There is a good treatment of the subject in the Wikipedia article on RTGs. I've talked with people at the Nuclear & Emerging Technologies for Space (NETS) conferences who have given me other useful tidbits. Some background helps understand why so few nations have undertaken the task of producing RTGs, and for those who have, why they might choose a particular isotope.
There are some very useful properties of candidate isotopes for RTGs:
- The half-life must be both long enough to provide useful power for a reasonable amount of time, but short enough that the specific power (thermal power per mass of isotope) is reasonably high
- Alpha decay is preferable to beta decay because you get much more energy per decay from alpha decay (typically a factor of ~10), and you need less shielding because the alpha particles are less penetrating and generate less gamma from bremsstrahlung
- Low neutron emission rates and energies are preferable, either by the primary decay or by emission from the decay chain (decay of the sequence of daughter products from the primary decay)
For space applications plutonium 238 seems to be the best fit, at least for long-duration missions. But because producing specific isotopes in useful quantities generally requires reactors that produce the proper feedstock (such as neptunium 237) and a reactor with a particular neutron energy spectrum to transmute, say, Np to Pu (Np237 + n -> Np238; Np238 beta-decays to Pu238), sometimes producing the best isotope is very expensive (such as, designing and building a dedicated reactor) and institutions will opt for something else. For instance, early RTGs used polonium 210, and the Soviets even used strontium 90, a beta-decay isotope.
As mentioned in Hobbes's answer to this question, the Soviets launched multiple fission reactors into Earth orbit, but much less is known about their RTGs used in space. The Chinese have also joined the club.
ESA currently is studying production of a European space RTG and is considering Americium 241, with a lower specific power than Pu238. Building the infrastructure to produce RTGs is a long process, so it might take a decade or more, but they might be joining the club. http://www.unoosa.org/pdf/pres/stsc2012/tech-18E.pdf