HAL Id: jpa-00246702
https://hal.archives-ouvertes.fr/jpa-00246702
Submitted on 1 Jan 1992
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.
Crossover from equilibrium to nonequilibrium dynamics in a reentrant AuFe ferromagnet
P. Mitchler, R. Roshko, W. Ruan
To cite this version:
P. Mitchler, R. Roshko, W. Ruan. Crossover from equilibrium to nonequilibrium dynamics in a reentrant AuFe ferromagnet. Journal de Physique I, EDP Sciences, 1992, 2 (12), pp.2299-2309.
�10.1051/jp1:1992282�. �jpa-00246702�
Classification
Physics
Abstracts75.508 75.50L 75.60L
Crossover from equilibrium to nonequilibrium dynamics in
areentrant Aufe ferromagnet
P.
Mitchler,
R. M. Roshko and W. RuanDepartment
ofPhysics, University
of Manitoba,Winnipeg,
Manitoba, Canada R3T 2N2(Received 26 May 1992, accepted in
final
form Ii August 1992)Abstract. Measurements of the
decay
of the low field thermoremanentmagnetization,
over four decades of observation time, 6 sw t
w10~
s. as a function of temperature T and wait time t~, have been
performed
on two Aufealloys
located on either side of the multicriticalpoint
: a Au~ofejospin
glass and astrongly
bond-disordered reentrant Au83Fej~ ferromagnet. Thespin glass
relaxation isotherms exhibit the canonical symptoms ofnonequilibrium
ordering belowTso,
the isotherms are characterized by an inflectionpoint,
on alogarithmic
timeperspective,
which is sensitive to the wait time t~,
display
a memory for ahistory
of field reversals, and are allcompatible
with ananalytical representation
based on a stretchedexponential
function,M~(i)=Mo+Miexp[- (i/r)~~"],
with r =t,~~ and a weakly temperature and wait time dependent exponent n m 0.7. The reentrant ferromagnet exhibits atemperature-driven
crossoverbetween two distinct relaxation
regimes
a «high
temperature»regime
coincident with theferromagnetic
phase, characterizedby
an extremelyrapid
approach toequilibrium, negligible
wait time effects, and a weak power lawdecay, M~(t)
=
Mo
+Mi
t~~, with a weakly temperaturedependent
exponent 0.04 w m w 0.08, consistent with thepredictions
of droplet fluctuation models of randomIsing ferromagnets,
and a « low temperature »regime
coincident with the reentrantphase,
withnonequilibrium, age-dependent
dynamics which are virtuallyindistinguishable
from those in the« pure » spin glass
phase
and describable by the same stretchedexponential
relaxation function. Thisstudy
represents the firstsystematic comparison
of relaxation dynamics in thereentrant and
ferromagnetic
phases of a system withsequential
transitions, andprovides
compelling
evidence for theequivalence
of the reentrant andspin
glass phases.1. Introduction.
Ferromagnets
with substantialexchange
bonddisorder,
due to some forni ofquenched
structuralrandomness, frequently exhibit,
uponcooling, apparently sequential magnetic
transitions from paramagnet to
ferromagnet
to a low temperaturephase
with many of the elements(strong irreversibility, displaced hysteresis loops) commonly
associated withspin glass freezing.
The Aufe system, inspite
of its well documentedmetallurgical complexities [I],
iswidely recognized
as thearchetype
of such « reentrant » behaviour and exhibits theclassic
symptoms [2]
: the lowfrequency dynamic
response first increasesrapidly
withdecreasing
temperature, then passesthrough
atypical ferromagnetic Hopkinson
maximum(if
the
susceptibility
is not limitedby
thedemagnetizing factor)
followedby
aweakly
temperaturedependent
«plateau
», and thenabruptly
decreases(in
a manner reminiscent of a purespin glass
below itsfreezing
temperatureTso)
as theferromagnetic phase collapses
into the« reentrant »
phase.
Infiniteranged
vectorspin
models formagnetic
systems withcompeting
exchange
interactions[3] interpret
the reentrant transition as the onset of transversespin glass freezing
in the presence oflongitudinal ferromagnet ordering
; while localprobes,
such as M6ssbauer spectroscopy[4]
and inelastic neutronscattering [5],
appear to support thisidentification, they
do not indicate that thisphenomenon
is critical in the sense of agenuine cooperative thermodynamic phase change,
nor has there been anysystematic attempt
to characterize the relaxationdynamics
within the reentrantphase,
which areexpected
to exhibitnonequilibrium
behaviour similar to that which dominates the relaxation response of purespin glasses.
According
to aphenomenological theory [6-8] developed by
Fisher and Huse forIsing
systems withshort-range interactions,
the lowlying
excitations which dominate thelong-
distance andlong-time
correlations in the orderedphase
of systems with discrete broken symmetry aredroplets (or domains)
ofcoherently flipped spins
enclosedby
asingle
wall.Droplet
formationrequires
activation over energy barriers B which grow withlength
scale L as B~L*, leading
todroplet
lifetimes rexp(flL*),
and hence tologarithmic droplet growth, L(t)~ (Tlnt)~~~.
If aspin glass
issubjected
to an idealquench
from infinitetemperature to a temperature T ~
Ts~
in zeromagnetic
field,then,
after a time t~ haselapsed,
the system will consist of domains oftypical
linear dimensionL(t~)
within which thespin
configuration
islocally equivalent
to one of the two pureequilibrium
states[6]
characteristic of thetemperature T,
and twotemporal
relaxationregimes
aredistinguishable [7]
: an «early epoch
»(In
t « Int~)
where relaxation occurs onlength
scales L «L(t~) through
activatedequilibrium droplet fluctuations,
and the correlationsC(t) decay
asC(t)~L(t)~Y~
(In t)~Y~*,
and a « lateepoch
»(In
t»Int~)
where thedynamics
are a consequence ofactivated
nonequilibrium
domaingrowth
and thedecay
isgovemed by
anonequilibrium
exponent A ~y which leads to a more
rapid logarithmic
powerlaw, C(t)~L(t)~~
(ln
t)~
~~*. Thermoremanent relaxation measurements allow bothregimes
to be observed in thesame
experiment [9]:
thesample
is cooledrapidly
in a field H to a temperatureT~TS~,
and, after a time t~ haselapsed,
the field isabruptly
reducedby
an amountAH which is small
enough
that theoverlap length
between theequilibrium
states(T, H)
and(T,
HAH)
is muchlarger
than thetypical (T, H)-domain
sizeL(t~).
On alogarithmic
timeperspective,
the relaxation curves exhibit an age(t~ )-dependent
inflectionpoint [9],
whichrepresents
the crossover between the two relaxationregimes,
andwhich,
inpractice,
is often modelledby
a stretchedexponential
functionM(t )
exp[- (t/r
)~ ~],
withr = t,n~mt~. The theoretical
justification
for this choice isprovided by
themesoscopic
domain model ofKoper
and Hilhorst[10]
whichpostulates
that, if anonequilibrium spin glass
is located in a heat bath of constant temperature T and in a constant
magnetic
fieldH
then,
for anyarbitrary
choice of(Ti, Hi ),
the system will becomposed
of(Tj, Hi )-domains
which are characterizedby
aspecific
set of(Tj, Hi )-spin correlations,
and which will growwith time but never exceed the
overlap length I (Tj T, Hi
H),
which defines the maximumlength
scale over which thespin
correlations for the twothermodynamic equilibrium
states(Ti, Hi )
and(T, H)
areindistinguishable.
The relaxation response function of aspin
in a domain with atime-varying
sizes(t)
is acomplicated
functional ofs(t)
and, as in the Fisher and Huseapproach,
this model features aninterplay
between twolengths,
a characteristic size I(t
for agrowing
domain and theoverlap length
I(AT,
AH ). In a fieldjump experiment
of thetype described
above,
where the system isquenched
toT~TS~
at t =0 in a fieldHo
which issubsquently
turned off at t= t~, the model
predicts that,
when AH=
Ho
issufficiently
small thatI(t~)~i(0, AH),
the relaxation rate exhibits a maximum at r'w t t~ m t~,signalling
a crossover between a slowequilibrium decay
and a morerapid nonequilibrium
stretchedexponential decay.
In this paper, we present a detailedcomparison
of thermoremanent relaxation in thespin glass
and reentrantferromagnetic configurations
ofAufe,
and we show that the reentrantferromagnet
is characterizedby
two distinct relaxationregimes corresponding
to the twosequential
orderedphases
: a low temperaturenonequilibrium regime,
coincident with the reentrantphase,
in which the relaxationdynamics
areessentially
identical to those in the pure
spin glass phase,
and ahigh temperature equilibrium regime,
coincident with the
ferromagnetic phase,
with anentirely
different functionalrepresentation.
2.
Experimental
details.Two
alloys
of Aufe, one aspin glass containing nominally
10 at. fGFe and the other a reentrantferromagnet containing nominally
17at.fGFe wereprepared by
arcmelting appropriate
amounts of 99. 99 fG pure Au wire and 99.99 9b pure Fe wire on the water cooled copper hearth
of a
titanium-gettered
argon arc fumaceusing
a tungsten electrode. The choice of thesecompositions
was based on the definitivedynamic magnetic
characterizations of SarkissianII
], and,
as our static measurementssubsequently confirmed,
thesimilarity
of the fabricationand heat treatment
techniques ultimately
ensured verycomparable magnetic products.
Eachalloy
wasrepeatedly
inverted and remelted in order to achieve ahomogeneous consistency,
and calculations based on total
melting
losses indicated that the true concentrations were(at worst)
within ± 0.3 at.fG of the nominalvalues,
in each case. Each of theoriginal ingots
wascold rolled into a
sheet,
and onesample
wasprepared
from each sheet: the reentrantAu~3Fej~ ferromagnet
was athin, needle-shaped sample,
with dimensions 0.08 mm x 0.8 mm x10.4 mm, while theAu~ofejo spin glass
was athicker, rod-shaped sample (for
enhancedsensitivity),
with dimensions 0.9 mm x0.9 mm x5.2 mm. Eachsample
wasencapsulated
in a qualtz tube in an argonatmosphere,
annealed for 3days
at900
°C,
and thenquenched rapidly
into water, in order togenerate
anatomically
disordered state, and an EDAXanalysis
was consistent with ahigh degree
ofhomogeneity
andnegligible clustering.
The staticmagnetization
and remanent relaxation measurements wereperformed
with a variable temperature, variable
frequency SQUID
susceptometer,operating
in themagnetometer mode
(w
=
0),
as described in detail elsewhere in the literature[12].
3. Data
analysis
and discussion.Figure
I shows the temperaturedependence
of the low field staticmagnetization
of theAu~ofeio spin glass
and the reentrantAu83Fej7 ferromagnet,
measured under both fieldcooled
(PC)
and zero field cooled(ZFC)
conditions. Thespin glass
exhibits the canonicalcusp-like
structure atTs~
=
32K,
indicative of randomspin freezing,
and thegradual
bifurcation of the
M~c
andMz~c
curves belowTs~ symptomatic
ofirreversibility
; the reentrantferromagnet
exhibitsmultiple
structure(identified by
the vertical arrows inFig. lb)
consistent with thedynamic
measurements ofSarkissian[11],
and thus with the formation of aferromagnetically
ordered state below about T~ m100K,
whichsubsequently collapses
nearT m 40 K into a low
temperature
reentrantphase,
within which the temperaturedependence
of the PC and ZFCmagnetizations
mimic those in the purespin glass phase.
In order to
fully appreciate
thesignificance
of the relaxationdynamics
of the Aufe system in its reentrantconfiguration,
it is first essential to characterize the thermoremanentdecay
within the purespin glass phase,
andfigure
2 summarizes the essential features of thisdecay
at atypical
measurement temperatureT~
=
22K. Each relaxation isotherm in
figure
2a wasobtained
by cooling
thesample
in anapplied
fieldH~
= 5.0 Oe from a referencetemperature
la)
Au~fe~~
Ha=500e 16)Au~fe~~
Ha.05/~
~OO~OOOOOO~~
Fcoo°°
'~
~o j
°~ ~~i
°~°°~
~
~~
Cl
, j LO ,
f I '.
.°
i
w O 'O ©
~ O*
2 ~
OO ~
*
°
' O
° O. ZFC ~~'
.' e
5o o ioo
T(K)
Fig. I. (al The temperature dependence of the field cooled (FC) and zero field cooled (ZFC) static
magnetization
of theAu~ofejo spin
glass measured in anapplied
field H~ = 5.0 Oe. (b) The temperaturedependence
of the FC and ZFC static magnetization of the reentrant Au83Fe17ferromagnet
measured inan
applied
field H~ = 0.5 Oe. The vertical arrowsidentify
the structural features which also characterizethe measurements of the
dynamic
response in reference [11].(~) T~= 22 K (bl
fl
6O ~i
'I
w Q~
7
9 _.~ 9
~~ .~ m
~
s
s
2 3 4 2 3 4
log t
Fig. 2. (a) The
decay
of the therrnoremanentmagnetization
of the Au~ofeio spin glass at a typicalmeasurement temperature
T~
= 22 K,plotted
on alogarithmic
time scale, for five different wait timest~. The vertical arrows mark the locations of the inflection
points
as determined from the maxima in the relaxation rate S(t ), and the solid curves illustratetypical
fits toequation
(I ) for two extreme wait times.(b) The
numerically
calculated relaxation rate S(tw
am~lafn
t for four of the five relaxation curves infigure
2a,plotted
on alogarithmic
time scale.T~
=
60K in the
paramagnetic regime,
where relaxation effects werenegligible,
to themeasurement temperature
T~
=
22K
(the cooling procedure
was veryreproducible
andyielded cooling
timesconsistently
close to t~ = 500 s),waiting
for apredetermined
time t~ toelapse
at constant temperatureT~,
thenabruptly removing
theapplied
field andrecording
the
decay
over four decades of observation time 6s~ t
w10~s.
Ona
logarithmic
timeperspective,
the relaxation isotherms all exhibit well defined inflectionpoints
which coincide, withinexperimental
error, with the effective age of the system, so that t,~~ = t~ + t~ this agedependent
behaviour is a manifestation of thenonequilibrium
nature of thespin glass
state, and translates into a maximum in the relaxation rateS(t)w %M~/%fn
t at t~ = t~ + t~ which propagates towardslonger
observation times withincreasing
t~, as shown infigure
2b. The existence ofmacroscopically long
relaxation times in theAu~ofeio spin glass ultimately implies
that thedecay
of the thermoremanentmagnetization depends
on thehistory
of thesample
and should thusdisplay
a memory for past events. This isvividly
illustrated infigure 3,
which shows the time
dependence
of the thermoremanentmagnetization
measured atT~
= 22 Kafter
thesample
wassubjected
to the sequence of field reversalsdepicted
in the upper half of thefigure
; the oscillations observed in zeroapplied
field are a manifestation of theprincipal
ofsuperposition [13]
which assumes that the memory of a fieldchange
is noterased
by
asubsequent reversal,
and which describes the net relaxation response as the sum ofa sequence of
essentially independent
responses..
,' P~ i
~ ~ ~~
'~
,m
°., ~/
~
2 3 4
log
Fig.3.-The
memory effect: the timedependence
of the therrnoremanentmagnetization
of theAu~ofeio spin glass
after thesample
was cooled in zero field to the measurement temperature T~= 22 K and then subjected to the sequence of field pulses shown in the upper half of the
figure.
In order to
permit quantitative comparisons
with the reentrantdata,
it was instructive to establish a functionaldescription
of the relaxation isotherms infigure
2a. Of the variousempirical
and theoreticalrepresentations proposed
in theliterature,
thesimplest
formulation which wascapable
ofconsistently
andaccurately replicating
theprincipal
structural features of theexperimental
data overvirtually
the entire observational timewindow,
consisted of thesuperposition
of a stretchedexponential
and a constant termMR(t)
=
Mo
+Mi
expi- (t/r
)~~~i (I)
(It
should be mentioned here that doublelogarithmic plots
ofM~(t)
as a function ofin
t, constructed toverify
the Fisher and Huseprediction
thatM~ (in t)~
Y, do notyield
straight
lines over any extendedtemporal interval,
but rather exhibit continuous curvature witha
monotonically increasing negative slope, although
thesystematic
trend towardslarger
values of y withincreasing
observation time isgenerally
consistent with the crossover from slower tofaster
dynamics expected
within thedroplet
fluctuationformalism.)
Table I presents acomprehensive
list of the best fit values of theparameters Mo,
n, and r,along
with a number of other relevantexperimental
parameters, for a sequence of relaxation isotherms(including
those in
Fig. 2)
with measurement temperaturesT~ spanning
much of thespin glass phase,
and the solid curves infigures
2a illustratetypical
fits for two extreme wait times. For wait times t~ w 240 s,equation (I) provides
acomplete description
of the measured isotherms over the entireexperimental
time interval(6
s ~ t w10~ s),
and the characteristic time r agrees, withinexperimental
error, with the location of the inflectionpoint
t~n~ deduced from the maxima in the relaxation rate S(t ).
Thiscorrespondence
continues to hold forlonger
wait times t~ ~ l 000 s,although
the range ofvalidity
of the stretchedexponential
fit becomesprogressively
narrower,with
systematic
deviations apparent at short observation times(solid
curve for t~ =3 600 s in
Fig. 2a).
Such behaviour is consistent withprevious investigations [14],
which suggest that theapplicability
of the stretchedexponential
is restricted toexperimental
time intervals in the immediatevicinity
of the effectiveexperimental
age of the system, where the relaxation rate is dominatedby
the influence of theaging
process, but that it isincompatible
with the short time(t
«t~) equilibrium
response, which isexpected
to be a weak power lawdecay [14].
Themagnitude
of the exponent n istypical
ofspin glasses,
and exhibits aslight
increase withmeasurement temperature for
T~/Ts~
~
0.8,
as well as a weak inversedependence
on the wait time t~, which isjust resolvable, experimentally.
The introduction of the constant termMo
is difficult tojustify
onphysical grounds,
and isprobably symptomatic
of a further limitation inherent in the stretchedexponential representation,
which vanishes toorapidly
toprovide
a properdescription
of the actuallong
termdecay
to zeromagnetization [9]
;nevertheless,
such a term is an essentialingredient
of otherempirical representations [9, 15]
ofspin glass relaxation,
even those[15]
which do notexplicitly
invoke a stretchedexponential
component, and exhibits similar temperaturedependent systematics [9].
Table I.
Au~ofejo spin glass
parameters.Tel) ~(s) ~(8) n (s)
13 60 640 600+200 730+10 0.68+0.01 1.63+0.01 6-10'
20 60 610 600+200 630+8 0.71+0.01 1.16+0.01 6-10'
22 60 610 360+100 440+6 0.67+0.01 0.96+0.01 6-10'
22 240 720 700+100 690+8 0.67+0.01 0.93+0.01 6-10'
22 1200 1680 2000+600 2050+60 0.65+0.01 0.88+0.02 10-10~
22 1800 2280 3000+600 2060+70 0.62+0.01 0.98+0.02 20-10'
22 3600 4050 4000+1000 4006+240 0.69+0.01 0.98+0.04 160'-10'
24 60 570 250+100 290+3 0.70+0.01 0.70+0.01 6-10'
26 60 780 350+100 330+4 0.72+0.01 0A7+0.01 6-10'
28 60 630 250+100 185+3 0.73+0.01 0.30+0.01 6-10'
The characterization of the thermoremanent relaxation in the reentrant
Au~~fei~ ferromagnet proceeded by adapting
theexperimental
andanalytical techniques developed
for theAu~ofejo spin glass.
Thetendency
for residual relaxation effects topersist
well above thepeak
in the ZFC
magnetization (see Fig. lb)
necessitated the choice of acomparatively high
reference
temperature T~
= 140 K forestablishing
the zero level ofmagnetization,
so that thereentrant
study
wasseverely
constrainedby
theanomalously large cooling
intervalsT~-T~ (~100K
for the reentrantphase)
andcorrespondingly long cooling
times t~(typically
l 000s),
which allowed thecooling
process to assume a dominant role indefining
the age of the system.Figure
4 summarizes thethermorenianent decay
fora sequence of
representative
measurement temperaturesT~ spanning virtually
the entire orderedphase (15
KmT~w100K),
with all isothermscorresponding
to an identicalcooling
fieldH~
= 2 Oe and wait time t~ =60 s. While the structure of these isotherms
(particularly
at the lowertemperatures)
is much broader andconsiderably
more subtle than that associated with« pure »
spin glass
relaxation(undoubtedly
a consequence of thecomparatively
slowcooling rates),
there is nevertheless anunmistakable, progressive change
in curvaturethroughout figure 4,
from concave down to concave up withincreasing
measurement temperatureT~. Furthermore, quantitative analysis
of the isotherms reveals that there are nvo distinct typesof relaxation
behaviour, corresponding
to two distincttemperature regimes
each of which coincides with one of the two orderedphases
which characterize the reentrant sequence, and the isotherms infigure
4 have beengrouped accordingly
in order to reflect thisdichotomy
:(a)
a low temperature (« reentrant
») regime
infigure
4a withT~
~ 35K,
where the isotherms all possess inflectionpoints
and exhibit a wait timedependence
indicative ofnonequilibrium
processes and describable
analytically by
thesuperposition
of a stretchedexponential
and aconstant term as in
equation (I),
and(b)
ahigh
temperature («ferromagnetic ») regime
infigure
4b withT~~35K,
whereaging
effects arenegligible
and where thedecay
isincompatible
with a stretchedexponential representation,
but is consistent with a weak power lawdependence
of the formM~(t)
=
Mo
+Mj t-m, (2)
(a) (b)
~~ Tm(K)
, so
~ ~~~ ~~
f
~70
80
uo
io ioo 1
tog t
Fig.
4.-(a)Decay
of the thermoremanentmagnetization
at severalrepresentative
temperaturesT~
within the reentrantphase
of theAug~fej~ ferromagnet,
for a wait time t~ = 60 s,plotted
on alogarithmic
time scale. The vertical arrows mark the locations of the inflectionpoints.
(b) Decay of the thermoremanentmagnetization
for t~ = 60s at Severalrepresentative
temperatures T~ within theferromagnetic phase
ofAug~Fej7.
symptomatic
of relaxation fron~ afully equilibrated
field cooled state.Figure
5provides
atypical
illustration of thequality
of th± fits for selected isotherms in each of the two relaxationregimes.
and tables II and III list all the relevantexperimental
parameters,including
the best fit values of the constantsMo,
n, r, and m inequations (I
and(2).
Based on acomparison
with the relaxationdynamics
of a purespin glass,
as definedby figure
2 and table I, themagnitude
and variation of the stretched
exponential
parameters n and r in table II arecompletely
consistent with
orientationally
randomspin freezing
within the lowtemperature
reentrantregime. However,
the constraintsimposed
on theexperimental cooling
ratesby
thenecessity
to cool the system over
relatively large
temperature intervalsT~ T~
areresponsible
fur some apparentsystematic discrepancies
; inparticular,
while theexperimental cooling technique
is unable todistinguish
bet~N.eencooling
intervalsdiffering by typically
A(TR T~ )/T~
0. I, sothat the nominal
experimental
age t~+t~ of the system isessentially independent
ofmeasurement temperature
T~
within the reentrantphase,
the location of the inflectionpoint
t~n~ = r
(see
vertical arrows inFig. 4a),
and hence the true age of the system, exhibits anobvious inverse correlation to
T~, shifting
from t,n~ ~ t~ + t~ to t,n~ ~ t~ + t~ withincreasing
measurement temperature. This behaviour
is,
however,entirely
consistent with the absence ofaging
observedthroughout
theferromagnetic phase,
andmerely
confirms that the system iscontinuously
inequilibrium throughout
the initialportion
of thecooling interval,
and thataging
processes do not commence until the system has been cooled
through
some effective reference temperatureTi
which lies well belowT~
but close to theferromagnetic-reentrant phase boundary (although
finetendency
for t~n~ to exceed all of the obviousexperimental
time(a) (b)
5emu/g
M~
05emu/g
70K
90K
~2 3 4 2 3 4
log t
Fig.
5. -(a)Typical
fits (solid curves) ofEq.
(I) to some of the reentrant relaAtion isotherms ofAug~fej~
for two different wait times t~ = 60 s(.) and t~ = 7 200 s (A). The vertical arrows mark the locations of the inflect:on points, and the zero magnetization levels of the relaxation curves have been shifted by arbitrary amounts for clarity ofpresentation,
so that only, the scale of the relaxation has been indicated in the lower left of the figure. (b)Typical
fits (solid curves) ofequation
(2) to some of theferromagnetic
relaxation isotherms ofAug~fej~.
As infigure
a,only
the scale of the relaxation has been indicated in the upperright
of thefigure.
Table II.
Au~~Fei7
reentrantphase.
t~(S) t~+t~(s) tan(s) ~s) n (s)
16 60 900 3500+1000 2860+l10 0.77+0.01 2.98+0.01 6-10'
17.5 60 900 3000+1000 2330+l10 0.83+0.01 2.92+0.01 5~10'
20 60 960 2500+500 2560+140 0.88+0.01 2.37+0.01 5-10'
20 7200 8220 > 10' (1.9+0.9)x10~ 0.87+0.01 2.75+0.03 6~10'
21.6 60 1200 1600+600 1800+180 0.91+0.01 2.82a0.01 6~10'
22.5 60 1260 1200+300 1000+70 0.92+0.01 2.80a0.02 5-10'
22.5 7200 8260 4000+1000 4800+300 0.89+0.01 2.7l+0.02 5~10'
23 60 l160 1200+500 1760+50 0.93+0.01 2.78+0.01 5~2.7x10'
23 7200 8220 > 10' (5.4+0.3)x10' 0.89+0.01 2.77a0.03 5~l.5x10'
23.5 60 1200 < 10 7.0a2.0 0.96+0.01 2.68+0.03 5-10'
Table III.
Au83Fe17 ferromagnetic phase.
T~O~) ~(s) m
40 60 0.06+0.01 1.69+0.01
45 60 0.06+0.01 1,64+0.01
60 60 0.04+0.01 lA7+0.02
65 60 0.04+0.01 lA6+0.01
60 60 0.04+0.01 1.38+0.01
60 7200 0.04+0.01 1.38+0.01
65 60 0.06+0.01 1.37+0.01
70 60 0.06+0.01 1.30+0.01
70 7200 0.04+0.01 1.30+0.01
75 60 0.08+0.01 1.27+0.01
80 60 0.07+0.01 1.18+0.01
80 7200 0.06+0.01 1.18+0.01
86 60 0.06+0.01 1.12+0.01
90 60 0.08+0.01 1.05+0.01
100 60 0.08+0.01 0.85±0.01
constants at the lower measurement
temperatures
is further evidence that slowcooling
ratesdistort the
anticipated correspondence
between t~ + t~ and tin~,and,
infact,
accelerate theaging process).
As mentioned
earlier,
the zero fieldequilibrium (t~
=co)
relaxation inspin glasses
isexpected,
on the basis of both Monte Carlo simulations[16]
as well as critical fractal-cluster model calculations[17],
to be characterizedby
a slowpower-law decay
belowTs~,
and thisexpectation
issupported by
both static anddynamic
measurements of the relaxation rate S in various canonicalspin glass systems [17].
Inferromagnets,
remanent relaxation hastraditionally
been attributed to domain rotation and wallmotion,
and models[18]
which invoke activation processes over a smooth distribution of barrierheights predict logarithmic
timedependences. However,
more recent theoreticalinvestigations
ofIsing ferromagnets, specifi- cally
those withquenched
randomexchange disorder,
suggest[8]
that power lawdecay
is alsoa feature of the
equilibrium dynamics
of randomferromagnets. Apparently,
the lowfrequency, long length
scale fluctuations in the orderedphase
of such systems withspontaneously
broken discrete symmetry aredroplets
ofcoherently
reversedspins
surroundedby
domain walls whichare
pinned
to the disorder, so thatdroplet
annihilationrequires
thermal activation overlarge free-energy
barriers which grow withlength
scale. Infact,
thelong-time
behaviour of thespatially averaged temporal spin
autocorrelation function C(t)
w
(S,(0) Sj(t)) (S,)~)
,
where
(.. )
denotes an infinite-time average and(.. )
~
a
configuration
average, is dominatedby droplet
excitations withanomalously long
relaxation timeswhich,
in randomferromagnets (through
a combination of activateddroplet
lifetimes and Boltzmanndroplet
excitationprobabilities [8]),
areexponentially
i-are inin
t,yielding
a power lawdecay
:c
(t j
~- kiln ti~ ~- tiT)
(3j
for random
exchange
where y = I.(This
result appears not to bespecific
toIsing spins,
sincerecent models
[15]
for activated relaxation ofdispersive
excitations on apercolation
distribution of fixed finite-sized
domains,
which are based onquite general geometric
considerations
independent
of suchspecifics
asspin dimensionality,
and which have beenapplied successfully
to pureHeisenberg ferromagnets [19], yield
the sameasymptotic behaviour.)
Moreover, Fisher and Huse havepointed
out[8]
that theregion
ofvalidity
of thepredicted asymptotic equilibrium decay (3) expands
to includeprogressively
shorter obser- vation times as the bond disorder becomes more extreme, so thatnearly
tricriticalferromagnets
should represent the best
experimental
candidates forassessing
theapplicability
ofequation (3).
Since these areprecisely
the conditions which favour reentrantferromagnetism,
it follows that the behaviour of the relaxation isotherms within the
ferromagnetic phase
of thereentrant
Aug~fei~ alloy (viz.,
a weak power lawdecay
with an exponent m 0.040.08,
and
negligible
wait timeeffects)
may beinterpreted
as direct confirmation of the theoreticalpredictions.
In summary, measurements of the low field thermoremanent relaxation on either side of the tricritical
point
in the Aufe system have revealed the existence of atemperature-driven
crossover between two distinct relaxation
regimes
inferromagnets
where theexchange
bonddisorder is sufficient to generate
sequential (or
« reentrant»)
behaviour one of theregimes
is coincident with the «high
temperature »ferromagnetic phase,
and, within theslow-cooling
constraints of the current
investigation,
is characterizedby
anextremely rapid approach
toequilibrium
andby
a weak power lawtemporal decay
of the thermoremanentmagnetization
consistent with the
predictions
ofdroplet
fluctuation models[8]
of randomIsing ferromagnets,
while the other
regime
coincides with the« low
temperature
» reentrantphase,
and exhibitsnonequilibrium, age-dependent
relaxationdynamics
which arevirtually indistinguishable
from those in the « pure »spin glass phase,
and which are describableby
a common relaxationfunction which has been linked to a
mesoscopic picture [10]
ofgrowth
andfragmentation
ofspin glass
domains.Acknowledgements.
This work has been
supported
in partby
a grant from the Natural Sciences andEngineering
Research Council of Canada.
References
[I] CRANE S. and CLAUS H., Solid State Commun. 35 (1980) 461.
[2] COLES B. R., SARKISSIAN B. V. B. and TAYLOR R. H., Philos. Mag. B 37 (1978) 489.
[3] GABAY M. and TouLousE G.,
Phys.
Rev. Lett. 47 (1981) 201.[4] LANGE S., ABD-ELMEGUID M. M. and MICKLJTz H.,
Phys.
Rev. B 41 (1990) 6907.[5] MIREBEAU I., HENNION M. and LEQUIEN S., J.
Appl.
Phys. 63 (1988) 4077.[6] FISHER D. S. and HusE D. A.,
Phys.
Rev. B 38 (1988) 386.[7] FISHER D. S. and HusE D. A., Phys. Rev. B 38 (1988) 373.
[8] HusE D. A. and FISHER D. S.,
Phys.
Rev. B 35 (1987) 6841.[9] NORDBLAD P., SVEDLINDH P., LUNDGREN L. and SANDLUND L., Phys. Rev. B 33 (1986) 645.
[10] KOPER G. J. M. and HILHORST H. J., J.
Phys.
France 49 (1988) 429.[I I] SARKISSIAN B. V. B., J.
Phys.
F : MetalPhys.
ii (1981) 2191.[12] KUNKEL H., RosHKo R. M., RUAN W. and WILLIAMS G., Philos. Mag. B 63 (1991) 1213.
[13] LUNDGREN L., NORDBLAD P. and SANDLUND L., Europhys. Lett. 1 (1986) 529.
[14J NORDBLAD P., LUNDGREN L., SVEDLINDH P., SANDLUND L. and GRANBERG P.,
Phys.
Rev. B 35(1987) 7181.
[15] CHAMBERLIN R. V. and HAINES D. N.,
Phys.
Rev. Lett. 65 (1990) 2197.[16] OGIELSKI A. T.,
Phys.
Rev. B 32 (1985) 7384.[17] LUNDGREN L., NORDBLAD P. and SVEDLINDH P.,
Phys.
Rev. B 34 (1986) 8164.[18] STONER E. C. and WOHLFARTH E. P., Philos. Trans. Roy. Soc. 240 (1948) 599.
[19] CHAMBERLIN R. V, and HOLTzBERG F.,
Phys.
Rev. Lent. 67 (1991) 1606.JOURNAL DE PHYSIQUE I T 2, N' 12, DECEMBER J992 83
regus
auThe
properties
of nuclei G. A. JONESSdrie :
O~fiord Physics
Series12(Oxford University Press, 1987)
ISBN :0-19~851869~2,
193 p., £20.Monographie
d'introduction h laphysique
du noyau;7chapitres: propridt£s gdndrales,
forces nud6aires, moddles, d6croissancealpha
et bdta, transitionsdlectro-magndtiques,
rdactions nuddaires, fission et fusion,appendices,
index.Aspects
of nuclear science(in
honour of Prof. A. C.Pappas)
E.HAGEBO,
B. SALBU Eds.(Norwegian University Press,
distribu6 par OxfordUniversity Press, 1987)
ISBN : 82-00~18343-2,
177 p., £ 20.Comptes
rendus d'unsymposium
tenu h Oslo (10/1985), 14 contributions.Nuclear
chemistry
A.V#RTES,
I. KISSSdrie :
Topics
inInorganic
and GeneralChemistry
22(Elsevier, 1987)
ISBN :0-444-99508~0,
619 p., Dfl.275, $
122.25.Monographie spdcialis6e
12chapitres
de la thdorie auxapplications,
index.Frontiers of
Heavy
IonPhysics
N.
CINDRO,
W.GREINER,
R. CAPLAR EdS.(World Scientific, 1987)
ISBN :9971-50~392-1, 541p.,
£ 62.55.Comptes
rendus d'une confdrence tenue h Dubrovnik, 15-19/06/87.TeV
Physics
andBeyond
R.
DELBOURGO,
J. R. Fox Eds.(World Scientific, 1987)
ISBN :9971-50-301~8,
309 p., £ 66.85.Comptes rendus d'une confdrence tenue h Launceston, Australie, 2-7/02/87.
Particle accelerators ;
applications
intechnology
and researchW. SCHARF
(Wiley, 1989)
ISBN :0-471-92206-4,
663 p., £ 77.50.Monographie sp6cialis6e,
traduite dupolonais
; 11chapitres,
index.Elementary Particles,
3rd Edition I. S. HUGHES(Cambridge University
Press,1991)
ISBN :0-521-40739-7,
431 p., £50/16.95,
$80/29.95 (relid/brochd).
Monographie
d'introduction 16chapitres
niveau secondcycle
;premibre
Edition en 1972.Neutrino
Physics
K. WINTER Ed.
Sdrie :
Cambridge Monographs
on ParticlePhysics,
NuclearPhysics,
andCosmology,
Vol. I(Cambridge University Press, 1991)
ISBN :0~521-36452-3,
670 p., £75,
125.Monographie sp£cialis6e; 6chapitres:
histoire,propridt£s intrinsdques
de neutrinos, interaction neutrinos-matidre, Etudeexpdrimentale
de l'interaction faible, Etude de la structure des nucldons h l'aide des neutrinos, neutrinos enastrophysique
et encosmologie.
Elementary
Particles and theUniverse, Essays
in Honor ofMurray
Gell-MannJ. H. SCHWARTz Ed.
(Cambridge University
Press,1991)
ISBN :0-521-41253-6, 212p.,
£24.95, S
49.95.Collection de 18 courts articles; compte rendu
partiel
d'unsymposium
tenu au CalTech 27- 27/01/1989.The
interacting
Boson-Fermion Model F.IACCHELLO,
P. VAN ISACKERSdrie :
Cambridge Monograph
on MathematicalPhysics (Cambridge University Press, 1991)
ISBN :
0-521-38092-8, 312p.,
£40, $
75.Monographie
spdcialis6e 4 parties, 12chapitres,
index.Electromagnetism (2e Edition)
I. S. GRANT, W. R. PHILLIPS
Sdrie : Manchester
Physics
Series(Wiley, 1990)
ISBN :0-471-92712-0,
525 p., £ 14.95.Livre de cours, contenu
classique,
I* ddition 1975.Electromagnetics
B. B. LAUD
(Wiley-Halsted Press, 1987)
ISBN :0-470-20746-9,
335 p., £ 24.20.Livre de cours, niveau
2ndcycle.
An introduction to
applied electromagnetism
C. CHRISTOPOULOS
Sdrie :
Wiley
Student Series in Electronic and ElectricalEngineer (Wiley, 1990)
ISBN : 0-47l-92761-9,
183 p., £ 14.95.Livre de cours, niveau
lercycle.
Differential
equations,
their solutionsusing symmetries
H. STEPHANI
Edited
by
M. MACCALLUM(Cambridge University Press, 1990)
ISBN :0-521-36689-5,
260 p.,£12.95, S
19.95.Monographie sp6cialis6e.
Introduction to
Group
CharactersW. LEDERMANN
(Cambridge University
Press,1987)
ISBN :0-521-33781-X,
227 p., £8.95, S
14.95.Monographie spdcialis£e, math6matiques grin£rales.
I* Edition en 1977.Les mdthodes tensorielles de la
Physique
: Calcul tensoriel dans un continuumstructurd,
Vol. 2J. WINOGRADZKI
(Masson, 1987)
ISBN :2-225-80575-X,
270 p., 235 F.Math6matiques gdndrales, monographie spdcialisde,
analyse tensorielle, th60rie des espaces h connection affine et espaces de Riemann.Contrblabilitd
Exacte,
Perturbations et Stabilisation deSystkmes
distribuds Tome I : Contrblabilitd ExacteTome 2 : Perturbations J. L. LioNs
Sdrie : Recherche en
Mathdmatiques Appliqudes
RMA 8-9(Masson, 1988)
ISBN : 2-225-8l477-5 et
81474-0,
540 + 270 p., 450 F les deux tomes.Monographie. Equations
aux ddrivdespartielles
de lam6canique
et de laphysique.
On numerical
approximation
in bifurcationtheory
M.CROUZIEX,
J. RAPPAZSdrie Recherches en
Mathdmatiques Appliqudes
RMA 13(Masson/Springer, 1989)
ISBN : 2-225-81794-4,
163 p., 160 F.Monographie spdcialis£e
: bifurcation des solutions des£quations
aux ddriv£espartielles, approxima-
tion et mdthode des 616ments finis.
Cyclic Cohomology,
within the differentialenvelope (an
introduction to Alain Connes' non- commutative differentialgeometry)
D. KASTLER
Sdrie : Travaux en cours 30
(Hermann, 1988)
ISBN :2-7056-6044-5,
183 p., 180 F.Monographie spdcialisde.
Feuilletages riemanniens, quantification gdomdtrique
etmdcanique
P.
DAZOR,
N.DESOLNEUX-MOULIS,
J.-M. MORVANSdrie Sdminaire sud-rhodanien de
gdomdtrie
VII(Hermann, 1988)
ISBN :2-7056-6076-3,
184 p., 160 F.Monographie spdcialisde.
TD et TD de
Topologie gdnkrale
A. FAISANT(Hermann, 1988)
ISBN :2-7056-5745-2,
304 p., 98 F.Livre d'exercices.
Chapitres topologie
fonctions continues espacessdpards,
compacts, connexesespaces
mdtriques
espaces nomads, espaces fonctionnels.Chaque chapitre
comporte un rappel de cours, des dnoncds d'exercices et deprobldmes puis
leurs solutions. Niveau 2ndcycle.
Almost
Complex Homogeneous Spaces
and their SubmanifoldsK. YANG
(World Scientific,
distributedby Wiley, 1987)
ISBN :9971-50-377-8,
l12 p., £ 22.35.Monographie spdcialisde.
Matrix
Logic
;theory
andapplications
A. STERN
(North-Holland, 1988)
ISBN :0-444-70432-9,
218 p.,Dfl.130,
68.50.Monographie spdcialisde. Chapitres
: Calcul matriciellogique
conventionnelle formulation matri- cielle application h laphysique
fondamentale, au calcul.Differential
Equations
C. M.
DAFERMOS,
G.LADAS,
G. PAPANICOLAOU EdS.Sdrie : Lecture Notes in Pure and
Applied
Mathematics Series, V. II 8(Marcel Dekker, 1989)
ISBN :0-8247-8077-9, 787p., $150.
Comptes
rendus de la confdrenceEQUADII~F
1987.Ondelettes et
opkrateurs
I : OndelettesY. MEYER
Sdrie : Actualitds
Mathdmatiques (Hermann, 1990)
ISBN :2-7056-6125-0,
220 p., 186 F.Monographie spdcialisde.
Perturbation Methods in
Optimal Control
A. BENSOUSSAN
Sdrie : Series in Modern
Applied
Mathematics(Gauthier-Villars, 1988)
ISBN : 2-04-016469-3,
573 p.Monographie spdcialis6e.
Mathematical
physics
P. K. CHATTOPADHYAY
(Wiley, 1990)
ISBN :0-470-21719-7,
352 p., £ 28.30.Cours de 2nd
cycle. Chapitres:
Variablescomplexes; dquations
diffdrentielles, du 2ndordre,probmmes
aux limites ; fonctionsspdciales
fonctions de Green ; espaces vectoriels et matrices thdorie des groupes.The
nonequilibrium
statistical mechanics of open and closed systems K.LINDENBERG,
B. J. WEST(VCH, 1990)
ISBN :3-527-26638-0,
448 p., DM88,
£34.50.Monographie
portant surquelques sujets
choisis enmdcanique statistique classique
etquantique
dessystdmes
ferrnds et ouverts. Huitchapitres
en deuxparties.
Horizons of
Physics
A. W.
JOSHI,
Ed.(Wiley Eastem, 1989)
ISBN :0-470-21329-9,
383 p., £ 32.15.Collection de