Mb. R . C h a d w i c k ,f M.A. (M em ber): I have always felt that the literature dealing with inverse segregation describes two distinct types o f phenomenon, and these the author fails to differentiate clearly in this review.
The first o f these types I would typify by the kind o f inverse surface structure found in chill-east ingots o f the tin bronzes and Duralumin. Here the ingot commonly has a surface layer o f low melting point constituent almost sharply differentiated from the mass o f the ingot. Such an ingot may be of quite fine equiaxed structure, the body o f the ingot after removal o f a thin surface layer showing very little variation in composition. The ingot during solidification reaches a stage at which it consists of primary crystals, wetted by,
t Research Metallurgist, I.C.I. Metals, Ltd., Birmingham.
and embedded in, a continuous network o f low melting Uf m it has been shown that this constituent flows outwards a^ . aohdliie1s,“ ” the outermost layer of the ingot; to some extent g a s e v o lu t io n would appear to be responsible" for this, for gas cavities, resulting from the displacement o f the low meRing point network, may be clearly seen in the cast bronzes and Duralumin, whilst the considerable diminution m the extent o f segregation by reduction o f the gas content o f the molten alloy has been f% , second type I would describe as inversion throughout the whole ingot
thickness, probably in conjunction with a columnar type o f structure. Un
fortunately investigators have devoted little attention to macrostruc , in many cases detailed observations on density, composition, microstructure,
&c , have been made, whilst the macrostructure o f ingots has not been described. In the case o f Rosenhain’ s experiment * on a 12 per cent. c°ppei 88 per cent, aluminium allov, it would appear that such a^ columnar structure w as present, as otherwise it is difficult to see how the sharp liq u id -soM division occurred in the freezing mass. It would appear to be o f considerable interest, therefore, to show whether primary columnar crystals themselves show an inversion in composition, the outer ends being o f lower melting point than the inner ends. Watson may have been dealing with such an inversion, bu one cannot deduce from the data given whether the primary dendrites formed portions of long columnar crystals. Watson attempted to explain su ci an inversion, but could give no scientific reason for the migration o f his primaries, I have, however, previously put forward a theory J (unnoted y the author o f this review) which explained the migration o f p rim a ries described by Watson, bv assuming that the primary crystals grew only at the side adjacent to the mould wall, and were therefore pushed inward by the mechanical force o f the growing crystals. Such a mechanism requires o n y the assumption o f a liquid layer wetting the outside o f the solidifying ing ■, whilst a further characteristic would be a very small temperature gradient in the solidifying ingot. The formation o f beads o f low melting point con
stituent on the surface layer o f prematurely-stripped ingots o f bronze and Duralumin have been frequently described, with chill moulds one can fre
quently observe a liquid layer wetting the ingot surface m these and other alloys,'and if the mould wall is smooth, an unbroken surface layer normally IL The two types o f inverse segregation postulated have in common the outflow o f liquid during solidification. In the former, however, tho outflow occurs only during the solidification o f the last low melting point traction, whilst in the latter the whole o f the primary columnar growth is assumed to occur bv an outflow o f liquid. From a study o f equiaxed and columnar structures, it should not be difficult to show whether two such types do in lact occur. It is also worth noting that an outflow o f liquid would permit ot a more rapid heat transfer than the conventional conduction through a- solid outer shell, and some evidence might be obtained from the observation ox
rates o f solidification and heat transfer. .
Previous investigators appear to have been rather shy o f the idea ot a partlv solidified ingot consisting o f primary crystals completely separated by films"of liquid alloy. One has only to think o f the strong adhesion between flat plates o f glass or metal separated by a liquid film, to realize the possible strength of such a structure, and the author should not therefore regard such
a n idea as too improbable. , f
Finally, I should like once again to make a plea for the examination ol large ingots, for it is then possible to observe more clearly and leisurely the intermediate stages o f solidification.
* ./. Inst. Metals, 1927. 37, 275. f J Metals, 1035, 57, 89.
Mr. A. R . Ubbelohde,* M .A., B .S e .: Mechanical theories o f inverse segregation (Class B in Mr. Vaughan’s paper) assume that the composition o f solid deposited at any point from the liquid agrees with the usual solidus- liquidus curves, and that any departure from normal segregation, in which the purest solid is th atjirii deposited, is due to a mechanical transport o f residual liquid to abnormal parts o f the ingot. Direct evidence o f such transport would bo o f great value. A method which does not seem to have been tried is to “ mark ” a layer o f the liquid in the mould with minute quantities o f an indicator element. Although certain experimental difficulties suggest them
selves, the range o f radio elements available makes it possible to add small amounts o f a suitable element without seriously disturbing the physico
chemical behaviour o f tho layer o f “ marked ” liquid. A radiation photo
graph t o f sections o f the solidified ingot would show what movement, if any, had taken place with the marked liquid, under different conditions o f cooling.
In the absence o f conclusive information on possible mechanical causes o f inverse segregation, it is useful to review the thermodynamical implications o f certain physico-chemical theories (Class A). These arc all based in some way on the observation that inverse segregation is only marked when rapid chilling is used, so that there are large temperature gradients in the liquid, and crystallization takes place under irreversible conditions.
(a) Types o f substance which might give an appreciable I/udwig-Soret effect.
It has been suggested that when large temperature gradients are present in the liquid, a concentration gradient may be set up, as in the Soret effect in salt solutions. It may be useful to indicate in what circumstances (if ever) such gradients might become sufficiently largo to account for inverse segrega
tion. According to fairly generally accepted theory,% a temperature gradient dTldx in a liquid mixture o f two components A and B leads to a concentration gradient dNA/dx, where N A is the mol fraction o f A , and
- SA {dTldx) = (8FJSN A)T(dN Jdx).
In this equation — SA is the “ entropy o f transfer ” o f the component .4 in solution, and [SFaI ^a)t18 the change in partial molal free energy o f tho component A with concentration. Owing to the fact that the magnitude and sign o f the entropy o f transfer can only be determined experimentally from measurements o f the Soret effect, this equation is o f little use in predicting the magnitude o f the concentration gradients, except in one class o f systems. It is clear that when (SFA/SNA)T tends to become very small, even quite small values o f the entropy o f transfer may lead to large concentration gradients.
Well-established experiment and theory show that (S-P^/SN A)T tends to become very small in the region o f concentration and temperature where the liquid is about to separate into two liquid phases. (The system phenol + water is a well-known example.) This may possibly account for the fact that in steel ingots containing a number o f impurities, the elements C, S, P , whose miscibility in the liquid is probably limited, show largo inverse segregations in suitable circumstances, whereas elements closely similar to iron, such as nickel or chromium, show much smaller segregations. Unfortunately, it must he emphasized that the Soret effect in liquid metals has not been sufficiently investigated to warrant any firm conclusions on its practical importance in ingot casting.
(b) Irreversible crystallization. Solidus-liquidus curves refer to true
* Tho Davy-Faraday Laboratory, Tho Royal Institution, London.
t Investigations on a limited number of radio elements have been published by Tammann and co-workers ; 7j. anorg. Ckem., 1932, 205, 145; 7. Metallhunde, 19 33 gij J53 9Q7
t E. Eastman, J. Amer. Chcm. Soc., 1928, 50, 283; 1937, 33, 1198 C. Wagner, Ann. Physih, 1929, [v], 3, 629.