Please cite this article in press as: Boschi, L. et al., Petascale computing and resolution in global seismic tomography, Phys. Earth Planet. Interiors (2007), doi:10.1016/j.pepi.2007.02.011 ARTICLE IN PRESS +Model PEPI-4794; No. of Pages 6 Physics of the Earth and Planetary Interiors xxx (2007) xxx–xxx Petascale computing and resolution in global seismic tomography L. Boschi a,∗ , J.-P. Ampuero a , D. Peter a , P.M. Mai a , G. Soldati b , D. Giardini a a Institute of Geophysics, E.T.H. H¨ onggerberg-HPP, 8093 Z¨ urich, Switzerland b Istituto Nazionale di Geofisica e Vulcanologia, via di Vigna Murata 605, 00143 Roma, Italy Received 11 January 2007; accepted 23 February 2007 Abstract Will the advent of “petascale” computers be relevant to research in global seismic tomography? We illustrate here in detail two possible consequences of the expected leap in computing capability. First, being able to identify larger sets of differently regularized/parameterized solutions in shorter times will allow to evaluate their relative quality by more accurate statistical criteria than in the past. Second, it will become possible to compile large databases of sensitivity kernels, and update them efficiently in a non-linear inversion while iterating towards an optimal solution. We quantify the expected computational cost of the above endeavors, as a function of model resolution, and of the highest considered seismic-wave frequency. © 2007 Elsevier B.V. All rights reserved. Keywords: Tomography; Finite-frequency; Resolution; Computational seismology; Inverse theory 1. Introduction In the past decade there have been a number of claims by tomographers that lead to a re-consideration of cer- tain aspects of the theory of the Earth. Those claims have been justified by improvements in tomographic resolution. van der Hilst et al. (1997), for example, inverted a very large database on an unprecedentedly dense voxel grid, making use of an inversion algorithm that exploited the inherent sparsity of the linear inverse problem. They found very sharp images of fast, deep heterogeneities, that, because of their geographic distri- bution, were explained in terms of subducted material, sinking into the lower mantle. This finding, while subject of debate (is “resolution” really as high as claimed?), has been a strong argument in favour of whole-mantle versus ∗ Corresponding author. E-mail address: lapo@erdw.ethz.ch (L. Boschi). layered convection. More recent examples are the work of Ishii and Dziewonski (2003), who mapped an “inner- most inner core” of only 300 km in radius, suggesting that it be “the oldest fossil left from the formation of Earth”, and the controversial article of Montelli et al. (2004), who improved global resolution by means of a more accurate approach to the calculation of sensitiv- ity functions, and found “clear evidence that a limited number of hotspots are fed by plumes originating in the lower mantle”; this claim is clearly relevant to the current debate on the nature of mantle plumes (e.g., Anderson, 2000; Courtillot et al., 2003; Schubert et al., 2004; Sleep, 2006; Boschi et al., 2007), involving all disciplines in the Earth sciences, and stirred a very animated discus- sion (de Hoop and van der Hilst, 2005a, 2005b; Dahlen and Nolet, 2005; van der Hilst and de Hoop, 2005, 2006; Boschi, 2006; Boschi et al., 2006; Montelli et al., 2006a; Trampert and Spetzler, 2006). The controversy originated by these publications, and, in general, the lack of correlation at short spatial 0031-9201/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.pepi.2007.02.011