Morphology-Dependent Room-Temperature Ferromagnetism in Undoped ZnO Nanostructures

Since Dietl et al. predicted that Co-doped ZnO may show room-temperature ferromagnetism (RTFM) in 2000, researchers have focused on the investigation of ferromagnetic ZnO doped with various transition metals. However, after decades of exploration, it has been found that undoped ZnO nanostructures can also show RTFM, which in general is dependent on ZnO morphologies. Here, we will give an overall review on undoped ZnO nanomaterials with RTFM. The advanced strategies to achieve multidimensional (quasi-0D, 1D, 2D, and 3D) ferromagnetic ZnO nanostructures and the mechanisms behind RTFM are systematically presented. We have successfully prepared ferromagnetic nanostructures, including thin films, horizontal arrays and vertical arrays. The existing challenges, including open questions about quantum-bound ZnO nanostructures, are then discussed.

In fact, undoped ZnO can also exhibit ferromagnetic ordering. The spin state related with Zn vacancy (V Zn ) was identified first by its electron spin resonance (ESR) spectrum in electron-irradiated ZnO single crystals [34]. It is worthwhile noting that recent experimental results have shown that nano-engineering can also induce ferromagnetic behavior in non-ferromagnetic bulk materials [35][36][37]. The magnetic ordering obtained by the minimum number of vacancies was not successfully detected at that time. Amazingly, it was found later that the miniature of the bulk materials in nanoscale, such as nanoparticles (NPs), could exhibit novel magnetic behaviors in undoped ZnO. In 2006, RTFM (M s , 0.0005 emu/g) was observed in sol-gel ZnO NPs [38], which could be attributed to oxygen vacancies (V O ). Subsequently, Banerjee et al. [39]. found that thermal annealing could induce RTFM in ZnO white powder, owing to the formation of V O clusters during the thermal treatment. Then, RTFM with a saturation magnetization up to 35.65 emu/g was also observed in undoped ZnO thin films (TFs) by pulsed laser deposition (PLD) and V Zn located at the surface or the interface of the samples could be the source of RTFM [40]. More importantly, various ZnO nanostructures with RTFM were obtained by various approaches, such as ionic layer epitaxy [41], polymer-assisted deposition (PAD) [42][43][44], electrochemical deposition method [45], sol-gel [36,46,47], ball milling (BM) [48][49][50], and PLD [51], etc.
Here, we will give an overview of the timeline of undoped ZnO nanostructures with RTFM ( Figure 1).
In addition, V Zn can also be produced in ZnO NRs with RFTM ( Figure 4B), which were synthesized by COP [68]. The as-prepared NWs [69] ( Figure 4C) show an obvious RTFM (M s ,~0.007 emu/g). However, highly crystalline ZnO NWs [69] (Figure 4) prepared by the pulsed laser vaporization (PLV) process show no RTFM. X-ray fine structure spectroscopy study confirms that V Zn can cause RTFM in ZnO NRs.
Furthermore, ZnO NRs have been obtained in aqueous conditions [70]. As-prepared NRs show RTFM behavior. Strangely, when the size of the NRs decrease, the magnetization of ZnO NRs increases. Singly charged oxygen vacancies (V O + ) localized on the sample is the origin of RTFM. Single-crystalline ZnO NWs with RTFM (M s ,~0.001 emu/g) ( Figure 4D,E) have been obtained by the vapor transport method (VTM) [71]. RTFM related with V O can be modulated by selecting different catalysts and changing the growth temperature. Jana et al. [72] has reported ZnO NRs (M s ,~0.059 emu/g) using hydrolysis of zinc acetate. Moreover, polycrystalline ZnO NRs with RTFM have also been prepared by a wet chemical method (WCM) [73]. RTFM in ZnO NRs ( Figure 4) is attributed to V O + .

Nanostructure Arrays
ZnO nanostructure arrays ( Figure 5) have been fabricated by various approaches, such as HYT [74], colloidal template method (CTM) [75], chemical vapor deposition (CVD) [76], chemical bath deposition (CBD) [76,77], and polymer-assisted deposition (PAD) [42][43][44]. ZnO NWs are usually synthesized by a seed method. Firstly, ZnO TFs are prepared by magnetron sputtering and used as seed layer, then ZnO NRs arrays are obtained by HYT [74]. The defect states are controlled by annealing in O 2 and H 2 atmosphere at 500 • C, respectively. It has been found that V O is closely related to RTFM by PL spectrum and vibrating sample magnetometer (VSM). Furthermore, vertically oriented NRs arrays (VNRA) and randomly oriented NRs arrays (RNRA) have been synthesized by CVD and CBD [76]. These results show that RTFM is originated from the V O + .  [77] in Figure 5. The low temperature method can generate more defects, and then induce the higher M s in nanocactuses. ZnO HNWA [43] and VNPA [44] have also been obtained by PAD in our lab. Our results show that more V Zn can potentiate RTFM [43,44]. Thiol-capped ZnO NWs/NTs arrays can present RTFM [78]. The M-H loops of the thiol-capped samples reveal the height-dependent and morphology-dependent RTFM ( Figure 6). Combing magnetic measurements and calculations show that the origin of ferromagnetism is mainly attributed to spin-polarized 3p electrons in S sites and, therefore, has a strong correlation with Zn-S bond anisotropy. In the nanowire system, the change of diameter has no obvious effect on RTFM ( Figure 7). However, the height-dependent RTFM has been observed in capped ZnO NWs ( Figure 6A).

Other Nanostructures
ZnO nanoflowers with RTFM (M s ,~0.069 emu/g) have been obtained by HYT [79]. TEM images show that this flower sample is composed of closely packed NRs. XPS indicates that isolated vacancy clusters is closely related to the RTFM. In addition, microdiscs and porous nanoassembly of ZnO have been synthesized by soft-chemical approaches [80]. Furthermore, various types of ZnO nanostructures have been obtained by a microwave as-sisted HYT [81]. Spectrum measurements show that the defects may mediate RTFM [80,81]. The morphology-dependent RTFM is shown in Figure 8. ZnO HNWA has the highest M s is~7.1 emu/g.

Nanosheets/Nanoplates
Two-dimensional (2D) nanosheets/nanoplates have been synthesized by HYT [45,82,83] and ionic layer epitaxy (ILE) [41] in Figure 9. The average thickness of ZnO Nanoplates is estimated to be~20 nm by the SEM images and XRD peak. Interestingly, it is only when the thickness is reduced to 5-8 nm, the samples exhibit RTFM. Interestingly, first principles calculation results show that when the thickness of the NSs decreases to a certain range, the magnetic moment will be generated; however, when the size of the NWs decreases along the a-axis and b-axis, the magnetic moment will not be generated. The distorted bands may be responsible for the RTFM in ZnO nanoplates.
ZnO single-crystalline nanosheets with RTFM have been obtained [45]. TEM images present that ZnO sheets and dodecyl sulfate bilayers compose the lamellar structure. The results show the cluster of spin-polarized defects induce RTFM in single-crystalline ZnO nanosheets.
Two-dimensional ZnO nanosheets with high V Zn concentration have been grown by ILE [41] (Figure 9E,G). The nanosheets annealed in Ar at 400 • C for 1h show M s of 50.9 emu/g in Figure 9G. Significantly, DFT calculations and experimental results suggest that V Zn could be associated with the RTFM.
By contrast with 2D materials, ZnO nanosheets do not show obvious thickness dependence. However, ZnO nanosheets with saturation magnetization up to 50.9 emu/g have been obtained recently by ILE ( Figure 10).

Thin Films
In 2007, RTFM has been first observed in undoped ZnO TFs by PLD [40]. The results suggest that V Zn can generate RTFM, where the ferromagnetism is related to the thickness of ZnO TFs. Soon after, calculations [84] not only showed that undoped ZnO TFs and NWs can be ferromagnetic but also indicated that V Zn is responsible for RTFM in lowdimensional magnetic ZnO nanostructures. Similar to other nanostructures [38,40,67,80,81], V Zn prefers to reside on ZnO NWs, resulting in stronger RTFM.
The thickness-dependent ferromagnetism in ZnO TFs has been investigated by Kapilashrami et al. [85]. With increasing the film thickness, the ferromagnetism first increases; when the film thickness exceeds 480 nm (M s ,~0.62 emu/g), the ferromagnetism then decreases; with increasing of film thickness, the ferromagnetism will transform into paramagnetism (PM) or even diamagnetism (DM). The defect is mainly responsible for the observed RTFM.
Grain boundaries can also mediate RTFM. ZnO TFs were also obtained by liquid ceramics method [86]. The HRTEM images show that ZnO grains is surrounded by the amorphous area. Furthermore, RTFM origins from the amorphous regions related the defects. By decreasing the grain size, the magnetic volume fractions will increase [87]. Tietze et al. [87]. built a magnetic shell model related to the grain boundaries, which clarified the origin of RTFM in ZnO TFs.
In the ZnO TFs system, V Zn [40,84,[88][89][90] can also induce the RTFM. The existence of V Zn is often confirmed by photoluminescence and positron annihilation spectrum. Xing et al. [88] has found that the sol-gel derived samples contain more V Zn , compared with the MBE-ZnO TFs, which shows much stronger RTFM. Similarly, V Zn is also the origin of RTFM in polycrystalline ZnO TFs prepared by PAD in our lab [89,90].
Additionally, undoped ZnO TFs [91,92] have been fabricated with CVD [93]. Interestingly, a nanosized structure can induce RTFM in ZnO TFs. Similar to ZnO NRs [71,76], V O can generate RTFM in undoped ZnO TFs obtained by PED [94] and undoped ZnO NPs prepared by an electrospinning method [95]. Moreover, Zhang et al. [96] indicated that RFTM in pure ZnO TFs should be related to Zn i . The shallow donor caused by Zn i defects might modify the electronic structure of ZnO TFs, leading to the RTFM.
The thickness dependence of RTFM has been observed in films prepared by different experimental methods (Figure 11). The RTFM of ZnO TFs is affected by the defects on the surface of the sample or at the interface between the sample and the substrate.

Three-Dimensional Materials
In 1970, Galland et al. [34] irradiated ZnO single crystals with 3MeV electrons. ESR spectra showed that V Zn can generate the high spin state. Furthermore, the intrinsic defect just like Zn i and V O may give rise to the RTFM in single crystalline ZnO [97].
Interestingly, the metal-insulator transition has been found in the ZnO supercells adsorbed with hydrogen atoms by first-principles calculations [98]. This phase transition can affect the magnetic transformation. Experimentally, ZnO single crystals have been obtained by HTY [99]. The hydrogen penetration depth is confirmed at 20 nm by stopping and range of ions in matter (SRIM) simulation. SQUID measurements indicate that the M s increases with the treatment time and the concentration of hydrogen. Subsequently, hydrogen atoms may trigger the RTFM in ZnO single crystals in Khalid et al. [100]. The M s is up to~4 emu/g. Strangely, the vacancies defects and the interstitial Zn atoms cannot induce the RTFM in ZnO monocrystal [101].
HYT [102] has been developed to grow Zn 5 (OH) 8 Ac 2 ·2H 2 O microspheres. In fact, the microspheres are grown from the nanosheets, which are curved and connected to each other. ZnO porous microspheres [103] have been obtained from Zn 5 (OH) 8 Ac 2 ·2H 2 O microspheres annealing at 500 • C. The RTFM in pure ZnO microspheres origins from V Zn and shallows donors.

Influence of Precursor Selection on Room-Temperature Ferromagnetism (RTFM)
The preparation of materials has always been optimized by selecting more suitable precursors. In 2019, Zn(NO 3 ) 2 ·6H 2 O and ZnCl 2 were selected as zinc sources to prepare ZnO NPs [104]. The difference of M s was nearly four times, which may have been due to V O in the samples produced by selecting different precursors.

Substrate Effects on Ferromagnetism
Substrates have played an important role in material synthesis [4,93,94,[105][106][107][108]. ZnO TFs have been grown on single-crystalline substrates by MOCVD [93]. M-H curves show that ZnO TFs on sapphire substrate are non-ferromagnetic ( Figure 12A,B). However, the structural defect located on the substrate-film interface can cause RTFM in ZnO TFs. Interestingly, different substrates have been used to fabricate ZnO TFs with the same thickness by PED [94], and their ferromagnetic behaviors are very different. The ZnO TFs on silicon substrate show stronger ferromagnetism than the others, because more V O can be created in the TFs [94]. The origin of RTFM may be V O .
ZnO TFs grown on sapphire substrates have been synthesized by PLD [105]. Only the samples on R-plane sapphire substrate exhibit RTFM. Furthermore, the results suggest that V Zn is probably the reason for the RTFM. In addition, quartz, glass, and silicon have been selected as substrates to prepare ZnO TFs by PLD, and the RTFM in the samples may come from the coupling of unpaired electron spins induced by the oxygen 2p orbitals around the V Zn and are enhanced by the in-plane compressive strain [106].
The modulation of RTFM by changing the substrate temperature has also been studied in ZnO TFs [107]. However, ZnO films grown at low substrate temperatures exhibit a larger M s . M s decreases with increasing the temperature. A similar result is found by Xu et al. [108]. Notably, the effect is different from that in TM-doped ZnO TFs [4].

Effect of Growth Conditions on RTFM
The RTFM can also be mediated by growth temperature. In 2012, Xu et al. [52] found that ZnO QDs with different sizes exhibited an obvious RTFM. Furthermore, the surface-volume ratio was closely related with the size-dependent ferromagnetism.
The size of QDs can also be controlled by milling time, which affects RTFM of QDs. Furthermore, it has been found that milling time is a superior way to mediate RTFM [49,57,58]. The grain size does not change monotonously with milling time, but M s will increase monotonously in a certain range. In fact, V Zn can induce RTFM in ZnO QDs.
Similarly, the grain size can be changed by milling time, as reported by Kisan et al. [57]. Interestingly, the transition from PM to FM is realized by this strategy. UV-vis spectra [57] and positron annihilation spectra [109] show that V O can affect the RTFM.

Influence of Post-Annealing on RTFM
Zhan et al. [110]. observed that RTFM can be produced by annealing in Ar. The magnetic field strength will affect the observed RTFM. When a magnetic field of 7T is applied, M s is as high as 2.7 emu/g. Furthermore, PL spectra indicate that RTFM originates from V O + . Undoped ZnO TFs with RTFM have been obtained by PAD [89]. The magnetization of the sample can potentiate by annealing in O 2 or H 2 gas. Positron annihilation spectroscopy indicates that V Zn can induce the RTFM. In addition, we have also fabricated ferromagnetic Zn 0.97 Co 0.03 O TFs by PAD [90]. The O i and V Zn can be mediated with the annealing. The PL spectra show that the RTFM is closely related to O i and V Zn .
Changing annealing temperature is another strategy to mediate RTFM [53,54] The H c , M r and M s increase with decreasing the annealing temperature of the samples. PL spectra show that Zn i and V O related defects are closely related to the RTFM.
Interestingly, we have found that as-grown ZnO TFs can be transformed into centimeterscale ZnO horizontal NWs arrays [43] by annealing in O 2 . With the extension of annealing time, the length becomes longer. After ZnO NWs arrays have been produced, the magnetic properties of the films transform PM into FM.

Dopping with Non-Magnetic Atoms
Notably, doping is also an effective way to modulate the RFTM of ZnO. Many research groups have attempted to look for compounds with RTFM since 2000, which can be obtained with doping transition metal elements into metal oxides such as Cr [4,[111][112][113], Mn [2][3][4], Fe [114], Ni [115], Co [90,[116][117][118], Sc [4,119,120], Ti [120], V [120], and Cu [120,121], and codoping such as CoFe [122], and MnCo [123]. However, experimental studies on TM-doped ZnO have produced inconsistent results and the mechanism of RTFM remains unclear. This promoted search for DMS based on alternative dopants. Manifestation of RTFM has also been undertaken by doping with non-magnetic dopants such as H [26,[98][99][100][101][102][103][104][105][106][107][108][109][110]124,125], N [126,127], C [128][129][130][131], and S [47,78]. Interestingly, the observation of RTFM strongly suggests the presence of a spin-split band with a non-zero spin-orbit coupling in H-ZnO single crystals [100]. In addition, incorporation of N at 60 • in ZnO induces the stronger FM, higher band gap reduction, larger reduction in surface roughness and higher reduction in the grains at RT [126]. Carbon-doped ZnO films show an intrinsic RTFM, which originates from the Zn-C system in the ZnO environment [128]. Meanwhile, thiol-capped nanotube arrays have exhibited a higher M s than the nanowire arrays due to its larger surface-to-volume ratio and thus higher density of Zn-S bond spins. The observed dependence of the magnetic anisotropy of ZnO NTs haves suggested that the change in the ordering of Zn-S spins on the crystal facets can affect the extent of magnetic ordering and alter the preferred magnetization direction [78].

Capping
Capping organic molecules is a common method to mediate RTFM of non-magnetic materials. In 2007, Garcia et al. [36] controlled the ferromagnetism by capping ZnO NPs with three different organic molecules. The band structure was mediated with capping molecules, and then the RTFM was induced from DM.

Conclusions and Outlook
The present literature analysis reveals that the amount of research focused on undoped ZnO nanostructures with RFTM has been increasing over the past two decades. We have successfully prepared ferromagnetic nanostructures, including thin films, horizontal arrays and vertical arrays. RTFM could be readily modulated by non-doping means including precursor selection, substrate effects, growth conditions, post-annealing, and capping. Some novel ways to drive the transition from DM to FM in undoped ZnO nanostructures have delivered interesting results, as shown in Figure 12.
The size-dependent RTFM in ZnO QDs/NPs ( Figure 3) can be observed [52]. M s is increased gradually by decreasing the diameters of ZnO QDs/NPs. However, in the nanowire system, the change of diameter has no obvious effect on RTFM (Figure 7). Interestingly, RTFM can be induced from non-magnetic ZnO nanosheets and TFs by reducing the thickness (Figures 10 and 12). Notably, the magnetic anisotropy of ZnO nanostructures should be further studied.
Through theoretical calculation, it has been found that there is a driving force of phase transformation in ZnO. Therefore, fabrication of ultrathin (thickness < 2 nm) freestanding ZnO structures remains challenging. Amazingly, Taniguchi et al. [45] have reported that organic layers could allow the formation of ZnO nanosheets (~1.5 nm) with RTFM. In addition, Zhao et al. [132]. have presented the phase transitions of quantum-confined ZnO NWs by in-site TEM. Experimentally, it is still a great challenge to study RTFM in quantum-bound ZnO nanostructures.
Even though a variety of magnetic nanostructures have been prepared and studied, there are still many problems worthy of further study. (1) Development of new ferromagnetic nanostructures, such as nano-octagonal stars [133]. In the past, the morphology characterization mainly focused on 2D characterization. Recently, 3D electron tomography has emerged to analyze the 3D structure of nanostructures. In addition, the self-assembly of nanostructures and the evolution of ferromagnetism were further studied by an in situ method [133,134]. (2) Development of some emerging methods to prepare and study the nanostructure. (3) New characterization methods have been employed to investigate the nanostructures with RTFM [135,136]. The crystal nucleation process of FePt NPs in four dimensions [135] has been studied by atomic electron tomography (AET) [135]. In addition, electron tomography (ET) has revealed FeCo nano-octopods with RTFM. (4) The mechanism of undoped ZnO with RTFM needs further study.
Author Contributions: Writing-original draft preparation, H.R.; writing-review and editing, H.R. and G.X.; supervision, G.X. All authors have read and agreed to the published version of the manuscript.

Conflicts of Interest:
The authors declare no conflict of interest.