There are a number of significant issues in LFTR Design. The two most important issues are the graphite moderator issue, and the one or two fluids issue. The graphite moderator issue has to do with neutron energy in the reactor core. There are two major moderators that are commonly used in reactors, heavy water and graphite. Although there have been proposals to use heavy water as a moderator in LFTRs, to do so would introduce a complication in core design that would lead inevitably to higher R&D expenses. My view is that early LFTR design should go with options that require less research when ever possible. That would suggest then that graphite would be the moderator of choice, even though graphite carries some liabilities.
There is a school of thought that holds that the best solution to the moderator problem is to have none. I might find this approach valid if were were going to be dealing with a limited number of reactors. We are not. I calculated the number of 100 MW "Big Lots" Reactors that would be needed to replace fossil fuel powered electrical generators in the United States. My estimate would be in the neighborhood of 10,000. Using a graphite moderated core design means that only about a fifth as much fissionable materials will be needed to start a moderated LFTR as are needed to start an unmoderated LFTR. The problem is one of scalability. We can start 5 times as many graphite moderated LFTRs with a given amount of fissionable material as we can start if the LFTRs are unmoderated.
Thus a graphite moderated LFTR makes large scale LFTR deployment more easy to manage. This decision is not a difficult one to make.
The second major design decision is between two and one fluid core designs. In a two fluid core design the thorium carrying is segregated from the uranium carrying salt. As thorium-232 absorbs neutrons, it almost always goes through a 2 stage transformational process that produces fissionable U-233. On the first stage Th-233 emits an electron and becomes protactinium-233. Pa-233 then emits an electron and becomes U-233. In a one fluid Th-232, Pa-233 and U-233 would all be in a single fluid. There is a problem with this because Pa-233 is a neutron eater, and that does two bad things. First every Pa-233 atom that absorbs a core neutron is one less U-233 atom that will be produced by the LFTR breeding process. Secondly every neutron that is absorbed by Pa-233, is one less neutron that is available to breed Th-232 into U-233. So every time Pa-233 absorbs a neutron, two U-233 atoms are lost to the breeding process.
The obvious solution to this little problem is to process Pa-233 out of the core salts as quickly as possible. But as my father discovered in the 1960's when he tried to find how to do this, processing Pa-233 out of the core salts of a single fluid MSR is quite a challenge. The added chemical processing equipment required by a single fluid design would add to reactor cost.
In contrast, Pa-232 poses less of a problem in a two fluid core and blanket design. In a core and blanket design the core is surrounded by an outer donut that contains thorium salts, and other carrier salts. The blanket serves a double purpose. Thorium in the blanket captures the neutrons that leak from the core. Thus the blanket is a radiation shield. Because blanket thorium is not mixed with with the core salts, it can be present in the blanket at a much higher concentration that would be the case in a single fluid design. The higher concentration of Th-232 would prevent the Pa-233 neutron absorption problem from becoming significant in the blanket, because there would be many more Th-232 than Pa-233 atoms in the blanket.
Solving the Pa-233 problem is likely to involve more R&D and increase the cost of the "Big Lots" reactor.
We are of course left with problems with our design choices. The primary problem is that when energetic neutrons encounter carbon atoms in graphite they tend to knock those atoms around. This process causes graphite to swell, and makes it a poor material for core structures. It has been suggested, for example that core graphite could be present in the form of balls that float in the core salt. Periodically the graphite balls would be floated out of the core to be replaced by fresh undamaged balls. A more likely solution would be to shut a graphite cored LFTR down every few years, remove the damaged core, and replace it with a fresh core. The damaged core would be, of course, highly radioactive, and it most likely would have to go into long term storage. Graphite does not pose much of a long term storage problem, but used LFTR cores would be classified as nuclear waste. It would be highly desirable to have a low cost long term core graphite disposal plan before the "Big Lots" reactor went into production.
The "Big Lots" concept was born because I saw that the requirements of load following/peak load power generation would lead to longer graphite core life, as well as permit the use of lower cost materials like stainless steel in reactor manufacture. The use of stainless steel would require a lower reactor temperature, and the decrease in neutron exposure that a reactor that might be operated on a part time, part power basis would increase the life span of graphite and metal structural components.
It is clear that I have shamelessly advocated what might be called marketing based design process over a purely technological. Under ordinary circumstances I would not dream of second guessing the scientists and engineers, but my interest is not in building the best possible reactor but in building a reactor that is good enough and highly scalable.