Hello, and welcome to Beyond NERVA! Today (and in the next series of posts, I’m trying to keep it a bit shorter!) we’re going to begin looking at another NASA nuclear program that’s been in the news a lot recently, Nuclear Thermal Propulsion (which is NASA’s preferred term for nuclear thermal rockets, or NTR). This is a system that is really misunderstood by the majority of those that I’ve seen comment on various articles on the subject, so in this series of (shorter and more frequent) blog posts we’re going to look at this system, and what makes it different from the NERVA-derived engines that have been proposed over the years.
If you’ve found your way to this blog, you are probably already familiar with nuclear thermal propulsion as a concept, and Project Rover was the most famous of the nuclear thermal programs that has been carried out worldwide. Project Rover is also the subject of the second video that I’m working on, and will be the example I use to teach the basics of NTRs as a complete engine. As such, I’m not going to go into either with much depth here.
In a nutshell, a thermal rocket uses heating from an outside source to produce an expansion of a propellant (usually cryogenic), which is then ejected out of a nozzle. This differs from most rockets, which use chemical combustion to produce expansion in the propellant(s, fuel and oxidizer), in many ways, but perhaps the most significant is in the fact that since combustion is not needed, the plumbing of the engine can be greatly simplified. The fact that fuel mixing, combustion efficiency, and leakage through fuel/oxidizer systems (with resultant explosive dangers) are not an issue also greatly helps matters from an engineering point of view. These subjects have been delved into beautifully by W Greene at “Inside the LEO Doghouse,” with two wonderful posts: first, looking at chemical engine cycles, and second looking at nuclear thermal rockets (I recommend reading both, available here for chemical and here for nuclear, even if you’re only marginally interested in how rockets work, because they really are wonderful pieces of writing, and the nuclear post builds directly on the chemical engine post). If you want a more general explanation of NTRs, please check out the NTR page.
What the US has Done Before
Project Rover was a program that went through many reorganizations and changes of sponsors during its early years: it started as a program at Los Alamos (LASL/LANL) and Lawrence Radiation (LRL, later Lawrence Livermore) Labs, and included aircraft reactors as well. It was decided in 1957 that the aircraft program would be transferred solely to LRL (and subsequently renamed Project Pluto), leaving nuclear thermal rocket development as LASL’s project under the Rover program. The Air Force had been a partner since the beginning in this program, but with the creation of NASA in 1958 they handed off their stake to the new space agency. Then, in 1977, the AEC became the Department of Energy (DOE). The program ended in 1977, despite meeting many of its test objectives and having one NTR design ready to fly. Many other designs waited in the wings, facilities had been built for all aspects of design, manufacture, and testing of these engines… and then the political winds changed in two different directions: the US canceled the manned Mars missions that were meant to follow the Apollo lunar missions (and thus the reason for the engine’s existence), and the growth of anti-nuclear activism led to the ending of not just Project Rover, but all US nuclear thermal testing powered by nuclear fission all the way to the present day (what testing has been done recently has been electrically heated, a process that we’ll look at more in depth later). This timeline is based on the in-depth overview offered by JD Finseth of Sverdup Corporation, working as a contractor for NASA. His Rover Nuclear Rocket Engine Program Engine Test Final Report, published in 1991, is the bible of the hot fire tests conducted through the various stages of engine development. This and a summary of historical NTR fuels by Kelsea Benensky are the primary sources for this post.
Looking back at Rover, the engine that was designed for flight as part of the Nuclear Energy for Rocket Vehicle Applications (or NERVA) program, the NRX-XE, was a solid core NTR using hydrogen propellant and graphite composite (GC) fuel clad in niobium carbide (NbC). This engine (as tested for the XE-PRIME test) sucked in 32 kg/s of cryogenic H2 to produce 244 kilonewtons (kN, 55 klbf) of thrust at a specific impulse of 710 s by heating it to 2475 C (2550 K fuel element outlet temperature) using the 1100 MWt nuclear reactor. This flight qualification test reactor burned at full power continuously for over an hour, and based on NERVA test data (and additional testing done at Westinghouse Astronuclear) However, the engines had a few problems that I’ll delve more into in detail below, because the potential solutions lead directly to design changes in the current LEU NTR that NASA is looking to test. Other, often flashier, problems that occurred during Rover were related to some of the first experiments with using cryogenic hydrogen as a propellant; happily, these issues have been resolved for the most part in the use of hydrogen propellant for (mostly upper stages) for chemical rockets. Issues that remain I hope to cover in the future, and the hydrogen zero-boil-off system will be featured in an early Dealing with Physics video, but we will touch on them briefly as they come up in this proposed engine as well.
The engines for the NERVA program were based on a design that was come up with at Los Alamos, the Phoebus reactor. This particular engine type (sometimes referred to as a Westinghouse A-type reactor) went through three major iterations (for hot-fire tests) and the XE-PRIME flight prototype alone was tested 24 times. Other concepts were proposed at the time as well, and a good part of the beginning of Beyond NERVA is going to be looking at those different concepts, especially PEWEE (a much smaller engine that’s closer to the engine size proposed more commonly today), which had a direct impact on what a NERVA-legacy engine that was built today would look like.
This reactor used hexagonal rods of graphite (or prisms) arranged in a roughly cylindrical reactor core, interspersed with tie tubes (more on these later). This was surrounded by a set of control drums, made of beryllium, with a coating of boron along one side to act as a neutron poison. These drums would be rotated to reflect or absorb more neutrons, and therefore control the reactivity of the core – and the power of the engine.
Graphite composite fuel is a good choice for a beginning NTR, because it has a high thermal capacity, is moldable and millable, and has fairly well-understood thermal expansion characteristics. However, it is highly susceptible to erosion in the propellant stream, so the propellant tubes must be clad to avoid major damage to the fuel element, and release of both fission products and unburned fuel into the propellant stream. As a practical matter, this clad is a milled tube, and a similar material is used to coat the outside of the fuel element as well.
As the name implies, this fuel is a composite of multiple materials: the fuel itself is uranium oxide (UO2, 95%+ enriched 235U), which is dispersed through a matrix of graphite which is pyrolitically deposited to form the basic fuel form shape. The details of ensuring even packing of this graphite consumed a lot of time and study at Los Alamos, and the problem has become well-understood. The fissile material itself came in multiple forms, the most common one being spheres of UO2 clad in pyrolitic graphite before FE manufacture to ensure a clad that would resist the release of fission products into the surrounding graphite, and other particle clad variations were experimented with as well. However, the issue of power distribution through the fuel element remained a technical challenge, and one that is exacerbated by the different isotopic and chemical compositions that occur throughout the more complex fuel element. More advanced FE designs focused on a more controlled distribution of fissile materials, resulting in increased performance out of a GC fuel element that would be used in a modern “rebuilding” of a NERVA engine.
This entire graphite hexagonal rod (called a prism) was then clad, the propellant tubes with niobium carbide (NbC) that was milled from bar stock to minimize erosion and thermal loading issues in the flow of hot hydrogen, and welded Mo clad surrounding the end caps and outside surfaces to prevent midband erosion and damage from the graphite-to-graphite interactions in the heavy vibrations that can occur in an NTR core. Again, the last 60 years of materials science have offered improvements to clad materials and manufacturing, and graphite deposition that requires extreme accuracy is now a relatively mundane task, as opposed to the major technical challenge it was during Rover.
Later versions used an unclad matrix of UO2 in the graphite, to increase the homogeneity (consistency) of the distribution of fissile fuel. This requires better cladding for the fuel elements, since the particle coatings are not available to catch fission products in the case of fuel element failure. However, the advances in clad materials allowed for this possibility, and it does improve the functioning of the engine.
The Problems with NERVA
Looking briefly at the problems that were most often encountered with the tested engines, common problems for the reactor itself were the fuel elements cracking due to sheer stresses and a phenomenon known as mid-band corrosion, where the clad (and then the fuel element itself) would be eroded by the combination of the intense heat and radiation in the core interacting with the hot hydrogen propellant and clad material.
Cracking across the narrow part of the fuel elements (transverse cracking) was a constant problem with the graphite composite fuel elements, leading to a number of hot fire tests aborted due to molten fragments of fuel elements being ejected into the Nevada desert. Because of their graphite composite construction, it’s very easy to shear the fuel elements along the line of deposition, i.e. along the short axis of the prism. This problem is seen in plastic extrusion 3D printing, as well, where the orientation of the printed model often has to take into account the structural needs of the final model to make sure it won’t be too fragile. Add in the relative brittleness of graphite and the vibrational and thermal extremes that the NTR required the tie tubes to deal with, and cracking seemed almost inevitable. Shorter lengths of GC stacked together in the clad had been experimented with, but rejected, before the initiation of NERVA as a potential solution. Having the clad support the fractured fuel element was another strategy that was used, but was complicated by the need to account for the different thermal expansion profiles the various parts of the reactor core had.
Many different clad types were experimented with, and it was found that milled cylinders that were then inserted into the pre-drilled fuel elements were the best option for erosion during Rover. This is an area that constant improvement has occurred in, and new manufacturing methods, materials, and dimensional trade-offs are proposed regularly (often for other high temperature gas core systems), and is looked at more in-depth in the fuel element page.
The Versatile Tie Tube
The shear force (vibrationally-caused transverse cracking) problems were solved using a very clever device known as a tie-tube. Beloved by many astronuclear geeks, this is a device that performs many different functions in the reactor: it provides structural support for the fuel elements, it moderates the neutron spectrum to reduce the required fuel loading, and it also collects thermal energy to power the turbopumps used for the propellant
A consistent problem with the Phoebus-derived reactors was that they were always starved for hydrogen, both as propellant and as moderator, and while the number and size of the holes in the fuel elements was constantly being increased—and the loading of fissile material in the fuel elements was constantly being tweaked – the problem remained. It was realized that whatever structural solution for the breaking fuel elements was going to be found would have to reside within the core, and therefore required its own cooling system. Hydrogen was the natural choice, as a good moderator that was already being used as propellant.
This is pumped first from the propellant tanks into the nozzle of the engine to cool it, then splits into two streams. The first stream enters the reactor vessel at the top and flows through the propellant channels in the fuel elements, to be ejected out the nozzle, but the second takes a longer route, traveling from the nozzle end of the reactor up to the top through the interior of the tie tube, then down the outer part of the tie tube to the bottom again before being fed into a turbine to drive the turbopumps. This now-cooler hydrogen is then used to provide roll control thrust through a smaller nozzle (hot bleed cycle). While in the core, this hydrogen both cools the tie tubes and provides neutron moderation, pushing this reactor into the epithermal spectrum. This has changed in modern NERVA-legacy engines, however, and the more efficient expander cycle is now the common design choice
However, while the tie tubes perform many roles in this engine, the primary reason for their existence is due to the nature of the fuel elements being used, namely the graphite composite elements. Here, the 235U that made up the nuclear fuel is spread through a graphite matrix. Since the graphite is built up by vapor deposition, distribution of the fissile fuel can be highly controlled with this method, and the form that it takes can vary from uranium in irregular patterns to pyrolitic graphite clad micro-granules similar to minuscule TRISO fuel elements that are sometimes used in either gas- or salt-cooled pebble-bed reactors. The graphite provides moderation for the reactor within the fuel element itself, and the ability to adjust the fuel loading was extensively experimented with, but there are drawbacks to graphite as a fuel for NTRs using hydrogen: it’s incredibly susceptible to hydrogen erosion, exposing both the unburned 235U and the fission products bound in the graphite to the exhaust stream and stripping them away. Cladding was the obvious solution.
Unfortunately, national interest in space waned as more and more astronauts landed on the Moon. First, the Mars missions were cut, then the last three Apollo lunar missions were dropped as well. Because the NERVA engine was specifically funded for the manned Mars missions, all NASA funding was stopped. LASL did two more years of work, closing the project up on the DOE end, and the program went into mothballs.
This meant that the engineers involved were able (now mostly as unpaid hobbyists, unfortunately) to reassess various engineering decisions that were made in deference to a schedule that was no longer there. Incremental, small improvements were suggested, and NASA’s plans were updated over time as dribbles of money could be found.
Other inherent engineering challenges in the engine were able to be addressed in a more leisurely fashion after the cancellation of the manned Mars missions, as well. In any major experimental engineering endeavor, the system as a whole isn’t necessarily optimized to function as well as it could for a variety of reasons. In order for the design to continue moving ahead, and for other decisions about the engine to be reached, certain specifications have to be defined earlier than others, and sometimes this leads to a limitation in the final design that is difficult or impossible to change later. For instance, the turbopump specified for the NRX flight engine was capable of pumping about 40 kg/s at 1360 psi, with a total mass of 243 kg. As is the case with any equipment that is commercially produced, the companies involved are constantly working to improve their product, and a more modern turbopump (for the Space Shuttle Main Engine, the usual benchmark for rocket engine comparison) can deliver 7206 kg/s at 7040 psi for only 350 kg. Since the hydrogen doesn’t just affect the thrust provided, and the cooling, but the neutron spectrum as well, the nuclear engineers needed to know exactly how much H2 was going through the pump… so a design was chosen, and the decision was frozen. Any upgrades to the turbopump are going to have to be analyzed to verify the effect of the increased flow rate on reactor dynamics, both neutronic and thermal, and any associated plumbing would have to be upgraded to the higher pressures that a more powerful turbopump would cause in the system. As in anything to do with either space or nuclear reactors, there are many things that go into an engine than the combustion chamber (or the reactor core).
The Rebirth of NTR at NASA
Research into nuclear thermal propulsion waxes and wanes. After the cancellation of NERVA, some research continued for a short time at LANL, but overall the program had ended. Some work continued on the high-temperature gas-cooled reactor being designed by General Electric (which was also seen as a possible NTR FE, as we’ll see later), but for actual NTR designs, NASA entered a drought.
The next NTR to be proposed was not for NASA, but for the Department of Defense, as part of the Strategic Defense Initiative (SDI, also known as Star Wars). This is a reactor that we’ll look at more in a later post, but in short it used fuel pebbles, held in place with centrifugal force by spinning the reactor core, in order to increase fuel element surface area, and therefor make the rocket more efficient. This engine design was not hot-fire tested (the only other one besides the NERVA engines that has is the Russian RD-0410), and as far as I can tell almost no hardware was built for it, either. The project was canceled when all SDI funds were methodically stripped out of any budget bill before congress, and the engineers involved moved on to other projects, still thinking about what could be done with an NTR.
During this time, NASA conducted some smaller-scale tests of NTR components, mostly chemical and thermal analyses of various materials that could be used to build the next-generation NTR. At this point, with no immediate mission, some of the basic assumptions and architectures could be relatively easily changed, so it was decided that a conference would be held to try and jump-start NTR development in the US again. This was the “Nuclear Thermal Propulsion Joint NASA/DOE/DOD Conference held in Albuquerque, NM from July 10th to 12th, 1990. The proceedings of the workshop can be found here.
To determine if something’s better than it was, though, you have to accurately assess where it would be today. Dr. Stanley Borowski (arguably the father of modern nuclear thermal rocketry at NASA) prepared a paper in 1990 for that same conference in Albuquerque, NM, titled “Nuclear Thermal Rocket Workshop Reference System -Rover/NERVA,” looking at these considerations. While this paper is now 27 years old, it still provides valuable insight into the impact of design decisions and the impact that new subsystems have on existing rocket engine designs. In it, he lays out the “baseline legacy NERVA engine” that is still referenced today (although it could still likely be improved using modern materials understanding and manufacturing techniques).
This design is based on the NERVA NRX-XE flight-configuration rocket, tested at Jackass Flats in 1969. This engine was a hot-bleed cycle (i.e. the hydrogen used to power the turbopumps was then vented outside the engine in a separate nozzle for roll and spin control) engine, although with modern computational analysis an expander cycle could be used to increase performance (for a more in-depth look at the different engine cycles, check out W Greene’s writing on “Inside the LEO Doghouse;” you can find the links here and here, as well as a great perspective on the Space Shuttle Main Engine in other posts of his). This reactor mostly used the same type of fuel as the NERVA A6 reactor, although many of the challenges that were presented during the manufacturing of these fuel elements have been solved (chemical vapor deposition for clad application, for example), and so the expected performance out of even the baseline graphite composite (with pyrolitic graphite coated fuel pebbles of various levels of fissile fuel content carefully distributed through the reactor) is likely higher than the numbers gathered in the 1960s would seem to indicate. Given these modernizations, it can reasonably be expected that a modernized, expander cycle GC NERVA engine designed for 75 klbf thrust would be able to operate at a chamber temperature of 2500 K, and a chamber pressure of 500 psia, with a 200:1 expansion ratio for the nozzle (rather than the 100:1 used in the 1970s). This results in an expected specific impulse of about 875 (which is very nice for high-thrust systems, but not astounding).
However, much of the benefit of modern materials and manufacturing gets hidden in the improved thrust to weight ratio, which jumped from 3.0 to 4.4 (with the internal shield). Turbopumps are lighter and more efficient, titanium pressure vessels offer much more strength for much less mass, and are commonplace today, and the regeneratively cooled nozzles for the Space Shuttle’s main engines had to deal with greater extremes than this engine will be able to produce, and did so with far less mass and a far higher degree of reliability. Many of the challenges that seemed to beset Project Rover constantly were non-nuclear in nature; those problems have mostly cropped up in other parts of aerospace development over the decades and have been addressed. This technology was already at Technology Readiness Level 6 (for an explanation of NASA’s TRL system, you can find a good one here), and that’s before taking these advances into account
There were also NTR programs in the USSR under different design bureaus, at different times; the design that ended up being tested is the RD-0140 and -0411, by Rosatom, Roscosmos, and NPO Luch. Instead of using graphite composite (GC), this reactor uses a carbide fuel element because it is far more temperature-resistant. Another thing that the Russians have researched extensively is an on-board effluent cleanup system to deal with radiological release in the rocket exhaust rather than cladding (the original design used unclad fuel elements, but apparently a contract was signed a number of years ago with Westinghouse to develop a clad for the FE’s – still trying to find it, though, I’ve just heard of it from a couple nuclear engineers I have met). Carbides are far harder, and more temperature and hydrogen resistant, than GC is; in this case a mix of two ternary carbides, UC-ZrC-NbC and UC-ZrC-C are used at different places in the reactor. This is an awesome system, with many advantages to the design from both a testing and an operations point of view. The Soviets weren’t the only ones interested in ternary carbide fuel elements: they were also a major area of study in the US, particularly for the nuclear ramjet under development at what was then Lawrence Radiation Lab (now Lawrence Livermore National Lab, LLL) and Idaho National Lab, working with Vought Aircraft, under Project Pluto from 1957 to 1964. Several different fuel elements were manufactured and tested in this program, and we’ll look at them more in-depth when we look at the Russian carbide designs. At the same 1991 conference, carbides were discussed for NTRs, and in the slide above the expected (at the time) operating parameters were very attractive. There’s a little more information on my NTR Fuel Elements page, but digging into Pluto isn’t something that I’ve had a chance to do much of yet. If anyone has more papers, especially about fuel element design and manufacture, core geometry, etc., that they could send me (or even better, link to in the comments!) I’d greatly appreciate it.
One of the other most attractive materials solutions was a curious mix of two materials that we deal with in everyday life: metals and ceramics. The CERMET fuel form was first proposed in the late 1950s at both Argonne National Labs and General Electric’s nuclear division in Cincinnati, OH, and is a composite of a ceramic fissile element (in this case highly enriched uranium oxide) and a metal matrix, and plays with the advantages that each material offers: The fact that it’s metal allows for alloying techniques that are well-understood for fuel element manufacture, and the ceramic fuel form was already well-understood in terms of radiochemistry and microstructure degradation when in direct contact with the fissioning uranium atoms. Even better, the fuel wouldn’t have the problem with transverse (sheer) cracking that the graphite composite fuel elements were constantly challenged by, because the majority of it was a metal, and therefore better at handling the vibrational effects due to its greater toughness… but that led to a complication: all of a sudden, the power source for your turbopumps wasn’t there anymore. There are two ways around this problem: First, you can use an external power source to drive your turbopumps (either chemically or electrically, as Rocket Lab does with their Electron rocket), or you find another way to extract energy out of the reactor core for your turbopumps. The second option is the way the designers of most CERMET-fueled reactors choose to solve the problem, often replacing fuel elements around the periphery with hydrogen feed tubes similar to tie tubes. This can cause challenges for a reactor designer, as we’ll see in a later blog post, in terms of evenly distributing power across the reactor, but by placing them around the periphery of the core the problem can be dealt with more easily. Other challenges reared their heads as well, perhaps the biggest of which was that the fuel elements swelled quite significantly when at operating temperature. This is something that happens with all nuclear fuel elements, and one that can be addressed for most designs (although some designs are able to handle this better than others, and it does provide challenges to reactor startup and shutdown).
This fuel form does allow a very interesting capability, though: it allows the possibility of low-enriched uranium fuel, by varying where in the fuel element the fissile material is, and what metals (each with different moderation and reflection properties) are distributed through the main body of the matrix in what densities. From a regulatory standpoint, this is a big deal, since virtually all in-space reactor designs up to this point have used highly enriched uranium fuel, and so to work on them requires extensive (and expensive) security procedures, and in order to do this the controlled distribution of moderator and fissile element within the matrix is a key enabling factor. Since 2012, the US government has been working to eliminate the use of highly enriched uranium wherever possible in the American nuclear fleet. Initially, shifting regulatory priorities (and pressure from the International Atomic Energy Agency) drove the push toward using low enriched uranium fuel, but since then the idea has gained traction in the industry (and academia) as well, because then universities with nuclear engineering programs could actually perform many of the tests required to fully verify a fuel form (and many Master’s and doctorates flow out of said testing). This fuel form is exciting because of the amount of flexibility it allows as far as chemical composition, fissile fuel distribution, and other factors that we’ll look at in the next post, so having this available more widely to research institutions across the country will lead to more innovation in this field, with more testing occurring to validate the systems that will fly, or could fly.
Certain reactors (such as the High Flux Isotope Reactor at Oak Ridge NL) need a very high neutron flux, either for irradiation or for feeding beamlines. Others, mainly the US Navy’s propulsion reactors, are designed to only be fueled once in their lifetime, which makes going to low enriched fuel essentially a non-starter (although research is currently being done into possibly developing a LEU naval reactor fuel by, among others, BWXT’s naval reactor program). Development of CERMET fuel began in the 1960s at both General Electric and Argonne National Labs for both NTR and aircraft propulsion reactors using a clad tungsten and pyrolitic clad UO2 particle composites, and then was stopped because carbide fuel seemed (with good reason) to be the best option for NTR fuels, and the aircraft reactor programs were canceled not long after.
For the majority of the history of in-space nuclear power, HEU has been the norm. Most of the numbers I’ve found say that 95-97% 235U is standard in both power and thermal reactors (TOPAZ, BES-5, NERVA/Rover, RD 0410, and KRUSTY/Kilopower are all 95+% enriched). It’s been widely assumed, in fact, that designing a LEU NTR is an exercise in futility, because the fission poisons would overwhelm the available fissile fuel, and the process of breeding 238U to 239Pu takes too long, is to slow, and too picky to be relied upon during a mission. Why rely on breeding when you can just make sure your uranium is all already fissile? It seems like bringing along only 20% 235U is a waste of mass. However, breeding of fuel occurs in all fuel elements, even in terrestrial reactors. By the end of a fuel pellet’s life, the vast majority of what’s being burned isn’t uranium, but plutonium that has been bred from the 238U in the initial fuel load. In a conversation on a Facebook group, a nuclear engineer with experience designing in-space systems ball-parked an estimate of a breeding ratio of approximately 1.01 as what would be needed to maintain power distribution and overcome the fission product buildup over the life of the fuel element (in this case a few dozen hours for a pure NTR, a number of months or years with a bimodal design, and about 18 months or so with current light water reactors). Since then, an interesting paper by Vishal Patel et al of the Center for Space Nuclear Research has come out that found a mass savings may be possible for a CERMET-fueled LEU system compared to an HEU one! The recent development history of CERMET fuels is a discussion for the next blog post, however.
CERMET gives us a few options to challenge that assumption, which we’ll look more at in the next blog post on these new fuel elements. The post after that will look at testing options that have been proposed over the years, and comparing them to what I’ve been able to find on the refit of NASA’s Stennis Space Center to allow nuclear thermal rockets to be tested at an already-existing test stand. Finally, we’ll look at the ships and missions that have been proposed for this NTR, and some variations on the design that have been proposed over the years, such as bimodal options.
All writing copyright Beyond NERVA, 2017. Images used with permission. Contact for reprint.