The theoretical background to the application of linear viscoelasticity in describing the anelastic behavior of rocks is developed in this review. The constraints imposed on the behavior of the impulse response and system function of a linear system by the requirements that the system be causal and passive are examined. These include the existence of dispersion relations for the system function for causal systems and the role of the concepts of positive reality and minimum phase in passive systems. The uses of these linear system concepts in linear viscoelasticity and in the propagation of plane waves in an isotropic viscoelastic medium are considered. Finally, applications to seismology are presented, in particular the significance of the dispersion associated with a quality factor Q that is constant or varies as a power of frequency in comparisons of seismic data for different frequency ranges and in the evolution of the shape of propagating pulses.
INTRODUCTIONIn studies relating to the anelasticity of rocks one of the most important considerations has been whether the anelastic behavior is or is not linear, in the sense that the principle of superposition of signals is valid. In seismological applications it has been common practice, if only for the obvious convenience, to assume that the anelasticity is linear, and a large body of theoretical seismology is founded, apparently successfully, on this assumption. Until recently, the experimental evidence was not convincing, and the validity of this approach has been questioned in the past [e.g., Knopoff,, 1964;Stacey et al., 1975]. Laboratory experiments by McKavanagh and Stacey [1974], which directly observed the stress-strain hysteresis loops for rocks subjected to a sinusoidal axial load, yielded loops that were cusped at the ends and were thus indicative of nonlinear artelasticity. On the other hand, the creep experiments of Pandit and Savage [ 1973] were consistent with the linear assumption. Less direct methods such as the analysis by Wuenschel [1965] of pulse degradation in Pierre shale and the observations of ultrasonic pulse rise times by Gladwin and Stacey [1974] also yielded results that were consistent with that assumption. Recent results indicate that at the strain levels involved in seismic waves the anelasticity of rocks is likely to be linear. Brennan and Stacey [1977] conducted experiments, similar to those of McKavanagh and Stacey [1974], in which granite and basalt rocks at room temperature and atmospheric pressure were subjected to sinusoidal shear. At the strain amplitudes involved (less than 3 x 10 -6) it was found that the hysteresis loops were elliptical, as is expected for linear anelasticity, and in addition, the variation of the real part of the compliance was of the form predicted by linear theory. Experiments conducted since then have demonstrated that similar results are observed in a sandstone specimen at a strain amplitude of 10 -6 . The implication is that the previous observations of hysteresis in rocks by McKavanagh and Stacey ...