Recently, diluted magnetic semiconductors (DMSs) [1-3] have attracted much interest for their potential applications in spintronics. Among them, Mn-doped ZnO is the most favorable candidate because the Mn ion possesses the highest magnetic moment among 3d transition metals and can create a fully polarized stable state due to halffilled 3d bands. There are many reports on room temperature ferromagnetism (RTFM) in the system since the theoretical prediction of RTFM by Dietl et al. in 2000 [4]. However, the origin of ferromagnetism (FM) is still unclear. Some studies indicate that FM might come from the precipitates (e.g., clusters and/or a secondary phase) [5-7]. On the other hand, some studies think that FM might originate from exchange interactions between the localized magnetic moment of the substitutional ions on Zn sites and free charge carriers generated by doping [8-10]. Another origin of FM might be defect-mediated coupling of spins [11-13]. In Mn-doped ZnO, the defects are mainly oxygen/zinc vacancies or interstitials. The defect like O vacancies or Zn interstitials might trap an electron, forming an H-like structure. These electrons confined in O vacancies or Zn interstitials would interact with the delectrons of a Mn atom within their orbits, yielding a bound magnetic polaron (BMP) [14]. The BMPs overlap with each other and hence inducing FM. Recently, FM is also observed in n-type ZnO:Mn films at or above RT, and the magnetization monotonously depends on the electron carrier concentration [15,16]. The experimental result contradicts the theoretical prediction which requires a high concentration of holes for FM [4,17]. Therefore, to further investigate and clarify the origin of FM, it is very important to investigate microstructrure of Mn doped ZnO, especially occupation sites and valent states of Mn atoms.

X-ray absorption near-edge structure (XANES) spectroscopy is a powerful probe for providing a “fingerprint” of chemical states and local electronic structure of incorporated atoms in the host compounds even in a dilute concentration. In particular, the absolute energy position of the edge spectra contains information about the valence state of the absorbing elements. In this paper we employ XANES combining other analysis techniques to investigate the Mn local atomic and electronic structures as well as magnetic interactions in Mn-doped ZnO film.

Mn-doped ZnO films were deposited on (0001) sapphire substrates (99.999%) by radio frequency (RF) magnetron sputtering, with a composite target of ZnO (99.99%) and Mn (99.99%). According to the area ratio of ZnO and Mn, the Mn-doped content could be adjusted. High purity Ar (99.999%) was introduced into the sputtering chamber at the base pressure of ≤6.0 × 10⁻⁴ Pa. The Arflow rate was 20 SCCM (SCCM denotes cubic centimeter per minute at STP (standard temperature and pressure)). The working pressure was 0.5 Pa, the sputtering power 90 W and the substrate temperature 500˚C. Prior to a deposition, a pre-sputtering cleaning was performed for about 20 minutes to eliminate possible contaminants from the target surface.

The Mn K-edge XANES (6.539 keV) spectra were measured at the U7C beamline of National Synchrotron Radiation Laboratory of China. High resolution XPS spectra were detected using a Kratos AXIS Ultra DLD spectrometer with a high resolution of 0.48 eV. The monochromatic Al Kα X-ray (λ = 0.8339 nm) was used as the incident light. The binding energy scale was calibrated using the C 1s line at 284.8 eV. Electrical properties were carried out at room temperature by Hall effect measurements using a Van der Pauw four-point method. An In electrode was made by soldering indium at four corns of the sample surface. The linear I-V behavior for all samples indicated a good Ohmic contact between In electrode and film layer. The magnetic property at room temperature was measured by the Quantum Design MPMSXL7 SQUID magnetometer.

To detect Mn cluster and any secondary phases in the ZnO:Mn samples, we have measured Mn K-edge XANES of all the DMS samples and Mn metal. For clarity, Figure 1 only shows the measured spectra of the Zn_0.89Mn_0.11O film and Mn metal. The FEFF 9.0 [18] simulations of Mn oxides (MnO, MnO₂, Mn₂O₃, Mn₃O₄) are also shown in Figure 1. The spectrum of the DMS sample is quite different from Mn oxides and metallic Mn. The spectrum of MnO has the same features as that of the DMS film in a high energy region (Peak B, C and D), while there is no agreement in the low energy region

(Peak A). It implies that there are no Mn oxides or metallic manganese clusters in the films. The same results have been obtained using an extended X-ray absorption fine structure (EXAFS) and synchrotron radiation X-ray diffraction (SR-XRD) [19]. Additionally, the Mn absorption edge position of the DMS samples is close to that of MnO, suggesting the existence of Mn²⁺.

In order to further investigate Mn occupation sites, we have attempted to perform XANES calculations with the FDMNES 2007 code [20] at Mn K-edge for a lot of possible configurations such as the substitutional Mn_Zn, the interstitial Mn_I, the Mn_Zn-Mn_I dimer, and the Mn_Zn-OMn_Zn. For the interstitial Mn_I model, Mn is placed at the void in wurtzite ZnO structure locating at the center of the Zn tetrahedron. The Mn_Zn-Mn_I dimer is based on the interstitial structure, where one Zn atom of the Mn-containing tetrahedron is substituted by a Mn atom. For the Mn_Zn-O-Mn_Zn, two subtitutional Mn atoms on Zn sites are separated by an O atom. All XANES functions are calculated within a sphere of 6 Å radius, whose center is the absorber atom Mn. Figure 2 shows the spectra of the DMS samples and the calculated spectra from the above-mentioned models. Obviously, the spectra of the DMS samples are different from that of the interstitial Mn_I, which means that Mn could not exist at the interstitial site. We also explore the clustering tendency of the Mn atoms in ZnO. From the obvious difference between the spectra of the Mn_Zn-Mn_I dimer, the Mn_Zn-O-Mn_Zn and the DMS samples, we can exclude the phenomena. In the XANES spectra of the DMS samples, there are four characteristic peaks A (6541 eV), B (6546 eV), C (6555 eV) and D (6567 eV). The preedge peak A can be interpreted as the transition of Mn 1s core electron to the unoccupied Mn 3d and O 2p hybridized states. In the sub stitutional model, “Mn_Zn” represent the model of no op-

timizing the geometry of the studied configurations, i.e., the bond lengths of Mn_Zn-O and Mn_Zn-Zn is 1.97 and 3.20 Å, respectively, while “Mn_Zn expand” represent the model in which the bond lengths of Mn_Zn-O and Mn_Zn-Zn is expanded to 2.03 and 3.28 Å, respectively. In the Mn_Zn model, A, C and D peaks can be well reproduced except the B peak. On the other hand, the height of the Peak A in the Mn_Zn model is much larger than that in the DMS samples. However, in the Mn_Zn expand model, four characteristic peaks can be well reproduced and the calculated spectrum resembles those of the DMS samples. It implies that Mn atoms are incorporated into the ZnO crystal lattice by the substitution on Zn sites, accompanying with an expansion of the Mn_Zn-O and Mn_Zn-Zn bond lengths due to the larger atomic radius of Mn compared with Zn. The same results have been observed by an extended X-ray absorption fine structure (EXAFS) [19].

The valent state of Mn in the DMS films is detected by high resolution XPS. In Figure 3, the Mn 2p_3/2 peak for the DMS samples is located at 640.73 eV, which is very close to Mn²⁺ in MnO (Mn 2p_3/2: 640.7 eV) according to the XPS handbook. It well agrees with those reported in the literature, indicating a divalent state of the Mn ions in ZnO:Mn films [21,22]. Moreover, from the XPS handbook we also know that the Mn 2p_3/2 peaks for metallic Mn, Mn₃O₄, Mn₂O₃ and MnO₂ are located at 639, 641.2, 641.6 and 642.2 eV, respectively. So we can exclude the existence of Mn oxides or metallic manganese clusters in the films, in good agreement with the XANES analysis.

Figure 4 shows the magnetization curves at room temperature. These magnetization curves are obtained with the applied field parallel to the plane of the sample. The diamagnetic background of ZnO and sapphire substrate are subtracted. The hysteresis loops show the clear ferromagnetic behavior of these Mn-doped ZnO films at room temperature. From Figure 4, the saturation mag-