TABLE OF CONTENTS Historically, models for "structures" of glasses have been based on a desire to explain trends in properties. The original 'random network model' attributed to Zachariasen was based solely on property data, i.e., glass formation behavior. While the work of Warren and co-workers which soon followed the proposal of the concept of a random network supported this model, Warren's work was not the source of the model. In fact, virtually all of the basic models for the structures of oxide glasses were originally proposed as means of justifying the trends observed in the properties of simple compositional series of glasses. Although modern spectroscopic methods (Raman and infrared spectroscopy, nuclear magnetic resonance, neutron diffraction, electron spin resonance, EXAFS, XPS, etc.) now allow researchers to propose models for new glasses solely on the basis of spectroscopic data, these models must still pass the time-honored test of explaining all of the trends observed in the properties of those glasses, before such models can be accepted as truly representing the structure of the material. It should also be recognized that these spectroscopic methods are best suited for examining the local structure about an ion, with some extension to the intermediate range of order, and, in general, are quite incapable of application to materials where the properties are controlled by microstructure rather than short to intermediate range atomistic order.
As a further guide to the testing of any proposed model, Shelby has proposed the Fundamental Law of Structural...
