1
A Comparison of Energy Conversion Technologies for Space
Nuclear Power Systems
Lee S. Mason
1
NASA Headquarters, Washington D.C., 20056, U.S.A
A key element of space nuclear power systems is the energy conversion subsystem that
converts the nuclear heat into electrical power. Nuclear systems provide a favorable option
for missions that require long-duration power in hostile space environments where sunlight
for solar power is absent or limited. There are two primary nuclear power technology options.
Radioisotope Power System (RPS) utilize the natural decay heat from Pu238 to generate
electric power levels up to about one kilowatt. Fission Power System (FPS) rely on a sustained
fission reaction of U235 and offer the potential to supply electric power from kilowatts to
megawatts. Example missions for nuclear power include Mars science rovers (e.g. Curiosity,
Mars 2020), lunar and Mars surface landers, crewed surface outposts, deep space planetary
orbiters, Ocean World science landers, and robotic space probes that utilize nuclear electric
propulsion (NEP). This paper examines the energy conversion technology options that can be
used with RPS and FPS, and provides an assessment of their relative performance.
I. Introduction
Nuclear systems provide a favorable option for missions that require long-duration power in hostile space
environments where sunlight for solar power is absent or limited. Example missions include Mars science rovers (e.g.
Curiosity, Mars 2020), lunar and Mars surface landers, crewed surface outposts, deep space planetary orbiters, Ocean
World science landers, and robotic space probes that utilize nuclear electric propulsion (NEP). There are two primary
nuclear power technology options: Radioisotope Power Systems (RPS) and Fission Power Systems (FPS).
RPS utilize the natural decay heat from Pu238 to generate electric power levels up to about one kilowatt. The
NASA Science Mission Directorate (SMD) RPS Program works in partnership with the Department of Energy (DOE)
to produce the Pu238 heat sources, supply RPS and related services to flight missions, and develop new power
conversion technologies. RPS have been a staple in NASA missions since the 1969 Nimbus III mission with a
portfolio that includes Apollo, Pioneer, Viking, Voyager, Cassini, Pluto New Horizons, and most recently, Mars
Curiosity. The current class of RPS utilize General Purpose Heat Source (GPHS) modules that supply approximately
250 Watts-thermal at Beginning-of-Life (BOL). Each GPHS module includes four fuel pellets that contain about 0.6
kg of plutonium-oxide. This fuel form has been in production since the late 1980s when the first GPHS Radioisotope
Thermoelectric Generator (RTG) was flown on the Galileo mission. The largest RPS mission ever flown was Cassini
(1997) which used three GPHS RTGs to supply nearly 900 Watts at launch using a total of 54 GPHS. The current
version is the Multi-Mission RTG (MMRTG) designed to produce about 110 Watts at launch using eight GPHS
modules. The MMRTG was first used on Mars Curiosity (Figure 1) and is slated for use on the upcoming Mars 2020
rover.
FPS rely on a sustained fission reaction of U235 and offer the potential to supply electric power from kilowatts to
megawatts. The U.S. has flown only one FPS, in 1965. The 500 Watt SNAP 10A (Figure 2) operated for 43 days
before a spacecraft malfunction (unrelated to the FPS) caused a premature ending to the mission. NASA and DOE
have attempted to develop FPS multiple times since the SNAP 10A, including SP-100 in the late 1980s and
Prometheus in the early 2000s. In general, NASA’s efforts to develop a space-qualified FPS fell short due to technical
complexity, high development costs, and aggressive performance claims. These past attempts were typically
accompanied by the need to develop new reactor fuel, structural materials and balance-of-plant components for a
system that was bound by mission needs to produce high power at low mass with long operational life. This is a risky
combination that undoubtedly contributed to the poor record of success in past programs.
1
Principal Technologist, NASA Space Technology Mission Directorate.
2
Fig. 1 MMRTG Concept and Installation on Mars Curiosity Rover
Fig. 2 SNAP-10 Reactor and Flight System Concept
Fig. 3 Kilopower Reactor Concept and Test Unit Assembly
The current space FPS development effort under the NASA Space Technology Mission Directorate (STMD) is
Kilopower, which is aimed at FPS that could produce a modest power output between 1 and 10 kW-electric (kWe).
The Kilopower Project recently completed a successful nuclear-heated reactor prototype ground test (Figure 3) and is
being considered for a possible flight Technology Demonstration Mission in the mid-2020s. The primary mission
applications under consideration include lunar and Mars surface power systems. Future versions could be adapted to
outer planet science missions including those that use NEP to bring orbiters and landers to the far reaches of the solar
system.
While both RPS and FPS are nuclear power systems, they are distinctly different in terms of processing and
operations. The U.S. has a very limited supply of the Pu238 used in RPS and the DOE has only recently begun to
produce new material after an extended hiatus. The cost and complexity of making Pu238 is significant, and currently
depends on NASA funding as the only recognized user. Once the Pu238 fuel is loaded into an RPS at DOE, that
system is operational and must be handled carefully through launch due to its elevated temperature and need for
thermal control. At launch, the Pu238 has a significant radiological inventory requiring specific safety measures to
assure containment should there be a launch accident. After the launch, the RPS has no means of heat source
shutdown, so missions must accommodate the constant thermal load. As the Pu238 fuel decays, the thermal output
and corresponding RPS electric output power gradually decreases. A distinct advantage of RPS is the benign radiation
environment produced by the alpha-emitting Pu238 resulting in minimal radiation effects on equipment or personnel.
3
The enriched U235 used in FPS is available from dismantled nuclear weapons and maintained in large quantities
by the DOE for purposes that include space reactors. Since NASA is a minor user, there is no funding commitment
and no cost for the allocated raw material. However, because the enriched U235 can be used to make weapons, it
requires special security measures to safeguard it from proliferation threats. During fabrication and launch processing,
the U235 reactor core is not radioactive nor does it produce heat until the reactor is turned on. To start the reactor, a
neutron-absorbing control rod (or rods) is removed from the core to allow the fission reaction to occur. Should there
be a launch accident, the U235 core is not a radiation hazard unless the system experiences an inadvertent criticality,
which can be avoided by careful design and simple safety systems. After the reactor is started in space, the nuclear
reaction and gradual buildup of fission products produce gamma and neutron radiation that requires shielding to
protect sensitive equipment and humans. A key discriminator relative to RPS is that the FPS can be stopped and
restarted using the control rod, as needed during the mission. The reactor thermal output can also be maintained at a
fixed level through occasional (perhaps yearly) control rod adjustments.
II. Study Methodology
The premise of this study is to compare energy conversion options based on their power output and specific power
(W/kg) assuming three different fixed nuclear heat sources. The heat sources considered are as follows: 1) an array
of eight GPHS modules, supplying approximately 2 kW-thermal (kWt), 2) the smallest Kilopower reactor, supplying
4.3 kWt, and 3) the largest Kilopower reactor, supplying 43 kWt. The heat sources will be treated as fixed thermal
supplies while accounting for thermal insulation losses and End-of-Mission (EOM) power degradation. System
performance generally improves with increasing heat source temperature, so the study will evaluate the benefits of
several different hot-end temperatures for each heat source. While GPHS-based heat sources have been shown to
operate as high as 1275 K, the Kilopower reactor heat sources are limited to about 1075 K based on the current cast
uranium-molybdenum (UMo) fuel form.
The key variable in determining system power output for a fixed heat source is the conversion cold-end
temperature, or more specifically, the converter temperature ratio (Thot/Tcold). The conversion technologies in this
study generally behave as thermodynamic heat engines with energy conversion efficiencies that are proportional to
their fraction-of-Carnot efficiency. For each heat source and hot-end temperature, an analysis will be performed by
varying cold-end temperature to examine the effect on power output, radiator area, and system specific power. In
sweeping through the cold-end temperatures for each option, an optimum Tcold occurs as the result of balancing the
power produced versus the size of the radiator. The system will have the highest efficiency and produce the maximum
power output at the lowest Tcold. However, the low cold-end temperatures result in larger radiators whose area is
inversely proportional to the cold-end temperature to the fourth power (Tcold^4).
Determining system mass is a complicated process, which will be greatly simplified in this analysis. The mass of
the three heat sources are fixed and taken from published values. A heat source assembly containing 8-GPHS modules
weighs about 13 kg, not including structural support or insulation. The smaller 4.3 kWt Kilopower reactor including
core, reflector, control rod, and heat pipes weighs about 136 kg and requires a 148 kg radiation shield (284 kg total
mass). The corresponding, larger 43 kWt Kilopower reactor weighs about 235 kg with a 547 kg radiation shield (782
kg total mass). The converter, controller (if needed), housing/heat rejection, thermal insulation and integration masses
are derived from historical systems and concepts based on the author’s judgement and calculated using appropriate
scaling methods. All EOM power values are based on a hypothetical 10-year mission (with 3-year storage for the
RPS). The radiator mass is based on Stefan-Boltzmann area calculations with reasonable assumptions on radiator
temperature drop, fin effectiveness, thermal emissivity, sink temperature, and aerial density (kg/m
2
).
III. Power Conversion Options
The currently available RPS is the MMRTG containing eight GPHS modules. It uses state-of-the-art PbTe-based
2
thermoelectric (TE) couples that operate at a hot-end temperature (Thot) of approximately 810 K and a cold-end
(Tcold) of 485 K. The MMRTG weighs approximately 45 kg and produces 110 Watts at launch (2.5 W/kg) in a design
that was intended for use in either planetary atmospheres (like Mars) or the vacuum of space. The total generator
efficiency is 110/(8*250) or 5.5%, but the TE conversion efficiency is more like 6.5% after accounting for thermal
and electrical losses. Given the operating temperatures, the equivalent TE fraction-of-Carnot efficiency is 6.5%/(1-
485/810) or about 16%. Using Curiosity performance data, the output power decreases at a rate of about 4.8% per
year from a combination of fuel decay (~0.8%/year) and thermoelectric degradation (~4%/year). The source of the
2
The MMRTG uses PbTe “N” Leg and TAGS (Tellurium-Antimony-Germanium-Silver)/PbSnTe “P” Leg
thermoelectric couples. For simplicity, this paper uses “PbTe” to designate this combination.
4
high thermoelectric degradation is related to material thermal stability and sublimation. For reference, the former
vacuum-only GPHS RTG produced about 285 Watts at launch and weighed approximately 56 kg (5 W/kg) while
operating at Thot of 1273 K and Tcold of 573 K. That system used SiGe couples with converter efficiency of about
7.5%, equivalent to 13.5% fraction-of-Carnot.
The RPS Program is working on advanced TE converters that operate at higher hot-end temperatures and greater
efficiencies to increase power output using materials that promise lower degradation. As a first step, the enhanced
MMRTG (eMMRTG) uses Skutterudite (SKD) TE couples instead of PbTe. Efficiency and operating temperature
can be increased further by adding additional thermoelectric segments to the SKD couples, including Zintl and LaTe
compounds (Figure 4). A segmented TE conversion system using this combination could operate at a hot-end
temperature of 1075 K with a Carnot fraction comparable to the PbTe-based systems (16%). The new segmented
thermoelectric couples are also predicted to have lower degradation rate (2.5%/year) compared to MMRTG, resulting
in higher EOM power output. With additional segments and/or variants of the Zintl and LaTe materials, the TE
converters could a achieve hot-end temperature up to 1275 K, similar to the SiGe couples used in past GPHS RTGs,
with better Carnot fraction (16%) and low predicted degradation (1.9%/year).
Fig. 4 Representative Segmented Thermoelectric Couple
The application of fixed Carnot fractional efficiencies to represent thermoelectric converter performance is
somewhat unorthodox, and there are better methods to use. However, the author has compared results from more-
sophisticated TE analyses in the literature and found the Carnot method to be simple and accurate.
The RPS Program is also developing power conversion technologies for a dynamic RPS. A recent procurement
resulted in four contractor studies exploring free-piston Stirling with gas bearings, free-piston Stirling with flexure
bearings, thermoacoustic Stirling and closed Brayton cycle. Dynamic conversion technologies have been under
development by NASA for decades but no converter has ever flown in space. The most recent flight development
attempt was the Advanced Stirling Radioisotope Generator (ASRG) which utilized two GPHS modules and two free-
piston Stirling units (Figure 5) with gas bearings to produce 140 Watts at launch with a system mass of 31 kg (4.5
W/kg, 28% generator efficiency). The Stirling converters operated at a hot-end temperature of 1033 K and cold-end
of 313 K with a converter efficiency of about 38%, after accounting for thermal and electrical losses. The equivalent
fraction-of-Carnot for the ASRG converters was 38%/(1-313/1033) or 54%. ASRG’s ultra-high efficiency may have
contributed to its demise as it resulted in converter manufacturing difficulties and test unit reliability issues. Future
Stirling converter developments need not push efficiency so hard, but rather focus on simplicity and robustness. A
conservative Carnot fraction of 50% is assumed for the Stirling conversion options in this study. Two hot-end
temperatures are considered: 925 K and 1075 K, representing a low-risk and higher-risk implementation while staying
in the class of Ni-based superalloy materials commonly used in recent converter development efforts.
Fig. 5 ASRG Concept and Stirling Converter
5
Space Brayton technology has seen many incarnations since the late 1960s including the 1.3 kWe Brayton Isotope
Power System (BIPS) and the 2.5 kWe Dynamic Isotope Power System (DIPS). The current effort by the RPS
Program is exploring a 0.5 kWe-class system with two 100% redundant converters. In simple terms, the Brayton
cycle uses constant pressure heat addition and heat rejection, rather than the constant temperature processes in the
Stirling cycle (and the Rankine cycle). This results in a reduced fraction-of-Carnot for space Brayton systems. Stating
it differently, Brayton systems require a greater temperature ratio to achieve the same conversion efficiencies as
Stirling or Rankine systems. Brayton does offer an advantage in specific power at higher power levels compared to
Stirling, due to the high power density of turbomachinery and the use of distributed heat exchangers. In evaluating
the past concepts and considering the current one, a conservative fraction-of-Carnot for Brayton is 35%. Like the
Stirling options, the Brayton concepts studied here will consider two hot-end temperature values of 925 K and 1075
K, bounded by the family of possible Ni-based superalloy materials.
Finally, despite it not being part of the RPS dynamic conversion options, the organic Rankine cycle (ORC) is
considered here. Space-based ORC systems have received relatively little attention in recent years. In the 1970s, the
Kilowatt Isotope Power System (KIPS) utilized the ORC and DOE’s Multi-Hundred Watt (MHW) heat source to
produce about 1.3 kWe as a competitor to the Brayton-based BIPS (Figure 6). Interestingly, the two concepts were
projected to have about the same mass (just over 200 kg), but the KIPS version required three heat source assemblies,
instead of two for BIPS, and a larger radiator. These same two technologies were pitted against each other again in
the 1990s for a 25 kWe solar dynamic power module on Space Station Freedom (SSF). Both the KIPS and SSF ORC
systems used toluene working fluid. The primary technical hurdles inhibiting the use of ORCs in space are two-phase
fluid management in zero-g and limited hot-end temperature related to toluene decomposition. The maximum hot-
end temperature for a toluene-based ORC system is about 675 K, which is a blessing for structural materials but a
curse for efficiency and radiator area. The advantage of ORC is the high Carnot fraction which is nearly as good as
Stirling, and conservatively assumed to be 45% for this study.
As stated previously for the TE options, the use of fixed Carnot fraction to determine the performance of the
dynamic converters is a simplification compared to more sophisticated methods. However, this approach provides a
good first-order approximation to compare all the technologies on a relative basis. Regarding degradation, all of the
dynamic conversion technologies considered here have non-contacting moving parts (either pistons or rotors)
supported by gas and/or mechanical bearings. The resulting converter degradation is negligible, but assumed to be
0.5%/yr on top of any heat source degradation.
Fig. 6 BIPS (Left) and KIPS (Right) Flight Concepts
IV. RPS Performance Comparisons
An Excel model was developed to estimate system performance for each heat source and power conversion option
described above. The Excel model sweeps through the cold-end temperatures and uses Carnot fraction to estimate
power output and the Stefan-Boltzmann equation to estimate radiator area. A system mass is calculated for each
design using scaling methods derived from historical systems, and the cold-end temperature that maximizes system
specific power (W/kg) is determined. Separate worksheets were generated for each of the three fixed heat source
options considered: 1) 2 kWt, 8-GPHS array, 2) 4.3 kWt Kilopower reactor, and 3) 43 kWt Kilopower reactor.
The RPS power output and radiator area estimates are presented in Figure 7 assuming the 8-GPHS heat source.
The cold-end temperature range considered was 350 to 650 K. The data curves show the relative performance of the
TE, ORC, Brayton, and Stirling conversion options assuming a conservative, low hot-end temperature (solid curves)
representing the current state-of-the-art and the more advanced, higher hot-end temperature (dashed curves). For the