More specific types of horizon include the related but distinct absolute and apparent horizons found around a black hole.  Still other distinct notions include the Cauchy and Killing horizons; the photon spheres and ergospheres of the Kerr solution; particle and cosmological horizons relevant to cosmology; and isolated and dynamical horizons important in current black hole research.
Event horizon of a black hole

Far away from the black hole a particle can move in any direction. It is only restricted by the speed of light.
Closer to the black hole spacetime starts to deform. In some convenient coordinate systems, there are more paths going towards the black hole than paths moving away.

Inside the event horizon all paths bring the particle closer to the center of the black hole. It is no longer possible for the particle to escape.
One of the best-known examples of an event horizon derives from general relativity's description of a black hole, a celestial object so massive that no nearby matter or radiation can escape its gravitational field.  Often,  this is described as the boundary within which the black hole's escape velocity is greater than the speed of light.  However,  a more accurate description is that within this horizon, all lightlike paths (paths that light could take) and hence all paths in the forward light cones
 of particles within the horizon, are warped so as to fall farther into 
the hole. Once a particle is inside the horizon, moving into the hole is
 as inevitable as moving forward in time, and can actually be thought of
 as equivalent to doing so, depending on the spacetime coordinate system
 used.
The surface at the Schwarzschild radius acts as an event horizon in a non-rotating body that fits inside this radius (although a rotating black hole
 operates slightly differently). The Schwarzschild radius of an object 
is proportional to its mass. Theoretically, any amount of matter will 
become a black hole if compressed into a space that fits within its 
corresponding Schwarzschild radius. For the mass of the Sun this radius is approximately 3 kilometers and for the Earth
 it is about 9 millimeters. In practice, however, neither the Earth nor 
the Sun has the necessary mass and therefore the necessary gravitational
 force, to overcome electron and neutron degeneracy pressure. The minimal mass required for a star to be able to collapse beyond these pressures is the Tolman–Oppenheimer–Volkoff limit, which is approximately three solar masses.
Black hole event horizons are widely misunderstood. Common, 
although erroneous, is the notion that black holes "vacuum up" material 
in their neighborhood, where in fact they are no more capable of seeking
 out material to consume than any other gravitational attractor. As with
 any mass in the universe, matter must come within its gravitational 
scope for the possibility to exist of capture or consolidation with any 
other mass. Equally common is the idea that matter can be observed 
falling into a black hole. This is not possible. Astronomers can detect 
only accretion disks
 around black holes, where material moves with such speed that friction 
creates high-energy radiation which can be detected (similarly, some 
matter from these accretion disks is forced out along the axes of spin 
of the black hole, creating visible jets when these streams interact 
with matter such as interstellar gas or when they happen to be aimed 
directly at Earth). Furthermore, a distant observer will never actually 
see something reach the horizon. Instead, while approaching the hole, 
the object will seem to go ever more slowly, while any light it emits 
will be further and further redshifted.
Cosmic event horizon
In cosmology, the event horizon of the observable universe  is the largest comoving distance from which light emitted now can ever reach the observer in the future. This differs from the concept of particle horizon, which  represents the largest comoving distance from which light emitted in the past
 could have reached the observer at a given time. For events beyond that
 distance, light has not had time to reach our location, even if it were
 emitted at the time the universe began. How the particle horizon 
changes with time depends on the nature of the expansion of the universe.
 If the expansion has certain characteristics, there are parts of the 
universe that will never be observable, no matter how long the observer 
waits for light from those regions to arrive. The boundary past which 
events cannot ever be observed is an event horizon, and it represents 
the maximum extent of the particle horizon. 
The criterion for determining whether a particle horizon for the universe exists is as follows. Define a comoving distance dp as
In this equation, a is the scale factor, c is the speed of light, and t0 is the age of the Universe. If dp → ∞ (i.e., points arbitrarily as far away as can be observed), then no event horizon exists. If dp ≠ ∞, a horizon is present. 
Examples of cosmological models without an event horizon are universes dominated by matter or by radiation. An example of a cosmological model with an event horizon is a universe dominated by the cosmological constant (a de Sitter universe). 
A calculation of the speeds of the cosmological event and particle horizons was given in a paper on the FLRW cosmological model, approximating the Universe as composed of non-interacting constituents, each one being a perfect fluid.
Apparent horizon of an accelerated particle
Spacetime diagram showing a uniformly accelerated particle, P, and an event E that is outside the particle's apparent horizon. The event's forward light cone never intersects the particle's world line.
If a particle is moving at a constant velocity in a non-expanding 
universe free of gravitational fields, any event that occurs in that 
Universe will eventually be observable by the particle, because the 
forward light cones from these events intersect the particle's world line.
 On the other hand, if the particle is accelerating, in some situations 
light cones from some events never intersect the particle's world line. 
Under these conditions, an apparent horizon is present in the particle's (accelerating) reference frame, representing a boundary beyond which events are unobservable. 
For example, this occurs with a uniformly accelerated particle. A spacetime diagram of this situation is shown in the figure to the right. As the particle accelerates, it approaches, but never reaches, the speed of light with respect to its original reference frame. On the spacetime diagram, its path is a hyperbola, which asymptotically approaches
 a 45-degree line (the path of a light ray). An event whose light cone's
 edge is this asymptote or is farther away than this asymptote can never
 be observed by the accelerating particle. In the particle's reference 
frame, there appears to be a boundary behind it from which no signals 
can escape (an apparent horizon).
While approximations of this type of situation can occur in the real world (in particle accelerators,
 for example), a true event horizon is never present, as this requires 
the particle to be accelerated indefinitely (requiring arbitrarily large
 amounts of energy and an arbitrarily large apparatus).
Interacting with an event horizon
A misconception concerning event horizons, especially black hole
 event horizons, is that they represent an immutable surface that 
destroys objects that approach them. In practice, all event horizons 
appear to be some distance away from any observer, and objects sent 
towards an event horizon never appear to cross it from the sending 
observer's point of view (as the horizon-crossing event's light cone never intersects the observer's world line).
 Attempting to make an object near the horizon remain stationary with 
respect to an observer requires applying a force whose magnitude 
increases unboundedly (becoming infinite) the closer it gets. 
In the case of a horizon perceived by a uniformly accelerating 
observer in empty space, the horizon seems to remain a fixed distance 
from the observer no matter how its surroundings move. Varying the 
observer's acceleration may cause the horizon to appear to move over 
time, or may prevent an event horizon from existing, depending on the 
acceleration function chosen. The observer never touches the horizon and
 never passes a location where it appeared to be. 
In the case of a horizon perceived by an occupant of a de Sitter universe, the horizon always appears to be a fixed distance away for a non-accelerating observer. It is never contacted, even by an accelerating observer. 
In the case of the horizon around a black hole, observers 
stationary with respect to a distant object will all agree on where the 
horizon is. While this seems to allow an observer lowered towards the 
hole on a rope (or rod) to contact the horizon, in practice this cannot 
be done. The proper distance to the horizon is finite,
 so the length of rope needed would be finite as well, but if the rope 
were lowered slowly (so that each point on the rope was approximately at
 rest in Schwarzschild coordinates), the proper acceleration (G-force)
 experienced by points on the rope closer and closer to the horizon 
would approach infinity, so the rope would be torn apart. If the rope is
 lowered quickly (perhaps even in freefall),
 then indeed the observer at the bottom of the rope can touch and even 
cross the event horizon. But once this happens it is impossible to pull 
the bottom of rope back out of the event horizon, since if the rope is 
pulled taut, the forces along the rope increase without bound as they 
approach the event horizon and at some point the rope must break. 
Furthermore, the break must occur not at the event horizon, but at a 
point where the second observer can observe it.
Observers crossing a black hole event horizon can calculate the 
moment they have crossed it, but will not actually see or feel anything 
special happen at that moment. In terms of visual appearance, observers 
who fall into the hole perceive the black region constituting the 
horizon as lying at some apparent distance below them, and never 
experience crossing this visual horizon.
 Other objects that had entered the horizon along the same radial path 
but at an earlier time would appear below the observer but still above 
the visual position of the horizon, and if they had fallen in recently 
enough the observer could exchange messages with them before either one 
was destroyed by the gravitational singularity. Increasing tidal forces
 (and eventual impact with the hole's singularity) are the only locally 
noticeable effects. Tidal forces are a function of the mass of the black
 hole. In realistic stellar black holes, spaghettification occurs early: tidal forces tear materials apart well before the event horizon. However, in supermassive black holes,
 which are found in centers of galaxies, spaghettification occurs inside
 the event horizon. A human astronaut would survive the fall through an 
event horizon only in a black hole with a mass of approximately 10,000 solar masses or greater.
Beyond general relativity
The description of event horizons given by general relativity
 is thought to be incomplete. When the conditions under which event 
horizons occur are modeled using a more comprehensive picture of the way
 the Universe works, that includes both relativity and quantum mechanics, event horizons are expected to have properties that are different from those predicted using general relativity alone.
At present, it is expected that the primary impact of quantum effects is for event horizons to possess a temperature and so emit radiation. For black holes, this manifests as Hawking radiation, and the larger question of how the black hole possesses a temperature is part of the topic of black hole thermodynamics. For accelerating particles, this manifests as the Unruh effect, which causes space around the particle to appear to be filled with matter and radiation.
According to the controversial black hole firewall hypothesis, matter falling into a black hole would be burned to a crisp by a high energy "firewall" at the event horizon.
An alternative is provided by the complementarity principle,
 according to which, in the chart of the far observer, infalling matter 
is thermalized at the horizon and reemitted as Hawking radiation, while 
in the chart of an infalling observer matter continues undisturbed 
through the inner region and is destroyed at the singularity. This 
hypothesis does not violate the no-cloning theorem
 as there is a single copy of the information according to any given 
observer. Black hole complementarity is actually suggested by the 
scaling laws of strings
 approaching the event horizon, suggesting that in the Schwarzschild 
chart they stretch to cover the horizon and thermalize into a Planck length-thick membrane.
A complete description of event horizons is expected to, at minimum, require a theory of quantum gravity. One such candidate theory is M-theory.  Another such candidate theory is loop quantum gravity.

