On the topic of erosion, we actually have a lot of good data from nuclear weapon testing programs, and it is actually surprising to the intuition how little erosion occurs, but upon reflection, the thermal pulse is so quick that even if the surface of the plate reaches melting temperatures, the heat is immediately conducted away (provided you have enough thermal mass behind it) and very little actual erosion or evaporation occurs, especially in the space environment where the thermal pulse is not amplified and mediated by an atmosphere. The basic feasibility of pulse production were validated by the use of subscale models (using conventional chemical explosives) and with data from nuclear weapons performance and effects testing. The capability promoted by enthusiasts (including Dyson himself) to achieve performance in the I[SUB]sp[/SUB] > 20,000 seconds area is a little dubious, but would certainly be a feasible proposal for interplanetary propulsion of large vehicles.
No argument there (I don’t think we’re really disagreeing at all). I suppose my point is this: one might look at the radiators on the ISS and prematurely conclude that we’ll never have a nuclear reactor in space. The EATCS (one of the ISS radiator systems), for instance, dissipates 70 kW and while I can’t find figures on the area, it clearly must be hundreds of square meters. How could you ever build radiators big enough to dump the heat from a 10 MW reactor?
The answer, of course, is that the ISS radiators are at more or less room temperature, while any nuclear reactor will be much hotter. If you run at 1300 K instead of 300 K, you dissipate heat at 350 times the rate due to the T[sup]4[/sup] factor in the Stefan–Boltzmann law. If you can run at 2000 K, you dissipate heat at nearly 2000 times the rate.
These high temperatures are only possible for a nuclear system in the first place. There is also a range of reactor designs with different natural temperatures. The “nuclear lightbulb”, for instance, would have temperatures in the 25,000 K range. Even if the “cool” side was 3,000 K, a tiny radiator could dissipate a fantastic amount of heat.
So I posit that the real limiting factor is materials and engineering. The Stefan–Boltzmann law is a theoretical limit and thus can’t be worked around, but at the same time you can always make it hotter and increase the rate. Making it hotter is a hard problem, though (even ignoring a heat pump cycle).