The local electronic structure and magnetic properties of GaAs doped with 3d transition metal (Sc, Ti, V, Cr, Mn , Fe, Co, Ni) were studied by using discrete variational method (DVM) based on density functional theory. The calculated result indicated that the magnetic moment of transition metal increases first and then decreases, and reaches the maximum value when Mn is doped into GaAs. In the case of Mn concentration of 1.4%, the magnetic moment of Mn is in good agreement with the experimental result. The coupling between impure atoms in the system with two impure atoms was found to have obvious variation. For different transition metal, the coupling between the impure atom and the nearest neighbor As also has different variation.diluted magnetic semiconductor, density functional theory, 3d transition metal, cluster model, electronic structure Modern electronic devices utilize the charge degree of freedom of electrons to process information in semiconductors and the spin degree of freedom to store information in magnetic materials.If both the charge and spin of electrons can be combined, we may be able to accomplish mass storage and information processing at the same time. In this way the performance of electronic devices will be enhanced. In general, semiconductors used for integrated circuits, such as GaAs and GaN, do not contain magnetic ions and are nonmagnetic. Some magnetic elements can be doped into non-magnetic semiconductors (GaN, GaAs) to make them magnetic. This category of semiconductors is called diluted magnetic semiconductors (DMS).DMS has been thought to be of scientific and technological importance in exploring new physics and new functionality toward future electronics. In 1998, Ohno [1] and co-workers successfully introduced Mn into nonmagnetic GaAs host, and they found this material had a Curie temperature as high as 110 K. Their work has aroused intense interest in the study of DMS.In this research, the local electronic structure and magnetic properties of GaAs doped with 3d transitional metals were studied by using cluster models, and the calculated results were compared