tenna, the open-circuit voltage cannot exceed
the peak ac voltage amplitude. Thus, these bi-
polar devices far exceed the limit of direct
rectification, suggesting that PAT and charge
separation can be further improved through
device and process optimization.Outlook
Efficient conversion of moderate-temperature
radiative thermal sources represents a large-
ly untapped resource for energy harvesting.
Microsystem-based radiative thermal-to-
electrical energy converters built on our bipolar
grating-coupled tunneling device represent a
scalable compact energy-harvesting technol-
ogy. These devices can be used as stand-alone
energy converters or in conjunction with ther-
moelectric power generators, where they need
only view the thermal source.
We have presented an alternative mecha-
nism for thermal photovoltaic conversion that
does not rely on square-law absorption but
uses PAT to pump charge into n- and p-type
wells in a bipolar grating-coupled device. A
spatially and temporally varying enhanced
transverse infrared electric field is confined in
the tunnel barrier that results from resonant
broadband coupling into the oxide LO phonon
resonance. A simple device model for the gate-
shunted photocurrent produced by PAT is de-
veloped, and the open-circuit voltage measured
across the pn junction is seen to exceed the
simulated peak ac voltage across the device.
The charge separation by PAT in the bipolar
device is akin to the Seebeck effect in a ther-
moelectric couple. The confined transverse
infrared field drives PAT in a similar fashion
as the temperature gradient in each leg of a
thermoelectric couple creates the charge sepa-
ration due to the internal electron and hole
currents. In general, the ideal diode picture for
the pn junction does not represent the ob-
served IV characteristics across the pn junc-
tion owing to the nature of the diffused pn
junction under the metal electrode and the
abrupt junction in the field. Electrical power
density improvements of several orders of
magnitude (× 10^4 to 10^5 ) are seen experimen-
tally in the three-terminal bipolar device rela-
tive to recent direct unipolar rectenna devices
( 20 ). The best-performing device (device 2)
had a measured electrical power density of
61 mW/cm^2 from a 350°C thermal source. This
results in an estimated conversion efficiency
for a single conversion bandwidth of 1 THz of
0.4%, which is approaching TPV conversion
efficiencies but at substantially cooler source
temperatures ( 1 ) (see supplementary mate-
rials). Further improvements in power gen-
eration can be achieved using structured
emitters on the thermal source with matched
polarization-insensitive antenna designs and
alternative gate dielectric materials in the
antenna-coupled tunnel diode for ENZ reso-nance matching to the thermal source. The
full power of the modified CMOS device de-
sign and process optimization requires fur-
ther exploration.REFERENCES AND NOTES- W. R. Chanet al.,Proc. Natl. Acad. Sci. U.S.A. 110 , 5309– 5314
(2013). - E. L. Dereniak, G. D. Boreman,Infrared Detectors and Systems
(Wiley Series in Pure and Applied Optics, John Wiley & Sons, 1996). - A. Fiorinoet al.,Nat. Nanotechnol. 13 , 806–811 (2018).
- M. Limet al.,J. Quant. Spectrosc. Radiat. Transf. 210 , 35–43 (2018).
- B. Zhaoet al.,Nano Energy 41 (suppl. C), 344–350 (2017).
- S. P. Kwok, G. I. Haddad, G. Lobov,J. Appl. Phys. 42 , 554– 563
(1971). - A. Sanchez, C. F. Davis Jr., K. C. Liu, A. Javan,J. Appl. Phys. 49 ,
5270 – 5277 (1978). - A. Sanchez, S. K. Singh, A. Javan,Appl. Phys. Lett. 21 ,
240 – 243 (1972). - J. O. McSpadden, L. Fan, K. Chang,IEEE Trans. Microw. Theory
Tech. 46 , 2053–2060 (1998). - J. O. McSpadden, T. Yoo, K. Chang,IEEE Trans. Microw. Theory
Tech. 40 , 2359–2366 (1992). - Y.-H. Suh, K. Chang,IEEE Trans. Microw. Theory Tech. 50 ,
1784 – 1789 (2002). - M. Dagenais, K. Choi, F. Yesilkoy, A. N. Chryssis, M. C. Peckerar,
Proc. SPIE Int. Soc. Opt. Eng. 7605 , 76050E (2010). - P. S. Davidset al.,Nat. Nanotechnol. 10 , 1033–1038 (2015).
- A. Sharma, V. Singh, T. L. Bougher, B. A. Cola,Nat. Nanotechnol.
10 , 1027–1032 (2015). - D. R. Ward, F. Hüser, F. Pauly, J. C. Cuevas, D. Natelson,
Nat. Nanotechnol. 5 , 732–736 (2010). - S. Grover, O. Dmitriyeva, M. J. Estes, G. Moddel,IEEE Trans.
NanoTechnol. 9 , 716–722 (2010). - P. Maraghechi, A. Foroughi-Abari, K. Cadien, A. Y. Elezzabi,
Appl. Phys. Lett. 99 , 253503 (2011). - P. Maraghechi, A. Foroughi-Abari, K. Cadien, A. Y. Elezzabi,
Appl. Phys. Lett. 100 , 113503 (2012). - E. A. Kadlec, R. L. Jarecki, A. Starbuck, D. W. Peters,
P. S. Davids,Phys. Rev. Appl. 6 , 064019 (2016). - J. Shanket al.,Phys. Rev. Appl. 9 , 054040 (2018).
- M. Z. Alam, I. De Leon, R. W. Boyd,Science 352 , 795–797 (2016).
- A. Alù, M. G. Silveirinha, A. Salandrino, N. Engheta,Phys. Rev. B
75 , 155410 (2007). - D. W. Berreman,Phys. Rev. 130 , 2193–2198 (1963).
- S. Vassant, J.-P. Hugonin, F. Marquier, J.-J. Greffet,Opt.
Express 20 , 23971–23977 (2012). - P. S. Davids, J. Shank,Phys. Rev. B 97 , 075411 (2018).
- P. K. Tien, J. P. Gordon,Phys. Rev. 129 , 647–651 (1963).
ACKNOWLEDGMENTS
P.S.D. thanks D. B. Burckel and R. Sanchez of Sandia for many useful
discussions of this work.Funding:This work was funded by Sandia’s
Laboratory Directed Research and Development (LDRD) program.
Sandia National Laboratories is a multimission laboratory managed
and operated by National Technology and Engineering Solutions
of Sandia, a wholly owned subsidiary of Honeywell International Inc.,
for the United States Department of Energy’s National Nuclear
Security Administration under contract DE-NA0003525. This paper
describes objective technical results and analysis. Any subjective
views or opinions that might be expressed in the paper do not
necessarily represent the views of the U.S. Department of Energy
or the United States government.Author contributions:P.S.D.
developed the device concept and design. A.S. and R.J. developed
the process flow and fabricated the devices. J.K. and J.S.
performed all experiments. D.P., J.S., and P.S.D. developed
theory and simulated device performance. All authors
contributed to the preparation of the manuscript.Competing
interests:None declared.Data and materials availability:
All data needed to evaluate the conclusions in the paper
are present in the paper or the supplementary materials.
SUPPLEMENTARY MATERIALS
science.sciencemag.org/content/367/6484/1341/suppl/DC1
Materials and Methods
Supplementary Text
Figs. S1 to S11
References ( 27 – 30 )
13 November 2019; accepted 6 February 2020
Published online 20 February 2020
10.1126/science.aba2089gate. Figure 3B shows two transmission elec-
tron microscope (TEM) cross-sectional images
of two different device fabrication runs taken
under the gate metal. Device 1 has a nominal
targeted gate oxide of 4 nm, and device 2 has
a nominal gate oxide thickness of 3.5 nm. For
device 1, a nominal oxide thickness was tar-
geted at 4 nm (4.2 nm measured), and we see
a thin alumina layer formed at the SiO 2 /Al
interface. This alumina layer arises from
contact silicide formation during thermal
processing. Its makeup as alumina has been
confirmed by energy-dispersive x-ray spec-
troscopy (EDS). Figure 3C shows the mea-
sured power density and the measured voltage
as a function of load resistance at various
source temperatures. The peak power den-
sity of ~27 mW/cm^2 occurs at source tempera-
tures of 250° and 400°C, with an open-circuit
voltage in excess of 2 mV for the two high-
power cases. The peak power generation trend
is not monotonic with increasing source tem-
perature, and this can be attributed to the
formation of an additional alumina layer.
The impact of the thick alumina layer shifts
the peak power density to lower source tem-
peratures because the alumina LO phonon
resonance occurs at roughly 200°C or lLO =
10.1 mm—compare this with SiO 2 LO pho-
non mode at 400°C or lLO = 8.1 mm. There is
therefore a complex interaction of the gate
oxide LO phonon resonance and the device
design parameters that determines the output
power for this newly realized form of TPV
conversion. This presents an opportunity to
adjust the operational temperature of the de-
vice by tuning the LO phonon resonance.
Device 2 shown in Fig. 3B, from a different
process run, targets an oxide thickness of
3.5 nm nominal (3.4 nm pictured). Figure 3D
shows the measured power density generated
and the measured voltage as a function of
the load resistance measured across the pn
junction for the nominal 3.5-nm device. The
three-terminal nature of the device allows for
measurement in two different wiring config-
urations. The default configuration has been to
leave the metal contact floating. This was the
condition in the previous measured device and
in the current 3.5-nm device shown in orange.
Alternatively, we can ground the metal contact
by removing any buildup of charge on the gate,
and the measured power density increases by
a factor of 10×, shown in black (Fig. 3D). The
peak power density for the 350°C source is
61 mW/cm^2 and occurs at a load resistance of
~250 ohms. The peak measured open-circuit
voltage for the grounded case is 1.6 mV com-
pared with 0.5 mV for the ungrounded case.
In both the 3.5- and 4-nm device cases, the
open-circuit voltage is obtained at large load
resistance and is shown to greatly exceed the
predicted ac voltage amplitude shown in Fig.
2C. By comparison to a direct unipolar
rec-
SCIENCE
20 MARCH 2020•VOL 367 ISSUE 6484 1345RESEARCH | RESEARCH ARTICLES