70 SCIENCE PROGRESS 



tartrate at a high temperature, namely to the region QRST in 

 Fig. 6, then the dispersion might be expected to be of the same 

 type in both cases, namely positive — the rotation for violet 

 should be greater in an absolute sense than for red. But if the 

 minimum were such as might be expected by continuing the 

 ethyl tartrate curves towards low temperatures, the rotation 

 for red light should have a higher absolute value than the 

 rotation for violet light, the dispersion being thus negative. 

 This point was therefore investigated by an examination of 

 homogeneous z'sobutyl dibenzoyltartrate, for six colours of 

 light and for temperatures between o° and i8o° {Proc. R.S.E., 

 191 8-19, 39, 26, 32), when curves were obtained corresponding 

 very closely to those between A and C, Fig. 6. There was a 

 minimum in each curve, and the absolute value of the rotation 

 was greater for red than for violet. The dispersion is thus 

 negative, and the minimum observed in this region could 

 therefore hardly correspond to the minimum in the neighbour- 

 hood of S in Fig. 6. It may therefore be concluded, until some 

 better suggestion can be made, that the general relationship 

 between temperature and rotation for ethyl tartrate, and 

 therefore, with minor modifications, for other tartrates, over a 

 very wide range of temperature is represented by the curves 

 in Fig. 6. At quite a low temperature the dispersion is visibly 

 normal ^ but negative ; at temperatures between about zero 

 and 90° the dispersion is visibly anomalous ; beyond that ,the 

 curves rise to reach a maximum ; they then fall to a 

 minimum, which apparently does not go below zero, and then 

 rise again, presumably to reach another maximum, all without 

 actual intersection, and therefore without visible anomaly in 

 the dispersion. 



Two other points may be dealt with in this connection. It 

 has long been obvious that the so-called dispersion coefficient 

 for any given substance has no practical value. This coefficient, 

 which is simply the value of the rotation for one colour of light 

 divided by the value of the rotation for another colour of light — 

 usually for violet divided by yellow or green divided by red, etc. 

 — naturally does not show any very marked variation at differ- 

 ent temperatures for any substance which has a high rotation, 

 positive or negative. Thus, for example, the dispersion co- 

 efficient measured in the ordinary way on the curves of Fig. 6 

 at A, B, or C would probably not differ very greatly. But it 

 will be obvious that dispersion coefficients taken from E to H 



^ This means merely that to the eye the magnitude of the rotation is in 

 the same order (or the inverse) of the magnitude of the wave-length. As 

 nobody knows what normal rotation dispersion is, it is difficult to define 

 the term except in a quite arbitrary manner. Consequently anomalous and 

 abnormal rotation dispersion present equal difficulty in the way of definition. 



