* Corresponding author: tel fax: address:

Transcription

1 Liquid-liquid phase separation and re-mixing in the Cu-Co system S. Curiotto abc*, N. H. Pryds d, E. Johnson ab, L. Battezzati c a Materials Research Department, Risø National Laboratory, Frederiksborgvej 399, DK-4000 Roskilde, Denmark b Niels Bohr Institute, Nanoscience Centre, University of Copenhagen, Copenhagen, Denmark c Dipartimento di Chimica IFM, Centro di Eccellenza NIS, Università di Torino, Via P. Giuria 9, Torino, Italy d Fuel Cell and Solid State Chemistry Department, Risø National Laboratory, Frederiksborgvej 399, DK-4000 Roskilde, Denmark * Corresponding author: tel fax: address: stefano.curiotto@unito.it (S. Curiotto) Keywords: Cu alloys, metastable miscibility gap, differential scanning calorimetry (DSC) Abstract This paper deals with the metastable liquid-liquid separation in the Cu-Co system. Several samples with different compositions have been investigated by Differential Scanning Calorimetry. High undercooling with respect to the liquidus has been reached by means of the glass fluxing technique. The alloys have been cycled with several heating and cooling runs in order to determine the temperature of the liquid-liquid separation and of the re-mixing. For each composition demixing and re-mixing temperatures have been found to be equal. Nucleation rate calculations of the liquid phase separation have been carried out to explain the experimental results. The liquid-liquid separation in the Cu-Co system has been found to be a nucleation process occurring with no detectable undercooling below the binodal line. 1

2 1. Introduction Most metal elements are completely miscible in the liquid state. However some systems where the elements are characterized by large positive enthalpy of mixing exhibit, below a critical temperature, a separation of the melt in two liquid phases. The critical temperature can be lower than the liquidus temperature: the miscibility gap is then metastable and occurs in the undercooled melt. This is the case for some Cu-based alloys, like Cu-Co. The liquid-liquid separation in this system has been studied for the first time by Nakagawa [1] who measured the change of magnetic susceptibility as a function of temperature and composition; he also examined the microstructure of quenched specimens. He found a miscibility gap nearly symmetrical about the equi-atomic composition and a minimum difference of 90 K between liquidus and critical temperature. In order to avoid contact of the sample with a solid crucible thus obtaining high undercooling with respect to the liquidus, Elder and co-workers [2] investigated the liquid miscibility gap using electromagnetic levitation and monitoring the temperature during the cooling by means of a pyrometer. Measuring the compositions of the phases after the solidification in different samples, they determined the position of the binodal line in the phase diagram. Munitz et al. [3-5] used electromagnetic levitation, splat cooling and electron beam surface melting to study the effects of undercooling and cooling rate on the solidification of Cu-Co alloys at different compositions. They reported demixing temperatures below the peritectic for alloys at the edges of the miscibility gap. Yamauchi and coworkers [6] measured the demixing temperature of some Cu-Co alloys by inserting a thermocouple directly in the melt. The cooling rates obtained were of the order of 300 K/min, high undercooling below the liquidus temperature were achieved. They also found that the degree of undercooling increased with the number of heating and cooling cycles. Another group [7] investigated the Cu-Co system monitoring by a pyrometer the change in temperature as a function of time during continuous cooling. They used a fluxing technique to avoid nucleation and to ensure undercooling below the liquidus line. The authors determined experimentally many points of the liquid miscibility gap, with an uncertainty in the order of 40 K. Cao et al. [8] employed differential 2

3 thermal analysis in combination with glass fluxing to determine with high precision the separation temperature. They showed some solidification microstructures obtained after liquid separation and ascertained that, for a certain composition and at a constant cooling rate of 20 K/minute, the demixing temperature is always the same. Therefore, they suggested that there should be no undercooling of the liquid phase separation. However, no investigation has addressed the re-mixing after the separation in the liquid state in the Cu-Co system. It has been done in the Cu-Fe system [9], which displays a metastable miscibility gap in the liquid phase like Cu-Co, but at higher temperature. In their work Wilde and Perepezko found undercooling of the demixing reaction for compositions at the edges of the miscibility gap and no undercooling at the maximum. They explained the results supposing a difference in the wetting behaviour of the liquids with the glass flux according to the composition of the alloys. Recently we have carried out two studies on the solidification of Cu-Co alloys [10,11] under different conditions. The first deals with the solidification of a Cu 58 Co 42 alloy in a wedge shaped mould. The observed microstructures have been related to the cooling rates achieved. An explanation of the coagulation mechanism of the droplets of the same phase has been proposed. The second paper investigates the effect of rapid solidification on the formation of microstructures. It is further suggested that, at very high cooling rates, the most likely demixing process is spinodal decomposition. Based on these two papers, the main aim of the present work is to determine the demixing-mixing mechanism in the Cu-Co system. The liquid phase separation has been studied in details by Differential Scanning Calorimetry (DSC) and microstructure observations. In particular, the melt has been cycled with several heating and cooling runs around the demixing temperature while avoiding the solidification of the Co-rich phase, in order to detect and compare the demixing and re-mixing signal. 3

4 2. Experimental details High purity copper (99.999%) and cobalt (99.9%) have been pre-alloyed in the desired proportion to attain different compositions in an arc-melting furnace (table 1). The weight of the pure elements was checked with a precision of 0.1 mg. Before melting the chamber has been evacuated and purged several times with high purity Ar and using lumps of Zr as getter. Each sample had a weight between 150 and 250 mg. In order to check for any loss of the elements (e.g. by evaporation) and consequent deviation from the chosen composition, after the melting the alloys have been weighed. The loss of mass has always been less than 0.3 mg, i.e. well below 1%. To avoid oxidation, each specimen has been inserted in an alumina crucible with about 50 mg of crushed Duran glass as fluxing agent. Duran has a glass transition at 798 K and is therefore liquid below the Cu-Co peritectic temperature (occurring at 1382 K): it covers the surface of the sample and avoids contact between the alloy and the atmosphere. Besides, borate and silicate glasses can dissolve metal oxides and other impurities; so, when the encapsulated sample is molten, the alloy is cleaned from possible heterogeneous nucleants. The DSC experiments have been carried out by means of a Netzsch DSC 404 C instrument calibrated with the melting temperatures of pure In, Zn, Ag, Au and the transition of pure Fe. For each composition, one to four samples have been processed several times, with heating and cooling rates ranging between 10 and 40 K/min. In order to allow better homogenization of the alloy in the liquid state, an isothermal step of 5 minutes has been conducted at the end of each heating run above the liquidus temperature; this temperature depends on the composition of the specimen. After the DSC experiments, the solidified samples have been mounted in resin, cross-sectioned and polished (without etching) to be observed by Scanning Electron Microscopy (SEM) in backscattering mode. 3. Results In Fig. 1 a DSC curve for a Cu 25 Co 75 alloy showing the various transitions is given as an example of the full processing cycle. In the insert of Fig. 1, the calculated Cu-Co phase diagram 4

5 [12] is shown; the transformations of the alloy during the DSC experiment can be followed along the vertical line at composition x Cu (molar fraction)=0.25. On heating, the first endothermic peak in the DSC trace at 1382 K is due to the peritectic transformation. The second signal, with peak at 1720 K, corresponds to the completion of the melting of the sample. After homogenization at a temperature above the melting, the alloy was cooled and the transformations during cooling are shown in the upper curve. The first small exothermic deviation from the baseline at 1493 K is due to the demixing between a Cu-rich and a Co-rich liquid. At 1395 K, the intense peak signals the solidification of the Co-rich fcc phase. The subsequent composite peak with maximum at 1364 K is due to the solidification of the Cu-rich phase starting with the peritectic transformation. The magnetic transition provides a small signal in both heating and cooling runs, it occurs at 1315±2 K. It has never been possible to undercool the liquid below the peritectic, as achieved by means of rapid solidification techniques in [5] Peritectic reaction The peritectic peak is shown in Fig. 2, curve a. Occasionally, a shoulder before the main peak is observed (Fig. 2, curve b). The onset temperature of the shoulder varies between 1360 and 1370 K. This behaviour is mainly displayed after cooling where demixing occurred, as verified by DSC cycling a Cu 25 Co 75 alloy many times around the peritectic before and after demixing. During cooling, when demixing occurs, there is in fact a sharp liquid-liquid separation, as explained below, because of coalescence of droplets of the minority phase [10]. Then the Co-rich phase solidifies, retaining a spherical shape (see section 3.5), at higher temperatures with respect to the peritectic. In the inner area of the solid Co-rich sphere, far from the interface with the Cu-rich liquid, the diffusion and the rearrangement of the composition according to the phase diagram are slow. Thus, as explained in [13], the peritectic transformation does not go to completion. The Cu-rich liquid does not solidify completely at the peritectic and, below that temperature, follows the liquidus line while precipitating Cu-rich solid. Crystals of the last solidified Cu-rich phase are richer in Cu with 5

6 respect to the equilibrium composition. Therefore, during heating, they start melting before the expected peritectic, resulting in the shoulder of the DSC signal. However the main peritectic signal occurs at the same temperature for all the compositions. Taking the onset of the transformation for every alloy as shown in Fig. 2, the average value has been found to be 1382±3 K Liquidus temperatures The dependence of the liquidus temperature on the heating rate was also investigated. The melting peaks appear shifted to higher temperature with increasing the heating rate. This wellknown effect is due to the thermal inertia of the instrument. For each composition and each heating rate the temperature of the crossing point between the tangents at the end of the liquidus peak and at the baseline has been taken. The liquidus temperature for a defined composition has then been obtained by extrapolating the measured data to 0 K/min by means of a linear fit. The procedure is explained in [14]. The error of the liquidus temperatures has been considered to be equal to the standard deviation of the fit at the intersection with the y axis. For every composition it has always been less than ±5 K. The liquidus temperature for each composition has been reported in table Demixing and mixing High undercoolings, with a maximum value of 303 K, have been obtained due to the glass flux. The extent of the value of the undercooling was random and did not depend on the number of cycles, as reported by Yamauchi et al. [6]. The demixing signal observed in the cooling curves is shown in Fig. 3 for different compositions. The onset temperature of demixing for a certain sample has always been found to occur at the same temperature in different DSC runs and does not depend on the cooling rate. The values are given in table 1. Finding the liquid separation in the sample with composition Cu 17 Co 83 has been more difficult than for all the other alloys, it occurred only once. In fact its demixing temperature is very low, 1419 K, while the liquidus temperature is 1708 K, with a difference between them of 289 K: often the undercooling below the liquidus temperature was not 6

7 sufficient and the Co-rich phase solidified before the liquid separation. The demixing signal of Cu 25 Co 75 is shown in Fig. 4 for different cooling rates. In the first cooling (full line) the signal is sharp and continuous. The second (dashes) and third (dots) runs correspond to cooling rates of 30 and 20 K/min respectively; the demixing is not sharp and it is difficult to determine a definite temperature. The fourth cooling (dash-dot, 10 K/min) presents two definite onsets, the first at the same temperature of the first run, while the second, sharper, is 10 K lower. Furthermore, from the second run a maximum in the demixing signal is evident. The same effect has been observed in Cu 20 Co 80. In order to detect the re-mixing temperature, some samples have been cooled about 20 K below the demixing temperature and then reheated. Finding the re-mixing temperature has not always been possible: in some cases the Co-rich fcc phase solidified before reaching it. Fig. 5 shows the signals of demixing and re-mixing for Cu 25 Co 75 (dashes), Cu 58.5 Co 41.5 (composition corresponding to the maximum of the calculated miscibility gap [12], full line) and Cu 80 Co 20 (dots). The full circle indicates the temperature of the programmed beginning of the heating. In all the experiments, the mixing signal ends always at higher temperature with respect to the demixing. Such temperature increases with the heating rates, but the signal is often very noisy, so it was not always possible to find a definite temperature of the re-mixing. Therefore the samples have been processed by DSC with a different cooling/heating profile and the results are shown in Fig. 6. The samples have been cooled around 20 K below the demixing temperature (curve a, full line) and reheated a few degrees (2 K) above the demixing temperature (dashes, curve b); then an isothermal step of 10 minutes has been carried out (c), followed by continuous heating (d) to the liquidus temperature. These cycles have been carried out on three compositions: Cu 25 Co 75, Cu 58.5 Co 41.5 and Cu 80 Co 20. After the instrumental transient (curve indicated as b), in the heating trace following the isothermal step no endothermic signal of mixing is seen, contrary to the finding shown in Fig. 5 for continuous heating experiments. 7

8 3.4. Solidus temperature in Cu 17 Co 83 Fig. 7 shows a DSC heating run for the composition Cu 17 Co 83. During heating, after the first endothermic peak corresponding to the peritectic, a very small exothermic signal is observed in the DSC trace (enlarged in the insert 2 of Fig. 7). From the phase diagram shown in Fig. 8, it is inferred that the offset of the small peak indicates the end of the solidification of the alloy during heating, with average measured temperature equal to 1503±11 K. The vertical dashed line shows the path of the alloy while the temperature increases. At low temperature the sample is solid; at the peritectic there is equilibrium between Co-rich, Cu-rich and Cu-rich liquid. Above the peritectic the Curich phase dissolves in Co. When the temperature increases, the amount of liquid phase diminishes and the Co-rich solid increases. The solidification ends in the point (a), where the dashed line crosses the solidus curve; between (a) and (b) the alloy presents a homogeneous solid phase, then the melting of the fcc phase starts in the point (b) where the dashed line crosses the solidus curve at high temperature. Above the liquidus temperature the sample is completely molten. In the DSC heating trace of Fig. 7 the onset of the melting of the Co-rich phase is indicated by an arrow, the average is 1647±8 K. In Fig. 8 the experimental points corresponding to (a) and (b) are indicated by full circles. The existence of a homogeneous fcc phase above the peritectic supports the retrograde solubility in the Cu-Co system, as calculated in [12] at variance with [15] Microstructures In Fig. 9 the cross-sections of three alloys processed by DSC are shown. The images are taken by SEM with a backscattered electrons detector, so the bright regions are Cu-rich and the dark are Co-rich. For all the compositions the Co-rich phase is mostly of spherical shape and the surrounding Cu-rich phase contains Co-rich dendrites. The image of Cu 75 Co 25 (Fig. 9a) displays several Co-rich droplets with a size varying between ~ 5 and 1000 m. In the insert figure Co-rich droplets are seen close to coagulating with a bigger particle in order to reduce the interface area with the Cu-rich phase. In this Cu-rich alloy, small Co-rich particles formed during the demixing in 8

9 the entire sample, quite far from each other and they then coagulated to form bigger spheres. The time before solidification was probably not enough to allow complete coagulation of the Co-rich spheres. The microstructure of Cu 60 Co 40 is shown in Fig. 9b. In this case a large Co-rich sphere appears encased in a Cu-rich matrix. The composition of the alloy is slightly richer in Cu with respect to the maximum of the miscibility gap. However, since the Co percentage is high, the number of Co-rich droplets per unit volume is also high and they can coagulate in only one big sphere. Cu-rich sphere tend to escape from it and few small Co-rich particles approach the large sphere. Here also, solidification occurred before complete coagulation. In Cu 25 Co 75 a large Co-rich sphere is completely surrounded by a thin layer of Cu-rich phase (see Fig. 9c). Moreover, small Curich droplets seem to migrate out of the Co-rich sphere, toward the surface of the sample, as shown in the insert of Fig. 9c. In some Cu-rich samples, after DSC processing with liquid separation, it is sometimes possible to see a large Co-rich sphere composed of many small droplets still separated, as shown in Fig. 10 for Cu 75 Co 25. The droplets have a dimension of about 5-10 m. 4. Discussion The peritectic, liquidus and demixing temperatures found experimentally for the compositions investigated in this work are shown in Fig. 8 and compared with the phase diagram [12] and with the set of data reported in [8]. Our data points are in good agreement with the previously measured ones and with the calculated equilibria. As reported in section 3.3 two onsets have been found in the demixing of Cu 25 Co 75 and Cu 20 Co 80. They have already been observed but not completely explained in Cu 18.8 Co 81.2 by Cao et al. [8]. Also in the present work the reason for the two onsets is not completely understood, however a tentative explanation is given. In Co-rich alloys the Cu-rich phase constitutes a thin layer around the large Co-rich sphere (see Fig. 9c). As it is always the minority phase which separates from the unmixed liquid [10], it is likely that the demixing started with heterogeneous nucleation at 9

10 the interface between alloy and glass with the formation of the Cu-rich liquid which has lower interface energy with the glass [8]. The result of this process is the first onset in the demixing in the DSC experiments. In regions far from the surface, in order to reach the equilibrium described by the binodal, separation of the Cu-rich liquid must take place as well. This occurs at slightly lower temperature, with undercooling of the not separated liquid below the binodal (second onset in the demixing signal). The sharpness of the second onset in Fig. 4 (dash-dot trace) supports this hypothesis. Droplets can therefore separate in the middle of the sample, as shown in the insert of Fig. 9c, and then move to the Cu-rich external layer. In the first cooling run, the signal never displays the second onset, perhaps because the sample touches the alumina crucible at the bottom. The liquid-liquid separation can occur continuously at the interface with the alumina crucible without formation of a Cu-rich layer. Meanwhile the Cu-rich liquid separation has started also at the interface with the glass. Then, in order to minimize interfacial and surface energy, the Cu-rich liquid moves to cover completely the surface of the sample, dragging also a thin glass layer. A certain time is necessary to complete this process. In the following run the sample is then completely surrounded by the glass and the separation mechanism goes on as described above. In the DSC signal of demixing of Cu-rich alloys it is not possible to see any change of slope because Co-rich droplets nucleate at the interface with the glass but then move to the interior to allow the Cu-rich liquid to reach the surface. The variation of compositions of the two separated liquids is therefore continuous inside the bulk: there is not a uniform Cu-rich region where a new demixing process can start. The Co-rich particles coagulate to reduce interface energy, but sometimes the coagulation is not complete and the resulting microstructure is that shown in Fig. 10. This microstructure is not completely understood. Presumably the droplets continuously coagulate while the temperature drops, until the complete solidification of the sample take place. During this time each single Corich sphere continues to separate a Cu-rich liquid which is segregated to the surface of the small droplet, delaying the complete coagulation. Then the solidification freezes the microstructure. 10

11 In order to check for the nature of the mixing effect shown in Fig. 5, the DSC experiment illustrated in Fig. 6 was performed. Here the signal associated with the mixing is not visible in the presented curves because the energy uptake due to this process is very small and spread out in the long time during the isothermal annealing. Since the liquids mix isothermally at a temperature very close to that found for demixing on cooling, it is concluded that the difference in temperature between demixing and re-mixing is so small that it cannot be detected in the DSC signal. The signal obtained in continuous heating at a higher temperature than that of demixing is then due to a kinetic effect. In fact, during liquid separation on cooling, there is coagulation of the two liquids in an inner Co-rich region and an external Cu-rich volume; heating the alloy, a certain time is necessary to reach full mixing of the liquids again. The mixing signal is then dragged to higher temperatures. Noisy signals appear in the DSC traces (see Fig. 5) likely due to presence of spheres with different sizes, which will dissolve in different times. In the Cu-Co system it has been reported that the difference between binodal and spinodal in the liquid-liquid separation varies between 0 (at the maximum of the miscibility gap) and more than 100 K (at the edges) [12], as predicted by the phase diagram shown in Fig. 8. On heating, the two liquids must mix at the binodal temperature. It has been shown here that demixing and re-mixing occur at the same temperature also for compositions at the edges of the miscibility gap. Therefore it is possible to conclude that the demixing occurs at the binodal: at the edges of the miscibility gap the separation cannot be spinodal, as suggested in [6], otherwise the demixing should take place at lower temperature with respect to the re-mixing. The demixing mechanism is then nucleation and growth with no detectable undercooling below the binodal. The prediction given by Predel et al. [16] is therefore confirmed. Only at the maximum of the miscibility gap the liquid phase separation might occur by spinodal decomposition. These results are different from those obtained for Cu-Fe alloys by Wilde and Perepezko [9]. In order to support the experimental results, the nucleation rates for the nucleation of Curich liquid phase and Co-rich solid phase in a Cu 25 Co 75 alloy have been calculated. The aim was to verify if undercooling of the liquid separation is possible and compare it with the undercooling 11

12 necessary for the nucleation of the Co-rich solid phase. The steady-state nucleation rate (I s ) for a condensed system can be written [17] as: 1/ 2 2 cos 1 cos 2 / Dn* N G a G * n I exp 2 6 * s (1) k BTn k BT with Gn * 2 3G v (2) Here D is a diffusion coefficient taken as that of Cu in liquid Cu for the liquid-liquid separation: H D D D0 exp ; as regards the solidification of the Co-rich phase, D has been RT calculated by means of the Stokes-Einstein law: H k BT D, where d 0 exp is the Co 6 RT 2 viscosity of liquid cobalt. n* is the critical cluster size: n* v G v 3 with v atomic volume; is the atomic jump distance, considered equal to the average atomic diameter, d a = x Cu d Cu +x Co d Co = Å (with x Cu =0.25, x Co =0.75); G is the difference in Gibbs free energy of the new phase and of the initial phase, per atom; N a is Avogadro constant; k B is the Boltzmann constant; R is the gas constant; T is the temperature; G v is the free energy difference between the nucleus of the new phase and the unmixed liquid, calculated using the description of G for liquid and fcc phases, as assessed in [12]; is the wetting angle between the newly formed phase and the glass (for the phase separation) or with the Cu-rich phase (for the solidification); is the interface energy between nucleus of the new phase and liquid phase. The interfacial energy between Co and Cu rich liquids is taken from [18]. The interfacial energy between fcc Co-rich phase and mixed liquid has been approximated to that between pure Co solid and pure Co liquid. It has been calculated according to Turnbull s approximation: H solidlquid 0.43 (3) N V 1/ 3 a m 2 / 3 m 12

13 WhereH m is the enthalpy of fusion and V m is the molar volume. solid-liquid has been calculated to be J m -2. Values, units and references for the different quantities are given in table 2. The variation of nucleation rates with temperature for the liquid-liquid separation and the Co-rich fcc phase solidification in a Cu 25 Co 75 alloy are shown in Fig. 11. Since the wetting angle between the various phases is not known, three curves have been calculated for heterogeneous nucleation with 10, 45, 90degrees and for the homogeneous nucleation rate i.e. =180 degrees. The heterogeneous I s curves show that the solidification of the Co-rich fcc phase is strictly dependent on the wetting angle. In our DSC experiments very often the solidification occurred at 1400 K. As our samples have a volume of around 10 mm 3, the I s needed to have at least one nucleus per second is The nearest curve to these values of temperature and nucleation frequency is the homogeneous one. Therefore the nucleation of the solid phase could have been homogeneous or heterogeneous with high wetting angle at the interface with the Cu-rich liquid phase. According to the calculation for the liquid-liquid separation in Cu 25 Co 75, a few degrees below the demixing temperature, the nucleation rate of the Cu-rich liquid sharply increases to high values. This result is due to the very low interface energy between the liquid phases. Considering heterogeneous nucleation with wetting angles lower than or equal to 45 degrees, the undercooling necessary for the demixing is much less than 1 K. Homogeneous nucleation of the phase separation should occur with an undercooling of around 20 K. These calculations support the DSC experiments showing no undercooling of the demixing. Moreover they suggest that the nucleation of the liquid-liquid separation is heterogeneous and that the wetting angle between Cu-rich liquid and glass is low. 5. Conclusions DSC experiments in the whole range of compositions of the Cu-Co system have been carried out. Investigations have been focused on the metastable liquid miscibility gap. High undercooling, up to 300 K, has been reached by means of the glass fluxing technique. At fixed 13

14 composition, the temperature of the liquid phase separation for different cooling runs is the same and does not change with the cooling rate. The samples have been cycled in the liquid state with heating, isothermal and cooling runs around the demixing temperature: the mixing temperature obtained on heating is equal to the demixing temperature. These experiments and nucleation rate calculations support the conclusion that the liquid-liquid separation in the Cu-Co system, for composition at the edges of the miscibility gap, is nucleation controlled and occurs without detectable undercooling below the binodal line. 6. Acknowledgments The work has been supported by the European Space Agency within the project CoolCop (ESA-MAP AO ) 14

15 Reference List [1] Y. Nakagawa: Acta Metallurgica, 1958, vol. 6, pp [2] Elder S.P., Munitz A., Abbaschian G.J.: Mater.Sci.Forum, 1989, vol. 50, pp [3] A. Munitz and R. Abbaschian: J.Mater.Sci., 1998, vol. 33, pp [4] A. Munitz and R. Abbaschian: Metallurgical and Materials Transactions A-Physical Metallurgy and Materials Science, 1996, vol. 27, pp [5] A. Munitz, S.P. Elderrandall, R. Abbaschian: Metallurgical Transactions A-Physical Metallurgy and Materials Science, 1992, vol. 23, pp [6] I. Yamauchi, N. Ueno, M. Shimaoka, I. Ohnaka: J.Mater.Sci., 1998, vol. 33, pp [7] M.B. Robinson, D. Li, T.J. Rathz, G. Williams: J.Mater.Sci., 1999, vol. 34, pp [8] C.D. Cao, G.P. Gorler, D.M. Herlach, B. Wei: Materials Science and Engineering A- Structural Materials Properties Microstructure and Processing, 2002, vol. 325, pp [9] G. Wilde and J.H. Perepezko: Acta Materialia, 1999, vol. 47, pp [10] S. Curiotto, N.H. Pryds, E. Johnson, L. Battezzati: in press on Materials Science and Engineering: A, 2006, presented at the 12 th International Conference on Rapidly Quenched and Metastable Materials [11] L. Battezzati, S. Curiotto, E. Johnson, N.H. Pryds, in press on Materials Science and Engineering: A, 2006, presented at the 12 th International Conference on Rapidly Quenched and Metastable Materials 15

16 [12] M. Palumbo, S. Curiotto, L. Battezzati: Calphad-Computer Coupling of Phase Diagrams and Thermochemistry, 2006, vol. 30, pp [13] D.A. Porter and K.E. Easterling, Phase Transformations in Metals and Alloys, Chapman & Hall, London, 1992, 231- [14] R.I. Wu and J.H. Perepezko: Metallurgical and Materials Transactions A-Physical Metallurgy and Materials Science, 2000, vol. 31, pp [15] J. Kubista and J. Vrest'al: Journal of Phase Equilibria, 2000, vol. 21, pp [16] B. Predel, L. Ratke, H. Fredriksson: Fluid Science and Materials Science in Space: a european perspective, H.U. Walter editions, Springer-Verlag, Berlin 1987, pp [17] K.F. Kelton: Solid State Physics, H.Ehrenreich and D.Turnbull eds., Academic Press, New York, 1991, vol. 45, p.75 [18] J. Brillo, I. Egry, D. Herlach, L. Ratke, M. Kolbe, D. Chatain, N. Tinet, C. Antion, L. Battezzati, S. Curiotto, E. Johnson, N.H. Pryds, presented at ELGRA Biennial Symposium and General Assembly Santorini, Greece: September 21st-23rd, 2005, to appear in Microgravity Science and Technology [19] T. Iida and R.I.L. Guthrie: The physical Properties of liquid Metals, Clarendon Press, Oxford, 1988 [20] M. Shimoji and T. Itami: Atomic Transport in Liquid Metals, Trans Tech Publications, Aedermannsdorf, CH, 1986 [21] Table of interatomic distances and configuration in molecules and ions, L.E. Sutton editions, Supplement , Special publication No.18, Chemical Society, London, UK,

17 Figures and tables captions Fig. 1. a) DSC signals of a Cu 25 Co 75 alloy on heating and cooling (see arrows). Heating and cooling rate: 40 K/min. b) Cu-Co phase diagram. Fig. 2. Peritectic peaks on heating; a: sample solidified without previous demixing in the DSC, the DSC signal displays only the peritectic onset at 1380 K; b: sample solidified after demixing in the DSC, showing two onsets of the peak, a pre-peritectic at 1369 K and the peritectic at 1380 K. Fig. 3. DSC cooling curves for alloys of different compositions. The lines are truncated above the solidification temperature of the Co-rich phase. The numbers on the lines are the Cu at.% in the sample. Fig. 4. Demixing signal for Cu 25 Co 75 at different cooling rates. At low cooling rate there is a sharp second onset. Fig. 5. DSC mixing signals of alloys at different compositions for continuous heating; (a): cooling curves at 30 K/min; (b): heating curves at 30 K/min. : Cu 25 Co 75 ; : Cu 58.5 Co 41.5 ; : Cu 80 Co 20. The black circles mark the programmed temperature for the end of cooling and the start of heating. The mixing signal during heating is complex except for Cu 80 Co 20. Fig. 6. DSC signals of alloys at three different composition: Cu 25 Co 75, Cu 58.5 Co 41.5, Cu 80 Co 20. After the demixing the samples have been heated and held few degrees above the demixing temperature with an isothermal step; letters indicate the DSC step: (a) 30K/min, (b) 30 K/min, (c) isothermal anneal 15 min, (d) 30 K/min. 17

18 Fig. 7. DSC heating curve of Cu 17 Co 83 at 10 K/min. The peak due to the solidification of the Corich phase is enlarged in the insert; the onset of the melting (b) is indicated by an arrow. Fig. 8. Lines: Cu-Co phase diagram as reported in [12]. Points: experimental temperatures for peritectic, miscibility gap, solidus and liquidus. This work: : liquidus temperatures; : demixing temperatures; : peritectic temperatures; : solidus temperatures. Measured by Cao et al. [8]: : demixing temperatures; : liquidus temperatures. The vertical dashed line is a guide for the eye at composition Cu 17 Co 83. Fig. 9. SEM images obtained by backscattered electrons. The micrographs show the microstructure of samples at different compositions processed by DSC. Dark regions are Co-rich; bright regions are Cu-rich. Every sample experienced liquid phase separation: Co-rich spheres are inside a Cu-rich phase. (a): Cu 75 Co 25 ; (b): Cu 40 Co 60 ; (c): Cu 25 Co 75 Fig. 10. SEM image obtained by backscattered electrons of a Co-rich sphere in Cu 75 Co 25 processed by DSC. Dark regions are Co-rich; bright regions are Cu-rich. The sample experienced liquid phase separation. The sphere is constituted by small Co-rich droplets separated one from the other by thin boundary layer of Cu. Fig. 11. Nucleation rate calculations for a Cu 75 Co 25 alloy. Thick lines: fcc Co-rich solidification; thin lines: liquid-liquid separation. : homogeneous nucleation rate; : heterogeneous nucleation rate, wetting angle 90 degrees; : heterogeneous nucleation rate, wetting angle 45 degrees; : heterogeneous nucleation rate, wetting angle 10 degrees. Table 1. Demixing and liquidus temperatures measured by DSC. The liquidus temperatures have been calculated with an extrapolation to nil rate of the measured offsets of the liquidus peaks at 18

19 different heating rates. The error has been taken equal to the standard deviation of the linear fit at the intersection with the temperature axis. Table 2. Values, units and references of the different quantities used in the nucleation rate calculations. The driving forces have been calculated in the temperature range K. 19