Thermoelectric Materials, Phenomena, and Applications: A Bird's

188 MRS BULLETIN • VOLUME 31 • MARCH 2006

IntroductionThermoelectric Phenomena:Background and Applications

Over the past decade, there has beenheightened interest in the field of thermo-electrics, driven by the need for more effi-cient materials for electronic refrigerationand power generation.1,2 Some of the re-search efforts focus on minimizing the lat-tice thermal conductivity, while otherefforts focus on materials that exhibit largepower factors. Proposed industrial andmilitary applications of thermoelectric

(TE) materials are generating increasedactivity in this field by demanding higher-performance high-temperature TE mate-rials than those that are currently in use.The demand for alternative energy tech-nologies to reduce our dependence on fossil fuels leads to important regimes of research, including that of high-temperature energy harvesting via thedirect recovery of waste heat and its con-version into useful electrical energy. Thus,the development of higher-performance

TE materials is becoming ever more im-portant. Power-generation applicationsare currently being investigated by the au-tomotive industry as a means to developelectrical power from waste engine heatfrom the radiator and exhaust systems foruse in next-generation vehicles. In addi-tion, TE refrigeration applications includeseat coolers for comfort and electroniccomponent cooling. Of course, the deep-space applications of NASA’s Voyagerand Cassini missions using radioisotopethermoelectric generators (RTGs) are wellestablished (see Reference 3 and the articleby Yang and Caillat in this issue). A keyfactor in developing these technologies isthe development of higher-performanceTE materials, either completely new mate-rials or through more ingenious materialsengineering of existing materials.

Thermoelectric refrigeration is an envi-ronmentally “green” method of small-scale, localized cooling in computers,infrared detectors, electronics, and opto-electronics as well as many other applica-tions. However, most of the electronicsand optoelectronics technologies typicallyrequire only small-scale or localized spotcooling of small components that do notimpose a large heat load. If significant eco-nomical cooling can be achieved, the re-sulting “cold computing” could producespeed gains of 30–200% in some computerprocessors based on complementarymetal oxide semiconductor (CMOS) tech-nology. Cooling of the processors is per-ceived by many to be the fundamentallimit to electronic system performance.4Thus, the potential payoff for the develop-ment of low-temperature TE refrigerationdevices is great, and the requirement forcompounds with properties optimizedover wide temperature ranges has led to amuch expanded interest in new TE mate-rials. Recent utilization of Peltier coolers(see next section) for the refrigeration ofbiological specimens and samples is anemerging TE application.

The development and potential of bulkmaterials for TE applications is an activearea of research. High-temperature bulkmaterials such as skutterudites, clathrates,half-Heusler alloys, and complex chalco-genides are being investigated (see the ar-ticle by Nolas et al. in this issue). Thesematerials possess complex crystal struc-tures and exhibit properties that are favor-able for potential thermoelectric materials.For example, skutterudites and clathratesare cage-like materials that have voids inwhich “rattler” atoms are inserted to sig-nificantly lower the thermal conductivitydue to the rattling atoms’ ability to scatterphonons. Recently, ceramic oxide mate-rials have also shown potential as high-

ThermoelectricMaterials,Phenomena, andApplications: ABird’s Eye View

Terry M.Tritt and M.A. Subramanian,Guest Editors

AbstractHigh-efficiency thermoelectric (TE) materials are important for power-generation

devices that are designed to convert waste heat into electrical energy.They can also beused in solid-state refrigeration devices.The conversion of waste heat into electricalenergy may play an important role in our current challenge to develop alternative energytechnologies to reduce our dependence on fossil fuels and reduce greenhouse gasemissions.

An overview of various TE phenomena and materials is provided in this issue of MRSBulletin. Several of the current applications and key parameters are defined anddiscussed. Novel applications of TE materials include biothermal batteries to powerheart pacemakers, enhanced performance of optoelectronics coupled with solid-state TEcooling, and power generation for deep-space probes via radioisotope TE generators. Anumber of different systems of potential TE materials are currently under investigation byvarious research groups around the world, and many of these materials are reviewed inthe articles in this issue.These range from thin-film superlattice materials to large single-crystal or polycrystalline bulk materials, and from semiconductors and semimetals toceramic oxides.The phonon-glass/electron-crystal approach to new TE materials ispresented, along with the role of solid-state crystal chemistry. Research criteria fordeveloping new materials are highlighted.

Keywords: energy, thermal conductivity, thermoelectricity.

www.mrs.org/bulletin

Thermoelectric Materials, Phenomena, and Applications: A Bird’s Eye View

MRS BULLETIN • VOLUME 31 • MARCH 2006 189

temperature TE materials (see Koumotoet al. in this issue). The potential ofnanomaterials and their role in TE re-search are an emerging area of interest(see Rao et al. in this issue). Bulk materialapplications are demanding new break-throughs in both materials and device en-gineering (see Yang and Caillat in thisissue). The role of thin-film properties, ap-plications, and recent results is also veryimportant (see Böttner et al. in this issue).A more complete overview of state-of-the-art materials, a theoretical and experimen-tal discussion of the basic principles, andan overview of some of the recent devel-opments and materials are given in textsby Tritt2 and Nolas.5

Seebeck and Peltier EffectsA discussion of thermoelectric effects

and devices should start with one of themost fundamental TE phenomena, theSeebeck effect, or thermopower.6–8 In theearly 1800s, Seebeck observed that whentwo dissimilar materials are joined to-gether and the junctions are held at differ-ent temperatures (T and T ∆T), a voltagedifference (∆V) develops that is propor-tional to the temperature difference (∆T).6The ratio of the voltage developed to thetemperature gradient (∆V/∆T) is relatedto an intrinsic property of the materialscalled the Seebeck coefficient, α. The Seebeck coefficient is very low for metals(only a few V/K) and much larger for semiconductors (typically a few hun-dred V/K).9 A related effect—the Peltiereffect—was discovered a few years laterby Peltier,10 who observed that when anelectrical current is passed through thejunction of two dissimilar materials, heatis either absorbed or rejected at the junc-tion, depending on the direction of thecurrent. This effect is due largely to thedifference in Fermi energies of the twomaterials. These two effects are related toeach other, as shown in the definition ofthe Peltier coefficient, Π αT. The rate atwhich the Peltier heat is liberated or re-jected at the junction (QP) is given by QP αIT, where I is the current through thejunction and T is the temperature inkelvin. There are also a number of ther-momagnetic effects such as the Hall,Ettingshausen, and Nernst effects that arebeyond the scope of this article. Thereader is referred to the text by Nolas et al.5 for a discussion of these effects.

Definition and Description of theFigure of Merit and ThermoelectricPerformance

The potential of a material for TE appli-cations is determined in large part by a

measure of the material’s figure of merit,ZT:*

(1)

where α is the Seebeck coefficient, σ is theelectrical conductivity, ρ is the electricalresistivity, and κ is the total thermal con-ductivity (κ κL κE, the lattice and elec-tronic contributions, respectively). Thepower factor, α2σT (or α2Τ/ρ), is typicallyoptimized in narrow-gap semiconductingmaterials as a function of carrier concen-tration (typically 1019 carriers/cm3),through doping, to give the largest ZT.9High-mobility carriers are most desirable,in order to have the highest electrical con-ductivity for a given carrier concentration.The ZT for a single material is somewhatmeaningless, since an array of TE couplesis utilized in a device or module.

There are two materials in the TE cou-ple, which is shown in Figure 1, an n-typeand a p-type. Ignoring parasitic contribu-tions that reduce the device performance,such as contact resistance and radiation ef-fects, the resulting figure of merit for thecouple (based solely on the TE materials)is given by

(2)

The coefficient of performance φ (refriger-ation mode) and the efficiency η (power-generation mode) of the TE couple aredirectly related to the figure of meritshown in Equation 3 for the efficiency. Theefficiency (η) of the TE couple is given bythe power input to the load (W) over thenet heat flow rate (QH), where QH is posi-tive for heat flow from the source to thesink:

(3)

where TH is the hot-side temperature, TC isthe cold-side temperature, and TM is theaverage temperature. Thus, one can see

TH TC

TH 1 ZTM1/2 1

1 ZTM1/2 TCTH,

η W

QH

ZT αp αn2T

ρnκn1/2 ρpκp1/2.

ZT α2σT

κ

α2Tρκ

,

that η is proportional to (1 ZTM)1/2 andthat the efficiency would approach theCarnot efficiency if ZT were to approachinfinity.

Thermoelectric Modules: DevicesThe Peltier effect is the basis for many

modern-day TE refrigeration devices, andthe Seebeck effect is the basis for TEpower-generation devices. The versatilityof TE materials is illustrated in Figure 1,which shows a TE couple composed of an n-type (negative thermopower andelectron carriers) and a p-type (positivethermopower and hole carriers) semicon-ductor material connected through metallic electrical contact pads. Both re-frigeration and power generation may beaccomplished using the same module, asshown in Figure 1. A TE module or deviceis built up of an array of these couples,arranged electrically in series and ther-mally in parallel. Thermoelectric energyconversion utilizes the Seebeck effect,wherein a temperature gradient is im-posed across the device, resulting in avoltage that can be used to drive a currentthrough a load resistance or device. This isthe direct conversion of heat into electric-ity. Conversely, the Peltier heat generatedwhen an electric current is passed througha TE material provides a temperature gra-dient, with heat being absorbed on thecold side, transferred through (or pumpedby) the TE materials, and rejected at thesink, thus providing a refrigeration capa-bility. The advantages of TE solid-state en-ergy conversion are compactness,quietness (no moving parts), and localizedheating or cooling. In addition, energy inthe form of waste heat (0% efficiency) thatwould normally be lost may be convertedinto useful electrical energy (7–8% effi-ciency) using a TE power-generation de-vice.

The best TE materials currently used indevices have ZT 1. This value has beena practical upper limit for more than 30years, yet there are no theoretical or ther-modynamic reasons for ZT 1 as anupper barrier. As seen from Equation 1,ZT may be increased by decreasing κL orby increasing either α or σ. However, σ istied to the electronic thermal conductivity,κE, through the Wiedemann–Franz rela-tionship, and the ratio is essentially con-stant at a given temperature.

Some of the goals of current research ef-forts are to find new materials that eitherraise the current efficiency of TE devices(i.e., increase ZT) or have the capability ofoperating in new and broader tempera-ture regimes, especially at lower tempera-tures (T 250 K) and higher temperatures(T 400 K). Over the past 30 years,

*The expressions for figure of merit, Z and ZT,are used interchangeably in the field of thermo-electrics. Z is the figure of merit with units of1/K (1/T), and ZT is the dimensionless (unit-less) figure of merit. Both must specify the tem-perature at which the quoted value wasobtained.



alloys based on the Bi2Te3 system[(Bi1–xSbx)2(Te1–xSex)3] and the Si1–yGey sys-tem have been extensively studied andoptimized for their use as TE materials toperform in a variety of solid-state TE re-frigeration and power-generation applica-tions.11,12 These traditional TE materialshave undergone extensive investigation,and there appears to be little room for fu-ture improvement in the common bulkstructures. However, recent results onnanostructures of traditional TE materialshave shown a promising new direction forthese materials. In addition, entirely newclasses of compounds will have to be in-vestigated. Figure 2 shows ZT as a func-tion of temperature for the Bi2Te3 andSi1–yGey materials as well as many of themore recent bulk materials that have beendeveloped over the last decade. The ZT ofmore exotic structures such as superlat-tices and quantum dot structures are notshown here but are addressed in the ar-ticle by Böttner et al. in this issue.

Transport PropertiesThe thermopower, or Seebeck coeffi-

cient, can be thought of as the heat per car-rier over temperature or, more simply, theentropy per carrier, α C/q, where C isthe specific heat and q is the charge of thecarrier.7 For the case of a classical gas, eachparticle has an energy of 3/2(kBT), wherekB is the Boltzmann constant. The ther-mopower is thus approximately kB/e,where e is the charge of the electron. Formetals, the heat per carrier is essentially aproduct of the electronic specific heat andthe temperature divided by the number of

carriers (N), that is, α CelT/N, and thenα is approximately

(4)

where EF is the Fermi energy (related tothe chemical potential of the material).

α Cel

q kB

e kBTEF

,

The Fermi energy is basically the energysuch that at T 0, all the states above this energy are vacant and all the states below are occupied. The quantitykB/e 87 V/K is a constant that repre-sents the thermopower of a classical elec-tron gas. Metals have thermopower valuesof much less than 87 V/K (on the orderof 1–10 V/K) and decrease with decreas-ing temperature, that is, EF kBT).

In a semiconductor, a charged particlemust first be excited across an energy gapEg. In this case, the thermopower is ap-proximated by

(5)

Thus, the thermopower is larger than thecharacteristic value of 87 V/K and in-creases with decreasing temperature.Semiconductors can exhibit either electronconduction (negative thermopower) orhole conduction (positive thermopower).The thermopower for different carriertypes is given by a weighted average of their electrical conductivity values (σn and σp):

(6)

It is necessary to dope the semiconductorswith either donor or acceptor states to

α αnσn αpσp

σn σp.

α Cel

q kB

e Eg

kBT.

Figure 1. Diagram of a Peltier thermoelectric couple made of an n-type and a p-typethermoelectric material. Refrigeration or power-generation modes are possible, dependingon the configuration. I is current.

Figure 2. Figure of merit ZT shown as a function of temperature for several bulkthermoelectric materials.



allow extrinsic conduction of the appro-priate carrier type, electrons or holes, re-spectively. It is apparent that the totalthermopower will be lower than that of ei-ther of the individual contributions, unlessthe direct bandgap is large enough—typically on the order of 10(kBT)—to effectively minimize minority carrier con-tributions. Typical thermopower valuesrequired for good TE performance are onthe order of 150–250 V/K or greater.

For high-temperature applications, it isimportant to minimize the contribution ofminority carriers in order to maintain ahigh thermopower. In addition, the ther-mal stability of the materials is an essentialaspect. Atomic diffusion within the mate-rials and interdiffusion of contacts can se-riously deteriorate the properties of agiven material at high temperatures. As-pects of this are discussed elsewhere.2,5

These materials and devices are expectedto operate at elevated temperatures forlong periods of time without deteriorationof their properties or performance. The ef-fects of diffusion and thermal annealingare important to thoroughly investigateand understand in any set of potential TEmaterials over the expected operatingtemperature range of the materials.

The description of electrical conductiv-ity for metals and semiconductors hasbeen covered extensively in many texts onsolid-state physics, and the reader is re-ferred there.13 There are a significant num-ber of carriers and states available forconduction in metals, typically n 1022

carriers/cm3. The electrical conductivity isthen very large for metals, on the order of106(Ω cm)–1. Again, for semiconductors,the carriers must be thermally excitedacross a gap for conduction to occur, asshown from the activated behavior that isderived for the temperature-dependenceof the electrical conductivity [σσ0exp(–Eg/kBT)]. There are two primaryways to achieve a high conductivity in asemiconductor, either by having a verysmall gap to excite across (Eg/kBT) or byhaving very high-mobility carriers, as dis-cussed later. Typical values of the electri-cal conductivity for a good TE material areon the order of about 103(Ω cm)–1.

The thermal conductivity κ is related tothe transfer of heat through a material, ei-ther by the electrons or by quantized vi-brations of the lattice, called phonons,such that κ κL κE, as mentioned ear-lier. The electrical conductivity and thethermal conductivity are interrelated, inthat σ is tied to κE through the Wiedemann–Franz relationship: κE L0σT, where theLorentz number L0 2.45 × 10–8 W Ω/K2

[or L0 2.45 × 10–8(V2/K2)]. The latticethermal conductivity is discussed later

in this article, in the section on minimumthermal conductivity. Typical thermalconductivity values for a good TE mate-rial are κ 2 W m–1 K–1, and typically,κL κE.

Investigating New ThermoelectricMaterialsThe “Phonon-Glass/Electron-Crystal”Approach

Slack has described the chemical char-acteristics of candidates for a good TE ma-terial.14 He states that the candidatesshould be narrow-bandgap semiconduc-tors with high-mobility carriers. Mahanhas also described the characteristics ofgood TE materials,15,16 agreeing with Slackthat the candidate material is typically anarrow-bandgap semiconductor [Eg10(kBT), or 0.25 eV at 300 K]. Also, the mobility of the carriers must remain high( 2000 cm2/V s), while the lattice ther-mal conductivity must be minimized. Insemiconductors, the Seebeck coefficientand electrical conductivity (both in the nu-merator of ZT) are strong functions of thedoping level and chemical composition.These quantities must therefore be opti-mized for good TE performance. The ther-mal conductivity of complex materials canoften be modified by chemical substitu-tions, and the lattice thermal conductivityneeds to be as low as possible. Under-standing these various effects and select-ing optimization strategies can be anexceedingly difficult problem, because incomplex materials there are often manypossible degrees of freedom. Slack sug-gested that the best TE material would be-have as a “phonon-glass/electron-crystal”(PGEC); that is, it would have the electri-cal properties of a crystalline material andthe thermal properties of an amorphousor glass-like material. Materials engineer-ing and the crystal chemistry approach togood TE materials are discussed later.

Minimum Thermal ConductivityIn many areas of research related to new

TE materials, attempts are being made toreduce the lattice part of the thermal con-ductivity to essentially its minimumvalue, that is, where a minimum latticethermal conductivity is achieved (whenall the phonons have a mean free path es-sentially equal to the interatomic spacingof the constituent atoms). This is being at-tempted by scattering phonons in differ-ent frequency ranges using a variety ofmethods such as mass fluctuation scatter-ing (a mixed crystal, in ternary and qua-ternary compounds), “rattling” scattering,grain-boundary scattering (due to the sizeof the grains), and interface scattering inthin films or multilayer systems.

The lattice thermal conductivity is givenby κL (1/3)(vsCLph), where vs is the ve-locity of sound, C is the heat capacity, andLph is the mean free path of the phonons.At high temperatures (T 300 K), thesound velocity and the heat capacity areessentially temperature-independent intypical materials. Therefore, the magnitudeand the temperature-dependence of κL arebasically determined by the mean free pathof the phonons. Slack defined the minimumthermal conductivity (κmin) as the thermalconductivity when the mean free path isessentially limited by the interatomic dis-tance between the atoms within the crys-tal.17 Typical analysis of κmin results invalues of κmin 0.25–0.5 W m–1 K–1.14,17

Minimum ThermopowerThere are certain practical limits for

each of the parameters used to calculateZT. These practical limits must be possiblein order to achieve a material viable forthermoelectric applications. For example,in Bi2Te3, in order to achieve a ZT 1 at T 320 K, σ 1 mΩ cm, α 225 V/K,and κ 1.5 W m–1 K–1. We have alreadydiscussed the ZT “barrier,” which in effectis given by minimizing the thermal con-ductivity. It is practical to investigate ma-terials where the electronic and latticeterms are comparable, on the order of0.75–1 W m–1 K–1. Let us look at the hypo-thetical situation of a material in which the lattice thermal conductivity is zero (κL 0). We will also assume the scatter-ing in this system is elastic and that theWiedemann–Franz relationship, slightlyrearranged [κE/σ L0T], is well behavedin this material. Then we can rewriteEquation 1 as

ZT α2T/ρκE α2/L0. (7)

Therefore, for a material to be a viable TEmaterial, it must possess a minimum ther-mopower that is directly related to thevalue of ZT and L0. Given this description,in order to achieve a certain value of ZT,the material would require that α (L0)0.5

157 V/K for ZT 1, and α (2L0)0.5 225 V/K for ZT 2. Of course, any“real” material will possess a finite κL, andthese values for the thermopower willhave to be higher to achieve the projectedvalues of ZT.

Solid-State Crystal ChemistryApproaches to AdvancedThermoelectric Materials

Thermoelectrics has always been a ma-terials design problem involving intricatetuning of structure–property relationshipsin complex solids through principles ofsolid-state chemistry and physics. The dis-



cussion thus far indicates that new mate-rials must be able to eventually achievecertain minimum values of important pa-rameters in order to be considered as a po-tential TE material. It does not matter if amaterial has a κL κmin; if it cannot be“tuned” or doped in order to attain a min-imum thermopower (150 V/K), it willnot be able to achieve ZT 1.

Classical Approach: Bulk BinarySemiconductors

Within the framework of simple elec-tronic band structure of solids, in general,metals are poor TE materials. Hence, mostof the early TE work put much emphasison semiconductors.18 As stated earlier, inorder to have a maximum ratio of electri-cal to thermal conductivity, the materialshould have a low carrier concentration,on the order of 1018–1019 cm3, with veryhigh mobilities. Crystal structure andbonding strongly influence the mobility.Materials with diamond or zinc-blendestructures with a high degree of covalentbonding frequently have high mobilities(e.g., Si, Ge, InSb), but also exhibit highthermal conductivity values. On the otherhand, low lattice thermal conductivitiesare found in conjunction with low Debyetemperatures and high anharmonic latticevibrations. These conditions are best satis-fied by highly covalent intermetallic com-pounds and alloys of the heavy elementssuch as Pb, Hg, Bi, Tl, or Sb, and S, Se, or Te.Once a material system has been selectedwith a favorable electrical-to-thermal con-ductivity ratio, one optimizes the compo-sition to further enhance the ZT by dopingthe material to maximize the density ofstates at the Fermi level and achieve a highSeebeck coefficient.

The most studied TE material, Bi2Te3,crystallizes in a layer structure (Figure 3)with rhombohedral–hexagonal symmetrywith space group R m (D5

3d). The hexa-gonal unit cell dimensions at room tem-perature are a 3.8 Å and c 30.5 Å. Thelayers stacked along the c-axis are

··· Te–Bi–Te–Bi–Te ··· Te–Bi–Te–Bi–Te ···.

The Bi and Te layers are held together bystrong covalent bonds, whereas the bond-ing between adjacent Te layers is of thevan der Waals type. This weak binding be-tween the Te layers accounts for the easeof cleavage along the plane perpendicularto the c-axis and the anisotropic thermaland electrical transport properties ofBi2Te3. For example, the thermal conduc-tivity along the plane perpendicular to thec-axis (1.5 W m–1 K–1) is nearly twice that ofthe value along the c-axis direction (0.7 W

3

m–1 K–1). When grown from a melt or byzone refining, the Bi2Te3 crystals are al-ways nonstoichiometric and show p-typebehavior. On the other hand, n-type mate-rials could be grown from the melt con-taining excess Te, iodine, or bromine. Thethermal conductivity values of both p- andn-type Bi2Te3 are 1.9 W m–1 K–1, giving aZT of about 0.6 near room temperature.Ioffe9 suggested that alloying could fur-ther reduce the lattice thermal conductiv-ity of Bi2Te3 through the scattering ofshort-wavelength acoustic phonons. Theoptimum compositions for thermoelectriccooling devices are normally Bi2Te2.7Se0.3(n-type) and Bi0.5Sb1.5Te3 (p-type) with ZT 1 near room temperature.

In contrast to Bi2Te3, PbTe crystallizes ina cubic NaCl-type crystal structure, andthe TE properties are isotropic. Both p-type and n-type thermoelements can beproduced by doping of acceptors (e.g.,Na2Te or K2Te) or donors (PbI2, PbBr2, orGe2Te3). In analogy with the Bi2Te3, thesolid-solution compositions (e.g., PbTe-SnTe) have been made to lower the latticethermal conductivity.19 The ZT value ofPbTe solid solutions is low near room tem-perature but rises to ZT 0.7 at 700 K,

making PbTe a prime candidate for powergeneration in that temperature range. It is possible to achieve ZT in excess ofunity at 700 K in structurally related solid-solution compositions, AgSbTe2 (80%)-GeTe(20%), known as TAGS (alloyscontaining Te, Ag, Ge, Sb). However, dueto high-temperature stability issues, thesecompositions are not currently favored inTE devices.

Neither Si nor Ge is a good TE material,as the lattice thermal conductivity is verylarge (150 W m–1 K–1 for Si and 63 W m–1

K–1 for Ge). The lattice thermal conductiv-ity can be substantially reduced by alloyformation between the two elements. Thebest alloy composition is Si0.7Ge0.3; its ther-mal conductivity is about 10 W m–1 K–1,and the reduction relative to Si and Ge is apparently due to the increasedphonon–phonon and phonon–electronscattering.19 Remarkably, such a large re-duction does not unduly reduce the car-rier mobility, and ZT 0.6–0.7 could berealized at elevated temperatures. Due totheir exceptional stability at high temper-atures (1200 K), these alloys are of inter-est to NASA for use in RTGs in deep-spaceprobes.

Figure 3. Crystal structure of the state-of-the-art thermoelectric material, Bi2Te3.The blueatoms are Bi and the pink atoms are Te.



Modern Solid-State ChemistryDesign Concepts for High-ZTMaterialsComplex Inorganic Structures. Most ofthe earlier investigations mentioned so farfocused on binary intermetallic semicon-ductor systems. Recent approaches tohigh-performance bulk TE materials focuson ternary and quaternary chalcogenidescontaining heavy atoms with low-dimensional or isotropic complex struc-tures to take advantage of the large carriereffective masses and low lattice thermalconductivity associated with such sys-tems.20 Along these lines, CsBi4Te6 pos-sessing the layered structure has beenidentified as a material showing a ZT of0.8 at 225 K, which is 40% greater than thatof the Bi2–xSbxTe3–ySey alloys.21 Other po-tential low-temperature TE materials currently under investigation are low-dimensional semiconducting or semimetal-lic doped layered pentatellurides (ZrTe5and HfTe5).22 These compounds have astructure similar to Bi2Te3, with van derWaals gaps between the individual layers.Although doped pentatellurides exhibitvery high power factors (exceeding theoptimally doped Bi2Te3 solid solutions) inthe low-temperature range (250 K), theirthermal conductivity is relatively high(4–8 W m–1 K–1), and the materials needto be compositionally tuned further tomake them useful as thermoelectrics.

Recently, cubic quaternary compoundswith a complex formula AgnPbmMnTem+2n

(M Sb, Bi), crystallizing in the PbTestructure, have been reported.23 The com-position AgPb10SbTe12 shows an excep-tionally high ZT value (2) at elevatedtemperature (shown in Figure 2). This isdue to the very low total thermal conduc-tivity of the bulk material, possibly arisingfrom compositional modulations (seen as“nanodots”) similar to the one found insuperlattices. If this is verified, it providesan additional “knob” to turn to achievehigh ZT in bulk materials. Another group of materials under investigation are half-Heusler alloys, with the generalformula MNiSn (M Zr, Hf, Ti). Acomplex composition of the typeZr0.5Hf0.5Ni0.5Pd0.5Sn0.99Sb0.01 shows a ZT of0.7 at T 800 K, highlighting the intricatebalance in structure, composition, and prop-erty relationships in these compounds.24

The β-Zn4Sb3 system has been reinvesti-gated for TE power-generation applica-tions at the Jet Propulsion Laboratory.25

Crystal Structures with “Rattlers.” Themethod of lowering the lattice thermalconductivity through mixed-crystal orsolid-solution formation does not alwaysproduce enough phonon scattering to

lower the lattice thermal conductivity toκmin. Slack’s concept of a “phonon-glass/electron-crystal,” described earlier,avoids this limitation. The concept of κminis successfully verified in crystal struc-tures with large empty cages or voidswhere atoms can be partially or com-pletely filled in such a way that they “rat-tle,” resulting in the scattering of theacoustic phonons. This approach espe-cially works well in highly covalent semi-conductor materials based on clathrates(Si, Ge, or Sn) and void structures formedby heavy elements of low electronegativ-ity differences (e.g., CoSb3-based skutteru-dites). Some doped skutterudites showexceptionally high ZT values at elevatedtemperatures (ZT 1.5 at 600–800 K).The structure–property relationships ofthese materials are discussed in the articleby Nolas et al. in this issue.

Oxide Thermoelectrics. There are nu-merous oxides with metal atoms in theircommon oxidation states that are stable atelevated temperatures and show electricalproperties ranging from insulating to su-perconducting. Nevertheless, oxides havereceived very little attention for TE appli-cations. This is due to their strong ioniccharacter, with narrow conduction band-widths arising from weak orbital overlap,leading to localized electrons with lowcarrier mobilities. This situation changedwith the unexpected discovery of good TEproperties in a strongly correlated layeredoxide, NaCo2O4.26 This oxide attains ZT 0.7–0.8 at 1000 K. Inspired by the strikingTE performance of NaCo2O4, most of thecurrent studies are focused on Co-basedlayered oxides, such as Ca3Co4O9 andBi2Sr3Co2Oy, crystallizing in “misfit”(lattice-mismatched) layered structures.26

Among the n-type oxides, Al-doped ZnO(Al0.02Zn0.98O) shows reasonably good TEperformance (ZT 0.3 at 1000 K).27

Rare-Earth Intermetallics with HighPower Factors. As mentioned earlier,metallic compounds are not suitable forTE applications. The exceptions to thisrule are intermetallic compounds contain-ing rare-earth elements (e.g., Ce and Yb),with localized magnetic moments wherethe Seebeck coefficient can approach100 V/K with metal-like conductivi-ties.28,29 In these compounds, the 4f levelslie near the Fermi energy and form nar-row non-parabolic bands, resulting in alarge density of states at the Fermi leveland large Seebeck coefficient values. Thehighest Seebeck values are found in cubicYbAl3 (n-type) and CePd3 (p-type). YbAl3shows a very high power factor(120–180 W/cm K2, or 3.6–5.4 W m–1 K–1)

at room temperature (300 K), which isnearly 4–5 times larger than that observedin optimized Bi2Te3-based thermo-electrics.30 Unfortunately, the large ther-mal conductivity (15–22 W m–1 K–1)lowers the ZT to about 0.2 at room tem-perature. Recently, the lattice thermal con-ductivity of this system has been loweredby doping Mn in the interstitial positions,resulting in the increase of ZT to about 0.4at room temperature.31 As mentioned ear-lier, ZT 1 requires a minimum Seebeckcoefficient value of 156 µV/K. The corre-lated metal with the highest Seebeck coef-ficient is CePd3, which has a maximum of125 V/K at 140 K.29 Future investigationsshould focus on increasing the Seebeck co-efficients of these materials above150 V/K through compositional andstructural tuning.

Engineered Crystal Lattices. The ap-proaches in bulk materials research relyheavily on the thermodynamic stability ofthe phases at a given condition, whereasthin-film deposition can yield metastable“designer” phases with unique properties.Quantum well systems (0D, 1D, and 2D)take advantage of their low-dimensionalcharacter through physical confinementsin quantum dots, nanowires, and thin-filmstructures to enhance the electronic prop-erties of a given material.32 In addition,nanostructured semiconductor materialscould scatter mid- to long-wavelengthphonons and thereby reduce the latticethermal conductivity to κmin.

Researchers at the Research Triangle In-stitute (RTI) have demonstrated a signifi-cant enhancement in ZT through theconstruction of Bi2Te3/Sb2Te3 superlat-tices.33 These materials exhibited ZT 2.4at T 330 K. The enhancement is attrib-uted to creating a “nanoengineered” ma-terial that is efficient in thermal insulationwhile remaining a good electrical conduc-tor. The thermal insulation arises from acomplex localized behavior for phonons,while the electron transmission is facilitatedby optimal choice of band offsets in thesesemiconductor heterostructures. Also, therehave been reports on PbTe/PbTeSe quantumdot structures that yield ZT 1.3–1.6.34

These materials have been grown as thickfilms that are then “floated off” the sub-strate to yield freestanding films, whichwere measured to yield these results. Theenhancement in ZT in the superlattice ma-terials appears to be more from a reduc-tion in the lattice thermal conductivitythan an increase in power factor.

SummaryCurrently, there are no theoretical or

thermodynamic limits to the possible



values of ZT. Given the current need foralternative energy technologies and mate-rials to replace the shrinking supply offossil fuels, the effort is becoming more ur-gent. Energy-related research will growrapidly over the next few years, andhigher-performance thermoelectric mate-rials and devices are direly needed. Slackestimated that an optimized phonon-glass/electron-crystal material could pos-sibly exhibit values of ZT 4.14 This givesencouragement that such materials maybe possible and could address many ofour energy-related problems. Thus, a systematic search and subsequent thor-ough investigation may eventually yieldthese much-needed materials for the nextgeneration of TE devices.

Although many strategies are being em-ployed in hopes of identifying novel TEmaterials, the PGEC approach appears tobe the best, as will become apparent in thefollowing articles. One has to decidewhether “holey” semiconductors (mate-rials with cages, such as skutterudites orclathrates) or “unholey” semiconductors(such as SiGe or PbTe) are the best to pur-sue, and which tuning parameters areavailable to improve these materials.35 Todate, none of the new materials has dis-placed the current state-of-the-art mate-rials (Bi2Te3, PbTe, or SiGe) in acommercialTE device. These materials have held thatdistinction for more than 30 years.

However, given the many materials yetto be investigated, there is certainly muchmore work ahead and promise for devel-oping higher-efficiency thermoelectricmaterials and devices. While the resultsare very exciting, thin films may be mostappropriate for small-scale electronic andoptoelectronics applications where smallheat loads or low levels of power genera-tion are more appropriate. To addresslarge-scale refrigeration (home refrigerators)or power-generation (automotive or in-dustrial) requirements, higher-perform-ance bulk materials will have to bedeveloped.

Certainly, theoretical guidance, in termsof band structure calculations and model-ing, will be essential to identifying themost promising TE materials. In addition,rapid yet accurate characterization of ma-terials and verification of results are alsoessential in order to effectively advancethis field of research. A multidisciplinary

approach will be required to develophigher-efficiency thermoelectric materialsand devices. The techniques used to de-velop “designer materials” needed forthermoelectrics will most likely prove im-portant in other areas of materials re-search as well.

References1. The reader is referred to the many MRS Symposium Proceedings volumes on the topicof thermoelectric materials and energy-conversion technologies: Thermoelectric Materi-als—New Directions and Approaches (Mater. Res.Soc. Symp. Proc. 478, 1997); Thermoelectric Mate-rials 1998—The Next Generation Materials forSmall-Scale Refrigeration and Power GenerationApplications (Mater. Res. Soc. Symp. Proc. 545,1999); Thermoelectric Materials 2000—The NextGeneration Materials for Small-Scale Refrigerationand Power Generation Applications (Mater. Res.Soc. Symp. Proc. 626, 2001); Thermoelectric Mate-rials 2001—Research and Applications (Mater.Res. Soc. Symp. Proc. 691, 2002); ThermoelectricMaterials 2003—Research and Applications(Mater. Res. Soc. Symp. Proc. 793, 2004); andMaterials and Technologies for Direct Thermal-to-Electric Energy Conversion (Mater. Res. Soc.Symp. Proc. 886, 2006) in press.2. T.M. Tritt, ed., “Recent Trends in Thermo-electric Materials Research,” Semiconductors andSemimetals, Vols. 69–71, treatise editors, R.K.Willardson and E. Weber (Academic Press,New York, 2000).3. Jet Propulsion Laboratory ThermoelectricScience and Engineering Web site,http://www.its.caltech.edu/jsnyder/ther-moelectrics/ (accessed February 2006).4. A.W. Allen, Laser Focus World 33 (March1997) p. S15.5. G.S. Nolas, J. Sharp, and H.J. Goldsmid,Thermoelectrics: Basic Principles and New Materi-als Developments (Springer, New York, 2001).6. T.J. Seebeck, Abh. K. Akad. Wiss. (Berlin, 1823)p. 265.7. D.T. Morelli, in Encyclopedia of AppliedPhysics, Vol. 21 (1997) p. 339.8. P.M. Chaiken, in Organic Superconductors, ed-ited by V.Z. Kresin and W.A. Little (PlenumPress, New York, 1990) p. 101.9. A.F. Ioffe, Semiconductor Thermoelements andThermoelectric Cooling (Infosearch, London,1957).10. J.C. Peltier, Ann. Chem. LVI (1834) p. 371.11. H.J. Goldsmid, Electronic Refrigeration (PionLimited, London, 1986).12. D.M. Rowe, ed., CRC Handbook of Thermo-electrics (CRC Press, Boca Raton, FL, 1995).13. P.L. Rossiter, The Electrical Resistivity of Met-als & Alloys (Cambridge Press, New York, 1987);F.J. Blatt, Physics of Electronic Conduction in Solids(McGraw-Hill, New York, 1968); L. Solymar

and D. Walsh, Electrical Properties of Materials,6th Ed. (Oxford Press, New York, 1998).14. G.A. Slack, in CRC Handbook of Thermo-electrics, ed. by D.M. Rowe (CRC Press, BocaRaton, FL, 1995) p. 407. 15. G.D. Mahan and J.O. Sofo, Proc. Natl. Acad.Sci. USA 93 (1996) p. 7436.16. G.D. Mahan, B. Sales, and J. Sharp, Phys.Today 50 (3) (1997) p. 42. 17. G.A. Slack, in Solid State Physics, Vol. 34, ed-ited by F. Seitz, D. Turnbull, and H. Ehrenreich(Academic Press, New York, 1979) p. 1.18. R.R. Heikes and R.W. Ure Jr., Thermoelectric-ity: Science and Engineering (Wiley Interscience,New York, 1961) p. 405.19. C. Wood, Rep. Prog. Phys. 51 (1988) p. 459.20. M.G. Kanatzidis, S.D. Mahanti, and T.P.Hogan, eds., Chemistry, Physics and MaterialsScience of Thermoelectric Materials: Beyond Bismuth Telluride (Plenum, New York, 2003)p. 35. 21. D.Y. Chung, T. Hogan, P. Brazis, M. Rocci-Lane, C. Kannewurf, M. Bastea, C. Uher,and M.G. Kanatzidis, Science 287 (2000) p. 1024.22. R.T. Littleton IV, T.M. Tritt, M. Korzenski, D.Ketchum, and J.W. Kolis, Phys. Rev. B. Rap. Com-mun. 64 121104 (2001).23. K.F. Hsu, S. Loo, F. Gao, W. Chen, J.S. Dyck,C. Uher, T. Hogan, E.K. Polychroniadis, and M.Kanatzidis, Science 303 (2004) p. 8181.24. Q. Shen, L. Chen, T. Goto, T. Hirai, J. Yang,G.P. Meisner, and C. Uher, Appl. Phys. Lett. 79(2001) p. 4165.25. T. Caillat, J.P. Fleurial, and A. Borshchevsky,J. Phys. Chem. Solids 58 (1997) p. 1119.26. I. Terasaki and N. Murayama, eds., OxideThermoelectrics (Research Signpost, Trivan-drum, India, 2002).27. M. Ohtaki, T. Tsubota, K. Eguchi, and H.Arai, J. Appl. Phys. 79 (1996) p. 1816. 28. R.J. Gambino, W.D. Grobman, and A.M.Toxen, Appl. Phys. Lett. 22 (1973) p. 506.29. G.D. Mahan, in Solid State Physics, edited byF. Seitz, H. Ehrenreich, and F. Spaepen (Acade-mic Press, New York, 1997) p. 51. 30. D.M. Rowe, G. Min, and V.L. Kuznetsov,Philos. Mag. Lett. 77 (1998) p. 105; D.M. Rowe,V.L. Kuznetsov, L.A. Kuznetsova, and G. Min,J. Phys. D: Appl. Phys. 35 (2002) p. 2183.31. T. He, T.G. Calvarese, J.-Z. Chen, H.D.Rosenfeld, R.J. Small, J.J. Krajewski, and M.A.Subramanian, Proc. 24th Int. Conf. on Thermo-electrics, edited by T.M. Tritt (IEEE, Piscataway,NJ, 2005) p. 434.32. L.D. Hicks and M.S. Dresselhaus, Phys. Rev.B 47 (1993) p. 12727.33. R. Venkatasubramanian, E. Siivola, T. Colpitts, and B. O’Quinn, Nature 13 (2001)p. 597.34. T.C. Harman, P.J. Taylor, M.P. Walsh, andB.E. LaForge, Science 297 (2002) p. 2229.35. T.M. Tritt, Science 283 (1999) p. 804.


Terry M. Tritt, GuestEditor for this issue ofMRS Bulletin, is a pro-fessor of physics atClemson University. Hereceived both his BA de-gree (1980) and his PhDdegree (1985) in physicsfrom Clemson Univer-sity and then went to theNaval Research Labora-tory (NRL) under a Na-tional Research Councilpostdoctoral fellowship.He subsequently becamea staff scientist at NRL,where he remained for11 years before joiningthe faculty at Clemsonin 1996. His primary re-search expertise lies inelectrical and thermaltransport properties andphenomena, and espe-cially in measurementand characterizationtechniques in novel ma-terials. He has extensiveexpertise in thermoelec-tric materials and meas-urement science and hasbuilt an internationallyknown laboratory for the measurement andcharacterization of ther-moelectric materialparameters, particularlythermal conductivity. Hehas recently become in-volved in the synthesisand characterizationof thermoelectricnanomaterials.

Tritt has served aslead organizer of threeMaterials Research Soci-ety symposia on ther-moelectric materials,edited the three-volumeRecent Trends in Thermo-electric Materials Research(Academic Press, 2000),and also recently editeda book published byKluwer Press on ther-mal conductivity. Hehas been a member ofthe executive board ofthe International Ther-moelectric Society (ITS)since 1999 and served aschair and host of the24th International Conference on Thermo-electrics (ICT-2005)

at Clemson in June 2005. He has writtenmore than 140 journalpublications and regu-larly gives invited pre-sentations at nationaland international meetings.

Tritt can be reached atClemson University, De-partment of Physics, 103Kinard Laboratory,Clemson, SC 29634,USA; tel. 864-656-5319and e-mail [email protected].

M.A. Subramanian,Guest Editor for thisissue of MRS Bulletin, isa research fellow atDuPont Central Re-search and Develop-ment. He holds BS andMS degrees in chemistryfrom the University ofMadras in India and re-ceived his PhD degreein solid-state chemistryin 1982 from the IndianInstitute of Technologyin Madras, where he fo-cused on synthesis andsolid-state studies of ox-ides with pyrochloreand perovskite struc-tures. He subsequentlyjoined the Departmentof Chemistry at TexasA&M University as anNSF postdoctoral fellow,where he worked on de-signing new fast ionconductors for solid-state batteries. He joinedDuPont in 1985 as a sci-entist and was recentlyappointed to researchfellow. Subramanian’scurrent interests include

the design and under-standing of structure–property relationships innew solid-state inorganicfunctional materials re-lated to superconductivity,colossal magnetoresistivematerials, high-κ andlow-κ dielectrics, ferro-electrics, multiferroics,oxyfluorination, andthermoelectrics.

Subramanian is a vis-iting professor at the Institut de Chimie de laMatière Condensée deBordeaux (ICMCB),University of Bordeaux,France. He serves as edi-tor for Solid State Sciencesand Progress in SolidState Chemistry, andserves or has served onthe editorial boards ofthe Materials ResearchBulletin, Chemistry of Ma-terials, and the Journal ofMaterials Chemistry. Hewas awarded theCharles Pedersen Medalby DuPont in 2004 forhis outstanding scien-tific, technological, andbusiness contributionsto the company. He hasauthored more than 200 publications andholds 42 U.S. patents,with 10 applicationspending.

Subramanian can bereached at DuPont Cen-tral Research and Devel-opment, ExperimentalStation, E328/219,Wilmington, DE 19880-0328, USA; tel. 302-695-2966 and [email protected].

Harald Böttner isdeputy head of theComponents and Micro-systems Department ofthe Fraunhofer Institutefor Physical Measure-ment Techniques inFreiburg, Germany, andis currently responsiblefor its thermoelectric ac-tivities. He graduatedwith a diploma degreein chemistry from theUniversity of Münsterin 1974 and received hisPhD degree in 1977 atthe same university forhis thesis on diffusionand solid-state reactionin the quaternary semiconductorII–VI/IV–VI materialsystem. He joined theFraunhofer Institut fürSilicatforschung in 1978and accepted his presentappointment in 1980. Hedeveloped IV–VI in-frared semiconductorlasers in parallel withactivities in thermo-electrics until 1995,when he became re-sponsible for the development of semi-conductor gas sensorsand thermoelectric materials until 2003.Böttner’s current re-search activities focuson thin-film andnanoscale thermo-electrics and microelec-tronics-related devicetechnology. He is the author or co-author ofmore than 100 publica-tions, holds more than10 patents, and is amember of the board

of the International Thermoelectric Society.

Böttner can bereached at FraunhoferIPM, Research FieldThermoelectrics, Hei-denhofstrasse 8, 79110Freiburg, Germany; tel.49-761-8857121 and e-mail [email protected].

Thierry Caillat is a sen-ior member of technicalstaff with the Jet Propul-sion Laboratory at theCalifornia Institute ofTechnology. He receivedhis PhD degree in mate-rials science from theNational PolytechniqueInstitute of Lorraine,France, in 1991. He thenreceived a National Re-search Council fellow-ship to study newmaterials for thermo-electric applications atJPL. He joined the per-manent staff at JPL in1994. Caillat’s primaryresearch interests havefocused on the identifi-cation and developmentof new thermoelectricmaterials and devices.In the last ten years, hehas played a key role in identifying severalfamilies of promising compounds for thermo-electric applications, including skutteruditesand β-Zn4Sb3–based materials. More recently,he has been involved in the development of advanced thermoelec-tric power-generationdevices for both


Terry M. Tritt M.A. Subramanian Harald Böttner Thierry Caillat



space and terrestrial applications.

Caillat has authoredor co-authored morethan 100 publications,given more than 40 in-vited presentations, andserved on numerous na-tional and internationalorganization committeesand panels. He was alsoa board member of theInternational Thermo-electric Society from1996 to 2005.

Caillat can be reachedat the California Insti-tute of Technology, JetPropulsion Laboratory,Mail Stop 277/207, 4800Oak Grove Drive,Pasadena, CA 91109,USA; tel. 818-354-0407and e-mail [email protected].

Gang Chen is a profes-sor of mechanical engineering at the Massachusetts Instituteof Technology. He re-ceived his BS (1984) andMS (1987) degrees fromthe Power EngineeringDepartment atHuazhong University ofScience and Technology,China, in 1984 and 1987,respectively, and hisPhD degree in mechani-cal engineering from theUniversity of California,Berkeley, in 1993. Hewas an assistant profes-sor at Duke Universityfrom 1993 to 1997 andan associate professor atthe University of Cali-fornia, Los Angeles,from 1997 to 2001. Hisresearch interests centeron microelectronics ther-mal management andnanoscale transport phenomena, particularlynanoscale heat transfer,and their applications in energy storage andconversion.

Chen is a recipient ofa K.C. Wong EducationFoundation fellowship,an NSF Young Investi-gator Award, and a JohnSimon Guggenheim

Foundation fellowship.He serves on the edito-rial boards of Annual Re-view of Heat Transfer, theJournal of Computationaland TheoreticalNanoscience, the Journalof Nanomaterials, and Microscale ThermophysicalEngineering. He alsoserves as the chair of theAdvisory Board of theASME NanotechnologyInstitute and on the ad-visory boards of several organizations.

Chen can be reachedat the Massachusetts Institute of Technology,Department of Mechani-cal Engineering, Room3-158, 77 MassachusettsAvenue, Cambridge,MA 02139-4307, USA;tel. 617-253-0006 and e-mail [email protected].

Ryoji Funahashi hasbeen a senior researcherat the National Instituteof Advanced IndustrialScience and Technologyin Japan since 1992. Hereceived his BS (1989)and MS (1992) degreesin physical chemistryand his PhD degree(1998) in crystallinematerial science fromNagoya University. His research interestsinclude novel synthetictechniques for high-performance superconducting oxidematerials and, recently,the exploration of newthermoelectric oxidematerials.

Funahashi is a recipi-ent of the Thermoelec-tric Conference ofJapan’s Best PaperAward, the Japan Journalof Applied Physics PaperAward, and a NEDO Industrial TechnologyResearch and Develop-ment Project grant. Hehas authored or co-authored more than 150papers, conference pro-ceedings, invited talks,and reviews. He is alsoa board member of theInternational Thermo-electric Society and theThermoelectric Society of Japan.

Funahashi can bereached at AIST,UBIQEN, MolecularMaterials and Devices,1-8-31 Midorigaoka,Ikeda, Osaka 563-8577,Japan; tel. 81-727-51-9485 and [email protected].

Xiaohua Ji is a postdoc-toral researcher in theDepartment of Physicsand Astronomy atClemson University. In2005, she received herPhD degree in materialsphysics and chemistryfrom Zhejiang Univer-sity in China under theguidance of XinbingZhao. For her graduatework, she developedsolvothermal and hydrothermal methodsfor synthesizing nano-structured thermoelec-tric materials. Shereceived the Best Scien-tific Paper Award at the

2004 ICT meeting inAdelaide, Australia.Under the guidance ofTerry M. Tritt at Clem-son, her current researchinvolves the synthesisand characterization ofnanostructured thermo-electric materials andnanocomposite bulkthermoelectric materials.

Ji can be reached atClemson University, De-partment of Physics andAstronomy, Clemson,SC 29634, USA; tel. 864-656-4596 and e-mail [email protected].

Mercouri Kanatzidis is aUniversity DistinguishedProfessor of Chemistry atMichigan State Univer-sity, where he has servedon the faculty since 1987.He received his BSdegree from AristotleUniversity in Greece,followed by a PhD de-gree in chemistry fromthe University of Iowa in 1984. He was then apostdoctoral research as-sociate at NorthwesternUniversity from 1985 to1987. He has generatedseminal work in metalchalcogenide chemistrythrough the develop-ment of novel syntheticapproaches aimed atnew materials discovery.His research interests include novel chalco-genides, thermoelectricmaterials, and the designof framework solids, in-termetallic phases, andorganic–inorganicnanocomposites.

Kanatzidis is a recipi-ent of the PresidentialYoung InvestigatorAward, the ACS MorleyMedal, the ACS ExxonSolid State ChemistryAward, and the Hum-boldt Prize. He has beena Guggenheim fellow aswell as a visiting profes-sor at the University ofNantes, the Universityof Münster, and the University of Munich.The bulk of his work is described in more than 450 research publications. He holdssix patents and is editorin chief of the Journal ofSolid State Chemistry.

Kanatzidis can bereached at MichiganState University, Depart-ment of Chemistry, 406Chemistry Building,East Lansing, MI 48824-1322, USA; tel. 517-355-9715 and [email protected].

Kunihito Koumoto is aprofessor in the Gradu-ate School of Engineer-ing at NagoyaUniversity, Japan. He re-ceived BS, MS, and PhDdegrees in appliedchemistry from the Uni-versity of Tokyo in 1974,1976, and 1979, respec-tively. He served as anassistant professor andassociate professor atthe University of Tokyobefore joining NagoyaUniversity as a full pro-fessor in 1992. His re-search focuses onthermoelectric materials

Gang Chen Ryoji Funahashi Xiaohua Ji Mercouri Kanatzidis


and bio-inspired pro-cessing of inorganic materials.

Koumoto received theRichard M. FulrathAward in 1993 and theAcademic AchievementAward of the CeramicSociety of Japan in 2000.He became a fellow ofthe American CeramicSociety and received theChinese Ceramic SocietyAward in 2005. He alsoserved the InternationalThermoelectric Societyas president from 2003to 2005 and is the authoror co-author of morethan 320 scientific papers and 38 books.

Koumoto can bereached at Nagoya Uni-versity, Graduate Schoolof Engineering, Furo-cho, Chikusa-ku,Nagoya, 464-8603,Japan; tel. 81-52-789-3327 and [email protected].

George S. Nolas is anassociate professor ofphysics at the Univer-sity of South Florida,where he has been since2001. He received hisPhD degree fromStevens Institute ofTechnology in 1994 andconducted pioneeringstudies on the synthesisand thermal propertiesof filled skutterudites as a postdoctoral associ-ate with Glen Slack atRensselaer PolytechnicInstitute. Before accept-ing his current position,

he spent five years as aphysicist and seniormember of the technicalstaff at Marlow IndustriesInc., a thermoelectricsmanufacturer in Dallas,Texas. His research inter-ests are in the synthesisand structure–propertyrelationships of newmaterials, and his cur-rent research focuses onnew materials for powerconversion and alterna-tive energy applications,including thermo-electrics, photovoltaics,and hydrogen storage.

Nolas is author ofThermoelectrics: BasicPrinciples and New Materials Developments,published by Springerwith co-authors JeffreySharp and JulianGoldsmid. He holdsthree patents, withanother two pending,on new materials forpower-conversion appli-cations. He has editedfour Materials ResearchSociety proceedings vol-umes (two as lead edi-tor) on thermoelectricmaterials research, or-ganized symposia forthe American PhysicalSociety and the Ameri-can Ceramic Society,given numerous invitedconference presentationsand seminars, and iscurrently in his secondterm as a board memberof the InternationalThermoelectric Society.

Nolas can be reachedat the University ofSouth Florida, Depart-

ment of Physics, PHY114, 4202 E. Fowler Av-enue, Tampa, FL 33620-5700, USA; tel.813-974-2233 and [email protected].

Joseph Poon is theWilliam Barton RogersProfessor of Physics atthe University of Vir-ginia, where he joinedthe Physics Departmentin 1980. He receivedboth his BS and PhD de-grees from the Califor-nia Institute ofTechnology and was aresearch associate in ap-plied physics at Stan-ford University. Hisareas of research haveincluded amorphoussuperconductors, quasi-crystals, bulk amorphousmetals, and thermoelec-tric alloys. He is a fellowof the American Physi-cal Society and wasnamed one of the “Sci-entific American 50” in2004 for the creation ofamorphous steel. He haspublished more than200 papers.

Poon can be reachedat the University of Virginia, Department ofAstronomy-Physics, POBox 400714, Jesse BeamsLab, Room 167, Charlottesville, VA22901, USA; tel. 434-924-6792 and [email protected].

Apparao M. Rao is aprofessor in condensed-matter physics at Clemson University. He

obtained his PhD degreein physics from the University of Kentuckyin 1989 and held a postdoctoral appoint-ment with Mildred S.Dresselhaus at MIT until 1991. His currentresearch focuses on understanding and con-trolling the synthesis of1D nanostructured organic and inorganicmaterials. He has pub-lished extensively on thesynthesis, characteriza-tion, and applications ofcarbon nanotubes.

Rao can be reached atClemson University, De-partment of Physics andAstronomy, 107 KinardLaboratory, Clemson,SC 29634-0978, USA; tel.864-656-6758 and [email protected].

Ichiro Terasaki is a pro-fessor in the AppliedPhysics Department atWaseda University inTokyo. He received hisBE, ME, and PhD de-grees in applied physicsfrom the University ofTokyo in 1986, 1988, and1992, respectively. Hebegan his research ca-reer as a research associ-ate at the University ofTokyo, moving on to be-come chief researcher atthe International Super-conductivity TechnologyCenter. He became anassociate professor atWaseda in 1997 and ac-cepted his current posi-tion as a full professorin 2004. His research in-

terests include experi-mental studies in con-densed-matter physics,especially charge trans-port properties of transi-tion-metal oxides,organic conductors, andintermetallic com-pounds.

Terasaki is the recipi-ent of the Sir MartinWood Prize and ICT’sBest Scientific PaperAward. He has pub-lished more than 150 papers, conference pro-ceedings, invited talks,books, and reviews.

Terasaki can bereached at Waseda Uni-versity, Department ofApplied Physics, 3-4-1Okubo, Shinjuku-ku,Tokyo 169-8555, Japan;tel. 81-3-5286-3854 ande-mail [email protected].

Rama Venkatasubra-manian is director of theCenter for Thermoelec-tric Research at RTI International in NorthCarolina and thefounder and chief tech-nology officer of Nex-treme Thermal SolutionsInc., a company spun offfrom RTI to commercial-ize its unique thin-filmsuperlattice thermoelec-tric technology. Heearned his PhD degreein electrical engineeringfrom Rensselaer Poly-technic Institute and is aNational Talent Scholarand a graduate of theIndian Institute of Tech-nology in Madras. Hisresearch interests in-clude photovoltaics, heteroepitaxy of novelmaterials, photonic materials, the study ofnanoscale thermalphysics, thermal management in high-performance electronics,and direct thermal-to-electric conversion de-vices. He is well knownfor pioneering thermo-electric superlattice materials and devices.Venkatasubramanian’s


Kunihito Koumoto George S. Nolas Joseph Poon Apparao M. Rao



work on superlatticeshas been recognized as asignificant breakthroughin thermoelectrics usingnanoscale engineeredmaterials. This technol-ogy has won an R&D100 Award (2002) andthe Technology of theYear Award (2005) from the Council for Entrepreneurial Devel-opment in North Carolina.

Venkatasubramanianis a recipient of theAllen B. Dumont Prizefrom Rensselaer, RTI’sMargaret Knox Excel-lence Award in Researchin 2002, and the IEEEEastern North CarolinaInventor of the Year in2003. He has severalpatents issued in thermoelectrics, has authored more than 100refereed publications,

and has contributed totwo book chapters.

Venkatasubramaniancan be reached at RTIInternational, Center forThermoelectricResearch, PO Box 12194,Research Triangle Park,NC 27709-2194, USA;tel. 919-541-6889 and e-mail [email protected].

Jihui Yang is a staff re-search scientist in the

Materials and ProcessesLaboratory at GeneralMotors Research andDevelopment Center.He received a BS degreein physics from FudanUniversity of China in1989, an MS degree in physics from the University of Oregon in1991, an MS degree inradiological physicsfrom Wayne State University in 1994, anda PhD degree in physicsfrom the University ofMichigan in 2000. Hisresearch interests in-clude low-temperaturetransport properties ofintermetallic compoundsand semiconductors,magnetism, thermoelec-tric materials, and thedevelopment of thermo-electric technology forautomotive applications.

Yang has publishedseveral book chapters

and more than 30 papers. He was the re-cipient of the GM DEGSfellowship in 1997 andthe Kent M. TerwilligerPrize for Best DoctoralThesis from the PhysicsDepartment of the Uni-versity of Michigan in2001. He has served onvarious committees forAPS and the MaterialsResearch Society and or-ganized several sym-posia for MRS andACerS. He was alsoelected to the board ofdirectors of the Interna-tional ThermoelectricSociety in 2005.

Yang can be reachedat General Motors Research and Develop-ment Center, Mail Code480-106-224, 30500Mound Road, Warren,MI 48090, USA; tel. 586-986-9789 and [email protected].

Rama Venkatasubramanian

Jihui YangIchiro Terasaki

For more information, see http://www.mrs.org/bulletin_ads


IntroductionAs the search for promising bulk ther-

moelectric materials intensifies, certainmaterial systems stand out as possessingthe highest potential for achieving ther-moelectric figures of merit well above unity.These materials possess relatively goodelectrical properties while maintainingvery low thermal conductivities. In somecases, enhancements in the electrical prop-erties have been realized. These resultshave attracted great attention, and manyresearch laboratories worldwide are nowworking on one or more of these materialsystems in order to achieve further im-provements in their thermoelectric prop-erties. We discuss some of the bulkthermoelectric materials of primary inter-est in this article.

SkutteruditesThe physical properties of skutterudites

depend sensitively on their compositions.This compositional dependence not onlyprovides a means to investigate the struc-ture–property relationships in this mate-rial system but also allows the optimizationof transport properties for thermoelectric(TE) applications. The diversity of poten-

tial compositional variants allows for richvariation in physical properties and is oneof the key reasons this material systemcontinues to be investigated by many re-search groups. One approach for optimiz-ing these materials is void-filling.

The skutterudite-type (CoAs3-type)structure is a cubic structure with thespace group composed of eight corner-shared TX6 (T Co, Rh, Ir; X P, As, Sb)octahedra. As depicted in Figure 1, thelinked octahedra produce a void, or va-cant site, at the center of the (TX6)8 cluster,occupying a body-centered position in thecubic lattice. This is a large void that canaccommodate relatively large metalatoms, resulting in the formation of filledskutterudites. Many different elementshave been introduced into the voids ofskutterudites, including lanthanide, ac-tinide, alkaline-earth, alkali, thallium, andGroup IV elements.1–3 The compositioncan be written as 2X8Y24 (typically X Co, Rh, or Ir; Y P, As, or Sb), with thesymbol illustrating the two voids percubic unit cell.

The concept of introducing “guest”atomsinto these voids to act as strong phonon-

scattering centers, thus greatly reducing thelattice thermal conductivity of these com-pounds, has resulted in improvements inthe TE properties of skutterudites.1,2 Thesmaller and heavier the ion in the voids, thelarger the disorder that is produced andthus the larger the reduction in the latticethermal conductivity. Skutterudite anti-monides possess the largest voids and aretherefore of particular interest for TE appli-cations. Results from Sales et al.4 andFleurial et al.5 show high ZT values (thecommon figure of merit for comparing dif-ferent TE materials) at elevated tempera-tures in LaFe3CoSb12 and CeFe3CoSb12 forboth p-type and n-type specimens. ZT val-ues approaching 1.4 above 900°C for thesematerials have been reported,5 indicatingtheir successful optimization for TE power-conversion applications.

It should be noted, however, that asmall concentration of void-fillers resultsin a large reduction in thermal conductiv-ity. Five percent of La6 or Ce,7 for example,in the voids of CoSb3 results in a thermal-conductivity reduction of 50%, as com-pared with CoSb3. In certain cases, higherpower factors have also been obtainedwith partial filling, as compared withmore fully filled, charge-compensatedcompositions. The aim in investigatingpartially filled skutterudites is realizing anoptimum electron concentration while re-


Recent Developmentsin BulkThermoelectricMaterials

George S. Nolas, Joe Poon, and Mercouri Kanatzidis

AbstractGood thermoelectric materials possess low thermal conductivity while maximizing

electric carrier transport.This article looks at various classes of materials to understandtheir behavior and determine methods to modify or “tune” them to optimize theirthermoelectric properties. Whether it is the use of “rattlers” in cage structures such asskutterudites, or mixed-lattice atoms such as the complex half-Heusler alloys, the abilityto manipulate the thermal conductivity of a material is essential in optimizing itsproperties for thermoelectric applications.

Keywords: alloy, compound, thermal conductivity, thermoelectricity.

Figure 1. Schematic illustration of askutterudite crystal, where the guestatom is inside a 12-coordinated “cage”(green) surrounded by yellow pnicogen(family of Bi, Sb, As, P, or N) atoms.Themetal sites are depicted in blue.Theoctahedral environment surrounding themetal sites is shown in the lower portionof the figure, also in blue.


ducing the thermal conductivity, therebymaximizing ZT. This is possible, since par-tially filled, uncompensated skutteruditesobey a rigid-band model, with the voidfillers being the electron donors. Thus, im-proved TE properties in partially filledskutterudites are achievable, owing to thecombination of a low thermal conductiv-ity with higher power factors as comparedwith those of nearly fully filled skutteru-dites. Figure 2 in the introductory articleby Tritt and Subramanian in this issueshows one example8 (Yb0.19Co4Sb12) ofhow partial filling can result in higher ZTvalues than full filling. Ba0.24Co4Sb12 hasalso demonstrated ZT 1 above 800 K.9Although these results are encouraging,and of technological interest for waste-heat recovery, still higher ZT values can beachieved by further optimization. Most re-cently, for example, a small amount of Nidoping for Co in Ba10 and Ca11 partiallyfilled skutterudites has resulted in higherZT values than the ZT values ofBaxCo4Sb12

9 and CaxCo4Sb12.12

The success of Slack and Tsoukala’s13

“rattling-atom” approach to skutteruditeresearch has lead to ZT 1 in a bulk TEmaterial. The research into skutteruditecompounds has also led to a greater un-derstanding of thermal transport proc-esses, revealed novel phonon-scatteringmechanisms, and resulted in efforts to ex-plore other materials with similar proper-

ties to obtain low thermal conductivitieswhile maintaining high power factors. Infact, the field is now focusing on low-thermal-conductivity compounds as amajor component in high-ZT thermoelec-tric materials, as we will see in several ex-amples to follow.

ClathratesClathrates are a class of novel materials

that can be thought of as periodic solids inwhich tetrahedrally bonded atoms (typi-cally Ge, Si, or Sn) form a framework ofcages that enclose relatively large metalatoms. Clathrates have demonstrated in-teresting properties that are rare incondensed-matter physics. One of themore interesting of these properties for TEapplications is a very low “glass-like”thermal conductivity in Type I clathrates.14

The Type I structure (Figure 2) can berepresented by the general formulaX2Y6E46, where X and Y are guest atomsencapsulated in two different polyhedra,while E represents the elements Si, Ge, or Sn.

There have been little data reported onthe high-temperature TE potential of thesematerials, and even less on the optimiza-tion of the TE properties of these materialsfor power-conversion applications. Blakeet al.15 used density functional calculationsto theoretically investigate the transportproperties of Type I clathrates. Their find-

ings showed room-temperature ZT valuesof up to 0.5 for optimized Sr8Ga16Ge30 andBa8In16Sn30 Type I clathrates, while ZT 1.7 at 800 K was predicted for optimizedcompositions. Kuznetsov et al.16 were firstto measure the Seebeck coefficient and re-sistivity above room temperature for theType I clathrates. Estimating the high-temperature thermal conductivity frompublished results of low-temperaturethermal transport in Type I clathrates re-sulted in ZT 0.7 at 700 K for Ba8Ga16Ge30and ZT 0.87 at 870 K for Ba8Ga16Si30.

Recently, research was initiated towardoptimizing the TE properties of Type Iclathrates above room temperature.17

Power factors approaching 1 W m–1 K–1 forGe clathrates were obtained at 650 K. To-gether with the low thermal conductivityachieved in these materials, these data in-dicate the potential for high ZT values atelevated temperatures. As outlined earlier,attempts at optimizing the TE propertiesof these interesting materials have only re-cently begun. Future studies, however,should also focus on other clathrate struc-ture types—Type VIII clathrates, for ex-ample. In n-type clathrates, optimizationhas reached ZT 0.4 at room tempera-ture,18 with theoretical predictions for p-type clathrates to be ZT 1.2 at 400 K.19

The Type II clathrate structure (Fig-ure 3) is particularly interesting, as it al-lows for partial filling of the polyhedra, incontrast to the totally filled Type I struc-ture. Thus, the electrical properties of TypeII clathrates can be more readily “tuned,”


Recent Developments in Bulk Thermoelectric Materials

Figure 2. Crystal structure of the Type I clathrate. Frameworkatoms are shown in blue, guest atoms inside the tetrakaidecahedraare orange, and guest atoms inside the pentagonal dodecahedraare purple.

Figure 3. Crystal structure of the Type II clathrate. Frameworkatoms are shown in blue, guest atoms inside the tetrakaidecahedraare orange, and guest atoms inside the pentagonal dodecahedraare purple.

allowing for better control of the dopinglevel.17 In compounds with the Type IIstructure, the interstitial “guest” atomsthat reside inside the polyhedra simulta-neously act as electrical dopants andphonon-scattering centers. These proper-ties allow an investigation into the interre-lationship between the guest atoms andtheir atomic cages and how this relation-ship affects their TE properties.

Half-Heusler AlloysHalf-Heusler (HH) intermetallic alloys

have recently received increasing atten-tion as potential TE materials for high-temperature applications. HH phases havethe MgAgAs (space group F 3m),20 or“half-stuffed GaAs,” crystal structure,consisting of three interpenetrating fccsublattices. Their chemical formula isXYZ, where X, Y, and Z can be selectedfrom many different elemental groups.21

Figure 4 shows the unit cell of TiNiSn, inwhich Ti and Sn occupy a NaCl lattice andNi occupies an fcc sublattice. HH phasesare semiconductors21–24 when the valencecount per formula unit is 8 (e.g., LiMgPhas a bandgap of 2.4 eV) or 18 (e.g., TiNiSn has a bandgap of 0.3 eV). SomeHH alloys are half-metallic ferromagnetsand heavy-fermion metals.22,25 The narrowbands give rise to a large effective massthat in turn leads to a large thermo-power.26 Because of the large thermopowerand spin-polarized band structure, HH al-loys have been investigated as thermo-electric and spintronic materials.21,25–28

The chemistries of the three sublatticescan be tuned independently. For example,in TiNiSn, doping the Sn site provides thecharge carriers, while doping the Ti andNi sites causes mass fluctuations that canlead to the reduction of thermal conduc-tivity. Many of the refractory metal26,27,29,30

and lanthanide metal28 thermoelectric HHalloys exhibit large room-temperatureSeebeck coefficients of 100 µV K–1 andmoderate electrical resistivities of 1–10µΩ m. The alloys reported are attractive ashigh-temperature TE materials, becausethey are relatively easily synthesized as100% dense samples. In particular, therefractory-based alloys exhibit high melt-ing points of 1100–1300°C as well aschemical stability and essentially zerosublimation at temperatures near 1000°C.

The effectiveness of doping in achievingimproved TE properties has been demon-strated in several HH alloys.26–32 Sb-dopedTiNiSn alloys exhibit power factors ashigh as 4.6 W m–1 K–1 at 650 K (380°C).29

Despite the large power factor values,there exists the challenge of reducing therelatively high thermal conductivity (κ 10 W m–1 K–1) that is evident in the HH al-

4

loys. The effect of disorder induced by dif-ferences in mass and atomic size is evidentin the reduction of κ.26–32

Notable progress was made when Shenet al. reported a maximum ZT 0.7 at800 K in n-type Sb-doped ZrNiSn alloyspartially substituted with the heavier Pdatoms in the Ni sublattice.31 More recently,Sakurada and Shutoh reported maximumZT values near 1.4 at 700 K in n-type(Zr0.5Hf0.5)0.5Ti0.5NiSn1–ySb.32 Despite failedattempts33 to reproduce the results ofSakurada and Shutoh, their report has al-ready generated considerable interest inHH alloys, as evidenced by the number ofpapers on HH alloys presented by re-searchers at the 2005 International Confer-ence on Thermoelectrics.

Meanwhile, recent approaches haveprovided new impetus for evaluating theZT of thermally stable, multicomponentHH alloys measured up to 1100 K. Indeed,the latest research has raised the prospectof achieving a ZT value near 1. In recentstudies33 performed jointly by two groups(University of Virginia and Clemson Uni-versity), the HH compositions were dopedwith Sb contents (up to 4%), which is sig-nificantly higher than previously reported.This is because it is recognized that higherdopant content can partly mitigate thecompensated behavior characterized by a

rollover in the Seebeck coefficient α ob-served in n-type alloys at high tempera-tures. It is believed that the shift of α(T)can lead to a larger power factor andtherefore higher ZT. The idea has beenverified in some (Ti,Zr,Hf)NiSn quinaryalloys when indeed it was found that athigher Sb doping content, the maximumin α2σT , where σ is the electrical conduc-tivity, is shifted to a higher temperature. Inaddition to doping the Sn sites with Sb,the refractory-metal sublattices are simul-taneously and randomly occupied by Hf,Zr, and Ti, providing additional tuning foroptimizing the power factor as well asfluctuations in the atomic mass and inter-atomic force for reducing the lattice ther-mal conductivity.34

Using the laser-flash method, κ(T) for three complex alloys have been meas-ured up to 1160 K by Tritt. The results are shown in the lower-right inset in Fig-ure 4. ZT for the quinary HH compositionHf0.75Zr0.25NiSn0.975Sb0.025 is also shown inFigure 4. It is noted that the ZT values ofHH alloys are already comparable withthose of state-of-the-art SiGe alloys near800°C.

While the prospect exists of achievingeven higher ZT values in complex, ther-mally stable, n-type HH alloys, p-type HHalloys such as MM’X (M Ti, Zr, and Hf;



Figure 4. Figure of merit ZT versus T for Hf0.75Zr0.25NiSn0.975Sb0.025 obtained from measured( ) and extrapolated ( ) thermal conductivities. (inset, upper left) Unit cell of TiNiSnhalf-Heusler phase. (inset, lower right) Thermal conductivities obtained from laser-flashthermal diffusivity and specific heat, with data points starting at 300 K:Hf0.75Zr0.25NiSn0.975Sb0.025 ( ), Hf0.6Zr0.25Ti0.15NiSn0.975Sb0.025 ( ), andHf0.75Zr0.25Ni0.9Pd0.1Sn0.975Sb0.025 ( ).

M’ Co, Ni, and Pt; X Sb and Sn, for ex-ample) should also be investigated. In-deed, TiCoSb exhibits a high Seebeckcoefficient of 400 µV K–1 at 300 K. Fur-thermore, one can exploit grain-boundaryscattering effects35 through grain-size re-finement. A bottom-up approach has alsobeen used to construct bulk nanostruc-tured samples that result in a lower latticethermal conductivity. 36

ββ-Zn4Sb3 AlloysA p-type intermetallic compound most

suitable for use as a high-ZT material atmoderate temperatures is one of the threemodifications of the Zn4Sb3 phase, namely,β-Zn4Sb3 (the other two modifications arethe α and γ phases).37 The β phase has thehexagonal rhombohedric crystal structure(space group R C, unit cell dimensions a 12.231 Å and c 12.428 Å) and is sta-ble between 263 K and 765 K. The highestZT value reported for β-Zn4Sb3 is 1.4 at400°C.38 Apparently, β-Zn4Sb3 starts todecompose into ZnSb and Zn phasesabove 400°C. Single-phase, polycrystallinesamples were made by melting andhomogenizing Zn and Sb in sealed quartzampoules. The obtained ingots wereground into powders. The alloy powderswere then hot-pressed to form crack-freesamples. In other reports, direct synthesismethods involving hot pressing or sparkplasma sintering were also reported.39–41

Both single-crystal and polycrystallinesamples have been investigated. Thepower factor in Zn4Sb3 is rather low(0.87 W m–1 K–1 at 400°C) for a high-ZTmaterial. For comparison, HH alloys havepower factors of 3–5 W m–1 K–1 at 400°C.What makes β-Zn4Sb3 a remarkable TEmaterial is its “phonon-glass” behavior,characterized by an unusually low ther-mal conductivity of 0.9 W m–1 K–1 at300 K.42 The already low thermal conduc-tivity in Zn4Sb3 can be further reduced bydoping the Zn and Sn sites.42,43 However,the doping concentrations are found to bequite limited; only a few percent of In orSn can be added to substitute Sb withoutresulting in a two-phase material. Even inthe case of substitution with the isostruc-tural compound Cd4Sb3, the solid-solutionrange of (Zn,Cd)4Sb3 is limited to only6 mol% Cd4Sb3 at 400°C.39 However, noimprovement in ZT above the highestrecorded value of 1.4 at 400°C has beenreported for the doped samples.

The detailed crystal structure of β-Zn4Sb3has been determined by employing bothsingle-crystal and powder x-ray diffractionmethods coupled with maximum entropyanalysis.42,44 The study has enabled theidentification of valence states of Sb atomsthat fully occupy the nonequivalent

==3

Sb(1) and Sb(2) sites as Sb3– and Sb24–

dimers, as well as the valence states of Znatoms that occupy 90% of the Zn(1) sitesas Zn2+. Overall, β-Zn4Sb3 is a valencecompound belonging to a similar class ofcompound semiconductors such as Bi2Te3and PbTe. The study has also uncoveredthree different interstitial sites for Znatoms in addition to the main Zn(1) sitesthat are only 90% occupied. Refinement ofthe structure based on this discovery of interstitial sites reveals a composition ofZn12.8Sb10. The mass density and composi-tion of the new crystal structure are reported to be in agreement with meas-urements. The Zn interstitials exhibit largethermal displacements. The glass-like in-terstitials are largely responsible for thephonon damping that suppresses the lat-tice thermal conductivity. There is appar-ently also Sb disorder along the c-axis thatcan contribute to the glass-like thermalconductivity. Electronic structure andtransport calculations using the crystalstructure obtained have identified thecompound as a p-type semiconductor,44 inagreement with experimental findings ofthe doping trend. It was demonstratedthat the Zn interstitial atoms play a dualrole as electron donors and thermopowerenhancers. First-principles calculations ofelectronic structure and thermoelectricproperties have also been performed.45

The band structure results reveal the cova-lent tendency of the compound, consistentwith the good carrier mobility measured.The high Seebeck coefficient can be attrib-uted to the strong energy-dependence of theFermi surface topology near band-filling.Recently, study of the low-temperaturephase of β known as α-Zn4Sb3 (composi-tion Zn13Sb10) reveals an electronic struc-ture similar to that of the β phase with abandgap of 0.3 eV.46 Further detailed in-vestigation of the structural and transportproperty changes across the α ↔ β transi-tion will shed light on the nature oforder–disorder transition in this complexcrystal system.

ChalcogenidesChalcogenide compounds comprise a

large class of materials that are predomi-nantly semiconductors. Most of the com-pounds are environmentally stable andcan have high melting points. They com-bine a set of properties that make themsuitable for practical thermoelectric appli-cations. Because of the versatility in thechoice of element and chalcogen and thesuitable electronegativity of the chalcogenelements, which varies little from sulfur totellurium, it is possible to obtain semicon-ductors with energy gaps appropriate forTE applications over a wide range of tem-

perature (typically 100–1400 K). Chalco-genide materials have had a prominentposition in the field of thermoelectricitygoing back to its early stages. For ex-ample, the cornerstone of today’s TE cool-ing has been the compound Bi2Te3 and itssolid solutions Bi2–xSbxTe3 and Bi2Te3–xSex.47

Another chalcogenide of importance isPbTe, which has a maximum ZT 0.8 at500°C; it has been suitable for powergeneration at intermediate temperatures.The germanium-based TAGS (Te-Ag-Ge-Sb) is more efficient than PbTe but hasfound little use, due to its high sublima-tion rate and low-temperature phase transition.48

The past decade has seen a strong resur-gence of interest in achieving superior TEperformance, and these “old” chalco-genides now have found new forms thathave led to more than incremental in-creases in ZT.49–51 Recent reports indicatethat nanostructured thin-film superlat-tices of Bi2Te3 and Sb2Te3 have ZT 2.4 atroom temperature,50 whereas PbSe0.98Te0.02/PbTe quantum dot superlattices grown bymolecular-beam epitaxy have ZT 1.6.50

These thin-film PbSe0.98Te0.02/PbTe sys-tems feature nanodots of PbSe embeddedin a PbTe material. At 550 K, thesesamples were reported to exhibit ZT 2.5.The primary reason for the high ZT valuesappears to be a very marked depression ofthe thermal conductivity, presumably bythe strong phonon scattering imposed bythe nanodots of PbSe.

One approach to searching for newthermoelectrics is to build exotic newstructures from scratch through solid-stateexploratory synthesis. This has affordedthe promising compounds BaBiTe3

52 andK2Bi8Q13 (Q Se,53,54 S55). These possesshighly anisotropic structures, low symme-try, and large unit cells, with “loosely”bound electropositive cations in channelsformed by extended Bi/Q frameworks, re-sulting in a low lattice thermal conductivity.

CsBi4Te6One noteworthy tellurium compound

is CsBi4Te6, which is in fact mixed-valentin Bi and can be viewed as containing Bi2+

ions that form Bi–Bi bonds. This com-pound features a strongly anisotropicstructure with one of the directions beingprominent in terms of its charge transportproperties. It possesses a lamellar struc-ture with slabs of (Bi4Te6)1– alternatingwith layers of Cs ions (Figure 5a).CsBi4Te6 is very responsive to the type andlevel of doping agent used. Low dopinglevels significantly affect the charge trans-port properties of CsBi4Te6. Appropriate p-type doping of CsBi4Te6 with SbI3 or Sbgives rise to a high ZTmax of 0.8 at 225 K,



which is among the highest values ever reported below room temperature.56 Atthis temperature, CsBi4Te6 is the best-performing TE material and raises newhope for extending the use of TE materialsto lower temperatures than ever before.Different dopants can allow the ZTmax tobe shifted to even lower temperatures, inthe neighborhood of 180 K.57 Band calcu-lations suggest that CsBi4Te6 has very ad-vantageous electronic structural featuresfor a promising TE material, characterizedby a great deal of anisotropic effectivemass. The electrical conductivity and ther-mopower are directly attributed to theelaborate electronic structure of a materialnear the Fermi level. The presence of Bi–Bibonds in the structure is responsible forthe material having a very narrow energygap, nearly half that of Bi2Te3. The nar-rower bandgap is related to the fact thatthe ZT maximum in CsBi4Te6 is achievedat lower temperatures than in Bi2Te3.

Tl9BiTe6 and Tl2SnTe5In the last decade, it has been more fully

appreciated that thallium chalcogenides

tend to possess very low thermal conduc-tivities. Tl9BiTe6 and Tl2SnTe5 are two suchnoteworthy phases. Tl9BiTe6 belongs to alarge group of ternary compounds whichcan be derived from the isostructuralTl5Te3 (Figure 5b).58 Tl9BiTe6 can be opti-mized to exhibit ZT 1.2 at 500 K, mainlydue to its extremely low lattice thermalconductivity of 0.39 W m–1 K–1 at 300 K.This value is nearly the same as the onereported for the PbSe0.98Te0.02/PbTe quan-tum dot superlattices grown by molecular-beam epitaxy.

The Tl2SnTe5 is a tetragonal phase, withchains of (SnTe5)2– running parallel to eachother and charge-balancing Tl ions situ-ated in between (Figure 5c). The Tl ionsare in an eightfold-coordinated site withrelatively long Tl–Te bonds. These longbonds produce very low-frequency phonons,which is one of the main reasons that thecompound has a very low thermal con-ductivity (0.5 W m–1 K–1). This compoundcan be optimized to a ZT of 1 at 500 K.59

Another interesting Tl-containing com-pound is Ag9TlTe5. It is isostructural toAg2Te, which has an even lower lattice

thermal conductivity than Tl2SnTe5, asshown in Figure 5d.60 Ag9TlTe5 combinesextremely low thermal conductivity andrelatively low electrical resistivity to give ZT 1.23 at 700 K. Unfortunately,thallium-containing compounds are un-likely to be accepted for practical use, dueto significant toxicity and environmentalissues. Nevertheless, they are interestingfrom a scientific standpoint, as they couldteach us a great deal about thermal con-ductivity and ZT optimization.

AgPbmSbTem+2Recently, the family of chalcogenide

lead-based compounds, AgPbmSbTem+2, orLAST-m materials (“LAST” for “lead anti-mony silver tellurium”), were revisited,and several suitably doped members werereported to exhibit large ZT values from1.2 (LAST-10) to 1.7 (LAST-18) at700 K.61

A large number of compositions can begenerated by changing m. This family ofcompounds is compositionally complex,yet its members possess an average NaClstructure. Research work conducted on



Figure 5. (a) Structure of CsBi4Te6.Yellow circles are Te atoms, blue circles are Bi atoms, and green circles are Cs atoms. (b) Structure ofTl9BiTe6.Yellow circles are Bi atoms, blue are Te atoms, and black are Tl atoms. (c) Structure of Tl2SnTe5. Black circles are Tl atoms, blue are Teatoms, and yellow are Sn atoms. (d) Thermal conductivity of Tl9BiTe6 and Tl2SnTe5. (e) High-resolution transmission electron microscopy imageof a typical nanoscale inhomogeneity (indicated by the red outline) found in Ag1–xPb18SbTe20.

this system in the 1950s and 1960s re-ported that the AgPbmSbTem+2 are solid so-lutions of the type (AgSbTe2)1–x(PbTe)x,and therefore the metal cations are ran-domly disordered in the NaCl structure.62

The metals Ag, Pb, and Sb were reportedas sitting on Na sites, while the chalcogenatoms occupied the Cl sites. However, re-cent work has shown that this is not thecase.63 Electron microscopic examinationof AgPb18SbTe20 samples revealed signifi-cant compositional modulations on the1–10-nm scale manifested in a variety ofways, including endotaxially dispersedquantum dots (see Figure 5e). This indi-cates that they are not solid solutions, rais-ing important new questions as to thesignificance and impact of these nano-structured features on TE properties. Thenanocrystals are endotaxially embeddedin the matrix, with a good lattice match onall surrounding surfaces (mismatch of2–4%). This endotaxy occurring in thesematerials could be highly conducive tofacile carrier transport throughout thesample.

The very high number of nanocrystal–matrix interfaces could provide a formida-ble barrier to phonon transmission in thebulk sample, thereby setting the foundationfor a marked enhancement in ZT relativeto that of the classical solid-solution sys-tem. The LAST materials show exception-ally low lattice thermal conductivity of0.45 W m–1 K–1 at 700 K (actual value de-pends on m). Experimental results indicatethat the formation of the nanostructuresand the TE performance of these materialsare very sensitive to the synthesis condi-tions and small changes in chemicalcomposition.

SummaryInvestigations into materials with low

thermal conductivity, and the reduction inthermal conductivity of other materialssystems, are now key aspects of researchinto advanced thermoelectric materials.An understanding of thermal transport innew materials can provide a means to in-corporate good electrical properties withlow thermal conductivity and result in im-proved TE performance. Whether it is theuse of “rattlers” in cage structures such asskutterudites or mixed-lattice atoms suchas the complex half-Heusler alloys, theability to manipulate the thermal conduc-tivity of a material is essential in obtaininggood TE materials. For example, the lowthermal conductivity of chalcogenides is akey property that leads to high ZT valuesfor these materials. A detailed under-standing of the phonon propagation inthese systems is thus desirable. Undoubt-edly, additional work is needed to better

understand these materials systems. How-ever, as proposed by Slack in the early1990s, the phonon-glass/electron-crystalapproach continues to be the most fruitfulmethod of research in our search for higher-performance thermoelectric materials.

References1. G.S. Nolas, D.T. Morelli, and T.M. Tritt,Annu. Rev. Mater. Sci. 29 (1999) p. 89 and refer-ences therein.2. C. Uher, in Semiconductors and Semimetals,Vol. 69, edited by T.M. Tritt (Academic Press,New York, 2000) p. 139 and references therein.3. B.C. Sales, in Handbook of the Physics andChemistry of Rare Earths, Vol. 33 (Elsevier Sci-ence, Amsterdam, 2002) p. 1.4. J.-P. Fleurial, A. Borshchevsky, T. Caillat, D.T.Morelli, and G.P. Meisner, in Proc. 15th Int. Conf.Thermoelectrics (IEEE, Piscataway, NJ, 1996)p. 91.5. B.C. Sales, D. Mandrus, and R.K. Williams,Science 272 (1996) p. 1325.6. G.S. Nolas, J.L. Cohn, G.A. Slack, and S.B.Schujman, Appl. Phys. Lett. 73 (1998) p. 176.7. D.T. Morelli, G.P. Meisner, B. Chen, S. Hu,and C. Uher, Phys. Rev B 56 (1997) p. 7376.8. G.S. Nolas, M. Kaeser, R. Littleton IV, andT.M. Tritt, Appl. Phys. Lett. 77 (2000) p. 1822.9. J.S. Dyck, W. Chen, C. Uher, L. Chen, X. Tang, and T. Hirai, J. Appl. Phys. 91 (2002)p. 3698.10. X. Tanga, Q. Zhang, L. Chen, T. Goto, and T. Hirai, J. Appl. Phys. 97 093712 (2005).11. M. Puyet, A. Dauscher, B. Lenoir, M. Dehmas, C. Stiewe, E. Müller, and J.Hejtmanek, J. Appl. Phys. 97 083712 (2005).12. M. Puyet, B. Lenoi, A. Dauscher, M. Dehmas, C. Stiewe, and E. Müller, J. Appl.Phys. 95 (2004) p. 4852.13. G.A. Slack and V.G. Tsoukala, J. Appl. Phys.76 (1994) p. 1665.14. G.S. Nolas, G.A. Slack, and S.B. Schujman,in Semiconductors and Semimetals, Vol. 69, editedby T.M. Tritt (Academic Press, San Diego, 2001)p. 255.15. N.P. Blake, S. Latturner, J.D. Bryan, G.D.Stucky, and H. Metiu, J. Chem. Phys. 115 (2001)p. 8060.16. V.L. Kuznetsov, L.A. Kuznetsova, A.E. Kaliazin, and D.M. Rowe, J. Appl. Phys. 87(2000) p. 7871.17. G.S. Nolas, Thermoelectrics Handbook: Macro- to Nano-Structured Materials, edited byD.M. Rowe (CRC Press, Boca Raton, FL) inpress.18. A. Bentien, V. Pacheco, S. Paschen, Y. Grin, and F. Steglich, Phys. Rev. B 71 165206(2005).19. G.K.H. Madsen, K. Schwarz, P. Blaha, andD.J. Singh, Phys. Rev. B 68 125212 (2003).20. W. Jeischko, Metall. Trans. A 1 (1970) p. 3159.21. S.J. Poon, in Recent Trends in ThermoelectricMaterials Research II, edited by T.M. Tritt, Semi-conductors and Semimetals, Vol. 70, Chap. 2, trea-tise editors, R.K. Willardson and E.R. Weber(Academic Press, New York, 2001) p. 37.22. J. Tobola, J. Pierre, S. Kaprzyk, R.V. Skolozdra, and M.A. Kouacou, J. Phys. Condens.Matter 10 (1998) p. 1013.

23. F.G. Aliev, N.B. Brandt, V.V. Moschalkov,V.V. Kozyrkov, R.V. Scolozdra, and A.I. Belogorokhov, Phys. B: Condens. Matter 75(1989) p. 167.24. S. Ogut and K.M. Rabe, Phys. Rev. B 51(1995) p. 10443.25. W.E. Pickett and J.S. Moodera, Phys. Today54 (2001) p. 39.26. C. Uher, J. Yang, S. Hu, D.T. Morelli, andG.P. Meisner, Phys. Rev. B 59 (1999) p. 8615.27. H. Hohl, A.P. Ramirez, C. Goldmann, G.Ernst, B. Wolfing, and E. Bucher, J. Phys. Con-dens. Matter 11 (1999) p. 1697.28. S. Sportouch, P. Larson, M. Bastea, P. Brazis,J. Ireland, C.R. Kannenwurf, S.D. Mahanti, C.Uher, and M.G. Kanatzidis, in ThermoelectricMaterials 1998—The Next Generation Materialsfor Small-Scale Refrigeration and Power GenerationApplications, edited by T.M. Tritt, M.G.Kanatzidis, G.D. Mahan, and H.B. Lyon Jr.(Mater. Res. Soc. Symp. Proc. 545, Warrendale,PA, 1999) p. 421.29. S. Bhattacharya, A.L. Pope, R.T. LittletonIV, T.M. Tritt, V. Ponnambalam, Y. Xia, and S.J.Poon, Appl. Phys. Lett. 77 (2000) p. 2476.30. Y. Xia, S. Bhattacharya, V. Ponnambalam,A.L. Pope, S.J. Poon, and T.M. Tritt, J. Appl.Phys. 88 (2000) p. 1952.31. Q. Shen, L. Chen, T. Goto, T. Hirai, J. Yang,G.P. Meisner, and C. Uher, Appl. Phys. Lett. 79(2001) p. 4165.32. S. Sakurada and N. Shutoh, Appl. Phys. Lett.86 (2005) p. 2105.33. S.R. Culp, S.J. Poon, N. Hickman, T.M. Tritt,and J. Blumm, Appl. Phys. Lett. 88 042106 (2006).34. Y. Yang, G.P. Meisner, and L. Chen, Appl.Phys. Lett. 85 (2004) p. 1140.35. J.W. Sharp, S.J. Poon, and H.J. Goldsmid,Phys. Status Solidi A 187 (2001) p. 507.36. S. Bhattacharya, T.M. Tritt, Y. Xia, V. Ponnambalam, S.J. Poon, and N. Thadhani,Appl. Phys. Lett. 81 (2002) p. 43.37. H.W. Mayer, I. Mikhail, and K. Schubert,J. Less-Common Metals 59 (1978) p. 43.38. T. Caillat, J.-P. Fleurial, and A. Borshchevsky, J. Phys. Chem. Solids 58 (1997)p. 1119.39. V.L. Kuznetsov and D.M. Rowe, J. AlloysCompd. 372 (2004) p. 103.40. S.C. Ur, I.H. Kim, and P. Nash, Mater. Lett.58 (2004) p. 2132.41. K. Ueno, A. Yamamoto, T. Noguchi, T. Inoue, S. Sodeoka, H. Takazawa, C.H. Lee,and H. Obara, J. Alloys Compd. 385 (2004) p. 254.42. G.J. Snyder, M. Christensen, E. Nishibori, T. Caillat, and B.B. Iversen, Nature Mater. 3(2004) p. 458.43. M. Tsutsui, L.T. Zhang, K. Ito, and M. Yamaguchi, Intermetallics 12 (2004) p. 809.44. F. Cargnoni, E. Nishibori, P. Rabiller, L. Bertini, G.J. Snyder, M. Christensen, and B.B. Inversen, Chem. Eur. J. 20 (2004) p. 3861.45. S.G. Kim, I.I. Mazin, and D.J. Singh, Phys.Rev. B 57 (1998) p. 6199.46. J. Nylen, M. Andersson, S. Lidin, and U. Haeussermann, J. Am. Chem. Soc. 126 (2004)p. 16306.47. G.S. Nolas, J. Sharp, and H.J. Goldsmid,Thermoelectrics: Basic Principles and New Materi-als Developments (Springer, New York, 2001).48. E. Skrabeck and D.S. Trimmer, in CRCHandbook of Thermoelectrics, edited by



D.M. Rowe (CRC Press, Boca Raton, FL, 1995)p. 267.49. R. Venkatasubramanian, E. Siivola, T. Colpitts, and B. O’Quinn, Nature 413 (2001)p. 597.50. T.C. Harman, P.J. Taylor, M.P. Walsh, andB.E. LaForge, Science 297 (2002) p. 2229.51. L.D. Hicks, T.C. Harman, and M.S. Dresselhaus, Appl. Phys. Lett. 63 (1993) p. 3230.52. D.Y. Chung, S. Jobic, T. Hogan, C.R. Kannewurf, R. Brec, J. Rouxel, and M.G.Kanatzidis, J. Amer. Chem. Soc. 119 (1997) p. 2505.53. T.J. McCarthy, S.P. Ngeyi, J.H. Liao, D.C.DeGroot, T. Hogan, C.R. Kannewurf, and M.G.Kanatzidis, Chem. Mater. 5 (1993) p. 331.54. T. Kyratsi, J.S. Dyck, W. Chen, D.Y. Chung,C. Uher, K.M. Paraskevopoulos, and M.G.

Kanatzidis, J. Appl. Phys. 92 (2002) p. 965.55. M.G. Kanatzidis, T.J. McCarthy, T.A.Tanzer, L. Chen, L. Iordanidis, T. Hogan, C.R.Kannewurf, C. Uher, and B. Chen, Chem. Mater.8 (1996) p. 1465.56. D.Y. Chung, T. Hogan, P. Brazis, M. Rocci-Lane, C.R. Kannewurf, M. Bastea, C.Uher, and M.G. Kanatzidis, Science 287 (2000)p. 1024.57. D.Y. Chung, T.P. Hogan, M. Rocci-Lane, P.Brazis, J.R. Ireland, C.R. Kannewurf, M. Bastea,C. Uher, and M.G. Kanatzidis, J. Am. Chem. Soc.126 (2004) p. 6414.58. B. Wolfing, C. Kloc, J. Teubner, and E.Bucher, Phys. Rev. Lett. 86 (2001) p. 4350.59. J.W. Sharp, B.C. Sales, and D.G. Mandrus,Appl. Phys. Lett. 74 (1999) p. 3794.

60. K. Kurosaki, A. Kosuga, H. Muta, M. Uno,and S. Yamanaka, Appl. Phys. Lett. 87 061919(2005).61. K.F. Hsu, S. Loo, F. Guo, W. Chen, J.S. Dyck,C. Uher, T. Hogan, E.K. Polychroniadis, andM.G. Kanatzidis, Science 303 (2004) p. 818.62. S. Sportouch, M. Bastea, P. Brazis, J. Ireland,C.R. Kannewurf, C. Uher, and M.G. Kanatzidis,in Thermoelectric Materials 1998—The Next Gen-eration Materials for Small-Scale Refrigeration andPower Generation Applications, edited by T.M.Tritt, M.G. Kanatzidis, G.D. Mahan, and H.B.Lyon Jr. (Mater. Res. Soc. Symp. Proc. 545, War-rendale, PA, 1999) p. 123.63. E. Quarez, K.F. Hsu, R. Pcionek, N. Frangis,E.K. Polychroniadis, and M.G. Kanatzidis,J. Amer. Chem. Soc. 127 (2005) p. 9177.



For more information, see http://www.mrs.org/bulletin_ads



IntroductionWaste heat from automobiles, factories,

and similar sources offers a high-qualityenergy source equal to about 70% of thetotal primary energy, but it is difficult toreclaim because its sources are small andwidely dispersed. Thermoelectric (TE)generation systems offer the only viablemethod of overcoming these problems byconverting heat energy directly into electri-cal energy irrespective of source size andwithout the use of moving parts or theproduction of environmentally deleteriouswastes. The requirements placed on mate-rials needed for this task, however, are noteasily satisfied. Not only must they pos-sess a high conversion efficiency, but theymust also be composed of nontoxic andabundantly available elemental materialshaving high chemical stability in air, evenat temperatures of 800–1000 K. Oxide ma-terials, such as those used in the present

study, are particularly promising for TEapplications because of their stability evenat high temperatures in air.

The challenge of creating novel TE oxides has recently motivated investiga-tions from various materials viewpoints.As it is difficult to control an electronicsystem and a phonon system simultane-ously in a simple crystalline field, a com-plex crystal composed of more than twonanoblocks with different compositionsand structural symmetries, the so-calledhybrid crystal, is considered to be effectivein controlling electron and phonon trans-port separately, thus enhancing the totalconversion efficiency.

Layered cobalt oxides, such as NaxCoO2and Ca3Co4O9, can be regarded as consistingof complex crystalline fields.1–4 In theseoxides, CoO2 nanosheets possessing astrongly correlated electron system serve

as electronic transport layers, while sodiumion nanoblock layers or calcium cobaltoxide misfit layers serve as phonon-scattering regions to achieve low thermalconductivity.5,6 This fact inspired us togenerate high-performance TE materialsfrom natural superlattices or hybrid crys-tals that are composed of the periodicarrangement of nanoblocks or nanosheetspossessing different TE functions.

TE modules using oxide materials havebeen reported;7,8 however, their perform-ance is much lower than expected, con-sidering the properties of their startingmaterials. This is thought to be becausethe contact resistance is very high at elec-trodes in which oxide/metal junctions areusually formed, thus severely limiting themagnitude of the output power. More-over, cracking and exfoliation due to thegreat difference in thermal expansion be-tween oxides and metals are also seriousproblems. Given the high temperatures(773 K) of applications with TE oxides,conventional materials and methods forconstructing electrodes cannot be used.Accordingly, the preparation of electrodespossessing good mechanical and electricalproperties is considered to be one of themost important issues in realizing TEpower generation. Here, we describe TEproperties of p-type layered CoO2 and n-type Srn1TinO3n1 materials. The fabrica-tion of modules composed of TE oxideswill be demonstrated, along with theirpower-generating properties.

Nanoblock IntegrationCrystals, in general, can be regarded as

being composed of coordination poly-hedra as structural units, and they are connected with each other to form nano-blocks or nanosheets as physical propertyunits. So if more than two kinds of unitnanoblocks, with different compositionsand symmetries, are integrated into super-lattices or hybrid crystals, each block canplay its own role in generating a specificfunction, and hence, electron and phonontransport can be independently controlled(Figure 1). These functions are combinedto give rise to high TE performance. Inter-facial effects may be generated, leading tofurther performance enhancements.

p-Type Oxides:ThermoelectricProperties of the Layered Co Oxides

At present, more than ten layered Cooxides have been discovered since the discovery of the good p-type TE proper-ties of NaxCoO2.1 The crystal structure ofNaxCoO2 is schematically shown in Fig-ure 2a, in which the CdI2-type CoO2 layerand the Na layer alternately stack along

Complex OxideMaterials forPotentialThermoelectricApplications

Kunihito Koumoto, Ichiro Terasaki, and Ryoji Funahashi

AbstractLayered CoO2 materials are excellent candidates for potential thermoelectric

applications.Their single crystals show good p-type thermoelectric properties attemperatures higher than 800 K in air. Recently, the mechanism of thermoelectricproperties was clarified through a discussion of electronic and crystallographic structure.In order to fabricate thermoelectric modules possessing good power-generationproperties, thermoelectric materials and metallic electrodes must be connected with lowcontact resistance and high mechanical strength. It has been found that good junctionscan be formed using Ag paste including p- and n-type oxide powders.The role of spinentropy contributions to thermopower will be presented, in connection with strongelectron correlation and triangular lattices.

Keywords: oxide, thermal conductivity, thermoelectricity.


Complex Oxide Materials for Potential Thermoelectric Applications


the c-axis. The Na content x is varied from0.3 to 1.0, which concomitantly changesthe crystal structures indexed as the α, α,β, and γ phases.9–11 In particular, the hy-drated x 0.35 sample exhibits a super-conducting transition at 5 K.12 Good TEproperties are observed near x 0.55–0.7(the γ phase), which has been calledNaCo2O4 after Jansen and Hoppe,13 whofirst synthesized this compound in the1970s.

Figure 2b corresponds to the Ca-basedCo oxide that has been known asCa3Co4O9.2 Recently, its crystal structurewas precisely analyzed and found to be amisfit-layer compound composed of alter-nate stacks of the hexagonal CoO2 layer ofCdI2 type and the square Ca2CoO3 layer ofNaCl type.3,4 Such a “hexagon-on-square”structure causes a highly distorted inter-face and a b-axis lattice misfit. Figure 2cshows the crystal structure of the Bi-basedCo oxide Bi2Sr2Co2Oy that was thought to an isomorphic material to the high-temperature superconducting Cu oxideBi2Sr2CaCu2O8.14 This material is the sec-ond example of the misfit-layered oxide,consisting of the CoO2 layer and theBi2Sr2O4 layer.15,16 In this particular com-

pound, a modulated structure is super-imposed in the Bi2O2 plane, giving an ex-tremely complicated structure.

Figure 3a shows the resistivity ρ of the three Co oxides along the in-planedirection. Note that the misfit oxides ex-hibit substantial in-plane anisotropy dueto the hexagonal lattice being distorted bythe square NaCl-type layer (the blocklayer), and only the b-axis data are shownhere. The most conducting compound isNaxCoO2, where a low ρ of 200 µ cm at300 K decreases in a metallic-like fashionwith decreasing temperature down to1.5 K.1 This is as conductive as the super-conducting Cu oxides, meaning that thelayered Co oxide is one of the most con-ductive layered oxides. One thing to noteis that there is no indication of localizationat low temperatures, which is quite un-usual in a quasi two-dimensional conduc-tor. Furthermore, the Na layer is highlydisordered by randomly distributed va-cant sites, which could act as a strong scat-terer for the conduction electrons. Thus,the conduction electrons in the CoO2 layerseem to be free from the disorder in the Nalayer, like the “charge confinement” in thehigh-temperature superconducting Cu

oxides. The other two compounds show afairly large ρ of 1–10 m cm at 300 K, withan upturn below 50 K.17,18 This upturn isdue to a pseudogap gradually opening inthe density of states at low temperatures.19

Figure 3b shows the Seebeck coefficientα of the three Co oxides along the in-planedirection (the b-axis data for the misfitoxides). The magnitude of α is as large as100–150 µV K1 at 300 K, comparable withthat of conventional TE semiconductors.Koshibae and co-workers20 have proposedthat the low-spin-state Co4 ion brings alarge entropy of kB ln 6 in the backgroundof the low-spin state Co3 ion of zeroentropy. The evaluated α in the high-temperature limit is kB log 6/e 150 µVK1, which satisfactorily agrees with the experiment. At low temperatures, the αvalues are essentially proportional totemperature, which reminds us of a mass-enhanced metal, such as a valence-fluctuation or heavy-fermion material. Infact, thermodynamic quantities such asspecific heat and susceptibility are verysimilar to those of the valence-fluctuationmaterial CePd3.20

Figure 3c shows the thermal conductiv-ity κ of the three Co oxides along the in-plane direction (the b-axis data for themisfit oxides).5 Clearly, a more compli-cated block layer shows lower κ, implyingthat is predominantly determined by the

Figure 1. Design of new functional oxides according to the concept of nanoblock integration.

Figure 2. Schematic illustration of the crystal structure of CoO2-based TE oxides:(a) NaxCoO2, (b) a Ca-based Co oxide known as Ca3Co4O9, and (c) the Bi-based Co oxideBi2Sr2Co2Oy.

Figure 3. (a) Temperature-dependenceof in-plane resistivity ρ, (b) Seebeckcoefficient α, and (c) thermalconductivity κ of the three Co oxidesshown in Figure 2.



block layer. Thus far, high κ was expectedfor oxides, because oxygen is a light atom.However, this is not always true: in thelayered Co oxides, the lattice thermal con-ductivity is determined by cations in theblock layer rather than by oxygen anions.

The layered structure works quite wellin two ways. The first is that the electriccurrent and the thermal current flow indifferent paths in space.5,21 This enables usto control the lattice thermal conductivityby properly choosing the insulating blocklayer. This is a manifestation of so-called“phonon-glass/electron-crystal” behavior,22

and we call this type of material design“nanoblock integration.” The other way isthat the CdI2-type CoO2 block favors thelow spin state of Co3 and Co4, whichholds the large α at high temperatures. Inthe case of the perovskite cobalt oxideR1–xMxCoO3 (R rare-earth, M alkaline-earth), the high spin state of Co3 seri-ously suppresses α.23 These favorableconditions rarely occur in other transition-metal oxides, most of which show poor TEproperties. One exception is the layeredRh oxides, where a nanoblock integrationworks well.24,25

n-Type Oxides:Thermoelectric Performance ofHeavily Doped n-Type SrTiO3

Strontium titanate (STO), heavily dopedwith donor dopants, has recently beenfound to show fairly good TE properties atroom temperature.26 STO has an isotropiccubic crystal structure (perovskite-type)and is known to have an electronic bandstructure in which the conduction band ismade of the Ti 3d orbitals.27 Band degener-acy Nc is 6 including the spin degeneracywith the bandgap energy of 3.2 eV. 26

Possibly due to the d-band nature, the ef-fective mass m* of carrier electrons is quitelarge; m* (1.1610)m0, where m0 is thefree electron mass.26,28

The values for m* estimated from thetemperature-dependence of α were 6m0and 7m0 for La-doped STO and Nb-doped STO, respectively.29 This large effec-tive mass gives rise to a large α 0.1 mVK–1 at 300 K and increases monotonicallywith increasing temperature. La- and Nb-doped (1020–1021 cm3) STO single crys-tals behave as degenerate semiconductors,and their electrical conductivity σ becomesas high as 103 S cm1 at 300 K. It decreasesgradually with increasing temperature,which is consistent with the temperature-dependence of Hall mobility µ (10 cm2

V1 s1 at 300 K), but it decreases propor-tionally to T2.0 below 750 K and T1.5

above 750 K. The temperature-depend-ence of Hall mobility indicates that the car-riers are scattered by polar optical phonons

below 750 K, while acoustic phonon scatter-ing becomes predominant above 750 K.29

It should be noted that the power factorα2σ of the heavily donor-doped STOreaches 2 103 W m1 K2 at 300 K,which is comparable with that of conven-tional TE materials such as bismuth tel-luride. This fact strongly indicates that theelectron system of STO has high potentialfor TE performance.

Accordingly, we have attempted toclarify the practically achievable maxi-mum thermoelectric figure of merit ZTof STO by carefully evaluating the transport properties of the single crystaland epitaxially grown thin-film samplesof good quality.30 Figure 4 shows thetemperature-dependence of ZT for La-and Nb-doped STO.31 It can clearly beseen that the maximum ZT 0.37 is ob-tained from 20% Nb-doped STO (Nb con-centration 4.0 1021 cm3) at 1000 K.Actually, Nb doping gives rise to larger ef-fective mass, m*, of carrier electrons thanLa doping, because Nb with a larger ionicradius than Ti expands the lattice and en-hances the carrier localization, while Ladoping leads to a smaller effective mass ofelectrons. The large effective mass keeps values from greatly decreasing, even athigh carrier concentration, and this factmakes Nb-doped STO a good base mate-rial for TE application.

However, the maximum ZT value of0.37 is still too low compared with state-of-the-art materials whose ZT exceeds 1.0 or even 2.0.32 The factor most respon-sible for the low ZT of STO is its high κ,10 W m1 K1 at 300 K, which cannotbecome lower than 3 W m1 K1 at1000 K, even though it decreases with in-creasing temperature.29

Muta et al. doped STO with variousrare-earth ions substituting for strontiumsites and found that Dy is the most effec-tive in reducing κ down to 3.4 W m1 K1

at 300 K and in increasing the figure ofmerit up to 3.84 104 K1 at 573 K.33

They further tried to reduce κ throughsubstituting Ba for Sr and found that theintermediate composition gives the lowestκ of 3.3 W m1 K1 at 300 K and thehighest figure of merit of 3.0 104 K–1

at 400–600 K.34 However, conventionalsolid-solution approaches do not appearto be very effective in suppressing thethermal conduction in STO.

Natural Superlattices:Hybrid Crystals

Based on the concept of nanoblock integration, we are interested in theRuddlesden–Popper phases, Srn1TinO3n1[SrO(SrTiO3)n, n integer] that exist as ahomologous series between SrTiO3 andSrO.35 These phases have layered perovskitestructures in which SrO and (SrTiO3)n layersare alternately stacked periodically (Fig-ure 5).35,36 These oxides can be regarded asnatural superlattices, and therefore sublat-tice interfaces are expected to change thephonon behavior to suppress thermal con-duction and improve the TE performance.

In our preliminary experiment, Nb-doped STO polycrystal and STO-327 poly-crystal (n 2, Sr3Ti2O7) that possess similarcarrier densities were employed for themeasurement to compare their TE proper-ties.37 It is clearly seen in Table I that σ ofSTO-327 is slightly larger than that of STO,while α is slightly smaller. This finding isconsidered to be reasonable, if it is takeninto consideration that the effective massof carrier electrons in STO-327 is lowerthan that in STO.38

The κ value of STO-327 is about halfthat of STO. This finding firmly suggeststhat the formation of a hybrid crystal with

Figure 4. Dimensionless thermoelectricfigure of merit ZT versus temperaturefor La-doped STO and Nb-doped STO;dashed curves were estimated fromexperimental data for lightly doped STOsingle crystals.31

Figure 5. Crystal structures of theRuddlesden–Popper homologousseries.36 Each octahedron consists ofTiO6.The solid black dots denote Srions.



Table I:Thermoelectric Properties of Polycrystals at 300 K.36

Electrical ThermalRelative Hall Conductivity, Seebeck Conductivity, Figure of ThermoelectricDensity n Mobility, Coefficient, Merit, Z Figure of

Polycrystal (%) (cm3) (cm2 V1 s1) (S cm1) (V K1) (W m1 K1) (K1) Merit, ZT

Nb-doped STO 94 9.1 1020 1.2 170 97 7.3 2.2 105 7 103

STO-327 92 9.0 1020 1.4 198 86 3.7 4.0 105 1.2 102

a layered structure leads to enhanced scat-tering of phonons that are responsible forthermal conduction. As shown in Figure 6,the κ values of STO-327, STO, and SrO alldecrease with increasing temperature(proportional to T1), indicating that thephonon–phonon scattering is dominantfor thermal conduction in all cases.39 A re-duction in the κ value of STO-327 wouldhave been caused by suppression of themean free path of phonons possibly asso-ciated with the formation of sublatticeinterfaces. This should be further verifiedby a more careful experiment employinghigh-quality single-crystal samples.

Thermoelectric Oxide ModulesConstructing Good Oxide/MetalJunctions41

Bulk materials of p-type Ca2.7Bi0.3Co4O9(Co-349) and n-type La0.9Bi0.1NiO3 (Ni-113)were prepared using a hot-pressing tech-nique in order to align the Co-349 grainsand to densify. Both p- and n-type hot-pressed plates were cut to provide a legelement area of 3.7 mm 4.5 mm and aleg element length of 4.7 mm. The lengthdirection was made perpendicular to the

hot-pressing axis for the purpose of in-ducing a temperature gradient.

An alumina plate 5.0 mm wide, 8.0 mmlong, and 1.0 mm thick was used as a sub-strate. Ag paste was applied to one side ofthe alumina plate and solidified by heat-ing to achieve electrical conduction. Agpaste was mixed with p-type oxide pow-der for connection of the p-leg and withn-type oxide powder for connection of then-leg. The compositions of the oxide pow-ders used were identical to those of thep- and n-legs. The same precursor pow-ders were pulverized by ball milling to ob-tain a grain size smaller than 10 m andthen were mixed in varying ratios withthe Ag paste (0 wt%, 1.5 wt%, 6 wt%, or10 wt% oxide/Ag paste). After connectionof the legs and substrate, the wet Ag pastewas dried at 373 K and solidified by heat-ing at 1073 K in air.

Figure 7 shows the temperature-dependence of internal resistance RI fordifferent versions of the unicouple con-structed using the Ag paste containingvarious weight percentages of oxide pow-ders. Since the difference in ρ values foreach leg among the samples is 10%, theincorporation of the oxide powders is seento be effective in reducing RI. This is due tothe reduction of contact resistance Rc be-tween the oxide leg elements and the Agpaste. Considering the resistance of Ag onthe substrate, Rc at 950 K accounts for 12.5%and 56.8% of total RI for the unicouples

connected using the 6 wt% and 0 wt%oxide/Ag pastes, respectively. An increasein RI, however, is observed when the10 wt% oxide/Ag paste is used.

Output power Pmax of the unicouple with6 wt% oxide/Ag paste increases with hot-side temperature TH and reaches 177 mWat TH 1073 K and a temperature differ-ence T between hot- and cold-side tem-peratures of 500 K. One of the strongpoints of unicouples fabricated in thismanner is their high power density, whichwould allow a large amount of electricalpower to be generated by a small, light-weight TE module. In the same oxide TEunicouple, the volume power density is0.96 W cm3 at TH 1073 K.

Fabrication and Properties of a 140-Couple Module42

The p-type Co-349 and n-type Ni-113bulks were prepared by hot-pressing, asmentioned previously. The resulting hot-pressed plates were cut into p- and n-typelegs with a height of 5.0 mm and a cross-sectional area of 1.3 mm 3 mm. The legs(140 couples) were inserted alternatelyinto the Al2O3 meshes for electrical contactin series. An Ag paste containing 6 wt% ofthe Co-349 powder was used to reduce theRc and the differential of thermal expan-sion between the oxide legs and the Agelectrodes. The Ag paste was solidified byannealing at 1123 K under uniaxial pres-sure of 6.4 MPa for 5 h in air. The dimen-sions of this module are 53 mm long,32 mm wide, and 5.0 mm thick.

The power-generating property of themodule at TH 1072 K and a T 551 Kis shown in Figure 8a. Open-circuit volt-age Vo, which corresponds to the interceptof current (I) versus voltage (V) lines,reaches 4.5 V. In the fabrication of this TEmodule, a critical issue is how to makeelectrodes that will provide low Rc be-tween the oxide and Ag materials as wellas a mechanically strong connection. Inthe module presented here, this concernhas been overcome through the use of aAg paste containing oxide powder to con-nect the legs and Ag electrodes.

A lithium-ion battery mounted in a mo-bile phone was charged by a 140-couple

Figure 6.Thermal conductivity versustemperature for STO-327 [SrO(SrTiO3)2]compared with STO39 (SrTiO3) andSrO.40

Figure 7.Temperature-dependence ofinternal resistance in oxide unicouples.



module using a small burner and pan ofwater (Figure 9). This phenomenon indi-cates that applications of oxide modulescan extend not only to generators for therecovery of high-temperature waste heatemitted by automobiles, factories, and simi-lar sources, but also to portable generators

for charging portable phones, personalcomputers, and other electronic devices.

SummaryIt has been eight years since the first re-

port of the thermoelectric properties of theTE oxide material NaxCoO2. Finally, TEmodules consisting of oxide legs possessinggood electrical and mechanical propertieshave been fabricated. Unfortunately, theirconversion efficiency is lower than conven-tional TE modules. In order to enhance thethermoelectric properties of oxide modules,new materials with higher ZT values, espe-cially n-type materials, are indispensable.The discovery of such oxides according tothe concept of nanoblock integration isearnestly desired.

References1. I. Terasaki, Y. Sasago, and K. Uchinokura,Phys. Rev. B 56 (1997) p. R12685.2. R. Funahashi, I. Matsubara, H. Ikuta, T.Takeuchi, U. Mizutani, and S. Sodeoka, Jpn.J. Appl. Phys. Pt. 2 39 (2000) p. L1127.3. A.C. Masset, C. Michel, A. Maignan, M.Hervieu, O. Toulemonde, F. Studer, B. Raveau,and J. Hejtmanek, Phys. Rev. B 62 (2000) p. 166.4. Y. Miyazaki, K. Kudo, M. Akoshima, Y. Ono,Y. Koike, and T. Kajitani, Jpn. J. Appl. Phys. Pt. 239 (2000) p. L531.5. A. Satake, H. Tanaka, T. Ohkawa, T. Fujii,and I. Terasaki, J. Appl. Phys. 96 (2004) p. 931.6. M. Shikano and R. Funahashi, Appl. Phys.Lett. 82 (2003) p. 1851.7. I. Matsubara, R. Funahashi, T. Takeuchi, S.Sodeoka, T. Shimizu, and K. Ueno, Appl. Phys.Lett. 78 (2001) p. 3627.8. W. Shin, N. Murayama, K. Ikeda, and S.Sago, J. Power Sources 103 (2001) p. 80.9. C. Fouassier, G. Matejka, J. Reau, and P.J. Hagenmuller, J. Solid State Chem. 6 (1973) p. 532.10. Y. Ono, R. Ishikawa, Y. Miyazaki, Y. Ishii, Y.Morii, and T. Kajitani, J. Solid State Chem. 166(2002) p. 177.11. M. Mikami, M. Yoshimura, Y. Mori, T.Sasaki, R. Funahashi, and I. Matsubara, Jpn.J. Appl. Phys. Part 2 41 (2002) p. L777.12. K. Takada, H. Sakurai, E. Takayama-Muromachi, F. Izumi, R.A. Dilanian, and T.Sasaki, Nature 422 (2003) p. 53.13. M. von Jansen and R. Hoppe, Z. Anorg.Allg. Chem. 408 (1974) p. 104.14. J.M. Tarascon, R. Ramesh, P. Barboux, M.S.Hedge, G.W. Hull, L.H. Greene, M. Giroud, Y.LePage, W.R. McKinnon, J.V. Waszczak, andL.F. Schneemeyer, Solid State Commun. 71 (1989)p. 663.15. H. Leligny, D. Grebille, O. Pérez, A.-C. Masset, M. Hervieu, C. Michel, and B. Raveau, C.R. Acad. Sci. Paris t.2, Série IIc(1999) p. 409.16. T. Yamamoto, I. Tsukada, K. Uchinokura,M. Takagi, T. Tsubone, M. Ichihara, and

K. Kobayashi, Jpn. J. Appl. Phys. Pt. 2 39 (2000)p. L747.17. T. Fujii, I. Terasaki, T. Watanabe, and A.Matsuda, Jpn. J. Appl. Phys. Pt. 2 41 (2002)p. L783.18. I. Terasaki, Frontiers in Magnetic Materials(2005) p. 327.19. T. Itoh and I. Terasaki, Jpn. J. Appl. Phys. Pt.1 39 (2000) p. 6658.20. I. Terasaki, Proc. 18th Int. Conf. Thermoelectrics(ICT’99) (IEEE, Piscataway, NJ, 1999) p. 569.21. K. Takahata, Y. Iguchi, D. Tanaka, T. Itoh, and I. Terasaki, Phys. Rev. B 61 (2000)p. 12551.22. G.A. Slack, CRC Handbook of Thermoelectrics(CRC Press, Boca Raton, FL, 1995) p. 407.23. H. Masuda, T. Fujita, T. Miyashita, M. Soda,Y. Yasui, Y. Kobayashi, and M. Sato, J. Phys. Soc.Jpn. 72 (2003) p. 873.24. S. Okada and I. Terasaki, Jpn. J. Appl. Phys.44 (2005) p. 1834.25. S. Okada, I. Terasaki, H. Okabe, and M. Matoba, J. Phys. Soc. Jpn. 74 (2005) p. 1525.26. T. Okuda, K. Nakanishi, S. Miyasaka, and Y.Tokura, Phys. Rev. B 63 113104 (2001).27. L.F. Matheiss, Phys. Rev. B 6 (1972) p. 4718.28. H.P.R. Frederikse, W.S. Thurber, and W.R.Hosler, Phys. Rev. 134 (2A) (1964) p. A442.29. S. Ohta, T. Nomura, H. Ohta, and K.Koumoto, J. Appl. Phys. 97 (2005) p. 34106.30. S. Ohta, T. Nomura, H. Ohta, M. Hirano, H.Hosono, and K. Koumoto, Appl. Phys. Lett. 87(2005) p. 92108.31. S. Ohta, T. Nomura, H. Ohta, M. Hirano, H.Hosono, and K. Koumoto, in Extended AbstractsNo. 1 of the 52nd Spring Meeting, Jpn. Soc. Appl.Phys. Related Soc. (2005) p. 254.32. F.J. DiSalvo, Science 285 (1999) p. 703.33. H. Muta, K. Kurosaki, and S. Yamanaka,J. Alloys Compd. 350 (2003) p. 292.34. H. Muta, K. Kurosaki, and S. Yamanaka,J. Alloys Compd. 368 (2004) p. 22.35. S.N. Ruddlesden and P. Popper, Acta Crys-tallogr. 10 (1957) p. 38; S.N. Ruddlesden and P.Popper, Acta Crystallogr. 11 (1958) p. 54.36. J.H. Haeni, C.D. Theis, D.G. Schlom, W.Tian, X.Q. Pan, H. Chang, I. Takeuchi, and X.-D.Xiang, Appl. Phys. Lett. 78 (2001) p. 3292.37. K. Koumoto, S. Ohta, and H. Ohta, in Proc.23rd Int. Conf. Thermoelectrics (IEEE, Piscataway,NJ, 2005).38. W. Wunderlich, S. Ohta, H. Ohta, and K.Koumoto, in Proc. 24th Int. Conf. Thermoelectrics(IEEE, Piscataway, NJ, 2005) p. 237.39. K. Kato, S. Ohta, H. Ohta, and K. Koumoto,in Abstracts of the 43rd Symp. on Basic Science ofCeramics (2005) p. 20.40. H. Szelagowski, I. Arvanitidis, and S.Seetharaman, J. Appl. Phys. 85 (1999) p. 1.41. R. Funahashi, S. Urata, K. Mizuno, T. Kou-uchi, and M. Mikami, Appl. Phys. Lett. 85 (2004)p. 1036.42. R. Funahashi, T. Mihara, M. Mikami, S. Urata, and N. Ando, in Proc. 24th Int. Conf. Thermoelectrics (IEEE, Piscataway, NJ, 2005) p. 292.

Figure 9. Demonstration of thermoelectriccharging of a portable phone.

Figure 8. (a) Power-generating propertyof the oxide module shown in (b).

MRS Materials Connectionsw w w. m r s . o r g / c o n n e c t i o n s /

Your source for Materials Research-Related Information!

Research news, materials data sources,materials Web site database, meetings calendar, and more. Check it out today!


IntroductionIdeas in using superlattices to improve

the thermoelectric figure of merit (ZT)through the enhancement of electronicconductivity and reduction of phononthermal conductivity were first discussedin a workshop by M.S. Dresselhaus, T. Harman, and R. Venkatasubramanian.1Subsequent publications from Dresselhaus’sgroup on the quantum size effects on elec-trons drew wide attention and inspired in-tense research, both theoretical andexperimental, on the thermoelectric prop-erties of quantum wells and superlattices.2Several groups reported in recent yearsenhanced ZT in various superlattices such as Bi2Te3/Sb2Te3 and Bi2Te3/Bi2Se3,3 and PbSeTe/PbTe quantum dotsuperlattices4 (Figure 1). The large im-provements observed in these materialssystems compared with their parent mate-rials are of great importance for both fun-

damental understanding and practical applications.

Superlattices are anisotropic. Differentmechanisms to improve ZT along direc-tions both parallel (in-plane) and perpen-dicular (cross-plane) to the film planehave been explored. Along the in-planedirection, potential mechanisms to in-crease ZT include quantum size effectsthat improve the electron performance bytaking advantage of sharp features in theelectron density of states,2 and reduction ofphonon thermal conductivity throughinterface scattering.5 Along the cross-planedirection, one key idea is to use interfacesfor reflecting phonons while transmit-ting electrons (phonon-blocking/electron-transmitting),6 together with other mecha-nisms, such as electron energy filtering7

and thermionic emission,8 to improveelectron performance. These mechanisms

have been explored through a few superlattice systems whose constituentmaterials have reasonably good thermo-electric properties to start with, V–VI ma-terials such as Bi2Te3/Sb3Te3,3,9 IV–VImaterials such as PbTe/PbSe,4,9 and V–Vmaterials such as Si/Ge10–12 and Bi/Sb,13

with the most impressive results obtainedin Bi2Te3 superlattices3 and PbTe-basedquantum dot superlattices.4

The large ZT improvements observedin these superlattices shattered the ZT 1ceiling that persisted until the 1990s,opening new potential applications incooling and power generation using solid-state devices. Much research is needed inmaterials, understanding, and devices tofurther advance superlattice thermoelec-tric technology. In this short article, wewill give a summary of the past work, em-phasizing the materials aspects of super-lattices, while commenting on currentunderstanding or lack of it, and some as-pects of the device research. We refer toother review articles for more in-depthdiscussions on these topics.5,6,14–19

Materials and PropertiesThe work on quantum well and

superlattice-based thermoelectric mate-rials mostly focused on perfect (i.e., epi-taxial) layer systems. So it was notsurprising that, due to the extensiveworldwide experience in IV–VI epi-taxy,20,21 approaches were taken to use thismaterial system to prove the quantumconfinement as well as the acousticphonon scattering.4,9 As Bi2Te3-based ma-terials have the highest ZT around roomtemperature, successful efforts werestarted to develop suitable epitaxial sys-tems for the V–VI compound family.9,22 It is worth mentioning that both IV–VI andV–VI semiconductor material familieshave a useful structural relationship (Fig-ures 2a and 2b).23 The current thin-film de-vice technologies for IV–VI and V–VIcompounds use either one or the other ofthese two material systems. Mixed stag-gered IV–V/V–VI superlattice thin-filmdevices are not known so far.

V–VI SuperlatticesVenkatasubramanian and co-workers

reported Bi2Te3-based superlattices grownby metallorganic chemical vapor deposi-tion (MOCVD) on GaAs substrates.22 TheGaAs substrates were chosen for theirease of cleaning prior to epitaxial deposi-tion and the fact that substrates with 2–4misorientation with respect to ⟨100⟩ can beconveniently obtained. It is important tonote that these trigonal-structured Bi2Te3materials are grown on GaAs with fccstructure. The misorientation allows the


Aspects of Thin-Film SuperlatticeThermoelectricMaterials,Devices, andApplications

Harald Böttner, Gang Chen, and Rama Venkatasubramanian

AbstractSuperlattices consist of alternating thin layers of different materials stacked

periodically.The lattice mismatch and electronic potential differences at the interfacesand resulting phonon and electron interface scattering and band structure modificationscan be exploited to reduce phonon heat conduction while maintaining or enhancing theelectron transport.This article focuses on a range of materials used in superlattice formto improve the thermoelectric figure of merit.

Keywords: thermal conductivity, thermoelectricity.


initiation of the epitaxial process at thekink sites on the surface, thereby allowingthe growth of mismatched materials. Thegrowth of Bi2Te3-based materials, with therather weak van der Waals bonds alongthe growth direction, requires a low-temperature process. A low-temperaturegrowth process leads to high-quality,abrupt superlattice interfaces with mini-mal interlayer mixing, and also allows thegrowth of highly lattice-mismatched materials systems without strain-inducedthree-dimensional islanding. Figure 3shows a high-resolution transmissionelectron micrograph of a Bi2Te3/Sb2Te3 su-

perlattice on a GaAs substrate, delineatingthe two very different crystalline orienta-tions. In situ ellipsometry has been used togain further nanometer-scale control overdeposition.24

As V–VI epitaxy is rather a scientific“virgin soil,” it is not surprising that evenfor the single homogeneous V–VI layersof the central compound Bi2Te3, only mini-mal information regarding thin-film dep-osition and thermoelectric propertycharacterization can be found. The prob-lem of a low Te sticking coefficient is dis-cussed in Reference 25. Different filmgrowth methods based on MBE,9,25–27

MOCVD,22 flash evaporation,28–30 and co-evaporation31,32 have been used to growsingle layers and superlattices on varioussubstrates. Nurnus et al.23 used a ratherhigh deposition temperature comparedwith that used in MOCVD but were stillable to obtain high-quality (Bi2Te3)/Bi2(Te,Se)3 superlattices using elementsources. The power factor (PF α2/ρκ) of50 µW cm−1 K−2 reported in Reference 33for Bi2Te3 is close to that of the best singlecrystals, which is 57 µW cm−1 K−2.34 Unfortu-nately, the mobility is limited to 150 cm2

V−1 s−1. A critical item in maintaining theoutstanding ZT of superlattices is theirstability against cation (p-material) andanion (n-material) interdiffusion. Resultsreported by Nurnus et al.23,33 strongly in-dicate a dependence of the superlatticestability against diffusion on perfection ofthe layer structure.

Recently, sputtering as a new deposi-tion method for forming V–VI superlat-tices35 was tested. Starting with alternatingelement layers, the corresponding super-lattice thermoelectric compounds wereformed by a subsequent annealing proce-dure. Here, at lower temperatures, the an-ions in the n-(“Se/Te”) alloy system or thecations in the p-(“Bi/Sb”) alloy systemtend to interdiffuse while compoundingthe thermoelectric material. Taking intoaccount the results by Johnson,36 who suc-ceeded in forming superlattices in theV–VI materials system using “modulatedelemental reactants,” it can be concludedthat besides alternating layers, a necessarycondition for the formation of superlat-tices is to deposit layers that are as per-fectly oriented as possible in order toobtain optimum diffusion stability for reli-able final devices. For the case of perfect c-oriented layers, the fast diffusion paths are


Aspects of Thin-Film Superlattice Thermoelectric Materials, Devices, and Applications

Figure 1.Thermoelectric figure of merit ZT for Bi2Te3 /Sb2Te3 superlattices3 (SL),PbSnSeTe/PbTe quantum dot superlattices (QDSL), and PbTeSe/PbTe quantum dotsuperlattices.4,45

Figure 2. (a) Crystallographic data for IV–VI and V–VI compounds, highlighting the structural relationship between both material systems.(b) Epitaxial map of semiconductor materials (at room temperature) that could be suitable in combination with Bi2Te3, based on their relevantatomic distance with respect to the a-plane of Bi2Te3.

blocked, even in the case of the non-epitaxially arranged layers, according toJohnson’s work.36 If the interdiffusion isblocked in the a-plane, superlattices willbe stable until the intrinsic interdiffusionin the c-direction is activated at signifi-cantly higher temperatures. For the V–VIcompounds, it is well known that the in-terdiffusion coefficients in the c-directionare normally smaller, by decades, than inthe a-direction.37

The ZT values of V–VI-based superlat-tices can be measured in either in-plane orcross-plane directions. So far, the largestenhancement is in the cross-plane direc-tion, with the major gain coming from thethermal conductivity reduction. Venkata-subramanian reported a cross-plane ZT 2.4 at room temperature for the p-typeBi2Te3-Sb2Te3 superlattices with a period of6 nm.4 In such superlattices, the elec-tronic power factor is in the range of 40µW cm–1 K–2 to 60 µW cm–1 K–2, compara-ble to or higher than standard bulk p-typeBi2Te3 solid-solution alloys measuredalong the a–b axis. However, the phononthermal conductivity (κp) dropped to 0.22W/m K, about a factor of 5 lower than thatof bulk alloys along a–b axis. Lambrechtet al.38 determined an in-plane reductionof the total thermal conductivity (electronsand phonons) down to 65%, comparedwith homogenous Bi2Te3 for n-typeBi2Te3/Bi2(Se0.12Te0.88)3 superlattices with a10-nm period.

IV–VI SuperlatticesIV–VI nanolayers have been success-

fully grown for more than a decade forIV–VI infrared lasers.39 Because of itsphysical and chemical properties, theIV–VI materials system is relatively easyto handle, compared with the V–VI com-pounds, particularly for epitaxial growth.

Thus, a lot of literature can be found onthe growth details and layer properties ofIV–VI nanolayer stacks.40,41 Here, we sum-marize results exclusively focused onthermoelectric applications.

The initial effort in IV–VI systems wasfocused on electron confinement effects.Using MBE-grown PbTe/EuxPb1–xTe, Har-man and co-workers showed an increasedelectron power factor (α2σ) inside thequantum wells along the in-plane direc-tion,42,43 as predicted by Hicks and Dresselhaus. However, the barriers inmultiple quantum wells (MQWs) degradethe overall ZT, because they conduct heatwithout contributing to electron perform-ance. Harman and co-workers further ex-plored various IV–VI superlattices. Theyfound that in PbTe/Te superlattices, ob-tained by the addition of a few nanome-ters of Te above the PbTe layer, ZTincreased from 0.37 to 0.52 at room tem-perature, and this increase was associatedwith the formation of quantum dot struc-tures at the interface.44 The Harman groupfurther discovered experimentally thatquantum dot superlattices based onPbTe/PbSexTe1–x (with x 0.98) have aneven higher ZT.45 The quantum dot for-mation is due to the lattice mismatch be-tween PbTe and PbSeTe40 (Figure 4).PbTe-based quantum dot superlatticeswith a total thickness of 100–200 µm havebeen grown with good thermoelectricproperties along the in-plane direction.4PbTe/PbSeTe n-type quantum dot super-lattices were obtained by Bi doping, and p-type quantum dot superlattices wereobtained through Na doping. The bestbulk PbTe-based alloys have a room tem-perature ZT of 0.4. Harman et al.4 re-ported n-type PbTe/PbSeTe quantum dotsuperlattices with ZT 1.6 at room tem-perature, compared with ZT 0.4 in the

best bulk PbTe alloys, and inferred thatquaternary superlattices based on PbTe/PbSnSeTe have a room temperature ZT of2. For the ternary PbTe/PbSeTe super-lattices, the factor of 4 increases comemainly from a large reduction in the thermalconductivity, while the power factor re-mains similar to that of the bulk, albeit atdifferent optimum carrier concentrations.The combined electron and phonon ther-mal conductivity (κe κp) drops frombulk values of 2.5 W/m K to 0.5 W/m K.Considering that the electronic contribu-tion to thermal conductivity for both superlattices and bulk materials is 0.3W/m K, a significant phonon thermalconductivity reduction is obvious.

Böttner and co-workers studied the in-plane thermoelectric properties of n- andp-doped PbTe/PbSe0.20Te0.80 systems.9,46

Compared with corresponding bulk Pb-SexTe1–x material, a significant reduction inthe thermal conductivity parallel to thegrowth direction was measured. Togetherwith nearly unchanged power factors, anin-plane ZT enhancement of up to 40% at



Figure 3. (a) Transmission electron micrograph of a Bi2Te3/Sb2Te3 (10 Å/50 Å) superlattice. (b) Image contrast oscillations through thesuperlattice. (c) Fast Fourier transform of the image in (a), showing the superlattice reflections of order n.22

Figure 4. Quantum dot structures in thePbTe/PbSeTe system.42

a temperature of 500 K was estimated. Re-cently, Caylor et al.47 reported their effortin growing PbTe/PbSe superlattices.

Other Superlattice SystemsNurnus and co-workers studied

IV–VI/V–VI heteroepitaxial layers toevaluate quantum confinement in V–VIlayers using suitable “wide-bandgap”IV–VI alloys such as Bi2Te3/Pb1–xSrxTe,33,48

including results involving stabilityagainst annealing. The practical use of theconcept (Figure 2) of a structural relation-ship between IV–VI and V–VI com-pounds was recently proved by Cayloret al., who deposited Bi2Te3 on a standardGaAs substrate as a buffer layer, followedby a IV–VI superlattice.47 They found out,surprisingly, that superlattices in (111) and(100) orientations grow simultaneously atlower temperatures.

Other superlattice systems have beenstudied for their thermoelectric proper-ties, such as Si/Ge superlattices,10–12 Bi/Sb superlattices,13 and skutterudite-based su-perlattices.49 Si/Ge and Si/SiGe alloy su-perlattices have shown a large reductionin thermal conductivity compared withthat of homogeneous alloys in the cross-plane direction,10,11 while in the in-planedirection, thermal conductivity values arecomparable with that of the homogeneousalloy with equivalent composition to thesuperlattices.50 Despite the reduction inthermal conductivity, there are no conclu-sive results on the figure of merit becauseof difficulties in measuring the thermo-electric properties of very thin films. Reported measurements of the thermo-electric properties of Bi/Sb and skutteru-dite superlattices are scarce and notconclusive.51,52

Characterization of ThermoelectricProperties

Thermoelectric property measurementsin many cases have been the bottleneck inthe development and understanding ofsuperlattice-based materials.53 Because ofanisotropy, all thermoelectric properties,including the Seebeck coefficient α, electri-cal conductivity σ, and thermal conduc-tivity κ, should be measured in the samedirection and, ideally, on the same sample.Along the in-plane direction, thermal con-ductivity is usually the most difficult parameter to measure. However, the sub-strate and the buffer layers can also easilyoverwhelm the Seebeck coefficient andelectrical conductivity measurements. Theneed to isolate the properties of the filmfrom those of the substrate and the bufferlayer often influences the choice of thesubstrate and the film thickness in thegrowth of superlattices. In the cross-plane

direction, the 3ω method and the pump-and-probe method are often used to measure the thermal conductivity of su-perlattices.54,55 However, measuring theSeebeck coefficient and the electrical con-ductivity in the cross-plane direction canbe even more challenging. Venkatasubra-manian et al.3 adapted the transmissionline model (TLM) technique used for themeasurement of specific electrical contactresistivities (ρc) to determine the cross-plane electrical resistivities in Bi2Te3-basedsuperlattices, which is feasible when ρc issmaller than the specific internal resis-tance of the thermoelectric film d/σ,where d is the thickness of the superlatticefilm.

Besides individual property measure-ments to determine ZT, other methods fordirect ZT determination have been suc-cessful. Venkatasubramanian et al.3adapted the Harman method56 to deter-mine ZT in the cross-plane direction ofBi2Te3-based superlattices with a maxi-mum thickness of the superlattice up to 5µm. Yet the most unambiguous measure-ment of the enhanced ZT comes from di-rect measurements of the cooling effect.Harman used this method to characterizethe performance of his PbTe/PbSeTequantum dot superlattices.4 He measuredcooling based on a thermocouple with oneleg made of the superlattice and the otherleg made of a section of Au wire, properlymatched in length. The maximum coolingmeasured from such a thermocouple was43.7 K, while a similar couple made of thebest bulk thermoelectric material onlyreached 30.8 K. To make such a couple, thetotal thickness of his superlattice samplewas 100 µm.

Current UnderstandingExperimental results so far have shown

that the thermal conductivity reductionwas mainly responsible for ZT enhance-ment in the superlattices. Theoretical stud-ies on the thermal conductivity have beencarried out.16,17 These models generally fallinto two different camps.

The first group treats phonons as inco-herent particles and considers interfacescattering as the classical size effect that is analogous to the Casimir limit at lowtemperatures in bulk materials andFuchs–Sonderheim treatment of electrontransport.57–59 These classical size effectmodels assume that interface scattering ispartially specular and partially diffuse,and can explain experimental data for su-perlattices in the thicker period limit.

The other group of models is based onthe modification of phonon modes in su-perlattices, considering the phonons as to-tally coherent.60,61 In superlattices, the

periodicity has three major effects on thephonon spectra: (1) phonon branches fold,owing to the new periodicity in thegrowth direction; (2) mini-bandgaps form;and (3) the acoustic phonons in the layerwith a frequency higher than that in theother layer become flat or confined be-cause of the mismatch in the spectrum.Comparison with experimental data,however, shows that the group velocityreduction alone is insufficient to explainthe magnitude of the thermal conductivityreduction perpendicular to the film plane,and it fails completely to explain the ther-mal conductivity reduction along the filmplane.61,62 The reason is that the lattice dy-namics model assumes phase coherenceof the phonons over the entire superlatticestructure and does not include the possi-bility of diffuse interface scattering, whichdestroys the perfect phase coherence pic-ture. Partially coherent phonon transportmodels can capture the trend of thermalconductivity variation in both the in-planeand the cross-plane directions over the en-tire thickness range.63,64 Molecular dynam-ics simulations considering interfacemixing can generate trends similar to thatobserved experimentally on GaAs/AlAssuperlattices, which is consistent with themodeling.65 Past models of thermal con-ductivity focused on III–V and IV–IV superlattices. There are no detailed mod-els on IV–VI and V–VI superlattices.Venkatasubramanian et al.66 observed aminimum in thermal conductivity forBi2Te3-based superlattices at a periodicthickness of 6 nm. Although similartrends can be obtained from partially co-herent phonon-transport models, the min-ima based on such models typically occuraround 3–5 monolayers (i.e., 1–2 nm).63,64

The discrepancy could be due to the un-usually large unit cell in the c-axis direc-tion and potentially to interface mixing,which is not well included in currentmodels, and to phonon localization.66 Themodeling conclusion that coherent statesof phonons cannot reproduce experimen-tal data has significant implications for materials synthesis, suggesting thatother nanostructures can lead to similarresults.67

While the thermal conductivity reduc-tion has been largely responsible for thereported high ZT so far in IV–VI andV–VI superlattices, the importance ofmaintaining the electronic power factorcannot be overemphasized. Although inboth these systems, the maximum powerfactors are close to those of their bulkcounterparts, the optimal dopant concen-trations between bulk and superlatticesamples differ, at least in PbTe-based sys-tems.44,45 The bandgaps of the constituent



materials in the IV–VI and V–VI superlat-tices with high ZT are similar, suggestingthat quantum size effects may not be im-portant. However, small band-edge off-sets can have an effect on the electronscattering mechanisms and shift the opti-mal carrier concentration. Although quan-tum size effects may not dominate in thesematerials systems, the principle of usingquantum size effects to improve electronperformance is sound. Minimizing inter-face scattering of electrons is crucial for realizing a high power factor. Better mate-rials synthesis can potentially lead tostructures that can take advantage of bothincreased electron performance and re-duced phonon thermal conductivity.

Devices and ApplicationsThe fabrication of superlattice-based

devices can take advantage of many of the standard tools of semiconductor device manufacturing, such as photoli-thography, electroplating, wafer dicing,and pick-and-place systems. This allowsscalability of the module fabrication, from simple modules that can pump mil-liwatts of heat to multiconnected modulearrays. Both in-plane and cross-plane de-vices are under development, and eachhave their unique advantages, challenges,and applications.

Cross-plane superlattice-based devicestypically have configurations similar tothose of bulk thermoelectric modules, al-beit with significantly shorter legs andsmaller leg cross sections.3,68 Such deviceshave extremely rapid cooling or heatingcharacteristics, and fully functional de-vices can be built using 1/40,000 of the active material required for state-of-the-art bulk thermoelectric technology.

Venkatasubramanian and co-workershave developed wafer-bonding technol-ogy to fabricate Bi2Te3 superlattice thermo-electric devices.68 It is clear from suchdevice development that significant chal-lenges exist in translating the intrinsicallyhigh ZT of the materials to the high per-formance of the devices. Some of the is-sues are related to the significant electricaland thermal parasitic resistances in amodular assembly. First, the specific elec-trical contact resistance ρc at both ends ofthe p-type and n-type devices must beminimized such that ρc is much smallerthan the specific resistance d/σ of the leg.Another significant challenge is the ther-mal management at both the hot and thecold sides, as the heat flux through eachleg can be as large as 1000 W/cm2. Sucha high heat flux cannot be handled withusual convective cooling techniques. Heatspreading by using sparsely spaced ele-ments or advanced thermal managementmethods is necessary.

Cross-plane superlattice-based devicesare being considered for a variety of appli-cations. Thermoelectric coolers have longbeen used for the wavelength stability ofsemiconductor lasers. Currently usedthermoelectric coolers are based on bulkmaterials machined down to small sizes(2 mm × 2 mm × 1 mm). Superlattice-based devices can better match the foot-print and heat flux of semiconductorlasers with a lower profile, which is ex-tremely important for fitting into the exist-ing packages such as metal transistoroutline cans. Superlattice thermoelectrictechnology is also now being actively con-sidered for the thermal management ofhot-spot and transistor off-state leakagecurrent in advanced microprocessors.

Besides cooling applications, superlattice-based thermoelectric devices can also beused for power-conversion applications.Figure 5 shows an example of local heat-ing and cooling that can be realized withsuperlattice-based devices. Early studiescarried out by Venkatasubramanian andco-workers of power-conversion effi-ciency using single p–n couples haveshown a significant correlation with themeasured ZT in the “inverted” p–n cou-ples69 by the Harman method.

In-plane device configurations are used mainly for sensors, and most of thepast work has been based on polycrys-talline thermoelectric material.70,71 Super-lattices with high ZT can improve theperformance of these devices. For sensorapplications, thermal bypass through the substrate must be minimized by removing the substrate, transferring thesuperlattice film to another low-thermal-conductance substrate, or depositing the film directly on a low-thermal-conductivity substrate.71

One big question regarding superlattice-based thermoelectric coolers and powergenerators is their stability and reliability.These devices operate under high heatand current fluxes, and both thermo- andelectromigration are of great concern. Atthis stage, only a little work has beendone. Venkatasubramanian’s group72 car-ried out initial power-cycle testing on rel-atively simple superlattice couples usingPb37Sn63 bonding for flip-chip attachment.No degradation in ∆T was observed after more than 100,000 power cycles, suggesting an intrinsic reliability in the su-perlattice material. However, the high-temperature reliability of superlatticematerials has not been studied.



Figure 5. Infrared image of a row of five Bi2Te3-based thermoelectric superlattice microcoolers. (a) Discrete heating; (b) discrete cooling.(c) Combined discrete cooling devices for larger-area cooling.The scale marker in (c) applies to all three images.

Summary and Research NeedsThe large figure-of-merit enhancements

observed in V–VI- and IV–VI-based su-perlattices and quantum dot superlatticeshave an impact on both fundamental un-derstanding and practical applications.For a long time, the maximum ZT for allbulk materials was limited to ZT 1, andas a consequence, applications have beenlimited to niche areas. Progress made insuperlattice-based thermoelectric mate-rials show that ZT 1 is not a theoreticallimit. With the availability of high-ZTmaterials, many new applications willemerge. The progress made also calls formore effort in materials development, the-oretical understanding, and device fabri-cation, concurrent with the pursuit ofpractical applications of these materials.

Materials-wise, research in both en-hancing ZT and reducing cost is needed.Practical thermoelectric devices need bothn-type and p-type materials with compa-rable figures of merit. So far, p-typeBi2Te3/Sb2Te3 superlattices have muchhigher ZT values than n-type Bi2Te3/Bi2Se3superlattices, while PbTe/PbSeTe-basedn-type and p-type quantum dot superlat-tices have comparable ZT values. Contin-uous improvements in ZT for differentmaterials in different temperature rangesare needed. In addition to reducing thephonon thermal conductivity, the princi-ple of increasing ZT through quantumconfinement of electrons should be ex-ploited, including the exploration of one-dimensional nanowires and nanowiresuperlattices.73 Further reductions in ther-mal conductivity may be possible inaperiodic superlattices. Similar effects thatlead to a reduction in phonon thermalconductivity may be observed in othernanostructures that are more amenable tomass production. In addition to materialsdevelopment, theoretical studies areneeded to further understand the electronand phonon thermoelectric transport. Particularly, quantitative tools capable ofpredicting thermoelectric transport prop-erties are needed. While ZT has reachedhigh values in superlattices, devices madeof these materials have not reached thebest performance of bulk thermoelectriccoolers, due to difficulties in electrical con-tacts, heat spreading, materials matching,and fabrication. Continued progress in thedevice area is critical for translating thelaboratory work successfully into practi-cal applications.

AcknowledgmentsH. Böttner’s work was partially sup-

ported by the German Federal Ministry ofEducation and Research (BMBF), grant03N2014A. G. Chen’s work on superlat-

tices was supported by the National Sci-ence Foundation and the Office of NavalResearch MURI program. R. Venkatasub-ramanian’s work was supported byDARPA/DSO through the Office of NavalResearch (1997–present) and the ArmyResearch Office (2000–2003).

References1. Proc. 1st Natl. Thermogenic Cooler Workshop,edited by S.B. Horn (Center for Night Vision and Electro-Optics, Fort Belvoir, VA,1992).2. D. Hicks and M.S. Dresselhaus, Phys. Rev. B47 (1993) p. 12727.3. R. Venkatasubramanian, E. Siivola, T. Colpitts, and B. O’Quinn, Nature 413 (2001)p. 597.4. T.C. Harman, P. Taylor, M.P. Walsh, and B.E.LaForge, Science 297 (2002) p. 2229.5. G. Chen, Semicond. Semimetals 71 (2001)p. 203.6. R. Venkatasubramanian, Semicond. Semimet-als 71 (2001) p. 175.7. B. Moyzhes and V. Nemchinsky, Appl. Phys.Lett. 73 (1998) p. 1895.8. A. Shakouri and J.E. Bowers, Appl. Phys. Lett.71 (1997) p. 1234.9. H. Beyer, A. Lambrecht, E. Wagner, G. Bauer,H. Böttner, and J. Nurnus, Physica E 13 (2002)p. 965.10. S.M. Lee, D.G. Cahill, and R. Venkatasubra-manian, Appl. Phys. Lett. 70 (1997) p. 2957.11. T. Borca-Tasciuc, W.L. Liu, T. Zeng, D.W.Song, C.D. Moore, G. Chen, K.L. Wang, M.S.Goorsky, T. Radetic, R. Gronsky, T. Koga, andM.S. Dresselhaus, Superlattices Microstruct. 28(2000) p. 119.12. G.H. Zeng, A. Shakouri, C. La Bounty, G.Robinson, E. Croke, P. Abraham, X.F. Fan, H.Reese, and J.E. Bowers, Electron. Lett. 35 (1999)p. 2146.13. S. Cho, Y. Kim, S.J. Youn, A. DiVenere,G.K.L. Wong, A.J. Freeman, J.B. Ketterson, L.J.Olafsen, I. Vurgaftman, J.R. Meyer, and C.A.Hoffman, Phys. Rev. B 64 235330 (2001).14. M.S. Dresselhaus, Y.M. Lin, S.B. Cronin, O.Rabin, M.R. Black, G. Dresselhaus, and T. Koga,Semicond. Semimetals 71 (2001) p. 1.15. J. Nurnus, H. Böttner, and A. Lambrecht, inHandbook of Thermoelectrics, edited by M. Rowe,Chapter 46 (CRC Press, Boca Raton, FL, 2005)p. 1.16. D. Mahan, Semicond. Semimetals 71 (2001)p. 157.17. M.S. Dresselhaus, G. Dresselhaus, X. Sun,Z. Zhang, S.B. Cronin, T. Koga, J.Y. Ying, and G. Chen, Microscale Thermophys. Eng. 3 (1999)p. 89.18. G. Chen, M.S. Dresselhaus, J.-P. Fleurial,and T. Caillat, Int. Mater. Rev. 48 (2003) p. 45.19. G. Chen and A. Shakouri, J. Heat Transfer124 (2001) p. 242.20. G. Springholz, A. Holzinger, H. Krenn, H.Clemens, G. Bauer, H. Böttner, P. Norton, andM. Maier, J. Cryst. Growth 113 (1991) p. 593.21. A. Lambrecht, H. Böttner, M. Agne, R. Kurbel, A. Fach, B. Halford, U. Schiessl, and M. Tacke, Semicond. Sci. Technol. 8 (1993)p. 334.

22. R. Venkatasubramanian, T. Colpitts, B.O’Quinn, M. Lamvik, and N. El-Masry, Appl.Phys. Lett. 75 (1999) p. 1104.23. J. Nurnus, H. Beyer, A. Lambrecht, and H.Böttner, in Thermoelectric Materials 2000—TheNext Generation Materials for Small-Scale Refriger-ation and Power Generation Applications, editedby T.M. Tritt, G.S. Nolas, G.D. Mahan, D. Man-drus, and M.G. Kanatzidis (Mater. Res. Soc.Proc. 626, Warrendale, PA, 2000) p. Z2.1.1.24. H. Cui, I. Bhat, B. O’Quinn, and R. Venkata-subramanian, J. Electron. Mater. 30 (2001)p. 1376.25. A. Mzerd, D. Sayah, G. Brun, J.C. Tedenac,and A. Boyer, J. Mater. Sci. Lett. 14 (1995) p. 194.26. A. Mzerd, D. Sayah, J.C. Tedenac, and A.Boyer, Int. J. Electron. 77 (1993) p. 291.27. Y.A. Boikov, V.A. Danilov, T. Claeson, andD. Erts, in Proc. ICT’97 (IEEE, New York, 1997)p. 89.28. A. Foucaran, A. Giani, F. Pascal-Delannoy,A. Boyer, and A. Sackda, Mater. Sci. Eng., B 52(1998) p. 154.29. F. Völklein, V. Baier, U. Dillner, and E.Kessler, Thin Solid Films 187 (1990) p. 253.30. V.D. Das and P.-G. Ganesan, in Proc. 16thInt. Conf. Thermoelectrics (IEEE, New York, 1997)p. 147.31. H. Zou, M. Rowe, and G. Min, in Proc.ICT’00 (IEEE, Piscataway, NJ, 2002) p. 251.32. L.W. Da Silva, M. Kaviany, A. DeHennis,and J.S. Dyck, in Proc. ICT’03 (IEEE, Piscataway,NJ, 2003) p. 665.33. J. Nurnus, H. Böttner, H. Beyer, and A.Lambrecht, in Proc. ICT’99 (IEEE, Piscataway,NJ, 1999) p. 696.34. J.P. Fleurial, L. Gailliard, and R. Triboulet, J.Phys. Chem. Solids 49 (1988) p. 1237.35. H. Böttner, A. Schubert, H. Kölbel, A.Gavrikov, A. Mahlke, and J. Nurnus, in Proc.ICT’04, CD-ROM, Paper No. 009 (IEEE, Piscat-away, NJ, 2004).36. F.R. Harris, S. Standridge, C. Feik, and D.C.Johnson, Angew. Chem. Int. Ed. Engl. 42 (2003)p. 5295.37. M. Chitroub, S. Scherrer, and H. Scherrer,J. Phys. Chem. Solids 62 (2000) p. 1693.38. A. Lambrecht, H. Beyer, J. Nurnus, C.Künzel, and H. Böttner, in Proc. ICT’01 (IEEE,Piscataway, NJ, 2001) p. 335.39. A. Lambrecht, N. Herres, B. Spanger, S.Kuhn, H. Böttner, M. Tacke, and J. Evers,J. Cryst. Growth 108 (1991) p. 301.40. G. Springholz, V. Holy, M. Pinczolits, andG. Bauer, Science 282 (1998) p. 734.41. H. Zogg and M. Hüppi, Appl. Phys. Lett. 47(1985) p. 47.42. T.C. Harman, D.L. Spears, and M.J. Manfra,J. Electron. Mater. 25 (1996) p. 1121.43. T.C. Harman, D.L. Spears, D.R. Calawa,and S.H. Groves, in Proc. 16th Int. Conf. Thermo-electrics (IEEE, New York, 1997) p. 416.44. T.C. Harman, D.L. Spears, and M.P. Walsh,J. Electron. Mater. Lett. 28 (1999) p. L1.45. T.C. Harman, P.J. Taylor, D.L. Spears, andM.P. Walsh, J. Electron. Mater. 29 (2000) p. L1.46. H. Beyer, J. Nurnus, H. Böttner, A. Lambrecht, T. Roch, and G. Bauer, Appl. Phys.Lett. 80 (2002) p. 1216.47. J.C. Caylor, K. Coonley, J. Stuart, S. Nangoy,T. Colpitts, and R. Venkatasubramanian, in



Proc. 24th Int. Conf. Thermoelectrics (IEEE, Piscat-away, NJ, 2005).48. N. Peranio, O. Eibl, and J. Nurnus, in Proc.23rd Int. Conf. Thermoelectrics CD-ROM, PaperNo. 1059 (IEEE, Piscataway, NJ, 2004).49. J.C. Caylor, M.S. Dander, A.M. Stacy, J.S.Harper, R. Gronsky, and T. Sands, J. Mater. Res.16 (2001) p. 2467.50. W.L. Liu, T. Borca-Tasciuc, G. Chen, J.L. Liu,and K.L. Wang, J. Nanosci. Nanotechnol. 1 (2001)p. 39.51. D.W. Song, G. Chen, S. Cho, Y. Kim, andJ. Ketterson, in Thermoelectric Materials 2000—The Next Generation Materials for Small-Scale Re-frigeration and Power Generation Applications,edited by T.M. Tritt, G.S. Nolas, G.D. Mahan, D. Mandrus, and M.G. Kanatzidis (Mater. Res. Soc. Proc. 626, Warrendale, PA, 2000)p. Z9.1.1.52. D.W. Song, W.L. Liu, T. Zeng, T. Borca-Tasciuc, G. Chen, C. Caylor, and T.D. Sands,Appl. Phys. Lett. 77 (2000) p. 3854.53. G. Chen, B. Yang, W.L. Liu, T. Borca-Tasciuc, D. Song, D. Achimov, M.S. Dressel-haus, J.L. Liu, and K.L. Wang, Proc. 20th Int.Conf. Thermoelectrics (IEEE, Piscataway, NJ,2001) p. 30.

54. S.M. Lee and D.G. Cahill, J. Appl. Phys. 81(1997) p. 2590.55. W.S. Capinski, H.J. Maris, T. Ruf, M. Cardona, K. Ploog, and D.S. Katzer, Phys. Rev. B59 (1999) p. 8105.56. T.C. Harman, J. Appl. Phys. 29 (1958)p. 1373.57. C.R. Tellier and A.J. Tosser, Size Effects inThin Films (Elsevier, Amsterdam, 1982).58. G. Chen, J. Heat Transfer 119 (1997) p. 220.59. G. Chen, Phys. Rev. B. 57 (1998) p. 14958.60. P. Hyldgaard and G.D. Mahan, Phys. Rev. B56 (1997) p. 10754.61. S. Tamura, Y. Tanaka, and H.J. Maris, Phys.Rev. B 60 (1999) p. 2627.62. B. Yang and G. Chen, Microscale Thermo-phys. Eng. 5 (2001) p. 107.63. M.V. Simkin and G.D. Mahan, Phys. Rev.Lett. 84 (2000) p. 927. 64. B. Yang and G. Chen, Phys. Rev. B 67 195311(2003).65. B.C. Daly, H.J. Maris, K. Imamura, and S.Tamura, Phys. Rev. B 66 024301 (2002).66. R. Venkatasubramanian, Phys. Rev. B 61(2000) p. 3091.67. B. Yang and G. Chen, in Chemistry, Physics, and Materials Science for Thermoelectric

Materials: Beyond Bismuth Telluride, edited byM.G. Kanatzidis, T.P. Hogan, and S.D. Mahanti(Kluwer Academic/Plenum, NY, 2003) p. 147.68. R. Venkatasubramanian, E. Siivola, B.O’Quinn, K. Coonley, P. Addepalli, C. Caylor,A. Reddy, and R. Alley, Proc. 24th Int. Conf. Ther-moelectrics (IEEE, Piscataway, NJ, 2005).69. R. Venkatasubramanian, E. Siivola, B.C.O’Quinn, K. Coonley, P. Addepalli, M. Napier,T. Colpitts, and M. Mantini, Proc. 2003 ACSSymp. Nanotechnol. Environ., ACS SymposiumSeries 890 (American Chemical Society, Wash-ington, DC, 2004) p. 347.70. F. Völklein, M. Blumers, and L. Schmitt,Proc. 18th Int. Conf. Thermoelec. (IEEE, Piscat-away, NJ, 1999) p. 285. 71. J. Nurnus, H. Böttner, C. Künzel, U. Vetter,A. Lambrecht, J. Schumann, and F. Völklein,Proc. 21st Int. Conf. Thermoelectrics (IEEE, Piscat-away, NJ, 2002) p. 523).72. R. Alley, J. Canchhevaram, K. Coonley, B. O’Quinn, J. Posthill, E. Siivola, and R.Venkatasubramanian, Proc. 24th Int. Conf. Ther-moelectrics (IEEE, Piscataway, NJ, 2005).73. Y.-M. Lin and M.S. Dresselhaus, Phys. Rev. B60 075304 (2003).



Advertisers in This IssuePage No.

Active Nanophotonic Devices 167

Advanced Metallization Conference(AMC) 2006 168

A & N Corp. 172

* Asylum Research 182

* Carl Zeiss SMT, Inc. 174

* Chemat Technology, Inc. 175

* Gatan, Inc. 173

Goodfellow Corp. 176

High VoltageEngineering Inside front cover

Page No.

* Huntington MechanicalLabs, Inc. Outside back cover

* Hysitron, Inc. 184

* INTRINSIC Semiconductor Corp. 180

* Janis Research Co., Inc. 172

* JEOL USA, Inc. 178

* Kurt J. Lesker Co. back cover

Lehigh Microscopy School 177

* MDC Vacuum Products 170

* MMR Technologies, Inc. 258

* MTS Systems Corp. 161

Page No.

* NanoInk, Inc. 162

National Electrostatics Corp. 198

* PANalytical, Inc. 205

Shiva Technologies, Inc. 256

Sigma-Aldrich 166

* Springer 265

Super Shuttle 229

* Ulvac Technologies, Inc. 231

* Veeco Instruments, Inc. 165

* ZIRCAR Ceramics, Inc. 169

For free information about the products and services offered in this issue, check http://www.mrs.org/bulletin_ads*Please visit us at the Exhibit, April 18–20, during the 2006 MRS Spring Meeting in San Francisco.



BackgroundThe discovery and development of

thermoelectric (TE) materials with a highfigure of merit (ZT 1) has proven to be achallenging task in materials science andengineering. Worldwide research effortshave culminated in somewhat of a two-pronged strategy for identifying this classof TE materials: engineered bulk materialsand nanostructured materials.

In the first approach, complex crystalstructures have been designed and syn-thesized with the aim of attaining phonon-glass/electron-crystal characteristics (lowlattice thermal conductivity as in glass,with high electrical conductivity as in

metals) for high ZT. Recent progress onbulk thermoelectric materials is discussedin the article by Nolas et al. in this issue.

The second approach is motivated bythe presence of increased electron densityof states at the Fermi level in nanostruc-tured materials and the possibility of ex-ploiting boundary scattering to reduce thethermal conductivity.

As indicated by its prefix, a nanostruc-ture has one of its crucial dimensions onthe order of 1–100 nm. Various forms ofnanostructures, such as nanotubes, nano-wires, nanorods, nanoparticles, ultrathinfilms, quantum wells, and superlattices,

have been prepared in the laboratory andare providing valuable insights for engi-neering materials with improved TE prop-erties. In this issue, Böttner et al. providean overview of the progress achieved in2D nanostructured TE materials, specifi-cally, superlattices and quantum well ma-terials. Nanocomposites that involve theincorporation of any of the aforemen-tioned nanostructures in the correspon-ding bulk material are also gaining muchinterest, since enhanced TE properties areexpected to result from interfacial reflec-tion and scattering of phonons in thenanocomposites. This article focuses onthe bulk synthesis of 1D thermoelectricnanomaterials and their TE properties. Inparticular, the Seebeck coefficient will bediscussed in detail for comparing differentforms of nanostructured materials, sincethis parameter has proved to be a verypowerful probe.

Carbon NanotubesWe begin with a discussion of the TE

properties of carbon nanotubes, which arethe latest molecular form of carbon, dis-covered by S. Iijima in 1991.1 Researchershave recently begun to synthesize othernanostructured materials, such as thosebased on state-of-the-art TE materialssuch as PbTe and Bi2Te3, which will be dis-cussed later in this article. Much can belearned from the TE properties of thesecarbon nanotubes, and therefore, we willfocus the early parts of this article on thesematerials.

A nanotube can be viewed as a graphenesheet rolled into a seamless cylinder withan aspect ratio (length/diameter) exceed-ing 1000. Carbon nanotubes are typicallyclassified into two groups: single-wallednanotubes (SWNTs) and multiwallednanotubes (MWNTs), which are basicallya set of concentric SWNTs.2 Nanotubescan be synthesized by one of three com-monly used methods:3 electric arc dis-charge, pulsed laser vaporization, andchemical vapor deposition (CVD).

In carbon nanotubes, the Seebeck coeffi-cient α is often used to determine the signof the dominant charge carrier4–8 and alsoto probe sensitivity to gas adsorption andmolecular collisions with the nanotubes.9Since in most metals, α is sensitive to thecurvature of the band structure near theFermi level through the Mott relation,10

(1)

it is a good indicator of dominant carriertype. In this expression, kB is the Boltz-mann constant, T is temperature, e is the

αd π2kBT

3e lnσε εEF

,

Properties ofNanostructuredOne-Dimensionaland CompositeThermoelectricMaterials

Apparao M. Rao, Xiaohua Ji, and Terry M.Tritt

AbstractOver a decade ago, Dresselhaus predicted that low-dimensional systems would one

day serve as a route to enhanced thermoelectric performance. In this article, recentresults in the thermoelectric properties of nanowires and nanotubes are discussed.Various synthesis techniques will be presented, including chemical vapor deposition forthe growth of thermoelectric nanostructures in templated alumina. Electrical transportmeasurements of carbon nanostructures, such as resistivity and thermopower, haverevealed some very interesting thermoelectric properties. Challenges still remainconcerning the measurement of individual nanostructures such as nanowires. Muchwork has been performed on the thermoelectric properties of carbon nanotubes, andthese results will be highlighted. In addition, routes for enhanced thermoelectric mate-rials have focused on incorporating nanostructures within the bulk materials.The role ofthese “hybrid composite structures” based on nanomaterials incorporated into the bulkmatrix and the potential for enhanced performance are discussed.

Keywords: composite, nanostructure, thermal conductivity, thermoelectricity.


Properties of Nanostructured One-Dimensional and Composite Thermoelectric Materials


electron charge, ∂lnσ/∂ε is the energy de-rivative of the electrical conductivity σ,and EF is the Fermi energy. The subscript dindicates the contribution of the diffusionthermopower to the total α.

A general comparison between α(T) ofSWNTs and MWNTs with highly orientedpyrolytic graphite (HOPG) is quite useful.Figure 1 shows α(T) for four different car-bon samples, which include as-preparedSWNT bundles produced by the electricarc discharge (black circles) and thepulsed laser vaporization (PLV) (bluesquares) techniques, an as-prepared film ofMWNTs grown by CVD (red diamonds),and HOPG (green crosses). All foursamples were exposed to room air androom light for an extended period of timeand are therefore considered to be suffi-ciently oxygen-doped. Clearly, SWNTs havea much higher room-temperature α of ap-proximately 45 µV/K than the MWNTs(approximately 17 µV/K). Notice that thesign of α for both the SWNTs and theMWNTs is positive, indicating p-type be-havior in oxygen-doped nanotubes. Thebulk graphite displays a room-temperaturevalue of approximately −4 µV/K. This figure serves as a quick reference for α(T)behavior in all three of the air-exposedcarbon forms.

Exposure to oxygen (or air) has beenshown to have a dramatic effect on α inSWNTs4–6 and MWNTs.7 The sign of α in a sample of purified SWNTs switches re-versibly from positive to negative in atime frame of 15–20 min as the sample iscycled between O2 and vacuum.4 This signreversal in α is indicative of a change inthe dominant charge carrier switchingfrom n-type (in vacuum) to p-type (in O2).In Figure 2a, α(T) data for a film of puri-fied SWNTs in its O2-saturated state (p-type) as well as after being completelydeoxygenated (n-type) are shown.5 Inter-estingly, this transformation is reversibleover the entire temperature range, andfield-effect transistors based on isolatedsemiconducting single-walled carbon nano-tubes (CNTFETs) also exhibited p-type(when exposed to O2) and n-type (an-nealed under vacuum) device characteris-tics.11 The data in Figure 2a suggest thatinteraction with O2 is responsible for the p-type character of CNTFETs in air. How-ever, detailed studies of doping-inducedFET characteristics of SWNTs conductedby the Avouris group at IBM do not seemto fully support this conclusion. In theSWNTs exposed to air or oxygen, the Fermienergy level (EF) is extrinsically pinnednear the valence-band maximum.11 Consis-tent with this model, little difference in theelectrical resistance R of oxygen-saturatedand annealed SWNTs has been observed.6

Alternatively, SWNTs can also be p-doped to a greater extent by immersion inconcentrated nitric or sulfuric acids. R(T)of such p-doped SWNTs decreases by afactor of 10 or more compared with thecorresponding value in pristine SWNTs,implying that EF is well into the semicon-ductor valence bands. The low-temperatureα(T) data of these acid-treated SWNTs ex-hibited distinctly different behavior fromoxygen-saturated and annealed SWNTs.This unusual low-temperature behaviorhas been attributed to phonon drag orelectron–phonon scattering in 1D elec-tronic systems.12,13

As in the case of acid-treated SWNTs,molecular adsorbates that engage incharge transfer with the nanotube wall are

expected to significantly impact the trans-port of charge and heat along the wall.However, from systematic studies of α orR or bundled SWNTs, Eklund and co-workers found evidence for new scatteringchannels for the free carriers in metallicnanotubes. An exponential increase in thetime-dependence of α(t) and resistanceR(t) of thin films of bundled SWNTs wasobserved, caused by physisorption of vari-ous alcohol molecules (CnH2n+1OH; n 1–4) when a sudden change in the molec-ular pressure occurs.14 The trends in thechanges in α and R with adsorption ofthese polar molecules were explained onthe basis of the interplay between the ad-sorption energy Ea (measured on graphiticcarbons) and the molecular coverage1/A, where A is the projection area of aparticular alcohol molecule on the nan-otube surface. In another systematic studyof α and R of six-membered ring hydro-carbons C6H2n (n 3–6) physisorbed onSWNTs, further evidence for scatteringchannels in nanotubes was found.15 Ad-sorption of benzene (n 3) induced thelargest increase in α and R. As the numberof π electrons per molecule was reduced(molecules with increasing n), the impactof the molecular adsorption on α and R ofthe SWNT diminished (Figure 2b). Theseobservations revealed that the magnitudeof scattering scaled with the coupling ofthe π electrons in the adsorbed moleculesto π electrons in the metallic nanotubewall. Similarly, small increases in α ofSWNTs due to adsorbed H2, He, and N2have also been observed and identifiedwith the creation of additional scatteringchannels. At fixed gas temperature andpressure, changes in α and R of thinSWNT films were experimentally foundand theoretically confirmed to scale as

Figure 1. Seebeck coefficient α versustemperature for air-saturatedsingle-walled nanotubes (SWNTs),multiwalled nanotubes (MWNTs), andhighly oriented pyrolytic graphite(HOPG). EA and PLV refer to SWNTsprepared by electric arc and pulsedlaser vaporization methods,respectively.

Figure 2. (a) Seebeck coefficient α versus temperature for a SWNT film exposed to O2 andthen deoxygenated.5 (b) In situ α( t) during successive exposure of a degassed SWNT filmto vapors of six-membered ring molecules C6H2n (n 3–6).The dashed lines in (b) areguides to the eye.15



M1/3, where M is the mass of the collid-ing gas molecules (He, Ar, Ne, Kr, Xe, CH4,and N2).9

The CNTFETs are inferior to analogoussilicon devices from the standpoint oflong-term device operation, since origi-nally n-type CNTFETs acquire p-type de-vice characteristics when exposed to airunder ambient conditions. To enable car-bon nanotubes to function as a basicbuilding block in future nanoelectronicdevices, it is crucial that the oxygen-induced reversal of electronic properties(discussed in Figure 2a) be arrested. Oneapproach is to dope nanotubes with eithernitrogen or boron to render them perma-nently n-type or p-type, respectively. Fig-ure 3 compares α(t) obtained for purifiedpristine SWNTs (crosses) with those of as-prepared boron-doped SWNTs8 andnitrogen-doped MWNTs.16 The α(t) dataof both SWNT samples were collected asthey were degassed in 10–6 Torr vacuum at500 K. It is clearly seen that, as adsorbedoxygen is removed from the pristineSWNT sample, α(t) changes sign from ap-proximately 20 µV/K (p-type) to –20µV/K (n-type), which is similar in behav-ior to data depicted in Figure 2a. How-ever, in the case of the boron-dopedSWNTs, α(t) remains constant at around60 µV/K, implying a greater stability ofthe doped nanotubes with respect to thedegassing procedure. Substituted boron inthe sample would be expected to donateholes to the SWNT lattice, making it a permanently p-type material and result-ing in no change in α(t) as the sample is degassed. Likewise, nitrogen-doped nan-otubes are expected to exhibit permanentn-type characteristics, as exemplified inFigure 3.

Finally, we note other interesting TEproperties in carbon nanotubes. Recent ad-vances in mesoscopic TE measuring deviceshave enabled measurements of α in asingle SWNT.17 A room-temperature α 200 µV/K, which is much larger than α 60 µV/K measured in SWNT films, wasobtained. As discussed in the introductoryarticle to this issue by Tritt and Subraman-ian, “rattlers” introduced in bulk cage-likematerials have the effect of reducing thethermal conductivity κ and thereby en-hancing the figure of merit, ZT α2σT/κ.The rattling motion of these loosely boundatoms effectively scatters phonons, thussignificantly reducing the lattice thermalconductivity of these materials. Onemight expect that C60 molecules inside thehollow core of a SWNT would act like rat-tlers.18 Regular SWNTs showed a room-temperature α value of 60 µV/K after airsaturation, whereas the C60@SWNTs(the @ symbol designates C60 molecules

encapsulated inside a nanotube) saturateat 40 µV/K at 300 K.

Evidence for the existence of the Kondoeffect has been reported through measure-ments of α in SWNTs synthesized using a variety of catalyst particles.19 The Kondoeffect arises from the interactions be-tween a single magnetic atom, such as a transition-metal atom, and the manyelectrons in an otherwise nonmagneticmetal, such as a carbon nanotube. A broadpeak in the α(T) data near 70–100 K has been associated with the Kondo effect and attributed to the presence ofmagnetic catalyst impurities in the SWNT films. The Kondo effect is com-pletely suppressed when the magnetic im-purities are converted to correspondingiodides by treating the SWNT samplewith iodine.

Other Nanostructured MaterialsSince the discovery of carbon nano-

tubes, there have been rapid advances inthe synthesis and understanding of thephysical properties of these interestingmaterials. As mentioned, the structure of acarbon nanotube is derived from the 2Dnature of the graphene sheet that is essen-tially “rolled up” into a seamless 1Dnanostructure, forming either a single-walled or a multiwalled nanotube. Recently,it has been found that other 2D electronsystems, such as layered dichalcogenides(WS2, MoS2, and TiS2), and other layeredstructures, including Bi2Te3, have beensynthesized in a 1D nanostructure form.We will not make an effort to survey thevast amount of high-quality research thathas been performed on the synthesis anddevelopment of 1D nanostructures, butwill instead attempt to highlight a fewareas that may be of importance to the de-velopment of higher-performance TE ma-terials that may be eventually based onthese nanostructures. We will try to also

highlight some of the broad synthesistechniques that are used to achieve thesenanostructures.

MoS2 NanotubesAs far back as 2001, nanostructures of

other 2D layered compounds have beengrown as 1D nanostructures.20 Remskaret al. used a novel type of catalyzed reac-tion including C60 as a growth parameterto grow subnanometer-diameter single-walled MoS2 nanotubes. The SWNTs ofMoS2 were grown by adding C60 (5 wt%)to MoS2 powder in a transport tube as acatalyst. The reaction was run for over 20days at 1010 K in an evacuated silica am-poule. The pressure was 10–3 Torr with atemperature gradient of 6 K/cm. TheMoS2 nanotubes grew as twisted chiralbundles. These bundles were held to-gether by interstitial iodine, which is easily removed, thus allowing for free-standing MoS2 nanotubes. These bundlescontain more than 500,000 ordered nan-otubes. The tubes have various lengths,up to several hundred nanometers, but arerather uniform in diameter (0.3–0.9 nm). Itis predicted that these materials shouldexhibit transport properties that are metal-lic; however, because of the narrow diam-eters of the tubes, energy quantization isexpected to occur and yield electron-hopping perpendicular to the tube axis.However, no electrical transport measure-ments were reported. One would expectthat large thermopower might be ob-served in the MoS2 nanotubes, because ofthe sharp variation in the density of statesaccording to the Mott formula, as dis-cussed previously.

WS2 NanotubesChen et al. recently reported the synthe-

sis of high-purity WS2 nanotubes.21 Thesematerials were synthesized on a ratherlarge scale by the thermal decompositionof amorphous/nano (NH4)2WS4 in a float-ing hydrogen and thiophene atmosphere.The heating temperature was between360C and 450C. This technique yieldedlarge quantities of open-ended WS2 nano-tubes with an average length of 5 µm anddiameters between 25 nm and 50 nm.They performed an extensive ball-millingprocedure for about 1 h on the (NH4)2WS4powder, which they emphasize is a veryimportant part of the process in order toobtain the nanotubes. They call the pow-ders amorphous/nano because of the extensive ball-milling. They investigatedvarious growth parameters such as temperature, flow rate, and thiopheneconcentration in order to obtain the opti-mum growth conditions for obtaining thenanotubes.

Figure 3. Seebeck coefficient α versustime in doped SWNTs and MWNTscompared with pristine SWNTs.



TiS2 NanotubesThese same authors also were able to

synthesize TiS2 nanotubes.22 They used alow-temperature gas reaction and ob-tained nanotubes with an outer diameterof 20 nm and lengths of several micro-meters. The TiS2 nanotubes were derivedfrom the following reaction:

TiCl4 + 2 H2S → TiS2 + 4 HCl. (2)

The nanotubes are multiwalled andopen-ended, with an outer diameter of20 nm, an inner diameter of 10 nm,and fringe spacing of roughly 0.57 nm.The crystal structure for the bulk TiS2 isshown in Figure 4. Notice that the 2Dlamellar structure is very similar to thegraphene sheets that result in the growthof carbon nanotubes. It is apparent thatthese 2D structures are very susceptible tothe growth of nanostructures from the cor-responding 2D parent bulk material.

So the question may be asked, why isthe synthesis of materials such as these ofimportance to improved TE materials?The dimensionality can play a crucial rolein the electronic transport processes in amaterial. Do the electrons interact in a 3D,2D, or 1D manner? The dimensionalitycan even change a system from metallic to

semiconducting behavior. Therefore, letus recall the Mott relation as described byEquation 1 but in a slightly different form.We observe that the thermopower is pro-portional to a derivative of the density ofstates (DOS) at the Fermi level, as shownin Equation 3:

(3)

Here, n(ε) is the density of states. Thus, aspredicted by Hicks and Dresselhaus,23 thethermopower could be enhanced in agiven set of materials simply by reducingthe dimensionality. Bulk TiS2 already pos-sesses some very interesting TE proper-ties. It has been reported that a series oftransition-metal dichalcogenides (e.g.,TiS2) exhibit very high thermopower val-ues (α 250 µV/K at T 300 K).24 Thematerials are relatively large single crys-tals (10 mm × 5 mm × 0.1 mm) that are ofnearly stoichiometric composition. Thesematerials are grown by chemical vaportransport using I2. The transition-metaldichalcogenides are a quasi 2D systemand are considered electronically as a 2Delectron gas. There should be considerablelatitude for further doping in the TiS2 sys-tem, and in addition, the samples can beintercalated with other atoms within the2D layers, which are separated by weaklybonded van der Waals planes. Whenformed as nanotubes, they may be able tobe intercalated within the fringe layers.

Here, again, exists the “layered struc-ture” that appears so common among po-tential TE materials. It is noted in the mostrecent reference22 that the carrier concen-tration n of the TiS2 system is an order ofmagnitude larger (n 1021 carriers/cm3)than the state-of-the-art TE material Bi2Te3,which has an optimum carrier concentra-tion. However, the power factor of the TiS2system is nearly the same as Bi2Te3 at T300 K and remains relatively large downto as low as 50 K, thus showing some po-tential as a refrigeration material.

Bi2Te3 Nanostructures:Electrodeposition

Nanostructures of state-of-the-art TEmaterials such as Bi2Te3 have also beensynthesized.25,26 For example, Stacy andco-workers have been successful in form-ing Bi2Te3–xSex nanostructures using elec-trodeposition techniques.27 These materialsare synthesized by array fabrication meth-ods using porous anodic alumina tem-plates. Very good control of the samplediameter, as well as large aspect ratios, canbe achieved using the alumina templates.This can be observed in Figure 5, where an array of electrodeposited Bi2Te3–xSex

αd 1

nεnεεEF

.

nanostructures is shown. A cross section isshown in Figure 5a, and a bottom viewafter removal of the Pt electrode is shownin Figure 5b. High-quality crystalline ma-terials of 50 nm and 200 nm have beensynthesized. The Bi2Te3–xSex crystal struc-ture has been confirmed by x-ray diffrac-tion, which shows that the wires aretextured along the [110] direction. Stacyand co-workers have been able to achievea relatively high density of wires, on theorder of 5 × 109 cm–2.28 These wires had afinal composite thickness of 55–70 µm andan average wire diameter of 45 nm,which gives a wire aspect ratio of greaterthan 1000. Stacy et al. conclude that thesearrays could be easily incorporated intodevices if the properties prove superior tothose of existing bulk Bi2Te3 materials.

Bi2Te3 Nanostructures:Hydrothermal Techniques

Zhao et al. and his group at ZhejiangUniversity have also been quite successfulin forming nanostructures of Bi2Te3 and itsalloys using hydrothermal techniques.29–32

These authors contend that the figure ofmerit of TE materials could be significantlyimproved if the materials are nanostruc-tured, since the thermal conductivity maybe decreased more significantly than theelectric conductivity, in addition to a po-tential enhancement of the thermopower.

In their work, novel solution chemicalroutes such as solvothermal synthesis,33,34

hydrothermal synthesis, low-temperatureaqueous35 chemical synthesis, and even ultrasonic chemical synthesis36 have beendeveloped to prepare Bi2Te3-based nano-powders. Various morphologies of nanoBi2Te3, including nanotubes andnanocapsules, have been obtained, andrare-earth elements (REs) have also beenalloyed in Bi2Te3-based compounds.37 BulkTE materials and nanocomposites havebeen prepared by vacuum hot pressing.The TE transport properties of hot-pressed samples have been measured.

The most important result of Zhaoet al.’s solution-chemical-synthesizednano-Bi2Te3 work is the first report ofBi2Te3 compounds with hollow shellnanostructures.29,30,38 The nanotubes, hy-drothermally synthesized at 150°C, havediameters of 25–100 nm and lengths of upto several micrometers. It was found thatthe tube walls are spiral and tilted, withthe (003) plane of the Bi2Te3 rhombohedrallattice at an angle of about 20° to the tubeaxis, as shown in Figure 6. Irregular hol-low nanocapsules synthesized by a low-temperature aqueous chemical methodare 10–50 nm in size, with some short,nanotube-like nanocapsules 100–200 nmin diameter and 300–1500 nm in length.

Figure 4. Crystal structure of thedichalcogenide TiS2, also showing the TiS6 octahedral unit that builds upthe layered dichalcogenide structure.



Another important result of the Zhaoet al.’s work is the preparation of ad-vanced Bi2Te3-based nanocomposites with high TE properties.30,31 They first proposed the concept of “coessential”nanocomposites, which means thatnanostructures of a specified material co-exist in a matrix of the same material, forexample, using commercial Bi2Te3-basedalloys as the matrix and Bi2Te3 nanotubesas the additive. The objective is to reducethe thermal conductivity of a TE material

without deteriorating its electric proper-ties. A very low thermal conductivity of0.8 W m−1 K−1 has been obtained for thehot-pressed bulk nanocomposites. Thehighest dimensionless figure of meritreaches 1.25 at about 420 K, which is alsoone of the highest values ever reported forbulk Bi2Te3-based TE materials (Figure 7).

Carefully designed experiments havebeen performed to study the nucleation andgrowth of Bi2Te3 nanocrystals during hy-drothermal synthesis.29,32 Various nucle-

ation mechanisms, including a moleculegrowth model, a continuous nucleationmodel, and a nucleus saturation model,have been proposed for the different syn-thesis conditions. It was experimentallyfound that lateral growth, includingsurface-nucleation lateral growth, spirallateral growth, and mis-layered lateralgrowth, dominates the Bi2Te3 crystalgrowth during hydrothermal synthesisbecause of the anisotropic lattice structure.

PbTe and SkutteruditeNanostructures

Recently, nanostructures of other TEmaterials have been fabricated. For example, nanostructures of PbSe39 andalso of CoSb3 skutterudites have been ac-complished.40,41 The PbSe materials wereinitially grown as sphere-like structures,which after additional reaction time weregrown in a wire-like structure with a typi-cal width of 60–150 nm and length of1–5 µm. Recently, the group at Clemsonhas used a CVD technique to synthesizePbTe nanoparticles.42 Their CVD tech-nique resulted in a high yield of cubicPbTe nanoparticles of sizes ranging be-tween 50 nm and more than 200 nm. TheCoSb3 skutterudites were grown viasolvothermal methods that were able togive a very high yield; grams of CoSb3could be collected from several growths.Xie and co-workers used a solvo-thermalmethod employing CoCl2 and SbCl3 asprecursors. This required a two-step process

Figure 5. Scanning electron micrographs of Bi2Te3–xSex, showing an array of50-nm-diameter templated nanowires: (a) cross section; (b) bottom view after removal ofPt electrode.Taken from Figure 6 in Reference 25.

Figure 6.Transmission electron microscopy (TEM) and high-resolution TEM (HRTEM) images of Bi2Te3 nanotubes synthesized viahydrothermal methods in a closed system at 150C. (a) TEM image of a cluster of hydrothermally synthesized Bi2Te3 nanotubes. (b) TEM imageof three parallel nanotubes; the inset in (b) shows an electronic diffraction image of these three nanotubes. (c) HRTEM image of a Bi2Te3

nanotube, showing that the tube wall is spiral and tilted. (d) HRTEM image of another smaller Bi2Te3 nanotube with inclined c-axis and (003)planes with reference to the tube axis.Taken from Reference 29.



that resulted in large quantities of CoSb3nanoparticles of less than 30 nm in size. TheXCo4Sb12 system (where X is a rare-earthfiller atom for the rattling motion) is muchmore interesting in relation to thermo-electrics since this system has alreadyshown very favorable TE properties in bulkmaterials. Initially, the goal of this workwasthe synthesis of nanoparticles of CoSb3.

One may not need to have filled skut-terudite nanowires. It is possible that thefilled skutterudite matrix materials, alongwith the incorporation of nanostructuresof CoSb3 as phonon scattering centers,may be enough to effectively minimizethe lattice thermal conductivity κL in theseskutterudite materials.

Concluding RemarksThe low-dimensional materials dis-

cussed in this article offer new ways tomanipulate the electron and phononproperties of a given material. The keyidea is to use quantum size effects to in-crease the electron density of states at theFermi level and in this way to optimizethe power factor. Even when the quantumsize effects are not dominant, it is possibleto utilize classical size effects to alter thetransport processes, for example, the exploitation of boundaries to scatterphonons more effectively than electrons.However, it may be that the incorporationof these nanostructured materials into abulk matrix of similar materials may serveto reduce the lattice thermal conductivityvia nanoscale phonon scattering centersand thus improve the overall ZT in thecomposite. Therefore, once grown asnanostructures, these materials may be in-corporated into other bulk structures toenhance the properties of the original

parent materials. They could exhibit en-hanced electronic properties for thermo-electric applications, or they may just actas nanostructured phonon scattering cen-ters. Thus, the birth of hybrid or compos-ite bulk thermoelectric materials, where abulk TE matrix can be incorporated withTE nanostructures and densified into a newbulk composite material, may be the newdirection of research in thermoelectrics.

References1. S. Iijima, Nature 354 (1991) p. 56.2. M. Dresselhaus, G. Dresselhaus, and Ph.Avouris, eds., Carbon Nanotubes: Synthesis,Structure, Properties, and Applications (Springer-Verlag, Berlin, 2001).3. M. Meyyappan, ed., Carbon Nanotubes: Sci-ence and Applications (CRC Press, Boca Raton,FL, 2005).4. P.G. Collins, K. Bradley, M. Ishigami, and A.Zettl, Science 287 (2000) p. 1801.5. K. Bradley, S.-H. Jhi, P.G. Collins, J. Hone,M.L. Cohen, S.G. Louie, and A. Zettl, Phys. Rev.Lett. 85 (20) (2000) p. 4361.6. G.U. Sumanasekera, C.K.W. Adu, S. Fang,and P.C. Eklund, Phys. Rev. Lett. 85 (5) (2000)p. 1096.7. T. Savage, B. Sadanadan, J. Gaillard, T.M.Tritt, Y.-P. Sun, Y. Wu, S. Nayak, R. Car, N.Marzari, P.M. Ajayan, and A.M. Rao, J. Cond.Matter. 15 (2003) p. 1915; M. Grujicic, S. Nayak,T. Tritt, and A.M. Rao, Appl. Surf. Sci. 214 (2003)p. 289.8. K. McGuire, N. Gothard, P.L. Gai, M.S. Dresselhaus, G. Sumanasekera, and A.M. Rao,Carbon 43 (2005) p. 219.9. H.E. Romero, K. Bolton, A. Rosen, and P.C.Eklund, Science 307 (2005) p. 89.10. F.J. Blatt, P.A. Schroeder, C.L. Foiles, and D.Greig, Thermoelectric Power of Metals (PlenumPress, New York, 1976).11. P. Avouris, Acc. Chem. Res. 35 (2002) p. 1026.12. V.W. Scarola and G.D. Mahan, Phys. Rev. B66 205405 (2002).13. J. Vavro, M.C. Llaguno, J.E. Fischer, S.Ramesh, R.K. Saini, L.M. Ericson, V.A. Davis,R.H. Hauge, M. Pasquali, and R.E. Smalley,Phys. Rev. Lett. 90 065503 (2003).14. H.E. Romero, G.U. Sumanasekera, S.Kishore, and P.C. Eklund, J. Phys. Condens. Mat-ter 16 (2004) p. 1939.15. G.U. Sumanasekera, B.K. Pradhan, H.E.Romero, K.W. Adu, and P.C. Eklund, Phys. Rev.Lett. 89 166801 (2002).16. B. Sadanadan, T. Savage, J. Gaillard, S.Bhattacharya, T. Tritt, A. Cassell, Z. Pan, Z.L.Wang, and A.M. Rao, J. Nanosci. Nanotech. (Spe-cial Issue) 3 (2003) p. 99.17. J.P. Small, L. Shi, and P. Kim, Phys. Rev. Lett.91 256801 (2003).18. J. Vavro, M.C. Llaguno, B.C. Satishkumar,D.E. Luzzi, and J.E. Fischer, Appl. Phys. Lett. 80(8) (2002) p. 1450.19. L. Grigorian, G.U. Sumanasekera, A.L.Loper, S.L. Fang, J.L. Allen, and P.C. Eklund,

Phys. Rev. B 60 (1999) p. R11309.20. M. Remskar, A. Mrzel, Z. Skraba, A. Jesih,M. Ceh, J. Demsar, P. Stadelmann, F. Levy, andD. Mihailovic, Science 292 (2001) p. 479 and ref-erences therein.21. J. Chen, S.-L. Li, F. Gao, and Z.-L. Tao, Chem.Mater. 15 (2003) p. 1012.22. J. Chen, S.-L. Li, Z.-L. Tao, and F. Gao, Chem.Commun. (2003) p. 980.23. L.D. Hicks and M.S. Dresselhaus, Phys. Rev.B 47 (1993) p. 12727.24. H. Imai, Y. Shimakawa, and Y. Kubo, Phys.Rev. B 64 241104-R (2001) and references therein.25. P. Magri, C. Boulanger, J.M. Lecuire,J. Mater. Chem. 6 (1996) p. 773.26. J.P. Fleurial, A. Borschevsky, M.A. Ryan, W.Phillips, E. Kolawa, T. Kacisch, and R. Ewell, inProc. 16th ICT (IEEE, Piscataway, NJ, 1997)p. 641.27. M. Martin-Gonzalez, G.J. Snyder, A.L. Pri-eto, R. Gronsky, T. Sands, and A.M. Stacy, NanoLett. 3 (2003) p. 973 and references therein.28. M.S. Sander, A.L. Prieto, R. Gronsky, T.Sands, and A.M. Stacy, Adv. Mater. 14 (2002)p. 665.29. X.H. Ji, X.B. Zhao, Y.H. Zhang, T. Sun, H.L.Ni, and B.H. Lu, in Proc. 23rd ICT (IEEE, Piscat-away, NJ, 2004).30. X.B. Zhao, X.H. Ji, Y.H. Zhang, T.J. Zhu, J.P.Tu, and X.B. Zhang, Appl. Phys. Lett. 86 062111(2005).31. X.H. Ji, X.B. Zhao, Y.H. Zhang, B.H. Lu, andH.L. Ni, J. Alloys Compd. 387 (2005) p. 282.32. X.H. Ji, “Syntheses and Properties ofNanostructured Bi2Te3-Based ThermoelectricMaterials,” PhD dissertation, Zhejiang Univer-sity, Hangzhou, China (2005).33. X.B. Zhao, X.H. Ji, Y.H. Zhang, and B.H. Lu,J. Alloys Compd. 368 (1–2) (2004) p. 349.34. X.H. Ji, X. Zhao, Y. Zhang, B.H. Lu, and H.Ni, in Thermoelectric Materials 2003—Researchand Applications, edited by G.S. Nolas, J. Yang,T.P. Hogan, and D.C. Johnson (Mater. Res. Soc.Symp. Proc. 793, Warrendale, PA, 2004) p. 21.35. X.B Zhao, T. Sun, T.J. Zhu, and J.P. Tu,J. Mater. Chem. 15 (2005) p. 1621.36. Y.Y. Zheng, T.J. Zhu, X.B. Zhao, J.P. Tu, andG.S. Cao, Mater. Lett. 59 (2005) p. 2886.37. X.H. Ji, X.B. Zhao, Y.H. Zhang, B.H. Lu, andH.L. Ni, J. Alloys Compd. 387 (2005) p. 282; X.H.Ji, X.B. Zhao, Y.H. Zhang, B.H. Lu, and H.L. Ni,Mater. Lett. 59 (2005) p. 682; X.B. Zhao, Y.H.Zhang, and X.H. Ji, Inorg. Chem. Commun. 7(2004) p. 386.38. X.B. Zhao, X.H. Ji, Y.H. Zhang, G.S. Cao,and J.P. Tu, Appl. Phys. A 80 (2005) p. 1567.39. A. Sashchiuk, L. Amirav, M. Bashouti, M.Krueger, U. Sivan, and E. Lifshitz, Nano Lett. 4(2004) p. 159.40. J. Xie, X.-B. Zhao, J.-L. Mi, G.-S. Cao, and J.-P. Tu, J. Zhejiang Univ., Sci. (JZUS), Lett. 5(2005) p. 1504.41. J. Xie, X.B. Zhao, G.S. Cao, M.J. Zhao, andS.F. Su, J. Power Sources 140 (2005) p. 350.42. B. Zhang, J. He, and T.M. Tritt, Appl. Phys.Lett. 88 043119 (2006).43. T.M. Tritt, Science 283 (1999) p. 804.

Figure 7.Thermoelectric figure of meritZT of the Bi2Te3 nanocomposite, shownwith those of the zone-melted sample.Adapted from Reference 43.

The Materials Gateway — www.mrs.org



IntroductionOf the various static energy-conversion

technologies considered for radioisotopepower systems for space applications,thermoelectric (TE) energy conversion hasreceived the most interest. Radioisotopethermoelectric generators (RTGs) generateelectrical power by converting the heat re-leased from the nuclear decay of radioac-tive isotopes (typically plutonium-238)into electricity using a TE converter (Fig-ure 1). RTGs have been successfully usedto power a number of space missions, in-cluding the Apollo lunar missions; theViking Mars landers; Pioneer 10 and 11;and the Voyager, Ulysses, Galileo, andCassini outer-planet spacecrafts. Thesegenerators have demonstrated their reliability over extended periods of time(tens of years) and are compact, rugged,radiation-resistant, and scalable. Theyproduce no noise, vibration, or torqueduring operation. These properties havemade RTGs suitable for autonomous mis-sions in the extreme environment of spaceand on planetary surfaces. Converterunits use TE materials, which, when oper-ating over a temperature gradient, pro-

duce a voltage called the Seebeck voltage1

(see the introductory article by Tritt andSubramanian in this issue). The systemconversion efficiency for state-of-practiceRTGs is about 6%.

The most widely used TE materials, inorder of increasing operating tempera-ture, are bismuth telluride (Bi2Te3); leadtelluride (PbTe); tellurides of antimony,germanium, and silver (TAGS); lead tintelluride (PbSnTe); and silicon germanium(SiGe). All of these materials, except Bi2Te3,have been used in RTGs on space missions.

A wide variety of physical, thermal, andTE property requirements must be met forthe design of reliable RTG converters.These properties are summarized in thisarticle; the discussion will also includelifetime power-output degradation mech-anisms in state-of-practice generators.

The number of motor vehicles on U.S.roads and the number of miles driven bythose vehicles continue to grow, resultingin increased air pollution and petroleumconsumption, and reliance on foreignsources of that petroleum, despite im-provements in vehicle emissions control

and fuel efficiency. To counter thesetrends, new vehicle technologies must beintroduced that can achieve better fueleconomy without increasing harmfulemissions.

For a vehicle powered by a typicalgasoline-fueled internal-combustion en-gine (ICE), only about 25% of the fuel en-ergy is used for vehicle mobility andaccessory power; the remainder is lost inthe form of waste heat in the exhaust andcoolant, as well as friction and parasiticlosses such as tire rolling resistance, winddrag, brake drag, and wheel misalign-ment (Figure 2).2 Furthermore, in order tomeet increasing safety requirements, im-prove engine performance, and reduce exhaust emissions, original equipmentmanufacturers (OEMs) are incorporatingincreased electronic content, such as sta-bility controls; telematics for accuratesatellite positioning of vehicles; collision-avoidance systems; OnStar communicationssystems; navigation systems; steer-by-wire systems, in which the steering col-umn is replaced with electronic controlunits linked in a fault-tolerant network forenhanced safety and fuel economy; elec-tronic braking; additional powertrain andbody controllers; and sensors that can au-tomatically optimize performance, im-prove fuel economy, and enhance vehiclesafety. All of these additional electronicdevices require more energy from the en-gine, or use enhanced energy-managementschemes.

Unfortunately, current engine designsmay not meet all of our future needs. Inorder to meet the ever-increasing require-ments of electrical power, automotive OEMsare considering several alternatives, suchas 42 V systems (versus the 14 V systemsin current automobiles), hybrid vehicles,fuel cell vehicles, and auxiliary electricpower units. One such alternative is theuse of TE technology.

Several proposed applications of TEwaste-heat recovery devices in the auto-motive industry are reviewed here. To as-sess the feasibility of these applications atthe vehicle level, the effect of electrical loadand weight on fuel economy for a repre-sentative midsize truck is discussed. Themost challenging areas of commercializa-tion of these technologies are identified.

Radioisotope ThermoelectricGeneratorsThermoelectric Converters

The RTGs used by the U.S. space pro-gram have utilized converter TE materialsmade from alloys of lead telluride, TAGS,or SiGe. Figure of merit (ZT) values forstate-of-practice TE power-generation ma-terials are listed in Reference 1. The TE ef-

ThermoelectricMaterials for Spaceand AutomotivePower Generation

Jihui Yang and Thierry Caillat

AbstractHistorically, thermoelectric technology has only occupied niche areas, such as the

radioisotope thermoelectric generators for NASA’s spacecrafts, where the low coolingcoefficient of performance (COP) and energy-conversion efficiency are outweighed bythe application requirements. Recent materials advances and an increasing awarenessof energy and environmental conservation issues have rekindled prospects forautomotive and other applications of thermoelectric materials.This article reviewsthermoelectric energy-conversion technology for radioisotope space power systems andseveral proposed applications of thermoelectric waste-heat recovery devices in theautomotive industry.

Keywords: thermal conductivity, thermoelectricity.


Thermoelectric Materials for Space and Automotive Power Generation


ficiency is proportional to ZT and theCarnot efficiency as

ε (1)

where TH and TC are the hot-end and cold-end temperatures of the thermoelectricmaterials, respectively, and T is the aver-age temperature. The Carnot efficiency isgiven by the term (TH –TC)/TH. Equation 1shows that increasing efficiency requiresboth having high ZT values and maximiz-ing the temperature gradient across the TEmaterials. There is, however, no single TEmaterial with a high average ZT between1275 K and 300 K. Each material is best ina limited temperature range.

Earlier RTGs used TE materials that in-cluded telluride alloys, while more recentRTGs have used SiGe alloys. Telluride-based TE materials provide better per-formance at lower temperatures (900 Kor less), while SiGe alloys can operate upto a maximum temperature of about1275 K. While it is important to maximizeefficiency to produce more power, a bettermeasurement of performance than effi-ciency for space power systems is howmuch power is produced per unit of mass,which is called specific power (typicallyexpressed in W/kg). It is therefore impor-tant when designing and optimizing anRTG to optimize the heat-rejection tem-perature for minimizing the size and massof the radiator used to reject the excessheat from the power converter into space(the size of the radiator varies as Tc

4).The TE efficiency of past and current

RTGs used on U.S. space missions is about8%. The system efficiency for these RTGsis further decreased because of thermaland electrical losses that typically amountto 10–15%, depending on the genera-

TH TC

TH

1 ZT 1

1 ZT TC

TH

,

tor’s cold and hot operating temperaturesand design.

Past and Current RadioisotopeThermoelectric Generators

The United States began the develop-ment of RTGs in the 1950s. As previouslymentioned, the earlier RTGs used tel-luride alloys, while the more recent RTGshave used SiGe alloys. The telluride-basedgenerators generally required the use of acover gas (typically a mixture of argonand helium) to minimize sublimation ofthe TE materials at high temperatures.SiGe materials can operate without usinga cover gas in the vacuum of space. Gen-erators using SiGe alloys must, however,be sealed with an inert cover gas on theground to prevent oxidation of the multi-foil thermal insulation (typically molyb-denum). Nearly all telluride-basedgenerators have used pressure contact be-tween the TE elements and the heat source.Because SiGe alloys can operate up to1275 K, a radiation coupling between theTE elements and the heat source was used.

RTGs that used telluride-based TE materials include SNAP-3A, SNAP-9A,SNAP-19, SNAP-19B, SNAP-27, andTransit-RTG. SNAP is an acronym for“systems for nuclear auxiliary power.”The Multi-Mission Radioisotope Thermo-electric Generator (MMRTG) being devel-oped by NASA and the U.S. Departmentof Energy (DOE) will also used telluride-based TE materials. The beginning-of-lifeelectric power output of MMRTG andSNAP systems ranges from 2.7 W to 125 W.A detailed list and characteristics of thesegenerators was compiled by G. Bennet.3Much improvement in the design of thesegenerators has been achieved, and whilethe initial RTGs had a specific power onthe order of only 1.3 W/kg, the specificpower for the SNAP-19 generators used

on the Pioneer and Viking missionsreached 2.8 W/kg.

The MMRTG is based on the SNAP-19RTG. The TE converter utilizes PbTe forthe n-type thermoelement and TAGS/PbSnTe for the p-type segmented thermo-element. It uses eight general-purposeheat sources (GPHSs),3 each of which de-livers approximately 250 W at beginningof life. The MMRTG is designed to operateboth in the vacuum of space and in the ox-idizing atmosphere of Mars. The MMRTGelectrical output is a function of the cold-side temperature of its TE modules and isdriven by the ambient environment. Thenominal power output in deep space iscurrently predicted to be 125 W at the be-ginning of the mission. This power is pro-duced through 768 thermoelectric couplesoperating between 823 K and 423 K. TheMMRTG measures approximately 66 cmlong and has a fin-to-fin tip width of 64 cmand a specific power of 2.8 W/kg.

SiGe alloys were first used in the Multi-Hundred Watt Radioisotope Thermoelec-tric Generator (MHW-RTG). The MHW-RTG was designed to provide over 150 Wof electric power at the beginning of mis-sion. Because SiGe alloys are able to oper-ate at a temperature of up to 1275 K, thegenerator was designed to operate the TEelements between 1275 K and 575 K witha thermoelectric efficiency of about 8%. Al-though a similar efficiency was achievedfor the telluride-based thermoelementsoperating between 823 K and 483 K, oper-ating SiGe alloys at higher temperaturesallowed a significant reduction in the finsize and, in turn, an increase in the specificpower to 4 W/kg for the MHW-RTG. Fur-ther improvements were introduced in theGPHS-RTG, the current standard RTGunit used in the United States. The GPHS-RTG is illustrated in Figure 3. It is de-signed to operate in the vacuum of space

Figure 2.Typical energy path in gasoline-fueled internal-combustionengine vehicles.2

Figure 1. Key components of a space-based radioisotopethermoelectric generator.



and uses 18 GPHS modules with a totalnominal thermal power of 4500 W. The572 SiGe unicouples are radiatively cou-pled around the 18 GPHS modules andconnected in a series-parallel network toprevent single-point failure of the genera-tor. The overall diameter of the RTG withfins is 42.2 cm, and it is 114 cm long. Thespecific power of the generator is about5.1 W/kg, nearly four times that of thefirst RTGs flown. GPHS-RTGs have beenused on the Galileo, Ulysses, and Cassinispacecrafts, providing years of continuouselectric power.

Physical and ThermoelectricProperty Requirements forThermoelectric Materials Used in RTGs

The power-output decrease for each ofthe three RTGs used on the Voyager space-crafts is 22% over 14 years, which repre-sents a 1.6% degradation in poweroutput per year. Out of this 1.6%, about0.8% is due to isotopic decay of the PuO2fuel. As a result of the reduced thermalinput over time, the hot-junction tempera-ture of the TE couples also decreases, re-sulting in a diminution of the coupleconversion efficiency that accounts forabout 0.47% a year of the total decrease ofthe generator power output. The rest ofthe power-output decrease, 0.33% peryear, is due to (1) variations in the trans-port properties of the TE materials overtime; (2) sublimation of the TE materialsnear the hot junctions of the couple, re-sulting in a cross-sectional reduction ofthe thermoelements and an increase inthermoelement resistance; (3) depositionof sublimation products, resulting in anincrease in electrical and thermal losses;and (4) an increase in the interfacial con-

tact resistance of the couples. This sectiondescribes the most important potentialdegradation mechanisms in RTGs relatedto TE materials.

SublimationSublimation is one of the primary

degradation mechanisms for TE power-generation devices. As mentioned earlier,sublimation of the TE materials near thehot junction of the couples can result in across-sectional reduction of the thermoele-ments that can increase the resistance ofthe TE elements, which in turn will de-crease the power output of the RTG. Ifsubstantial sublimation occurs over time,this could lead to mechanical failure of thejunction between the TE material and themetallization at the couple hot junction.The sublimation rate requirement to mini-mize the cross-sectional reduction to anacceptable value (typically up to 5–10%reduction over 10 years) can be estimatedas a function of the TE element diameter.Sublimation rates are typically expressedin g/cm2/h. Figure 4 shows that, as the di-ameter of the thermoelements diminishes,the sublimation rate required to limit thecross-sectional reduction to within 5–10%of the nominal cross-section area rapidlydecreases. The sublimation rate desiredfor the MMRTG and GPHS-RTG thermo-electric elements are also shown inFigure 4.

For the MMRTG, the desired sublima-tion rate ranges between 1 × 10–6 g/cm2/hfor a 10% reduction and 6 × 10–7 g/cm2/hfor a 5% reduction. As previously men-tioned, a 1 atm mixture of argon and he-lium is maintained in the convertercompartment of the generator to achievethose sublimation rates. In addition, the p-thermoelement of the MMRTG couple is

segmented, because although the TAGSmaterial is phase-stable up to 900 K, itssublimation rate above 673 K is above thedesired value and its operation is there-fore limited to 673 K in the couples. It is segmented to a top PbSnTe segment operating between 973 K and 673 K. ThePbSnTe material has a much lower subli-mation rate than the TAGS material in thistemperature range.

The sublimation rate requirement forthe SiGe alloys used in the GPHS-RTG is 6 × 10–7 cm2/g/h for a 10% reduction and1.5 × 10–7 cm2/g/h for a 5% reduction. Thedesired sublimation rates for SiGe arelower than those for the TE materials usedin the MMRTG couples, not only becausethe nominal cross section of the TE ele-ment is smaller but also because the mate-rial density is also lower. To achieve thisdesired rate, a Si3N4 coating was success-fully developed, which, in conjunctionwith wrapping the TE elements with aquartz yarn, has been typically used onSiGe thermoelements used in RTGs.

In addition to cross-sectional reductionthat can lead to substantial power outputdecrease and even mechanical failure ofthe TE couples if not controlled, sublima-tion products can condense within thethermal insulation between the couplesand on the cold-side interconnects, poten-tially creating electrical and thermal shortsin the converter area of the generator. Thiswould typically reduce the thermal effi-ciency and power output of the deviceand further justify the need for controllingthe sublimation of the TE materials.

Device ConsiderationsDesigning and fabricating mechanically

stable TE power-generation devices re-quires, to a great extent, taking into con-sideration the thermal expansion andmechanical strength of the TE materials.The joins between the TE materials andthe cold-side and hot-side interconnectmaterials are exposed to stress during fab-rication, and the choice of interconnectmaterials is critical to achieving a mechan-ically stable couple. TE materials withlower coefficients of thermal expansion(CTEs) tend to be more suitable for deviceintegration, as most refractory metalsoften used as interconnect materials tendto have relatively low CTEs.

Achieving long-lasting joins betweenthe TE materials and the couple intercon-nects requires not only minimizing ther-momechanical stresses resulting from thedissimilar nature of the joined materials,but also ensuring that chemical interac-tions at the joins will not increase the ther-mal and electrical contact resistance. Theelectrical contact resistance value for the

Figure 3. Configuration of a general-purpose heat source radioisotope thermoelectricgenerator (GPHS-RTG) using SiGe thermoelectric elements.



joins should typically be equal to or lessthan 5 µΩ cm2 in order to have no impacton the generator power output. Joins between the TE materials and the inter-connects are typically achieved by diffu-sion bonding or brazing. The diffusion-bonding or brazing process should lead tosome chemical reaction at the join toachieve the desired electrical contact resis-tance value, but this reaction should belimited over time to prevent the growth ofreaction layers that may have differentCTEs and could cause degradation andeven mechanical failure of the bonds.

Thermoelectric PropertiesThe obvious requirement for a good TE

material to be considered for integrationinto a power-generation device is to pos-sess a high average ZT over the projectedtemperature range. This is a beginning-of-life requirement to maximize the effi-ciency of the generator. For practicaloperations and in order to predict thepower output of the RTG over its lifetime,it is important to understand how the TEproperties vary as a function of time andtemperature.

A number of mechanisms can con-tribute to substantial variations in TE prop-

erties with time. In the case of SiGe alloys,the materials are saturated with dopantsto achieve the highest possible ZT valuesat beginning of life. Lifetime studies of theTE properties of SiGe alloys have shownthat dopants tend to precipitate from thesolid solutions and form localized precip-itates, resulting in a decrease in the carrierconcentration of the materials.4 The pre-cipitation is slower at lower temperaturesand has a more profound effect on theelectrical resistivity than on the Seebeckcoefficient and thermal conductivity. Theoverall effect on the ZT of SiGe alloys overtime is relatively small.

In the case of TAGS alloys, the materialstend to undergo a phase transformationover time, especially near maximum oper-ating temperatures.5 This results in achange in the overall TE properties thathas not been fully documented to date, al-though extensive life testing of the poweroutput of PbTe/TAGS couples has beenperformed.

Potential Automotive Applicationsof Thermoelectrics

Advances in TE technology can have a significant impact on the U.S. automo-tive industry in terms of fuel economy

improvements by generating electricityfrom waste heat and enhancing air-conditioning efficiency. First, TE technol-ogy has the ability to utilize the tens ofkilowatts of heat losses in vehicles2 to gen-erate electricity without added engineload. Second, TE technology could lead toan all-solid-state, extremely reliable, re-versible automotive air-conditioning sys-tem that does not use refrigerants withgreenhouse gas concerns and that can besimpler, easier to package, and more effi-cient to operate. TE coolers will almostcertainly require more electrical powerthan current mechanical systems, and thisshould be considered. The increased relia-bility of batteries and selected compo-nents is generally due to being able to useTE devices to control the temperature ofthe battery or other device.

TE power generators could help to in-crease the ability of ICEs to convert fuelinto useful power. By converting wasteheat into electricity, engine performance,efficiency, reliability, and design flexibilitycould be improved significantly. TE gener-ators could also be used to eliminate sec-ondary loads from the engine drive train,thus reducing torque and horsepowerlosses from the engine. This would helpreduce engine weight and direct the fullpower to the drive shaft, which would inturn help improve the performance andfuel economy. Furthermore, TE powergenerators could help improve fuel effi-ciency (through waste energy recovery)by supporting engine-off operation withminimum battery needs, and could in-crease electric power for new features. TEpower generators can function for only ashort period of time after the engine isturned off using available exhaust orcoolant heat. A fuel burner added to theTE subsystem will use fuel, but at a muchlower rate than a fully operating engine,thereby improving fuel economy withouta large battery pack.

For almost a half-century between the1940s and the early 1990s, the highest ZTvalues of all materials remained low, with ZT ≤ 1. Substantial federal fundingsince the early 1990s as well as private enterprise research and development efforts have led to significant increases in ZT in recent years,6–10 reinvigorating interest in TE technology. Figure 5a shows that ZT 3 would lead to the efficiency of a TE generator approaching50% of the Carnot efficiency (a thermo-dynamic limit), and Figure 5b shows a coefficient of performance in TE coolersthat outperforms those of mechanical air-conditioning units. Recent materials research breakthroughs that provide ZT 1, and some as high as ZT 3.6,10

Figure 4. Calculated desired sublimation rates required to keep the cross-sectionalreduction of thermoelectric (TE) elements between 5% and 10% over 10 years at themaximum operating temperature.The calculations were made for different types ofthermoelectric couples and TE materials as a function of the thermoelectric elementdiameter and thermoelectric material density. Each TE maximum operating temperature islisted, and the filled ellipses represent the desired sublimation rate targets for the two typesof thermoelectric couples. MMRTG is the Multi-Mission Radioisotope ThermoelectricGenerator being developed by NASA and the U.S. Department of Energy; TAGS stands fortellurides of antimony, germanium, and silver.



motivate significant interest for automotiveapplications of TE technology.

Efficiencies of ThermoelectricWaste-Heat Recovery Units

Because of the strong temperature-dependence of ZT for most materials, weestimate the average energy-conversionefficiency εave of a TE waste-heat recoveryunit by

εave

(2)

A typical vehicle exhaust gas temperatureis approximately 500°C. The TH of a TEgenerator at vehicle exhaust is generallylower than 500°C, because it is difficult toextract heat energy from flowing gas tothe solid surface of the heat exchanger.Optimistically, we assume TH 400°Cand TC 100°C, such that the cold side ofthe generator is cooled by radiatorcoolant. The best materials for such a tem-perature range would be the n- and p-typefilled skutterudites. Using their reportedZT values and Equation 2, εave is approxi-mately equal to 6.7%.6 Based on recentlydeveloped models and test data for TEunicouples, a TE generator made of seg-mented unicouples with the same valuesof TH and TC and two to three differentmaterials on each thermoelement wouldhave εave 7.5%,11 which provides an additional increase in the thermal-to-electrical energy-conversion efficiency.

For a radiator TE generator, one shouldhave TH ≈ 100°C and TC ≈ 27°C. The mate-rials choices in this case are the thin-film-

TH TC

THTH

TC

ZT 1 1ZT 1 TCTH

dT

TH TC.

based Bi2Te3/Sb2Te3 superlattices andPbSe1–xTex/PbTe quantum dots.9,10 Eventhough these materials have the highestZT values ever reported, a small tempera-ture gradient for the radiator TE generatorlimits the Carnot efficiency (thermody-namic limit) to only 19.6%. Assuming thatthe ZT values for n- and p-type Bi2Te3/Sb2Te3 superlattices are identical and canbe extrapolated9 into the temperaturerange under consideration, εave ≈ 6.9 % isestimated using Equation 2.

Methods of Improving FuelEconomy

Recently, the U.S. Department of En-ergy initiated several programs on auto-motive thermoelectric waste-heat recoverytechnologies. The overall objective is toachieve a 10% improvement in fuel econ-omy. A typical vehicle electrical load, ac-cording to the Environmental ProtectionAgency’s federal test procedures, is about300 W. Figure 6 shows the dependence offuel economy on electrical load at the al-ternator for a representative midsize truckat various weights to be in the 0–600 Wload range. These results were generatedby the Overdrive simulation tool inter-nally developed by General Motors.12

Overdrive solves systems of ordinary dif-ferential equations that approximate therigid-body dynamics of the vehicle. Theoutput from the simulation is a set of pa-rameters that reflects vehicle performance,fuel economy, drive quality, and energymanagement.12 These data, along withthose presented in Reference 2, illustratethat if the 10% fuel economy goal isachieved, then more electrical power willbe produced than is consumed by the ve-hicle electrical system under most drivingscenarios, including EPA fuel economyand emissions test procedures.

This leads to several choices: (1) reduc-ing electrical accessory load on the alter-nator using thermoelectrically generatedpower; (2) shifting some of the engine-driven accessories to electrical drive toraise the electrical accessory load con-sumption; or (3) attempting to use the ex-cess electrical power for something otherthan the vehicle electrical load, such aspropulsion. Examples of choice (2) includeelectric power steering, an electric coolantpump, an electric cooling fan, an oil pumpassist, electric valve actuation and timing,and electric heating for catalytic convert-ers. The use of electric power steering oran electric coolant pump alone, for ex-ample, could result in 2–3% or 3–5% fueleconomy improvements, respectively.13

Choice (3) is easily adaptable for hybridvehicles. It is difficult to estimate the fueleconomy gains for a TE-augmented hy-brid before detailed information on pack-aging, electrical interface, mechanicalinterface, control interface, and power-train are available. Because the excess elec-trical power is used directly for vehiclepropulsion, this choice has the potentialfor the largest fuel economy gains amongthe three. In order to achieve the 10% fueleconomy improvement goal, a combina-tion of the three choices mentioned shouldbe used. It should be pointed out that thefuel economy improvement technologiesare not necessary additive.

It is also important to keep in mind thatthe addition of a TE waste-heat recoveryunit would increase the overall vehicleweight and therefore reduce fuel econ-omy. The truck represented in Figure 6weighs 4417 lb, with an 18.8-mpg EPA

Figure 5. (a) The ratio of thermoelectric power-generation efficiency ε and the Carnotefficiency εcarnot versus the figure of merit ZT for a thermoelectric power generator operatingbetween 30C and 500C. (b) Coefficient of performance (COP) at room temperature as afunction of ZT and temperature gradient.The COP of mechanical air-conditioning systemsover a similar temperature gradient is also shown.

Figure 6. Dependence of fuel economy(FE) in miles per gallon (mpg) onelectrical load (EL) at the alternator fora representative midsize truck atvarious weights.The symbols are datasimulated by Overdrive, and the linesare guides for the eye.



composite rating; a 10% fuel economy im-provement would mean a 1.88-mpg im-provement. The mass penalty shouldgenerally not exceed 5% of the 1.88-mpggain. Data in Figure 6 indicate an 1190lb/∆mpg mass penalty; the total mass of aTE generator for this truck should bebelow 112 lb.

Challenges for ThermoelectricAutomotive Waste-Heat Recovery

Despite many new materials discover-ies in the past decade, the development oflarge-scale automotive TE waste-heat re-covery technologies remains extremelychallenging. There are major challenges inthe areas of materials, processes, and inte-gration. In order to obtain higher thermal-to-electrical energy-conversion efficiencyand coefficient of performance, one needshigher-ZT materials that cover a broadtemperature range. The thermal stabilityof new materials should be addressed andevaluated. Additionally, there have beenfew studies on assessing TE performanceat a module level using the advanced TEmaterials, designing optimum automotiveheat exchangers, integrating TE subsys-tems into vehicle electrical power man-agement, and examining fuel economyimpact at a vehicle level. The great uncer-tainty in material, module, and subsystemcosts, and in the OEM market size, is alsoa major factor inhibiting this technologydevelopment. Until both performance andcosts are better understood, the ability toselect the best thermoelectric materials forautomotive waste-heat recovery remainsdifficult.

ConclusionsThermoelectrics has proven to be a vi-

able technology for space applications, asdocumented by the SiGe- and telluride-based couples that have accumulated millions of hours of operation in space

without a single thermoelectric-relatedfailure. Radioisotope thermoelectric genera-tors with higher specific power wouldresult in more on-board power for thesame RTG mass, or less RTG mass for thesame on-board power. This could enablethe use of high-power-demand instru-ments or increase the mass available forthe science payload.

The requirements for the choice of TEmaterials used in advanced couples basedon new TE materials under development(and described elsewhere in this issue) arederived from the development of success-ful RTGs over the past 55 years. They include temperature stability and subli-mation rates, thermomechanical proper-ties, and stable thermoelectric propertiesas a function of time and temperature.

It is also clear that TE waste-heat recov-ery technology could potentially offer sig-nificant fuel economy improvements. Ifthis is demonstrated feasibly on large-scale applications such as automobiles, asignificant savings in worldwide fuel con-sumption can be achieved by applying thetechnology across the board to conven-tional and hybrid vehicles. However,many challenges remain for large-scaledevelopment. The key to the realization ofthis technology is the continued develop-ment of new materials with increasedthermoelectric efficiency.

AcknowledgmentsPart of the work described in this article

was carried out at the Jet Propulsion Lab-oratory/California Institute of Technol-ogy, under contract with the NationalAeronautics and Space Administration.J. Yang would like to thank F. Stabler formany fruitful discussions, K. Deer for as-sistance with the Overdrive simulation,and J.F. Herbst and M.W. Verbrugge forcontinuous support and encouragement.

The work is also supported in part by theU.S. Department of Energy under corpo-rate agreement DE-FC26–04NT42278.

References1. D.M. Rowe, ed., CRC Handbook on Thermo-electrics (CRC Press, Boca Raton, FL, 1995).2. F. Stabler, “Automotive Applications of HighEfficiency Thermoelectrics,” presented atDARPA/ONR Program Review and DOE HighEfficiency Thermoelectric Workshop (SanDiego, CA, March 24–27, 2002).3. G. Bennet, in Encyclopedia of Physical Scienceand Technology, Third Ed., Vol. 15 (2002) p. 537.4. C.B. Vining, in CRC Handbook on Thermo-electrics, edited by D.M. Rowe (CRC Press, BocaRaton, FL, 1995) p. 329.5. E. Skrabek, in CRC Handbook on Thermo-electrics, edited by D.M. Rowe (CRC Press, BocaRaton, FL, 1995) p. 267.6. J.-P. Fleurial, A. Borschevsky, T. Caillat, D.T.Morelli, and G.P. Meisner, in Proc. 16th Int. Conf.Thermoelectrics (IEEE, Piscataway, NJ, 1997)p. 91.7. T. Caillat, J.-P. Fleurial, and A. Borschevsky,J. Phys. Chem. Solids 58 (1997) p. 1119.8. Q. Shen, L. Chen, T. Goto, T. Hirai, J. Yang,G.P. Meisner, and C. Uher, Appl. Phys. Lett. 79(2001) p. 4165.9. R. Venkatasubramanian, E. Siivola, T. Colpitts, and B. O’Quinn, Nature 413 (2001)p. 597.10. T. Harman, “Quantum Dot SuperlatticeThermoelectric Unicouples for Conversion ofWaste Heat to Electrical Power” presented atthe 2003 MRS Fall Meeting (Boston, MA, De-cember 1–5, 2003).11. T. Caillat, J.-P. Fleurial, G.J. Snyder, and A.Borschevsky, in Proc. 21st Int. Conf. Thermo-electrics (IEEE, Piscataway, NJ, 2001) p. 282.12. D.G. Evans, M.E. Polom, S.G. Poulos, K.D.Van Maanen, and T.H. Zarger, SAE TechnicalPaper No. 2003-01-0085 (Society of AutomotiveEngineers, Warrendale, PA, 2003).13. Kolbenschmidt Pierburg AG, “Thermo-management: possible contest strategies,”www.kolbenschmidt.de/pdfdoc/elektrische_kuehlmtelpumpen_e.pdf (accessed February2006).


Documents

Thermoelectric Materials, Phenomena, and Applications: A Bird's