526 
Proceedings of the Royal Society of Edinburgh. [Sess. 
After the results given in the above table were obtained, a determination 
of A was made for more concentrated solutions. It is, of course, well 
known that the absorption spectra of some cobalt salts undergo changes 
when the solutions become very concentrated. For more concentrated 
solutions thinner layers of solution must be used, 1 mm. and thereabouts. 
There is difficulty in measuring this thickness, and I thought at first that 
it might be possible to obtain A without knowing the thickness by using 
the arrangement illustrated in the diagram. M and M' are mirrors. The 
solution is contained in a glass box, and the light passes vertically down 
through the bottom of the box. At first the box is filled with concentrated 
solution of strength c', depth d', and molecular extinction coefficient A', and 
the deflection is noted. Some water is then added, the strength, depth, 
and molecular extinction coefficient change to c, d, and A, and the deflection 
is again taken. Any change in the deflection must be due to a change in 
A for cd — c'd'. It is easy to measure d, as it is much larger than d ' ; c can 
be obtained by comparing the tint of the solution spectrophotometrically 
with solutions of known strength ; hence At This method failed, however, 
owing to a curvature of the surface of the liquid which was produced by 
capillary forces. The liquid thus acted as a concave lens of changing 
power, the curvature altering when more water was added. 
I therefore used cells of 1 cm. inside thickness, putting parallel- sided 
pieces of plate glass inside them to diminish the thickness of the liquid, and 
determined the thickness of the layer with a reading microscope. The 
value of the smallest thickness could not be relied on to less than 4 per 
cent, by this method ; consequently, the values of A obtained with it are 
uncertain to a constant factor — that is, they may have to be multiplied 
