The Phase D'. The great similarity in the character of this phase from the moment when it first appears in the C' eutectic of Sna, to the moment when it begins to appear as fern leaf in Snig, and gradually grows in amount until it fills the alloy at Sn-2o, strongly inclines us to think that it is everywhere Cu 4 Sn, and not a solid solution isomorphous with this compound. Perhaps the best argument for the view that D' is pure Cu 4 Sn can be found in the fact that Roberts-Austen and Stansfield's line of eutectics (XD 2 in our figure) extends to Sn 20 .
The Phase E'. Between the percentages D' and E' there is, as we have said, some uncertainty as to the composition of this phase; indeed, at one time we thought that there was another singularity between these points. But on the right of Sn 2 5 E' must be pure, or nearly pure, Cu 3 Sn. The following experiments bear on this question.
We found that E' and H were but little attacked by hydrochloric acid in the cold, and not at all attacked so long as free tin was present. We therefore chilled the alloys Sn 35 , Sn 50 , and Sn 65 at 450, a tem- perature at which they consist only of plates of E' and liquid. They were then exposed for a week or more to strong hydrochloric acid, the acid being tested and changed every day. At first large quantities of tin, but not a trace of copper, dissolved, but gradually less tin came into solution, and when no more tin would dissolve a little copper was found in solution. At this stage the alloy consisted of separate plates of E' and a good deal of black mud. The mud was thoroughly washed away arid the plates of E' analysed. Two analyses of each alloy were made.
The Sn 35 gave 39'1 39'1 per cent. Sn. The Sn 50 gave 38-538-8 The S nti5 gave 41 -7 42-2 while Cu 3 Sn requires 38*35 per cent. Sn.
Considering the great difficulty of avoiding the presence of H in the plates of E' analysed, we think the above conclusive.
By similarly treating the alloy Sn s ,o, in which there should theoreti- cally be no E', we obtained a beautiful and clean sample of H in brilliant spear-like leaflets. This contained 61 per cent, of tin, while CuSn requires 65 per cent., so that we have not yet obtained CuSn in a state of purity.
One reason of some weight in support of the formula CuSn is that H must be the body in solution in the liquid alloys of the branch IK of the freezing-point curve ; and as we showed in 1890, the value of the depression of the freezing point of the tin points to a molecule in solu- tion containing only one atom of copper.
We have to thank Miss D. Marshall, Lecturer of Girton College, and Mr. W. Fearnsides, B.A., of Sidney College, for much help during the experiments.
