• Aucun résultat trouvé

Pressure dependence of the thermoelectric power of TTF-TCNQ

N/A
N/A
Protected

Academic year: 2021

Partager "Pressure dependence of the thermoelectric power of TTF-TCNQ"

Copied!
7
0
0

Texte intégral

(1)

HAL Id: jpa-00209491

https://hal.archives-ouvertes.fr/jpa-00209491

Submitted on 1 Jan 1982

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Pressure dependence of the thermoelectric power of TTF-TCNQ

C. Weyl, D. Jérome, P.M. Chaikin, K. Bechgaard

To cite this version:

C. Weyl, D. Jérome, P.M. Chaikin, K. Bechgaard. Pressure dependence of the thermoelectric power of TTF-TCNQ. Journal de Physique, 1982, 43 (7), pp.1167-1172. �10.1051/jphys:019820043070116700�.

�jpa-00209491�

(2)

Pressure dependence of the thermoelectric power of TTF-TCNQ

C. Weyl, D. Jérome, P. M. Chaikin (*)

Laboratoire de Physique des Solides, Université de Paris-Sud, 91405 Orsay, France

and K. Bechgaard

H. C. Oersted Institute, Universitetsparken 5, DK 2100 Copenhagen, Denmark (Reçu le 2 avril 1981, révisé le 2

mars

1982, accepté le 22

mars

1982)

Résumé.

2014

Nous

avons

mesuré le pouvoir thermoélectrique de TTF-TCNQ dans

un

intervalle de température

allant de 300 à 4 K pour des pressions comprises entre

une

atmosphère et 15 kbar. Le pouvoir thermoélectrique,

à la température ambiante, décroît de ~ - 30 03BCV/K à ~ - 10 03BCV/K dans

ce

domaine de pression indiquant que la conductivité relative de la chaîne de TTF par rapport à celle de la chaîne de TCNQ croît

en

même temps que la

pression. Dans le domaine où la transition métal-isolant

a

lieu, le pouvoir thermoélectrique change de signe; la plus haute température de transition correspond donc à la disparition des porteurs de charges

sur

les deux chaînes.

Lorsque la pression augmente cette transition affecte de plus

en

plus la chaîne de TTF par rapport à la chaîne de

TCNQ.

Abstract.

2014

We have measured the thermoelectric power of TTF-TCNQ in the temperature region from 300 K- 4 K for pressures from atmospheric to 15 kbar. The

room

temperature thermopower decreases from ~ - 30 03BCV/K

to ~ - 10 03BCV/K in this pressure range indicating that the relative conductivity of the TTF chain to the TCNQ

chain increases with pressure. In the temperature regime of the metal-insulator phase transition

we

find that the

sign of thermopower changes. The highest temperature transition therefore corresponds to the

«

freezing-out

»

of carriers

on

both chains. As pressure increases this transition increasingly effects the TTF chain relative to the

TCNQ chain.

Classification

Physics Abstracts

72.15J

-

72.20P

Measurements of the properties of highly conduct- ing quasi-one-dimensional organic crystals

as

a func-

tion of hydrostatic pressure have proven particularly

useful [1-3]. This results from the high compressibility

of the crystals and the fact that pressure is the only

method for continuously varying the most funda-

mental quantities of physical interest in these narrow-

band systems namely : the longitudinal and transverse

bandwidths and the cation-anion charge transfer.

Most of these compounds undergo metal-insulator transitions due to the instability of one-dimensional electronic systems against Peierls’ distortions [4].

Application of pressure strongly effects this phase

transition. In the

case

of TTF-TCNQ pressure studies have indicated

a

complicated set of phase transitions

(*) Permanent address : Department of Physics, Univer- sity of California, Los Angeles, California 90024, A. P. Sloan Foundation Fellow, work partially supported by NSF

DMR-79-08560.

which at high pressure (19 kbar) evolve to a single

first order commensurate transition [5]. More recently

studies

on

(TMTSF)2-PF6 have shown that the Peierls transition can be suppressed completely by the application of 10 kbar and that the material then becomes superconducting [6].

The organic charge transfer compound which has

been most widely studied is TTF-TCNQ. At atmo- spheric pressure TTF-TCNQ remains metallic from above 300 K to 53 K at which temperature it under-

goes the first of at least three transitions which even-

tually leave it insulating in its low temperature ground

state [7]. The phase diagram as

a

function of pressure has been obtained from conductivity measurements and more recently from neutron scattering studies [ 8].

In order to further explore the metallic state as

well as the phase diagram of TTF-TCNQ under

pressure, we have measured the thermoelectric power.

Along with the Hall effect [9] the thermopower is capable of distinguishing the sign of the charge

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:019820043070116700

(3)

1168

carriers. This is particularly important in linear charge transfer materials where both chains are

conducting with carriers of opposite sign [10]. It

allows the investigation of the relative conductivity

contributions of the two chains and in the case of metal-insulator transitions provide information about which chain is predominantly involved.

The thermopower was measured in an apparatus which allows simultaneous measurement of four

probe conductivity. The apparatus is described in detail elsewhere [11]. The temperature gradient is

established through the pressure medium, in this

case isopentane. The pressure was measured with a

manganin strain gauge, but often the pressure depen-

dence of the conductivity of TTF-TCNQ served as a

very good pressure gauge.

In figure 1 we show the conductivity and thermo- power for TTF-TCNQ measured at room tempera-

ture 291 K as a function of pressure. The conducti-

vity is in very good agreement with previous measu-

rements.

The absolute thermopower was measured on four

different samples at separate times to study different

pressure regimes as

a

function of temperature. In

Fig. 1.

-

a) Normalized conductivity and b) absolute ther- mopower

as a

function of pressure at

room

temperature;

solid line is guide to the eye only.

figure one, two samples are plotted up to 20 kbar, however, all samples measured fell within the limits of those shown.

The room temperature thermopower falls quite rapidly with pressure attaining one-third of its ambient pressure value by 20 kbar. If TTF-TCNQ were a simple carrier system or if the relative conductivities of the two carriers remained the same this would

imply a change in bandwidth of a factor of 3. The thermopower of a one-dimensional material with a

tight binding band is given by

where p is the number of conduction electrons per

molecule, W is the bandwidth and

T

is the scattering

time. The first term clearly varies as the reciprocal

bandwidth as long as there is no strong variation of p. The second term involves an energy derivative of the scattering time and has been estimated to be

as significant as the first term [10]. However, as far

as the metallic behaviour is concerned at room tempe-

rature for the elastic scattering T can be assumed as having only

a

moderate variation with

8

(kB T/EF 1).

Its energy dependence makes

a

similar contribution to S as the band term : S oc (4 t) - ’, where t is the longi-

tudinal transfer integral. Thus S

oc

W - 1.

For two conducting chains the thermopower is

the conductivity weighted average or

We would

now

like to explore how much of the observed change in thermopower can be accounted for by the known changes in charge transfer between chains and the bandwidth change as a function of pressure. Within tight binding theory we can estimate

the pressure dependence of the thermopower of each

chain by using equation (1). Although the particular

form of the energy dependence of the scattering rate

is important for this calculation it should typically

vary as the characteristic energy for the system (sF)

and hence as the « band term » or first term in equa- tion (1). Conwell [12] has argued that for one phonon

processes the scattering term is in fact identical to the band term. For two phonon (or libron processes)

it has the same dependence on bandwidth and slower

dependence on charge transfer [13, 12].

The neutron scattering data [14] shows

a

charge transfer of 0.59 electron/molecule at one bar and

0.616 electron/molecule at 5 kbar or

and confirms [15] an average value of 0.90 %/kbar up

to 14.5 kbar. The change in the bandwidth can be

calculated from the compressibility and theoretical

(4)

values for the transfer integral as a function of lattice

constant [16,

,

171.

.

This yields variations for a ln W

of 2-3 %/kbar for the TCNQ chain. A less reliable method is fitting the pressure dependence of the

Drude edge which gives OP

= -

1 %/kbar [ 18].

The initial dependence of thermo power on pres-

sure can be obtained for each chain by taking the logarithmic derivative of equation (1), with respect

to pressure. We take the band term as representing

the functional dependence of the total thermopower

on p and W. This will produce the most rapid varia-

tion for each chain. Using an initial charge density

of p

=

0.59 electron/molecule we have :

The charge transfer term then reliably gives a

decrease of 2.3 %/kbar. The bandwidth contribution is from - 1 to - 3 % kbar and the total reduction from these effects is

- -

(3-5) %/kbar. From the experimental data of figure 1 the initial pressure

dependence is

- -

10 %/kbar. The conclusion is that a uniform change of the properties of the two

chains is not sufficiently rapid to explain the obser-

vations in the low pressure domain.

The conductivity of the TTF chain must be increas-

ing relative to that on the TCNQ chain. Since

the signs of the thermopowers of the two chains are opposite from equation (1). The TCNQ chain domi- nates the conductivity at room temperature and atmospheric pressure. As pressure increases the con-

ductivity of the TTF chain becomes relatively more important. This same conclusion was reached from

previous studies of the Hall effect in TTF-TCNQ [9].

One theory postulated before these experiments

were done attempted to predict the thermopower

as

a

function of pressure in terms of

an

electron-electron

scattering model. However, the model could not forsee the large relative conductivity change between

the two chains which dominate the pressure depen-

dence [J9].

If we look at the change in thermopower up to 8 kbar

we see

that it has been reduced by 50 %.

About 30 % of this change is due to the bandwidth and charge transfer term already discussed. Previous estimates of the relative conductivity at atmospheric

pressure show O’Q 1’-1 5 Op. Assuming that the thermo- powers of SQ and SF are comparable the ratio of the conductivities at 8 kbar is UQ/ CF - 2.5. If the change

were entirely associated with the more rapid increase

in bandwidth of TTF compared with TCNQ the

ratio would be

even

smaller due to the smaller TTF

thermopower. A similar analysis of the Hall effect data also gives - 2 at 8 kbar [9].

Fig. 2.

-

Absolute thermopower

as a

function of tempera-

ture for several values of pressure. (A - 12.5 kbar, e - 8 kbar,

x -

4.6 kbar, D - 2.5 kbar, 0 -1 kbar).

The temperature dependence of the thermopower

at various pressures is shown in figure 2. In the temperature region above 60 K the behaviour is

quite similar at all pressures shown. At high tempe-

rature (from 100 K-300 K) the thermopower is small

and approximately linear in temperature indicating

the metallic state. Between 100 K-60 K there are

deviations from this simple behaviour which pro-

bably arise from the competition between the sign

of the thermopower of the two chains. The relative conductivities are changing with temperature with the TTF chain becoming more important. This change in relative conductivity may arise simply from

different temperature dependent mobilities or from fluctuation effects preceding the metal-insulator transition [20]. Either the thermopower or the conduc-

tivity on the TCNQ chain is decreasing with respect

to the TTF chain.

At low temperatures (below 20 K) the rapid increase

in the thermopower to large values indicates the low temperature insulating state. Note that the low tem-

perature sign is negative for all these pressures. Both

signs have been previously found in samples from

different sources [21].

The most interesting temperature region is that of the cascade of phase transitions from - 60 K to

-

20 K [7]. Figure 3 shows

our

results in this transi-

tion region in more detail. At 1 kbar the thermo-

power is similar to what has been observed at 1 atm.

(5)

1170

Fig. 3.

-

Absolute thermopower

as a

function of tempera-

ture in the region of the metal-insulator transition for several pressures.

[21 ]. At 1 atm. the thermopower is flat with

a

value

-

+ 30 pV/K between 50 K and 40 K and then monotonically rises to high values in either positive

or negative

sense

depending on sample origin. In

the present case the o flat » region is from 43 K to

33 K and is followed by

a

positive dip before it becomes

strongly negative and semiconducting at - 20 K.

The usual interpretation of the ambient pressure

thermopower is that there is an upper transition

(TH - 53 K) which predominantly effects the TCNQ

chain [10]. As we have seen the TCNQ chain has a negative thermopower and when these carriers are

frozen out by the opening of a gap the thermopower quickly becomes

more

positive. Between 50 K and 40 K the conductivity is dominated by carriers

on

the TTF chain which has positive thermopower.

Finally at - 40 K

a

well defined gap develops on

the TTF chain and the material is a traditional semi- conductor. The thermopower then becomes large

and its sign is determined by what impurities are present to slightly shift the Fermi energy closer to the valence or conduction band. ESR susceptibility

measurements have led to similar conclusions con-

cerning the claims involved in the transitions [22].

This picture is not what one finds in the inter-

pretation of the X-ray and neutron scatterings stu-

dies. The model of Bak and Emery suggests that the 53 K transition involves a distortion associated with

one order parameter (usually and probably mista- kenly associated with one chain-TCNQ) with trans-

verse order period 2

a.

A second transition at 49 K associated with a second order parameter on the

TTF chain with incommensurate transverse perio- dicity and a third transition at 38 K in which both order parameters lock at a commensurate value of wavevector transverse to the conducting axis namely

4 a [23]. Transport and susceptibility studies tend to show electronic gaps forming at only the first and

last transitions although there is

a

small anomaly

at 49 K in the resistivity [24].

For the present

we

will stay with the conventional

interpretation of the thermopower. As pressure is increased the temperature of the upper transition remains practically constant. Below this transition however, the thermopower becomes increasingly more negative as pressure is increased, crossing zero at

-

7 kbar. This implies that the carriers are changing

from predominantly TTF at low pressure to predo- minantly TCNQ at 12.5 kbar. If the upper transition

were

previously associated with TCNQ at 1 atm., at 12.5 kbar it must be associated with TTF.

A consequence of this model is that it would be necessary for the lower transition of the TTF chains to « pass through » the TCNQ chain transition. The

problem is that such

a

picture would involve the

crossing of two phase transition lines with four

phases meeting at

one

point in P-T space. This is forbidden for a one component system by the Gibbs phase rule [25]. The only tenable solution is that the

simple decomposition of the transitions as being

associate with each chain is incorrect. The order parameter for the upper transition must involve distortions and gaps

on

both conducting chains but

not equally. At low pressure the TCNQ is mainly

effected and at high pressure the TTF is mainly

effected.

Again there is

no

evidence of an intermediate transition from the thermoelectric power at different pressures. The lower transition is apparent in the data shown at 1 kbar and at 12.5 kbar. For many of the other pressures it is clear that immediately below

the upper transition the compound is not insulating

but becomes insulating (from the thermopower) over

a

temperature region which extends over 20 K without any obvious transitions. The thermopower

in the semiconducting state is given approximately by

where we merely wish to point out the simplest

variation with gap A and temperature. By analogy

with the resistivity, if we wish to look for the pre-

sence of a phase transition, we must look for rapid changes of A with temperature. For resistivity this is

done by Y looking g at O(IIT) ðlnR For the thermo power the

a(1/T) p

analogous g derivative is as

O(IIT)

Since we have taken both resistivity and thermo- power measurement

on

each sample at each pressure

we can plot the phase diagram as obtained by the approximate derivative of the two measured quanti-

ties. The phase diagram which results is shown in

figure 4. The resistively defined curve is in good

agreement with previous measurements. The thermo-

power defined curve is similar except for the region

between 2.5 kbar and 8 kbar where no lower transi-

(6)

Fig. 4.

-

Phase diagram of TTF-TCNQ obtained from the thermopower and resistivity anomalies. Circles

-

from

analysis of resistivity data. Squares

-

from thermopower

data.

tion has been observed as a peak in .O(IIT) . This

is the pressure region in which the upper transition is crossing over from predominantly TCNQ to pre-

dominantly TTF and hence at these pressures the transition affects both chains comparably. The non-

observation of the lower transition may simply be

an experimental artifact due to the fact that neither chain dominates the transport. On the other hand the fact that the upper transition is comparable

on

both chains may change the nature of the lower transition which is still observed in resistivity.

Recent neutron scattering studies have revealed a

quite complicated phase diagram for pressures up to 5 kbar [26]. These studies indicate that the transition

temperature associated with the TTF chains and

a

longitudinally polarized CDW increases with pres-

sure. This is not in disagreement with thermopower

measurements presented here. Below 54 K the beha- viour of the absolute thermopower in the range 3.5-8 kbar does not allow

an

accurate determination of the lower transition as already mentioned. Addi-

tionally they find that at 5 kbar there is a single phase

transition in possible agreement with the thermo- power measurements. The resistively determined lower

phase transition is not seen.

The suggestion was made that at 5 kbar TTF-

TCNQ is very similar to TSeF-TCNQ both having

the same value of 2 k and

a

single neutron deter-

mined phase transition. While our data suggests that the conductivities of both chains are comparable

at the transition temperature for the two compounds,

the thermopowers measured through this region are quite different. TSeF-TCNQ shows a large gap and hence large thermopower forming directly below the

transition. TTF-TCNQ retains a small thermopower

for - 20 K below the transition.

In conclusion the thermopower measurements in the metallic state show that the conductivity of the

TTF chain increases faster with pressure than that of the TCNQ. At ambient conditions the ratios are

aQl aF - 5 while at 8 kbar the ratio is O’Q/O’F -...; 2.5.

In the region of the metal-insulator transitions the effect of pressure is to change the upper transition from predominantly on the TCNQ chain at 1 atmo- sphere to predominantly

on

the TTF chain at 10 kbar.

This implies a more complicated phase diagram and requires the order parameters to involve distortions

on both chains. The usual simplified picture of

Peierls distortions

on

each chain associated with different transition temperatures is no longer appro-

priate.

We would like to acknowledge useful discussions with S. Barigid, M. Weger, R. Comes, S. Megtert,

J. P. Pouget, E. Conwell and P. Seiden.

References

[1] JÉROME, D. and WEGER, M., Chemistry and Physics of One-Dimensional Metals, H. J. Keller editor

(Plenum Press, N. Y.) 1977.

[2] JÉROME, D., Proceedings NATO Advanced Study Insti-

tute

on

Low Dimensional Solids, Portugal, 1979,

edited by L. Alcacer (D. Reidel, N. Y.) 1980.

[3] COOPER, J. R., WEGER, M., JÉROME, D., LE FUR, D., BECHGAARD, K., BLOCH, A. N. and COWAN, D. O., Solid State Commun. 19 (1976) 749.

[4] DENOYER, F., COMÈS, R., GARITO, A. F. and HEEGER,

A. J., Phys. Rev. Lett. 35 (1975) 445 ;

COMÈS, R., SHAPIRO, S. M., SHIRANE, G., GARITO,

A. F. and HEEGER, A. J., Phys. Rev. Lett. 35 (1975) 1518 ;

WEYL, C., ENGLER, E. M., BECHGAARD, K., JEHANNO,

G. and ETEMAD, S., Solid State Commun. 19

(1976) 925.

[5] FRIEND, R. H., MILJAK, M. and JÉROME, D., Phys.

Rev. Lett. 40 (1978) 1048.

[6] JÉROME, D., MAZAUD, A., RIBAULT, M. and BECH- GAARD, K., J. Physique-Lett. 41 (1980) L-95.

[7] FERRARIS, J. P., COWAN, D. O., WALATKA, V. V. and

PERLSTEIN, J. H., J. Am. Chem. Soc. 95 (1973) 948;

COLEMAN, L. B., COHEN, M. J., SANDMAN, D. J., YAMAGISHI, F. G., GARITO, A. F. and HEEGER,

A. J., Solid State Commun. 12 (1973) 1125 ; Highly Conducting One-dimensional Solids, edited by

J. T. Devreese, R. P. Evrard and V. E. Van Doren

(Plenum Press, N. Y.) 1979.

[8] FINCHER Jr., C. R. and SHIRANE, G., COMÈS, R. and GARITO, A. F., Phys. Rev. B 21 (1980) 5424.

[9] COOPER, J. R., MILJAK, M., DELPLANQUE, G., JÉROME, D., WEGER, M., FABRE, J. M. and GIRAL, L.,

J. Physique 38 (1977) 1097.

(7)

1172

[10] CHAIKIN, P. M., KWAK, J. F., GREENE, R. L., ETEMAD, S. and ENGLER, E., Solid State Commun. 19

(1976) 1201 ;

CHAIKIN, P. M., GREENE, R. L., ETEMAD, S. and ENGLER, E., Phys. Rev. B 13 (1976) 1627.

[11] CHAIKIN, P. M., WEYL, C. and JÉROME, D., Rev. Sci.

Instrum. 52 (1981) 1397.

[12] CONWELL, E. M., Phys. Rev. B 19 (1979) 2409, also CONWELL, E. M., Proc. NATO Advanced Study Institute

on

Low Dimensional Solids, Portugal 1979, edited by L. Alcacer (D. Reidel, N. Y.) 1980 and CONWELL

E. M., Phys. Rev. B 22 (1980) 1761.

[13] GUTFREUND, H. and WEGER, M., Phys. Rev. B 16 (1977) 1753 ;

GUTFREUND, H., WEGER, M. and KAVCH, N., Solid State Commun. 27 (1978) 53 ;

WEGER, M., Proc. NATO Advanced Summer Institute

of Low Dimensional Solids, Portugal, 1979, edited by L. Alcacer (D. Reidel, N. Y.) 1980.

[14] MEGTERT, S., COMES, R., VETTIER, C., PYNN, R. and GARITO, A. F., Solid State Commun. 31 (1979)

977.

[15] MEGTERT, S., COMÈS, R., VETTIER, C., PYNN, R. and GARITO, A. F., Solid State Commun. 37 (1981)

875.

[16] DEBRAY, D., MILLET, R., JÉROME, D., BARI0160I0106, S., FABRE, J. M. and GIRAL, L., J. Physique Lett. 38 (1977) L-227-231.

[17] COOPER, J. R., Phys. Rev. B 19 (1979) 2404.

[18] WELBER, B., SEIDEN, P. E., GRANT, P. M., Phys. Rev.

B 18 (1978) 2692.

[19] SEIDEN, P. E. and GRANT, P. M., Proc. of the Inter-

national Conference

«

Quasi One-Dimensional Con- ductors I » (Dubrovnik), edited by S. Bari0161i0106,

A. Bjeli0161, J. R. Cooper and B. Leonti0107 (Springer- Verlag, Berlin, Heidelberg, N. Y.) 1979.

[20] HORN, P. M. and RIMAI, D., Phys. Rev. Lett. 36 (1976)

809.

[21] CHAIKIN, P. M., KWAK, J. F., JONES, T. E., GARITO,

A. F. and HEEGER, A. J., Phys. Rev. Lett. 31

(1973) 601.

[22] TOMKIEWICZ, Y., Proc. of NATO Advanced Summer Institute of Low Dimensional Solids, Portugal, 1979, edited by L. Alcacer (D. Reidel, N. Y.)

1980.

[23] PER BAK and EMERY, V. J., Phys. Rev. Lett. 36 (1976)

978.

[24] CHU, C. W., HARPER, J. M. E., GEBALLE, T. H. and GREENE, R. L., Phys. Rev. Lett. 31 (1973) 1491.

[25] LANDAU and LIFSHITZ, Statistical Physics (Pergamon Press, N. Y.) 1969, p. 275.

[26] MEGTERT, S., COMÈS, R., PYNN, R., VETTIER, C. and GARITO, A. F., Summer Institute of Low Dimen-

sional Solids, Portugal, 1979, edited by L. Alcacer

(D. Reidel, N. Y.) 1980.

Références

Documents relatifs

All the parameters describing the ESR spectra of the salts under pressure, especially the exchange frequencies (Fig. 7) and the activation energies (Figs. 8 and 9),

il and Institute of Solid State Physics (Fig. The TEP is very sensitive to temperature under rather low magnetic field, following the field dependence of the superconducting

For liquid Cu-Sn alloys [13] a detailed comparison between the Faber-Ziman theory of the thermopower and the experimental data is possible because the three

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

Water at the cavitation limit: Density of the metastable liquid and size of the critical bubble.. Infra-red imaging of bulk water and water–solid interfaces under stable and

(The transition at TLow was not reported by them.) The small pressure dependence of TLow is in.. disagreement with the interpretation suggested

An experimental investigation into the variation of the mechanical properties (yield stress, yield strain, elastic moduli, hardness) from sub-ambient temperature (77 K) to that

Contrary to other organic metals, of which the FS can be regarded as a network of orbits, no frequency combinations are observed for the studied salt, likely due to high