Ferromagnetism in undoped ZnO grown by pulsed laser deposition

Undoped ZnO films grown on sapphire by pulsed laser deposition are magnetic at room temperature. A comprehensive study involving x-ray diffraction, positron annihilation spectroscopy, and superconducting quantum Interference device-vibrating sample magnetometer is performed to study the origin of the observed magnetization. Correlations between the saturation magnetization, VZn−2VO concentration and surface to volume ratio of the grain found experimentally show that the magnetization is associated with the vacancy cluster and probably VZn−2VO residing on the grain surface.


Introduction
Since the theoretical prediction of room temperature (RT) ferromagnetism (FM) of transition metal doped ZnO [1], there were extensive research devoted to study the RT FM in transition metal doped ZnO materials [2,3]. RT FM was reported in ZnO doped with a variety of transition metals [4], namely Mn [5][6][7][8][9][10], Co [7,8,[11][12][13][14][15][16][17], Fe [18], and Cu [19][20][21] etc, and defects played important role in mediating the ferromagnetic coupling [6,[20][21][22]. However, there were concerns that the measured ferromagnetism was originated from the secondary phase [23]. RT FM was also reported in undoped ZnO materials [24][25][26][27][28][29][30][31], which thus makes ZnO a feasible potential material for spintronic applications [24]. The report of bias voltage controlled FM switching in undoped ZnO multi-layer structure device also reveals the potential magneto-electrical device applications [32,33]. As compared to transition metal doped ZnO, there were relatively fewer RT FM study in undoped ZnO. The origin of the RT FM of undoped ZnO is still controversial. The RT FM has been associated to the surface of nanoparticle, single crystal and grain boundary materials [24][25][26][27][28][29][30][31], though the exact spices contributing for the FM is not unambiguously known. Tietze et al [27] suggested that the RT FM was originated from the dangling electrons on the nanoparticle surface. Many other experimental and theoretical studies reported that vacancy type defects could contribute for the RT FM, thought there was no agreement on the identity of the vacancy. In a hybrid density functional theoretical DFT study, Chakrabarty and Patterson [34] reported that the negatively charged (V Zn V O ) − divacancy and -V Zn monovacancy are possible origins for RT FM in undoped ZnO. Wang et al [35] carried out DFT study in undoped ZnO and found that FM could be originated from the V Zn , arising from the unpaired 2p electrons at the O sites surrounding the V Zn . Their calculation also shows that the V Zn prefers to cluster and resides on the surface of the thin film. Performing density functional theory study, Wu et al [36] reported that V Zn in twin grain boundary favours ferromagnetic coupling. The density functional theory within the GGA exchange functional study performed by Lin et al [37] showed that V O could not contribute for net magnetic moment, and the FM was likely originated from the V O cluster and V Zn complex defects. However, based on the results of experimental and theoretical studies, Sun et al [29] suggested that the FM was associated to V O .
Obtaining the unambiguous origin or the defect microstructure associated with the magnetization observed in undoped ZnO is not easy. A comprehensive study on the correlation between the magnetic property and the stoichiometry of Zn and O during growth could offer the stoichiometric information of the defect associated with the observed magnetization. PAS is probe selectively sensitive to defect containing V Zn which can offer information of the microstructure of the defect [38][39][40]. The Doppler broadening of the annihilation radiation (which is characterized by the S-parameter) originated from the incident positron and the electron around the defect can reveal the electronic momentum around the defect. The S-parameter depends on the defect microstructure, and for defect having the same microstructure an increase of S-parameter implies a higher defect concentration. Khalid et al [26] observed correlation between the magnetization and the S-parameter of the PLD grown undoped ZnO samples grown with different nitrogen pressure, and suggested that Zn vacancy is probably responsible for the magnetic order. However, the S-parameter is a combined effect of the defect microstructure and concentration, and solely the S-parameter study cannot be used to reveal the microstructure of the defect.
In the present study, the undoped ZnO films grown by PLD are all magnetic at room temperature. To study the origin of the magnetization of these samples, a comprehensive approach is adopted. x-ray diffraction (XRD) study is conducted to study the crystalline quality of the film and also the grain surface area to grain volume ratio s GB . PAS study was carried out to reveal the origin of the RT FM. As compared to the PAS study in Khalid et al [26] for which the S-parameter approach is adopted, the coincidence Doppler broadening approach is used in the present study, in which the defect microstructure of the involved defect can be identified while comparing the experimental spectrum with the theoretical simulation. The magnetic properties of the samples were studied by the Superconducting Quantum Interference Device-Vibrating sample Magnetometer (SQUID) measurement.

Experimental
The undoped ZnO films were grown on c-plane sapphire substrate using the PLD method, whereas laser pulse with wavelength, energy, spot fluence, and repetition rate of respectively 248 nm, 300 mJ, 7.4 Jcm −2 and 2 Hz from a krypton fluoride (KrF) excimer laser (COMPexPro 102 by Coherent Inc.) was used to ablate the ZnO ceramic (purity of 99.999%). The background pressure was 10 −4 mbar. The samples were grown on the c-sapphire substrate with different oxygen pressures P(O 2 ) of 0 Pa, 1 Pa and 3 Pa for the fixed growth time of 3 h. This method of growth yields undoped ZnO films with the polarity of O-face [41]. The annealing was carried out in Ar atmosphere for a period of 45 min.
The film thickness was determined by inspecting the sample cross section using the scanning electron microscope (JOEL JSM-7001F). The XRD measurement was conducted using the Bruker D8 Advance. The magnetic measurement was conducted using the Quantum Design SQUID with a sensitivity of 10 −7 emu. The PAS study was conducted using a monoenergetic positron beam with positron energy varying up to 30 keV. The annihilation gamma energy spectra of the annihilation photons were detected by a high purity Ge (HPGe) detector and the corresponding nuclear electronic system which had an energy resolution of 1.1 keV for the 514 keV gamma radiation peak. The coincidence Doppler broadening of the annihilation radiation lineshape was monitored by the S-parameter and W-parameter, respectively defined as the ratios of the count of the central window (511.00±0.76 keV) and the ratio of the counts of the wings (511.00±3.4 keV and 511.00±6.8 keV) divided by the total spectrum count.

Results
The area and the thicknesses of ZnO films (determined by the cross sectional SEM images) are tabulated in table 1. The film thickness was ∼240 nm and ∼330 nm for the films grown without and with oxygen respectively. XRD study was performed on the ZnO samples grown with P(O 2 ) = 0, 1 and 3 Pa (Samples A-0Pa, A-1Pa and A-3Pa). The XRD spectrum of the as-grown undoped ZnO films grown with P(O 2 ) = 0 Pa (i.e. A-0Pa) is shown in figure 1(a) with the intensity shown in log scale, which shows that the film has wurtzite structure with (002) as the preferential orientation. Although another ZnO (100) peak is also identified, its intensity is much lower (more than 200 times) than the ZnO (002) peak. The other samples grown with different P(O 2 )'s have similar XRD pattern. Figure 1 where f b and f di are respectively the fractions of positron annihilating in the delocalized bulk state and the i-th vacancy states, and S b and S di are respectively the corresponding characteristic S-parameter [38][39][40]. If a single type of defect exists, it then yields ( ) ( implying that the W-S plot is a stragiht line and the slope is the fingerpring of the vacancy whereas the S-parameter is proportional to the defect concentration [38][39][40]. The experimental (W,S) parameter data of a ZnO single crystal was also measured and included in     For the samples grown with the same P(O 2 ), the W-S data moves up after annealing at 750 o C, implying the reduction of the defect concentration upon the annealing. It would be worthy to explore the microstructure of this V Zn -related defect found in these samples. Makkonen et al [42] carried out theoretical study on the Doppler broadening of annihilation gamma radiation coming from the V Zn -related defects in ZnO. Their W-S characteristic straight line of V Zn monovacancy and 4V Zn vacancy cluster are included in figure 2. It is also found the W-S characteristic straight lines of V Zn V O and 2V Zn −V O is very close to that of V Zn , and beyond the resolution of the Doppler broadening technique. The vacancy clusters V Zn −2V O is also studied and its characteristic straight line is shown in figure 2. It is noticed that the characteristic straight line for the defect as found in the current ZnO samples is very close to the theoretical straight lines of the V Zn −2V O and clearly distinguishable from the other theoretical W-S lines. It is thus plausible to infer that the V Zn −2V O is the single type V Zn -related defect dominated in these ZnO samples.
Magnetic study was carried out on the as-grown ZnO samples grown with different P(O 2 )'s (0 Pa, 1 Pa and 3 Pa) which is dominated by V Zn −2V O . The M-H curves of the samples normalized against the film volume (in unit of emu cm −1 ) are shown in figure 3, and the corresponding magnetization are tabulated in table 1. One of the measurements is shown in the insert low right hand corner of figure 3 in emu without normalization. Hysteresis loop is revealed for all the three samples, though the hysteresis is small. Their saturated magnetization (M s ) are tabulated in table 1, for which the M S increases from 0.108 to 0.119 emu g −1 while P(O 2 ) increases from 0 Pa to 3 Pa. The coercivities H C of the 0Pa, 1Pa and 3Pa samples are respectively 119 Oe, 99 Oe and 84 Oe, while their remanences are 0.0148 emu g −1 , 0.0129 emu g −1 and 0.0118 emu/g. The experimental saturated magnetization of undoped ZnO observed from previous literatures diverges over a wide range, say 0.2 emu g −1 for ZnO nanowire [31], and ∼0.01 emu g −1 for PLD grown ZnO film on sapphire [26]. The saturated magnetization observed in the present study is within the range reported in the literatures of undoped ZnO film grown by PLD. Note that we have also carefully checked the Al 2 O 3 substrates used for growing ZnO films. As shown in the inset to figure 3 (the upper left-hand corner), the field-dependent magnetization of the Al 2 O 3 substrate shows a linear diamagnetic behavior, but the ZnO film shows a clear deviation from the linear behavior due to the ferromagnetic component.
It is worthy to explore the temperature dependence of the magnetic properties, and the correlation between the saturated magnetization, V Zn −2V O abundance and surface to volume ratio of the grain. Figure 4(a) shows the zero field cooling (ZFC) and field cooling (FC) temperature dependent magnetization for the as-grown sample grown with P(O 2 ) = 0 Pa. To perform the measurements, the sample was cooled down from room temperature to 5 K. Then a field of 100 Oe was applied. The ZFC curve was measured during warming up. The FC curve was measured after the sample being cooled down from room temperature to 5 K under a field of 10000 Oe. Then the field was set to 100 Oe. The FC magnetization was also recorded during warming up. The irreversibility between ZFC/FC curves would be due to the ferromagnetic component, which persists well above room temperature in this sample. Figure 4(b) shows the temperature dependent magnetization for this sample measured under 10000 Oe which insures the saturation. Below 50 K, the saturation magnetization shows a fast drop with increasing temperature, which might be due to a small paramagnetic component arising from uncoupled defects. Above 50 K, the saturation magnetization shows mild decrease with increasing temperature, indicating the Curie temperature for the ferromagnetic component is well above room temperature.
The fraction of positron annihilating at V Zn −2V O ( f defect ) is related to its concentration and can be found by if single type of the vacancy type defect exists. Figure 5 summarizes the saturated magnetization M S , fraction of positron annihilating at V Zn −2V O , and the grain surface area to volume ratio S GB as a function of P(O 2 ) for the as-grown samples grown with different P(O 2 )'s. While P(O 2 ) increases from 0 Pa to 3 Pa, the corresponding M S , S GB and f defect increase from 0.108 emu g −1 to 0.119 emu g −1 , 3.3×10 7 m −1 to 3.6×10 7 m −1 , and 0.33 to 0.37 respectively. These correspond to the fractional changes of ΔM S /M S ∼10%, Δs GB /s GB ∼9% and Δf defect /f defect ∼12%. These fractional changes are close, thus suggesting that the correlations between the saturation magnetization, grain boundary surface and the V Zn −2V O concentration. This implies that the observed RT FM in the current undoped ZnO films is associated with the V Zn −2V O residing in the vicinity of the grain boundary.
Khalid et al's [26] previous experimental magnetic studies on the PLD grown undoped ZnO films correlated with results of PAS using the S-parameter approach revealed that the magnetization is associated with defect or defect complex consisting of V Zn . As compared to the present CDBS approach which identified defect structure

Conclusion
The undoped ZnO samples grown by PLD at oxygen pressures of 0 Pa, 1Pa and 3 Pa are magnetic at RT and the hysteresis is small. The dominating V Zn -related defect in these samples is the V Zn −2V O vacancy cluster. Experimental correlations between the magnetization, the abundance of V Zn −2V O and the surface to volume ratio of the grain unambiguously show that the magnetization is associated with the V Zn −2V O residing on the grain surface. However, the physics of how the magnetism is arisen from this defect is not known and requires theoretical study.