Appendix 5 Pool vs. Loop Controversy

One of the options available in LMFBR design involves whether or not to enclose the major components of the primary system fully within a single tank. Designs that do so are referred to as “pool-type” and those that do not are “loop-type”. One of the earliest LMFBRs, EBR-2 was of the pool design. SEFOR, Fermi, and FFTF were all loop-type reactors. In the 60s and 70s in the U.S., there was the belief that while a pool might be a practical approach for a small experimental reactor, as the plant size is increased, loop-type reactors were considered more economic. However, outside the U.S., the pool-type design was preferred. While the German and Japanese demonstration plants incorporated the loop concept, the French, British, and Russians all moved quickly towards the pool concept. Superphénix, the largest LMFBR built worldwide to date, was a pool. Even the Italians had once started construction on a demonstration plant invoking the pool concept. As of the date the CRBRP Project was cancelled in 1983, the U.S., Japan, and Germany were in the loop camp while the U.K., France, and Russia were pool advocates. Since that time, what little work continued in the U.S., viz. EPRI’s Large Scale Prototype Breeder (LSPB), Rockwell’s Sodium Advanced Fast Reactor, and GE’s Power Reactor Integral Safe Module all incorporated the pool concept although a loop-type version of the LSPB was developed at the conceptual design level. When SNR-300 was cancelled, the Germans cast their lot with France joining a European consortium focused on the pool. LMFBR work undertaken in India, China, and Korea is all focused on the pool concept. Only the Japanese remain interested to the loop concept to the extent that there remains any significant LMFBR program in Japan today.1

Figures 49 & 50 are cutaways of the Superphénix reactor pool. Sodium from the pool enters the upper end of the IHX tube bundle and exits at the bottom. To provide a driving force for flow in the IXH, the pool is separated into two regions by a horizontal baffle also know as a “redan”. The primary pumps take suction on the cooler region below the redan. The pressure difference across the redan will be equal to the pressure drop across the IHX tube bundle. The primary pump discharges into piping that enters a plenum below the reactor core in the lower internals structure. Sodium then flows through the fuel assemblies and back to the pool.

This image has an empty alt attribute; its file name is 49.jpg

Figure 53 Superphénix pool & dome

Although most of the primary system is contained within the pool vessel, there are of necessity some systems which must operate outside the pool. Systems involved in refueling result in transfer of primary system sodium outside the pool tank. The primary sodium must be cold trapped to remove impurities, and those associated components along with plugging temperature indicators are outside the pool. In addition, cover gas processing must be performed outside the pool. If it were desired to have decay heat removal directly off the primary system, all those associated components would be outside the pool. Of course, all these components require piping to connect the pool with the components.2

The weight of the core, the shield assemblies, and the internal structure below the core as well as the total primary sodium inventory is carried by the tank wall.3 Because of the load carried by the tank wall, it is necessary to protect it from the hotter sodium at the core outlet. To accomplish this, an inner tank is installed above the redan. Cold leg sodium flow is directed to this region directly off the pump discharge. The upper internals structure which houses the control rod drive mechanisms and provides backup hold down for the fuel assemblies, the IHXs, and the primary system pumps are supported by the tank closure. The periphery of the tank closure also supports the tank itself as well as its guard vessel. This closure must therefore be a very massive structure.

There are no penetrations of the vessel wall. All penetrations are made through the closure (which is sometimes referred to as the “roof”). Eliminating vessel penetrations eliminates a source of failure and makes the vessel stress designer’s task simpler.

The tank closure or deck actually consists of two offset rotating plugs, one inside the other. This feature is necessary to permit refueling. By manipulating the rotating plugs and using the transfer machine, fuel assemblies may be moved to a refueling station near the tank wall then outside the tank using, as in the case of the Superphénix plant, an A-frame device.

This image has an empty alt attribute; its file name is 50-1.jpg

Figure 56 Superphénix pool

Although it may be appear intuitive that the loop concept is simpler than the pool, the pool does have the advantage of allowing the primary system components, the intermediate heat exchangers (IHXs) and the primary pumps to be brought in much closer to the reactor than is possible in a loop since there is no need for piping expansion loops to accommodate the thermal expansion attendant with heat-up from ambient conditions to operating temperatures. This means that the containment diameter can be made considerably smaller. The Superphénix containment is about 85 ft. in diameter which contrasts starkly with the 186 ft. CRBRP containment – a plant with just slightly more than ¼ the electric output of Superphénix.4 Since space within the containment comes at a very high premium, the pool concept is presumed to be more economic than the loop, all other things being equal. Moreover, only a single guard vessel is required (as opposed to seven for example on the CRBRP loop plant) and the reactor vessel is essentially eliminated (or becomes the pool vessel itself depending on one’s point of view) along with the boundary vessels for the IHXs and the primary pumps.

The sodium inventory of pool-type plants is considerably greater than for equivalently sized loop plants, by a factor of three or more. This larger inventory tends to reduce the effect of thermal transients on components both within and outside the pool. Since transients are generally more severe in LMFBRs than LWRs owing to the greater temperature rise across the reactor and the higher outlet temperature, features that reduce thermal transient stresses on components are potentially desirable.

The pool concept enables a very compact reactor unit, which is reliable with regard to cooling of the core and confining radioactivity. Pipelines with high temperature coolant, operating under stress are excluded, as well as the cumbersome electric trace heating cables and the sealed concrete cells for location of the primary equipment. The whole issue of protection from loss of primary coolant accidents is essentially eliminated. The hot leg vs. cold leg pump controversy is eliminated in favor of the preferable cold sodium location. Less metal is used for the components, and the total amount of construction work is greatly reduced. There are no nozzles on the tank wall with all penetrations being directly through the head. The surface area of load-bearing walls separating radioactive sodium from the external environment is significantly reduced. Absolute leak-tightness of the main primary circuit pipes is not required, as leaks would be confined within the pool vessel. The repairs on loop type plants of primary system leaks are time consuming and potentially hazardous. The loop may require draining, the cell in which it is contained must be de-inerted to permit personnel entry, care must be taken to ensure air doesn’t get into the unaffected parts of the primary system and that adjacent cells remain inerted. Provisions to accommodate draining a primary loop include tanks with cover gas, connecting piping, valves, pumping equipment, trace heating and instrumentation. Each of the above is a very powerful argument favoring the pool and taken together, they make a good case that the matter is decided, there is no contest.

But hold on a minute. The pool concept does have certain important disadvantages. In the case of Superphénix, the main tank diameter is nearly 69 ft. Since Superphénix was at commercial size, one would expect it to be representative of what is to be expected for any commercially sized pool concept plant. In fact, one might expect larger pool sizes to emerge in the quest for ever larger plant sizes. This compares with a reactor vessel diameter for a commercially sized loop plant as described in this monograph of about 28-32 ft. A vessel of the pool size can only be assembled on site from pieces fabricated elsewhere. The same statement applies to much of the vessel internals including the inner vessel and lower baffle. Of course, the guard vessel would require on site assembly as well. Fabrication on site is always more difficult (and riskier) than shop fabrication where the environment is better controlled, better skilled labor and engineering support are available and appropriate machinery is close at hand. It may be possible to make the argument that the EVST leak on Superphénix was a direct consequence of on-site fabrication.

As is obvious from Figure 50, the tank, its internal structures, and the deck are massive and complicated. Complex structures which will be exposed to a challenging environment must be carefully engineered into the system with considerable attention to the plant duty cycle. There are fewer options available for instrumentation and failed instrumentation is potentially less accessible. Since it is unlikely that all the events to which the plant will be exposed during its lifetime will not have been anticipated by the designer, questions arise as to the response of all this complexity to the unknown. The complexity arises from all the tasks the pool is obliged to perform. When too much is asked of a single machine, there will eventually come the point where the machine fails all its tasks. The question is how much is too much?

Support is another issue. Seismic response becomes more challenging as the vessel size is increased. Thickening of the vessel wall to address seismic requirements complicates the structural response to thermal transients, increases costs, and makes on site fabrication more difficult. Seismic isolation is an option, but was not chosen for the Superphénix plant for reasons that are not widely known, but possibly because seismic isolation was not a developed technology at the time.

Accommodation of thermal expansion of the coolant is another issue. In a loop type design the vessel is provided with an overflow nozzle to an overflow tank. As the sodium heats up the overflow tank fills. Since there are no nozzles on a pool, all thermal expansion must be accommodated by the pool vessel itself in the form of increased vessel height.

In order to prevent activation of the intermediate sodium passing through the IHXs, it is necessary to shield the core in a pool reactor to a greater extent than is required in a loop reactor. In fact, in a loop reactor, the only equivalent assemblies are the reflectors, part of whose purpose is to conserve neutrons with its shielding function being to protect the core barrel from excessive neutron fluence, a much less demanding requirement. This requirement for shielding would increase the diameter of the pool by about ten feet in the case of Superphénix had it not already been set by the IHXs.

Although the pool may involve less total construction, there is tremendous construction activity centered on the pool area. The scheduling of activities to prevent interferences during construction will inevitably lead to a longer critical path than would be the case for a loop design. The critical path on a pool type plant is almost certainly established by the pool itself. For a capital intensive construction project such as a power plant, there is a strong incentive to reduce construction time to a minimum so as to reduce the carrying cost of the financing required to support construction. Time is money when it comes to nuclear power plant construction. The figure below shows the upper part of the Superphénix tank during a phase of the construction

This image has an empty alt attribute; its file name is 50.jpg

Figure 57 Superphénix during construction

Combined with the complexity of this area during construction is the complexity of the end product that results from imposing so many functions to be performed in such a small area. Figures 49 & 50 omit all the auxiliary functions than need to be crammed into this space. It would be revealing to place a photograph here to make the point, if one could be readily located. Suffice it to point out that complexity in the head area was a huge problem for CRBRP and was much greater for Superphénix.

The IHX is yet another issue for the pool. Since it is located within the pool vessel itself and directly impacts the pool diameter, there is an incentive to minimize its heat transfer area. For the case of Superphénix, the IHX log mean temperature difference (LMTD) was 59°F. Lowering the LMTD to say 40°F enables either better steam conditions, lower HTS flow rates, or a combination of the two, both of which have economic impact. An IHX with a LMTD of 40°F would have a heat transfer area 47.5% greater than an IHX with a 59°F LMTD. This greater heat transfer area is likely to pose a much smaller economic impact for a loop plant than it would for a pool.

If one were motivated to use an EM pump for primary flow, it could not be buried in the pool and would probably need to be mounted on the head requiring long discharge piping. The incentive for adopting an EM pump for the PHTS is probably non-existent in a pool reactor in contrast to the loop where the incentive is great.

Earlier in this section the greater sodium inventory of the pool was given as one of its advantages. In fact, pool advocates routinely advance this argument. Mainly, this argument is centered on the response of the plant to a reactor trip after which the hot leg experiences a significant down temperature transient owing to the long coastdown time of the primary system centrifugal pumps. If the pressure drop across the core is reduced and the primary pumps are replaced with EM pumps, this issue vanishes. Following a reactor trip, if EM pumps are used the primary system flow rate declines promptly to natural circulation flow rate and the hot leg transient is greatly ameliorated. Other transients such as loss of heat sink and transient overpower are terminated with reactor trips and tend to be benign. A large sodium inventory is a liability rather than an asset. It results in components that are larger, heavier, and more expensive. Another problem with a large sodium inventory relates to air intrusion events such as the one that resulted in a two year outage at Superphénix. The large sodium inventory on that plant made it more difficult to detect that air intrusion was occurring. Once it was detected much time was required to clean up the system because of the large quantities of sodium involved. It would be much preferred for the designer to be motivated to minimize sodium inventory rather than maximizing it. The argument about increasing sodium inventory to mitigate transients crept into the CRBRP design with unfortunate results.

Another factor weighing on the pool vs. loop controversy relates to trends in the development of the LMFBR concept. There has long been an interest to push core burnup up to ever higher levels for economic reasons. Since the core creates new fuel in the process of operation, there is no neutronics reason why peak burnup of 30% or even higher could not be achieved. Combining heterogeneous core designs with high burnup capability, it is possible to design a core that would require refueling only infrequently. If, for example, one could design a core that would require refueling only once every ten years, a totally different approach to refueling system design is created as described in section 6. There is less incentive for exploring such options in a pool since all pool reactors have two rotating plugs and can accommodate frequent refuelings.

Earlier (Section 3), an Argonne National Laboratory feasibility study of a 10,000 MWth plant was discussed. Although this study could not be identified as anything more than a concept, nonetheless, it was embodied in a loop type design. The reactor vessel was 40 ft. in diameter and 64 ft. high. The point to be made here is one of limits. Eventually, it will become impractical to increase the size of components further in the quest for ever increasing plant sizes. In this regard, there is much more maneuvering room for plant size increases if one is starting with a 28-35 ft. diameter reactor vessel as opposed to a 69 ft. diameter pool.

To achieve capital cost improvement, there is no reason why a loop-type plant can’t be designed with fewer loops. For the pool, since it is desirable to minimize the pool diameter, there is an incentive to adopt small IHXs and fit them tightly inside the vessel. Superphénix had eight IHXs, which is relatively typical. Two IHXs were connected to a loop resulting in four primary loops. Loop type plants have no particular incentive to hold down the IHX size, opening the possibility of two loop plants.

The loop concept does not require that the reactor vessel be in the center of the containment. The vessel can be offset if there is a design advantage in doing so. In Japan, there has been interest in integrating the IHXs and the pumps into a single component. Doing so would eliminate the crossover pipe between the pump and the IHX. Another cost saving measure would be to eliminate the elevated loop concept and accommodate piping system expansion in a downward vertical loop contained within double-walled piping as was done on the Fermi-1 reactor. Interestingly, when Fermi-1 was being designed, consideration was given to the pool but the loop was chosen because of better access for maintenance of components, flexibility of design, and expectation that the loop would be less costly.5 These are examples of the design flexibility afforded by loop-type plants.

In a relatively recent IAEA conference on the breeder6, the Japanese representative present stated that Japan is continuing to develop the loop focusing on a 1500 MWe two loop concept. There is no a priori reason why loop type reactors must have three or four loops. The CRBRP had three loops because the original design called for the decay heat removal system to be taken off the steam generators through the Steam Generator Auxiliary Heat Removal System (SGAHRS). If one loop were inoperative, there had to be two more to provide decay heat removal redundancy using SGAHRS – thus three loops. As things turned out, the original CRBRP decay heat removal concept was not accepted by the NRC and a second system with air cooled heat exchangers off the primary system was installed (see section 8). Provided that decay heat removal is directly off the reactor or primary system, it is even possible that a single loop concept would be workable if it were to prove to be more economic. Both the SRE and SEFOR were single loop plants.

There is a design approach called the top entry concept7 which provides for decay heat removal directly off containment vessels surrounding the reactor and loop component vessels. This concept would be workable for two loop plants and has been, in fact, adopted for the JSFR-1500. The top entry concept with separate redundant decay heat removal loops could be used on a single loop plant. Even without top entry, a single loop concept with separate reactor vessel nozzles for a redundant decay heat removal system would be workable. If the pump and IHX are integrated to fit into a single vessel, and a single loop concept is adopted, one would be left with just two vessels rather than seven as on CRBRP, each of a much more manageable size than is the case with a pool type plant. The spacing of these two vessels could be chosen in such a way so as to optimize cost and improve constructability. Since there is a single reactor and a single turbine, having single heat exchange components in between, if feasible, could be simpler than having multiple IHXs and steam generators, although for plant reliability reasons, it may be desirable to retain multiple steam generators. The economic incentive for reducing the number of steam generators is not as great as is the case for the primary loops.

Another example is the reactor vessel height. On CRBRP, the interior dimension of the reactor vessel was 59 feet – all for housing a core with an active length of just over 5 ft. There are several steps that can be taken to reduce the length of the reactor vessel, some of which were described in section 6. Does a similar opportunity apply to the pool as well? The answer is, yes but probably not to the same extent, since it would be more difficult to eliminate in-vessel transfer in a pool and the pool must accommodate the IHXs and primary pumps. All of the foregoing suggests that there is considerable unrealized potential for improvement of the loop-type design which is less obvious for pool-type plants.

The following table trades off the advantages of each of the concepts.

Pool AdvantagesLoop Advantages
Eliminates separate vessels for IHXs and pumpsOn-site fabrication minimized
Eliminates thermal expansion loopsFewer critical path interferences
Eliminates overflow vesselBetter scale-up to larger sizes
Single guard vesselBetter IHTS separation
Close-in containmentBetter T/H optimization of IHXs
Reduced volume to be shieldedUse of PHTS EM pumps accommodated
No side penetrationsAir intrusion easier to detect
Maximizes sodium inventoryMinimizes sodium inventory

Note that the last item shows up on both sides — it is an advantage for the pool in providing greater thermal inertia and an advantage for the loop when EM pumps are deployed and a fast acting natural circulating DHRS is available. In-vessel storage of spent fuel could be considered an advantage for the pool. CRBRP had some limited in-vessel storage, but the “design concept” does not. In-vessel storage allows recovery of the decay heat from spent fuel assemblies and makes such assemblies easier to handle once they are transferred outside the reactor vessel. The first six pool advantages could be summarized as “compact PHTS” and the seventh advantage is not quantifiable. The loop advantages are better constructability, better operability, and better scale-ability.

The loop appears to be generally more adaptable to evolving design approaches than is the case with the more greatly constrained pool. Moreover, the economic argument favoring the pool appears to be vulnerable and certainly didn’t materialize on Superphénix. Because of the above considerations, the “design approach” is based on a loop design, albeit one very different from CRBRP.

Before leaving this subject, it needs to be acknowledged that the Russian BN-800, a pool reactor, was reported to have been completed (in 2016) for the equivalent of $2B, which would be competitive if the same could be accomplished in the U.S.  Of course, a report in a technical meeting is not the same as an audit to some acceptable accounting standard.  The applicable differences in labor, material, energy, and manufacturing costs would all need to be accounted for.  The ability of the Russian design to pass licensing requirements in the U.S. is an uncertainty.


1 During the 1980s, senior representatives of the Central Research Institute for the Electric Power Industry (CRIEPI) were known to favor the pool concept.  The government-sponsored entity Power Reactor & Nuclear Fuels Corporation (PNC) tended to favor the loop.

2 The Russians planned to install cold traps and a decay heat removal heat exchanger directly into the pool in their BN-1600 design.

3 The Russians support their pool vessels at the bottom using a skirt welded to the vessel.  This approach takes much of the load off the vessel wall, but results either in a movable deck or a requirement to accommodate motion between the pool and the deck.  In the case of the Russian design, the shroud surrounding the pumps and IHXs must be provided with a bellows where it attaches to the pool structure.

4 CRBRP containment compares poorly with Fermi-1 whose containment diameter was 72 ft.  Scale-up from Fermi-1 to CRBRP size would suggest the CRBRP containment diameter should have been 135 ft.

5 Fermi-1 – New Age for Nuclear Power, E. P. Alexanderson, ed., American Nuclear Society, 1979

6 Liquid Metal Cooled Reactors: Experience in Design and Operation, IAEA-TECDOC-1569, December, 2007

7 Passive cooling system for loop-type top entry liquid metal cooled reactors, Patent Application EP 0533351 A2, C. E. Boardman et al, General Electric Co., March 24, 1993