Segregation controls many microstructure phenomena and is the main mechanism for solute based grain boundary engineering. This includes grain boundary chemistry, grain growth, grain boundary strengthening, and recrystallization phenomena. Although numerous studies computed segregation energies for different matrix-solute pairs, there exists no consistent data set covering all technically relevant combinations, because most studies focus on selected matrix-solute systems only. This poses a significant obstacle to grain boundary engineering approaches.

In the present work, we present the first consistent ab initio data set for a large set of matrices. The data set includes grain boundary segregation energies, grain boundary strengthening energies, and surface segregation energies in each matrix for all transition metals and many s and p elements as solutes. For each matrix, a representative grain boundary and surface were chosen and the segregation to multiple sites at the interface calculated, which allows to observe general trends between different matrices. The obtained data set is compared to a large bulk of literature data and phenomenological models resulting in a general agreement but highlighting also the importance of a consistent data set. In a next step, we subjected the data to machine learning approaches, which gives further insights into how this already vast data set can be extended to all possible matrix-solute combinations. The data is made available within the SEGROcalc software that offers already tools for grain boundary engineering.