The Inversion of Specific Heat Curves

It is shown that the methods proposed by Montroll and Kroll for deriving the lattice frequency spectrum from the experimental specific heat curve will in practice yield only a `smoothed' spectrum, averaged over a range of 1.6: 1 or more in frequency. Any practicable improvement in experimental accuracy will reduce this smoothing only slightly. The same smoothed spectrum can be derived more simply by a procedure involving only one quadrature, and a still simpler trial-and-error method is also described and illustrated.