Abstract: As a kind of common nature disaster, acid rain will not only pollute and damage the environment, but also trigger the failure of rock and soil mass such as slope collapse, while scholars study less on the latter. The failure of slope of mafic igneous/ultramafic igneous rocks is not unusual in the acid rain area, but the specific mechanism of instability remains to be studied by considering the acid rain effects. In this paper, gabbro samples from a cutting slope in Ya'an city of typical acid rain area were processed, and the simulated acid rain treatment of pH 3.0 in the laboratory was carried out. The shear strength parameters of structural plane of samples before and after the action of acid rain were compared by direct shear tests. Then the slope stability before and after the action of acid rain was studied and evaluated. The result shows that the slope stability decreases with the increase of acid rain duration. In order to obtain the internal reasons of acid rain significantly weakening the slope stability, samples before and after the action of acid rain were tested by X-ray Diffraction(XRD), Polarizing Microscope and Scanning Electron Microscope(SEM). The results indicate that the mechanism of acid rain weakening on the gabbro slope is mainly reflected in two aspects. Part of the chain silicate mineral(pyroxene) and the framework silicate mineral(calcic plagioclase) convert into layered silicate minerals(chlorite, illite and kaolinite) under acidic conditions, which is inferred to be one of the major causes of the decrease of the shear strength parameters. With soluble salts and part of clay minerals being gradually dissolved and removed out by acid rain, cracks gradually grow and propagate on the structural plane of samples. Newly formed cracks leading to the less effective shear area of structural plane probably also accelerates the reduction of strength. Conclusion of this paper may be taken as references for the stability analysis of rock slopes in acid rain area and their reinforcement design methods.