• Aucun résultat trouvé

Low-temperature thermodynamic properties of amorphous sputtered Zr 100-xCux alloys. Effect of structural relaxation

N/A
N/A
Protected

Academic year: 2021

Partager "Low-temperature thermodynamic properties of amorphous sputtered Zr 100-xCux alloys. Effect of structural relaxation"

Copied!
26
0
0

Texte intégral

(1)

HAL Id: jpa-00210992

https://hal.archives-ouvertes.fr/jpa-00210992

Submitted on 1 Jan 1989

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.

Low-temperature thermodynamic properties of

amorphous sputtered Zr 100-xCux alloys. Effect of

structural relaxation

F. Zougmoré, J.C. Lasjaunias, O. Béthoux

To cite this version:

(2)

Low-temperature

thermodynamic properties

of

amorphous

sputtered Zr100-xCux

alloys.

Effect of structural relaxation

F.

Zougmoré,

J. C.

Lasjaunias

and O. Béthoux

Centre de Recherches sur les Très Basses

Températures,

C.N.R.S., B.P. 166 X, 38042 Grenoble

Cedex, France

(Reçu

le 10 octobre 1988, révisé le 3

janvier

1989,

accepté

le 9

janvier

1989)

Résumé. 2014 Nous

rapportons sur des mesures de chaleur

spécifique

à basse

température

d’alliages

Zr100-xCux

(19~

x~

64)

préparés

par

pulvérisation cathodique,

que nous comparons aux

alliages

correspondants

obtenus par la

technique

d’ultra-trempe

de l’état

liquide.

Alors que la

température

de transition

supraconductrice Tc

est très voisine pour les deux types

d’alliages

amorphes,

indication

qu’elle

est

pratiquement

insensible au

degré

de désordre structural

plus

élevé induit par la

technique

de

pulvérisation,

par contre à la fois le coefficient

électronique

y et la contribution de réseau 03B2 T3 sont

plus

élevés pour les

alliages

«

pulvérisés

». Le caractère commun aux deux

alliages

d’une croissance de 03B3 avec la concentration en Zr est considérablement accentué dans le cas des

alliages

«

pulvérisés

».

Cependant,

ces fortes valeurs de 03B3 assez surprenantes ne

conduisent pas à un comportement anormal pour le processus

supraconducteur,

ainsi que le

prouve la condensation

électronique

totale en dessous de

Tc.

Tous les

paramètres

thermodynami-ques sont sensibles à la relaxation structurale, au contraire des

alliages trempés

du

liquide,

tandis que la diminution de la

Tc

est similaire dans les deux types

d’alliages.

Abstract. 2014 We

report on

low-temperature specific

heat measurements of

superconducting

amorphous Zr100-xCux

(19 ~

x ~

64)

alloys prepared by sputtering

that we compare to

corre-sponding alloys

obtained

by

fast

liquid-quenching technique.

Whereas the

superconducting

transition temperature

Tc

is very close for these two kinds of

amorphous alloys, indicating

that it is almost insensitive to the

higher degree

of structural disorder inherent to

sputtering,

both the electronic coefficient 03B3 and the

lattice 03B2T3

contribution are

larger

for the

sputtered alloys.

The common character of an

increasing

value of 03B3 with the Zr content is

considerably

enhanced for the

sputtered alloys.

However, such

surprisingly high

y values do not lead to any anomalous behaviour for the

superconductivity

process, as

proved by

the

complete

electronic condensation below

Tc.

All

thermodynamic

parameters are sensitive to structural relaxation, at variance with the

liquid-quenched

alloys,

whereas the

Tc depression

is of the same

magnitude

in both kinds of

alloys.

Classification

Physics

Abstracts 65.40 - 63.50 - 74.20F

1. Introduction.

There is a

growing experimental

evidence for a

dependence

of numerous

physical properties

of

amorphous

metallic

alloys

upon their conditions of

preparation,

e.g. either

by

vapor-quenching

(V.Q.)

or

liquid-quenching

(L.Q.)

techniques.

Moreover,

upon these

conditions,

they

are

differently

sensitive to structural relaxation

[1].

(3)

It is now demonstrated

by X-ray

measurements, differential

scanning calorimetry

(DSC)

[2],

extended

X-ray

absorption

fine structure

(EXAFS) [3]

and small

angle X-ray scattering

[4]

that

sputtered samples

exhibit a more disordered structure than

liquid-quenched

ones with a

degree

of disorder

probably

intermediate between thin films

vapor-quenched

onto a cold

substrate and

melt-spun

ribbons.

Furthermore,

properties

of materials

prepared by

vapor-quenching

onto a cold substrate can be

significantly

modified

by

a

subsequent

thermal

treatment,

contrary

to the case of the

L.Q.

ones which have been

already

« stabilized »

during

quenching

from the melt. We had

previously

studied the effect of structural relaxation on the

low-temperature

thermal

properties

of Zr-based

sputtered alloys,

at Zr

composition

close to

75 at% : nominal

Zr76Ni24,

Zr76CU24, Zr80CU20 [5, 6].

In that case,

thermodynamic properties

such as the electronic coefficient y, the

Debye

temperature

BD,

the

density

of

low-energy

excitations

(or

two-level

systems,

T.L.S.)

tend toward the values of the

corresponding

L.Q.

materials with initial densities for every kind of excitations

(electron,

phonon,

and

T.L.S.)

which

largely

exceed those of

L.Q.

alloys ;

in the same

time,

the

superconducting

transition

temperature

T,

decreases in a similar way as in

L.Q.

alloys.

We have used the

possibility

by

varying continuously

the concentration and

by

relaxing

the

structure to extend a similar

thermodynamic

study

to a

large

concentration range

(19 --

x , 64

at%)

of the

Zr,oo - xcux

system,

that we

report

here. Most of the

previous

conclusions are

confirmed,

concerning

the

Debye

temperature,

the

superconducting

transition and the

low-energy

T.L.S. excitations.

However,

the behaviour of the electronic

y coefficient

is

quite

unexpected.

Its

magnitude

remains

always larger

than for

L.Q.

Zr-Cu

alloys,

especially

at

high

Zr content where the structural

stability

of the

alloy

decreases.

Moreover,

y is very sensitive to thermal treatments, even to

ageing

at room

temperature,

with senses of

variation which

depend

on the concentration range of the

alloy.

The paper is subdivised as

following :

in section 2 we describe

samples preparation

and

characterization,

and the

acquisition

of

specific

heat data

technique,

i.e. the

analysis

of the transient responses to an heat

pulse ;

in section

3,

we describe results and

analysis

in the whole

temperature

range,

including

the determination of

7c

and the electronic contribution

Ce

in both normal and

superconducting

states ; in section

4,

we discuss the normal state

properties

and the relation of

Tc

with the electronic and

phonon

densities of states. In a final section

5,

we discuss the

superconducting

state, with the electronic condensation process, the

phonon

contribution at very low

temperature

and the onset on the T.L.S contribution.

Elsewhere is

reported [7]

the

analysis

of the TLS

excitations,

which are known to be a

characteristic of these

amorphous

materials. 2.

Experimental.

2.1 SAMPLES PREPARATION AND CHARACTERIZATION. - The

amorphous samples

of

nominal concentration

varying

between 20 -- x -- 65 were

prepared by

a

high-rate

(10 jim/h)

D.C.

magnetron

sputtering

technique [6]

at a

deposition

temperature

of 77

K,

in the form of foils about 2 x 4

cm2,

80 to 100 m in

thickness,

about 0.5 g

weight.

In all the

series,

hydrogen

which could be

present

at rather

large

concentration in

sputtered

films

[8],

is not

detected

by

chemical

analysis

[9]

within the limit of

detection,

i.e. 10-2

weight

% or about

1 at%. The absence of

hydrogen

in these bulk

samples

is confirmed

by

the excellent

agreement

of

7c

with

L.Q.

ZrCu,

since

7c

is known to be very sensitive to the presence of

hydrogen

in Zr-based

alloys

[ZrPd,

Zr-Ni,

...] [10].

The exact

composition,

obtained

by

absorption

chemical

analysis,

indicates that in

comparison

to the nominal concentration of the

target,

there is a

depletion

in Cu which

increases from about 1 at% for

ZrSOCu20

to about 2 at% for

Zr5oCu5o

and

Zr3SCu6S.

In the

(4)

Characterization results.

1)

The

homogeneity

of the

samples

is

proved by

the mass

density

measurements

systematically performed

[by

Archimede method with toluene as

acting

fluid]

on each of the

five or six foils of a same

deposition

batch,

and which

generally

differ

by

less than

0.03

g.cm- 3.

These fluctuations of mass

density,

between different

foils,

can be accounted for

by

those of

concentrations,

of the order or less than 1 at%. We

verify

the very

good

agreement

between our values

(Tab.

1 and

Fig. 1)

and those for

L. Q.

alloys

from two groups

[11, 12],

which differ between them

by

about 2 %. The effect of densification

(-

0.25

%)

on

thermal

annealing, previously pointed

out on a same foil of

Zr77CU23 [6]

could not be detected for other

compositions, being

within the

uncertainty

of the measurements

(about

0.01

g.cm- 3).

As well as for densification

effects,

a decrease of

length

could be

expected

due to the

annealing

of the frozen-in free volume. Irreversible contraction

¥

of about 2 x

10- 4

has been measured

by

Hillairet et al. in a similar

sputtered Zr76Ni24 alloy,

which

Table I. - Characterization data

of sputtered

ZrlOO-xCux

samples.

* See also reference

[6].

Fig.

1. - Mass

density

versus the copper concentration x of

sputtered samples

(symbol 0)

and two

series of

liquid-quenched

from reference

[11] (Symbol e)

and reference

[12] (symbol Âà, ).

Data of

(5)

corresponds

to an

(almost undetectable)

densification effect of the order or less than

10- 3

[13].

2)

Similar

good

agreement

with

L.Q.

ZrCu

alloys

[11,

12]

is obtained for the

position

(2 () m)

of the main

X-ray

di f fraction

halo

(Tab.

1 with the

corresponding

transfert momentum

Qp = 4 sin

()m

(À-

with À = 1.542

À

for

Cu-Ka).

These

6max

values are not

notably

affected

by

heat treatment either for

L.Q.

or

sputtered alloys.

But one

observes,

for the

sputtered

ones, a

narrowing

(-10%)

of the width of the halo which indicates a

reorganisation

of the

short-range

structural order

[4b].

When

comparison

is made

(e.g.

Zr6Ni24

prepared

in the same way

[4c]),

the first halo is a little bit broader

(-

10

%)

for the

sputtered

samples.

In a

general

way, these results for

() m (Q p)

and

density

confirm the

similarity

of the average

structural

parameters

for both kinds of materials.

3)

Differential

thermal

analysis

data,

generally

obtained at a

heating

rate of 20

K/min,

are

shown on

figures

2 and 3. At

high

Zr content i.e. x -- 28 at%

(Fig. 2),

recrystallization

occurs

in two

stages,

as

already

discussed

[7] :

the first broad exothermic

peak

(symbol A

in

Fig.

3)

is the formation of metastable

co-Zr,

which thereafter transforms in

equilibrium

cy-Zr above

400 °C

[4b].

We note that this w

crystallization

occurs at a

rapidly decreasing

temperature

for an

increasing

Zr content. The second

sharp peak

(symbol e

in

Fig.

3)

is the formation of

Zr2Cu

at a rather constant

temperature

of 340 °C. For copper content

larger

than 28

at%,

the w-Zr

crystallization

stage

is no more detected.

Instead,

the

unique crystallization peak

Fig.

2. -

D.S.C.

thermograms

of parts of

sputtered samples

(exact

composition

reported)

at a

heating

rate T = 20 K

min- l.

The dashed line is the base-line

corresponding

to the

recrystallized

state. For

(6)

(T.,

symbol

0 in

Fig.

3)

is

preceeded

by

a broad endothermic

signal

characteristic of a

glass

transition

( TG,

symbol

0 for x =

32,

38,

48

at%)

defined as the

beginning

of the endothermic

effect.’

Also in that case a very

good

agreement

for both

TG

and

T.

is obtained between these

sputtered alloys

and the

L.Q.

ones

(Fig. 3).

Fig.

3.

- Sputtered samples : crystallization

(T.,

symbol

D)

and

glass-transition

(TG,

symbol

0)

temperatures versus x. Below x = 28 at%, 8 and. represent the two

crystallization

stages

(see

Fig.

2 and

text) ;

symbols

(D and M represent

TG

and

Tx

of

sputtered Zr40Cu60

from

Walmsley

et al.

[16]

(T =

20 K

min-l) ;

the dashed-lines for

melt-spun

from reference

[12] ( T

= 10 K

min- 1).

A characteristic difference between both kinds of

alloys

lies in the

larger

irreversible exothermic effect for the

sputtered alloys,

measured

by

sensitive differential

scanning

calorimetry

(DSC)

on the first

heating

above room

temperature

[2].

For

Zrg1Cu19

and

Zr77CU23,

the heat release is of the order of 1 kJ/mole

[14].

This effect has been ascribed to a

decrease of the

configurational enthalpy,

which is

larger

in the initial state of

sputtered alloys,

due to

higher

disorder

[15].

4)

Sensitive structural

investigations

such as extended

X-ray

absorption fine

structure

(EXAFS)

and

small-angle

X-ray

scattering

(SAXS)

were

performed

on some

samples.

Both

techniques

confirm the

higher degree

of structural disorder in

comparison

to L.Q.

alloys.

In

SAXS,

the flat Laue

scattering

of the

« as-sputtered » samples

is the indication of an

homogeneous

medium

[4b].

Thermal treatments cause

tiny,

but still

detectable,

variations of the structural

parameters.

There is a trend for

clustering

of the Zr atoms in

Zr77Cu23

as shown

by

EXAFS

[3].

5)

Another characteristic difference between the two kinds of

alloys

is the irreversible decrease of the electrical

resistivity

on first

heating

(by

about 1

%)

in

sputtered alloys

(Zr76Ni24, [13]),

whereas it is almost absent

($

lo- 4)

in

L. Q.

This relaxation effect is ascribed

to a

topological reorganization

on a local scale

[13].

In

conclusion,

sputtered alloys

are in a

higher degree

o f structural

disorder than

L.0.

on the

(7)

recent

study

of their

dependence

upon the conditions of

low-temperature deposition

for

amorphous

H20

or

D20

films

[17].

The effects of structural relaxation are

consequently larger

in

sputtered alloys

than in

L.Q.

However,

no structural

inhomogeneities

such as

phases

separation

have been detected in these

samples,

either

by

X-ray

investigations

[4, 18]

or from

their

superconducting properties by

the

unicity

of the transition.

At

last,

EXAFS

investigations

[3, 19]

do not

point

out chemical short range order

(C.S.R.O.)

in the Zr-Cu

system,

either for

sputtered

(Zr77CU23, Zr8lCU19-II)

or

L. Q. ,

except

at

high

Cu content

(Zr40CU60-L.O.).

This is at variance with the Zr-Ni

system.

Thermal Treatments.

Samples

have been measured

firstly

in their «

as-prepared »

state, after

being

removed from the

deposition

apparatus

at room

temperature,

and

secondly generally

after an

isothermal

annealing

of one

hour,

under an

atmosphere

of

ultra-pure

argon, at

temperature

Ta

well blow the

crystallization

temperature

( Ta

= 200 °C for 19 -- x --

38,

Ta

= 250 °C for

x =

48),

or

ageing

at room

temperature.

Between the measurements,

samples

are stored in

liquid nitrogen.

2.2 SPECIFIC HEAT TECHNIQUE. - The

specific

heat of about 1 to 1.5 g of material

(two

or

three

foils)

was measured on a

He3-He4

dilution

refrigerator by

means of a transient heat

pulse technique.

The

temperature

varies

step

by

step,

over a range

(0.08

K to 7.5

K)

which includes the

superconducting

transition

T,.

The

validity

of the transient

technique

is based on

the condition that the internal time constant due to the thermal

coupling

between the

sample

foils must remain small

compared

to T which characterizes the

exponential

temperature

decay

back to thermal

equilibrium,

and which is

equal

to the

product

of the total heat

capacity

times the thermal resistance

Re

of the link to the cold sink

[20a].

In the

present

case, this is obtained

by

a

sample

holder

arrangement

well

adapted

to the

sample

geometry

and which ensures a

good

thermal

diffusivity, especially

at very low

temperatures

(T« Tc)

where no more

electronic

diffusivity

is

present :

the

sample

foils are

pressed

between two silicon

plates

of similar

surface,

one of those is

equipped

with the

heater,

and the

opposite

with the

thermometer and the thermal

link,

with an

adapted

value of

Rp .

Such an

arrangement

has been tested and used in numerous

samples

of bad thermal

diffusivity

[6,

20a-b,

21].

The accuracy of determination of

C p

has been

improved by

an automatic data

acquisition

and

analysis

of the

exponential

transients

(Figs.

4a and

b).

During

a set of

acquisition,

the reference

temperature

of the copper screen around the

sample

holder is

regulated by

a

four-wire

probe bridge

within

10- 4

of the fixed

temperature.

The

signal

of the measurement

resistance is also

amplified by

an a.c.

( f

= 90

Hz)

bridge.

The

procedure

is the

following :

a)

for the definition of the base

line,

corresponding

to the reference

temperature,

10 data

points

are taken. Each of them is the average of 4

points

taken at

microcomputer

clock pace

(every

10

ms).

So

they give

a well-defined base

line,

parallel

to the x axis

(time) ;

b)

at the 10th

point,

a heat

pulse

of energy

R 2 t

is

automatically

sent on the

sample

during

a

laps

of time t.

Throughout

the 40th data

point

after the maximum

(kl

on

Fig.

4a),

data

points

are taken without

averaging,

at a pace which is

generally

10 times

greater

than for the

averaged points.

So one

gets

experimental

data

(S,,,,p)

very close to the real

signal

in the critical

region

for the

specific

heat calculation :

heating

up,

decay

of the

signal

down to or

beyond

the inflexion

point

(k2

in

Fig.

4a),

which

corresponds

to the onset of the

exponential

decay regime

of the

temperature,

when thermal

equilibrium

is established within the whole

sample

and also between the

sample

and the different

addenda ;

(8)

Fig.

4.

- a)

shows the temperature

decay

after a heat

pulse

and its

analysis by

means of a

microcomputer

for determination of the

exponential

variation : + for

experimental decay

(Se.,P) ;

dashed-line for theoretical one

(Sth).

K2

and

K3

can be moved for

fitting

as close as

possible

the

experimental

data ;

b)

is a

plot

of AS =

Sth -

S,,,p

versus times

(arbitrary units).

It shows that the fit of

figure

4a is within about 1 % of the

signal.

d)

between k2

and

k3,

the

point

at which the

signal equals

20 % of the maximum

deflection,

the

exponential decay

is fitted to

exp (- ak

+ b

),

called

Sthe,,, ; a

and b

being

determined

by

a

linear

regression ;

e)

thereafter,

we determine

kI, point

of the idealized instantaneous heat

jump

of the

sample, by equalization

of areas A and B of

figure

4a,

corresponding

to the conservation of energy sent into the

sample :

or

where the

integration

of

Sexp

is

performed by

a

trapezoidal approximation

method in this

region

with

large density

of data

points.

Then,

from abscissa

kI,

one

gets

the final

R

f

( oc Sf = exp ak, + b ])

and initial

R;

(

oc S;

of

Fig.

4a)

resistance values of the

thermometer,

and from the calibration law the initial and final

temperatures

(Ti, T f) .

The three coefficients

Ai, Bi

and

Ci

i of the standardization law of the thermometer

[Log Ri = Ai (Log. T)2 + Bi (Log., T) + Ci]

are

determined for each value of the resistance

Ri by fitting

the calibration curve to a

parabola

throughout

the i -

3, i - 2, i - 1, i, i

+

1, i

+

2, i

+

3,

points

of calibration

[20a].

This determination of 3

parameters

Ai, Bi

and

Ci,

well

adapted

to the

Si-doped

thermometers,

is

more

precise

and continuous than the usual method of

approximation by polynomials

of

higher degree

over a

larger

number of calibration

points.

The thermometers

presently

used are Si sheets

doped

with P or B

by

ionic

implantation.

They

show a remarkable

stability against

the thermal

cycles.

In the

present

case,

along

this

series of

experiments

(about 20),

the resistance varied

by

less than 0.05 Q over 80 at 4.2

K,

corresponding

to a variation in

temperature

less than 10 mK.

At

last,

if W is the electrical power

(Rf t)

used for

heating,

the total heat

capacity,

including

W

addenda,

is

mC

= Tf - Ti

The

specific

heat of the

sample

is obtained after

subtracting

the

(9)

3.

Spécifie

heat data

analysis

and results. 3.1 DATA ANALYSIS. - We

can

distinguish

three

temperature

ranges, characterized

by

different

weights

for

phonon,

electronic or T.L.S. contributions :

i)

between

T,

and

7 K,

there is a

good

agreement

of

C p

with the usual

y T +

,8

T3 law,

as shown

by

the

plot

C/T

versus

T 2in

figure

5 for some

samples

of the series. Such a

variation,

including

sometimes

higher

power terms

(oc

T5),

is

generally

also

obeyed

for Zr-based

melt-spun alloys

in similar

T-range

[22-24],

and more

generally

for numerous

amorphous

metallic

alloys

[25].

From the electronic coefficient y one obtains the

density

of

states at Fermi level :

N, (EF)

=

3 y

(states.

e v- 1 .

atom- 1,

with y in mJ.

mole - 1 .

K- 2) ;

,7r 2 k2.- .-lu

and from the lattice

T3

term, the

Debye

temperature :

Fig.

5. -

Cp/T

versus

T2 for sputtered amorphous samples.

All data, but for x = 19 at%

(sample II),

correspond

to annealed

samples.

The continuous lines represent the fit to the

l’ T + f3 T3 law

above

Tc.

ii)

the electronic

specific

heat

Ce,

obtained from

Cp

after

subtracting

the

phonon

ceT

Fi

12 . The

contribution {3

T3,

can be

plotted

as

T

versus T

(Fig. 6)

or

log

Ce

versus

T

(Fig. 12).

The

T

’Y Tc

T

first

diagram

enables the determination of

7c by

the criterion of

equalization

of

entropy

for the

experimental

transition and for the idealized

specific

heat

jump

at T =

Tc ;

it also enables

an

improved

determination of y in

comparison

to

figure

5. The second

diagram allows

one to

verify

the

exponential

decay

of the electronic contribution

C es

below

7c

(starting

at

7"

20132).

This determination is also

improved

if one subtracts the T.L.S

contribution,

as

T

)

p

discussed in section 5.

iii)

this T.L.S. contribution becomes

predominant

for

temperatures

below 0.5

K,

or more

7"

precisely

when

Ces

has

vanished,

i.e.

for ) ±

6

(Fig. 12).

h Y es

(10)

Fig. 6.

against

T. This

diagram

allows the determination of

T, by

the

equalization

of entropy for both

experimental

and idealized

specific

heat

jump

at

T,.

It also allows a

good

determination

of the electronic coefficient y above

Tc;

; here y = 11.7 ±

0.2 mJ.mole-l.K-2 for

ZrS1Cu19

(II)

and

7.45 ± 0.1

rnJ.mole-l.K-2

for

ZrCu32 (annealed).

3.2 RESULTS FOR THE NORMAL STATE.

3.2.1 Electronic

density

of

states. - In table Il and

figure

7 are collected values of

N y(EF).

These values determined from yn above

T,,

have been submitted to the criterion of

equality

of

entropy

in

superconducting

and normal states. If we assume the same lattice

/3 T3

term in the two states, one must have

Sn (Tc)

=

Ss(Tc),

and therefore :

There is a

good

agreement

for the whole series : calculated ys agrees with the measured Yn

above Tc

within 1 to 6 %

(with Ys> Yn),

except

for

Zrg1Cu19 (I)

and

(II)

where ys

exceeds

yn

by

11-14 %. But we note that at this

high

Zr content, there is an

uncertainty

concerning

the lattice contribution in the

superconducting

state which is very

probably

smaller than above

Tc

[Ref. [6]

and Sect.

5]

and which could

parly explain

the

discrepancy.

For

Zr52Cu48,

there is a

slight discontinuity

at T = 1.9 K

(see

Fig.

5)

which can be

analysed

as an increase in y

by

about 5 % below this

temperature,

feature which is not modified

by

subsequent annealing.

A similar feature has been observed in

Zr60Cu40

[23]

and in

Zr62.9Ni37.1 [29],

which,

for the second case, has been ascribed to the presence of a second

amorphous phase,

but undetectable

by X-rays.

The main results for

N y (EF)

are the

following :

a)

the electronic D.O.S. is very sensitive to heat treatment or to

ageing

effects.

Figure

8a shows the effect of a

stay

of a few

days

at room

temperature,

between two

successive

experiments.

One observes a variation of 10 % of the electronic

y T

term, without any

change

either for

0 D or

for

T,.

(11)

Table II. -

Superconducting

and

thermodynamic

parameters

of sputtered Zr,oo - xcux

samples.

(1)

From reference

[5].

(2)

From reference

[6].

It seems that there is a cross-over effect for

Zrg1Cu19

(I)

between two «

equilibrium

» values

(Fig. 8b) :

firstly,

on

ageing

at room

temperature

over

period

of a few

months, y decreases

toward a value rather similar to

L.Q.

alloys

(points

1,

2,

3),

and also close to heat-treated

Zr77CU23,

about 5

mJ/mole.K2.

After a thermal treatment at 200 °C and

during subsequent

storage

at -

20 °C,

y tends toward a much

higher

value of about 11

mJ/mole.K2 (points

4,

5,

6).

Note also in

figure

8b that in the same time there is a small variation of

Tc, following

the initial decrease which appears as a

general

consequence of structural relaxation

(see below).

This

strong

sensitivity

to thermal treatments is at variance to

L.Q.

Zr-Cu

alloys,

where no

systematic

and much smaller variations

(increase)

can be detected

(in

Zr54Cu46

[26]

and

Zr72CU28

[27]) ;

b)

values of

Ny (EF)

considerably

exceed those of

L.Q.

alloys,

obtained

by melt-spinning

[22, 23].

There is an overall trend of

Ny (EF)

to increase at

high

Zr content. For

L. Q .

alloys,

the variation is almost linear with concentration. It is also the case for the extreme maximum values of the

sputtered alloys,

which concern the concentrations x = 19

(1),

38,

48 at%

(relaxed state)

and x = 19

(II), 27, 63.5

at%

(as-prepared state).

The differences of

amplitude

between these linear variations decrease when x

increases,

as does the difference between the

values of

crystalline

and

amorphous phases,

with a trend to a common value at

large

Cu

content. This confirms the

predominant

role of Zr in the electronic D.O.S. of these

(12)

Fig.

7. - Electronic coefficient

y and

corresponding

D.O.S. versus the concentration for :

sputtered

samples :

ZrSlCu19 (I),

see

figure

8b ; for other

samples :

(0)

as-prepared,

(A)

annealed state.

Upper

dashed line is drawn

through

the

high

y values for both annealed

(Zr81Cu19 (I), Zr62Cu3g,

Zr,2CU48)

and

as-prepared

state

(Zrg1Cu19 (II),

Zr73CuZ7,

Zr36.5Cu63.5)·

Melt spun :

(x )

from reference

[23] ; (0)

from reference

[22].

Dashed line : mean value of electronic D.O.S. from

Hc2

and p measurement of both

sputtered

samples

(either

as-prepared

or annealed

state)

and L. Q.

samples

(Refs. [11,

23,

28]).

The effect of

annealing

on

Zr54Cu46 (Ref. [26])

is indicated.

Crystalline :

(8.)

from reference

[22].

c)

within the G.L.A.G.

(Ginzburg,

Landau,

Abrikosov,

Gor’kov)

theory

for

weak-coupling superconductors

in the

dirty

limit,

one can also determine the D.O.S. at the Fermi level from electrical

resistivity

p and upper critical field

Hc2

measurements

[28]

by :

with

These determinations show that there is no effect of thermal treatment on

’Y He2

and that the

D.O.S. determined in that way decreases

linearly

with x as

NH (EF)

=

2.5 (1 - x)

[states.eV-1,

atom -1

],

in

good

agreement

with the values of

L.Q.

alloys

(shown

by

the lower dashed line in

Fig.

7).

Therefore,

there is a

discrepancy

between

N ’B1 1 (EF)

and

N H e2 (EF)

for the

sputtered alloys,

which is much

larger

than that

previously reported

for

L.Q.

Zr-Cu or Zr-Ni

[23, 29].

3.2.2

Debye

temperature.

- In table II and

figure

9 are also

reported

the values of the lattice

coefficient

/3

and

OD.

However,

we have to take care in

using

the

Debye

model for the

(13)

Fig. 8a.

Fig.

8b.

Fig.

8. -

Sensitivity

of the electronic coefficient y on thermal treatments :

a)

a

subsequent experiment

following

a stay of a few

days

at room temperature indicates an increase of y without any

change

either for

0 D

or

7c ;

b)

a cross-over effect of y is observed for

Zr81Cu19 (I)

with the thermal

history : points

(1),

(2), (3) :

as-prepared

state then

ageing

of 45

days

and 4 months at room temperature,

point

(4) :

effect of

annealing

(200

°C-1

h) ;

points

(5)

and

(6) :

ageing

at around - 20 °C after heat treatment. On the

right

side the

corresponding

superconducting

transition temperature

Tc

(0

and

A,,&

respectively aged,

annealed and

highly-relaxed

states).

Note the

quasi-invariance

of

7c

for the annealed state while y had

largely

increased.

Fig. 9. - Debye

temperature

0 D

(calculated

from the cubic term of the

specific

heat)

versus x.

Sputtered samples :

(0)

as-prepared,

(A)

annealed,

(À)

highly-relaxed : ZrSlCu19 (I).

Melt-spun :

(x)

from reference

[23], (0)

from reference

[22] ;

the effect of thermal

annealing

on

Zr54CU4

[26]

is indicated.

Crystalline

(0)

from reference

[22] ;

the

drop

of

6D for Zr40Cu60

has been ascribed to the

(14)

description

of vibrational

spectra

at low

frequencies

in disordered solids :

despite

a

T3 specific

heat contribution in these

amorphous alloys,

there is some

experimental

evidence

of additional excitations to the actual

ùj 2 phonon

contribution.

Hitherto,

published

acoustic data are

only

available in

L.Q.

Zr4oCu6o [30]

which indicate an acoustic determination

Oacoustic

= 272 K

higher

than the calorimetric one :

(Jcalor.

= 230 K

[22].

We did not

presently

get

systematic experimental

values of sound velocities which allow us to determine

exactly

80

[by

-4 oc

(p,

the mass

density)].

However,

we intend to define a calorimetric

(JD

PVD

OD

value that we can compare to those of

L.Q.

alloys

determined in the same way.

The main results are :

a)

OD

of

sputtered samples

are lower than for

L. Q. ,

even after structural relaxation

by

subsequent

heat treatment ;

b)

like for y, at variance to

L. Q.

alloys,

OD iS

sensitive to structural relaxation. One observes a

systematic

increase,

whereas such variations are not

systematic

in

melt-spun

samples :

e.g.

only

in

Zr60Cu40 (increase

of 10 %

[27])

and in

ZrS4Cu46

(surprisingly

a decrease

of 2 %

[26]).

3.2.3

Superconducting

transition.

a)

The

superconducting

transition

tempe rature Tc

(Tab.

Il and

Fig.

10)

exhibits universal

properties

for the different kinds of Zr-Cu

amorphous

alloys. Firstly,

similar values in the

as-prepared

state,

despite

the different

degree

of structural disorder. The calorimetric transition width is somewhat

larger

for the

sputtered samples

(0.30

to 0.45

K)

than for

L.Q.

(0.1

to 0.4

K).

Notice that this rather

large

width cannot be

entirely

accounted for

by

the chemical concentration fluctuations which are of about 1 at% between the mean values of different

foils,

or up to 2 at% within a same foil of a

sputtered sample,

and which would

correspond

to

widths of 0.1 to 0.2 K.

Fig.

10. -

Superconducting

transition temperature versus x.

Sputtered :

same

symbols

as in

figure

9 ;

melt

spun

(as-prepared state) : (e)

from reference

[11] [resistive measurement] ;

(x)

from reference

[23].

The effect of thermal

annealing

is indicated for

melt-spun

Zr54CU46 [0,

Ref.

[26]].

Despite

the very different numerical values and behaviour of the electronic D.O.S., the

Tc

seems to be universal with a

dT

relative

decrease

à T,:

of about 0.1 K/at%.

(15)

Secondly,

after heat treatment,

Tc drops by

about 0.3-0.4 K

independently

of the

alloy

concentration,

but the width remains almost

unchanged. Again

this behaviour is universal for the different kinds of

alloys, despite

the different behaviour on relaxation of the

physical

parameters

which are

supposed

to govern

Tc,

mainly (JD

and N

(EF).

For

example,

the increase of

0 D

on

annealing,

which could

explain

the

7c

depression,

is not

systematic

for

L.Q.

alloys.

As

previously reported

[28, 31],

this calorimetric determination is in

good

agreement

with the electrical

resistivity

transition which

naturally corresponds

to the

higher

side of the calorimetric one, and with a width of the order of 100

mK,

in better

agreement

with the concentration fluctuations for a much smaller

piece

of

sample

used in the

resistivity

experiments

(about

1 mm x 15

mm).

b)

Electron-phonon coupling strength.

In absence of

tunneling

measurements on these

alloys,

we use the

general

McMillan numerical formula

[32]

for transition metallic

alloys,

which determines the

coupling

parameter À

from

Te and OD :

with

* = 0.13 for transition metals .

(1)

Values of À

[Tab. II]

indicate that the

coupling strength

deviates

progressively

from weak to

intermediate when the Zr concentration increases from 35 to 80 at%. This

progressive

and very continuous behaviour has also been

pointed

out

by

the

precise analysis

of the

thermodynamic

critical field

Hc(T)

determined from

Ces (T)

below

Tc,

and estimation of the B.C.S.

parameters

such as the condensation energy, or the deviation function of

H (T) [31].

After heat

treatment, À

decreases

by

0.03-0.04.

For the «

as-prepared

»

alloys, À

for the

sputtered

is

higher

than for

L. Q.

by

about 0.05 in

the whole concentration range. Since their

Tc’s

are similar

(cf.

Fig.

10),

we suppose that the use of McMillan formula

implies

that the increase of À is

compensated by

the decrease of

0 D.

However,

in absence of

tunneling

measurements which would

give directly

the value of

À,

this

hypothesis

remains to be confirmed.

c)

At the

transition,

one can determine as indicated in

figure

6 the

specific

heat

jump

AC and estimate the value

of T âC

=

Ces (Tc) - Cen(Tc)

which

are

reported

in table II.

They

Y TC

Tc C en (Tc)

are

systematically

somewhat

larger

(for x --

50

at%)

than the B.C.S. value of 1.43 for weak

coupling superconductors,

and almost similar to

L.Q.

alloys,

without a clear trend to increase with the Zr content, as one could

expect

for an

increasing coupling strength.

4. Discussion of the normal state and

T,.

4.1 CONCERNING

N (EF)

AND

OD-

- In a

general

way, it has been shown before that

sputtered alloys

are in a

higher degree

of structural disorder than

L. Q.

ones on every scale of

the structure :

short,

medium and

long-range

order. We intend to

try

to connect this

specificity

to the different

contributions,

mainly

those of

phonons

and

electrons,

whereas this connection appears the most obvious for the

low-energy

excitations,

as discussed elsewhere

[7].

a)

With this

property,

are consistent the

Debye

temperature

(OD)

values,

which are lower

than in

L.Q.

alloys,

due to a more

loosely

connected lattice. This is

supported

by

a

tigth-binding

model

proposed by Cyrot-Lackmann

[33]

which

explains

semi-quantitatively,

for

(16)

the shear

modulus,

which is related to the lattice sound

velocity by

v,

= ( 2013 ) ,

p

being

the

P

mass

density.

The

theory predicts

a variation of

OD

of the order of 10 %

betwèen

the

amorphous

and the

crystalline

states. But

experiments

in Zr-Cu

(Fig. 9)

indicate that

indeed,

it can be more

higher ;

the

drop

of

OD

increases with an

increasing

structural disorder of the

amorphous

state as shown

by

the

following

results for

Zr68Cu32 :

crystalline

state :

0 D

= 315

K ;

liquid

quenched

(amorphous) :

189 K

(40

% of relative

variation) ;

sputtered :

annealed

(amorphous)

(JD

= 163 K

(48 %)

and

as-prepared, OD

=

156 K

(50

% of relative

variation).

Annealing

induces in the

sputtered alloys

a

systematic

increase of

6D,

due to an increase of stiffness of the material. A rather similar behaviour occurs for the

low-energy

(TLS)

excitations

[7] :

a much

larger

density

of states

comparatively

to

L.Q.

alloys

and its

systematic

reduction on

annealing.

Both

correspond

to a much

higher

overall D.O.S. of the

low-energy

vibrational

spectrum

(including

the

TLS,

defined as

configurational

localized

excitations)

which is

depressed by

the structural relaxation.

b)

The consequence of a less stable structure is

particularly striking

in the case of the electronic D.O. S. at the Fermi level :

N y(Ep)

values are much

larger

than for

L.O.

alloys,

especially

at

high

Zr content. For

ZrSlCu19,

a

composition

which cannot be

prepared by

melt-spinning,

the variation of

N,(EF)

between two different

samples

(I

and

II)

or for a same

sample

(I)

upon the thermal

history,

reaches a factor of two ! As

reported

previously,

there is a cross-over effect of

NY(EF)

at this

high

Zr content. Note that

larger

effective

Ny

values reflect very

probably higher

D.O.S. of the band structure at

Ep,

N(0),

since renormalization

effects,

supposed

here to be limited to

electron-phonon

interactions,

are

similar for both kinds of

materials,

sputtered

and

L.Q.

A first

explanation

for these differences could be in differences in the chemical

short-range

order

[C.S.R.O.].

For

example, studying

the Zr-Ni

system,

Kroeger et

al.

[29]

found a

peculiar

behaviour of

NY(EF).

They

observed

rapid

variations of

NY(EF)

between 60 and 65 at% Zr. These values exceeded

by

20-25 % those observed

by

Altounian and Strom-Olsen

[11],

obtained from

H,2

and resistive measurements, at about the same concentrations. These features have been ascribed to

rapid changes

of the C.S.R.O. due to a

competition

of two

amorphous phases

Zr3Ni2

and

Zr2Ni.

But in our case, this

interpretation

is not

expectable :

EXAFS measurements show that C.S.R.O. is absent in the

sputtered Zr77Cu23

and

Zrg1Cu19 ;

instead,

there is a

slight tendency

to

clustering

for Zr atoms

[3].

More

generally,

it

seems that chemical

short-range

order does not characterize the ZrCu

system

when Zr

content is

higher

than 50 at%

[19],

at variance with ZrNi. It is therefore difficult to test the role of C.S.R.O. in the ZrCu

alloys,

which has been

suggested

[34]

to lead to a decrease of N

(EF).

On the other

hand,

our data agree with the

general

ideas of Morruzi et al.

[35]

about the

relationship

between the electronic d-band

properties

and the

stability

of transition-metal

glasses. They

argue that

high

D.O.S. at Fermi level are characteristic of the relative

instability

of their

short-range

atomic

arrangements.

At

variance,

energetically

favorable atomic

arrangements

will lead to small

N (EF).

That is to say that a

high

N (EF )

implies

a

high

free energy, less-stable

configuration.

These

arguments

were tested on the y values of

crystalline

and

amorphous

Zr-Ni

alloys

by

Kuentzler

[36].

Characterization

data,

particularly

the irreversible

decay

on

annealing

of the

enthalpy

and

of the electrical

resistivity

(Zr77CU23),

much

higher

in

sputtered

than in

L.O.

alloys,

are

consistent with this less stable

configuration

in

sputtered alloys.

This

hypothesis

is also consistent with both the smaller

Debye

temperature,

indicative of less cohesive local

(17)

excess free-volume.

Hence,

are understandable the

general

higher

values of

N,(EF).

However,

the

unexpectable

sense of variation of y on

annealing

cannot agree with a

systematic

trend toward a more « stabilized » state.

According

to these

ideas,

one can also account for the differences in

NY(EF)

between different series of

L.Q.

alloys

and the

sensitivity

of some of them to thermal treatment,

features which have still not been matter of discussion. Since

L.Q.

alloys,

due to their much lower effective

quenching

rate than for

sputtering,

have

already

reached a more « stabilized »

state, one could

expect

almost similar values of

N,(EF).

However,

experimental

data show

that,

even among

L. Q. ,

there exist differences which exceed

largely

the uncertainties of the

experiments.

For

example,

in the ZrNi

system,

values of

N y

from references

[29]

and

[24]

exceed

largely

those of Onn et al.

[37] ;

and the

striking

behaviour of

N y

at Zr content around 60-65 at% described

by Kroeger et

al.

[29]

has not been seen

by

other groups

[11, 37].

In the case of

reference

[29],

samples

were

prepared by

arc-hammer

technique

in

comparison

to

melt-spinning

for two other groups. For

ZrCu,

results of Garoche et al.

[22]

indicate

higher

values than Samwer and

Lôhneysen

[23],

both groups

using

the

melt-spinning technique.

Within the scheme of Moruzzi et

al.,

one could

explain

these differences

by

differences in the

cooling

rates

during

the

glass-forming

process, which lead to more or less « stabilized » structures,

and therefore to different electronic

D.O.S. ;

and some of them will be

significantly

sensitive

to thermal treatment. This is

supported, firstly, by

the results of

Kroeger et

al. on

ZrNi,

and

secondly

for

ZrCu,

by

the observable effects

(increase

by

5-10

%)

of structural relaxation on

ZrS4Cu46 [26], Zr72Cu28

and

Zr6gCu32

[27]

whereas no effect was

reported

in

Zr6oCu4o [27]

and

Zr70CU30 [38].

c)

An

unexpected

feature is the

discrepancy

which exists between the values of

N y (EF)

and

N H,:2

(EF),

the electronic D.a.S. determined within the GLAG

theory, by

the

resistivity

p and the

slope

of

Hc2 at T,,

with

N y

always exceeding

N H,:2*

This

disagreement

could result from these two different ways

(local

electronic

transport property

or

global

thermodynamic

property)

used to

study

the

superconductivity,

as

already

discussed

by

Laborde et al.

[31]

and in reference

[29].

But,

it is

striking

that the

amplitude

of this

discrepancy

varies

accordingly

to the

technique

of

preparation

of the

amorphous

state :

- for the series ZrCu obtained

by melt-spinning,

this

amplitude

varies between 6 % and 17%

[11,

22,

23] ;

- for ZrNi obtained

by

the arc-hammer

technique,

it is about 20 %

[11, 29] ;

- for

sputtered

ZrNi

[31]

or ZrCu

[28],

it varies between 20 %

[Zr77CU23 ]

and 100 %

[Zr8lCU19].

Note that for

Zr-Cu,

NH "2(EF)

is similar for both kinds of

alloys.

An extended discussion

[40]

about the determination of y from p and the

slope

of

Hc2 at Tc,

shows that one must be careful when

using

this method for calculation of the D.O.S.

at

EF.

Indeed,

experimental

uncertainties are

typically

of about 10

%,

mainly

due to the

resistivity

measurements.

But,

in the case of Zr-based

alloys,

the

discrepancy

exceeds the uncertainties. It can no more be due to

inhomogeneity

of the

sputtered

samples,

because all characterization

measurements indicate that the whole series is

homogeneous

[41].

It appears that the

discrepancy

is actual and

probably

related to the

degree

of structural disorder of the material. In the case of the

sputtered

ZrCu

(or ZrNi)

alloys,

one can assume that the electronic

D.O.S.

thermodynamically

determined,

largely

in excess to that of

L.Q.

alloys,

and which does not intervene in the

superconducting properties,

is of localized nature. A structural

origin,

which is now

being

tested

by

transmission electron

microscopy,

could be in clusters of

Références

Documents relatifs

Following these authors bulk effect represents a reversible conductivity change and surface effects can be attributed to a variable charge density. at the surface as

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

Here we focus our attention to the difficulty in obtaining mutually consistent results for the residual resistivity (p ), the temperature coefficient of resistivity

The ferromagnetic state begins to disappear below a critical concentration correspond- ing to an average distance between impurity atoms of 15-20 A. It is possible that

For many of these alloys, particularly those without metalloid components such as YFe /17/ ZrCu 1181 and TiZrBe 16,191, there is never an extended range li- near in temperature but

In the pre- sent study we report the temperature dependences of the magnetic anisotropy and the composition depen- dence of the distribution of the magnetic hyperfine fields

Introduction.- A recent experimental work on the concentration and temperature dependence .of the ma- gnetic moment in rapidly quenched amorphous Fe-B alloys /I/ has shown that

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