A method of far-field super-resolution infrared microscopy was presented by using vibrational sum-frequency generation(VSFG) and donut-beam illumination. To achieve this, one Gaussian-shaped visible beam and one donut-shaped visible beam with different wavelengths were combined with an infrared beam coaxially to excite the sample. When the frequency of the infrared light was as the same as the resonant frequency of the molecules, the molecules absorbed the energy of the infrared photon and were excited to the vibrational excited state. The photons in the donut-shaped and the Gaussian-shaped visible beams both interacted with the excited molecules, and generated useful and useless SFG signal respectively. Simulations based on the vectorial field of the three beams and rate equations demonstrated that, when the visible intensity was improved to a certain level, the SFG signal tended to be saturated, then the donut-shaped visible photons and the Gaussian-shaped photons competed with each other. By increasing the photon flux density of the donut-shaped visible light to be larger than the saturated value, and reducing the photon flux density of the Gaussian-shaped visible light, the useful SFG signal in the donut-shaped area was surpressed effectively, which means the effective PSF was shrinked. With an objective which has a small numerical aperture (NA) 0.6, a simulated resolution as high as 56 nm was obtained.