SynopsisThe broad-line proton NMR spectra of polyethylene are separated into three components (broad, medium, and narrow) corresponding to the crystalline and two kinds of amorphous protons, respectively. All amorphous protons are found to be mobile above 210°K. The unusually low molecular mobility in the amorphous regions of polyethylene compared to purely amorphous and other partially crystalline polymers is considered to be adequately described by the network model of Edwards and De Gennes. In this model the chain motions are anisotropic and restricted to small tubular volumes. The medium and the narrow component are believed to result from two different modes of chain motion within these tubes, depending on the free volume available. Two motions in the crystalline regions are observed. One a t temperatures below 110°K involves 2%5% of the protons, depending on the crystallinity of the material, and the other beginning a t 290°K involves all crystalline protons (a-process). Coupling of the amorphous and crystalline motions is found above 290°K. Several line shapes have been tested as representations of the three components. Of these the low-temperature spectrum, the Gauss-Lorentz product curve, and the Lorentz curve proved to be the most suitable shapes for the broad, medium, and narrow component, respectively. Using these line shapes, the best fit of the experimental spectra and the expected agreement of the broad and the crystalline fraction are obtained over a very broad temperature range. Above 310°K the low-temperature spectrum must be replaced by the convolution of a Gauss curve and a rectangular function to take into account the line-shape transition of the a-process. The modulation broadening of all components is considered, and this allows line-shape analysis close to the melting point.It was soon recogni~ed,~-~ however, that the Wilson-Pake method yields crystallinities which deviate greatly from the values obtained from independent measurements and, in addition, are found to depend on temperature in regions where they are expected to be constant. These discrepancies were shown to originate in the greatly simplified assumptions of the Wilson-Pake method.