Intermetallics 18 (2010) 2109e2118
Contents lists available at ScienceDirect
Intermetallics journal homepage: www.elsevier.com/locate/intermet
Magnetism in DyNi1xCuxAl pseudoternary series J. Prchal a, *, P. Javorský a, J. Poltierová Vejpravová a, O. Isnard b, B. Detlefs c, S. Danis a, V. Sechovský a a
Charles University, Faculty of Mathematics and Physics, Department of Condensed Matter Physics, Ke Karlovu 5, 12116 Prague 2, The Czech Republic Institut Néel, CNRS / Université J. Fourier, Boîte F, BP 166, 38042, Grenoble cedex 9, France c European Synchrotron Radiation Facility, 6 rue Jules Horowitz, BP 220, F-38043 Grenoble, Cedex, France b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 November 2009 Received in revised form 17 June 2010 Accepted 24 June 2010
The evolution of magnetism in the DyNi1xCuxAl pseudoternary series was studied. The results of magnetization, magnetic susceptibility, heat capacity and neutron diffraction experiments are summarized and discussed in context of other RNi1xCuxAl pseudoternaries. The compounds with x 0.2 adopt the coexistence of antiferromagnetic and ferromagnetic order similar to that one in the parent DyNiAl. The compounds with 0.3 x 0.6 exhibit antiferromagnetic order only while the compounds with x 0.9 are ferromagnets. The compounds with 0.6 x 0.9 exhibit signs of shortrange correlations, especially the compound with x ¼ 0.8 does not show any long-range magnetic order at all. It is in accordance with previously studied series. Ó 2010 Elsevier Ltd. All rights reserved.
Keywords: A. Rare-earth intermetallics A. Ternary alloy systems B. Electronic structure of metals and alloys B. Magnetic properties F. Diffraction
1. Introduction DyNiAl and DyCuAl belong to a large family of 1:1:1 rare-earth ternary RTX (R ¼ rare-earth, T ¼ transition-metal, X ¼ p-metal) compounds crystallizing in the hexagonal ZrNiAl-type structure, space group P-62m, no. 189. Existence of such a large group of isostructural materials allows systematic studies of magnetism of rare-earth ion in various chemical environments of T and X ions, which however maintain the symmetry of the R-ion neighborhood conserved. In the past, several pseudoternary series with alloying Ni and Cu in the transition-metal sublattice e TbNi1xCuxAl [1] and ErNi1xCuxAl [2], in particular e were studied. In both the systems the RNiAl parent compounds possess the antiferromagnetic (AF) ground state whereas the RCuAl counterparts become ferromagnetic (F) at low temperatures. The evolution of magnetism between antiferromagnetism and ferromagnetism with substituting Cu for Ni was found to be quite complex. The antiferromagnetism of the RNiAl compounds is rapidly destabilized with the Cu substitution. Ferromagnetic order has been observed in TbNi1xCuxAl and ErNi1xCuxAl already for x ¼ 0.1 and x ¼ 0.5, respectively. The ferromagnetism is, however, not maintained for all the higher concentrations of Cu. In both series the
* Corresponding author. Tel.: þ420 221911653; fax: þ420 224911061. E-mail address:
[email protected] (J. Prchal). 0966-9795/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.intermet.2010.06.019
long-range magnetic order (LRO) is lost within a narrow concentration range around x ¼ 0.8. This effect was discussed in terms of two competing exchange interactions of different type being controlled by the 3d-electron concentration [1,3]. Contrary to the two above-mentioned systems, the magnetism in the parent compounds of the Dy-based series, DyNiAl and DyCuAl, is rather alike. A neutron diffraction experiment performed on DyNiAl indicated ferromagnetism below TC ¼ 31 K with Dy magnetic moments aligned parallel to the crystallographic c-axis. At temperatures below T1 ¼ 15 K an antiferromagnetic component of the Dy moment perpendicular to c appears [4e6]. The magnetism in DyCuAl seems to be analogous; the only difference is observed in the temperatures of the magnetic phase transitions: TC ¼ 28 K and T1 ¼ 12 K. The bulk magnetization data fit well with this scenario [7]. The aim of the present study is to follow the evolution of magnetism within the Dy(Ni,Cu)Al series with respect to the Cu for Ni substitution within the transition-metal sublattice with special emphasis on the possible loss of the long-range magnetic order by studying bulk material properties (magnetization, AC magnetic susceptibility, heat capacity) and also microscopic aspects of magnetism by neutron diffraction. Some preliminary results were already published [8,9]; here the complete results are presented and discussed in the context of physics of the other RNi1xCuxAl systems. Further motivation of this work was the discontinuity in the concentration dependence of the lattice parameters a and c in the DyNi1xCuxAl series for x between 0.3 and 0.4 and its consequences in magnetic
2110
J. Prchal et al. / Intermetallics 18 (2010) 2109e2118
properties. Results of the structural study of this series were described in Ref. [8].
2. Experimental The measurements on the studied compounds were performed on polycrystalline DyNi1xCuxAl samples with composition x ¼ 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.8, 0.9 and 1.0 (DyCuAl). These were prepared by arc melting in a monoarc furnace under protection by the argon atmosphere after previous evacuation of the reaction chamber. The initial materials with the purity of 3 N for Dy, 4 N for Ni, 4 N for Cu and 5 N for Al were used in the nominal stoichiometric ratio. The samples were remelted five times to achieve better homogeneity. The samples were then checked for the phase purity by the X-ray powder diffraction and four of them were also checked for the composition by an electron-microprobe. The results for selected compounds (x ¼ 0.2, 0.4, 0.6, 0.8) are shown in Table 1. Fig. 1 displays the microphotograph of one of the samples (x ¼ 0.6). The stoichiometric composition of the main phase is in agreement with expected values e it is close (within the error of the method w2e3%) to the ratio of R:T:X ¼ 1/3:1/3:1/3 composition (resulting in the 1:1:1 stoichiometry). The ratio of the Ni and Cu content determines the real substitution parameter x. One can see that the real Cu content is slightly lower in comparison to the nominal one. Nevertheless, this shift is in the range of few %, which compares to the accuracy of the X-ray experiment. Moreover, in all the samples, one minority phase has been indicated corresponding approximately to a stoichiometry Dy2T4.5Al3.5. Analogous compounds were observed also within the EreCueAl systems [10]. A similar impurity phase was observed also in previously studied compounds but no observable influence on the magnetic properties of studied samples was detected by the neutron diffraction and the bulk property measurements, respectively [2,7]. The heat capacity and magnetization were measured in the temperature range of 1.8e330 K and under an external magnetic field up to 14 T using the Physical Property Measurement System (PPMS), Quantum Design, installed in the Joint Laboratory for Magnetic Studies (DCMP) in Prague. The temperature dependencies of magnetization were measured in regime “field-cooled (fc)” and “zero-field cooled (zfc)” depending if the external magnetic field was turned on above the ordering temperature (40 K) or at the lowest temperature (2 K), respectively. The AC susceptibility was measured using the same PPMS instrument in the temperature range 2e40 K with excitation field of 103 T and with various frequencies. The powder neutron diffraction was measured in the Institute LaueeLangevin in Grenoble, using the D1B diffractometer equipped by a standard banana-type multidetector, having a resolution of A. 0.2 . The wavelength of the incident neutron beam was 2.44
Table 1 The composition of the selected DyNi1xCuxAl compounds studied by the electronmicroprobe experiment. x
Phase type
Atomic fraction (in %) Dy
Ni
Cu
Al
Real x
0.2
Main phase Impurity phase Main phase Impurity phase Main phase Impurity phase Main phase Impurity phase
33.8 25.2 34.1 20.7 34.1 18.4 34.3 17.9
26.6 28.3 19.8 16.3 13.7 12.0 7.2 5.1
5.5 11.1 12.2 29.6 18.1 33.8 22.5 43.6
33.9 35.4 33.9 33.4 34.2 35.8 36.0 33.4
0.17
0.4 0.6 0.8
Fig. 1. Photographic zoom of the surface of the DyNi0.4Cu0.6Al sample studied by the electron-microprobe experiment.
A standard orange cryostat with temperature range 1.6e300 K was used for cooling of the sample. Program FullProf [11] was used for the refinement of the measured diffraction patterns. To reduce the high absorption of neutrons by Dy, an annular double-wall cylindrical vanadium container with the sample space restricted by diameters of 9 mm and 8 mm, respectively, was used for the experiment. Before starting the fitting procedure the obtained diffraction data were corrected using the program described in Ref. [12]. 3. Results and discussion In the paramagnetic region, the temperature dependence of the susceptibility of all the studied compounds follows the modified CurieeWeiss law.
c ¼
NA m0 m2B m2eff M þ c0 ¼ H 3kB T qp
(1)
where: meff is the effective magnetic moment, qp is the paramagnetic Curie temperature. The other symbols are constants usually used: NA and kB are the Avogadro’s and the Boltzmann’s number, respectively, mB stands for the magnitude of the Bohr’s magneton and m0 is the permeability of the vacuum. The parameter c0 is temperature independent. The values of c0 have been found close to zero in all our fits. The fitted values of the effective moment meff for all compositions are close to the Dy3þ free-ion value of 10.65 mB (Table 2 and Fig. 2). The values of the paramagnetic Curie temperature qp are positive in all cases, indicating major ferromagnetic interactions. The temperature independent term c0 was found to be of the order of 109 m3/mol in all cases, comparable to the values of the temperature independent susceptibility of YCuAl and LuCuAl [7]. When considering the behavior in the magnetically ordered state, the series can be divided into several concentration regions.
0.38 0.57 0.76
3.1. x ¼ 0.1 and 0.2 The low-temperature magnetization data of DyNi0.9Cu0.1Al document very similar behavior to parent compound DyNiAl with
J. Prchal et al. / Intermetallics 18 (2010) 2109e2118
2111
Table 2 Summary of magnetic properties for the DyNi1xCuxAl series. The results displayed for the compound with x ¼ 0.0 (DyNiAl) were obtained from the DyNiAl-single crystal measurements. The magnetic structure of the DyNi0.7Cu0.3Al compound was not determined due to low-quality diffraction patterns for this compound. The parameters obtained from the bulk data of DyCuAl were taken from [7], whereas the parameters of the magnetic structure were obtained from our results of the powder neutron diffraction experiment at 1.8 K. Signs of short-range order could be visible on samples with 0.6 x 0.9. T > Tord
x
Tord > T > T1
T¼2K
meff (mB)
qp (K)
Tord (K)
p.v.
Type
T1 (K)
mDy (mB)
p.v.
Type
Rmag(%)
0.0 [32]
10.65
30.0 (0.5)
(0 0 0)
Fkc
15.0 (0.5)
0.1
10.76 (1)
11Hkb 46Hkc 16.6 (2)
25.0 (1.0)
(0 0 0)
Fkc
13.0 (0.8)
0.2
10.75 (1)
13.3 (2)
21.0 (0.5)
(½ 0 ½)
AFkc
11.5 (0.2)
0.3 0.4 0.5 0.6 0.8 0.9 1.0
10.72 11.08 10.96 11.02 11.06 10.87 10.65
10.2 (3) 6.5 (2) 7.1 (3) 6.8 (2) 11.3 (2) 18.2 (2) 25.9 [7]
19.0 24.0 20.5 20.0 e 18.5 26.0
? e e e e e e
AF? e e e e e e
14.5 (0.5) e e e e e e
7.1 (1) 4.2 (1) 7.1 (3) 1.7 (4) 5.1 (3) 3.0 (3) ? 4.7 (2) 5.7 (5) 4.1 (2) & SRO SRO 3.8 (6) 5.8 (4)
(0 0 0) (½ 0 ½) (0 0 0) (½ 0 ½) (0 0 0) (½ 0 ½) ? (½ 0 0.448) (½ 0 ½) (½ 0 ½) e (0 0 0) (0 0 0)
Fkc AFkc Fkc AFkc Fkc AFkc AF? AFkc AFkc AFkc
6.2 10.0 5.4 ea 5.0 ea e 18 10 13
a
(2) (1) (2) (2) (2) (1) [7]
(0.5) (0.5) (0.5) (1.0) (1.0) (1.0)
Fkc Fkc
3.7 7.5
Due to low intensities of the AF reflections, the precise structure of the AF components cannot be refined.
ferromagnetic ordering below TC ¼ 25 K and appearance of the antiferromagnetic component below T1 ¼ 13 K (see Figs. 3a and 4a). The magnetization curves measured at temperatures below T1 exhibit a field-induced phase transition around 2 T. This is most probably due to the rotation of the antiferromagnetic component e that we expect to be located within the hexagonal basal plane in zero magnetic field, in analogy to the magnetic structure of DyNiAl e since this transition is seen only at temperatures below T1 where the antiferromagnetic component exists. At temperatures between T1 and TC one can observe magnetization curves typical for ferromagnets. The temperature dependencies of the real part of the AC susceptibility (c0 ) at selected frequencies (f) are shown in Fig. 5. The character of the c0 (T) curves coincides with the ZFC magnetization data shown in Fig. 3 and do not show any significant frequency dependence as expected for a spin glass, cluster-glass or superparamagnetic system. The curve related to the sample with x ¼ 0.1 exhibits a doubled anomaly at w25 K, and an additional small peak, found also on the ZFC magnetization (Fig. 3a) at w5 K. The expected
Fig. 2. The concentration dependence of the effective magnetic moment (meff), ordered moment (mtot), paramagnetic Curie temperature (qp) and ordering temperatures (Tord and T1) for the DyNi1xCuxAl compounds. Data for parent DyNiAl and results from the paramagnetic region for DyCuAl were taken from Refs. [7,32].
anomaly corresponding to the F-to-AF order may originate a weakly f-dependent decay from 20 K to 10 K. The doubled character of the anomaly on the real part of the AC susceptibility is probably caused by local fluctuation of composition of the measured piece of the material, that can occur namely at the grains boundaries. The small deviation of the composition, which comes from minority part of the sample (corresponding anomaly is clearly weaker than the main maximum), causes the slightly shifted transition temperature. For the refinement of the peak position, the dominant maximum corresponding to the static magnetization data was used. The c0 (T) dependence of the sample with x ¼ 0.2 shows a similar character, however, symmetric anomaly, corresponding to the TN depicts at 20 K and a kink appears at 10 K, respectively. A tiny f-dependent tail can be observed below 5 K. The neutron diffraction study of DyNi0.9Cu0.1Al confirmed close similarity of the magnetic structures to these found in DyNiAl in corresponding temperature intervals. Below TC a ferromagnetic order is observed with the Dy magnetic moments parallel to the crystallographic c-axis (Fkc). Below T1 an additional antiferromagnetic component (described by the propagation vector (½ 0 ½)) arranged within the basal plane (AFkc) appears (see Fig. 3a). Nevertheless, the exact arrangement of the AF component could not be determined due to very low intensities of the corresponding reflections (see Fig. 6). The sample with x ¼ 0.2 exhibits entirely different features e when cooling from paramagnetic state a transition to antiferromagnetic phase occurs first, at TN ¼ 21 K e as seen in the temperature dependence of magnetization (see Fig. 3b). Also the magnetization curve at 16 K, i.e. above T1 ¼ 11.5 K, is typical for an antiferromagnet with a metamagnetic transition at low magnetic field (see Fig. 4b and c). Below T1, the Dy moments most probably order ferromagnetically. The neutron diffraction data confirm the scenario of the strikingly different evolution of magnetic ordering in the compounds with x ¼ 0.1 and 0.2, respectively. The antiferromagnetic reflections with a compatible propagation vector (½ 0 ½) emerge around 22 K and the ferromagnetic intensities increase at lower temperatures around 11 K, as it can be seen on the Fig. 3b. The magnetic moments above T1 are oriented along the hexagonal c-axis and magnetic moments on the position (0, XDy, ½) are frustrated and consequently almost vanish. The magnetic structure is then similar to the high-temperature structure (T1 T Tord) of TbNiAl [13] (R14K mag ¼ 20%). Any other model e especially such with moment arrangements away from the c-axis direction brings rather worse fit. At 1.6 K the magnetic structure is similar to the ground state magnetic
2112
J. Prchal et al. / Intermetallics 18 (2010) 2109e2118
Fig. 3. Temperature dependence of the heat capacity, DC‑magnetization under field of 0.01 T and intensities of selected reflections measured by powder neutron diffraction for the DyNi0.9Cu0.1Al and DyNi0.8Cu0.2Al compounds. The vertical dashed lines point the temperatures of phase transitions as derived from the heat capacity.
structures of DyNiAl and DyNi0.9Cu0.1Al e the antiferromagnetic component lies in the basal plane and the ferromagnetic component then appears along the c-axis. The agreement factors for the antiferromagnetic component for the compounds with x ¼ 0.1 and 0.2 as displayed in the Table 2 are enhanced due to low intensities of the AF reflections at 1.6 K (see Fig. 6). For the same reason we could not refine the exact magnetic structure of the AF components within the basal plane. We can only tentatively conclude that they lie in the basal plane. Fit with the antiferromagnetic components along the c-axis leads to worse agreement, especially the calculated intensity on the position of the (½ 0 ½) and (½ 1 ½) reflections would be zero in such a case, while the measured one is clearly present. An example of the magnetic structure fit is displayed on the Fig. 7.
3.2. 0.3 x 0.5 The set of samples with x ¼ 0.3e0.5 exhibits very similar temperature dependences of the magnetization (as an example see Fig. 8). The data indicate a transition to an antiferromagnetic order at an appropriate TN and some anomalies at lower temperatures. The latter features, however, find no correspondence in the heat capacity and neutron diffraction data, which rules out suspicions on the ordereorder transition. The magnetization curves indicate suppression of the ferromagnetic component from x ¼ 0.1e0.4 (Fig. 9). The weakening of the ferromagnetic component of the Dy moment (with probable simultaneous enhancement of the antiferromagnetic component, i.e. the total Dy moments deviate from the c-axis) in this concentration region coincides with the minimum of qp (see Table 2). Since qp reflects the effective result of
Fig. 4. The magnetization curves of the DyNi0.9Cu0.1Al and DyNi0.8Cu0.2Al compounds. Opposite sequence of the F and AF order with respect to temperature can be visible also on these data.
competing ferromagnetic and antiferromagnetic exchange interactions in the material, the decrease of this parameter is compatible with enhancement of the antiferromagnetic exchange. The AC susceptibility dependence generally follows the ZFC magnetization, as demonstrated for the sample with x ¼ 0.4 in the Fig. 5. The anomaly at the TN w 22 K shows no f dependence. Below 15 K, a significantly f-dependent doubled anomaly takes place, which can be attributed to the DyNiAl-like AC response of Ni-rich fraction (s) of a slightly inhomogeneous sample. The neutron diffraction pattern of the DyNi0.7Cu0.3Al compound also revealed an antiferromagnetic order. The weak antiferromagnetic reflections and the low quality of this sample, that was caused probably by vicinity of the structural transition and possible existence of two different crystal phases at low temperature in this compound (different a, c parameters due to forbidden c/a ratio; see e.g. [8,14]), prevented us from determining the magnetic structure in more detail. Therefore only the information about the AF order without any other detail of magnetic structure is given in Table 2. The DyNi0.6Cu0.4Al compound exhibits an antiferromagnetic incommensurate structure with the propagation vector k0 ¼ (½ 0 q), where q ¼ 0.448, while the sample with x ¼ 0.5 orders again with the commensurate propagation (½ 0 ½) and moments adjusted within the basal plane (see Fig. 6). Again, we could not refine the exact magnetic structure of the AF components within the basal plane due to low intensities of the AF reflections. 3.3. x ¼ 0.6 and 0.8
Fig. 5. The real part of the magnetic susceptibility for all studied compounds except for the compound with x ¼ 0.5 which posses the same shape (only slightly shifted maximum in temperature) as the compound with x ¼ 0.4.
The samples with x ¼ 0.6 and 0.8 exhibit rather broader maximum in the M/H vs. T dependence and also very broad anomaly in the temperature dependence of heat capacity (see Fig. 10) that can indicate loss of long-range magnetic order (see below). The AC susceptibility of the x ¼ 0.6 and x ¼ 0.8 sample, respectively, lacks the typical features observed for the samples with x close to 0 or 1. Both curves show evident maxima with positions depending on the frequency (f) of the applied AC magnetic field (shown in Fig. 5). The upward-shift of the peak position in the c0 (T) curve with rising frequency is a typical feature of a spin glass (SG) material, which can be considered as an important evidence for the existence of random spin freezing effect in the sample. Concerning a competitive character of the ferromagnetic and antiferromagnetic interaction in the samples, the formation of the spin glass like state is highly probable. The imaginary part of the AC susceptibility (c00 ) demonstrates a sensitive response to any spin inhomogeneities and coincides rather well with the real part (Fig. 11), corroborating the SG scenario. Moreover, the f-dependent anomaly for f < 1 kHz cannot be observed in long-range ordered systems with well-defined ordering temperature because of MHz to GHz frequencies required for such shift of the corresponding maxima. However, beside the dominating anomaly a small shoulder appears at a temperature slightly lower than the peak temperature for the x ¼ 0.6 sample. It suggests that an inhomogeneous magnetic structure originating from inhomogeneous CueNi substitution exists in the sample to a certain extent. In order to inspect spin dynamics within the collapse of the long-range magnetic ordering for the critical concentration range (x ¼ 0.6e0.8), the AC susceptibility at varying frequency was investigated using a standard approach used for similar spin glass and cluster-glass intermetallic systems [15e17]. For a qualitative comparison of the freezing temperature, Tf vs. f dependence in various systems, a simple criterion (2) is often used. We obtained the values dTf ¼ 0.074 and 0.020, for the sample with x ¼ 0.6 and x ¼ 0.8, respectively. The values are comparable to other
2114
J. Prchal et al. / Intermetallics 18 (2010) 2109e2118
Fig. 8. Temperature dependence of the heat capacity, DC-magnetization under field of 0.01 T and intensities of selected reflections measured by powder neutron diffraction for the DyNi0.6Cu0.4Al compound. The vertical dashed line points the temperature of phase transition as derived from the heat capacity. The anomaly visible on the heat capacity data in DyNi0.6Cu0.4Al around 5 K is probably caused by the impurity phase.
dTf ¼ Fig. 6. Difference neutron patterns between the ordered state (or lowest studied temperature) and the paramagnetic state of DyNi1xCuxAl compounds. The data are shifted for better view.
concentrated SG or so-called cluster-glass systems, which are typically non-magnetic atom-disorder compounds like Ce2AgIn3 0.022 [18] or CeNixCu1x [16], but they are significantly higher than a typical dTf value observed in canonical SG or ferromagnetic cluster-glass compound U2IrSi3 (dTf ¼ 0.005) [15].
DTf Tf Dlogð2fp
Because the dTf value for the x ¼ 0.6 is rather close to 0.1, which is a value typical for a non-interacting superparamagnetic system, the data were subjected to the fit of the NéeleArrhenius law [19,20] using a characteristic relaxation time, s0, and a corresponding energy barrier, Ea/kB. The resulting values are: s0 ¼ 3.3 107 s and Ea/kB ¼ 86 K, which point to unrealistic assumption of the noninteracting spins due to extremely large Ea/kB value. The AC susceptibility data were further analyzed using the VogeleFulcher (VeF) law (3) with the characteristic parameters: T VF e VeF temperature as a measure of inter-spin or inter-cluster interaction strength, f0 e characteristic relaxation frequency, and Ea e activation energy of the relaxation barrier [21,22]. According to similar analyses published previously [23,24], the f0 value (related to the characteristic relaxation time as s0 ¼ 2p/f0) was kept fixed: 1013 s. To ensure a proper extraction of the Tfdependence, the c00 curves were used for the absolute maxima definition. The obtained values of the T VF are 5.3 0.5 K and 3.3 0.5 K, and Ea/kB ¼ 12.5 1.4 K and 2.8 0.3 K for the sample with x ¼ 0.6 and 0.8, respectively. The estimated activation energy, Ea is then proportional to the VeF temperature as follows 2.4 kBT VF and 0.9 kBT VF, for the x ¼ 0.6 and 0.8, respectively.
Ea f ¼ f0 exp kB Tf TVF Fig. 7. The refined neutron pattern of the DyNi0.8Cu0.2Al compound in the magnetically ordered state. Only the most significant reflections are signed.
(2)
(3)
The results of the VeF fit suggest, that both samples exhibit a typical cluster-glass behavior. The very low value of the x ¼ 0.8-related Ea
J. Prchal et al. / Intermetallics 18 (2010) 2109e2118
2115
Fig. 9. Comparison of magnetization curves of all studied DyNi1xCuxAl compounds at T ¼ 2 K. The strongest AF character can be visible for samples around x ¼ 0.4 and 0.5.
and its ratio to the T VF suggest either unique spin dynamic in the sample, or rather incorrect consideration of the fixed s0 ¼ 10‑13 s. The expected strong frustration of the FM and AF interaction for the critical x ¼ 0.8 may cause a considerably ‘faster’ fluctuation of the spins or cluster superspins resulting in a several order larger relaxation time. The analysis of neutron diffraction data collected on DyNi0.4Cu0.6Al also reveals an antiferromagnetic order. The magnetic reflections are quite weak; nevertheless the Dy magnetic moments (mAF ¼ 4.1(2) mB) could be determined. The values of moments are very reduced with respect to the free Dy3þ ion value (10 mB). Besides signs of short-range magnetic order can be observed (see the enhanced intensity at low angles in Fig. 6). These results may be interpreted by assuming a coherent antiferromagnetic ordering of the unified very reduced Dy magnetic moments throughout the entire sample. Alternatively, one may imagine another, for the substitution system rather probable, scenario involving a coherent antiferromagnetic ordering of Dy moments realized within (probably Ni-rich) clusters whereas the rest of the sample may consist of paramagnetic regions, and regions exhibiting a short-range order of Dy moments coexist. In any case, the ordered Dy magnetic moments are confined within the basal plane and propagate with a commensurate vector k ¼ (½ 0 ½). Appropriate low-angle neutron scattering experiments combined with another suitable microscopic experiment e.g. mSR spectroscopy or rather the 161 Dy are strongly desirable in order to resolve this complex problem. When spanning the Cu concentration x from 0.5 to 0.8, the ordered magnetic moment is decreasing and the long-range magnetic order gradually disappears resulting in the total loss of the long-range order for x ¼ 0.8, which was clearly manifest in the neutron diffraction experiment. This can be seen in the difference
Fig. 10. Temperature dependence of DC-magnetization and heat capacity for the DyNi0.2Cu0.8Al compound.
diffraction pattern between the lowest studied temperature of 1.6 K and paramagnetic state (40 K; see Fig. 6). The difference is practically zero except for the enhanced intensity at low angles, which indicates short-range ferromagnetic correlations around the (0 0 0) reflection. The low-temperature pattern of DyNi0.2Cu0.8Al at T ¼ 1.6 K can then be described by the nuclear structure only (with RBragg ¼ 5.7%; see Fig. 12). 3.4. x ¼ 0.9 and DyCuAl Bulk magnetization data measured for the compounds with x 0.9 with respect to temperature (Fig. 10) and magnetic field (Fig. 9) are rather typical for a ferromagnet as observed for DyCuAl before [7]. Strengthening of the ferromagnetic component from x ¼ 0.6e1.0 can be deduced from high-field parts of
Fig. 11. The imaginary part of the magnetic susceptibility for compounds with composition close to the critical concentration around x ¼ 0.8. The various marks for different frequencies correspond to the same values as in Fig. 5. Inset for each compound displays result of fit to the VogeleFulcher law (see text for details).
2116
J. Prchal et al. / Intermetallics 18 (2010) 2109e2118
Fig. 12. The neutron diffraction pattern of the DyNi0.2Cu0.8Al compound can be fitted just with the crystal structure at 1.6 K. No contribution coming from the long-range magnetic ordering can be found in this pattern.
the magnetization curves in Fig. 9. There is a small bend on the magnetization curves visible in low fields (w0.2 T), which was also observed earlier in DyCuAl [7]. This could be in accordance with expected antiferromagnetic component in the lowtemperature phase (in analogy with DyNiAl) but the antiferromagnetic component was not confirmed by the microscopic method (neutron diffraction). The real part of the AC susceptibility almost coincides with the low-field ZFC magnetization data, however the imaginary part shows a pronounced f-dependent response with two maxima at w8 K and w17 K for the lowest f value. The data were also analyzed using the VeF law yielding the T VF ¼ 4.6 0.5 K and Ea/kB ¼ 10.0 1.4. The values are very similar to those obtained for the samples with x ¼ 0.6. Moreover, dTf for the sample with x ¼ 0.9 is 0.072, which also coincide with that for the x ¼ 0.6. The behavior of the imaginary part of the AC susceptibility can be interpreted as follows. As determined by neutron diffraction, the sample orders ferromagnetically below w18 K. Due to local composition fluctuations and presence of grain boundaries, part of the sample exhibits in addition to the frequency-independent peak at 18 K also a clear anomaly around 8 K. The considerable frequency dependence of the 8 K-anomaly can be attributed to relaxation phenomena of spins in Cu-rich clusters and/or dangled spins on the grain boundaries. The neutron diffraction revealed simple ferromagnetic structures in DyNi0.1Cu0.9Al and DyCuAl (for DyCuAl see Fig. 13) with moments aligned along the c-axis and with magnitudes mDy ¼ 3.8 (6) mB and 5.8(4) mB for x ¼ 0.9 and 1.0 (DyCuAl), respectively. The magnetic reflections e and thus the magnetic moments e increase with increasing x (see Fig. 6). The apparently reduced value of the ordered magnetic moment was obtained under consideration that the whole sample orders ferromagnetically. However, just part of the volume of the sample is being ordered due to local composition fluctuations. The inner parts of the grains (‘core’) most probably have slightly different composition from that of grain boundaries region (‘shell’). Thus the more Cu-rich parts can exhibit ferromagnetic ordering while the behavior of the shell shows spin glass character with considerable interaction between neighboring shells what is in agreement with the observed AC susceptibility behavior discussed above. A simple ferromagnetic order in DyCuAl is in contradiction with the bulk magnetization data that indicated similar behavior to DyNiAl, i.e. with two magnetic phases and the AF component in the ground state. The strong dependence of magnetic properties on the sample quality could be an explanation as it was discussed e.g. for the isostructural HoNiAl compound [25]. The evolution of the magnetic entropy is presented in Fig. 14. It was derived from the magnetic contribution to the specific
Fig. 13. Temperature dependence of the heat capacity, DC-magnetization under field of 0.01 T and intensities of selected reflections from powder neutron diffraction for the DyCuAl compound.
heat, which was obtained by subtracting the specific heat of a non-magnetic analogue LuNiAl or LuCuAl from the specific heat of the studied compound. Although the difference between the specific heat of these two non-magnetic compounds is almost negligible, we follow the same procedure as in the ErNi1xCuxAl system: considering the abrupt change of the lattice parameters between x ¼ 0.5 and 0.6, we take LuNiAl (LuCuAl) as an estimation for Cph for compounds with x 0.3 (x > 0.3) because of their similar c/a ratio. The resulting data are discussed later. The summary of the refined magnetic moments, types and directions of the moments as well as the propagation vectors are shown in the Table 2. The agreement factors for the AF component for the compounds with x ¼ 0.1 and 0.2 are enhanced due to low intensities of the AF reflections at 1.6 K.
Fig. 14. The temperature evolution of the magnetic entropy for selected DyNi1xCuxAl compounds.
J. Prchal et al. / Intermetallics 18 (2010) 2109e2118
2117
4. General discussion Contrary to the fact that the parent compounds DyNiAl and DyCuAl exhibit similar bulk magnetic behavior (the dominant ferromagnetic order, the same type of magnetocrystalline anisotropy), the development of magnetic behavior across the entire series is quite complicated. Namely the change of the sequence of the magnetic-ordering types between x ¼ 0.1 and 0.2 as confirmed by both e bulk and neutron scattering data e is surprising. It expresses the importance of the density of conduction electrons that mediate the ReR exchange interaction (in fact the RKKY-type) as the key parameter influencing the magnetic behavior in this family of compounds. This fact is also confirmed by the loss of LRO in the region around x ¼ 0.8, similarly to the other RNi1xCuxAl pseudoternaries with R ¼ Tb and Er. The loss of LRO was discussed as a result of increasing the number of 3d electrons when increasing the Cu content which can lead to a complex development of exchange interactions [1,3]. The possibility of no LRO in compounds with the ZrNiAl-type structure is also theoretically predicted for a certain combination of nearest and next-nearest interactions [26]. As an answer to the question of influencing the magnetic properties by the discontinuity in the evolution of the lattice parameters a and c between 0.3 < x < 0.4 we can conclude that no abrupt change of magnetic structures was observed. This finding is in agreement with previously studied series with R ¼ Er [2] and Tb [1] exhibiting similar structural discontinuity. Only in case of the DyNi0.7Cu0.3Al sample the proximity to the critical a and c values probably caused inferior sample quality and thus the magnetic structure could not be refined into details. Although the specific-heat data alone do not allow reasonable determination of the crystal-field (CF) level scheme, certain indications can be deduced from the magnetic entropy represented in Fig. 14. The qualitative trends are the same as in the ErNi1xCuxAl [27]. When substituting Ni by Cu in DyNiAl, we first (up to x ¼ 0.3) observe a decrease of the entropy at temperatures below 40 K what indicates a shift of the excited CF levels to higher energies. In analogy to ErNi1xCuxAl, this behavior can be tentatively related to the increase of the lattice constant a. The tendency is interrupted between x ¼ 0.3 and x ¼ 0.4, probably due to the structural transition [8]. The increase of the magnetic entropy below 40 K for higher Cu concentrations (x > 0.4) then indicates a shift of the excited CF levels back to lower energies. We shall emphasize the fact that the same qualitative change of the CF level scheme occurs in the two investigated systems, DyNi1xCuxAl and ErNi1xCuxAl, in the Cu concentration, where the structural discontinuity happens, but these concentrations are different in these two systems. It strengthens our conviction about the relation of the CF change and the structural transition. The magnetocaloric effect (MCE) can also be determined from comparison of specific heat data measured in magnetic field and in zero field. The relative cooling power, calculated as the product of maximum DSM and the full width at half maximum of the DSM vs T plot, is shown in Fig. 15. In the Ni-rich part, MCE decreases compared to that in DyNiAl and the sign of the entropy change varies with temperature and Cu concentration. Such behavior reflects the complex magnetic structures that involve both ferromagnetic and antiferromagnetic components. In the Cu-rich part, MCE gradually increases with increasing Cu content. The relatively large magnetocaloric effect in the compound (x ¼ 0.8) without long-range magnetic order is an interesting observation, which indicates strong ferromagnetic short-range correlations. The existence of short-range correlations is also corroborated by the AC susceptibility measurements, where the samples with x ¼
Fig. 15. The specific heat and magnetocaloric effect for selected DyNi1xCuxAl compounds close to x ¼ 0.8.
0.6, 0.8 and 0.9 exhibit a pronounced frequency dependent response, namely on the imaginary part of the AC susceptibility. The results suggest formation of a cluster-glass state. There are two important phenomena to point out. First, the reference value of the dTf is about three times lower for the x ¼ 0.8 and identical for the other two considered samples. It should suggest stronger interaction of the spins (or clusters of spins), however the interaction strength characterized by the VeF temperature is much lower. The discrepancy can be explained either by fixing the unrealistic value of the relaxation time or formation of a different SG state closer to the canonical case, probably due to more homogeneous NieCu substitution. Second, the fit of the VeF law was applied on the data extracted from the imaginary part, which is more sensitive to spin disorder and inhomogeneities in the probed samples. On the real part, the slightly doubled maximum appears at almost the same temperature for the sample with x ¼ 0.6 and x ¼ 0.9, respectively; in contrast a single peak is on the c0 curve for the x ¼ 0.8. Using the f-shift extracted from the c0 curve, the corresponding dTf increases to 0.045 and the VeF parameter Ea/kB varies from 7.5 K to 13.0 K for the fixed s0 from 107 s to 1015 s, respectively, with almost zero T VF. Such result would corroborate the proposed very fast relaxation without significant internal interaction.
2118
J. Prchal et al. / Intermetallics 18 (2010) 2109e2118
The magnitudes of the ordered Dy magnetic moments in the studied compounds are smaller than the Dy3þ free-ion value of 10 mB. It is not so surprising, reduced moments are reported also in other isostructural Dy-based compounds, e.g. in DyPtSn (mDy ¼ 7.1 mB [28]) or DyAuIn (mDy ¼ 5.4 mB [29]). Generally, the magnetic moments can be reduced due to the effect of the crystal field, due to the geometrical frustration of the moments that occurs quite often in compounds with the hexagonal symmetry and triangular arrangement of the magnetic ions ([30]) or due to competition and strength of the exchange mechanisms. The crystal-field effect is mostly used in the literature to explain the reduced magnetic moments [31]. It certainly plays an important role also in our case. The geometrical frustration should be considered in case of the antiferromagnetic order, i.e. for x < 0.6 in our case. This effect might explain the reduction of the Dy moments between x ¼ 0.0 (DyNiAl) and x ¼ 0.5 where the reduction of the ferromagnetic component is not compensated by enhancement of the antiferromagnetic component. The largest total Dy moment is indeed observed in compounds with a large ferromagnetic component (total mDy ¼ 8.3 mB in DyNiAl). The third factor mentioned e the origin of the exchange interactions is probably the main reason of the strongly reduced magnetic moments or even the loss of LRO in compounds with x around 0.8. As we see, the size of the Dy moments throughout the DyNi1xCuxAl series is probably influenced by all the abovementioned factors. These results may be interpreted by assuming a coherent antiferromagnetic ordering of the unified very reduced Dy magnetic moments throughout the entire sample. Alternatively, one may imagine another, for the substitutional system rather probable, scenario involving a coherent antiferromagnetic ordering of Dy moments realized within (probably Ni-rich) clusters whereas the rest of the sample may consist of paramagnetic regions and regions exhibiting a short-range order of Dy moments coexist. In any case, the ordered Dy magnetic moments are confined within the basal plane and propagate with a commensurate vector k ¼ (½ 0 ½). Appropriate low-angle neutron scattering experiments combined with another suitable microscopic experiment e.g. mSR spectroscopy or rather the 161Dy are strongly desirable in order to resolve this complex problem. 5. Conclusions The development of the magnetic properties in the DyNi1xCuxAl series is very complex. The dominant ferromagnetic and smaller antiferromagnetic components are present in DyNiAl. With increasing the copper content, first we observe the weakening of the ferromagnetic component resulting in a pure antiferromagnetic order for x > 0.3. The long-range magnetic order then becomes gradually suppressed for x > 0.5 and completely vanishes for x ¼ 0.8. The simple ferromagnetism then again gradually appears for x > 0.8. We show that there is no strong relation between the magnetic properties and the structural transition around x ¼ 0.3.
Acknowledgments We acknowledge ILL for allocation of beamtime for performing the neutron diffraction experiment and the ILL staff for support during the experiment. This work is a part of the research plan MSM 0021620834 and the project LA339 both financed by the Ministry of Education of the Czech Republic. The work was also supported by the Czech Science Foundation, grants # 202/07/P153 and 202/08/0711. References [1] Ehlers G, Ahlert D, Ritter C, Miekeley W, Maletta H. Europhysics Letters 1997;37:269. [2] Prchal J, Javorský P, Sechovský V, Dopita M, Isnard O, Jurek K. Journal of Magnetism and Magnetic Materials 2004;283:34. [3] Prchal J, Naka T, Mísek M, Isnard O, Javorský P. Journal of Magnetism and Magnetic Materials 2007;316:e499. [4] Prchal J, Javorský P, Prokes K, Ouladdiaf B, Andreev AV. Physica B: Condensed Matter 2006;385e386:346. [5] Ehlers G, Maletta H. Zeitschrift für Physik B 1996;101:317. [6] Singh NK, Suresh KG, Nirmala R, Nigam AK, Malik SK. Journal of Applied Physics 2006;99. [7] Javorský P, Havela L, Sechovský V, Michor H, Jurek K. Journal of Alloys and Compounds 1998;264:38. [8] Prchal J, Javorský P, Danis S, Jurek K, Dlouhý J. Czechoslovak Journal of Physics 2004;54:D315. [9] Prchal J, Javorský P, Dopita M, Isnard O, Sechovský V. Journal of Alloys and Compounds 2006;408e412:155. [10] Kuz’ma YuB, Pan’kiv TV. Russian Metallurgy; 1989:208. [11] Rodriguez-Carvajal J. Physica B: Condensed Matter 1993;192:55. [12] Schmitt D, Ouladdiaf B. Journal of Applied Crystallography 1998;31:620. [13] Javorský P, Burlet P, Sechovský V, Andreev AV, Brown J, Svoboda P. Journal of Magnetism and Magnetic Materials 1997;166:133. [14] Prchal J, Javorský P, Rusz J, de Boer F, Divis M, Kitazawa H, et al. Physical Review B: Condensed Matter 2008;77:134106. [15] Li DX, Nimori S, Shiokawa Y, Haga Y, Yamamoto E, Onuki Y. Physical Review B: Condensed Matter 2003;68:172405. [16] Marcano N, Gomez Sal JC, Espeso JI, De Teresa JM, Algarabel PA, Paulsen C, et al. Physical Review Letters 2007;98:166406. [17] Marcano N, Gomez Sal JC, Espeso JI, Fernández Barquín L, Paulsen C. Physical Review B: Condensed Matter 2007;76:224419. [18] Nishioka T, Tabata Y, Taniguchi T, Miyako Y. Journal of the Physical Society of Japan 2000;69:1012. [19] Néel L. Annales de Géophysique 1949;5:99. [20] Wohlfarth EP. Physica BþC 1977;86e88:852. [21] Vogel H. Physikalische Zeitschrift 1921;22:645. [22] Fulcher GS. Journal of the American Ceramic Society 1925;8:339. [23] Tholence JL. Solid State Communications 1980;35:113. [24] Dho J, Kim WS, Hur NH. Physical Review Letters 2002;89:27202. [25] Javorský P, Sugawara H, Rafaja D, Bourdarot F, Sato H. Journal of Alloys and Compounds 2001;323e324:472. [26] Gondek q, Szytu1a A. Journal of Alloys and Compounds 2007;442:111. [27] Javorský P, Daniel P, Santavá E, Prchal J. Journal of Magnetism and Magnetic Materials 2007;316:e400. [28] Szytu1a A, Penc B, Kolenda M, Leciewicz J, Stüsser N, Zygmunt A. Journal of Magnetism and Magnetic Materials 1996;153:273. _ [29] Szytu1a A, Bazela W, Gondek q, Jaworska-Go1a˛ b T, Penc B, Stüsser N, et al. Journal of Alloys and Compounds 2002;336:11. [30] Dönni A, Ehlers G, Maletta H, Fischer P, Kitazawa H, Zolliker M. Journal of Physics: Condensed Matter 1996;8:11213. [31] Morosan E, Bud’ko SL, Canfield PC. Physical Review B: Condensed Matter 2005;71. [32] Andreev AV, Mushnikov NV, Goto T, Prchal J. Physica B: Condensed Matter 2004;346e347:201.
