1181

Superconductivity, localization and crystallographic phase transition inLa2 - XSrXCuO4 - y

W. Kang, G. Collin, M. Ribault, J. Friedel, D. Jérôme, J.M. Bassat+, J.P. Coutures+ andPh. Odier+

Laboratoire de Physique des Solides, Université Paris-Sud, 91405 Orsay, France

+ CRPHT-CNRS, 45071 Orléans Cedex 2, France

(Reçu le 11 mai 1987

Résumé.2014 Nous présentons une étude par résistivité de la supraconduction de la série La2 - xSrxCuO4 - y à l’aided’échantillons cractérisés par analyse thermogravimétrique en ce qui concerne la teneur en oxygène. Onobserve la supraconductivité d’une part pour des valeurs de x non nulles avec un maximum de

Tc au voisinage de la transition de phases tetragonal-orthorombique et d’autre part dans le composé non-substitué La2Cu O4 - y. Une localisation importante des électrons est observée lorsque x ~ 0,04. Ces résultatssont en accord avec un modèle BCS simple pour lequel le remplissage de bande est déterminé par le dopage austrontium. L’existence de localisation pourrait provenir de la compétition entre les effets des lacunes

d’oxygène et du dopage au strontium. Lorsque x ~ 0,04 la situation de bande demi-remplie permet lesdiffusions de type Umklapp. Nos résultats suggèrent une contribution des fluctuations antiferromagnétiques aumécanisme de "pairing" supraconducteur.

Abstract.- We report a resistive investigation of superconductivity in the system La2 - xSrxCuO4 - y on sampleswhich have been analysed by thermogravimetric techniques for oxygen deficiency determination. A maximumsuperconductive transition temperature is observed in La2 - xSrxCuO4 - y near the tetragonal-orthorhombicphase transition, together with a resurgence of high temperature superconductivity in non-substituted

La2CuO4 - y. Drastic localization effects are observed in the vicinity of x = 0.04 with no sign of superconductivi-ty at low temperature. These observations are coherent with a simple BCS scheme with a band fillingdominated by Sr-doping and localization effects dominated by an interplay between 0 vacancies and Sr-doping.This situation leads to half-filling of the band near x = 0.04 with enhanced electron localization via Umklappscattering through a near nested Fermi surface. Moreover, our data suggest some contribution of

antiferromagnetic fluctuations to the superconducting pairing mechanism.

J. Physique 48 (1987) 1181-1186 JUILLET 1987,

Classification

Physics Abstracts74.70 - 71.55J

The recent report of superconductivity at a tem-perature higher than 30K in a ternary lanthanum-alkaline earth copper oxide by Bednorz and Muller[1] has stimulated a very impressive amount of workand has raised several fundamental experimentaland theoretical problems which are a challenge forsolid-state chemists and physicists.The high Tc superconducting phases belong to a

class of materials which have heen around for quitesome time : the lanthanum copper oxides La2CU04[2]. The structure of La2CU04 studied in details byLongo and Raccah [3] shows it derives from the

tetragonal structure of the 2D magnetic perovskiteK2NiF4 but also exhibits a small orthorhombic distor-tion. Later, metallic conductivity was reported in thesolid solution Laz-xMxCu04 (M=alkaline earth) [4]and the group of Caen has studied in details the

oxygen intercalation in this class of materials andshown they usually are oxygen deficient, namely

La2-xMxCu04-y [5]. In the present work we dealwith materials which are either non-substituted or Srsubstituted. There exists a consensus in the recentliterature regarding the optimum amount of Srsubstitution. It is around x = 0.15 where both

T, is the highest (=35K) and the diamagnetic expul-sion is the largest (=50%) .However, the La2CU04 family exhibits a wealth of

interesting features which are worth mentioning asthey could be related to the occurrence of supercon-ductivity.

(i) A small orthorhombic distortion exists in

La2Cu04 at room temperature and vanishes at a

temperature of 533K [3] above which the tetragonalK2NiF4 structure becomes stable.

(ii) Substitution of Sr for La preserves the dis-torted orthorhombic structure of lower temperaturebut increases the average valence charge of the

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:019870048070118100

1182

copper ions from the charge Cu2+ of La2CU04, inorder to maintain charge neutrality.

(iii) Oxygen is known to deintercalate and reinter-calate easily in La2-xSrxCu04-y (y -- 0) even at

moderate temperature and pressure, with a strongconcomitant influence on transport properties [5].

Also, several experiments [6-9] on these materialshave recently proved the dominant role played bythe oxygen stoichiometry in the occurrence or disap-pearence of the superconductivity phenomena. Lowoxygen content, i.e. high oxygen vacancy concen-tration, which promotes a decrease in the net holeconcentration, favors carrier localization and poss-ibly suppresses superconductivity phenomena. It hasthus been considered as an important point to

evaluate quantitatively the net oxygen content ofseveral materials in which superconductivity proper-ties have been studied.

Hence, the main purpose of this work is a study ofthe superconductivity occurrence in the

Laz - xSr xCuO 4 - y series in relation with Sr sub-

stitution, content of oxygen vacancies and the prox-imity of the tetragonal to orthorhombic (T ---o. 0)distortion.Our results confirm that divalent alkaline-earth

substitution is not a prerequisite for the observationof superconductivity in the La2CU04 series [9].Moreover, a detailed investigation at various sub-stitution levels suggest the possibility of an interplaybetween the T --o. 0 phase transition and supercon-ductivity which is reminiscent of the interplay be-tween superconductivity and charge density waveformation observed in 2D layered dichalcogenides[10].

All oxides studied in this work were prepared inthe same way : mixing of freshly calcined lanthanumoxide, copper oxide and strontium carbonate withsubsequent firing in air, followed by several grindingand annealing procedures. Final treatment consistsin overnight annealing of pressed pellets at 1100°C inair and spontaneous cooling after switching off thefurnace. X-ray powder patterns performed on

Guinier-Laine camera (CuK ) indicate that thematerials are perfectly single-phased, no trace ofimpurity can be detected even on overexposedphotographs. The materials exhibit at room tempera-ture the orthorhombic distortion up to x = 0.09 andare strictly tetragonal for higher strontium concen-trations. It is the same source of materials which hasbeen used for an X-ray determination of theT 0 phase diagram [11].Small parallelepipedic samples of typical size

=0.5 x0.sx6mm were cut in the sintered pellets witha diamond saw and used for resistivity measurementswith four silver paint contacts. Either low fequencyAC or DC currents were used to measure resistivity.The oxygen content has been determined by the

weight loss of the sample under reducing conditions.

A symmetrical micro-balance from SETARAM

(1500) has been used between 20 and 1000°C. Thesintered samples were grinded in an agath mortarjust before their introduction in the thermobalance.A typical weight of 100 to 200 mg has been used.The sample is then out gassed for an hour undervacuum (10- 1 torr) at room temperature in order toremove part of the adsorbed species. Argon is thenintroduced at a flow rate of a few litres per hour. The

temperature is then raised up to 1000°C at a rate of300°C/h. An initial weight loss is observed between20 and 550°C, followed by a range of temperatures inwhich no weight reduction is observed. At 1000°Cthe weight remains constant (under Ar) at least formore than an hour. After that initial step,Ar - 5%H2 is introduced in the thermobalance atthe same flow rate as previously. The sample is thusreduced rapidly, in 30min, in its constituants

La203, SrO and Cu metal. It was checked that the

weight remains stable for more than one hour aftercomplete reduction.The resistivity data at various substitution levels

below 60K are presented in figure 1. All samples for0.08:!!!E;x--0.20 exhibit superconductivity. The criticaltemperature in figure 2a is defined by the tempera-ture corresponding to the mid-point of the resistivitytransition. The width AT of the superconductingtransition is characterized by the temperature incre-

Fig.l.- Temperature dependence of the resistivity versusthe Sr-substitution ratio. The insetshows the current-volt-

age characteristics of the x = 0.15 sample at 4.2K.

1183

ment between 10% and 90% of the resistive transi-tion.

Furthermore, we may notice from figure 1 a

drastic dependence of Tc and of the normal stateresistivity on x. The value x = 0.15 provides aboutthe lowest normal state resistivity which goes alongwith the optimum range for superconductivity. Thiswas also derived from the concentration dependenceof the Meissner effect [12] which shows the maximumfraction of ideal diamagnetism (about 50%) aroundx °--° 0.15 - 0.20. The inset of figure 1 shows thebehaviour of the resistance at 4.2K versus DCcurrent for x = 0.15. The low value of the criticalcurrent may be attributed to the granular characterof the material but the critical current is still much

larger at x = 0.15 than in the case of "pure"La2CU04 - y which will be discussed later in thisarticle.

Data of Tc and transition width versus x are

reported in figure 2a. A maximum of Tc is clearyobserved at x = 0.15. A minimum width at x = 0.15

Fig.2.- a) Critical temperatures and transition widths forvarious samples, 0 denotes the temperature of the anoma-ly for the x = 0.06 sample as explained in the text. b)Phase diagram of La2-xSrxCuO4-y’

is expected as Tc is passing through a maximum atthe same value of x. However, figure 2a unambigu-ously shows that the drop of Tc observed on eithersides of x = 0.15 is a genuine effect. It cannot be

attributed to some kind of smearing of the transition.The Tc = 6.6K(D) point corresponds to a small butsharp drop of the resistivity occurring at this tem-perature in a x = 0.06 sample. However, this sampledoes not show zero resistivity below Tc. This anomalyis probably related to superconductivity but it is byno means a bulk phenomenon at x = 0.06. At theconcentration x = 0.04 the temperature coefficientof the resistance is negative at low temperature withno sign of any superconducting anomaly down to4.2K (Fig. 3).

Figure 2b deals with the interplay between theT --0 0 phase transition and superconductivity. Thex = 0.10 and 0.12 data points have been taken froma recent X-ray study [11]. The dashed line is a guide

Fig.3.- Resistivity behaviour in x = 0.02 and x = 0.04

samples. Localization is evident at low temperature.

for the eye. Since the slope of the crystallographicphase transition line is likely to go to -oo as thetransition temperature approaches zero temperaturefor thermodynamical reasons we feel confident thatfigure 2b is a fair representation of the phasediagram in the vicinity of x = 0.15. Moreover, thediagram of figure 2b is in fair agreement withneutron diffraction studies performed on powdersamples [13] which imply that the T 2013> 0 transitionof the x = 0.15 sample is located above 60K. It alsoindicates that the vanishing of the orthorhombicdistortion represents the optimum situation for theonset of superconductivity.The amount of additional holes (electrons) and

relative weight losses determined from thermo-

gravimetric studies are given in table I. We notice

that the additional hole concentration is alwayssmaller than the amount inferred from the Sr

concentration but increases more or less monoton-

ously with x. In the case of Sr-substituted La2Cu04the oxygen vacancy concentration is in the range of afew percent. The hole concentration estimated bythe charge neutrality is thus in the range of 3 to9 x 102ocm-3 from x = 0.08 to x = 0.15. These

measurements are in general agreement with previ-ous results from Raveau et al. [14]. As far as pureLa2Cu04 is concerned, the sample slowly cooledfrom 1000°C in air gives a small but significantvacancy concentration of about 4% excess electrons

per formula unit. One has thus to account for oxygenvacancies which decrease the hole concentration.Two possibilities can be considered: Lanthanumvacancies coming from structural basis considera-

tions [15] and/or negative charge carrier excess.As shown in figure 2a, we confirm the existence of

a superconducting transition in unsubstituted

La2CU04 - y which is usually reported as exhibiting asemiconducting behaviour at low temperature

1184

[8,16]. In our sample, the onset of superconductivityin pure La2CU04 - y is quite different from what it isin Sr-substituted materials. The resistivity of

La2CU04 - y is fairly high at room temperature andincreases further by a factor about 50 down to thesuperconducting transition (Fig. 4). Given the sensi-tivity of our measuring system we can say that a zeroresistance is achieved below Tc as long as the currentpassing through the sample does not exceed about10RA (see insert of Fig.4). The critical current of4.2K is roughly 103 times lower than that of thex = 0.15 samples..

Fig.4.- Temperature dependence of the resistivity of

La2Cu04 - y under three different magnetic fields. The twoinsets show the current-voltage characteristics at 4.2K andthe critical field temperature dependence.

Furthermore, the shift of the transition under

magnetic field (dHc/dT == 1. 12T/K ) displayed in

figure 4 supports the superconducting nature of theresistive anomaly of "pure" La2CuO 4 - Y’ Consideringthe quite similar values of T, and dHc/dT values inpure and Sr-substituted La2CU04 - y . we may inferthat the electronic structure should not be thatdifferent in both cases. However, superconductivityin the present La2CuO 4 - y sample does not seem tobe a bulk phenomenon. The critical current is

extremely low and a preliminary attempt to measurethe Meissner effect has failed to detect more than0.1% flux expulsion [17]. Also in this sample, weobserved a weak resistivity anomaly at around 230Kwhich seems to correspond with that from suscepti-bility measurements [18,19].

Thermogravimetric results reported in table I dorequire some comments. The oxygen deficiency is

small in all Sr-substituted samples in agreement withthe early work of Michel and Raveau [5] and with aneutron diffraction study [16] establishing an upperlimit of 1% for oxygen vacancies in

Table I.- Results jrom the thermogravimetry.- Am/m measured from 1 000 °C. The relative weightloss between 20 and 550°C is smaller than 4 x 10- 3for the Sr-substituted sample.

Lal.85Bao.15CU04-y- On the other hand, "pure"La2CU04-Y exhibits a non zero content of oxygenvacancies which amounts to a concentration of 4%additional electrons in samples A and B. Sample Awhich shows superconducticity was slow cooled inair after firing at 1100°C whereas sample B quenchedin air exhibits a more insulating character and nosuperconductivity at low temperature. Quite similar-ly Cava and van Dover [8] could not find supercon-ductivity in a quenched powder of La2CU04-Taken in isolation, the results reported here are

not necessarily contradictory with the various

theories of strong coupling superconductivity basedon a non conducting pure and stoichiometric

La2CU04 compound and on superconductivity as-sociated with doping [20-23] : in this light,Laz - xSr xCuO 4 - y would be superconductive becauseof Sr doping ; La2CU04 - y because of 0 deficiency.However, two points must be stressed : (i) thebehaviour of the conductivity above Tc, which is verysensitive to 0 deficiency, does not look like beingdue to the approach of a metal-insulator phasechange by gap opening ; (ii) the rather small sensi-tivity of Tc on 0 deficiency [9]. They show thatsimple minded strong coupling scheme should bemodified to take such observations into account.One can also refer to the simple BCS scheme [24-

26], treating Sr doping and 0 deficiency effects asonly changing the band filling in a rigid band model.In such a picture, Tr should increase in the stronglydoped tetragonal phase when x - 2y is reduced,because the Fermi level EM would approach thecentral van Hove anomaly Eo (Fig. 5a). Below acritical value of (x - 2y ), the orthorhombic phaseshould become more stable, with the Fermi levelnear to but above the lower van Hove anomalyEl (Fig. 5b). Further reduction of (x - 2y) wouldpush the Fermi level away from El and this couldreduce Tc. Hence, a maximum of Tr could occur nearthe boundary of the phase change, as observed

experimentally. Analogous interplays between

1185

superconductivity and phase change are known :cubic to tetragonal phase change in quasi 1DV3Si type of compounds with doping [27], suppres-sion of a CDW under pressure in quasi 2D transitiondichalcogenides such as 2H-NbSe2 or 4Hb-TaS2 [10].In all cases, the phase change is associated with areduction in the density of states near the Fermilevel, which lowers Tc. In the same rigid bandmodel, the Tc observed in La2CU04 - y could beattributed to a raising of the Fermi level near to thesecond van Hove anomaly E2 (Fig. 5b).

Fig. 5.- Band structure in the simple BCS scheme : a)Tetragonal phase, b) Orthorhombic phase.

We believe that such a simple minded descriptionalso requires some modifications.Because the T ---,> 0 distortion is small (about

1%), the energy splitting of figure 5b is modest. As aresult, for any position of the Fermi level betweenthe two van Hove anomalies, they should add theireffects, leading to a Tc not very sensitive to bandfilling in this range [28]. To explain the total absenceof superconductivity in the vicinity of x = 0.04, onemust invoke the presence, in this range of Sr

concentration and in our samples (with y = 0.02), ofsome new insulating phase due to a nesting process.Indeed, because x - 2y - 0 in this range, the conduc-tion band should be about half filled. In the simplestapproximation for the copper (x2 - y2) d band,where Cu-O-Cu transfers are treated as pertur-

bations, exact nesting conditions for Q = ± !!.. ±2a

7r are realized in the orthorhombic [28] as well as in2bthe tetragonal phase (Fig. 6a,b). For such nestingwave vectors commensurate with a reciprocal latticevector, Umklapp terms in electron scattering couldbecome relevant and promote the suppression ofcharge degrees of freedom with a concomitant

Fig.6.- Nesting conditions for half filled band : a) tetragon-al phase ; b) orthorhombic phase.

vanishing of superconductivity. This situation is

reminiscent of superconductive-insulator transitionin organic superconductors of the (TMTTF )zXseries, where a spin density wave or spin Peierls statedevelop at low temperature even at moderate valuesof the interchain Coulomb repulsion when g3 Umk-lapp terms become pertinent [29,30]. There is indeedsome evidence of a SDW state below 240 K for smallvalues of x in the La2-.,,Sr.,CuO4-y series [19,31].More work is needed to check whether our sampleswith x :. 0.04 ± 0.02 show some evidence of such a

phase change.For x = 0, the practical absence of Meissner effect

in our samples suggests a very inhomogeneoussuperconductivity, which probably only developsalong the grain boundaries with an inhomogeneousrepartition of 0 vacancies. If the grains are in-

sulating, the preceeding discussion suggests that theyare nearly stoichiometric in 0, and that 0 vacanciesconcentrate near the grain boundaries. Indeed theweak resistivity anomaly observed by us and others[18,19] at about 240K in our undoped samplessuggests that the center of the grains has the

expected SDW phase. The finding of the same 0deficiency in A and B samples by thermogravimetricstudy, which is a volume determination of 0 vacan-cies, supports the picture for a surface only supercon-ductivity of sample A.

If 0 vacancies are concentrated, in our undopedsamples, near the grain boundaries, this means thatthe local shift of Fermi level due to 0 vacancies is

larger in the superconductive phase than that due tothe average value y = 0.02. One could thus be

locally nearer to a van Hove anomaly of the or-thorhombic phase. This would lead to a higherexpected Tc. However a second effect might thenhave to be taken into account : if a sizeable fraction

of the 0 vacancies are in the CU02 planes, each 0vacancy and its neighbouring Cu+ ions provide astrong scattering center for the conductive electrons,which could produce 2D localization effects on theseelectrons. This would explain the difference of

p (7) behaviour of our samples between x = 0 andx = 0.08.

Data on annealing superconducting samples undervacuum [32] are in fair agrement with our picture asan increase of 0 vacancies should shift the Fermi

level towards the centre of the band, promotelocalization and suppress superconductivity.

In conclusion, the purpose of this note was todraw attention on two major experimental evidences(i) the role of oxygen vacancies on the superconduc-tivity phenomenon of La2CU04-y or lightly Sr

substituted samples (x : 0.08) and (ii) the role of Srsubstitution on both the T, 0 transition and the

occurrence of a maximum in the superconductingTC -The observations are coherent with a simple BCS

1186

scheme and a band filling governed by a balancebetween 0 vacancies and Sr substitution. Transportproperties data suggest that disorder induced locali-zation plays a crucial role on the behaviour of theresistivity above the onset of superconductivity inLa2CU03-98- We also propose that the band becomeshalf-filled at a very low concentration of Sr-substitu-tion (x 0.04) leading to a carrier localization viaUmklapp scattering in near nested 2D bands with noloss in the spin degrees of freedom some whatsimilar to the observation in 2 :1 organic conductors.More experiments on varying and controlling thecontent of 0 vacancies, especially in

Laz-xSrxCu04-y with x small or equal to zero, areunder progress to check the validity of such a model[33].

Finally, a realistic model should analyse in somedetails the coupling term leading to the e-e pairing(thus the weak phonon mode associated with theT --o. 0 transition near x, could be partly responsiblefor the maximum of Tc observed at xc) [26].But the failure to observe any isotope effect on

oxygen [34] could suggest e-e coupling throughexchange of antiferromagnetic fluctuations [35]. Awealth of experimental and theoretical evidencessupport the existence of such spin-fluctuationmediated interchain pairing in quasi-lD organicsuperconductors up to temperatures of about 30-40K[36,37].

This work has been partly supported by theDRET contract 86/057. We thank J.C. Ameline andT. Pourthie for their skillful technical help.

References

[1] BEDNORZ J.G. and MULLER K.A., Z. Phys. B64(1986) 189.

[2] FOEX M., MANCHERON A. and LINE L. C. R. Heb.Acad. Sci. Paris 250 (1960) 3027.

[3] LONGO J. and RACCAH P.M., J. Solid State Chem. 6(1963) 526.

[4] SHAPLYGIN I.S. KAKHAN B.G. and LAZAREV V.B.,Zh. Neorg. Khim. 24 (1979) 1478 [J. Inorg.Chem. 24 (1979) 820].

[5] MICHEL C. and RAVEAU B., Rev. Chim. Miner. 21(1984) 407.

[6] SHAMOTO S., HOSOYA S., ONODA M. and SATO M.,submitted to Japan J. Appl. Phys.

[7] TAKAGI H., UCHIDA S., KITAZAWA K. and TANAKAS., preprint.

[8] CAVA R.J., van DOVER R.B., BATLOGG B. and

RIETMAN E.A., Phys. Rev. Lett. 58 (1987) 408.[9] BEILLE J., CABANEL R., CHAILLOUT C., CHEVALLIER

B., DEMAZEAU G. DESLANDES F., ETOURNEAUJ., LEJAY P., MICHEL C., PROVOST J., RAVEAUA., SULPICE A., THOLENCE J.L. and TOURNIERR., C.R. Heb. Acad. Sci. Paris, under press.

[10] FRIEND R.H., JÉROME D., FRINDT R.F., GRANT A.J.and YOFFE A.D., J. Phys. C10 (1977) 1013.

[11] MORET R., POUGET J.P. and COLLIN G., submittedto Europhys. Lett.

[12] van DOVER R.B., CAVA R.J., BATLOGG B. and

RIETMAN E.A., Phys. Rev. B35 (1987) 5337.[13] CAVA R.J., SANTORO A., JOHNSON D.W., Jr. and

RHODES W.W., preprint.[14] NGUYEN N., CHOISNET J., HERVIEU M. and RAVEAU

B., Solid State Chem. 39 (1981) 120.[15] LEBLANC, CHOISNET J., BASSAT J.M. and ODIER

Ph., to be published.[16] JORGENSEN J.D., SCHUTTLER H.-B., HINKS D.G.,

CAPONE Jr. D.W., ZHANG K., BRODSKY M.B.and SCALAPINO D.J., Phys. Rev. Lett. 58 (1987)1024.

[17] SENOUSSI S. and MONOD P., private communication.[18] JANOSSY A., private communication to FRIEDEL J.[19] GREENE R.L., MALETTA H., PLASKETT T.S., BED-

NORZ J.G. and MULLER K.A., submitted to

Solid State Commun.

[20] BASKARAN G., ZON Z. and ANDERSON P.W., SolidState Commun. , under press.

[21] ALEXANDROV A.S., RANNINGER J. and ROBASKIE-QICZ S., Phys. Rev. B33 (1986) 4526.

[22] PRELOBSEK P., RICE T.M. and ZHANG F.C., preprint.[23] CYROT M., submitted to Solid State Commun.[24] SCHUTTLER H.D., JORGENSEN J.D., HINKS D.C.,

CAPONE Jr. D.W. and SCALAPINO D.J., pre-print.

[25] LABBE J. and BOK J. Europhys. Lett. 3 (1987) 1225.[26] BARISI0106 S., BATISTI0106 I. and FRIEDEL J. Europhys.

Lett. 3 (1987) 1231.[27] LABBE J., Phys. Rev. 172 (1968) 451.[28] FRIEDEL J., to be published.[29] COULON C., DELHAES P., FLANDROIS S., LAGNIER

R., BONJOUR E. and FABRE J.M., J. Physique43 (1982) 1059.

[30] EMERY V.J., BRUINSMA R. and BARISI0106 S., Phys.Rev. Lett. 48 (1982) 1039.

[31] BEDNORZ J.G., TAKASHIGE M. and MÜLLER K.A.,Europhys. Lett. 3 (1987) 379.

[32] SHAMOTO S., HOSOYA S., ONADA M. and SATO M.,Jpn. J. Appl. Phys. 26 (1987) 493.

[33] KANG W. et al., to be published.[34] Communication of AT&T at MRS, Conference April,

1987.

[35] EMERY V.J., Synthetic Metals 13 (1986) 21.[36] BOURBONNAIS C. and CARON L.G., Phys. Rev.

Lett. (1987) to be published.[37] JÉROME D., CREUZET F. and BOURBONNAIS C.,

Proceedings of the Yamada Conference XV,p. 329, Lake Kawaguchi, Yamanashi, Japan,May 26-30, 1986.