3.1. Solidification Process

Figure 2 illustrates the measured cooling profiles of three Cu-15 wt.% Fe samples at different undercooling states. On the curve (a), there was a recalescence corresponding to the growth of primary γ-Fe dendrites (at about 1543 K) and then led to thermal arrest, which determined the liquidus temperature, T_L ≈ 1578 K. The measured T_L was consistent with the value in the equilibrium phase diagram [2], thus confirming the present temperature calibration. Moreover, it is also confirmed in Figure 3, since γ-Fe dendrites are clearly visible. For all the curves in Figure 2, after further cooling, the peritectic reaction can be observed at T_P (at about 1369 K). At the higher undercooling, additional anomalies appear before the major recalescence, as marked by T_S in Figure 2(b). This unusual phenomenon neither emerged at low undercooling, nor arose from the flux agent due to the low-transition temperature (lower than 800 K) of the glass. Accordingly, it can be deduced that the anomalous behavior on the curve (b) can be attributed to a precursor reaction [12], namely, a phase separation in the sufficiently undercooled melts. The angular point on the cooling curve before the recalescence onset is hereafter referred to as T_S, the temperature at which the separation process into two liquid phases L1 (Fe-rich) and L2 (Cu-rich) begins. As the undercooling increases, liquid-phase separation and recalescence are more obvious, as shown in Figure 2(c).

3.2. Microstructural Selection

Figure 4(a) shows the liquid-phase separation microstructure of Cu-15 wt.% Fe at the undercooling of 75 K. The lever rule and the metastable miscibility gap of Cu-Fe phase diagrams (Figure 1) indicate that the Cu-15 wt.% Fe melt at the undercooling of 75 K is separated into a Fe-rich melt (L1) and a Cu-rich melt (L2). Obviously, the spheres (Figure 4(a)) are the L1 (Fe-rich) minority phase [13]. Furthermore, there are a few of γ-Fe dendrites formed due to segregation of the L2 (Cu-rich) phase [14].

During the further cooling process, the solute of one phase in another phase tends to be supersaturated. It should be noted that the solute is a relative concept. Cu in the L1 (Fe-rich) phase is a solute, and Fe in the L2 (Cu-rich) phase is a solute. In general, the composition of each liquid is expected to follow the miscibility phase boundary if adequate time is allowed for the complete transfer of atoms between the two phases [16]. At the same time, within a metastable miscibility gap of salient nonequilibrium features, the viscosity rises and diffusion coefficient declines with the increase in undercooling [12]. In these circumstances, complete diffusion was absent, and secondary separation (even multiphase separation) in liquid might occur. An example is shown in Figure 4(b).

3.3. Microstructural Evolution

The microstructures at different undercooling states were observed at the same magnification (Figure 5). The droplet size increases with the increase in the undercooling because the higher the undercooling is, the larger the interval ∆t from T_S to T_N is, that is, the longer coarsening time. By assuming that the largest droplet radius, r_m, can be used to quantify the extent of droplets coarsening [12], a relationship between r_m and ΔT is shown in Figure 6. Corresponding to two-characteristic undercooling states, that is, ΔT_c1 = 103 K and ΔT_c2 = 142 K, the increase of r_m with ΔT can be divided into three regimes: small r_m accompanying with a slow increasing rate (i.e., dr_m/dΔT) for ΔT < ΔT_c1, and medium r_m with a steep dr_m/dΔT for ΔT_c1 < ΔT < ΔT_c2. This result is consistent with the previous analysis result of undercooled Cu-Co melt [12]. However, in the later stage of liquid-phase separation, there is a slight deviation between the results in this study and previous results [2, 12] because the maximum undercooling achieved in the study is not large enough.

As the undercooling continues to increase, the size of droplets increases significantly and the effect of Stokes movement gradually becomes prominent. According to (4) and (5) when the droplet size is larger, V_S is much larger than V_M, and Ostwald ripening and Brownian motions are significantly weakened. Therefore, Stokes movement plays a dominant role in the midliquid-phase separation. Furthermore, a higher undercooling tends to facilitate Stokes movement [23]. Migration and coagulation of droplets are intensified significantly due to the gradually enhanced Stokes movement, thus resulting in the rapid increase in the droplet size. In turn, the progressively larger droplet sizes further promote Stokes movement, as shown in Figure 7. The two coagulating spheres are not in the same color because of differences in the composition between two spheres in Figure 7(b). As mentioned above, the larger sphere with the darker color is the Fe-rich sphere, and the smaller sphere with the lighter color is the Cu-rich sphere. It is assumed that the main coagulation mechanism of the two spheres is the Ostwald ripening, but the smaller sphere with the low content of Fe coagulates with the larger sphere with the high content of Fe. The above mechanism is contrary to the Ostwald ripening mechanism due to the concentration gradient. The result confirms the above conclusion that the Stokes movement is the main mechanism. Consequently, the droplet size presents the linear growth with the increase in the undercooling in the medium stage of the liquid-phase separation. The result is consistent with that shown in Figure 6.