The Modified Geometry 2 Fluid Molten Salt Breeder

Dr. David LeBlanc
Physics Department
Carleton University
Ottawa, Canada

In this presentaion I will attempt to explain a completely new concept for a Liquid Fluoride or Molten Salt Breeder Reactor. The basic idea may seem almost trivial in nature, given its simplicity, but I hope its importance will be evident once all aspects of the design are examined. I am not aware if this idea is completely original as I think it likely to have been at least discussed during early days at Oak Ridge. Regardless, I believe it represents a major step forward in reactor design. I will first to give enough background such that its advantages and disadvantages can be appreciated.

This power reactor concept is based on a 2 Fluid molten salt reactor. A 2 Fluid reactor has separate carrier salts for the fertile Thorium and fissile U233. A Single Fluid reactor has both fissile and fertile mixed in the same carrier salt. The 2 Fluid reactor was the main focus of Oak Ridge National Laboratories (ORNL) for much of the 1960s before they switched to a Single Fluid reactor which has dominated the field ever since. My proposed reactor attempts to solve the problems that led to them abandoning the 2 Fluid concept, despite its numerous advantages. In order to understand the proposed solution, it is best to first briefly review the basics of molten salt reactor design. A description of the new design in some if its many possible forms will then be presented.


Fuel Processing

There are 2 types of fuel processing necessary to achieve break even breeding in molten salt reactors. A 3rd type of processing helps the breeding ratio but introduces some proliferation concerns and should be avoided if possible.

Fluoride Volatility

The first process is to remove uranium from the fuel salt. This is known as the fluoride volatility process and has been well known since the 1950s. It is one of the main advantages of working with these salts is that by simply bubbling fluorine gas through the salt, the uranium content in the salt will convert from UF4 to UF6 which then comes out of the salt as a gas. This UF6 can be later converted back to UF4 and reintroduced into the reactor as needed.

There are two main areas that this process is needed. First is that the fissile uranium typically needs to be removed from the fuel salt before it is processed to remove fission products. The second is that if a separate blanket salt carrying fertile thorium is used in the reactor, this process is used to take the produced U233 from the blanket and transfer it to the fuel salt.

Fission Product Processing

A big advantage of the molten salt system in general is that approximately 1/3 of all fission products are removed almost immediately after formation. These include the noble gases, most importantly Xenon which come directly out of the fuel salt, assisted by bubbling through helium gas. Also, noble and semi-noble metals will not stay in the salt but will plate out onto exposed surfaces, ideally onto an added high surface area sponge type metal that is periodically replaced.

The remaining fission products need to be removed from the carrier salt if a breeding ratio of 1.0 or greater is desired. This can be done on a reasonably slow timescale if a high breeding ratio is not desired. In fact, cycle times up to a year or more to process the entire carrier salt may be possible in many designs, especially those with a harder neutron spectrum. Of these, the rare earth fission products are the most important, if these are removed, the others may stay in the salt for many years without stopping the ability to breed.

Historically there have been many proposed methods, with only the last two being seriously considered.

Pre 1964

Various methods included slowing discarding the carrier salt (after recycling the fissile U233), a method called CeF3 substitution which would only process the rare earth fission products and several aqueous processes. None of these were ever shown to be truly practical on an engineering scale or due to economic reasons.

1964 Vacuum Distillation

A major breakthrough happened in 1964 with the discovery of the vacuum distillation technique. In this method the Uranium is first stripped from the carrier salt by the fluoride volatility process and then the carrier salt and the associated fission products are heated to 1000 C under low pressure. This effectively boils off the carrier salt, leaving the majority of fission products (FP) behind. A few FPs do remain with the salt but not enough to affect the ability to breed as the most important rare earth fission products are indeed removed. The clean carrier salt after condensing is recombined with the separated Uranium and sent back to the reactor. This technique does not work however if large amounts of thorium are present in the carrier salt as it would be left behind with the fission products and wasted. Thus the system is practical only for a 2 Fluid design which has most or all of the fertile Thorium in a separate blanket salt.

The idea of this technique being used for a Single Fluid design is sometimes mentioned. However, this is basically fractional distillation which would be orders of magnitude more difficult to engineer than simply boiling off the carrier salt in the 2 Fluid case.

1968 Liquid Bismuth Reductive Extraction

While ORNL struggled with what was known as the “Plumbing Problem” of the 2 Fluid design, they believed they had a breakthrough with the development of a technique that could remove fission products while Thorium was present. If proven viable, this would allow consideration of Single Fluid designs.
The technique involves counter flowing the salt with a stream of liquid bismuth that contains reductive material such as LiF. The reducing agent can trade with the fission products and be effectively removed. The main problem with this concept is that Thorium behaves chemically almost identically with the rare earth fission products that are the most important to remove. The term “delicate” is often used to describe the elaborate stages necessary to separate out fission products while leaving the Thorium in the salt. Thus for a Single Fluid design the process is possible but highly difficult. For a 2 Fluid design it is much more feasible as Thorium is not present in the fuel salt.

Protactinium Separation and Holdup

A 3rd processing step is not necessary, especially in a 2 Fluid design but does help the breeding ratio. This step is to remove Pa233 as it is produced and to store is separately to allow it to decay to U233. Protactinium 233 is the 27 day half life intermediate between fertile Thorium and fissile U233. The problem is that it has a moderately high cross section for absorption, such that if it stays in the reactor loop, it may capture a neutron and not become U233. The average neutron flux the Pa sees is the main factor on how many neutrons will be lost. By increasing the volume of salt that carries the Thorium, one can lower these losses and skip this processing step. However, in a Single Fluid design if one increases the salt volume (by a lower power density core etc.) then this also increases the amount of fissile U233 needed at the same time.

In a 2 Fluid design with Thorium only in the blanket salt, one can increase its volume which does not increase how much U233 is needed. This is a financial burden in terms of carrier salt and Thorium but it is by no means excessive.

Skipping the Pa removal step is important for two main reasons. First, the process must be done very quickly, on the order of processing the whole volume of carrier salt in3 to 10 days. This is an enormous t
echnical and economic challenge. Second is that this introduces a unique proliferation risk. When U233 is produced while in the reactor, significant quantities of U232 are also present. This is highly radioactive and leads to an extremely strong gamma ray being emitted. This would make working with the material extremely difficult if not impossible and also make detection of illicit material easily detectable. However, if the Pa is removed and allowed to decay outside the reactor it produces relatively clean U233.

Thermal, Epithermal or Fast…

In Liquid Fluoride or Molten Salt reactors it is fairly simple to adjust the degree of neutron moderation in the design. The percentage of graphite moderator can be adjusted or can be removed all together. The carrier salt itself, typically LiF-BeF2 known as Flibe, provides a modest amount of moderation as well. How thermal or how fast, also termed a softer or harder spectrum has many effects.

In general a softer or thermal spectrum substantially increases the fission cross section and thus typically means a lower percentage of U233 in the salt and smaller startup fissile load. However the cross section for absorption in the carrier salt atoms, any fission products and any graphite will also rise substantially. To get a very soft or thermal spectrum a good moderator should be used such as graphite. Another important issue is the the prompt neutron lifetime will be longer in these designs which makes reactor control easier.

A harder or faster spectrum will mean the fission cross section will effectively be lower. This means either raising the fissile concentration and/or substantially increasing the neutron flux. Thus the starting fissile load is typically seen to be higher in these designs. A positive effect is that the absorption cross section for salt and fission products will also drop and at an even faster rate. This means that for quite hard spectrums, alternative carrier salts might be considered or much longer processing times to remove the fission products. A negative effect is that the prompt neutron lifetime becomes shorter.

One way to make a substantially harder spectrum is to completely remove graphite from the core. This has the great advantage or removing the whole problem of graphite replacement due to fast neutron swelling.

Another important issue is the number of neutrons emitted per absorption in U233. While it remains above 2.0 for any molten salt system, it has a U shaped profile with average neutron energy. For a very thermal spectrum it is fairly high, above about 2.20. For more intermediate or epithermal spectrum’s it tends to drop, perhaps down to 2.17 or lower. For hard spectrum’s it tends to rise up again, not nearly as dramatically as seen in Pu239 but still much higher, well above 2.2.

Having a very hard spectrum is desirable and often ends up being a design goal but this comes with many inherent problems. The higher the fissile concentration the harder the spectrum and better the breeding BUT this leads to worries about accidental criticality in fuel handling or potential spills. If a significant amount of moderation is found, such as water within concrete, the spill could reach criticality. Thus in my opinion it is desirable to harden the spectrum but only to a modest degree such that these issues are far more manageable.

For the concepts presented here, it is proposed that two alternatives be focused on. First is a graphite moderated thermal design that brings starting fissile load as small as possible while still having a breeding ratio = 1.0. The second is a harder spectrum without the complications of graphite but keeping the fissile concentration and starting load as low as possible while still having a breeding ratio = 1.0

Single Fluid vs 1 and 1/2 Fluid vs 2 Fluid Designs

As mentioned, the design to be presented is a 2 Fluid Design. It is important then to highlight the differences between the 2 Fluid and Single Fluid design. There is also a sort of in between design called the 1 and 1/2 Fluid design which has some good aspects of both designs but also the many disadvantages of both as well. Almost all designs since the 1970s have been Single Fluid designs with the exception of recent French work that is modelling what might be thought of as a partial 1 and 1/2 Fluid design.

Single Fluid

In this design both the fertile and fissile material is within the same carrier salt. This introduces many advantages but some serious disadvantages. One obvious area is in increased neutron leakage. In a graphite design, this can be partially minimized by having an outer annulus of the core that has a higher percentage of salt. This makes the region under moderated and will be absorbing more neutrons than it produces. Also, larger cores will minimize leakage losses but they will always be higher than in 2 Fluid or 1 and 1/2 Fluid designs. The list below is primarily for graphite moderated designs but many aspects are also true for designs without graphite.


  1. Simple core design and can learn from MSRE experience directly
  2. Slightly lower neutron losses in the carrier salt (one salt instead of two)
  3. Ability to run as simple converter reactor if fuel processing not employed


  1. Fission product processing greatly complicated by the presence of Thorium
  2. Higher neutron leakage
  3. Weakly positive temperature coefficient, can be fixed but at large cost
  4. Pa removal needed unless both thorium and 233U loading increased substantially

Point 3 above is important to discuss. A positive temperature feedback coefficient is generally a bad thing for any reactor design. It is not as serious as may be thought however since the positive term results from effects of the graphite which will lag behind any temperature increase in the salt by tens of seconds at least. Original ORNL work thought it to be slightly negative, recent French studies have shown that to be mistaken. This was mainly due to older calculations treating the graphite and salt mix as homogeneous. In order to solve this problem without destroying the ability to breed, French proposals have gone the route of having an extra Thorium blanket around the core (radial only, not axial). This make it a partial 1 and 1/2 Fluid reactor.

A “standard” 2 Fluid Reactor

In a 2 Fluid design there are separate carrier salts for both the fertile Thorium and fissile U233. The Thorium blanket salt surrounds the core to all but eliminate neutrons leakage. These designs will typically be neutronically superior and able to reach a higher breeding ratios but perhaps the largest advantages are in terms of simplifying the fission product processing and the ability to avoid Protactinium separation. In what will be termed the “standard” 2 Fluid design, the blanket salt is not only surrounding the core but also is interlaced within the core region with the fuel salt. This very complex arrangement gives rise to what was known as the “Plumbing Problem” which eventually led to ORNL abandoning this design despite its many advantages. It is a proposed solution to this plumbing problem that is the basis of the new reactor concept.


  1. Much more practical fission product processing without thorium in fuel salt
  2. Have choice of Vacuum Distillation or Simplified Liquid Bismuth Extraction
  3. Strongly negative temperature/void fuel salt reactivity constants
  4. P
    a removal easily avoided by simply increasing volume of blanket salt
  5. Neutron leakage near zero


  1. Interlacing of fuel and blanket salt within core is the “Plumbing Problem”
  2. Blanket salt has positive temperature/void coefficients
  3. Need for extra heat transfer loop for the blanket salt (5-10% of heat load)

1 and 1/2 Fluid Design

In a 1 and 1/2 Fluid design the reactor core has a single mixed fluid containing both fissile plus fertile material and surrounding the core is a blanket salt of fertile only. This is a way to limit neutron leakage and improve breeding ratio. This type design was very common in early ORNL work on graphite free or Homogeneous designs. A core salt with both Thorium and fissile U233 or U235 would be surrounded by a blanket salt. Some early graphite designs also looked at this route. Recent work from France models what might be called a partial 1 and 1/2 Fluid as they have a blanket salt radially around the core but not above and below axially.


  1. Reasonably simple core design
  2. Improved neutron economy due to low neuton leakage
  3. Improved ability to have a negative temperature coefficient
  4. Potential to avoid Pa removal, depending on Thorium worth in core region


  1. Same complex fuel processing as with Single Fluid design for core salt
  2. Adds need to also process a blanket salt by fluorination
  3. Extra heat transfer loop needed for blanket salt
  4. Need of barrier to separate core and blanket region, potential neutronic parasite

Molten Salt Development Timeline

With a background in design issues in place, it is useful to show a condensed timeline of how Molten Salt Reactors were developed at ORNL. It is important to remember that goals at the time would have been very different than today. At the time, it was viewed that world supplies of Uranium would not last more than a few decades. As such the primary goal was a breeder reactor with as short a doubling time as possible. This is a factor of both Breeding Ratio and Specific Inventory (how much fissile material is needed for startup). Molten Salt Breeder Reactors (MSBR) were in competition with another breeder design with far greater funding and more nationwide support, the liquid sodium fast breeder.

Current goals for a reactor would be far different. Having a low specific inventory is important to allow widespread startup but with the glut of enrichment capability, weapons grade material and fissile actinides available as waste, this problem is not as severe. As well, a breeding ratio above 1.0 should actually be looked upon as a drawback. With a plant designed and built to operate at B.R. = 1.0 there is no entry or exit of fissile material to the plant after startup. Also, any hostile take over of such a plant could not produce excess fissile material without rebuilding the plant infrastructure. Passive safety and proliferation resistance are also goals well matched in Molten Salt Reactor designs. The inclusion of the standard 1970s MSBR in the Gen IV program should be a testament to those aspects.

Early 1950s
Concept of operation based on fluoride fuels takes shape. Obvious ability of very high temperature operation leads to the construction of the Aircraft Reactor Experiment which uses canned BeO moderator and runs at 860 Celsius. Shows the system’s general prospect as power reactor. No fuel processing steps established beyond the fluoride volatility process for removing uranium. Compatability of graphite in the salts unknown

Late 1950s
Focus is on what are called Homogeneous designs that have no extra moderation beyond that of the salt itself. Studies are all in spherical geometry with a central core salt surrounded by a blanket salt and separated by a 1/3 inch Hastelloy N shell. In most cases the core salt also contains Thorium making most 1 and 1/2 Fluid designs. The only true 2 Fluid designs practical are for very small diameter cores. Work shows the ability to breed if spectrum is hardened somewhat by increasing fissile concentration along with Thorium in the core salt.

Early 1960s
The ability of graphite to have long term compatibility with the salts is now established. Using graphite is seen as a way to significantly lower the specific fissile inventory and most development work now goes in this direction. Some work on Homogeneous designs continues, highlighted by the 2 Fluid MOSEL design of ORNL and Germany that has an elaborate internal plumbing scheme to keep fuel and blanket salts separated in the core and to cool the core by the blanket salt. Design and construction of the highly successful Molten Salt Reactor Experiment (MSRE) which is a 8 MW (th) reactor based on the Single Fluid model for simplicity.

1964 to 1967
Even as the Single Fluid MSRE nears startup, breeder design focus is still on the 2 Fluid model. The golden era of 2 Fluid, graphite moderated designs is marked by the 1964 development of the Vacuum Distillation process. For the first time it appears a practical method of fission product processing is available. The “standard” 2 Fluid design is the focus which has the inherent difficulty of trying to keep fuel and blanket salts separated within the core. This problem proves more and more a difficulty such that even a single leak between the two within the core may mean the entire core and vessel would need replacing. Also the swelling effect of graphite during irradiation is found to be stronger than previously thought.

Molten Salt Reactor Experiment operates throughout this time and runs very smoothly on all three fissile materials separately (U235, U233, Pu). Several operational issues are encountered but solutions are subsequently established.

The advent of the Liquid Bismuth Reductive Extraction method results in a change of ORNL`s focus to a Single Fluid design. Regardless of the difficulties of this process due to the presence of Thorium in the salt, the great difficulties of the standard 2 Fluid “Plumbing Problem” leads ORNL to abandon this avenue and all major studies since have been Single Fluid or occasionally 1 and 1/2 Fluid designs. In the early 1970s though, the government decides that it does not wish to keep a second breeder option and cuts off funding in favor of the liquid sodium fast breeder, a major mistake in the view of many.

1976 to mid 1990s
Modestly funded studies continue at ORNL and worldwide. Russia’s work in the field has been and continues to be relatively unreported outside their borders. Highlights from ORNL include a converter design with much better resources utilization than LWR and a 30 year core lifetime without fuel processing. As well, a denature cycle with added U238 is shown to be able to have a break even breeding ratio without Pa removal. Most other countries however are reluctant to pursue a design that has been abandoned by its country of origin. A small but vocal group of supporters keep the pressure on to return to these design, such as Charles Forsberg of ORNL and Kazuo Furukawa of Japan.

Late 1990s to the Present
Changing needs and new technology starts to see renewed interest in the systems potential. From a technological standpoint, new compact heat exchangers are thought to be able to significantly reduce the amount of salt outside the core which is important for many reasons. As well, closed cycle gas turbines are seen to be a much better match to the salts high temperature operation compared to steam cycles. The systems ability to help with the great quantities of actinide wastes in spent fuel is also desirable
. Fuel elements need not be fabricated and inputs of different isotopic blends are not a problem. The basic molten salt design’s inherent abilities lead it to being included in the Gen IV program. Major studies begin in France which highlight the main problems of the Single Fluid MSBR design which are the complex fuel processing and poor temp and void reactivity coefficients.

2007 The Return of the 2 Fluid Reactor

As discussed, there are numerous advantages to the standard 2 fluid design in terms of reprocessing and temperature reactivity coefficients. The main drawback leading to its abandonment was the complex plumbing presumed to be needed to keep the fuel and blanket salts separate but intermixed within the core region. This presumed need for fluid intermixing is often explained in early ORNL documents. The following quote is from ORNL 4518 which focuses on graphite moderated designs but a similar argument holds for graphite free designs.

“In order for the graphite to have an acceptable radiation life, we estimate that the maximum power density should not exceed 100 kW per liter of core volume. With this limit on power density, the core of a central-station power reactor would have a volume of several hundred cubic feet. This size is too large for the core for to consist of graphite bars and highly enriched fuel salt contained in a thin metal shell and surrounded by a region of blanket salt. The critical concentration of 233U in the fuel salt would be so low that the absorptions in the carrier salt and graphite would be excessive. Absorption of neutrons by the shell would further degrade the performance.
The concentration of 233U in the fuel salt can be raised to the desired level by dispersing blanket salt throughout the core…”

On the surface this statement appears to be valid. A 1000 MW(e) or 2200 MW (Th) plant at a high power density of 80 kW per litre would need approximately 30 m3 (1000 ft3) of core volume. This 80 kW/l is core average which relates to about 400 kW/l in the fuel salt. If one assumes the traditional spherical core or short cylinder (z ~ d) this equates to close to a 4 meter (12 foot) diameter core. Even a smaller 100 MW(e) unit would be about 2 meters in diameter. In order to have even the smaller 2 meter core be just critical, then U233 concentration would need to be extremely low. This would entail too high parasitic absorptions in the salt and graphite.

Typical U233 concentrations in ORNL work were on the order of 0.2% to 0.4% molar % in the carrier salt. With an approximate 80/20 ratio of graphite to fuel salt, the critical diameter of a spherical core would be on the order of 1 meter in this concentration range. This size core would only produce about 30 MW(e) and a power plant consisting of 30 or more such units would not be seen as practical.

In terms of a graphite free design, the critical diameter of a pure salt core with similar U233 concentrations is approximately the same. With the same 400 kW/l in the fuel salt now being 100% of the core volume a core of this size might generate 150 MW(e) but it was still thought necessary to intermix fuel and blanket salts for 2 Fluid designs as well.

Solving the Plumbing Problem

It is the premise of this presentation that a surprisingly simple solution to the plumbing problem may be found by a basic rearrangement of the core geometry. To have a practical 2 fluid design without the plumbing required of intermixing within the core region is dependent on two factors. First, that the inner core diameter must be small enough that it is just critical and that over half of produced neutrons will leak into a surrounding blanket salt and be predominately captured by thorium atoms. Second, that the core have enough volume to have a practical total power output.

These two requirements can not be met by a spherical core or the typical short cylinder core. However, they can indeed be met by a core geometry with one or two extended dimensions. Namely an elongated cylinder (z >>d) core or a thin rectangular slab core.

If a spherical core has a critical diameter of about 1 meter, then a long cylindrical core will have a critical diameter of significantly smaller. As a first approximation the critical diameter will be the ratio of the Buckling constants between the given geometries. Thus, for the same graphite and/or fuel salt combination, an infinite cylinder will have a critical diameter approximately 76.6% that of a sphere. If a specific combination of fissile concentration, graphite percentage and carrier salt gives a critical diameter of 1 meter for a bare sphere, then for comparison, a 5 meter long cylinder would have critical diameter of 0.77 m and a 4m by 4m slab would be 0.51 m thick.
The great advantage of going to a long cylinder or thin slab is the fact that a practical total power can now be obtained without intermixing but by simply extending the length of the core. While a barrier will be needed to be maintained between the core and blanket regions, this will be far less complex than the intimate intermixing of fuel and blanket salts in a standard 2 fluid design. In terms of end plenums on these cylindrical cores, the simplest arrangement would be to taper the ends to a sub critical diameter while still surrounded by the blanket salt (see figure). This should all but eliminate leakage of neutrons.

In terms of core diameter, this will be dictated by the desired mix of fissile concentration, carrier salt moderation and absorption properties along with whether graphite is also incorporated. The surrounding blanket salt will function somewhat as a reflector to slightly lower the critical diameter. Due to the strong absorption to thorium it will behave only weakly as a reflector unless graphite is also present in the blanket region as was occasionally proposed in ORNL work. Core diameters on the order of 0.75 to 2 meters are expected for typical fissile concentrations. Core diameters beyond 2 meters would entail such low fissile concentrations as to make parasitic losses to graphite and/or salt potentially dominant. Using high fissile concentrations may result in critical diameters much less than 1 meter but this would be counter productive as it would imply higher fissile start-up loads, increased length and/or number of cores and also complicating salt handling due to accidental criticality concerns.

tric studies of critical diameters for various combinations is planned but data available from previous ORNL and other studies gives a good indication of working parameters as a starting point for optimal reactor design speculation.

The needed thickness of blanket salt surrounding the core is for now assumed to be at least 60 cm (2 feet) as this was commonly chosen in ORNL studies. The prime candidate for a blanket salt is the 73% LiF + 27% ThF4 which has a reasonably low melting point of 560 oC. Other alternatives should of course be examined. It is planned that enough blanket volume and thus Thorium will be employed such that Pa removal will not be necessary.

Another important aspect of this modified geometry is in terms of temperature and void coefficients. In common with the standard 2 fluid design, there will be a strong negative reactivity coefficient for the fuel salt. In terms of the blanket salt, as this salt now acts somewhat as a reflector, lowering its density by void or temperature should lead to a small but negative temperature and/or void reactivity coefficient for the blanket as well. This alone is a very important improvement over the standard 2 fluid design. To increase this negative term, the blanket could be made of a minimal thickness to capture leakage neutrons but then surrounded by a good neutron absorber material. Thus, if the blanket density falls, more neutrons will scatter into the absorber and be taken out of the cycle before being able to reflect back into the core region.

Before listing some possible design variations in the coming sections a few final general features of the design should be mentioned. One is that in operation, the blanket salt should be run at a higher pressure than the fuel salt. This was also the mode of operation for the standard 2 fluid design as one wishes any breakdown or leak in the barrier to lead to blanket salt flowing into the fuel salt as opposed to the reverse. Also, it should be mentioned that some proposed designs may see a spike in the thermal neutron flux and thus power peaking just inside the barrier between core and blanket. This would result from reflected neutrons that have experienced extra moderation in the blanket before returning to the core or also if the barrier material is graphite of appreciable thickness. In any sort of solid fuel core this would be of much concern but with a fluid fuel that can quickly mix, this should not be a problem unless power peaking is extreme.

Another issue of a general nature is in the orientation the long core (or thin slab) whith horizontal or vertical being the main choices. While there are some advantages to a vertical design it is currently preferred by the author to adopt a horizontal design for the cylindrical core. One reason being the lack of significant hydrostatic head pressures as no part of the system will need to be substantially elevated over any other. Another reason is in terms of response to accidental core rearrangement such as by external explosion or sabotage. Any reactor can have imaginable ways to reconfigure into a slightly more critical arrangement during extreme and highly unlikely circumstances. This is even more of a concern for these extended geometry designs and perhaps represent the only disadvantage compared to previous designs. In the horizontal arrangement though, gravity will not act to concentrate core material in an accident scenario. Having a strong neutron absorber material just beyond the blanket salt should also be of assistance if any fuel salt finds its way into the blanket area. Fast shutdown systems by injectable poisons or control rods might also be employed.

A central control rod is not depicted in these simple figures. As molten salt reactors undergo very little reactivity change during operation, they are typically not needed for safety as long as there are negative temp and void coefficients. They are a convenient way to moderate the temperature of the core. Most ORNL designs had control rods and if it proves advantageous, they can be added to these designs. Given the close proximity of the blanket to all parts of the core, it may be possible to located control mechanism in the blanket region alone.

In designs employing graphite moderation that may require periodic core replacement due to fast neutron damage, removal and replacement by overhead cranes for a slender horizontal core should prove simpler than in a vertical design or indeed simpler than any previous molten salt design. Replacing the surrounding vessel itself during this operation might prove a simplier and faster operation.

Core Design Proposals

In this new elongated geometry there are of course many possible versions to examine using the studies of ORNL and elsewhere as a guide. Using graphite versus a Homogeneous design is perhaps the most important issue. While extending the length of the cylindrical core or slab gives a very useful total power, it will still be necessary in many designs to have multiple cores adding up to a typical 1000 MW(e) plant. This may go against economies of scale somewhat but allows many other benefits such as having one or more cores down for maintenance or perhaps graphite replacement while still producing power.

Barrier material between the fuel salt and blanket is another major issue effecting designs. For proposals with a more thermal spectrum, use of Hastelloy N or other metal barrier will need to be very limited. If only acting as a barrier such as simply being wrapped around a graphite core it would seem the barrier could be quite thin as it would need little structural strength. Hastelloy N being mostly nickel will have a thermal neutron cross section of approximately 4 barns. While not extreme, this is more than a hundred times the thermal absorption cross section of graphite with about the same atomic density. On a 1 meter diameter core, a 1 mm thick Hastelloy N barrier would be 0.4% of the core volume (1/250). Sub millimetre non structural barriers in thermal spectrums should be acceptable but barriers beyond 2 mm may prove detrimental to the neutron economy. For comparison purposes, neutron loses to graphite in standard 2 fluid graphite designs were on the order of 0.03 to 0.05 per fissile absorption while breeding ratios were 1.05 to 1.07.

Using Hastelloy N as a barrier between fuel and blanket salts in graphite free designs is a completely different situation. Absorption cross sections of structural materials, carrier salt or fission products will drop off more rapidly than the fission cross section. Early ORNL computation studies of homogenous salt reactors with core and blanket regions employed a Hastelloy N barrier of 1/3 of an inch or about 8 mm and losses were typically not extreme. This thick barrier was envisioned as some spherical cores were up to 12 feet in diameter in the study. For modified geometry designs in which the core is a simple tube of Hastelloy N, on the order of 1 meter in diameter, it should be possible to lower this thickness substantially. It is entirely likely that a graphite or carbon-carbon composite will form the barrier, in this case the thickness can be as much as desired as it has a very low absorption cross section.

The following sections are meant to act as early models to guide design and as well to show the great versatility of the general molten salt system especially when combined with this modified geometry. All proposals require much further calculations but are offered as a guide to expedite design feasibility studies.

Fuel and Blanket Salt Data

The primary candidate of carrier salt the fuel salt is the LiF-BeF2 (67-33) eutectic known as Flibe. Its melting point is 460 Celsius and it has a volumetric heat capacity of 4.69 J/cm3 C. As a fuel salt carrier for 233U, its properties will be only weakly affected by the small percentage of UF4. For a more thermal spectrum this salt is perhaps the only choice given this sa
lts very low absorption cross section. For harder spectrums there are many other candidates that should be considered. NaF-BeF2 has many advantages including a much reduced melting point (340 C). NaF-ZrF4 is also attractive especially on cost issues, low moderation and the fact that tritium production will be curtailed. For the present though, the Flibe carrier salt will be assumed for the fuel salt in calculations in all designs.

The blanket salt is expected to be the eutectic of 27% ThF4 and 73% LiF with a melting point of 560 C proposed in most ORNL designs. Heat production in the blanket salt will be only a small fraction of the overall thermal load and would be sent to separate heat exchangers.

Cylindrical Graphite Core

Employing graphite in a molten salt reactor introduces many advantages and disadvantages. Lowering the neutron energy spectrum typically reduces the fissile concentration needed and at the same time improves reactor control due to a lengthening of prompt neutron lifetime. On the other hand, fast neutron damage induces first shrinkage then swelling of the graphite such that an expected lifetime in the core conditions may be limited. High power densities of 80 kw/l suggested in early ORNL work may only see a lifetime of 2 to 4 years depending on flux flatting and graphite temperature. Going to much lower power densities and thus larger core volumes can extend this. The extending geometry proposed, especially a long slender cylinder, should make core replacement a much more practical endeavour which may mean even higher power densities than 80 kw/l might be worth considering.

Monolithic cylinders of graphite with appropriate fuel channels drilled out are a simple starting point for speculation. Due to such issues as differential thermal expansion and irradiation damage it will most likely entail the use of standard hexagonal graphite elements. At the radial edge of such a cylinder, a barrier of graphite or other carbon based tube is a possibility as well as using limited amounts of Hastelloy N. The figure shows a possible arrangement showing the tapering of end plenum space by using Hastelloy N.

For example purposes we might consider a 1 meter diameter graphite cylinder with 20% fuel channels by volume which was typical of ORNL 2 fluid work. It is estimated that a fissile concentration of about 0.3% 233UF4 would be needed for criticality based on ORNL studies such as ORNL-TM-1851, Table 21.

As a starting point, a flow rate of through the core of 4.5 m/sec is chosen. The single fluid MSBR design had a somewhat lower velocity of 2.6 m/s but values up to 6 m/s appear within reason. Pumping power will be increased somewhat as a penalty. A temperature increase of 140 K is also assumed which was the typical value for ORNL studies.


(a) A 100 cm diameter core which is 80% graphite and 20% fuel salt.
(b) A flow rate through the core of 4.5 m/sec.
(c) A delta T of 140 K for the salt temperature increase (same as MSBR).

Volumetric Flow Rate = Salt Velocity x Fuel Salt Cross Sectional Area
= 450 cm/s x core area x 20%
= 450 cm/s x 3.14 x (50 cm)^2 x 0.20
= 7.07 x10^5 cm3/sec

Thus 0.707 m3/sec will be flowing into and out of the cylindrical core.

Heat Production = Volumetric Flow Rate x pCp x deltaT
= 7.07 x10^5 cm3/sec x 4.69 J/cm3 K x 140 K
= 464 MJ/sec
= 464 MW(th)
= 200 MW(e) at 43% efficiency

Efficiencies of 48% are thought possible with a Brayton closed gas cycle so 43% is a conservative value. If two such cylinders are operated in tandem with heat exchangers on each end to minimize out of core salt volumes, then the base unit is 400 MW(e). This is a good multiplier size for any desired plant, i.e. 400, 800, 1200, 1600 MW(e).

How long this single tube core needs to be is determined by the power density. As core fabrication and replacement is projected to be easier than historic designs, it should prove acceptable to have quite high power densities. A moderately high power density of 80 kW/l would require a core length of 7.4 meters. Core replacement may be needed every 2 to 4 years at this power density depending upon graphite temperature and flux flattening.

This fuel concentration of 0.3% is the same range as both the ORNL standard 2 fluid design and the single fluid MSBR. The high power density afforded by the ease of graphite replacement should lead to a low fissile inventory. The standard 2 fluid, 80 kw/l design of ORNL had a specific inventory around 800 kg per Gw, reaching this should be straight forward. In fact this value may be substantially reduced. The originally proposed heat exchangers contained about 50% of the total salt volume. New compact heat exchanges may lower this by up to 80%. Also, standard 2 fluid designs had 20% of salt volume tied up in end plenums, very little will be needed in this current design. Finally, fissile density might be lowered from 0.3% as a breeding ratio of only 1.0 is acceptable. Given the possible savings, a much lower limit of 300 kg U233 per 1000 Mw(e) may be attainable. While lower salt volumes and specific inventory are important goals, it needs to be remembered that this also lowers the total heat capacity of the salt.

Start up on reactor grade plutonium requires an even smaller inventory as it has a higher fission cross section in these softer neutron spectrums and high temperatures. 200 kg for a 1000Mw(e) plant may prove possible, thus a mere 1% or 2% of that needed for a liquid metal cooled fast reactor. Start up on plutonium requires adding small amounts for the first few months of operation. For comparison the annual Pu waste of a 1000 Mw(e) LWR is about 266 kg while a sodium or lead cooled fast reactor requires 12 to 20 tonnes of Pu.

Hastelloy N Tube Homogeneous Design

At the other end of the design spectrum is a simple Hastelloy N tube to act as barrier between fuel salt and blanket. Without graphite for extra moderation the spectrum will be hardened unless fissile concentration is lowered. Going the route of increasing the fissile concentration will further harden the spectrum and improve the overall neutron economy due to a better ratio of fission cross section to parasitic capture of carrier salt and structural material. A harder spectrum also increases the number of neutrons per absorption in 233U. However this also goes against the goals of low specific inventory which is one of the single most important design issues.

Much work on such graphite free systems was performed at ORNL in the 1950s before it was confirmed that graphite and fluoride salts were truly compatible. All studies assumed a spherical geometry but still provide a good guide for estimation purposes. In all computational studies a fairly thick Hasteloy N barrier between the inner core fuel salt and the outer blanket salt was employed. A full 1/3 inch or 8 mm was standard since some designs were up to 12 feet in diameter. Lowering this thickness to improve the neutron economy should be possible.
Another issue is that smaller cores, with higher fissile content, led to a harder spectrums and the 60 cm blanket was not sufficient to prevent significant leakage. This can be corrected by simply increasing the blanket thickness. In the blanket over 95% of the absorptions will be in Thorium such that almost all previously leakage neutrons could potentially go into breeding.

In the late 1950s, practical fission product processing methods such as vacuum distillation had not yet been discovered. As such, the
y examined various situations of either having thorium in the fuel salt (a 1 and ½ fluid) or no thorium and thus a true 2 fluid for small spheres. Representative data shown below are for clean starting cores on pure 233U. Additional ORNL work on long term breeding ratios typically showed a drop of 0.03 to 0.04 in the long term and this was with a long 1 year processing time for fission products. Thus if a breeding ratio of 1.04 is calculated for a clean core then break even long term should not be of difficulty. These studies assume a slightly different fuel carrier salt of 69% LiF and 31% BeF2. The blanket salt was 25% ThF-75%LiF. In some calculations, such as the case with 7 foot diameter employed a blanket salt with lower thorium content which lowered regeneration ratios but improved the critical mass due to the blanket acting as a better reflector.

TABLE I. Initial State Nuclear Characteristics of Spherical Two Region, Homogeneous, Molten Fluoride Salt Reactor with U233. Data taken from ORNL 2751 with added projections by author.

Inner Core Diameter 3 feet 4 feet 4 feet 6 feet 8 feet
Thorium in Fuel Salt 0 % 0 % 0.25 % 0 % 7%
U233 in Fuel Salt 0.592% 0.158% 0.233% 0.048% 0.603%
Neutrons per absorption in U233
Be, Li and F in Fuel Salt 0.0639 0.1051 0.0860 0.318 0.078
Hastelloy Core Wall 0.0902 0.1401 0.1093 0.1983 0.025
Li and F in Blanket Salt 0.0233 0.0234 0.0203 0.0215 0.009
Leakage 0.0477 0.0310 0.0306 0.016 0.009
Neutron Yield 2.1973 2.1853 2.1750 2.2124 2.200
Median Fission Energy 174 ev 14.2 ev 19.1 ev 0.326 ev 243 ev
Initial Breeding Ratio 0.9722 0.8856 0.9288 0.6586 1.078
Projected B.R. Thinner Wall* 1.060 0.9836 1.011 0.7722 1.099
Projected B.R. Graphite Wall** 1.105 1.054 1.066 0.8714 1.112

* Projected assuming a thinner Hastelloy core wall of 1/6 inch (4.2 mm) and 90% leakage reduction by using a thicker blanket salt

** Projected assuming a Graphite or Carbon-Carbon core wall with negligible absorptions and 90% leakage reduction by using a thicker blanket salt

Projected initial Breeding Ratios are estimated for potential neutron savings coming from a thinner Hastelloy N core wall of 1/6 inch (4.2 mm) along with lower leakage using a thicker blanket. Also a second projection is given if the core wall is graphite or carbon-carbon composite (a strong likelihood). It is unknown if the neutron bonus from (n,2n) reactions in beryllium was included in these 1950`s computations. If it was not, this factor typically adds up to 0.02 to the breeding ratio which would further improve the situation. It is of course realized that neutron cross sections may not have been very accurate in the 1950s but even a large increase in true neutron losses can be accomodated by increasing the fissile concentration.

Many trends can be observed in this small sampling of data. Evident is the poor neutronic results for the 6 foot core as these larger diameter reactors require extremely low fissile concentrations if thorium was not present. This highlights the rational for mixing in thorium for designs greater than 3 of 4 foot spherical diameter cores as such small cores would not have been seen to be of useful total power. Also noteworthy is the U shaped trend of neutron yield with median fission energy. For hard spectrums it rises above 2.20 and for quite soft spectrums it rises as well.

The data for a 3 foot sphere shows obvious potential to have a long term breeding ratio above 1.0 even with a slow fission processing cycle of up to a year. The fissile concentration being 0.6% is low enough that specific inventory for a 1000 MW(e) reactor should still be quite reasonable. It is projected that a fissile concentration down to 0.3% may represent the lower range of the ability to break even with a Hastelloy N wall, while 0.15% should be possible with a graphite core wall. These two lower concentrations will have a fairly soft spectrum and a longer prompt neutron lifetime. For example in the 0.158%, 4 foot case above, it had a median fission energy of 14.2 ev and 7.95% of fissions being thermal.

If the case of a 3 foot sphere is converted to elongated cylindrical geometry, the diameter should change to approximately 2.3 feet or 70 cm. The power density of the core is much higher as it is 100% salt as opposed to 20% salt, thus, the total power of even such a small tube reactor can be impressive. Flow rates through such a simple core can also be much higher if desired, up to 10 m/s would seem reasonable as the core is simply a pipe.


(a) A 70 cm diameter core which is 100% fuel salt.
(b) A flow rate through the core of 4.5 m/sec.
(c) A delta T of 140 K for the salt temperature increase (same as MSBR).

Volumetric Flow Rate = Salt Velocity x Fuel Salt Cross Sectional Area
= 450 cm/s x core area
= 450 cm/s x 3.14 x (35 cm)^2
= 1.732 x10^6 cm3/sec

Thus 1.732 m3/sec will be flowing into and out of the cylindrical core.

Heat Production = Volumetric Flow Rate x pCp x deltaT
= 1.732 x10^6 cm3/sec x 4.69 J/cm3 K x 140 K
= 1137 MJ/sec
= 1137 Mw(th)
= 489 Mw(e) at 43% efficiency
= 546 Mw(e) at 48% efficiency

If one assumes the same power density in the fuel salt as graphite example, this is 400 kW/l. This is somewhat lower than many early ORNL homogeneous designs. With this smaller diameter core, the needed length to obtain 1137 MW(th) is slightly longer at 7.4 meters. Two such tubes, run in tandem with heat exchangers at either end already form enough for a large power plant.

The Hastelloy N barrier was expected to have a lifetime of 10 to 20 years according to early ORNL work. Even if the lifetime in full neutron flux proves shorter, it should be very simple to replace this single tube at regular intervals.

Graphite or Carbon-Carbon Composite Tube

A graphite or carbon-carbon composite tube is also be a strong possible design option. In this case it either provides the barrier itself between fuel and blanket salts or perhaps merely structural support to afford a thinner Hastelloy N barrier.

If the tube is of appreciable thickness, then it will also add a degree of neutron moderation. This may be beneficial in several ways. It would slightly lower the fissile concentration needed in the fuel salt. It would slow down neutrons leaking into the blanket region and thus make a thinner blanket region more effective. Also it would improve the prompt neutron lifetime somewhat.
Having little or no Hastelloy N barrier material will significantly improve the neutron economy. Thus the fissile concentration and/or processing rate may be lowered while still maintaining a positive breeding ratio.

A good example is for the 2 Fluid case of ORNL data for the 4 foot core. If the core wall is carbon based, then this example should be able to break even on breeding in the long term with only 0.158% of U233. The 4 foot (122 cm) sphere would equate to a 95 cm diameter tube in elongated cylindrical geometry. The calculation below shows its great potential to form an ideal power core.


(a) A 95 cm diameter core which is 100% fuel salt.
(b) A flow rate through the core of 4.5 m/sec.
(c) A delta T of 140 K for the salt temperature increase (same as MSBR).

Volumetric Flow Rate = Salt Velocity x Fuel Salt Cross Sectional Area
= 450 cm/s x core area
= 450 cm/s x 3.14 x (47.5 cm)^2
= 3.19 x10^6 cm3/sec

Thus 3.19 m3/sec will be flowing into and out of the cylindrical core.

Heat Production = Volumetric Flow Rate x rCp x DT
= 1.732 x10^6 cm3/sec x 4.69 J/cm3 K x 140 K
= 2094 MJ/sec
= 2094 Mw(th)
= 900 Mw(e) at 43% efficiency
= 1000 Mw(e) at 48% efficiency

At a typical power density of 400 kw/l, this would be a core of 7.4 meters in length. The intermediate heat exchanger could run parallel this length to form the basis of an extremely attractive power plant. Alternatively, the flow rate and power density could both be halved to make a tandem arrangement of two 3.7 meter long cores.

This low U233 concentration of approximately 0.15% will result in a quite soft spectrum even for this case of a Homogeneous core. The main advantages to this are the fact that prompt neutron lifetime will be long and accidental criticality in fuel handling or spills should not be an issue.

Other Configurations

In this modified geometry, numerous other designs are possible and under investigation but beyond the scope of this presentation of basic principles. Examples include a central graphite moderator within an otherwise homogeneous design to give a hybrid neutron spectrum. Also the use of external moderation, perhaps even by D2O, to simplify core internals whilst lowering fissile inventory and increasing prompt neutron lifetime should be examined.

Another possibility of interests is allowing small amounts of thorium to be present in the core salt. This would allow larger diameter cores to be critical on the same fissile concentration. In this mode, in order to retain the fission processing advantages of the 2 Fluid system, this Thorium would be discarded with the fission products during processing. For example, a thorium concentration of about 1% in the fuel salt and a one year processing cycle time would waste about 4 tonnes per year per Gw while 1 tonne of Thorium is burned. With the large reserves of Thorium and a cost of only 30$ per kg, this is a minor overall cost.

Denatured Operation

Another option to investigate will be operating on a denatured cycle in which the uranium within the reactor never goes beyond a weighted average of 20% U235 or 13% U233. This was shown possible for the Single Fluid design so it should be even more likely to succeed in this 2 Fluid design. As the flow of material entails adding fertile material to the blanket and then transferring fissile material from blanket to fuel salt, this would mean adding a combination of U238 and Thorium to the blanket salt such that the transferred Uranium is in a denatured state. This has the many problems associated with processing of Plutonium and higher actinides in the fuel salt but if it is deemed necessary it is certainly an option. It is the author’s opinion that the proliferation resistant properties of U232 combined with a break even design for the entire plant should make such a denatured cycle unnecessary.

Basic System Advantages

This system has the benefits of the standard 2 fluid design without its plumbing problem.

-Fission product processing by Vacuum Distillation or Simplified Bismuth Extraction
-Strongly negative temperature or void fuel salt coefficient.
-Avoid Pa removal simply by increasing thorium blanket salt volume.
-Extremely low fissile inventories possible
-300 kg U233 per 1000 Mw(e) a likely goal, even lower on Pu

Benefits beyond the standard 2 fluid ORNL design include

-Blanket salt temp/void will also be negative as this region acts as neutron reflector
-Much reduced salt volumes needed in plenum spaces
-Ability to also breed in Homogeneous designs with low specific inventories

Basic System Disadvantages/Limitations

May require non-graphite barrier material such as Hastelloy N which is parasitic load.

Multiple cores may be required for full 1000 Mw(e) power plant.

This elongated geometry is more susceptible to rearrangement to increase reactivity. A horizontal arrangement to counter effects of gravity along with the traditional method of keeping the blanket salt at higher pressure should address this concern. This represents an aspect that must be closely examined for all possible reactivity transients.



Leave a Reply