An inverse scattering technique is developed to enable a three-dimensional
sample reconstruction from the diffraction figures obtained for different s
ample orientations by electronic projection microscopy, thus performing a d
iffraction tomography. In its Green's-functions formulation, this technique
takes account of all orders of diffraction by performing an iterative reco
nstruction of the wave function on the observation screen and in the sample
. In a final step, these quantities enable a reconstruction of the potentia
l-energy distribution, which is assumed real valued. The method relies on t
he use of singular values decomposition techniques, thus providing the best
least-squares solutions and enabling a reduction of noise. The technique i
s applied to the analysis of a three-dimensional nanometric sample that is
observed in Fresnel conditions with an electron energy of 40 eV. The algori
thm turns out to provide results with a mean relative error around 3% and t
o be stable against random noise.