Would the heating have to be done while exposed to the surface?
No, and it essentially never is. In any solid, the atoms have a certain probability of moving from place to place. That probability is
strongly dependent on the temperature, getting higher with higher temperature. When the temperature gets high enough for each atom to have a high probability of moving during the time that the material is exposed to that temperature, then most of the atomss
will move at some time or another, and many will move multiple times.
If we are looking at "simple accumulation" dating, like potassium-argon (beloved of creationists because it
can come out wrong), and the argon atoms that are produced by decay of potassium escape form the material, then the clock is reset to zero. This doesn't happen all that much in solids, 'cause the argon atoms have to move a really long distance (compared to their size) to get out of the material.
What happens more often (but not very often) is "homogenization". For example, in isochron dating we look at how isotopes of elements are
distributed within the material. In rubidium-strontium isochron dating, the material starts out (just after solidification) with the
ratio of
87Sr to
86Sr the same throughout the material, although the
amounts of the materials vary from place to place. As time goes on, the amount of
87Rb decreases due to radioactive decay to
87Sr, and (of course) the amount of
87Sr increases, but the amount of
86Sr stays the same; so the ratio of
87Sr to
86Sr increases by different amounts in different parts of the material
because the produced 87Sr atoms are trapped in place, and we can use this to obtain an age and an indication of the reliability of that age. But, if the material is hot enough, the produced
87Sr atoms can move around so the concentration of Sr is the same everywhere, and that zeros out the clock. Clear as mud? .
(That's much simplified but reasonably accurate).
And would the properties of the atmosphere matter?
Not really.
Do they notice xenoliths before testing the sample?
Almost always yes, although they can be tricky to spot. But they are often difficult to separate out even if you can see them.
Another creationist favorite dating episode is the KBS Tuff. A tuff is a particular product of a volcanic eruption, and the KBS tuff is named for the initials of the woman who first studied it. Initially (in 1969) it was dated at 2.9 million years by potassium-argon (K-Ar) dating (and this was pushing the envelope at the time, that's pretty young for successful K-Ar dating). But the pig fossils under it didn't seem right for that age, some other groups dated it and found younger ages, still other groups dated it and found the same older age, and a hominid fossil (skull KNM-ER 1470) was found under it. If the KBS tuff were really 2.9 million years old, then that hominid fossil was about 3 million years old, and that would push back the timeline of human evolution significantly. Eventually Ian McDougall (a Grand Old Man of radioisotope dating) and his group determined that the KBS Tuff contained significant xenoliths. The tuff itself is 1.8-1.9 million years old, the xenoliths are much older. The differing age results were a result of samples that contained differing amounts of difficult-to-see xenoliths. After McDougall's results were confirmed by other researchers and other methods, the KBS tuff episode was over in 1980. (Of course the creationists claim that the date was adjusted to fit the presumed age of the hominid fossil, but that's just silly; it was the pig fossils, well known and understood from all over Africa, that really signalled something was wrong, and the scientists beat at the problem until they figured out its source, established the mechanism, and replicated the results many times). There's a good discussion of this incident at
Claim CD031 and some abstracts of the final age determinations at
Fission track age of the KBS Tuff and associated hominid remains in northern Kenya,
K’Ar age estimate for the KBS Tuff, East Turkana, Kenya, and
KBS Tuff dating and geochronology of tuffaceous sediments in the Koobi Fora and Shungura Formations, East Africa.