Mass bends spacetime, so a star passing almost exactly in front of a more distant star acts as a lens: the background star brightens smoothly over days or weeks, then fades as the alignment passes.
If the lens star has a planet, the planet's own gravity adds a second, much sharper spike lasting hours to a couple of days, riding on top of the main curve.
The catch is unforgiving: the alignment happens once and never again. There is no confirming observation, no second look. Everything must be measured while it is happening, which is why these events are watched by networks of telescopes spread across longitudes.
The arithmetic
Magnification
A(u) = (u² + 2) / (u √(u² + 4))
u is the separation of the two stars on the sky in Einstein-radius units. As u → 0 the background star can brighten a thousandfold.
Einstein radius
θE = √(4GM / c² · (1/DL − 1/DS))
Sets the scale of the whole event. The planet's spike appears when it sits near this radius, which for typical distances corresponds to a few AU.
What the spike gives you
q = Mp / M*
A ratio, not a mass. Converting it to kilograms needs the lens star's mass, which is often only estimated — so masses here carry large error bars.
Why this method reaches where others cannot
Typical event duration
~30 days for the star
the planet's spike: hours to 2 days
Distance probed
1–8 kpc toward the galactic bulge
thousands of light years, not dozens
Sensitivity sweet spot
1–10 AU from the lens star
the cold outer region other methods miss
This is the only technique that finds planets at Jupiter-and-beyond distances around distant stars, and the only one that detects free-floating planets bound to no star at all.
What it does well
Reaches planets thousands of light years away, across the galaxy
Sensitive to cold, wide-orbit planets that transits and wobbles both miss
Finds rogue planets — no host star required
Sensitive down to roughly Earth mass
Where it cannot help
Each event happens once and cannot be revisited or confirmed
Yields a mass ratio and a projected distance, little else
The host star is often too faint to study
Requires continuous monitoring of millions of stars to catch the alignments
Landmark discovery
OGLE-2005-BLG-390Lb — a cold ~5 Earth-mass planet, at the time the least massive found around a normal star.
Instruments
OGLE, MOA, KMTNet; Nancy Grace Roman Space Telescope will industrialise it
Archive name
Microlensing — the value in NASA's discoverymethod field