The First US RTG

Hello, and welcome back to Beyond NERVA! I’m sorry it’s taken so long to get back into the groove of writing, but this blog series will (I hope) mark the rebirth of this blog.

Additionally, I will continue to pick away at updating the website. Finally, I am now offering the (publicly available) papers I use for reference with my handwritten notes, as well as the initial, hand-written draft of each post and major page, to my Patrons. Y’all have been so patient, I’m looking forward to being able to begin properly saying thank you. For those that would like to read about the reasons for my hiatus, I cover it here.

Now, let’s begin a look into the long, fascinating history of not just the well-known RTG, but the entire family of RPS. There are many fascinating and forgotten ideas that are well worth remembering.Today, though, we will begin at the beginning, with the first US RTG.

Mound Meets the Us Signals Corps

While the use of RTGs in certain space missions is now de rigeur, everything has to start somewhere: In the case of RTGs, it wasn’t even in space! This post was actually a happy accident, while looking into fuel options, but the story of the first American RTG offers many lessons about the wide range of options in this often derided technology.

Our story starts in New Jersey, at the US Army Signal Corps Research Laboratory at Fort Monmouth. While work into RTGs started at Mound Laboratory as early as 1953, and the first patents were filed in 7954, this work was largely theoretical. It did catch the attention of the signals Corps, though, which led to a contract in 1957 for a demonstration design. Being a proof of concept, rather than an operational device, it was modest in its requirements, but with the expectation that it would lead to a useful device a short time afterward.

The design challenges the team faced, as well as the limitations in both technology and infrastructure, offer a fascinating glimpse into not only the early history of this fascinating bit of technology, but may show glimpses of possible futures as well!

The Contract

The Signals Corps were intrigued at the possibility of powering equipment using nuclear power, and an RTG offered both more mobility and safety than a reactor – not to mention a much more reasonable level of power. With this in mind, they issued a contract and set of requirements to Mound Laboratory to 

"... conduct research into radioactive materials and thermocouples suitable for the direct energy conversion of heat to electricity." 

US Army Signals Corps, 1955

Unfortunately, the details of the contract itself (R-65-8- 99877-S C-03-97), as well as the supporting requirements documents (58-E L S/R-3765 and 58-ELS/386757) from the signals Corps, as well as the Army documentation on the project (3-99-15-102) do not seem to be available online.

Design Requirements

In order to fulfill their customers’ needs, a number of considerations had to be addressed by the team at Mound.

All of these decisions would have a profound impact on the device. Additionally, care was taken to look at the future of this technology as well; what good would an overly expensive, impractical device be to either the U.s. Army or Mound? The sheer number of questions, design options, limitations, and unknowns was daunting, just as it is at the beginning of any new technology’s development. Let’s begin with the heart of any RTG: the radioisotope fuel.

Fuel Selection and Composition

Being incredibly early, the number of possible isotopes to use was staggering. The team began by assessing about 1300 isotopes, rapidly down- selecting until 47 were left.This was done based on the specific activity of the isotope (defined as the radioactivity per quantity of a particular isotope) as well as the isotope having a useful half-life (too short and the device is too short-lived as a result, too long and the amount of fuel required becomes unwieldy, and finally the prevalence of easily absorbed radiation (primarily alpha and beta emission.The threshold values they decided on were:

With these criteria met, the remaining isotopes were assessed for production, rarity of precursor isotopes, isotopic enrichment requirements, fuel chemical composition, and other practical matters. Some, like 749 Eu and 148 Gd, lacked feasible – or even known – production methods for the isotope. Others, such as 42Ar and 194Os, are second order irradiation products, meaning the seed isotope must absorb two neutrons, greatly reducing yield.

Many were fission products of various likelihoods of production, from the very common (like 1446, whose mass makes up 5.7% of fissioned uranium to the very rare(755 Eu is a good example, whose mass makes up only 0.03% of fissioned uranium). Most were the result of reactor irradiation of particular isotopes, meaning that enrichment of the seed material is a major concern, as well as neutron cross-section (to determine yield as well as impacts on local reactor criticality and final isotopic enrichment challenges. Still others were only able to be produced in a particle accelerator. Some, such as the Ca and Sn isotopes, lacked a way to enrich them to a useful purity. To say that this research was challenging is an understatement. especially given the few technical sources available.

In the end, due to its many advantages in an experimental, short term technology demonstrator, 210Po was selected as a fuel. This short-lived alpha emitter is easily and cheaply produced. It has a high specific energy (Z), and can be used (at low temperatures) as a pure metal. At the time, the most common method of production was through milking Ra. This has relatively low yields, however, which means that a method involving neutron bombardment in a reactor was developed. While the discussion of the isotopic fuel options was fascinating, we’ll save that for another time.

Perhaps the biggest downside to 210 Po is its toxicity, both chemical and radiological. Certain governments are even known to take advantage of this fact. For our designers, though, this meant that additional measures were needed to ensure safe handling and operation of the power source.

Initially, the 210Po metal was acquired through Oak Ridge National Laboratory, which was refined at Oak Ridge (the metallurgical laboratories were not yet fully developed), and then milled to size at Mound. This was then canned in a 0. 75″ long, 0. 35″ diameter steel cylinder, with a smaller cylinder within providing additional containment. Later containment vessels would often use Haynes 25, a corrosion-resistant nickel alloy.

The final consideration, while simple in most cases, is how the fuel elements are arranged in the RTG. As is the standard today, the elements were arranged in what was called a “common core”: a central cylinder designed to manage fuel outgassing (caused by not radioactive decay and subsequent chemical byproducts but unavoidable material contamination while providing a central location for the fuel. With the isotope selected, produced, and refined, as well as clad to prevent the release of the highly reactive and toxic metal, the rest of the generator remained to be designed and tested.

Power Conversion System

The next challenge was how to convert the heat to electricity using the thermoelectric effect. While it has the significant advantage of providing solid state power conversion, efficiencies were pitifully low (-1%) was considered acceptable in this case. Just getting there was a major undertaking. requiring far more analysis than had been applied before. Thermoelectric generators (TEGs) operate on the principle that an electric current across two different conductive materials will generate heat at one end of the juncture and cold at the other (the Peltier effect), and that the reverse-a heat gradient turning into an electric current-is also possible (the Seebeck effect). These were further refined by that inescapable giant of thermodynamics, Lord Kelvin, into the Thompson Effect.

In practice, two differently conductive materials cause electrons to behave slightly differently in the two materials; with the input of one type of energy it is converted to the other.The metric used to determine the level of difference is called the Seebeck coefficient (S) , and is usually measured in mS. S can be positive or negative, with S>0 meaning that the hot side of the junction has higher charge, and the reverse is true as well. Two materials are formed into the two different sub-components, called “shoes.” These come in two varieties: p-type and n-type semiconductors. These materials are specialized to be the positive (p-type) and negative (n-type) ends of the electric circuit. The Seebeck coefficient defining the charge orientation across the junction combined with the placement of the feet determines the direction of current flow through the generator.

Sadly, nothing can be that simple.The Seebeck coefficient changes based on temperature, which is variable based on the heat source, the thermal conductivity of the material, the steepness and distribution of the thermal gradient-it’s a whole thing that we still are learning about-and improving-to this day. Because of all these effects, many different combinations of materials are used depending on temperature and application.


The most common application for the thermoelectric effect is in temperature monitoring. the ubiquitous thermocouple. This device is highly insulated at the outlet, with two wires running out the cold end. The power generated is directly related to temperature, providing highly accurate temperature data based on small voltage changes.

While not individually significant, large numbers of thermocouples can generate small but significant electrical power – the question is how exactly to go about it.

Since the power is derived from the junction between the materials, a large block of an n-type foot with a p-type one would be useless, but the impracticalities of stringing a careful grid of thermocouples are significant. The solution is simple: take a whole bunch of thermocouples, set the circuit to be in series (each thermocouple passes on the current of all the preceding ones in addition to its own) with one positive and one negative connection for the bundle. This bundle is called a thermopile.

The thermopiles were interesting in this design: thin mica sheets held Constantan-Chrome thermocouples in a flaring array, widening to the cold junction.The hot junction rested in a groove milled into the exterior of the fuel casing, with later versions using thermally conductive cement to secure the joint and increase thermal cohesion. The cold junction was held clamped between two electrically insulated aluminum bars, which in turn would mount to two support rings while also forming most of the surface of the generator which would mount to the external case.

Mound/Signals Corps RTG Generator 3
Partially exploded view of Generator 3, the prototype for the rest of the generators in this program. Image credit AEC/DOE

The way I think of it is that the thermocouple uses electricity to measure temperature, while the thermopile uses heat to generate electricity.

With every (still small) thermopile, the overall power output increases, but the question becomes how do you string them together? If you go straight from one to the next from one side of the generator-a simple series circuit-just one thermocouple in one thermopile could cause the entire electrical system to fail. At the same time, the other extreme is just as distasteful; a purely parallel circuit, which would ensure that any individual component failure wouldn’t have any impact on any other component, is incredibly messy to build, and losses due to electrical resistance become very significant at these low conversion efficiencies. Clearly, a compromise will be required.

In this case, most systems used 37-thermocouple thermopiles – the design for the third generator was replicated for the rest of the program. A dozen of these thermopiles surrounded the fueled area. The feet were made of Chromel (a nickel-chromium alloy) and Constantin (a copper-nickel alloy). These materials are still used in oxidizing environments for thermocouple components.

This gave an incredibly flexible system to account for both changes in fuel diameter as well as thermocouple external diameter for any reason, with a low thermal conductivity material that’s both low enough cost and effective enough in this particular application to be a phenomenal example of practical experimental engineering.

In the case of our RTG, a whole host of mathematical, theoretical, and experimental headaches awaited – after all, this was hopefully only the beginning. While the Seebeck effect was discovered by Volta (and rediscovered by Seebeck in 1827), and the revisions by Kelvin were also far from new, the subject saw new scrutiny in the early 20th century. This led to the team laying out basic models for a wide range of temperature, power output, and operating regime variability considerations. More immediately applicable were calculations of the effect of each design component in the generator. Thermal flow circuit routing, optimized load resistance (in general and at a specified power level), relative wire gauge. and a host of other considerations were addressed for the first time. We’ll look into these tradeoffs later when we focus specifically on thermoelectric system design as a whole in a future post.

Thermal Management

The thermal management system was simple: conductive heat transfer was used to remove heat from the cold side of the thermopile, and then the exterior case were used to remove the heat from the system.

A primary concern with this system was balancing the heat production and rejection capacities of the system. Like in all thermal-to-electric conversion systems, the most important thing for the system is the maximum temperature difference in the core of the generator. In this case, the key spots are the hot and cold legs of each thermocouple. This means that everything that ISN’T a thermocouple is insulated, and the cold junction is the only exit point for waste heat. In order to assist with this, the gaps between the mica cards were often insulated with a special foam.

As is common with RTGs today, demonstration versions of the device used metal fin-type radiators. These were sized to allow the rejection of at least 10 W (I wasn’t able to find the exact number), the energy generated by the fuel.

Interestingly, there is a note that the fins would help reduce the effective dose not only due to additional shielding, but because the fins were in effect a standoff device.

In addition to the demonstration assembly, a lead shipping and storage cask was used for the fueled RTGs to provide a biological shield against the ~1000 curies of 210Po (and its resulting brehmsstrahlung radiation).

Testing History

A total of 17 RTGs were built during the course of the program, with the third being the prototype for the others.

While the second generator had a cruder thermocouple design, it was used extensively in the test program along with Generator 3 to analyze the operational behaviors and optimization of a number of other key characteristics. After these tests were complete, one generator went its own way: the 3rd generator went to the customer. The Signal Corps kept it for demonstration and testing purposes at Ft. Monmouth in New Jersey. The rest were tested at Mound, although they had their own hiccups. The fifth generator was rebuilt, and subsequently renamed the Generator 6.

One significant challenge was the newness of the fuel manufacturing method. A “completely engineered” radioisotope power source was not achieved immediately (it’s unclear what that actually means), finally occurring on the sixth example. Even then, progress was far from smooth,

The 8th generator had a less powerful 880 curie power source, rather than the design load of 1100 curies. This led to the removal of two mica card assemblies from the high voltage section to better match the lower powered RTG. The fifth generator was rebuilt completely during construction, partly due to poorly machined parts. Impurities like metal oxides contributed to the neutron flux. However, it seems like the challenges were minimal for such a groundbreaking program.

Program History

By a stroke of luck, I came across six of the nine quarterly progress reports at the very end of writing, with many interesting details. Unless you’re interested in the ins and outs, small successes and failures, I would recommend skipping this section. 

In the beginning of 1957, much of the mid-to long term viability of various isotopes was studied, as well as work on the design of the first prototype. We’re going to skip this one for now, but if you’re interested it’s linked below.

Our story picks back up in June (I wasn’t able to find the 2nd QPR). In the interval the ACTUAL first RTG was built, but isn’t used for any reported data collection.

The next big focus was on the design and fabrication of the second generator – which would be the forebearer of all subsequent generators in the program (in a slightly modified form).

This was the first time a wheel type thermopile was used, with soldered or spot welded wire connecting the heated central capsule. Each ring made up a single thermopile. In order to ease construction (since like with a bike wheel, the wires of the thermocouple form the entire structure from the center to the outside edge, styrene blocks had the wires pushed through them as a sort of welding jig, and once all the junctions were in place benzine was used to dissolve the foam.

For the first twelve tests, three thermocouples were used, with 122 junctions. Various tests were conducted using an electric heater while encased in a brass housing which was subsequently filled with foam insulation. Finally, the entire assembly was submerged in a bath that also served as a calorimeter (with temperature regulation capability of 540-s C) to both control ambient temperature and to measure thermal output of the device. Tests with and without electrical loading (and at a variety of loads) were conducted to evaluate maximum theoretical efficiency.

The ninth test was the first that the brass enclosure was evacuated to form a vacuum in the generator. The efficiency-already at 0. 457%- dropped even further, to 0. 392%.

The difference was accounted for due to higher operational temperature effects, or increased thermal resistance from the cold junction to the outer housing. Three more (non- submerged) tests at various temperatures were conducted, before the first refit of the generator occurred.

The thirteenth test was the first one to use four thermopiles, this one with 39 junctions (yes, if you’re following at home, two have 41, one has 40, and now one has 39…), additionally thermocouples to directly measure hot and cold junction temperatures were mounted. Tests much like the three thermopile devices were conducted.

As expected, efficiency of the generator dropped, although not by as much as expected. This is due to a number of factors, including a lower core temperature while also having greater heat losses through the more numerous junctions.

The detailed results of the tests on Generator 2 are below (click to expand).

While these designs were being built and tested, discussions with Site K-10 (a reprocessing program at Oak Ridge) on the availability of fuel isotopes. It appears this was a “we’d love to help you, but we’re tapped out and at capacity” situation. Additional discussion with outside experts convinced the team to stick to commercially available thermocouples, especially considering the lack of metallurgical capabilities. It’s likely this is when the Chromel-Constantan was finalized.

With another gap in reporting, we pick the project up again in January 1958. By the first of the year, the new generator (#3) was either nearly or fully complete.

This was the first generator to use the mica card type thermocouples which were so distinctive of the program. Each card was its own thermopile, with the wire junctions radiating out over the card. Each initial thermopile had 40 junctions (nominally, many had one or two fails. The hot shoes were pressure fit to the container in specially machined grooves, while the cold junction was held in place by two thin Teflon sheets. The same insulation filling in atmosphere remained.

Initial testing with four thermopiles showed much higher thermal resistance in this new generator at the hot junction, likely the result of not cementing the cards in place. Subsequent testing, this time in a vacuum, showed that the cement would be necessary at the hot junction.

This lesson learned, the generator continued to evolve. The next tests, using 11 cards (with a total of 437 junctions), did not go as hoped. Thermal isolation issues, especially at the cold junction/case interface, had to be addressed with a special grease. A significant drop in fuel casing temperature was recorded (from 206 to 181C), despite power increasing by more than 1kW (1428 W to 2567W). This is due, of course, to the almost tripling of the number of thermopiles. During testing, two Failed, always the outermost on the card on the hot side. Not only were these junctions most stressed during installation, but the spot welds used to make the wire-to-case connections were another weak point that a solution (flame welding) was found for.

Results for this round of testing can be found below. Additional tests to study Thompson and Joule effects were conducted, but they fall outside the scope of this blog. [Figs 3, 4,8,Tables D

During the spring and summer of 1959, discussions surrounding acquiring more radioisotopes continued, this time with Brookhaven National Lab. Their Liquid Metal Fuel Reactor Experiment (LMFRE) used bismuth coolant, which would produce 210 Po-not much, but about 25 grams per day (although more was possible through design tweaks on the unfinished reactors in about three years. More pressing fuel concerns, such as for the 50mW generator, were addressed with other organizations. On the construction and experimentation side of things, development of the heat source for the next generator was proceeding apace, with a 440 Ci (1.640 GBq) 210 Po heat source being prepared, in order to simulate the activity of a 6 month old unit. The generator this heat source was destined for was also under construction.

Pretty much identical to previous generators, this was designed to provide  about 18 mVe after 6 months. The number of thermopiles increased from 11 to 20, with each having 37 thermocouples in series. The groove to mount the hot junction to the fuel case was kept, as was the cement to ensure good thermal connectivity. On the cold junction, Teflon-insulated bars made of anodized aluminum were screwed into two mounting rings of the same metal, and were then turned on a lathe to a uniform size (in what sounds like a nerve-wracking procedure) before being placed in a thin aluminum can. On the can, a ceramic rod was held in place by spring steel supports to align the fuel element. No mention was made of the insulation.

Testing was largely the same as with the previous generator, with the notable exception of vacuum testing. Reporting on the performance of this generator was minimal:

"At a temperature differential of 2490C induced by a thermal input of 15. 612 watts, an efficiency of 0.32% was obtained."

AEC/DOE 1958

Sadly, this does not line up with data from Generator #3 – both the lack of insulation and evacuation means we’re left with no good points of reference for a good comparison especially with one sentence and a single data point.

At this point, the program switched focus for a time, to examination of radioisotopes selection. As I did in the introduction, I’m going to skip this subject for now.

On the programmatic side, the Signal Corps authorized another 50-mW generator. Alternatives to the Chromel-Constantan thermocouples were apparently still under consideration: senior program leadership visited a semiconductor manufacturer to discuss thermocouple options in the field. Notably, they discussed bismuth telluride, the choice for ESA’s 141 Am RTG (for more on the system, see my post here: ESA’s RTG Program: Breaking New Ground)

Another generator (#5) was also begun… which would become the problem child of the program. This unit had a different, more versatile purpose: as a demonstration unit. This seems to be the first time that transportation of a fueled RTG was addressed. a modified outer canister and new biological shield were required.This shield was 1″ of lead. From the initial dose of 2.26 R/hr at the case exterior, the shield would reduce the dose to 2.5 MR/hr. This generator also introduced another common characteristic for RTG cases-the finned case. Not only would this reduce the surface temp. from 89-92° C above room temp (by differing amounts depending on geometry and composition), but

 "Introduction of a finned case cover would decrease this temperature and also make it more difficult for a person handling the battery to receive an excessive radiation dose."

AEC/DOE 1958

This is an interesting, if relatively minor, side effect of the design, and one I had not considered before.

The assembly and testing of this generator were beset with difficulties. to begin with, the grooves for the hot junction were too narrow, and the sleeve they were to mount in twisted significantly. This degraded electrical insulation of the system, as well as I’m sure much cursing.

Sadly, the problems didn’t stop there. After the assembly of the generator itself was complete, it was placed within its outer container, pressure tested, fueled, filled with insulation, and sealed. Both submerged and air cooled tests were conducted. While conducting one test the generator was shorted out, leading to the need to repair the generator. The solution was simple: pull out the bad cards.

Unfortunately, at this point… it was dropped. Subsequently it was found to be noisy, so it was disassembled yet again.

In the process of fixing this short, another was discovered – but only after the insulation had been replaced and was most of the way outgassed. This means re-disassembly, then hunting for the short. It was then re-reassembled, worked for a few days… then died. Leads and thermo-couples were oxidized.

At this point, a decision was reached: scrap and remake the thermopiles, and rebuild the generator. Because of this, they decided to rename the device from Generator 5 to Generator 6. To minimize future oxidation issues, the cement was reduced – meaning greater care had to be taken to ensure good thermal continuity. The value was shown in subsequent tests, though; the reagent acted as flux on the wires in the presence of air at operating temperatures.

While this experience was I’m sure frustrating (based on the increased labor hours reported), the lessons were taken to heart.

After this, the matter of transportation and demonstration moved back to the fore. A finned case was made, as well as a transport case- however, the Bureau of Explosives (in charge of the transport of radioactive material)had yet to approve it. (Sadly, I was unable to find a picture of the case; see above for images of the finned case and storage cask.

Sadly, this ends our progress report availability, but one more generator was built.


While never even considered for field applications, this RTG had a profound impact on the space program. As we shall see later in this series, the potential applications reach deep into the sea and to all corners of the globe, but sadly that future never came to be.

While NASA and the U.S. Air Force would use RTGs (and their little cousins, radioisotope heating units), the devices never became widely applied, as was the case in the Soviet Union. In fact, it appears that the Signal Corps would not pursue their initial interest; on the other hand, several other organizations would continue to support Mound Laboratory as they continued exploring the design envelope of this fascinating technology


Blanke, Birden, Jordan, and Murphy, “Nuclear Battery Thermocouple Type Summary Report,” MLM-1127, AEC 1960

Blanke, B. C. “Nuclear Battery-Thermocouple Type, First Quarterly Report, Jan, 1 1957-Mar 31, 1957 MLM-CF-57-4-34, AEC 1957

Blanke, B. C. “Nuclear Battery-Thermocouple Type, Second Quarterly Report, Apr 1, 1957-June 30, 1957,” MLM-CF-57-8-27, AEC 1957

Blanke, B. C. “Nuclear Battery-Thermocouple Type, Third Quarterly Report, Jul 1, 1957-Sept 30, 1957,” MLM-CF-57-10-31, AEC 1957

Blanke, B. C. “Nuclear Battery-Thermocouple Type, Fourth Quarterly Report, Oct 1-Dec 31 1957,” MLM-CF-58-1-40 AEC 1958

Blanke, B. C. “Nuclear Battery-Thermocouple Type, Fifth Quarterly Report, Jan 1-Mar 31 1958” AEC 1958

Blanke, B. C. “Nuclear Battery-Thermocouple Type, Sixth Quarterly Report, Apr 1-Jun30, 1958” AEC 1958

Blanke, B. C. “Nuclear Battery-Thermocouple Type, Seventh Quarterly Report, Jul 1-Sept30 1958.” AEC 1958

What's Next At Beyond NERVA?

As I’ve hinted at, we will be continuing our look into the history of RPS as our focus for the blog for a while more. What’s more, we will be looking at some fascinating forgotten generator technology, as well as a look into the missions that this technology has-and continues to-enable. We will also look into the (far larger in scope) Soviet history with RPS, both good and bad. Finally, we’ll talk about what lies ahead for this fascinating but oft-overlooked technology.

Table of Contents


Share on facebook
Share on twitter
Share on pinterest
Share on linkedin

Leave a Reply

Your email address will not be published.

On Key

Related Posts

Fusion Ignition

To build a fusion rocket we first need to “ignite” the reaction, which is what this article will cover. We will dive into what fusion

Fusion Fuels

Nuclear fusion is often seen as the “Holy Grail” of power generation and rocketry. It powers stars and our most powerful weapons, and is ~4X