NET-N0--7D1
ARNE KRISTIAN DAHLE MO 9 70518 2
MUSHY ZONE PROPERTIES AND CASTABILITY OF ALUMINIUM Arne Kristian Dahle aj£x- fVo— nsi
Mushy Zone Properties and Castability of Aluminium Foundry Alloys
INSIBiSUTlON OF 1HIB DOCUMENT IS UNL&ffnSD m
Thesis submitted to the Norwegian University of Science and Technology (NTNU) in partial fulfilment of the requirements for the degree Doktor Ingenipr. DISCLAIMER
Portions of this document may be Illegible in electronic image products. Images are produced from the best available original document " <§Jm Afifimg war, die QdCfpetese" C. Dexter ACKNOWLEDGEMENTS
The work presented in this thesis was started in January 1993 and finished in January 1996.1 am very grateful to theNorwegian Research Council, Elkem Aluminium ANS and Hydro Aluminium for funding this work.
My advisor, actually for the last four years, has been Professor Lars Amberg. I would like to thank him for his stimulating and energetic support and advice, for providing opportunities, and for listening to my, sometimes, disputable ideas during these years.
Stationed at the Department of Metallurgy, at the former Norwegian Institute of Technology, I have gained much insight from discussions with some of my predecessors doing their doctor ’s degree on casting and solidification. I would especially like to mention: Dr. Axel Kolsgaard, Dr. Stian Sannes and Dr. Per Arne Tpndel.
Casting and moulding is not a straightforward task, and I am grateful to Mr. Arne Nordmark, Mr. Freddy Syvertsen and Mr. Alf Sandberg for sharing their knowledge and for their practical support. Without them I would have made spirals without a sprue....
I spent two months at the Aluminum Casting Research Laboratory at Worcester Polytechnic Institute, Worcester (MA), between February and May of 1995. The hospitality and involvement of Dr. Alauddin Ahmed, Professor Diran Apelian, Professor Makhlouf M. Makhlouf and Professor Peder C. Pedersen is very much appreciated. In addition to these, I would like to thank all my friends in Worcester for making the time there so enjoyable and memorable.
The support of alloys from Fundo a.s., Hpyanger, was also very important, and I would like to thank Mr. Asbjpm Prestmo for providing an inexhaustible source of material and for some interesting discussions. Some material was supplied by Elkem Aluminium Mosjpen.
I would like to thank my previous and present office mates, Dr. Egil Trpmborg and Mr. Jan Anders Saster, for providing interesting discussions and friendly surroundings. Dr. Axel Kolsgaard is acknowledged for helping me keeping track of time.
- i - All my colleagues and friends at theDepartment of Metallurgy and SINTEF Materials Technology are acknowledged for creating a stimulating and friendly atmosphere.
Finally I would like to thank my parents and my brother for their continued support.
In addition to the articles included in this thesis, parts have been presented in the following publications:
A.K. Dahle and L. Amberg, "Investigation of dendrite coherency in two commercial Al-Si foundry alloys", SINTEF Report, STF34 A93248, December 1993
A.K. Dahle and L. Amberg, "Relationship between dendrite coherency and castability of some commercial foundry alloys", SINTEF Report, STF24 A94558, April 1994
A.K. Dahle and L. Amberg, "Investigation of the Dendrite Coherency Point in Solidifying Al-Si Foundry Alloys", 4’th International Conference on Aluminium Alloys, Atlanta (Ga), 1994, T.H. Sanders Jr. and E.A. Starke Jr. (Eds.), pp. 91- 98
A.K. Dahle and L. Amberg, "Dendrite Coherency in Aluminium Alloys", TMS Materials Week 1994, Rosemont, Chicago (111.), 2-6.0kt. 1994, p. 112
A.K. Dahle, F. Syvertsen, A. Nordmark and L. Amberg, "Relationship Between Rheological Properties of the Mushy Zone and Feeding Mechanisms of Solidifying Aluminium Foundry Alloys", TMS Materials Week 1995, Cleveland (Oh), 29.Okt-2.Nov. 1995, p. 50
L. Amberg, A.K. Dahle, CJ. Paradies and F. Syvertsen, "Factors Affecting the Castability of Aluminium Foundry Alloys", Proceedings from the International Conference on Casting & Solidification of Light Alloys, Institute of Metals and Materials Australasia Ltd, Gold Coast (Qld.), Australia, 30-31 August, 1995, pp. 49-54
A. Ahmed, A.K. Dahle, D. Apelian, L. Amberg and M.Makhlouf, "Modeling of Solidification and Feeding of Solute-Rich Alloys", ACRL Spring Report,
- it - Aluminum Casting Reseach Laboratory, Worcester Polytechnic Institute, Worcester(MA), May 1995, pp. 32-57
A.K. Dahle and L. Amberg, "The Effect of Grain Refinement on the Fluidity of Aluminium Alloys", accepted for publication at The 5’th International Conference on Aluminium Alloys, Grenoble- France, 1-5. July 1996
Trondheim, January 1996
Arne Kristian Dahle
- m - - IV - ABSTRACT
The expected growth in application and market share of aluminium castings is demanding an increased understanding of the mechanisms of defect formation during casting. Casting is a cost-effective production route, but the restricted utilization of castings is often a result of the inadequate reproducibility and quality of the cast structure. Today, much effort is concentrated on the development of a more complete understanding of how to produce the optimum microstructure at an optimum cost, involving the development of advanced computer models.
The objective of this thesis was threefold: First to determine how the solidification conditions affect the rheological behavior in the partially solidified state. Second, to measure how alterations in solidification variables influence castability, and third to investigate the relationship between mushy zone rheology and castability.
The development of mechanical strength in the mushy zone was measured as a function of chemical composition; content and type of main alloying element, grain refinement and eutectic modification. Measurements of the dendrite coherency point provided accurate determination of the point where the dendrite network is established. The strength measurements confirm that the dendrites are largely independent and free- floating before dendrite coherency. Both the point and rate of strength development in the subsequently established interdendritic network are strongly dependent on the size and morphology of the dendrites and fraction solid.
The castability investigation was limited to evaluations of fluidity and feeding. Fluidity measurements showed a complex effect of increased grain refinement; first decreasing and then increasing or stabilizing the flow length. No direct correlation between dendrite coherency and fluidity was found for the effect of grain refinement, since dendrite coherency is continuously postponed with grain refinement. Alterations of the concentration and type of main alloying element gave a direct relationship between mushy zone rheology and fluidity.
The range of the operating feeding mechanisms during solidification is directly related to the rheological properties of the mushy zone. Radiographs of cast Al-Si plates indicate that the amount and distribution of porosity may be influenced by burst feeding. The dendritic network collapses when the stresses upon it exceed its strength.
- v - The mechanism and importance of burst feeding deserve future attention since it in combination with interdendritic feeding may be the most important feature controlling porosity formation. Some relationship between rheological properties and porosity was detected, but the quantification of porosity is difficult and puts a restriction on the significance of the results. This needs future focus.
The data and new knowledge gained in this thesis have subsequently culminated in a comprehensive micro-macro model of shrinkage porosity formation, incorporating the basic principles of solidification and feeding. The model is currently in the state of numerical solution and programming, and no simulations are available at present.
- vi - - Vll CONTENTS
Acknowledgements i
Abstract v
Introductory Summary 1
Article #1 25 A.K. Dahle and L. Amberg: "On the Assumption of Additive Effect of Solute Elements in Dendrite Growth", submitted to Metallurgical and Materials Transactions
Article #2 43 A.K. Dahle and L. Amberg: "Rheological Properties of Solidifying Aluminum Foundry Alloys", accepted for publication in Journal of Metals(JOM), March, 1996
Article #3 55 A.K. Dahle and L. Amberg: "Development of Strength in Solidifying Aluminum Alloys", submitted to Acta Metallurgica et Materialia
Article #4 79 A.K. Dahle, P.A. Tpndel, C.J. Paradies and L. Amberg: "Effect of Grain Refinement on the Fluidity of Two Commercial Al-Si Foundry Alloys", accepted for publication in Metallurgical and Materials Transactions A, 1996
Article #5 97 A.K. Dahle, S. Karlsen and L. Amberg: "Effect of Grain Refinement on the Fluidity of Some Binary Al-Cu and Al-Mg Alloys", accepted for publication in Cast Metals, 1996
- viii - Article #6 121 A. Ahmed, A.K. Dahle, D. Apelian and L. Amberg: "Modeling the Feeding of Aluminum Alloy Castings", accepted for publication in Light Metals 1996, TMS, Warrendale (PA), 1996
Article #7 139 L. Amberg, A.K. Dahle, C.J. Paradies and F. Syvertsen: "Feeding Mechanisms in Aluminum Foundry Alloys", AFS Transactions, vol. 103, 1995, Paper 95-115 also published in Materials Australia, vol. 27, No. 7&8, 1995
Article #8 155 A.K. Dahle, A. Nordmark and L. Amberg: "Measuring Feeding and Porosity in Al-Si Plate Castings", SINTEF Report, STF 24 A96501, January, 1996
»
- IX - INTRODUCTORY SUMMARY
This introductory summary is intended to put the work reported in the articles comprising this thesis into a wider perspective and to show how the articles are related. Basically the summary should show how the work fit into and complement our understanding of solidification and defect formation during casting. Although the summary has a large scope and touches upon many aspects of solidification and casting it is not intended to give a complete coverage, but merely to provide the reader with enough information to relate and hopefully understand the fundamental principles that this thesis rely on. For more thorough and introductory descriptions of the topics discussed in the summary, the reader is advised to consult the textbooks that exist on the field[l-3].
1. Brief history
Compared to the other important structural metals, aluminium was discovered rather late. The Dane H.C. 0rsted was the first to produce a few milligrams of pure aluminium on the laboratory scale in 1825. The development and production was on a small scale for many years. Napoleon was very fascinated by aluminium, which at that time was even more expensive than gold. Ordering aluminium armour for his soldiers, all he received were some military helmets, an aluminium dinner set and some aluminium toys for the children.
With the invention of the dynamo (by von Siemens in 1866), followed the patent of the electrolytic production of aluminium, which was patented nearly simultaneously by P.L.T. Heroult in France and C.M. Hall in U.S.A. in 1886. Their process formed the basis of the process that is used today. From a production of 16 metric tons in 1886, the world wide production of aluminium turned 20.3 million tons in 1992. Today, aluminium is the second most widely used metal in the world.
The history of casting is much older, and started with the noble metals. Simple cast objects appeared by 4000 B.C[4-5], The development of casting from here on is a combination of the development of mould materials and techniques, and the discovery of new compositions that could be more easily cast or that would give sounder, stronger, cheaper or more decorative castings. Although casting has a long history, the
- 1 - developments made after World War II has surmounted the preceding years of development.
Among the first aluminium castings that were made is the Eros statue on Piccadilly Circus in London, cast in 1893.
Since all aluminium today is produced by electrolysis, all virgin aluminium passes through a casting operation, usually DC casting. However, the largest tonnage of cast aluminium today is subsequently wrought to final shape. Approximately 25 % of the world consumption of aluminium is in the form of shaped castings.
Historically the foundry industry has been rather small-scale, often family driven, companies where the experience and expertise have been carried from generation to generation. Developments were based on trial-and-error and luck, combined with the insight of the foundryman. Making good castings has for a long time been regarded as an art.
Pure aluminium is a soft and ductile material with little strength, and adding alloying elements such as silicon, copper, magnesium and manganese is necessary to achieve the desired properties. The optimization of alloy selection is based on different criteria for wrought and casting alloys, and casting alloys usually contain more alloying elements than the wrought alloys. Figure 1 show how important casting properties vary with composition in the Al-Si phase diagram. Aluminiums relative strength, lightness, corrosion resistance and ability to be easily recycled ensure its widespread application in many industries. The combination of high strength and low weight has made aluminium alloys attractive to the aircraft industry, and about 75% of aircraft construction is in aluminium.
A development of particular importance for the aluminium foundry industry was the the aluminium silicon range of alloys. The good fluidity imparted by silicon makes it especially suitable for casting purposes, however the embrittling silicon impairs mechanical properties. This was later remedied with the discovery of modification by A. Pacz in 1920[7]. Al-Si alloys are by far the most widely used aluminium casting alloys today.
The advantage of casting as a manufacturing route is that molten metals are easily shaped to net-shape or near-net shape products with complex geometries and fine
-2- detail. Compared to parts made from wrought material, castings have certain disadvantages. Along with lower strength values governed by structure and alloy, large inclusions in castings limit their applicability. Achieving a dense structure is a basic problem in the production of castings for high strength requirements. On the other hand, forging techniques have definite shape limitations, besides being more energy intensive.
Figure 1. Casting properties ofAl-Si alloys as a function of composition[6], (a) Fluidity; (b) Filling capacity; (c) Shrinkage sensitivity; (d) Uniformity of shrinkage; (e) Hot-tearing tendency.
The traditional, low-cost, method of producing aluminium castings is the die cast route in its pressure, gravity and low-pressure variants. Sand castings retains its markets for small series castings and investment casting techniques are used extensively for aerospace and speciality applications.
- 3 - 2. Casting today and future perspectives
In a society that is showing increasing environmental as well as energy-saving awareness, there is a strong incentive towards making environmentally sustainable products which are recyclable and energy-efficient.
Legislative requirements and marketplace driven forces are demanding changes in the materials selection[8]. Today there is a strong incentive and need from the automotive industry in developing aluminium castings. The demands for lesser emissions of greenhouse gases (C02, NOx etc.) is forcing the car-manufacturers to reduce the weight of the cars. GM’s goal is 3 times more fuel efficient cars, which are lighter and safer, and at no additional cost[9]. Other sectors, such as consumer goods show a similar trend. Much of this weight reduction is expected to occur by an increasing use of aluminium, both wrought and cast. Today, about 75 % of the aluminium used in cars is castings.
The stricter requirements on consistency, reliability and high-performance, as well as mass production of large volumes are bringing changes to the foundry industry. Foundries are currently showing much interest in the field of computer modeling, where the aim is to be able to simulate the casting throughout the production and in this way improve mould design, gating practice, alloy selection etc. This involves simulating mould-filling, thermal history, solidification and feeding, end microstructure and mechanical properties of the final product. The process of developing software to do these tasks has already begun, and foundries are using the models. One important feature is to be able to predict the formation of defects, and this field will most probably receive increasing attention in the next few years. The work described in the present thesis is a contribution to this area, namely understanding the underlying principles of defect formation during solidification. Let us first focus on the background features of the solidification process.
3. What happens when a metal solidifies ?
Solidification is a crystallization process driven by the cooling of a molten metal or alloy. The main event that occurs upon solidification is the establishment of a cast structure. This structure is characterized by the size, shape and orientation of the crystals, the distribution of the chemical species (i.e., micro- and macrosegregation),
- 4 - and the internal and external perfection (i.e., porosity, tears, surface finish etc.).
Many variables exert an influence on the cast structure, but they may be systematized into a wide definition of chemical composition and rate of solidification. Included in chemical composition are (in addition to the nominal composition) the inclusions and gaseous species that are present, and the solidification rate accounts for the convection that may be present as well as the fact that the rate of freezing may vary in different locations in the casting.
Depending on chemical composition and rate of freezing the solid/liquid interface may adopt different morphologies, classed as planar, cellular and dendritic. Aluminium foundry alloys solidify with the dendritic morphology, which was also the first one to be identified. Figure 2 shows an early drawing made by Tschemoff in 1868. Dendrites are characterized by a linear, branched structure, with dendrite arms parallel to specific crystallographic directions.
Figure 2. An early drawing of a dendrite[10].
Two different types of dendrite growth may be encountered, depending on chemical composition and rate of cooling; columnar and equiaxed. Due to the high thermal conductivity of aluminium, the high solute content (large interval of solidification) and the presence of good, potent nuclei, aluminium foundry alloys usually exhibit equiaxed dendritic growth.
- 5 - 3.1 Nucleation
Nucleation is the term used to describe the process of formation of the first clusters of atoms that may later develop into stable growth sites. Usually a certain undercooling is required for nuclei to be stable in the melt; an effect related to the capillary effect of the radius of curvature of the solid/liquid interface. When foreign particles act as nucleation sites, the term heterogeneous nucleation is used, and for most practical casting situations, this is the mechanism by which the first crystals form.
Foreign particles may act as nucleation sites when they reduce the activation energy for crystal formation. The main criterion for a particle to be favourable for nucleation is that it must provide a low surface tension to the solid, i.e., good wetting.
Heterogeneous nucleation is a favourable event, and is utilized in daily foundry practice by addition of grain refiners, or inoculants. In aluminium it is widely accepted that titanium, boron, or both titanium and boron provide grain refinement. However, one must realize that it is not only important to form the nuclei, they also have to survive to form grains.
Effective heterogeneous nuclei are one requisite, and explanation, as to why equiaxed dendritic growth occurs. Another source to grain formation is the survival and growth of detached dendrite fragments.
3.2 Dendritic growth
Figure 3 a) illustrates part of the phase diagram for a binary alloy. The equilibrium partition coefficient k (=CS/C„ where Cs is the composition of the solid and C, is the composition of the liquid) defines how the solute separates between solid and liquid, and for the diagram sketched in Figure 3 a), k - 6 - growth. Chalmers and his coworkers were thefirst to describe and quantify this effect, in 1953[11-12]. Composition sol.+liq. Ck c C/k distance Composition Temperature Actual ___ i { temperature Freezing temperature Undercooled distance solid/liquid interface Figure 3. Schematic of the principle of constitutional supercooling, (a) Part of a binary phase diagram; (b) Concentration profile of solute ahead of the solid/liquid interface; (c) Resultant temperature profile ahead of the solid/liquid interface. Figure 4 a) and b) illustrates the two types of dendritic growth that may occur, columnar and equiaxed, respectively. In conditions where steep thermal gradients may be established, columnar dendritic growth will be the dominant growth mechanism. - 7 - Columnar dendrite growth is characterized by the dendrites growing opposite to the direction of the heat flow. The solute-rich, grain refined aluminium foundry alloys, however, usually solidify with a breakdown into the equiaxed manner of growth. Due to alloy constitution, convection and the high thermal conductivity, the equiaxed crystals nucleate and grow in the undercooled liquid ahead of the interface. Cq > 0 a) mGc > G > 0 b) 6<0 Figure 4. Illustration of geometry and concentration and thermal profiles of the two dendrite growth modes: (a) columnar, and (b) equiaxed[2]. An extra complication arises when metals solidify in that heat is generated, i.e., release of latent heat. Since the equiaxed dendrites are growing in an undercooled melt, the latent heat of fusion is rejected into the liquid in an approximately radial fashion. For - 8 - solidification to proceed it is therefore important that heat is carried away from the solidification front since the latent heat tends to reduce the supercooling. The rate of removal of latent heat therefore controls the rate of solidification and dendrite growth rate. The total undercooling ahead of the growing dendrites is given by: AT = ATC + AT, + AT, + ATk (1) where ATC is the constitutional supercooling, AT, is the thermal undercooling, AT, is the undercooling due to curvature (Gibbs-Thomson undercooling), and ATk is the undercooling due to atom attachment kinetics (usually neglected). Hunt[13] has proposed that the dendrite growth rate is related to the total undercooling by: A'DAT2 (2) TmC0(k-l) where A’ is a constant, D is the diffusion coefficient in the liquid, F is the Gibbs- Thomson coefficient, C0 is the initial alloy composition, k is the partition coefficient and m is the slope of the liquidus line. In a multicomponent alloy several elements are rejected by the interface simultaneously, and it has been suggested[14-15] that this may be accounted for by assuming an additive effect of the elements. For the dendrite growth rate this involves summing the effect of the elements, i.e., A1 DAT2 (3) r 5>;<2o,A- i) i for all elements i that are present in dilute amounts. This equation hereby predicts a decrease of dendrite growth rate with increasing concentration of dilute elements. The effectiveness of the solute atoms, i.e., their constitutional effect, depends on thevalues of k and m, which determines the degree of separation of the element. This has been investigated in Article #1 of this thesis. - 9 - During the initial stages the equiaxed dendrites can grow and reject heat and solute relatively independently of each other with a growth rate that is largely governed by the rate of latent heat removal, i.e., thermal undercooling (the others being relatively constant). In an aluminium casting, especially when sand cast, grains will soon start to form throughout the whole cross-section, so that the casting is entirely in a mushy condition. In general, the size of the mushy zone is determined by the solidification range (AT0) and the thermal gradient (G) at thelocation. Aluminium foundry alloys are classified as long-freezing range alloys, and they will therefore have a relatively large mushy zone. After the initial period of growth of the primary dendrite arms, the dendrites impinge on their neighbours and form a coherent network in thepartially solidified alloy. This is called dendrite coherency. 3.3 Dendrite coherency Chai et al.[16-17] developed two techniques to precisely measure the dendrite coherency point, identical to the techniques used in this thesis. Two factors determine when a dendrite network is formed; dendrite growth rate and grain size. Equation (3), which has been suggested to describe the effect of alloying elements in dilute concentrations indicates that the dendrite growth rate is decreased when the concentration of dilute elements increases. This would indicate a delayed coherency point, and was the focus of the investigation reported in Article #1. The effect of iron and magnesium was elucidated in AlSi7 and AlSil 1 alloys. It was observed that the grain size was altered with the iron addition, and that assuming an additive effect of solute is not correct unless the effect of grain size is taken into consideration. The coherency measurements in Articles #2, 4 and 7 was focussed on the effect of cooling rate and grain refinement. An increase in cooling rate will increase the rate of latent heat removal, and the dendrite growth rate is increased. Larger undercoolings may also arise when the cooling rate is increased, giving an increase in the nucleation rate and a decrease of grain density. Dendrite coherency occurs earlier. The effect of adding grain refiner is to cause a larger number of dendrites, each growing at a lower rate, and a postponed coherency point. - 10 - 3.4 Growth after dendrite coherency After theprimary dendrite arms have impinged onto their neighbours the major growth mechanisms will be growth of the secondary dendrite arms and coarsening. In cases where the dendrites are highly branched, the secondary dendrite arms may still experience a constitutional supercooling that forces them to grow and fill out the interdendritic spaces. However, as solute is rejected by the solidification front and enriched in the liquid, the concentration in the liquid rises and smooths out the concentration gradient that creates the constitutional supercooling. Coarsening will then be the rate controlling solidification mechanism. Coarsening is a diffusion controlled growth process driven by the energy reduction achieved by reductions of the solid-liquid interfacial area. A ripening process causes an increase of the secondary dendrite arm spacing, Due to the large surface tension associated with smaller dendrite arms due to a large curvature, the larger dendrite arms will grow on the expense of the smaller which remelts. The dendrite arm spacing is approximately proportional to the cube root of time[2], i.e., X2 = 5.5 (Mtf)113 M = (4) where tf is the solidification time and Ce is the eutectic composition. The secondary dendrite arm spacing is important since it determines the spacing of precipitates and porosity and thus have a strong effect on the mechanical properties of the as-solidified material. Coarsening is the final stage of solidification in alloys that only solidify with precipitation of a primary phase. However, in the solute rich aluminium foundry alloys, precipitation of a eutectic phase usually ends the solidification process, beginning when the concentration in the liquid is equal to the eutectic composition. 3.5 Eutectic solidification The eutectic solidification in a binary alloy is an invariant reaction in the phase - 11 - diagram, and therefore proceeds without changes in composition and temperature. The bulk of the experimental work reported in this thesis has been focussed on the Al-Si alloys, and it is hereby appropriate to narrow ourselves to describing the growth of the binary Al-Si eutectic. Silicon has a high entropy of fusion and therefore belong to the group of elements capable of faceting. Due to anisotropy in the growth of the atomic planes, the growing crystals are bounded by the slowest growing faces. Eutectic solidification is characterized by the simultaneous growth of two or more phases from the liquid, i.e., aluminium and silicon. Normally Al-Si alloys grow with an irregular lamellar eutectic structure. Ahead of each lamellae, be it A1 or Si, there exists a short diffusion layer of the opposite element that controls the growth. The lamellae rejects solute due to a low solubility. Since the solubility of A1 in Si is very low, but not vice versa, the constitutional supercooling ahead of the Si-lamellae will be larger than in front of the Al-lamellae, and Si will lead the growth front, causing a highly irregular solidification interface. Due to the anisotropy in the growth directions of the Si-phase, the interlamellar spacing operates within a range defined by the branching or termination of growth of the lamellae, due to constitutional effects. The Si-phase is therefore normally observed as a flake like structure. Additions of certain elements, such as Na, Sr or Sb, or rapid cooling may have a so- called ’modifying ’ effect on the eutectic, changing the silicon morphology into a rod like, fibrous, appearance. This is accompanied by a reduction of the eutectic temperature and an increase of eutectic composition. Without going too much into detail, this effect is probably attributed to a restricted growth of silicon, smoothing the solidification front. Growth of silicon becomes more isotropic. However, the exact explanation as to how and why this change in growth kinetics of silicon yields a fibrous structure is still being discussed[18]. The formation of the eutectic structure is, similarly to the primary phase formation, a nucleation and growth process. It has been suggested! 19] that the nuclei are different from those for a-Al, and that they instead aid the formation of the Si-phase upon whichthe eutectic grow in a radial fashion. In hypereutectic alloys it is established that phosphorous (P) helps nucleation of primary silicon. Figure 5 illustrates the complete solidification process for a binary hypoeutectic Al-Si alloy. The eutectic grain boundaries, which are the last to solidify, may differ from those of the primary phase. - 12 - In multicomponent alloys, phases may precipitate both pre- and proeutectically. A more complete treatment of eutectic growth is given by Elliott[20-21], Figure 5. Schematic illustration of the solidification sequence of a binary hypoeutectic Al-Si alloy[19]. 3.6 Phenomena accompanying solidification 3.6.1 Porosity Since most structural metals, either in thepure state or as alloys, have a closely packed crystal structure, i.e., face-centered cubic, close-packed hexagonal or body-centered cubic, the liquid will have a density lower than the solid. Consequently, all metals and their alloys contract upon solidification. Pure aluminium has a shrinkage of about 6.5 %[22], In aluminium alloys the volumetric shrinkage can range from 3.5 to 8.5 %. Silicon expands about 4 % upon solidification[23]. The obvious consequence of solidification shrinkage is that a quantity of metal that completely fills a mould cavity when it is molten will not do so after solidification. However, the problem is not solidification shrinkage in itself (although things would be easier without it). The problem is rather the location, i.e., position and distribution, of the shrinkage, which is determined by the mode and direction of solidification. As concluded earlier, most elements are enriched in the liquid during solidification - 13 - (k 500 550 600 650 700 750 800 850 900 Temperature [°C] Figure 6. Temperature-dependence of hydrogen solubility in pure aluminium at 1 atm hydrogen pressure[24]. Both shrinkage and gas causes porosity in the microstructure. It is, however, not realistic to distinguish between ’shrinkage porosity ’ and ’gas porosity ’, as is sometimes done, as if they were two entirely different kinds of defect. They are usually combined in the term ’microporosity ’ and their occurrence is related through their pressure dependence. - 14 - 3.6.2 How does the material behave in the partially solidified state ? Articles #1 through 4 are dealing with aspects of the rheological behavior of solidifying aluminium alloys. To briefly summarize the results, presented more thoroughly in the articles, the rheological behavior is strongly dependent on the solidification conditions. The boundaries of therheological behavior is, of course, the complete liquid above the liquidus temperature and the complete solid below the solidus temperature. Measurements of dendrite coherency as well as strength show that the dendrites are free-floating before dendrite coherency is reached and the material therefore still has a high mobility and low resistance to deformation. Then, when a dendrite network is formed at the dendrite coherency point, the dendrites interlock each other. Measurements of the strength of the dendrite network showed that the strength is strongly dependent on fraction solid, and at a given solid fraction it is significantly altered by changes of the dendrite size and morphology. In a dendrite network consisting of small, globular dendrites, strength develops more slowly since the deformation may be concentrated to the interdendritic regions where the strain is restricted to sliding of the dendrites past each other without extensive deformation and fragmentation of the dendrites themselves. The conclusion is that the point of and rate of development of mechanical strength may be altered by the solidification conditions, through dendrite size and morphology, solid fraction gradient etc., and that grain refinement both postpones the point where strength develops as well as the rate of strength development. 4. What is castability ? Castability is a very vaguely defined and comprehensive term describing the ability of the alloy to be cast without forming defects, and its ability to reproduce the mould pattern with the desired properties of the final casting. Since castability is so wide-ranging, it is usually separated into subgroups that incorporate one or more sources of defects and thereby is measured by certain laboratory experiments. Figure 7 summarizes the subgroups usually considered as a part of the definition of castability. - 15 - (a) Mould-filling (b) Fluidity (d) Macrosegregation (e) Hot-tearing Figure 7. Schematic illustration of the subgroups usually considered as parts of the definition of castability. 4.1 Mould-filling Mould-filling, as the word says, describes the ability of the metal to fill out the mould pattern and reproduce fine details of the mould. Mould-filling is strongly dependent of the surface tension between the mould and the metal, and is thereby affected by the mould characteristics and the melt composition and melt-cleanliness. Molten aluminium is always enclosed by an oxide film, which therefore is the surface actually in contact with the mould. The pressure in the metal (hydrostatic etc.) will have some effect on mould-filling. 4.2 Fluidity Fluidity is the ability of themolten metal to flow before it is stopped by solidification. Fluidity is evaluated by measuring the length a molten metal has flowed before it is stopped by solidification when it is poured into a channel with a small cross-sectional - 16 - area. The small cross-sectional area provides conditions for rapid cooling and relatively large temperature gradients. Fluidity hereby characterizes the ability of the alloy to be cast into thin sections. Articles #4 and 5 of this thesis focussed on the effect of grain refinement additions and the concentration of main alloying element on fluidity, which was evaluated by two different tests; spiral casting in sand and vacuum fluidity testing in Pyrex glass tubes. In the previous section it was noted that grain refinement postpones the point where the material stiffens with the formation of a dendrite network. Similarly, an increase of the content of the main alloying element will increase the constitutional supercooling in the melt and thereby establish a dendrite network earlier. Flemings [26] suggested on the basis of structural and visual observations that the flow is stopped when a coherent dendrite network is formed. Thus, an improved fluidity with increasing grain refinement and decreasing alloy content would be expected. With the limitations due to scatter in the measurements, the average fluidities agree with the theory, increasing with grain refinement, above a certain level of grain refinement. However, instead of a continuous improvement in fluidity with grain refinement, fluidity first decreased and then gave an increase in average fluidity. The results hereby showed no direct relationship to the rheological properties. However, the effect of varying the concentration main alloying element showed a direct relationship to the rheological properties, i.e., postponed strength development giving improved fluidity. Several explanations to this complex behavior obtained have been discussed, including the effect of grain refinement being on the limit of what is detectable in these tests. 4.3 Feeding Articles #2, 6 and 8 are considering feeding of Al-Si alloys. Feeding is the process where material movement occurs to compensate for solidification shrinkage. Figure 8 shows the feeding mechanisms in a directionally solidifying plate casting. The feeding process may be divided into five separate mechanisms, as observed in the Figure, and the transitions are coincident with the transitions in rheological behavior (which subsequently is determined by solidification conditions). The features of the mechanisms are discussed in greater detail in the articles. - 17 - liquid Mushy zone —►.solid Liquid Mass Interdendritic Solid Feeding Feeding Feeding Feeding Burst Feeding Figure 8. Schematic illustration of the five feeding mechanisms. At increasing solid fractions: liquid feeding (movement of a entirely liquid material), mass feeding (transport of a mixtureof solid and liquid), interdendritic feeding (transport of liquid through a stationary solid skeleton), burst feeding (collapse of solid skeleton resulting in movement solid and liquid and compaction of solid), solid feeding (high-temperature yielding of the solid). The measurements reported in Article #8 were directed towards investigating the relationship between rheological properties on the feeding process and the subsequent porosity amount and distribution in plate castings. Although there was big scatter in the measurements, some trends are indicated. The effectivity of the interdendritic feeding mechanism, where melt is transported through the dendrite skeleton is also strongly affected by the factors that determine the rheological behavior which adds extra complexity. The results indicate that dendrite coherency is not the decisive factor that strongly controls feeding and porosity formation. Dendrite coherency determines the extent of mass feeding, and a postponed coherency point would be expected to improve feeding, prolonging the range of mass feeding. - 18 - The balance between stress development, largely affected by the inherent resistance to interdendritic flow, and the strength development determines the occurrence of burst feeding; network collapse. The significance of the burst feeding mechanism is today not very well established. The results of the investigation presented in Article #8, as well as Article #7, show porosity distributions that indicate burst feeding. This suggests that burst feeding may be a mechanism that exerts a controlling influence on porosity formation in aluminium alloy castings. Today it is not possible to accurately predict burst feeding, which now (with strength development reported in Article #3) is mainly due to lack of accurate permeability data of the mushy zone. In addition it is necessary to develop a theory of the evolution of specific surface area of the solid during solidification and a fracture criterion of the dendrite network. Campbell [27] stated when he suggested the possibility of burst feeding that it would not be neither useful nor particularly controllable, but the measurements of rheological properties oppose this conclusion. Article #6 presents the general framework for a new model for predicting feeding and shrinkage porosity formation in addition to giving an overview of the existing efforts and models that are reported in the literature. The model is based on the results presented in this thesis and on the fundamental processes occurring during solidification. It has the potential of predicting shrinkage porosity formation and with later extensions include the effect of dissolved hydrogen and burst feeding. 4.4 Hot-tearing tendency The linear contraction upon cooling of the solid, in addition to solidification shrinkage, may cause significant stresses to develop in situations where the contraction of the casting is restricted. The casting may tear in regions wherethe strength is still low, i.e., regions which are not completely solidified (hot-spot). Measurements indicate that tearing may occur in regions where the solid fraction is 90-99 %[3], Similarly to burst feeding, strength measurements with a model of the stress development in the solid and a fracture criterion should allow predictions of hot-tearing. In principle there are many similarities between burst feeding and hot-tearing since they are both a question of the balance between strength and stress, as all fracture processes are. Contractionally induced stresses may develop in both cases, however being more important for hot-tearing, which thereby probably occurs at a higher level of stress. The tear itself develops in a region of low strength where stress concentration - 19 - is larger, and where the feeding mechanisms are barely operating. (Note: Regions in the microstructure where the dendrites have been tom apart in a similar fashion as in a hot-tear, but later filled with eutectic may be observed[3]. These are not classed as a hot-tear since there is obviously no tear. In this case no burst occurs due to the higher strength of the surrounding regions, which are at a lower temperature.) An important difference between hot-tearing and burst feeding is that hot-tears arise due to strain, while burst feeding is a compression induced fracture. This will give a difference in the fracture criterion. The conclusion to be drawn from this discussion is that a model of burst feeding, i.e., network fracture, should also be able to predict hot-tearing. Grain refinement is observed to reduce hot-tearing susceptibility [3], in good accordance with our measurements of the rheological behavior of grain refined materials. However, the ’nucleating ’ effect of inclusions may be more difficult to predict. 4.5 Macrosegregation Macrosegregation is areas of segregation on the scale of the casting itself. The mechanism of macrosegregation formation is a relative movement of solid and liquid, which are at different compositions. It is hereby related to the density differences caused by solidification and segregation, convection and interdendritic fluid flow. Burst feeding, clogging the interdendritic feeding path, may reduce the interdendritic flow- caused macrosegregation. However, an increase in porosity or reduced surface quality may result. 5. Conclusions This thesis has focussed on the relationship between the development of the equiaxed cast structure in aluminium foundry alloys, the factors that influence that, and the mechanical behavior of the material in the partially solidified state. The results obtained in this part has subsequently been related to some of the castability features, namely fluidity and feeding, of the identical alloys. Measurements of the dendrite coherency point has confirmed that dendrite network establishment may be rationalized through a consideration of grain size and dendrite growth rate, well fundamented in the theory of equiaxed dendrite growth. A vane-method has been developed to reproducibly and quantitatively measure the strength development in the mushy zone. The results indicate that the size and morphology of the developing dendrites and the fraction eutectic have a strong influence on the strength development. No strength is measured before dendrite coherency, and the coherency measured by the vane method corresponds well to that measured in the coherency experiments. It has been suggested that the strength data may be used to predict burst feeding. Measurements of fluidity give no direct relationship between fluidity and dendrite coherency, but a better correlation is observed when fluidity is compared to the rate of development of mechanical strength after coherency. Measurements of feeding indicate that burst feeding may be very significant to the development and distribution of porosity in Al-Si foundry alloys. No direct relationship between the rheological properties and porosity has been observed, most likely due to the fact that the interdendritic permeability is dependent on similar variables as the rheological properties. Dendrite coherency alone is not enough to predict porosity, although increasing the range of mass feeding. A model has been developed to predict porosity in Al-Si alloys. The model uses the information from this thesis, and will allow numerical testing of our hypotheses. - 21 - 5. References 1. M.C. Flemings: "Solidification Processing ", McGraw-Hill Inc., London, 1974 2. W. Kurz and DJ. Fisher: "Fundamentals of Solidification", Third Edition, Trans Tech Publications Ltd., Aedermannsdorf, 1989 3. J. Campbell: "Castings ”, Butterworth-Heinemann Ltd., Oxford, 1991 4. C.S. Smith: "The Early History of Casting, Molds, and the Science of Solidification", Inf. Proc. Second Buhl Int. Conf. on Materials: Metal Transformations, W.W. Mullins and M.C. Shaw (Eds.), Gordon and Breach Science Publishers Inc., London, 1968, pp. 3-52 5. J.W. Meier, "Nonferrous Metals Castings - Past and Future", Modem Casting, vol. 55, No. 6, 1969, pp. 97-112 6. D. Altenpohl: "Aluminum Viewed from Within", 1st Edition, Aluminium-Verlag, Dusseldorf, 1982 7. A. Pacz, U.S. Patent, No. 1387900, 1920 8. D. Apelian, "The Aluminum Casting Industry: From Foundry Art to Microstructure Controlled Processing", Processing Materials for Properties Conference, Hawaii, 7-10. Nov., 1993 9. V. Laxmanan, "Semi-Solid Forming for Lightweight Automotive Applications", TMS Materials Week 1995, Cleveland (Oh.), 1995, p. 117 10. C.S. Smith: "A History of Metallography ”, Paperback Edition, The MIT Press, London, 1988 11. J.W. Rutter and B. Chalmers, Can. J. Phys., vol. 31, 1953, pp. 15-39 12. W.A. Tiller, K.A. Jackson, J.W. Rutter and B. Chalmers, Acta Met., vol. 1, 1953, pp. 428-437 13. J.D. Hunt, Mat. Sci. Eng., vol. 65, 1984, pp. 75-83 14. L. Nastac and D.M. Stefanescu, Modeling of Casting Welding and Advanced Solidification Processes VI, T.S. Piwonka, V. Voller and L. Katgerman (Eds.), The Minerals, Metals and Materials Society, Warrendale (Pa.), 1993, pp. 209- 217 15. J. Zou and R. Doherty, Modeling of Casting Welding and Advanced Solidification Processes VI, T.S. Piwonka, V. Voller and L. Katgerman (Eds.), The Minerals, Metals and Materials Society, Warrendale (Pa.), 1993, pp. 193- 200 16. G. Chai, Ph.D. thesis, Chem. Comm., No. 1, Stockholm University, Stockholm, 1994 17. L. Backerud, L. Amberg and G. Chai, "Dendrite Coherency in Aluminum -22- Foundry Alloys", AFS, Des Plaines (111.), 1996 18. S.-Z. Lu and A. Hellawell, J. Met. (JOM), 1995, February, pp. 38-40 19. G. Laslaz, 4 ’th Int. Conf. on Molten Aluminum Processing, AFS, Des Plaines (111.), 1995, pp. 459-486 20. R. Elliott, Int. Met. Rev., 1977, September, pp. 161-186 21. R. Elliott: "Eutectic Solidification Processing. Crystalline and Glassy Alloys", Butterworths & Co Ltd., London, 1983 22. Metals Handbook: vol. 2 Properties and Selection: Nonferrous Alloys and Pure Metals, Ninth edition, ASM International, Metals Park (Oh), 1988 23. Properties of Silicon, INSPEC, The Institution of Electrical Engineers, 1988 24. E.L. Rooy: "Aluminum and Aluminum Alloys", Metals Handbook, vol. 15 Casting, Ninth Edition, ASM International, Metals Park (Oh), 1988, pp. 743-770 25. T.A. Engh: "Principles of Metal Refining", Oxford University Press, Oxford, 1992 26. M.C. Flemings, 30’th Int. Foundry Congress, Praha, 1963, pp. 63-81 27. J. Campbell, British Foundryman, 1969, April, pp. 147-158 - 23 - Article #1 A.K. Dahle and L. Amberg "On the Assumption of Additive Effect of Solute Elements in Dendrite Growth" submitted to Metallurgical and Materials Transactions On the Assumption of Additive Effect of Solute Elements in Dendrite Growth A.K. Dahle and L. Amberg Norwegian Institute of Technology, Department of Metallurgy, N-7034 Trondheim, Norway The individual effect of some common alloying elements in aluminium-silicon casting alloys has been investigated by measuring the dendrite coherency point. In the existing theories for dendrite growth it is usually assumed that the effect of the solute elements is additive, and one purpose of the present investigation is therefore to investigate this approach. The effect of successive additions of Mg, Fe and two levels of grain refinement, by AlTi5Bl, on dendrite coherency has been investigated in two AlSi alloys; AlSi7 and AlSill. By changing the order of addition of these elements, the assumption of additive effect is verified. The results show that two separate parameters are altered with the alloying additions: the grain size and the dendrite growth rate, both being very important for the solidification characteristics. Unless the effect on grain size is taken into account, the assumption of additive effects is not correct. However, when the effect on grain size is considered separately, there is a good qualitative correspondence between the theoretical predictions and the experimental results. L INTRODUCTION Adding alloying elements and controlling the chemical composition is an important step in the production of aluminium alloy castings. The alloying elements affect important properties such as the casting characteristics, strength, ageing behavior and corrosion resistance. Alloys based on the Al-Si system, such as 356, are the most important commercial casting alloys today. Iron is usually present in most aluminium alloys, since it is virtually impossible to produce iron-free primary aluminium during the electrolysis and oxide production. A large tonnage of casting alloys in the foundries is produced from recycled metal, and iron is also here usually among the main impurities. Magnesium may be added to improve the alloys ’ response to thermal treatment. In most aluminium casting alloys, especially those grain refined, equiaxed dendrite - 25 - growth is frequently observed. These dendrites nucleate separately in the melt, and grow and move independent of each other in the first stages of solidification. Later a point in the solidification interval is reached, termed the dendrite coherency point, where the primary dendrite arms impinge onto the neighbouring dendrites and forms an interdendritic network. Solidification after this point is mainly by growth of secondary arms, coarsening, and often a final eutectic precipitation. Modeling of the development of microstructures and cooling curves in alloys with dendrite growth has been performed by several investigators! 1 -10]. In these models the dendrite coherency point presents itself as a suitable measure, being a detectable point in the solidification interval related to a physically significant change in solidification character. Since the dendrite coherency point provides information on the kinetics of primary dendrite growth, models predicting equiaxed dendrite growth should show a correlation with the experimentally determined coherency points. The dendrite coherency point can be explained by considering the dendrite growth conditions, giving an inherent link to these models. In the proposed models for equiaxed dendrite growth, the effect of each solute element is assumed to be additive. The individual effect of alloying additions, by successive alloying, on dendrite coherency has been investigated in this paper. The effects of adding magnesium, iron and two levels of grain refinement have been investigated in two important Al-Si casting alloys, AlSi7 and AlSill. The experimental results are then compared to the predictions of a recently proposed theory for dendrite coherency. H. PREDICTION OF DENDRITE COHERENCY Since the solubility of most alloying elements in aluminium alloys is larger in the liquid than in the solid, they are rejected by the solidification front when growing under nonequilibrium conditions. This effect produces a thin diffusion layer in front of the solidification interface, assumed to have an extent in the range of 5C~D/V [11], where 5C is the thickness of the diffusion layer, D is the diffusion coefficient in the liquid and V is the growth rate. This layer is responsible for creating the constitutional supercooling in front of the growing solid, which is the driving force for instability during growth. The diffusion layer follows the contour of the solidification front, i.e., in front of dendrite tips, arms and depressions. From a microscale theoretical viewpoint, dendrite tips and arms advancing towards each other should decrease their - 26 - growth rate and finally stop growing as the diffusion layers start to overlap. This is a consequence of eliminating the constitutional supercooling and creating a layer with a uniform and high concentration of solute. However, in the practical solidification situation convection currents in the liquid decreases the size of the diffusion layers, hydrostatic pressures and mechanical stresses are often operating during solidification and the dendrites often move due to convection or gravity. These effects may bring the dendrites into thermal and mechanical contact, making them coalesce through a reduction of surface energy. In this way a dendrite network is created in the partially solidified alloy melt. Dendrite coherency is strongly dependent on the two primary processes involved in solidification - nucleation and growth. The nucleation process determines the number of dendrites and the distance between them, i.e., the growth distance for dendrite coherency. This also affects the second important process, dendrite growth. The dendrite growth rate, as affected by constitutional and cooling conditions, determines the point of impingement. Chai et al. [12-16] have suggested that dendrite coherency is a result of a competition between these processes. Fraction solid at dendrite coherency has been shown to be proportional to the ratio d/V, where d is the average grain size and V is the dendrite growth rate. It is now well established that the primary aluminium phase in commercial aluminium castings nucleates on heterogeneous substrates added through master alloys containing Ti-B or B, although the exact mechanisms are still discussed[16-18]. Addition of sufficient quantities of good, potent, grain refiners are expected to give simultaneous formation of all the dendrites, possibly without any further significant nucleation and recalescence. When no extra nucleating additions are made, dendrites are expected to nucleate over a range of temperatures, as the energy barrier of more substrates are overcome, displaying a large period of recalescence. The last few years have seen the development and improvement of several analytical and theoretical models on dendrite growth, and especially the complex case of growth of equiaxed crystals has received increased attention[l-10], A common way to test the predictions of these models is to compare calculated and experimental cooling curves for alloys with different compositions. The first models of equiaxed dendrite growth used the concept of spherical grains or grain envelopes[l-8]. The early model of Maxwell and Hellawell[l] assumed - 27 - spherically growing dendritic grains, and managed to reproduce the early stages of nucleation and recalescence. Later Hunt[2] relaxed this assumption by suggesting that the internal fraction solid was given by the Scheil equation. Dustin and Kurz[3] assumed a constant internal fraction solid of 0.3 and ignored overall solute balances. This model gave a quite good reproduction of cooling curves and also predicted features such as number of grains and recalescence. The models of Rappaz and Thevoz[4-5] included overall thermal and solute balances, neglecting back diffusion and assuming complete mixing in the interdendritic liquid. In this model the internal fraction solid was suggested to depend on the Peclet number of the dendrite and the melt supersaturation, i.e., (1) 4P where f, is the internal fraction solid within the grains, fs is total fraction solid, fg is fraction grains, Q. is the supersaturation, Pc is the Peclet number(=VR g /2D), Rg is the instant radius and D is the diffusion coefficient in the liquid. Treating the multicomponent aluminium alloys as dilute mixtures, and assuming no interaction between solute atoms, it has been suggested[8,19] that f; can be calculated as the sum of the contribution to theinternal fraction solid from each solute atom. Mathematically thisinvolves summing the constitutional term of each component. For the condition of growth with a low Peclet number, i.e., 8 « Rg , or neglecting the solute diffusion layer ahead of the dendrites, the internal fraction solid can be approximated to £2, or equivalently XQ for a dilute multicomponent alloy. At the dendrite coherency point it is expected that the grain size is given, i.e., fg =l and Rg =d/2. The fraction solid at the coherency point would hereby be given as fs=fj. This would lead to a constant fraction solid at the coherency point of 0.3 from the model of Dustin and Kurz[3], Based on the equation for the undercooling ahead of the growing dendrite tip, the morphological stability criterion and assumptions of constant temperature gradient ahead of the growing dendrite and growth under steady-state conditions at a given undercooling, Chai et al.[13,15] derived equations for the growth rate and dendrite grain radius. Fraction grains and the internal fraction solid could hereby be calculated from: - 28 - VAT v = and (2) and (3) (i+4cr2) 3T Fraction grains is given by: 4pAT3 3 (4) 3(1 +4a* I)77? m( with the alloy constant p, for all elements i, defined as -i Zm,Cj(t,-l) i (5) D and the internal fraction solid could be calculated from equation (1). a* equals (2tu)"2, F is the Gibbs-Thomson coefficient, m is the slope of the liquidus-Iine, C,* is the composition at the dendrite tip, k is the equilibrium partition coefficient, Lv is the volumetric latent heat and K is the thermal conductivity of the liquid. Table I gives the physical properties used in the calculation for Al-Si alloys. Table I. Physical Constants used for Al-Si alloys. [4,25-26] Symbol Meaning Value m slope of liquidus line [K/wt%] -6.6 k equilibrium partition coefficient 0.117 D diffusion coefficient in the liquid[m 2/s] 3-10"9 K thermal conductivity of the liquid[W/m K] 95 r Gibbs-Thomson coefficient [K m] 1.96*10 ‘7 Lv volumetric latent heat [J/m3] -9.5*10" Figure 1 shows how the internal fraction solid(fj) and fraction grains(f g ) changes with temperature in an Al-7Si alloy, calculated from the above equations (1-5) and values from Table I under equilibrium conditions. The grain size was taken from T0ndel[2O], Nastac and Stefanescu[8] converted growing equiaxed dendrites into spheres with -29- equivalent volume and number of instabilities. A volume-averaged melt composition was used to deter mine the growth of the instabilities of the sphere. The evolution of fraction solid was directly related to the radius of the sphere, the number of instabilities and the tip growth rate. Growth was calculated until the average composition equals the eutectic composition, Ce, or fs=l, i.e., Rg =Rtot. This gave: % = 3 A_y (6 ) dt where Rg is the instant grain radius, Rtot is the radius of the grain and ^ denotes the average growth velocity. Putting ^ = the evolution of fraction solid with time could be calculated. Unfortunately, this model cannot predict dendrite coherency. At the dendrite coherency fraction solid it is expected that Rg =R,ot , which is an impossible assumption with this model, since this is only possible for fs=l. fgorfi 600 602 604 606 608 610 612 614 Temperature (°C) Figure 1. Calculated evolution of the internal fraction solid (f) and fraction grains (fg ) vs. dendrite tip temperature for an Al- 7wt% Si alloy. Cooling rate lK/s and grain size d= 3.1 mm. Equilibrium conditions. Wang and Beckermann[9] have proposed a rigorous model of equiaxed dendrite growth with a multiphase approach, relaxing the spherical dendrite constraint and accounting - 30 - for the different length scales in the dendrite structure. This model converges towards the models of Rappaz and Thevoz[4-5] if adjusted to the same diffusion approximations, i.e., no back diffusion in the solid and complete interdendritic mixing, and a spherical dendrite envelope. This model has been further developed by Wang and Beckermann[10] to include the effect of convection and relative movement of both solid and liquid, neglecting the effect of solidification shrinkage, for an Al-4Cu alloy. III. EXPERIMENTAL PROCEDURE Two experimental methods have been utilized to measure dendrite coherency; rheological measurements and thermal analysis. Two pure binary Al-Si alloys, with 7 and 11 wt% Si, were made by mixing high-purity aluminium and commercial purity silicon. To these alloys were successively added magnesium, by commercial purity Mg, iron, by an Al-7.5 wt% Fe master alloy, and two fractions of grain refiner added by Al-5Ti-lB master alloy rods. The level of addition of each element was chosen such that the end composition reproduced the composition of two commercial alloys previously investigated[21,22]. The compositions of the two commercial alloys are shown in Table II. An additional factor to be investigated was if the order of addition of Mg and Fe had any influence on dendrite coherency. Table II. Main constituents in the commercial Al-Si alloys[wt%]. Alloy Si Mg Fe Ti Al Al- 7% Si 6.80 0.174 0.15 0.129 bal. Al- 11% Si 11.20 0.172 0.13 0.117 bal. About 3 kilos of Al-Si alloy was made in an electric resistance furnace, and all alloying additions were made directly to the melt. This served as the melt reservoir. Two minutes after the melt had been stirred for 30 seconds, a melt sample was taken from the charge in the holding furnace by a cylindrical graphite crucible with sample dimensions, diameter 030 mm, height 40 mm and wall thickness 10 mm. For the rheological measurements the graphite crucible was put on a stand within an electric resistance furnace. A preheated steel paddle was inserted in the centre and a - 31 - type K thermocouple, previously calibrated towards pure aluminium, was inserted adjacent to the wall, both 10 mm into the melt sample. Figure 2 a) shows this experimental setup. The furnace was removed and the system cooled naturally in air, giving a cooling rate of ~0.7 K/s. A fiberfrax lid was put on top of the crucible to reduce heat radiation from the melt surface. During cooling, the torque necessary to rotate the paddle at a constant rotation speed of 0.05 rpm, and the temperature was monitored by a computer. Figure 2 b) shows an example of the measured parameters during one experiment. (%J Torque furnace Figure 2. a) Experimental setup for rheological measurements. b) Example of the lapse of the recorded parameters, temperature and torque, during rheological measurements for an Al-7%Si alloy, grain refined with 0.05% Ti as AlTi5Bl-rods. The dendrite coherency point is identified as the point where the torque increases sharply. For the thermal analysis experiments the graphite crucible was placed on a stand. Two thermocouples, one adjacent to the wall and the other in the centre, were immersed 10 mm from the bottom. Fiberfrax felt was put below and on top of the crucible. The system cooled naturally in air, and the temperatures from the two thermocouples were continuously recorded with a computer. Figure 3 a) and b) show the experimental setup -32- and measured parameters, respectively. Fraction solid was calculated from the temperature readings by a procedure developed by Tamminen[23]. Tw-Tc Figure 3. a) Experimental setup for thermal analysis experiments. b) Example of the measured parameters during such an experiment. Tw is the temperature close to the wall and Tc is the temperature in the centre. The first minimum in the Tw-Tc curve after nucleation is identified as the dendrite coherency point. The dendrite coherency point was in the rheological measurements identified by the point where the torque increased sharply due to the paddle being frozen in, and locked, by the developing interconnected dendrite network. In the thermal analysis experiments it was defined by the first minimum in the curve for the temperature difference between wall (Tw) and centre (Tc), i.e., aT=Tw-Tc. At the dendrite coherency point the heat flow-rate increases sharply when heat can be conducted through the more or less continuous solid network, since the thermal conductivity of the solid is larger than that of the liquid. The coherency data reported are the average of at least three tests, each performed with a new melt sample. Grain sizes were determined by the linear intercept method on anodized samples cast at similar conditions as in the coherency experiments. Samples were cast for chemical analysis by optical emission spectroscopy from each melt charge after every new alloying addition. - 33 - IV. RESULTS AND DISCUSSION The purpose of the experiments was to investigate the individual effect of alloying elements, i.e., changes in the constitutional conditions, on dendrite coherency, and whether the effect of each element was additive. Finally the predictions of the previously described model should be tested with the same alterations in composition and grain size. Figures 4 a) and b) show how fraction solid at dendrite coherency and grain size change as the pure binary Al-Si alloys are successively added iron, magnesium and two levels of grain refinement. The fraction solid at dendrite coherency are from rheological measurements (fsc0h~rheology) and thermal analysis (fscoh-therm.an.). The resulting effects are very similar in both alloys. The iron addition increases the grain size, and seems to give a small decrease or an unchanged fraction solid at dendrite coherency. The successive magnesium addition gives a slight reduction in the grain size and an increase in fraction solid at dendrite coherency. As expected, grain refinement decrease the grain size and postpones impingement. a) b) AlSi7 AlSill — --o - 1600 —j— fscoh-rheology —O— fscoh-therm.an. —0 - Grain size Al-7%Si +0.14%Fe +0.20% Mg +0.05% Ti +0.12% Ti Al-11% Si +0.14%Fe +0.16% Mg +0.06% Ti +0.15% Ti Figure 4. Effect of successive additions of iron, magnesium and two levels of grain refinement on the grain size and fraction solid at dendrite coherency, determined by rheology and thermal analysis in a) Al- 7% % and 6; AZ- 11% A". f0.7 AAf. - 34- Figures 5 a) and b) presents results where the same experiments were repeated, except for a change in the order of addition of iron and magnesium. The magnesium addition increases fraction solid at dendrite coherency and gives a reduction in grain size in the Al- 11% Si alloy. The following iron addition increases the grain size significantly in both alloys and the dendrites meet earlier, at a lower fraction solid. Grain refinement has the same effect as above. a) b) AlSi7 AlSill - 1800 [urn] size 800 a Grain —|— fscoh-rheology - 600 —Q— fscoh-therman. - —|— f s coh- rheology —()• - Grain size ■ —0— fscoh- therm.an. . —0 - Grain size V1 I I I------low Al- 7% Si -HX18% Mg +0.16%Fe +0.07% Ti +0.10% Ti Al-11% Si +0.16% Mg +0.13%Fe +0.04% Ti +0.09% Ti Figure 5. Effect of successive additions of magnesium, iron and two levels of grain refinement on the grain size and fraction solid at dendrite coherency, determined by rheology and thermal analysis in a) Al- 7% Si, and b) Al- 11% Si. [0.7 K/s]. Compared to Figure 4, the order of addition of iron and magnesium has been changed. Grain refinement is an effective way to increase grain density, as observed in Figures 4 and 5. The larger number of grains grow at a lower rate because of the large growth restricting effect of titanium and because of larger supersaturations of alloy elements in front of the dendrite tips. The diffusion fields of the dendrites are larger and overlap at an earlier stage of growth, and the dendrites impinge later. In the specifications of commercial aluminium alloys, compositions of individual alloying elements may vary within a limited range. These variations, of course, have an effect on types and amounts of precipitates, but also the solidification behavior is affected. Backerud et al.[24] investigated composition variations in 380-type alloys and - 35 - observed significant effects on liquidus and eutectic temperature. This also suggests that the point where the dendrite network is established would be shifted. Nastac et al.[8] and Zou et al.[19] have suggested that the growth restricting effect of alloying elements in front of the growing dendrite tips are additive when the elements are present in dilute amounts. In the experiments reported in Figures 4 and 5, both iron and magnesium fulfil this requirement. According to equation 2, addition of these elements should hereby, due to the additive effect, result in a reduced dendrite growth rate since both elements have a growth restricting effect. If the results of iron and magnesium additions in Figures 4 a) and 5 a), for Al- 7% Si, and in Figures 4 b) and 5 b), for Al- 11 % Si, are compared, they seem to be contradictory. The iron addition does in both cases reduce the coherency fraction solid, while magnesium postpones coherency, independent of the sequence of addition. The results do therefore as a first approximation not support the theory of additive effects of these elements, since in that case fraction solid at dendrite coherency should also be increased by the iron addition. Table III shows the growth restricting factors of the elements relevant for this investigation. It may be observed that iron has the smallest growth restricting factor, while Ti has a very large effect. Figure 6 shows how the growth restricting factor, with the additive approximation, changes for each alloying addition. For every alloying addition an increase in the growth restricting factor is observed, which indicates a decrease in the dendrite growth rate. A very likely explanation to the observed discrepancy between the additive approximation and the experimental results is the increase in grain size observed with the iron addition. Dendrite coherency is as mentioned earlier a compromise between two mutually dependent processes; nucleation and growth. Both the grain size and the dendrite growth rate determines the point of impingement, and ignoring one of these variables makes it impossible to explain the coherency point. The growth restricting effect of iron is less than that of magnesium, which suggest that the growth rate is less changed with the iron addition. The effect of magnesium, on the other hand, agrees better with the assumption. Fraction solid at dendrite coherency is increased with its addition, probably also because of the increased grain density. Table HI. Alloy dependent constant in the growth restricting factor: m(k-l). Element m(k-l) Fe 2.9 Mg 4.0 Ti 246 b) AlSill Al-11%Si -K).14%Fe +0.16% +0.06% Ti +0.15% Ti Figure 6. The effect of each alloying addition on the growth restricting factor, 'LmCfk-l), in a) Al- 7% Si, and b) Al- 11% Si. A significant difference in fraction solid at dendrite coherency(fsroh) between Al- 7% Si and Al- 11% Si may be observed. This effect is mainly due to the difference in grain size, and the constitutional and growth conditions that determine the dendrite tip radius. Based on the chemical analysis of the samples and the main alloying elements Si, Fe, Mg and Ti, the experimentally determined coherency fraction solid (fscoh) is plotted vs. the inverse product of the grain size (d) and the growth restricting factor (£mC0(k-l)) in Figure 7. According to equation 2 the growth rate is inversely proportional to the growth restricting factor on the dendrite tips. But, as observed in Figures 4 and 5, addition of these elements also have a strong effect on the grain size. Figure 7 show - 37 - that the results correlate with the parameter l/XmC0(k-l)d. The reason for this behavior is probably that grain size has a very strong effect on dendrite coherency, and that the dendrite growth rate should be related to the effective concentration at the dendrite tips (Ci"), which is different from the initial content (C0). Also some of the changes measured with alloying in theseexperiments are on the limit of what is detectable with the techniques employed. 30 25- o 0 * 9 - O o + 20- 0+ 0 b. 5 -r - + + 1 15- =v ° *2 5 10- - . O o' 5- “T fscoh (rheology) - O fscoh (therm an.) 0 " 0.000000 0.000010 0.000020 0.000030 0.00004 1 Figure 7. Relationship between the measured fractions solid at dendrite coherency vs. the inverse product of the growth restricting factor and grain size. The model predictions, eqs. 1-5, of the experiments presented in Figures 4 and 5 are shown in Figure 8 and 9, respectively. The calculations are based on the measured grain sizes, a cooling rate of 1 K/s and the concentration of the elements Si, Fe, Mg and Ti. When the model predictions and the experimental results are compared, it may be observed that the model consistently predicts a lower fraction solid at dendrite coherency than what is experimentally observed. Some of the changes observed with alloying are qualitatively reproduced by the model, but the predicted numerical values are far from the ones observed. This investigation therefore indicates that the present theory for prediction of dendrite coherency is far from complete and needs further development. -38- a) f s coh (%) fscoh(%) Al- 7% Figure Figure Si +0.14%Fe 8. 9. AlSi7 40.20% Calculated Calculated before added Mg before magnesium 40.05% effect effect Ti iron 40.10%Ti of of presented presented the the - 39 successive successive - b) Al- in in Figure 11% Figure alloy alloy Si 40.14%Fe additions additions 5. 4. AlSill 40.16%Mg with with magnesium iron 40.06% added Ti 40.15% Tf V. SUMMARY AND CONCLUSIONS The individual effect of alloying elements on primary dendrite growth has been investigated by determining the dendrite coherency point, the point where the primary dendrite arms impinge on the surrounding dendrites. The effects of iron, magnesium and two levels of grain refinement were investigated in two Al-Si alloys with 7 and 11 wl% Si. Successive additions of alloying elements, with alternated order, have an effect on dendrite coherency that correspond to their impact on the grain density and dendrite growth rate. The addition of magnesium fit better than iron the approximation of additive effect of alloy elements giving a reduced dendrite growth rate, since iron affects the grain size more than magnesium. The important conclusion to be drawn from this investigation is that the assumption of additive effect of alloying elements is an incorrect approximation unless the effect on the grain size is considered. The model proposed for predicting dendrite coherency reproduces some of the tendencies observed experimentally, but is best suited for giving a qualitative understanding instead of exact predictions. There is still a need for good models to predict dendrite coherency more accurately. ACKNOWLEDGEMENTS The financial support from the Norwegian Research Council, Elkem Aluminium and Hydro Aluminium is gratefully acknowledged. REFERENCES 1. I. Maxwell and A. Hellawell, Acta Met., 1975, vol.23, pp. 901-909 2. J.D. Hunt, Mat. Sci. Engng., 1984, vol.65, pp. 75-83 3. I. Dustin and W. Kurz, Z. Metall., 1986, vol.77, pp. 265-273 4. M. Rappaz and P. Thevoz, Acta Met., 1987, vol.35, pp. 1487-1497 5. M. Rappaz and P. Thevoz, Acta Met., 1987, vol.35, pp. 2929-2933 6. C.S. Kanetkar and D.M. Stefanescu, AFS Trans., 1988, vol.60, pp. 591-598 7. M. Rappaz, Int. Mat. Rev., 1989, vol.34, No.3, pp. 93-123 - 40 - 8. L. Nastac and D.M. Stefanescu, Modeling of Casting, Welding and Advanced Solidification Processes VI, Eds. T.S. Piwonka, V. Voller and L. Katgerman, The Minerals, Metals & Materials Society, 1993, pp. 209-217 9. C.Y. Wang and C. Beckermann, Met. Trans. A, 1993, vol.24A, pp. 2787-2802 10. C.Y. Wang and C. Beckermann, Light Metals 1995 11. R. Trivedi and W. Kurz, Acta Met., 1994, vol.42, pp. 15-23 12. G. Chai, T. Rplland, L. Amberg and L. Backerud, Processing of Semi-Solid Alloys and Composites, Eds. S.B. Brown and M.C. Flemings, MIT, 1992, pp. 193-201 13. G. Chai, Ph.D. thesis, Stockholm University, Chem. Com., No.l, (1994) 14. L. Arnberg, G. Chai and L. Backerud, Mat. Sci. Engng., 1993, A173, pp. 101- 103 15. G.Chai, L. Backerud, T. Rplland and L. Amberg, Met. Mat. Trans. A, 1995, vol.26A, pp. 965-970 16. D.G. McCartney, Int. Mat. Rev., 1989, No.5, vol.34, pp. 247-260 17. P. Schumacher and A.L. Greer, The 4th Int. Conf. on Aluminum Alloys, Eds. T.H. Sanders Jr. and E.A. Starke Jr., Atlanta, GA, 1994, pp. 730-737 18. P.S. Mohanty and J.E. Gruzleski, Acta Met. Mat., 1995, vol.43, pp. 2001-2012 19. J. Zou and R. Doherty, Modeling of Casting, Welding and Advanced Solidification Processes VI, Eds. T.S. Piwonka, V. Voller and L. Katgerman, The Minerals, Metals & Materials Society, 1993, pp. 193-200 20. P.A. Tpndel, Ph.D. thesis, Norwegian Institute of Technology, 1994 21. A.K. Dahle and L. Amberg, The 4th Int. Conf. on Aluminum Alloys, Eds. T.H. Sanders Jr. and E.A. Starke Jr., Atlanta, GA, 1994, pp. 91-98 22. A.K. Dahle and L. Amberg, Sintef Report, No. STF34 A93248, Trondheim, Norway, 1993 23. J. Tamminen, Ph.D. thesis, Stockholm University, Chem. Com., No.2, 1988 24. L. Backerud, G. Chai and J. Tamminen: Solidification characteristics of aluminium alloys, vol. 2, Foundry alloys, AFS/SkanAluminium, 1990 25. W. Kurz and D.J. Fisher: Fundamentals of solidification, Third Edition, Trans Tech Publications, 1989 26 . M. Gundtiz and J.D. Hunt, Acta Met., 1985, vol.33, pp. 1651-1672 - 42 - Article #2 A.K. Dahle and L. Amberg "Rheological Properties of Solidifying Aluminum Foundry Alloys" accepted for publication in Journal of Metals(JOM), March, 1996 Rheological Properties of Solidifying Aluminum Foundry Alloys Ame K. Dahle and Lars Amberg Increasing the fundamental understanding of the mechanisms of defect formation is essential for further optimization of cast products. The origin of most casting defects in aluminum alloys is usually explained by the solidification, segregation and material transport processes that occur while the casting is in a semi solid state. The semi solid region, often termed the mushy zone, is in most aluminum foundry alloys characterized by equiaxed dendrites and interdendritic liquid. The rheological properties of the mushy zone, which describes the deformation and flow properties, is determined by the microstructural development, and plays an important role for the efficiency of the transport processes occurring to compensate for density differences, including solidification shrinkage. INTRODUCTION The foundry industry today is experiencing a growing demand for optimizing existing alloys and processes as well as developing new and improved castings. This is a result of an increasing market for high-quality, cost and energy efficient products. To be able to meet these expectations there is a need, and room, for improvements at several levels in the production, such as casting design, molding and melt treatment. This has necessitated further fundamental elaborating investigations of the inherent characteristics of the mechanisms that cause defect formation. Material movement and fluid flow are important and critical features of all casting processes. The shrinkage upon solidification of the aluminum foundry alloys imposes a movement of material (liquid and/or solid) within the casting to avoid porosity. Porosity is an undesirable defect in aluminum alloys since it is detrimental to the mechanical properties, such as tensile strength and fatigue resistance, and because it causes a reduced reproducibility of the castings, increasing the scrap-rate and costs for the foundry. The process where material moves to compensate for shrinkage is termed feeding. A pressure differential driving flow toward the solidifying regions is generated during solidification, and the transport mechanisms occurring in the mushy-zone are the most critical to control and understand. Figure 1 illustrates the five mechanisms that are - 43 - envisioned to occur to feed metal in a mushy-freezing alloy, as summarized by Campbell1,2. Liquid feeding denotes transport of the molten alloy; Mass feeding is transport of a melt with dispersed equiaxed dendrites; Interdendritic feeding denotes the flow of liquid through a solid skeleton of interconnected crystals; Solid feeding covers the high-temperature deformation of fully solid regions; Burst feeding is a mechanism that involves thebreakdown of the dendritic network, subsequently giving further transport of a collapsed or fragmented solid phase with the liquid. Porosity is also caused by liberation of dissolved hydrogen during solidification, often on dispersed inclusions in the melt. However, the formation of gas induced porosity requires a certain pressure, and this is determined by the efficiency of the feeding mechanisms. Feeding Feeding Feeding Feeding Burst Feeding Figure 1. The feeding mechanisms of a mushy freezing alloy in a plate casting with a chill on the far end imposing a positive temperature gradient towards the feeder. To obtain a proper understanding of these processes, both conceptually and for subsequent modeling purposes 3, a detailed knowledge of the properties and characteristics of each mechanism is required. This article discusses the evolution of the rheological properties during solidification with reference to the feeding mechanisms illustrated in figure 1. - 44 - The Liquid Region Above the liquidus temperature of the alloy the material is essentially a pure melt (neglecting dispersed exogenous particles). Molten metals behave like Newtonian fluids, i.e., t = t| — . The viscosity for pure aluminum at the melting point is about dx 1.2xl0" 3 [Pa-s]4. Viscosity decreases with increasing temperature according to an Arrhenius type of relationship, i.e.4, where T| is viscosity and T is the absolute temperature. The viscosity of aluminum alloys is not expected to deviate much from that of pure aluminum, however this needs to be verified. It must be emphasized that this description does not cover the mold filling stage, where flow is also limited by the properties of the enclosing oxide film. The Mass Feeding Region As temperature drops below the nucleation temperature of most aluminum foundry alloys, especially those that are grain refined, equiaxed dendritic crystals start to form. The mixture of liquid and dispersed crystals will exhibit an apparent viscosity which is larger than of the pure melt due to interactions between the dendrites, and between the dendrites and the liquid. The viscosity of this semisolid slurry increases with decreasing temperature, since the volume fraction solid increases and the dendrites interact more with each other. The slurry can be expected to show a shear-rate dependent behavior, i.e., non-Newtonian, probably with a decrease of apparent viscosity with increasing shear-rate as in pseudoplastic materials. This is a result of fragmentation of the dendrites, transforming them into a more globular morphology. Mori and Ototake 5 have proposed the following equation: d(T)Sr(T) •ncCD = 1 + 2(l/fs(T)-llfsoh) - 45 - where T|a is the apparent viscosity, p, is the viscosity of the liquid, d is the grain size, Sr is the specific surface area of the solid, fs is fraction solid and fscoh is the fraction solid at dendrite coherency (see below). Dendrite Coherency The growing equiaxed dendrites soon start interacting more strongly with each other as the space between them decreases and they become more entangled with increasing solid fractions. A characteristic, measurable, alteration occurs at the so-called dendrite coherency point. Here a dendrite network consisting of bridged equiaxed crystals is formed. The establishment of this more or less continuous network is strongly dependent on the solidification conditions. Previous work has shown that there is a very strong link between the kinetics of dendritic growth and the temperature or fraction solid where coherency occurs 6'9. Fraction solid at dendrite coherency usually falls between 10 and 50% solid. Dendrite coherency occurs at about 20% solid for an unrefined 357-type alloy, and it may be postponed to -30% with good grain refinement as observed in Figure 2 a). Dendrite coherency has been shown to depend on both the dendrite growth rate, V, and the grain size, d. The effect of altering the solidification conditions on dendrite coherency is hereby explained by considering the effects on grain size and dendrite growth rate. (Details about the experimental techniques for coherency determination are described elsewhere6'9). Dendrite growth rate. V: There are several ways to change the dendrite growth rate. Increasing the cooling rate increases the dendrite growth rate since the rate of latent-heat removal increases, but the grain size is also affected. Figure 2 b) shows that increasing the cooling rate in the 357-type alloy decreases the coherency fraction solid. Dendrite growth rate has been suggested 10,11 to be inversely proportional to a growth restricting factor ( ^ mffa-Y) ), where m is the i slope of the liquidus line, C is the concentration, and k is the partition coefficient of the i’th element. With m<0 and k - 46 - Grain size, d: Decreasing the grain size by adding grain refiners results in a larger number of dendrites, each growing at a lower rate. Although the separation of the crystals is decreased, the effect of solute enrichment dominates, and fraction solid at dendrite coherency increases. This is observed in figure 2 a). a) b) 1400 Model alloys 25 - ------700 1200 1000 [nm] 15 - _ 600.= size —H— fscoh (rheology). -©- fscoh(n*xm.ST 800 —0 - Grain size Grain 600 f s coh- rheology 400 1.0 1.5 2.0 0.00 0.05 0.10 0.15 0.20 0.25 0.30 SiBloy Cooling rate (K/s) Ti-content [wt%] Figure 2. (a) The effect of increasing grain refinement, by addition of AlTiSBl or ~150 ppm boron only (SiBloy), on grain size and fraction solid at dendrite coherency, determined by thermal analysis and rheological measurements, in an A357-type alloy; (b) The effect of cooling rate on grain size and coherency fraction solid in an A357-type alloy, grain refined with 0.05% Ti. Interdendritic Feeding The resistance to mass transport is significantly increased by the formation of a dendrite network. From this point on flow has to occur between the dendrites. Dendrite coherency hereby defines the transformation from mass feeding to interdendritic feeding. Flow through the dendrite network is controlled by the interdendritic permeability which depends on solid fraction and solid surface area, where the solid fraction gradient and dendrite morphology are strongly dependent on the cooling conditions, nucleus density and constitutional conditions. Experiments have confirmed that interdendritic flow is similar to flow through other porous media 12,13. The - 47 - permeability, K, may be approximated by the Kozeny-Carman equation 14: r- 5(S,// and flow through this region is calculated from the Darcy equation 15: dP _ vti ,/, a % where v is the flow rate of the liquid, f, is the liquid fraction and P is pressure. However these equations are only approximations, and reliable measurements of the permeability and its influencing factors are requisite. Burst Feeding The interdendritic permeability decreases with increasing solid fraction, hereby increasing the pressure required for further flow. In this way, the stresses on the dendrite network are also increased, and the network will collapse if the generated stresses exceed the network strength. This may occur several times during solidification, as observed microstructurally by Amberg et al.16 Network collapse is termed burst feeding. The significance and mechanism of burst feeding have not been investigated in detail. Basically it depends on the compromise between the decrease of permeability upon solidification, determining the stress development in the solid skeleton, and the increase of dendrite network strength with increasing solid fraction. Strength development The vane method 17 '19, illustrated in figure 3, has been used to measure the strength development in themushy zone. The molten alloy is poured into a cylindrical graphite crucible with grooves on the inner face. A thermocouple is inserted close to the crucible wall, and a four-bladed vane is immersed in the center, 10 millimeter from the bottom. The system is cooled naturally in air, and at a certain temperature and fraction solid the rotation of the vane is started, at 5 rpm. Temperature and torque values are collected with a computer. Figure 4 shows an example of one such experiment. The torque measurement in the figure shows that the torque increases rapidly once the stirrer is started. It then reaches a maximum, corresponding to the stresses around the - 48 - Figure 3. Apparatus for determination of Figure 4. Example of one strength the strength development during measurement. Several experiments solidification. like this are required to obtain a description of the development of strength during solidification. vane exceeding the network strength. At this point the dendrite interconnections are broken, and the physical rotation of the vane begins and torque is decreased. The measured maximum torque can be converted into shear strength from knowledge about the yield surface and shear stress distribution around the vane blades 17 "19. This procedure is repeated with a new melt sample at different temperatures, to finally give a description of the build-up of strength throughout the solidification interval. The apparatus can measure strengths up to 51.5 kPa. Figure 5 a) and b) shows how the microstructure of an AlCu4 alloy changes with an addition of 0.05% Ti, added as a AlTiSBl grain refiner. The structure changes from large, highly branched dendrites to very small globular dendrites. Figure 5 c) shows the measured strength development vs. temperature for thesetwo alloys. A description of the relationship between temperature and fraction solid is determined from thermal analysis experiments with the two-thermocouple method 20. Figure 5 d) shows strength vs. fraction solid for the alloys. The figure shows that no significant strength is observed in any alloy at low fraction solid. This is, of course, before the dendrite coherency point, and the dendrites are still free-floating. Coherency is established first in the unrefined alloy, and strength increases very rapidly in thecoherent structure. The grain refined alloy has a later coherency point and strength increases more slowly in the coherent network. The difference between the alloys is explained by the structural differences. Since the dendrites are larger and more branched in the unrefined alloy they are more entangled in each other, thereby forming stronger interconnections. In - 49 - the grain refined alloy, the stress is more easily concentrated to shear the interdendritic regions without extensive deformation and fragmentation of the dendrites. 1 mm i------1 c) d) 0%Ti 0.05% Ti 0.05% Ti TT|T"TJTTT 645 643 641 639 637 635 63 0 5 10 15 20 25 30 35 40 45 50 55 60 Temperature [°C] fraction solid [%] Figure 5. Results of the investigated AlCu4 alloys; (a) and (b) show micrographs of the alloy in the unrefined and with addition of 0.05% Ti as AlTiSBl, respectively; (c) Strength development and fraction solid vs. temperature; (d) Effect of grain refinement on the strength development vs. fraction solid. - 50 - 1 mm c) 0.05% Ti 0.20% Ti ! 0 10 20 30 40 50 60 70 80 90 100 fraction solid [%] Figure 6. Results on the effect of grain refnement on a commercial AlSillMg alloy, (a) and (b) shows micrographs after addition of 0.05% Ti and 0.20% Ti as AlTiSBl, respectively; (c) Strength development in these alloys. Figure 6 a) and b) shows the microstructure for a commercial, near-eutectic, AlSil lMg alloy with 0.05% Ti and 0.20% Ti, respectively, added as AlTiSBl. This commercial alloy initially contains about 0.1% Ti from remelted metal, and this explains the minor structural difference with the grain refinement addition. The strength development vs. fraction solid for these alloys are shown in Figure 6 c). Since the structures are very similar, there is not any big difference in the strength development. However, - 51 - comparing the strength development for AlCu4 alloys and AlSillMg alloys, i.e., Figure 5 d) and 6 c), a very big difference is observed. The AlSillMg alloys develop strength very late during solidification, and only reach full strength when it is fully solid. (The reason why this is not observed in the figure is an error in estimating the solidus temperature). However, although strength develops late in the AlSil lMg alloys, dendrite coherency is measured very early (fs~0.1). This indicates that strength development and dendrite coherency is independent of each other, and that an early coherency point not necessarily is followed by a rapid increase in strength. However both coherency and strength are important for the mushy-zone rheology. Concluding Remarks We are in the process of developing a good understanding and obtaining quantitative measurements of the rheological properties of the mushy zone, both as a function of fraction solid and the morphology of the solid. This article has shown that the rheological behavior through the mushy zone depends strongly on the microstructural evolution, where both the size and morphology of the developing dendrites as well as dendrite growth rate are important. Dendrite coherency alone may not be the only parameter controlling the alloy castability, since both the permeability and strength development of the established interdendritic network are also depending on the solidification conditions. Measurements of the strength development in the mushy zone shows that grain refinement and constitutional conditions are both important for how the rigidity of the dendrite network changes with fraction solid. The apparatus gives reproducible quantitative measurements, and from these we may next evaluate the unknown importance of burst feeding as a mechanism for mass transport in the mushy zone. This requires knowledge of the stress development in the dendrite network, determined both by the linear contraction of the casting and by the pressure gradient for flow, which depends on shrinkage and interdendritic permeability. The compromise between stress and strength development with fraction solid will then indicate how important burst feeding is for the formation of porosity in aluminum castings. We believe that this may also provide new insight into the mechanism of hot-tearing of metals. The more applied side of this is that determination of the mushy zone properties provides a link between the microscopic solidification models, predicting the - 52 - microstructural development, and the macroscopic casting software, which now generally identify problematic areas likely for formation of defects. Knowledge about the relationship between the rheological properties and the solidification characteristics provides the opportunity to quantitative defect predictions by incorporation of the solidification models into the casting software. Acknowledgements Financial support from the Norwegian Research Council, Elkem Aluminium and Hydro Aluminium is gratefully appreciated. References 1. J. Campbell, AFS Cast Met. Res. 1969, pp. 1-8 2. J. Campbell, Castings , (Butterworth-Heinemann Ltd., Oxford), 1991, pp. 191- 200 3. A. Ahmed, D. Apelian, A.K. Dahle and L. Arnberg, to be published in Light Metals 1996, Anaheim (CA) 4. Smithells Metals Reference Book, E.A. Brandes and G.B. Brook (Eds.), ASM International, 1990 5. Y. Mori and N. Ototake, Chem. Eng., vol. 20, 1956, pp. 488- 6. L. Arnberg, G. Chai and L. Backerud, Mater. Sci. Eng., vol. A173, 1993, pp. 101-103 7. G. Chai, L. Backerud, T. Rplland and L. Arnberg, Met. Mat. Trans., vol. 26A, 1995, pp. 965-970 8. G. Chai, Ph.D. thesis, Stockholm University, Stockholm, Chem. Comm., 1994, No. 1 9. A.K. Dahle and L. Arnberg, 4th Int. Conf. on Aluminum Alloys, T.H. Sanders, Jr. and E.A. Starke, Jr. (Eds.), Atlanta (GA), 1994, pp. 91-98 10. J. Zou and R. Doherty, Modeling of Casting, Welding and Advanced Solidification Processes VI, T.S. Piwonka, V. Voller and L. Katgerman (Eds.), IMS, 1993, pp. 193-200 11. L. Nastac and D.M. Stefanescu, ibid., pp. 209-217 12. D. Apelian, M.C. Flemings and R. Mehrabian, Met. Trans., vol. 5, 1974, pp. 2533-2537 -53 - 13. N. Streat and F. Weinberg, Met. Trans., vol. 7B, 1976, pp. 417-423 14. P.C. Carman, J. Soc. Chem. Industry, vol. 57, 1938, pp. 225-234 15. H. Darcy, Les Fontaines Publiques de la Ville de Dijon, Victor Dalmont, Paris, 1856 16. L. Arnberg, A.K. Dahle, CJ. Paradies and F. Syvertsen, AFS Trans., vol. 103, 1995 17. Q.D. Nguyen and D.V. Boger, J. Rheology, vol. 27, 1983, pp. 321-349 18. Q.D. Ngyuen and D.V. Boger, Ann. Rev. Fluid Mech., vol. 24, 1992, pp. 47-88 19. S. Sannes, H. Gjestland, L. Arnberg and J.K. Solberg, 3rd International Conference on Semi-Solid Processing of Alloys and Composites, M. Kiuchi (Ed.), Tokyo, Japan, 1994, pp.271-280 20. L. Backerud, E. Krol and J. Tamminen, Solidification Characteristics of Aluminium Alloys, Vol.l, SkanAluminium, Oslo, 1986 - 54 - Article #3 A.K. Dahle and L. Arnberg "Development of Strength in Solidifying Aluminum Alloys" submitted to Acta Metallurgica et Materialia Development of Strength in Solidifying Aluminium Alloys A.K. Dahle and L. Arnberg The Norwegian Institute of Technology, Department of Metallurgy, N-7034 Trondheim, NORWAY Abstract The efficiency of the feeding mechanisms is crucial for the porosity in aluminum alloy castings. One of these mechanisms, burst feeding, occurs when the developing dendrite network collapses. However, little is known about the occurence and importance of this stress-relieving collapse. It is suggested that the compromise between the development of shear strength in the dendrite network and the increase of resistance to material transport during solidification determines the extent of burst-feeding. Quantitative measurements of the strength development in the mushy-zone during solidification have been obtained in two commercial Al-Si alloys, AlSi7Mg and AlSillMg, an AlCu4 alloy and an AlMg5 alloy with a vane method. The results show that the material does not have any strength prior to dendrite coherency, confirming that the dendrites are free-floating and independent. Grain refining the AlSi7Mg and AlCu4 alloy postpones strength development, and strength increases slowly in the grain refined AlCu4 alloy. Adding Sr to the AlSi7Mg alloy postpones dendrite coherency and strength develops more slowly. Fading is observed to change the strength back towards that of the unmodified alloy with time. Adding 1% Si to the AlMgS alloy postpones dendrite coherency, but the strength increases rapidly and makes thestrength of the AlMg5Sil alloy converge towards that of the AlMgS alloy at increasing solid fractions. Strength develops very slowly in the AlSillMg alloy. The results are shown to be related to the size and morphology of the growing dendrites. 1. Introduction Casting as a production route imposes restrictions on the design of theproduct and the casting conditions, since the melt, chemistry and solidification conditions must be controlled to achieve the desired product quality. As the demands for high- performance, reproducible castings are increasing, there is a need for understanding the underlying principles of the origin of the defects, to form the basis for developing new and improved techniques. - 55 - Modeling and understanding the solidification process has received much attention in recent years, and especially the most common solidification morphology, i.e., dendrite growth, has been emphasised. Today there exist rigorous models for predicting the evolution of the dendrite structure, encompassing both the thermal and solutal effects. However, using these models to predict the casting properties and defect generation is the next big challenge. Most of the alloys commonly used for producing castings in the foundries today are highly alloyed and, except for the eutectic alloys, solidify over a temperature interval. Solidification over a range of temperatures produces a mushy-zone in which solid and liquid coexists, i.e., a part of the casting that is only partially solidified. The width of this zone is determined by the solidification range (AT) and the temperature gradients in the casting. Casting in relatively insulating mould materials, i.e., sand, without chilling, creates small temperature gradients in the casting, and the solid and liquid may coexist throughout the whole casting during most of the solidification process 111. Defects such as porosity, macrosegregation and hot-tears are generated in the mushy- zone. To prevent shrinkage from appearing as porosity, material movement, feeding, has to occur through the mushy-zone, and is therefore critically dependent on the properties of the partially solidified material. The knowledge of the strength development in the mushy-zone is today very limited. An increase in fraction solid will increase the strength, but which parameters affect the development of strength and in what manner is more difficult to predict. This has also led most investigators to assume a stationary and rigid solid skeleton in their models involving the mushy-zone. 2. Background Grain-refined and/or solute rich hypoeutectic aluminium alloys often solidify with growth of equiaxed dendrites dispersed in the liquid. These dendrites initially grow more or less independently of each other, but impinge and form a continuous skeleton at the dendrite coherency point, usually between 10-50% solid 12"51. The dendrite coherency point marks the point in the mushy-zone where the material starts to develop strength, and where the resistance to material movement increases drastically. A transfer from complete volume mobility to interdendritic transport of liquid through the interconnected solid skeleton is observed. This increase in resistance to material transport increases the driving pressure for fluid flow. However, an increase in the pressure differential may also have other consequences. The barrier to formation of gas - 56 - porosity may be exceeded, or the dendrite network may collapse when the developing stresses, from shrinkage and pressure differential driving fluid flow, exceed its strength. The mushy-zone has a very characteristic transformation in behavior at the dendrite coherency point 161. At higher temperatures, the dendrites are free to move. The dendrites may easily rearrange by sliding and rolling, and do not transfer any forces. However dendrite coherency marks the formation of a dendrite network, and the material develops resistance to deformation. After dendrite coherency, bonds have to be broken to accommodate strain. Porosity is a problem in aluminium castings related to incomplete compensation of solidification shrinkage and liberation of dissolved hydrogen. Solidification shrinkage creates a driving pressure for flow in the casting, and depending on the local solid fraction five different mechanisms of material transport may be envisaged. One of these, burst-feeding, was suggested by Campbell171 to occur when the strength of the mushy zone is exceeded by the stresses. Two factors may contribute to the strength of an alloy above the solidus temperature; narrow ’bridging ’ (solid-solid contact) between the dendrites and interlocking of the dendrites as a tangled mass. The ductility and toughness in the solidification range can be expected to be related to the solidification kinetics and the size and shape of the dendrites, the liquid-solid surface energies and the liquid viscosity, which governs the flow through the interdendritic network, and finally the amount of eutectic. In the case of solid ’necking ’, the resistance to fracture of the solid contacts between grains has a controlling influence on the strength of the alloy. The dendrite coherency point and its determining factors have been extensively investigated in recent years. Network establishment can be rationalized through a consideration of the dendrite growth rate and the grain size12,81 . However, recent investigations 191 have indicated that dendrite coherency may not be the only rheological mushy-zone feature determining alloy castability. The development of strength in solidifying alloys has mostly been related to discussions of the tendency and mechanism of hot-tearing of alloys. Hot-tears have been explained by strains of mechanical and thermalorigin, i.e., restrained contraction, creating stresses exceeding the strength of the semi-solid structure 1101. However, the mechanical properties in the solidification range are fundamental material properties - 57 - as they are for solids, and are important for characterizing the material behavior and processes occurring in the partially solidified state. There exist various experimental techniques for measurement of mechanical properties in the solidification range, including the viscosity 1113, compressive 1121 and shear strengths113"151, and ultimate tensile elongation 116,171 and strength118 "231 of solidifying alloys. Singer and Cottrell 1181 argued that the temperature range between dendrite coherency, where the material starts to develop strength and simultaneously have negligible ductility, and the solidus temperature is of great importance withrespect to hot-tearing characteristics. A large coherency temperature interval was related to a susceptibility towards hot-tearing. However, Singer and Cottrell 1181 did not report data on the strength development in this zone in the hypoeutectic Al-Si alloys tested in a tensile test. Williams and Singer 1241 showed that the strength above the solidus was significant in remelted Al-Sn samples and suggested a mechanism for the deformation of alloys above the solidus temperature during tensile testing. Kubota and Kitaoka 1201 determined the point where solidifying alloys started to develop measurable contraction and also performed horizontal tensile tests in the mushy-zone. Grain refinement was shown to lower the temperature where contraction was measurable, hereby reducing the temperature range for development of contraction and also decreasing the extent. The tensile tests showed that the rate of increase of tensile strength after dendrite coherency was significantly larger in the grain refined alloy. This was related to an increase in resistance to hot-tearing, since the rigidity of the material is increased in combination with a reduced contraction. The strength increases rapidly and lesser stresses may develop. This has also been confirmed by experimental hot-tear tests. Forest and Bercovici 1221 performed tensile tests on partially solidified samples. An increase in cooling rate was related to an increase in strength for a given temperature or fraction solid. Ackermann et al.1251 investigated the mechanical properties perpendicular to the growth direction of columnar dendrites in Al-Mg alloys by tensile testing. A significant effect of the size and morphology of the mushy-zone was observed, and virtually no strength was detected at solid fractions below 0.95. Wisniewski and Brody 1231 performed tensile tests on as-solidified Al-Cu alloys at solid fractions above 0.9. Grain refinement was shown to give a lower flow stress than the unrefined alloy, and also a lower strain-rate sensitivity. -58 - Ohm and Engler 126,271 made horizontal tensile tests of macroscopic directionally solidifying shells of wrought aluminium alloys at high solid fractions (> 0.9). Grain refining, transforming the structure from columnar to equiaxed, gave a lower tensile strength and larger elongation at fracture compared to the unrefined alloys at the same temperature or fraction solid, over a wide range of temperatures. Increasing the strain rate gave a considerable increase in the tensile strength and a relatively unchanged elongation. Decreasing the temperature could generally be related to an increase in tensile strength and a reduced ductility, and the properties of the grain refined, equiaxed, alloy approach those of the unrefined, columnar, alloy due to formation of agglomerates by bridging and welding of the dendrites. Metz and Flemings 110,131 strained partially solidified samples in tension, compression and shear, and observed that the strength at a certain strain was strongly dependent on fraction solid and grain size, but not directly on composition. The deformation mechanism was investigated metallographically, and deformation in shear was accomplished by sliding of grains over one another, complemented by fracture in interdendritic regions. The dendrites in the shear zone break free of each other and subsequent deformation is primarily through sliding and rolling of the dendrites. When the alloys are grain refined the structure is less branched, and the dendrites are more ’isolated ’ from each otherand less tangled. Lack of extensive branching gives a slower increase in shear strength with fraction solid in the grain refined castings™. The rapid increase in strength with increasing fraction solid is a result of interactions of thedendrites with each other. Mechanical hindrance and interlocking may occur and also welding of the dendrites. New portions of the dendrites may be brought into contact and form additional welds. The deformation may brake off dendrite arms that may interact mechanically or weld neighbouring dendrites. Metz and Flemings 1101 observed in castings tested in shear, that the rate of strain hardening was much lower at low shear rates than at higher strain rates. Unfortunately the strength levels reported are not true strength levels, and can only be used to discuss general trends. Sannes et al.™ determined the yield point during remelting of fine-grained thixotropic magnesium alloys with a hot-torsion test and a vane method. The yield point was explained by solid necks connecting the grains in a three-dimensional skeleton 1281 . A significant influence of microstructure and composition was reported. Tensile testing and the resulting strength measurements is very close to the deformation -59 - mechanism during hot-tearing, where the dendrites are tom apart. But for feeding, and the burst-feeding mechanism in particular, this is not the case. The mechanism is definitely not tension, but rather compression or shear, as the dendrites are compacted until they break. Elaborating the extent and duration of each feeding mechanism from the structures and porosity of real castings is inherently difficult, which is one reason for the veil of fog covering the concept of burst feeding. Its extent will in addition to alloy selection be affected by the casting method and the coupled fluid flow and thermal gradients. Casting processes that utilize high pressures during feeding, such as squeeze casting and pressure die casting, will certainly be more prone to burst-feeding. There is a need for reliable data on the strength development during solidification, especially with respect to the burst-feeding mechanism. This is important both with respect to evaluating if burst-feeding is likely to occur, under what circumstances, and to what extent. This article presents data on the influence of grain size and dendrite morphology on the strength development in some common aluminium casting alloys. 3. Experimental method A four-bladed vane was used to measure the fracture strength of the dendrite network during solidification, and Figure 1 shows an illustration of the apparatus used to experimentally determine the strength development. The shearing zone is approximated to be cylindrical, defined by the dimensions of the vane. The melt was poured into the preheated graphite crucible with an outer diameter and height of 50 mm. The sample had a diameter of 30 mm and a height of 40 mm. Semi circular shaped grooves with a depth of 2 mm were machined on the inner face of the crucible at every 45°, to prevent slip between the sample and the crucible. A preheated steel vane with four blades was inserted in the centre of the melt, 10 mm from the bottom, and a thermocouple was inserted close to the crucible wall. The thermocouple was of type K (alumel-chromel), calibrated against pure aluminium with a melting point defined to be 660.0°C. After stabilizing the system at about 720°C, the furnace was removed, and the system cooled naturally in air at about 1 K/s. The relationship between fraction solid (fs) and temperature (T) was determined from a cooling-curve by a procedure developed by Tamminen [29]. - 60 - Jross-section Thermocouple Furnace Figure 1. Schematic description of the experimental apparatus for determination of mushy-zone strength. At a certain temperature, or fraction solid, below the nucleation temperature, rotation of the vane was started, and the torque necessary to rotate it at a constant rotation speed of 5 rpm was continuously monitored. Figure 2 shows an example of the general shape of a torque-time curve for a material at a temperature where it exhibits a yield point. Initially the torque (or stress) increases continuously in an elastic manner, attributed to elastic deformation of the part of the material close to the edges of the vane blades, stretching the bonds constituting the dendrite network. Eventually a point is reached when the interconnections start to break, and this threshold torque, or yield point, has to be overcome for the vane to physically start rotating. After network collapse the torque decreases rapidly. No sign of plasticity was observed in the alloys and solid fraction range reported in this investigation. The yield stress can be calculated from the measured maximum torque (Tmax) and the known surface area of the cylindrical yielding surface, defined by a diameter, d, and height, h. Four vanes, with dimensions given in Table 1, were used in this investigation to increase the accuracy in the measurements. The thickness of the blades (t) was 2.5 mm. - 61 - Temperature |,,iT“T"T",r Exp. time (sec.) Figure 2. Example of the measured parameters, torque and temperature, in one strength measurement at a temperature where the alloy has developed a dendrite network. Table 1. Characteristics of the vanes. Dimensions, geometric factor, K^,,, and maximum measurable shear stress, Tmax. 1Z Vane diameter, height, h ■“‘vane ^max No. d[mm] [mm] [x 10"6 m3] [kPa] A 10 10 1.34 51.5 B 10 20 2.42 28.7 C 15 20 6.77 10.2 D 20 20 13.76 5.0 Sannes [301 refined the equations developed by Nguyen and Boger [3L32] to relate the maximum torque, Tmax, to the yield stress, Ty. The torque balance: djl Tmra = (Tidh-4th)^T. : +2 f [(2nr-4t)rx e(r)]dr 2 0 where Te(r) is the distribution of shear stress along the flat circular end surfaces, can be integrated to give the relationship between yield stress and maximum torque: - 62 - T = K where Kvane is a geometric factor: td(2h+d) 2 \d 3 The rheometer can measure a maximum torque of 69.2TO"3 Nm. By performing such experiments for several points in the whole solidification range (i.e., temperatures or fractions of solid), the strength development can be determined. Four sets of alloys have been investigated. Two commercial Al-Si alloys, AlSiVMg and AlSillMg, were supplied by Fundo a.s. Two pure alloys, AlCu4 and AlMg5, were made by melting high-purity aluminium with electronic grade copper or commercial purity magnesium. The effect of grain refinement was investigated in the Al-Si alloys and the AlCu4 alloy by addition of an A1T15B1 rod-type master alloy. A commercial AlSi7Mg alloy, SiBloy, grain refined with boron was also included. In addition to investigating the effect of grain refinement, theeffect of eutectic modification, by rod- type AlSrlO, and subsequent holding time was evaluated in the AlSi7Mg alloy. The AlMg5 alloy was tested with and without addition of 1% Si. Chemical analyses of the alloys are given in Table 2. Metallographic samples of the alloys were solidified under similar conditions as in the torque experiments and sectioned at the same level as the bottom of the vane, i.e., 10 mm from the bottom. They were anodized and investigated in an optical microscope. Table 2. Concentrations of the main constituents in the base alloys [wt%]. Alloy Si Mg Fe Cu Ti B 7Si 6.87 0.179 0.114 0 0.107 0 7SiBloy 7.11 0.244 0.086 0 0 0.0153 11 Si 10.78 0.146 0.115 0 0.0998 0 4Cu 0 0 0 3.77 0 0 5Mg 0 4.663 0 0 0 0 -63 - 3. Experimental results 3.1 Effect of grain refinement 1 mm ------1 Effect of grain refinement on the yield-point behavior of AlSiTMg alloys (Ti-B or B) 0.05% Ti ’j_ri 1" I"] I i I I | I I M | 615 610 605 600 595 590 585 580 575 Temperature [°C] Figure 3. Micrographs of the AlSi7Mg alloy at increasing levels of grain refinement, (a) 0.05% Ti; (b) 0.20% Ti; (c) SiBloy, and (d) Measurements of strength development vs. temperature for these alloys. Figure 3 a)-c) shows the microstructure of the AlSi7Mg alloy at increasing levels of - 64 - grain refinement. The reduction of grain size is relatively small since the alloy initially contains a high level of titanium from remelted metal. Moreover, it has been demonstrated that the effect of adding Ti-B type grain refiner to Al-Si type alloys is limited. The measured strength development for these alloys is shown in figure 3 d). Increasing the grain refinement postpones the development of strength to lower temperatures. 1 mm i------1 Effect of grain refinement on the yield-point behavior of AlSillMg alloys + 0.05% Ti A 0.20% Ti 20 30 40 50 60 70 80 time from 580°C [s] Figure 4. (a) and (b) Micrographs of the AlSillMg alloy at 0.05% Ti and 0.20% Ti, respectively; (c) Experimental results of strength vs. cooling time from 580'C for the AlSil lMg alloys. - 65 - Micrographs of the AlSil lMg alloy at two levels of grain refinement, 0.05 and 0.20% Ti, are shown in figure 4 a) and b). This alloy also contains titanium from remelted metal, and the reductions of grain size are therefore lesser than what otherwise would have been observed. AlSil 1 Mg is a near-eutectic alloy, and the fraction of primary phase is about 15%. Fraction solid changes markedly with small changes in temperature, and time from a reference temperature of 580°C was therefore used for the experiments, as observed in figure 4 c). From the figure it may be observed that the spread in the results is larger than in the AlSi7Mg alloy, possibly due to the inaccuracy of the temperature measurement. The alloy with the highest level of grain refinement seems to develop strength earlier and more rapidly than the less grain refined alloy. Effect of grain refinement on the yield-point of AlCu4 alloys. + 0.05% Ti 11 pi 111 nil 1111111 m I j 11111 645 643 641 639 637 635 633 Temperature [°C] Figure 5. (a) and (b): Micrographs of the AlCu4 alloy at no grain refinemer addition and 0.05% Ti, respectively; (c) Effect of grain refinement on the strength vs. temperature relationship of AlCu4. - 66 - Figure 5 a) and b) shows micrographs of an AlCu4 alloy without grain refiner and with 0.05% Ti added as AlTi5Bl. A drastic reduction of grain size is observed with the grain refiner addition. The dendrites in the unrefined alloy are large, highly branched and tangled into each other, while the grain refined structure consists of small, globular dendrites. The strength development for these two alloys are shown in Figure 5 c). The unrefined alloy is observed to gain strength rapidly once the dendrites start to interact, whereas the grain refined alloy exhibits a very slow development of strength. Fraction solid vs. temperature d) Effect of adding 1 % Si on the yield-point AlMgS + 0.05% Ti alloys behavior of AlMgS + 0.05% Ti 640 -g ------1------1------1------1------1 ! without Si 630 4 0%Si A 1% Si added 6204 u -1 2^ 6104 2 4 l%Si I 600 i ^ 5904 £ i 5804 \ 570 4 560 im in u luuimiii'n'rrnrrrnTrp'ii ii iu|iin|ui) |mi|irii|Uii|iii)|iiinini|ii'ti|iiinrm 0 10 20 30 40 50 60 70 80 90 100 628 624 620 616 612 608 604 600 596 592 588 fraction solid [%] Temperature [°C] Figure 6. (a) and (b) Micrographs ofAlMg5, grain refined with 0.05% Ti, without and with 1% Si, respectively; (c) Fraction solid vs. temperature of these alloys; (d) Measured strength development vs. temperature. - 67 - 3.2 Effect of adding silicon Figure 6 a) and b) shows the microstructure of the AlMg5 alloy, grain refined with 0.05% Ti, before and after the addition of 1% Si. Adding 1% Si produces a eutectic that covers the grain boundaries, and the grain size is also somewhat decreased. The cooling curve, as shown in figure 6 c), changes significantly with and addition of 1% Si. The liquidus temperature is decreased and a eutectic precipitation is observed. Figure 6 d) shows the strength measurements, and the difference in liquidus temperature is responsible for changing the temperature range where strength is developing. - , . - ? , /■ 100 pm i------1 c) Effect of Sr-modification and holding time on the yield-point behavior of AISi7Mg+0.05% Ti alloys. 1 l~T A 0.02% Sr added, long holding time ^ 0.02% Sr added, short holding time 608 604 600 596 592 588 584 580 Temperature [°C| Figure 7. (a) and (b) Micrographs of AlSTJMg, grain refined with 0.05% Ti before and after addition of200 ppm Sr, respectively; (c) Strength development in the AlSi7Mg alloy, grain refined with 0.05% Ti and modified with 200 ppm Sr at short and long holding times. - 68 - 3.3 Effect of eutectic modification Figure 7 a) and b) shows the effect of adding 200 ppm strontium on the microstructure of AlSi7Mg, grain refined with 0.05% Ti. Adding strontium changes the eutectic silicon structure from lamellar to fibrous. The strength development of this alloy is shown in Figure 7 c). The experimental results have been separated into those made at holding times lesser than one hour and those at more than 1.5 hours. Separating the results in this way shows that the holding time has a strong effect on the results, and that the strength at a given temperature increases with increasing holding time. 4. Discussion Fraction solid vs. temperature was determined by thermal analysis, and in Figures 8 a)- e) the strength (%,,) vs. fraction solid (fs) have been plotted for all alloys. One feature common to all alloys is that strength does not develop until a certain fraction solid is reached, corresponding to the dendrite coherency point. No strength before dendrite coherency hereby confirms that the dendrites are free-floating and independent of each other in the first part of the solidification process. The coherency fraction solid indicated by these measurements corresponds well to those determined by other techniques, such as thermal analysis and rheological measurements with the paddle- method 14,51. The effect of the structural refining additions confirms earlier observations that grain refinement postpones dendrite coherency, and that eutectic modification slightly postpones coherency151. An interesting effect is that addition of 1% Si to the AIMg5 alloy seems to postpone coherency, possibly due to the decrease of grain size. Figure 8 a) shows that grain refinement of the AlSi7Mg alloy postpones the fraction solid at which strength develops, but the shape and slope of all curves are very similar. The reason for this may be that the dendrite morphology does not undergo a major change with grain refinement compared to thegrain size. At an equal solid fraction the network strength is lower with increasing grain refinement. The AlSillMg alloy exhibits a very interesting development of strength. Dendrite coherency in these alloys is usually measured between 10 to 15% solid 151, and figure 8 b) shows that strength develops very slowly. Whereas the AlSi7Mg alloy, at all tested grain refinement levels, reached a strength exceeding the maximum measurable strength at solid fractions lower than 50%, strength in the AlSillMg alloy is still very low at this solid fraction. At higher solute contents the dendrites are highly branched with thin dendrite arms. The dendrite network is established at a low solid fraction and is therefore rather tenuous and weak. Eutectic precipitation starts soon after dendrite - 69 - i a) Strength development during solidification b) Strength development during solidification of grain refined AlSiTMg alloys. of grain refined AlSillMg alloys. 55- 1 1 1 1 1, ,1 1 / + 0.05% Ti + 0.05% Ti 45. ^ 0.20% Ti I - A 0.20% Ti / / / _ A SiBloy 5 & s 25 •o + A, tjnrr[rrrr[im|iiiipm|iiii|iiii|iif 10 15 20 25 30 35 40 45 50 55 0 10 20 30 40 50 60 70 80 90 100 faction solid [%] fraction solid [%] c) Strength development during solidification d) Effect of Si-addition on the strength development ot gram relined AlCu4 alloys. during solidification of AlMg5+0.05% TL l%Si added 0 5 10 15 20 25 30 35 40 45 50 55 60 0 10 20 30 40 50 60 70 fraction solid [%] fraction solid [%] Effect of Sr on the strength development during solidification of AlSi7MgH).05% TL -j- Without Sr addition A 0.02% Sr added, long holding time -f. 0.02% Sr, short holding ti me 0 5 10 15 20 25 30 35 40 45 fraction solid [%] Figure 8. Calculated relationship of strength vs. fraction solid for all investigated alloys, (a) Effect of grain refinement in commercial AlSi7Mg; (b) Effect of grain refinement in commercial AlSil lMg; (c) Effect of grain refinement inAlCu4; (d) Effect of Si-addition on AlMg5+0.05% Ti; (e) Effect of Sr-modfication and holding time in commercial AlSi7Mg, grain refined with 0.05% Ti. - 70 - coherency. The AlSillMg alloy grain refined with 0.20% Ti seems to develop a slightly stronger dendrite network at the same fraction solid as the alloy with 0.05% Ti. AlSil lMg contains a high volume fraction of eutectic which has a high latent heat of crystallization. Chai[4] reported observations of partial remelting of dendrites during eutectic precipitation, but the strength measurements show no evidence of this. The small difference between the two levels of grain refinement is most likely caused by a slightly lower Si content in the alloy with 0.20% Ti addition. The reason that full strength is not observed at 100% solid is due to an error in estimating the solidus temperature. The micrographs of the AlCu4 alloy in figure 7 show a dramatic change in microstrucure with an addition of 0.05% Ti. The strength vs. fraction solid relationship for the two levels of grain refinement are also very different, as shown in figure 13c). Since the dendrites are large and highly branched in the unrefined alloy they are very entangled in each other, and the structure cannot encompass much strain by rearrangement of the dendrites. Instead strain must be adopted by deformation of the dendrites themselves, and extensive dendrite fragmentation must occur. Strength therefore increases rapidly. In the grain refined alloy, however, the dendrites are smaller and more rounded, allowing them to rearrange more easily, by sliding past each other during straining. Deformation is more concentrated to the interdendritic areas, without the extensive deformation required in the unrefined alloy. Strength therefore develops more slowly in the grain refined alloy. The large difference in strength development between the two levels of grain refinement is seen by the unrefined alloy reaching the maximum measurable strength before any significant strength is developed in the grain refined alloy. 1% Si is often added to AlMg5 alloys to increase theresistance to hot tearing. Figure 13 d) shows that strength after dendrite coherency develops very rapidly in the alloy with the silicon addition, subsequently converging to the values for the alloy without silicon. Similar morphologies may be the reason why the strength of the two alloys converge at higher solid fractions. Addition of eutectic modifiers to Al-Si alloys, such as Sr, displaces the eutectic composition to higher Si contents and reduces the eutectic temperature thereby increasing the solidification range. The nucleation temperature is also increased somewhat and combined with the growth restricting effect this is responsible for a small increase in the coherency fraction solid. The effect of the Sr addition fades with time. The results shown in figure 13 e) have therefore been divided into those performed less than one hour after addition, and those tested after more than 1.5 hours. -71 - In addition, the curve for the unmodified alloy has been plotted in the figure. The results clearly indicate that Sr-modification postpones the point where strength develops and also decreases the slope of the strength curve. The strength of the modified alloy converges towards the unmodified alloy at long holding times. This is related to the fading of the modifier, which is also responsible for the scatter in the measurements. Another interesting general observation from figures 13 a)-e) is that there does not seem to be a unique relationship between dendrite coherency and the slope of the strength-curve. For example, a late coherency point is not necessarily related to a slow development of strength. This indicates that both alloy constitution as well as grain size affect the strength development. Table 3 shows exponentially curve-fitted equations of the strength vs. fraction solid for each alloy and addition. Table 3. Equations for exponential curve-fitting of these results: = exp (Afs)*B Alloy A B AlSi7Mg+0.05%Ti 0.17453 0.100897 AlSi7Mg+0.20%Ti 0.191167 0.0227583 SiBloy-7Si 0.148664 0.0239577 AlSil lMg+0.05%Ti 0.0719598 0.0247321 AlSil lMg+0.20%Ti 0.0541909 0.219893 AlCu4 0.45771 0.000116557 AlCu4+0.05%Ti 0.138481 0.0118422 AlMg5+0.05%Ti 0.113197 0.0143243 AlMg5Sil+0.05%Ti 0.191583 3.9893E-5 AlSi7Mg+0.05%Ti 0.17453 0.100897 AlSi7Mg+0.05Ti+0.02%Sr 0.136605 0.119309 Figure 9 illustrates the possible feeding mechanisms in a plate casting of an alloy solidifying with development of equiaxed dendrites, where an end-chill imparts a positive temperature gradient towards the feeder. Detailed descriptions of the features of each feeding mechanism have been given by Campbell[7,33] and Dahle and Arnberg 1343. The volume deficiency caused by the shrinkage experienced by most - 72 - alumin ium alloys during solidification creates a pressure differential within the casting that enforces material movement towards the solidifying regions. From the point where a dendrite network is established in the mushy zone, i.e., dendrite coherency, subsequent material transport must occur by melt flow through theinterdendritic solid skeleton. Interdendritic feeding has been designated the most crucial and controlling of the feeding mechanisms for the subsequent porosity formation. Flow through the dendrite network is controlled by the interdendritic permeability, K, which depends strongly on the size and morphology of the dendrites and the solid fraction. The Kozeny-Carman 1351 relationship is often used to estimate the permeability: 5 v where v is liquid flow rate, p is pressure, T|, and p, are the viscosity and density of the liquid, respectively, and g is gravity. liquid Mnsbv 7nnp solid Liquid Feeding Feeding Feeding Feeding Burst Feeding Figure 9. Principal feeding mechanisms in a plate casting with an end chill. Interdendritic permeability decreases with increasing solid fraction, and consequently - 73 - the pressure gradient required for further melt flow will increase as the flow resistance increases. As the pressure for flow in the interdendritic channels increases, the surrounding dendrite network will develop larger stresses. At some locality in the mushy zone the stresses may exceed the strength of the dendrite network, and if this occurs the dendrite network will collapse in a burst. If the network collapses it is very likely that it will also cause increased mobility at the lower solid fractions since the network consists of connected dendrites and the feeding process is continuous. As the fragments are compacted into regions of higher solid fraction they may clog the feeding paths to the more distant areas and cause a pocket of porosity behind the compacted network. In this way burst feeding may have a significant influence on the amount and distribution of porosity. Burst feeding events may occur several times in a casting. Evidence of such a mechanism has been reported by Amberg et al[37] . (stress) Region likely to burst (strength) coherency Figure 10. Illustration of stress-strength compromise that determines susceptibility to burst feeding. This article has presented measurements of the strength development in the mushy zone. To be able to use this information to predict the probability of burst feeding it is necessary to model the development of stresses in the permeable solid skeleton during interdendritic flow. Models exist to estimate the pressure in the liquid during interdendritic feeding, but the solid stresses have been largely overlooked 1381 . Describing the stress development requires knowledge about the variation of fraction solid and specific surface area at any location in the casting as a function of fraction solid, and comparing the stress and strength will then indicate if the network can resist the flow -74- pressure. Figure 10 shows a schematic diagram of the way in which the stress-strength compromise would be expected to evolve for an alloy susceptible to burst-feeding. In this context it should be realized that if the pressure barrier for formation of gas porosity is exceeded, the formation of a gas pore will relieve the stresses in the mushy- zone. Evidently there is a need for improving the understanding of the behavior of the mushy-zone, especially with respect to strength development and burst feeding. Future work should be directed towards measuring the strength development, with different variables, throughout the mushy-zone. The apparatus used herein has restrictions on the maximum measurable strength. Determining the strength development throughout the solidification interval will also generate information that is relevant to hot-tearing, which arises in the later stages of solidification. Experimental investigations of the deformation mechanisms in the mushy-zone are also necessary. The importance and occurrence of burst-feeding must be established in real castings, for example by a combination of externally loading the mushy-zone and microstructural investigations of quenched samples. Furthermore, the stress and strength development of the two- phase zone should be modeled, including a close coupling to solidification models with both coarsening and eutectic precipitation. 5. Conclusions A vane method has been used to obtain quantitative measurements of the development of strength during solidification of some common aluminium foundry alloys. No strength is found at low solid fractions; the equiaxed dendrites are mobile and independent. Strength starts to develop at the dendrite coherency point, and the solid fraction at dendrite coherency found with the vane method correlates well to that measured by other techniques. A significant impact of the investigated variables, namely grain refinement, eutectic modification, and Si content, on the strength development was observed. Generally the strength of a grain refined (AlSiVMg or AlCu4) or eutectic modified sample (AlSiVMg) was lower than with less structurally modifying elements at the same solid fraction. Strength was found to develop very late and slowly in the AlSillMg alloy. It is suggested that the strength development in the mushy-zone is important for the feeding and subsequent porosity content and distribution in castings, especially with respect to the occurrence of the burst-feeding mechanism. - 75 - Acknowledgements Financial support from the Norwegian Research Council, Elkem Aluminium and Hydro Aluminium is gratefully appreciated. Ms. Natalia L.M. Veldman is acknowledged for proof-reading the manuscript. References 1. M.C. Flemings: Solidification Processing, Mc-Graw Hill Book Co., 1974, p. 164 2. L. Amberg, G. Chai and L. Backerud, Mat. Sci. Eng., 1993, vol. A173, pp. 101- 103 3. G. Chai, L. Backerud, T. Rplland and L. Amberg, Met. Mat. Trans. A, vol. 26A, 1995, pp. 965-970 4. G. Chai, Ph.D. thesis, Stockholm University, Chem. Comm., No. 1, 1994 5. A.K. Dahle and L. Amberg, 4 ’th Int. Conf. on Aluminium Alloys, T.H. Sanders, Jr. and E.A. Starke, Jr. (Eds.), Atlanta (GA), 1994, pp. 91-98 6. M.C. Flemings, Met. Trans. A., vol. 22A, 1991, pp. 957-981 7. J. Campbell, AES Cast Met. Res. J., March 1969, pp. 1-8 8. A.K. Dahle and L. Amberg, submitted to Met. Trans. A 9. A.K. Dahle, P.A. Tpndel, C.J. Paradies and L. Amberg, accepted for publication in Met. Mat. Trans. A 10. S.A. Metz and M.C. Flemings, AFS Trans., vol. 78, 1970, pp. 453-460 11. R. Ichikawa and K. Miwa, J. Jpn. Inst. Met., 1978, vol.42,pp. 1023-1028 12. M. Suery and M.C. Flemings, Met. Trans. A, vol. 13A, 1982, pp.1809-1819 13. S.A. Metz and M.C. Flemings, AFS Trans., 1969, vol.77, pp.329-334 14. S. Sannes, H. Gjestland, L. Amberg and J.K. Solberg, 3rd International Conference on Semi-solid Processing of Alloys and Composites, Eds. M. Kiuchi, Tokyo, Japan, 1994, pp.271-280 15. C.L. Martin, S.B. Brown, D. Favier and M. Suery, 3rd International Conference on Semi-solid Processing of Alloys and Composites, Eds. M. Kiuchi, Tokyo, Japan, 1994, pp. 27-36 16. N.N. Prokhorov, Russian Castings Product., 1962, vol.4, pp.176-179 17. I.I. Novikov, Russian Castings Product., 1962, vol.4, pp. 167-172 18. A.R.E. Singer and S.A. Cottrell, J. Inst, of Metals, vol. 73, 1947, pp. 33-54 19. H.F. Bishop, C.G. Ackerlind and W.S. Pellini, AFS Trans., vol. 65, 1957, pp. 247-258 20. M. Kubota and S. Kitaoka, AFS Trans., 1973, vol.106, pp.424-427 21. T. Isobe, S. Kitaoka and M. Kubota, J. Japan Foundrymen’s Soc., 1978, vol. - 76 - 50, pp. 235-239 22. B. Forest and S. Bercovici, Solidification Technology in the Foundry and Casthouse, The Metals Society, London, 1983, pp. 607-612 23. P. Wisniewski and H.D. Brody, Modeling of Casting, Welding and Advanced Solidification Processes V, Eds. M. Rappaz, M.R. Ozgu and K.W. Mahin, The Minerals, Metals & Materials Society, Warrendale (PA), 1991, pp. 273-278 24. J.A. Williams and A.R.E. Singer, J. Inst, of Metals, vol. 96, 1968, pp. 5-12 25. P. Ackermann, W. Kurz and W. Heinemann, Mat. Sci. Eng., vol. 75, 1985, pp. 79-86 26. L. Ohm and S. Engler, Aluminium, vol. 64, No. 5, 1988, pp. 513-520 27. L. Ohm and S. Engler, Metall, vol. 43, No. 6, 1989, pp. 539-543 28. S. Sannes, H. Gjestland, L. Amberg and J.K. Solberg, 3rd International Conference on Semi-solid Processing of Alloys and Composites, Eds. M. Kiuchi, Tokyo, Japan, 1994, pp.75-84 29. J. Tamminen, Ph.D. thesis, Stockholm University, Chem. Com., No.2, 1988 30. S. Sannes, Ph.D thesis, The Norwegian Institute of Technology, 1994 31. Q.D. Nguyen and D.V. Boger, J. Rheology, 1983, vol. 27, pp. 321-349 32. Q.D. Nguyen and D.V. Boger, Ann. Rev. Fluid Meek, 1992, vol.24, pp. 47-88 33. J. Campbell: Castings, Butterworth-Heinemann Ltd., Oxford, United Kingdom, 1991 34. A.K. Dahle and L. Amberg, submitted to J. Met, April, 1996 35. P C. Carman, J. Soc. Chem. Industry, vol. 57, 1938, pp. 225-234 36. H. Darcy, Les Fontaines Publiques de la Ville de Dijon, Victor Dalmant, Paris, 1856 37. L. Amberg, A.K. Dahle, C.J. Paradies and F. Syvertsen, AFS Trans., paper 95- 115, 1995 38. Q. Yu, M. Makhlouf and D. Apelian, Int. J. Heat and Mass Transfer, vol. 38, 1995, pp. 31-38 - 77 - Article #4 A.K. Dahle, P.A. T0ndel, CJ. Paradies and L. Amberg "Effect of Grain Refinement on the Fluidity of Two Commercial Al-Si Foundry Alloys" accepted for publication in Metallurgical and Materials Transactions A, 1996 Effect of Grain Refinement on the Fluidity of Two Commercial Al-Si Foundry Alloys A.K. Dahle1, P.A. T0ndel 2, CJ. Paradies 1 and L. Amberg 1 1) The Norwegian Institute of Technology, Department of Metallurgy, N-7034 Trondheim, Norway 2) Elkem Aluminium ANS, P.O. Box 566, N-8651 Mosjoen, Norway ABSTRACT The effect of grain refinement on the fluidity of AlSi7Mg and AlSillMg has been investigated by spiral tests. Two different types of grain refiners have been evaluated. An AlTi5B 1 master alloy was added to different Ti contents. Since the commercial alloys had a high initial content of titanium, model alloys were made to investigate the fluidity at low grain refiner additions. Commercial alloys grain refined only by boron additions have also been investigated. The results from the fluidity measurements have been verified by measuring the dendrite coherency point of the different cast alloys. Although different, the two methods show similar trends. The spirals from each fraction grain refiner cast were subsequently investigated metallographically at the tip of the spirals and at a reference point a distance behind, but no obvious difference in structure was observed. For both alloys, an increase in fluidity is observed as the content of grain refiner increase above 0.12 pet Ti, while the fluidity is impaired with increased grain refinement below 0.12 pet Ti. The alloys grain refined with -0.015 pet B show the highest fraction solid at dendrite coherency, the smallest grain size, and the best fluidity. I. INTRODUCTION During solidification of a solute-rich aluminium foundry alloy, equiaxed dendrites start to grow into the undercooled liquid. In the first stages, these dendrites can grow relatively independently of each other, but after a certain range of growth, depending - 79 - on dendrite growth rate and grain density, the primary dendrite tips impinge onto the neighbouring dendrites. This is identified as the dendrite coherency point, thepoint at which a more or less continous three-dimesional dendrite network is formed in the semisolid material. Previous work[l-5] has shown that several parameters affect the point where the dendrite network is formed, and dendrite coherency is expected to have significant influence on the castability of these alloys [6]. The castability is an important feature of all aluminium foundry alloys, restricting the applicability of the alloys for casting purposes. Castability is a wide concept including characteristics such as form-filling ability, fluidity, hot-tearing susceptibility, and feeding behavior. One of the important factors that may limit the quality of a casting is the ability of the liquid alloy to flow and fill the mould pattern as cooling and solidification progress. Fluidity is empirically defined as the length the metal flows in a channel with a small cross-sectional area while solidifying. Fluid flow is also important with respect to feeding, i.e., material transport to compensate for shrinkage. Therefore some relationship between fluidity and feeding is expected since both depend on the ability of the melt to flow during solidification]! -8]. The solidification mode, morphology, and rate and the grain density are important, determining the size and irregularity of the developing solid, which, combined with the strength and physical properties of the material, define the point of flow inhibition in these channels. Two well-known test methods to investigate the fluidity are the spiral test, often in sand moulds, and the vacuum-fluidity test, in glass tubes. Many other parameters have been identified to affect the fluidity and feeding capability of aluminium alloys. Among these are melt superheat, temperature gradients, latent heat, alloy composition, eutectic modification, and grain refinement]!- 15]. The formal definition of fluidity is the inverse of viscosity, but it has come to reach another meaning in casting terminology. In fact, casting fluidity is independent of the viscosity of the liquid, according to the work by Bastien et al.[13]. This indicates that interdendritic fluid flow is of low significance in the fluidity test and that the length of flow is determined by the earlier stages of solidification. However, a mixture viscosity characterising the suspension of crystals and melt may be more significant with respect to fluidity. Fluidity is a complex parameter that is affected by the properties of the metal and the mould and the pouring conditions, the solidification condition being the underlying controlling factor. Flemings [16] predicted that the mechanism by which the flow is stopped changes as the solidification mode is altered. Pure metals and dilute alloys solidify with a smooth, compact growth front, continously tapering the central flow - 80 - channel. The flow is stopped somewhere behind the flow tip, where the growth fronts meet and obstruct the flow path, closing the channel entrance. Solute-rich alloys, on the other hand, solidify in a mushy manner. Here, equiaxed dendrites are expected to be nucleated at and transported by the flow tip. These dendrites finally stop the flow when they reach a critical fraction solid. Fluidity is expected to level off, or even increase slightly, with increasing alloy content. Pai and Jones [12] showed that solidification in the flow tips stops the flow in mushy freezing alloys and confirmed that extensive interlocking and bridging between the dendrites near the tip had taken place. The mould acts as a perfect heat sink during flow, since the flow tip is continously exposed to a cold mould channel during its advancement. Campbell [15] also noted that the semisolid mixture at the tip stops flowing at the fraction solid where the dendrites impinge, giving a significant increase of viscosity of the slurry. Flemings et al.[ll] calculated the critical fraction solid for a fluidity test vs. pressure head, and found that it increases with increasing pressure head, converging towards about 35 pet solid. The relationship between this mechanism of flow stoppage and dendritic coherency is obvious; in fact, it is dendrite coherency - network establishment. This suggests that alloys solidifying with an equiaxed structure should be expected to exhibit a firm relationship between dendrite coherency and fluidity. ■i i i i i i I- "|11 i i i T-r-|-r 10 = 15 Silicon content [wt%] Figure 1. Fluidity of binary Al-Si alloys, poured at a constant pouring temperature of 800 °C. (after Lang[9]) Lang[9] has measured the effect of increasing the silicon content in binary Al-Si alloys - 81 - with a bar casting. Figure 1 illustrates the results. From the figure, it is observed that fluidity decreases in the range of solid solubility and increases slowly as fraction eutectic increases. This can also be interpreted as a confirmation of the contribution from interdendritic flow to the fluidity being relatively small. A curious feature is that maximum fluidity lies in the hypereutectic region, around 18 to 19 pet Si, and not at the eutectic composition as for other systems, e.g., Al-Mg[9]. This effect is partly described by an enthalpy consideration, i.e., HSi>HA1 [7]. Grain refinement of aluminium alloys, usually achieved by additions of master alloys based on Al-Ti or Al-Ti-B, often improves the mechanical properties and porosity distribution of the final product. A new method to effectively grain refine Al-Si foundry alloys has recently been introduced[17], showing important and commercially attractive characteristics, such as no fading (reduced grain refining efficiency with time). The essence of this method is that boron is added as a silicon-based master alloy containing about 1 to 2 wt pet boron. Since the solid solubility of boron in silicon is about 1 wt pet, dissolving the addition is expected to give a uniform dispersion of boron in the liquid. Only a small amount of boron is necessary to give a small grain size. The method is being introduced commercially under the brand name SiBloy. Although the method of grain refinement has been known for decades, its influence on the castability has still not been established, possibly because test methods affect experimental results. Loper[7] and Loper and Prucha[8] states that adding grain refiners containing Ti reduces the fluidity of Al-Si casting alloys, although experimental confirmation of this statement is missing. Mollard et al.[10] reports a reduction in fluidity of an Al-4.5 pet Cu alloy(195) when adding 0.15 pet Ti using a vacuum fluidity test apparatus. Lang[9], on the other hand, expected an increase in fluidity when grain refining agents were present because of the reduced possibility of growing large dendrites. Lang [9] also measured a significant increase in fluidity with boron additions in the range of 0.04 to 0.07 pet to Al-Si alloys, tested with a bar casting. Alsem et al.[18] took a different approach to measuring fluidity using a mold shaped like a tuning fork. An addition of 0.04 pet Ti as AlTi5Bl to an AlSi7 alloy gave a considerable increase in the shortest running lengths, as compared to the alloy without additions. The other lengths were reported to be unchanged. This article presents an experimental investigation of the effects of grain refining - 82 - additions on dendrite coherency and fluidity and is a first step in investigating the correlation between dendrite coherency and castability of aluminium alloys. Although the solidification conditions during spiral testing are quite different from those of dendrite coherency measurements, especially with respect to fluid flow and cooling conditions, the underlying solidification mechanisms are expected to be crucial. The objective of this experimental work consist of two parts: first, to investigate the effect of grain refinement on the fluidity of two hypoeutectic Al-Si alloys, AlSi7Mg and AlSillMg, obtained with addition of AlTiSB 1 or SiBloy; and second, to investigate if there is a relationship between fluidity and dendrite coherency and grain size of the alloys. II. EXPERIMENTAL A. The Experimental Alloys Two hypoeutectic Al-Si alloys, AlSi7Mg and AlSil lMg, have been investigated in this study. The AlSi7Mg alloy is very similar in compositon to a 357-type alloy in the alloy designation system of the Aluminum Association[21]. Table I gives the content of the major alloy elements in all base alloys investigated. Table I. Main constituents [wt pet]. M = Model alloys, C = Commercial alloys, and B = SiBloy. Alloy Si Mg Fe Ti B A1 M7 6.86 0.187 0.150 0.007 0 bal. Cl 6.80 0.174 0.150 0.129 0 bal. B7 7.11 0.244 0.086 0 0.0153 bal. Mil 11.17 0.185 0.131 0.013 0.0001 bal. Cll 11.20 0.172 0.130 0.117 0 bal. Bll 10.89 0.117 0.100 0 0.0160 bal. The basis of this investigation was two commercial alloys (C), supplied by Fundo a.s. Three grain refiner additions (AlTiSB 1) were investigated: 0, 0.05, and 0.20 pet Ti, giving total titanium contents of 0.12, 0.17, and 0.32 wt pet. Due to the relatively high initial content of titanium in the commercial alloys, Ti-free model alloys (M) with similar nominal compositions were made. In these alloys, three grain refiner additions - 83 - were investigated: 0, 0.05, and 0.12 pet Ti, the final composition of the model alloy being similar to that of the initial commercial alloy. In addition to investigating the effect of grain refining by the AlTi5B 1 master alloy, alloys grain refined with boron (B) were studied. Thesealloys were supplied by Elkem Aluminium ANS, and no extra additions were necessary since boron was added during the industrial production. B. Test Apparatus and Procedure for Fluidity Measurements The spirals were moulded in quartz sand, with a gating system developed to minimize problems with entrapped gas in composite materials[19]. Figure 2 shows the mould design. The gating system consists of a pouring cup, a rectangular tapered sprue, a runner, an open riser, and a choke. The riser is intended to trap bubbles formed by turbulence during filling of the mould, before they enter the spirals. The two archimedian spirals, each with a cross section of 4 X 10 [mm], consisted of 3.5 turns, giving a maximum running length of 120 cm each. Both of the spiral ends were vented. The running length was defined as the length of the spiral from the tip to the riser. Pouring basin Riser acting as a bubble trap Runner Choke Figure 2. Principal sketch of the spiral mould. The alloys were melted in an induction furnace. Alloy additions were made directly to the melt, and the contact time for the AlTiSBl grain refiner was 30 minutes. The melt was poured when the contact time was reached and the temperature was correct. -84- Three double spirals were cast for each alloy, i.e., a total of six spirals for each heat. The maximum temperature was measured with a digital thermometer located in the pouring cups of the first and last moulds during pouring. The experiments were performed at a superheat of about 75 °C, relative to the equilibrium liquidus, giving pouring temperatures of 700 °C and 670 °C for AlSi7Mg and AlSillMg, respectively. C. Determination of the Dendrite Coherency Point Small pieces (1 to 2 kg) of the alloy ingots were separately melted and stabilized at 750 °C to 800 °C in a resistance furnace. Except for the alloys grain refined with boron only, grain refinement was achieved by adding an AlTiSBl master alloy followed by vigorous stirring using a graphite rod. The contact time for the first sample was 30 minutes. Two minutes before sampling, the melt was stirred for approximately 30 seconds. Two experimental techniques have been used to determine the dendrite coherency point; rheological measurements and thermal analysis. Figure 3. Sketch of the setup for the Figure 4. Sections of the spirals rheological measurements. investigated metallographically. For the rheological measurements, a sample of the melt was transferred from the holding furnace to the preheated graphite crucible in the rheological setup. The experimental setup is shown in Figure 3. The thermocouple adjacent to the wall and the boron-nitride covered paddle-type steel stirrer in the centre were immersed 1 cm into the melt, as shown in the figure. When the temperature had stabilized, the furnace - 85 - was removed, and the crucible was cooled naturally in air. At a temperature of about 700 °C, the rheometer and paddle were started, with a rotation speed of 0.05 rpm. The temperature and torque values were collected simultaneously by a computer. The dendrite coherency point was identified as the fraction solid where the torque increased sharply[l,3-5]. For thermal analysis, a preheated graphite crucible was filled by immersing it into the melt. The crucible with specimen was placed on an 8-mm-thick fiberfrax felt, and two thermocouples were placed in the melt, 1 cm above the bottom, one close to the wall and the other at the center of the crucible. The thermocouples of type K (alumel- chromel), covered with a thin layer of boron-nitride to prevent reactions with the melt, were calibrated against high-purity aluminium (99.998 pet) witha melting temperature defined to 660 °C. The temperatures were monitored with a computer via amplifiers and an A/D converter. This arrangement gave a precision of ±0.7 K in the temperature measurements. The dendrite coherency point was identified by the first minimum in the difference between the temperatures at the wall and in the centre after nucleation[3,20]. In both experiments, a fiberfrax lid was placed on top of the crucibles to reduce heat loss from the melt surface. This setup gave a cooling rate of about 1 K/s just before the start of solidification. All experiments were reproduced at least three times with a new melt sample. The fraction solid was calculated from the temperature measurements using a procedure developed by Tamminen[20]. D. Metallographical Investigation Metallographic samples were made at similar experimental conditions as in the coherency measurements. All samples were anodized, and grain sizes were measured with the linear intercept method in an optical microscope. Samples from the cast spirals were cut for metallographic investigation, as shown in Figure 4. III. EXPERIMENTAL RESULTS A. The AlSi7Mg Alloys Figure 5 shows the coherency parameters for the AlSi7Mg alloys as a function of the content of titanium, with an extra point for the boron-refined alloy. As may be observed in the figure, increasing the grain refinement, i.e., increasing the grain density, gives a decrease in grain size and an increase in fraction solid at dendrite -86- coherency (fscoh) determined by rheology and thermal analysis. The alloy grain refined with boron only has the smallest grain size and the highest fraction solid at dendrite coherency. Figure 6 shows the change in running length with increasing grain refinement in the AlSi7Mg alloys. The temperature indicated for each fraction of grain refiner is the average of the temperatures recorded in the first and the last pouring cups. Table II summarizes the fluidity results for the AlSi7Mg alloys. The results indicate that the running length is reduced with increasing Ti content below 0.12 pet Ti, whereas an improvement in fluidity is observed with grain refinement above this limit. The alloy grain refined with boron has a significantly larger running length than all other alloys. AlSi7Mg AlSiTMg - 1400 - 700 Commercial alloys [°C] at r —0— Model alloy Temperature —A— Commercial alloy - + SiBloy “ —^— Temperature — 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.00 0.05 0.10 0.15 0.20 0.25 0.30 SiBloy Ti-content [wt%] Ti-content [wt%] Figure 5. Fraction solid at dendrite Figure 6. The effect of grain refinement on coherency, measured by rheology and the running length of the AlST7Mg alloys. thermal analysis, and grain size plotted vs. The average pouring temperature is shown increasing Ti content, with an extra point by the second y-axes. for the alloy grain refined with boron, for the AlSi7Mg alloys. Figure 7 (a) through (f) show the micrographs from the center of the spiral tips and at a position 20 cm from the riser for the commercial AlSi7Mg alloys, with 0, 0.05, and 0.20 pet Ti added, respectively. There are no significant differences between the samples with increased grain refinement; i.e., the grain size is relatively constant. A somewhat finer structure can be observed at the spiral tip, probably because thetip has a higher cooling rate. -87- a) b) Figure 7. (a) through (f) Micrographs of the commercial AlSi7Mg alloy and the three grain refiner levels: (a) and (b) 0, (c) and (d) 0.05, and (e) and (f) 0.20 pet Ti. The sections shown are numbers (a), (c) and (e) 3 and (b), (d) and (f) 5 in Figure 4. Section 3 is in the center of the spiral tip, and section 5 is 20 cm from the riser. Magnification 200 times. - 88 - Table II. Fluidity data for the alloys with 7 pet Si. L is the average running length of all experiments, SD denotes the standard deviation of the running lengths, Tcast is the average pouring temperature, and AT is the temperature difference between the first and the last experiment. Alloy Running Length [cm]: Average Temperature No.l No.2 No.3 No.4 No.5 No.6 L SD Tcast AT M+OTi 45.3 44.3 44.3 42.5 37.7 40.2 42.4 2.9 701 1 M+0.05Ti 44.2 43.5 46.4 44.9 26.8 28.1 39.0 9.0 708 3 M+0.12T1 41.3 41.5 40.8 40.7 42.1 41.5 41.3 0.5 702 6 C+OTi 30.8 31.2 24.2 24.6 30.4 30.1 28.6 3.2 684 6 C+0.05T1 41.2 38.8 44.1 45.3 41.0 39.0 41.6 2.7 692 2 C+0.20Ti 35.3 34.5 41.5 39.8 42.3 44.2 39.6 3.9 696 1 B 48.9 51.9 52.1 - - 49.3 50.6 1.7 706 - B. The AlSillMg Alloys AlSillMg AlSillMg Model alloys Commercial alloys - 1400 [°C] Temperature T— f s coh (rheology) _ ■■0-- Model alloy - 0- fscchCIhomaui.) _ -A— Commercial alloy _ —£>- Grain size _ + SiBloy i------— Temperature ~ 111111 i iTj m 11 j 1111 |'i i n | 1111111 1 " j " i11 ' fl'H'i 111 | " 1 1 | i i 1 ' | " i 1 | 1 i------1------0.00 0.05 0.10 0.15 020 0.25 0.30 SiBloy 0.00 0.05 0.10 0.15 0.20 0.25 0.30 SiBloy Ti-content [wt%] Ti-content [wt%] Figure 8. Fraction solid at dendrite Figure 9. The effect of grain coherency, measured by rheology and refinement on the running length of thermal analysis, and grain size plotted vs. the AlSi7Mg alloys. The average increasing Ti content, with an extra point pouring temperature is shown by the for the alloy grain refined with boron. second y-axes. AlSillMg alloys. - 89 - Figure 8 shows the effect of grain refinement on the coherency parameters of the AlSillMg alloys. The same trends as observed in Figure 5 are evident, but the reduction of grain size is much larger and thefraction solid at dendrite coherency less for AlSillMg than for AlSi7Mg. Figure 9 shows the effect of grain refinement on the running length of the AlSillMg alloys. The temperature is again the average of the pouring temperatures. As may be observed in the figure, increasing the content of titanium by adding grain refiner clearly increases the running length of the material. The alloy grain refined with boron has the best fluidity of all alloys. Table III summarizes these results and the length of each spiral. Table III. Fluidity data for the alloys with 11 pet Si. L is the average running length of all experiments, SD denotes the standard deviation of the running lengths, Tcast is the average pouring temperature and AT is the temperature difference between the first and the last experiment. Alloy Running Length, [cm]: Average Temperature No.l No.2 No.3 No.4 No. 5 No.6 L SD Tc= AT M+OTi 42.5 45.4 42.3 42.6 42.6 45.4 43.5 1.5 663 1 M+0.05T1 42.5 42.9 45.7 46.1 46.7 46.8 45.1 1.9 668 - M+0.12Ti 53.9 54.3 46.8 45.3 48.1 47.2 49.3 3.9 680 0 C+OTi 42.1 42.3 44.8 44.9 - - 43.5 1.5 673 7 C+0.05Ti 49.9 48.6 48.6 48.9 48.8 47.9 48.8 0.6 673 3 C+0.20T1 49.3 53.8 51.6 52.0 49.9 49.1 51.0 1.8 675 1 B 55.1 57.6 49.7 54.2 53.6 - 54.0 2.9 676 - In both the AlSiVMg and the AlSillMg alloy the individual difference between the two spiral lengths in the double-spiral mould was calculated to be 1 to 1.5 cm on average. When comparing the results in Figures 6 and 9, it is observed that the running length of AlSillMg is larger than for AlSiVMg, confirming the observations by Lang[9] in Figure 1. In both Figures 6 and 9 a difference in behavior is observed at 0.12 pet titanium. The initial Ti content of the commercial alloy was at this level, and the last grain refinement addition brought the model alloy to 0.12 pet titanium. The observed difference may be explained by the fact that commercial A1T15B1 grain refiners lose - 90 - some of their performance upon remelting, due to settling and agglomeration of TiB2 nucleants. At 0.12 pet Ti, the largest running length and the highest fraction solid at dendrite coherency belong to the model alloy. This can also be interpreted as an indication of the effect of grain refinement on fluidity, i.e., effective grain refinement increasing the running length. These considerations are confirmed by Tables II and III. Figure 8 shows that the alloy grain refined with boron reaches dendrite coherency significantly later and has a smaller grain size than all the other alloys. The fluidity of this alloy is considerably better, as shown in Figure 9. IV. DISCUSSION Addition of grain refiners decreases the grain size. The cooling rate also has a strong effect upon the final grain size. When the cooling rate is increased, larger undercooling may be achieved, whichcan accelerate the nucleation rate, resulting in a reduced grain size. These considerations apply when the coherency experiments and the spiral tests are compared. In the spiral test, the spiral tip is continuously exposed to the cold mould channel at room temperature as it flows, producing a high cooling rate and melt undercooling, due to the small cross-sectional area. In the coherency experiments, on the other hand, the cooling rate is about 1 K/s, which is probably much lower than that experienced by the tips of the spirals. This implies that the differences in grain size effected by grain refinement are less profound at the higher undercoolings in the spirals. The cooling rate also influences the dendrite coherency point, since the large dendrite growth rate obtained at a high cooling rate makes the dendrites impinge earlier, decreasing the fraction solid at dendrite coherency [3-4], The effect of material flow and deformation is also very important to the development of microstructure in the spirals. The solid microstructure starts to develop as soon as the liquidus temperature is reached and continues as the material flows along the spiral channel. This must ultimately lead to significant stresses being imposed on the growing dendrites as they move and interact with each other. Fragmentation and deformation of the dendrites result from mutual mechanical contacts and by remelting because of the turbulent flow. As the dendrites become coherent, two things can be pictured. The dendrite network which forms may be strong and stiff enough to resist the imposed stresses, hereby stopping further macroscopic flow. Alternatively, the dendrite network may be deformed by the flow pressure. As has been shown here and in the literature[l-6], the dendrite coherency point is affected by several parameters. As an example, the development of a coherent dendrite network by grain impingement is postponed if the alloy is well grain refined and/or -91 - is slowly cooled. Of course, the structure and morphology at the dendrite coherency point is also very different, depending on when the network was established. An early coherency point can usually be related to a highly branched structure, suggesting very weak and thin dendrite interconnections, while a late coherency point is related to a more globular structure with broader contact areas between the dendrites and only small interdendritic areas. The micrographs in Figure 7 show that there are no characteristic, observable structural differences in the spirals with increasing grain refinement. The results clearly show that the running length is increased with increasing grain refinement without giving any obvious change in structure after the end of solidification. A very likely explanation of this tendency is the effect of fragmentation and network breakage. Breaking the dendrite network requires some stress, which ultimately leads to energy dissipation, reducing the driving force for material transport. This mechanism for material transport is termed "burst feeding"[15]. Since the structure becomes more globular with increased grain refinement, the likelihood of fragmentation decreases. In the non-grain-refined alloy, the structure is highly branched, especially at high growth rates and when fluid flow alters the diffusion fields ahead of the growing dendrite tips. Another factor which is also important is how the strength develops when the dendrite network is established, i.e., if it increases slowly or rapidly. A slower increase in strength in AlSillMg compared to AlSi7Mg provides an explanation of the slightly higher fluidity reported for AlSil lMg. This effect can be explained by the higher heat of crystallization liberated during solidification of the AlSillMg alloy, due to the higher silicon content, which can lead to softening and partial remelting of the dendrite network, as has been observed by Chai[3], The difference in latent heat between AlSi7Mg and AlSillMg is in the range of about 50 J/g (13 pet). In addition, the dendrites are longer, thinner, and more branched in the case of AlSil lMg. Therefore, the coherent network of the AlSil lMg alloy is susceptible to shearing, which helps to explain the longer spiral length. - 92 - AlSiTMg AlSillMg Ti-content [wt%] Ti-content [wt%] Figure 10 (a) Temperature adjusted running lengths for the AlSi7Mg alloys, corrected to a pouring temperature of 700 °C. (b) Temperature adjusted running lengths for the AlSillMg alloys, corrected to a pouring temperature of 670 °C. The effect of pouring temperature also has to be considered when evaluating fluidity. In Figure 6, for AlSi7Mg, and in Figure 9, for AlSillMg, the average pouring temperature for each grain refiner content has been plotted. Kolsgaard[19] has reported that a 10 °C increase in pouring temperature increases the running length of AlSiVMg by 7 cm, for the identical spiral mould. This is a very important factor, being about 15 pet of the experimental running lengths and even larger than most of the measured changes with grain refining additions. In Figures 10 (a) and (b) the temperature- adjusted running lengths are shown for the AlSi7Mg and AlSillMg alloys, respectively, where the identical temperature correction was assumed for the AlSillMg alloy. Although this correction is only an approximation, it gives an idea of the influence of the pouring temperature. From both Figures 10 (a) and (b) it can be observed that the fluidity is somewhat reduced up to a titanium content of 0.12 pet, but an increase of the titanium content above this limit, or grain refinement with -0.015 pet B, increases the fluidity. There seems to be a transition at around 0.12 pet Ti. If the results in Figure 10 are compared to the coherency results for thealloys, in Figures 5 and 8, a relationship between dendrite coherency and fluidity is observed above 0.12 pet Ti for both alloys. An increased fraction solid at dendrite coherency corresponds directly to increased fluidity. Below 0.12 pet Ti, the tendencies are opposite, i.e., postponed coherency and reduced running length. No satisfactory explanation of this - 93 - relationship is known, but it might be structurally and morphologically related, changing the shear strength of the developing solid network. The liquidus temperature of the alloys might also have an influence on the fluidity. Grain refinement is related to an increase in nucleation (or liquidus) temperature, which means that the material starts to solidify at a higher temperature. This temperature increase is in the range of about 3 °C to 5 °C, depending on the amount and type of grain refiner[17], and corresponds to an increase in running length of about 2 to 3 cm. It may be concluded that the pouring temperature has a much stronger effect on fluidity than grain refinement in the spiral test. However, casting of thin-walled sections may be expected to be improved by efficient grain refinement. As mentioned in Section I, Mollard et al.[10] reported a decrease in fluidity with an addition of 0.17 pet Ti to an AlCu4.5 alloy. As shown by Hoefs et al.[22] grain refinement of AlCu4 by addition of AlTiSBl only produces a small reduction in grain size, and the initial grain size is comparatively much smaller than in AlSi7Mg (about half) and AlSil 1 (about one third) alloys. The small change in grain size and the very globular structure in AlCu4.5 compared to the AlSiMg alloys provide an explanation of this disagreement, because the coherency and strength of the network do not change significantly for the AlCu alloys. The effects of the oxide skin and dissolved hydrogen on the fluidity are uncertain. The strength of the oxide skin enclosing the leading tip of the flowing stream is a factor that may restrict the movement of the material, since the energy barrier to break this skin has to be exceeded for flow to continue. In the spiral test, the driving pressure for flow is high, giving a relatively high flow rate of the metal. This pressure could also influence the liberation of dissolved gas from the liquid during solidification. During pouring and filling of the moulds, oxidation of the metal will take place and the oxides will be mixed with the flowing stream. In this way, nuclei for gas bubbles are present and become active as soon as the pressure barrier for nucleation is exceeded. As soon as gas bubbles are formed in the material, the driving pressure for flow may be reduced by an amount corresponding to the back pressure from the bubbles. This is an effect which is difficult to evaluate, since the gas contents of thematerials were not measured. No systematic investigations of the effect of dissolved gas on fluidity have been reported in the literature, but a reduction would be expected. - 94 - V. CONCLUSIONS Increasing the grain refinement postpones dendrite coherency and reduces grain size. The dendrites impinge at a higher fraction solid. However, the fluidity does not show a unique relationship to grain refinement in the alloys investigated, AlSi7Mg and AlSillMg. Grain refinement additions with an AlTi5Bl master alloy indicate a transition point at 0.12 wt pet Ti. Above this level fluidity is improved with grain refinement, whereas it is impaired with titanium additions below 0.12 pet Ti. AlSi7Mg and AlSillMg alloys grain refined with boron show superior properties compared with those grain refined with A1T15B1, since they have a significantly higher fraction solid at dendrite coherency, a smaller grain size, and a better fluidity. Some of the changes in fluidity observed with grain refinement can be related to the developing rheological properties of the material, such as coherency and strength, and to the increase in liquidus temperature with grain refinement, reducing the melt superheat. ACKNOWLEDGMENTS The financial support from the Norwegian Research Council, Elkem Aluminium, and Hydro Aluminium is gratefully acknowledged. REFERENCES 1. G. Chai, T. Rplland, L. Amberg and L. Backerud, Processing of Semi-Solid Alloys and Composites, Eds. S.B. Brown and M.C. Flemings, MIT, 1992, pp. 193-201 2. L. Amberg, G. Chai and L. Backerud, Mat. Sci. Eng., vol. A173, 1993, pp. 101- 103 3. G. Chai, Ph.D. thesis, Stockholm University, Chem. Com., No.l, (1994) 4. A.K. Dahle and L. Amberg, The 4th Int. Conf. on Aluminum Alloys, Eds. T.H. Sanders Jr. and E.A. Starke Jr., Atlanta, GA, 1994, pp. 91-98 5. A.K. Dahle and L. Amberg, SINTEF Report No. STF34 A93248, Trondheim, Norway, 1993 6. G. Chai, L. Backerud and L. Amberg, Z. Metall, vol. 86, 1995, pp. 54-59 7. C.R. Loper Jr., AFS Trans., vol. 100, 1992, pp. 533-538 8. C.R. Loper Jr. and T.E. Prucha, AFS Trans., vol. 98, 1990, pp. 845-853 9. G. Lang, Aluminium, vol. 48, No. 10, 1972, pp. 664-672 - 95 - 10. F.R. Mollard, M.C. Flemings and E.F. Niyama, J. Met., November, 1987, pp. 34- 37 11. M.C. Flemings, E. Niyama and H.F. Taylor, AFS Trans., vol. 69, 1961, pp. 625- 635 12. B.C. Pai and H. Jones, Solidification Processing 1987, The Institute of Metals, London, 1988, pp. 319-322 13. P. Bastien, J.C. Armbruster and P. Azou, AFS Trans., vol. 70, 1962, pp. 400-409 14. F.L. Arnold, J.L. Jorstad and G.E. Stein, Current Eng. Pract., July 1963, pp. 10- 15 15. J. Campbell: Castings, Butterworth-Heinemann Ltd, Oxford, 1991 16. M.C. Flemings: Solidification Processing, Mc-Graw Hill Book Co., 1974 17. P.A. Tpndel, Ph.D. thesis, Norwegian Institute of Technology, 1994 18. W.H.M. Alsem, P.C. van Wiggen and M. Vader, Light Metals 1992, Ed. E.R. Cutshall, The Minerals, Metals and Materials Society, 1991, pp. 821-829 19. A. Kolsgaard, Ph.D. thesis, Norwegian Institute of Technology, 1993 20. J. Tamminen, Ph.D. thesis, Stockholm University, Chem. Com., No.2, (1988) 21. Registration Record of Aluminum Association Alloy Designations and Chemical Composition Limits for Aluminum Alloys in the Form of Castings and Ingot, The Aluminum Association, July 1985 22. P. Hoefs, W. Reif and W. Schneider, Giesserei, vol. 84, No. 12, 1994, pp. 398- 406 Article #5 A.K. Dahle, S. Karlsen and L. Amberg "Effect of Grain Refinement on the Fluidity of Some Binary Al-Cu and Al-Mg Alloys" accepted for publication in Cast Metals, 1996 Effect of Grain Refinement on the Fluidity of Some Binary Al-Cu and Al-Mg Alloys A.K. Dahle, S. Karlsen and L. Amberg Department of Metallurgy, The Norwegian Institute of Technology, N-7034 Trondheim, Norway The effect of grain refining, by A1T15B1 master alloy additions, on the fluidity of two binary Al-Cu and Al-Mg alloys has been assessed by a vacuum fluidity test apparatus. Five levels of grain refinement were investigated in each alloy; AlMgl, AlMg5, AlCu2 and AlCu4. In AlCu4 and AlMg5, fluidity initially decreased with increasing grain refinement, reaching a minimum at 0.03wt% Ti. Increasing the grain refinement above this level did not give any significant increase in fluidity. In AlMgl, the minimum was displaced to 0.05wt% Ti. Fluidity decreased continuously with increasing grain refinement in the AlCu2 alloy. However, for all alloys, the fluidity was lower at the highest grain refiner content than in the unrefined alloy. Increasing the content of the major alloy element decreased the fluidity. Introduction Performing fluidity testing was once a necessity for mapping the casting properties of alloys used in the foundries, but is now more important for the production of thin- section castings and infiltration of fiber-preforms. The length a solidifying metal flows in a channel with a small cross-sectional area is defined as the fluidity. Over the years many different moulds and geometries have been developed to measure fluidity. In general it has been established that the thermal conditions, such as superheat, latent heat and cooling rate, have a controlling influence on the length of flow. Despite the long history of fluidity testing, the effect of grain refinement on fluidity is still controversial. Very contradictory experimental results exist in the literature. Compared to other variables, grain refining does not give large changes in fluidity, and this can obscure the measurements. This paper is therefore aimed at investigating the effect of increasing grain refinement in some binary Al-Cu and Al-Mg alloys, since it is of both practical and theoretical interest to know its effect. The theoretical side of it is concerned with the mechanism by which the flow is stopped. - 97 - Background Many reviews of the factors controlling fluidity have been presented in theliterature, 1"3 most recently by Campbell.4 The factors determining fluidity can basically be divided into metallurgical and mould/casting variables. Among the metallurgical factors are the composition, superheat, latent heat, surface tension of the melt (including oxide film) and mode of solidification. With respect to the metallurgical factors it is very important to realize that fluidity has been shown to be independent of the viscosity of the liquid, 5 i.e., the definition of the term fluidity in casting science is different from the formal meaning, which is to be the inverse of viscosity. Grain refinement is an effective way to improve the quality of cast aluminiu m products. Usually this structural refinement is achieved by additions of master alloys containing titanium and boron. The grain refiners contain effective nuclei for growth of solid crystals during the cooling and solidification process, which may effect a transformation from a directionally growing columnar solidification front to independently growing equiaxed crystals (GET). Hunt 6 analyzed and modeled the conditions for this transformation, concluding that the conditions for equiaxed dendrite growth were more dependent on the number than on the efficiency of the nucleation sites at large temperature gradients (G) and growth rates (V). Grain refiners will also reduce the grain size and increase the structural uniformity, which is generally believed to improve the product performance. Among the improved properties are increases in resistance to hot-tearing, improved feeding ability, refined distribution of porosity and precipitates and improved mechanical properties. 7,8 The solidification conditions, i.e., the constitutional and cooling conditions, producing either a directionally growing solidification front or separately growing equiaxed dendrites, were shown by Flemings et al.1,9 to be of major importance for the mechanism of flow stoppage in the fluidity test. In the case of a directionally solidifying front, flow is stopped somewhere behind the flow tip, where the fronts meet and freeze off the flow path. This is illustrated for the case of columnar dendrites in Figure 1 a). When the alloy solidifies with equiaxed dendrites, the crystals are proposed to be carried with the flow, and since the flow tip experience the fastest cooling, crystals will form first here and grow fastest. The flow is expected to stop when these crystals impinge and form a coherent dendrite network at the tip, as illustrated in Figure 1 b). Of course, thisis an idealized consideration, and intermediate situations have been observed, where a combination of equiaxed crystals and columnar dendrites stop the flow somewhere between the flow tip and the channel entrance, as illustrated in Figure 1 c). - 98 - a) Columnar dendrites b) Equiaxed dendrites Fig. 1 Mechanisms for flow stoppage in the fluidity test for different solidification modes; (a) columnar dendritic growth; (b) equiaxed dendritic growth; and (c) mixed case. Fluidity generally shows a relationship inversely proportional to the solidification range. 1,3,10 This is related to several transitions in solidification morphology, i.e., planar front, columnar and equiaxed growth, and also a transition in the mechanism of flow stoppage. When equiaxed dendrites form at, and flow with, the tip, the viscosity of the flowing mixture increases and the fluidity is decreased. Similarly, with the transition from planar front growth to dendritic growth, the resistance/friction to flow increases and fluidity decreases. Dendritic growth is the most common mode of solidification in commercial casting - 99 - processes. 11 The dendrites grow at a rate that is determined by the undercooling ahead of the tips (aT), consisting of a thermal (aT,), a constitutional (aTc) and capillary (aT,.) contribution. Adding alloying elements will increase the constitutional contribution, while the thermal contribution is determined by the cooling rate. The aluminium casting alloys used in the foundries today are highly alloyed long- freezing range alloys, usually solidifying with equiaxed dendrites, as also predicted by Hunt ’s model. 6 The equiaxed dendrites initially grow separately, releasing their latent heat into the liquid. However, independent growth cannot continue indefinitely, and they therefore impinge on each other at a certain temperature, called the dendrite coherency point. 12 The point of impingement, of course, depends on the separation of the crystals, i.e., the grain size, and the growth rate of the dendrite tips.13"15 A method hasbeen developed, allowing accurate and reproducible determination of the dendrite coherency point based on utilizing the change in mechanical properties associated with the establishment of a dendrite network. 13'15 With thismethod the effect of alloying, cooling rate, modification and grain refinement has been established. Dendrite coherency usually occurs at solid fractions between 10 and 50%, and more often in the lower range. Amberg et al.13-15 have shown that grain refinement postpones dendrite coherency, i.e., giving network formation at lower temperatures (T) and larger fraction solid (fs). From these results it would be expected from the proposed mechanism for flow stoppage with equiaxed growth, that grain refinement would improve fluidity, since the flow should stop when the dendrites impinge. 1,9 However, contradictory results have been reported in the literature. Mollard et al.2 reported a reduction in fluidity when 0.15% Ti was added to an Al- 4.5% Cu alloy, tested with a vacuum fluidity apparatus. Tiryakioglu et al.16 found no effect of grain refinement in 319 and A356 alloys, adding approximately 0.04 wt% Ti as A1T15B1, tested in a sand spiral test. Lang 10 found a significant increase in fluidity with boron additions in the range of 0.04-0.07% to Al-Si alloys, tested with a bar die casting. Alsem et al.17 used a tuning fork-type of casting, and a considerable increase in the shortest running lengths were observed with a 0.04% Ti addition as AlTi5Bl to an AlSi7 alloy. Dahle et al.18 observed a more complex variation in fluidity with successive grain refining additions, as AlTi5B 1, in AlSi7Mg and AlSil lMg tested with sand spirals. Fluidity was reduced with grain refinement below 0.12% Ti, while it increased with additions above 0.12% Ti. Alloys grain refined only by boron showed the smallest grain size and the best fluidity. A mathematical model for estimating the fluidity of an alloy solidifying with equiaxed - 100 - dendrites flowing with the tip was proposed by Flemings et al.1,9 It was based on the assumptions that the crystals form at the tip and are carried with it, that flow stops when a critical solid fraction is reached and a constant flow velocity until flow stoppage. This gave the following equation for the length of flow (Lf), i.e., the fluidity 9: (1) where (2) This equation was shown to reproduce the general effects of superheat, latent heat, flow velocity and channel geometry. The critical solid fraction was expressed by:9 (3) Flemings, Niyama and Taylor 9 plotted the critical solid fraction vs. pressure head, and observed that it was increasing, converging towards about 35% solid. This range of solid fractions compares well to those where dendrite coherency has been measured experimentally. Experimental Procedure A vacuum fluidity test device was built, similar in principle to equipment described in the literature.19"21 This method was selected since it has been shown to be the most controllable and reproducible way of measuring fluidity. Since the grain refining additions are expected to produce relatively small changes in fluidity, it is crucial that the other variables are kept constant and under control. Figure 2 shows a schematic of the apparatus. The principle of the method is that the melt is sucked into a small glass tube, and the length of flow in the tube is a measure of fluidity. - 101 - Fig. 2 Schematic illustration of the vacuum fluidity test apparatus. Base alloys were prepared from commercial purity aluminium (99.85%), pure magnesium and electronic grade copper. Two Al-Cu alloys, AlCu2 and AlCu4, and two AlMg alloys, AlMgl and AlMg5, were made. Grain refinement was achieved by adding AlTi5B 1 master alloy rods to the melt of approximately 3 kg. To avoid the fading effect, each grain refiner addition was made to a new, unrefined, base alloy. Four different levels of grain refinement were investigated in addition to the pure alloy: 0.01, 0.03, 0.05, and 0.12 wt% Ti. The alloy was melted in a resistance furnace, and the alloy additions were made directly to the melt. Nitrogen gas was purged over the melt surface to reduce oxidation. A lid reduced melt radiation. The flow tube was made from Pyrex glass, with an inner diameter of 7 mm and an outer diameter of 10 mm. The tubes were reproducibly bent on one end to an angle of 90° around a radius of 20 mm, and finally stress relieving annealed. Just before use, the tubes were carefully cleaned with soap, and finally dried in pure nitrogen. A rotary vacuum pump was connected to a glass assembly with several valves. The desired pressure was set by evacuating a vacuum reservoir of 1 liter, which was large enough to prevent a significant drop in pressure when the tubes were evacuated. The pressure was read with a pressure gauge with an accuracy of 0.005 bars. - 102 - The tube was mounted in a tilter to ensure that it was level when put down into the melt. The tilter was made from stainless steel, with alumina pieces closest to the glass tubes. The alumina pieces were used to avoid the tube holding system from affecting the cooling rate and thermal gradients during testing. All experiments were performed at a constant superheat of 40°C above the equilibrium liquidus of the pure binary alloys (assumed to have 1 or 5 wt% Mg or 2 or 4 wt% Cu), and a pressure difference of 0.05 bars was used. The melt was stirred for 30 seconds two minutes before the experiments were started, and the tubes were immersed 1 cm below the melt surface two seconds before they were evacuated. At least five separate experiments were performed for each percentage of grain refiner. Metallographic examination was performed on longitudinal sections of the solidified samples. All alloys were investigated at the tip, and some of them also close to the entrance. Samples were also cast for chemical analysis by optical emission spectroscopy. Experimental Results 1. Al-Mg alloys Table I shows the measured contents of Mg and Ti and the running lengths, including standard deviation, for each grain refiner percentage in the AlMgl alloy. The average running length (Lf) varies from 26.8 cm at 0.00 wt% Ti to 23.0 cm at 0.063 wt% Ti. The largest standard deviation was 0.7 cm, and the smallest 0.4 cm, i.e., 3.0 and 1.6%, respectively. The Mg contents vary between 0.819 and 1.196 wt%. Thereare also some differences between the nominal grain refinement and the analyzed Ti-content. Table I. Measured fluidity (Lf) and Mg and Ti concentrations in the AlMgl alloys. Nominal grain Lf ± c wt% Mg wt% Ti refinement [wt% Ti] [cm] 0 26.8+0.8 0.960 0.000 0.01 25.5±0.5 0.819 0.013 0.03 25.6±0.4 1.014 0.038 0.05 23.0+0.7 0.950 0.063 0.12 23.4±0.4 1.196 0.132 Figure 3 shows micrographs from the tip of the flow of the fluidity test samples for - 103 - eachfraction of grain refiner in the AlMgl alloy. A significant difference in grain size between the samples is observed. The decrease in grain size with increasing grain refinement is most obvious between the samples going from 0.013 to 0.038 wt% Ti and from 0.063 to 0.132 wt% Ti. Shrinkage porosity is readily observed at the tip in the grain refined samples, decreasing in size with increasing grain refinement. Figure 4 shows micrographs moving towards the entrance in the unrefined AlMgl alloy. A coarser structure with larger grains is observed towards the entrance, indicating that the tip experiences the most rapid cooling. The tendency towards columnar growth from the channel walls increases with increasing distance from the tip. Porosity is observed along the centerline. Fig. 3 Micrographs of the flow tips of the AlMgl alloy with increasing grain refinement; (a) 0.000 wt% Ti; (b) 0.013 wt% Ti; (c) 0.038 wt% Ti; (d) 0.063 wt% Ti; and (e) 0.132 wt% Ti. Table II shows the measured contents of Mg and Ti and therunning lengths, including standard deviation, for each grain refiner percentage in the AlMg5 alloy. In this alloy the average running length varies between 20.7 cm at 0.005 wt% Ti to 18.1 cm at 0.035 wt% Ti. The largest standard deviation is 1.0 cm (4.8%), and the smallest is 0.2 cm (1.1%). The Mg-content varies between 4.759 and 5.240 wt%. - 104 - Fig. 4 Micrographs moving sequentially from the flow tip towards the channel entrance in the unrefined AlMgl alloy. The distance from the flow tip is given on the right hand side in cm. - 105 - Table II. Measured fluidity and Mg and Ti concentrations in the AlMg5 alloys. Nominal grain Lf ± c wt% Mg wt% Ti refinement [wt% Ti] [cm] 0 20.7+1.0 5.032 0.005 0.01 19.9+0.7 4.759 0.015 0.03 18. 1±0.2 5.190 0.035 0.05 18.6±0.9 5.240 0.060 0.12 18.9+0.4 4.870 0.139 Figure 5 shows micrographs from theflow tip at each grain refiner level in the AlMg5 alloy. A gradual decrease in grain size with increasing grain refinement is observed. A marked morphological transformation appears to occur between 0.015 and 0.035 wt% Ti, changing from highly branched to more globular dendrites. Fig. 5 Micrographs from the flow tips at increasing grain refinement in the AfMg5 afloy; (a) 0.005 Ti; [6] 0.075 wf% Ti; W 0.055 Ti; (WJ 0.060 wt% Ti; and (e) 0.139 wt% Ti. The running lengths of the AlMg alloys are shown vs. Ti-content in Figure 6. The general trend for the AlMgl alloy seems to be a reduction in fluidity with grain - 106 - refinement up to a Ti-content of 0.063wt% Ti, where the fluidity is 14.2% less than that of the unrefined alloy. The fluidity in the AlMg5 alloy is reduced with increased grain refinement up to a Ti-content of 0.035 wt% Ti, where fluidity is 12.6% less than in the unrefined alloy. Following the initial decrease in fluidity with increasing grain refinement the fluidity seems to stabilize. For both alloys, a slight increase in the average fluidity with grain refinement was observed in this region. However, the effect is within the range of the standard deviation. In both AlMg alloys the fluidity at the highest level of grain refinement is less than in the unrefined alloy. AlMgl - AlMg5 - 0.000 0.020 0.040 0.060 0.080 0.100 0.120 0.14 wt% Titanium Fig. 6 Fluidity vs. grain refinement (wt% Ti) for the two AlMg alloys. 2. Al-Cu alloys Table III shows the measured contents of Cu and Ti and the running lengths, including standard deviation, for each grain refiner percentage in the AlCu2 alloy. The running lengths vary between 26.0 cm at 0.002 and 0.036 wt% Ti to 22.6 cm at 0.132 wt% Ti. The largest standard deviation is 1.1 cm (4.2%), and the smallest is 0.3 cm (1.2%). The copper content varies between 1.679 to 1.965 wt%. Figure 7 shows how increasing grain refinement affects the structure at the flow tip of the AlCu2 alloy. The decrease in grain size seems most marked between 0.036 and 0.053 wt% Ti. The amount of microporosity seems less than in the AlMg alloys. Figure 8 shows micrographs with increasing distance towards the entrance in the unrefined AlCu2 alloy. The grain size increases towards the entrance, and little columnar growth is observed. - 107 - Table III. Measured fluidity and Cu and Ti concentrations in the AlCu2 alloys. Nominal grain Lf ± a wt% Cu wt% Ti refinement [wt% Ti] [cm] 0 26.0±0.9 1.782 0.002 0.01 24.9+0.5 1.749 0.014 0.03 26.0+1.1 1.704 0.036 0.05 24.2±0.3 1.965 0.053 0.12 22.6±0.4 1.679 0.133 Fig. 7 Micrographs from the flow tips at increasing level of grain refinement in the AlCu2 alloy, a) 0.002 wt% Ti, b) 0.014 wt% Ti, c) 0.036 wt% Ti, d) 0.053 wt% Ti and e) 0.133 wt% Ti. Table IV shows the measured contents of Cu and Ti and the running lengths, including standard deviation, for each grain refiner percentage in the AlCu4 alloy. The average running length varies between 22.9 cm at 0.006 wt% Ti to 21.5 cm at 0.019 and 0.031 wt% Ti. The largest standard deviation is 0.8 cm (3.7%) and the smallest is 0.5 cm (2.2%). The copper content varies between 3.86 and 4.39 wt%. - 108 - Fig. 8 Micrographs of the flow channel in the unrefined AlCu2 alloy. The numbers on the right hand side indicate the distance from the flow tip in cm. 109 - Table IV. Measured fluidity and Cu and Ti concentrations in the AlCu4 alloys. Nominal grain Lf + c wt% Cu wt% Ti refinement [wt% Ti] [cm] 0 22.9±0.5 4.35 0.006 0.01 21.5+0.5 4.30 0.019 0.03 21.5+0.5 4.39 0.031 0.05 22.1+0.7 186 0.047 0.12 21.7+0.8 4.25 0.122 Micrographs from the tips of each grain refiner fraction in the AlCu4 alloy are shown in Figure 9. A decrease in grain size accompanies increasing grain refinement. Shrinkage porosity is observable. Fig. 9 Micrographs of the flow tips at increasing fractions of grain refinement in the AlCu4 alloy, a) 0.006 wt% Ti, b) 0.019 wt% Ti, c) 0.031 wt% Ti, d) 0.047 wt% Ti and e) 0.122 wt% Ti. Figure 10 shows the results for the AlCu alloys. The fluidity of the AlCu2 alloy shows a general reduction in fluidity with increasing grain refinement. The smallest fluidity is observed at the highest level of grain refinement, 0.133 wt% Ti, where the fluidity is 13% less than in the pure alloy. A strange increase, attaining the same fluidity as the unrefined alloy, is observed at 0.036 wt% Ti. In the AlCu4 alloy, fluidity initially drops with grain refinement, giving equal fluidities at 0.019 and 0.031 wt% Ti, 6.1% - 110 - less than that of the unrefined alloy. Fluidity seems to reach a plateau above this level, and average values indicates an improved fluidity. However, the increase is probably not significant since it is within the standard deviation. In both AlCu alloys, fluidity is less at the highest level of grain refinement than in the unrefined alloy. AlCu2— - -A - AlCu4 — 1 I 1 ' I 0.020 0.040 0.060 0.080 0.100 0.120 0.14 wt% Titanium Fig. 10 Fluidity vs. grain refinement (wt% Ti) in the AlCu alloys. Discussion All alloys investigated in this paper show a tendency to equiaxed dendrite growth, evident by a clearly equiaxed structure at the tips of all castings. With a possible exception of the unrefined AlMgl alloy, none of the investigated alloys exhibit a large macroshrinkage at the flow tip that is often linked to the mechanism of flow stoppage by columnar growth, illustrated in Figure 1 a). Porosity at the tips instead appears as micropores on the dendrite boundaries, similar to the case illustrated in Figure 1 b). With the equiaxed mode of solidification, the theory of Flemings et al.1,9 for flow stoppage illustrated in Figure lb) should apply. According to this theory the dendrites at the tip stop the flow when they reach dendrite coherency. At this point the dendrites impinge on each other and form continuous dendrite interconnections, transforming the mixture into more solid-like behaviour. With the formation of a dendrite network the dendrites are no longer free to move, and flow stops since interdendritic flow has a negligible impact on fluidity. Campbell4 has suggested that this critical solid fraction for most practical situations lies between 20 and 50%, which agrees well with the solid fractions at dendrite coherency reported in the literature.13"15 - Ill - In the micrographs from the flow tips it is readily observed that the grain sizes decrease with increasing grain refinement in all alloys. Measurements of dendrite coherency have shown that grain refinement is an effective way to increase the solid fraction where the dendrites impinge. 13"15 Within thealloy system the reduced grain size promotes a reduction of the dendrite growth rate since the volumetric transformation rate is similar and governed by the rate of latent heat removal. Solute elements are enriched in the interdendritic areas during solidification. Since dendrite coherency is postponed with grain refinement it would be expected from the proposed mechanism of flow stoppage that fluidity would increase with increased grain refinement, but such a clear relationship does not seem to the case. The results could agree with the theory above a certain limit of grain refinement where the average values indicate a small increase of fluidity. The increase, though, is within the standard deviation of the measurements. Below this limit the experimental results indicate just the opposite trend, i.e., increased grain refinement giving reduced fluidity. The same tendency was observed by Dahle et al.18 in Al-Si alloys. Fluidity has been shown to be inversely proportional to the interval of solidification. 1,2,10 In the experiments performed in this investigation, the content of the main alloying element, i.e., Mg or Cu, is somewhat different in each experimental level of grain refinement. However, these compositional variations are quite small, and correcting the fluidity length with an assumption of a linear relationship between fluidity and composition in the investigated composition range would not affect the results significantly. The fluidity test is a dramatic casting event. The flow velocity is high, giving turbulent flow in the test channel. In addition the cooling rate is also high, producing high undercoolings in the metal, and also a short solidification time. The flow tip would experience these effects the most, being continuously exposed to new and cool parts of the tube. There may therefore be several sources of new grains, in addition to those created by the grain refiner. The metal may contain impurities that may become nucleation sites at certain undercoolings, and fragmentation of already growing crystals may also produce new grains. All these mechanisms of grain formation may operate at the flow tip, since the undercooling may be large and since shearing forces may be exerted on the growing dendrites flowing with the tip. Since there are many favourable conditions for grain formation at the tip, the grain size is small. Behind the flow tip, towards the channel entrance, the solidification conditions are not that extreme and the grain size is larger, as illustrated in Figure 11. Here the grain refiner will play a more important role in reducing the grain size. Adding grain refiners would therefore be expected to reduce the difference in grain size along the fluidity channel. More - 112 - effective nuclei would be activated at smaller undercoolings when grain refining additions are made, and saturation levels may be reached sooner. Since the dendrites are less branched at high levels of grain refinement, the importance of fragmentation is probably reduced. In other words, with increasing grain refinement, the grain formation process is becoming more controlled by the grain refiner. Grain size Fig. 11 Schematic illustration of the qualitative variation in grain size along the fluidity channel. The morphological transition and change of grain size with grain refinement will change the apparent viscosity of the partially solidified suspension. Mori and Ototake 22 have suggested an equation for calculating the viscosity of the mixture: (4) where, (5) It can be expected that the grain size (d) and the solid surface area per unit volume (Sr) will decrease with increasing grain refinement since grain refinement often produces - 113 - a more globular, less-branched structure, and the coherency fraction solid (fscoh) will increase. Equation (4) then predicts that the viscosity of the less grain refined suspension will be larger than a more grain refined, at the same solid fraction. The initial reduction of fluidity with grain refinement is hereby not explained by a suspension viscosity consideration. The mobility of the dendrites may also affect the fluidity. The dendrites are in theory1,9 expected to be stationary relative to theflow tip, i.e., with negligible relative velocities between solid and liquid. However, this is not necessarily the case. The mobility of the dendrites may depend on their size and surface area. In the experiments performed here, the unrefined alloy had a larger grain size at the tip and a better fluidity than the other alloys. One possible explanation to this tendency is that the mobility of the larger, branched, dendrites is lower than of the small, grain refined, dendrites. If this is the case, the unrefined alloy would flow longer with only liquid and the smaller crystals at the tip (the larger crystals being left behind), while the grain refined alloys would have to carry more dendrites with the tip (since a larger number of crystals have a high mobility), giving an increase in viscosity. At higher levels of grain refinement, this effect is not as pronounced, since all grains below a certain size may be transported, and the postponed dendrite coherency comes into effect, increasing the fluidity. However, such a hypothesis is not confirmed by the fluidity tests of Pai and Jones 23 on stirred semi-solid SnPbl4.5 alloys. Increasing theholding time between the stirring and pouring resulted in formation of agglomerates, and a reduced fluidity. Additions of grain refiners produce more effective nucleation sites during solidification, which are activated at smaller undercoolings. The maximum growth temperature of a metal with good grain refinement can often be 3-4°C larger than of the unrefined alloy. 24 Fluidity has been reported to be proportional to superheat,2,25 and the experiments in this investigation were performed at a constant superheat relative to the equilibrium liquidus of the pure alloys. This implies that the superheat of the grain refined samples would be less than of the unrefined alloys, which might cause a reduced fluidity. Since the maximum growth temperature changes the most with the first grain refining additions, 24 the cause of the stabilization of fluidity observed at higher levels of grain refinement can be that the maximum growth temperature is relatively unchanged. The strength of a dendrite network formed at low solid fractions can also be of significance. After dendrite coherency, the strength of the network will, of course, increase as the solid fraction increases. This effectively means that the network that stops further flow has developed sufficient rigidity to withstand the effect of the - 114- pressure differential driving the flow. The fraction solid at which flow stops depends both on the degree of interconnection of the dendrite network and its strength. Little is known about the strength development during solidification, and it is also difficult to speculate and predict the effect of grain refinement on strength development, although it is known that measurable strength will develop later in the solidification process. 14 Important information about the solidification behavior of alloys can be gained by considering the equilibrium phase diagrams. Some characteristics of the two alloy systems are given in Table V . Table V. Phase diagram characteristics of the two alloy systems. Alloy k m Ce [K/wt%] [wt%] Al-Mg 0.47 -5.8 36 Al-Cu 0.17 -3.4 33 k is the equilibrium distribution coefficient, m is the slope of the liquidus line and Ce is the eutectic composition. The liquidus and solidus lines in the phase diagrams have been assumed to be straight, giving constant values for m and k. From these relationships some interesting solidification properties may be calculated, as shown in Table VI. Fraction of eutectic can be estimated by the Scheil equation: cj (6 ) , where C0 is the initial composition, Ce is the eutectic composition and k is the equilibrium distribution coefficient. From Table VI it is interesting to notice that the AlMg5 and AlCu2 alloys have very similar solidification ranges and eutectic fractions, but a very different constitutional effect. The size of the mushy zone, given by the solidification range, would increase in the order AlMg 1 -AlMg5-AlCu2-AlCu4. The constitutional undercooling, proportional .to mC(k-l), would decrease in the order AlMg5-AlCu4-AlCu2-AlMg 1, and dendrite growth rate would be expected to decrease in the same order, since it has been suggested 6 that V°caT2. Indeed the fluidity also increases in the order AlMg5- - 115 - AlCu4-AlCu2-AIMg 1, i.e., correlating well to the trend in growth rate. Alloys with a high dendrite growth rate have a small fluidity, and vice versa. These results confirm that alloys with a wider solidification range have lesser fluidity, within the alloy system. Table VI. Solidification characteristics of the alloys.AT 0 is the equilibrium interval of solidification, feSchei1 is the eutectic fraction calculated from the Schell equation, Tliq is the equilibrium liquidus temperature, and mC0(k-l) is a constitutional factor. Alloy jr Schei) Tn q mC0(k-l) [%] [°C] AlMgl 6.54 0.12 654.5 3.074 AlMgS 32.70 2.41 631.3 15.37 AlCu2 33.20 3.41 653.5 5.644 AlCu4 66.40 7.87 646.7 11.288 From the mathematical equations developed by Flemings et al.1-9, Eq. 1-3, the individual differences in the AIMg and AlCu alloys may be compared. Assuming that the heat flow is controlled by heat transfer at the tube-metal interface, Flemings 1 presented the following simplified equation: L , gp,K 0or»VVrO (7) ' 2HT-TJ If the properties of the alloys are calculated by assuming mechanical mixing and that the critical solid fraction is proportional to the fraction of primary phase, thedifference in fluidity within each alloy system can be calculated from equation (7). This gives: L^msl) - 1.00 and ■ 1.04 L/AIMgS) Lf(AlCu4) The experimental results gives: Lf(AlMgl) LXAlCu2) and ------= 1.10 Lf{AlMg5) Lf(AlCu4) The mathematical model predicts the observed difference in fluidity between the AlCu alloys fairly well, while there is a big difference between theoretical prediction and experimental results for the AIMg alloys. - 116 - Conclusions The effect of five levels of increasing grain refinement, by AlTi5B 1 master alloy additions, have been investigated in two AlCu and two AIMg binary alloys with a vacuum fluidity test. Fluidity shows a complex relationship to grain refinement. With increasing grain refinement, fluidity seems to go through a minimum before stabilizes. In the two alloys with the highest alloy content, AlMg5 and AlCu4, the minimum is observed at -0.03 wt% Ti, while it is postponed to -0.05 wt% Ti in the AlMgl alloy. In the AlCu2 alloy a gradual decrease of fluidity with grain refinement is observed. Fluidity at the maximum level of grain refinement, -0.12 wt% Ti, is less than that of the unrefined alloy for all alloys. There are several possible causes of such a complex relationship, but it is difficult to conclude which ones are most important and responsible for the observed effects. There is a weak correlation between the grain size at the flow tip and the fluidity, since the grain size is continuously decreased with increased grain refinement. Increasing the content of the main alloy element reduces fluidity, in accordance with observations that fluidity is inversely proportional to the solidification range. Acknowledgements This research was sponsored by Elkem Aluminium ANS, Hydro Aluminium AS and the Norwegian Research Council. Nomenclature A - Mould surface area cp - Specific heat of metal cP,m - Specific heat of mould Ce - Eutectic composition C0 - Initial composition d - Grain size fs - Fraction solid fscoh - Fraction solid at dendrite coherency fscnt - Critical solid fraction f/chetf „ Eutectic solid fraction, as predicted by the Scheil equation - 117 - h - Heat transfer coefficient between mould and metal k - Equilibrium distribution coefficient ly, - Thermal conductivity of mould K - Constant Lf - Fluidity length m - Slope of liquidus line R - Radius of fluidity test channel S - Circumference of mould channel Sr - Surface area per unit volume T - Temperature in liquid Tliq - Liquidus temperature T0 - Room temperature V0 - Flow velocity am - Thermal diffusivity of mould (km/cp mpm) AHf - Latent heat AT - Undercooling AT’ - Superheat AT0 - Equilibrium solidification range Ax - Choking length r|a - Apparent viscosity rji - Viscosity of liquid p, - Density of liquid pm - Density of mould References 1. M.C. Flemings, "Fluidity of metals - Techniques for producing ultra thin section castings", 30th International Foundry Congress, Praha, 1963, 61-81 2. F.R. Mollard, M.C. Flemings and E.F. Niyama, "Understanding Aluminum Fluidity: The Key to Advanced Cast Products", AFS Trans., 1987, 95, 647-652 3 C.R. Loper, Jr., "Fluidity of Aluminum-Silicon Casting Alloys", AFS Trans., 1992, 100, 533-538 4. J. Campbell, "Review of fluidity concepts in casting", Cast Metals, 1995, 4, 227-237 5. P. Bastien, J.C. Armbruster and P. Azou, "Flowability and viscosity", AFS Trans., 1962, 70, 400-409 6. J.D. Hunt, "Steady State Columnar and Equiaxed Growth of Dendrites and Eutectic", Mat. Sci. Engng., 1984, 65, 75-83 - 118 - 7. D.G. McCartney, "Grain refining of aluminium and its alloys using inoculants", Int. Mat. Rev., 1989, 5, 247-260 8. D. Apelian, G.K. Sigworth and K.R. Whaler, "Assessment of Grain Refinement and Modification of AlSi Foundry Alloys by Thermal Analysis ”, AFS Trans., 1984, 92, 297-307 9. M.C. Flemings, E. Niyama and H.F. Taylor, "Fluidity of aluminum alloys, An experimental and quantitative evaluation", AFS Trans., 1961, 69, 625-635 10. G. Lang, "Giesseigenschaften und Oberflachenspannung von Aluminium und binaren Aluminiumlegierungen", Aluminium, 1972, 48, 664-672 11. R. Trivedi and W. Kurz, "Dendritic growth", Int. Mat. Rev., 1994, 39(2), 49-74 12. M.C. Flemings, "Behavior of Metal Alloys in the Semisolid State, Met. Trans. A, 1991, 22A, 957-981 13. L. Amberg, G. Chai and L. Backerud, "Determination of dendritic coherency in solidifying melts by rheological measurements", Mat. Sci. Engng., 1993, A173, 101-103 14. A.K. Dahle and L. Amberg, "Investigation of the dendrite coherency point in solidifying Al-Si foundry alloys", The 4 ’th International Conference on Aluminum Alloys, Eds. T.H. Sanders, Jr. and E.A. Starke, Jr., The Georgia Institute of Technology, Atlanta (GA), 1994, 91-98 15. G. Chai, L. Backerud, T. Rplland and L. Amberg, "Dendrite Coherency during Equiaxed Solidification in Binary Aluminum Alloys", Met. Mat. Trans., 1995, 26A, 965-970 16. M. Tiryakiouglu, D.R. Askeland and C.W. Ramsay, "Fluidity of 319 and A356: An Experimental Design Approach", AFS Trans., 1994, 102 ,17-25 17. W.H.M. Alsem, P.C. van Wiggen and M. Vader, "The combined effect of grain refinement and modification in aluminium alloys", Light Metals 1992, (Ed. E.R. Cutshall, The Minerals, Metals and Materials Society, Warrendale (PA), 1991), 821-829 18. A.K. Dahle, P.A. Tpndel, C.J. Paradies and L. Amberg, "Effect of grain refinement on the fluidity of two commercial Al-Si foundry alloys", accepted for publication in Met. Mat. Trans. A 19. D.V. Ragone, C.M. Adams and H.F. Taylor, "Some factors affecting fluidity of metals", AFS Trans., 1956, 64, 640-652 20. D.V. Ragone, C.M. Adams and H.F. Taylor, "A new method for determining the effect of solidification range on fluidity", AFS Trans., 1956, 64, 653-657 21. S. Venkateswaran, R.M. Mallaya and M.R. Seshadri, "Effect of Trace Elements on the Fluidity of Eutectic Al-Si Alloy Using the Vacuum Suction Technique", AFS Trans., 1986, 94, 701-708 22. M. Hirai, K. Takebayashi, Y. Yoshikawa and R. Yamaguchi, "Apparent - 119 - viscosity of Al-10mass%Cu Semi-solid Alloys", ISIJ Int., 1993, 33, 405-412 23. B.C. Pai and H. Jones, "Casting fluidity of partially solid Sn-Pb alloy slurries made by stirring within the freezing range", Solidification Technology in the Foundry and Casthouse, (The Metals Society, London, 1983), 126-130 24. P.A. Tpndel, "Grain Refinement of Hypoeutectic Al-Si Foundry Alloys", Ph.D. thesis, The Norwegian Institute of Technology, 1994 25. A. Kolsgaard and S. Brusethaug, "Fluidity of aluminium alloy AlSi7Mg-SiC particulate composite melts", Mat. Sci. Tech., 1994, 10, 545-551 - 120 - Article #6 A. Ahmed, A.K. Dahle, D. Apelian and L. Amberg "Modeling the Feeding of Aluminum Alloy Castings" accepted for publication in Light Metals 1996, TMS, Warrendale (PA), 1996 Modeling the Feeding of Aluminium Alloy Castings A.Ahmed1, A.K. Dahle2, D. Apelian 1 and L. Amberg 2 1) Worcester Polytechnic Institute Aluminum Casting Research Laboratory 100 Institute Road, Worcester, Massachusetts 2) Norwegian Institute of Technology Department of Metallurgy N-7034 Trondheim, Norway ABSTRACT Aluminum alloy castings, in general solidify with the growth of equiaxed dendrites followed by eutectic precipitation. Dendrite growth and eutectic precipitation are influenced by a complex combination of macroscopic transport phenomena and microscopic solidification kinetics. The feeding process, i.e., the shrinkage driven material transport and solidification mechanism, is influenced by the state of the mushy zone - the local solid fraction and its distribution within the zone. Consequently, the mushy zone must be divided into several distinct regions in order to successfully model the solidification and feeding process. These regions have different mechanical and solidification characteristics and are governed by different transport mechanisms and solidification kinetics. A mathematical model for solidification and feeding is developed coupling the macroscopic transport process with microscopic solidification kinetics considering the various regions of the mushy zone. An outline of the numerical solution of the coupled system is presented. Upon implementation, this model will predict several important solidification variables such as dendrite coherency, grain size distribution and porosity. The phenomenology of feeding during solidification and the framework of the model are reviewed. Introduction Casting a part in aluminium is not a straightforward task, especially when there are restrictions on the tolerances of dimensions, properties and appearance of the product. A major reason for these difficulties in the casting operation is the difference in density between solid and liquid. The increase of density upon solidification causes the - 121 - material to contract, and for most aluminium based alloys the shrinkage is in therange of 3 to 8 %. Unless the correct precautions are undertaken to compensate for the volume contraction, the casting will exhibit undesirable defects such as porosity, surface sinks and dimensional inaccuracies. Porosity in castings is undesirable for several reasons: The cast parts may not be pressure-tight, they may be more susceptible to corrosion, and the mechanical properties, such as strength and fatigue resistance, may be impaired. In addition the castings do not have a reproducible quality, which leads to increased scrap rate and costs. Strict control of product and mould design, proper melt treatment and controlled solidification conditions are therefore crucial for attaining the optimum quality assurance. The mechanism where liquid moves to compensate for the shrinkage is termed ’feeding ’, and a controlled feeding process is necessary to achieve the highest quality. Normally one or more feeders are placed on the casting to supply metal to compensate for the volume deficit. Depending on the solidification conditions, i.e., alloy composition, grain refinement, cooling rate etc., porosity due to shrinkage may appear as macropores or more evenly distributed micropores. Another cause of porosity in aluminium castings is dissolved hydrogen, which is liberated during solidification due to its very small solubility in the solid. However, it is virtually impossible to separate shrinkage and gas induced porosity since they are working synergistically. Poorly designed castings, such as those where parts of the castings are prematurely frozen off from the feeders by solidification in intervening regions, will also exhibit porosity since the feeding path is not kept effective throughout the solidification range. Generally, the aim is to concentrate all shrinkage to the feeder, and by careful control of the solidification direction and -rate, by chilling, insulation or exothermic additions, a desirable opposite movement of solid and liquid may be achieved. (The solid, or solidification front, moving towards the feeder.) In this paper we discuss the feeding of aluminium alloys, with focus on the mechanisms and influencing factors of the feeding process and porosity formation in aluminium alloys. The predictability and limitations of the existing models and software are evaluated, and the tools and framework for a new mathematical model of feeding are presented. - 122 - Feeding Mechanisms During Solidification Aluminium alloys solidify over a temperature interval, and thus interdendritic liquid and dendritic solid coexist in the mushy zone. The size of the mushy zone is determined by the solidification range of the alloy (AT) and the temperature gradient (G) in the section. The volume change experienced during solidification creates a pressure differential inside the casting that enforces mass transport towards the solidifying regions. Five different mechanisms of mass transport may be envisioned to occur to supply extra material[l,2]. Figure 1 illustrates these possible transport processes. Above theliquidus temperature the melt is transported, and this is the so-called liquid feeding mechanism. Solid crystals start to form as the temperature drops below the nucleation temperature, and movement of the semi-solid slurry, where the dendrites are carried with the flow, is termed mass feeding. The dendrites impinge and form an interconnected solid network when the dendrite coherency point is reached[3]. Further mass transport therefore has to occur by transport of liquid through the skeleton, i.e., interdendritic feeding. Flowever, burst feeding may occur if the pressure for flow exceeds the strength of the dendrite interconnections. Since the permeability of the interdendritic network is continuously decreasing with increasing solid fraction, the stress on the dendrites increases and the network may collapse, giving rise to transport of solid fragments and liquid. Solid feeding denotes the feeding of solid material, by deformation, and may occur just below the solidus temperature where the yield strength of the material is relatively low. Both liquid and mass feeding are expected to occur without significant resistance, and the interdendritic mode of mass transport is expected to be the main limiting feeding mechanism. The significance of the burst feeding mechanism is virtually unknown since strength data of the solid network, as well as reliable permeability data, have not been obtained throughout the solidification interval. Work is currently in progress to obtain these properties, preliminary work indicates that burst feeding has significant impact on porosity formation[4]. In a short-freezing range alloy, columnar dendrites grow opposite to the heat flow direction, and the mushy zone is between the dendrite stems. In this case the mass feeding mechanism is less important, except for transporting pinched off dendrite fragments. . - 123 - Feeding Feeding Feeding Feeding Burst Feeding Figure 1: A schematic of the feeding mechanisms. The criterion for the formation of porosity is governed by the following pressure consideration^, 5]: Pg+P|5>Pa,m+Ph+P0 where Pg is the total gas pressure in the pore, Pp is the pressure due to shrinkage, Patm is the ambient pressure, Ph is the metallostatic pressure, and Pc is the capillary pressure (=2a/r) necessary to create pore interface. - 124 - Factors Affecting Feeding of Aluminium Alloys The following paragraph is not intended to be a complete coverage of all the possible variables that affect the feeding process, since an overwhelming amount of work has been performed investigating feeding and porosity formation. In general, there are, in addition to the physiochemical properties, two separate, but related, factors that must be understood and controlled to avoid porosity: 1) The size and properties of the mushy zone, and 2) the formation of gas induced porosity In practise these are determined by the interplay between several factors: alloy composition, cooling conditions, grain refinement, eutectic modification (Al-Si alloys), hydrogen content, inclusions in the melt and external pressure. Feeding, and the resulting porosity, is usually evaluated by casting plates or bars, or in more controlled directional solidification experiments. The extent and properties of the mushy zone is affected by theevolving microstructure and dictates the resistance towards feed metal transport by determining the extent of each feeding mechanism. The effects of these variables are briefly reviewed. Alloy Composition: A change of alloy composition changes the solidification range (AT) of the alloy, and increasing the solidification range increases the size of the mushy zone. Other changes caused by changing the alloying is the possible formation of intermetallic phases. One example is iron in Al-Si alloys that may precipitate large, long AlFeSi(Mg,Mn)-phases that may block the feeding path. At low cooling rates these (3-phase particles are large and have a coarse needle-like morphology, but become smaller and more dispersed at higher cooling rates[6]. Sigworth[7] hasreported that good modification suppresses the formation of large intermetallic Fe-phases. Alloying may also affect the hydrogen solubility in the liquid[8], and changing the constitutional conditions may change the grain size and the dendrite growth rate. It must be noted that the physiochemical properties also depend on composition. Temperature Gradient. G: An increase of the thermal gradient is related to a decrease of the size of the mushy zone, i.e., a reduced length in the casting between the liquidus and solidus temperature. An increase of the temperature gradient has been observed to reduce the amount of - 125 - porosity [9-15]. However, steep thermal gradients are difficult to establish in aluminium alloys due to the high thermal conductivity. Cooling Rate. R: In general, an increase of cooling rate refines the dendritic structure, with reductions both of grain size and dendrite arm spacing. This results in a reduced size, but increasing number, of interdendritic channels. The size of the mushy zone is also reduced, and these effects cause a decrease in porosity content[ll-12, 15-17], On the other hand, Lee et al.[15] observed two different regions as the solidification time was increased. Porosity first increased with increasing solidification time, possibly due to an increase in the size of the mushy zone. However, at increasing solidification times, porosity decreased. Due to a longer time available for interdendritic flow through a more permeable dendrite network as a result of an increased dendrite arm spacing. Grain Refinement: Grain refining, usually obtained by additions of AlTiB-master alloys, is an effective way to postpone dendrite coherency. In this way a larger part of the feeding process is carried out by the more open feeding mechanism, mass feeding. Fang and Granger! 16] showed that grain refinement reduced both volume fraction porosity and pore size in A356, also giving a more uniform distribution of porosity. This was confirmed by LaOrchan and Gruzleski[18], Shivkumar et al.[6] observed a similar tendency in 319, whereas Irani and Kondic[19] observed increased porosity with grain refinement in an alloy similar to 319. Eutectic Modification: Na, Sr or Sb may be added to Al-Si alloys to transform the silicon structure from acicular to fibrous. The modifying addition is related to a suppression of the eutectic temperature, depending on the degree of modification. Theresult is that the size of the mushy zone increases and, more specifically, the length of the interdendritic feeding path increases. Eutectic modification has been shown to make the alloys more prone to porosity, and many explanations have been suggested[6, 16-18,20-22], Emadi et al.[22] recently showed that additions of 0.01 wt% Sr or 0.005 wt% Na to A356 reduce the surface tension of the liquid by 19 and 10 %, respectively, and may increase the volume shrinkage about 12%. Both of these effects may promote porosity formation early on in the solidification process, and longer time for pore growth, since the solidification time is increased. Sigworth et al.[20] suggested that Sr modification increases the density of the solid in A356. The combination of grain refinement and eutectic modification usually results in - 126 - properties between those of the additions made separately, i.e., less porosity than in the modified alloy, but more than in the grain refined[6,18]. Hydrogen: Hydrogen concentration in the remaining liquid increases during solidification due to its very low solubility in the solid. However, a nucleation barrier has to be exceeded for a pore to form. Campbell[2] calculated that a pressure of 30800 bar is necessary for a pore to nucleate homogeneously in the melt. Chen and Engler[23] have shown that very low supersaturation pressures are necessary to form stable pores, indicating that there usually exists effective heterogeneous substrates for porosity nucleation, such as gas filled holes or gaps at interfaces and inclusions. The content of non-wetted inclusions, such as oxides, in the melt plays a crucial role when the alloy contains dissolved hydrogen. When shrinkage pores form, the nucleation barrier is overcome, and the rapidly diffusing hydrogen contributes to the growth of the pores. Of course, hydrogen diffusion becomes more important with increasing solidification time. It has also been suggested that a threshold, or critical, hydrogen concentration exists, depending on the cooling rate, below which negligible porosity forms[6,ll,16]. Shivkumar et al.[6] approximated this concentration to 0.05 to 0.1 cmVlOOg in untreated A356.2 alloys. Generally, porosity is observed to increase with an increase of hydrogen content[6, 11,16-18]. The evolution of gas induced porosity has an adverse effect on the feeding process, since the pressure differential driving the material transport is reduced. In addition, gas pores may act as physical barriers to fluid flow through the interdendritic network and transport of gas pores does not represent effective feeding. External Pressurization: Few investigations have been reported on the effect of external pressurization[19,24]. It would be expected that an extra pressure on the feeder would increase the driving pressure for flow and thereby reduce porosity, and this is supported by the reported experiments. The effect of external pressurization needs further clarification. - 127 - Mathematical Treatment and Predictability of Feeding The development of powerful computers and computer simulation methodologies are facilitating the prediction of feeding and porosity formation. Traditional approaches to predict the feeding behavior and resulting porosity levels in casting alloys are Modulus Method, Criteria Functions, Macro/micro Mathematical Models, and Empirical Models. The effectiveness and limitations of each of these methods are discussed in the following subsections. Modulus Method The modulus method uses Chvorinov ’s rule to determine the size of the feeder/riser in order to obtain efficient feeding of shrinkage cavities. It has been demonstrated experimentally [25] that it is not possible to predict feeder/riser size for optimum feeding based on shape alone. Chvorinov ’s rule has been used, however, to determine an approximation of the feeder/riser size because it is easy to use and computationally it is inexpensive. Criteria Functions The criteria functions refer to a combination of the relevant solidification variables such as temperature gradient, cooling rate, solidification rate etc. Some of these numbers are derived using Darcy’s law to calculate pressure drop in the mushy region. These functions appear to give a qualitative indication of feeding efficiency and porosity. Suri et al.[26] and Huang et al.[27] studied theapplicability of these functions to predict feeding and porosity formation. Irani and Kondic [19] introduced a function called the Feeding Efficiency Factor (FEF) using experimental results from a plate casting in a tapered mold. This function is defined as FEF=(tb-ta)/ta, where ta and ty are the solidification times at two locations a and b, with b being closest to the riser. When measured porosity data were plotted against FEF, the scatter in the data was significantly reduced and it was observed that porosity decreased as FEF increased. Irani and Kondic concluded that the FEF could be used as an indication of feeding efficiency and porosity. Davies[28] developed an equation involving the solidus velocity to calculate the capillary feeding distance, but this methodology ignored the structural effect of the dendritic network. Niyama et al.[29] and Minakawa et al.[30] introduced the ratio of thermal gradient to the square root of cooling rate ( G/ffi ), popularly known as the Niyama function. This parameter predicts that shrinkage decreases with increasing solidification time. - 128 - Lecomte-Beckers [31] proposed a microporosity index taking into consideration the alloying elements of a Nickel-based superalloy. Lee et al.[15] developed a feeding efficiency parameter integrating all individual thermal variables, denoted as (Gt2/3)IVS , where G is the thermal gradient, t is the local solidification time and Vs is the velocity of the solidus isotherm. The criteria functions discussed above unfortunately ignore the effects of the microstructure of the casting. Feeding Resistance Number (FRN), introduced by Sufi et al. [32,33], is a nondimensional number which incorporates the effects of microstructure such as grain size and shape and provides a qualitative indication of porosity. Correlations need to be developed for quantitative prediction. Macro/Micro Mathematical Models The criteria functions described above, with the exception of the Feeding Resistance Number (FRN) are all based on macroscopic heat transfer parameters. These criteria functions do not include the effects of microscopic functions such as grain size, secondary dendrite arm spacing etc. on the formation of porosity. This has been addressed by several investigators. Kubo and Pehlke[12] developed a mathematical model for the formation of shrinkage and gas porosity in steel and Al-Cu alloys considering the interdendritic fluid flow and evolution of dissolved gases. Interdendritic fluid flow was modeled using Darcy’s law. The numerical results predicted by the model agreed with experimentally measured porosity in Al-4.5Cu plate castings. However, the model ignored the effects of liquid and mass feeding on interdendritic feeding and assumed an arbitrary fraction solid at which point interdendritic feeding begins. Poirier et al.[ll] followed the same approach as Kubo and Pehlke to predict porosity formation in Al-Cu alloys. They used a different approach for calculating the permeability of the interdendritic network and estimation of the radii of the gas bubbles that form in the interdendritic space. The major variables studied were the effects of thermal gradient and solidification rate and of the concentration of dissolved hydrogen and copper concentration. Zou et al.[34] developed a mathematical model for microstructure evolution and microporosity formation in Al-Si alloys. Although the thermal gradient in the casting was ignored, thepredicted results agreed with available experimental results. Suri and Paul[35] developed a mathematical model for porosity prediction combining theeffects of shrinkage and dissolved gases. The volume fraction of porosity is determ ined by coupling the local fluid pressure with the conservation of - 129 - dissolved gases in the melt as well as the gas-liquid surface tension. The porosity results predicted by this model agrees with the experimental observation. Empirical Models Due to limitations of the criteria functions to adequately predict feeding and porosity, empirical models have been and are being developed applying the statistical data reduction technique on a large volume of experimental results [36,37], These models take all factors that affect feeding and porosity (thermal parameters, liquid metal quality, inclusion level and gas content) into account and predict porosity and pore size distribution. The advantage of these empirical models is that they use a large volume of experimental results and the predictions fall within the limits of the experimental results. The development of these relationships is extremely expensive and time consuming. In addition, these relationships are not general and do not represent the physics of heat transfer and phase transformation, and as such empirical models are not applicable to a variety of systems. Existing Commercial Software There are several commercial software packages available and which are being used by foundries as a design tool[38]. These software packages solve the governing equations for fluid flow, heat transfer and phase transformation and predict the thermal fields in the casting and mold. In order to predict feeding and porosity formation, these packages use one or more of the methods discussed earlier. Although these software are quite efficient and accurate in predicting thethermal fields in thecasting and mold, their capability to predict feeding is subject to the same limitations as the methods themselves. Despite the development of the criteria functions, macro/micro models and empirical models for porosity prediction, a comprehensive model should consider all the physical phenomena (macroscopic transport processes and microscopic solidification kinetics) and feeding mechanisms during solidification. Proposed Mathematical Framework to Predict Feeding Regions of the Mushy Zone For modeling purposes, a domain of solidifying alloy can be divided into three zones, viz. Liquid, Mushy and Solid. In the present development, we further subdivide the mushy zone into several distinct subregions depending on the solidification and feeding mechanism. For equiaxed dendritic solidification, these are: - 130- Mass Feeding: In this subregion, grains of primary solid are freely suspended in the pool of liquid alloy. Solidification progresses via dendritic growth. Feeding occurs by the flow of the bulk mixture. Interdendritic Feeding I: The free-floating dendrites grow and move closer to each other. At some point, these dendrites touch each other to form a coherent dendritic network. Solidification changes from tip growth to coarsening and growth of secondary arms. Feeding takes place by flow of the liquid through the interdendritic network. Interdendritic Feeding II: When the precipitation of the primary phase is complete, growth of the dendrites ceases and precipitation of the eutectic phase begins. This subregion is characterized by a constant temperature of the liquid. Feeding continues to take place by flow through the network of liquid channels. Critical Feeding Zone: At some point during eutectic precipitation, liquid channels become narrow and flow resistance due to capillary effects overcome the pressure gradient caused by solidification shrinkage. Feeding ceases to take place, and it is in this subregion where shrinkage pores begin to nucleate and grow. The different regions and subregions discussed above have distinct mechanical and thermal characteristics. The physical phenomena taking place in these regions are different and are governed by specific mathematical equations. However, it is possible to formulate a general form of the governing equations and simplify them according to the characteristics of various regions of the mushy zone. Hydrogen Evolution As solidification progresses, hydrogen gas is released at the solid-liquid interface due to the large difference in solubility of the gas in liquid and solid phases. The hydrogen rejected at the interface diffuses into the liquid and increases its gas concentration. Since hydrogen diffuses relatively fast in the liquid, previous investigators]! 1,12,34] have ignored the concentration gradient in the liquid and have used simple algebraic equations such as the Brody-Flemings equation to calculate the hydrogen concentration in the liquid and solid phases. In reality, concentration gradients in the liquid may not be negligible. There are two competing factors that determine the concentration of hydrogen gas in the liquid; namely, diffusion through the liquid and the level of convection in the liquid. Gas released at the interface diffuses into the liquid at a certain rate determined by the diffusivity of hydrogen in the liquid alloy. Due to the feeding mechanism, diffused hydrogen gas flows back with liquid metal towards the solid-liquid interface. The concentration of hydrogen gas in this boundary layer may - 131 - be much higher than the average concentration in the liquid. Therefore, while it is appropriate to use the algebraic equations to get an approximation of the gas effects at a lower computational cost, a complete model for predicting feeding should be able to take these phenomena into consideration. In this work we express the hydrogen balance using an advection-diffusion equation similar to the one used for species balance in order to take into account the variation of gas concentration in the liquid. Macroscopic Transport Equations The macroscopic transport processes are the transport of mass, momentum and thermal energy. For a two-dimensional system, the general form of the equations governing these processes are presented in this section. These equations are similar to the ones used by Reddy and Beckermann[39] to simulate the DC casting of Al-Cu ingot and by Wang and Beckermann[40] to simulate microstructure development. The effects of gravity have been neglected in order to reduce the complexity of the model. This assumption does not affect the generality of the model. The general form of the equations governing the macroscopic transport processes are: (1) Continuity: -|(P,/, + P,/,)+'£(P,/A + Pz/A) + -|(P,/,v s + P,/,v z) = 0 (2) Momentum: ^(p,JX+p;.W + ",-g;(p,.W + + «,^(p,.W + v,-§(p,W = ^(p,.^,+p,.W + ^(p,.W + v,-^(p,.W + - 132- (3) Energy: (p,/, + Pi + ( Ps fsus+ Pi Aui) "^ + (P, fsvs + P| fivi) dH 6y d &T dT dx dx dy ay dx. In the above equations, the components of the stress tensor are given by x(j = p dXj where p is the viscosity and enthalpy (H) is given by: # = (/X+^T+(i.o-#Aar(c,-c;r^ Microscopic Equations Transport of chemical species, nucleation, growth and transport of grains, eutectic precipitation and evolution of hydrogen gas are the microscopic phenomena taking place during solidification. The equations governing the microscopic processes are: (4) Species Balance: ^(P,.tC, + p,/C,) + K, + ^(P^CJ + ac. dC, Ur^(PlflCl) + vi^(PifiC[) - DsPsfs vDiPi fi- ay dx. ' ' ' a% a% ac. dCt DsPs +DiPifr dy dy (5) Grain Transport: dn dn dn — + u — + v. — a 'ay - 133 - (6) Hydrogen Evolution The above equations are simplified according to the physical phenomena taking place in the different regions of the domain. The fully coupled system is solved using the finite element method. Summary and Future Steps The basic framework of a mathematical model for predicting feeding in aluminium alloy castings has been presented. The model couples macroscopic transport phenomena (heat, mass and momentum transport) with microscopic solidification kinetics (grain nucleation, grain growth, grain transport, chemical species balance, hydrogen gas evolution). Future steps are: develop and implement a numerical solution methodology to solve this coupled system, design and develop appropriate experimental verification procedures for the numerical solutions, generalize the model to simulate more general three-dimensional problems, and test the validity of the model on the shop floor. - 134 - Nomenclature C Concentration of chemical species CH Hydrogen concentration c Specific heat D Diffusivity of chemical species f Phase volume fraction G Temperature gradient H Enthalpy K Diffusivity of hydrogen gas k Thermal conductivity k(, Equilibrium partition coefficient n Grain density n 0 Initial grain density R Cooling rate T Temperature Tn Nucleation temperature aT Solidification range t Time u Velocity in the x-direction v Velocity in the y-direction x, y Coordinate system &Hf Latent heat p Viscosity ty Components of the stress tensor p Density Subscripts s Solid phase 1 Liquid phase eu Eutectic phase - 135 - References 1. J. Campbell, AFS Cast Met. Res. J., March, 1969, pp. 1-8 2. J. Campbell: "Castings", Butterworth-Heinemann Ltd., Oxford, 1991 3. A.K. Dahle and L. Amberg, The 4 ’th Int. Conf. on Aluminum Alloys, Eds. T.H. Sanders Jr. and E.A. Starke Jr., Atlanta(GA), 1994, pp. 91-98 4. L. Amberg, A.K. Dahle, C.J. Paradies and F. Syvertsen, AFS Trans., 1995, Paper No. 95-115 5. J. Campbell, British Foundryman, April, 1969, pp. 147-158 6. S. Shivkumar, L. Wang and D. Apelian, J. Met., January, 1991, pp. 26-32 7. G.K. Sigworth, Modem Casting, vol. 77 (7), 1987, pp. 23-25 8. G.K. Sigworth and T.A. Engh, Met. Trans. B, vol. 13B, 1982, pp. 447-460 9. K. Kubo and R.D. Pehlke, AFS Trans., vol. 94, 1986, pp. 753-756 10. E.N. Pan, C.S. Lin, C.R. Loper Jr., AFS Trans., vol. 98, 1990, pp. 735-746 11. D.R. Poirier, K. Yeum and A.L. Maples, Met. Trans. A, vol. 18 A, 1987, pp. 1979-1987 12. K. Kubo and R.D. Pehlke, Met. Trans. B, vol. 16B, 1985, pp. 359-366 13. E. Chang and Y.S. Kuo, AFS Trans., vol. 102, 1994, pp. 167-172 14. S.-T. Kao and E. Chang, Cast Metals, vol. 7 (4), 1995, pp. 219-225 15. Y.W. Lee, E. Chang and C.F. Chieu, Met. Trans. B, vol. 21B, 1990, pp. 715-722 16. Q.T. Fang and D A. Granger, AFS Trans., vol. 97, 1989, pp. 989-1000 17. D. Emadi and J.E. Gruzleski, AFS Trans., vol. 102, 1994, pp. 307-312 18. W. LaOrchan and J.E. Gruzleski, AFS Trans., vol. 100, 1992, pp. 415-422 19. D.R. Irani and V. Kondic, AFS Trans., vol. 77, 1969, pp. 208-211 20. G.K. Sigworth, C. Wang, H. Huang and J.T. Berry, AFS Trans., vol. 102, 1994, pp. 245-261 21. D. Argo and J.E. Gruzleski, AFS Trans., vol. 96, 1988, pp. 65-74 22. D. Emadi, J.E. Gruzleski and J.M. Toguri, Met. Trans. B, vol. 24B, 1993, pp. 1055-1063 23. X.-G. Chen and S. Engler, AFS Trans., vol. 102, 1994, pp. 673-682 24. J.T. Berry and T. Watmough, AFS Trans., vol. 69, 1961, pp. 11-22 25. P.N. Hansen and P R. Sahm, Modeling of Casting and Welding Processes TV, A.F. Giamei and G.J. Abbaschian (eds.), The Minerals, Metals and Materials Society, 1988, pp. 33-42 26. V.K. Suri et al., AFS Trans., vol. 100, 1992, pp.399-408 27. H. Huang and J.T. Berry, AFS Trans., vol. 101, 1993, pp. 669-675 28. V. de L. Davies, AFS Cast Metals Research Journal, vol. 11, 1975, pp. 33-44 - 136- 29. E. Niyama et al., AFS Int. Cast Metals Inst. J., vol. 7, 1982, pp. 52-63 30. S. Minakawa, I,V. Samarasekera and F. Weinberg, Met. Trans. B, vol. 16B, 1985, pp. 823-829 31. J. Lecomte-Beckers, Met. Trans. A, vol. 19A, 1988, pp. 2341-2348 32. V.K. Suri et al., Light Metals 1994, The Minerals, Metals and Materials Society, 1994, pp. 907-912 33. V.K. Suri, Ph.D. thesis, University of Alabama, Tuscaloosa, 1993 34. J. Zou, S. Shivkumar and D. Apelian, AFS Trans., vol. 98, 1990, pp. 871-878 35. V.K. Suri and A.J. Paul, Proceedings of the 1992 ASME Winter Annual Meeting, HTD.Vol. 218/AMD- Vol. 139, Micro/Macro Scale Phenomena in Solidification, ASME, 1992, pp. 101-110 36. K. Tynelius, J.E. Major and D. Apelian, AFS Trans., vol. 101, 1993, pp. 401- 413 37. K.E. Tynelius, Ph. D. thesis, Drexel University, 1992 38. A. Ahmed and D. Apelian, 3’rd Int. Conf. on Aluminum Processing, 1994, pp. 207-224 39. A.V. Reddy and C. Beckermann, Materials Processing in the Computer Age II, V.R. Voller, S.P. March and N. El-Kaddah (eds.), The Minerals, Metals and Materials Society, 1995, pp. 89-102 40. C.Y. Wang and C. Beckermann, Materials Processing in the Computer Age II, V.R. Voller, S.P. March and N. El-Kaddah (eds.), The Minerals, Metals and Materials Society, 1995, pp. 129-143 - 137 - - 138 - Article #7 L. Amberg, A.K. Dahle, C.J. Paradies and F. Syvertsen "Feeding Mechanisms in Aluminum Foundry Alloys" AFS Transactions, vol. 103, 1995, Paper 95-115 Feeding Mechanisms in Aluminium Foundry Alloys L.Amberg 1, A.K. Dahle1, C.J. Paradies 1 and F. Syvertsen 2 1 Norwegian Institute of Technology Department of Metallurgy Trondheim, Norway 2 SINTEF Materials Technology Trondheim, Norway ABSTRACT Solidification shrinkage during casting of aluminum foundry alloys requires feeding of melt through the mushy zone in order to avoid porosity and surface defects. The mechanisms responsible for feeding the volume contraction include liquid feeding, mass feeding, interdendritic feeding, burst feeding and solid feeding. Three separate experiments have been conducted with the goal of developing a complete understanding of the feedability of aluminum foundry alloys. Results of experiments to measure the coherency point, thepermeability and the porosity are presented in the context of developing the tools necessary to fully explore feeding mechanisms in long freezing range alloys. Coherency measurement is a mature technique permitting laboratory measurement of the point corresponding to the initial establishment of an interconnected network of dendritic grains. Results show that this typically occurs at fractions of solid between 15-30%. A method of measuring the permeability of the mushy zone during solidification promises to provide the permeability for a variety of aluminum alloys over a wide range of solid fractions. Recent observations of porosity in plate castings provide a foundation for discussing the burst feeding mechanism. The results indicate that burst feeding occurs discontinuously, when the strength of the developing network can no longer resist the stress induced by the flow of melt through the increasingly impermeable mushy zone. Burst feeding appears both to feed the solidification shrinkage and to establish an impermeable boundary to further interdendritic feeding. Experiments are being contemplated to measure the strength of the developing dendritic network permitting burst feeding to be predicted, based on the stress produced by resistance to interdendritic feeding of the solidification shrinkage. This could be compared to observed porosity in plate castings. - 139 - Introduction The solidification of most commercial alloys involves significant volume contraction. In long freezing range alloys, melt must be fed through a partially solidified, coherent dendritic network, if porosity and surface defects are to be avoided. Insufficient feeding is one of the most serious causes of casting defects in aluminum alloy castings. Commercially pure metals, influenced by a temperature gradient ahead of the interface, solidify as a planar front or a short columnar zone, and melt can easily feed the solidification shrinkage. However, as alloying elements are added to the melt, the freezing range increases and dendritic solidification begins to occur throughout a large portion of the casting. A description of the feeding mechanisms for long freezing range alloys is provided by Campbell1. He outlines five distinct feeding mechanisms occurring throughout the casting: - liquid feeding - mass feeding - interdendritic feeding - burst feeding - solid feeding Figure 1 illustrates the sequence of the feeding mechanisms as the solid fraction increases. Campbell describes liquid feeding as melt flow through large channels within the casting without the hindrance of any appreciable fraction solid. Mass feeding begins after solid crystals are nucleated within the melt. The solid crystals flow with the melt as they continue to grow, usually dendritically. This two-phase flow can be treated as a slurry and may be responsible for feeding solidification shrinkage until the solid develops into a rigid, interconnected network. For example, an alloy of the type A356.1 at a cooling rate of 0.6°C/s develops a coherent network as early as 23% fraction solid 2. At this point the dendrite tips of the individual crystals begin to form connections between themselves and their surrounding neighbors, and a coherent network of dendrites forms. After coherency the structural integrity of the dendritic network prohibits further mass feeding. Instead the melt must flow through the solid network of dendrites to feed solidification shrinkage. At low solid fractions, the developing network does not provide much resistance to melt flow. The permeability of the network is high. As fraction solid continually increases, the strength of the network increases, but the permeability decreases, increasing the stress on the network. Burst feeding occurs when the strength of the network is insufficient to resist the increasing pressure induced by continued solidification shrinkage. Campbell claims that this point occurs at a fraction solid of about 68% solid 3, but alloy composition and - 140 - liquid ------=> solid feeding ------5* mass feeding ------=> interdendrltic feeding ------5*. burst feeding ------> Figure 1. Illustration of the five feeding mechanisms that may occur when shrinkage occurs during solidification of a long freezing range alloy. casting conditions probably play a significant role in determining this point. Both a knowledge of the permeability of the mushy zone and the yield strength of the dendritic network must be known as a function of the fraction solidified to estimate a likely point for burst feeding to be activated. The final feeding mechanism, solid feeding, results from plastic deformation or creep of the completely solidified casting. This can be observed as sinks on the surface of the casting and results in poor surface quality or incorrect dimensions of the finished casting. It is usually desirable to avoid solid feeding by insuring that one of the other feeding mechanisms continues to feed the solidification shrinkage to very high solid fractions. This paper will present some combined experimental results investigating the parameters important to interdendritic feeding and burst feeding. These are the last of two feeding mechanisms occurring before either porosity or surface sinks develop. Therefore, a more detailed understanding of the parameters that affect the initiation and continuation of feeding by these mechanisms for commercial aluminum alloys will be a significant contribution to alloy design and selection. Results from experiments measuring the point of dendritic coherency, the interdendritic permeability of a mushy zone and porosity within a plate casting will be presented and discussed. The goal of - 141 - the research is to develop the necessary data for prediction of feedability for particular aluminum alloys, allowing prediction of porosity and surface quality using computer modeling and simulation and direct comparison between alloys with varying compositions and levels of grain refiner additions. Recent results indicate that most of thecomponents needed for predicting feeding are now available: reliable measurement of the coherency point, a method of measuring permeability during solidification and an empirical measurement of feedability. Work continues to develop a method to measure the strength of the mushy zone. Experimental Each of the three experiments was conducted independently; therefore, each will be described separately. A detailed description and validation of the coherency measurement technique used in this study has been presented elsewhere4; therefore only a brief description of the experimental method is provided here. The software used to evaluate the fraction solid from thermal analysis of the cooling curves was developed by Tamminen 5 and is used for determining the fraction solid for both the coherency and the permeability measurements. Figure 2. Illustration of the experimental apparatus used to measure the coherency point rheologically. Coherency Measurement An apparatus has been constructed that permits the direct rheological measurement of the coherency point, the point corresponding to the initial development of an - 142 - interdendritic network. Figure 2 illustrates the experimental apparatus used for rheological measurements. The boron-nitride coated steel paddle located in the center of the graphite crucible and the K-type thermocouple located next to the wall are both immersed 10 mm into an alloy that completely fills the crucible. The paddle is rotated very slowly, 0.05 rpm, and the torque is measured using a rheometer to measure the coherency point. Thecoherency point is defined as the point when the torque rises very rapidly. Three different cooling rates are reported for the alloy described as A1 in Table 1, grain refined by adding A15%Ti-l%B master alloy rod. Thermal analysis was used to provide an alternative coherency point and to determine the fraction solid during solidification. The thermal analysis was conducted under similar conditions as the rheological measurement, except the paddle was replaced by a second thermocouple centered in the crucible. The point corresponding to the maximum difference in temperature between the wall and the center yields the initial point of thermal contact between the dendrites and can be used as an alternative coherency point. The fraction solid calculated from the thermal analysis was used with the temperature and torque measured during therheological experiment to specify the fraction of solid at the point of coherency for the rheological measurement. Table 1. Aluminum Alloy Compositions [wt%]. Exp. Alloy Si Mg Fe Ti Sr B A1 AlSiTMg 6.800 0.174 0.140 0.130 0.001 0.0004 A2 AlMg5 0.034 4.960 0.016 0.008 - - A3 AlSiTMg 7.030 0.130 0.150 0.110 0.001 0.0004 Heat resistant, insulating felt insulated the top and bottom portions of the crucible to reduce heat loss in the axial direction of the cylindrical crucible. The resulting temperature distribution approximates the two-dimensional symmetry of a cylinder of infinite length, reducing any error that might be introduced by heat flow in the axial direction. Reported grain sizes were determined from the thermal analysis specimens after metallographic preparation and measurement using the linear intercept method. Permeability Measurement A graphite crucible hasbeen designed that permits a pressure to be applied to a column of molten metal as it is solidifying, Figure 3. The two halves of the graphite crucible allow for removal of the aluminum from the channel for metallographic examination, following the experiment. The pressure is applied by opening a valve to a fixed nitrogen pressure. The pressure is monitored by a pressure transducer, ±1.0 mbar. A small graphite float in contact with the free surface of the column is attached to a - 143 - displacement transducer, ±0.01 mm, to measure the volume flow rate. The probe of the transducer acts as a counterweight to the float. Fraction solid is calculated from temperature measurements within the mushy zone, using K type thermocouples accurate to ±1.0 K. The pressure, flow rate and solid fraction are used to calculate permeability. The aluminum for the result published here was alloyed in the laboratory by adding 5 wt% magnesium to refined aluminum, 99.997 %. Spark emission testing was used to evaluate the alloy described as A2 in Table 1. The alloy was melted, cast into the prepared graphite crucible, and the ceramic tube that delivers the nitrogen gas was inserted. The assembly was cooled, the float was inserted and the crucible was sealed to prevent leakage of nitrogen pressure during the test. The mould was inserted into a circular, resistance furnace and heated until the melt temperature reached 973 K. The alloy was then cooled in the furnace at a cooling rate of 0.24 K/minute. At a fraction solid of 60 %, a pressure of about 150 mbar above ambient was applied. The temperature, pressure and position data was recorded and will be used in later modeling of solidification processes. Pressure Position Sensor Graphite I Crucible Upper Float Half 71 Lower J Half Thermocouples Figure 3. Illustration of the experimental apparatus used to measure the permeability of the mushy zone of an aluminum alloy during solidification. Porosity Measurement The porosity of a plate cast in a permanent die was examined using X-rays and density measurements. The geometry of the die is shown in Figure 4. The die was coated with a thin protective coating, and the thickness of the coating was adjusted by sand grinding. Compressed air was used to remove any residual, loose particles. The - 144 - temperature of the die was controlled by both the cooling channels visible in Figure 4 and by electrical resistance heating elements. The oil in the cooling system was maintained at 350 °C during the duration of the experiment. The resistance heaters were used to bring the initial temperature of the die to about 400 °C. The distance from the surface of the cavity to the cooling channels increased towards the feeder at an angle of 3 degrees. This caused a temperature gradient in the 195 mm long plates that insured that the end of the plate would solidify first during casting. A commercial AlSi7Mg alloy, A3 in Table 1, was prepared by melting in a crucible contained in a resistance furnace. No additional grain refiner was added, and no degassing was attempted. The hydrogen content of the melt, 0.24 ml H2/100 g, was measured with an AlScan probe prior to casting. The casting temperature was measured with type K thermocouples in the pouring basin during filling and with a hand-held temperature logger. This hand held device gave a maximum temperature for the pouring temperature that is used to compare between individual experiments. The die itself provided 8 slots for inserting type K thermocouples into the parting plane of the die. Four additional holes were provided in the die for thermocouples that could be inserted to a point 2 mm from the cavity surface. This gave the researcher a complete picture of the die temperature distribution during the experiment. Figure 4. Illustration of the die used to cast the plates in the porosity investigation. After the die and the melt were prepared, the die coolant would begin circulation through the die and the temperature of the die would be brought to 400 °C using the resistance heaters. The melt temperature was reduced to 720 °C, and the melt was poured into the pouring basin of the die. Temperature was logged during the solidification of the plate and the casting temperature, die temperature and - 145 - solidification time were recorded. Plate porosity was first examined nondestructive^ by X-ray. The density of the plates was determined using Archimedes’s principle. The surface tension of the distilled water was measured using a pyknometer to 0.997714 g/cm 3 at 20 °C. A reference casting was made by pouring the alloy into an insulated sleeve cooled by a steel chill at the bottom. A section of the casting located 5 mm above the chill was removed as a density reference. This part of the casting was expected to be fully dense, because feeding and temperature conditions were optimal. The theoretical density based upon spectrographic analysis of the alloy was also calculated as an additional reference. Finally, a value of the density for AlSi7Mg was taken from the literature6 for comparison. The density variation along the plates was measured by cutting parallelepipeds from the plate as shown in Figure 5. The parallelepiped dimensions were approximately 15x15x8 mm. They were cut, marked and ground slightly to remove any burr remaining from the cutting. Figure 5. Location of the parallel-epipeds cut from the plate that were used for measuring the density variation along the plate axis. Results Results are reported for each of the experiments. The results of the permeability measurement is from a preliminary test of the apparatus. Additional experiments for coherency, permeability and porosity measurements are planned for a variety of alloys. Dendrite Coherency Figure 6 shows the results of coherency measurements on alloy Al, an AlSi7Mg alloy. Adding additional grain refiner increases the grain density, reduces dendrite growth rate - 146 - b) —0— f s coh- rheology —□ - Grain size [uni] -0— fs coh-rheology _ azis A - fs coh-thermal _ uibj rn - Grain size 500. £ 0 ■S 55 0.00 0.05 0.10 0.15 0.20 0.0 0.5 1.0 1.5 2.0 2.5 3.0 % Titanium added Cooling rate [K/s] Figure 6. Graphical resultsof the coherency measurements as a function of (a) grain refinement additions, and (b) cooling rate with 0.05% Ti grain refiner added. and increases the solid fraction at coherency. Over grain refining tends to produce spherodized grains that may tend to decrease the fraction solid at the rheological coherency point. Increasing the cooling decreases the grain size, increasing the grain density, but also increases the dendrite growth rate, reducing the fraction solid at coherency7 . Permeability Early experiments prompted a redesign of the permeameter to its current configuration. Only very preliminary results are available with the new system, but it hasbeen tested, and the result can be compared to other published results. The measured temperatures, pressures and position can be used to estimate a permeability from Darcy’s law8 , where K=pg,tfL/P=(3xlCr 3 Nsm"2)(0.6)(3.6xl0 -3 ms'!)(0.05 m)/( 1.106x10 5 Nm'2)= 3xl0 42 m2. and p is the viscosity, g, is the volume fraction of liquid, is the displacement over time, L is the length of the mushy zone and P is the average of the applied pressure. This point is plotted in Figure 7. Porosity Measurements The results from three experiments are reported here. The pore areas in Table 2 was measured by circumscribing the region of porosity visible in the X-ray images and then measuring the area with a planimeter. Table 3 presents the results of the density measurements on the three plates. The measured density of the reference casting was - 147 - 4 E 3 b x O D I2 5 O ©(0 E © 1 soa a o c o 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 volume fraction of liquid o Ganesan O Ocansey • Current Figure 7. Graphical comparison of the experimental result obtained in the current study with the previous results reported for permeability of aluminum alloys. Table 2. Experimental Temperature and Porosity Measurements. Exp. Temperature Solid, time Feeding Length Pore Area Die Casting No. (°Q (°Q (sec.) (mm) (mm2) 1 388 720 21 50 0 2 394 720 26.5 59 565 3 389 720 23 75 150 Table 3. Detailed Summary of Plate Density. Plate Density Porosity Literature Measured Theoretical No. (g/cm 3) (%) (%) (%) Ref. 2.6999 -0.62 0.00 -0.67 1 2.6781 0.19 0.81 0.15 2 2.6742 0.34 0.95 0.29 3 2.6773 0.22 0.84 0.17 - 148 - higher than both the theoretical and the literature values. The die temperatures varied slightly between 388 °C for plate 1, and 394 °C for plate 2. The pore area and porosity were larger for the casting with higher die temperatures, but porosity was low for each of the plates examined. The results of the density measurements of the parallelepipeds cut from along the axis of the plate are summarized in Figure 8 . The variation in density measurement for all the plates is small ranging only from no porosity near the ends to about 1 % in the middle of the plate. The horizontal line in Figure 8 represents a threshold value of 0.5 % porosity thatwas used to define the feeding length reported in Table 2. 1.20 - -■ - 2 -0.40 Distance from the feeder [mm] Figure 8 . Summary of the results of the density variations for the test specimens cut from plates 1, 2 and 3. Discussion The combination of coherency, permeability and porosity measurements permits investigation of thefundamental feeding mechanisms. First, eachof thesetools will be discussed in detail. Then, the feeding mechanisms will be analyzed, and the utility of the three experiments will be explained. Dendrite Coherency The measured coherency values for alloy A1 indicate that earlier predictions of coherency, in the range of 50-65 %, do not apply to commercial aluminum alloys. This means that the interdendritic feeding mechanism may start earlier than expected. However, it is possible thatmass feeding continues after the initial coherency point by - 149 - the remelting and deformation of thedeveloping dendritic network. The strength of the initial network is not known at present. Attempts to determine the strength of the network are underway, but results are not yet available. The coherency point does correspond to the beginning of a transition between mass feeding and interdendritic feeding, at least. From thepoint of coherency and onward, the feeding mechanism can no longer be treated as an increasingly viscous slurry. The flow of themelt through the solid network must be considered. Permeability Currently, the value determined for K is only an approximation (± lxlO "12 m2), because of an estimated length of the mushy zone, averaging of the applied pressure, time averaging of the displacement, temperature variations within the crucible and uncertainties in the calculated fraction liquid. Most of these sources of error have already been addressed for future experiments by combining the empirical measurements with a computer simulation of the experiment that incorporates more temperature measurement points, varying pressure with time and varying displacement rates. Quenching experiments are planned that will verify g, and will determine a value for the surface area per unit volume (Sv ), allowing direct comparison with an empirical equation developed by Poirier and Ocansey 8 . Figure 7 compares the estimated value of K in this experiment with other reported permeabilities, and the value appears to be of similar magnitude. The advantages of the current method are the reduced setup time, ease of data collection and collection of data during solidification rather than during remelting of a sample. Collecting data during solidification, instead of during remelting, allows measurement of permeability for highly ramified aluminum alloy microstructures, eliminating the coarsening that was unavoidable during earlier experiments that measured permeability during remelting. Experiments to determine the effect of coarsening may be possible by measuring the permeability at successively longer holding times. However, the permeability of the mushy zone also depends on the shape of the channels and the size of the dendrites 9,10. An analysis of measured permeabilities of columnar-dendritic alloys was conducted by Poirier et al.8,11,12 . They found thatthe permeability for flow normal to the mushy zone varied only with fraction solid but also with both the primary and secondary dendrite arm spacings. The cooling rate, the melt superheat, the alloy chemistry, coarsening within the mushy zone and convection in the melt alters the interdendritic and secondary arm spacing 1314. Reducing the distance between secondary dendrite arms and the interdendritic spacing by increasing the growth rate13'14, for example, decreases the - 150 - permeability of the mushy zone 11. Coarsening increases the distance between the secondary and tertiary arms, increasing the permeability. Increased cooling rates tend to limit coarsening, reducing the secondary dendrite arm spacing, decreasing the permeability of the mushy zone 11. A detailed study of permeability is planned for an AlMg5 alloy to determine the role and significance of some of theseeffects. Shrinkage porosity probably develops when the permeability increases sufficiently to stop the molten metal from feeding the solidification shrinkage. Therefore, efforts to understand the development of shrinkage porosity within a casting requires the ability to predict the permeability of the mushy zone under various casting conditions. Porosity Measurements The X-ray, plate density and density variation results generally agree quite well; however it appears that the density variation measurements resolve the porosity better than the X-ray images. A careful examination of the X-ray images indicates that there might be some porosity visible in plate 1, but most appears to be on the opposite side from the side cut for the test specimens. It is possible that finely dispersed porosity was not resolved by the X-ray for this plate. The density of the plate compares favorably to plate 3, which shares a similar die temperature and porosity variation. Porosity is detectible on the X-ray images for plate 1 and 3, but the density variation measurements appear to resolve porosity better than the X-ray images for both plates. However, plate 2 shows very good agreement between the X-ray image and the density variation, Figure 9. The three areas of porosity within the plate are clearly delineated in both the X-ray image and the density variation. Discontinuous burst feeding is suspected as the mechanism producing the variation along the axis of the casting. This will be discussed in detail in the next section. Feeding Mechanisms The ability of molten metal to flow through the semisolid region of a casting may have a profound effect upon the porosity and surface quality of cast aluminum alloys. The coherency measurements indicate that a coherent network of interconnected dendrites may form early in the solidification process. After this network forms, it is thought that the dendrites are no longer free to flow with the molten metal. Consider the stages of the feeding process. Liquid feeding feeds the solidification shrinkage without significant resistance. During mass feeding, the viscosity of the slurry increases as the volume fraction solid increases. However, the fluidity of the melt with additions of SiC particles has been shown to be only slightly affected by a fraction solid below 20 %15, and the effect of coherency is far greater. As the individual dendrites begin to interact, a slow rise in the torsion curve is observed during coherency measurements, but as thedendrites begin to form a coherent network, the torsion increases sharply7 . During - 151 - 1.2 •• 3.0 Porotity along pot» 2 — 2.0 0.6 ••• — 1.0 — 0.0 Distance from the feeder [mm] Figure 9. Graphical comparison of the density of the test specimen cut from plate 2 and the pore area fraction measured from the x-ray image of plate 2. interdendritic feeding, themelt is forced by the solidification shrinkage to pass through a stationary, interconnected network of dendrites. The transition to interdendritic feeding can be predicted by measuring the coherency point for an alloy. Although some mass feeding may continue beyond this point, the coherency point provides a convenient point of demarcation between the two mechanisms, as a first approximation. After coherency a loose network of interconnected solid grains has developed. The high permeability of the network allows melt, and probably some solid, to easily flow through the solid structure to feed the solidification shrinkage. Little deformation of the solid occurs, although the solid structure may be greatly influenced by remelting and enhanced coarsening by local variations in the solute content of the melt. As solidification continues, the permeability of the mushy zone decreases as the solid fraction increases, increasing the stress on the network. However, the strength of the network also increases as a function of the solid fraction. If the stress increases faster than the strength of the network, burst feeding will occur when the "yield stress" of the network is exceeded by the pressure induced by solidification shrinkage. The collapse - 152 - of the solid network will temporarily reduce the stress on the network and provide mass to the solidification shrinkage. The pressure will then gradually increase, but now the collapsed network would act as a barrier to continuing interdendritic feeding. The barrier might result in the development of discontinuous, shrinkage induced, porosity or deformation of the solid. Alternatively, the increasing pressure required for continued feeding of the solidification shrinkage eventually becomes too great and comparatively evenly distributed porosity and solid feeding begin to develop. The alternating low and high porosity zones in the casting shown in Figure 9 indicate that burst feeding might have occurred. This discontinuous porosity is a particularly important problem. It is difficult to predict, and usually leads to a higher rejection rate or a redesign of the gating and risering of the mould itself. A comparison between the strength and permeability of the network might be able to predict the transitions between interdendritic and burst feeding. The initial test of the permeameter indicates that it might be capable of measuring the flow rate of the melt through a solidifying mushy zone as a function of the applied pressure, temperature and alloy composition for a variety of casting conditions and solid fractions. A technique of measuring the strength of the network is currently being devised. Some promising measurements have been obtained using an apparatus similar to that used in coherency measurements, but they are too preliminary to report. The combination of strength and permeability measurements could be used in a casting simulation, similar to the plate casting geometry, using temperature data measured during the casting of the plates. If the areas of discontinuous porosity could be reproduced, the technique could be applied to increasingly complex casting simulations to predict the occurrence of burst feeding. The simulations could then be used to eliminate burst feeding or minimize porosity by optimizing the casting geometry. Conclusions The preliminary experiments conducted during this investigation indicate that nearly all of the tools necessary to investigate the interdendritic feeding and burst feeding mechanisms are now available. The coherency measurements provide a point corresponding to the transitions between mass feeding and interdendritic feeding. The permeameter is able to measure the permeability of the solidifying dendritic network as a function of the applied pressure, temperature and alloy composition. Only a method of measuring the strength of the dendritic network needs to be developed to permit a direct comparison between the stress applied to the mushy zone by melt flowing through the mushy zone and the strength of the network. - 153 - The discontinuous porosity in the plate castings might be caused by the burst feeding mechanism. As the fraction solid increases in the mushy zone, the stress on the solid network increases. If this stress becomes too great for the strength of the network, burst feeding will occur. The burst feeding mechanism would provide a temporary reduction in pressure and would supply mass to solidification shrinkage, but might act as a barrier to continued feeding of solidification shrinkage, causing localized porosity development or solid feeding in the isolated portion of the casting. Plate 2 indicates three distinct areas of porosity. If the strength and permeability of the mushy zone were known as a function of solid fraction, the temperature data from the plate casting could be used to simulate the burst feeding mechanism. This information could be used to design an experiment to confirm the role of the burst feeding mechanism in the development of discontinuous porosity in a casting. The results could then be applied to more sophisticated casting simulation that could be used to minimize porosity by optimizing riser placement and feeding length, for example. References 1. J. Campbell, Castings, Butterworth-Heinemann Ltd., Oxford, 1991, p. 191 2. L. Backerud, G. Chai and J. Tamminen, Solidification Characteristics of Aluminum Alloys, AFS/SkanAluminium, Stockholm, Sweden, 1990, p. 129 3. J. Campbell, AFS Cast Met. Res. J., March, 1969, pp. 1-8 4. G. Chai, Chem. Comm., No. 1, Stockholm University, Stockholm, Sweden, pp. 12-21 5. J. Tamminen, Chem. Comm., No. 2, Stockholm University, Stockholm, Sweden, 1988, pp. 7-34 6. X.G. Chen and S. Engler, Metall, vol. 45, No. 10, 1991, pp. 995-1000 7. A.K. Dahle and L. Amberg, The 4 ’th Int. Conf. on Aluminum Alloys, T.H. Sanders Jr. and E.A. Starke Jr. (Eds.), Atlanta(GA), 1994, pp. 91-98 8 . D.R. Poirier and P. Ocansey, Mat. Sci. Eng., vol. A171, 1993, pp. 231-240 9. R.P. Lowell and G. Bergantz, NATO ASI Series E, D.E. Loper (Ed.), No. 125, 1987, p. 385 10. F.A.C. Dullien: Porous Media Fluid Transport and Pore Structure, Academic Press, New York(NY), 1979, p. 418 11. D.R. Poirier, Met. Trans. B, vol. 18B, 1987, pp. 245-255 12. D.R. Poirier and S. Ganesan, Mat. Sci. Eng., vol. A157, 1992, pp. 113-123 13. C.M. Klaren, J.D. Verhoeven and R. Trivedi, Met. Trans. A, vol. 11A, 1980, pp. 1853-1861 14. R. Trivedi and K. Somboonsuk, Mat. Sci. Eng., vol. 65, 1984, pp. 65-74 15. A. Kolsgaard, Ph.D. thesis, Norwegian Institute of Technology, Trondheim, Norway, 1993, pp. 13-18 - 154 - Article #8 A.K. Dahle, A. Nordmark and L. Arnberg "Measuring Feeding and Porosity in Al-Si Plate Castings" SINTEF Report, STF 24 A96501, January, 1996 Measurements of Feeding and Porosity in Al-Si Plate Castings Arne K. Dahle1, Arne Nordmark 2 and Lars Amberg 1 1 Norwegian Institute of Technology Department of Metallurgy Trondheim, Norway 2 SINTEF Materials Technology Trondheim, Norway Abstract Porosity formation during solidification is a major drawback of the casting processes since it decreases theyield of the foundry and impair the mechanical properties of cast products. Consequently, it is necessary to understand the mechanisms of porosity formation as well as knowing the factors that affect it. The present work was directed towards investigating the relationship between the rheological properties of the mushy zone and porosity content and distribution. Two commercial Al-Si foundry alloys, AlSi7Mg and AlSil lMg, were cast as plates in a preheated steel die. The effect of extra pressure for feeding, hydrogen content, level of grain refinement, eutectic modification, and plate thickness were investigated. Porosity was evaluated by doing density measurements on small cubes along the centre line and X-ray photography. A method for quantifying porosity by automatic image analysis of X-ray images was developed and compared to visual examination. A new, simple method for pressurization of the feeder was developed. The results show a complex influence of the investigated variables on the porosity. The results indicate that burst feeding may be an important mechanism for porosity formation. Introduction The most frequently encountered defect that is created during casting is porosity, and the porosity content of cast materials is one reason for selecting other manufacturing - 155 - operations for products which have high requirements on internal structure and mechanical properties. It has been known for decades that there are two basic reasons for porosity to arise during casting; liberation of dissolved gases and solidification shrinkage. However, the exact mechanisms of porosity formation are still being discussed, and it has been established that the solidification conditions play a leading role in determining the final amount and distribution of porosity. The cooling characteristics, growth rate and alloy composition, which alters the structural features of the solidification front, such as direction!s) of growth and morphology (dendrite arm spacing, specific surface area), and the physical characteristics of the material (solidification range, surface tension, shrinkage, gas solubility, eutectic fraction etc.) together determine the susceptibility to porosity formation. The shrinkage driven material transport that occurs during solidification is termed feeding, and understanding its inherent characteristics and dependence of the solidification conditions is crucial for the ability to limit porosity. Feeding has been subdivided into five separate mechanisms that may occur[l], and the extent of each mechanism depends on fraction solid, composition and rheological properties. As fraction solid increases, the resistance to feeding increases, and the feeding mechanisms occurring at higher solid fractions have therefore been designated to be the most crucial. Interdendritic feeding and burst feeding are coupled through their pressure dependence. Many investigations of feeding and porosity are reported in the literature, and the most common approach is to investigate plate castings. Methods have also been developed to investigate the fundamental characteristics of each feeding mechanism more thoroughly. Engler and coworkers performed ’pour-out ’ experiments to evaluate the variables that determine the duration of mass feeding[2-7], Interdendritic feeding has been investigated to measure permeability by pushing interdendritic liquid through the dendrite network at various solid fractions by means of gas pressure or a melt of different density [8-13]. Mai and Drossel[14], and later Michels and Engler[15], used a beam and scales principle to directly, and nondestructively, determine the duration of all feeding mechanisms in a plate casting. Several empirical, semi-empirical and theoretical criteria functions to predict porosity have been proposed in the literature. Generally these functions are based on thermal conditions, and are therebybased on the assumption that the local thermal gradient and solidification rate determines the porosity. It is known that melt composition (alloying, - 156 - grain refinement and eutectic modification), inclusions and dissolved gas have a significant impact on the amount and distribution of porosity in aluminium alloy castings. However, these variables are not accounted for in the criteria functions. This report summarizes the results of an investigation of porosity formation, and variables that affect porosity formation, in two Al-Si alloys, AlSi7Mg and AlSil 1 Mg, used commercially by Fundo a.s. to cast car wheels. Plate castings were made in a steel die, preheated to 300°C, and the variables that have been investigated are: pressurization of feeder, thickness of plates, hydrogen content, grain refinement, and eutectic modification. The porosity content and distribution was subsequently compared to the measured rheological properties of the mushy zone, i.e., dendrite coherency and strength development. Experimental Techniques Casting Experiments Four commercial hypoeutectic Al-Si alloys have been investigated. One AlSi7Mg alloy and one AlSil lMg alloy was supplied by Fundo a.s.. Two other commercial alloys grain refined with boron, SiBloy, was supplied by Elkem Aluminium ANS, and no extra chemical additions were made to these alloys. Chemical compositions of the base alloys are given in Table 1. Table 1. Concentration of the main constituents in the base alloys [wt%]. Alloy Si Mg Fe Ti B A1 AlSiTMg 7.29 0.310 0.099 0.119 0 bal. 7 SiBloy 7.11 0.244 0.086 0 0.0153 bal. AlSil 1 Mg 11.20 0.172 0.130 0.117 0 bal. 11 SiBloy 10.89 0.117 0.100 0 0.0160 bal. Approximately 60-70 kg of each base alloy was melted in an electric resistance furnace and kept at a temperature of ~750°C. Alloy additions were made to the melt; Grain refinement was obtained by additions of A1T15B1 master alloy rods, and eutectic modification by additions of rod-type AlSrlO. The hydrogen level in the melt was measured with an AlScan analyzer, with proper calibrations to the effect of alloy composition and temperature. A rotor impeller degasser was used to reduce the hydrogen level in the melt, usually for about 30 minutes, using argon gas. Two - 157 - hydrogen levels were investigated; a low gas level corresponding to <0.10 ml H2/100g, and a high gas level corresponding to >0.20 ml H2/100g. Nitrogen gas was purged over the melt surface to limit oxidation and hydrogen absorption. A sample of the alloy was taken from the holding furnace with a ladle and the temperature adjusted to the casting temperature of 715°C. The alloy was subsequently poured into the sprue in the die. Plate castings were made in the steel die illustrated in Figure 1 a) and b). Figure 1 a) shows the cross-section along the centre line of the casting with feeder, runner and sprue. Figure 1 b) shows an illustration of thedie. It consisted of three parts, and could be opened and closed by movement of one side plate and the bottom plate, which were hydraulically controlled. The die was heated by circulating oil to a temperature of about 300°C, and the oil-injection to the three heating channels, i.e., in each side plate and the bottom plate, were in parallel. Oil entered on the side closest to the feeder. The plates had a length of 250 millimeter and a width of 100 millimeter. Two different plate thicknesses were investigated, i.e., 10 and 20 millimeter, adjusted by the level of the bottom plate. An extra pressure for feeding was added on the feeder by increasing the internal pressure within the feeder by an air pressure of 3 bar through a glass tube that was put into the top of the feeder through a hole in the die. The small gap between the glass and the die was sealed with glue. Silica glass tubes with an outer diameter of 9 millimeter and 1 millimeter thickness were used. During mould filling the tip of the tube was sealed by a small plug of heat resistant wool and a piece of exothermic material to ensure that it was not sealed by frozen metal during filling. The pressure was switched on as soon as the mould was filled and the running system frozen off by extra cooling. When the pressure was switched on, the metal was pushed aside by the air and flowed towards sides of the feeder. At the top of the feeder, the metal froze a solid shell around the glass tube, making it pressure tight. No audible leakages were registered when the pressurization was working properly. Table 2 summarizes the experimental variables tested in this investigation. Samples for chemical analysis by spark emission analysis were cast from the alloys at each change in chemical composition. - 158 - Sprue Feeder t<25 Figure 1. Schematic illustration of the die used to make the plate castings in this investigation; (a) Cross section of the die cavity (casting); (b) Illustration of the three tool parts. Table 2. Experimental variables in the feeding investigation of AlSi7Mg and AlSillMg. Variable Low High Plate-thickness 10 mm 20 mm Extra Pressurization None 3 bar Hydrogen Content < 0.10 cm3/100 g > 0.20 cm3/100 g Sr-modification 0 200 ppm Grain refinement 0, 0.05 and 0.20% Ti and SiBloy - 159- Experiments with all five variables were performed on the commercial AlSi7Mg and AlSillMg alloys, where the grain refinement variation was 0 and 0.20% Ti. One alloy with 0.05% Ti was cast for comparison. These experiments were performed in such a way that they later could be subject to a factorial evaluation, quantifying individual and cooperative effects, although this is not reported here. In the alloys grain refined by boron, SiBloy, with 7 and 11% Si, only the effects of the first three variables were investigated. Temperature recordings were made at four points in the die, 14 mm above the centre line of the plate casting, for all castings, to ensure that the castings were made at similar thermal conditions. In addition, it was possible to make five temperature recordings in the casting, along the parting line. Thermocouples were of type K (alumel-chromel), and temperature was recorded at 0.1 seconds intervals. Porosity Evaluation Two methods have been used to evaluate the porosity in the plates: X-ray investigation and density measurements along the centre line. The surface of the plates was somewhat rough due to an uneven coating thickness in the die, and the plates were therefore machined to the same thickness to ensure that the surface roughness did not affect the porosity results. Approximately 0.1 millimeter was scalped on each side. However, the plates were X-rayed both before and after machining. Cubes for density measurements were cut along the centre line of the AlSi7Mg alloy plates as illustrated in Figure 2. Each cube was 15x15 millimeter cross-section and plate thickness. About 1 millimeter of material was lost between each cube in the cutting operation. The cubes were weighed in air and water, and the density was calculated according to Archimedes’ principle. The percent porosity is defined by the relationship: (Pn™”P) % Porosity = 100 where p is the measured density and pmM is the density of the fully dense material. - 160 - Cubes: 15 15xrv=k Feeder / Marks from the ejector pins Figure 2. Cubes cut for density measurements along the centre line, and labelling. The feeder was on the left hand side, and marks from the ejector pins of the die are indicated on the bottom. Table 3 shows the settings of the X-ray device that were used for the two plate thicknesses. Table 3. Settings for X-ray radiography. Plate thickness Voltage Current Exposure time Film fokus [mm] [kvolt] [mA] [min] distance [cm] 10 70 4 0.5 75 20 90 4 0.5 75 Radiographs were taken on an AGFA Gevaert D4 film, which is a fine-grained X-ray film. To obtain quantitative data from the X-ray images, the negatives were scanned by a UMAX PowerLook II-scanner with a diasadapter. The scanner had an optical resolution of 600x1200 dpi. It was connected to a computer with an SCSI. Radiographs were scanned with the software MagicScan 1.3.3, and pictures were processed in Adobe Photoshop. The scanner was programmed to a greyscale setting independent of the contents of the radiographs, i.e., a constant scale. This allowed the radiographs to be compared directly. Files were made in Windows Bitmap format, where the number 0 is completely black and 255 is completely white. The scanned pictures had a resolution of 0.14 millimeter, giving files of approximately 1.1 Mb. A custom program was written to evaluate the blackening in the pictures, both for the whole picture as well as for the same cubes along the centre line as were taken for density measurements, allowing for loss of material during cutting. Comparing the - 161 - density measurements with the average blackening of each cube could then give a relationship between greyscale and density for the two plate thicknesses. Experimental Results and Discussion Thermal conditions of the die The X-ray pictures show that the porosity in all castings is displaced from the centre line. This indicates that the thermal conditions are different in each half of the die, and warmer on the side where porosity is concentrated. Figure 3 shows an example of a temperature recording of the thermocouples in the die during one experiment. It shows that the heating channels causes the initial temperature profile in the die to have a maximum in the middle. When the molten metal is poured into the die, the metal in the feeder heats up the die and creates a positive thermal gradient towards the feeder. The initial thermal profile is not the optimum situation, since a uniform temperature or a positive thermal gradient towards the feeder would have been preferred. However, a positive thermal gradient towards the feeder seems to be established during casting. Close to plate end 2 (middle) 3 (middle) 4 (middle) Close to feeder Time [sec.] Figure 3. Example of a temperature recording of five thermocouples in the die. In the literature, it is common to quantitatively report porosity by density measurements along the centre line. The centre line is often the thermal centre and symmetry plane of the casting. However, as discussed above, the thermal conditions of the plates cast in this investigation do not fulfil this requirement. The density - 162 - measurements were taken along the centre line anyway, since this is ’standardized ’, and the intention was to establish a relationship between porosity and the greyscale. This would make possible later predictions of density in any location in the plate. Extra feeding pressure Figure 4 shows an illustration of the cross-section of the feeders of unpressurized and pressurized feeders along the centre line after the end of solidification. Without the extra pressure, Figure 4 a), the shrinkage in the feeder appears as a large centre line macroshrinkage withpeak in the centre. In the pressurized feeder, Figure 4 b), a large internal shrinkage cavity was observed instead, showing indications of a certain internal ’collapse ’ in the lower part of the feeder, probably due to mass/burst feeding. The volume of the cavity is bigger in the pressurized feeder, and the difference in appearance between the feeders indicates that the feeders have been working differently. Figure 4. Illustration of the appearance of the feeders after the end of solidification for: (a) without extra pressure, and (b) pressurized. During pressurization of a casting it is very important that there are no leakages in the casting itself. If this occurs, the gas will penetrate into and through the casting, creating channels of porosity. A solid shell of material enclosing the casting is therefore a requirement for this type of pressurization to work properly. It may be imagined that the introduction of the pressurizing air at room temperature - 163 - into the feeder would have a cooling effect, and that it in this way would reduce the solidification time and thereby the efficiency of the feeder. An estimate of the cooling effect is shown in the appendix, and shows that this cannot be expected to be a dominating effect of the pressurizing gas in aluminium alloys. Other metals, for example Mg, could show a different tendency since the latent heat is lower. In instances where larger volumes of gas are required for pressurization it may be useful to preheat the gas to reduce the cooling effect. However, some cooling is positive since it promotes sealing of the gas within the feeder. X-rav imaging, scanning and interpretation of pictures Figure 5 shows printed X-ray pictures of scanned plates at the same plate thickness. It is readily observed that there is a big difference in the base greyscale in these pictures, and this was confirmed upon visual examination of the negatives. Since the pictures were scanned onto the same greyscale, this indicates that there are differences in the X-ray plates themselves. There may be several reasons for a difference in the exposure of the X-ray images. The exposure time during X-raying may be different (operator could set the exposure time with an accuracy of 5 sec.), the quality and reproducibility of the X-ray film, the temperature, composition and purity of the developing bath etc. This is clearly the main weakness in this procedure, and needs to be optimized. One way to go to optimize the reproducibility of the X-raying of the plates would be to digitize the picture of a fluorescent screen that works the same way as the negatives do, similar to those that are used for medical purposes. In this way the whole step with film and development would be eliminated. In addition, a more accurate adjustment of exposure time is necessary. Figure 6 shows greyscale vs. density for the thin, a), and thick plates, b), obtained from the centre line cubes. The difference in base greyscale of the images makes the relationship very weak and scattered. However, for each individual plate - X-ray image pair, the relationship between density and greyscale is more credible and reasonable, as observed in Figure 7. The trends observed in density are reproduced by the greyscale of the X-ray images. This indicates that there is a unique relationship between greyscale and density if the X-ray images are obtained withan equal greyscale of the background. The radiographs do not seem to miss any significant information observed in the density measurements, i.e., microporosity. The thick plates are at a lower greyscale range than the thin plates, as may be observed in Figs. 6 and 7. This effect is a result of the radiographs of the thick plates being darker than those of the thin plates. The resolution of the thick plates would have - 164- Figure 5. Printed radiographs showing the difference in greyscale of the images. - 165 - Density [g'cm3] a) Density 2.50- 2.70- Feeder 2.48- 2.52- 2.56- 2.64- 2.60- 2.68 80 [g/cm3] — 161514131211109 Figure Figure Density 90 Sample a ---- ' % 100 - 6 J 7. i No., - Density Greyscale greyscale 110 i t hin i decreasing Greyscale A28-3 i Relationship and 120 Greyscale plates 876543210 i i relationship. 130 (b) : from i thick 140 i feeder vs. i 150 ------ i between plates. Greyscale density 160 - -55 -35 -45 40 60 50 - 166 for density Density Density - separate Feeder 2.60- 2.70- 2.50- 2.68 2.60- 2.64- 2.52 2.56- 2.62- 2.63 2.64- 2.65 2.66 2.67 and 30 [g/cm3] [g/cm3] ------— 16151413121110 55 : ------t Sample greyscale 40 Density plates. T — i — 60 : 50 ' — No., : I 65 — - 60 i greyscale r Thick ~1 i decreasing B29-6 for: Greyscale B29-2 Greyscale 9 70 70 i — t 876543210 i r 1 plates (a) 80 ' i 75 relationship. : 1 from - — 90 thin : r - 80 i Density Greyscale 100 feeder 9 i plates, s ------i 85 Greyscale 110 B - i r j_ r- - - 120 90 110 115 120 125 135 become better if the darkness was compensated during the scanning operation, increasing the applied greyscale. The difference in base greyscale, as well as difference in X-ray settings, makes it impossible to directly compare the radiographs of thin and thick plates. To be able to directly compare and quantify porosity in the thick and thin plates it was decided to do a greyscale compensation. This was done by first calculating the average greyscale in two small cubes, 5x5 mm, at locations where the plates were assumed to be fully dense, i.e., close to the feeder and on the far end. The average of the greyscale of the cubes was subsequently used as a greyscale reference for fully dense material. (It should be noted that the greyscale values obtained at these two points showed that all plates are brighter at the feeder end than at the far end, indicating that there is a gradient in the greyscale along the plate. This effect is probably due to an uneven irradiation of the plates during X-raying or that the plates actually are denser close to the feeder.) A ’cal-value ’ equal to the difference between average total greyscale and greyscale reference, divided by the greyscale reference was calculated to characterize the total porosity content in the plates. This corresponds to the percentage deviation from the reference. A completely dense plate would then have a cal-value equal to zero. Reproducibility in the measurements The reproducibility in the casting operation was tested by making several castings at similar conditions, machining, X-raying and scanning the radiographs. Similarly, several radiographs were taken of some plates to check the reproducibility of the X- raying. Radiographs were scanned several times to check the reproducibility of the scanning operation. The reproducibility of the scanning operation was very good, and identical cal-values were obtained from each scanned image. The reproducibility of the X-raying was also quite good, at least if it was done on the same day, i.e., with the same settings. However, later results indicated that the settings varied according to when the X-raying was performed, and this was one reason whyit was decided to do the ’calibration’ . For castings made under similar conditions and X-rayed at different points of time, the deviation in the cal-values was usually below 10%, but a deviation of up to 50% was observed for some castings. The casting operation is certainly the most difficult to reproduce since there are so many factors that have to be kept under control, and that may cause errors in the - 167 - a) Density [g/cm3] factors, reproducible Effect pressurized for Figure porosity. Visual concluded a examination. the that AlSi7Mg the additon, Figure condition, plate 15 pressurized unpressurized cal-values, this ------ Sample 14 of Figure 8 9 inspection A- as where 13 1 ----- the conclusion a) shows extra and — confirming much that 12 and plates fluctuation manner No., A B29-2 B29-1 11 AlSillMg. 8 porosity pressurization but . printed b) the and plate 10 as decreasing plates clearly (without (Pressurized) shows 9 is this pressurization, Effect unpressurized possible, is B29.2/Z, with and 8 higher the not radiographs result in 7 is is P) two However, measuring indicates observations a porosity 6 always of ' eliminated. more from large 5 than pressurization and examples under is 4 feeder also evenly area correct. 3 castings. of when of is that the control, some #08-2/3. 2 an less, some the of calculated calculated made - of extra indication porosity distributed. 168 working density same Of of and plates The i.e., on b) above. - these pressurization course, the plate density figures by in may cal-values measurements properly, large variables. of the performing Pressurization All 15 there without the be ----- pressurized and -A Sample 14 along peak show judged in subjective 13 ----- are reduces alters printed all - 12 in that the the many No., B08-2 B08-3 11 it the porosity along as the may 10 extra centre seems the more in experiments decreasing porosity (without (Pressurized) nature uncertainties 9 distribution Table the density unpressurized definitely 8 pressure. porous observed to line centre 7 P) of compact 4 6 for in from shows of visual 5 both than in line the be a) 4 of of In in feeder a 3 2 1 Figure 9 shows printed radiographs of some plates in the pressurized and unpressurized condition, confirming the observations made above. Pressurization seems to compact the porosity area. (Figure 9 has been put at the end of this report.) Very few investigations of the effect of pressurizing the feeder are reported in the literature (one reason being the practical difficulties), but those that exist reports an advantageous effect of pressurization[16-19]. Examples of processes where the application of extra pressure for feeding is used to produce complex shapes with improved mechanical properties and integrity are squeeze casting, centrifugal casting! 1] and the Castyral process[18]. Pressure die casting and similar methods are not considered in this group since they also use a high pressure during filling, and are known to contain porosity[l]. Increasing the pressure on the feeder will increase the pressure driving the flow to compensate for shrinkage: Mass feeding will generally proceed at a higher rate, and maybe giving some more fragementation of the dendrites, but its extent would not be expected to change significantly. Pressurization is more important for the interdendritic and burst feeding mechanisms, which are also the most critical ones since they occur at higher solid fractions. When a pressure is applied to the feeder, the pressure gradient for flow through the dendrite network increases, and flow becomes more efficient. The formation of porosity is governed by the following criterion]!]: Pg + Pp > Patm + Ph + P0 where Pg is the equilibrium gas pressure of dissolved gas, Pp is the pressure drop due to shrinkage, Patm is the atmospheric pressure over the system, Ph is the metallostatic pressure, and Pc is the pressure from surface tension at the gas-liquid interface. This equation shows that pressurizing the feeder, i.e., increasing Patm, will impede the formation of gas and shrinkage pores by increasing the resistance to porosity nucleation. In addition, increasing the efficiency of the feeding mechanisms will reduce the shrinkage pressure, Pp. - 169 - Table 4. Alloys and casting conditions where pressurization is the only variable, and the calculated cal-values. Alloy Grain ref. Eut.mod. Hydrogen Plate No P Extra P AlSi7 No No High Thick 0.1548 0.1402 AISi7 0.20 Ti Sr High Thick 0.1515 0.1803 AISi7 No No High Thick 0.1618 0.1556 AlSi7 No No Low Thick 0.1471 0.1585 AlSi7 No Sr High Thin 0.0698 0.0465 A1S17 Boron No Low Thin 0.0755 0.0526 AlSi7 Boron No . Low Thick 0.1967 0.1571 AlSi7 Boron No High Thin 0.0917 0.0847 AlSi7 Boron No High Thick 0.1250 0.1846 A1S17 No No Low Thick 0.1644 0.1684 AlSill No No Low Thick 0.1471 0.1687 AlSill No Sr High Thin 0.0841 0.0305 AlSill 0.20 Ti No High Thin 0.0750 0.0677 AlSill 0.20 Ti Sr Low Thick 0.1818 0.1493 AlSill Boron No Low Thin 0.0446 0.0800 AlSill Boron No Low Thick 0.1429 0.2266 AlSill Boron No High Thin 0.1102 0.1150 AlSill Boron No High Thick 0.1449 0.1728 Effect of thickness of cast plates: The effect of plate-thickness was only uniquely investigated in the SiBloy alloys. Visual examination of the radiographs indicates that the effect is more pronounced in the alloys with 7% Si. Table 5 summarize the results. The cal-values hereby indicate that the thick plates always are markedly more porous than the thin plates in all casting conditions for these alloys. The visual observations indicate that the distribution of porosity is different for the two plate thicknesses, and the thin plates of AlSi7 were interpreted as more porous than the thick. The region of higher porosity is longer and wider in the 10 millimeter than in the 20 millimeter plates. Although the porosity may be higher in the thick plates, it is - 170 - concentrated to a relatively narrow region. In addition there is a significantly higher content of microporosity in the thick plates (which may be so small in the thin plates that they are not picked up by the radiographs, i.e., less than 0.14 mm). Table 5. Alloys and casting conditions where the hydrogen content is the only variable. Cal-values indicates porosity as deviation from base greyscale. Alloy h2 Pressure Thin Thick thin cal thick cal A1S17+B low Not A28-1 A28-3 0.0755 0.1967 A1S17+B low Pressure A28-2 A28-4 0.0526 0.1571 A1S17+B high Not A28-5 A28-7 0.0917 0.1250 A1S17+B high Pressure A28-6 A28-8 0.0847 0.1846 AlSill+B low Not A19-1 A19-3 0.0446 0.1429 AlSill+B low Pressure A19-2 A19-4 0.0800 0.2266 AlSill+B high Not A19-5 A19-7 0.1102 0.1449 AlSill+B high Pressure A19-6 A19-8 0.1150 0.1728 Reducing the plate thickness implies reducing the volume of the casting, and the total shrinkage and feed metal requirement will therefore be less. An increase in cooling rate (reduction of solidification time) and thermal gradients accompanies the thickness reduction. This results in a smaller mushy zone and a reduction of dendrite arm spacing. The Niyama criterion G/y'R , where G is thermal gradient and R is cooling rate, is often used as a porosity predictor. It has been proposed that the value required to produce sound castings increases with increasing section thickness. For thick castings the Niyama function will assume a small value, and a large amount of porosity is predicted. The higher porosity content observed in the thick plates is probably due to the slower cooling rate and lesser thermal gradients in the thick plates, which gives longer time for hydrogen diffusion into the interdendritic spaces, resulting in larger pore size, and a larger mushy zone that makes feeding more difficult. Porosity formation therefore becomes more difficult at increasing cooling rates, and thin castings tend to contain less and more finely dispersed porosity. However, the increased cooling rate with reducing plate thicknesses will reduce the dendrite arm spacing. Interdendritic fluid flow is dependent on the specific surface area of the solid, which determines the inherent flow resistance and required pressure - 171 - gradient for flow. The specific surface area is increased when the dendrite arm spacing is reduced, and due to a lower permeability, interdendritic flow will be more difficult. Changing the plate thickness is hereby related to several opposing effects. Figure 10 a) shows printed pictures of an alloy cast under similar conditions, except for a difference in plate thickness (A28-1/3), and figure 10 b) shows the density measurements along the centre line for these plates. The observations made above are largely confirmed. A28-1 (thin) - A- A28-3 (thick) 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Sample No., decreasing from feeder Figure 10 (a) Radiographs illustrating the effect of plate thickness have been put at the back of this report; (b) Effect of plate thickness on density along the centre line for the same two plates The thin plates would be expected to be more susceptible to burst feeding than the thick ones. The small dendrite arm spacing and early coherency point, combined with large flow resistance would provide the conditions for network collapse, which is also observed on the radiographs. However, measurements on the effect of cooling rate on strength development should be performed to substantiate these speculations. The literature[ 17,20-27] confirms the observations made in this investigation, i.e., thinner plates, at higher cooling rates, are more dense than thicker plates, or lower cooling rate. Generally it is found that increasing the cooling rate gives an increase in porosity content and average pore size. - 172 - Effect of structurally refining additions: Grain refinement was varied under the conditions of low hydrogen content, thick plate and no extra pressure. Table 6 summarizes the results for the AlSi7 alloy. Table 6. Effect of grain refinement and Sr-modification on porosity in AlSi7. Alloy Marking Observation Cal-value A1S17 B29-3 4 0.1644 A1S17+ 0.05 Ti B30-1 3 0.1648 A1S17+ 0.20 Ti B30-2 1 0.1739 AlSi7+0.2Ti+Sr B08-4 2 0.1863 AlSi7+ B A28-3 5 0.1967 An attempt was made to rank the porosity in the plates from visual inspection of the radiographs. Ranking 1 is lowest in porosity, and 5 highest (i.e., the porosity level is largest in the alloy grain refined with boron, and least in the alloy with 0.20% Ti). This ranking indicates that grain refinement reduces porosity, with exception of the 7SiBloy alloy, and that eutectic modification with Sr increases porosity. The calculation from the greyscale of the pictures indicates that porosity is continuously increased with increasing level of grain refinement, i.e., increasing in the order: unrefined - 0.05Ti - 0.20Ti - 0.20Ti0.02Sr - Boron. The calculation hereby shows unequivocal trend that addition of grain refiner and Sr-modification increases the porosity content. Visual inspection of the radiographs shows that the length of the area of higher porosity is decreasing in the order: unrefined - 0.05Ti - Boron - 0.20Ti0.02Sr - 0.20Ti. In the unrefined alloy there are indications of burst feeding, by regions of high density separating the porous regions, and this tendency is reduced, or not so easily identified, when the size of the region is decreased. Similarly, for the AlSillMg alloys: - 173 - Table 7. Effect of grain refinement and Sr-modification on porosity in AlSil 1. Alloy Marking Observation Cal-value AlSil1 7112-3 4 0.1471 AlSil 1+0.05 Ti 7113-1 5 0.1385 AlSil 1+0.20 Ti 7113-2 1 0.1818 AlSil l+0.2Ti+Sr A13-6 3 0.1493 AlSil 1+ B 7119-3 2 0.1429 The effect of structural refinement is more scattered for AlSil 1, both in the visual observation and calculation. Upon visual inspection of the radiographs, no significant difference in porosity distribution is discernible. Porosity is generally evenly distributed, except for the alloy with0.05% Ti where a larger porosity area is observed. In both Table 6 and Table 7 it may be noticed that the observed differences and ranking of porosity differ from what is calculated from the greyscale of the images. Density measurements were only performed for the AlSi7 alloys, and Figure 11 shows how the density variation along the centre line changes with grain refinement and Sr- modification. The alloys which have the largest porosity peak are: unrefined, 0.05Ti and SiBloy, and the smallest is 0.20Ti. The alloy with 0.20Ti has a large area with a reduced density, but not the type of characteristic peak observed in the other alloys. The unrefined alloy has the largest length of the porosity region. Addition of AITi5Bl or B promotes an accelerated nucleation rate and a decreased grain size during solidification of the alloys. The dendrites of a grain refined casting are often observed to have a more globular morphology due to a lower growth rate. These effects promote a delayed impingement of the dendrites, and the dendrite network is hereby established later in the solidification process. Dendrite coherency would be expected to have a significant effect on feeding and subsequent porosity, since it defines the size of the mushy zone where mass feeding (transport of a slurry with dispersed dendrites) is the main mechanism of material transport. Since mass feeding is more efficient than the subsequently occurring feeding mechanisms, enhanced mass feeding is expected to reduce porosity. Grain refinement, and to a small extent Sr-modification, postpones dendrite coherency, and therefore increases the region of mass feeding. However, the structural alteration will affect the subsequently occurring transport mechanisms. The evolution of specific surface area with fraction - 174 - // A g 2.60 ------1----- AlSi7 (B29-3) % \\ -A - AlSi7+0.05Ti (B30-l)\ \ C AlSi7+0.20Ti(B30-2) —AlSi7+0.20Ti+Sr(B08-4)\ — Q — AlSi7+B (A28-3) 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 Sample No., decreasing from feeder Figure 11. Effect of grain refinement and eutectic modification on density along the centre line of AlSUMg. solid of the small, globular dendrites characterizing the grain refined alloy will be different from that of the unrefined structure. This causes a difference in the inherent resistance to interdendritic feeding, i.e., the permeability, which therefore requires a different pressure gradient for flow at a similar fraction solid. The load and stresses on the solid skeleton increases concomitantly. Measurements have shown that grain refinement gives a slower increase in strength of the mushy zone in AlSi7, and only minor changes in AlSill. The effectiveness of grain refinement in reducing porosity in casting alloys has not clearly been demonstrated in the literature. In the few papers published in this area it has been indicated that the grain size reduction in hypoeutectic Al-Si alloys does not have significant effects on mechanical properties [28], but can rather lead to a decrease in ductility [29]. Michels and Engler[15] found no effect of grain refinement on feeding efficiency in hypoeutectic Al-Si alloys. Irani and Kondic[17] observed increased porosity with grain refinement in AlSi5Cu3, but reduced porosity in AlSil2 plate castings. Several investigators report small increases in cast sample density with grain refinement[30-33] . - 175 - One reason for the ambiguous effect of grain refinement on porosity reported in the literature may be that grain refinement, although improving mass feeding, may have a deleterious effect on the subsequent feeding mechanisms. The small, globular dendrites of the grain refined alloy may be compacted more at the end of mass feeding so that the interdendritic feeding paths are smaller and more tortuous. Apelian et al.[9] reported a consistently lower specific permeability of grain refined samples at equivalent solid fractions in AlSi4. The importance of this effect may depend on the casting conditions and flow rate (pressure gradient), which may therefore explain why results are scattered. Strontium is added to hypoeutectic Al-Si alloys to improve mechanical properties by changing the silicon morphology from acicular to fibrous. It has been shown extensively in the literature that Sr-modification increases porosity and the average pore size[20-21,23,25-26,34-36]. Emadi and Gruzleski[26] found that the increase in porosity was higher at low cooling rates and high hydrogen contents. Although the effect of Sr in increasing porosity has long been recognized, the explanation of the mechanism is still under dispute. The following factors have been suggested[20- 21,23,25-26,34-36]: Increase in the hydrogen content of the melt, or the rate of regassing. Oxidation of Sr during melting changes the structure of the surface oxide, which becomes more permeable to hydrogen Increase in both the inclusion content and the amount of hydrogen absorbed on these Reduced surface tension of the liquid Decreased hydrogen solubility in the solid, or increased in the liquid Increased volumetric shrinkage Reduced interdendritic feeding by 1) changing the shape of the solid-liquid interface 2) increasing the length of the mushy zone 3) increasing melt viscosity Increase in time for hydrogen diffusion and porosity growth due to an increase in freezing range by reductions in the eutectic growth temperature. Which one or combination that is most important is not known. - 176 - Effect of melt hydrogen concentration Table 8 shows the alloys and casting conditions where the hydrogen content of the melt is the only variable. Porosity in the thin plates increases with increased hydrogen content, observed by a higher cal-value of the high-hydrogen plates. In the thicker plates the results are scattered, and there may be several reasons for this. Since the total variation in greyscale of the thicker plates was lower, due to settings in the scanning, and a low base greyscale, the hydrogen effect may not be picked up. Hydrogen may also be present as microporosity, which is not so easily picked up on X-ray pictures. Table 8. Alloys and casting conditions where the hydrogen content is the only variable. Cal-values indicates porosity as deviation from base greyscale. Alloy Plate Pressure Low H2 High H2 low cal high cal A1S17+B Thick Not A28-3 A28-7 0.1967 0.1250 A1S17+B Thick Pressure A28-4 A28-8 0.1571 0.1846 A1S17 Thick Not B29-3 B29-2 0.1644 0.1548 A1S17 Thick Pressure B29-4 B29-1 0.1684 0.1402 AlSi7+0.2Ti+Sr Thick Not B08-4 B08-2 0.1683 0.1515 A1S17+B Thin Not A28-1 A28-5 0.0755 0.0917 A1S17+B Thin Pressure A28-2 A28-6 0.0526 0.0847 AlSill+B Thick Not A19-3 A19-7 0.1429 0.1449 AlSill Thick Not A12-3 A12-2 0.1471 0.1923 AlSill Thick Pressure A12-4 A12-1 0.1687 0.1594 AlSill+B Thick Pressure A19-4 A19-8 0.2266 0.1728 AlSill+B Thin Not A19-1 A19-5 0.0446 0.1102 AlSill+B Thin Pressure A19-2 A19-6 0.0800 0.1150 The following observations are made: 1) Thin plates: hydrogen increases porosity. 2) Pressurized thick plates: Hydrogen seems to reduce porosity. 3) Thick, unpressurized AlSi7Mg: hydrogen reduces porosity. The density measurements along the centre line, some examples which are shown in Figure 12, indicate that the general density is lower, but that the size and extent of the large porosity peaks are reduced in the plates with a high hydrogen content. Porosity - 177 - Density [g/cm3] in P Increasing is be nucleation The increase nucleation zone, with experiments, be obtained dominating formation particles wetting etc. 2.66 Adding 2.67- 2.68 g +Pp>P\ available these calculated several - -A The 15 increasing formation giving Figure Sample 14 plates hydrogen porosity towards in oxide are 13 where of barrier the and reasons this role for 12 measuring porosity dependent an from 12. No., content 11 is growth hydrogen these content investigation on of enhanced P more 10 since the to g decreasing to porosity, for Effect porosity Sieverts is 9 porosity processes. the of - melt[l]. and the by 8 evenly the this. hydrogen, A of on dissolved content, 7 melt reducing of equilibrium the equilibrium - the growth 6 the One law: since melt from is do formation. B29-1 B29-4 5 distributed resistance will melt, Similarly thermal generally not 4 uncertainty P and density hydrogen feeder inclusions (High) (Low) g hydrogen period. shrinkage 3 affect =(C(/K) always as porosity 2 gas, gas - also history The to 178 measurements, as the the governed pressure Efficient P 2 feeding. conform with content . micropores. may act p affected equilibrium would b) is - pressure, The efficiency is size may experimental of the 2.50- 2.55- 2.60- 2.65- equilibrium therefore the and is from on shrinkage as feeding by 15 by increased. melt, density so nuclei Sample 14 morphology radiographs feeding the pouring gas this this that 13 form and 12 pressure can following error equation expectation. for a pressure, gas along No., B29-2 B29-3 11 smaller However, mechanisms. porosity, may earlier and suppress 10 pressure in decreasing and (High) (Low) 9 the of performing in filling, be also 8 interpretation driving and requirement: in the inter-metallic centre expected 7 There being the the both affect increases melt P* 6 from plays mushy As results 5 is force line. pore non may can 4 the the the the feeder to a 3 2 pressure gradient is increasing with increasing solid fractions in the mushy zone, burst feeding and gas porosity nucleation are two competing events. Hydrogen pore formation may therefore reduce the significance of burst feeding. Since burst feeding is expected to be more important for AlSi7Mg, the effect of hydrogen would be more important for this alloy, which seems to be the case. Hydrogen evolution affecting burst feeding, may also serve as an explanation for the difference in response to hydrogen in thin and thick plates. The literature in general shows that larger hydrogen contents increases porosity[15,26], AlSi7Mg vs. AlSillMg In most cases it is evident that there is a marked difference in the porosity distribution between AlSi7Mg and AlSillMg alloys. While porosity in the AlSi7Mg alloys is observed as relatively large areas of higher porosity, porosity is more evenly distributed in the AlSillMg alloys. The solidification range of thesetwo alloys is different, which helps somewhat to explain these results, since the size of the mushy zone is smaller in the AlSillMg alloy. However, more important is the difference in strength development in the mushy zone of the two alloys[37]. It was observed that the mushy zone had a very low strength up to fractions solid of about 90% in AlSil lMg, whereas strength started to develop earlier and more rapidly in the AlSi7Mg alloy [37], Burst feeding, i.e., network collapse and compaction, will occur whenever the strength of the dendrite network is exceeded. If burst feeding occurs in the AlSillMg alloy, the strength measurements indicate that the whole mushy zone may collapse, and the subsequent porosity, due to compacted fragments clogging the feeding path to the areas beyond, may therefore be of a relatively small extent. In the AlSi7Mg alloy, the network may only collapse at lower solid fractions, and the pockets of porosity may therefore be larger. This is in good accordance with the observations. These results conform well to those reported in the literature. Chen and Engler[38] found that as the eutectic composition is approached, the tendency to porosity formation is decreased due to a decrease of freezing range and the resultant lack of dendrite structure. Increasing the silicon content leads to a reduced dendrite arm spacing, and, thus, porosity is more finely dispersed. Michels and Engler[15] measured a reduced feed metal requirement with increasing Si-contents. This is consistent with what could be expected. Since pure silicon expands upon solidification, an increase in silicon content of the alloy would reduce the total shrinkage. Emadi and Gruzleski[26] found that AlSi7 was more susceptible to porosity than AlSi5.7 and AlSiS.l. - 179 - Conclusions Feeding and porosity have been investigated in two commercial Al-Si foundry alloys; AlSi7Mg and AlSillMg. Plates were cast in a steel die, preheated to 300°C. The effects of extra pressure on the feeder, level of grain refinement, eutectic modification, plate thickness, and melt hydrogen concentration were investigated. Porosity was evaluated by performing density measurements on small cubes along the centre line and X-ray photography. A method to obtain quantitative data from the radiographs was developed, involving scanning and subsequent determination of greyscale on the images. There was a good correlation between density along the centre line and greyscale for each plate. However, due to inaccuracies in the X-raying operation, the greyscale of each plate was not comparative and a normalization based on the greyscale of fully dense material was necessary. The normalized results were not always in accordance with the visual impression of the amount of porosity. A simple method for pressurizing the feeder was developed, involving an increase of the internal pressure within the feeder by applying an air pressure through a glass tube immersed into the feeder. The results show a very complex influence of the investigated variables on porosity. Porosity content and distribution were compared to the measured dendrite coherency point and strength development in the mushy zone. Since the results were scattered it is difficult to reach any straightforward conclusions about the relationship between mushy zone rheology and porosity. It is speculated that interdendritic feeding may be impaired when dendrite coherency is postponed, an effect that is important both for burst feeding as well as for the precipitation of dissolved hydrogen. Indications of burst feeding was observed, being more pronounced in AlSi7 than in AlSil 1. This conforms with what could be expected from measurements of the strength development in the mushy zone. One reason for the scatter in the results may be that the occurrence of burst feeding has a big impact on the level of porosity. References 1. J. Campbell: "Castings", Butterworth-Heinemann, Oxford, 1991 2. W. Patterson and S. Engler, Giesserei, vol. 48, No. 21, 1961, pp. 633-638 3. S. Engler and L. Heinrichs, Aluminium, vol. 48, No. 5, 1972, pp. 351-356 - 180 - 4. S. Engler and L. Heinrichs, Aluminium, vol. 49, No. 6, 1973, pp. 426-429 5. S. Engler and M. Dette, Giesserei, vol. 61, No. 26, 1974, pp. 769-775 6. W. Kallmeyer and S. Engler, Giessereiforschung, vol. 37, No. 4, 1985, pp. 131- 144 7. W. Kallmeyer, W. Michels and S. Engler, Giessereiforschung, vol. 38, No. 1, 1986, pp. 1-10 8. T.S. Piwonka nd M.C. Flemings, Trans. Met. Soc. AIME, vol. 236, 1966, pp. 1157-1165 9. D. Apelian, M.C. Flemings and R. Mehrabian, Met. Trans., vol. 5, 1974, pp. 2533-2537 10. D.R. Poirier and S. Ganesan, Mat. Sci. Eng., vol. A157, 1992, pp. 113-123 11. D.R. Poirier and P. Ocansey, Mat. Sci. Eng., vol. A171, 1993, pp. 231-240 12. P. Ocansey, M.S. Bhat, D.R. Poirier and T.L. Finn, Light Metals 1994, U. Mannweiler (ed.), The Minerals, Metals and Materials Society, 1994, pp. 807- 812 13. CJ. Paradis, L. Arnberg, HJ. Thevik and A. Mo, Modeling of Casting, Welding, and Advanced Solidification Processes VI, TMS, London, 1995 14. R. Mai and G. Drossel, Giessereitechnik, vol. 29, No. 2, 1983, pp. 46-49 15. W. Michels and S. Engler, Giessereiforschung, vol. 41, No. 4, 1989, pp. 174- 187 16. J.T. Berry and T. Watmough, AES Trans., vol. 69, 1961, pp. 11-22 17. D.R. Irani and V. Kondic, AES Trans., vol. 77, 1969, pp. 208-211 18. M. Garat, S. Guy and J. Thomas, The Foundryman, January, 1991, pp. 29-34 19. N.A. Mufti, P.D. Webster and T.A. Dean, Mat. Sci. Eng., vol. 11, 1995, pp. 803-809 20. D. Argo and J.E. Gruzleski, AFS Trans., vol. 96, 1988, pp. 65-74 21. Q.T. Fang and D.A. Granger, AFS Trans., vol. 97, 1989, pp. 989-1000 22. Y.S. Kuo, E. Chang and Y.L. Lin, AFS Trans., vol. 97, 1989, pp. 777-782 23. H. Iwahori, K. Yonekura, Y. Yamamoto and M. Nakamura, AFS Trans., vol. 98, 1990, pp. 167-173 24. Y.S. Kuo, E. Chang, R.R. Jeng and H.H. Ho, AFS Trans., vol. 98, 1990, pp. 801-811 25. S. Shivkumar, L. Wang and D. Apelian, J. Met., January, 1991, pp. 26-32 26. D. Emadi and J.E. Gruzleski, AFS Trans., vol. 102, 1994, pp. 307-312 27. M. Tiryakioglu, C.W. Ramsay and D.R. Askeland, AFS Trans., vol. 102, 1994, pp. 737-748 28. M. Tsukuda, T. Suzuki, I. Fukui and M. Harada, J. Japan Inst. Light Metals, vol. 29, 1979, pp. 437-444 29. H.T. Wu, L.C. Wang and S.K. Kung, J. Chinese Foundryman’s Assoc., vol. 29, - 181 - 1981, pp. 10-18 30. G. Drossel, R. Mai and O. Liesenberg, Giessereitechnik, vol. 27, 1981, pp. 167- 180 31. G.K. Sigworth, Proc. Conf. Int. Molten Metal Processing, AFS, Des Plaines (IL), 1986, pp. 75-99 32. W. LaOrchan and J.E. Gruzleski, AFS Trans., vol. 100, 1992, pp. 415-424 33. L. Wang, S. Shivkumar and D. Apelian, 2nd Int. Conf. on Molten Aluminum Processing, AFS, Des Plaines (IL), 1989, pp.5.1-5.20 34. E.N. Pan, H.S. Chiou and GJ. Liao, AFS Trans., vol. 99, 1991, pp. 605-621 35. G.K. Sigworth, C. Wang, H. Huang and J.T. Berry, AFS Trans,, vol. 102,1994, pp. 245-261 36. D. Emadi, J.E. Gruzleski and J.M. Toguri, Met. Trans. B., vol. 24B, 1993, pp. 1055-1063 37. A.K. Dahle and L. Amberg, submitted to Acta Met. Mat. 38. X.G. Chen and S. Engler, Metall, vol. 45, Oct., 1991, pp. 1-17 - 182 - Appendix Cooling effect of the pressurizing air: The feeder of one pressurized casting was sectioned and the volume of the gas-filled internal cavity was approximated to V= 39.2 cm3. Approximating the air to pure nitrogen, the heat capacity is Cp= 29 J/Kmol. At STP, the no. mol of nitrogen in the cavity is: 39.2 1.749 mol 22.414 The energy required to heat the air to the temperature of the metal, i.e., from 25 to 680°C is: h;;; = 29(953-298 ) = 18995 j If the energy to heat the air should be supplied by solidifying some aluminium, the required weight and volume would be (Ahp 10 kJ/mol, M= 27 g/mol, p= 2.7g/cm 3): ~-!!'27 '513s -19otS This seems like a relatively small volume compared to the volume of the feeder, and heat is also supplied by cooling conduction through the metal. It may hereby be concluded that the cooling effect of the pressurizing gas would not be expected to result in a reduced efficiency of the feeder. A2S-3 Figure 9-1. Radiographs showing the effect of pressurization. 7SiBloy, low H2 content, Thick plate. Top picture: No extra pressure; Lower picture: Pressurized. (cont.) 184 Figure 9-2. Radiographs showing the effect of pressurization. AlSi7, low H2 content, Thick plate. Top picture: No extra pressure; Lower picture: Pressurized. (cont.) - 185 - BG8-2 Figure 9-3. Radiographs showing the effect of pressurization. AlSi7+0.20Ti+Sr, High H2 content, Thick plate. Top picture: No extra pressure; Lower picture: Pressurized. (cont.) 186 Figure 9-4. Radiographs showing the effect of pressurization. HSiBloy, low H2 content, Thick plate. Top picture: No extra pressure; Lower picture: Pressurized. (cont.) - 187 - Figure 9-5. Radiographs showing the effect of pressurization. AlSil 1, low H2 content, Thick plate. Top picture: No extra pressure; Lower picture: Pressurized. (cont.) 188 - AT 3 -4 Figure 9-6. Radiographs showing the effect of pressurization. AlSill+0.20Ti, High H2 content, Thin plate. Top picture: No extra pressure; Lower picture: Pressurized. 189 Figure 10 (a) Radiographs showing the effect of plate thickness. 7SiBloy, Low H2 content, No extra pressure. Top picture: Thin plate; Lower picture: Thick ISBN 82-7119-890-4 ISSN 0802-3271