HAL Id: jpa-00208849
https://hal.archives-ouvertes.fr/jpa-00208849
Submitted on 1 Jan 1978
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Nuclear antiferromagnetism in lithium hydride
Y. Roinel, V. Bouffard, P. Roubeau
To cite this version:
Y. Roinel, V. Bouffard, P. Roubeau. Nuclear antiferromagnetism in lithium hydride. Journal de
Physique, 1978, 39 (10), pp.1097-1103. �10.1051/jphys:0197800390100109700�. �jpa-00208849�
NUCLEAR ANTIFERROMAGNETISM IN LITHIUM HYDRIDE
Y. ROINEL, V. BOUFFARD and P. ROUBEAU CEN Saclay, B.P. N° 2, 91190 Gif sur Yvette, France
(Reçu le 22 mai 1978, accepté le 20 juin 1978)
Résumé.
2014L’antiferromagnétisme nucléaire a été créé dans l’hydrure de lithium (LiH) par la méthode désormais usuelle : 1) Polarisation dynamique nucléaire, dans un champ de 5,5 T ; 2) Désai-
mantation adiabatique dans le référentiel tournant. Un réfrigérateur à dilution d’un type spécial
a été construit, pour obéir aux contraintes particulières de l’expérience. L’état ordonné subsiste
plus d’une heure après la désaimantation. On a obtenu par RMN trois types d’évidences expé-
rimentales de la structure ordonnée : 1) Forme de la raie d’absorption dipolaire des spins nucléaires
de 7Li; 2) Comportement de la susceptibilité perpendiculaire de ces spins en fonction de l’énergie dipolaire; 3) Dédoublement de la raie d’absorption Zeeman des spins nucléaires de 6Li. A la lumière de ces résultats a été entreprise une étude de LiH par diffraction de neutrons, qui a porté ses premiers
résultats.
Abstract.
2014A nuclear antiferromagnetic structure has been produced in lithium hydride (LiH) by the usual method : 1) Dynamic nuclear polarization in high field (5.5 T); 2) Adiabatic demagne-
tization in the rotating frame. A dilution refrigerator of special design has been built for the particular
features of the experiment. The ordered state persists for more than one hour after the demagne-
tization. Three types of evidence have been obtained experimentally by NMR on the ordered struc- tures : 1) Shape of the dipolar absorption line of the 7Li nuclear spins; 2) Behaviour of the per-
pendicular susceptibility of these spins as a function of dipolar energy; 3) Splitting of the Zeeman absorption line of the 6Li nuclear spins. Following these results, a neutron diffraction study of LiH
has been undertaken and has already yielded some results.
Classification
Physics Abstracts
76.50
-75.50E - 75.25
-76.70E
1. Introduction.
-It has been pointed out a long
time ago [1] ] that an assembly of nuclear spins in a crystal with dipolar interactions should reach an
ordered state at a sufficiently low field and tempera-
ture. An order of magnitude of the critical field Hé
and temperature Te is given by the linewidth of the NMR line of those spins, which is precisely due to dipolar interactions. This yields typical values of :
Such low nuclear spin temperatures can be obtained in insulators through the well known two-step process
proposed by Abragam in 1962 [2]. The first step consists in a dynamic polarization of the nuclei in a high field H, and brings the nuclear temperature into the millikelvin range. The second step is an adiabatic
demagnetization which eliminates the magnetic field, and reduces the nuclear spinjjempcrature by a factor :
possibly below Te. The two main conditions the
samples must fulfil are, obviously, a long nuclear spin-lattice relaxation time, and the possibility of obtaining high nuclear polarizations by means of the
solid effect. These conditions, along with the necessity
of having a simple crystalline structure, were met in
CaF2 doped with Tm2+ impurities, and since the first successful experiment in 1969 [3] the ordered structures
of 19F nuclear spins have been studied extensively in CaF2 [4]. The need for studying other substances and
particularly LiH arose from the desire to ascertain the nuclear ordered structure by neutron diffraction.
It has long been recognized [5] that this was not
possible in CaF2 because of the very small spin- dependent amplitude of the neutron-fluorine scatter-
ing in contrast to the case of protons [6]. In addition,
it was interesting to study a structure slightly more complicated than that of CaF2. LiH contains prin- cipally two nuclear species, protons and ’Li, with spins respectively 1/2 and 3/2, having comparable magnetic moments, and forming two imbricated f.c.c.
lattices. The theoretical calculations due to Gold-
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:0197800390100109700
1098
man [4] predict that each species contributes to the ordered phase, and that both protons and lithium are present in each sublattice. Nuclear antiferromagne-
tism studies had been made previously on lithium
fluoride (LiF) whose structure is identical to that of LiH [7]. We wish to report here recent experiments
on nuclear antiferromagnetism in LiH as observed by NMR. During the writing of this article, a nuclear superstructure line has been observed by neutron
diffraction in LiH, and this preliminary result is reported elsewhere [14]. We restrict ourselves in this article to the description of the NMR observations made on antiferromagnetic LiH in August and Sep-
tember 1977, before the diffraction experiment.
2. Dynamic polarization.
-The principle of dyna-
mic nuclear polarization (DNP) consists in saturating
off-centre the ESR line of some paramagnetic impu-
rities present in the sample. The result obtained
depends very much on the nature and the concentra- tion of the paramagnetic impurities, as well as on
the experimental conditions such as : magnetic field, lattice temperature, etc. A recent and complete review
of DNP can be found in ref. [8]. We simply specify
here the features relative to our samples. The mecha-
nism of polarization in LiH as in LiF containing a
few 10-4 of F-centres, is well established and consists in a cooling of the nuclei by thermal mixing with the
electronic spin-spin interactions reservoir, itself cooled
by the effect of the microwave irradiation of the ESR line of the F-centres. However, some spurious effects,
not well understood up to now, limit the nuclear
polarization to values lower that what could be
expected from theory [8] and dependent upon the
sample preparation. Samples of pure LiH, kindly
furnished by Dr. Bédère (1), in the shape of thin
slices of dimensions 5 x 5 x 0.5 mm, were irradiated with 1.5 MeV to 3 MeV electrons in order to create the F-centres. Previous studies [9] have shown that the temperature of the sample during this irradiation
plays an important role, and a special temperature control apparatus has been built [10]; however, obtaining a good sample is still subject to uncertainties.
The best results have been obtained with sample n° 1
of ref. [9] which reached after 50 hours of polarization
a nuclear temperature of 3 mK in a field of 5.5 T;
this corresponds to a proton polarization of 95 % (80 % for ’Li). Another of the samples studied, labelled # 17, reaches smaller polarizations (80 % for protons) but in only 15 hours. The F-centre concen-
tration Ne and relaxation time Tle of this sample
have been measured by the methods described in ref. [9] and are respectively :
(1) Service de Métallurgie Physique, B III Bruyères-le-Châtel,
B.P. n° 61, 92120 Montrouge, France.
Most of the apparatus is the same as in ref. [10].
However, we introduced in July 1977 two important
modifications. The first is the use of a dilution refri- gerator (Fig. 1) of the type new geometry described
by Roubeau in ref. [11] ; the second is a lower part of the ’He dewar which is composed of two coaxial cylindrical Kel-F tails glued to the upper parts by
means of stycast. The Kel-F-metal bounding is found
to be leak-proof over a period of several months, and
to be quite unsensitive to thermal shocks and to vibrations. The inner tail has an internal diameter of 6 mm and contains only the sample and its Kel-F
holder, plus a small 0.7 x 1 mm teflon tube driving
the concentrated phase down to below the sample.
The RF coils and the microwave system are now rejected to the outside of the outer tail (of external
diameter 10.5 mm) in the ’He bath at 2.5 K. This
geometry, well adapted to the perspective of neutron diffraction, has also the advantage of minimizing the
losses due to microwave during the DNP, and to the radiofrequency during the adiabatic demagnetization.
FIG. 1.
-The lower part of the dilution refrigerator cryostat.
a) Insulating vacuum ; b) Still ; c) Insulating vacuum ; d ) 06 mm guide filled with the sample holder ; e) Spirally wound heat exchan- ger ; f ) Leak flow impedance; g) Coiled heat exchangers;
h) Pseudo-mixing chamber ; i) Metal-Kel-F stycast glued joints ;
j) Kel-F tails ; k) Concentrated phase tubing.
At the optimum polarizing microwave power, the temperature of the sample is about 200 mK whereas
in ref. [10] in the same conditions it was 500 mK and the ’ Li polarization could not exceed 35 to 40 %.
The lowering of the sample temperature seems thus
to be essential for these experiments, and a new refri- gerating system taking these features into account has been built for neutron diffraction and is presently being tested. Another particularity of this refrigerator
is the presence along its axis of a 6 mm diameter passage for introducing the samples at low temperature (the F-centres are not stable at room temperature).
It takes about one hour to remove a sample, introduce
a new one and cool it from 77 K to 100 mK.
No NMR measurements were made on the protons because of the presence of a very strong spurious signal due to protons probably located in the cryostat tails and in the sample holder (the Kel-F available to us does contain protons). The polarization of the
’Li nuclei is measured by comparing the area of the absorption signal of those spins with that obtained at thermal equilibrium at 4.2 K. The polarization of
the two other nuclear spin species (’ H and 6Li) is calculated, using the well-verified [15, 16] hypothesis
of equal spin temperature.
3. Adiabatic demagnetization.
-For the second step, we use the so-called Adiabatic Demagnetization
in the Rotating Frame (ADRF). This consists in a
familiar fast passage stopped exactly at the nuclear
resonance. More details about this technique can be
found in ref. [4]. The difference with the case of CaF2
lies here in the fact that both ’Li and 1 H must be
demagnetized simultaneously. The NMR frequencies
chosen in our experiment are respectively
and
The amplitudes of the two RF fields are of the
)rder of 50 mG, and the sweep rate dH/dt is 1 G/s.
B. well-known advantage of the ADRF as compared
:o conventional demagnetization is that the ordered
;tates produced can be observed by the usual techni- lues of NMR, since the high magnetic field is still présent. Furthermore the Hamiltonian responsible for
;he ordering in high field being the truncated dipolar Hamiltonian, which is orientation-dependent, it is possible to produce different kinds of ordering by -hoosing the orientation of the field with respect to
;he crystalline axes. Positive and negative temperatures
iave been investigated. With a given initial nuclear
)olarization, this can be done by simply starting the
ADRF from above or below the resonance. In both -ases, with H ç [100], the structure predicted by ,heory is antiferromagnetic and consists in (110) )lanes in which the spins are alternatively parallel and mtiparallel to H (Fig. 2). At T > 0 the spins of protons ind lithium nuclei are parallel in a given (110) plane,
FIG. 2.
-Stable structure expected at low temperature with Ho 1’[100] at negative and positive temperatures.
whereas they are antiparallel at T 0. A very impor-
tant feature for NMR is the presence of lithium in both sublattices, as will appear later.
4. Observations on the ordered state.
-The NMR
absorption signal of lithium nuclei is recorded with
a voltage Q meter, during linear sweeps of the magnetic
field. The methods of detection, using a multichannel
analyzer, and the general geometry of the experiment
have already been described elsewhere [10, 7]. After
the ADRF, the spin temperature is of the order of 1 gK, and then begins to increase slowly under the
effect of the dipolar spin-lattice relaxation, at a rate depending on the lattice temperature and varying
somewhat from one sample to another [9]. Typically,
the lattice temperature is 100 mk during this phase
of the experiment, and the dipolar relaxation time is several hours.
4. 1 SHAPE OF THE ’Li ABSORPTION LINE. - The NMR absorption signal of the ’Li nuclei is shown on
figure 3 at different times after the ADRF. The first
signals strongly suggest the existence of two species
of ’Li nuclei, with opposite polarizations, and seeing opposite local fields, in accordance with the Weiss field description of antiferromagnetism. As the time elapses, the disorder of the spins increases, and the
two components broaden and merge together. The shape of the line evolves smoothly towards the usual, paramagnetic, high temperature dipolar line (Fig. 3f),
and it is difficult to detect accurately by this method the antiferro-paramagnetic transition.
4.2 TRANSVERSE SUSCEPTIBILITY OF 7Li NUCLEI.
-Another evidence for the existence of an antiferroma-
gnetic state is the near constancy of the transverse
susceptibility xl of the spins for the lowest spin tem- peratures T. However, the temperature in this highly
non-linear domain is not related in a simple manner
to quantities measured by NMR. A better parameter is given instead by the dipolar energy of the spins,
which is an increasing function, of 1/ T (proportional
1100
FIG. 3. - Negative temperature. Sample n° 1. NMR absorption signal of ’Li nuclei versus magnetic field at various times t after
the ADRF (t = 0 : beginning of the ADRF). a) t = 2’ ; b) t = 10’;
c) t = 20’ ; d) t = 30’; e) t = 40’ ; f) t = 80’.
to 1/T at high temperature). As shown in ref. [4] and [12] the first moment of the NMR dipolar absorption
line is proportional to the dipolar energy, when
only one nuclear spin species is present. With two nuclear species, 1 H and ’Li, the first moment is pro-
portional to
where JC’LiLi is the part of the truncated dipolar Hamil-
tonian relative to interactions between’Li nuclei, and
£’LiH the part relative to interactions between ’Li and iH nuclei. The transverse susceptibility xi can be extracted from the dipolar absorption signal through
a Kramers-Krônig transform [4]. Both xl and the
first moment, (5) can be calibrated by comparison
with a Zeeman signal corresponding to a known polarization [12], so that it is possible to have absolute
values for these quantities, for comparison with theory.
The computation of the first moment and the Kramers-Krônig transform are performed on the
recorded dipolar signals by means of a HP 9820 elec-
tronic calculator. The experimental results are shown
on figure 4 (points). We have also plotted on the same figure the theoretical curves 1) in the paramagnetic
FiG. 4.
-Negative temperature. Sample no 1. Perpendicular susceptibility xi as a function of the first moment for ’Li nuclei, in the demagnetized state at T 0. X, is defined as in ref. [12] and
is expressed in G-1 , MLi1 is expressed in G.
high temperature limit neglecting the effect of the
impurities and 2) in the Weiss-field approximation, according to calculations given in ref. [7]. As a func-
tion of time, the experimental plot is to be read from the top right to the bottom left. The trend to a plateau
is noticeable at low temperature. However, this effect is not as pronounced as in CaF2 [4]. The discrepancy
of 30 % with the Weiss-field plateau, comparable to
that observed in CaF2, is far outside the experimental
error, and can perhaps be attributed to the crude-
ness of the theoretical model. A discrepancy also
exists in the high temperature domain : it can be shown that the reciprocal slope of the curve in the paramagnetic, high temperature limit is equal to the
second moment of the Zeeman absorption line. The
excess by a factor of two observed experimentally (Fig. 4) with respect to the theoretical second moment of the 7Li absorption line is presumably due to the
presence of the F-centres, whose dipolar fields add a
Lorentzian broadening in the wings of the NMR line.
The impurities could also be responsible for the dis-
crepancy observed at low temperature.
4. 3 SHAPE OF THE 6 Li ABSORPTION LINE. - The
most spectacular effect of the antiferromagnetic tran-
sition observed by NMR is the splitting of the 6Li absorption line. This isotope of lithium, with a spin 1
and a NMR frequency of 34.303 MHz in the field Ho
of the experiment, has in our samples an abundance
c = 3 % (lower than its natural value 7 %). It is used
here as a probe that should not perturb too much the
ordered state, and gives local details on the structure.
A similar experiment has already been performed in CaF2, using the 43Ca spin as a probe [4]. However, in antiferromagnetic CaF2 the Weiss-field on calcium sites is zero, and one could observe a splitting of the
line only in a different phase called ferrosandwich,
obtained when the demagnetization is performed with
H parallel to a [111] ] direction. In antiferromagnetic
LiH on the contrary, the Weiss field on lithium sites
HW is non-zero, and one can expect a splitting by
an amount 2 HwL’ of the 6Li Zeeman absorption line
into two components. The two components are not
expected to be identical in general because of the strong tendency to dipolar ordering. Let a be the inverse
Zeeman temperature of the 6Li nuclei, and fi their
inverse dipolar temperature, which is common to all the nuclear species of the sample, as has been shown
experimentally [13]. Under the effect of the Zeeman interactions alone, we should observe two lines of the
same intensity proportional to :
(where B1 is the Brillouin function for spins 1 and y6
the gyromagnetic ratio of 6Li nuclei) and whose
centres are respectively Ho + HwLi and Ho - HWLi .
On the contrary, under the effect of dipolar interac-
tions alone, the Weiss-field theory predicts that the
two lines will be of opposite intensities, namely :
and
The shape of the ’Li absorption line in figure 3a is compatible with a formula of this type.
A straightforward generalization of the Weiss for- mulae where both Zeeman and dipolar interactions
are taken into account yields for the two intensities :
and
In order to be able to see two distinct lines of com-
parable intensities, it is very useful to lower the Zeeman temperature so as to make the term aHo dominant
in (8). The technique, as in [17], consists in mixing in
the rotating frame the Zeeman reservoir of 6Li with the dipolar reservoir, by application of a strong RF field at a distance Li = - y6 Heff from the resonance frequency of the 6Li. In a frame rotating at a fre-
quency - yHo - d, the inverse Zeeman temperature of 6Li :
will then become equal to fl, and equations (8) can
be written :
and
In practice, the mixing in the rotating frame is expected to become exceedingly slow if the effective field Heff is too large. Experimentally in our conditions
the limiting value of Heff is 45 G. A rough estimate,
of the heat capacity Cf of the effective Zeeman reser-
voir of 6Li in the rotating frame and Cdip of the dipolar
reservoir yields :
where y6 and 77 are the gyromagnetic ratios. With c = 3 X 10-2, Heff = 45 Gand H!} = 10 G, one
has
and the heating of the dipolar reservoir induced by
this operation is very small. In order to minimize this heating, we have made the mixing reversible, by sweeping slowly the effective field from an initial value
H° f to 45 G (at positive temperature) or to - 45 G (at negative temperature). The initial value H ° f is
determined in such a way as to introduce no irreversi-
bility at the switching on of the RF power : this means
that the inverse Zeeman temperature a;n of 6Li in the effective field H0eff before the thermal mixing must be equal to the inverse temperature fl of the dipolar
reservoir : -
P is not known but can be guessed approximately,
and also deduced from the polarization of the 6Li
nuclei after the mixing as in [18]. ain is known from the initial ’Li polarization. In our conditions, this polarization, due to the DNP, is 30 % and H.Of f 1 - 10 G.
The results are shown on figure 5. Figure 5a shows
the signal of 6Li after the DNP and before the ADRF, and corresponds to a polarization of 30 %. In figure 5b,
we have saturated the 6Li resonance before the ADRF :
no signal is thus detectable at this time. Figure 5c
FIG. 5. - Negative temperature. Sample no 1. NMR absorption signal of 6Li nuclei versus magnetic field for different polarizations P.
a) At the end of the ADRF : P = 30 % ; b) After saturation of the 6Li : P = 0 ; c) After the ADRF, when P = 0 ; d) After
the ADRF, when P = 30 % ; e) After the ADRF + mixing with
the dipolar reservoir, when P = 65 % ; f ) Theoretical signal
expected after the ADRF for P = 100 %.
1102
shows the 6Li signal after the ADRF : the Zeeman
temperature is still infinite (a = 0) but a dipolar signal
appears, which is very similar to that of ’Li (Fig. 3).
In another experiment we do not saturate the ’Li at the end of the DNP, and we get, after the ADRF of ’Li and 1 H, the signal of figure 5d, which is a mixture of a Zeeman plus dipolar signals. It is not possible to distinguish two components. In figure 5e
the polarization of 6Li have been raised up to 65 % by mixing with the dipolar reservoir, and one can
now see two components, but the one on the right
is smaller than the one on the left. With a 6Li pola-
rization of 100 % one could expect two components of equal intensity, as sketched by the (theoretical) figure 5 f :
The experiments shown above correspond to a negative dipolar temperature. We can have an esti-
mate of its value, since a polarization of 65 % in a
field of - 45 G corresponds to a temperature of
-