I realize that I am both going against the post-1970 ORNL liquid-fluoride reactor work and the small amount of “conventional wisdom” on the topic by advocating two-fluid (separate blanket and core salts), but I find nuggets in some of these older documents that tell me I am on the right track.

This is from Reactor Physics and Fuel Cycle Analysis, a paper written by A.M. Perry and H.F. Bauman of ORNL in 1970.

On page 209:

We have previously given serious consideration to, a two-fluid reactor in which the fissile and fertile materials are carried in separate salt streams, the bred uranium being continuously stripped from the fertile stream by the fluoride volatility process. Blanket regions contain only the fertile salt, while the core contains both fissile and fertile streams; these streams must be kept separate by a material with a low-neutron cross section, that is, by the graphite moderator itself. This approach appears to yield the best nuclear performance, owing primarily to a combination of maximum blanket effectiveness and minimum fuel inventory. It also exhibits attractive safety characteristics because expansion of the fuel salt, upon heating, removes fissile material from the core while leaving the thorium concentration unchanged. The concept does, however, involve important questions regarding the reliability of the graphite “plumbing” in the core, the adequate proof of which may require a good deal of time and testing.

Yes, and I think that a new core geometry will be required to make the two-fluid scheme work. But consider the alternative:

On page 218:

Expansion of the single fissile-fertile salt in the one-fluid reactor reduces the density of most of the absorbing materials in the same proportion, so that one might expect a very small prompt temperature coefficient of reactivity. The density coefficient of the salt is in fact very small. In addition to this, however, there is both a positive contribution to the salt-temperature coefficient associated with the shift in thermal-neutron spectrum with increasing salt temperature, and a negative contribution associated with the Doppler broadening of resonance capture lines in thorium.

The latter predominates, resulting in a prompt negative coefficient of -2.4×10^-5 ?k/k/°C . The graphite moderator contributes a positive component to the overall temperature coefficient, attributable to an increase in the relative cross section of 233U with increasing neutron temperature. This results in an overall coefficient which is very small, though apparently negative, i.e., -0.5×10^-5 /°C.

The prompt negative salt coefficient will largely govern the response of the reactor for transients whose periods are several seconds or less. The small overall coefficient will provide little inherent system response to impressed reactivity changes, and it will consequently be necessary to provide control rods to compensate any reactivity changes of intermediate duration. Long-term reactivity effects, such as those associated with the fuel cycle, are compensated by adjustment of the fuel concentration in the salt.

So from the way I see it, we MUST make the two-fluid LFR work. It has all the advantages: strong negative temperature coefficient, vastly simplified reprocessing, and minimum fuel and fertile inventory. The primary drawback was their approach to laying out the core. That part must be changed, along with perhaps a change in moderator material.