J. Phys. F: Met. Phys., 12(1982)2393-41 I . Printed in Great Britain
Magnetic moments in manganese containing intermetallic compounds
M J Besnust, A Herr?, K Le Dang$, P Veillet$, A S Schaafsmag, I Vincze§*. F van der Woudet, F Mezei
11
and G H M CalisBt Laboratoire de Magnetisme et de Structure Electronique des Solides (associe au CNRS No 306). Universite Louis Pasteur. lnstitut de Physique, 67084 Strasbourg, France
$+ Institut d’Electronique Fondamentale, Universite Paris Sud. 91405 Orsay, France
#Solid State Physics Laboratory, Materials Science Centre, University of Groningen, The Netherlands
;I lnstitut Max von Laue-Paul Langevin. Grenoble, France
7 Laboratory for Physical Chemistry, Catholic University of Nijmegen, The Netherlands
Received 14 December 1981. in final form 5 April 1982
Abstract. Magnetisation, diffuse neutron scattering, 5 7 F e Mossbauer measurements and nuclear magnetic resonance (NMR) studies were performed on the pseudobinary (Fe, -.rMn,),Y and ( F e , -xMn,),B intermetallic compounds. It is shown that in both sys- tems the Mn atoms possess magnetic moments and that all these magnetic moments are ferromagnetically coupled. The concentration dependence of the individual Fe and M n moments has been derived from a combination of magnetisation. neutron and Mossbauer data. The mean Fe moments show a strong decrease with increasing M n content; the Mn moments appear to be less concentration dependent.
1. Introduction
In recent years much research has been carried out on the magnetic properties of intermetallic compounds incorporated in transition metals. The magnetic behaviour was found to depend sensitively on the nature of the transition metal involved, and on the local environment of the transition-metal atom within a given structure.
These among other conclusions were obtained from combined Mossbauer and magnetisation studies of the A1 substituted binary YFez and ternary (Y -xGd,)Fe, alloys (e.g. Besnus et a1 1978, Buschow 1977. Steiner 1979). In these investigations it was rather important that A1 be a non-magnetic diluent because direct determination of the individual magnetic moments by neutron scattering measurements is difficult due to the effects of large nuclear scattering and possible sublattice disorder. It has also been shown that the combination of Mossbauer experiments and bulk magnetis- ation measurements can be very useful in deducing the behaviour of the 3d magnetic moments in these systems.
In this paper we extend this work to the pseudobinary (Fel-.Mnx),Y and (Fel -xMn,)2B systems by using bulk magnetisation, diffuse neutron scattering, Mossbauer and NMR measurements. The crystal structures of Fe,Y and Fe2B are quite
* On leave from the Central Research Institute for Physics. Budapest. Hungary.
0305-4608/82/302393
+
19$01.50 @ 1982 The Institute of Physics 23932394 M J Besrius et ul
different, as discussed below, but they remain isostructural when Mn is substituted.
Both types of alloy are ferromagnetic, but Mn2Y and Mn,B are enhanced Pauli paramagnets, which leads us to ask whether or not manganese atoms possess magne- tic moments in the Mn-diluted systems. Thus the study of these crystallographically different systems may yield important information concerning the effect of different local environments on the formation and value of individual magnetic moments. Some very preliminary results of this study have already been reported (Schaafsma et al 1980).
The crystal structure and the local neighbourhood of the transition metals are quite different in these two systems (Pearson 1967). Fe,Y crystallises in the cubic C15 Laves phase structure. In this structure, the Y atoms situated on a diamond lattice have a cubic site symmetry (43m) and are surrounded by 12 Fe neighbours at 3.05
A.
The Fe atoms occupy the corners of regular tetrahedra, i.e. the crystallographically equivalent
5,n
sites with the threefold axes being parallel to the ( 1 1 1 ) directions.These sites may become inequivalent in the magnetic state depending on the angle between the magnetisation direction M and the CEF gradient, which is axially sym- metric and directed along the ( 1 1 1 ) axes. The number of distinct Fe sites and their population ratio is determined by the M direction. In Fe,Y the easy axis is along ( 1 1 I ) , leading to t w o subspectra in the Mossbauer pattern (Bowden et a / 1968). A gradual change towards a (100) direction is observed in the (Fe, Mn),Y ternaries (Van der Kraan et al 1980). In the C15 structure each Fe atom is surrounded by 6 Fe atoms at 2.60
A.
12 Fe atoms at 4.51A
and 6 Y atoms at 3.05A.
In a similar way to Fe,Y. Mn,Y crystallises in the MgCu, type structure and a solid solution exists throughout the whole concentration range. Fe,B has a body-centred tetragonal C16 structure with 42 point symmetry for Fe atoms and mm point symmetry for B atoms.Each Fe atom has 4B nearest neighbours at 2.18
A
and 11 Fe neighbours between 2.41A
and 2.72A;
each B atom has 8 Fe neighbours and 2B nearest neighbours at 2.12A.
The crystal structure remains unchanged when Mn is substituted.2. Experimental procedure
(Fe,-,Mn,),Y and (Fe,-,Mn,),B samples with Mn contents ranging from 0 to 100 atyo were prepared by induction melting the constituent metals in an argon atmos- phere. They were then homogenised in evacuated quartz tubes at 800 "C for several days, The purity grade of the starting materials was 3N for Y and B and better than 4N for Fe and Mn. Any possible loss of Mn during the preparation of (Fe, _,Mn,),Y was determined from the weight loss during preparation (assuming the total loss to be due to the evaporation of Mn) and also by neutron activation analysis. The actual Mn concentrations were found to be a maximum of 0.05-0.1 times lower than the nominal ones. The experimental results of the present paper are presented as functions of the nominal Mn concentration. A 5% loss of Mn was taken into account when estimating the error in the calculated values of the magnetic moments.
X-ray powder diffraction measurements, with either Fe or Cu Ksc radiation, were performed in order to standardise the phase homogeneity. In the Y series, the cubic C15 structure is observed throughout the concentration range and no foreign phases could be detected. Lattice parameters (shown in table 1) derived by the standard extrapolation procedure of Nelson and Riley (1945) were found to increase monotoni- cally with increasing x. with a slight deviation from Vegard's law.
Magnetic inonients in M n containing interrnetallics 2395
Table I . Lattice constants a at 300K. Curie temperatures, saturation magnetisation. M , . per mean atom and specific susceptibility x at 4.2 K of (Fe,Mn),Y.
At% Mn a(A) T,(K) M ( p B ) x ( x lo6 EMU g-I) T,(K)t M (pB)t
3.5 7.357
5.0 7.358
5.93 7.364
10.0 7.369
14.80 7.379 16.92 7.378
20.0 7.395
28.1 7.4 I6
30.0 7.404
40.0 7.442
50.0 7.465
60.0 7.505
69.8 7.527
75.0 7.556
80.0 7.564
85.0 7.602
92. I 7.631
97.5 7.642
100.0 7.652
513 0.904
- 0.922
487 0.897 484 0.878 406 0.796
- 0.745
391 0.720 327 0.594 310 0.619 245 0.504 160 0.406 75 0.243 35 0.107
- 0.032
- - - -
- -
- - - -
3.00 5.60 6.2 7.2 9. I 13.1 13.9 14.8 21.3 26.0 25.0 22.6 16.6 11.6 9.6 7.7 -
-
-
t Data related to (Fe, Mn),B alloys. from Cadeville (1965).
The magnetisation measurements were made on powdered samples by an induc- tion method in fields of up to 27 kOe between 4.2 K and 300 K, and in fields of up to 150 kOe at 4.2 K t . Curie temperatures, which also proved that no detectable amount of YFe, could be present in the alloys, were obtained from measurements in a low field. In the case of the (Fe, Mn),B series, the results of Cadeville (1965) were used.
The diffuse neutron scattering measurements were performed on the (Fe, Mn),Y series and were carried out using the D7 spectrometer at the high-flux reactor of the ILL at Grenoble with 4.75
A
neutrons. The samples were studied in powdered form and held at 4.2 K ; the applied field was 14.4 kOe.The Mossbauer spectra were recorded at 5 K and at room temperature with a conventional constant acceleration spectrometer. The external field dependence of the Mossbauer spectra of (Fe,,8,Mn,,20)2Y was recorded in longitudinal external fields ranging from 0 to 6.2 T at 4.2 K. In this experiment the 25 mCi "Co/Rh source was also held at 4.2 K in the zero field region. All spectra required a measuring time of about one day.
In the NMR study, the resonance signals of 55Mn, "B and "Y were observed at 4.2K in zero external field by a spin-echo method. In some cases, the external field dependence of the 5sMn and 89Y signals was measured.
3. Results
3.1. Magnetisation measurements
The results of the magnetic study are summarised in table 1. Upon Mn substitution the Curie temperatures of both systems decrease almost linearly and extrapolate to
t SNCI. Grenoble.
2396 R.1 J Besnrrs et a1
zero near x 1 0.6 for (Fe, -xMnx)2B (Cadeville 1965) and x
=
0.7 for (Fe, -xMn,)2Y (Schaafsma et a / 1980). In the case of the Y alloys, our results agree well with the recent magnetic measurements of Hilscher and Kirchmayr (1979). In this series no ferromagnetic order is observed down to 4.2 K for the alloys with x 2 0.75 as is shown by their plots of M 2 against H I M , though the M ( H ) curves show curvatures up to about 92.5 at% Mn. This observation is in very good agreement with the result from Mossbauer data that no magnetic ordering was observed until 4.2 K in this composition range. The ferromagnetic alloys show well-developed saturation behav- iour though one observes an increased high-field susceptibility with decreasing Fe content. Saturation magnetisations were derived from experimental data by fitting to the classical expression M = Ms(l - a H - ' )+
1 H . The field dependence in the para- magnetic concentration range may be satisfactorily fitted in the whole field range by the sum of a single Langevin function and the additional susceptibility term. In the high-field limit. the approach to saturation yields the magnetisation values of the remaining moments which may be more or less independent and persist as superpara- magnetic entities. Typical values of about 7-8 pB are observed.The M , values given in table 1 show an almost linear decrease with increasing x, with a correlated increase in the high-field susceptibility which exhibits the usual maximum in the critical concentration range. In both systems the average magnetic moments per transition-metal atom decrease faster than is explained by simple dilu- tion, though no drastic variation such as that due to A1 in YFe, is observed (Besnus et al 1978). As d,ii/dx + 0, the initial slopes are given as - 2.7 p B per Mn atom in the B series and - 1.8 p g per Mn atom in the Y series.
3.2. Difluse scattering of polarised and unpolarised neutrons
The measurements were performed at 4.2 K on five polycrystalline samples with Mn contents of x = 0, 0.041, 0.123, 0.253 and 0.393 at 32 scattering angles in the K range between about 0.03 and 2.45
A - '.
The measured diffuse intensity, corrected for instru- mental background, geometrical effects, incomplete incident polarisation, sample absorption and flipping efficiency, was converted to absolute cross sections by calibra- tion with a standard vanadium scatterer.The angular dependence of the nuclear intensity deviates from the expected behav- iour of the magnetic form factor only in the restricted K range below about 0.4
A-,.
The increase of the magnetic diffuse scattering effect at small K appears to be due mostly to nuclear correlations between second nearest or further neighbours. In ad- dition, we had to assume that the Mn atoms occupy only Fe sites and that no transition metal atoms occupy Y sites. This assumption is justified later in the dis- cussion. Thus the Fe-Mn moment difference is determined from the high K region of the spectra which appears to be negligibly affected by ordering effects, assuming a single form factor for the alloy, close to that of iron (Mezei 1976). The results for p F e and pMn obtained from a combination of the polarised neutron scattering data and the bulk magnetisation measurement results are plotted in figure 12. One observes a nearly linear decrease of individual moments from 1.37 to 0.87 pB for Fe and from 0.99 to 0.56 p B for Mn for the 4 and 40 at% samples. By comparison the Mn moment in Fe is found to be 0 . 6 5 ~ ~ at low temperature (Mezei 1976). Similar values for the Mn moment were also reported for the x Fe-Mn solid solution by Nakai and Kunitomi (1979, Child and Cable (1976) and Radhakrishna and Livet (1978).
Magnetic inoments i n M n containing intermetallics 2397
G
$
e 0 i 3 5; 0 1
c c 0 w U
-
-
,,.,,),,A
0 3 0 1 (bl. .
Figure 1. "Mn W R spectra in ( a ) (Fe,-,Mn,),Y and ( h ) (Fe,_.Mn,),B T h e dotted curbe indicates the signal in residual MnB.
0 2
3.3. N M R ineas ureiii en t s
The resonance signals of "Mn, "B and "Y nuclei were observed at 4.2 K. The nuclear spectra of Mn at Fe sites are shown in figure 1. In an applied magnetic field of 12 kOe the Mn spectrum is shifted towards lower frequencies, showing the ferromag- netic coupling between Fe and Mn moments. In the case of the (Fe, -,Mn,),Y com- pounds, only compositions with less than or equal to 20 at% Mn were studied because of the very strong broadening of the spectra.
Other resonance signals observed in (Fe, -,Mn,x),Y are shown in figure 2. In external fields of more than 3 kOe, the resonance frequencies of the observed lines decrease with increasing external field with a slope of about - 1 MHz per kOe. The threshold field of 3 kOe was also found for the Y line, so these extra signals should arise from the same phase and must be attributed to Mn impurities at Y sites, denoted Mn(Y). The corresponding concentration is about 5% of that of the Mn atoms at Fe sites. The lines at 380, 365 and 352 MHz may arise from Mn(Y) atoms having respect- ively 0, 1 and 2 Mn neighbours at the Fe sites.
The N M R spectra of "Y in (Fe,-,Mn,),Y are shown in figure 3. The field depen- dence of the Y resonance frequency has shown that the transferred hyperfine field is negative and it is not very different from the Y field in Fe (Marest et al 1978),
-226 kOe compared with -293 kOe. The intensity of the satellite lines in the Y spectra (figure 3) increases with Mn concentration. These lines may arise from the Y sites having 1, 2 . .
.
Mn neighbours with statistical weights corresponding to a random distribution of Mn atoms at Fe sites. However. their intensity corresponds to a Mn concentration which is systematically lower than the nominal one, giving 2.3, 5 and 7% instead of 3.5, 10 and 15%. if random distribution is assumed. The frequency shift due to the substitution of one Fe atom by one Mn atom is about 5.7 MHz, while the contribution of one Fe atom among twelve neighbours to the Y resonance frequency0 5
2398 M J Brsrius et a1
I
o ' 2300 350 400 450
Frequency F l MHz)
Figure 2. " M n spectra of Mn atoms at Y sites in ( F e , -IMn,),Y. So is the 'main' line: the low-frequency satellite lines are due to Mn neighbours at Fe sites and the high-frequency line is due to Fe neighbours at Y sites.
is only 3.87 MHz when the next nearest neighbours are neglected. As the Mn and Fe moments are parallel, the Mn moment should give a transferred field in the same direction. The observed shift indicates that the Fe moments around Mn atoms should be smaller than the undisturbed moments. In the cubic Laves phase structure one Mn atom disturbs six Fe moments, three of which are also nearest neighbours to the Y atom. The decrease in the Fe atomic moment would be 0.23 pB if the Mn contribution was to be neglected.
The Y spectrum shows that the transferred hyperfine field is about eight times the shift due to one Mn neighbour. By extrapolating to the Mn case, the contribution of Fe moments to the Mn hyperfine field would be - 100 kOe. The Mn(Y) moment may be estimated as 2.8 pB parallel to the host magnetisation since its total hyperfine field is - 360 kOe. It should be noted that the relative intensity of the satellite lines is larger for the Mn than for the Y spectrum, indicating preferred environments. A small signal proportional to the Mn(Y) line was observed at about 420 MHz. This can be attri- buted to Mn(Y) atoms having one Fe atom at a Y site as its nearest neighbour. No Y satellite line was detected between 50 and 80 MHz; it may be obscured by the high- frequency wing of the main line. Assuming a random distribution of the Fe atoms among the Y sites, we evaluated the Fe concentration from the relative intensity between the Mn satellite and main lines to be 5% with regard to Y.
The "B NMR spectra in (Fe,_,Mn,),B are shown in figure 4. From the M n hyperfine field, it could be concluded that the Mn moment is greater in Fe,B than in Fe,Y. The unresolved B spectra (figure 4) show that the contribution of the Mn moment to the B hyperfine field is not much smaller than that due to the Fe moment.
iMuqrzrtic moments irt M n coiitaininy interrnetallics 2399
r.
Lc
c 2.
C C 0 Y
+ .- 5
d 3
O I 5 , ~
,\
0 1
. . .
, , ..
*. .
, ...'I. 9.
,
.,.... ;
. ...
....
20 30 40
Frequency F IMHz)
50
Figure 3. ''Y N M R spectra in (Fe,-,Mn,J,Y. The .Y values are the nominal ones. The dotted curves represent a possible decomposition of the spectrum into lines with different near-neighbour environments.
X
0.1
N
Lc c 1.
C
E e
0.3Y
30 40
Frequency f IMHzI
Figure 4. "B NMR spectra in ( F e l _ x M n , ) , B . The arrows represent the assumed positions and the statistical weights of the lines arising from the sites with 0. 1. 2 . . . Mn neighbours.
The broken curves were computed using individual Gaussian lines of 3.3 MHz width.
2400 M J Besnus et a1
This hyperfine field, about 30 kOe, was also found to be negative from its frequency shift under an applied magnetic field. The resonance spectrum can be decomposed into individual lines corresponding to B sites having 0, 1, 2 . .
.
Mn neighbours by assuming a random Mn distribution and a frequency shift proportional to the number of Mn neighbours. The correlation between different samples show that the effective Mn concentrations, in this case, are close to the nominal ones.The Mn hyperfine field and consequently the Mn moment is very small in (Feo,,Mn,,,)2B. Thus both Fe and Mn moments decrease noticeably at this Mn concentration. The Mn spectrum at 220 MHz is ascribed to the residual MnB com- pound whose nuclear resonance frequency has been reported previously (Hihara and Hirahara 1965).
3.4. Mosshauer measurement.$
Typical magnetic "Fe Mossbauer spectra are shown in figures 5 , 6 and 7.
The fluctuation in the strength of the hyperfine interactions due to the fluctuating environments in these disordered systems can be described by the distribution of the hyperfine parameters. Information about the magnetic behaviour of Fe is contained in the magnetic hyperfine splitting of the Mossbauer spectra and can be given by the distribution of the hyperfine fields, p ( H ) .
Three different methods were used to describe the present Mossbauer spectra.
(i) A Window-type Fourier analysis (Window 1971) in which p(H) is expanded into a truncated Fourier series where the coefficients of the expansion are determined by fitting the spectra. Constant (i.e. independent of environment) isomer shift and zero quadrupole splitting are assumed (the continuous curves in figures 5 and 6 were obtained in this way).
(ii) A binomial approximation (Vincze 1978) in which p(H) is approximated by a limited number of equidistant discrete hyperfine fields, the relative intensities of which are given by a binomial distribution. Both the spacing of the hyperfine fields and the shape of the binomial distribution are determined by a least-squares fit to the spectra.
A product of two or more binomial distributions can be used depending on the lineshape of the measured spectra. The isomer shift was held constant but fluctuation in quadrupole splitting was allowed. (The histograms of p(H) in figures 5 and 6 were obtained by this method.)
(iii) Depending on the shape of the spectra two to seven independent six-line patterns with unrestricted intensities, hyperfine fields and quadrupole splittings were fitted. The value of the isomer shift was kept constant.
All of these evaluations have given the same average value (within experimental error) of the iron hyperfine fields,
HFer
throughout the investigated concentration range. In an earlier paper (Schaafsma et al 1980) we have reported average iron hyperfine field values in the same system obtained by the common practice of fitting a single six-line pattern with free linewidth and intensity parameters in order to repro- duce the broadening of the spectra with increasing Mn concentration. The values of HFc obtained in this way were systematically too high compared to the present results and correspond approximately to the maximum probability hyperfine field values. The reason for the discrepancy is that neither the asymmetrical shape nor the low-field part of p(H) are properly accounted for by these assumed symmetric lineshapes. The difference between the values is rather large; for example, in the case of (Fe0.7,Mn0.29)2Y the average six-line fit results inHFe
= 193 kOe, whereas the fitMugnrtic r?lor~~er~t.s in M n containing irzternietullics 240 I
I , 1
Y
I
(bl
m3 .
40 -
3 2 -
24 -
16 -
8 -
80 160 240
0 80 160 240
x10‘3
i
.6 -4 -2 0 2 4 6
v(mm i’) 0 80 HlkOe) 160 240
Figure 5. Mossbauer spectra of (Fe,_,Mn,),B compounds measured at 5 K : ( a ) (Fe,,,Mn,,,),B: ( h ) (Fe,,7Mno,,)2B: ( c ) (Feo,,Mno,,),B. The continuous curves were calcu- lated from the hyperfine field distributions, p ( H ) was fitted to the spectra. In the case of (Fe, .Mn,,,),B the p ( H ) obtained by the binomial distribution method (histogram) and also by the Fourier series method (dotted curve) are shown.
using the p ( H ) distribution gives
HFe
= JHp(H)dH = 171 kOe. Thus all the conclu- sions of Schaafsma et uI (1980) based on this erroneous evaluation are mistaken.A serious problem in the evaluation of highly asymmetric hyperfine field distribu- tions is the a priori unknown value of intensity ratios of line pairs in the six-line pattern, Z 1 . 5 : 1 2 . 5 : Z 3 . 4 = 3 : h : l . An improperly chosen h value may result in spurious features of the determined p ( H ) (Schaafsma 1982). Measurements were carried out in which the angle between the 7 direction and the plane of the sample was different (namely 90” and 45’) and the spectra obtained were identical. This effectively rules out any possible magnetic texture effect. i.e. the magnetisation directions are distributed
2402 M J Brsnus et a1
t
n
i i
Figure 6. Mossbauer spectra of (Fe, -xMn,),Y compounds measured at 5 K for various values of x : ( a ) 0; ( h ) 0.035; (c) 0.10; ( d ) 0.15; ( e ) 0.20;
(1)
0.29; (y) 0.40; ( h ) 0.60. The continuous curves were calculated from the hyperfine field distributions. p ( H ) was fitted to the spectra by two different methods: ( a t ( f ) binomial distribution method: ( g H h ) Fourier series method. The vertical scale for the histograms is p ( H ) kOe- ’). Note the change of scale of p ( H ) for Y > 0.20.Magnetic moments in M n conruining ititrrrnetullics 2403
100
9 9
90
91
1 '
- 6 -4 - 2 0 2 4 6
v (mm s-')
Figure 7. Mossbauer spectra of (Fe, 8Mn, z ) z Y recorded at 4.2 K in longitudinal external fields and in zero field. The full curves represent theoretical spectra consisting of Lorent- zian lines.
randomly corresponding to b = 2. This value agrees well with the description of the Fe,Y spectrum where only two six-line patterns are necessary because of the magnetic structure of the sample and the 3:2:1 intensity ratio was used in all evaluations reported here. The hyperfine field distributions obtained by different methods agreed quite well as illustrated in figure 5(b).
The evaluated hyperfine field distributions of (Fe, -xMnx)2Y show a very interest- ing feature for .Y = 0.20 and 0.29. A well-resolved satellite is observed at 80-100 kOe.
2404 M J Bcsnus et al
This value of the hyperfine field roughly corresponds to the second and fifth line position of the six-line patterns belonging to the main components and therefore their shape (or existence at all) is very much dependent on the value of b. The spectrum subtraction method of Vincze (1978) has given h = 2 and the existence of the low-field satellite in p ( H ) depends very much on this value. The validity of this evaluation was controlled in measurements by constraining the applied field to be parallel to the direction of the y rays. The magnetic moments of the sample are all aligned parallel to the external magnetic field above 0.52 T, thus reducing the number of lines to four per hyperfine spectrum, i.e. h = 0 (figure 7). These spectra have been fitted with Lorent- zian lines, assuming a distribution of hyperfine fields, characterised by five different subspectra, each having a weight factor. The best results were obtained taking 0.41.
0.27, 0.15, 0.10 and 0.07 as the weight factors throughout the series of spectra. One isomer shift was taken for all five hyperfine patterns, but the quadrupole splitting was variable for each hyperfine pattern, as in (iii) above. The effective fields are plotted in figure 8 as a function of the external magnetic field. For all components of the distribution of hyperfine fields the effective field decreases linearly as a function of Be,, with a slope of approximately one. The results for the components with the lowest effective fields are less accurate because of their small weight factors. However. it can
Figure 8. Effective magnetic fields obtained for the five components of the hyperfine field distribution from fits with Lorentzian lines. The circles in parenthesis are characteristic for the error of the evaluation of the nearest component.
Mugnetic niovlients in M n containing intermetallics 2405 be concluded that the magnetic fields of all the components of the hyperfine distribu- tion are aligned antiparallel to the external field.
These measurements in applied fields are important for two reasons: (i) they ex- clude the possibility of the presence of antiferromagnetically coupled Fe magnetic moments in a similar way to the NMR measurements; and (ii) they confirm the exist- ence of the small-field satellite deduced from the fitting of the spectra without external field.
Figure 9 shows the room-temperature paramagnetic spectra. The small asymmetry in the quadrupole doublets suggests a distribution of correlated isomer shift and quadrupole splittings, i.e. it corresponds to Fe atoms in different environments. This may arise from sublattice disorder or from the distribution of surrounding Mn neigh- bours.
I C 1
-2 -1 0 1 2
v ( m m s-’1
Figure 9. Typical room-temperature paramagnetic Mossbauer spectra of (Fe, -*Mn,)*B and (Fe, -IMn,)2Y compounds. ( a ) (Feo.sMno,s)zB; ( h ) (Feo,,Mn0 & Y ; (d
(Feo.4Mno.6)2Y; (4 (Feo 2MnO.&Y.
2406
-
& 150 5I E"
100
50
M J Besnus et a1
-
-
-
"
IC 0
0 0 . 4 0.8
X
Figure 10. Concentration dependence of the average Fe isomer shift with respect to pure Fe. Open circles. (Fe,-,MnJ2B; full circles, (Fe, -xMn,)2Y. All data refer to room temperature.
0 la1
0 0
* * O
0
0
0 0 4 0.6
2 5 0 r o
f
P €
0
0 0 4 0.8
X
Figure 1 1 . Concentration dependence of (a) the average transition-metal moments extrapolated to O K and ( b ) the average Fe hyperfine fields measured at 5 K in (Fe, -xMn,),B (open circles) and (Fe, -xMn,),Y (full circles). The p T V data of (Fe, -SMn,),B were taken from Cadeville (1965).
Figure 10 shows the concentration dependence of the average isomer shift with respect to pure Fe. The weak concentration dependence is characteristic of these types of intermetallic compounds and can probably be attributed to the change in lattice parameters and/or to the presence of some sublattice disorder (or impurity phase like that observed by NMR in (Feo.sMno,s)2B).
The average quadrupole splitting is rather small, A E Q is 0.21 m m s - ' for (Feo,sMno,5)2B at room temperature. In the case of (Fe, -xMn,),Y the room-tempera- ture average quadrupole splitting is slightly decreasing with increasing Mn concen- tration from a value of 0.34 m m s - ' at .Y = 0.40 to 0.27 m m s - ' at .Y = 0.92. This relatively small decrease can be attributed both to the direct (via overlap) and indirect (via change in lattice parameter) effects of the Mn substitution.
Figure 1 1 shows the concentration dependence of the average transition metal moments and average iron hyperfine fields.
4. Discussion
4.1. Mosshaurr, neirtron arid hiilk magnetisation results
From the concentration dependence of the shape of the Fe hyperfine field distribution p ( H ) we may conclude the following.
(1) The changes in the Fe hyperfine field due to the Mn neighbours are not additive in these systems. i.e. the effect of I I Mn first nearest neighbours is considerably larger than I I times the effect of a single Mn nearest neighbour.
(ii) The absence of additivity in hyperfine field changes means that no definite conclusion can be drawn about the random or non-random distribution of Mn atoms on the transition metal sublattice. Using an ad hoc model in which it is assumed that the second nearest neighbour transition metal atom also contributes to the Fe hyper- fine field, Van der Kraan er al (1980) concluded that the distribution of Mn on the transition-metal sites is not fully random in (Fe, -,Mn,),Y because the evaluated probability of Fe atoms having no Mn first and second nearest neighbours did not follow a binomial distribution. However the lack of resolution of the Mossbauer spectra at high Mn concentrations as well as the presence of a relatively strong magnetic anisotropy at low Mn concentrations (as reflected by the asymmetry of the lines in the spectra) result in rather inaccurate identification and intensity determi- nation of satellites with a specific environment.
(iii) In (Fe, -xMn,)2Y a low-field peak appears in p ( H ) for x 2 0.20. The origin of this peak may be sublattice disorder. i.e. Fe atoms sitting on Y sites; and/or a sudden decrease in the Fe magnetic moment (and hyperfine field) to about 40% of its original value when Fe has about three or more Mn neighbours.
This second type of assumption is often used to describe the concentration depen- dence of average magnetisation in pseudobinary rare-earth-transition-metal com- pounds (Buschow 1980). The critical number of 3 Mn nearest neighbours was obtained from the comparison of the low-field peak area to the main peak area and random distribution (i.e. binomial distribution) of Mn neighbours was assumed.
Detailed evidence is available in the case of the present type of compounds for the proportionality - of the hyperfine field and the individual magnetic moment, i.e.
H,, = In several Fe-Y compounds having rather different crystal structures the proportionality constant was found to be a = 147 kOe per pB (Gubbens et a1 1974). In
2408 M J Besnus et a1
the (Fe, C O ) ~ B system, which is isostructural to (Fe, Mn),B, the average magnetis- ation, the average Fe and CO hyperfine fields have been measured and were shown to be consistent over the whole concentration range with a strict proportionality between
HFe
and jiFe, and betweenHco
and jiCo. respectively (Cadeville and Vincze 1975, Takacs er a1 1975). The proportionality constant HFe/jiFe was 130 kOe per pB in this system.Figure 12 shows the calculated jiFe and jiMn values as a function of Mn concen- tration using the proportionality discussed above with a = 130 and 147 kOe per pB, respectively. ,GFe decreases strongly with increasing Mn concentration while jiMn is an approximately constant value of (0.5 & 0.2) p B in the concentration range investi- gated. The possible existence of a small overlap or conduction electron polarisation contribution does not influence the above conclusion. In this case
RFe
= ajiFe+
bjiTM,and if
I
b1
5 0.1 a as the data discussed earlier show, the calculated values of ,EFe change very little and even ,GMn remains constant within 0 . 2 ~ ~ . For example, in the case of (Feo,,Mno,3)2B with a = 130 kOe per pB and b = 0 we have ,GMn = 0.57pB, whereas for a = 115 kOe per pB. h =+
15 kOe per pB and a = 145 kOe per pB, b = - 15 kOe per pB we obtain jiMn = 0.49 and 0.64 pB, respectively.The agreement between the Fe moments deduced from neutron and Mossbauer effect data as well as their concentration dependence is good (as shown in figure 12), whereas it is worse for the Mn moments especially in the low Mn concentration range.
The basic assumption in the interpretation of the neutron scattering experiments on
\ 0
1
\ \, 1 \ \
,
0 0 2 0 4 0.6
X
Figure 12. Concentration dependence of the average magnetic moments of Fe and Mn, pFe and pMn. respectively as calculated: from the average transition-metal moment given in table 1, pTM (full and broken curves) and the neutron scattering data (pFc, open triangle;
pMn. open inverted triangle); from pTM and the average hyperfine field R,, assuming the proportionality of the hyperfine field and magnetic moment a s described in the text.
(pFc. circles: pMn. squares. Open symbols and the full curve relate to the borides, full symbols and the broken curve relate to the Y series.)
Magnetic rnoriients in M n containing interriietallics 2409 these ternary alloys is that the Y and transition-metal atoms occupy only their re- spective crystallographic sites. However, the nuclear disorder scattering was found to be unexpectedly high at low Mn concentration and to change less than expected between different concentrations. This may be due partly to the Mn/Y disorder as supported by the NMR study. Taking account of these results and of the fact that Mn moments on Y sites may have fairly large values, the Ap = ,EFe -
iiMn
values derived from the neutron experiments may be considered as the lower limit of the actual ones.Higher Ap values would scarcely affect the Fe moment values but could yield substan- tially lower Mn moment values, especially in the low Mn concentration range. As little as an 8% fraction of Mn atoms occupying Y sites could explain why pMn appears to be 0.4 p B higher in the neutron scattering experiment than in the Mossbauer effect data.
Unfortunately, the presence of Mn and/or Fe atoms at Y sites and the correlated possibility of Y atoms located at transition-metal sites introduces an additional number of parameters to the problem in excess of the number of independent measured parameters. Therefore more precise conclusions cannot be drawn.
4.2. Comparison of Mossbauer and N M R results
Because of the rather broad hyperfine field distributions the spin-echo measurements were limited for relatively low Mn concentration. Also the overlap of low-frequency signals originating from different nuclei may cause difficulties in the estimation of the average values. The ratio of the "Mn hyperfine field to its magnetic moment was found to be ( 1 18
i
4) kOe per pB in ferromagnetic (CO, Mn)B compounds (Lemius and Kuentzler 1980) and about 100 kOe per pB in other compounds containing Mn (Kawakami and Hihara 1968). The former value will be used during the present discussion.Despite the limitation in concentration and frequency range of NMR the results are in reasonable qualitative agreement with the Mossbauer findings.
4.2.1. (Fe,-,Mn,),B. The Mn hyperfine fields of maximum probability are about 190 kOe and 140 kOe for x = 0.1 and 0.3, respectively. These values correspond to magnetic moments for Mn of 1.6 pB and 1.2 pB respectively. They are larger than the
iiMn
values of 0.6 pB deduced for the x = 0.3 composition but the low-field (i.e. low magnetic moment) contribution may decrease the 1.2 pB maximum probability value of the Mn moment mentioned before. The high-frequency signal at around 220 MHz in (Feo,,Mn0,,),B was caused by a small amount of MnB impurity in the sample.B frequency approximately follows that of the average magnetisation. This observation excludes the possibility of appreciable ordering of Fe and Mn atoms.
The concentration dependence of the
4.2.2. ( F e , - x M n x ) 2 1 : Again the N M R results are in good qualitative agreement with the Mossbauer results. For example, the most probable "Mn hyperfine field in (Feo,9Mno,l)2Y is about 125 kOe, corresponding to a Mn moment of about 1.1 pB, close to the value deduced from the neutron data. While this is obviously larger than the jiMn value of 0.5 pB deduced from the Mossbauer study, the average Mn moment is rather difficult to estimate from the NMR measurements because of the overlap and decrease in sensitivity at low frequencies. The existence of the high-frequency satellites (between 320 and 400 MHz) is very remarkable. These were attributed to sublattice disorder-Mn occupying the Y sites. The satellites seem to be related to the low-field
2410 M J Besnus et a1
peak observed in the p ( H ) of Fe (the value is too small below x = 0.20 to be resolved in the Mossbauer spectra). These Mn hyperfine fields correspond to rather large (about 3 lB) magnetic moments. Since the Y sites are surrounded by 12 transition- metal nearest neighbours it seems reasonable for the Mn moments on these sites to be about the same as in the FCC transition-metal alloys containing Mn (e.g. Ni-Mn). O n the other hand, if the low-field Fe atoms correspond to the same type of sublattice disorder (i.e. Fe atoms occupying Y sites) and if the proportionality between hyperfine field and magnetic moment holds with the same proportionality constant then these Fe atoms have rather low (80-100)/147 = (0.54.7) pB magnetic moments. In FCC
alloys (e.g. Ni-Fe) the iron atoms have magnetic moments of about 3 p B when they are ferromagnetically coupled, but the moment values may depend sensitively on the
lattice parameter and on the local environments.
5. Conclusions
From the work described above, several conclusions may be drawn.
It is clear on the basis of neutron and NMR data that the Mn atoms possess magnetic moments in the (Fe, Mn),Y and (Fe, Mn),B systems despite their absence in Mn2Y and Mn2B.
All these magnetic moments are ferromagnetically coupled.
The value and concentration dependence of the individual magnetic moments of Fe and Mn have been determined.
A strong decrease of the mean Fe moments with increasing Mn concentration is found in the two systems. Good quantitative agreement is achieved for the Fe moments deduced from neutron and Mossbauer data. On the basis of the NMR data only a qualitative estimation of the value of the Mn moment can be made due to the absence of random distributions and additivity in neighbourhood effects, and it is difficult to achieve definite conclusions because of the signal overlap from the different nuclei. The determination of p M n from neutron data may be affected mainly in the low Mn concentration range by the presence of a few per cent of the Y-transition-metal disorder. Nevertheless we believe that the Mn moment values deduced from all the experimental data are in reasonable qualitative agreement and that, apart from minor quantitative discrepancies, this thorough study gives a coherent picture of the behav- iour of transition metals in these Fe-based intermetallics.
Acknowledgments
Part of this work belongs to the research programme of the Foundation for Funda- mental Research on Matter (FOM), with financial support from the Netherlands Organ- isation for the Advancement of Pure Research (ZWO).
References
Besnus M J . Bauer P and Genin J M 1978 J . Phps. F: Mer. P h p s 8 191
Bowden G J. Bunbury D St P. Guimaraes A P and Snyder R E 1968 J . Phys. C: Solid State Phys. 1 1376
Magnetic tnoments in M n containing intertnetallics 241 1
Buschow K H J 1977 Rep. Prog. Phys. 40 1179
~ 1980 Ferromagnetic Materials vol 1, ed E P Wohlfarth (Amsterdam: North Holland) p 344 Cadeville M C 1965 P h D Thesis University of Strasbourg, France
Cadeville M C and Vincze I 1975 J . Phys. F : Met. Phys. 5 790 Child H R and Cable J W 1976 Phys. Rev. B 13 227
Gubbens P C M. Van Apeldoorn J H F, Van der Kraan A M and Buschow K H J 1974 J. Phps. F : M ~ r t . Hihara T and Hirahara E 1965 J. Phys. Soc. Japan 20 879
Hilscher G and Kirchmayr H 1979 J . Physique Coll. 40 C5 196 Kawakami M and Hihara T 1968 J . Phys. Soc. Japan 25 1733 Le Dang K. Veillet P and Campbell 1 A 1977 Physica B 86-88 413 Lemius B and Kuentzler R 1980 J . Phys. F : M e t . Phys. 10 155 Marest G. Harotonian R and Berkes I 1978 Hyperfine Interactions 4 4
Mezei F 1976 Proc. Conf. on Neutron Scattering (Gatlinburg) vol 2. ed R M Moon. p 670 Nakai Y and Kunitomi N 1975 J . Phq's. Soc. Japan 39 1257
Nelson J B and Riley D P 1945 Proc. Phys. Soc. 57 160
Pearson W B 1967 Lattice Spacings and Structures of Metals and Alloys vol2 (Oxford: Pergamon) Radhakrishna P and Livet F 1978 Solid State Commun. 25 597
Schaafsma A S 1982 Phjis. Rec. B 23 4784
Schaafsma A S. Besnus M J. Vincze I and Van der Woude F 1980 J . Magn. M a p . Mater. 1 5 1 8 1149 Steiner W 1979 J . Magn. Magn. Mater. 14 47
Takacs L. Cadeville M C and Vincze I 1975 J . Phys. F : Met. Phys. 5 800
Van der Kraan A M, Gubbens P C M and Buschow K H J 1980 J. Physique Coll. 41 C1 197 Vincze I 1978 Solid Stare Commun. 25 689
Window B 1971 J . Phys. E : Sci. Instrum. 4 401 Phps. 4 921
