Several groups of astrophysicists are searching for "Gravity Waves" to test our fundamental understanding of physics. The only gravity waves that we are likely to detect in the near future are those generated by rare, violent astronomical events: our best bet is the collision of two neutron stars. ATLAS cannot detect gravitational waves itself, but will be on the lookout for the expected flashes of light from such events, providing data that will complement those from gravitational wavel detector systems.
Einstein's General Theory of Relativity states that space and time are linked in a space-time "fabric" with "dents" where there are massive objects like stars.
Imagine a taut painter's canvas lying parallel to the ground with a cannonball in the center. Now imagine rolling another cannonball towards the first one and as they collide and bounce back and forth the canvas will bounce around as well and anything attached to the canvas will experience the effect of the cannonballs' collision. The animation to the left illustrates the rippling of space-time in two dimensions. Of course, space is 3-dimensional so it's a little more difficult to imagine.
Neutron stars are the remains of a supernova explosion in which the entire mass of a star like the Sun gets crushed into a volume no bigger than 12 km across (that's about 8 miles in diameter). A single teaspoon of neutron star material weighs as much as an entire mountain like Mt. Everest. Now imagine that two neutron stars collide (and coalesce) - they are our cannonballs and their collisions will create ripples in the fabric of space-time that may be detected by sensitive gravity wave detectors like the Laser Interferometer Gravitational Wave Observatory (LIGO), Advanced LIGO and the European Gravitational Observatory (EGO).
The ripples in space-time that are caused by the mind-blowing mergers of neutron stars travel vast distances before being detected even by Advanced LIGO. The typical merger will occur at distances on the order of 1.5 billion light years! That distance corresponds to seeing things 10% of the way to the edge of the observable universe! Having never actually seen a neutron star-neutron star merger taking place we rely on calculations and computer simulations to inform us how much energy would be released. Furthermore, we don't know how much of that energy will be emitted in the form of visible light rays. But if just 1 part in 100,000 of the energy appears in visible light then ATLAS will be able to just detect the typical events observed by Advanced LIGO. Of course, ATLAS will have no difficulty detecting the many events that are brighter than typical and these detections combined with the Advanced LIGO results will provide important constraints on the formation of gravitational waves and on the General Theory of Relativity.