Due to their negative impacts on environment and human health, future regulations on soot emissions are expected to become stricter, in particular by controlling the size of the emitted particles. Therefore, the development of precise and sophisticated models describing the soot production, such as sectional methods, is an urgent scientific and industrial challenge. In this context, the first objective of this work is to use for the first time a sectional model to perform an LES of a sooting turbulent flames in order to demonstrate its capacities. For this, the whole LES formalism for this approach is developed. It includes state-of-art models for the description of the gaseous phase and an extension of a soot subgrid intermittency model to the sectional approach, originally proposed for the hybrid method of moments. Then, the LES is used to analyze a turbulent non-premixed ethylene-air jet diffusion flame and results are validated by available experimental data. The quality of results for the gaseous phase is satisfactory and results for the solid phase show a reasonable agreement with the experimental results in terms of localization, intermittency and soot volume fraction magnitude. Once the coupled LES-sectional approach validated, having access to the full information on the spatial and temporal evolution of the soot Particle Size Distribution (PSD), the second objective of this work is to provide a new fundamental insight on soot production in turbulent non-premixed flames. First, it is observed that a one-peak and a two-peak PSD shapes are observed at the bottom and downstream of the flame, respectively. Second, high fluctuations of the PSD distribution is observed all along the flame. In particular, a time bimodal behavior is observed with the presence of a zone with regular transitions between one-and two-peak PSD shapes. By analyzing soot particles Lagrangian paths, these high fluctuations are shown to be linked with the wide range of history paths of soot particles, which are mainly driven by turbulence.