Effect of annealing on Co2FeAl0.5Si0.5 thin films ... - APS Link Manager

PHYSICAL REVIEW B 83, 104412 (2011)

Effect of annealing on Co2 FeAl0.5 Si0.5 thin films: A magneto-optical and x-ray absorption study Simon Trudel,1,* Georg Wolf,1 Jaroslav Hamrle,1,† Burkard Hillebrands,1 Peter Klaer,2 Michael Kallmayer,2 Hans-Joachim Elmers,2,‡ Hiroaki Sukegawa,3 Wenhong Wang,3 and Koichiro Inomata3 1

Department of Physics and Research Center OPTIMAS, Technische Universität Kaiserslautern, Erwin-Schrödinger-Straße 56, D-67663 Kaiserslautern, Germany 2 Institute for Physics, Johannes Gutenberg-Universität, Staudinger Weg 7, D-55128, Mainz, Germany 3 Magnetic Materials Center, National Institute for Materials Science (NIMS), 1-2-1 Sengen, Tsukuba, 305-0047, Japan (Received 18 September 2010; revised manuscript received 10 December 2010; published 21 March 2011) A series of Al and MgO-capped Co2 FeAl0.5 Si0.5 epitaxial thin films grown on MgO with various levels of L21 ordering was obtained by in situ annealing. The films were studied by means of x-ray absorption spectroscopy, x-ray magnetic circular dichroism (XMCD), magneto-optical Kerr effect magnetometry, and Brillouin light scattering. We find the anisotropy constants decrease, while the spin wave stiffness increases as the samples are annealed to higher temperatures. The magnetization as determined by Brillouin light scattering reveals a maximum value at intermediate annealing temperatures. Surprisingly, the orbital-to-spin-moment ratio (as seen from XMCD) is essentially stable through the sample series and does not change upon annealing, despite the observed changes in anisotropy and exchange. DOI: 10.1103/PhysRevB.83.104412

PACS number(s): 75.30.Et, 75.30.Ds, 75.50.Cc, 78.35.+c

I. INTRODUCTION

Heusler compounds with composition Co2 MZ, where M and Z are a transition metal and a main-group element, respectively, are gathering considerable attention due to their predicted half-metallic character.1,2 Materials endowed with this type of electronic structure only have electrons of a given spin at the Fermi level, and as such, their current is fully spin-polarized.3 This is highly advantageous for spintronic applications, where the performance of devices is directly related to the spin polarization of the current emitted and collected by the ferromagnetic electrodes.2 In addition to high spin polarizations, Co2 MZ are attractive due to their high Curie temperatures, low magnetic damping, and tunable magnetic and electronic properties.1,4 However, demonstration of true half-metallic behavior (i.e., 100% spin polarization) has yet to be done. A major issue consists of preventing the population of electronic states in the minority spin gap.4–6 An approach to mitigating this effect is engineering the material such that the Fermi level lies directly in the middle of the minority spin gap. This makes the creation of states, for example, through band smearing near the valence band maximum and the conduction band minimum, less likely. Balke et al. and Fecher and Felser discussed how the Fermi level of quaternary compounds Co2 MZ1−x Zx is easily tuned by modifying the Z-to-Z ratio, where Z and Z have different numbers of valence electrons.7,8 For example, ab initio calculations show that the Fermi level of Co2 FeAl lies just above the minority valence band maximum, while for Co2 FeSi it lies near the minority conduction band minimum. Over the Co2 FeAl1−x Six substitutional series, the minority band gap remains approximately 760 meV. Meanwhile, the Fermi level itself shifts within this band gap.7 For Co2 FeAl0.5 Si0.5 (x ∼ 0.5), the Fermi level lies in the middle of the minority band gap,7,8 a result that was also observed by Nakatani and coworkers.9 In addition to tuning the Fermi level position, Nakatani also showed that the spin polarization was maximum for x = 0.5 for the bulk Co2 FeAl1−x Six quaternary 1098-0121/2011/83(10)/104412(10)

compounds, as determined by point-contact Andreev reflection at liquid Helium temperature.9 The preparation of high-quality Co2 FeAl0.5 Si0.5 thin films was demonstrated by magnetron sputtering by some of the current authors.10,11 Particularly interesting toward the mass production of practical devices is the demonstration of tunneling magnetic junctions built on MgO-buffered thermally oxidized silicon wafers. These junctions showed tunneling magnetoresistance (TMR) ratios of 196% at 7 K, corresponding to a spin polarization of 0.71.12 Recently, it was reported that Co2 FeAl0.5 Si0.5 thin films can have a high spin polarization, as high as 0.91 at room temperature, and show a weak dependence on temperature.13 Furthermore, the Inomata group has also recently used Co2 FeAl0.5 Si0.5 to achieve spin-transfer switching in an epitaxial spin-valve nanopillar.14 The critical current density required for switching was low, owing to the high spin polarization of Co2 FeAl0.5 Si0.5 . Co2 FeAl0.5 Si0.5 is thus an attractive and promising material for future spintronic devices. Annealing Co2 FeAl0.5 Si0.5 has been shown to be an efficient way to optimize several important properties for the production of high-quality tunneling junctions. For example, the surface roughness,10,15,16 saturation magnetization,10,12,15 and TMR ratios,15–18 and the impact annealing has on these, have been investigated. To study the effect annealing has on the crystal structure and local environment, Co2 FeAl0.5 Si0.5 samples annealed at various temperatures have been characterized using x-ray diffraction,10,15,16 reflective high-energy electron diffraction,11 and nuclear magnetic resonance.16 As-prepared deposited Co2 FeAl0.5 Si0.5 thin films are B2 ordered,10,12 as (002) reflections associated with the B2 superstructure are observed in θ − 2θ x-ray diffractograms. Pole figures do not show the (111) reflections characteristic of L21 -ordered structures. Upon annealing to temperatures up to 480 ◦ C, no or only very weak (111) reflections are observed,10,15,16 indicating that the sample remains B2 ordered. When annealing is conducted at or above 500 ◦ C, (111) reflections can be observed, and their relative intensity will increase with increasing annealing temperature.10,15,16 This indicates the improvement of the B2

104412-1

©2011 American Physical Society

SIMON TRUDEL et al.


ordering toward L21 . It must be noted that one is limited as to the maximum annealing temperature the sample can be subjected to. As Balke7 has shown, bulk Co2 FeAl1−x Six (x 0.4) will undergo an L21 → B2 phase transition between 752 and 852 ◦ C (852 ◦ C for x = 0.5). The improvement of local structure was also shown using nuclear magnetic resonance.16 For thin-film samples annealed at 600 ◦ C, the 59 Co spin echo intensity spectra could be well fit to an L21 -ordered model incorporating 16% of Fe(Al) antisites on the Al(Fe) sublattice. The tendency for Si to be well ordered, in comparison to Al, which participates in the formation of antisite defects, is attributed to the stronger Co–Si bonding.19 In this article, we investigate the magnetic anisotropy, thermal magnon spectra, and microscopic magnetic properties of Co2 FeAl0.5 Si0.5 and their dependence on local order. To accomplish this, we study a series of Co2 FeAl0.5 Si0.5 thin films subjected to various annealing temperatures and use magneto-optical Kerr effect (MOKE) magnetometry, Brillouin light scattering (BLS) spectroscopy, and x-ray absorption spectroscopy (XAS) and magnetic circular dichroism (XMCD) at the metal L2,3 edges to probe how this affects the resultant properties.

at 525 ◦ C and higher) samples were obtained. The effect of annealing on the the crystalline ordering of Co2 FeAl0.5 Si0.5 was reported previously (consult Refs. 10, 12, 15, and 16), and the reader is directed to these for more details.

II. SAMPLE PREPARATION AND CHARACTERIZATION

The reference spectrum Iref (hν) was measured by the bare substrate crystal and found to increase linearly with the photon energy. Iref (hν) was then normalized at the pre-edge region of the corresponding element. This normalization corresponds to an infinitely large penetration depth at the pre-edge. An external magnetic field of μ0 H = 1.2 T was applied perpendicular to the film surface to saturate the sample magnetization. The incident x-ray beam’s circular polarization was maintained the same throughout all spectra acquisition. The magnetic field orientation was switched to determine the XMCD signal parallel and antiparallel to the photon polarization (see details in Ref. 20). After subtracting a linearly varying background from the raw data, the TM and TEY XAS spectra were normalized at the postedge intensity. The XMCD spectrum is calculated as k + (hν) − k − (hν) for TM and I + − I − for TEY. Element-specific magnetic moments were derived by a sum-rule analysis. In the following we exclusively discuss relative changes of magnetic moments. Assumptions regarding the numbers of d holes are not necessary in this case.

A. Sample preparation and structural details

The samples studied were grown on single-crystal Mg(001) substrates. The stacking structure of the sample is as follows: MgOsub (001)/MgObuffer (20 nm)/Co2 FeAl0.5 Si0.5 (30 nm)/MgObarrier (0 or 3 nm)/Alcap (3 nm). Samples were prepared in a magnetron sputtering system with a base pressure of 8 × 10−8 Pa. The MgO(001) substrates were annealed under vacuum at 700 ◦ C for 1 h in situ prior to deposition. MgO was deposited by RF magnetron sputtering at room temperature, from a sintered MgO target under an Ar pressure of 1.3 Pa. Co2 FeAl0.5 Si0.5 was deposited by DC sputtering at room temperature from a stoichiometric Co-Fe-Al-Si target (Co, 50.0%; Fe, 25.0%; Al, 12.5%; Si, 12.5%), under an Ar pressure of 0.13 Pa and a typical deposition rate of 2 nm/s. Samples were annealed in situ for 30 min after Co2 FeAl0.5 Si0.5 deposition at temperatures of 480, 525, 550, 575, and 600 ◦ C. An as-prepared, unannealed sample was also investigated. Two series of samples were prepared. In the first series, the Co2 FeAl0.5 Si0.5 layer was capped by an Al layer. In the second series, a 3-nm MgO layer was deposited on the Co2 FeAl0.5 Si0.5 layer before Al capping. These samples are more representative of magnetic tunneling junction devices. Further details on sample preparation can be found elsewhere.10,11 The x-ray diffraction pole figures were measured for each sample for the (220) reflections. These confirmed that the sample grew with a Co2 FeAl0.5 Si0.5 (001)[100] MgO(001)[110] epitaxial relationship, consistent with previous reflective high-energy electron diffraction measurements.11 It was also previously shown that annealing improves the degree of L21 ordering in the sample, as inferred from the increasing intensity of (111) reflections observed in pole figures.10,15 The annealing temperatures were chosen so that B2-ordered (as prepared and at 480 ◦ C) as well as L21 -ordered (annealed

B. XAS and XMCD measurements

All MOKE, BLS, and XAS measurements described below were performed at room temperature. XAS experiments were performed at the UE56/1-SGM beamline at the German synchrotron light source BESSY II (Berlin). The incident photon flux was monitored by a Au mesh. The total electron yield (TEY) I + ,I − was measured via the sample current. The transmitted photon flux (TM) through the film was determined by x-ray luminescence in the MgO substrate, measured by a photodiode attached to the backside of the substrate. Assuming that the luminescence signal of the ± (where ± stands for x-ray photon helicity parallel substrate Ilum or antiparallel to magnetization direction) is proportional to the transmitted x-ray intensity, the x-ray absorption coefficient k ± for both magnetization directions can be calculated using ± Ilum (hν) ± . (1) k (hν) ∝ −ln Iref (hν)

C. MOKE and QMOKE measurements

For the Kerr rotation measurements as a function of applied field strength, the sample was mounted on a stage rotating about the sample’s normal. The measurements were carried out in the longitudinal configuration (i.e., the magnetic field is applied in the sample plane and is parallel to the plane of light incidence). The incident light (laser diode, λ = 670 nm) was s-polarized. The sample was oriented such that the edge of the MgO(001) substrate (i.e., an MgO 100 direction) is defined as the 0◦ in-plane orientation (i.e., 0◦ is along the Co2 FeAl0.5 Si0.5 [110] direction). For a given in-plane orientation α (with respect to the magnetic field) of the sample, the Kerr rotation is measured as the applied magnetic field is cycled between ±35 mT. Another Kerr rotation vs. field loop

104412-2

EFFECT OF ANNEALING ON Co2 FeAl0.5 Si . . .


is measured after rotating the sample by 1◦ , and this procedure is repeated until a loop has been measured for every in-plane orientation through a full rotation of the sample (360◦ ). Kerr rotation loops as a function of applied field were calculated based on a Stoner-Wohlfarth model22 as described in the Appendix. The expected Kerr rotation for a cubic material, when carrying out a series expansion of the dielectric tensor up to the second order of the magnetization,23 is the real part of s = A · KML K2 G ML MT + B − 2 + 2G44 + n˜ 2 −

2

G G cos 4α · ML MT − sin 4α ML2 − MT 2 4

, (2)

H270 H315

0]

O Mg H180

[10

H0

MT H135 ML

1 sat ML = 2 [(H0 ) − (H180 )], 1 sat ML MT = 4 [(H45 ) + (H225 ) − (H135 ) − (H315 )],

sat = 14 [(H0 ) + (H180 ) − (H90 ) − (H270 )] M 2 −M 2 L

T

and yield the LMOKE signal in saturation sat ML , as well as sat and . The measurements two QMOKE signals, sat 2 2 ML MT ML −MT were carried out using a MOKE magnetometer equipped with a quadrupole magnet, which is described elsewhere.25 D. BLS measurements

where A and B are (complex) optical weighting factors,24 K and Gij are components of the linear and quadratic magnetooptical tensors, respectively, G = G11 − G12 − 2G44 embodies the magneto-optical anisotropy, and n˜ is the complex refractive index of the material. ML is the in-plane component of the magnetization in the direction of the incident light, MT is the in-plane component perpendicular to the plane-of-light incidence, and the angle α gives the orientation of the sample with respect to its crystallographic axes. Due to the strong demagnetizing field in thin films, the out-of-plane component of the magnetization can be neglected. Besides the linear contribution, two quadratic contributions occur in Eq. (2), one proportional to the product of ML and MT and one proportional to the difference (ML2 − MT2 ). The ML MT contribution has a constant part and a part that is changing with cos(4α), while the (ML2 − MT2 ) contribution changes with sin(4α). The constant offset and the amplitude of the cosine (sine) function are given by the dielectric tensor components. To determine the magnitude of the different contributions to the total signal, the Kerr rotation n is measured when the applied field is along each of the eight field directions Hn shown in Fig. 1. The field strength μ0 H = 200 mT is chosen to saturate the sample in each direction. This procedure is repeated for every incremental rotation of 5◦ through a full rotation of the

H225

sample. The eight measured Kerr rotations n are combined according to (see Fig. 1 for geometry)

H45 H90

FIG. 1. (Color online) Geometry for the MOKE and QMOKE measurements in saturation (for the latter, each field direction n, μ0 Hn = 50 mT). The plane of incidence is along the H180 -H0 direction, and the angle shown is α in the text.

The BLS measurements were performed in the magnetostatic surface mode geometry, that is, the magnetic field H was applied parallel to the film surface and perpendicular to the plane of light incidence. A diode-pumped, frequency-doubled Nd:YVO4 laser (λ = 532 nm) was used as a light source. The angle of incidence ϕ (defined by the direction of the incident light and the sample’s normal) and the magnetic field were changed. A more detailed description of this instrument is found in Ref. 26. For a given angle of incidence the in-plane wave vector q of the detected magnons is given by 4π sin(ϕ). (3) λ Except where specifically indicated, the BLS spectra were recorded at an angle of incidence ϕ = 45◦ , which corresponds to q = 1.67 × 105 cm−1 . q =

III. RESULTS AND DISCUSSION A. Fe and Co L 2,3 edge XAS and XMCD

Representative XAS Fe and Co L2,3 edge spectra are presented in Figs. 2(a) and 2(b) for samples annealed at 550 ◦ C, with and without an MgO barrier deposited on the Co2 FeAl0.5 Si0.5 layer. We plot the area of the Co XAS satellite peak indicated in the inset of Fig. 2(b) as a function of annealing temperature in Fig. 3. It has been demonstrated that the spectral weight of this satellite peak at 3.8 eV above the Co L3 absorption edge directly correlates with the degree of atomic ordering in Co-based Heusler alloys,27,28 where higher intensities are observed for well-ordered (L21 ) samples. The satellite peak is more pronounced at the L3 edge than at the L2 edge because of the Coster-Kronig lifetime broadening and was explained by a hybridization of Co d-band states with sp states of the main group (nonmagnetic) element29 [see Fig. 2(b)]. The local atomic order increases with increasing annealing temperature, in agreement with the results from x-ray diffraction10,15,16 and nuclear magnetic resonance characterization.16 For the TEY data the maximum occurs at Tanneal = 575 ◦ C, and for the TM data the spectral weight is maximized at 600 ◦ C. Note that the TEY data are surface sensitive, with an information depth limited by the electron escape depth of 2.5 nm, while the TM signal integrates along the film normal, thus representing bulk properties of the samples. The fact that the satellite peak is more pronounced for the transmission

104412-3

SIMON TRUDEL et al.


(a)

(b)

(c)

(d)

FIG. 2. (Color online) X-ray absorption spectra (XAS) of Co2 FeAl0.5 Si0.5 annealed at 550 ◦ C at (a) the Fe and (b) the Co L2,3 edge averaged from the total electron yield (TEY) and transmission intensities (TM) measured at 300 K for the magnetization direction parallel (antiparallel) to the x-ray polarization, (I + + I − )/2. (c,d) Corresponding XMCD spectra, I + − I − . The inset in (b) indicates the determination of the spectral weight of the satellite peak in the Co XAS spectra. Solid lines, TM spectra; dashed lines, TEY spectra. Blue represents Al-capped samples; orange, MgO-capped samples.

the optimization of high-quality tunneling magneto-resistance device fabrication. Relative spin magnetic moments and the ratio of orbitalto-spin magnetic moments (ml /ms ) were calculated using the sum rules.21 The saturation magnetization of the investigated films exceeds the applied external field as noted in Fig. 11. The magnetic anisotropy is dominated by the shape anisotropy as deduced from the comparatively small values of the in-plane anisotropy constants. In this case the out-of-plane anisotropy is mainly given by the shape anisotropy. For external fields smaller than the saturation field, one then expects a constant magnetization component parallel to the field independent of the saturation magnetization. This is indeed what we observed. In particular, the spin moments derived from the transmission data are equal for the two series. However, interesting information can be gained from the comparison of surface and volume contributions to the magnetic moment of the films. Considering the small crystal anisotropy constants as discussed below, we assume a homogeneous perpendicular magnetization component throughout the film. Moreover, the large exchange coupling results in parallel Fe and Co moments. Therefore, we consider the ratio ms (TEY)/ms (TM) of the surface and volume contribution to be equal to the ratio of the corresponding perpendicular magnetization components measured below saturation. These values are summarized in Fig. 4. For both film series, we observe the same values for surface and bulk in the case of the magnetic spin moments located at the Fe atoms, irrespective of annealing temperature

(a)

(b)

(c)

(d)

signal indicates that the bulk of the film shows a higher degree of local atomic order than the surface region. For the surface of the series of samples capped with an MgO barrier, we observe a particularly small satellite peak. This observation indicates that the local atomic order at the film surface is degraded by the interfacial MgO layer, thus suggesting a route toward

FIG. 3. (Color online) Spectral weight of the satellite peak after linear background subtraction vs. annealing temperature, indicating the increase in local atomic order with increasing annealing temperature. , Al-capped, TM; •, MgO-capped, TM; , Al-capped, TEY; ◦, MgO-capped, TEY. Lines are guides for the eye.

FIG. 4. (Color online) Ratio of surface (TEY)- and bulk (TM)related spin magnetic moment ms (TEY)/ms (TM) for (a) Fe and (b) Co. Data for the Al-capped series are indicated by ; data for the MgO-capped series, by ◦. The ratio of orbital-to-spin magnetic moment ml /ms for (c) Fe and (d) Co as a function of annealing temperature. Labels for (c) and (d) as in the captions to Fig. 3. Lines are guides for the eye.

104412-4



[Fig. 4(a)]. Comparable ratios are seen for the the Co spin moments in Al-capped films [Fig. 4(b)]. This is in contrast with previous observations for Co2 Cr0.6 Fe0.4 Al films,20 but consistent with improved deposition parameters30 for the same composition and also for Co2 FeSix Al1−x films.31 In contrast, the MgO-capped films show an increased Co spin moment for annealing temperatures below 550 ◦ C [Fig. 4(b)]. The enhanced Co moment near the surface is a direct consequence of the increased disorder in this region. While a Co moment of 1.3 μB is expected for an ideally ordered Heusler compound, the moment for bulk Co metal is considerably larger (1.6 μB ). Therefore, disorder tends to increase the Co moment, in good agreement with previous observations.32 This explanation is also consistent with the observed smaller satellite peak in the Co XAS spectra shown in Fig. 3. For an annealing temperature of 575 ◦ C the difference between the surface and the bulk moment nearly vanishes, thus indicating a better ordering at the MgO interface. Upon further annealing at 600 ◦ C the Co interface moment rises again with respect to the bulk moment, indicating that order on the Co sites is reduced again by transposition of Co and Fe. The existence of such a mechanism was demonstrated for the Cr-containing related compound Co2 Cr0.6 Fe0.4 Al with the use of Mössbauer spectroscopy.33 The orbital-to-spin ratio ml /ms does not show a significant trend with increasing annealing temperature, for either bulk or interface values [Figs. 4(c) and 4(d)]. However, values resulting from the TEY data are systematically lower than the TM data. Decreased orbital moments could be an indication of improved local order with a better approximation to the perfect cubic symmetry. However, this explanation can be excluded, as the XAS data for the same series shows smaller satellite peaks in the TEY data compared to the TM data. An alternative explanation could be the influence of spin-orbit coupling, that is, magnetic anisotropy. Orbital moments are suppressed if the magnetization is rotated into the hard magnetic axis.34 The interface magnetic anisotropy in both series might prefer an in-plane magnetization. In our measurement the magnetization is forced out of the film plane, hence into the hard axis, thus causing a decrease in the orbital moment in the interfacerelated TEY data. The XMCD results demonstrate that bulk characterization methods are of limited significance for the optimization of Heusler thin films incorporated into tunneling junctions or spin valves. The measurements of the annealing-temperaturedependent surface magnetic moment of the Co2 FeAl0.5 Si0.5 films are consistent with the observation that the maximum TMRs of TMR junctions with Co2 FeAl0.5 Si0.5 electrodes are obtained with annealing temperatures in the intermediate range.17 B. MOKE magnetometry

Kerr rotation loops were measured as a function of inplane sample orientation. Representative measurements are presented for the as-prepared, Al-capped sample in Fig. 5. As shown, the magnetization reversal does not occur in a single switching event for all orientations. For example, for angles 0 ◦ to 25 ◦ , two-step magnetization switching is observed. This type of switching has been observed elsewhere

FIG. 5. (Color online) Measured Kerr rotation loops at various in-plane orientations for the as-prepared sample. The vertical scale is the same for both graphs.

for Heusler compound thin films, as reported, for example, by Yang,35 Ambrose,36,37 and coworkers [who studied GaAs(001)/Co2 MnGe(001) samples], and was recently reviewed by Trudel et al.4 This behavior can be explained by a superposition of a uniaxial and cubic in-plane anisotropy, resulting in an overall twofold anisotropy. At 90 ◦ (corresponding to the [110] direction in Co2 FeAl0.5 Si0.5 ), the magnetization is along the easiest axis of magnetization, and a square magnetization reversal is seen. The 0 ◦ orientation (the [110] direction in Co2 FeAl0.5 Si0.5 ) corresponds to the second-easy axis. When the magnetic field is along this direction, the magnetization will coherently rotate not to the opposite direction but, rather, toward the easiest axis, which is perpendicular to the second easy axis (i.e., along 90 ◦ ). This corresponds to the first switching event in the loop measured at 0 ◦ in Fig. 5. As the magnetization is perpendicular, no Kerr rotation is recorded by the MOKE detector, and this corresponds to the plateau in the loop shown in Fig. 5. When a sufficiently high field is applied, the magnetization will then reorient in a second sudden switching step, to align with the second easy axis. The hard axis lies between the two easy axes, namely, around 45 ◦ (the Co2 FeAl0.5 Si0.5 [100] direction). This is consistent with the rounded character of the magnetization reversal loop shown in Fig. 5 for 45 ◦ and 135 ◦ . In an attempt to quantify this behavior, we have used a Stoner-Wohlfarth model22 to extract cubic and uniaxial anisotropy constants (KC and KU , respectively). Typical results are shown in Fig. 6(a), where the calculated loop (solid line) is compared to experimental data. The determined constants, shown in Fig. 6(b), show an overall decrease in both anisotropy constants as a function of annealing temperature. Such a systematic decrease in the anisotropy constant was previously deduced from BLS measurements for

104412-5

SIMON TRUDEL et al.


(a)

(b)

FIG. 6. (Color online) (a) Experimental data ◦: Al-capped sample, Tanneal = 600 ◦ C, in-plane orientation α = 90 ◦ ) and best fit (solid line). (b) Determined cubic (KC ; •) and uniaxial (KU ; ) constants for the Al-capped series of samples, as a function of annealing temperature.

a series of Co2 MnSi samples annealed at various temperatures, where lower KC values were observed with increasing L21 ordering.38,39 The observed values are within the range of those reported for similar systems. For a compilation of values, see Ref. 4. The polar plots of the measured coercive fields Hc are shown in Fig. 7 for the Al-capped samples. The curves undergo a distinct change as the samples are annealed. However, the uniaxial symmetry in the magnetization reversal is clearly seen here for all samples. While in the case of Heusler samples grown onto GaAs,35–37 an anisotropy may be induced by the asymmetry of the substrate’s surface,40 no such anisotropy

is expected for cubic MgO(001). As such, the microscopic origin of this anisotropy and inequivalence between the Co2 FeAl0.5 Si0.5 [110] and [110] directions remains unclear. While the shape of these polar plots is similar for all the annealed samples, a systematic decrease in the coercive fields is observed as the annealing temperature is increased. As stated above (see Fig. 6), this is mostly due to a decrease in the anisotropy constants. The results obtained for the MgO-capped samples were comparable to those obtained for the Al-capped series shown in Fig. 7. The exception to this were the samples annealed at 480 ◦ C, where the shape was similar for the two samples, but the measured coercive fields for the Al-capped sample were on average roughly twice that of the MgO-capped film. It is worth noting the satellite peak intensity (Fig. 3) determined from transmission (i.e. bulk) XAS measurements are significantly different (in comparison to the other samples) at this annealing temperature. The lower degree of L21 ordering therefore correlates with the higher coercivities observed.

C. QMOKE magnetometry

Following the procedure described in Sec. II C, the QMOKE signals were obtained for each sample. Exemplary results for the QMOKE contributions as a function of the sample orientation α are given in Fig. 8 for the Al-capped sample annealed at 600 ◦ C. According to Eq. (2), the QMOKE sat signals sat ML MT and ML2 −MT2 are fit with a cosine and a sine function, respectively. It is obvious that the measured data fits the theoretical prediction very well. The free fit parameters were the amplitude of the sine for the sat signal and M 2 −M 2 L

(a)

(b)

(c)

(d)

(e)

(f)

T

FIG. 7. (Color online) Polar plots of the coercive field measured for (a) as-prepared sample and samples annealed at (b) 480 ◦ C, (c) 525 ◦ C, (d) 550 ◦ C, (e) 575 ◦ C, and (f) 600 ◦ C. 104412-6



(a)

FIG. 8. (Color online) QMOKE contributions obtained at normal incidence as a function of the sample orientation for an annealing temperature of 600 ◦ C . Lines are fit to Eq. (2).

(b)

FIG. 9. (Color online) Amplitude and offset of QMOKE components as a function of the annealing temperature (a) for the Al-capped sample series and (b) for the MgO-capped sample series. , ML MT offset; •, ML MT amplitude; ◦, ML2 − MT2 amplitude. Lines are guides for the eye. D. Brillouin light scattering spectroscopy

the amplitude of the cosine and the constant offset for the sat ML MT signal. The amplitude and the offset of the cosine and the amplitude of the sine function give information about the magnitude of the G factor. Thus the signals were obtained at normal incidence, where the linear contribution MLsat is small enough to be neglected. The values gathered from the fitting procedure allow one to give a trend in the QMOKE strength as a function of the annealing temperature for Al-capped samples [Fig. 9(a)]. For the as-prepared sample the values are rather small, which indicates that the QMOKE signal is not (or only weakly) present. For the sample annealed at 480 ◦ C the QMOKE signal increases sharply. The QMOKE amplitudes and offset then decrease for the sample annealed at 525 ◦ C, from which point they increase again to the maximum value for the sample annealed at 600 ◦ C. Since all three parameters follow this trend, it can be assumed that the G factor increases with the annealing temperature. This can be attributed to the improved crystallographic ordering of the samples. The values for the offset are smaller than those of the amplitudes of the cosine (sine) function. This can be understood since not only does the G factor contribute to the offset, but also, 2G44 is added and the linear magneto-optical constant K contributes with a negative sign. The QMOKE values increase with the annealing temperature for the MgO-capped series as well [Fig. 9(b)]. The absolute values are approximately halved compared to the values determined for the Al-capped series. However, a direct comparison cannot be made, as the light passes in the two-layer systems with a different optical path, resulting in different absolute Kerr rotations. Furthermore, the amplitudes of the two QMOKE contributions are different for the samples annealed above 500 ◦ C. This is not expected from Eq. (2).23 From this we conclude that the cubic model we applied to the system is no longer valid. We assume that the additional MgO layer disturbs the cubic ordering. This is in good agreement with the observations we have made with XAS, where the local atomic ordering at the interface is seen to be disturbed by the presence of the MgO layer.

BLS spectra were collected for all Al-capped samples at various fields and angles of light incidence. Representative spectra are shown in Fig. 10, where the applied field is 130 mT, and the angle of incidence is 45 ◦ . For both the Stokes (negative frequency shifts, associated with the creation of magnons) and the anti-Stokes (positive frequency shifts, associated with the annihilation of magnons), two peaks were observed. The peak at lower absolute frequencies is attributed to the Damon-Esbach mode (DE), while the peak at higher frequency is attributed to the first-order perpendicular standing spin-wave mode (PSSW). It can be seen that the peaks shift to higher frequencies as the samples are annealed to higher temperatures. The observed spin-wave frequencies were modeled as a function of incidence angle and applied magnetic field strength, using a model developed by Hillebrands41 based on a continuum-type magnetostatic theory. This procedure allows for the determination of the saturation magnetization Ms , the exchange constant A, and the related spin-wave stiffness D. As shown in Fig. 11(a), for the as-prepared sample a saturation magnetization of 5.33 ± 0.13 μB /fu was found. The saturation magnetization of half-metallic Heusler compounds is expected to follow a Slater-Pauling behavior, Ms = Nv − 24, μB

(4)

where Nv is the number of valence electrons per formula unit. For Co2 FeAl0.5 Si0.5 , Nv = 29.5, and as such, Ms = 5.5 μB /fu is expected, which is close to our determined value. As the samples were annealed at increasing temperatures, Ms is seen to increase, reaches a maximum value of 5.78 ± 0.14 μB /fu for samples annealed at 550 and 575 ◦ C, and decreases to 5.63 ± 0.14 μB /fu when annealed at 600 ◦ C. The trend of increasing saturation magnetization with annealing temperature reproduces the authors’ previous work, where similar Co2 FeAl0.5 Si0.5 thin films were studied by means of vibrating sample magnetometry (VSM).10 The values obtained by BLS spectroscopy presented here are slightly higher from the figures reported in Ref. 10. However,

104412-7

SIMON TRUDEL et al.


(f)

(e)

(a) (d) (c)

(b)

(b)

(a)

FIG. 10. (Color online) Measured BLS spectra for the asprepared sample (a) and for samples annealed in situ at (b) 480 ◦ C, (c) 525 ◦ C, (d) 550 ◦ C, (e) 575 ◦ C, and (f) 600 ◦ C. All spectra measured under an applied field μ0 H = 130 mT and angle of incidence ϕ = 45 ◦ . Only the anti-Stokes part of the spectrum is shown.

considering previously reported discrepancies between VSM- and BLS-derived magnetizations measured on the same samples (see Ref. 43), the experimental errors associated with both methods, and the fact that these consist of independent sample sets, the results are in fairly good agreement. The difference may be due to the probing depth of BLS spectroscopy, which will not measure the full thickness of the film. We have shown above that the top MgO (barrier)/Co2 FeAl0.5 Si0.5 interface presents some disorder, with an enhanced magnetic moment. Due to the limited depth sensitivity of BLS spectroscopy, the increased moments near the MgO barrier have a weighted predominance with respect to VSM measurements that probe the full sample, which may explain the higher moments observed by BLS. It might be expected that the lower MgO(buffer)/Co2 FeAl0.5 Si0.5 interface will also exhibit some form of disorder. However, this interface is buried within the sample stack and not available for independent analysis. In addition to modulating the saturation magnetization, annealing also impacted magnetic exchange. For example, the spin-wave stiffness D [shown in Fig. 11(b)] was 331 meV ˚ 2 for the as-prepared sample and increased to 493 meV A ˚ 2 for the sample annealed at 600 ◦ C. These values are in A the range expected for Heusler compounds.4,38,42–46 The spinwave stiffness D for the as-prepared sample is comparable to ˚ 2 ),43 while the values we reported for Co2 FeAl (370 meV A highest D value obtained here lies between values reported for ˚ 2 )45 and Co2 MnSi nonstoichiometric Co2 MnGe (413 meV A 2 46 ˚ ). The maximum value obtained here, 493 (580 meV A ˚ 2 , lies somewhat between the values for the parent meV A ˚ 2 ,42 ternary compounds Co2 FeAl and Co2 FeSi (715 meV A the record D for Co2 MZ compounds), which would be in

(c)

FIG. 11. (Color online) (a) Determined saturation magnetization per formula unit (fu), (b) spin-wave stiffness D, and (c) exchange constant A determined from fitting the spin-wave frequencies measured by BLS. Lines are guides for the eye.

agreement with the observation that D scales with Nv .4,47 The observed increase in exchange with increasing L21 ordering is also congruent with the behavior observed for Co2 MnSi by Gaier et al.4,38 A similar increase with increasing annealing temperature (and thus L21 ordering) was observed for the related exchange constant A [see Fig. 11(c)].

IV. CONCLUSION

In conclusion, we have studied two series of Co2 FeAl0.5 Si0.5 thin films with various degrees of L21 ordering. A comparison of the TEY and transmission XAS/XMCD data reveals different structural and magnetic properties for Co2 FeAl0.5 Si0.5 in the vicinity of an interface with an MgO or Al layer. It appears that the Co2 FeAl0.5 Si0.5 /MgO barrier interface can be improved and that such interface engineering would likely lead to better device performances. We find that a large QMOKE signal is indicative of L21 ordering. Surprisingly, there seems to be an antirelation to the magnetic anisotropy. Contrary to the expectation of a dependence on the spin-orbit coupling and orbital moment, we find that ordering is a far more important parameter. Good ordering also leads to smaller coercive fields, which is good for applications.

104412-8



ACKNOWLEDGMENTS

We thank Dr. Gaier for useful discussions during this work. We thank Dr. J. Lösch (IFOS, TU Kaiserslautern) for x-ray measurements and S. Cramm for support at BESSY. This project was financially supported by DFG Research Unit 559, “New Materials with High Spin Polarization” (Germany), BMBF (Grant No. ES3XBA/5), and CREST-JST. S.T. gratefully acknowledges the Alexander von Humboldt foundation for a postdoctoral research fellowship. APPENDIX: MAGNETIZATION REVERSAL CALCULATION

We use an extended Stoner Wohlfarth model22 to calculate the magnetization reversal process. The model is based on the minimization of the free enthalpy density g as a function of the magnetization orientation αM with respect to the externally applied field H . The Zeeman energy density and the anisotropy energy density contribute to the free enthalpy density: g = −MS H cos αM + KU cos2 (αM − α) + KC cos2 (αM − α) sin2 (αM − α) .

(A1)

The anisotropy energy density consists of a uniaxial and a biaxial contribution. The angle α is the direction of the crystallographic axis with respect to the externally applied magnetic field. The algorithm differentiates ∂g/∂αM and identifies the minima and maxima of g. To model a hysteretic magnetization reversal process properly, we use a perfect delay convention.48 This means that the direction of the magnetization not only is given by the global minimum of g, but also can be given by a local minimum, depending on the minimum of the previous field step. The magnetization can remain in the local minimum until it is unstable and then moves to the next lower minimum. With this model we can calculate the magnetization reversal process as a function of the orientation of the crystallographic axis. The calculated coercive field is an upper limit for a given anisotropy. A similar model has been used elsewhere by Mewes et al.49–51 1. Thermal activation

We have also considered thermal activation to improve the model. If the magnetization is in a local minimum, it can overcome the barrier (value of the difference between

*

[email protected]; Present address: Department of Chemistry, University of Calgary, 2500 University Drive NW, Calgary, Alberta T2N 1N4, Canada. † Present address: Department of Physics, VSB—Technical University of Ostrava, 17 Lisopadu 15, CZ-708 33 Ostrava-Poruba, Czech Republic. ‡ [email protected] 1 C. Felser, G. H. Fecher, and B. Balke, Angew. Chem. Int. Ed. Engl. 46, 668 (2007). 2 K. Inomata, N. Ikeda, N. Tezuka, R. Goto, S. Sugimoto, M. Wojcik, and E. Jedryka, Sci. Technol. Adv. Mater. 9, 014101 (2008).

local minimum and neighboring maximum) and move to the next lower minimum as a result of the thermal energy in the system.52 The probability for this process is determined by the Boltzmann factor p = exp[ kEBbarrier ] (here V is the thermally T /V active volume). The ratio of the enthalpy density barrier to the thermal energy density is crucial for this process. For barrier heights larger than the thermal energy density, the magnetization remains in the local minimum. For barrier heights smaller than the thermal energy density, the magnetization moves to the next minimum. Reducing this process to a “do” or “do not” scenario is a very strong approximation;, see discussion below. Additionally this process is time dependent.52 2. Fitting experimental data

We calculate magnetization reversal curves for a set of values of the anisotropy constants (both KU and KC ). We calculate a fit quality parameter q, defined as

2 1 expt q= (A2) Mi − Micalc , N i expt

where Mi is the normalized experimental Kerr signal for a given field value i, Micalc is the calculated magnetization orientation for the field value i, and N is the number of data points. The combination of the anisotropy constants KU and KC of the smallest fit quality parameter are considered to be the best-fit values. The saturation magnetization from the literature is used. 3. Limitations

The model has certain limits. First, there is the limited description of the thermal activation; see above. Here we emphasize that the activation volume plays an important role, as it determines the thermal energy density. The value of V influences the absolute values of the best-fit anisotropy constants. Furthermore, the model does not consider the formation or movement of magnetic domains, which may in some cases be a driving mechanism in the magnetization reversal process. The formation of magnetic domains is strongly connected to the thermal activation, as parts of the magnetic volume change their orientation while others do not, which will lead to an averaged signal in the experiment. In the calculation it is treated as one single domain.

3

J. Kübler, Theory of Itinerant Electron Magnetism, rev. ed. (Oxford University Press, New York, 2009). 4 S. Trudel, O. Gaier, J. Hamrle, and B. Hillebrands, J. Phys. D Appl. Phys. 43, 193001 (2010). 5 P. Klaer, M. Kallmayer, C. G. F. Blum, T. Graf, J. Barth, B. Balke, G. H. Fecher, C. Felser, and H. J. Elmers, Phys. Rev. B 80, 144405 (2009). 6 P. Klaer, E. Arbelo Jorge, M. Jourdan, W. H. Wang, H. Sukegawa, K. Inomata, and H. J. Elmers, Phys. Rev. B 82, 024418 (2010). 7 B. Balke, G. H. Fecher, and C. Felser, Appl. Phys. Lett. 90, 242503 (2007).

104412-9

SIMON TRUDEL et al. 8


G. H. Fecher and C. Felser, J. Phys. D Appl. Phys. 40, 1582 (2007). 9 T. M. Nakatani, A. Rajanikanth, Z. Gercsi, Y. K. Takahashi, K. Inomata, and K. Hono, J. Appl. Phys. 102, 033916 (2007). 10 W. Wang, H. Sukegawa, R. Shan, T. Furubayashi, and K. Inomata, Appl. Phys. Lett. 92, 221912 (2008). 11 W. Wang, H. Sukegawa, R. Shan, and K. Inomata, Appl. Phys. Lett. 93, 122506 (2008). 12 W. Wang, H. Sukegawa, R. Shan, and K. Inomata, Appl. Phys. Lett. 93, 182504 (2008). 13 R. Shan, H. Sukegawa, W. H. Wang, M. Kodzuka, T. Furubayashi, T. Ohkubo, S. Mitani, K. Inomata, and K. Hono, Phys. Rev. Lett. 102, 246601 (2009). 14 H. Sukegawa, S. Kasai, T. Furubayashi, S. Mitani, and K. Inomata, Appl. Phys. Lett. 96, 042508 (2010). 15 N. Tezuka, N. Ikeda, A. Miyazaki, S. Sugimoto, M. Kikuchi, and K. Inomata, Appl. Phys. Lett. 89, 112514 (2006). 16 K. Inomata, M. Wojcik, E. Jedryka, N. Ikeda, and N. Tezuka, Phys. Rev. B 77, 214425 (2008). 17 N. Tezuka, N. Ikeda, S. Sugimoto, and K. Inomata, Appl. Phys. Lett. 89, 252508 (2006). 18 H. Sukegawa, W. Wang, R. Shan, T. Nakatani, K. Inomata, and K. Hono, Phys. Rev. B 79, 184418 (2009). 19 H. C. Kandpal, G. H. Fecher, and C. Felser, J. Phys. D Appl. Phys. 40, 1507 (2007). 20 M. Kallmayer, H. Schneider, G. Jakob, H. Elmers, K. Kroth, H. Kandpal, U. Stummn, and C. Cramm, Appl. Phys. Lett. 88, 072506 (2006). 21 R. Nakajima, J. Stöhr, and Y. U. Idzerda, Phys. Rev. B 59, 6421 (1999). 22 E. C. Stoner and E. P. Wohlfarth, Phil. Trans. Roy. Soc. Lond. Ser. A 240, 599 (1948). 23 ˇ Viˇsnovský, and K. Postava, D. Hrabovský, J. Piˇstora, A. R. Fert, S. T. Yamaguchi, J. Appl. Phys. 91, 7293 (2002). 24 At normal incidence, A = (n˜ 2o ϕ)/[n˜ 2 (n˜ 2o − n˜ 2 )] and B = ˜ n˜ 2o − n˜ 2 )], where ϕ is the angle of incidence and n˜ and n˜ o n˜ o /[n( are the complex refractive indexes of the material and the incident medium (air), respectively. 25 S. Trudel, G. Wolf, H. Schultheiß, J. Hamrle, B. Hillebrands, T. Kubota, and Y. Ando, Rev. Sci. Instrum. 81, 026105 (2010). 26 B. Hillebrands, Rev. Sci. Instrum. 70, 1589 (1999). 27 O. Gaier, J. Hamrle, B. Hillebrands, M. Kallmayer, P. Pörsch, G. Schönhense, H. J. Elmers, J. Fassbender, A. Gloskovskii, C. A. Jenkins, C. Felser, E. Ikenaga, Y. Sakuraba, S. Tsunegi, M. Oogane, and Y. Ando, Appl. Phys. Lett. 88, 152508 (2009). 28 N. D. Telling, P. S. Keatley, G. van der Laan, R. J. Hicken, E. Arenholz, Y. Sakuraba, M. Oogane, Y. Ando, and T. Miyazaki, Phys. Rev. B 74, 224439 (2006). 29 H. J. Elmers, S. Wurmehl, G. H. Fecher, G. Jakob, C. Felser, and G. Schönhense, J. Magn. Magn. Mater. 272–276, 758 (2003). 30 M. Jourdan, E. Arbelo Jorge, C. Herbort, M. Kallmayer, P. Klaer, and H. Elmers, Appl. Phys. Lett. 95, 172504 (2009).

31

M. Kallmayer, P. Klaer, H. Schneider, E. Arbelo Jorge, C. Herbort, G. Jakob, M. Jourdan, and H. J. Elmers, Phys. Rev. B 80, 020406(R) (2009). 32 E. Arbelo Jorge, M. Jourdan, M. Kallmayer, P. Klaer, and H. J. Elmers, J. Phys. Conf. Ser. 200, 072006 (2010). 33 V. Ksenofontov, C. Herbort, M. Jourdan, and C. Felser, Appl. Phys. Lett. 92, 262501 (2008). 34 D. Weller, J. Stöhr, R. Nakajima, A. Carl, M. G. Samant, C. Chappert, R. Megy, P. Beauvillain, P. Veillet, and G. A. Held, Phys. Rev. Lett. 75, 3752 (1995). 35 F. Y. Yang, C. H. Shang, C. L. Chien, T. Ambrose, J. J. Krebs, G. A. Prinz, V. I. Nikitenko, V. S. Gornakov, A. J. Shapiro, and R. D. Shull, Phys. Rev. B 65, 174410 (2002). 36 T. Ambrose, J. J. Krebs, and G. A. Prinz, Appl. Phys. Lett. 76, 3280 (2000). 37 T. Ambrose, J. J. Krebs, and G. A. Prinz, J. Appl. Phys. 87, 5463 (2000). 38 O. Gaier, J. Hamrle, S. J. Hermsdoerfer, H. Schultheiß, B. Hillebrands, Y. Sakuraba, M. Oogane, and Y. Ando, J. Appl. Phys. 103, 103910 (2008). 39 J. Hamrle, O. Gaier, S. Trudel, G. Wolf, and B. Hillebrands, in Spintronics: New Materials with High Spin Polarization, edited by G. Fecher and C. Felser (Springer-Verlag, Heidelberg, 2011), (in press). 40 J. J. Krebs, B. T. Jonker, and G. A. Prinz, J. Appl. Phys. 61, 2596 (1987). 41 B. Hillebrands, Phys. Rev. B 41, 530 (1990). 42 O. Gaier, J. Hamrle, S. Trudel, B. Hillebrands, H. Schneider, and G. Jakob, J. Phys. D Appl. Phys. 42, 232001 (2009). 43 O. Gaier, J. Hamrle, S. Trudel, A. Conca Parra, B. Hillebrands, E. Arbelo, C. Herbort, and M. Jourdan, J. Phys. D Appl. Phys. 42, 084004 (2009). 44 M. Belmeguenai, F. Zighem, Y. Roussigné, S.-M. Chérif, P. Moch, K. Westerholt, G. Woltersdorf, and G. Bayreuther, Phys. Rev. B 79, 024419 (2009). 45 S. Trudel, J. Hamrle, B. Hillebrands, T. Taira, and M. Yamamoto, J. Appl. Phys. 107, 043912 (2010). 46 J. Hamrle, O. Gaier, S.-G. Min, B. Hillebrands, Y. Sakuraba, and Y. Ando, J. Phys. D Appl. Phys. 42, 084005 (2009). 47 T. Kubota, J. Hamrle, Y. Sakuraba, O. Gaier, M. Oogane, A. Sakuma, B. Hillebrands, K. Takanashi, and Y. Ando, J. Appl. Phys. 106, 113907 (2009). 48 S. Nieber and H. Kronmueller, Phys. Status Solidi B 165, 503 (1991). 49 T. Mewes, H. Nembach, M. Rickart, S. O. Demokritov, J. Fassbender, and B. Hillebrands, Phys. Rev. B 65, 224423 (2002). 50 T. Mewes, H. Nembach, J. Fassbender, B. Hillebrands, J. V. Kim, and R. L. Stamps, Phys. Rev. B 67, 104422 (2003). 51 T. Mewes, Ph.D. thesis, Technische Universität Kaiserslautern, Germany, 2002. 52 R. Street and J. C. Woolley, Proc. Phys. Soc. London 62, 562 (1949).

104412-10