A nuclear salt-water rocket (or NSWR) is a proposed type of nuclear thermal rocket designed by Robert Zubrin that would be fueled by water bearing dissolved salts of Plutonium or U235. These would be stored in tanks that would prevent a critical mass from forming by some combination of geometry or neutron absorption (for example: long tubes made out of Boron in an array with considerable spacing between tubes). Thrust would be generated by nuclear fission reactions from the nuclear salts heating the water and being expelled through a nozzle. The water would serve as both a neutron moderator and propellant.
In a conventional chemical rocket, chemical reactions of the fuel and oxidizer (e.g. Oxygen and Kerosene) heat the byproducts of the chemical reaction (e.g. CO2 and H2O) to high temperatures as they are forced through a rocket nozzle. The fast moving molecules in the exhaust focused in one direction create thrust. In a nuclear thermal rocket (or NTR) a nuclear fission reactor would serve as a source of heat which would be transferred to a propellant that is then exhausted through a rocket nozzle. The propellant in this case can be any material with suitable properties, it need not react during the operation of the rocket, it is simply a source of mass to be heated up and exhausted out of the rocket at high speeds. In an NSWR the nuclear salt-water would be made to flow through a reaction chamber and out an exhaust nozzle in such a way and at such speeds that the peak neutron flux in the fission reaction would occur outside of the vehicle. This has several advantages relative to conventional NTR designs. Because the peak neutron flux and fission reaction rates would occur outside of the vehicle, these activities could be much more vigorous than they could be if it was necessary to house them in a vessel (which would have temperature limits due to materials constraints). Additionally, a contained reactor can only allow a small percentage of its fuel to undergo fission at any given time, otherwise it would overheat and meltdown (or explode in a runaway fission chain reaction). Because the fission reaction in an NSWR is dynamic and because the reaction products are exhausted into space it doesn't have a limit on the proportion of fission fuel that reacts. In many ways this makes NSWRs like a hybrid between fission reactors and fission bombs.
Because of their ability to harness the power of what is essentially a continuous nuclear fission explosion, NSWRs would have both very high thrust and very high exhaust velocity, a rare combination of traits in the rocket world, meaning that the rocket would be able to accelerate quickly as well as extremely efficient in terms of propellant usage. One design would generate 13 meganewtons of thrust at 66 km/s exhaust velocity (compared to ~4.5 km/s exhaust velocity for the best chemical rockets of today). Another design would achieve much higher exhaust velocities (4,700 km/s) and use 2700 tonnes of highly enriched Uranium salts in water to propel a 300 tonne spacecraft up to 3.6% of the speed of light.
NSWRs share many of the features of Orion propulsion systems, except that NSWRs would generate continuous rather than pulsed thrust and may be workable on much smaller scales than the smallest feasible Orion designs (which are generally large, due to the requirements of the shock-absorber system and the minimum size of efficient nuclear explosives).
The vessel's exhaust would contain radioactive isotopes, but these would be rapidly dispersed after traveling only a short distance; the exhaust would also be traveling at high speed (in Zubrin's scenario, faster than Solar escape velocity, allowing it to eventually leave the Solar System). This is however, little use on the surface of a planet, where a NSWR would eject massive quantities of superheated steam, still containing fissioning nuclear salts. Needless to say, this is extremely dangerous from an ecological point of view, and such engines could not be used on the surface of a planet without terrible environmental damage occurring.