Thermoelectric properties of the semiconducting Chevrel phase Mo2Re4Se8

SOO22-3697(97)00183-2

Pergamon

THERMOELECTRIC

PROPERTIES OF THE SEMICONDUCTING PHASE Mo2Re4Seg T. CAILLAT*

Jet

I PA>\. C/w,,, So/,
CHEVREL

and J.-P. FLEURIAL

Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109 USA

(Received 28 March

1991; accepted 26 June 1997)

Abstract-We have prepared and measured the electrical resistivity, Seebeck coefficient, and thermal conductivity of the Chevrel material Mo2Re&3e8 in the 300-IGOCI K temperature range. The results show that Mo2Re4Ses is a semiconductor with a relatively low thermal conductivity. The semiconducting nature of this compound is consistent with theoretical predictions based on the number of valence electrons per [Mot,] cluster. The crystal structure of Chevrel phases contains several voids of different sizes and shapes which can be filled by a variety of atomic elements. The filling of these interstices may be used as a means to achieve low thermal conductivity values by increasing the phonon scattering. This represents an attractive possibility considering that low thermal conductivity is one of the requirements for a good thermoelectric material. Various approaches to determine the potential of Chevrel phases for thermoelectric applications are presented and discussed. 0 1998 Elsevier Science Ltd. All rights reserved. Keywords:

A. chalcogenides. A. semiconductors, D. transport properties

1. INTRODUCTION Ternary

chalcogenides

of formula

atoms and dividing the result by the number of MO atoms (M = Cu.

M,Mo~(~

Ag, Ni, Fe, rare earth, etc. and X = S, Se, or Te) have been

large critical magnetic

structure for 4 valence electrons

properties with

cluster [4]. The cluster-VEC

fields [I]. They were first syn-

from

thesized by Chevrel er nl. in 1971 [2] and therefore are often referred

to as Chevrel compounds.

The ternary

Mo6Xx unit is illustrated gens arranged

in Fig. 1 and consists

the semiconducting

of an

mixed-metal

cube. The rhombohedral

additional

MMobX8 ternary

feature of the Chevrel

compounds.

variety of elements

cluster-VEC

One

phases is the great

with a wide range of mass, size,

adding the valence electrons of the atoms M to the valence electrons of MO, by subtracting the number of the octets of the

such as Mo2RedSes and

number of 4. above, these compounds

mainly investigated for their superconducting

have been properties.

A recent interest in studying and developing new classes of thermoelectric materials has been motivated by the identification of several new promising materials, includ-

atom in the cluster. This quantity, often referred to as ‘cluster-valence-electron’ (cluster-VEC) is calculated by

‘fill’

was not

supporting the idea of an energy gap in

As we mentioned

and composition.

required to

‘magic’

the band structure of the Chevrel phases with a ‘magic’

Many of the physical properties of the Chevrel phases appear to be linked to the number of electrons per MO

electrons

the

Values of 4 are also attained in

cluster compounds

semiconductors,

flexibility of the crystal structure which allows all the constituents of the ternary phases to be substituted by a electronegativity

value of 4 for

particular,

ModRu2Ses [5,6], and these compounds were found to be

contains channels where additional metal atoms can be forming

In

nature of this compound

confirmed experimentally.

Chevrel phase consists of a stacking of MO&~ units and inserted,

compounds.

number of 4 is met in the compound Cu4M06S8, although

‘cluster’ surrounded by eight chalco-

in a distorted

per MO atom in the

of Chevrel phases varies

3.3 (Mo6Se8) to a maximum

M,Mo6Se8

phases have structures closely related to those of binary molybdenum chalcogenides Mo6X8 (X = S, Se, Te). The [Mo6] octahedron

numbers

between 3.3 and 4 [3]. In addition, band structure calculations results predicted an energy gap in the electronic

known since the 1970s. They have attracted considerable interest because of their superconducting

[3]. Chevrel phases are formed for cluster-VEC

ing skutterudites

[7, 81. Studies on skutterudites

focused on binary compounds that thermoelectric

initially

but it became soon clear

figures of merit (ZT) greater than

those obtained for state-of-the-art thermoelectric materials would not be achieved if the lattice thermal conductivity could not be significantly reduced. One of the features of the skutterudite compounds is the presence of holes in the structure. These holes can be filled by various atoms which have been shown to produce an important phonon scattering, resulting in a significant

chalcogen

*Corresponding author. II39

1140

T. CAILLAT and J.-P. FLEURIAL

0

polished

S,SeorTe

using

Microprobe

0 MO

standard

analysis

metallographic

techniques.

(MPA) was performed

samples to determine

on these

their atomic composition

JEOL JXA-733 electron superprobe

using a

operating at 20

X

IO’ V of accelerating potential and 25 X 10e9 A of probe current. Pure elements were used as standards and X-ray Fig. I. Illustration of the MohXg (X = S, Se, Te) building block of the rhombohedral Chevrel phase structure.

intensity

measurements

conducted

of peak and background

by wavelength

dispersive

were

spectrometry.

The

density of the samples was calculated from the measured reduction

in lattice thermal conductivity

thermoelectric

properties

[7-IO].

and superior

As we mentioned

weight and dimensions of the samples. Samples in the form of disks (typically a I .O mm thick,

before, the Chevrel phases also contain interstices which can be filled with a variety of atoms which may

6.35 mm diameter

effectively scatter phonons like in skutterudites, resulting

direction)

in low thermal conductivity

measurements.

which is one of the require-

ments for a good thermoelectric. synthesize

some

We have started to

of the presumably

semiconducting

slice) were cut from the cylinders

using a diamond

saw (perpendicular

for electrical

temperature electrical

All samples were characterized by Seebeck

resistivity

Chevrel phases to assess their potential for thermoelectric

resistivity,

applications. To the best of our knowledge, no data exists

to the pressing

and thermal transport property coefficient,

measurements.

Hall effect,

Seebeck

at room

Hall effect

and

High temperature coefficient,

thermal

properties of Chevrel phases. We

diffusivity, and heat capacity measurements were also conducted on selected samples between room temper-

have chosen to start with the compound MozRedSes, and

ature and about 1000 K. The electrical resistivity (p) was

report here on its thermoelectric

measured

on the thermoelectric

properties.

current

apparatus [ 131. The Hall coefficient

2. EXPERIMENTAL DETAILS

of -10400

Single phase,

polycrystalline

by mixing and reacting stoichiometric molybdenum (99.999%), rhenium

of

of MozRe$es

(99.9997%). and selenium (99.999%) powders. The powders were first mixed in a plastic vial using a mixer

(RH) was measured

1.O in a single carrier scheme, by p/n = I/Rg, and n are the densities

of holes and electrons,

where p respec-

tively, and e is the electron charge. The Hall mobility (pLH) was

evacuated

resistivity

were then heated

Gauss. The carrier density was calculated

from the Hall coefficient, assuming a scattering factor of

before being loaded into a quartz ampoule which was and sealed. The ampoules

with a

high temperature

in the same apparatus with a constant magnetic field value

were prepared amounts

samples

using the van der Pauw technique of 100 mA using a special

calculated

from

the

Hall

coefficient

and

the

values of pLH= RH/p. Errors were estimated

at 1473 K for 2 days. In agreement with HGnle et al. [I I],

to be

X-ray patterns showed that, after this first anneal, the

coefficient

samples contained

(a) of the samples was measured on the same samples

a significant amount of /3-MoSe*. A

2 0.5% and

total of three anneals at 1473 K for 2 days each and with

used

intermediate

measurements

crushing

and grinding

obtain single phase materials. were analyzed each anneal.

was necessary

The annealed

by X-ray diffractometry The powders

to

samples

(XRD) after

were then hot-pressed

in

for

-C 2% for the resistivity

data, respectively. electrical

resistivity

and Hall

The Seebeck coefficient and

Hall

coefficient

using a high temperature light pulse tech-

nique [ 141. The error of the Seebeck coefficient measurement was estimated to be less than ? 3%. The heat capacity and thermal diffusivity were measured using a

graphite dies into dense samples, 10 mm long and 6.35 mm in diameter. The hot-pressing was conducted

flash diffusivity technique [ 151.The thermal conductivity

at a pressure

capacity, and thermal diffusivity values. The overall error

between

of about 20000 psi and at temperatures

1123 and 1273 K for about 2 h under argon

atmosphere. XRD analysis was performed at room temperature on a Siemens D-500 diffractometer using CuK, radiation. Small additions of Si powders were made to the samples as an internal standard. Powder X-ray patterns were taken with scan steps of 20 = 0.05“ and counting time of 3 s. The lattice parameters were calculated from the experimental X-ray patterns using a modified Apple-NBS refinement program [ 121. The error in the determination of the lattice parameters was estimated at about 2 0.015. Selected samples cut from the hot-pressed bars were

(h) was calculated

from the experimental

in the thermal conductivity

measurements

density, heat was estimated

to be about +- 10%. 3. RESULTS AND DISCUSSION Some properties of two Mo2RebSe8 are listed in Table I. The density of the samples varies between 94 and 99.7% of the theoretical value, depending on the pressing temperature. MPA results showed that the samples were slightly different in composition, presumably because of the different temperature used for hot-pressing. Sample DA401 was composed of 95% in volume of a

The semiconducting chevrel phase

I. Properties of MozRedSes

Table

MozRedSes

samples at room temperature Sample

Property Units

DA401

DA402

4

9.667 10.738 1123 8.97 8.41 72 0.09 0.91 - 70 24 0.28

9.667 IO.738 1273 8.97 8.95 14 0.23 I .97 65 35 0.85

f65 ~LVK-’

at room

UH

A K g crne3 g cm-’ mfl cm cm? “-1

(‘H

Hot-pressing temperature X-ray density Experimental density Electrical resistivity Hall mobility Hall carrier concentration Seebek coefficient Thermal conductivity Figure of merit

phase

with

a composition

remaining

5%

MO 13,sRe29Se57,5

phase

had

a

s-I

IO*’ cm-’ gV K-’ mW cm-’ K-’ ,o-’ K-I

while the

composition

from

was MO 1s.6Re~6.$e57.8(about 80% in volume) while the phase had a composition

temperature

to

nearly

0 FV K-’ at 1073 K. The results show that Mo*Re$es

Mo1,.sRe2,.sSe54.+ For sample DA402, the major phase remaining

1141

is a semiconductor

in

agreement with previous results [S]. However, significant

MO 13.6Re2s.$!ie57,6. changes in the transport properties

are associated

with

These results show that it is essentially the MO to Re ratio

variations in the MO to Re ratio the samples. For sample

that changes in the samples indicating that there exists an

DA401, most of the sample is composed of a phase close

homogeneity

domain centered around the stoichiometric

compound MozRe$Ses. Similar results were obtained for the compound deviations

Mo4RuzSee

seem

samples.

Indeed,

ductivity

while

temperature

[4]. These

to influence sample sample

DA401

shows

DA402

of the

n-type

is p-type

con-

at room

(see Table I ). The data in Table I suggest

that MozRepSeB

is

a semiconductor

mobility. The high temperature Seebeck coefficient resistivity

with a low carrier

electrical resistivity and

values for Mo2Re4Ses

Figs 2 and 3, respectively. electrical

stoichiometric

the properties

For sample

decreases

are

shown in

DA401,

stoichiometry

while the dominant

DA401, n-type conductivity conductivity

is obtained

sess a more semimetallic nature of MozRerSes

behavior. The semiconducting

is consistent

with the theoretical

prediction of an energy gap in the band structure of Chevrel phases with a cluster-VEC of 4 [4]. Compounds with a cluster-VEC consistent

lower than 4 are metallic. This is

with our results indicating

sample has a more semimetallic

that the MO-rich

behavior, suggesting

decrease

in the energy

ature while the Seebeck coefficient increases, suggesting

tailoring

the energy

that both electrons and holes contribute to the conduction.

useful when considering

the optimization

The maximum

electric

Chevrel

coefficient

value for sample

while p-type

is observed for sample DA402 which pos-

temper-

Seebeck

with increasing

the

to the MozRe4Ses

phase in sample DA402 is slightly MO-rich. For sample

properties

band gap. The possibility

gap with composition of

phases

a of

might be

of the thermoin different

DA401 is -160 PV K-’ at 1000 K. For sample DA402,

temperature ranges and doping levels. Some other semi-

the electrical,

conducting

increasing

resistivity

temperature

decreases while

only slightly

the Seebeck

with

decreases

Chevrel phases are listed in Table 2. It is

interesting to point out that the compound MolRuzTes

is

60 a

70

$

60

.

50

q

50

z% 0 -&-DA401 +DA4Dz

-50

ii4 -

B

30

-100

20 1 w

-150

10

900c

0 0.5

1

1.5

2

2.5

1000/T

3

3.5

4

(K)

Fig. 2. Electrical resistivity versus inverse temperature for two polycrystalline Mo2Re4Ses samples (see text for the exact composition of the samples).

200

300

400

500 T-m

600

700

800

900 1ooo1100

00

Fig. 3. Seebeck coefficient versus temperature for two polycrystalline Mo2Re&es samples (see text for the exact composition of the samples).

1142

T. CAILLAT and J.-P. FLEURIAL Table 2. List of semiconducting Chevtel phases

metallic

Material

Cluster-VEC

Type of conductivity

Reference

MoIRedSw MozRe4Ses Mo4RuzSes MozRe&_,Se, (0 5 x 5 8) Mo2RerSeR_,Te, (0 5 x 5 1.2) Mo4Ru2Se8_,Te, (0 I x 5 1) MoPRuzTeR

4 4 4 4 4 4 4

Semiconductor Semiconductor Semiconductor

[Sl [Sl bl fS1 [Sl NY El

although

its cluster-VEC

MooRuzSes_,
is 4. It was found that

in

the energy gap is reduced

going from the Se-rich materials which are semi-metals

Semiconductor Semiconductor Semiconductor Semiconductor

to Te-rich materials

[6]. A similar result was reported

for Mo2Re4Xs materials, i.e. a reduction in energy gap is observed

going from the sulfides to the selenides

and

phonon scattering and

comparable

only. The values are relatively to

state-of-the-art

low

thermoelectric

materials. The value of a material for thermoelectric

applications

can be evaluated by a quantity called the thermoelectric figure of merit (Z7) defined as ZT = cy*pT/X. The larger

tellurides.

Otis, the better the material’s conversion efficiency. One

The thermal conductivity data for Mo*Re.,Ses samples are shown in Fig. 4. The room temperature value is

of the requirements

24 mW cm-’ K-’

for

sample

DA40 1

and

for a good thermoelectric

therefore a low thermal conductivity. by Yvon [I],

the

material is

As was described

stacking of MO,& units leaves a number

mW cm-’ K-’ for sample DA402. The higher value

of cavities in the lattice which can be filled by a variety of

for sample DA402 is mostly due to the fact that the sample has a higher density than sample DA401 (see

atoms with different sizes. These atoms may be efficient in scattering the phonons as it was established for

Table 1) and also to the fact that the sample presents a semimetallic behavior, resulting in an increased elec-

skutterudite materials [7-IO] and first suggested by Slack [16]. The cavities in the Chevrel structure are

35

tronic contribution Mixed conduction

to the total thermal

conductivity.

occurs for sample DA401 and some

of the heat is conducted by both types of carriers and this electronic

thermal conductivity

thermal conductivity

accounts

for the total

values shown in Fig. 4. For both

samples, the thermal conductivity

decreases steadily with

increasing temperature and varies as T -“.4. This indicates that point defect scattering (resulting from the substitution of Re for MO) leads to a weaker dependence

than

would

be

obtained

b

temperature for

299 309 400 SW 909 799 900 900

acoustic

1ooo11w

Tamparat=a (Q Fig. 4. Thermal conductivity versus temperature for two polycrystalline MolReBes sample (see text for the exact composition of the samples and Table 1 for density values). Values for state-of-art thermoelectric materials are also presented for comparison: BizTeJ and PbTe based alloys and TAGS (Te-Ag-Ge-Se alloys).

Fig. 5. Illustrations showing the positions of tilling atoms in the cavities of the Chevrel crystal structure (after Ref. [I]). The figures repreqt a projection of the crystal structure on the hexagonal (I 120) plane. The large atoms of Pb in PbMo,Ss occupy a cube shape chalcogen hole (a) whereas small Cu atoms can be distributed over I2 different sites in the chalcogen network of Cu,Mo$s (x,,.,, = 4) (b).

The semiconducting

empty

in the binary

compounds

filled

by

atoms

the

M,Mo6Xs.

M

The largest

compounds

of the voids in the Chevrel

structure has nearly a cubic shape formed by eight chalcogen atoms as illustrated in Fig. 5a. Large atoms such as Pb or La can exclusively occupy these large voids with a filling factor limit corresponding

to x-l.

Smaller

atoms such as Cu, Ni or Fe, for example can be inserted in I2 different

smaller holes with irregular shapes in the

chalcogen

channels

as illustrated

in Fig. 5b. For small

atoms, the upper occupancy limit was experimentally found to be corresponding to x = 4. It is also important to point out that the shapes and sizes of the interstices depend on the composition

and degree of filling. The

thermal behavior of M,Mo&s

has been studied by Yvon

[I]. One of the results of these studies is that it appears that the M atoms are systematically not confined in a fixed position, are weakly bound and small atoms in particular are able to move between the I2 different

lattice sites.

The motion of these atoms within the lattice may be particularly

efficient in scattering

low lattice thermal conductivity.

phonons, resulting in

produced

by mass and volume

fluctuations

and the

values are already relatively

low.

If metal atoms can be inserted into the existing voids in the lattice, very low thermal conductivity achieved.

However,

For the first time, the thermoelectric semiconducting

Chevrel,

properties

MozRedSes,

have

of a been

measured. The results confirmed the semiconducting nature of this material predicted by a valence electron count rule per [Moe] cluster. The thermal conductivity values

are relatively

measured

low and comparable

for state-of-the-art

The crystal

structure

to those

thermoelectric

of the Chevrel

materials.

phases

present

cavities which can greatly vary in size and accommodate various atoms ranging from large ones such as Pb to small ones such as Cu. These atoms are not localized

in the

structure and, depending on their size, can move between different

sites. We believe

significant

phonon scattering

that they can produce

a

and result in low lattice

thermal conductivity

Chevrel phases. Several composi-

tions were proposed

to study the impact of the void

fillers on the lattice thermal

conductivity

phases and continue their assessment

of Chevrel

for thermoelectric

applications.

The study of MxMogX8

ternary compounds would therefore be of great interest. In MozRe.$es, most of the scattering of the phonons is thermal conductivity

4. CONCLUSION

such as Mo6Tes and are

in the ternary

II43

chevrel phase MozRe4Ses

values could be

it was found that only very small

amounts of an additional element M such as Sn can be

Acknowledgemenrs-The work described in this paper was carried out at the Jet Propulsion Laboratory/California Institute of Technology, under contract with the National Aeronautics and Space Administration. We thank A. Borshchevsky and D. Singh for stimulating and valuable discussions. The authors would also like to thank Danny Zoltan and Andy Zoltan for thermoelectric property measurements, Paul Carpenter for microprobe analyses, and Jim Kulleck for XRD analyses. This work was supported by the US Defense Advanced Research Projects Agency, Grant No. E407.

introduced in the compound [S]. This might be explained by the fact that the cluster-VEC

is already at 4 and the

bands below the gap are completely filled, preventing the insertion of additional M elements. Other approaches must thus be considered

to study the effect of void fillers

on the thermal conductivity phases. In the Cu ,Mo$s

of semiconducting

Chevrel

system, compositions

with x =

4 have been reported [ 171 and, assuming that the Cu has an oxidation Cu:Mo~S~-.

state of + I, the ionic formula would be The ‘magic’ cluster-VEC

number of 4 is

met in this compound, which should be semiconducting. Another possibility to explore is to prepare semiconducting Chevrel phases combining void fillers and a mixed cluster (Mo,_,M’ ,, where M’ is a metal) to achieve ‘magic’ VEC number of 4. For example, an hypothetical compound CuzMolResSes

would have a cluster-VEC

4 and should be a semiconductor. would be particularly

of

Such a compound

attractive because phonons would

presumably be scattered by both point defects and the void fillers. The possibilities are obviously numerous, but efforts should focus first on determining the impact of the void fillers on the lattice thermal conductivity of semiconducting Chevrel phases. This constitutes the next step in determining

thermoelectric

the

potential

applications.

of Chevrel

phases

for

REFERENCES

1. Yvon. K., in Currenr

Topics in Materiufs Science, Vol. 3, ed. E. Kaldis. North-Holland, Amsterdam, 1979. 2. Chevrel, R., Sergent, M. and Pringent, J., J. So/id Srafe Chem., 1971.3.515. 3. Yvon, K. and Paoli, E., Solid State Communication, 1977, 24.41. 4. Matheiss, 1760.

L.F.

and

Fong, CF..

Phys. Rev., 1977, BE.

5. Perrin, A., Sergent, M. and Fischer, O., Mar. Res. Bul.. 1978, 13, 259. 6.

Perrin, A., Chevrel, R., Sergent, M. and Fischer, 0..

J. Solid

43. 7. Fleurial, J.-P., Borshchevsky, A., Caillat, T., Morelli, D.T. and Meisner, G.P., in Proceedings ofrhe XV fnternafional Conference on Thermoelectrics. Pasadena, CA, USA, IEEE Catalog Number 96TH8 169, 1996, p. 9 I, 8. Sales, B.C., Mandrus, D. and Williams, R.K., Science, 1996. State Chem., 1980.33,

272, 1352. 9. Morelli, D.T. and Meisner, G.P., J. Appl. Php., 3777.

1995. 77.

IO. Nolas. G.S., Slack, G.A., Morelli, D.T.. Tritt, T.M. and Ehrlich, A.C., J. Appl. Phys., 1995, 79, 4002. I I. Hiinle, W.. Flack, H.D. and Yvon, K., J. So/id State Chem.. 1983,49, 157. 12. Evans, H.T., Appleman,

D.E. and Handwerker, D.S., Report # PB216188, US Dept of Commerce, National Technical Information Center, 5285 Port Royal Road. Springfield, VA 22151, 1973. 13. McCormack, J.A. and Fleurial. J.-P.. in Modern

1144 Perspectives on Thermoelectrics

T. CAJLLAT and J.-P. FLEURIAL and Related Materials.

MRS Symp. Proc. 234. Materials Research Society, Pittsburgh, PA, 1991, p. 135. 14. Wood, C., Zoltan. D. and Stapfer, G., Rev. Sci. Instrum., I985,56 (5). 7 19. J.W., Wood, C., Zoltan. A. and IS. Vandersande,

Whittenberger, D., Thermal Conductivity. Plenum Press, New York, 1988,445. 16. Slack, G.A., Thermoelectric Handbook, ed. M. Rowe. CRC, Boca Raton, FL, 1995, p. 407. 17. Johnston, DC., Shelton, R.N. and Bugaj, J.J., Solid State Communication,

I977,21,

949.