Institute of Metals Division - Analysis of Molten-Zone Refining

The process of molten-zone refining is analyzed for long ingots and many zone passages. Formulas are derived which give the resultant impurity distribution in terms of finite series. A comparison with the approximate procedure of Hamming is given. HE physical principles and applications of an extremely physicalprinciple efficient form of metallurgical refinement has been described by Pfann. The purpose of the present paper is to describe a method of analyzing exactly the particular program used which enables the segregation effect to be predicted for any number of molten-zone passages in a long ingot. The method is applied to the particular case of refinement of an ingot whose impurity initially is uniformly distributed throughout its length. A number of molten zones of equal length are passed through the ingot effecting a radical redistribution of impurity. Pfann has indicated an approximate method, due to R. W. Hamming, of calculating the resultant concentration after each successive zone pass for a particular value of the segregation constant defined in his paper. Here a solution will be presented in terms of the number of zone passes and the segregation constant. The expression, though cumbersome, is exact and susceptible to ordinary numerical computation procedures. The results of a similar computation using the procedure of Hamming are presented in a table together with the exact results of the present method. The discrepancy in terms of absolute concentrations is tabulated for the first eight zone-lengths. To establish the notation (which follows that of Pfann1 as closely as possible) and physical basis of the analytical equations, the physical model and principal assumptions may be reviewed. An alloy of two elements, where there is formed a continuous range of solid solutions, usually does not melt as a simple compound. Rather, a temperature is reached where the solid solution is in heterogeneous phase equilibrium with a liquid solution of different composition. The temperature dependence of these equilibrium compositions forms part of the phase diagram. For very small concentrations of a solute B in a solvent A, this usually takes the form of Fig. 1. Sometimes the solidus and liquidus slope upward. This corresponds to a segregation constant (defined below) which is greater than unity. The segregation constant is now defined as k = Cn(x)/CnI(x) [1] where C,,(x) is the impurity concentration in the solid ingot at distance x during the nth passing of a molten zone and Cnl (x) is the impurity concentration of the liquid zone from which the solid at dis- tance x is formed (see Fig. 4 of ref. 1.) C (x) remains the same after passage of the zone. The constant k may be either greater or less than unity in general. Purification in the former case is effected only in a finite ingot and in the portion that is melted last. For k less than unity purification is effected even in an infinite ingot. The method which follows gives, in the former case, the successive increases in impurity concentration and, in the latter case, the successive decreases in concentration. The general case of impurity redistribution will be considered first, and purification will be discussed later on. The analysis rests on the following assumption: The movement of the zone is too rapid to allow appreciable atomic rearrangement in the solid sections and too slow to disturb the uniform impurity distribution in the liquid zone characteristic of equilibrium. Hence, the composition in the solid at the left solidifying interface will be determined by Eq. 1 while the impurity concentration of the liquid zone will be uniform throughout its length. The reasoning which follows closely parallels that of Appendix 11 in Pfann&apos;s paper. It is reviewed here for the case of the nth zone pass in order to make clear the meaning of an operator essential to the present method. Fig. 4 of ref. 1 shows the movement of a molten zone of length 1 in an ingot of total length d. Each Cn(x) can be determined from the condition that the amount of solute added to the zone during an incremental advance, dx, is due to the melting in of a solid portion C(x)dx and the freezing out of kCnl(x), that is d I —r- CnL (x) dx = Cn-1 (x+l)- kCnL (x) dx or, in terms of Cn(x) d k k —— C,(x) +—c.(x) =— Cn-1 (x + l). [2] dx l l This, of course, is derived from the main assumption, the fact that 1 is constant, and that the total impurity content previously present up to x + 1 is constant . A correction has to be made for the region (d — nl) < x < d. This is due to the zone length changing during the passage of the solidifying interface beyond x = d — 1. Since the general solution would be too complicated otherwise, only the region 0 < x < d — nl is considered. The general solution of Eq. 2 is