Résumé : Miombo woodlands are vast wooded savannas covering 2 million km2 in East and Southern Africa. The main feature of these landscapes with closed but not overly dense canopy is the dominance of Brachystegia, Isoberlina, and Julbernardia legume trees genera. The timing of the onset of such vegetation, mainly understood through pollen fossils remains unclear. Dated molecular phylogeny calibrated with fossils for the aforementioned genera has the potential to provide details regarding the origin of species that nowadays dominate the current woodlands and possibly of miombo vegetation as a whole. The Brachystegia is a taxonomically complex genus and is compared to the other aforementioned genera, rich in species with 21 savanna species and eight species in the African Guineo-Congolian rain forests. We aim through the thesis to identify the diversification history of Brachystegia using dated phylogenies.We first reconstruct the Brachystegia phylogeny using nearly full plastome sequences in addition to ribosomal DNA sequences. Both sequences were obtained using a genome- skimming approach. In plastid phylogeny, species represented by multiple specimens appear rarely monophyletic while plastid clades display strong geographical structuring, independently of the species. Ribosomal phylogeny conversely allowed to identify morphological clades, but the lack of DNA polymorphism prevents the reconstruction of a well-resolved Brachystegia phylogeny. The strong spatial structure detected in plastid phylogeny suggests hybridization among the different species leading to recurrent chloroplast captures. Plastomes proved very informative for tracking the past dynamics of the genus and suggest a historical westwards expansion of miombo Brachystegia during the Plio-Pleistocene. We subsequently reconstructed the evolutionary history of the genus using targeted enrichment sequencing. Phylogenetic inferences were conducted using supermatrix and summary-method approaches on a dataset encompassing around 200 individuals loci for more than 200,000 base pairs. Opposite to previous reconstructions, most species appear as monophyletic groups even if high levels of gene tree conflict between the species trees and individual gene trees are reported, suggesting either incomplete lineage sorting (ILS) and/or reticulate evolution. Introgressed plastomes, and signature of ILS and reticulation for nuclear genes when species are nevertheless relatively well delineated by nuclear genome support to some extent that Brachystegia may behave a group of interfertile but still relatively well-delineated species (i.e. syngameon). Molecular dating analysis supports a Pliocene origin for the genus, with most of the diversification events occurring during the Plio-Pleistocene. We also conducted preliminary investigations to explore the potential of the genomics approaches used in this thesis to delineate problematic species or to reconstruct the past spatial dynamic of Brachystegia in current miombo regions.Overall, through this thesis, we clarified many aspects of the taxonomically complex Brachystegia genus. Genomic data support hybridization and plastid introgression on large spatial scales, giving credit to a Brachystegia syngameon that remains yet to be furthered characterised and validated. Moreover, results indicate a fairly recent origin of dominant species of the miombo congruently with their spatial expansion documented by plastid data, giving possibly insights into the temporal and spatial evolution of the miombo woodlands