Local structure and magnetization of ferromagnetic Cu-doped ZnO films

Journal of Alloys and Compounds 678 (2016) 304e311

Contents lists available at ScienceDirect

Journal of Alloys and Compounds journal homepage: http://www.elsevier.com/locate/jalcom

Local structure and magnetization of ferromagnetic Cu-doped ZnO films: No magnetism at the dopant? c P.S. Vachhani a, b, *, O. Sipr , A.K. Bhatnagar a, d, R.K. Ramamoorthy b, d, 1, R.J. Choudhary e, D.M. Phase e, G. Dalba b, A. Kuzmin f, F. Rocca g a

School of Physics, University of Hyderabad, Hyderabad 500 046, India Department of Physics, University of Trento, 38123 Povo, Trento, Italy 10, Prague, Czech Republic Institute of Physics AS CR v. v. i., Cukrovarnicka d School of Engineering Sciences & Technology, University of Hyderabad, Hyderabad 500 046, India e UGC-DAE Consortium for Scientific Research, Khandwa Road, Indore 452 017, India f Institute of Solid State Physics, University of Latvia, LV-1063 Riga, Latvia g IFN-CNR, Institute for Photonics and Nanotechnologies, Unit “FBK-Photonics” of Trento, 38123 Povo, Trento, Italy b c

a r t i c l e i n f o

a b s t r a c t

Article history: Received 24 November 2015 Received in revised form 22 March 2016 Accepted 1 April 2016 Available online 4 April 2016

Relationship between magnetism and structure of Cu-doped ZnO was investigated at macroscopic and microscopic levels. Thin Zn1xCuxO films (x ¼ 0.02, 0.04, 0.07 and 0.10) were prepared by a pulsed laser deposition and characterized via superconducting quantum interference device (SQUID) magnetometry, high-resolution x-ray diffraction, and Cu K-edge and Zn K-edge x-ray absorption, x-ray linear dichroism and x-ray circular magnetic dichroism spectroscopy. Even though the samples exhibit room-temperature ferromagnetism with magnetization that increases with Cu concentration, we did not detect signatures of local magnetic moments associated with Cu atoms, as evidenced by the lack of any XMCD signal. The host ZnO wurtzite lattice is not significantly altered by the addition of Cu. At the same time, most of the Cu atoms are not incorporated into the wurtzite lattice but rather have a CuO-like coordination. These results indicate that ferromagnetism of the investigated Zn1xCuxO films is not directly linked to the doping atoms but rather is due to some other changes which have been introduced to the host ZnO by the dopants. © 2016 Elsevier B.V. All rights reserved.

Keywords: EXAFS NEXAFS Magnetization Crystal structure

1. Introduction The focus of current materials science is on versatile materials which exhibit more interesting properties at the same time. Among these materials, there are the magnetic semiconductors, where the main current issue is to design materials which would be ferromagnetic not only at low temperature but at room temperature as well. In this respect, doped ZnO attracted a lot of attention, with ample experimental evidences of room-temperature ferromagnetism (RTFM) in ZnO doped with transition metal (TM) ions such as Ti, V, Mn, Fe, Co, Ni, or Cu [1e3]. Despite the enormous effort that has been devoted to research on doped ZnO, the issue remains

* Corresponding author. Current address: D.K.V. Arts and Science College, P.N. Marg, Jamnagar 361008, India. E-mail address: [email protected] (P.S. Vachhani). 1 Current address: LIONS, CEA Saclay, IRAMIS/UMR CEA CNRS 3299 NIMBE, 91191 Gif-sur-Yvette, France. http://dx.doi.org/10.1016/j.jallcom.2016.04.002 0925-8388/© 2016 Elsevier B.V. All rights reserved.

controversial both as concerns the mechanism of the RTFM phenomenon and as concerns the phenomenon itself or, more precisely, the conditions under which it arises. This is at least partially due to the fact that the magnetic behavior of doped ZnO is strongly linked to how the material was prepared. More reliable data on well-defined systems characterizing as many aspects as possible are therefore needed. One of the major questions is whether the observed RTFM is an intrinsic property of the TM-doped ZnO or whether it is due to extrinsic causes such as nanoclustering of the dopant [4e7]. Therefore, when studying the intrinsic RTFM, it is convenient to turn to dopants which are non-magnetic in elemental form, to minimize the possibility that the magnetism is due to dopant precipitates. In this respect Cu-doped ZnO (Cu:ZnO) is an especially attractive material and indeed a lot of studies were devoted to it. However, the results are quite diverse, e.g. controversial conclusions were reached on whether the RTFM can be observed only in p-type Cu:ZnO or whether ferromagnetic Cu:ZnO can be also of ntype [8e12].

P.S. Vachhani et al. / Journal of Alloys and Compounds 678 (2016) 304e311

The large spread of Zn1XCuXO properties reported in the literature suggests that there may be in fact several different kinds of this material. To distinguish between them, more information on the structure is needed, both at the long-range order (to understand how the ZnO host is affected by dopants) and at the shortrange order (to learn about the local geometry around the doping ions), which can be probed by x-ray diffraction and x-ray absorption spectroscopy, respectively. The latter technique has an additional asset that it can also probe element-specific magnetic properties via x-ray magnetic circular dichroism (XMCD). This is especially attractive because then one can investigate the structure and magnetism simultaneously as a pair. X-ray absorption spectra probe transitions of core electrons to unoccupied states. By a proper choice of the absorption edge, one can further specify that these transitions occur only to states with a particular symmetry with respect to atoms of given chemical type. The choice of the absorption edge depends, among others, on whether one is interested in studying the geometrical structure or the electronic structure. In particular, the K edge spectra of transition metals (TM's) involve delocalized p states which make them sensitive to the geometry around photoabsorbing atom. They cover a wide energy range, so one can measure the extended x-ray absorption fine structure (EXAFS), the analysis of which has become a powerful technique for structure determination. The L2,3 edge spectra involve more localized d states and cover only a short energy range (for 3d TM's), so they are not very suitable for structural studies. On the other hand, the L2,3 edge-XMCD has been conveniently used to study magnetism because the magnetization of TM atoms is carried mostly by the d electrons. Nevertheless, magnetism can be studied also via the K-edge XMCD because it is linked to a small but still detectable p component of the orbital magnetic moment [13]. So the K-edge xray absorption spectroscopy can be used as a tool for studying both the structure and magnetism simultaneously. Let us recall that the local view on magnetization via XMCD may indeed bring different information than global magnetization measurement based on superconducting quantum interference device (SQUID) magnetometry, as it was demonstrated for Co:ZnO as well as for Cu:ZnO [14,15]. Earlier studies of Cu:ZnO via the K-edge x-ray absorption near edge structure (XANES) or EXAFS focused on the structure, without carrying out XMCD measurements [16e18]. On the other hand, when local magnetic properties were studied via L2,3 edge XMCD, it was not possible to do a full structural analysis simultaneously because the L2,3 edge x-ray absorption spectra allow only limited structural analysis via “XANES fingerprinting” [15,19e21]. In a previous work, we have presented a combined study of local structure and local magnetism via the Cu K-edge XANES, EXAFS and XMCD spectroscopy for the Cu:ZnO pellets that were used as a target for the pulsed laser deposition (PLD) of the films investigated in the present paper [22]. Since the pellets were characterized as paramagnetic, it would be very interesting to perform such a study also for the ferromagnetic Zn1xCuxO films, so that links between magnetization and structure can be drawn. Therefore, we undertook a study of a series of ferromagnetic Cu:ZnO films with varying Cu concentrations, characterizing the magnetic properties via SQUID and the Cu K-edge XMCD, and the structural properties via x-ray diffraction and the Zn and Cu K-edge XANES, x-ray linear dichroism (XLD), and EXAFS. We will demonstrate in the following that most of the Cu atoms are not incorporated in the wurtzite ZnO lattice but rather reside in a CuO-like environment while the host ZnO lattice is left mostly unperturbed. Even though the samples exhibit RTFM with magnetization that increases with Cu concentration, no detectable Cu K-edge XMCD signal can be identified. This implies that the measured

305

ferromagnetism of Zn1xCuxO films is not directly associated with the magnetic properties of the doping atoms. 2. Methods 2.1. Sample preparation Oriented Cu:ZnO films were prepared by employing the PLD technique to transfer the target material onto the substrate. First, powdered Zn1xCuxO (x ¼ 0.02, 0.04, 0.07, 0.10) was synthesized by mixing ZnO and CuO powders (purity 99.99%, Sigma Aldrich) in stoichiometric proportions and calcinating the mixture at 400 C for 12 h. Polycrystalline pellets were then obtained by grinding, pelletizing and sintering the powders at 1000 C for 12 h. The properties of these pellets were investigated in an earlier study [22]. We only recall here that these pellets are paramagnetic, and that there was a clear Cu K-edge XMCD signal proving that the Cu atoms are magnetic in the Zn1xCuxO pellets. By analyzing the XAS data, we found that at low Cu content most of the Cu atoms substitute for Zn inside the ZnO wurtzite lattice, while for higher Cu concentrations some unreacted CuO remains segregated from the Zn1xCuxO solid solution. To get oriented Zn1xCuxO films, the above mentioned pellets were used as targets for PLD to deposit the material onto the cplane of a sapphire substrate. Prior to deposition, the substrate was cleaned ultrasonically using trichloroethylene, acetone, methanol and distilled water. A pulsed KrF excimer laser with a wavelength of 248 nm was used for the PLD, with a repetition rate of 10 Hz. The laser beam had an energy density at the target of ~2 J/cm2. The deposition on substrates maintained at the temperature of ~500 C for 30 min was done under oxygen partial pressure of 2 mTorr, with a nominal deposition rate of 10 nm/min (±10%). Similar preparation technique was used in earlier study [23]. 2.2. Experimental procedures The magnetization of cleaned substrate as well as films was measured using a SQUID magnetometer (Quantum Design) in the temperature range of 10e300 K. Figs. 1 and 2 contain only the signal from the films, after correction. In addition, it should be noted here that similar ZnO films doped with 2% Co or Ti were also prepared with the same experimental parameters, but they were found to be non-magnetic. A high resolution x-ray diffraction (HR-XRD) characterization of the samples was performed at room temperature using a Bruker D8 advanced diffractometer, equipped with copper anode x-ray tube (Cu Ka radiation). The x-ray absorption experiments were performed at the ESRF

Fig. 1. Magnetization curves for the Zn1xCuxO films measured at 300 K.

306


3.2. High resolution x-ray diffraction (HR-XRD)

Fig. 2. Expanded view of the magnetization curves for the Zn1xCuxO film for x ¼ 0.02 (left panel) and x ¼ 0.07 (right panel) measured at 10 K and at 300 K.

(Grenoble). The XANES and EXAFS spectra were measured at the BM08-GILDA beamline, at 77 K, using the available 13-elements Ge multidetector. The XLD and XMCD spectra were measured at the ID 12 beamline, using multiple and single Si diode detectors, respectively. XLD measurements were done at 300 K, XMCD at 7 K. When performing the XANES and EXAFS measurements, the angle between the incident x-rays and the films, mounted vertically, was 45 . When performing the XLD and XMCD measurements, the angle between the incoming x-ray beam and the surface of the films mounted horizontally was 15 . The XLD was obtained as a difference between the spectra obtained for two perpendicular polarizations of the incoming radiation; a linearly polarized light (either vertically or horizontally) was obtained from a circularly polarized synchrotron radiation via a quarter wave plate. The polarization vector was thus either parallel (with a 15 deviation) or perpendicular to the c-axis of the Cu:ZnO films. The XMCD was obtained as a different between spectra recorded with right and left circularly polarized light, with the external magnetic field of 3 T being oriented parallel to the incoming x-rays. 3. Results 3.1. Magnetic characterization The magnetization versus magnetic field (M H) curves for the Zn1xCuxO films with x ¼ 0.02, 0.04, 0.07 and 0.10 measured at 300 K are shown in Fig. 1. One can see that all our Zn1XCuXO films are ferromagnetic at room temperature, with the maximum magnetization increasing linearly with increasing Cu concentration. Average values of 0.50 mb/Cu, for the maximum magnetization per Cu atoms were calculated using the nominal Cu content and thickness for the four films: within the estimated error bar (±0.08 mb), they can be considered identical. An expanded view of the magnetization curves for the x ¼ 0.02 and x ¼ 0.07 films measured at 10 K and 300 K is shown in Fig. 2, where we scaled the curves in mb/Cu units. The magnetization measurements at 10 K were performed after cooling down without applied field. Clear hysteresis can be seen in this detailed view. Even though full saturation has not been attained for the highest magnetic field we applied (3000 Oe), the ferromagnetic order is wellevidenced by the presence of hysteresis. The fact that the magnetization of our films scales nearly linearly with x indicates that the observed ferromagnetism is indeed connected to the Cu-doping. The coercive field at 300 K is 41, 16, 21 and 41 Oe for the x ¼ 0.02, 0.04, 0.07 and 0.10 films, respectively. These values are similar to coercive fields found in Zn1xCuxO samples investigated in most of the earlier x-ray absorption spectroscopy studies [15,18,19] but significantly less than the value of 260 Oe reported in the study of Herng et al. [20].

The structural characterization of the films at the long range was carried out by the HR-XRD. The corresponding diffractograms in logarithmic vertical scale are shown in Fig. 3. For films with low Cu concentration (x ¼ 0.02 and 0.04), only peaks belonging to the (0001) planes of wurtzite structure were observed (apart from the peaks belonging to the substrate). For higher Cu concentrations (x ¼ 0.07 and 0.10), however, additional peaks characteristic for CuO appear, albeit with a very small intensity. The q-2q rocking curves of the wurtzite (002) diffraction peaks were also measured: their full-widths at half-maximum (FWHM) are between 0.06 and 0.35 , which suggests a very good crystalline quality of all the films. As a whole, the data indicate that we have highly textured films with a preferred orientation along the c axis of wurtzite lattice. For higher Cu concentration, a small admixture of a secondary phase of monoclinic CuO appears. The emergence of CuO, nevertheless, does not cause significant damage to the parental wurtzite lattice. 3.3. Zn K-edge x-ray absorption spectra Weighted Zn K-edge EXAFS signals k2c(k) for Zn1xCuxO films and for a ZnO reference are presented in Fig. 4 (left panel). The Fourier transforms (FT's) of these signals taken within the range from 0 to 12 Å1 are shown in Fig. 4 (right panel). One can see a close similarity between the signals for Zn1xCuxO films and for the pure ZnO. This shows that the local geometry around Zn atoms is close in all these systems. A quantitative analysis of the EXAFS signals did not reveal significant differences between our Zn1xCuxO films and ZnO either. Another view on the local geometry around Zn atoms can be obtained from polarized spectra, in particular from near edge XLD. The Zn K-edge XANES spectra measured with the polarization vector E of the incoming x-rays perpendicular and parallel to the caxis of the Zn1xCuxO films are shown in Fig. 5, together with the corresponding XLD signal defined as the difference between the spectra for E ⊥ c and E k c. The XANES spectra measured with E ⊥ c do not show much variations when the concentration x is varied and resemble analogous spectra of pure ZnO [24]. On the other hand, spectra measured with E k c exhibit clear variations with Cu concentration, mostly concerning the peak around 9680 eV indicated by an arrow in Fig. 5. The spectra for x ¼ 0.07 and 0.10 concentrations resemble

Fig. 3. HR-XRD data for the Zn1xCuxO films, together with data for the sapphire substrate. Peaks belonging to the (0001) planes of the wurtzite structure are labeled by their respective indices. Peaks belonging to the substrate are marked by “S”. Peaks which are characteristic for the CuO are marked by an asterisk. The scale on the vertical axis is logarithmic.


Fig. 4. Experimental weighted EXAFS signal k2c(k) (left panel) and their Fourier transforms (right panel) at the Zn K-edge for Zn1xCuxO films and for pure ZnO, measured at 77 K. The curves are vertically displaced for easier comparison.

the spectrum of pure ZnO more than the spectra for the other two concentrations. This difference can be also quantified: maximum XLD signal with respect to the edge jump is 30%, 30%, 60% and 50% for x ¼ 0.02, 0.04, 0.07 and 0.10, respectively. This indicates that the films with higher Cu concentration are better ordered and more textured. 3.4. Cu K-edge x-ray absorption spectra The Cu K-edge XANES spectra of Zn1xCuxO films and of CuO are shown in Fig. 6 (upper panel). It is evident that apart from the sample with x ¼ 0.02, the spectra of the Zn1xCuxO films with x 0.04 resemble quite closely the spectrum of CuO. The spectrum for the x ¼ 0.02 film is quite different. When analyzing the K-edge XANES of TM's, the pre-peak is often used as a marker of a tetrahedral or distorted octahedral coordination of the photoabsorbing atom [25]. For tetrahedral coordination the pre-peak corresponds to dipole-allowed transitions

307

to states formed by hybridization of the 3d states of the TM photoabsorber with the ligand states [25]. The pre-peak at E ¼ 8977 eV in the Cu K-edge XANES of CuO is, however, due to quadrupole transitions [26]. If Cu is in a tetrahedral position (as it would be the case of a Zn-substitutional site in w-ZnO), the pre-peak appears at a slightly different energy e it is shifted by about 1.4 eV to the lower energy, as shown in our previous paper [22]. To allow a better comparison between different curves, we show the first derivatives of the spectra in the pre-peak range in Fig. 6 (bottom panel). Our data show that, for the x ¼ 0.07 and x ¼ 0.10 samples, the pre-peak is at the same energy as for CuO. So regarding also the agreement between the overall shapes of the spectra, we can deduce that Cu is locally in a CuO-like environment in these films. The situation for the x ¼ 0.04 sample differs a bit: there appears to be also a second component of the pre-peak corresponding to Cu in the Zn-substitutional site and some small deviations from the CuO XANES appear also around 9000 eV. Nevertheless, the overall shape of the spectra still resembles the spectrum of CuO closely so it is likely that most of Cu atoms are again in CuO-like environment and a small fraction of Cu atoms is in Zn-substitutional sites. The spectrum of the x ¼ 0.02 film clearly differs from the spectrum of CuO. It bears some resemblance to the spectra of Zn1xCuxO pellets which we have measured earlier and which correspond to Cu in Zn-substitutional sites [22]. A further structural analysis can be done by means of EXAFS, measured at the BM 08 beamline. The k2-weighted EXAFS c(k) signals are displayed, together with their FT's, in Fig. 7. The Zn Kedge EXAFS of a ZnO reference is included as well for comparison. One can see at a first sight that there is a close similarity between the data for the Zn1XCuXO films for x ¼ 0.04, 0.07 and 0.10 and the data for CuO. However, the data for the film with x ¼ 0.02 are clearly different. For better understanding, let us recall that in wurtzite ZnO, a Zn atom is tetrahedrally coordinated with four O atoms at distances 1.95 Å and 1.98 Å. In monoclinic CuO, a Cu atom is planarly coordinated also with four O atoms, now at distances 1.95 Å and 1.96 Å. So, the first shell in the FT of the ZnO and CuO signals should be

Fig. 5. XANES spectra at the Zn K-edge for Zn1xCuxO films for two orientations of the polarization vector of the incoming light measured at ID12, together with corresponding XLD signals (blue line). Red filled squares correspond to the polarization vector E parallel to the film surface, black empty circles correspond to E perpendicular to the film. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

308


Fig. 6. The Cu K-edge XANES spectra for Zn1xCuxO films (bottom panel) and their first derivatives (top panel), together with corresponding data for CuO. The curves are vertically displaced for easier comparison.

similar, as it is in Fig. 7 too. The differences appear in further coordination shells: the second coordination shell of Zn in ZnO is formed by 12 Zn atoms (six at 3.21 Å and six at 3.25 Å) and is represented by a peak in the FT around 2.9 Å. For Cu in CuO, on the other hand, there are two O atoms at 2.78 Å followed by a series of Cu shells (four atoms at 2.90 Å, four atoms at 3.08 Å and two atoms at 3.17 Å) which are represented in the FT as a complex structure between 2.5 Å and 3.5 Å (see Fig. 7 right panel). Regarding this second coordination shell, we can see that for the films with a high Cu concentration (x ¼ 0.07 and x ¼ 0.10), the FT's of the signals for Zn1xCuxO and for CuO are very similar. We thus conclude that practically all Cu atoms are in CuO-like environment in these films. For the x ¼ 0.04 film, some small differences with respect to CuO appear in the second FT peak e the middle of the three-peak structure gets more pronounced, which can be seen as arising from the second FT peak of ZnO signal. This is consistent with the picture that the majority of Cu atoms is in CuO-like environment but some small portion is in Zn substitutional sites, as suggested already by the XANES analysis. For the x ¼ 0.02, the EXAFS and its FT significantly differ from the CuO signal. Rather, they are more similar to the ZnO signal, even though there are clear differences here as well. In particular, the second FT peak differs from both CuO and ZnO references and it is likely that there will be multiple Cu sites, probably with high static disorder. For a quantitative analysis, we best-fitted the first FT peak to get

Fig. 7. Weighted EXAFS signals k2c(k) for the Cu K-edge of Zn1xCuxO films and of CuO and for the Zn K-edge of ZnO measured at 77 K (left panel) together with their Fourier transforms (right panel).

the CueO distances R, the coordination numbers N and the relative mean square displacements (MSRD) s2 for the first coordination shell. The fitting was performed using the EDAFIT software [27]. The phase shifts and back scattering amplitudes were obtained theoretically using the FEFF8.0 code, assuming Cu in Zn substitutional site of ZnO [28]. A comparison between the original and the bestfitted signal is shown in Fig. 8, the R, N and s2 quantities resulting from the best-fitting procedure are presented in Table 1. Within our estimated error bar, coordination numbers and nearestneighbor distances are close for all the Zn1xCuxO films with x 0.04 and they are also close to the respective values for CuO and for ZnO. On the other hand, for the x ¼ 0.02 film N is significantly lower and s2 is significantly larger than for the other systems. This indicates a larger disorder, as it would be the case for a multiplicity of Cu sites. We also measured Cu K-edge XLD at the ID 12 beamline for all the films. If some part of Cu atoms is in substitutional sites (as suggested at least for the x ¼ 0.02 film), there should be a non-zero XLD, similar to that measured for the Zn K-edge. However, the intensity of the signal was very low for the x ¼ 0.02 and x ¼ 0.04 films and we were not able to identify any clear XLD signal for them. We assume that a detectable XLD is missing for the x ¼ 0.02 and x ¼ 0.04 samples because of a poor statistics or signal-to-noise ratio. For the remaining two films with a larger Cu concentration we detected some small XLD signals, presented in Fig. 9. These signals are quite noisy, they have different shapes for different Cu concentrations and their maximum intensities are about 5% of the edge jump. They are any way different from what can be expected for Cu in Zn-substitutional sites. The absence of a clear trend in the measured XLD signals may look contradictory to the assumption that Cu atoms are in CuO-like environment because earlier calculations showed quite a large XLD signal for CuO, reaching up to about 50% of the edge jump [26,29]. However, one has to consider that the wurtzite ZnO and monoclinic CuO lattices are incommensurate, and it is thus very probable that each of the CuO-like regions formed in the Zn1XCuXO films will have a different orientation with respect to the ZnO host. Consequently, the XLD signal will be strongly suppressed, similarly as in polycrystals. A perfect orientational averaging would lead to identically zero XLD. The fact that we observe some very faint “residual” XLD signals indicates that there is some small preferential orientation of CuO regions in Zn1xCuxO films but this preference is very weak and different for different samples. 3.5. Cu K-edge XMCD spectra To explore the intrinsic magnetic nature of the Zn1xCuxO films,

Fig. 8. First shell Cu K-edge EXAFS k2c(k) functions of Zn1xCuxO films (dots) and their simulations (lines) based on best-fitting of the signal in the range 3e10.5 Å1.

P.S. Vachhani et al. / Journal of Alloys and Compounds 678 (2016) 304e311 Table 1 Structural parameters [coordination number N, interatomic distance R(CueO) and mean-squared relative displacement s2(CueO)] for the first coordination shell of copper in Zn1xCuxO (x ¼ 0.02, 0.04, 0.07 and 0.10), obtained from the best fit analysis of the Cu K-edge EXAFS spectra. Sample

N

R [Å]

s2 [Å2]

Zn0.98Cu0.02O Zn0.96Cu0.04O Zn0.93Cu0.07O Zn0.90Cu0.10O

3.4 4.1 3.9 4.1

1.952 1.941 1.951 1.945

0.0043 0.0022 0.0010 0.0015

Fig. 9. The Cu K-edge XLD signals for the films with x ¼ 0.07 and 0.10.

XMCD measurements at the Cu K-edge have been carried out for the x ¼ 0.04 and x ¼ 0.10 films. The experiment was carried out at the temperature of 7 K, in an external magnetic field of 3 T. No significant signal was obtained, as can be seen from the data shown in Fig. 10. This suggests that Cu atoms are non-magnetic in our ferromagnetic samples. Due to limited beamtime at ID12, we did not try to measure the XMCD signals for the films with x ¼ 0.07 Cu concentration. In fact, as concerns the x ¼ 0.07 film, our results indicate that the structure of this film is very similar to the structure of the x ¼ 0.04 and x ¼ 0.10 films. There are thus no reasons to expect that the XMCD for this film would differ from the XMCD for the x ¼ 0.04 and x ¼ 0.10 films, i.e. we assume it will be zero. As concerns the x ¼ 0.02 film, the situation is different, because some of the Cu atoms are now in Zn-substitutional positions, resembling the case of the Zn1xCuxO pellets where a non-zero XMCD was recorded before [22]. However, the films contain a much less material than

Fig. 10. The Cu K-edge XMCD signal normalized to the x-ray absorption edge jump for Zn1xCuxO films with Cu concentration x ¼ 0.04 and x ¼ 0.10, recorded at temperature of 7 K, for an external magnetic field of 3 T.

309

the pellets and, consequently, the fluorescence intensity of the Cu K-edge signal for the x ¼ 0.02 film was too low to allow us a reliable XMCD measurement e similarly as it was the case with the XLD. Therefore, we cannot say anything definite about the Cu K-edge XMCD of the “atypical” x ¼ 0.02 film. Despite the zero Cu K-edge XMCD signal for the x 0.04 films, the Cu atoms might be in principle magnetic. Namely, the biggest contribution to magnetism of TM atoms comes from the spin magnetic moments of the d electrons while here we probe just the p component of the orbital magnetic moment, which hypothetically might be zero for a magnetic atom [13]. However, we are not aware of any system where this would really be the case. Moreover, let us recall that a well-defined Cu K-edge XMCD signal was measured for paramagnetic Cu:ZnO pellets, demonstrating that this technique is able to probe the magnetic state of Cu atoms [22]. Considering solely the XMCD experiment, we cannot further rule out that there are magnetic moments on Cu atoms which are antiferromagnetically ordered: antiparallel alignment of magnetic moments would also yield zero XMCD signal and bulk CuO is antiferromagnetic only below TN ¼ 230 K, with magnetic moment of about 0.6 mB per Cu atom [30]. We have no knowledge about the size of individual CuO-like regions in the Zn1xCuxO films so we cannot estimate whether the conditions within these regions will be favorable for antiferromagnetic ordering or not. However, even if this was the case, it would remained to be explained where the ferromagnetism of our samples, as evidenced by SQUID measurements (see Figs. 1 and 2), comes from. 4. Discussion Our goal was to link local magnetic and structural properties of ferromagnetic Cu:ZnO thin films. We found that even though our samples exhibit RTFM, there are no local magnetic moments associated with Cu atoms which could be considered as an intrinsic source of the macroscopically observed ferromagnetism. The host ZnO wurtzite lattice is mostly unperturbed: however the Cu atoms are not incorporated in this lattice but rather have a CuO-like local coordination at least for x 0.04. Our results indicate that at least in some TM-doped ZnO systems, the measured ferromagnetism is not directly linked to magnetism of the doping atoms. In this respect, our work is an extension of some earlier experimental studies, which found that for ferromagnetic samples the TM atom is only paramagnetic and not ferromagnetic [15,31,32]. Some experimental studies did find ferromagnetically coupled magnetic moments associated with Cu in doped ZnO but without looking at local structure around Cu (by means of Cu K-edge EXAFS) one cannot be sure that, Cu which give magnetic moments is really at Zn site in hexagonal wurtzite structure or not. If it is not then claim from author remains just speculation [20,21]. Even though Cu atoms are non-magnetic in our Zn1xCuxO films, they clearly play a role in the mechanism how ferromagnetism arises in these materials because the magnetization per volume increases with Cu concentration (see Fig. 1). Presumably the dopant introduces some changes in the ZnO host which in turn are the cause of the ferromagnetism. Reckoning this as well as earlier studies of Cu:ZnO, we conjecture that if Cu atoms are in Zn-substitutional site, they are (usually) magnetic, as demonstrated by experiments as well as by calculations [8,22,29,33,34]. If Cu atoms are in a CuO-like environment, they are non-magnetic (consistently also with the results of Ma et al. [16]). It should be mentioned in this connection that RTFM occurs also for CuO nanoparticles e it has been linked to uncompensated spins on the surface of the nanoparticles [35]. However, the total

310


magnetization of our Cu:ZnO films increases linearly with number of Cu atoms which does not point to surface-dominated mechanism. Besides, it seems that the geometric arrangement of Cu atoms in our films does not simply mirror the CuO geometry but is rather CuO-like, with differences at the long-range which become more evident for low Cu concentration. There are no traces of CuO inclusions in the XRD data in Fig. 3 for x ¼ 0.02 and x ¼ 0.04 and the local coordination of Cu for the x ¼ 0.02 sample is totally different from that in CuO. It is thus improbable that the ferromagnetism of our Cu:ZnO films has the same origin as ferromagnetism of CuO nanoparticles. We should note, in this respect, that there are experimental studies which found RTFM in ZnO without any doping [36e38]. Some calculations support this concept as well [39]. We can also mention that a crucial role of Zn vacancies for ferromagnetism in undoped ZnO was indicated both experimentally and theoretically [38e42]. Unfortunately, it seems that the Zn K-edge XANES and XLD are not very sensitive to the presence of Zn vacancies in ZnO e we checked this by numerical simulations, using the procedures employed in our earlier works on doped ZnO [22,29]. Our analysis of the Zn K-edge EXAFS signals (Fig. 4) did not reveal any significant deviation from a perfect ZnO structure either. Also, one has to realize that a few percentage concentration of Zn vacancies would hardly affect the Zn K-edge XANES and/or EXAFS significantly unless the effect of the vacancies in the spectra was really substantial. Given all this, we cannot draw reliable conclusions about either presence or absence of Zn vacancies in our samples. Some other studies found that introducing oxygen vacancies close to the dopant strengthens the magnetism of doped ZnO [18,43]. Our data do not show any evidence for oxygen vacancies either around Cu or around Zn. However, one should note that the studies which demonstrated that oxygen vacancies promote magnetism did not include simultaneous XMCD measurements so it is not clear whether the TM dopants were really ferromagnetically coupled or whether they were only paramagnetic, in which case they would not be directly involved in the ferromagnetism, meaning that also the role of oxygen vacancies would be different in that case [18,43]. Interestingly, according to the Zn K-edge XLD spectra shown in Fig. 5, it appears that films with a higher Cu concentration are more similar to a perfect ZnO single crystal thin films with a lower Cu concentration. This may be due to changes in the texture of films, probably to be related to the size of CuO nanocrystals embedded in the ZnO wurtzite structure. It is interesting to note that our PLDprepared ferromagnetic films, with Cu mostly in a CuO-like environment, were produced from paramagnetic Cu:ZnO pellets in which Cu atoms were mostly found in Zn-substitutional positions (with some exception for the highest Cu content) [22]. This means that the CuO-like environments in which most of Cu atoms reside in the films cannot be just residua from not properly reacted building material. Reckoning the large amount of apparent inconsistencies and disagreements between various studies of doped ZnO in the past, it seems that the properties of TM-doped ZnO vary significantly depending on the method of preparation. It is thus difficult to transform partial results of one study to another because each study may in fact deal with a different material. Every study should thus provide information as much complete as possible about the system, because only then possible links to other studies can be made. More specifically, it appears that it is desirable to check whether the doping atoms participate actively in the observed “global” ferromagnetism as measured via SQUID or not. Otherwise one might be focusing on investigating details which are not truly important for the phenomenon we are interested in. To be able to link local magnetism (or its absence) to the local

structure, one should investigate the magnetic and structural properties simultaneously. From this perspective, it should be noted that while the L2,3 edge spectra are quite handy for studying magnetism of TM-doped oxides, they are not very suitable for structural analysis e one is mostly left with just indirect qualitative estimates based on the electronic structure of valence states, which is less informative and more ambiguous than EXAFS analysis. It seems that by working with the K-edge XANES, EXAFS and XMCD, the connection between magnetism and structure can be drawn with a better confidence. 5. Conclusions By analyzing x-ray diffraction and XANES, XLD, EXAFS and XMCD spectra of Zn1xCuxO films with Cu concentration x ¼ 0.02, 0.04, 0.07 and 0.10, we found that even though the samples exhibit ferromagnetism, with magnetization that increases approximately linearly with Cu concentration, there are no local magnetic moments associated with Cu atoms. The Zn atoms are located in a nearly perfect wurtzite lattice, as in pure ZnO. However, most of the Cu atoms are not incorporated in the host ZnO lattice but rather are in a CuO-like environment. Our results indicate that at least in some TM-doped ZnO systems, the macroscopically observed ferromagnetism is not directly linked to the magnetism of doping atoms. Rather, the doping introduces some yet unspecified changes in the ZnO host which, in turn, become the key factor behind the observed roomtemperature ferromagnetism. Acknowledgements This work has been carried out under the India-Trento Program for Advanced Research (ITPAR 2) funded by the DST and University of Trento (Trento, Italy). The European Synchrotron Radiation Facility (ESRF) is acknowledged for provision of synchrotron radiation facilities. This work was partially supported by the ESRF project HE-3444 at ID 12 and 08-01/893 at BM 08 Gilda. Authors are indebted with A. Rogalev and F. Wilhelm, Scientists in charge of the ID 12 Beamline at ESRF, for their scientific support during and after the experiment. We thank Ajay Gupta and V.R. Reddy for providing a HR-XRD facility and Alok Banerjee for help with magnetization measurements using a SQUID magnetometer at UGC-DAE Consortium for Scientific Research (CSR), Indore, India. Prof. A.K. Bhatnagar thanks National Academy of Sciences, Allahabad, India for support through its Senior Scientist Platinum Jubilee Fellow scheme. Support by a project 108/11/0853 is also gratefully acknowledged. GACR References [1] K. Ando, H. Saito, Z. Jin, T. Fukumura, M. Kawasaki, Y. Matsumoto, H. Koinuma, Magneto-optical properties of ZnO-based diluted magnetic semiconductors, J. Appl. Phys. 89 (2001) 7284. [2] P. Sharma, A. Gupta, K.V. Rao, F.J. Owens, R. Sharma, R. Ahuja, J.M.O. Guillen, B. Johansson, G.A. Gehring, Ferromagnetism above room temperature in bulk and transparent thin films of Mn-doped ZnO, Nat. Mater. 2 (2003) 673. [3] M. Venkatesan, C.B. Fitzgerald, J.G. Lunney, J.M.D. Coey, Anisotropic ferromagnetism in substituted zinc oxide, Phys. Rev. Lett. 93 (2004) 177206. [4] J.H. Park, M.G. Kim, H.M. Jang, S. Ryu, Y.M. Kim, Co-metal clustering as the origin of ferromagnetism in Co-doped ZnO thin films, Appl. Phys. Lett. 84 (2004) 1338. [5] J. Blasco, F. Bartolome, L.M. Garcia, J. Garcia, Extrinsic origin of ferromagnetism in doped ZnO, J. Mater. Chem. 16 (2006) 2282. [6] M. Snure, D. Kumar, A. Tiwari, Ferromagnetism in Ni-doped ZnO films: extrinsic or intrinsic? Appl. Phys. Lett. 94 (2009) 012510. [7] A. Ney, V. Ney, S. Ye, K. Ollefs, T. Kammermeier, T.C. Kaspar, S.A. Chambers, F. Wilhelm, A. Rogalev, Magnetism of Co doped ZnO with Al codoping: carrierinduced mechanisms versus extrinsic origins, Phys. Rev. B 82 (2010) 041202. [8] M.S. Park, B.I. Min, Ferromagnetism in ZnO codoped with transition metals:

P.S. Vachhani et al. / Journal of Alloys and Compounds 678 (2016) 304e311 Zn1x(FeCo)xO and Zn1x(FeCu)xO, Phys. Rev. B 68 (2003) 224436. [9] D.B. Buchholz, R.P.H. Chang, J.H. Song, J.B. Ketterson, Room-temperature ferromagnetism in Cu-doped ZnO thin films, Appl. Phys. Lett. 87 (2005) 082504. [10] D.L. Hou, X.J. Ye, H.J. Meng, H.J. Zhou, X.L. Li, C.M. Zhen, G.D. Tang, Magnetic properties of n-type Cu-doped ZnO thin films, Appl. Phys. Lett. 90 (2007) 142502. [11] A. Tiwari, M. Snure, D. Kumar, J.T. Abiade, Ferromagnetism in Cu-doped ZnO films: role of charge carriers, Appl. Phys. Lett. 92 (2008) 062509. [12] X.L. Li, X.H. Xu, Z.Y. Quan, J.F. Guo, H.S. Wu, G.A. Gehring, Role of donor defects in enhancing ferromagnetism of Cu-doped ZnO films, J. Appl. Phys. 105 (2009) 103914. [13] J.I. Igarashi, K. Hirai, Magnetic circular dichroism at the K edge of nickel and iron, Phys. Rev. B 50 (1994) 17820. [14] A. Ney, K. Ollefs, S. Ye, T. Kammermeier, V. Ney, T.C. Kasper, S.A. Chambers, F. Wilhelm, A. Rogalev, Absence of intrinsic ferromagnetic interactions of isolated and paired Co dopant atoms in Zn1xCoxO with high structural perfection, Phys. Rev. Lett. 100 (2008) 157201. [15] D.J. Keavney, D.B. Buchholz, Q. Ma, R.P.H. Chang, Where does the spin reside in ferromagnetic Cu-doped ZnO? Appl. Phys. Lett. 91 (2007) 012501. [16] Q. Ma, D.B. Buchholz, R.P.H. Chang, Local structures of copper-doped ZnO films, Phys. Rev. B 78 (2008) 214429. [17] S.Y. Zhuo, X.C. Liu, Z. Xiong, J.H. Yang, E.W. Shi, Defects enhanced ferromagnetism in Cu-doped ZnO thin films, Solid State Commun. 152 (2012) 257. [18] Q. Ma, J.T. Prater, C. Sudakar, R.A. Rosenberg, J. Narayan, Defects in roomtemperature ferromagnetic Cu-doped ZnO films probed by X-ray absorption spectroscopy, J. Phys. Condens. Matter 24 (2012) 306002. [19] P. Thakur, V. Bisogni, J.C. Cezar, N.B. Brookes, G. Ghiringhelli, S. Gautam, K.H. Chae, M. Subramanian, R. Jayavel, K. Asokan, Electronic structure of Cudoped ZnO thin films by x-ray absorption, magnetic circular dichroism and resonant inelastic x-ray scattering, J. Appl. Phys. 107 (2010) 103915. [20] T.S. Herng, D.-C. Qi, T. Berlijn, J.B. Yi, K.S. Yang, Y. Dai, Y.P. Feng, I. Santoso, nchez-Hanke, X.Y. Gao, A.T.S. Wee, W. Ku, J. Ding, A. Rusydi, RoomC. Sa temperature ferromagnetism of Cu-doped ZnO films probed by soft x-ray magnetic circular dichroism, Phys. Rev. Lett. 105 (2010) 207201. [21] T. Kataoka, Y. Yamazaki, V.R. Singh, A. Fujimori, F.-H. Chang, H.-J. Lin, D.J. Huang, C.T. Chen, G.Z. Xing, J.W. Seo, C. Panagopoulos, T. Wu, Ferromagnetic interaction between Cu ions in the bulk region of Cu-doped ZnO nanowires, Phys. Rev. B 84 (2011) 153203. [22] P.S. Vachhani, G. Dalba, R.K. Ramamoorthy, F. Rocca, O. Sipr, A.K. Bhatnagar, Cu doped ZnO pellets: study of structure and Cu specific magnetic properties, J. Phys. Condens. Matter 24 (2012) 506001. [23] P.S. Vachhani, J.H. Markna, R.J. Choudhary, D.M. Phase, D.G. Kuberkar, High field sensitivity at room temperature in p-n junction based bilayered manganite devices, Appl. Phys. Lett. 92 (2008) 043506. [24] T. Mizoguchi, I. Tanaka, S. Yoshioka, M. Kunisu, T. Yamamoto, W.Y. Ching, First-principles calculations of ELNES and XANES of selected wide-gap materials: dependance on crystal structure and orientation, Phys. Rev. B 70 (2004) 045103. [25] T. Yamamoto, Assignment of pre-edge peaks in K-edge x-ray absorption spectra of 3d transition metal compounds: electric dipole or quadrupole? X-

311

Ray Spectrom. 37 (2008) 572. [26] S. Bocharov, T. Kirchner, G. Drager, O. Sipr, A. Simunek, Dipole and quadrupole contributions to polarized Cu K x-ray absorption near-edge structure spectra of CuO, Phys. Rev. B 63 (2001) 045104. [27] A. Kuzmin, EDA: EXAFS data analysis software package, Phys. B 208&209 (1995) 175. [28] A.L. Ankudinov, B. Ravel, J.J. Rehr, S.D. Conradson, Real-space multiplescattering calculation and interpretation of x-ray absorption near-edge structure, Phys. Rev. B 58 (12) (1998) 7565. [29] O. Sipr, J. Vackar, P.S. Vachhani, R.K. Ramamoorthy, G. Dalba, A.K. Bhatnagar, F. Rocca, Local structure and magnetism of Cu-doped ZnO via Cu K-edge XAS and XMCD: theory and experiment, J. Phys. Conf. Ser. 430 (2013) 012128. €ssler, M. Schulz, Landolt-Bo €rnstein, Cupric Oxide (CuO) [30] O. Madelung, O. Ro Magnetic Properties, Heat Capacity, Density, Group III Condensed Matter, vol. 41C, Springer-Verlag, Berlin, 1998. [31] A. Barla, G. Schmerber, E. Beaurepaire, A. Dinia, H. Bieber, S. Colis, F. Scheurer, J.-P. Kappler, P. Imperia, F. Nolting, F. Wilhelm, A. Rogalev, D. Müller, J.J. Grob, Paramagnetism of the Co sublattice in ferromagnetic Zn1xCoxO films, Phys. Rev. B 76 (2007) 125201. [32] T. Tietze, M. Gacic, G. Schütz, G. Jakob, S. Brück, E. Goering, XMCD studies on Co and Li doped ZnO magnetic semiconductors, New J. Phys. 10 (2008) 055009. [33] L.H. Ye, A.J. Freeman, B. Delley, Half-metallic ferromagnetism in Cu-doped ZnO: density functional calculations, Phys. Rev. B 73 (2006) 033203. [34] L.M. Huang, A.L. Rosa, R. Ahuja, Ferromagnetism in Cu-doped ZnO from firstprinciples theory, Phys. Rev. B 74 (2006) 075206. [35] A. Punnoose, H. Magnone, M.S. Seehra, J. Bonevich, Bulk to nanoscale magnetism and exchange bias in CuO nanoparticles, Phys. Rev. B 64 (2001) 174420. , Observation of ferromagnetism at room tem[36] N.H. Hong, J. Sakai, V. Brize perature in ZnO thin films, J. Phys. Condens. Matter 19 (2007) 036219. [37] S. Banerjee, M. Mandal, N. Gayathri, M. Sardar, Enhancement of ferromagnetism upon thermal annealing in pure ZnO, Appl. Phys. Lett. 91 (2007) 182501. [38] M. Khalid, M. Ziese, A. Setzer, P. Esquinazi, M. Lorenz, H. Hochmuth, M. Grundmann, D. Spemann, T. Butz, G. Brauer, W. Anwand, G. Fischer, W.A. Adeagbo, W. Hergert, A. Ernst, Defect-induced magnetic order in pure ZnO thin films, Phys. Rev. B 80 (2009) 035331. [39] Q. Wang, Q. Sun, G. Chen, Y. Kawazoe, P. Jena, Vacancy-induced magnetism in ZnO thin films and nanowires, Phys. Rev. B 77 (2008) 205411. , M. Dubiel, Ch. Eisenschmidt, M. Khalid, P. Esquinazi, [40] J. Haug, A. Chasse Characterization of lattice defects by x-ray absorption spectroscopy at the Zn K-edge in ferromagnetic pure ZnO films, J. Appl. Phys. 110 (2011) 063507. [41] D. Kim, J. Yang, J. Hong, Ferromagnetism induced by Zn vacancy defect and lattice distortion in ZnO, J. Appl. Phys. 106 (2009) 013908. [42] W.A. Adeagbo, G. Fischer, A. Ernst, W. Hergert, Magnetic effects of defect pair formation in ZnO, J. Phys. Condens. Matter 22 (2010) 436002. [43] G. Ciatto, A. Di Trolio, E. Fonda, P. Alippi, A.M. Testa, A. Amore Bonapasta, Evidence of cobalt-vacancy complexes in Zn1xCoxO dilute magnetic semiconductors, Phys. Rev. Lett. 107 (2011) 127206.