Gravitational microlensing relies on chance events where from our viewpoint, one star passes in front of another star. The farther star is usually a bright star, and the near one is normally one we couldn't ordinarily see from Earth. When it passes in front of the farther star, however, its gravity causes the light from the farther star to bend and the star is magnified from our point of view. If, during the event, the background star appears to be magnified even more for a short time, that means a planet orbiting the smaller star is increasing the effect of the magnification.
For more, please see our Gravitational Microlensing Observing Program.
Einstein predicted that the gravitational field of any massive star will act as a gravitational lens and bend the path followed by the light rays originating from any bright star that happens to pass behind the lens. The effect of lensing at cosmological distances is practically observed as multiple distorted images of the background star around the edge of the gravitational influence of the lensing star. However, lensing also occurs on smaller scales in our galaxy and then the resulting images cannot be individually resolved. We call this phenomenon microlensing. What we see in this case instead, is a brightening of the background star that can last from a few days to several weeks. Then the star fades back to it's normal brightness. If the lensing star hosts a planetary companion, there is a chance that the planet can also act as a mini-lens and thereby reveal its presence.
Microlensing is unique in its capability to rapidly survey the population of cold planets, with a sensitivity to planetary mass that goes down to just below the mass of the Earth. The population of stars that it surveys are low-mass stars, typically M-dwarfs, between here and the centre of the Galaxy. Other methods are capable of detecting planets up to a few hundred light years away but microlensing is the only method that can probe the galactic population of planets. The planets discovered by this method are typically located between 0.6 and 6 AU from the host star, which corresponds to a cold zone that is more conducive to planet formation and which nicely overlaps the colder outer edge of the Habitable Zone. This region of parameter space is still largely inaccessible to other methods.
At least 80 planets have been discovered by this method (as of October 2018). Of these planets, most are Jupiter-analogs, but a few have masses comparable to that of Neptune and below. Theoretical predictions estimate that small, cold planets are abundant and these can be detected by microlensing surveys. Microlensing is also sentitive to multiple planet systems and free-floating planets.