Unlike piezoelectricity (proportionality between a uniform stress or strain and the electric polarization of a material), flexoelectricity (proportionality between stress or strain gradient and electric polarization of a material) is rarely considered for electromechanical transduction and energy harvesting at the sub-micron scale, though flexoelectricity does not require that the material unit cell be non-centrosymmetric (as volume piezoelectricity does). Indeed, the challenge is to find a material with flexoelectric effects strong enough to be interesting for applications. One strategy could be to use the increase of the deformation gradient for a reduction of scale and the corresponding transition from bulk effects to surface effects to allow for new ways to select materials at the nanoscale and get an interesting conversion ratio between electric and mechanical energies. In order to study these phenomena, we have improved on previous molecular dynamics simulations of the bending of various carbon single-wall nanotubes by non-longitudinal static electric fields [1,2] and compared the results with the previsions of a continuum mechanics model for the bending of a flexoelectric beam by an external electric field [3], in order to extract the relevant flexoelectric tensor components from a fit. Preliminary results will be given.