The aim of this work is to investigate the differences of effective connectivity of the default mode network (DMN) in Alzheimer's disease (AD) patients and normal controls (NC). The technique of independent component analysis (ICA) was applied to identify DMN components and multivariate Granger causality analysis (mGCA) was used to explore an effective connectivity pattern. We found that: (i) connections in AD were decreased than those in NC, in terms of intensity and quantity. Posterior cingulated cortex (PCC) exhibited significant activity in NC as it connected with most of the other regions within the DMN. Besides, the PCC was the convergence center which only received interactions from other regions; (ii) right inferior temporal cortex (rITC) in the NC exhibited stronger interactions with other regions within the DMN compared with AD patients; and (iii) interactions between medial prefrontal cortex (MPFC) and bilateral inferior parietal cortex (IPC) in the NC were weaker than those in AD patients. These findings may implicate a brain dysfunction in AD patients and reveal more pathophysiological characteristics of AD.