Contraction of Aluminum Alloys during and after Solidification D.G. ESKIN, SUYITNO, J.F. MOONEY, and L. KATGERMAN A technique for measuring the linear contraction during and after solidification of aluminum alloys was improved and used for examination of binary and commercial alloys. The effect of experimental parameters, e.g., the length of the mold and the melt level, on the contraction was studied. The correlation between the compositional dependences of the linear contraction in the solidification range and the hot tearing susceptibility was shown for binary Al-Cu and Al-Mg alloys and used for the estimation of hot tearing susceptibility of 6XXX series alloys with copper. The linear thermal contraction coefficients for binary and commercial alloys showed complex behavior at subsolidus temperatures. The technique allows estimation of the contraction coefficient of commercial alloys in a wide range of temperatures and could be helpful for computer simulations of geometrical distortions during directchill (DC) casting.

I.

INTRODUCTION

THE process of direct-chill (DC) casting is the most common way to produce ingots and billets for further deformation processing. Despite the fact that this technology has been used in the aluminum industry since the 1950s, the cause of some common defects is still under discussion. Hot tearing, porosity, macrosegregation, and distortion of billet geometry (e.g., butt curl) are the major defects that occur during casting. The feeding of the growing solid phase with the liquid (hence, the permeability of the mushy zone), the structure formation, the development of strength in the mushy zone, and the solidification shrinkage together with thermal contraction being in complex interaction may result in the formation of defects. Hot tearing or hot cracking is one of the most common problems encountered in DC casting of aluminum alloys. The main cause of this defect is that stresses and strains built up during solidification are too high compared to the actual strength of the semisolid material. This type of defects occurs in the lower part of the solidification range, close to the solidus, when the solid fraction is more than 0.9.[1] At this point, the mushy zone is definitely coherent, but the liquid film still exists between most of the grains. The term coherency (or coherency temperature) should be used with caution. If the coherency is understood as a temperature at which a continuous dendritic network is formed, and the material starts to develop strength and retain its shape,[2,3] then this point can be better defined as a rigidity point. At temperatures above the rigidity point, the grains are free to move with respect to each other and so do not transfer any forces. Moreover, before the rigidity temperature is reached upon solidification, the liquid phase can easily flow between grains and, therefore, the melt feeding and D.G. ESKIN, Senior Scientist, is with the Netherlands Institute for Metals Research, 2628AL Delft, The Netherlands. Contact e-mail: d.eskine@ tnw.tudelft.nl L. KATGERMAN, Professor, and SUYITNO, Postgraduate Student, are with the Department of Materials Science, Delft University of Technology, 2628AL Delft, The Netherlands. J.F. MOONEY, Graduate Student, is with the Department of Engineering Materials, University of Sheffield, S1 3JD Sheffield, United Kingdom. Manuscript submitted May 6, 2003. METALLURGICAL AND MATERIALS TRANSACTIONS A

the redistribution of solute elements occur without much difficulty. This terminology is the one we adopt in this article. The question is when hot tearing really occurs, and what are the driving forces for hot tearing. To answer the first question, the terms “the effective solidification range”[2,4] and “the vulnerable part of the solidification interval”[5] were introduced in the 1940s–1950s. The upper boundary of this range is the point where the stresses begin to build up,[2] and the lower boundary is the solidus (equilibrium or nonequilibrium, depending on the solidification conditions). As for the driving force, it is generally accepted that the inadequate feeding compensation of the shrinkage in the presence of thermal stresses is the major origin for the occurrence of hot tears in DC casting.[6] The necessary condition for hot tearing is the existence of thin, continuous, interdendritic liquid film alongside the low permeability of the mushy zone. This condition is usually fulfilled at large volume fractions of solid, 0.9 to 0.99.[1,7] Novikov[2] suggested determining the upper boundary of the effective solidification range by measuring a so-called linear shrinkage. Hence, the upper temperature of this solidification range is the temperature at which the “linear shrinkage” starts. Let us specify the terms we are using in this paper. The solidification shrinkage is the shrinkage (usually volume shrinkage) that occurs in the solidification range, from 100 pct liquid to 100 pct solid, as a result of the density difference between the liquid and solid phases. Solidification shrinkage of aluminum alloys amounts to 6 to 8 vol pct. The thermal contraction is the contraction of the solid phase resulting from the temperature dependence of the density. The thermal contraction is usually described by the linear or volume thermal expansion coefficient. The linear contraction (or shrinkage) is the horizontal change in linear dimensions of a casting during solidification and usually ranges from one hundredth to one percent.[2,8] Above the temperature of the linear contraction onset, the alloy is fluid because between opposite walls of the mold there is no continuous network of dendrites. In this stage of solidification, the solidification shrinkage of the melt and the thermal contraction of the solid phase cannot manifest themselves as the horizontal contraction of the casting. All volumetric changes appear as the decreasing level of the melt in the permanent VOLUME 35A, APRIL 2004—1325

mold (experimentally demonstrated in Figure 1(a)) or do not appear at all during DC casting, due to the continuous supply of melt to the mold. However, the linear contraction appears, and can be measured, when the fluidity of the alloy drops drastically, and the rigid skeleton of the solid phase forms. Starting from that moment, the alloy acquires the capability to retain its shape, and the volumetric changes display themselves as the linear contraction in the horizontal direction (Figure 1(b)). The understanding of the shrinkage and contraction phenomena occurring in the solidification range is very important for the analysis of stress-strain development and modeling of hot cracking. The temperature dependence of the linear contraction in the solidification range is used in some hot tearing criteria.[2,9,10] Below the solidus, the thermal contraction continues and frequently manifests itself as geometrical distortion of the billet shape. In DC casting practice, this phenomenon is known as a butt curl (lifting of the billet shell from the starting block). The occurrence of butt curl can reduce the stability of the ingot standing on the starting block and is therefore a potential safety hazard. Besides that, the partial loss of contact between the ingot and bottom block will initially reduce the heat transfer with the possible danger of remelting. In the worst case, butt curl can cause cracks and hot tears. A special technique was developed to measure the linear contraction (and preshrinkage expansion) upon solidification and applied to some binary and commercial alloys.[2,8,11]

Several designs of an experimental setup were suggested, all sharing the following features: graphite mold (providing low friction and high thermal conductivity) with one moving wall; water-cooled base (for high cooling rates comparable with those upon DC casting); and simultaneous temperature and displacement measurements. The aim of this article is to describe the development of an experimental technique, to discuss the experimental factors influencing the measured property, and to apply the experimental results obtained using the developed technique to the analysis of hot tearing susceptibility and the contraction development during and after solidification.

II.

EXPERIMENTAL PROCEDURE

The experimental setup used measures the linear contraction upon solidification and is based on the idea suggested by Novikov.[2] It consists of the following parts: a T-shaped graphite mold (Figure 2) with one moving wall; a watercooled bronze base; and a linear displacement sensor (linear variable differential transformer (LVDT)) attached to the moving wall from outside and aligned with the longitudinal axis of the mold. The reason behind the T shape, which is narrower than the main cavity, is to make the melt solidify faster there than in the rest of the mold, and so the solidifying sample can be fixed on that side. To attach the solidifying metal to the moving wall on the other side of the cavity, we use a metallic rod with a thread (screw) embedded into the moving wall, as shown in Figures 2 and 3. The metallic rod fixed in the moving head is frozen in the sample immediately after filling the mold with the melt. The cross section of the main cavity was 25 3 25 mm with a gage length of 100 mm. The linear displacement is measured by a SCHAEVITZ* *SCHAEVITZ is a trademark of Schaevitz Sensors, Hampton, VA.

(a)

(b) Fig. 1—(a) Experimentally measured melt level and (b) horizontal contraction of an Al-4.5 pct Cu alloy. Solidification shrinkage manifests itself in the decrease of melt level (measured by a laser sensor) up to the moment when the rigid solid skeleton is formed and the contraction begins (measured by an LVDT). Note the difference in scales on vertical axes. 1326—VOLUME 35A, APRIL 2004

DC-DC LVDT, which is accurate to 6 mm or 0.006 pct, and two to four samples are measured for each point. A low-friction mechanism of the LVDT and low friction of the moving wall/mold (graphite on graphite) contact provide minimum interference into the measured expansion/contraction. The temperature is measured by K thermocouples with 0.25-mm-thick wires and an open tip that enables quick response to the changing temperature. In the case of measuring the displacement, the reference thermocouple is placed either in the center of the mold or in the corner of the mold in the central cross section, the distance between the thermocouple tip and the bottom of the mold being about 1.5 mm (Figure 3(b)). When studying the temperature distribution in the sample during solidification, a grid of thermocouples was inserted into the mold and temperature was measured at six positions inside the cavity (Figure 3(a)). Accuracy of temperature measurements was within 2 K. During the experiments, the temperature and displacement are recorded simultaneously by a PC-based data acquisition system, at least 20 points per second per channel being registered.[8] The linear solidification shrinkage (contraction) is determined as follows: «s 5 {(ls 1 Dlexp 2 lf)/ls } 3 100 pct METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 2—Diagram of an experimental mold with a moving wall.

(a)

(c)

(b)

(d)

Fig. 3—Scheme of the mold cavity with applied insulation paint: (a) top view and (b) side view (circles show typical positions of thermocouples, gray color shows the paint, and distances are in millimeters) and the schematic propagation of solidification front (solidus isotherms) in the case of (c) bare and (d) painted walls.

where ls is the initial length of the cavity, lf is the length of the sample at a temperature of solidus, and Dl exp is the preshrinkage expansion. The preshrinkage expansion is mainly due to the evolution of gases and the pressure drop over the mushy zone, METALLURGICAL AND MATERIALS TRANSACTIONS A

and it depends on the alloying system, melting, and solidification conditions.[2,12] The effect of heat-transfer conditions was studied by applying a refractory paint (bone ash) onto the internal surface of the mold, as shown in Figures 3(a) and (b). VOLUME 35A, APRIL 2004—1327

In most cases, liquidus and solidus temperatures can be derived from the cooling curve. Note, however, that we determine the linear contraction in the entire solidification range, which, at the used cooling rates (5 to 10 K/s), extends to the lowest possible eutectic temperature—nonequilibrium solidus (NES). After acquiring the primary data, temperature, and displacement against time, the cooling curve is processed in order to obtain information about critical temperatures and cooling rates. After that, the data are reconstructed to find the direct dependence of displacement on temperature. From this dependence, the linear preshrinkage expansion, the linear solidification contraction, the temperature of its onset, and the linear thermal contraction coefficient (TCC) can be extracted (Figure 1(b)). Earlier experiments described elsewhere[8] showed the feasibility of the developed technique, but there were also some discrepancies with previously reported data,[2] mainly in the amount of the contraction accumulated during solidification and on the temperature of the contraction onset. It was clear that the temperature measurements and computer modeling of solidification in the experimental mold should be performed in order to understand the solidification pattern. The first results of computer simulations performed by finite element modeling with MSC.MARC software[13] show that the solidification front progresses along the longitudinal axis of the mold (as expected and desired), but, at the same time, the solidification along the walls occurs much faster than along the centerline, as schematically shown in Figure 3(c). As a result, two fronts meet in the central cross section of the mold, but rather near the walls than in the center where the reference thermocouple is conventionally placed. Therefore, the measured temperature does not reflect the real temperature at which the contraction starts. The temperature measurements along the longitudinal and cross sections confirmed that the cooling rate close to the wall is indeed much higher (2 to 4 times) than at the same height in the center. This is illustrated in Figure 4(a). Attempts have been made to place the reference thermocouple close to the wall, but, due to the filling problems and air gap formation, the open tip of the thermocouple is frequently detached from the sample and the readings from this thermocouple become unreliable. A way was found to reduce temperature gradients in the central cross section of the mold and to achieve the required progress of the solidification fronts, as schematically shown in Figure 3(d). The application of the refractory paint (bone ash), as shown in Figures 3(a) and (b), considerably reduces the heat flux through the side walls and to the corners of the mold, thus equalizing cooling rates and flattening the solidification fronts (Figure 4(b)). Figure 5 demonstrates the distribution of solidification times along the longitudinal axis of the mold for the cases of bare and thermally insulated walls. In fact, the longitudinal gradients increase with painted walls, but the cross gradient virtually vanishes in the center. It is important to note that the macrostructure of the sample also changes with the cooling conditions. Figure 6 shows examples of central cross sections of Al-3.7 pct Cu samples. When the mold walls are bare, the structure is mixed and well-pronounced columnar grains mark the direction of heat extraction to the corners and bottom of the 1328—VOLUME 35A, APRIL 2004

(a)

(b) Fig. 4—Cooling curves in two positions of thermocouples (near the wall and in the center) in the central cross section of the sample: (a) with bare graphite walls and (b) with thermally insulated graphite walls (Figs. 2(a) and (b)).

Fig. 5—Average times when the equilibrium solidus is reached along the longitudinal axis and a side wall of the mold for two cases (with bare and thermally insulated walls). Positions of thermocouples are given by points (scheme shown in Figs. 3(a) and (b)).

mold. After reducing temperature gradients in this section, the structure becomes much more uniform with mostly equiaxed grains. With this modification of the experimental technique, the point at the bottom in the center of the mold is the place where two solidification fronts (from the fixed wall and from the moving block) meet. Therefore, the readings from the thermocouple placed in this point can be reliably used as a reference temperature for displacement measurements. METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 6—Macrostructures of the center cross section of Al-3.7 pct Cu samples cast in the mold with bare walls (left) and thermally insulated walls (right).

Table I. Alloy Group Al-Cu

Chemical Composition of Experimental Alloys (in Weight Percent)

Cu 0.1; 0.2; 0.4; 0.8; 1.5; 2.5; 3.5; 4.5

Al-Mg

— 0.22; 0.33; 0.44; 0.70; 1.39; 2.05; 3.31; 3.79 ,0.01

60611 Cu 5182*

Mg

Fe

Si

Ti (as Al-3 pct Ti-1 pct B Rod)

— 0.2; 0.5; 1.0;1.5; 2.0; 3.0; 5.0; 6.0

,0.1

,0.15

—

,0.1

,0.15

—

0.52 to 0.67 0.21

0.01 0.02

0.83 to 0.96 3.6

0.01 0.26

*Contained also 0.16 pct Mn.

Several binary and commercial alloys were tested using the developed technique. Their chemical composition is given in Table I. III.

RESULTS AND DISCUSSION

A. Effect of Experimental Parameters on the Measured Contraction After the experimental technique was improved in terms of better heat-transfer conditions and the correct reference temperature, we used it to test the effect of the experimental parameters on the measured contraction. Previously,[8] we showed that the friction force applied to the moving block can considerably decrease the measured contraction in the solidification range, not affecting, however, the temperature of the contraction onset. In this work, two geometrical parameters were changed: the gage length of the mold and the height of the sample (amount of metal poured in the mold). METALLURGICAL AND MATERIALS TRANSACTIONS A

Experimental results show that variation of the gage length from 100 to 50 mm does not affect the resulting contraction. The amount of melt cast into the mold affects the measured parameters, which is evidently due to the combined influence of thermal gradients, dimensionality of the solidifying sample (unidirectional solidification, flat, or full threedimensional casting), and the mechanical and rheological properties of the mushy zone. The onset of the contraction and the amount of contraction accumulated in the solidification range can decrease with decreasing amount (level) of melt, as in the case of an Al-3.5 pct Cu alloy, or behave inversely in the case of a 5182 (Al-Mg-Mn) alloy. It is known from the casting practice that the observed geometrical changes in the size of a DC cast billet cannot be readily explained as a result of the thermal contraction of a solid sample of the same size. The observed contraction is usually considerably larger. The “real” casting contraction increases with increasing casting speed and varies from the theoretical value (calculated using the linear thermal expansion VOLUME 35A, APRIL 2004—1329

coefficient) at a zero speed to up to 50 pct larger values at high casting speeds.[14] The increase of the experimentally observed contraction with respect to the theoretical value originates from thermal gradients in the casting. Evidently, different layers of the casting are at different stages of solidification, contain different amounts of the solid phase, and contract at different rates. In addition, the internal, more liquid parts of the casting undergo, along with intrinsic contraction, tension imposed by the external, already solid and more contracted shell. The difference between the theoretical and observed contraction apparently depends on the mechanical properties of the mush, the thickness of the solid shell, and the volume of complete fluid slurry and melt (the depth and profile of the sump during DC casting). The “softness” of the liquid interior, which offers only negligible resistance to the contraction of the external shell, contributes to the experimentally observed contraction.[13,14] Moreover, the pressure drop that appears in the mushy zone due to the solidification shrinkage and the poor permeability of the mush could add some contraction by deforming the very weak solid network. The solidification of melt in the experimental mold used in this investigation is in many aspects (although on the other scale) similar to the solidification of an ingot or a billet cast from the top and cooled from the bottom. In this case, the change of the amount of melt is similar to the decrease of the sump depth. With taking all this into account, we can say that the tested Al-3.5 pct Cu alloy, which shows the decrease of the observed contraction with decreasing amount of melt, behaves normally. The question remains why does the 5182 (Al-Mg-Mn) alloy show an opposite behavior. Four factors may play a role here: (1) the amount of solidification contraction is smaller in the case of the 5182 alloy (0.13 to 0.17 pct against 0.15 to 0.28 pct for the Al-Cu alloy); (2) the effective solidification range is larger for the 5182 alloy (90 °C to 120 °C against 35 °C to 65 °C for the Al-Cu alloy); (3) the tested 5182-type alloy is a commercial alloy containing grain refiners and, therefore, having a much finer structure than the model binary Al-Cu alloy; and (4) the overall contraction behavior of Al-Mg alloys is different from that of Al-Cu alloys.[18] The decrease in the amount of melt poured into the experimental mold and, as a result, the decreased melt level and the depth of the molten pool in the mold reduce the experimental linear thermal contraction coefficient for both tested alloys. These results are discussed in more detail later in this article. B. Correlation between the Contraction during Solidification of Aluminum Alloys and the Hot Tearing Susceptibility In this work, we performed experiments and measured contraction parameters using alloys listed in Table I. The main aim of these experiments is to follow the effect of chemical composition on the contraction of binary Al-Cu and Al-Mg alloys and to compare the obtained results with the compositional dependence of hot tearing susceptibility reported elsewhere. The contraction is calculated in the temperature range between the temperature of contraction onset and the temperature of the nonequilibrium solidus (NES), the latter considered to be equal to the eutectic temperature for alloys containing more than 0.5 pct Cu or Mg, which is quite reasonable for the used cooling rates (5 to 10 K/s).[15] 1330—VOLUME 35A, APRIL 2004

Experimental results shown in Figures 7(a) and (b) demonstrate that (1) the contraction (under given solidification conditions) starts in the temperature range between the equilibrium and nonequilibrium solidus and (2) the compositional dependence of the contraction shows a maximum in the range of alloys with the largest effective solidification range. The comparison of these results with the results on hot tearing susceptibility (Figures 7(c) and (d)) reveals that the compositional range of maximum hot tearing coincides with the range of the largest contraction in the solidification range. These results can be interpreted as follows. During solidification, a semisolid sample undergoes contraction. This contraction starts at a certain temperature in the solidification range, which we agreed to call the rigidity point. At this temperature of contraction onset, the semisolid material acquires some strength and ability to transfer stress and strain through a coherent solid network of dendrites, the solid fraction being usually higher than 0.9.[2,8] The amount of free contraction accumulated in the solidification range determines the strain imposed onto the semisolid material. If the ductility of this material does not allow it to accommodate the contraction-induced strain, then failure or cracks may occur in the semisolid state. Under conditions of poor permeability of the semisolid mush and, hence, lack of feeding through the dense solid network, this crack may not be filled with new melt (healed) and will develop into a hot tear. This line of logic is in the basis of several hot-tearing criteria.[2,9,10] It is important to note here that our recent results[13] confirm the conclusion of Novikov[2] that the contraction observed in the solidification range below the rigidity temperature is caused by the thermal contraction of the solidphase network. Experimental results given in Figure 7 demonstrate that the contraction in the solidification range can be used as a measure of hot tearing susceptibility. The analysis of the contraction during solidification is quite obvious and straightforward for binary alloys, when the phase diagram with all temperatures and concentrations is known. The situation changes for commercial, multicomponent alloys. In most cases, the phase diagram and corresponding data are unknown. In this case, the measurement of the contraction in the solidification range can provide useful information about the temperature and concentration range where these alloys are prone to hot tearing. As an example, we tested a 6061-type alloy containing different amounts of copper (within and exceeding compositional range of the grade). The effect of grain refiner was also examined. The results are summarized in Figure 8. The nonequilibrium solidus is adopted as 500 °C based on an Alstruc estimation.[17] The following conclusions can be made: 1. contraction maxima in the solidification range correspond to 0.2 to 0.3 pct Cu and to about 4 pct Cu; 2. contraction starts at about 600 °C to 610 °C and the onset temperature decreases to 585 °C on increasing copper concentration from 0.2 to 3.8 pct; 3. grain-refined alloys exhibit much more pronounced preshrinkage expansion than non-grain-refined alloys, with the maximum between 0.7 and 1.5 pct Cu; 4. the addition of grain refiner (0.01 pct Ti/B) dramatically decreases (2 to 3 times) the amount of the contraction accumulated in the solidification range; METALLURGICAL AND MATERIALS TRANSACTIONS A

(a)

(c)

(b)

(d)

Fig. 7—Compositional dependences of contraction behavior superimposed on the phase diagram (a) and (b), dotted line shows the NES, and hot tearing susceptibility[16] (c) and (d) for binary Al-Cu and Al-Mg alloys.

Fig. 8—Effect of copper concentration and grain refining on the contraction/expansion behavior of a 6061-type alloy: e contraction onset, °C; r contraction onset with TiB, °C; s contraction to 500 °C, pct; d contraction to 500 °C with TiB, pct; h expansion, pct; and j expansion with TiB, pct.

5. the addition of grain refiner decreases the temperature of contraction onset by 20 °C to 60 °C depending on the composition of an alloy. METALLURGICAL AND MATERIALS TRANSACTIONS A

The experimental observations confirm the previously reported fact that grain refining decreases the hot tearing susceptibility of alloys.[2,16] Two factors act here. On one hand, the contraction and, therefore, strain accumulation starts at much lower temperature, thus reducing the vulnerable solidification range and total strain during solidification. On the other hand, the preshrinkage expansion is much more pronounced, compensating for some portion of the thermal contraction. The much larger preshrinkage expansion of grain-refined alloys can be explained in the following way. It is known that the expansion is caused by the evolution of gas (mainly hydrogen) during solidification[2] and the pressure drop across the mushy zone.[12] In the experimental setup used in this work, the solidification starts at the ends of the mold (Figure 3). The temperature gradient in the longitudinal direction (Figure 5) causes nonuniform volume shrinkage and, therefore, an uneven pressure drop in the two-phase zone. The pressure in the mushy zone near the ends of the solidifying sample is lower than in the center. As a result, there is a pressure-induced flow directed from the center toward the ends of the sample. Since the dendrites of the solid phase are mostly separated by liquid films (expansion occurs above the rigidity point), they are shifted outward from the center of the sample. This movement is registered as a preshrinkage VOLUME 35A, APRIL 2004—1331

expansion. In the case of grain-refined alloys (Figure 9), the structure consists of small, equiaxed grains that become bridged at relatively low temperature (shrinkage onset). Such a structure facilitates the relative movement of grains and, therefore, expansion. In addition, gas evolves from the melt more eagerly as fine grains provide more large-angle interfaces and more time before rigidity. The maximum of expansion frequently corresponds to the maximum distributed porosity,[12] as can also be seen from the structure in Figure 9. Therefore, alloys that are less vulnerable to hot tearing may exhibit larger interdendritic porosity. C. Contraction of Aluminum Alloys at Subsolidus Temperatures The analysis of geometrical changes of a billet in the last stages and immediately after solidification requires the knowledge of the thermal contraction coefficient. Because the thermal contraction/expansion is a physical property of the material, it is conventional to use the reference data on the linear thermal expansion coefficient (LTEC). There

(a)

(b) Fig. 9—Microstructures of a 6061-type alloy with 1.4 pct Cu: (a) without grain refiner and (b) with grain refiner. 1332—VOLUME 35A, APRIL 2004

are, however, two drawbacks of this situation. First, the linear thermal expansion coefficients are seldom available for commercial alloys at high, subsolidus temperatures. Second, the linear thermal expansion coefficient is usually determined by a dilatometer on long cylindrical (one-dimensional) samples under nearly isothermal conditions (i.e., no thermal gradients in the sample). In addition, these samples are carefully homogenized to achieve the equilibrium state of the alloy. This situation is very different from the real contraction conditions of a just solidified and cooling bulk sample in which the processes of excess-phase precipitation may well continue. We suggest using our technique for the estimation of linear thermal contraction coefficients at subsolidus temperatures. The average TCC is calculated as follows: TCC 5 [(LT 2 2 LT1)/Lgage ]/(T 2 2 T1) where T2 and T1 are the temperatures (T2 . T1) below the solidus; LT2 and LT1 are the positions of the displacement sensor at T2 and T1, respectively; and Lgage is the gage length of the sample. Figure 10 demonstrates the TCC for binary Al-Cu and Al-Mg alloys with respect to the composition. The coefficients were estimated in two temperature ranges: within 50 °C below the nonequilibrium solidus (adopted as 550 °C for Al-Cu alloys and 450 °C for Al-Mg alloys) and from 300 °C to the solidus. It is worth mentioning here that the results reflect the nonequilibrium situation of cooling after solidification. In binary Al-Mg alloys, the TCC decreases with an increasing concentration of magnesium, the coefficients being very close at subsolidus and lower temperatures. This trend is quite distinct from the previously reported results. According to the literature, the linear thermal expansion coefficient in Al-Mg alloys is believed to increase with Mg concentration and temperature.[2,18] In the given compositional range, the linear thermal expansion coefficient, as determined on annealed samples in a dilatometer, increases with magnesium concentration from 33.6 to 35 3 1026 K21 at 500 °C[18] and from 29.2 to 30 3 1026 K21 at 400 °C.[2] At the same time, the average LTEC (20 °C to 500 °C) slightly decreases with increasing magnesium, from 29.3 to 29 3 1026 K21.[18] Our experimental results give a much more pronounced change in the TCC (300 °C to 450 °C), from 29 to 25.5 3 1026 K21. The contraction behavior of Al-Cu alloys in the temperature range from 300 °C to 550 °C complies with the previously reported data. According to Novikov,[2] the LTEC at 500 °C decreases from 32.4 to 31.2 3 1026 K 21 on increasing the copper concentration from 0 to 4.5 pct. The average LTEC (20 °C to 500 °C) decreases from 29 to 28 3 1026 K 21[18] in the given range of copper concentrations. In our experiments, the average TCC (300 °C to 550 °C) decreases from 30 to 28.5 3 1026 K21 . This is in good agreement with the data by Ellwood and Silcock[19] on the subsolidus LTECs of binary Al-Cu alloys. They reported that the LTEC decreases from 32.4 3 1026 K 21 at 0.95 pct Cu to 28.2 3 1026 K 21 at 4.97 pct Cu.[19] However, the contraction behavior of these alloys at subsolidus temperatures observed in our work is quite different. After an initial decrease in the alloys containing up to 1 pct Cu, the measured TCC increases with copper concentration and METALLURGICAL AND MATERIALS TRANSACTIONS A

(a) (a)

(b) (b) Fig. 10—Effect of composition and temperature range on the linear thermal contraction coefficient in binary (a) Al-Cu and (b) Al-Mg alloys.

becomes higher than the TCC (300 °C to 550 °C) at copper concentrations greater than 2.5 wt pct. The application of the well-known Grüneisen equation[20] to the examined alloys shows that the thermal expansion coefficient of the solid solution should increase with copper concentration and slightly decrease with magnesium concentration, which is rather opposite to the experimental “dilatometric” results reported in the literature. On the other hand, the linear thermal expansion coefficients of the involved phases are quite different, being approximately 32 3 1026 K21 at 420 °C [21] or 38 3 1026 K 21 at 500 °C [18] for alloys close to Al8 Mg5 and 17.4 3 1026 K 21 in the range from 100 °C to 500 °C[22] for Al2Cu. Therefore, the application of a mixture rule gives the increase in the LTEC with concentration in heterogeneous Al-Mg alloys and the opposite trend in the case of Al-Cu alloys. When considering our experimental results and comparing them with those reported in the literature on LTECs, one should bear in mind that the amount of excess phase (Al2Cu or Al8Mg5) in the alloy and the amount of alloying element in the aluminum solid solution in our experiments are quite different from those in the references. A solidified sample contains a considerable amount of nonequilibrium eutectics METALLURGICAL AND MATERIALS TRANSACTIONS A

Fig. 11—Effect of the sample height on the linear thermal contraction coefficient of (a) an Al-3.5 pct Cu and (b) a 5182-type (Al-Mg-Mn) alloy.

and the depleted solid solution, while a homogenized sample of the same composition will comprise only the equilibrium solid solution containing either copper or magnesium. In our experiments, the ratio between the amount of excess phases and the amount of alloying element in the solid solution may play a decisive role in the measured thermal contraction. An additional argument in favor of our results is the report by Afanas’ev et al.,[23] where the authors examined the thermal expansion of cast and homogenized Al-Mg alloys and found that the LTEC at 400 °C decreases with magnesium concentration in cast Al-Mg alloys and increases in homogenized alloys. This “anomalous” behavior was observed only at temperatures higher than 200 °C and was correlated with the amount of hydrogen in the alloy—the higher the amount of hydrogen, the more pronounced the “anomaly.” The effect of sample height (amount of the melt cast into the mold) on the estimated contraction coefficient of two alloys, i.e., Al-3.5 pct Cu and the 5182-type alloy, was also studied. The results are given in Figure 11. In both cases, the TCC decreases with the decreasing height of the sample (or the depth of the sump), the estimated TCC being higher at subsolidus temperatures. This change of TCC reflects the inhomogeneity of the solidifying sample, as has been discussed earlier in this article. The obtained values of TCC can be readily used in computer simulations of contraction during and after solidification. It is VOLUME 35A, APRIL 2004—1333

IV.

Fig. 12—Butt curl simulated using thermal expansion coefficients based on Ref. 26 (e , s ) and determined in this work by contraction experiment (h , n ), the first benchmark case (a modified Ludwik model) (h , e ), and the second benchmark case (a Ludwik model for the solid and a rheological model for the mush) ( s , n ).

worth noting here that the modeling of shape distortion (butt curl) during DC casting showed that choosing a correct model is much more important than the knowledge of the exact thermal contraction coefficient. The simulation is performed for an AA5182 slab 1897-mm wide, 550-mm thick, and 620-mm high. A finite-element model was implemented in the MSC.MARC software. The procedure is similar to that reported elsewhere.[24] All thermophysical properties used in this simulation are taken either from References 25 and 26 or from our experiments. Thermal conductivity, specific heat, and Young’s modulus are implemented as functions of temperature, while the other properties are kept constant. Two benchmark cases are selected for the description of the thermomechanical behavior of the material over a wide range of temperatures, from solidification range down to subsolidus temperatures. In the first case, the behavior of the solid material is described by a modified Ludwik model,[27] and that of the semisolid mush, by extrapolating the prediction of the Ludwik model to the semisolid temperature range. In the second case, the Ludwik model is used only for the description of the thermomechanical behavior of the solid material at subsolidus temperatures, whereas the mushy zone is described by a viscoplastic model within which the plastic deformation of the mush depends on solid fraction, strain rate, and temperature.[28] The parameters in the equations used are fitted to experimental data, as described elsewhere.[29] The results show that the butt curl (Figure 12) computed using the tabular coefficient of thermal expansion[26] does not differ significantly from the value calculated using the experimentally determined coefficient of thermal contraction. The significance of taking into account the semisolid plastic flow behavior is evident. The results obtained for the second benchmark case are much closer to the available experimental results.[30] However, knowledge of the thermal contraction coefficient is still required for the modeling. Despite the observed complex behavior of TCC, the suggested technique can be used as a relatively simple means for the estimation of the real contraction coefficient at subsolidus temperatures, especially in cases when it is unknown. 1334—VOLUME 35A, APRIL 2004

CONCLUSIONS

1. The developed technique allows characterization and quantification of the contraction behavior of aluminum alloys during and after solidification under nonequilibrium conditions close to those of casting practice, by means of the following quantities: the temperature of the contraction onset, the preshrinkage expansion, the amount of contraction in the solidification range, and the thermal contraction coefficient at subsolidus temperatures. 2. The amount of the contraction accumulated in the solidification range is shown to correlate well with the hot tearing susceptibility of binary alloys. The effect of copper and grain refinement on the contraction and, hence, hot tearing susceptibility is examined for a 6061-type alloy. Two maxima of contraction are found at 0.2 to 0.3 and 4 pct Cu. The addition of a grain refiner to the 6061-type alloy decreases the solidification contraction by 2 to 3 times and the temperature of the contraction onset by 20 to 60 K. 3. The thermal contraction coefficient at subsolidus temperatures shows complex behavior with respect to the composition of binary alloys and the experimental parameters. However, our technique is highly reproducible and can be used for the determination of the contraction coefficient at high temperatures. ACKNOWLEDGMENTS The work is done within the framework of the research program of the Netherlands Institute for Metals Research (www.nimr.nl), project Nos. MP97014 and MC02135. The authors thank A. Stangeland and Y. van der Drift for their assistance in performing some of the experiments.

REFERENCES 1. J. Campbell: Castings, Butterworth-Heinemann, Oxford, United Kingdom, 1991. 2. I.I. Novikov: Goryachelomkost tsvetnykh metallov i splavov (Hot Shortness of Nonferrous Metals and Alloys), Nauka, Moscow, 1966. 3. L. Arnberg and L. Bäckerud: Solidification Characteristics of Aluminum Alloys, vol. 3, Dendritic Coherency, AFS, Des Plaines, IL, 1996. 4. A.A. Bochvar: Izv. Akad. Nauk SSSR, OTN, 1942, No. 9, p. 31. 5. W.S. Pellini: Foundry, 1952, vol. 80, pp. 125-33 and 192-99. 6. L. Katgerman: J. Met., 1982, vol. 34, pp. 46-49. 7. T.W. Clyne and G.J. Davies: Br. Foundryman, 1981, vol. 74, pp. 65-73. 8. D. Eskine, J. Zuidema, Jr., and L. Katgerman: Int. J. Cast Met. Res., 2002, vol. 14, pp. 217-24. 9. N.N. Prokhorov: Russ. Castings Prod., 1962, No. 2, pp. 172-75 10. D.G. Eskin, Suyitno, and L. Katgerman: Progr. Mater. Sci., 2004, in press. 11. I.I. Novikov, G.A. Korol’kov, and A.N. Yakubovich: Russ. Castings Prod., 1971, No. 8, pp. 333-34. 12. G.A. Korol’kov and G.M. Kuznetsov: Sov. Castings Technol., 1990, No. 6, pp. 1-3. 13. M. M’Hamdi, A. Pilipenko, and D. Eskin: AFS Trans., 2003, vol. 111, pp. 333-40. 14. V.I. Dobatkin: Nepreryvnoe lit’e i liteinye svoistva splavov (Continuous Casting and Casting Properties of Alloys), Oborongiz, Moscow, 1948. 15. I.I. Novikov and V.S. Zolotorevskii: Dendritnaya likvatsiya v splavakh (Dendritic Segregation in Alloys), Nauka, Moscow, 1966. 16. W.I. Pumphrey and J.V. Lyons: J. Inst. Met., 1947, vol. 74, pp. 439-55. 17. A.L. Dons: private communication. METALLURGICAL AND MATERIALS TRANSACTIONS A

18. A.E. Vol: Handbook of Binary Metallic Systems. Structure and Properties, Israel Program for Scientific Translations, Jerusalem, 1966, vol. 1. 19. E.C. Ellwood and J.M. Silcock: J. Inst. Met., 1948, vol. 74, pp. 457-67. 20. Modeling for Casting and Solidification Processing, Kuang-O (Oscar) Yu, ed., Marcel Dekker, New York, NY, 2002. 21. J. Timm and H. Warlimont: Z. Metallkd., 1980, vol. 71, pp. 434-37. 22. Y.S. Touloukian, R.K. Kirby, R.E. Taylor, and P.D. Desai: Thermophysical Properties of Matter, vol. 12, Thermal Expansion. Metallic Elements and Alloys, IFI/Plenum, New York, NY, 1975. 23. V.K. Afanas’ev, V.L. Ukhov, and A.N. Solopeko: Izv. Akad. Nauk SSSR, Metally, 1975, No. 5, pp. 189-91. 24. Suyitno, L. Katgerman, and A. Burghardt: Proc. IMECE 2002, 2002, vol. 1.

METALLURGICAL AND MATERIALS TRANSACTIONS A

25. W.M. van Haaften: “Thermophysical Properties of Aluminum Alloys,” Internal Report, Delft University of Technology, Delft, 1997. 26. ASM Handbook. Properties and Selection: Nonferrous Alloys and Special Purpose Materials, ASM INTERNATIONAL, 1997. 27. B. Magnin, L. Maenner, L. Katgerman, and S. Engler: Mater. Sci. Forum, 1996, vols. 217–222, pp. 1209-14. 28. M. Braccini, C.L. Martin, and M. Suery: Proc. MCWASP IX, P.R. Sahm, P.R. Hansen, and J.G. Conley, eds., Shaker Verklag, Aachen, 2000, pp. 18-24. 29. W.M. van Haaften: Ph.D. Thesis, Delft University of Technology, Delft, 2002. 30. J. Sengupta, D. Maijer, M.A. Wells, S.L. Cockroft, and A. Larouche: in Light Metals 2003, P.N. Crepeau, ed., TMS, Warrendale, PA, 2003, pp. 841-47.

VOLUME 35A, APRIL 2004—1335