cryogen

Hiding IR signatures is a lot easier than has been suggested. You don't need to reduce thermal radiation; only emit it in the opposite direction as the observer. You can do this passively. One way is by choosing surface materials with different emissivities (i.e. "black" on the radiating side, "white" on the stealth side). Another way is to stack layers of surfaces, with vacuum in between. At each surface, thermal radiation is as likely to go backwards as forwards, so heat flow through the whole stack is greatly reduced (by a factor of 1/N). This is the idea behind multi-layer insulation (the gold foil on spacecraft), and on a larger scale, the "sunshield" on the James Webb telescope: http://jwst.nasa.gov/sunshield.html The cold side is 40 Kelvin! With enough layers, you could probably vanish into the CMB noise. Reflected radiation (like radar) is particularly bad at interplanetary distances, since reflected signals fall off as 1/R4, not merely 1/R2.

You are right! The smaller ones are Falcon Heavy, not SLS. That adds up to about 8*50t + 2*130t = 560t in LEO.

It's not an SSTO actually. From page 30 of the PDF, they design for a two-stage ascent rocket, fueled by RP-1/LOX. The mass cost is 63 tons to lift a 2.2 ton crew capsule, and is the majority of the mass on Venus (96 tons). (The atmospheric habitat is only 5.1 tons. And most of the dirigible mass is (presumably) only there for lifting the ascent rocket. So it's really expensive. They want 10 SLS cargo launches (including some Block 2) to carry everything there).

I found out that Zubrin looked at NTR propellants other than hydrogen. His idea (1991) was to land an NTR on the surface of Mars, and use the NTR as an ascent rocket using Martian fuel (CO2). (NASA's modern concept is to bring a refrigerated zero-boil-off LH2 tank to Mars, and leave that in Mars orbit for the return trip. The surface ascent would be CH4/LOX, the CH4 from Martian ISRU). Some of the theoretical Isp's Zubrin cites are (at 2800 K): 283 s for CO2, 370 s for H2O, and 606 s for CH4. (So in particular, CH4 can be much better than chemical LH2/LOX at 450 s, but still worse than hydrogen NTR at ~900 s). Nuclear Rocket Using Indigenous Martian Fuel (Zubrin, 1991) Mars Design Reference Architecture 5.0 (NASA, 2009)

You're talking about nuclear thermal rockets specifically. Actually, there's no point in using any propellant other than LH2 -- that's the only reason they have high Isp. Small molecules move faster than heavy ones -- the speed scales as 1/sqrt(molecule mass) according to equipartition. This translates to exhaust velocity (Isp). NTR's propellant is H2, with a mass of 2 amu: that gives it its edge over chemical LH2/LOX rockets, whose propellant stream is mainly H2O with mass 18 amu. NTR's don't run hotter than chemical rockets, so with a heavier propellant (like CH4), the Isp advantage probably wouldn't be there. Barring exotic ideas like gas-core reactors, they're limited by the temperature of the solid reactor fuel, that's acting as a heat exchanger. It's actually a soluble problem though (LH2 boiloff): you can carry along a refrigeration system to keep the LH2 below boiling. In NASA's bimodal NTR concept, it'd cost about 1 ton of mass and 10 kWe power. If you're building large enough rockets, LH2 is a storable propellant. (There's already precedent for this: Hubble has its own cryogenic cooler. It has this ridiculously tiny gas turbine that keeps liquid neon below 70 K, to keep thermal noise out of the infrared camera).