Kenneth C. Creager

Professor of Earth and Space Sciences

Ph.D., Earth Sciences, University of California, San Diego, 1984.

Primary research interests: Seismology and geophysical inverse theory.

My research has focused primarily on exploring dynamical processes within the Earth's deep interior by probing the Earth using seismic waves.

Inner-Core Anisotropy and Rotation

We are analyzing differential travel times and amplitudes of PKP phases to constrain the elastic properties of the Earth's solid inner core. The inner core appears to be strongly anisotropic, characterized by compressional wave speeds being 3-4% faster for rays oriented parallel to the Earth's spin axis than those which are perpendicular. The anisotropy appears to be much stronger in the western hemisphere than the eastern hemisphere, suggesting that crystal alignment is much better organized in the western hemisphere. Several physical mechanisms can be invoked to align the crystals, but it appears that every mechanism which would lead to the observed symmetry with respect to the Earth's spin axis is associated with a strong toroidal magnetic field within the outer core. The toroidal part of the magnetic field in the inner core cannot be observed directly from the Earth's surface, but our analysis suggests that it may by much larger than the poloidal part of the field which can be inferred from surface magnetic observations. The travel times of compressional waves that go through the inner core from South Sandwich Island earthquakes to stations in Alaska increases systematically by about .1 s per decade. This observations, combined with estimates of the lateral gradient in wave speed within the inner core lead me to the conclusion that the inner core is rotating 0.2-0.3 degrees/year faster than the mantle.

Imaging discontinuities in the mantle

We have developed methods of stacking seismograms recorded at arrays of hundreds of seismic stations to look for small coherent phases that would be caused by energy reflected and refracted off of boundaries internal to the mantle that separated rocks of differing chemistry or polymorphic phase. We clearly image such a "discontinuity" at a depth ranging from 660 to 690 km beneath the Izu-Bonin subduction zone south of Japan. This is interpreted as the spinel to perovskite plus magnesiowustite phase change. The local depression of this phase boundary which normally occurs at 660 km depth is probably caused by the cold temperatures beneath this subduction zone. We also look for evidence of a deeper discontinuity that might be caused by a change in chemistry between the upper and lower mantle. At the 1% detection level, there is no nearly horizontal discontinuity in shear-wave velocity beneath Izu-Bonin. This suggests that the upper and lower mantle have the same chemistry. We do however, see evidence for a near vertical discontinuity at depths of about 1000 km that may be the remains of an ancient subducted slab.

Seismic Imaging of Subducting Slabs

We are also imaging the seismic velocity structure associated with the subduction of cold slabs of oceanic lithosphere. The primary focus of this research has been to answer the question, "Do slabs penetrate deep into the lower mantle?" In ongoing work we have presented strong evidence that the Tonga, Mariana, part of Izu-Bonin, Japan, and Kuril-Kamchatka slabs all penetrate deep into the lower mantle, in some cases at a steeper dip than in the upper mantle. The Aleutian slab, with seismicity terminating at a depth of less than 300 km, also appears to subduct into the lower mantle at an increased dip.

In order to obtain a high-resolution image of the discontinuities in seismic wavespeeds within the upper 100 km of the crust and mantle, we have deployed an array of 16 broad-band seismometers at 4 to 8 km spacing stretching from the Olympic Mountains through Seattle to the Cascade foothills. One discontinuity of particular interest is one that separates the crustal and mantle parts of the subducted plate. By comparing the depth of this dipping interface with the locations of deep (>40 km) earthquakes, we hope to determine whether the earthquakes lie within the mantle or crustal part of the subducted plate. This will help sort out theories regarding the cause of these earthquakes.

3-D Modeling of Slab Deformation

Oceanic lithosphere is thin (relative to its lateral dimensions), considerably stronger than the mantle below, and thus has very little internal deformation. Because of the canceling effects of increasing temperature and increasing pressure on rheology, we expect that after subduction into the upper mantle, a slab may maintain its relative strength advantage. We have developed a technique for calculating the 3-D flow field which gives the minimum amount of global in-plane strain rate (minimum dissipation power if we assume a linear-viscous or power-law rheology) associated with the geometric change in configuration from a spherical shell of oceanic lithosphere to a slab's observed geometry. Applied to the Aleutians, these 3-D flow calculations, along with the 3-D thermal conduction calculated from this flow field, provide a quantitative explanation for the long-arc variation in the depth to the seismicity cutoff (which varies from 50 to nearly 300 km), seismic moment release rates of intermediate-focus earthquakes, seismic moment tensor orientations, and variations in slab dip from the Aleutians to Alaska.

These flow calculations are also being applied to the Cascadia subduction zone, where the trench has a backwards curvature (oceanward concave) relative to most island arcs. Landward of this bend lie the Olympic Mountains (one of the few subareal accretionary prisms not associated with continent-continent collision), intense intraslab seismicity relative to regions to the north and south, and an arch in the slab geometry characterized by a shallow (10 deg) dip relative to the 20(deg) dip along profiles to the north and south. Each of these observations can be explained as a natural consequence of the trench geometry and the 3-D flow field that minimizes the total amount of in-plane strain rate.

In South America it appears that the concave oceanward bend in the trench associated with the Bolivian Orocline causes along-strike compression. This may explain the along-strike buckling apparent in the geometry of the slab as defined by deep seismicity. The subducted slab is seismically active at depths between 500 and 650 km along two nearly north-south segments under Argentina and under Peru. The Peruvian segment is offset about 500 km to the west. The region between the segments has very few earthquakes, but contains three very large earthquakes including the June 9, 1994 great Bolivian earthquake whos magnitude was Mw=8.3. This is the largest deep earthquake of the century and was felt as far away as Seattle, the furthest reported felt earthquake ever. In this region where the two straight segments are being connected the local slab strike is nearly east-west. Our calculations show that the particle trajectories during subduction are locally along strike in this area rather than the normal situation where particles are going down dip. The lower mantle (below 660 km depth) is thought to be more viscous (stiffer) than the upper mantle. If the subducted slab is anchored in the lower mantle and moving at near surface plate rates along-strike in the upper mantle the strain rates of the slab will be high leading to local slab thickening and conditions which may allow very large earthquakes to occur. A similar setting may explain the large Sakhalin deep earthquake in 1990 which also occurs in a region that is otherwise aseismic.

Coral: A matlab program for seismic waveform analysis

We have written a MATLAB (TM) program we call Coral that is designed for analyzing seismic waveform data. It includes a toolbox of matlab functions that are of general use for seismologists, as well as a program called ah2ml that reads waveform data into matlab. The code can be obtained by anonymous ftp to: and is in the file called pub/out/coral_tar.Z (for matlab version 5) and pub/out/coral4_tar.Z (for matlab version 4). To run coral you will need MATLAB and MATLAB's signal processing toolbox as well as FORTRAN and C compliers.

Click HERE to view the paper on South America.

Click HERE to view an earthquake location homework problem.

List of recent publications

Click HERE to get latest information and e-mail address.