HAL Id: jpa-00246452
https://hal.archives-ouvertes.fr/jpa-00246452
Submitted on 1 Jan 1991
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.
The low field response of a reentrant NiMn ferromagnet close to the tricritical point
R. Roshko, W. Ruan
To cite this version:
R. Roshko, W. Ruan. The low field response of a reentrant NiMn ferromagnet close to the tricritical point. Journal de Physique I, EDP Sciences, 1991, 1 (12), pp.1809-1821. �10.1051/jp1:1991241�.
�jpa-00246452�
Classification Physics Abstracts
75.40G 75.50L 75.60L
The low field response of
areentrant Nilsln ferromagnet close to the tricritical point
R. M. Roshko and W. Ruan
Department of
Physics, University
of Manitoba,Winnipeg,
Manitoba, Canada R3T 2N2(Received
9July
1991,accepted
5September 1991)
Abstract. Measurements of the low field
complex dynamic susceptibility
and of the relaxation of the thermoremanentmagnetization
of a reentrant Ni-23.6 at.9b Mnferromagnet,
located very close to the tricritical point, arepresented
as a function of temperature,magnetic
field, and observation time. The temperaturedependence
of the real part of thedynamic
response in fixed staticbiasing
field exhibits atriple-peaked
structure,consisting
of a «high
» temperatureferromagnetic
criticalpeak,
and two lower temperature reentrantpeaks.
A criticalanalysis
of theferromagnetic peak yields
a Curie temperature(= (104±2)K
and an effective criticalisotherm exponent 8 with both low field and
high
fieldregimes, BLF=5.6+0.3
and8~~
=
2.9 ± 0.I, while the effective Kouvel-Fisher exponent y*, extracted from the temperature
dependence
of the zero fieldsusceptibility,
exhibits a maximum as a function of reducedtemperature and approaches the mean field limit y
=
I for T» T~. The
decay
of the low field therrroremanent magnetization, observed over four decades of time (I s ~ t « 10~s),provides
direct evidence that the
ferromagnetic
state evolves into aspin glass-like configuration
at low temperatures : within theferromagnetic phase (65
K « T« 104
K),
the relaxation is describedby
the
superposition
of a weak power law and a constant term,M~(t)
= Mo+ «o t~~', while, within
the reentrant phase
(T
~ 65
K),
thedecay abruptly acquires typical spin glass
characteristics, and the functionalrepresentation changes
to thesuperposition
of a constant term and a stretchedexponential, M~(t)
= Mo +
Mj
exp[- (t/T)' "],
with temperaturedependent
parameters n and T indicative of canonicalspin glass
behaviour.1. Introduction.
The
possibility
ofsequential magnetic phase
transitions inferromagnets
with extreme bond disorder continues tointrigue
bothexperimentalists
and theorists. Thepreliminary diagnosis
of this so-called reentrant behaviour is
typically
based on the observation of aweakly temperature dependent plateau
» in the low fieldsusceptibility,
which terminatesabruptly
in
high
and low temperatureregimes
where thesusceptibility
decreasesrapidly
with temperature(although
theprecise shape
of the«plateau»
and theabruptness
of the transitions » are influencedsignificantly by
such technical considerations as the geometry of thesample).
The currentinterpretation,
based on thepredictions
of vectorspin
models of bond disordered systems[I],
identifies thehigh
temperatureregime
with theparamagnetic
1810 JOURNAL DE PHYSIQUE I bt 12
phase,
the intermediate«plateau
»regime
with aferromagnetic phase,
characterizedby
alongitudinal
spontaneousmagnetization,
and the lowtemperature
« reentrantregime
with the onset of transversespin glass freezing
andsignificant irreversibility.
Vfhile the « reentranttransition » in some
systems
isaccompanied by
anomalous(although definitely non-singular)
structure in the
temperature dependence
of the nonlineardynamic
response[2],
reminiscent of that observed at the directparamagnetic-spin glass
transition[3]
and consistent with thepredictions
of mean fieldIsing
models[4],
the existence of agenuinely critical, cooperative ferromagnetic-spin glass phase
transition in thethermodynamic
sense remainshighly
contentious.
The
magnetic phase diagram
of the NiAln system[5]
possesses all the features of theprototypical
bond disordered(or frustrated)
magnet : thesystem
is aspin glass
for Mn concentrations c> 24
at.9b,
and exhibitspredominantly ferromagnetic
character for concen- trations c<24at.9b,
with an associated «reentrant»phase
which shifts toprogressively
lower temperatures as the Mn concentration decreases and the
ferromagnetism
becomesmore canonical. lvhile
previous investigations [6, 7]
haveprovided
acomprehensive picture
of the field and
temperature dependence
of the staticmagnetization throughout
themagnetic phase diagram,
attention hasonly recently
been focused on the low fielddynamic
response ofNimn
[8],
which exhibits a wealth of structuralfeatures, particularly
when thesystem
is in its reentrantconfiguration,
and which fumishes a valuabletechnique
forexploring
the criticalproperties
of the system. In this paper, we present measurements of the lowfrequency complex dynamic susceptibility
of a reentrant NiAlnferromagnet
located very close to the multicriticalpoint (c~m24at.9bMn),
as a function of bothtemperature (4.2
Km T<170
K)
andmagnetic
field(0
«H~
« 46Oe),
as well as a detailedstudy
of the relaxation ofthe low field thermoremanent
magnetization
over four decades of observation time(ls< t«10~s),
and over a temperature interval which includesportions
of both theferromagnetic
and reentrantphases.
2.
Experimental
details.An
alloy
of Nimncontaining nominally
23.8 at.9bMn wasprepared by
arcmelting
theappropriate
amounts of 99.997 9b pure Ni foil and 99.99 9b pure Mn flake(both
obtained from Aldrich ChemicalCo., Milwaukee)
on the water cooled copper hearth of an argon arc furnaceusing
a tungsten electrode. Thealloy
wasrepeatedly
inverted and remelted in order toachieve a
homogeneous consistency. Melting losses,
which were attributedpredominantly
to Mnvaporization,
indicated that the truecomposition
was closer to 23.6at.9b,
and this wassubsequently
confirmedby
a concentrationanalysis
of thesamples using
EDAX. Portions of theoriginal ingot
were cold rolled into thinsheets, approximately
0.17 mmthick,
and a number oflong, thin, needle-shaped strips,
withtypical
dimensions 0.3 mm x 10.0 mm, and with a calculateddemagnetizing
factor D m 0.23g-Oelemu,
were cut from the sheets. All of the needles were etched in a solution of dilute nitric acid in order to eliminate surfacecontamination.
The Nimn system
undergoes
an atomic order-disorder transformation at about 500 °C for thecomposition prepared here,
with astrongly ferromagnetic atomically
orderedNi~mn phase (T~ m700K).
In order togenerate
theatomically
disordered reentrant state, the needles wereencapsulated
in aquartz
tube in an argonatmosphere,
annealed for 3days
at 900°C,
and thenquenched rapidly
into water. Thephysical configuration
of the furnace and thebath,
whichimposed
a lower limit ofapproximately
one second on the totalelapsed quenching time,
was identical to thatemployed
in ourprevious investigations [9].
The a-c-
susceptibility
measurements wereperformed
with aSQUID
susceptometeroperating
at an excitationfrequency
ofw = 16
Hz,
andconsisting
of a multi-functionSQUID
probe
and controlelectronics,
an a.c.impedance bridge,
and abi-phase detector,
all manufacturedby BTi,
SanDiego.
Details of the cryostat andpick-up
coilassembly
areavailable in the literature
[9].
The staticmagnetization
andmagnetic viscosity
data wereacquired using
the sameSQUID
systemoperating
in the static(w
=
0,
or d-c-magnetometer)
mode.
(The extremely high sensitivity
of the astaticpair secondary pick,up
coilspermitted
this conversion from a
dynamic
to a static function to beaccomplished
without any further modification to thewiring
of theinput
circuit to theSQUID probe).
3. Data
analysis
and discussion.3,I LOW FIELD DYNAMIC RESPONSE.
Figure
I summarizes the temperaturedependence
ofthe real
part x'
of thecomplex dynamic susceptibility
X=
X'
+iX
" of the Ni-23.6 at.9b Mnsample,
measured at an excitationfrequency
w= 16 Hz and
appropriately
corrected fordemagnetizing effects,
x[~~~=
x[~~~/(l Dx[~~~),
in avariety
ofrepresentative
staticbiasing
fields between
H~
=
00e and
H~
=
45.0 Oe. Each isofield was obtained
by following
anidentical
experimental procedure
: thesample
was cooled in zeromagnetic
field from atemperature
T= 170
K,
well within theparamagnetic regime,
to T= 4.2
K,
where both the excitation field and the static field wereapplied
to thesample,
and thesusceptibility
was thenmeasured upon incremental
warming.
Themagnetic
response in zero staticbiasing
field(Fig. la)
consists of asingle, relatively sharp peak, superficially
reminiscent of that observed intypical spin glass
systems.However,
this resemblance isdeceptive and,
asfigure
16shows,
the
magnetic
response infinite
staticbiasing
fields exhibitsmultiple
structure whichsuggests
amagnetically
ordered state of considerablecomplexity (*).
The structure in each isofield(a> Ni- 23.6 at.% Mn (b)
H~=o
Oet 7~
'~ ©
O °
7
~ Wi
~ E
E ©
" /
, ~
~
12.S 15.0 20.O 25.O 35
ioo o ioo 2oo
T(K) T(K)
Fig.
I.(a)
Temperaturedependence
of the real part X' of thedynanfic susceptibility
in zero static biasing field.(b)Temperature
dependence of X' in several representative staticbiasing
fields H~ between 10 Oe and 44 Oe. For purposes ofclarity,
successive isofields have beenseparated vertically by
+0.01emu/g-Oe,
so that the vertical scaleapplies only
to the 44 Oe isofield.(*)
In this respect, the behaviour of NiAfn close to the multicriticalpoint
is similar to that of theamorphous
systemPdgo_
~fe~si~o, in its «ferroglass configuration [10].
1812 JOURNAL DE PHYSIQUE I N 12
consists of three
peaks
: thehighest temperature component experiences
asystematic suppression
inamplitude
andupward
shift intemperature
withincreasing
field which are characteristic of the criticalpeaks
associated with aparamagnetic-ferromagnetic transition,
while the two lower temperaturepeaks,
which are alsosuppressed
inamplitude,
both exhibita downward shift in temperature,
symptomatic
ofspin glass ordering.
In theregion
between the two lowertemperature peaks,
andparticularly
in theneighbourhood
of the enclosedminimum,
themagnetic
response is timedependent
: when thetemperature
is increasedincrementally
and thenstabilized,
themagnetization (which
can be monitoredsimultaneously
with the
susceptibility
in theSQUID system)
continues to relaxupward,
while the real part of thesusceptibility
relaxesdownward,
at a rate which decreasesprogressively
with time for bothobservables.
Figure
2 summarizes the variation of the threepeak
temperatures T~ with internal fieldH;
at eachpeak temperature,
themagnetization
was evaluatedby integrating
H~
the measured
susceptibility
withrespect
to theapplied field,
M=
o x[~m( T~, H~) dH~,
and the internal field was then determined from the relationlfi
=
H~
DM. Arough extrapola-
tion of these three curves to
@
= 0
(dashed
fines inFig. 2)
indicated that thesystem
possesses three distinct «characteristic» temperatures,suggesting
thepossibility
of threesequential
« transitions ».
,
T~(K)
,
so
Hj(Oe)
Fig.
2. Internal fielddependence
of the temperatures of the threepeaks
in X" The dashed lines arean
extrapolation
to zero intemal field.Figure
3 shows theimaginary
componentx"
of thecomplex dynamic susceptibility
of the Ni-23.6 at.9bMnsample plotted
as a function oftemperature,
for the same set of staticbiasing
fields as infigure
I. Acomparison
of these twofigures
shows that the two lower temperaturepeaks
inx'
areaccompanied by
acomplementary dual-peaked
structure in theabsorption (or loss) signal X",
with similar fielddependent
and timedependent systematics,
~i (
~
~
Ha(Oe)
~
12.5 15.0 20.O 25.O 35.O
50 ioo
TlKl
Fig.
3.Temperature dependence
of theimaginary
part X" of thedynamic susceptibility
in several representative staticbiasing
fields H~ between 100e and 350e. With theexception
of the 35 Oe isofield, successive isofields have been separatedvertically by
+ 0.2 a-u- for purposes ofclarity.
while the
highest temperature peak
inx', tentatively
attributed toferromagnetic
criticalfluctuations,
has noapparent
counterpart inx".
While several features of the
magnetic
response infigures
Ithrough
3 are consistent with apicture
of reentrantferromagnetism,
the extremeproximity
of thesystem
to the multicriticalpoint,
and the attendant breadth of theexchange
bonddistribution,
areexpected
toyield
anomalous
ferromagnetic
criticalproperties.
In the absence of arigorous
criterion which correlates theferromagnetic
critical temperature T~ with aparticular
structural feature of thezero field
susceptibility,
an indirecttechnique
wasemployed
toidentify
T~.According
to staticscaling theory,
aferromagnet obeys
anequation
of state of the formII Ii m(h, t)
=
t~ F(h/tY+fl) (I)
where m is the
magnetization,
t=((T-T~)/((
is the reducedtemperature,
andh
= gpB
HJkT
is the uniform intemal field. Differentiation withrespect
to hyields
thecorresponding equation
of state for the differentialsusceptibility x(h, t)
m
amlah
: x(h,
i)
= h (1/8) G
(h/iY
+fl)
,
(2)
whichpredicts [12]
the existence of a maximum in the temperaturedependence
of thesusceptibility
at t =t~((aXlat)~,
j~ =
0),
which issuppressed
inamplitude
and shiftedupward
in
temperature
withincreasing
field h :X
(h, t~)
cc h ~~M~(3a)
t
(h
~z h I/(y + p ~~~~P
1814 JOURNAL DE PHYSIQUE I bt 12
t~(h)
is known as the crossoverline,
and(y +fl)
is the crossoverexponent.
Thesesystematics
arequalitatively
identical to those of thehighest temperature component
of thetriple-peaked
structure infigure16,
andequation (3b)
shows that aplot
of thispeak
temperature T~ as a function of H;~'~Y+ fl>extrapolates
to T~ -l~
asH;
- 0. Such a
plot
wasused to obtain a
preliminary
estimate for T~,assuming
a 3-DHeisenberg
value of(y+fl)~~
=
0.57,
and this estimate wassubsequently
refined until a linear doublelogarithmic plot
oft~ verses
q
was achieved. Thisanalysis yielded
a critical temperatureT~ =
(104
±2)
Kjust
below theprincipal
maximum in the zero fieldsusceptibility),
and an inverse crossover exponent(y
+fl )~
=
0.47 ±
0.07,
which is lower than the 3-DHeisenberg prediction.
This latter behaviour is consistent with numericalIsing
model simulations[13],
which show
that,
as thesystem approaches
the tricriticalpoint
and the ratioJo/Jof
the mean value to the width of theexchange
bond distributionapproaches unity (the
limit ofstability
of theferromagnetic ground state),
the size of the criticalregion
in the h tplane
becomesvanishingly
small andessentially
inaccessibleexperimentally,
sothat,
even inrelatively
smalllaboratory fields,
the corrections toscaling
aresignificant
and reduce the effective exponent(y
+fl )~
below its trueasymptotic
critical value.The behaviour of the critical exponent
8,
which govems the internal fielddependence
of the criticalpeak amplitudes,
also reflects theproximity
of the multicriticalpoint. Figure
4a showsa double
logarithmic plot
of the criticalpeak heights x'(H;, T~)
as a function of internal fieldH;.
Theplot
reveals two distinctregimes corresponding
to different values of 8: anasymptotic
low field
regime (l~<150e)
with8~~
=
5.6±0.3,
reminiscent of the 3-DHeisenberg prediction
of 8= 4.8
[14],
and ahigh
fieldregime ( l~~~
> 15 Oe
)
with8~~
=
2.9 ± 0.
I,
which isessentially
the mean fieldprediction
8= 3
ill].
The existence of tworegimes
for theexponent
8 is a feature of other bond disorderedferromagnets
like Pdmn[15],
and thesystematics
of the variation of 8 with intemal field are also consistent with the modelsimulations
II 3]
referred toearlier,
whichpredict
a reduction in8,
for systems with extreme bonddisorder,
as the internal field increases and corrections toscaling
becomeprogressively
more
appreciable.
The
temperature dependence
of the zero fieldsusceptibility
above T~ is describedby
the exponent y,according
tox'(0,
t> 0
)
~ t~ Y, and
figure
4b shows a doublelogarithmic plot
ofX'(0,
t as a function of reducedtemperature
t with T~ = 104 K. The twostraight
lines in thisfigure represent
the mean field Curie-Weissprediction,
y=
I,
for purposes ofcomparison.
While several features of this
plot, particularly
the inflectionpoint,
which translates into a distinct maximum in the temperaturedependence
of the effective Kouvel-Fisher exponent~
bin x' (0,
t)
~' ~
i
'(~)
near t =
0.5,
and theasymptotic approach
to the mean field limit for T» T~, aretypical
of bond disorderedferromagnets,
and aresuccessfully replicated by
correlated molecular field models[16],
the curvature isunusually pronounced
in the presentsystem, leading
to values of the effectiveexponent
ashigh
asy*
~10 in theneighbourhood
of the inflectionpoint.
Anomalously high
values fory*,
which exceed both the 3-DHeisenberg
and mean fieldpredictions,
have also beenreported
in the Pdmn[15]
and FeZrII?]
systems, and appear to be a characteristic of extensiveexchange
bond disorder. Such effects areexpected
to beparticularly
dramatic in the present system, which liesextremely
close to the multicriticalpoint. Nevertheless,
aninspection
of the insert infigure 4b,
which shows the behaviour of the effectiveexponent
over a restricted reducedtemperature
interval(t«0.4),
reveals thaty*
does assume its 3-DHeisenberg
valuey*
=1.38[14]
at a reducedtemperature
t =
0.27
(dashed
lines in theinsert),
close to the inflectionpoint
in the zero fieldsusceptibility
H,~,<Oe)
3 S 1d
1/
(a)
$~~=56+0.3
tL
~~
~'
~ lb) 1
-
f
~
i 'f
2 '~
)
01
t
1816 JOURNAL DE
PHYSIQUE
I bt 12lines
[1, 20]
which characterize vectorspin
models of bond disorderedferromagnets,
the correlation is difficult tojustify rigorously
since these models arecurrently incapable
ofreplicating
themultiple
structure observed in thesusceptibility,
andconsequently
the presentinvestigation
focused on thermoremanent relaxation effects as a mechanism forexposing
fundamental differences between the two
phases.
As mentioned
earlier,
themagnetic
response in the reentrantphase
contains asignificant time-dependent component
which is observable as a slow relaxation of thedynamic susceptibility
inducedby
incrementaltemperature changes
in fixed field. The presence of this component also means that some elements of the structure associated with thedynamic
response
(Fig. lb)
will beparticularly
sensitive to thecooling conditions,
and this is illustrated infigure 5,
which shows the differentialsusceptibility
at 16 Hz measured uponwarming
in astatic
biasing
fieldH~= 250e,
aftercooling
in various static fieldsH~
between 0wH~
w 25 Oe. Fortemperatures
belowapproximately
55 K(which
is also theregion
where relaxation effects are mostnoticeable),
the presence of thecooling
field inhibits thedynamic
response,
systematically suppressing
andultimately quenching
the lowest temperaturepeak.
The onset of strong
irreversibility
within the reentrantphase
is also visible in the temperaturedependence
of the field cooled staticmagnetization,
andfigure
6 shows atypical
set of FCdata,
cooled and measured in a field of 0.5 Oe(solid circles)
for purposes ofcomparison,
thesusceptibility
isofield measured in a staticbiasing
fieldH~
= 25 Oe has also been
reproduced
in this
figure (triangles).
The vertical arrows mark the three characteristictemperatures
whichidentify
theextrapolated
location of the threesusceptibility peaks
in the limit of zero internal field(the
dashed lines indicate thecorrespondence)
and the crossover from weak to strongirreversibility
isreadily
apparent inM~c
as achange
in curvature near 60K,
which coincides with the lowest characteristictemperature.
H~
= 25Oe
~©
i
m
E
© cv
~
,ZO
X
1.8
4.6
11.O
25.O
ioo T(Kl
Fig.
H~
. M~~
.
x~
i j
,
J -
' I ,
' , Q1
' , o
i '
'~ I
f i '
I Cn
§
i ~~
/ i
$ ,
'p~
~
ioo
TIK)
Fig. 6. Field cooled static magnetization
(.)
cooled and measured in a field H~ = 0.5 Oe, and the250eX'
isofield fromfigure I(b).
The vertical arrows mark the three zero field characteristic temperatures, and the dashed fines indicate thecorrespondence
with eachsusceptibility peak.
Figure
7a shows a sequence of twelve low field thermoremanent relaxation isotherms obtained over atemperature
interval 40 Km T « 95 K which spans both the reentrant andferromagnetic phases.
An identicalexperimental technique
wasemployed
to measure eachcurve : the
sample
was first warmed to a referencetemperature T~
=
170 K well within the
paramagnetic phase,
where relaxation effects werenegligible,
then cooled in a field of1.0 Oe to themeasuring temperature T~,
at an averagecooling
rate of 7K/min (corresponding
totypical
totalcooling
times of t~m 15
min) and,
afterwaiting
for a time t~ = I m at constant temperature, the field wasabruptly
removed and thedecay
was recorded over the time interval Is<t«10~s.
Thesample
was then warmed in zero field to the referencetemperature T~
in order to establish the zero ofmagnetization.
The most obvious feature of the relaxation curves infigure
7a is thegradual
andsystematic change
in curvature fromconcave down at low
temperatures
to concave up athigh temperatures,
with the data atintermediate
temperatures exhibiting
an inflectionpoint. Although
the latter structure is rather subtle when viewed from theperspective
offigure 7a,
it isreadily apparent
in the derivative of the relaxation curves, andfigure
7b shows thenumerically
calculated localslope S(t)
mam~lain
t of some of the isotherms infigure 7a, plotted
on alogarithmic
time scale ; in thisrepresentation,
the inflectionpoint
translates into a maximum in the relaxation rate,and a
comparison
of the various isotherms suggests that the maximumpropagates
withincreasing
temperature fromlong
to short times like the crest of a wave,passing rapidly through
theexperimental
time window attemperatures
near T= 55 K.
Since much of this behaviour is at least
qualitatively
reminiscent ofspin glass relaxation,
theanalysis
focusedinitially
on those functional forms which characterize the thermoremanentdecay
intypical spin glass systems. (Although
some of thesefunctions, particularly
the1818 JOURNAL DE
PHYSIQUE
I bt 12(a) 16)
..:..._ °°°°°°°.. 40 K
45 K
So K $
c
~i
~ '
2
c
~
7J
~
li
$
$ 0l
~
,.~ 65 K
'....
~~ 70K
... 75K
40 K
~
85 K
~»ww...~~ °°. 90 K
~~*~'°"°....
.. 95 K
2 3 4 2 3 4
log
tlog
tFig.
7.-la) Decay
of the thermoremanentmagnetization
MR as a function oflog
t for several temperatures between 40 K and 95 K. ~b) The relaxation rate S for each of the isotherms in partla).
Note the maximum in the 55 K isotherm.
stretched
exponential,
areregarded
asonly approximate representations
ofnonequilibrium phenomena [21],
their presence is nevertheless a valuable indicator of the processes which dominate therelaxation,
and hence assist inidentifying
the nature of the orderedspin configuration.)
Two distinct relaxationregimes
were identified:(a)
Fortemperatures
T « 60 K(I.e.
within the reentrantphase),
the data areaccurately
described over the entireexperimental
time windowby
thesuperposition
of a stretchedexperimental
and a constant term,M~(t)
=
Mo
+Mi exp[- (t/T)~ ~"]
this is illustrated infigure 8a,
which shows that doublelogarithmic plots
ofd/dt[in (M~(t) Mo)]
as a function of time tyield
excellentstraight
linesthroughout
the reentrantregime,
and table I summarizes the values of the variousparameters
extracted from thestraight
line fits infigure
8a. The timeindependent component
of the remanence(Mo)
isuncharacteristically large
incomparison
with purespin glasses, accounting
forapproximately
90-95 9b of themagnitude
of the totalsignal,
and decreasesmonotonically
withincreasing
temperature, thusresembling
the behaviour of aferromagnetic spontaneous magnetization by
contrast, thetemperature dependence
andmagnitude
of the exponent n and the characteristic time Tare bothtypical
of purespin glasses and,
inparticular,
the variation of T(which
locates the inflectionpoint
in the stretchedexponential,
and hence the maximum in the relaxationrate)
with temperature is consistent with theexperimental systematics
described in theprevious paragraph, viz,
apicture
in whicha fixed observation time window
samples
differentportions
of a stretchedexponential
curveas the characteristic time T passes
through
the window.(b)
Fortemperatures
65 KS T « 95 K(I.e.
within theferromagnetic phase),
the stretchedexponential description fails,
and the functional form of the relaxation
changes
to thesuperposition
of asimple
power law and a constant term,M~(t)
=
Mo
+ «o t'"'; as shown infigure 8b,
doublelogarithmic plots
la) 16)
Z -2
7J
7J
~
2i
~ ?
~ i_~
~ ~
~ if
+ 1
o
~
°°
4
.
~
2
og t log t
Fig.
8.(a)
Doublelogarithmic
plot ofd[fn (M~ Mo)I/dt
as a function of i for the reentrant isotherms (40 Km T« 60K).(b)
Double logarithmic plot of -dMR/dt as a function of t for the ferromagnetic isotherms(65
K « T « 95 K).Table I. Parameters
from
stretchedexponential fits
infigure
8a.T (K) 40 45 50 55 60
A( (a. u.) 552 ± 2 539 ± 509 ± 457 ± 2 352 ± 2
n 0.64 ± 0.02 0.65 ± 0.02 0.66 ± 0.01 O-gI ± a-al 0.92 ± 0.01
r (s) (8.02 ± 2.I I) x 10~ (5.00 ± 1.44) x 103 (2.22 ± 0.30) x 10~ (4.77 ± 0.40) x 102 2.4 ± 1.0
Table II. Parameters
from
power lawfits
infigure
8b.T
(K)
65 70 75 80 85 90 95Mo (a.u.) 330 ± 50 270 ± 30 215 ± 45 130 ± 60 168 ± 5 138 ± 6 123 ± 3
m 0.05 ± 0.02 0.06 ± 0.02 0.05 ± 0.02 0.05 ± 0.02 0.14 ± 0.02 0.10 ± 0.02 0.12 ± 0.02
of
dM~/dt
versus tyield
excellentstraight
line behaviour over three decades of observationtime,
with aweakly
temperaturedependent
exponent m(see
Table II for a list of the best fitvalues of the parameters
Mo
andm).
Both the functional form of thedecay
and themagnitude
of theexponent
aresymptomatic
of relaxation at thermalequilibrium [21, 22],
and indicateJOURNAL DE PHYSIQUE T I, M 12, D#CEMBRE lwi 71
1820 JOURNAL DE
PHYSIQUE
I N 12that
aging
processes, which dominate thedecay
in the reentrantphase
and areresponsible
for the stretchedexponential
timedependence,
areessentially
absent within theferromagnetic phase.
This conclusion is consistent with the behaviour of thedynamic susceptibility
infigure 5,
which exhibitsvirtually
nosensitivity
to thecooling
conditions for temperatures T > 65 K(field cooling
isexpected
toproduce
a state which isconsiderably
closer to thermalequilibrium
than thatgenerated by
zero fieldcooling).
It is instructive to compare the viscous
properties
of the reentrant Nimnferromagnet
described above, with a recent
study by
Chamberlin and Haines[23]
of thermoremanent relaxation in theAui
_~Fe_~ system, in which a model for activated magnon relaxation on apercolation
distribution of finite domains is used to derive twomesoscopically
exactrelaxation functions which are
capable
ofreproducing
many of theexperimental
features overnine decades of observation time
(-
5<
log
t«4). Although
the Aufestudy
did notexplicitly
include reentrantferromagnets,
a number ofintriguing
similarities nevertheless emerge from thiscomparison (a)
the existence of two distinct relaxationregimes
is also acharacteristic of Aufe
alloys
within the clusterglass region
of themagnetic phase diagram (x
m0.12)
for temperatures above the maximum in the ZFCmagnetization (which
defines the transition temperatureTc~),
the relaxation curves are concave up andrepresented by
arelaxation function which reduces
[23]
to asimple
power lawM~(t)
t~ " in the limit oflong
observation times,
while,
belowTc~,
the curvature of the data isinverted, necessitating
the introduction of a further relaxation function which reduces[23]
to the Kohlrausch-Williams- Watts stretchedexponential M~(t) exp(- tfl)
in a similar limit. Thus thevalidity
of the« power law » and stretched
exponential
»descriptions
of the reentrant NiA4n data is aconsequence of the
comparatively
more «restricted»experimental
time window(ls<
t < 10~
s) employed
in the currentinvestigation,
while the absence of a well-defined inflectionpoint (which
is known to be anonequilibrium
manifestation ofaging)
below TCG in the Aufestudy,
is related to therelatively long waiting
timesemployed
in thisstudy
(t~~10~s),
whichyield
a response closer to true thermalequilibrium. (b)
Within theferromagnetic phase
of the Aufe system, the relaxation exhibited severalcomplex
features whichrequired
notonly
thesuperposition
of the two relaxation functions mentionedabove,
but also the introduction of a constant baseline,
in order to achieve asatisfactory description
of these more concentrated
samples.
This timeindependent
component of the remanence is also an essentialingredient
of the currentanalysis
of reentrantferromagnetism
in Nimn(see
Tables I and II for values of the parameterMo), particularly
in the reentrantphase,
and supports the
hypothesis
that the reentrantphase
is ahybrid
of two orderedspin configurations
:longitudinal ferromagnetism,
which generates thelarge
static remanence, andspin glass freezing,
which isresponsible
for the stretchedexponential
thermoremanentdecay
with itsage-dependent
characteristics.In summary, measurements of the low field
dynamic
response and thermoremanentrelaxation on a reentrant Nimn
ferromagnet
located very close to the multicriticalpoint
reveal a system with
predominantly ferromagnetic
correlations and associated criticalexponents
indicative of extensiveexchange
bonddisorder,
but with two distinct relaxationregimes
: ahigh
»temperature regime
where thermalequilibrium
is establishedrapidly
and thedecay
is describedby
a weak powerlaw,
and a lowtemperature regime
where thedynamic
response is sensitive to thecooling
field and where thedecay
exhibitsnonequilibrium
behaviour which can be modelled
by
a stretchedexponential superposed
on a staticremanence, and is thus consistent with the
predicted
coexistence offerromagnetic
andspin
glass ordering.
Acknowledgments.
This work was
supported
inpart by
agrant
from the Natural Sciences andEngineering
Research Council of Canada.
References
[1] GABAY M. and TOULOUSE G.,
Phys.
Rev. Lett. 47 (1981) 201.[2] KUNKEL H. and WILLIAMS G., J. Magn. Magn. Mat. 75
(1988)
98.[3] ZASTRE E., RosHKo R. M. and WILLIAMS G.,
Phys.
Rev. B 32(1985)
7597.[4] KORNIK K. J., RosHKo R. M. and WILLIAMS G., J. Magn. Magt~. Mat. 81
(1989)
323.[5] ABDUL-RAzzAQ W. and KOUVEL J. S., J.
Appl. Phys.
55(1984)
1623.[6] ABDUL-RAzzAQ W. and KOUVEL J. S.,
Phys.
Rev. B 35(1987)
1764.[7] KOUVEL J. S., ABDUL-RAzzAQ W. and ZIQ Kh.,
Phys.
Rev. B 35 (1987) 1768.[8] KUNKEL H., RosHKo R. M., RUAN W. and WILLIAMS G., Phil. Mag. 864 (1991) 153.
[9] KUNKEL H., ROSHKO R. M., RUAN W. and WILLIAMS G., Phil. Mag. 863 (1991) 1213.
[10] GOLDFARB R. B., RAO K. V. and CHEN H. S., Solid State Comm~tn. 54
(1985)
799.[ll]
STANLEY H. E., Introduction to Phase Transition and Critical Phenomena, 1971, Oxford, Clarendon.[12] Ho S. C., MAARTENSE I. and WILLIAMS G., J.
Phys.
F: MetalPhys.
lI (1981) 699.[13] KORNIK K., KUNKEL H. P., RosHKo R. M. and WILLIAMS G., Solid State Comm~tn. 76 (1990) 993.
[14] LE GUILLOU L. C. and ZINN-JUSTIN J.,
Phys.
Rev. B 21 (1980) 3976.[15] Ho S. C., MAARTENSE I. and WILLIAMS G., J. Phys. F: Metal
Phys.
11 (1981) l107.[16] FXHNLE M., J. Magn. Magn. Mat. 65
(1987)
1.[17] REISSER R., FAHNLE M. and KRONMULLER H., J. Magn. Magn. Mat. 75
(1988)
45.[18] AvIRovIc M., ZIEMANN P., HASK B. and HESSE J.,
E~trophys.
Lett. 8(1989)
281.[19] KUNKEL H., RosHKo R. M., RUAN W. and WILLIAMS G., J.
Appi. Phys.
69(1991)
5060.[20] DE ALMEIDA J. R. L. and THOULESS D. J., J. Phys. A Math. Gen. 11
(1978)
983.[21] NORDBLAD P., LUNDGREN L., SVEDLINDH P., SANDLUND L. and GRANBERG P., Phys. Rev. B 35 (1987) 7181.
[22] LUNDGREN L., NORDBLAD P. and SVEDLINDH P., Phys. Rev. 834 (1986) 8164.
[23] CHAMBERLIN R. V. and HAINES D. N.,