The identification of room-temperature ferromagnetism (RTFM) in non-magnetic dopant (NM)-doped oxide has recently attracted considerable attention for use in semiconductor spintronics 1, 2. Finding a suitable NM dopant that makes non-magnetic oxides magnetic is important. Copper-doped ZnO has been demonstrated to be a diluted magnetic oxide (DMO) with RTFM evidenced by X-ray magnetic circular dichroism (MCD). Herng et al. 3 proposed a microscopic indirect double-exchange model of RTFM that explains their main experimental findings. However, according to recent studies, the magnetic properties are not exclusively related to the dopants, but are largely governed by defects 4–6. According to one investigation, the presence of Li in a ZnO matrix during the growth process reduces the formation energy of Zn vacancies, stabilizing them. A Zn vacancy can cause RTFM in intrinsic DMO 7. These results reveal that the RTFM that is caused by different NM dopants in the ZnO system may have different magnetic origins. Carbon-doped ZnO (C:ZnO) is also expected to be a DMS, and its ferromagnetism has been suggested to originate from carbon atoms which are incorporated at O (CO) sites in ZnO 8, 9. However, the origin of the RTFM of C:ZnO remains unclear. These films are sensitive to the deposition conditions and methods of preparation. Whether the magnetism arises from the C dopants in ZnO or the additional defects is important to determine 10, 11. This study investigates the variation of the electronic structure, magnetic properties and related MCD effects with annealing conditions, to clarify the role of C dopants in a ZnO DMO.
The C:ZnO thin film samples were prepared on quartz substrates at RT using an r.f. magnetron sputtering system with a base pressure at about ∼1 × 10–6 Torr. The sputtering target is a commercial Zn–C–O compound with 5 at% C (purity 99.95%, 2 in. in diameter, Materion Corp.) which was used for deposition. The working pressure of the sputter gas (Ar, 99.9995% purity) was 1.3–2.0 mTorr and the applied sputtering power was 100 W with a target-to-substrate distance of 20 cm. The substrate was rotated at 20 rpm to ensure uniformity of the films. The films were measured to be ∼100 nm using the cross-section scanning electron microscopy image. The actual concentration of C in the as-grown films is ∼3.4 at% estimated by X-ray photoelectron spectra (XPS). The intrinsic origin of RTFM is reportedly related to the ratio of the grain-boundary area to grain volume 4. There are also reports that thermal
oxidation of carbon occurs at T > ∼300 °C 12. To elucidate the role of C, the films were annealed in air and in vacuum (1 × 10–2 Torr) at a relatively low annealing temperature of 350 °C for 1 h. Therefore, the defect structures could be varied but similar grain sizes are maintained. The structure and crystalline quality of the film were characterized by X-ray diffraction (XRD) measurements. These films exhibited a prominent c-axis (002) texture, corresponding to a ZnO wurtzite structure. The grain sizes, which were estimated by XRD were nearly the same, at around 35 nm. The two samples resembled each other in their long-range structures. However, the two films had distinct magnetic properties. These were characterized by a commercial superconducting quantum interference device (SQUID). Whereas the vacuum-annealed C:ZnO samples had negligible RTFM, the SQUID hysteresis loop of air-annealed C:ZnO, displayed in Fig. 1, reveals RTFM. Notably, however, the SQUID loops of the air-annealed C:ZnO, after subtraction of the non-ferromagnetic signal at 15 K and 300 K reveal little hysteresis. Hence, the film did not exhibit the expected typical ferromagnetism. The anhysteretic magnetization curves and the order of magnetization are also similar to other defect-induced ferromagnetic materials 4, 7.
XPS data were obtained for further analysis of the states of the dopant C atoms. Figure 2 presents the XPS of air-annealed and vacuum-annealed C:ZnO films. Sputtering was carried out for 2 min (∼5 nm) and 4 min (∼10 nm), respectively, to remove C-contaminants before scanning. Thereafter, a peak at ∼285 eV was observed. This peak is probably attributable to the “sp2 orbit of the carbon”. However, the formation of Zn–C bonds with a C1s binding energy (280–284 eV) in C-doped ZnO films, which has been reported by others 9, was not observed. The absence of experimental proof of CO in the film herein reveals that the reported ‘Co-induced RTFM’ is inapplicable in our case. However, the absence of RTFM in air-annealed ZnO without C doping implies that C may play an important role in determining the observed magnetic properties. Notably, the intensities of air-annealed samples are smaller than those of vacuum-annealed samples, revealing that the reaction between carbon (in the matrix) and oxygen upon annealing in air can easily form the carbon dioxide by the interaction of oxygen atoms with the dangling bonds in the sp2 orbits of the carbons, reducing the carbon concentration. The C content decreases to ∼1.8 at% after air annealing near the surface. Therefore, more defects remain in the air-annealed C:ZnO film, directly corresponding to the observed RTFM, as discussed below.
Figure 3(a) presents absorption spectra of the studied samples. The band tails of the air-annealed C:ZnO can be related to the formation of more defects. Energy-dependent MCD and spectra can provide valuable information on the above magnetic phenomenon 13. MCD spectra were collected using a JASCO 815. The MCD and optical absorption measurements were made using a xenon lamp with alternating σ+ and σ– circularly polarized light (50 kHz), produced using a quartz stress modulator with a magnetic field normal to the plane of the film. Figure 3(b) shows the RT MCD spectra of the annealed C:ZnO samples at a magnetic field of 0.78 T. The spectra were obtained by subtracting out the substrate signals. A broad MCD structure at approximately the ZnO band edge can be observed in the vacuum-annealed sample. If the MCD signal is attributable entirely to an effective field that is proportional to the magnetization, then the vacuum-annealed samples can be reasonably speculated to be weakly ferromagnetic. However, magnetic measurements revealed no detectable RTFM in the vacuum-annealed sample. This finding suggests that this MCD signal is not ferromagnetic. Careful analysis of MCD is necessary to elucidate the origin of the ferromagnetism. On the other hand, the MCD signal of the air-annealed sample is more intense and is shifted to a lower energy. The air-annealed sample exhibits an anhysteretic magnetization curve. The observed MCD signal at the ZnO band edge in Fig. 3(b) normally arises from differential absorption, owing to the splitting of the ZnO band edge such that MCD is proportional to the derivative of the absorption coefficient, dK /dE, which has a maximum close to the band edge 14. The dK /dE spectrum should coincide closely with the MCD spectrum, where K is the optical absorption coefficient and E is the photon energy. Figure 3(c) also shows the derivative of optical density, dK /dE, in arbitrary units. (The optical density is proportional to the optical absorption coefficient.) Not only are peaks obtained at ∼3.4 eV from both films, but also an additional pre-edge peak at ∼3.3 eV is observed from the air-annealed C:ZnO sample, and this peak is related to the first derivative of the absorption band tail. On the other hand, the two films had similar long range structures, as indicated above. Therefore, the MCD signal from air-annealed C:ZnO can be further divided into two components. The first component, called feature A is a peak which is regarded as arising from a non-ferromagnetic material, and is also obtained from the vacuum-annealed sample. The other component can be obtained by subtracting out the curve from the vacuum-annealed sample. Figure 3(d) presents the result obtained after the subtraction. The component, feature B, is a peak at a low energy, and is obtained from the ferromagnetic phase.
The MCD–H curves provide further insight into the properties of both MCD components. As the magnetic field is increased, the intensity of feature A increases linearly, as would occur if the sample were paramagnetic, as shown in the inset of Fig. 4. Unlike the non-saturated feature mentioned above, the magnetic field-dependent feature B from the air-annealed C:ZnO sample gives a saturated field of ∼0.5 T, whose value equals that measured by the SQUID. The spectroscopic contribution is well correlated with the observed RTFM of air-annealed C:ZnO, indicating that the magnetism is intrinsic. This energy region corresponds to the defect states and constitutes the impurity band. The ferromagnetism might be induced by the Stoner excitation with strong orbital localization. No similar finding concerning MCD was observed for the pure ZnO films. We suggest some defects are generated during the loss of carbon after air annealing. Therefore, defect-related bands in air-annealed C:ZnO can induce RTFM.
In conclusion, RTFM was observed in a C:ZnO system even when most of the carbons were not incorporated at O sites. The anhysteretic magnetization curves of C:ZnO are attributable to defect-mediated ferromagnetism. RTFM was associated with a narrow ferromagnetic MCD peak, which can be attributed to the origin of the observed RTFM. The C dopants can promote the formation of defects by reacting with oxygen during annealing in air. The results reveal that even when the C dopants are not at CO sites, a combination of C doping and suitable post-annealing treatment has a strong impact on the defect-mediated ferromagnetism of C:ZnO.