Fracture is the process of crack propagation. When fracture occurs in a layer of snow underneath a slab sitting on a steep slope, a slab avalanche will occur.
How Snow Fails and fractures:

Avalanches don't "strike without warning", as we so often read in the press. They are only the most spectacularly visible event in a long series of precursors leading up to the grand finale.

It all begins many hours--or even days--before, usually when new snow or wind-blown snow begins to pile weight on top of a buried weak layer. Added weight causes the underlying snow to deform; rapidly added weight causes snow to rapidly deform. On an inclined slope, the deformation tends to concentrate within buried weak layers in the form of shear.
Inside of a weak-layer under stress, we can think of this as a race between bonds being broken and bonds being re-formed. Let's look at three different rates of deformation, slow medium and fast:

Slow deformation rate
If the weak-layer deforms slowly, it either deforms the bonds between the ice grains or more bonds form than break. This means that the weak-layer adjust to its load and actually gains strength. Snow can lazily drape over the terrain like a cat draped over the back of the couch—like a limp rubber band—and if you’ve every tried to cut a limp rubber band with a knife, you know what a stable snowpack is like.

Medium deformation rate
With an increasing rate of deformation, we reach a point where nearly as many crystalline bonds break as form and the strength of the weak layer remains about the same. With sensitive microphones we can actually hear the rupture of individual bonds between the ice grains, like the sound of slowly ripping Velcro.

Rapid deformation rate
If deformation occurs too rapidly--past a critical threshold--then more bonds break than form. The weak-layer inexorably looses strength and begins the slippery slide towards disaster. We call this "failure"--when the snow begins to progressively loose strength. We also call this "strain softening." To understand failure and strain softening, do this experiment: Take a paper clip and bend it in the same place repeatedly, and after about ten bends you'll notice that it is getting weaker (failure) and after about 15 bends, it snaps right off (fracture). Got that, the difference between failure and fracture?

Having said this, scientists still don't know exactly how avalanches fail and fracture because snow is such a devilishly difficult substance to study. First, large variations commonly exist over both distance and time and second, as you can imagine, catching a natural avalanche in the act is stupendously difficult and dangerous, or as Monty Atwater put it, "an occupation something like trailing a wounded African buffalo." So ten or twenty years from now, the following paragraphs--like so many of the "facts" we believe about avalanches--may seem like yet more quaint, geezer-ramblings. Nevertheless, this is what scientists generally believe about how avalanches fail and fracture.

Failure occurs slowly at perhaps centimeters per hour; whereas, fracture occurs catastrophically and has been measured at around 20 meters per second. Whamo! The slope shatters like glass.

When a person triggers an avalanche, it means that they have found a trigger point of the avalanche. Perhaps it’s a place where the slab is thinner allowing more of the victim’s weight to tickle the weak layer. Perhaps it’s a place where the weak layer is more poorly bonded than the rest of the slope. I don’t think anyone knows for sure. But we do know that snow is very sensitive to the rate at which it is deformed and the extremely rapid deformation caused by the weight of a person is exactly the kind of thump needed to intiate the fracture process.

Without this final trigger, unstable slopes can teeter on the brink of disaster for quite some time, giving us the illusion that all is well. After a storm, we never know how many slopes would come down if they just had a proper trigger. But with enough good snowmobiles and skilled riders, we can certainly find out. As avalanche researcher Rand Decker likes to say, "Avalanches are like a bar room brawl. No matter how much tension you have, you need to give somebody a thump to get things going."

Additional Terms:
Anchors Hard Slab Avalanche Slide
Aspect High Danger Sluff
Avalanche High Marking Snowpit
Avalanche Path Isothermal Soft Slab Avalanche
Avalanche Transceiver Layer, Snow Stability
Bed Surface Leeward Stability Test
Collapse Loading Starting Zone
Concave Slope Loose Snow Avalanche Stepping Down
Considerable Danger Low Avalanche Hazard Sun Crust
Convex Slope Melt-Freeze Snow Surface Hoar
Cornice Metamorphism, Snow Sympathetic Trigger
Corn Snow Moderate Danger Temperature Gradient
Couloir Persistent Weak Layers Terrain Trap
Cross Loading Point-Release Track
Crown Face Probe Trigger
Danger Ratings Propagation Trigger Point
Deep Slab Avalanche Rain Crust Upside-Down Storm
Density, Snow Remote Trigger Weak Layer
Depth Hoar Rime Weak Interface
Dry Snow Avalanche Runout Zone Wet Snow Avalanche
Extreme Danger Sastrugi Windward
Faceted Snow Settlement Wind Loading
Fracture Ski or Slope Cut Wind Slab
Glide Skinning, Skin Track Whumpf
Graupel Slab