On magnetic interactions in quasicrystalline Al86Mn14 and Al75Mn20Si5 alloys

On magnetic interactions in quasicrystalline Al86Mn14 and Al75Mn20Si5 alloys

~Solid State Communications, Vol.64,No.4, pp.425-429, 1987. Printed in Great Britain. 0038-1098/87 $3.00 + .00 Pergamon Journals Ltd. ON MAGNETIC IN...

371KB Sizes 0 Downloads 36 Views

~Solid State Communications, Vol.64,No.4, pp.425-429, 1987. Printed in Great Britain.

0038-1098/87 $3.00 + .00 Pergamon Journals Ltd.

ON MAGNETIC INTERACTIONS IN QUASICRYSTALLINE AI86Mn14AND AI75Mn20Si 5 ALLOYS J.C. Lasjaunias and J.L. Tholence Centre de Recherches sur les Tr~sBasses Temperatures, CNRS, BP 166 X, 38042 Grenoble C~dex. C. Berger and D. Pavuna* Laboratoire d'Etudes des Propri~t~s Electroniques des Solides, CNRS, BP 166 X, 38042 Grenoble C~dex. France. (Received 23 july 1987 by E.F. Bertaut)

Low temperature specific heat and a.c. susceptibility studies on quasicrystalline AI86Mn14 and AI75Mn20Si 5 show that most Mn-s are nonmagnetic. The fraction of magnetic Mn-s (about 5-10% of all Mn atoms) exhibit spin-glass behaviour which in the case of AIMnSi is comparable to that of canonical spin-glasses like A g-Mn or Cu-Mn. Nuclear hyperfine specific heat term, spinglass ordering temperature and magnetoresistance, all indicate an enhanced magnetic character in quasicrystalline AIMnSi as compared to AIMn14.

Several investigations of structural [1], magnetic [2,3] or electronic [4] properties of quasicrystals (QC) have recently been reported. In previous publications [5,6] we have compared the magnetic and calorimetric properties of AI86Mn14 and AI75Mn20Si5-QC as well as their electronic properties with those of other QC samples. We presented the evidence that the large residual resistivity and the linear term of the specific heat could be ascribed to an enhanced density of states (DOS) coming from Mn-s atoms which act as isolated "impurities" in an AI matrix. Moreover, the magnetic character of these two QC phases changes rapidly when Mn content increases from 14 at.% to 20 at.%, leading to an unambiguous spin-glass state for the highest concentration. In this Communication we present detailed low-temperature specific heat, a.c. susceptibility and magnetoresistance measurements on AI86Mn14 and AI75Mn20Si5-QC. Our main results are that in AIMn14, we observe feeble nuclear specific heat term, spin-glass ordering temperature around 1 K from a.c. susceptibility and positive magnetoresistance, while in AIMnSi nuclear specific heat term is ten times larger, characteristic spinglass ordering temperature appears at 4.7 K and magnetoresistance is negative. All this confirms an enhanced magnetic character in AIMnSi as compared to AIMn14. Most of Mn-s in both samples are nonmagnetic; the fraction of magnetic ones is about 3-13%. All samples were prepared by melt-spinning in widths of up to 6 mm; their average thickness was ~ 20 i~m. They were characterized in great detail by X-ray diffraction and transmission electron microscopy. AI75Mn20Si 5 consists almost entirely of the

QC phase, with a typical grain size ~ 0.5 t~m and with less than about 10 vol% of fcc AI. AI86Mn14 consists of ~ 70 vol% of QC phase with typical grain size ~ lpm. The remaining represents fcc AI which contains up to 3 at% Mn in solid solution. Because of this second phase, we have also studied the behaviour of AIMn(3%) alloy. Exact total Mn concentrations agree with the nominal ones within less than 0.5 at.%. Specific heat (Cp) measurements were performed between 0.07 K and 7 K on pieces of ribbon of total weight between 200 to 400 mg by means of a step by step transient heat-pulse technique [7]. We note that other Cp measurements at higher temperature on icosahedral systems :AI80Mn20 [8], AI84Mn16 [9] indicate contradictory results for both the electronic y coefficient and the Debye temperature e D. The a.c. susceptibility %ac (f = 113 Hz) and magnetization were measured simultaneously down to 40 mK in a small dilution insert [10], and the magnetoresistance by a standard 4 probe technique in a 3He cryostat equiped with a 8 T superconducting magnet. Firstly, we compare the results of specific heat to those of magnetic susceptibility. As can be seen in Fig. 1 a transition to nuclear hyperfine regime from an electronic (or magnetic) regime occurs for Cp at T ~ 0.1 and ~ 0.2 K for AIMn14 and AIMnSi, respectively. The large bump at 1-2 K in AIMn14 we ascribe to a magnetic ordering which, as shown below, is also observed as a maximum in %ac (at T = 1.1 K, Fig. 3). We have analysed very low T regime as the sum of two terms : an electronic term which varies as yT (with y = 10 m J/mole K 2) for AIMn14 (Fig. 2), or as a power T 1-2 law extrapolated from higher temperatures (Fig. 1 and 2) for AIMnSi [11], and an

* Present address: Physics Departement, EPFL-1015 Lausanne-Switzerland. 425

426

ON MAGNETIC I

I

I

i

i

INTERACTIONS

i

10

.I .'/t • /

100 /

//

E

10

N

5

/ / /,/

~l

%

I

/

2

• o. ;;,

I

\

I

5

• g~

X\(m-2)

O

F E

?

~

2

°/ 0

o

U

/ / •



/

o o,,o ,,,,,,=cp° ° • /

/ / %.

u

g

t

I

/

AI86Mn14 (QC-1) ~/, • '~'-AI75Mn20Si5 tp (OC-2) / / .~

~. 20

?"-

,-,

/

Vol. 64, No. 4

"\

/

50

INALLOYS

A%6Mn14

.2

//

\

// /
0

0.5 0.5

,,t

,

0.1

, , .... I 0.5 1

,

0.2

:~

0 o

\

, ~ .... , T(K) 10 ,

o

\

\ \

o (~-2) 0

0.2

Fia.1 Specific heat of AI75Mn20Si5 (full pointsQC2), compared to that of AI8sMn14 (dashed curveQC1). The dashed-dotted line is the power T 1.2 variation extrapolated from data above 0.3 K for AIMnSi. hyperfine T -2 term, O N. This nuclear C N regime, which is totally absent either in AIMn (3%) solid solution or AIMn14 recrystallized [5,6] can only be ascribed to the effective magnetic field Heft induced by the electronic moments of magnetic Mn atoms (concentration x*) close to their saturation on their own nuclei [12]. f ON

=

x'~ N k B

IJf'l

-~-

I

0.05

t

till

I

0.1

0.2

J

T(K)

i

0.5

Fig.2 Hyperfine nuclear contribution (o~ T-2) deduced from total Cp assuming an electronic contribution ?T (dashed line) for AI86Mn14 (0), and varying as T 1.2 for AI75Mn20Si 5 (A).

~2

/IJ*N Hoff /

Lk - - ~ - - )

0.1

= AT "2

(1)

with IJ.N the nuclear moment and I = 5/2 the nuclear spin of 55Mn, N the Avogadro number. In this case we neglect other contributions from the matrix, i.e. from nonmagnetic Mn atoms, or AI atoms, which could contribute via the conduction electrons. All the magnetic sites, ferromagnetically or antiferromagnetically coupled, contribute equally to the C N term, which is not true for magnetization measurements. The amplitude of CN is ten times larger in AIMnSi ( A = 3 . 7 5 x 1 0 -5 J.K/mole) than in AIMn14, which indicates an enhanced magnetic character. This can be further substantiated by magnetic susceptibility measurements. As shown in Fig. 3, there is a well defined cusp of Zac at "If = 4.7 K in AIMnSi as compared to the rounded maximum at 1.1 K for AIMn14. Moreover, the frequency dependence of the cusp temperature is very close to the values reported for canonical spin-glasses like Cu-Mn, A_.g.-Mn, A u - F e [5]. Other recent magnetic studies have confirmed this similarity [13].

Taking into account the susceptibility data, we can now estimate the concentration x* of localized magnetic moments. Above 4 K, susceptibility has been analysed in terms of a Curie-like term, C/T, plus a constant contribution, %o (see insert of Fig. 4). From C, we estimate a mean effective moment P-eff per Mn atom of 0.51 j.tB and 1.0 IJ.B for AIMn14 and AIMnSi respectively, in excellent agreement with previously reported values [2,3]. Assuming the spin value S of the magnetic atoms, we obtain x* from expression : 2

x * N g 2 S(S + 1) I-LB

2

X* N IJ.

C =

= (2) 3k 8 3k a C decreases from 25x10 -3 emu/mole in AIMnSi to 4.5x10 -3 emu/mole in AIMn14o Assuming g = 2 , f o r S = l (l~=2.8#B) O r S = 2 ( l ~ = 4 . 9 # B ) w e obtain respectively x* = (4.5 _+ 0.5).10 -3 or (1.5_+ 0.2).10 .3 for AIMn14, and x* = (25 + 4).10 -3 or (8.5 + 1.5).10 -3 for AIMnSi. Now we discuss the value of the spin. Since AIMnSi exhibit magnetic

,o

ON MAGNETIC INTERACTIONS IN 7~LOYS

Vol. 64, No. 4

700

1

2

3

4 T(K

X

'A|86Mn14

30

6

ool. ,

, .

, I A175Mn20Si

5

• "

.Y

0

\

80 •

~

J

_=120L

:.

180

8

1

50 ~ 40 E

427

. . . . . .

1,o I lo

t,

Z

60

O

40 ti

AI75Mn20Si5(b)

....

U

;x~ 170

o 0

20

40

60

80

100

T tK]

160 Fia.4 a.c. susceptibility as a function of T at f =-113 Hz for (a) AI86Mn14 and (b) AI75Mn20Si 5. In the insert, %.T versus T. Fiq.$a.c. susceptibility of A I 8 6 M n 1 4 and AI75Mn20Si5; in the lower part we present the frequency dependence at 11, 111, 1100 Hz and 11 kHz. characteristics equivalent to ~.g.-Mn or C u - M n canonical spin-glasses, we can ascribe, like in these spin-glasses, a value of S close to 2, which corresponds to x* = 8.5 x 10 -3 [or a fraction of 4.2 % of Mn-s atoms which are magnetic]. Introducing this value in expression (1), we find Heft = 270 + 25 KOe in excellent agreement with previously reported values from nuclear orientation or nuclear Cp experiments in C u - M n [14,15] or A_g.-Mn [16] spin-glasses. Hence this confirms the classical spin-glass behaviour of this QC alloy. Turning now to AIMn14 QC alloy, we have also estimated the concentration of magnetic free impurities from the entropy Sm -- x* Nk B Log (2S+1) of the magnetic specific heat related to the large bump around 2 K [5]. For S = 1, we obtain x* = (7 _+0.7).10 -3 [or (5 + 0.5)% of the Mn atoms], in rather good agreement with the susceptibility, and for S = 2, x* = (4.6 + 0.5).10 -3 . Since values of S larger than 1 result in greater discrepancy with susceptibility results, we believe that the correct spin value is probably on the lower end. On the other hand, we have outlined that, in comparison to AIMnSi, the nuclear hyperfine contribution, C N, decreases by a factor of 10, whereas the coefficient of the Curie-like law decreases only by a factor of 5.5. If we suppose that both types of measurements are sensitive to the same concentration x*N of magnetic centres (expressions 1 and 2), the only plausible explanation is that the effective field He(( related to the same value of electronic spin S is smaller in AIMn14, i.e. the ratio of Heft to the magnetic moment

p. (in kOe per P.B units) is smaller by about 30% in AIMn14 than in AIMnSi (= 55 kOe/l~B for S = 2). For a series of Mn metallic alloys and elemental Mn metal, this ratio increases from ~ 30 kOe/PB in antiferromagnetic 7--Mn metal [17] or Mn-Fe alloys [18], to a maximum value of about 70 in spin-glasses [19]. The lower value of this ratio for AIMn14 again confirms weaker magnetic character as compared to AIMnSi-QC. Values of x* estimated from C N in AIMn14 by varying the effective field value from 30 to 55 kOe/P8, either for S = 1 or S = 2, are generally lower than estimated from the magnetic entropy Sin. This agrees with the fact that C N is essentially sensitive to moments close to saturation, whereas the anomaly around 2 K probably takes into account a larger distribution of moments, including weaker values. At this point, we discuss the small fraction of Mn-s atoms that are magnetic, roughly of the order of 5%, determined from both the Curie-like susceptibility coefficient and the magnetic entropy Sm in AIMn14. A similar value is determined in AIMnSi from both CN and the Curie-like law. However in that case, a rather large constant susceptibility %o was found (Fig. 4), of (4.6 + 0.3).10 -s emu/g, one order of magnitude larger than in AIMn14. The Mn atoms which contribute to %0 can be either nonmagnetic or antiferromagnetically (A.F.) coupled at high temperatures. Using first hypothesis and the susceptibility of Kondo impurities with T K = 600 K [20] we estimate %Kondo = 1.4x10 -5 emu/g for 20% Mn. This is of the same order of magnitude as %o observed for AIMn14 (%0 = 0.4x10 -5 emu/g) but for AIMnSi

428

ON MAGNETIC

INTERACTIONS

IN

ALLOYS

Vol. 64, No. 4

with a InT Kondo type temperature dependence for T < 2 K. This is further indication that in AIMn-QC most of M n - s behave very much as isolated impurities and give rise to an enhanced DOS at the Fermi level. The negative MR in AIMnSi is most likely mainly due to freezing out of the spin-flip scattering of conduction electrons on well defined magnetic moments. As magnetoresistance, resistivity and Hall effect results will be discussed at length elsewhere [22], here we emphasize only that the MR behaviour is broadly in accordance with the differences between AIMn14 and AIMnSi that we observed in specific heat and susceptibility results. In conclusion our experimental studies of low temperature specific heat and a.c. susceptibility show that most Mn-s are nonmagnetic in both QC AIMn14 and AIMnSi. The fraction of magnetic Mn-s (~ 5-10% of all Mn-s) exhibits spin-glass behaviour which, in the case of AIMnSi, is comparable to that of canonical spin-glasses like A__g.-Mn, C u - M n . From the observed hyperfine nuclear specific heat term (< 300 mK) and assuming spin S = 2 per magnetic moment, we have calculated an effective field of 270 kOe in agreement with value for canonical spin-glasses. The hyperfine nuclear term, the spin-glass ordering temperature, as well as the change of sign of magnetoresistance (negative in AIMnSi) all indicate an enhanced magnetic character in AI75Mn20Si 5 as compared to AI86Mn14.

%0 is much larger and could be probably explained by somewhat lower Kondo temperature. However, the second hypothesis of A.F. coupled Mn-s moments is also consistent with the value of XO and cannot be excluded from the nuclear specific heat CN, where such a contribution could be considered if one supposes for Heft a lower value than that for classical spin-glasses (~ 270 kOe). For example, if one takes for free moments (S = 2, x* = 8.5x10 -3) a lower limit of ~ 150 kOe, which corresponds to a ratio of only 30 kOe/~B and for the A.F. coupled moments another low limit of 65 kOe, like in y - M n metal, we estimate a concentration of about 10% of A.F. coupled Mn-s compared to a nominal concentration of 20%. Therefore, with these assumptions there should remain less than half of the Mn atoms with no magnetic moment. This seems to be contradictory with our measured large electronic y contribution (= 9 + 1 m J/mole K2), which is similar to that of AIMn14, and that we have interpreted as being due to an enhanced DOS due to a majority of Mn atoms acting as isolated nonmagnetic "impurities". Moreover, either NMR [3] or other magnetization experiments [13] excluded fraction of magnetic Mn atoms larger than about 20-25%, in agreement with our value of y. So the origin of the large %0 remains to be clarified. The difference in the magnetic characteristics of AIMn14 and AIMnSi samples is also confirmed by low temperature magnetoresistance (MR) measurements. As presented in Fig. 5 we clearly distinguish two different regimes: a positive MR for AIMn14 and a negative MR for AIMnSi. AIMn14 behaves very much as an A__[Mndilute alloy (with up to 0.55 at%) [21] showing the same order of magnitude in the relative MR change (&p/p ~ 4x10 -3 at 6 Tesla), and

Acknowledgements - We acknowledge many stimulating discussions with F. C y r o t - L a c k m a n n , J. Souletie and experimental help and assistance of G. Fourcaudot and J.C. Gri~co. One of us (C.B.) is grateful to C~g~dur-P~chiney for financial support.

&S x103 SB=°

0,3 K')

3

3

4

5

z~m.~

4K / A[

6

7 B(Tesla)

Mn

-1 -2 -3

F i a . 5 T h e longitudinal magnetoresistance of AI86Mn14 and AI75Mn20Si 5 at different temperatures.

VOI. 64, No. 4

ON MAGNETIC INTERACTIONS IN QUASICRYSTALLINE

429

REFERENCES 1. D. Shechtman, I. Blech, D. Gratias and J.W. Cahn, Phys. Rev. Lett. 53, 1951 (1984). 2. J.J. Hauser, H.S. Chen and J.V. Waszczak, Phys. Rev. B 33, 3577 (1986). J.J. Hauser, H.S. Chen, G.P. Espinosa and J.V. Waszczak, Phys. Rev. B 34, 4674 (1986). 3. W.W. Warren, H.S. Chen and G.P. Espinosa, Phys. Rev. B 34, 4902 (1986). 4. C. Berger, D. Pavuna and F. Cyrot-Lackmann, J. Physique 47, C3-489 (1986). D. Pavuna, C. Berger, F. Cyrot-Lackmann, P. Germi and A. Pasturel, Solid State Comm. 59, 11 (1986). 5. C. Berger, J.C. Lasjaunias, J.L. Tholence, D. Pavuna and P. Germi, to published. 6. J.C. Lasjaunias, C. Paulsen, M. Godinho, C. Berger and D. Pavuna, to be published in 6th Int. Conf. on Rapidly Quenched Metals (1987). 7. J.C. Lasjaunias and P. Monceau, Solid State Commun. 41,911 (1982). 8. F.L.A. Machado, W.G. Clark, L.J. Azevedo, D.P. Yang, W.A. Hines, J.J. Budnick and M.X. Quan, Solid State Commun. 61, 147 (1987). 9. M. Maurer, J. Van den Berg and J.A. Mydosh, Europhys. Lett., 3, 1103 (1987) 10. A. Benoft, M. Caussignac, J. Flouquet and J.L. Tholence (unpublished). 11. Alternative analysis for AIMnSi is to assume the same 71- regime as for AIMn14; it modifies the amplitude of this hyperfine regime by less than 10%. 12. It is unlikely that electrical quadrupolar interactions make any appreciable contribution to

C N, since from NMR experiments (W.W. Warren, H.S. Chen and J.J. Hauser, Phys. Rev. B 32, 7614 (1985)), the mean quadrupolar frequency ~Q is about 2 MHz, which would contribute to a few 10 -7 J.K/mole in C N. 13. R. Bellissent, F. Hippert, P. Monod and F. Vigneron, to be published 14. I.A. Campbell, J.P. Compton, I.R. Williams and G.V.H. Wilson, Phys. Rev. Lett.19, 1319 (1967). 15. P. Costa-Ribeiro, B. Picot, J. Souletie, and D. Thoulouze, Rev. Phys. Appl., 9, 749 (1974) 16. J.A. Cameron, I.A. Campbell, J.P. Compton, R.A.G. Lines and G.V.H. Wilson, Phys. Lett.20, 569 (1966). 17. J.C. Ho and NE. Phillips, Phys. Lett. 10, 34 (1964). 18. P.Costa-Ribeiro, P.Radhakrishna, D. Thoulouze and Y. Ishikawa, J. de Phys., 32, C 1-76, (1971 ). 19. I.R. Williams, I.A. Campbell, C.J. Sanctuary and G.V.H. Wilson, Solid State Commun., 8, 125 (1970). 20. J.R. Cooper and M. Miljak, J. Phys. F 6, 2151 (1976). 21. A. Hamzi~, E. Babi~ and B. Leonti6, Mat. Sci. & Eng. 23, 271 (1976). See also :A. Hamzi(~, and E. Babi6, Solid State Commun. 21,607 (1977). 22. C. Berger, D. Pavuna, F. Cyrot-Lackmann (unpublished).