VOLUME 85, NUMBER 5 PHYSICAL REVIEW LETTERS 31 JULY 2000 Universal Slow Dynamics in Granular Solids James A. TenCate, Eric Smith, and Robert A. Guyer Los Alamos National Laboratory, Earth and Environmental Sciences Division, Los Alamos, New Mexico 87545 (Received 27 September 1999; revised manuscript received 22 December 1999) Experimental properties of a new form of creep dynamics are reported, as manifest in a variety of sandstones, limestone, and concrete. The creep is a recovery behavior, following the sharp drop in elastic modulus induced either by nonlinear acoustic straining or rapid temperature change. The extent of modulus recovery is universally proportional to the logarithm of the time after source discontinuation in all samples studied, over a scaling regime covering at least 10 3 s. Comparison of acoustically and thermally induced creep suggests a single origin based on internal strain, which breaks the symmetry of the inducing source. PACS numbers: 62.40. + i, 62.65. + k, 91.60. – x Relaxation processes in which some stress is relieved as the logarithm of time after a step function is applied are remarkably common: in mechanical response of rocks [1] and metals [2], thermoremnant magnetization relaxation in spin glasses [3], dc susceptibility in granular magnetic media [4], and even fluid invasion percolation in soils [5]. Though these systems have completely unrelated micro- scopic dynamics, they have in common that these forms of creep all monotonically minimize a free energy in response to a constant external stimulus, respecting the symmetry of the source. In studies of nonlinear elasticity of many rocks, it was found that, at strains 10 26 , retarded effects resembling creep appeared, which could not be explained with equi- librium elastic theory, either classical [6] or hysteretic [7]. Their universal feature was a persistent drop in elastic modulus, and increase in material damping, which could be induced by harmonic acoustic stressing, and we have since found also by thermal shocking. After the stress or shock is removed, the material properties recover toward their original values as the logarithm of elapsed time, over a featureless scaling regime lasting hours to days. An im- portant feature of this effect, which we call slow dynamics, is that the elastic modulus decreases in response to sym- metric stress cycling or temperature change of either sign, thus violating the symmetry of the inducing source. This Letter reports properties of slow dynamic response in a variety of sandstones, limestone, and concrete. A reso- nance method is used to enhance sensitivity to small shifts in material properties: reduced modulus causes resonant frequency to drop, and increased damping lowers the res- onant quality factor Q (defined so that successive cycles of an undriven oscillator decrease in magnitude by e 2pQ [8].) The materials studied are sufficiently different that there appears to be no universally shared chemistry, char- acteristic scale, or microstructure, suggesting that slow dy- namics may be an emergent form of creep not seen before. The symmetry breaking of slow dynamics resembles the quick loss of microscopic contact area, and its subsequent restoration as log (time), in the slipstick of a static friction bond [9]. Depending on geometry, however, slip may lead to irreversible change (damage accumulation) while slow dynamic recovery is (at least macroscopically) perfect, even over hundreds of cycles spanning more than a year. The combined resemblance to stress-relieving creep and bond rupture raises the question of whether slow dynam- ics may arise from a glasslike state, somehow intermediate between equilibrium elasticity and damage formation. Creep in all familiar systems [1–5,9] is understood as a result of thermal activation, and even in the absence of a full microscopic understanding of slow dynamics, this leads to a generic prediction of temperature dependence of the susceptibility. Therefore, slow dynamic recovery was measured as a function of temperature as well as sample type and, in one case, humidity. The suspected connec- tion to bond rupture suggests that modulus and damping changes are induced by internal straining. Therefore, the shifts from acoustic driving and thermal shocking were also compared. From the known thermal expansion an- isotropy of the underlying grain materials, the compari- son shows all slow dynamic effects to be consistent with a single strain-dependent origin. The experimental method, described in detail in Refs. [10,11], is to drive a suspended cylindrical sample of rock or concrete in the fundamental longitudinal elastic mode (Young’s mode), with a piezoelectric force transducer cemented between one end of the sample and a massive backload. Acceleration of the opposite end of the sample is measured with a lightweight accelerometer and processed with a lock-in amplifier referenced to the driving signal. The driving force is a harmonic acoustic wave, incremented through the fundamental resonance frequency of the bar, to produce the frequency-dependent lumped-parameter response function of the resonant barbackload system. These response functions can cur- rently be processed to resolve shifts in Young’s modulus of one part per million of the static value. Figure 1 shows the amplitude of a typical response func- tion, for frequency incremented several times per second, first upward and then downward through resonance. An elastic system described by an equilibrium equation of state, either single-valued (classical) [6] or multivalued 1020 0031-900700 85(5) 1020(4)$15.00 © 2000 The American Physical Society