An Introduction to the DARPA Program in Advanced Thermoelectric Materials and Devices

Lawrence H. Dubois
Director, Defense Sciences Office
Defense Advanced Research Projects Agency
Arlington, VA 22203-1714 USA

Military and Civilian Interest in Thermoelectronics

Introduction
The Defense Advanced Research Projects Agency Future military applications for improved thermoelectric
(DARPA) initiated its program in advanced thermoelectric devices will depend on their ability to cool or power larger
materials and devices in 1995 in response to a belief that these systems with greater efficiency and to reach lower operating
materials could play a significant role in the development of temperatures with fewer stages. Expanded military
new and enhanced military capabilities. The current and applications include the cooling of both short and long
future use of these all solid-state devices for both cooling and wavelength infrared detectors, microprocessors, multi-chip
power generation is driven by their inherent advantages: high modules, and high temperature superconducting filters,
reliability with no maintenance, silent operation with no oscillators and low-noise amplifiers. Temperatures below 80
moving parts or vibrations, and environmental compatibility K will be required for these latter applications. Larger-scale
(no compressed gases or chemicals). Even with the limited cooling applications include individual, man-portable
performance of current materials, the military Services have microclimate systems and air conditioning and refrigeration
found a number of critical applications for thermoelectric for ships, aircraft and land vehicles. While the temperature
devices. These include the cooling and temperature constraints on these latter systems are eased somewhat
stabilization of short wavelength infrared detectors and optics (typical minimum temperatures of only 250 K are required),
(operating temperature from 165 - 320 K) (see Figure l), the size, weight, efficiency and cost remain key drivers. At the
other extreme, applications for thermoelectric devices in
cooling of charge coupled devices (220 - 250 K), and the
active cooling of electronic enclosures (typically less than 100 power generation from both low and high temperature waste
heat sources, altering heat signatures of hot objects, and the
W). Expansion of the use of thermoelectric devices into other
efficient actuation of shape memory alloy devices, will
military systems will only occur if the properties of the
require materials capable of operating at elevated
materials and the resulting components can be improved
substantially. temperatures (in some cases up to 1000 K).
While the military has substantial interest in the
improvement of thermoelectric materials and devices, clearly
the key economic driver will be the vastly larger civilian
market (see Figure 2). The current market for thermoelectric
cooling devices is estimated to be in the hundreds of millions
of dollars (primarily for small refrigerators/picnic coolers)
and estimates for the use of enhanced thermoelectrics for both
cooling and power generation reach into the billions of
dollars. Key civilian applications include the replacement of
large-scale conventional cooling and refrigeration systems
(both stationary and mobile) and the cooling of
microprocessors and multi-chip modules as well as high
temperature superconducting switching and filter systems.
The market for power generating systems may be even larger.
This includes power for hybrid electric vehicles, the
replacement of both truck and automobile alternators, self-
powered engine pre-heaters, blowers for gas furnaces and
heaters, and remote power sources. Commercial industry will
drive the need for low-cost materials, materials processing,
and device fabrication. The Department of Defense will
clearly gain from the expanded market due to its ability to
purchase high performance components at more reasonable
prices.
Figure 1. Representative thermoelectric cooling devices
for military applications: (a) a range of packaged coolers and
infrared detectors and (b) six-stage thermoelectric cooler
capable of reaching -150 K (Marlow Industries,
www.marlow.com).

0-7803-5451-6/00/$10.0002000 IEEE 1 18th International Conferenceon Thermoelectrics (1999)

 ~

new materials and structures and building them into prototype

thermoelectric devices. These devices are being tested for
both cooling and power generation applications. This portion
of the work entails the development of novel devices and
systems and the coupling of these devices with specific
applications (e.g., energy scavenging from small temperature
differences to power networks of distributed sensors).
Ultimately the program will obtain true performance metrics
under realistic operating conditions. While future
applications will require addressing issues such as materials
and manufacturing costs and environmental concerns, these
are not playing a major role in the current DARPA-funded
efforts. The ultimate goal of our program is to quadruple the
figure of merit of thermoelectric materials, thus making the
resulting devices competitive with conventional phase change
systems.
The dimensionless figure of merit of a thermoelectric
material, ZT, is given by the familiar equation Z T = S 2 T o / ~ ,
where S is the Seebeck coefficient, o is the electrical
conductivity, and K is the thermal conductivity. Balancing
these parameters typically leads one to choose a semi-metal
(such as Bi,Te,) in order to maximize ZT. The traditional
methods for increasing the figure of merit include increasing
the Seebeck coefficient through the use of materials with
large effective masses and unusual band structures and
increasing the carrier mobility by using covalently bonded
solids to minimize electron-phonon scattering and barrier-
doped quantum wells to separate carriers from scattering
centers. In addition, decreasing both the phonon ( K ~ ~and )
electronic (K,) contributions to the thermal conductivity can
occur through the decoupling of K~~ and K,, the use of
materials and/or structures with large unit cells and large
Figure 2. Representative thermoelectric devices available effective masses to scatter phonons, and increased disorder to
in the commercial market: (a) thermoelectric picnic cooler decrease the phonon mean free path. The problem, of course,
(Igloo Products Corp.); (b) coolers for diode lasers (Marlow is that for simple materials an increase in the Seebeck
Industries, www.marlow.com), and (c) 15 W thermoelectric coefficient typically leads to a decrease in the electrical
power generator (Hi-Z Industries, www.hi-z.com). conductivity (and visa versa). And, in addition, an increase in
o almost always results in an increase in K, (the well-known
New Materials a n d Devices Wiedemann-Franz law).
With both the military needs, the huge potential civilian Despite these difficulties, there are a number of exciting
market, and DARPA’s willingness to take a significant risk if new materials and structures that may lead to substantial
the pay-off is large enough, four years ago we embarked on a increases in Z T (see, for example, Figure 3 ) . These include
program to create whole new classes of thermoelectric (but are not limited to) filled skutterudites and new
devices. The first step in this program was devoted to the skutterudites (e.g., XeIr,Sb,,), mesoporous materials (e.g.,
development of new materials through the synthesis and Ru/TiO,), thin film/quantum well/quantum wire/quantum dot
assenibly of novel thermoelectric compounds (including bulk structures (e.g., Bi,Te,/Sb,Te,, and “opals and inverse opals”),
materials, thin films, mesoporous materials, etc.) and the intercalation compounds (e.g., BE), heavy
construction of novel structures (e.g., superlattices, quantum fermion/hybridization gap systems (e.g., CeFe,Sb,,),
wells, quantum wires, quantum dots, graded layers, opals and intermetallic semiconductors (e.g., TiNiSn), doped polymeric
inverse opals, etc.). Keys to the success of this effort are the materials, functionally graded materials, and quasicrystals
development of enhanced measurement and analysis tools and (e.g., AI,, ,Pd,,,Mn, ,). There are always issues, of course,
a more complete, fundamental understanding of the when one develops a new material or combination of
underlying theory of thermoelectricity. Improved modeling materials. For the development of thermoelectric devices,
of materials, structures, and devices is also critically these issues include the ability to synthesize both n- and p-
important. type high-ZT materials, obtaining materials whose Z is high
While new materials are being developed, other team over a broad temperature range, developing materials tolerant
members are taking both conventional materials and these to repeated temperature cycling and with shock and vibration

2 18th International Conference on Thermoelectrics (1999)

resistance and finally, the ability to fabricate realistic (and calculated given its crystal or “amorphous” structure? Can
usefiil!) devices. In the future, costs of materials and the thermoelectric properties of amorphous, quasicrystalline,
production and environmental compatibility will also be and mesoporous materials be predicted before hand or
concerns. accurately modeled after the fact? Can the effects of
inhomogeneously distributed dopants be calculated and
therefore optimized before committing to an extensive
experimental program?

The Future
All of the above raises the question: Is the time right to
quadruple ZT? We at DARPA believe that the answer to this
question is a resounding yes for several key reasons. First, a
number of new synthetic techniques have become available to
make new superlattices (quantum wells), intercalation
compounds, “opals and inverse opals,” and mesoporous
materials at mild temperatures (e.g., using molten salt or
powder processing) and using a variety of novel molecular
precursors (e.g., derived from chemical vapor deposition,
liquid phase epitaxy, and sol-gel chemistry). In addition,
processing technology originally developed for the
semiconductor industry has now become available to grow
solid state materials and structures. One can now grow
macroscopic quantities of multi-layer structures rapidly and
economically. Finally, the fields of combinatorial materials
synthesis and combinatorial materials processing have opened
Figure 3. (a) Phenolic inverse opal fabricated by Allied- the possibility to rapidly explore the properties of vast arrays
Signal. Each hollow sphere is approximately 200 nm in of ternary, quaternary, etc. compounds.
diameter and may be filled with a variety of materials. (b) Advances in measurement and characterization tools now
Electrochemical deposition of a 2-dimensional array of allow researchers to rapidly screen large numbers of
bismuth telluride posts by the Jet Propulsion Laboratory. compounds. For example, high-speed x-ray crystallography
Each post in the array has a diameter of approximately 20 can be used to determine the crystal structure of a complex
microns. material in less than a day (whereas in the past it took several
months). High-resolution microscopy, scanning probe
True creativity is being shown in many of the DARPA- microscopy, and other surface analysis techniques have now
funded efforts through the development of novel become routine analytical tools to help characterize new
thermoelectric device structures. These include segmented materials. Finally, the development of rapid screening tools
thermoelectric converters that optimize performance through for ZT, the Seebeck coefficient, and the electrical and thermal
the use of combinations of materials whose properties are conductivities, have turned both thin film growth and
maximized over limited temperature ranges (JPL, Hi-Z combinatorial materials synthesis/processing into practical
Industries) and thermoelectric coolers operating in a transient tools for the development of new thermoelectric materials.
(or switched) mode where rapid cooling can occur before Theory has improved to the point where one can now
Joule heat diffuses to the cooled junction (MCC, IBM, calculate the band structure and many of the resulting
Marlow Industries). In addition, concepts for intimately properties of extended and complex materials. In addition,
connecting thermoelectric converters with micro-catalytic computational power has grown at such a rate that even now
combustors and shape memory alloy actuators may lead to one can tackle a material with a large unit cell with a desktop
new classes of high performance power generators and system.
compact, high power density actuators, respectively. Given the promise of these new developments, and despite
As noted above, theory and modeling can play important the potential pitfalls discussed above, contractors in the
roles in the discovery and optimization of thermoelectric DARPA program, in conjunction with those in efforts funded
materials and devices. Outstanding questions that will by the Office of Naval Research and the Army Research
hopefully be addressed in the DARPA program include: Is Office have demonstrated some major successes. This is
there an upper limit to ZT? What type of electronic band clearly illustrated by the data in Figure 4. For over 30 years
structure is necessary to maximize ZT? Are there conditions ZT was stalled out near I (in fact questions were raised as to
or materials in which 0 and K or CT and S can be more whether 1 was the theoretical maximum for ZT!). As the data
independent of each other? What is the lowest possible value in the figure clearly shows, demonstrable progress has been
of the thermal conductivity of an electrically conducting made. It is fair to say that ZTs near 2 are real and
material? Can the thermal conductivity of a material be reproducible and that ZTs approaching 3 may also have been

3 18th InternationalConference on Thermoelectrics(1 999)

observed. Proprietary materials, which may have an even and to the contributions from our technical agents John Pazik
higher ZT, have not been included in the figure. of the Office of Naval Research and John Prater of the Army
Key to the Success of this effort is the cooperative Research Office. Of course progress in this field would not
relationship established between organizations within the be possible without a diverse, exceptionally talented, and
Federal government and teaming between industry, extremely hard-working contractor base. Our hope is that
universities and national laboratories so that new results, they will reap the technical and financial rewards they so
technical hurdles and thought processes (both good and bad) richly deserve.
can be shared and collectively solved. Cooperation between
researchers synthesizing new materials and structures,
measuring their resulting properties, and
modelinghnderstanding their performance is also critical to
the success of the DARPA program.

DoD Goals

1950 1960 1970 1980 1990 2000

DARPA / ONR
Year Program
Start

Figure 4. History of the thermoelectric figure of merit

(ZT) for high temperature (solid), room temperature (dashed),
and cryogenic (dotted) materials. Proprietary materials,
which may have an even higher ZT, have not been included.

Still, two critical questions remain. (1) Is the slope of the

figure of merit vs. time curve in Figure 4 still strongly
positive or has the performance of thermoelectric materials
plateaued and (2) can the materials’ performance implied and
already demonstrated in this program be turned into useful
devices with enhanced cooling/power generating efficiency?
Whili: these questions may remain unanswered for some time,
it is clear that given the progress to date, we have entered into
a new era in thermoelectric materials research and
development. No longer will people question whether ZT is
pegged at 1 - a variety of interesting materials and structures
now reproducibly yield ZTs of 2 (and potentially above).
After a substantial funding hiatus, the time was clearly right
for DARPA and the DoD to reignite this field. The hope is
that tGe enthusiasm and excitement already generated will
continue into the development, demonstration, and
commercialization of novel thermoelectric devices of use in
both the military and civilian markets.

Acknowledgments
The success of the DARPA program is due in large part to
the outstanding program management skills of Stuart Wolf

4 18th International Conference on Themoelectrics (1999)