y. Phys. IVFrance 104 (2003) 647 @ EDP Sciences, Les Ulis DOI : K). 1051/jp4 : 20030163 X-ray fluorescence tomography of individual waste fly ash particles B. Gotosio', A. Simionovici\ A. Somogyi\ C. Camerani3 and B. M. Steenari /D22, SRF, BP. 220, 38043 Grenoble cedex, France 2 Dipartimento d/F Fisica, Università di Cagliari, Cagliari, Italy 3 Chalmers University of Technology, 41296 Goteborg, Sweden Abstract. Fluorescence tomography is a non-destructive, non-invasive technique that providesinformation aboutthe internai spatial distributionof each element that emits a detectable fluorescence signal from the measured slice of the sample. The reconstruction proMem for fluorescence tomography is much more difficult than it is for transmission tomography, due to the absorption of thé photons within theexcitation and détection paths. The present work présents a reconstruction technique that is based on the SART (Simultaneous Algebraic Reconstruction Technique) algorithm, which hasbeen modifiedin order to take into account absorption corrections. This technique has been applied to theanalysis of individual waste fly ash particles of about 80-150microndiameters, which have been scanned using an X-ray beamof V*H=2*5 pjn spot-size and placing an energy dispersive Si (Li) X-ray detector at 90 degrees to the incoming excitation beam. From previous scanning i-SRXRP measurements it is clear that the elemental distributionwithin individual fly ash particlesis highly inhomogeneous but no information could beobtained on the location of the différentinvestigated elements (within/onthe surface of the particle). On the other hand thé location of toxic elements within individual fly ash particles affects the possible fate of thèse elements during e. g. fly-ash recycling. Thus the aim of this study was to investigate the 2D interna) elemental distribution of the particles, with spécial attention to that of the toxic metals, such as Zn, Cd, Pb. 1. INTRODUCTION Conventional X-ray transmission tomography allows one to obtain a spatial map of the analyzed sample in terms of thé absorption coefficient distribution. This, however, does not provide information about the chemical composition of the sample. In order to obtain additional information, other tomographic techniques hâve been developed, based on the détection of the scattered/fluorescent X-ray photons. In particular, fluorescence tomography is based on the signal produced by the photons coming from fluorescent émission on a detector that is generally placed at 90 degrees to the direction of the incident beam. Fluorescence tomography présents two principal complications compared to the transmission tomography : usually the counting rate is several orders of magnitude less than for transmission tomography and the reconstruction calculations are more complicated. The first problem is usually dealt with by using intense X-ray sources such as synchrotron radiation [1]. The second problem has not yet been completely solved in général, even though gréât improvements have been done in the reconstruction techniques. Atthough the conventional filtered backprojection algorithm is often used for fluorescence tomography reconstruction, this technique is inappropriate when X-ray absorption effects are not negligible, i. e. when the sample dimension is comparable or greater than the fluorescent photon mean free path. Hogan et al. [2] proposed an adaptation of the filtered backprojection algorithm, modified in order to keep into account absorption corrections. This approach gives acceptable results only when the absorption inside the sample is low. In order to obtain good quality reconstructions even in strong absorption conditions, several itérative approaches have been proposed in the literature [3-5]. However, the computational time required by some of thèse techniques may be very large. Furthermore, the speed of convergence and even the ability to converge of such techniques often dépend on an initial guess for the solution. In this paper we present a reconstruction technique that is based on the SART (Simultaneous Algebraic Reconstruction Technique) algorithm [6], which has been modified in order to take into account absorption corrections.