RESEARCH ARTICLES
◥APPLIED PHYSICS
Electrical power generation from
moderate-temperatureradiative thermal sources
Paul S. Davids*, Jared Kirsch, Andrew Starbuck†, Robert Jarecki†, Joshua Shank, David Peters
Moderate-temperature thermal sources (100° to 400°C) that radiate waste heat are often the by-product
of mechanical work, chemical or nuclear reactions, or information processing. We demonstrate
conversion of thermal radiation into electrical power using a bipolar grating-coupled complementary
metal-oxide-silicon (CMOS) tunnel diode. A two-step photon-assisted tunneling charge pumping
mechanism results in separation of charge carriers in pn-junction wells leading to a large open-circuit
voltage developed across a load. Electrical power generation from a broadband blackbody thermal
source has been experimentally demonstrated with converted power densities of 27 to 61 microwatts per
square centimeter for thermal sources between 250° and 400°C. Scalable, efficient conversion of
radiated waste heat into electrical power can be used to reduce energy consumption or to power
electronics and sensors.
A
ll objects at finite temperature radiate
owing to thermal fluctuations of their
atomic constituents in a characteristic
spectrum that depends on the object’s
surface temperature and spectral emis-
sivity. Radiative heat transfer from the Sun
(6050°C effective solar blackbody spectrum;
air mass 1.5) is the dominant radiative energy
resource available on Earth. Photovoltaic power
generation is an effective and rapidly growing
technology for converting this incident radia-
tion into electrical power. However, radiative
processes from cooler terrestrial sources or
man-made waste heat can give rise to consid-
erable net energy exchange that is also readily
available as an electrical power source, pro-
vided that it can be efficiently converted.
Thermophotovoltaic (TPV) devices that con-
vert radiation from broadband thermal sources
into electrical power are promising technolo-
gies for solar energy conversion and waste
heat recovery. These devices work by heat-
ing a secondary thermal source that acts as
a selective emitter. The emission spectrum
is filtered and matched to a small bandgap
semiconductor device ( 1 ). The semiconductor
device is typically a p- and n-type (pn) junction
designed such that the absorption takes place
in the depletion width of the device. The ab-
sorption of a photon in the depletion region
of the semiconductor creates an electron hole
pair that is separated by the internal field,
resulting in the separation of charge and an
open-circuit voltage induced across the device.
These devices typically work in the range of
temperatures from 1000 to 2000 K or wave-
lengthslof 1.4 to 3.0mm, which corresponds
to bandgap energiesEgof 0.43 to 0.86 eV.
Wien’s law requires that, as the temperature
of the blackbody source is decreased, the wave-
length at peak power increases such that, for
source temperatures between 100° and 400°C,
the spectral range is in the thermal infrared
(7 to 12mm). Narrow bandgap semiconductors
at room temperature matched to this infrared
band have considerable photon generation
from thermal fluctuations and thermal gener-
ation across the bandgap resulting in appre-
ciable noise. For this reason, infrared detectors
are typically cooled below room temperature
to reduce background noise driven by these
thermal fluctuations in the narrow-gap semi-
conductor ( 2 ). Furthermore, the power den-
sity in the working bandwidth of the device
is decreased because of the exponential na-
ture of Planck’s radiation law, thus making
TPV conversion from a moderate-temperature
source very challenging. Improvements in TPV
conversion efficiency for moderate-temperature
sources using vacuum near-field enhancement
have been proposed ( 3 – 5 ). These near-field TPV
devices show great promise, but practical limits
of vacuum gaps and planarity from micrometer
to nanometer length scales remain a challenge
for larger-scale power generation.Approaches for energy conversion
Alternative approaches for thermal-to-electrical
conversion on the basis of direct rectification
of infrared radiation using ultrafast tunnel-
ing have been proposed ( 6 – 8 ). This approach
is not based on square-law absorption and
generation of carriers in a semiconductor, but
rather it relies on high-speed direct tunneling
within an asymmetric tunnel diode to separate
charge. In the microwave part of the spectrum,
antenna-coupled diode rectifiers are used as
efficient converters from high-speed gigahertzsignals to direct current (dc) ( 9 – 11 ). These de-
vices use standard antennas to channel micro-
wave radiation into a stripline guided mode,
which is loaded with a fast diode to rectify the
signal. Impedance-matching techniques are
used to minimize reflectance from the anten-
na and effectively maximize power transfer to
the load. This is inherently a narrow-band con-
version process, differing from that in infrared
and optical devices. These microwave rectify-
ing antenna (rectenna) devices can have con-
version efficiencies >85% ( 10 ), and attempts to
scale to infrared and optical frequencies (30 to
200 THz) have been explored ( 12 – 15 ). A key
component in these scaled devices is the ultra-
fast direct tunnel diode. Metal-insulator-metal
(MIM) or double insulator tunnel (MIIM)
diodes have been examined for this purpose
( 16 – 18 ). However, large diode asymmetry at
small voltages is required for rectification of
infrared radiation, and further progress is
needed in this area. A promising approach for
making highly asymmetric tunnel diodes on
the basis of advanced complementary metal-
oxide-silicon (CMOS) fabrication has recently
been developed, which brings new design tools
and scalable processing to large-area antenna-
coupled tunnel diode devices ( 13 , 19 ).
Electrical power generation from a direct
grating-coupled tunnel MOS diode rectifier
illuminated by a thermal source has recently
been experimentally observed ( 20 ). Peak power
densities of 1 to 8 nW/cm^2 have been measured
for thermal sources between 400° and 450°C
in a large area n+ MOS grating-coupled tunnel
diode. The peak power is seen to increase as
the source temperature is increased, and im-
pedance matching into the load resistance is
as predicted by the simple rectifier model.
These antenna-coupled tunnel diodes strongly
interact photonic resonances with longitudinal
optical (LO) phonon resonances in the polar
oxide tunnel barrier ( 13 , 19 ). The confined and
enhanced transverse optical field in the tunnel
barrier occurs because of the epsilon-near-zero
(ENZ) dispersion in the polar gate oxide, and
the resonant antenna or grating structure is
necessary to couple into the LO phonon mode
( 21 – 24 ). The transverse field in the tunnel bar-
rier is spatially varying and enhanced relative
to the incident field amplitude and can be
orders of magnitude larger ( 13 , 19 , 20 ). More-
over, the enhanced transverse field confined
in the tunnel barrier occurs over a finite band-
width near the ENZ material resonance and is
a result of the avoided crossing of the photonic
and LO phonon resonances. A mechanism for
the generation of a direct photocurrent from
the confined transverse field in the barrier is
based on a model of photon-assisted tunneling
(PAT) in a uniform barrier ( 25 ). PAT is obtained
as a perturbation expansion in the strength
of the electromagnetic interaction in the con-
fined tunnel barrier and leads to a multiphotonRESEARCH20 MARCH 2020•VOL 367 ISSUE 6484 1341Sandia National Laboratory, P.O. Box 5800, Albuquerque,
NM 87185-1082, USA.
*Corresponding author. Email: [email protected]
†These authors contributed equally to this work.
SCIENCE