|Zusammenfassung||Analog sandbox simulations have been applied to achieve qualitative and quantitative insight into geological processes occurring in compressional and extensional settings. A direct comparison of model and nature is possible, because suitable analog materials, such as sand or glass beads, exhibit a similar Mohr-Coulomb behavior as sediments and rocks of the upper crust. Thus, analog models are scaled geometrically to nature by the density and frictional properties of the material used in the experiments. For example, to study the evolution of accretionary wedges in subduction zones, a typical experimental apparatus consists of a fixed horizontal plate (few meters x few decimeters) on which a conveyor belt, representing the subducting oceanic plate, is dragged underneath a rigid back wall, acting as the rigid part of the continental margin. The sand, representing deposited sediments, is sieved in layers onto the conveyer belt, and upon convergence, accumulates in regular imbricates in front of the back wall. Internal structures of sandbox models and their temporal evolution can only be directly observed in 2-D profiles along the glass walls confining the experiment or indirectly by surface observations by means of particle imaging velocimetry (PIV). When investigating regimes with along-strike variations, 3-D information of the sand models is needed, but can only be obtained by either very expensive and very elaborate X-ray tomography on small models (few centimeters), or, after the deformation is finished, by solidifying the model with transparent resin and cutting slices. This method provides high resolution 2-D slices to analyze 3-D structures. However, after solidification, further deformation of the model is impossible.
To extend the simulations to three dimensions, I perform non-invasive seismic physical modeling on these analog sandbox models. The long-term objective of this approach is to image seismic and seismological events of static and actively deforming 3-D analog models. To achieve this objective, a new small-scale seismic apparatus, composed of a water tank, a PC control unit including piezo-electric transducers (PETs), and a positioning system, was built for laboratory use. To build the models, I use granular materials such as quartz sand, garnet sand and glass beads, so that brittle deformation can take place. Unlike typical analog sandbox models, the granular models now are required to be completely water saturated so that the sources and receivers are directly and well coupled to the propagating medium. Ultrasonic source frequencies (~500 kHz) corresponding to wavelengths ~5 times the grain diameter are necessary to be able to resolve small scale structures. When thus doing seismic physical modeling of granular models, two aspects besides the model scaling require particular attention to assess the feasibility of this setup and method: The transducer properties with respect to their use in seismic reflection surveys on mm-scale, and the acoustic material properties.
The properties of specially designed PETs with reduced directionality were tested to assess their feasibility for seismic profiling on millimeter-scale with respect to their frequency sensitivity, their directionality, and the change of waveform as a function of offset. The experiments show that the PETs produce the best quality data at frequencies around 350-550 kHz, which is sufficient to resolve structures of ~2-1.5 mm dimension within saturated granular material. However, to inhibit ringing, a better control over the emitted source signal should be achieved. For these frequencies, the amplitudes decay to ringing noise level at incidence angles of <35°; for a 10 cm deep reflector that results in a 14 cm source-receiver offset. Below this offset, the first and second phase of the recorded signals still coincide so that a normal-movout correction during seismic data processing improves the signal. This shows that the special design of the PETs amounts to a reduced directionality compared to traditional transducers while maintaining the energy output. However, the energy output is fairly low for a highly attenuative material such as sand, so that the penetration depth is only 5 cm. Nevertheless, to this date, these are the most suitable transducers available to bridge the gap between the unwanted directionality and the desired energy output.
The acoustic properties of various granular materials are reviewed and tested experimentally in order to identify materials of sufficient impedance contrast. However, the sound velocity of various granular materials, such as quartz sand, garnet sand and glass beads, under atmospheric pressure is difficult to obtain. Only the velocity measurements of glass beads produce reasonable results of 1.8 km/s. The extreme variability of quartz and garnet sand prevents, that the true velocity can be deduced. The reason for this variability is that sound velocity primarily depends on the coordination number, which is a measure of the nature of the grain-to-grain contacts. Therefore, the velocity and attenuation are highly sensitive to small changes in packing, which are difficult to control when building a model. Hence, a reflection of an interface cannot be coerced by different acoustic velocities above and below the interface, but by an interface that has a contrasting coordination number compared to the model material above and below. The clearest reflections are generated in glass bead models where the interface is sprinkled with glass powder filling the intergranular space, and then graded flat. Seismic sections over layers made of glass beads contain less internal noise and attenuation than those made of sand due to the better sorting and smoother surface of glass beads compared to sand grains. Hence, the use of very well-sorted materials consisting of well-rounded grains, independent of mineralogy, reduces the inhomogeneities in packing and therefore improves the data quality.
Since it is not only desired to seismically image layer interfaces, but also shear bands within a deforming model, I show seismic images of a model before and after a string has been pulled through to simulate the decompaction occurring at shear bands. The decompaction of grains due to the string produces a reflection that can be detected in seismic data. The shear band is better resolved in sand than in glass beads. Different to field surveys, laboratory surveys are able to resolve the shear zone itself.
Finally, seismic reflection processing of a multiple-offset survey over a two-layer structural model containing channels and a shear band enhances the data quality and resolution significantly. This result is an improvement to previous studies, in which zero-offset surveys were conducted under the assumption that the directionality of the transducers impedes the advantages of multiple-offset data. Here, this assumption does not hold true due to the advanced PETs and to the survey geometry which is optimized to the properties of these particular PETs. However, especially for more complex models, the clarity and penetration depth need to be improved to study the evolution of geological structures in analog models with this method. As long as no source with a considerably higher energy output and spherical wave emission is available, I suggest to do ultrasonic seismic surveys across rather shallow models.
Nevertheless, even with model thicknesses above the penetration depth, the 3-dimensional albeit shallow information gained by seismic imaging of the models is feasible and would be beneficial in combination with PIV imaging, which provides a 2-D image of high spatial and temporal resolution over the entire depth of the model.