Subduction is not the end
all-geo.org
all-geo.org
http://en.m.wikipedia.org/wiki/Isotopic_dating
By "formed" I guess this means "cooled down" so it's not magma or lava anymore.
But how do they know the initial isotope proportion? When an "old" rock with some ratio of isotopes melts and becomes magma or lava again, why does this "reset" the proportion of its isotopes? To put it another way: the isotope will have also been decaying when it was part of the magma wouldn't it? Why don't new rocks have the same isotopic ratios as old rocks?
In a nutshell, it's because we choose minerals whose crystal structure can only bond with the original element and not with it's "daughter" that's produced after radioactive decay.
For example, Uranium/Lead dating of zircons is one of the most common methods (particularly for very old rocks). There's a reason we use zircons for uranium/lead dating, and not just any mineral that contains trace amounts uranium.
In zircon (ZrSiO4), uranium can (and does, in small amounts) substitute for zirconium in the crystal structure and bond with the silica tetrahedra. However, lead cannot bond with the silica tetrahedra in zirconium's place. Therefore, the zircon crystal structure "rejects" lead, and any lead in the crystal lattice at the present day must be the result of radioactive decay.
If we know the rate of radioactive decay of the uranium isotopes in question (there are two uranium isotopes that decay to two different lead isotopes), all we need to date the rock is the present-day ratio of a particular uranium isotope to the corresponding lead isotope.
In practice, you also exploit the fact that there are two uranium isotopes that independently decay to two different lead isotopes. This gives two independent dates, and disagreements between them can be use to test for contamination, etc. (There are also a few other checks I won't get into... I'm not a geochronologist, and each different isotopic system and instrument used to measure it has slightly different protocols. Reality is messy, but this gives the basic picture.)
For other, longer-lived isotopes, keep in mind you are not just measuring the amount of radioactive material left, but are also measuring the amount of the daughter product mixed in some proportion with it. And the proportions in the material being dated might not be the same as what they are in the surrounding environment (in fact, can not be, if the dating method is going to work). As a simple example, you could date rocks with uranium and lead, since you would expect there to be more lead in old rocks because some of the uranium decayed to lead. However, if that rock melted and mixed with younger rocks, the proportion of lead and uranium would cease to be of any interest.
Banded iron formations are fascinating - while browsing around I noticed an example on the relevant Wikipedia page of a rock that has layers that are thought to correspond to a day-night-cycle. The rock is 3.5 billion years old!
The idea that we have things to correspond to individual days from 3.5 billion years ago is stunning.