Development and Testing Forgotten Reactors History Non-nuclear Testing Nuclear Thermal Systems Test Stands

Pebblebed NTRs: Solid Fuel, but Different

Hello, and welcome back to Beyond NERVA!

Today, we’re going to take a break from the closed cycle gas core nuclear thermal rocket (which I’ve been working on constantly since mid-January) to look at one of the most popular designs in modern NTR history: the pebblebed reactor!

This I should have covered between solid and liquid fueled NTRs, honestly, and there’s even a couple types of reactor which MAY be able to be used for NTR between as well – the fluidized and shush fuel reactors – but with the lack of information on liquid fueled reactors online I got a bit zealous.

Beads to Explore the Solar System

Most of the solid fueled NTRs we’ve looked at have been either part of, or heavily influenced by, the Rover and NERVA programs in the US. These types of reactors, also called “prismatic fuel reactors,” use a solid block of fuel of some form, usually tileable, with holes drilled through each fuel element.

The other designs we’ve covered fall into one of two categories, either a bundled fuel element, such as the Russian RD-0410, or a folded flow disc design such as the Dumbo or Tricarbide Disc NTRs.

However, there’s another option which is far more popular for modern American high temperature gas cooled reactor designs: the pebblebed reactor. This is a clever design, which increases the surface area of the fuel by using many small, spherical fuel elements held in a (usually) unfueled structure. The coolant/propellant passes between these beads, picking up the heat as it passes between them.

This has a number of fundamental advantages over the prismatic style fuel elements:

  1. The surface area of the fuel is so much greater than with simple holes drilled in the prismatic fuel elements, increasing thermal transfer efficiency.
  2. Since all types of fuel swell when heated, the density of the packed fuel elements could be adjusted to allow for better thermal expansion behavior within the active region of the reactor.
  3. The fuel elements themselves were reasonably loosely contained within separate structures, allowing for higher temperature containment materials to be used.
  4. The individual elements could be made smaller, allowing for a lower temperature gradient from the inside to the outside of a fuel, reducing the overall thermal stress on each fuel pebble.
  5. In a folded flow design, it was possible to not even have a physical structure along the inside of the annulus if centrifugal force was applied to the fuel element structure (as we saw in the fluid fueled reactor designs), eliminating the need for as many super-high temperature materials in the highest temperature region of the reactor.
  6. Because each bead is individually clad, in the case of an accident during launch, even if the reactor core is breached and a fuel release into the environment occurs, the release of either any radiological components or any other fuel materials into the environment is minimized
  7. Because each bead is relatively small, it is less likely that they will sustain sufficient damage either during mechanical failure of the flight vehicle or impact with the ground that would breach the cladding.

However, there is a complication with this design type as well, since there are many (usually hundreds, sometimes thousands) of individual fuel elements:

  1. Large numbers of fuel beads mean large numbers of fuel beads to manufacture and perform quality control checks on.
  2. Each bead will need to be individually clad, sometimes with multiple barriers for fission product release, hydrogen corrosion, and the like.
  3. While each fuel bead will be individually clad, and so the loss of one or all the fuel will not significantly impact the environment from a radiological perspective in the case of an accident, there is potential for significant geographic dispersal of the fuel in the event of a failure-to-orbit or other accident.

There are a number of different possible flow paths through the fuel elements, but the two most common are either an axial flow, where the propellant passes through a tubular structure packed with the fuel elements, and a folded flow design, where the fuel is in a porous annular structure, with the coolant (usually) passing from the outside of the annulus, through the fuel, and the now-heated coolant exiting through the central void of the annulus. We’ll call these direct flow and folded flow pebblebed fuel elements.

In addition, there are many different possible fuel types, which regulars of this blog will be familiar with by now: oxides, carbides, nitrides, and CERMET are all possible in a pebblebed design, and if differential fissile fuel loading is needed, or gradients in fuel composition (such as using tungsten CERMET in higher temperature portions of the reactor, with beryllium or molybdenum CERMET in lower temperature sections), this can be achieved using individual, internally homogeneous fuel types in the beads, which can be loaded into the fuel support structure at the appropriate time to create the desired gradient.

Just like in “regular” fuel elements, these pebbles need to be clad in a protective coating. There have been many proposals over the years, obviously depending on what type of fissile fuel matrix the fuel uses to ensure thermal expansion and chemical compatibility with the fuel and coolant. Often, multiple layers of different materials are used to ensure structural and chemical integrity of the fuel pellets. Perhaps the best known example of this today is the TRISO fuel element, used in the US Advanced Gas Reactor fuel development program. The TRI-Structural ISOtropic fuel element uses either oxide or carbide fuel in the center, followed by a porous carbon layer, a pyrolitic carbon layer (sort of like graphite, but with some covalent bonds between the carbon sheets), followed by a silicon carbide outer shell for mechanical and fission product retention. Some variations include a burnable poison for reactivity control (the QUADRISO at Argonne), or use different outer layer materials for chemical protection. Several types have been suggested for NTR designs, and we’ll see more of them later.

The (sort of) final significant variable is the size of the pebble. As the pebbles go down in size, the available surface area of the fuel-to-coolant interface increases, but also the amount of available space between the pebbles decreases and the path that the coolant flows through becomes more resistant to higher coolant flow rates. Depending on the operating temperature and pressure, the thermal gradient acceptable in the fuel, the amount of decay heat that you want to have to deal with on shutdown (the bigger the fuel pebble, the more time it will take to cool down), fissile fuel density, clad thickness requirements, and other variables, a final size for the fuel pebbles can be calculated, and will vary to a certain degree between different reactor designs.

Not Just for NTRs: The Electricity Generation Potential of Pebblebed Reactors

Obviously, the majority of the designs for pebblebed reactors are not meant to ever fly in space, they’re mostly meant to operate as high temperature gas cooled reactors on Earth. This type of architecture has been proposed for astronuclear designs as well, although that isn’t the focus of this video.

Furthermore, the pebblebed design lends itself to other cooling methods, such as molten salt, liquid metal, and other heat-carrying fluids, which like the gas would flow through the fuel pellets, pick up the heat produced by the fissioning fuel, and carry it into a power conversion system of whatever design the reactor has integrated into its systems.

Finally, while it’s rare, pebblebed designs were popular for a while with radioisotope power systems. There are a number of reasons for this beyond being able to run a liquid coolant through the fuel (which was done on one occasion that I can think of, and we’ll cover in a future post): in an alpha-emitting radioisotope, such as 238Pu, over time the fuel will generate helium gas – the alpha particles will slow, stop, and become doubly ionized helium nuclei, which will then strip electrons off whatever materials are around and become normal 4He. This gas needs SOMEWHERE to go, which is why just like with a fissile fuel structure there are gas management mechanisms used in radioisotope power source fuel assemblies such as areas of vacuum, pressure relief valves, and the like. In some types of RTG, such as the SNAP-27 RTG used by Apollo, as well as the Multi-Hundred Watt RTG used by Voyager, the fuel was made into spheres, with the gaps in between the spheres (normally used to pass coolant through) are used for the gas expansion volume.

We’ll discuss these ideas more in the future, but I figured it was important to point out here. Let’s get back to the NTRs, and the first (and only major) NTR program to focus on the pebblebed concept: the Project Timberwind and the Space Nuclear Propulsion Program in the 1980s and early 1990s.

The Beginnings of Pebblebed NTRs

The first proposals for a gas cooled pebblebed reactor were from 1944/45, although they were never pursued beyond the concept stage, and a proposal for the “Space Vehicle Propulsion Reactor” was made by Levoy and Newgard at Thikol in 1960, with again no further development. If you can get that paper, I’d love to read it, here’s all I’ve got: “Aero/Space Engineering 19, no. 4, pgs 54-58, April 1960” and ‘AAE Journal, 68, no. 6, pgs. 46-50, June 1960,” and “Engineering 189, pg 755, June 3, 1960.” Sounds like they pushed hard, and for good reason, but at the time a pebblebed reactor was a radical concept for a terrestrial reactor, and getting a prismatic fueled reactor, something far more familiar to nuclear engineers, was a challenge that seemed far simpler and more familiar.

Sadly, while this design may end up have informed the design of its contemporary reactor, it seems like this proposal was never pursued.

Rotating Fluidized Bed Reactor (“Hatch” Reactor) and the Groundwork for Timberwind

Another proposal was made at the same time at Brookhaven National Laboratory, by L.P. Hatch, W.H. Regan, and a name that will continue to come up for the rest of this series, John R. Powell (sorry, can’t find the given names of the other two, even). This relied on very small (100-500 micrometer) fuel, held in a perforated drum to contain the fuel but also allow propellant to be injected into the fuel particles, which was spun at a high rate to provide centrifugal force to the particles and prevent them from escaping.

Now, fluidized beds need a bit of explanation, which I figured was best to put in here since this is not a generalized property of pebblebed reactors. In this reactor (and some others) the pebbles are quite small, and the coolant flow can be quite high. This means that it’s possible – and sometimes desirable – for the pebbles to move through the active zone of the reactor! This type of mobile fuel is called a “fluidized bed” reactor, and comes in several variants, including pebble (solid spheres), slurry (solid particulate suspended in a liquid), and colloid (solid particulate suspended in a gas). The best way to describe the phenomenon is with what is called the point of minimum fluidization, or when the drag forces on the mass of the solid objects from the fluid flow balances with the weight of the bed (keep in mind that life is a specialized form of drag). There’s a number of reasons to do this – in fact, many chemical reactions using a solid and a fluid component use fluidization to ensure maximum mixing of the components. In the case of an NTR, the concern is more to do with achieving as close to thermal equilibrium between the solid fuel and the gaseous propellant as possible, while minimizing the pressure drop between the cold propellant inlet and the hot propellant outlet. For an NTR, the way that the “weight” is applied is through centrifugal force on the fuel. This is a familiar concept to those that read my liquid fueled NTR series, but actually began with the fluidized bed concept.

This is calculated using two different relations between the same variables: the Reynolds number (Re), which determines how turbulent fluid flow is, and the friction coefficient (CD, or coefficient of drag, which deptermines how much force acts on the fuel particles based on fluid interactions with the particles) which can be found plotted below. The plotted lines represent either the Reynolds number or the void fraction ε, which represents the amount of gas present in the volume defined by the presence of fuel particles.

Hendrie 1970

If you don’t follow the technical details of the relationships depicted, that’s more than OK! Basically, the y axis is proportional to the gas turbulence, while the x axis is proportional to the particle diameter, so you can see that for relatively small increases in particle size you can get larger increases in propellant flow rates.

The next proposal for a pebble bed reactor grew directly out of the Hatch reactor, the Rotating Fluidized Bed Reactor for Space Nuclear Propulsion (RBR). From the documentation I’ve been able to find, from the original proposal work continued at a very low level at BNL from the time of the original proposal until 1973, but the only reports I’ve been able to find are from 1971-73 under the RBR name. A rotating fuel structure, with small, 100-500 micrometer spherical particles of uranium-zirconium carbide fuel (the ZrC forming the outer clad and a maximum U content of 10% to maximize thermal limits of the fuel particles), was surrounded by a reflector of either metallic beryllium or BeO (which was preferred as a moderator, but the increased density also increased both reactor mass and manufacturing requirements). Four drums in the reflector would control the reactivity of the engine, and an electric motor would be attached to a porous “squirrel cage” frit, which would rotate to contain the fuel.

Much discussion was had as to the form of uranium used, be it 235U or 233U. In the 235U reactor, the reactor had a cavity length of 25 in (63.5 cm), an inner diameter of 25 in (63.5 cm), and a fuel bed depth when fluidized of 4 in (10.2 cm), with a critical mass of U-ZrC being achieved at 343.5 lbs (155.8 kg) with 9.5% U content. The 233U reactor was smaller, at 23 in (56 cm) cavity length, 20 in (51 cm) bed inner diameter, 3 in (7.62 cm) deep fuel bed with a higher (70%) void fraction, and only 105.6 lbs (47.9 kg) of U-ZrC fuel at a lower (and therefore more temperature-tolerant) 7.5% U loading.

233U was the much preferred fuel in this reactor, with two options being available to the designers: either the decreased fuel loading could be used to form the smaller, higher thrust-to-weight ratio engine described above, or the reactor could remain at the dimensions of the 235U-fueled option, but the temperature could be increased to improve the specific impulse of the engine.

There was als a trade-off between the size of the fuel particles and the thermal efficiency of the reactor,:

  • Smaller particles advantages
    • Higher surface area, and therefore better thermal transfer capabilities,
    • Smaller radius reduces thermal stresses on fuel
  • Smaller particles disadvantages
    • Fluidized particle bed fuel loss would be a more immediate concern
    • More sensitive to fluid dynamic behavior in the bed
    • Bubbles could more easily form in fuel
    • Higher centrifugal force required for fuel containment
  • Larger particle advantages
    • Ease of manufacture
    • Lower centrifugal force requirements for a given propellant flow rate
  • Larger particle disadvantages
    • Higher thermal gradient and stresses in fuel pellets
    • Less surface area, so lower thermal transfer efficiency

It would require testing to determine the best fuel particle size, which could largely be done through cold flow testing.

These studies looked at cold flow testing in depth. While this is something that I’ve usually skipped over in my reporting on NTR development, it’s a crucial type of testing in any gas cooled reactor, and even more so in a fluidized bed NTR, so let’s take a look at what it’s like in a pebblebed reactor: the equipment, the data collection, and how the data modified the reactor design over time.

Cold flow testing is usually the predecessor to electrically heated flow testing in an NTR. These tests determine a number of things, including areas within the reactor that may end up with stagnant propellant (not a good thing), undesired turbulence, and other negative consequences to the flow of gas through the reactor. They are preliminary tests, since as the propellant heats up while going through the reactor, a couple major things will change: first, the density of the gas will decrease and second, as the density changes the Reynolds number (a measure of self-interaction, viscosity, and turbulent vs laminar flow behavior) will change.

In this case, the cold flow tests were especially useful, since one of the biggest considerations in this reactor type is how the gas and fuel interact.

The first consideration that needed to be examined is the pressure drop across the fuel bed – the highest pressure point in the system is always the turbopump, and the pressure will decrease from that point throughout the system due to friction with the pipes carrying propellant, heating effects, and a host of other inefficiencies. One of the biggest questions initially in this design was how much pressure would be lost from the frit (the outer containment structure and propellant injection system into the fuel) to the central void in the body of the fuel, where it exits the nozzle. Happily, this pressure drop is minimal: according to initial testing in the early 1960s (more on that below), the pressure drop was equal to the weight of the fuel bed.

The next consideration was the range between fluidizing the fuel and losing the fuel through literally blowing it out the nozzle – otherwise known as entrainment, a problem we looked at extensively on a per-molecule basis in the liquid fueled NTR posts (since that was the major problem with all those designs). Initial calculations and some basic experiments were able to map the propellant flow rate and centrifugal force required to both get the benefit of a fluidized bed and prevent fuel loss.

Rotating Fluidized Bed Reactor testbed test showing bubble formation,

Another concern is the formation of bubbles in the fuel body. As we covered in the bubbler LNTR post (which you can find here), bubbles are a problem in any fuel type, but in a fluid fueled reactor with coolant passing through it there’s special challenges. In this case, the main method of transferring heat from the fuel to the propellant is convection (i.e. contact between the fuel and the propellant causing vortices in the gas which distributes the heat), so an area that doesn’t have any (or minimal) fuel particles in it will not get heated as thoroughly. That’s a headache not only because the overall propellant temperature drops (proportional to the size of the bubbles), but it also changes the power distribution in the reactor (the bubbles are fission blank spots).

Finally, the initial experiment set looked at the particle-to-fluid thermal transfer coefficients. These tests were far from ideal, using a 1 g system rather than the much higher planned centrifugal forces, but they did give some initial numbers.

The first round of tests was done at Brookhaven National Laboratory (BNL) from 1962 to 1966, using a relatively simple test facility. A small, 10” (25.4 cm) length by 1” (2.54 cm) diameter centrifuge was installed, with gas pressure provided by a pressurized liquefied air system. 138 to 3450 grams of glass particles were loaded into the centrifuge, and various rotational velocities and gas pressures were used to test the basic behavior of the particles under both centrifugal force and gas pressure. While some bobbles were observed, the fuel beds remained stable and no fuel particles were lost during testing, a promising beginning.

These tests provided not just initial thermal transfer estimates, pressure drop calculations, and fuel bed behavioral information, but also informed the design of a new, larger test rig, this one 10 in by 10 in (25.4 by 25.4 cm), which was begun in 1966. This system would not only have a larger centrifuge, but would also use liquid nitrogen rather than liquefied air, be able to test different fuel particle simulants rather than just relatively lightweight glass, and provide much more detailed data. Sadly, the program ran out of funding later that year, and the partially completed test rig was mothballed.

Rotating Fluidized Bed Reactor (RBR): New Life for the Hatch Reactor

It would take until 1970, when the Space Nuclear Systems office of the Atomic Energy Commission and NASA provided additional funding to complete the test stand and conduct a series of experiments on particle behavior, reactor dynamics and optimization, and other analytical studies of a potential advanced pebblebed NTR.

The First Year: June 1970-June 1971

After completing the test stand, the team at BNL began a series of tests with this larger, more capable equipment in Building 835. The first, most obvious difference is the diameter of the centrifuge, which was upgraded from 1 inch to 10 inches (25.4 cm), allowing for a more prototypical fuel bed depth. This was made out of perforated aluminum, held in a stainless steel pressure housing for feeding the pressurized gas through the fuel bed. In addition, the gas system was changed from the pressurized air system to one designed to operate on nitrogen, which was stored in liquid form in trailers outside the building for ease of refilling (and safety), then pre-vaporized and held in two other, high-pressure trailers.

Photographs were used to record fluidization behavior, taken viewing the bottom of the bed from underneath the apparatus. While initially photos were only able to be taken 5 seconds apart, later upgrades would improve this over the course of the program.

The other major piece of instrumentation surrounded the pressure and flow rate of the nitrogen gas throughout the system. The gas was introduced at a known pressure through two inlets into the primary steel body of the test stand, with measurements of upstream pressure, cylindrical cavity pressure outside the frit, and finally a pitot tube to measure pressure inside the central void of the centrifuge.

Three main areas of pressure drop were of interest: due to the perforated frit itself, the passage of the gas through the fuel bed, and finally from the surface of the bed and into the central void of the centrifuge, all of which needed to be measured accurately, requiring calibration of not only the sensors but also known losses unique to the test stand itself.

The tests themselves were undertaken with a range of glass particle sizes from 100 to 500 micrometers in diameter, similar to the earlier tests, as well as 500 micrometer copper particles to more closely replicate the density of the U-ZrC fuel. Rotation rates of between1,000 and 2,000 rpm, and gas flow rates from 1,340-1,800 scf/m (38-51 m^3/min) were used with the glass beads, and from 700-1,500 rpm with the copper particles (the lower rotation rate was due to gas pressure feed limitations preventing the bed from becoming fully fluidized with the more massive particles).

Finally, there were a series of physics and mechanical engineering design calculations that were carried out to continue to develop the nuclear engineering, mechanical design, and system optimization of the final RBR.

The results from the initial testing were promising: much of the testing was focused on getting the new test stand commissioned and calibrated, with a focus on figuring out how to both use the rig as it was constructed as well as which parts (such as the photography setup) could be improved in the next fiscal year of testing. However, particle dynamics in the fuidized bed were comfortably within stable, expected behavior, and while there were interesting findings as to the variation in pressure drop along the axis of the central void, this was something that could be worked with.

Based on the calculations performed, as well as the experiments carried out in the first year of the program, a range of engines were determined for both 233U and 235U variants:

Work Continues: 1971-1972

This led directly into the 1971-72 series of experiments and calculations. Now that the test stand had been mostly completed (although modifications would continue), and the behavior of the test stand was now well-understood, more focused experimentation could continue, and the calculations of the physics and engineering considerations in the reactor and engine system could be advanced on a more firm footing.

One major change in this year’s design choices was the shift toward a low-thrust, high-isp system, in part due to greater interest at NASA and the AEC in a smaller NTR than the original design envelope. While analyzing the proposed engine size above, though, it was discovered that the smallest two reactors were simply not practical, meaning that the smallest design was over 1 GW power level.

Another thing that was emphasized during this period from the optimization side of the program was the mass of the reflector. Since the low thrust option was now the main thrust of the design, any increase in the mass of the reactor system has a larger impact on the thrust-to-weight ratio, but reducing the reflector thickness also increases the neutron leakage rate. In order to prevent this, a narrower nozzle throat is preferred, but also increases thermal loading across the throat itself, meaning that additional cooling, and probably more mass, is needed – especially in a high-specific-impulse (aka high temperature) system. This also has the effect of needing higher chamber pressures to maintain the desired thrust level (a narrower throat with the same mass flow throughput means that the pressure in the central void has to be higher).

These changes required a redesign of the reactor itself, with a new critical configuration:

Hendrie 1972

One major change is how fluidized the bed actually is during operation. In order to get full fluidization, there needs to be enough inward (“upward” in terms of force vectors) velocity at the inner surface of the fuel body to lift the fuel particles without losing them out the nozzle. During calculations in both the first and second years, two major subsystems contributed hugely to the weight and were very dependent on both the rotational speed and the pellet size/mass: the weight of the frit and motor system, which holds the fuel particles, and the weight of the nozzle, which not only forms the outlet-end containment structure for the fuel but also (through the challenges of rocket motor dynamics) is linked to the chamber pressure of the reactor – oh, and the narrower the nozzle, the less surface area is available to reject the heat from the propellant, so the harder it is to keep cool enough that it doesn’t melt.

Now, fluidization isn’t a binary system: a pebblebed reactor is able to be settled (no fluidization), partially fluidized (usually expressed as a percentage of the pebblebed being fluidized), and fully fluidized to varying degrees (usually expressed as a percentage of the volume occupied by the pebbles being composed of the fluid). So there’s a huge range, from fully settled to >95% fluid in a fully fluidized bed.

The designers of the RBR weren’t going for excess fluidization: at some point, the designer faces diminishing returns on the complications required for increased fluid flow to maintain that level of particulate (I’m sure it’s the same, with different criteria, in the chemical industry, where most fluidized beds actually are used), both due to the complications of having more powerful turbopumps for the hydrogen as well as the loss of thermalization of that hydrogen because there’s simply too much propellant to be heated fully – not to mention fuel loss from the particulate fuel being blown out of the nozzle – so the calculations for the bed dynamics assumed minimal full fluidization (i.e. when all the pebbles are moving in the reactor) as the maximum flow rate – somewhere around 70% gas in the fuel volume (that number was never specifically defined that I found in the source documentation, if it was, please let me know), but is dependent on both the pressure drop in the reactor (which is related to the mass of the particle bed) and the gas flow.

Ludewig 1974

However, the designers at this point decided that full fluidization wasn’t actually necessary – and in fact was detrimental – to this particular NTR design. Because of the dynamics of the design, the first particles to be fluidized were on the inner surface of the fuel bed, and as the fluidization percentage increased, the pebbles further toward the outer circumference became fluidized. Because the temperature difference between the fuel and the propellant is greater as the propellant is being injected through the frit and into the fuel body, more heat is carried away by the propellant per unit mass, and as the propellant warms up, thermal transfer becomes less efficient (the temperature difference between two different objects is one of the major variables in how much energy is transferred for a given surface area), and fluidization increases that efficiency between a solid and a fluid.

Because of this, the engineers re-thought what “minimal fluidization” actually meant. If the bed could be fluidized enough to maximize the benefit of that dynamic, while at a minimum level of fluidization to minimize the volume the pebblebed actually took up in the reactor, there would be a few key benefits:

  1. The fueled volume of the reactor could be smaller, meaning that the nozzle could be wider, so they could have lower chamber pressure and also more surface area for active cooling of the nozzle
  2. The amount of propellant flow could be lower, meaning that turbopump assemblies could be smaller and lighter weight
  3. The frit could be made less robustly, saving on weight and simplifying the challenges of the bearings for the frit assembly
  4. The nozzle, frit, and motor/drive assembly for the frit are all net neutron poisons in the RBR, meaning that minimizing any of these structures’ overall mass improves the neutron economy in the reactor, leading to either a lower mass reactor or a lower U mass fraction in the fuel (as we discussed in the 233U vs. 235U design trade-off)

After going through the various options, the designers decided to go with a partially fluidized bed. At this point in the design evolution, they decided on having about 50% of the bed by mass being fluidized, with the rest being settled (there’s a transition point in the fuel body where partial fluidization is occurring, and they discuss the challenges of modeling that portion in terms of the dynamics of the system briefly). This maximizes the benefit at the circumference, where the thermal difference (and therefore the thermal exchange between the fuel and the propellant) is most efficient, while also thermalizing the propellant as much as possible as the temperature difference decreases from the propellant becoming increasingly hotter. They still managed to reach an impressive 2400 K propellant cavity temperature with this reactor, which makes it one of the hottest (and therefore highest isp) solid core NTR designs proposed at that time.

This has various implications for the reactor, including the density of the fissile component of the fuel (as well as the other solid components that make up the pebbles), the void fraction of the reactor (what part of the reactor is made up of something other than fuel, in this particular instance hydrogen within the fuel), and other components, requiring a reworking of the nuclear modeling for the reactor.

An interesting thing to me in the Annual Progress Report (linked below) is the description of how this new critical configuration was modeled; while this is reasonably common knowledge in nuclear engineers from the days before computational modeling (and even to the present day), I’d never heard someone explain it in the literature before.

Basically, they made a bunch of extremely simplified (in both number of dimensions and fidelity) one-dimensional models of various points in the reactor. They then assumed that they could rotate these around that elevation to make something like an MRI slice of the nuclear behavior in the reactor. Then, they moved far enough away that it was different enough (say, where the frit turns in to the middle of the reactor to hold the fuel, or the nozzle starts, or even the center of the fuel compared to the edge) that the dynamics would change, and did the same sort of one-dimensional model; they would end up doing this 18 times. Then, sort of like an MRI in reverse, they took these models, called “few-group” models, and combined them into a larger group – called a “macro-group” – for calculations that were able to handle the interactions between these different few-group simulations to build up a two-dimensional model of the reactor’s nuclear structure and determine the critical configuration of the reactor. They added a few other ways to subdivide the reactor for modeling, for instance they split the neutron spectrum calculations into fast and thermal, but this is the general shape of how nuclear modeling is done.

Ok, let’s get back to the RBR…

Experimental testing using the rotating pebblebed simulator continued through this fiscal year, with some modifications. A new, seamless frit structure was procured to eliminate some experimental uncertainty, the pressure measuring equipment was used to test more areas of the pressure drop across the system, and a challenge for the experimental team – finding 100 micrometer copper spheres that were regularly enough shaped to provide a useful analogue to the UC-ZrC fuel (Cu specific gravity 8.9, UC-ZrC specific gravity ~6.5) were finally able to be procured.

Additionally, while thermal transfer experiments had been done with the 1-gee small test apparatus which preceded the larger centrifugal setup (with variable gee forces available), the changes were too great to allow for accurate predictions on thermal transfer behavior. Therefore, thermal transfer experiments began to be examined on the new test rig – another expansion of the capabilities of the new system, which was now being used rigorously since its completing and calibration testing of the previous year. While they weren’t conducted that year, setting up an experimental program requires careful analysis of what the test rig is capable of, and how good data accuracy can be achieved given the experimental limitations of the design.

The major achievement for the year’s ex[experimentation was a refining of the relationship between particle size, centrifugal force, and pressure drop of the propellant from the turbopump to the frit inlet to the central cavity, most especially from the frit to the inner cavity through the fuel body, on a wide range of particle sizes, flow rates, and bed fluidization levels, which would be key as the design for the RBR evolved.

The New NTR Design: Mid-Thrust, Small RBR

So, given the priorities at both the AEC and NASA, it was decided that it was best to focus primarily on a given thrust, and try and optimize thrust-to-weight ratios for the reactor around that thrust level, in part because the outlet temperature of the reactor – and therefore the specific impulse – was fixed by the engineering decisions made in regards to the rest of the reactor design. In this case, the target thrust was was 90 kN (20,230 lbf), or about 120% of a Pewee-class engine.

This, of course, constrained the reactor design, which at this point in any reactor’s development is a good thing. Every general concept has a huge variety of options to play with: fuel type (oxide, carbide, nitride, metal, CERMET, etc), fissile component (233U and 235U being the big ones, but 242mAm, 241Cf, and other more exotic options exist), thrust level, physical dimensions, fuel size in the case of a PBR, and more all can be played with to a huge degree, so having a fixed target to work towards in one metric allows a reference point that the rest of the reactor can work around.

Also, having an optimization point to work from is important, in this case thrust-to-weight ratio (T/W). Other options, such as specific impulse, for a target to maximize would lead to a very different reactor design, but at the time T/W was considered the most valuable consideration since one way or another the specific impulse would still be higher than the prismatic core NTRs currently under development as part of the NERVA program (being led by Los Alamos Scientific Laboratory and NASA, undergoing regular hot fire testing at the Jackass Flats, NV facility). Those engines, while promising, were limited by poor T/W ratios, so at the time a major goal for NTR improvement was to increase the T/W ratio of whatever came after – which might have been the RBR, if everything went smoothly.

One of the characteristics that has the biggest impact on the T/W ratio in the RBR is the nozzle throat diameter. The smaller the diameter, the higher the chamber pressure, which reduces the T/W ratio while increasing the amount of volume the fuel body can occupy given the same reactor dimensions – meaning that smaller fuel particles could be used, since there’s less chance that they would be lost out of the narrower nozzle throat. However, by increasing the nozzle throat diameter, the T/W ratio improved (up to a point), and the chamber pressure could be decreased, but at the cost of a larger particle size; this increases the thermal stresses in the fuel particles, and makes it more likely that some of them would fail – not as catastrophic as on a prismatic fueled reactor by any means, but still something to be avoided at all costs. Clearly a compromise would need to be reached.

Here are some tables looking at the design options leading up to the 90 kN engine configuration with both the 233U and 235U fueled versions of the RBR:

After analyzing the various options, a number of lessons were learned:

  1. It was preferable to work from a fixed design point (the 90 kN thrust level), because while the reactor design was flexible, operating near an optimized power level was more workable from a reactor physics and thermal engineering point of view
  2. The main stress points on the design were reflector weight (one of the biggest mass components in the system), throat diameter (from both a mass and active cooling point of view as well as fuel containment), and particle size (from a thermal stress and heat transfer point of view)
  3. On these lower-trust engines, 233U was looking far better than 235U for the fissile component, with a T/W ratio (without radiation shielding) of 65.7 N/kg compared to 33.3 N/kg respectively
    1. As reactor size increased, this difference reduced significantly, but with a constrained thrust level – and therefore reactor power – the difference was quite significant.

The End of the Line: RBR Winds Down

1973 was a bad year in the astronuclear engineering community. The flagship program, NERVA, which was approaching flight ready status with preparations for the XE-PRIME test, the successful testing of the flexible, (relatively) inexpensive Nuclear Furnace about to occur to speed not only prismatic fuel element development but also a variety of other reactor architectures (such as the nuclear lightbulb we began looking at last time), and the establishment of a robust hot fire testing structure at Jackass Flats, was fighting for its’ life – and its’ funding – in the halls of Congress. The national attention, after the success of Apollo 11, was turning away from space, and the missions that made NTR technologically relevant – and a good investment – were disappearing from the mission planners’ “to do” lists, and migrating to “if we only had the money” ideas. The Rotating Fluidized Bed Reactor would be one of those casualties, and wouldn’t even last through the 1971/72 fiscal year.

This doesn’t mean that more work wasn’t done at Brookhaven, far from it! Both analytical and experimental work would continue on the design, with the new focus on the 90 kN thrust level, T/W optimized design discussed above making the effort more focused on the end goal.

Multi-program computational architecture used in 1972/73 for RBR, Hoffman 1973

On the analytical side, many of the components had reasonably good analytical models independently, but they weren’t well integrated. Additionally, new and improved analytical models for things like the turbopump system, system mass, temp and pressure drop in the reactor, and more were developed over the last year, and these were integrated into a unified modeling structure, involving multiple stacked models. For more information, check out the 1971-72 progress report linked in the references section.

The system developed was on the verge of being able to do dynamics modeling of the proposed reactor designs, and plans were laid out for what this proposed dynamic model system would look like, but sadly by the time this idea was mature enough to implement, funding had run out.

On the experimental side, further refinement of the test apparatus was completed. Most importantly, because of the new design requirements, and the limitations of the experiments that had been conducted so far, the test-bed’s nitrogen supply system had to be modified to handle higher gas throughput to handle a much thicker fuel bed than had been experimentally tested. Because of the limited information about multi-gee centrifugal force behavior in a pebblebed, the current experimental data could only be used to inform the experimental course needed for a much thicker fuel bed, as was required by the new design.

Additionally, as was discussed from the previous year, thermal transfer testing in the multi-gee environment was necessary to properly evaluate thermal transfer in this novel reactor configuration, but the traditional methods of thermal transfer simply weren’t an option. Normally, the procedure would be to subject the bed to alternating temperatures of gas: cold gas would be used to chill the pebbles to gas-ambient temperatures, then hot gas would be used on the chilled pebbles until they achieved thermal equilibrium at the new temperature, and then cold gas would be used instead, etc. The temperature of the exit gas, pebbles, and amount of gas (and time) needed to reach equilibrium states would be analyzed, allowing for accurate heat transfer coefficients at a variety of pebble sizes, centrifugal forces, propellant flow rates, etc. would be able to be obtained, but at the same time this is a very energy-intensive process.

An alternative was proposed, which would basically split the reactor’s propellant inlet into two halves, one hot and one cold. Stationary thermocouples placed through the central void in the centrifuge would record variations in the propellant at various points, and the gradient as the pebbles moved from hot to cold gas and back could get good quality data at a much lower energy cost – at the cost of data fidelity reducing in proportion to bed thickness. However, for a cash-strapped program, this was enough to get the data necessary to proceed with the 90 kN design that the RBR program was focused on.

Looking forward, while the team knew that this was the end of the line as far as current funding was concerned, they looked to how their data could be applied most effectively. The dynamics models were ready to be developed on the analytical side, and thermal cycling capability in the centrifugal test-bed would prepare the design for fission-powered testing. The plan was to address the acknowledged limitations with the largely theoretical dynamic model with hot-fired experimental data, which could be used to refine the analytical capabilities: the more the system was constrained, and the more experimental data that was collected, the less variability the analytical methods had to account for.

NASA had proposed a cavity reactor test-bed, which would serve primarily to test the open and closed cycle gas core NTRs also under development at the time, which could theoretically be used to test the RBR as well in a hot-fore configuration due to its unique gas injection system. Sadly, this test-bed never came to be (it was canceled along with most other astronuclear programs), so the faint hope for fission-powered RBR testing in an existing facility died as well.

The Last Gasp for the RBR

The final paper that I was able to find on the Rotating Fluidized Bed Reactor was by Ludewig, Manning, and Raseman of Brookhaven in the Journal of Spacecraft, Vol 11, No 2, in 1974. The work leading up to the Brookhaven program, as well as the Brookhaven program itself, was summarized, and new ideas were thrown out as possibilities as well. It’s evident reading the paper that they still saw the promise in the RBR, and were looking to continue to develop the project under different funding structures.

Other than a brief mention of the possibility of continuous refueling, though, the system largely sits where it was in the middle of 1973, and from what I’ve seen no funding was forthcoming.

While this was undoubtedly a disappointing outcome, as virtually every astronuclear program in history has faced, and the RBR never revived, the concept of a pebblebed NTR would gain new and better-funded interest in the decades to come.

This program, which has its own complex history, will be the subject for our next blog post: Project Timberwind and the Space Nuclear Thermal Propulsion program.


While the RBR was no more, the idea of a pebblebed NTR would live on, as I mentioned above. With a new, physically demanding job, finishing up moving, and the impacts of everything going on in the world right now, I’m not sure exactly when the next blog post is going to come out, but I have already started it, and it should hopefully be coming in relatively short order! After covering Timberwind, we’ll look at MITEE (the whole reason I’m going down this pebblebed rabbit hole, not that the digging hasn’t been fascinating!), before returning to the closed cycle gas core NTR series (which is already over 50 pages long!).

As ever, I’d like to thank my Patrons on Patreon (, especially in these incredibly financially difficult times. I definitely would have far more motivation challenges now than I would have without their support! They get early access to blog posts, 3d modeling work that I’m still moving forward on for an eventual YouTube channel, exclusive content, and more. If you’re financially able, consider becoming a Patron!

You can also follow me at for more regular updates!


Rotating Fluidized Bed Reactor

Hendrie et al, “ROTATING FLUIDIZED BED REACTOR FOR SPACE NUCLEAR PROPULSION Annual Report: Design Studies and Experimental Results, June, 1970- June, 1971,” Brookhaven NL, August 1971

Hendrie et al, “ROTATING FLUIDIZED BED REACTOR FOR SPACE NUCLEAR PROPULSION Annual Report: Design Studies and Experimental Results, June 1971 – June 1972,” Brookhaven NL, Sept. 1972

Hoffman et al, “ROTATING FLUIDIZED BED REACTOR FOR SPACE NUCLEAR PROPULSION Annual Report: Design Studies and Experimental Results, July 1972 – January 1973,” Brookhaven NL, Sept 1973

Cavity Test Reactor



Forgotten Reactors History Nuclear Thermal Systems

Dumbo: America’s First Forgotten NTR

Hello, and welcome back to Beyond NERVA! Today, in our first post in the Forgotten Reactors series, we’re going back to the beginnings of astronuclear engineering, and returning to nuclear thermal propulsion as well, looking at one of the reactors that’s had a cult following since the 1950s: the pachydermal rocket known as DUMBO.

In a nuclear thermal rocket, the path that the propellant takes has a strong impact on how hard it is to predict the way that the propellant will move through the reactor. Anyone who’s dealt with a corroded steam central heating system that won’t quit knocking, no matter how much you try, has dealt with the root of the problem: fluid behavior in a set of tubes only makes sense, and doesn’t cause problems, if you can make sure you know what’s going on, and that’s not only counter-intuitively hard, but it’s one of the subjects that (on the fine scale, in boundary conditions, and in other extremes) tends to lead towards chaos theory more than traditional fluid dynamics of ANY sort, much less adding in the complications of heat transport. However, if you can have gas flow longer through the reactor, you can get greater efficiency, less mass, and many other advantages.

This was first proposed in the Dumbo reactor at the beginning of Project Rover, alongside the far more familiar Kiwi reactors. Rather than have the gas flow from one end of the reactor to the other through straight pipes, like in Kiwi, the propellant in Dumbo would flow part of the way down the reactor core, then move radially (sideways) for a while, and then returns to flowing along the main axis of the reactor before exiting the nozzle. Because of the longer flow path, and a unique fuel element and core geometry, Dumbo seemed to offer the promise of both less volume and less mass for the same amount of thrust due to this difference in flow path. Additionally, this change offered the ability to place thermally sensitive materials more evenly across the reactor, due to the distribution of the cold propellant through the fuel element structure.

Dumbo ended up being canceled, in part, because the equations required to ensure that fatal flow irregularities wouldn’t occur, and the promised advantages didn’t materialize, either. None of this means that Dumbo was a bad idea, just an idea ahead of its time – an idea with inspiration to offer. Dumbo’s progeny live on. In fact, we’ve covered both the fuel element form AND the modern incarnation of the fuel geometry in the blog before!

With today’s knowledge of materials, advanced flow modeling, a cutting edge carbide fuel, and the beginnings of a Renaissance in nuclear design are breathing new life into the program even today, and the fundamental concept remains an attractive (if complex) one.

The First Forgotten Reactor

Early Dumbo cutaway drawing with flow path

In the early days of astronuclear engineering, there was a lot of throwing pasta at the wall, and looking to see what stuck. Many more slide rules than are involved in the average family’s spaghetti dinner preparations, to determine if the pasta was done enough, but a large number of designs were proposed, and eventually settled down into four potentially near-ish term useful: radioisotope power supplies, nuclear thermal propulsion, nuclear electric propulsion, and nuclear explosive propulsion (which we usually call nuclear pulse propulsion). Each of these ended up being explored extensively, and a number of other novel concepts have been proposed over the years as well. In the beginning, however, research tended toward either the NTR or NPP designs, with two major programs: ROVER and ORION. Orion was fairly narrowly focused from the beginning, owing to the problems of making an efficient, low-mass, easily deployable, reliable, and cheap shaped nuclear charge – the physics drove the design. Rover, on the other hand, had many more options available to it, and some competition as to what the best design was. Being the earliest days of the atomic era, which way to go, and the lack of knowledge in both nuclear and materials science often limited Rover as much as lack of fuel for their reactors did! This led to some promising designs being discarded. Some were resurrected, some weren’t, but the longest lived of the less-initially-preferred designs is our subject for today.

Dumbo was initially proposed in the literature in 1955. Two years later, a far more developed report was issued to address many of the challenges with the concept. The idea would continue to bounce around in the background of astronuclear engineering design until 1991, when it was resurrected… but more on that later. The concept was very different from the eventual NERVA concept (based on the Phoebus test reactor during Rover) in a number of ways, but two stand out:

1. Fuel element type and arrangement: The eventual Rover elements used uranium oxide or carbide suspended within graphite flour, which was then solidified, in Dumbo the fissile fuel was “metal.” However, the designers used this term differently than how it would be used today: rather than have the entire fuel element be metal, as we’ve seen in Kilopower, the fuel was uranium oxide pellets suspended in some type of high temperature metal. Today, we call this CERMET, for ceramic metal composite, and is the current favorite

2. Flow pattern: while both the initial Rover concepts (the Kiwi reactors) and the eventual NERVA engines used straight-through, axial propellant flow, which is simple to model mathematically, Dumbo’s flow path started the same (going from the nozzle end to the spacecraft end, cooling the reflectors and control components), but once it reached the top of the reactor and started flowing toward the nozzle, things changed. The flow would start going toward the nozzle through a central column, but be diverted through sets of corrugated fuel “washers” and spacers, making two 90 degree turns as it did so. This was called a “folded flow” system.

A host of other differences were also present throughout the reactor and control systems, but these two differences were the biggest when comparing the two nearly-simultaneously developed systems. The biggest advantages that were offered by the basic concept were the ability to go to higher temperatures in the core, and be able to have a more compact and less massive reactor for the same thrust level. Additionally, at the time it seemed like the testing would be far simpler to do, because it appeared that the number of tests needed, and the requirements of those tests, would make the testing program both simpler and cheaper compared to the competing Kiwi design concept. Sadly, these advantages weren’t sufficient to keep the project alive, and Kiwi ended up winning in the down-selection process.

In 1959, the Dumbo portion of Rover was canceled. There were two stated main reasons: first, there were no weight savings seen between the two systems upon in depth analysis; second, the manufacture of the components for the reactor required high precision out of at-the-time exotic materials. Another concern which was not mentioned at the time of cancellation but is a concern for certain variations on this reactor is the complex flow patterns in the reactor, something we’ll touch on briefly later.

Contrary to popular belief, Dumbo’s design isn’t dead. The fuel type has changed, and many of the nuclear design considerations for the reactor have also changed, but the core concept of a stacked set of fuel discs and a folded flow pattern through the core of the reactor remains. Originally revived as the Advanced Dumbo concept, proposed by Bill Kirk at LANL in 1990, which advocated for the use of carbide fuels to increase the reactor core temperature, as well as moving to a solid disc with grooves cut in it. This was proposed at the same time as many other concepts for nuclear thermal rockets in the bout of optimism in the early 1990s, but funding was given instead to the pebblebed NTR, another concept that we’ll cover. This in itself evolved into the Tricarbide Grooved Ring NTR currently under investigation at the Marshall Space Flight Center, under the direction of Brian Taylor and William Emrich, a concept we covered already in the carbide fuel post, but will briefly review again at the end of this post.

Is Dumbo Really a Metal Reactor?

At the time, this was called a metal reactor, but there’s metal and there’s metal. Metal fuels aren’t uncommon in nuclear reactor design. CANDU reactors are one of the most common reactor types in operation today, and use metal fuel. New designs, such as Kilopower in space and the Westinghouse eVinci reactor on Earth, also use metal fuels, alloying the uranium with another metal to improve either the chemical, thermal, or nuclear properties of the fuel itself. However, there are a few general problems (and exceptions to those problems) with metal fuels. In general, metal fuels have a low melting point, which is exactly what is undesirable in a nuclear thermal rocket, where core temperature is the main driving factor to efficiency, even ahead of propellant mass. Additionally, there can be neutronic complications, in that many metals which are useful for the fuel material components are also neutron poisons, reducing the available power of the reactions in the core. On the flip side, metals generally offer the best thermal conductivity of any class of material.

CERMET fuel micrograph, image NASA

Rather than a metal alloy fuel such as CANDU or Kilopower reactors, Dumbo used uranium oxide embedded in a refractory metal matrix. For those that have been following the blog for a while, this isn’t metal, it’s CERMET (ceramic-metal composite), the same type of fuel that NASA is currently exploring with the LEU NTP program. However, the current challenges involved in developing this fuel type are a wonderful illustration as to why it was considered a stretch in the 1950s. For a more in-depth discussion on CERMET fuels, check out our blog post on CERMET fuels in their modern incarnation here:

The metal matrix of these fuel elements was meant to be molybdenum initially, with the eventual stretch goal of using tungsten. Tungsten was still brand new, and remains a challenge to manufacture in certain cases. Metallurgists and fabricators are still working on improving our ability to use tungsten, and isotopically enriching it (in order to reduce the number of neutrons lost to the metal) is still beyond the technical capabilities of American metallurgical firms. The Dumbo fuel elements were to be stamped in order to account for the complex geometries involved, although there was a new set of challenges involved with this forming process, including ensuring even distribution of the fissile fuel through the stamped material.

Folded Flow Reactors: Why, and How Hard?

Perhaps the biggest challenge in Dumbo wasn’t the material the fuel elements were made of, but the means of transferring the heat into the propellant. This was due to a couple of potential issues: first, the propellant passed through a more convoluted than typical path through the reactor, and second, the reactor was meant to be a laminar flow heat exchanger, the first time that this would have been done.

Dumbo fuel stack flow pattern, original image DOE

Each Dumbo core had a number of sets of fuel washers, moderator spacers, and securing components stacked into cylinders. The propellant would flow through the Be reflector, into the central opening of the fuel elements, and then flow out of the fuel elements, exiting around the perimeter of the cylinder. This would then be directed out the nozzle to provide thrust. By going through so many twists and turns, and having so much surface area available for heat transfer, the propellant could be more thoroughly heated than in a more typical prismatic fuel element, such as we see with the later Kiwi and Phoebus reactors. As with folded optics in telescopes, folded flow paths allow for more linear distance traveled in the same volume. A final advantage is that, because of the shape and arrangement of the washers, only a small amount of material would need to be tested, at a relatively minor 1.2 kW power level, to verify the material requirements of the reactor.

Timber Wind NTR, image DOE
Timber Wind NTR, image DOE

This sort of flow path isn’t unique to Dumbo. TRISO fuel, which use beads of fuel coated in pyrolitic carbon, fission product containment materials, and others have a very complex flow path through the reactor, increasing the linear distance traveled from one end of the core to the other well beyond the linear dimensions of the reactor. The differences mainly arise in the fuel geometry, not the concept of a non-axial flow.

The challenge is modeling the flow of the coolant through the bends in the reactor. It’s relatively easy to have hot spots develop if the fluid has to change directions in the wrong way, and conversely cold spots can develop as well. Ensuring that neither of these happen is a major challenge in heat exchanger design, a subject that I’m far from qualified to comment on.

The unique concept at the time was that this was meant to be a laminar flow heat exchanger (the fuel elements themselves form the heat exchanger). Laminar fluid flow, in broad terms, means that all of the molecules in the fluid are moving together. The opposite of laminar flow is turbulent flow, where eddies form in the fluid that move in directions other than the main direction of fluid flow. While the term may bring up images of white water rapids (and that’s not a bad place to start), the level of turbulence varies depending on the system, and indeed the level of turbulence in a heat exchanger modifies how much heat is transferred from the hot surface to the coolant fluid. Since the molecules are moving together in the same direction during laminar flow, the eddies that are a major component of heat transfer in some designs are no longer present, reducing the efficiency of heat transport through the working fluid. However, in some designs (those with a low Reynolds number, a characteristic of heat transfer capability) laminar flow can be more efficient than turbulent flow. For more discussion on the efficiency of laminar vs turbulent flow in heat exchangers, check out this paper by Patil et al: .

For a rocket engine, the presence of laminar flow makes the rocket itself more efficient, since all of the molecules are moving in the same direction: straight out of the nozzle. The better collimated, or directional, the propellant flow is, the more thrust efficient the engine will be. Therefore, despite the fact that laminar flow is less efficient at transferring heat, the heat that is transferred can be more efficiently imparted as kinetic energy into the spacecraft.

In the case of Dumbo, the use of a large number of small orifices in the fuel elements allows for the complete transferrance of the heat of the nuclear reaction into the propellant, allowing for the efficient use of laminar flow heat exchange. This also greatly simplifies the basic design calculations of the fluid dynamics of the reactor, since laminar flow is easy to calculate, but turbulence requires complexity theory to fully model, a technique that didn’t exist at the time. However, establishing and maintaining laminar flow in the reactor was rightly seen as a major challenge at the time, and even over three decades later the challenges involved in this part of the design remained a point of contention about the feasibility of the laminar heat exchanger concept in this particular application.

Another distinct advantage to this layout is that the central annulus of each fuel element stack was filled with propellant that, while it had cooled the radial reflector, remained quite cool compared to the rest of the reactor. This meant that materials containing high hydrogen content, in this case a special form of plastic foam, could be distributed throughout the reactor. This meant that the neutron spectrum of the reactor could be more consistent, ensuring more uniform fissioning of the fuel across the active region of the reactor, and a material could be chosen that allows for greater neutron moderation than the graphite fuel element matrix of a Kiwi-type reactor. A variation of this concept can be seen as well with the Russian RD-0140 and -0411, which have all of their fuel around the outer circumference of the reactor’s pressure vessel and a large moderator column running down the center of the core. This allows the center of the core of the reactor to be far cooler, and contain far more themally sensitive materials as a result.

The Death of Dumbo

Sadly, the advantages of this reactor geometry weren’t sufficient to keep the program alive. In 1959, Dumbo gained the dubious distinction of being the first NTR concept that underwent study and development to be canceled in the US (perhaps even worldwide). Kiwi, the father and grandfather of all other Rover flight designs, was the winner, and the prismatic fuel element geometry remains the preferred design even today.

According to the Quarterly Status Report of LASL ROVER Program for Period Ending September 20, 1959, two factors caused the cancellation of the reactor: the first was that, despite early hopes, the reactor’s mass offered no advantages over an equivalent Kiwi reactor; the second was the challenges involved in the fabrication and testing of many of the novel components required, and especially the requirements of manufacturing and working the UO2/Mo CERMET fuel elements to a sufficiently precise degree, promised a long and difficult development process for the reactor to come to fruition.

Dumbo remained an interesting and attractive design to students of astronuclear engineering from that point on. Mentions of the concept occur in most summaries of NTR design history, but sadly, it never attracted funding to be developed. Even the public who are familiar with NTRs have heard of Dumbo, even if they aren’t familiar with any of the details. Just last month, there was a thread started on NASASpaceFlight Forum about Dumbo, and reviving the concept in the public eye once again.

The Rebirth of Dumbo: the Advanced Dumbo Rocket

Advanced Dumbo fuel element stack. Notice the change in fuel shape due to the different material properties. Image NASA

In 1991, there was a major conference attended by NASA, DOE, and DOD personnel on the subject of NTRs, and the development of the next generation NTR system for American use to go to Mars. At this conference, Bill Kirk of Los Alamos National Labs presented a paper on Dumbo, which he was involved in during its first iteration, and called for a revival of what he called a “folded flow washer type” NTR. This proposal, though, discarded the UO2/Mo CERMET fuel type in favor of a UC-ZrC carbide fuel element, to increase the fuel element maximum temperature. For a more in-depth look at carbide fuel elements, and their use in NTRs, check out the carbide fuel element post here. As we discussed in the carbide post, there are problems with thermal stress cracking and complex erosive behaviors in carbide fuel elements, but the unique form factor of the grooved disc allows for better distribution of the stresses, less continuous structural components to the fuel elements themselves, allowing for better thermal behavior and less erosion. Another large change from the classic Dumbo to the Advanced Dumbo was that the fluid flow through the reactor wasn’t meant to be laminar, and turbulent behavior was considered acceptable in this iteration. Other changes, including reflector geometry, were also incorporated, to modernize the concept’s support structures and ancillary equipment.

Timber Wind reactor, image Winchell Chung Atomic Rockets

Once again, though, the Dumbo concept, as well as the other concepts presented that had a folded flow pattern, were not selected. Instead, this conference led to the birth of the Timber Wind program, a pebble bed reactor design that we’ll cover in the future. Again, though, the concept of increasing the surface area compared to the axial length of the reactor was an inherent part of this design, and a TRISO pebble bed reactor shares some of the same advantages as a washer-type reactor would.

The Second Rebirth: the Tricarbide Grooved Ring NTR

Tricarbide Grooved Ring NTR fuel element stack. Notice the return of more complex geometry as materials design and fabrication of carbides has improved. Image NASA

Washer type reactors live today, and in many ways the current iteration of the design is remarkably similar to the Advanced Dumbo concept. Today, the research is centered in the Marshall Space Flight Center, with both Oak Ridge National Laboratory and the University of Tennessee being partners in the program. The Tricarbide Grooved Ring NTR (TCGR) was originally proposed in 2017, by Brian Taylor and Bill Emrich. While Bill Kirk is not mentioned in any of the papers on this new iteration of this reactor geometry, the carbide grooved washer architecture is almost identical to the Advanced Dumbo, so it’s reasonable to assume that the TCGR design is at least inspired by the Advanced Dumbo concept of 27 years before (Bill Emrich is a very old hand in NTR design and development, and was active at the time of the conference mentioned above).

The latest iteration, the TCGR, is a design that we covered in the carbide fuel element post, and because of this, as well as the gross similarities between the Advanced Dumbo and TCGR, we won’t go into many details here. If you want to learn more, please check out the TCGR page here: insert link. The biggest differences between the Advanced Dumbo and TCGR were the flow pattern and the fuel element composition.

The flow pattern is a simple change in one way, but in another way there’s a big difference: rather than the cold end of the reactor being the central annular portion of the fuel element stack, the cold end became the exterior of the stack, with the hot propellant/coolant running down the center of the core. This difference is a fairly significant one from a fluid dynamics point of view, where the gas flow from the “hot end” of the reactor itself to the nozzle turns from a more diffuse set of annular shaped gas flows into a series of columns coming out of each fuel element cluster; whether this is easier to work with or not, and what the relative advantages are, is beyond my understanding, but [take this with a grain of salt, this is speculation] it seems like the more collimated gas flows would be able to integrate more easily into a single gaseous flow through the nozzle.

Similarly to the simple but potentially profound change in the propellant flow path, the fuel element composition change is significant as well. Rather than just using the UC-ZrC fuel composition, the TCGR uses a mix of uranium, zirconium, and tantalum carbides, in order to improve both thermal properties as well as reduce stress fractures. For more information on this particular carbide type, check out the carbides post!

Funding is continuing for this concept, and while the focus is primarily on the CERMET LEU NTP engine under development by BWXT, the TCGR is still a viable and attractive concept, and one that balances the advantages and disadvantages of the washer-type, folded flow reactor. As more information on this fascinating reactor becomes available, I’ll post updates on the reactor’s page!

More Coming Soon!

This was the first of a new series, the Forgotten Reactors. Next week will be another post in the series, looking at the SP-100 reactor. We won’t look at the reactor in too much depth, because it shares a lot of similarities with the SNAP-50 reactor’s final iteration; instead we’ll look at the most unique thing about this reactor: it was designed to be both launched and recovered by the Space Shuttle, leading to some unique challenges. While the STS is no longer flying, this doesn’t mean that the lessons learned with this design process are useless, because they will apply to a greater or lesser extent to every reactor recovery operation that will be performed in the future, and well as the challenges of having a previously-turned-on reactor in close proximity to the crew of a spacecraft with minimal shielding between the payload compartment and the crew cabin.


Dumbo — A Pachydermal Rocket Motor, DOE ID LAMS-1887 McInteer et al, Los Alamos Scientific Laboratory 1955

A Metal Dumbo Rocket Reactor, DOE ID LA-2091, Knight et al, Los Alamos Scientific Laboratory 1957

Quarterly Status Report of LASL Rover Program for Period Ending Sept 20, 1959, LAMS-2363

Dumbo, a Pachydermal Rocket Motor [Advanced Dumbo], Kirk, Los Alamos National Laboratory 1992

Investigation of a Tricarbide Grooved Ring Fuel Element for a Nuclear Thermal Rocket, NASA ID 20170008951 Taylor et al NASA MSFC 2017
Conference paper:
Presentation slides: