Phasing problem in X-ray structure determination can be challenging. Several methods, such as Isomorphous Replacement and Molecule Replacement, are frequently used. But their success relies on the availability of either an isomorphous heavy-atom derivative or a high identity homologous model. In this thesis, a low resolution ab initio phasing method is proposed. A large number of trial phases value and their corresponding high density masks are generated and modified alternately in reciprocal and real space. Then, the output phase sets are averaged to give the estimated phases and figure of merit which are capable of capturing key feature of the molecules’ low resolution envelope by Fourier synthesis. Smoothed particle hydrodynamics is employed to animate the high density masks’ modification process. The method is tested and compared with Lunin’s connectivity-based phasing method which also takes advantage of geometric properties of high density masks.