Alzheimer's disease (AD) is a neurodegenerative disease characterized by progressive loss of memory. Impairment of working memory was typically observed in AD. The concept of brain functional connectivity plays an important role in neuroscience as a useful tool to understand the organized behavior of brain. Hence, the purpose of this study is to investigate the possible mechanism of working memory deficits in AD from a new perspective of functional connectivity. Rats were randomly divided into 2 groups: Aβ injection group (Aβ₁₋₄₂-induced toxicity rat model) and control group. Multi-channel local field potentials (LFPs) were obtained from rat prefrontal cortex with implanted microelectrode arrays while the rats performed a Y-maze working memory task. The short-time Fourier transform was utilized to analyze the power changes in LFPs and sub-bands (in particular theta and low gamma bands) were extracted via band filtering. Then the Directed transfer function (DTF) method was applied to calculate the functional connections among LFPs. From the DTF calculation, the causal networks in the sub-bands were identified. DTFmean (mean of connectivity matrix elements) was used to quantify connection strength as well as global efficiency (Eglob) was calculated to quantitatively describe the efficient of information transfer in the network. Our results showed that both connection strength and efficient of information transfer increased during the working memory task in the control group; by contrast, there was no significantly change in the Aβ injection group. These findings could lead to improve the understanding of the mechanism of working memory deficits in AD.