HAL Id: jpa-00212460
https://hal.archives-ouvertes.fr/jpa-00212460
Submitted on 1 Jan 1990
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.
C.M.B. van der Zon, G.D.F. van Velzen, W. Th. Wenckebach
To cite this version:
1479
Nuclear
magnetic ordering
in
Ca(OH)2.
III.
Experimental
determination of the
critical
temperature
C. M. B. van der
Zon,
G. D. F. van Velzen and W. Th. WenckebachKamerlingh
OnnesLaboratory,
P.O. Box 9504, 2300 RALeiden,
The Netherlands(Reçu
le 1 er décembre 1989,accepté
le 21 mars1990)
Résumé. 2014 On
présente
une étudeexpérimentale
de l’ordremagnétique
desspins
des protonsdans
(OH)2Ca.
L’état ordonné estproduit
parpolarisation dynamique
nucléaire suivie par désaimantationadiabatique
dans le référentiel tournant. Latempérature
despin
estnégative,
lorsque
lechamp
magnétique
extérieur estparallèle
à l’axe c du cristal. On a déterminé l’étatmagnétique
en utilisant les résultatsexpérimentaux
sur l’aimantationparallèle
auchamp
effectif. On a mesurél’énergie
etl’entropie
dusystème
despins
et déterminé latempérature
despin
àpartir
de ces résultats. On trouve que la transition dephase
seproduit
à - 0.9 03BCK, en bon accord avec laprédiction théorique
del’approximation
deschamps
moléculaires.Abstract. 2014 An
experimental study
of themagnetic ordering
of the protonspin
system inCa(OH)2
ispresented.
The orderedphase
is reached viadynamic
nuclearpolarization
followedby
an adiabaticdemagnetization
in therotating
frame. Thespin
temperature isnegative
while the externalmagnetic
field isparallel
to thecrystalline
c-axis. Measurements of themagnetization
parallel
to the effectivemagnetic
field are used to determine themagnetic
phase
of the nuclearspin
system. From a measurement of the energy of the nuclearspin
system as a function of theentropy, the nuclear
spin
temperature is derived. Thephase
transition is found to occur at- 0.9
03BCK,
which is ingood
agreement with the theoreticalprediction
of the molecular fieldapproximation.
J.
Phys.
France 51(1990)
1479-1488 1 er JUILLET 1990,Classification
Physics
Abstracts 75.30K - 76.601. Introduction.
Nuclear
spins interacting
viadipole-dipole
interactions show aphase
transition to amagnetically
ordered state at atemperature
ouf the order of 1...,K.
Thisextremely
lowtemperature
is reachedusing
amulti-step cooling
process introducedby Abragam
andco-workers
[1, 2]
and used laterby
Marks et al.[3]
to create aferromagnetic ordering
of theproton
spins
inCa(OH)2.
First,
thesample
is cooledby
cryogenic
methods to about 0.5 K andplaced
in astrong
magnetic
field.Then,
the nuclearspin
system
is cooled further to about 2 mKby
means ofDynamic
Nuclear Polarization(DNP) [4, 5]. Finally,
an AdiabaticDemagnetization
in theRotating
Frame(ADRF) [4] brings
thetemperature
of the nuclearspin
system
below the transitiontemperature.
This paper concems an
experimental study
of theordering
of theproton
spins
inCa(OH)2,
which has thecrystal
structure shown infigure
1. Thishexagonal
structure is of theCdI2-type,
with space
symmetry
groupD 3 -
P3m.
Thepositions
of the nuclei and the dimensions of theFig.
1.- (a) Hexagonal
unit cell ofCa(HO)2 showing
theposition
of the ions and the cell parameters.(b)
The structure of alayer
of protons shown in a 0,0, 1
cross section of the unit cell.Open spheres
2
are
0.355 Â
above and solidspheres
are 0.355 Â below0,
0, § ] .
unit cell were determined
by
neutron diffraction at lowtemperatures
[6].
Theprotons
arearranged
in almost twodimensional layers.
Thearrangement
of theprotons
in theselayers
is shown infigure
1 b. Theproton-proton
distance in thelayers
is 2.186Á,
while the distance between thelayers
is 4.880Á.
In order to describe the
proton
spin
system
ofCa(OH)2 during
and after theADRF,
we use a frame of reference with its z-axisparallel
to the external staticmagnetic
field. The ADRF isperformed using
asaturating
rf-field 2BI
cos w t, which is oriented in thexy-plane.
Then,
the nuclearspin
system
is in thermalequilibrium
if our frame of reference rotates about the z-axis with thefrequency w
of this rf-field. In thisrotating
frame,
its Hamiltonian isgiven by
In this
equation, d
=yBo - w
is the so-called effective field in therotating
frame ofreference,
whileBo
is theamplitude
of the staticmagnetic
field and y is the nucleargyromagnetic
ratio.Furthermore,
K§
is the truncateddipolar
Hamiltonian,
i.e. thepart
of thedipolar
interaction that commutes with the Zeeman interaction :where
1481
vector
connecting
thesespins
and the externalmagnetic
field.Furthermore,
I)
and1:
are thecomponents
of the i-thspin. Finally,
cc o = 4 r x 10 -
and h is Planck’s constant.At the start of the
ADRF,
we choose thefrequency w
of the rf-field such that 4 is manytimes the NMR line width
4NMR.
Then,
w isswept
until 4 s4NMR
and the rf-field is switchedoff. In the
special
case that we choose 4 =0,
we call the ADRFcomplete.
Otherwise,
it iscalled
incomplete.
Directly
after thisADRF,
the nuclearspin
system
is still in internal thermalequilibrium
in therotating
frame of reference. So it is describedby
adensity
matrix :where T =
h/kf3
is thetemperature
of the nuclearspin
system
after theADRF,
while k isBoltzman’s constant.
In our
experiments
4 « 0 at thebeginning
of theADRF,
so this finaltemperature
is alsonegative.
Furthermore,
theexternally applied
staticmagnetic
field is orientedparallel
to thecrystalline
c-axis.According
to theexperiments
ofSprenkels et
al.[7],
theproton
spins
then order as alongitudinal ferromagnet
with a domain structure. There are twotypes
ofdomains,
in one thepolarization
of theproton
spins
isparallel
to theexternally applied
staticmagnetic
field,
in the otheranti-parallel.
I.e. in onetype
it isparallel
and in the otheranti-parallel
tothe
crystalline
c-axis.Furthermore,
the domains have a flatpancake-like shape,
with theplanes
of the domainsperpendicular
to thecrystalline
c-axis,
so each domain consists of alarge
number ofadjacent layers
ofproton
spins
of thetype
shown infigure
lb.We studied this
ferromagnetic ordering
moreelaborately by measuring
theproperties
ofthe ordered
spin
system
as a function of itsentropy.
Wepresent
measurementsyielding
thetransverse
susceptibility
y, and of theparallel magnetization p,
=(Iz)
II
as a function of thelongitudinal
effective field A The results areinterpreted using
the restricted traceapproximation
[8]
as elaborated for the case of themagnetic ordering
of theproton
spins
inCa(OH)2 by
van Velzen et al.[9]. Thus,
we determine themagnetic phase
of the nuclearspin
system
as a function of itsentropy.
Subsequently,
we show measurements of the energy of the nuclearspin
system
as a function of itsentropy.
From theseresults,
the nuclearspin
temperature
is calculated.Then,
themagnetic phase being
known from theprevious experiment,
the criticaltemperature,
where the transition to theferromagnetic phase
occurs, is determined.2.
Experiments.
Our
experiments
areperformed
on acylindrically shaped single crystal
ofCa(OH)2,
with adiameter of 3.78 mm and a
length
of0.18 mm,
while the axis of thecylinder
is orientedparallel
to thecrystalline
c-axis. For theDNP,
thesample
is irradiated with a 1.5 MeV electron beamcreating 02 -centres,
at a concentration of 6 x10- 5
centres perproton,
[10].
The
sample
is cooled to 0.5 Kusing
a3He
evaporation
cryostat.
Asuperconducting
solenoid created amagnetic
field of 2.73 Talong
thecrystalline
c-axis. Theproton
spins
arepolarized
by
means of DNPusing
75 GHz microwave irradiation.Subsequently,
the externalmagnetic
field is raised to 5.63 T in order to reduce thespin-lattice
relaxation rate.Finally,
an ADRF isperformed
as described in theprevious
section and in such a way that the finaltemperature
isnegative.
2.1 THE ENTROPY. - All the
properties
of the nuclearspin
system
are measured for fourvalues of the nuclear
polarization
po
before theADRF,
which is determined from the areausing
the method of reference[11]
]
to determine thefactor g
for ourcylindrical sample.
The
entropy S
of theproton
spin
system
is almostcompletely
conservedduring
the ADRF.Furthermore,
thisentropy
is related to thepolarization
p °
before the ADRFby [4] :
Hence,
by performing
all measurements as a function of the initialpolarization
p °,
, weeffectively perform
them as a function of theentropy
of the nuclearspin
system.
We note that
during
theADRF,
theentropy
increasesby
about 9%,
as was checkedby
performing
ADRFs andsubsequent remagnetizations
in therotating
frame. In thefollowing
sections,
we will use valuesof p °
that are corrected for thisslight
increase ofentropy.
The four corrected valuesof p °
aregiven
in the first column of table I.Table I. - Results
of
the transversesusceptibility
measurements.2.2 THE TRANSVERSE SUSCEPTIBILITY. - We have also determined the transverse
suscepti-bility
of the
proton
spin
system
after acomplete
ADRF. It is defined in therotating
frame ofreference. In
equation (2.3),
Px is themagnetization
in thisrotating
frame due to a transversefield
B1 =
y -1
1 li)
1applied
in the x-direction of thisrotating
frame. We have obtainedX 1 from NMR
absorption signals
recordeddirectly
after acomplete
ADRF andcomputing
their
Kramers-Kronig
transforms[4] :
where the calibration
factor g
is the same as above inequation (2.1 ).
The results aregiven
inthe second column of table 1.
2.3 THE PARALLEL POLARIZATION. - After the
ADRF,
theproton
spin
system
is describedby
the Hamiltonian JCgiven by equation ( 1.1 )
and thedensity
matrixgiven by equation (1.4).
If the ADRF isincomplete,
the effectivelongitudinal
field A differs from zero.Then,
also the1483
We have measured these
polarizations
as a function of the effective fieldd,
by performing
incomplete
ADRFs andby measuring
the nuclearpolarizationpz
after these ADRFsusing
thesame method as described above for
p °.
The results for the four valuesof p °
given
in table 1 arepresented
as datapoint
infigures
2a,
b,
c and d. The curves are the result of theoreticalcalculations that will be treated in section 3.
Fig. 2. - pz plotted
as a function ofA/2 w,
for four different values of the initialpolarization
p °.
The curves are discussed in section 3.2.4 THE INFLUENCE OF THE DEMAGNETIZING FIELD. - The
crystal
structure ofCa(OH)2
is non-cubic and thesample
has anon-spherical shape.
Therefore thepolarized
proton
spins
create a
demagnetizing
field that should be added to the effective fields in therotating
frame.Thus,
theprotons
experience
a totallongitudinal
effective fieldFor our
cylindrical sample,
A 0
is calculated in reference[ 11 ] .
Its value isgiven
in the third column of table I.The influence of the
demagnetizing
field on Pz will be accounted for in the next section. Its influence on the observedperpendicular susceptibility
X 1 iseasily
calculated[12] :
where X 1, o is the
susceptibility
in asample
whereA ° - 0. Applying
this relation to theexperimental
values for X 1, we obtain X 1, o as a function of the initialpolarization
Po.
These results aregiven
in the last column of tableI,
and furthermoreplotted
infigure
3. The solid squaresrepresent
our results. Another result ofSprenkels [7]
is shown as a opensquare.
Finally,
small dotsrepresent
values for X 1, 0, obtainedby
Marks[3]
in acompletely
different way, i.e.
by
recording
theimaginary
component of thefrequency dependent
susceptibility during
the ADRF.2.5 THE DIPOLAR ENERGY. -
Finally,
we determined the firstmoments m
1 of the NMRabsorption signals directly
after thecomplete
ADRFs. From these first moments, thedipolar
energy T ! ri1 1
Fig.
3. - The transversesusceptibility
X 1.. o versus the initialpolarization
p 0.
(0)
results from this work.(e)
results of Marks.(D)
result ofSprenkels.
The curves are determinedby
the restricted trace1485
Fig.
4. - Thedipolar
energy perspin
Ed/2 ir N
versus the entropy perspin
SI Nk.
Theslopes
of the dotted lines 1 and 2 represent thedipolar
temperature at 1 and 2kHz/spin.
of the
proton
spin
system
was calculatedusing [4] :
where the calibration
factor g
is the same as above.As was
explained
above,
the initialpolarization
p °
is related to theentropy S
of theproton
spin
system
viaequation (2.2).
Henceusing equation (2.2),
one canplot
thedipolar
energyEd
as a function of theentropy
S. The result is shown infigure
4.As is well
known,
thethermodynamical
relationallows us to calculate the
temperature
T of the nuclearspin
system
after the ADRF from the resultspresented
infigure
4. Since the number ofexperimental points
issmall,
we restrictourselves to an estimate of the average
dipolar
temperature
for two values ofEd/2
’TT Nonly :
1 and 2
kHz/spin.
From theslope
of the dashed lines 1 and 2through
theexperimental points
infigure
4,
we find :3. Theoretical
interpretation.
applied
to the conditions of theexperiments given
above.In our
previous
paper, amajor
effort was concerned with the determination of themagnetic
phase diagram
of the nuclearspin
system
inCa(OH)2.
Threephases
areexpected
to occur, aparamagnetic phase,
where allspins
are oriented in the samedirection,
aferromagnetic phase
with domain structure as described in section
1,
and anantiferromagnetic phase,
where the nuclearspins
in successivecrystal planes
haveopposite polarizations.
We note that in our
experiments
all thesephases might
occur. This can be seen mostcléarly
in
figure
2,
showing
ourexperimental
results for theparallel polarization
p, as a function ofthe
longitudinal
effective field 4 and the initialpolarization
Po.
Using
the calculations of reference[9],
we have determined for each value of 4 andp °,
, which of the threephases
ismost favourable. For the most favourable
phase,
we havesubsequently
calculated theparallel
polarization
Pz as a function of d. The results areplotted
infigure
2by
solid curves. Pindicates the
paramagnetic phase,
F theferromagnetic phase
and AF theantiferromagnetic
phase.
It can be seen that for the casethat p 0 = 0. 19
only
theparamagnetic phase
isexpected
and no
ordering
should be observable.However,
forhigher
valuesof po
all
threephases
areexpected.
’-Then,
for small values ofd,
thetheory predicts
theantiferromagnetic
phase
with a smallparallel susceptibility
For
higher
values ofà,
itpredicts
theferromagnetic phase
with a muchhigher
value ofXII
Finally,
at stillhigher longitudinal
effectivefields,
thespin
system
ispredicted
to becomeparamagnetic
and p, reaches anearly
constant valueapproaching
po.
We note that the theoretical results are very
detailed,
while theexperimental
accuracy islimited. The accuracy
of pZ
ismainly
determinedby
ourprevious
determination of thecalibration
factor g
and is about 1 %[ 11 ].
However,
the accuracy of à is muchless,
because theexternally applied magnetic
field driftsslowly.
Therefore,
before eachADRF,
wedetermined the
frequency
yB°
of the rf-field for which 4 = 0. This determination isperformed by recording
the NMR line of thepolarized
proton
spins
andby fitting
yB°
in such a way that the ratio of the moments m°and ml
of this NMR linecorresponds
tothe value
given
in reference[11].
As can be seen from thatreference,
the accuracy of thisprocedure
is about 0.5 kHz.As a
result,
ourexperiments
are notprecise enough
to show thepossible
existence of thepredicted antiferromagnetic phase.
However,
in aprevious
paper[9],
weanalysed
an earlierexperiment
ofSprenkels [7]
which wasperformed
at ahigher
valueof p ° =
0.70. Thisanalysis
showed that the
antiferromagnetic phase
does not occur. As anexplanation,
weproposed
thatthe
spin
system,
which passesthrough
theferromagnetic phase during
theADRF,
remains frozen in thatphase
when 4 = 0.Hence,
we conclude that all datapoints
infigure
2a concern theparamagnetic phase,
whilethose in
figures
2c and 2d concern theferromagnetic phase.
The scatter in thedatapoints
infigure
2b is toolarge
to draw any conclusions. Fromfigures
2c and2d,
we also seethat,
for theferromagnetic phase,
theslope
of ourexperimental
curve for Pz versus 4corresponds
well with the theoreticalprediction
of the restricted traceapproximation.
Butfigure
2a showsthat,
for the
paramagnetic phase,
theexperimentally
foundslope
is less thanpredicted
theoretical-ly.
As theslope
of these curvescorresponds
to theparallel susceptibility
Yll, for the1487
restricted trace
approximation.
However,
theprediction
for theparamagnetic phase
seems tobe less
good.
3.2 THE TRANSVERSE SUSCEPTIBILITY.
- Using
the restricted traceapproximation,
we alsocalculated the transverse
susceptibility
of the nuclearspin
system
as a function of the initialpolarization
p °.
The results areplotted
infigure
3together
with theexperimental
results discussed above. The curve indicatedby
Frepresents
the result for theferromagnetic phase,
the curve indicatedby
P theparamagnetic phase.
Forp ° > 0. 3,
theexperimental
values ofX 1 are more or less
independent
of po
. which agrees with the theoreticalprediction
for theordered
phase.
At lower valuesof p 0
the
experimental
values diminishrapidly
aspredicted by
the theoretical curve for the
paramagnetic phase.
We notehowever,
that for these lower values ofp °,
the observed transversesusceptibility
seems to beconsiderably
lower thanpredicted by
the restricted traceapproximation.
This observation is inagreement
with earlierexperiments
onCaF2 [8],
where alsosystematically
lower transversesusceptibilities
wereobserved than
predicted by
the restricted traceapproximation.
3.3 THE TRANSITION TEMPERATURE. - As discussed in section
2.5,
thespin
temperature
isexperimentally
found to beequal
to - 1.1uK
when thedipolar
energy isequal
toEd/2
irN = 1kHz/spin,
while it is - 0.7uK
whenEd/2
irN = 2kHz/spin.
Furthermore,
ouranalysis
of our measurements of theparallel polarization
shows that the nuclearspin
system
isparamagnetic
in the first case, while it isferromagnetic
in the latter case. This conclusion is confirmedby
our measurements of the transversesusceptibility.
Hence,
the transitiontemperature
should lie between these two values :This value for the critical
temperature
agrees very well with the valuepredicted by
the molecular fieldapproximation
[7] :
4. Conclusions.
In this paper, we
presented
astudy
of theordering
of theproton
spins
inCa(OH)2
that occursat
negative spin
temperature.
Theexperimental
values of theparallel
nuclearpolarization
areinterpreted using
the restricted traceapproximation.
The results show that alongitudinal
ferromagnetic ordering
with domain structure is created. Theantiferromagnetic ordering,
which ispredicted
for very small values of the effectivelongitudinal
field could not be observed. In the present case, thisnegative
resultmight
very welloriginate
ininadequate
experimental
accuracy,contrary
to the case ofSprenkels
experiments
asanalyzed
in aprevious
paper[9].
From the measurements of the
dipolar
energy versus theentropy
we determined thedipolar
temperature
after the ADRF.Combining
these results with those for theparallel polarization,
we have
determined
theexperimental
value of the criticaltemperature.
Itcorresponds
very well to the valuepredicted by
the restricted traceapproximation.
Acknowledgments.
This work is part of the research program of the «
Stichting
Fundamenteel Onderzoek der Materie(FOM) »
and has been madepossible by
financialsupport
from the « NederlandseReferences
[1]
ABRAGAM A., C.R. Hebdo. Acad. Sci. Paris 251(1960)
225-227 ; ibid. 254(1962)
1267-1269.[2]
CHAPELLIER M., GOLDMAN M., VU HOANG CHAU, ABRAGAM A., C.R. Hebdo. Sci. Paris 268(1969)
1530-1533.[3]
MARKS J., WENCKEBACH W. Th., POULIS N. J.,Physica
B 96B(1979)
337-340.[4]
ABRAGAM A., GOLDMAN M., Nuclearmagnetism ;
Order and disorder(Clarendon
Press,Oxford)
1982.[5]
VAN DER ZON C. M. B., MARKS J., WENCKEBACH W. Th., POULIS N. J.,Physica
B 145B(1987)
153-164.
[6]
BUSING W. R., LEVI H. A., J. Chem.Phys.
26(1957)
563-568.[7]
SPRENKELS J. C. M., WENCKEBACH W. Th., POULIS N. J., J.Phys.
C. 15(1982)
L941-L947 ; ibid.16