I did a google a while back on the subject, found some interesting material. I think what you're asking about is what goes on in the media, right?
Here's what I understand of it; more knowledgeable members might have a few corrections in my narrative.
The "phase change" material is in a layer just next to the reflective aluminum plate. It's sensetive to temperate changes, but not so that any room temperature levels should alter it. When the material is heated to a certain target temperature and held there for a certain duration of time, it reverts to it's liquid crystal state, which makes it fairly transparent - the signature of a "1" - being reflective.
During writing, a higher temperature for a shorter duration changes this into a solid state, and it obstructs light, creating a zero.
Times are very short, but are much longer than what's required to "color" the ink on R media, which is one reason RW media is always slower.
Anyway, like a previous post said, the area representing the links to the directory are written with a "signature" indicating it's empty, and the directory section may be written over - but erasure is a series of "1"'s not "0"'s (though I suppose either would work).
Obviously the laser can be controlled to revert the material to liquid or solid states at will during a write operation, else it would have required a full scale erasure to be reusable. In either case, it would seem clear that an erased disc is not exactly the same as a unused one. It's likely that what was once a zero would have a tiny remnent of it's former self when converted into a 1, dependent on aim accuracy. This isn't enough to cause read failures (we hope), but perhaps enough that an "erased" disk might be recoverable in a lab (partially).