Pollution of the ocean by plastic litter has become a major environmental problem : when mismanaged, plastics enter the environment where they undergo degradation and fragmentation into microplastics that are now ubiquitous in all aquatic compartments. In addition to the fact that microplastics are impossible to remove from the marine environment, they are even more damaging than the macroscopic waste.The fate of polymers in the aquatic environment depends both on abiotic phenomena (UV, mechanical stress), and on biotic ones, due to the colonization of plastics by micro-organisms. A primary step for bio-degradation is the constitution of a biofilm and reduction of the polymer chain length via exo-enzymes produced by bacteria from the biofilm. Once polymer chains are short enough, they can be assimilated by bacteria. While abiotic phenomena lead to the damage and fragmentation of a polymer by oxidation and hydrolysis mechanisms, creation of structural defects and fracture propagation, it is generally admitted that only biotic phenomena will result into the complete bio-degradation of a polymer, i.e. its conversion into biomass, water and CO2. In the marine environment, many questions remain about the relative kinetics of abiotic and biotic degradation and their respective impact in terms of fragmentation. For instance, several papers have recently reported that the size distribution of particles collected in the ocean between 5mm and a few hundreds of microns, does not seem to correspond to a single-kinetic fragmentation process.We studied the enzymatic erosion process in semi-crystalline polymersto understand the potential fracture and fragments generation in relation to the formation of erosion patterns . Being specifically interested in the role of heterogeneities at the scale of a few nanometers to a few micrometers, we used a polymer of a given chemical composition and monitored its morphology through its change in crystallinity ratio, everything else remaining constant. We used the well-known model system PDLLA/proteinase KEnzymatic erosion kinetics were measured through weight loss experiments and erosion patterns were observed over time through atomic force microscopy (AFM) and SEM. In order to interpret the results, we combined a simple two-phase geometric erosion model with the well-known Michaelis-Menten model for enzymatic kinetics. Our geometric erosion model is based on the evolution of the erosion front with time induced by the erosion rate difference between crystalline and amorphous regions. This new model accounts very well for the experimental results and unexpectedly predicts that after a lag time, the final erosion rate will be the one of the fastest eroding phase. Moreover, we observed a morphology-dependent release of fragments, which the model is also able to predict. In particular, one observes the release of spherulites as long as they are smaller than a critical size determined in the model. Some important consequences relevant for the understanding of the formation of micro-plastics in the ocean can be drawn from these results.