• Aucun résultat trouvé

Nuclear relaxation and electronic correlations in quasi-one-dimensional organic conductors. II. Experiments

N/A
N/A
Protected

Academic year: 2021

Partager "Nuclear relaxation and electronic correlations in quasi-one-dimensional organic conductors. II. Experiments"

Copied!
32
0
0

Texte intégral

(1)

HAL Id: jpa-00246707

https://hal.archives-ouvertes.fr/jpa-00246707

Submitted on 1 Jan 1993

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.

Nuclear relaxation and electronic correlations in quasi-one-dimensional organic conductors. II.

Experiments

P. Wzietek, F. Creuzet, C. Bourbonnais, D. Jérome, K. Bechgaard, P. Batail

To cite this version:

P. Wzietek, F. Creuzet, C. Bourbonnais, D. Jérome, K. Bechgaard, et al.. Nuclear relaxation and

electronic correlations in quasi-one-dimensional organic conductors. II. Experiments. Journal de

Physique I, EDP Sciences, 1993, 3 (1), pp.171-201. �10.1051/jp1:1993123�. �jpa-00246707�

(2)

Classification

Physics

Abstracts

76.60E 74.70K 75.40E 75.50E

Nuclear relaxation and electronic correlations in quasi-one-

dimensional organic conductors. II. Experiments

P. Wzietek

('~~),

F. Creuzet

("*),

C. Bourbonnais

("~),

D.

Jdrome('),

K.

Bechgaard (3)

and P. Batail

(')

(1) Laboratoire de

Physique

des Solides, Universitd de Paris-Sud, bit. 510, 91405

Orsay,

France (2) Centre de Recherche en

Physique

du Solide, Universitd de Sherbrooke, Sherbrooke,

Qudbec,

Canada JIK-2Rl

(3) H. C. Oersted Institute, DK-2100

Copenhagen,

Denmark

(Received 22

April

J992,

accepted

in

final form

4 September 1992)

Abstract. In this paper we collect and detail recent results

conceming

~~Se and '~C nuclear relaxation rate measurements obtained on several representatives of the (TMTSF)~X and (TMTTF)2X series of

charge

transfer salts as a function of temperature and hydrostatic pressure.

From the temperature and pressure

dependences

of

Tj',

for the compounds (TMTTF)2PF6, (TMTTF)2Br,

(TMTSF)~PF~

and

(TMTSF)~Cl04,

we are in a position to extract the nature, the

amplitude,

the statics, the

dynamics

and finally the

dimensionality

of

spin

correlations that characterize the unified

phase diagram

of both series of materials.

1. Introduction.

In this second paper of the

series,

we review and

complete

the

analysis

of the

??Se

and

~~C nuclear relaxation data obtained for

typical

members of the series

(TMTSF)~X

(X

=

PF~

and

CIO~)

and their

sulphur analogs (TMTTF )2X (X

=

PF~

and

Br)

as a function of temperature and

hydrostatic

pressure. The

pronounced quasi-one-dimensional anisotropy

of these series of materials has been for

long

time considered as a

key ingredient

in the succession of

phase

transitions orchestrated

by

either the

hydrostatic

pressure, the anion or the

sulphur-

selenium substitution or even the

application

of a

high magnetic

field

[Ii.

The

spatial anisotropy

of the electronic structure

[2]

promotes the

building

up of

antiferromagnetic

correlations which are

omnipresent

in these systems. An

important

and related energy scale of these correlations is found for

sulphur compounds

at low pressure and it

corresponds

to the

temperature

T~(100 230K)

below which the

resistivity

loses its metallic character to

become

thermally

activated

[3] (see Fig, I).

In

compounds

like

(TMTTF)~Pf~and

(TMTTF )~ASF~, X-ray

measurements

[4] performed

below T~ have shown the existence of

(*) Present address : Unitd Mixte de Recherche CNRS-St-Gobain, 39 Quai L. Lefranc, B-P-135, 93303 Aubervilliers Cedex, France.

(3)

~i

p ~~(n cm)

~~ ~~~

~~"~~~~'~~'

i

(TMTTF),Br

~-i

~~-2

(TMTSF)~PF,

~-3

lO~'

~-5

~~-8

(TMTSF)~CIO,

~~-7

io ioo

T(K)

Fig,

I. Resistivity vs, temperature for

representatives

of (TMTTF )~X and

(TMTSF)2X

series. After [3] and [I Ii.

another energy scale

Tjp(

40 60

K)

that marks the onset of a ID lattice

instability

at the wavevector 2

k~

and which evolves to a true 3D

phase

transition at

Tsp(

lo 20

K)

with a static lattice distortion of the

spin-Peierls (SP)

character

[3-7].

With the

application

of pressure, T~ decreases and for

(TMTTF)~PF~,

lo kbar of pressure is sufficient to

yield

the SP state

unstable with respect to the formation of

antiferromagnetic (AF) long

range order below a Neel critical

point T~

S 17 K

[5, 8].

In these pressure

conditions,

the AF

phase

transition presents

the same features as those found for

(TMTTF)~Br

at ambient pressure where

T~m

loo K

(Fig. I)

and

T~

m 15 K

[3, 5, 9, 10].

At

higher

pressure, the temperature range above

T~

with metallic

properties

increases and

beyond

some critical pressure which

depends

on the

anion

X,

the transport

properties

remain metallic down to the

vicinity

of

T~

where AF critical fluctuations dominate (T

m

Tn). Therefore,

the characteristics of the AF transition

smoothly

evolve toward those found for the selenides series

(TMTSF [X

with

centrosymmetrical

anions

(X

=

PF~, ASF~, TaF~..

and for which

T~

~ 10 K at ambient pressure

(Fig. 1) [8].

As it is well known, the AF state becomes in tum unstable

beyond

a critical pressure P

~

ar 7

(20)

kbar

for selenides

(sulphurs) [8, II, 12] leading

for

(TMTSF )~X

to the observation of

superconduc- tivity

at the lower temperature scale T~ l K [I Ii-

Despite

the absence of firm evidence for

superconductivity

in

sulphur compounds

at very

high

pressure

[8, 13],

the unified

phase diagram

shown in

figure

2 can be constructed in which

antiferromagnetism

acts as the central

pivot

for both series of materials. Here it is w~rthwhile to note that members of

(4)

,

,~T

loo

' '

,

CONDUCTOR

,

~o ' ,

q SF -

~i

~ ~fi

$

lo

l$ SP

§

~ SOW

i

sc

O lo 20 30 40 50

PRESSURE

(kbar)

i i

i i

»

»~

_

» O

« « ji

#

b

C C /

« /

~ W ~ c

~

>

2

~

~

~

t

§ a

>

?

~ ~ t ~

Fig.

2. Unified

phase diagram

and energy scales of (TMTTF )~X and (TMTSF

~X

series as a function of

hydrostatic

pressure or anion X substitution. The

phase diagram

also includes the

sulphur-selenide

compound

(TMDTDSF)~PF~.

the

recently synthetized

mixed

sulphur-selenium

series

(TMDTDSF )~X (X

=

PF~, ...) [14]

are

easily incorporated

to this

phase diagram.

For

example, (TMDTDSF)~PF6 compound

with T~ m loo

K, Tjp

m

20 K and

T~

m

7 K is located in the

right

hand side of the SP-AF

boundary [14].

What catches the attention in this combined

phase diagram

is

certainly

the existence of

multiple

energy or temperature scales. One can also look at these as various

length

scales connected with electronic

(spin

or

charge)

or lattice correlations and which can be of different

spatial

dimensionalities. It is

clearly

of basic

importance

to

bring

out the

microscopic origin

of

the rich structure of this

phase diagram

and

especially

to assess the fundamental role

played by

electronic correlations when

extremely anisotropic

conditions

prevail [15, 16].

This can be

only

achieved

by

successive confrontations between theoretical

predictions

and

experiments.

It is

along

these lines that over the past ten years a great deal of efforts was devoted to the

comparison

of nuclear relaxation rate measurements in these materials with the

T/

'

predictions

obtained in the framework of the

quasi-one-dimensional

electron gas model

[5, 14b, 17-23].

As

extensively

reviewed and discussed in the first paper, hereafter noted as

I,

the great interest in

Tj

' measurements for the

understanding

of the

phase diagram

of

figure

2 follows from the basic

expression

T/

=

2

y( T(gp~ )~

~

ld~q

[A~(~ x

[ (q, w~)/w~, ill

(5)

which is

directly

related to the statics, the

dynamics

and the

dimensionality

of

spin

correlations at all

length

scales. These are connected to the temperature

dependence

of

spin

correlations and thus to the temperature variation of

T/

' itself.

Putting

the

emphasis

on the temperature

profile

of

Tj

rather than in its

frequency

or

field dependence

marks an

important

difference with

respect

to

previous

NMR

analysis

of low dimensional materials. In the

past,

a lot of

interest was indeed devoted to the

study

of

spin dynamics

in these systems as it can be obtained via the field

dependence.

From a series of works initiated

by

Boucher et a/.

[24],

it was found that diffusive

long

wave

length spin

excitations for one-dimensional

spin

chains lead to a

characteristic square root

frequency dependence Tj w/

"~ of the relaxation rate. The effect of diffusion was

previously

noted in connection with the EPR linewidth measurements

[25].

The observation of this law in several

insulating

materials

[26]

demonstrated that from the

T/

measurements it was

possible

to get clear information about the

dimensionality

of the

long

wavelength spin

fluctuations. Later on, Soda et al.

[27], presented

in the same

spirit

a detailed

analysis

of the

frequency dependence

of

'H

and

'DTj'

data of the two-chain

organic

conductor

TTF-TCNQ

and showed that for both chains, uniform

spin

fluctuations are diffusive and one-dimensional in character.

Following

these lines of

thought,

similar studies were

carried on several series of

organic

materials

[28-30] including

the

Bechgaard

salts

[31-32].

In order to establish the existence of a diffusive

component

of

spin

fluctuations from the field variation of

T/

'

however,

one must be in a

position

to

give

a

precise

evaluation of the relative

weight

of the uniform

(Tj

'

[q 0]

and the

staggered (Tj

'

[q

~

Qol) Parts

of the relaxation rate, which necessitates a full temperature

study

of each contribution. At a

given temperature,

diffusion of low-dimensional

long wavelength spin

fluctuations can be assessed

only

if

T/

'

[q

0 w

Tj

'

[q Qol,

which tums out to hold in the

high

temperature domain for

organic

conductors

[5, 21-23, 27, 30].

At T

m 300 K and for a field of 64 koe or so, Stein et al.

[32], reported

the observation of a weak

frequency dependence

for the

'H

relaxation of

(TMTSF)~PF~

and

(TMTSF)~CIO~

at ambient pressure which was attributed to a diffusive component. Within

experimental

error, all the

'~C

and the

??Se

data

presented

in this work

however,

do not show any

frequency dependence

at ambient temperature and for fields up to 9 tesla. A

possible explanation

for this could reside in the

amplitude

of the extemal field which is not

sufficiently large.

In this case, the time scale

w/

' set

by

the electronic

spin-flip

process

involved in a scalar

hyperfine

interaction like

(I),

is too

long

with respect to the characteristic

time scale r~

(w~

r~ ml

)

of processes that do not conserve

spin along

the chains

(e,g.

interchain

tunneling

of uniform

spin fluctuations)

and which cut-off ID diffusion

[24].

If the field necessary to make w~ r~

~ l is very

large,

one

approaches

the so-called collisionless or non-diffusive limit

[33]

where collisions have little influence on q 0

spin

fluctuations and can therefore be

neglected.

In the absence of field

dependence

for the observed

Ti ',

we will

therefore assume that the collisionless limit is achieved so that the whole

T/' analysis presented

here for the

sulphur

and the selenides series focuses

entirely

on the temperature and the pressure

dependences.

This

clearly

contrasts with other

interpretations [31-32]

which associate a diffusive component to the relaxation rate of the

Bechgaard

salts in the very low temperature domain

(T

4 K

).

Although,

in

principle,

relaxation time measurement

techniques

do not need to be

detailed,

in section 2 we comment on some

particular

aspects of our

experimental approach

which have

enabled us to obtain

high quality

data necessary for a

quantitative analysis.

Section 3 presents a collection of nuclear

spin-lattice

relaxation results on

(TMTTF)2PF6, (TMTTF)~Br, (TMTSF)~PF~

and

(TMTSF)~CIO~

as a function of temperature and

hydrosta-

tic pressure and which cover the essential features of the unified

phase diagram

of

figure

2. The

analysis

of the data is made in section 4. There,

predictions

of I for the

quasi-one-dimensional

electron gas model for lD correlation effects, critical behavior and crossover

phenomena

are

(6)

description

of these

complex

systems can come out from such an

analysis.

The last part of section 4 is devoted to a

quantitative analysis

of the temperature and pressure

dependences

of the

magnetic susceptibility

as it can be extracted from the relaxation rate

measurements in the

paramagnetic regime.

Such an

analysis

is based on the well established

empirical relationship (Tj T)~

' oz

X/(T)

at P

= I bar between the measured relaxation rate and

magnetic susceptibility

over a

large high temperature

domain.

Although

this way of

obtaining x~(T)

does not lead to its absolute value, there are clear

advantages

in

using

the NMR

approach

under pressure

compared

to other kinds of

magnetic

measurements.

Indeed,

two standard methods have been used so far for the

spin susceptibility investigation

in

organic

salts : the

Faraday

and the EPR

techniques

and these become less accurate when one reaches

high

pressure. The

Faraday

method

yields

to the absolute value of the electronic

spin

susceptibility,

but it

requires

corrections for the

diamagnetic

contribution from inner-shell

electrons

[34, 35].

Some

Faraday susceptibility

measurements under pressure have been

carried out

by

Forro et al.

[36]

on

(TMTSF)~X

but numerous corrections must be made for the pressure cell and the

transmitting

medium

diamagnetism

contributions. EPR is a more direct method for Xs measurement but its

sensitivity

decreases

rapidly

when the resonance line width increases. Under pressure, EPR is a rather difficult

technique

since the

sensitivity

of the

spectrometer

may vary under pressure and

actually only

few results for

organic

conductors are available

[37].

In contrast to other theoretical

interpretations

of

Xs(T)

in

organic

conductors

[38],

our

approach

is

uniquely

based on the effect of electron-electron interaction in the framework of the ID electron gaz model as

developed

in I. When constant pressure data are

compared

to the

results of a constant volume

theory,

it follows that for a

non-singular quantity

like

x~(T),

it is necessary to evaluate the effect of tliermal

expansion

on the

susceptibility.

For this, the full temperature and pressure

dependences

of

Xs(T)

need to be known and these can be

extracted from its relation to

(Tj T)~

"~

2.

Experimental.

All

experiments

have been

performed

on

single crystal samples prepared by

electrochemical

methods

[20].

For the

sulphur compounds, selectively

'~C 95 % enriched

(TMTTF)2PF6

and

(TMTTF)2Br samples

have been used.

Compared

to

powder samples,

this has the

advantage

of

eliminating

the

Tj anisotropy

and an

exponentional

relaxation can be recorded. Furthermore, the

powder

spectrum is broadened

by

the

Knight

shift

anisotropy.

The

experiments

are

performed

on ??Se in the 45 MHz range for the

(TMTSF)~PF~

and

(TMTSF)2Cl04 compounds.

For the

sulphur compound (TMTTF)~PF~

the central '3C sites have been

observed at 45 MHz in the low temperature range and 33 MHz at

high temperature. Finally, (TMTTF)2Br

data have been

acquired

at 60 MHz

(P

=

I

bar)

and 30 MHz

(P

=

13

kbar).

In contrast to

previous experiments

made on protons, for which

methyl

group rotation effects

strongly

contribute to the relaxation over a

large

temperature domain

[9, 32, 40, 41, 42],

these nuclei

only probe

the fluctuation field

produced by

the electrons since no other molecular motion at the Larmor

frequency

is

expected

to contribute to the relaxation. On the other

hand,

relaxation measurements on

naturally

abundant

??Se present

some technical difficulties due to

a very poor

signal-to-noise ratio, especially

at

high

temperature and under pressure. We are

usually

led to use

composite samples

made of a few

(2-4) aligned crystals.

The data at low temperatures except for

(TMTTF)~Br

were obtained on a conventional

broadband NMR spectrometer where the

wnplitude

of the

z-magnetization

was measured

by

integrating

the free induction

decay signal.

In other

experiments

a commercial Fourier

(7)

transform spectrometer has been used. The

spectra

of ??Se in

(TMTSF)~PF~

and

(TMTSF)~CIO~

are made up of 4 well resolved lines

corresponding

to 4

non-equivalent

Se sites

[43],

with a residual linewidth of 2 kHz due

mainly

to the

dipolar broadening by

the

methyl

group protons. For

Se,

the

knight

shift is known to be much

larger

than the chemical shift and it govems the

position

of the lines

[31].

The

Tj

is found to be site

independent

within the

experimental

error

lo

%)

which is consistent with a rather weak

spatial anisotropy

of

knight

shifts and with

spin density

calculations

[44].

The relaxation is

exponential

over

nearly

two decades for the

magnitude

of

magnetization.

We should comment here on some

particular points conceming

the

experiments

on ??Se at

high

temperature

(100-300 K).

Its relaxation rate is

larger

than that of '3C and

special

care

must be taken when

Tj

is of the order of the free induction

decay length Tf. Owing

to the time

consuming adjustment

of the

w-pulse

at every temperature, we have used the inversion- recovery sequence with

special

care on the

phase cycling

necessary to extract the z component of the

magnetization.

In

addition, during

the

signal averaging,

the coherent noise, e,g. the

acoustic

ringing,

has been

suppressed by altematively changing

the

length

of the detection

pulse (from approximately

w/2 to 3

w/2).

In both sulfur

compounds

studied it is the central carbons that were observed. These can take two

non-equivalent positions

in the unit

cell, giving

two resonance

frequencies separated by

the difference of

knight/chemical

shifts. In

highly

enriched

samples,

however, each line will

split

in

general

into a doublet due to C-C

dipolar coupling.

For central

carbon,

the

strength

of this

coupling (few kHz)

is

comparable

to the

separation

of Larmor

frequencies.

In such a case,

as it is well known for

strongly coupled spin

systems

[45a],

several difficulties appear in the relaxation measurements.

First, partially

relaxed spectra

depend

in a

complicated

way on the initial

non-equilibrium

state of the system, and thus, the

preparation procedure (pulse calibration,

etc. is

subject

to

severe

conditions,

to avoid the

mixing

of

population

differences

resulting

in

non-exponential

decays. Moreover,

for a

coupled pair

of

spins

the observed

Tj's

are

always weighted

averages of transition

probabilities

of the two

spins,

where the

weight

factors

depend

on the ratio :

coupling

constant

(J~/shift

difference

(A) (these

factors are :

cos~q~

and

sin~q~

with tan q~ m J/2 A which can be

easily

shown

by calculating

the transition

probabilities

between the

exact

eigenstates

of the

system).

This

point

is of

particular importance,

because it can, in

principle,

introduce

systematic

errors into the temperature

dependence

of

Tj.

In all

experiments

on

(TMTTF)2Br,

the

crystal

was oriented in a way to cancel the

dipolar interaction,

I-e- the b' axis

approximatively parallel

to the

applied

field

[45b].

We then observed two lines of 2 kHz width

separated by

about 100 ppm at 300 K. This is close to the value found

by

Bemier et al.

[45c],

for

(TMTSF)2Cl04.

The relaxation rates of the two lines

are different

by

a factor of 3 or so, which is also consistent with the

knight

shifts ratio of l.5 in

(TMTSF)2Cl04.

This ratio does not exhibit any variation with temperature and pressure.

A more detailed discussion is needed for the case of

(TMTTF)~PF~

where we used the same

crystal

orientation.

(Hoflc*)

as the one of a

previous

work at low temperature

[5]

in order to

get

a consistent set of data. The

typical

spectrum is shown in

figure

3b. Here, the

dipolar splitting

is close to the shift difference so that the two central lines

overlap.

We found that, within the

experimental

error, the relaxation is characterised

by

a

single Tj

for all the spectrum, and

this,

qt all temperatures. Since the

averaging

of transition

probabilities

cannot account itself for such a difference between these two

compounds (the weight

factors ratio is estimated to be of the order of

10),

we

performed

an additional

experiment

at

higher field,

where all lines

become resolved

(Fig. 3a).

Here the relaxation shows the same characteristics as for

(TMTTF)~Br.

This

clearly

shows that the

averaging

is due to

spin

diffusion between the

spin pairs

of

adjacent

TMTTF molecules. The

spin

diffusion rate, related to the

coupling strength

between

neighbouring spin pairs

is of several hundreds Hertz.

Therefore,

it is faster than the

(8)

(a)

(b)

20 lo -lo -20

f

(kHz)

Fig.

3.-'~C spectra of a single crystal of

(TMTTF)~PF~ (Ho flc*)

at room temperature for a)

Ho

9.4T, and b)

Ho

3.I T. At

high

field, the relaxation times of the two doublets are :

5 ms (left) and ~13

ms

(right).

At low fields, all the lines have the same Tj at about 8 ms.

relaxation at all temperatures

(the Tj inhomogeneity

over the

spectrum

is about lo % at 300 K and

disappears

at lower

temperature).

The relaxation rate has been calculated

by integrating

the

whole spectrum. As

long

as the

Tj's

of both nuclei are

governed by

the same

phenomena,

as it is

(TMTTF)2Br,

this

procedure

is reliable for the

study

of the temperature

dependence

of the

relaxation rate.

Measurements under pressure at

high

temperatures

(T

~ loo K

)

have been

performed

with a

CuBe 8 mm inner diameter cell cooled

by

a thermal

exchanger

with

liquid nitrogen.

The

liquid

flow was

automatically

controlled in order to minimize the temperature

gradients

in the cell.

The RF

circuity

was mounted as close as

possible

to the

sample (10 cm)

but was

thermally

isolated and temperature controlled to ensure the

long

time

stability required

for

averaging.

When

performing high

pressure

experiments

at low temperatures, it is

important

to circumvent any pressure loss on

cooling, especially

near the

freezing point

of the

pressurized

medium. To minimize this

effect,

we have used an

extemally

controlled pressure system.

Isopentane

is

pressurized

in a

primary

chamber where the pressure is controlled

by

a

manganine

gauge. A

capillary tubing

drives the pressure to the cooled cell where an

intensifying piston

creates the

high

pressure in the

experimental

chamber. Two different

liquids

were used to create

high

pressure freon in the

measuring

chamber and

isopentane

in the low pressure stage of the intensifier. Since the

freezing point

of freon

(300

K at 14

kbar) always

occurs well above the one of

isopentane,

losses could be

compensated

until

complete

freon solidification and therefore are

kept

within 500 bar at lo kbar.

All

experiments

under pressure and at

high

temperature have been

performed

at lower field

(about

3

Tesla)

but we have checked that the relaxation rate does not exhibit any field

dependence

up to the

highest

field we used at room temperature

(9 Tesla).

(9)

3. Results.

The '3C

Tj

' temperature

dependence

for

(TMTTF)2PF6 compound

at P

= I bar and 13 kbar

is shown in

figure

4. The low

temperature profiles clearly emphasizes

the different nature of the

ground

state at each pressure. At ambient pressure, the system

develops

a

spin-Peierls instability

at

Tsp=19K

below which

T/' drops rapidly

to zero. Above

T~~ and below T~~ m 40 K or so, there is also a sizeable

depression

of the relaxation rate. These results are

clearly

consistent with

X-ray

measurements

[4]

which show a lD lattice

softening

below 60 K and a true lattice distortion of the

spin-Peierls type

well below 19 K. Similar

depressions

take

place

for the EPR

spin susceptibility

at the same temperature as shown in

figure

4. It is of interest to

point

out here that

essentially

the same features in the

X-ray [4] resistivity [3, 6],

and

Xs(T) (Faraday) [6-7]

results were obtained in another

spin-Peierls compound, (TMTTF)2AsF6.

In

figure 6,

we report

Ti

' and

Xs(T)

data below

Tsp

on a

semi-log

scale

showing

an activated behavior with a gap

Asp

= 3.5

Tsp,

that

apparently

satisfies a BCS like relation for both

quantities.

Above 40 K,

Ti

' is found to be temperature

independent

up to

65K or so, where it starts to increase with an

upward

curvature. The

Tj'

and

x~(T)

temperature

profiles

do not show any

anomaly

near

T~

m 220 K where the

resistivity

minimum

occurs

(Fig. I).

This indicates the very

interesting property

of

separation

between

charge

and

spin degrees

of freedom.

Furthermore,

the enhancement of

Ti

' above 65 K is

directly

related to the one of

Xs(T) [5]. Indeed, figure

7

clearly

shows the existence of the

following scaling

relation

Tj

'

= C

j +

Co TX/(T) (2)

above T~~. This

emphasizes

that an

important

contribution to the relaxation comes from

long wavelength

uniform

spin

fluctuations. As for the constant term

Cj,

it will be show to come from ID AF correlations

(see

next

Sect.).

160

~,~

T,~'(s") (TIITTF)~PF,

~",

o

120

",

o

",~~

o

~$

~~~ ~~~" (K)

~

°

i ~

4

o

. ~

V

AA~

.. o ~

. , ~

~: ~

'eZtl:]...

~

50

emperature (K)

Fig. 4.

13

kbar

The low

data are from [5]. In

the

inset : power law analysis

T/~ in

the

cRtical domain of the AF transition at 13 kbar.

(10)

(imiifi~

PF~

.i

so i i o

0 ~ 20 30 10 50 60 10 T(Kl

Fig.

5. EPR

susceptibility

of (TMTTF )~PF~ as a function of temperature. The low temperature

region

is detailed iri the inset.

I

(Kj

20 0 7 5

to' io'

~~~0)

(TMTTF),PF,

~' ~~~~

10° io°

lo" to"

to-' to-'

10~~ lO~~

0.5 1.0 1.5 2.0 2.5

10/T

Fig.

6.-Therrnally activated behavior of

q' (right

scale) and Xs (left scale) as a function of I/T for (TMTTF )2PF~ below

Ts,

at I bar.

At P

= 13 kbar, the

spin-Peierls

state of

(TMTTF)~PF~

is no

longer

stable

[5, 8]. Instead, figure

4 shows that

T/' develops

a

singular

behavior which

corresponds

to the onset of

magnetic ordering

at

T~

=17 K. From the

log-log plot

in the insert, the power law

~i ~A/ (3)

(11)

I (K)

0 50 loo 150 200

Tl'(S~')

(TMTTF),PF, 60

40

20

0

0.0 0.5 1.0 1.5 2.0

X~T(a.u.)

Fig.

7. Linear

dependence

of the measured

Tj (Fig.

3) on

Tx)(T (Fig.

5 for (TMTTF )~PF~ at I bar.

takes

place

near

T~

where in the notation of

I, iA~

=

[(T T~ )/T~ ]~

° and

d

m

1/2 ± 0.05. It is attributed to an AF type of critical

ordering (see

next

Sect.).

The reduced width

Atn

m

(Tn T~)/T~

m

0.6 for the emergence of these critical fluctuations above

T~

is rather

large.

At

higher

temperature, the relaxation rate becomes temperature

independent

and the

plateau

of relaxation is seen up to 80

K,

where an increase of

Ti

' emerges with a much less

pronounced upward

curvature than the one found at I bar. If we assume that the

empirical

relation

(2)

is still

valid,

these data show that the

amplitude

of uniform fluctuations is

sizeably

reduced under pressure.

Looking

now at the effect of anion substitution in the

sulphur

series,

figure

8 shows the

temperature profiles

of the '3C relaxation rate for each NMR line of the

non-equivalent

center carbons of

(TMTTF)2Br

at I bar

[23].

At this pressure, the

compound

presents a

resistivity

minimum at

T~m

loo K

(Fig. 2)

and it is known to become

antiferromagnetic

below

T~m15K.

For both NMR

lines,

the relaxation rate shows a clear

singularity

near

T~

which is

quite

similar to the one found for the

PF6 compound

at 13 kbar

(Fig. 4).

The critical behavior can be best fitted to the same

expression (3)

with the critical exponents

jm

0.45 and 0.5

respectively

to each line. As for the critical

width,

its

amplitude

~t~

~ o,6 is consistent with the one found for

(TMTTF)~PF~

at 13 kbar. The

slight departure

(12)

=

worth

noting

that the above '3C data confirms the

previous

'H

T/

' temperature variation of reference

[9]

obtained up to 30 K and above which the

methyl-group

rotation contribution to

T/

' becomes

important.

The close

similarity

to the

PF~

salt under pressure is

again obviously

clear in the

paramagnetic regime. Indeed,

the

resistivity (Fig, I)

and relaxation rate

(Fig. 8)

data are consistent with a

complete separation

between

charge

and

spin degrees

of freedom.

The

plateau

of relaxation is well defined up to 75 K where the increase of

T/'

becomes apparent.

Furthermore, using

the EPR

susceptibility

data shown in

figure 9a,

we have checked in

figure

9b that for both NMR

lines,

relation

(2)

is well satisfied.

We now tum our attention to the

Bechgaard

salt

compound (TMTSF)~PF~, figure10a

shows the ??Se

T/

' vs. T

profiles

at P

=

I bar, 5.5 and lo kbar up to 300 K. From the data,

we see that the

upward

curvature of

Tj

' with T is still present in the

high

temperature

regime

but its

amplitude

decreases with pressure. There is no

resistivity

minimum

reported

for the

Bechgaard

salts

(T~

~

Tn,

see

Fig. I)

and the

paramagnetic regime

is well characterized

by

metallic

properties [I I].

At low

temperature

and for P ~ P

~ m

6.9 kbar

[I1, 46],

the

ground

state is

antiferromagnetic.

This is illustrated

by

the

figure 10a,

where at I bar and 5.5

kbar,

T/~

is

clearly singular

at

T~m12K

and

T~=8.7K. Figure

lob shows that the critical variation near the transition still

obey equation (2)

where

again

1Y = 1/2 ± lo % over a decade in temperature. The critical width

Atn

- 0.8

(1 bar), (5.5 kbar)

is also

large

and is similar to

those found in

sulphur compounds.

At lo

kbar~P~,

AF

ordering

is unstable and

(TMTSF)~PF~ undergoes

a

superconducting phase

transition near lK. Nevertheless,

figure10a,

shows that a rather

large

enhancement of the relaxation rate below 30K still remains. This enhancement is

independent

of the symmetry of the anion since very similar results are found for the ??Se

Tj

' data of

(TMTSF)~Cl04

in the so-called relaxed state at ambient pressure

(~

P

~

(Fig.

II

).

This

large

enhancement above

P~

has also been confirmed

by

other groups

[40, 47].

It has also been

reported

for

(TMTSF)2FS03

at 10kbar

~ P

~ =

7

kbar) [17].

From the data of

figures 4, 8,

lo and

II,

it is

interesting

to observe that this

temperature

scale for selenides is

clearly

lower than the one found for the emergence of the

Ti

'

plateaus

in the

sulphur compounds thereby suggesting

that the AF component of relaxation is less

developed

in the

Bechgaard

salts. As for the

high

pressure

region,

the

upward

curvature

of the relaxation

profiles

is rather well

pronounced

for both

(TMTSF)2PF6

and

120

T,~'(s~') (TMTTF)zBT

loo '

,

~ ,

'

'

~~ ~~~

~~~~

t

~

lo

"'~ '~[

.

~$

60 T-T»

(~~/

'

i '

4

,

*

.

20 °

, .

~ .

°

~

~

0 50 i

0

50 200 250 300

Temperature (K)

Fig.

8. Tj vs. T data for the two central carbon sites in (TMTTF

)~Br

at 56 kG. The insert shows the power law

divergence

near T~.

(13)

1-a

Xs (a.u.) (TMTTF),Br

to e

e

e .

O-S

~ .

~ ~

.. . . . .

~~

' ' 0.6

.

o 0.4

,

j~ j

O-Z

0 0.0

(14)

50

~

1500 (Q)

..

~ .

.

$°& ~Q

l ,*»»»

0~1000

~~ O

O ~

500 o

~ ~

o

a

o ~

a

o ~

o

~

, ~

loo 300(K)

(15)

so

T~~~(3~~) (TMT3F)~C10,

P - I bar

60 ,~

,'

. '

~.A

40

~A

20

O

0 lo 20 30

TEMPERATURE (K)

Fig,

ll. Low temperature ~?Se

Tj~

vs. T data of

(TMTSF)~Cl04

in the relaxed state full circles, 31.9 kG f b' full

triangles,

6.o kG

(powder)

open squares, 31.9 kG # c* open

diamond-shaped,

21 kG # b'.

~~Se

~~~'~~~~2 ~§ ~~~'~~~~2 ~~°4 /

%

o. 53 KGw

b'

A : 56.5

KG,I

~

/

. 31KGll

I

a 64

KG,/6

o

P =

I bar. The field

circles

and triangles

correspond to

the

(16)

T~~(3 ~)

/

T~~~(S'~)

60 is

(T~T~~)z~~o

/

~

C

(TVTTF) ~Br SOD

40

iooo

20

w

(TVT3F)~PF,

~~ ~~~

(TMT3F)~C10,

O-O O.5 1-O 1.5 2.0 2.5

x

( T(w,u.)

Fig.

13. Plots of

Tj (Fig.

12) vs. the measured TX

j(T)

of references [34, 35] for

(TMTSF)~PF~

and

(TMTSF)2Cl04

above 50K. The results of

figures

7 and 9b for

(TMTTF)~PF~ (squares)

and

(TMTTF)2Br

(dashed line) are

reported

for

comparison.

tT~Tl'~

,

SK

l~~ *

o/

a

o

~/~

/

a

~/

.

/

4

(17)

4.

Analysis.

4. I THE MODEL. It was soon inferred that for each

compound

of the

sulphur

series at low

pressure, the existence of a

resistivity

minimum at T~ w

T~

is a direct consequence of the relevance of lD electronic

umklapp

processes. Barisic and Brazovskii

[48]

and

Emery

et al.

[16],

indeed

suggested

that the combined influence of the 4

k~

anion

potential

and the apparent

slight

dimerization of the

organic

stacks must open a gap

A~

in the electronic

spectrum

of the

chains at ± 2

k~.

This leads to an effective half-filled band at low temperature and electronic

umklapp

effects will

strongly

influence the nature as well as the

amplitude

of correlations. If

one assumes that far from the critical

point

the

compounds

can be described

by

lD

physics,

repulsive

interactions between electrons will

invariably

lead to the creation of a correlation gap A~ and of an

insulating

behavior below T~ -

A~/w.

These

properties

are well known to

follow from the lD

repulsive

Hubbard model and its continuum

generalization, namely

the so- called lD electron gas model. As described in I and

largely

discussed in the literature

[49],

the direct electron-electron interaction in the latter model is

parametrized

in terms of four

couplings

constants g~ i ~ for left and

right moving

electrons.

They

all reduce to the one-site

repulsion

U in the Hubbard limit. This model is

particularly interesting

for the

present compounds

since the

amplitude

of g~ is

directly

related to the

amplitude

of electronic

umklapp

processes and in tum to the dimerization

responsible

for the

insulating

gap A~

[48].

From its

amplitude,

an effective bare value of g~ can be found as a function of pressure

[14, 50].

A very

peculiar

characteristic of the lD

physics

of this model is the

complete separation

between

long wavelength charge

and

spin

excitations. As a consequence of this

spin-charge separation,

the

absolute value of

Xs(T)

allows the determination of gi and g~

(see

Sect.

5)

and this

independently

of g~ and the combination

2g~-gj

which are connected to the

charge

excitations. The latters

strongly

influence the

strength

of AF correlations and the value of TN

lls, 50].

This property is

invariably

observed in all

sulphur compounds

that

present

a

resistivity

minimum. For

(TMTTF)2PF6

and

(TMTTF)2Br compounds

at ambient pressure for

example, figures

5 and 9a show that the

long wavelength spin degrees

of freedom which

contribute to

Xs(T)

remain

clearly

unaffected in the temperature range where the

resistivity

minimum occurs

[7].

The effect of

hydrostatic

pressure is of interest since it reduces the dimerization and this will diminish the

amplitude

of g~.

Therefore,

it will shift the

resistivity

minimum at lower

temperature.

This has been found to be consistent with the

resistivity

data of

Creuzet et al.

[8]

for

(TMTTF)~PF6

and

(TMTTF)~Br

under pressure and also with the

absence of

resistivity

minimum for all the

Bechgaard

salts at ambient pressure. These are indeed less dimerized than the

sulphur compounds [16].

It is from these

experimental

facts and the central role

played by

the

antiferromagnetism

in the combined

phase diagram

of both series that the ID electron gas model has attracted considerable attention for the theoretical

interpretation

of these materials

[5 II-

As shown in I, the

quasi-ID generalization

of this model with the inclusion of the small transverse

hopping integrals

is necessary to assure the existence of

long

range

ordering.

For both

sulphur

and selenides series the band structure calculations as well as several

experiments

lead to the

anisotropic

sequence of t~ m15 ti~

m 450 ti~ between the a, b and c direction

hopping integrals. Despite

similar

anisotropy

ratio

ti~/t~

between the two

families,

the

amplitude

of t~ differs

sizeably

however.

According

to Ducasse etal.

[2b]

at I bar,

E~

m 300 K for

(TMTTF)~PF~ compound

while

E~

m 2 500 K for the

(TMTSF)~PF~.

In presence of a lD correlation gap

A~,

transverse

single

electron band motion is

essentially

frozen below T~ and it is

through

an effective interchain

exchange

mechanism

(IEX)

which has

a kinetic

origin

that interchain

propagation

of AF correlations can lead to

long

range order at a

finite

T~ [15].

As we will discuss next, temperature

dependence

of NMR relaxation rate

(18)

especially conceming

the nature, the

amplitude,

the

statics,

the

dynamics

and the dimensionali- ty of

spin

correlations.

4.2 SCALING FEATURES.

Sulphur compounds.

Let's first look at the temperature variation of the

~~CTj'

in

(TMTTF)~PF~

at 13 kbar

(T~

m 17 K ; T~ m 75

K) [5]

and

(TMTTF)2Br

at P

=

I bar

(T~

=

15 K T~ m loo

K) [23]

which are

typical antiferromagnets

of the

sulphur

series at low pressure. As

already

mentioned in section

3,

the

log-log plots

of the

figures

4 and 8 for the cRtical behavior lead to the value

d

= 1/2 ± lo % in both cases.

According

to the results of the

section of 2, the AF contribution to the nuclear relaxation rate,- for a

quasi-

lD electron gas model with a lD correlation gap A~ near

T~,

is

given by

i~l

[~ Q01

~ ~

(Q0) ~Al'~ (~)

Here

©(Qo)

is a temperature

independent quantity (cf. Eqs.(1-6)

and

(1-57)).

This

Ti

'

expression

results from a RPA treatment of the interchain AF

exchange (IEX) coupling

while the lD part of the

problem

is treated

rigorously.

As for the cRtical index

d

=

1/2,

it is a 3D result that combines Gaussian exponents for the AF

susceptibility,

the correlation

length

and the relaxation time for fluctuations. The Gaussian cRtical width

Atn

m

(Tn T~)/T~

for the emergence of critical fluctuations above

T~

is obtained from the temperature

Tn

at which the cRtical contribution to

T/

' exceeds the

paramagnetic

one. This

quantity

reflects the non-universal features of the model used. Whenever T~ w

TN,

the transition is driven

by

the IEX

coupling

and from

(1-62),

it is

given by

~~fl "

"~[~~

Y~

("i~Nf~F) f0a f0b f0clT/ (5)

where

according to1,

6 S I and y

=

I. For the IEX

mechanism,

the transverse coherence

lengths lo

~ and

lo

~

are small

quantities (~

l while

(wT~/v~) to

~

l so that A tn l which is

large

in agreement with the results of

figures

4 and 8 for which one has

Atn

0.6 for

sulphur compounds.

It should be stressed however, that the

anisotropy

with respect to the third

direction should be

quite large

in these

compounds (fo~w fo~)

so that it may appear

somewhat

surpRsing

that no 2D critical effects are seen above

T~.

The reason for this is not clear so far.

As discussed in I, the choice of a different mechanism for the

phase

transition can lead to a

quite

different value for

Atn.

In this respect, a 3D

logarithmic nesting (NAF)

mechanism of

antiferromagnetism,

would lead to different precursors

(see Sect.1-6)

and to a smaller

amplitude

for

Atn [41].

In that sense, the

amplitude

of the precursors to the transition can

give

valuable indication on the

type

of

microscopic

mechanism involved in the

propagation

of AF correlations for the transition.

In the framework of the

quasi-ID

electron gas model, the presence of a correlation gap A~

wT~

300 K for both

compounds

in their

respective

pressure conditions would corre-

spond

to the temperature scale for strong electronic

umklapp scattering.

This

yields

a power law exponent y of the ID 2 k~ AF response function x

(2 k~, T)

T~ ?

)

that is

exactly equal

to

unity.

This exponent

together

witli the exact relation

v = z = I for the coherence

length

and

Références

Documents relatifs

the additional transverse integration in the triangular diagrams of figures 6 and 7, generated by the square diagrams under consideration, reduces both interac- tions

2014 We report an investigation of the temperature dependence of the conductivity and conductivity ani- sotropy in the single-chain quasi-one-dimensional organic

Abstract 2014 The Tomonaga model for a system of parallel chains with the long range Coulomb interactions is.. investigated by the bosonization

In the amorphous phase, since the chains are no longer constrained within a crystalline lattice, slow « out-of-lattice » motions of a great number of chain bonds or

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

By naive analogy with conventional semiconductors, one might say [10, iii that all electron states below trie "primary" gap, opened by trie FISDW potential at trie wave

However, despite adequately sensitive data considering the [C ii ] - SFR relation observed at lower red- shifts (e.g. The [O III] / [C II] emission line ratio for high

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