization (Fig. 3), and, 2) precise emission spectrometry analysis by means of POLYVAC 2000 apparatus (Hilger Co, UK) following FRI certified testing procedures. For the first method, the measurements were done in similar locations of the same regions of the as-received rods and the produced casings. For example, Figure 4 illustrates the setting of the XRF measurement points for the as-received rods, 3

4 marked as rod #1 and rod #2. The second method was used as a reference only while the measurements were done in 10 places of one cross-section at about a 10 cm height of each rod. Additionally, for selected rods and castings, the amount of gases (oxygen, nitrogen) were measured by means of the thermal extraction method using a TC600 analizer (Leco, USA). Fig. 3. Application of none-destructive XRF method for chemical composition measurements of charge material using NITON XL3t 900S GOLDD apparatus (Thermo Scientific, USA) Fig. 4. Schematic representation of the location of XRF measurements made for rod #1 and rod #2, whose values are collected in Table 2 and Table 3, respectively Castability Castability characterizes the flowing properties of molten metal and describes the case with which it can be cast to obtain a part with good quality minimizing cost, defects and lead-time. In foundry practice, castability is characterized by fluidity, i.e. an ability of metals and alloys to flow through the gating system filling the cavity of the casting mold of constant cross sectional areas and conforming to its shape. Quantitatively it is defined as the maximum length the metal flows in a given test mould before it is stopped or halted by the solidification process when cast at a given temperature [5]. The greater the length of the solidified metal, the greater is its fluidity and thus better its castability. 4

5 In this study, the castability of H282 alloy was characterized by a widely used fluidity spiral tests [5] assisted by the thermal analysis of a set of thermophysical properties of the alloy compared to experimentally based estimates and the examination of the alloy s tendency for thermal constriction and shrinkage cavity formation using stepped tests. In both cases, sand molds were used. The molds were produced using furan-based binder as being the most suitable for this kind of technological investigation since it is characterized by a high degree of flexibility. Thermal analysis In order to estimate the main temperature characteristics of the H282 alloy, thermal analysis was done on filtered molten alloy of the first heat poured at an initial customized temperature of 1450 o C into an alumina crucible of φ35x70 mm dimensions and with an inserted type B thermocouple. The STEELFERR filters of φ50x10 ppi (FERRO-TERM, Poland) were used. Two kinds of temperature changes were evidenced using the MrAC-15 multichannel recorder, i.e. basic cooling curve and differential curve (as a derivative from first basic function). Stepped test This technological test is widely used in foundry practice for characterization of casting contraction and tendency regarding the formation of shrinkage cavities. For this reason, the molten alloy was cast into the sand molds of a special stepped design with four thermocouples inserted in different locations as shown in Fig. 5a. The mold was prepared and dried in the same way as it is traditionally used using standard foundry practices for the casting of alloys (Fig. 5b). Figure 5c shows the 4- stepped casting with corresponding dimensions of each step. The tendency of shrinkage cavity formation during solidification of the molten alloy was determined by measurements of the defected areas in the perpendicular section of the stepped test castings in relation to the surface area of the ingate using data from three heats. a) b) HAYNES 282 g1 g2 g3 g4 5

6 H1: mm H3: mm c) Fig. 5. a) Ready-to-cast completed mold used for stepped technological tests showing the locations of the four thermocouples; b) dried two parts of mold in oven; c) stepped casting of H282 alloy with dimensions of gx35x150 mm (where g is thickness of each step) made in stepped tests for heats H1 and H3 Fluidity tests In fluidity tests, the thermocouple was located in each mold and the temperature measured before metal pouring and during its solidification (Fig. 6). Four superheat temperatures for pouring were used: 1440, 1470, 1510 and 1570 o C. For each temperature, a spiral casting ( spiral ) was made and its length was experimentally estimated. a) b) Fig. 6. The mold used in fluidity test: before (a) and after (b) pouring of H282 alloy Structural characterization Structural examinations by means of optical microscope (AxioOZm1) and scanning electron microscope (STEREOSCAN 420) equipped with energy dispersive spectroscopy analyzer (EDS LINK ISIS) were done for selected samples taken from cross-sections or on the surface of the H282 rods and castings. As complementary studies, the 3D distribution of structural defects was characterized by computed tomography (CT) by means of Nanotom and V/Tome/x L-450 apparatuses (Germany). RESULTS AND DISSCUSSION Figure 4 illustrates the setting of XRF measurement points for the as-received rods, marked as rod #1 and rod #2, with corresponding measurement values collected in Table 2 and 3, respectively. Their comparison with reference measurements of the chemical composition of the H282 rods made on their perpendicular section at about the 10 cm height by the spectrometric method using a certified testing procedure (e.g., Table 4) shows a good agreement that might be considered as experimental evidence for both high homogeneity of the as-received H282 nickel alloy and applicability of express XRF method for chemistry 6

7 characterization. Among all examined elements, the XRF measurements show a large scattering of only aluminum content that could have been caused by the selection of the place for analysis. For XRF, the measurements are done on the rod (casting) surface while the measurements by the emission spectroscopy method are done inside the rods (castings) in their cross sections. Therefore, the effect observed may have been caused by the segregation of aluminum. Table 2. Chemical composition by XRF analysis in selected area of H282 rod #1 Measurement point Content of elements, wt.% Mo Ni Co Fe Cr Ti Al On cylindrical surface Average In section # In section # Table 3. Chemical composition by XRF analysis in selected area of H282 rod #2 Measurement point Content of elements, wt.% Mo Ni Co Fe Cr Ti Al On cylindrical surface Average In section # In section # The detailed analysis of the changes in the chemical composition of the commercial H282 alloy after three heats (H1, H2, H3), compared to that of the synthetic alloy (H0) suggests that the total melting time is a key factor in the scattering of aluminum content while the aluminum segregation might be caused its 7

8 evaporation and partial consumption due to a reaction of the aluminum with the mold (Table 5). Table 6 shows the results of gas content measurements in the castings produced from synthetic H0 and Haynes produced H282 alloys using different melting procedures accompanied with liquid metal treatment with NiMg alloy as a deoxidizer. In both cases, the amount of oxygen and nitrogen is similar showing more than a twice lower value compared to that of the H282 alloy after melting without deoxidizing. Table 4. Chemical composition of commercial H282 rod recorded by spectrometric analysis taken from selected regions of one cross-section Measurement Content of elements, wt.% point Mo Co Fe Cr Ti Al C Ni base Table 5. Chemical composition of synthetic (H0) and commercial H282 alloy (heats H1, H2, H3) after melting and casting (the results of spectrometric analysis) Heat number Content of elements, wt.% Mo Co Fe Cr Ti Al C Ni H H H H base Table 6. Effect of deoxidizing of H282 alloy melt with NiMg on the amount of gases recorded in solidified castings by thermal extraction method Operations Gas content, wt.% Oxygen Nitrogen Melting of synthetic alloy with deoxidizing Melting of H282 alloy with deoxidizing Melting of H282 alloy without deoxidizing

9 1 2 3 T = f(t), o C t, s a) dt/dt = f(t), K/s t, s b) Fig. 7. Cooling curves of H282 alloy: a) basic b) differential, where is time, 1 beginning of solidification (t b ), 2 maximum solidification rate, 3 end of solidification (t e ) Figure 7 shows two kinds of cooling curves obtained for the H282 alloy using the MrAC-15 multi-channel recorder. Comparison of the basic cooling curve (Fig. 7a) with a differential curve, as a derivative from first basic function (Fig. 7b), allows distinguishing three characteristic temperatures as follows: 1350 o C - beginning of solidification, 1310 o C - temperature corresponding to maximal solidification rate, 1260 o C - end of solidification. The results of thermal analysis for heats H1 and H3, are shown in Figs. 8 and 9, respectively. 9

10 T = f(t), o C t, s a) g1 g2 g3 g4 dt/dt = f(t), K/s t, s b) Fig. 8. Cooling curves of stepped test for heat H1 recorded by four thermocouples located in different places of the mold (Fig. 5) and corresponding to four steps of thickness g1, g2, g3 and g4 (g1=6 mm, g2=11 mm, g3=23 mm, g4=40 mm): a) basic b) differential. Dark blue color corresponds to the temperature change in section of g1, green in section of g2, brown in section of g3 and light blue in section of g4 10

11 T = f(t), o C t, s a) g1 g2 g3 g4 dt/dt = f(t), K/s t, s b) Fig. 9. Cooling curves of stepped test for heat H3 recorded by four thermocouples located in different places of the mold (Fig. 5) and corresponding to four steps of thickness g1, g2, g3 and g4 (g1=10 mm, g2=15 mm, g3=28 mm, g4=44 mm): a) basic, b) differential Based on the results obtained the following thermal characteristics have been calculated: temperature of maximum solidification rate (Fig. 10a), time of maximum 11

12 solidification rate (Fig. 10b), solidification time (Fig. 10c) and thermal effect of solidification rate (Fig. 10d). a) b) c) d) Fig. 10. Temperature of maximum solidification rate (a), time of maximum solidification rate (b), solidification time (c) and thermal defect of solidification rate (d) estimated from stepped test as a function of wall thickness of casting made from H282 alloy. Data from heats H1 and H3 were used The thermal effect of the solidification rate following formula: was calculated using the where: and are temperatures of the end and the beginning of solidification, respectively. All thermal characteristics increase in value with increases in the wall thickness of the stepped test casting. The relation between thermal effect of solidification rate and solidification time has a strict rectilinear character with high correlation coefficient of R 2 = (Fig. 11): (1) (2) 12

13 This fact can be considered as evidence of reliable experimental thermophysical data obtained in this study. The tendency of molten H282 alloy regarding shrinkage cavity formation during solidification, experimentally determined by measurements of the defected area in the perpendicular section of the stepped test castings in relation to the surface area of the ingate, is shown in Fig. 12. The results obtained for the three heats suggests that the larger the surface area of the ingate the smaller the total area of cavities in the investigated sections. Figure 13 illustrates the results of the fluidity tests made with the H282 alloy at four temperatures for melt pouring: 1440, 1470, 1510 and 1570 o C. For each temperature, the length of spiral casting produced ( spiral ) was measured as a function of pouring temperature. Fig. 11. Thermal effect of solidification rate as a function of solidification time for H282 alloy. Data from heats H1 and H3 were used Fig. 12. Surface area of shrinkage cavities in relation to surface area of ingate for H282 alloy castings. Data from heats H1, H2 and H3 were used 13

14 Fig. 13. The photos of spirals cast from H282 alloy in four fluidity tests illustrating the effect of pouring temperature on spiral length For validation of the results obtained in the fluidity tests, experimental data were compared with a theoretical assumption by calculating the individual spiral length for each casting and pattern used when making the molds. Radius for given angle of Archimedean spiral can be calculated using the formula: where a is a constant. By measuring of a pattern and casting spiral radius at a given angle (e.g. ) we are able to calculate the constant a: (3) Having the data for a maximum radius of cast spiral at a given value of the constant and using the Archimedean spiral approach, the maximum angle and the radius of the pattern spiral corresponding to this angle were calculated and collected in Table 7. Experimental data were found to be in a good agreement with calculated values. The fluidity of molten H282 alloy increases significantly with increase of superheat from 130 K to 160 K and then grows slightly at the range from 160 K to 260 K (Fig. 14). Compared to a minimal superheat, the preheating of molten H282 alloy up to 1570 o C (superheat of 260 K) decreases its fluidity by almost twice the value. (4) 14

15 Table 7. Calculated constant a for given pouring temperature and for pattern used for mold Pouring temperature, o C Parameters Pattern Fig. 14. Comparison of experimental length of fluidity spirals made from H282 alloy with calculated results at different superheat degrees Using the calculated values of cast spiral length and its corresponding values for the pattern, we may determine casting contraction which is one of the most important characteristics in determining sound foundry practices and for product engineers. Knowing the spiral constant, the angle and the corresponding value of the radius, the length of Archimedean spiral can be calculated from the following equation: L = ½ a (ϕ (ϕ 2 + 1) ½ + arcsin hϕ) (5) Since each individual spiral has its own specific dimensions the length of spiral can be estimated by calculation. The comparison of the actual obtained values from the experimental lengths of pattern spiral gives the casting contraction for the selected pouring temperatures (Table 8). As shown in Fig. 15, for the overheat range of K, the H282 alloy demonstrates near liner dependence of measured casting constriction versus superheat, showing a maximum value of 4.8% at a pouring temperature of 1570 o C and corresponding superheat of 260 K. 15

16 Table 8. Calculated and measured lengths of fluidity spirals cast at different temperatures with corresponding values of casting contraction for H282 alloy Pouring temperature, o C Parameters, mm, mm, % Fig. 15. Casting contraction determined for H282 alloy in relation to superheat Structural characterization Structural characterization by optical microscopy (OM) and scanning electron microscopy (SEM) coupled with energy dispersive x-ray spectroscopy (EDS) evidenced significant difference in structure of commercial H282 alloy and those melted in cast in this study. The structure of as-received H282 rods is typical for wrought alloys and it is composed of γ phase grains with gray carbide and yellow nitride precipitates, well distinguished by the color on OM micrographs (Fig. 16). SEM+EDS analysis suggests the presence of two types carbides, MC and M6C, mainly dispersed along grain boundaries. Small amount of carbide precipitates are noted also within grains. Figures illustrate a comparison of microstructures of the H282 alloy after melting and casting under different magnifications, for both etched and none-etched samples taken from different sections of spiral (Figs ) and stepped (Figs ) castings. Independently on casting conditions, cast H282 alloy has heterogeneous dendritic structure with numerous carbide and nitride precipitates dispersed in interdendritic regions. However, the size of dendrites as well as the size and distribution of precipitates change from section to section due to difference in corresponding cooling conditions. 16

17 Fig. 16. OM microstructure of commercial as-received wrought H282 alloy at different magnifications (etched sample) a) b) c) d) Fig. 17. OM microstructures of cast H282 alloy (heat H3, T=260 K) in sample 0 of spiral casting with a= 4.02, ϕ=0 rad, L=0 mm: a-c) etched; d) none-etched 17

18 c) d) c) d) Fig. 18. OM microstructures of cast H282 alloy (heat H3, T=260 K) in sample 2 of spiral casting with a=4.02, ϕ=7.85 rad, L=345 mm: a-c) etched; d) none-etched As shown in Figures 17-23, a strong microsegregation of the alloying elements caused from the solidification process is well distinguished on the etched samples indicated by a difference in color making qualitative image analysis and experimental estimation of the dendrite arm size (DAS) possible. The measurements made in the cross sections of each step of the stepped H282 castings show that the DAS values increase with increase in the casting thickness. For example, in heat H3 (Fig. 22), the average DAS values measured are 83, 63, 54 and 44 µm for 44, 28, 15 and 10 mm casting thickness, respectively. Furthermore, microsegregation of the alloying elements can also be characterized by their distribution across a dendrite using EDS microanalysis in the center and at the periphery of dendrites. Using the example of heat H3, a comparison of the average values of the amounts of alloying elements identified in cast H282 alloy by EDS in two sections (A, D) corresponding to two steps located far from each other (the thinnest step g1 versus the thickest step g4) is shown in Table 9. For both steps, a significant difference in the chemical composition at the center (C o ) and at the periphery (C p ) of the dendrites is noted for titanium and molybdenum as a consequence of its location in carbides and nitrides in the interdendritic regions. A comparison of C o /C p values for each element recorded in the two sections g1 and g4 suggests that cobalt and molybdenum tend to also segregate along the casting length. 18

19 a) b) c) d) Fig. 19. OM microstructures of cast H282 alloy (heat H3, T=260 K) in sample 3 of spiral casting with a=4.02, ϕ= rad, L=601 mm: a-c) etched; d) none-etched a) b) 19

20 c) d) Fig. 20. OM microstructures of cast H282 alloy (heat H3, T=260 K) in sample 4 of spiral casting with a=4.02, ϕ=20.79 rad, L=920 mm: a-c) etched; d) none-etched a) b) c) Fig. 21. OM microstructures of cast H282 alloy (heat H3) corresponding to the sample taken from g1 step section of the stepped casting (step thickness g1=10 mm): a-b) etched; c) none-etched 20

21 a) b) c) Fig. 22. OM microstructures of cast H282 alloy (heat H3) corresponding to the sample taken from g2 step section of the stepped casting (step thickness g2=15 mm): a-b) etched; c) none-etched 21

22 a) b) c) Fig. 23. OM microstructures of cast H282 alloy (heat H3) corresponding to the sample taken from g3 step section of the stepped casting (step thickness g3=28 mm): a-b) etched; c) none-etched a) b) Fig. 24. OM microstructures of cast H282 alloy (heat H3) corresponding to the sample taken from g4 step section of the stepped casting (step thickness g4=44 mm, etched sample) 22

23 Fig. 25. Dendrite arm size (DAS) measured in four sections (g1, g2, g3, g4) corresponding sections of each step of stepped casting (heat H3) Table 9. Dendrite microsegregation in cast H282 alloy recorded in two sections (A, D) corresponding to two steps (g1, g4) of stepped casting (heat H3) Element Center of dendrite, C o Content of element, wt. % Periphery of dendrite, C p C o /C p A (g1) D (g4) A (g1) D (g4) A (g1) D (g4) Al Ti Cr Fe Co Ni Mo Additionally, detailed structural characterizations of the castings produced in the stepped and spiral fluidity tests were performed. Comparing the results obtained OM, SEM and computed tomography (CT) H282 alloy shows a strong susceptibility to casting contraction and the formation of shrinkage porosity and cavities (e.g. Figs a). The 3D-CT observations (Fig. 26a) indicate that besides shrinkage cavities some discontinuities are noted near the casting surface up to a few mm depth. Following SEM+EDS analysis (Fig. 26b), it was determined that they correspond to oxide-rich phases (mainly containing Al 2 O 3 and MgO), most probably formed due to interaction of the molten alloy with the mold and playing a role of nucleator for carbide precipitates. 23

24 a) b) Fig. 26. CT-image of stepped casting (heat H3, section of thickness g2=15 mm) illustrating the presence of discontinuities near casting surface (a) and its corresponding SEM analysis showing the formation of oxide-rich precipitates:1 - matrix; 2- oxide-rich phase; 3,4 carbides; 5 nitrides Conclusions The technology of induction melting of commercial wrought HAYNES 282 alloy in air under argon cover has been presented. Compared to a non-deoxidized melt, the use of deoxidizer (NiMg alloy) reduces oxygen content in H282 castings by over three times. Castability characterization using stepped and spiral fluidity tests shows strong susceptibility of the H282 alloy to the formation of shrinkage cavities and hence casting contraction. Fluidity of the H282 melt depends on the temperature of superheating. Reducing the superheat degree from 260 K to 130 K decreases the length of the cast spiral by over 50 %. Heavy-section castings can be poured from lower temperatures, which means that the casting contraction will be less than 2 %. Pouring castings with medium wall thicknesses requires the superheat of a minimum 150 K, and in this case, a casting contraction of about 2.3 % can be expected. Thinwall castings require a high degree of superheating, which means that the casting contraction can exceed even 3.5 %. The casting contraction of H282 alloy depends very strongly on the temperature of the melt superheating. Raising the superheat from 130 K to 260 K increased this contraction by almost three times. Acknowledgements The authors thanks Dr Henry White for fruitful discussion on the results of this study. The financial support from the Ministry of Science and Higher Education of Poland (Project No. 721/N-NICKEL/2010/0) is acknowledged. References R. Viswanathan, R. Purgert and U. Rao, Materials for Ultra-Supercritical Coal-Fired 3. Power Plant Boilers, in Proc. Materials for Advanced Power Engineering, Part II, Forschungszentrum Julich GmbH, 2002, pp Proc. Advances in Materials Technology for Fossil Power Plants, edited by R. Viswanathan, D. Gandy, K. Coleman, ASM International, R. Viswanathan, J.F. Henry, J. Tanzosh, G. Stanko, J. Shingledecker, B. Vitalis, and R. Purgert, U.S Program on Materials Technology for Ultra-Supercritical Coal Power Plants, Journal of Materials Engineering and Performance, 14 (2005), M.C. Flemings, Solidification Processing, McGraw-Hill Inc. London,