Background

Fe3O4 has been extensively studied because of its high Curie temperature, and because it has been predicted to have a half metallicity and a full spin polarization [1, 2], these properties make Fe3O4 a promising material for applications in data storage and spintronic devices such as memories or magnetic sensors [3]. Epitaxially grown Fe3O4 nanofilms have been extensively investigated and have been reported to contain a type of natural growth defect, namely, antiphase boundaries (APBs) caused by strains on the film from mismatch between the substrate and the film [4, 5]. The presence of APBs leads to unusual magnetic properties of epitaxial Fe3O4 films, such as nonsaturation of magnetization in high magnetic fields [6]. Moreover, an epitaxial Fe3O4 nanodot array is a possible candidate to meet the requirements of ultrahigh-density storage and reduce the size of spintronic devices. In addition, the magnetic property dependence on nanodot size is worth investigating. However, to our knowledge, up to now, no studies on epitaxial Fe3O4 nanoarrays have been reported.

One popular top-down method, namely, the lithography technique, has been used to pattern films into nanoarrays [7, 8]. However, this top-down method has a complex procedure and is cost prohibitive for mass production. In addition, covalent oxides of Fe3O4 being fragile are easily damaged during etching. In another technique, Fe3O4 nanoparticles are synthesized first by chemical methods and then self-assemble on a substrate to form nanodot arrays [9, 10]. However, in this case, the particles have random orientations, easily aggregate, and have weak bonding force with the substrate, which make the properties quite different from those of epitaxial nanodot arrays.

Recently, porous anodized aluminum (PAA) templates have been used to fabricate large areas of metal and oxide nanodot arrays because of their low production cost, pore size controllability, and ease of fabrication, which is called the bottom-up method [11–15]. The PAA-based method is a direct approach for growing epitaxial nanodot arrays, in which the atoms passing through the pore directly arrive at the substrate. After removing the PAA template, an epitaxial growth of an ordered nanodot array is obtained.

In this paper, we used the bottom-up patterning technique, precisely controlled the fabrication parameters, and ultimately obtained large-area epitaxial Fe3O4 nanodot arrays on a single crystal substrate of SrTiO3. The magnetic property dependence on dot size and the morphology of nanoarrays were investigated and subsequently compared with their corresponding films. The density and size of the nanodots controlled by the PAA template could tailor magnetic properties including coercivity (Hc), squareness, magnetic anisotropy, and Verwey transition temperature (Tv).

Methods

Ultrathin PAA templates with various pore diameters and inter-pore distances were prepared through a typical two-step anodization process [16, 17]. After the second anodization, the top surface of the PAA template was spin coated with a layer of polymethylmetacrylate (PMMA). The aluminum and thin nonporous barrier layer were then removed by immersing the template in an acid mixture of CuCl2 and HCl. A thin PMMA layer prevents the ultrathin ceramic membrane from undergoing mechanical deformations, such as folding, cracking, or ripping. After complete removal of aluminum, the template was transferred onto the SrTiO3 substrate and immersed in C3H6O at 60°C to dissolve the PMMA coating. Prior to the removal of PMMA, the templates were further thinned by immersing in H3PO4 for different durations to obtain the desired template thickness. In this work, three sizes of the PAA template with the same thickness of 200 nm were fabricated. The average pore diameter and inter-pore distance of the templates were (a) approximately 38 and approximately 67 nm, (b) approximately 67 and approximately 115 nm, and (c) approximately 137 and approximately 160 nm, respectively.

Fe3O4 was subsequently deposited on the PAA/SrTiO3 (001) substrate by laser molecular beam epitaxy (LMBE) using a KrF excimer laser (λ = 248 nm) with a repetition rate of 10 Hz and energy of 300 mJ on a ceramic target of Fe3O4. The PAA template was then removed using NaOH solution at 35°C, so that large-area Fe3O4 nanodot arrays were obtained. The substrate temperature was chosen between 400°C and 700°C, and the chamber oxygen pressure was 10−4 Pa. Some samples were annealed at 700°C in situ for 2 h before cooling to room temperature.

The crystal structure was characterized by X-ray diffraction (XRD) using Cu Kα radiation (λ = 0.15406 nm). The morphology of the PAA templates and nanodot arrays was investigated by scanning electron microscopy (SEM) and atomic force microscopy (AFM). Magnetization measurements were performed by a superconducting quantum interference device magnetometer (SQUID).

Results and discussion

Ultrathin PAA templates with pore diameters of 38, 67, and 137 nm were fabricated. Figure 1a shows a well-ordered array of circularly shaped holes with a pore diameter of 67 nm. Figure 1b is the cross section of this PAA template, with a thickness of approximately 200 nm. The high-quality ultrathin PAA template is the key in preparing epitaxial Fe3O4 nanodot arrays. When the PAA template was partially removed, the Fe3O4 nanodot arrays and the cover of the PAA template were clearly observed, as shown in Figure 1c. The resulting nanodot arrays have respective dot diameters (D) and inter-dot distances (Dint) of approximately 40 and approximately 67 nm (Figure 1d), approximately 70 and approximately 115 nm (Figure 1e), and approximately 140 and approximately 160 nm (Figure 1f). These D were a slightly larger than the pore diameter of PAA, which may be due to the diffusion of atoms during deposition and annealing. The nanodot density was estimated using the following equation [15]: . As shown in Figure 1d, the dot density can be as high as 0.18 Tb in.−2, using Fe3O4 nanodot arrays with D and Dint down to 40 and 67 nm.

Figure 1
figure 1

SEM images for the PAA templates and Fe 3 O 4 dot arrays. (a) Ultrathin PAA template, (b) cross section of the PAA, (c) Fe3O4 nanodot array together with a partially removed PAA template, and Fe3O4 nanodot arrays with various sizes - the average dot sizes and inter-dot periods are (d) approximately 40 and approximately 67 nm, (e) approximately 70 and approximately 115 nm, and (f) approximately 140 and approximately 160 nm, respectively.

Full size image

AFM images of the Fe3O4 nanodot array with D of approximately 70 nm and Dint of approximately 115 nm are shown in Figure 2. These images confirm that nanodots have a well-ordered hexagonal arrangement in agreement with the SEM results. Although the nanodots and the film were prepared using the same parameters, the average dot height of Fe3O4 is around 20 nm, which is slightly lower than the thickness (24 nm) of the film. Possibly, some of the atoms cannot arrive at the bottom of the pore, but rather are deposited on the surface of the pore wall and disappeared during the removal of the PAA templates.

Figure 2
figure 2

AFM images of Fe 3 O 4 dot array withD = 70 nm.

Full size image

The θ-2θ scans of the series of film and nanodot arrays deposited at different substrate temperatures (Ts), as well as the scans after annealing, are shown in Figure 3. The XRD patterns of the nanodot arrays without in situ annealing showed peaks that correspond to FeO (002), as well as the Fe3O4 (004) reflections. The intensity of the Fe3O4 (004) peaks increased, whereas the FeO (002) peaks decreased with an increase of Ts. These trends indicate that high substrate temperature favors the formation of the Fe3O4 (004) phase. The peak of FeO (002) disappeared after 2 h of in situ annealing at 700°C. In this condition, epitaxial growth of Fe3O4 nanodots with (004) orientation was obtained. The Fe3O4 (004) film could have epitaxial growth at a Ts of 600°C. Clearly, epitaxial growth of Fe3O4 nanodots is more difficult to achieve compared with that of the corresponding film. This difficulty may be attributed to the following reasons: (1) The edge of the nanosized pore may hinder deposition resulting in some of the atoms losing part of their energy so that atom hopping time and length are reduced, which is unfavorable for epitaxial growth. (2) The atoms enter the nanosized pores rather than on the flat substrates. Part of the atoms neighboring the wall of the pore obviously cannot move to the normal direction of the wall. This limitation also influences epitaxial growth. (3) The atoms in the pores may be separated by PAA from the activated oxygen atoms. Meanwhile, the energy for FeO formation was lower than that of Fe3O4[18], so part of the Fe3O4 may have been reduced. From the above analysis, we can conclude that with the aid of high Ts and in situ annealing, the deposited atoms which gained more thermal dynamic energy finally resulted in the good quality of Fe3O4 nanodot arrays. Following the optimal parameters of a 70-nm nanodot array, with D = 40, 140-nm-sized arrays all grow at Ts = 700°C and 2 h of in situ annealing at 700°C.

Figure 3
figure 3

XRD patterns for the film and dot arrays. The sample of D = 70 nm and film deposited at various substrate temperatures and after annealing. Except for the peaks of FeO (002) and Fe3O4 (004), the rest are peaks of the substrate.

Full size image

In-plane (H) and out-of-plane (P) room-temperature hysteresis loops of the Fe3O4 nanodot arrays and the film acquired are shown in Figure 4. The Hc, the squareness factor (S), and the saturation magnetization (Ms) of nanodot arrays and the film are shown in Figure 5. These show that the magnetic in-plane easy-axis anisotropy was observed, and the anisotropy is reduced with the dot size. Possibly, the magnetostrictive anisotropy was greatly reduced by the fast relaxation strain from the dot edges, whereas the shape anisotropy and crystalline anisotropy were released by forming noncontinuous dots [19]. As a result, the smallest nanodots exhibited a weak magnetic anisotropy.

Figure 4
figure 4

Hysteresis loops of Fe 3 O 4 film and dot arrays withD = 40, 70, and 140 nm. The H and P means the hysteresis loops of the samples measured at in-plane and out-of-plane to the film directions.

Full size image
Figure 5
figure 5

Variations in H c , M s , and S as functions of Fe 3 O 4 film and different nanodot diameters.

Full size image

The Ms of the film (310 emu/cm3) is smaller than that of the bulk (480 emu/cm3), which was attributed to the presence of APBs. The lattice constant (0.8398 nm) of the Fe3O4 film grown on SrTiO3 (0.3905 nm) is more than twice that of the substrate, indicating that the film undergoes a compressive strain. This strain induces some defects such as dislocations or APBs during film growth. Presences of APBs are the main reason for the lower saturation magnetic value compared with the bulk Fe3O4[20]. Moreover, the Ms of the nanodot arrays is smaller than that of the film. This characteristic could be explained by the accumulation of spin canting around the nanoparticle surface and the exchange coupling between the nanodots that reduce magnetic moment [21].

From Figure 5, the in-plane hysteresis loop of the film displayed an almost rectangular shape with S = 0.96 and a large Hc of 450 Oe, which is better than that reported for Fe3O4 films with comparable thickness [6]. The sample with D = 140 nm had a large Hc of 400 Oe and was easily saturated. The magnetization reversal may be dominated by the domain wall movement, similar to the film. However, the S of this sample is 0.50, which was half that of the film. The reduced S may be due to magnetization relaxation from dot edges. The sample with D = 70 nm exhibited smaller S = 0.28, large Hc of 300 Oe, and unsaturation at an external field of 2,000 Oe. At S = 0.05, Hc dropped to 50 Oe and unsaturation of the sample with D = 40 nm exhibited more or less soft magnetic properties similar to those of the bulk. The sample with D = 40 nm may reverse their magnetization by coherent rotation because of a single domain size of 38 nm for Fe3O4[22]. On the other hand, the sample with D = 70 nm with its relatively high Ms and Hc maybe in a transition state so that reversal is partly due to wall movement, impeded by defects, and rotation, which should be further investigated in the future.

Plots of field cooling-zero field cooling (FC-ZFC) magnetization versus temperature between 10 and 300 K are shown in Figure 6. A magnetic field of 50 Oe had been applied on the plane during measurement. The Tv obtained was around 120 K for the Fe3O4 film and arrays with D = 140 and 70 nm, close to that of the bulk Fe3O4[23]. However, Tv was undetectable at D = 40 nm. Normally, the presence of Tv is used as an evidence that the sample has a perfect stoichiometry of Fe:O = 3:4. The identity of the sample with D = 40 nm had been confirmed to be Fe3O4 based on the XRD pattern. The undetected Tv may be attributed to its small particle size and relatively large inter-particle spacing. The hopping energy barrier between different Fe sites within the nanodot was relatively small compared with the inter-nanodot tunneling barrier, so no Tv was observed in the sample [24].

Figure 6
figure 6

FC-ZFC curves of Fe 3 O 4 film and dot arrays withD = 40, 70, and 140 nm.

Full size image

Conclusions

In summary, well-ordered large-area arrays of epitaxial Fe3O4 nanodots were obtained on the SrTiO3 (100) substrate by LMBE at an elevated growth temperature (700°C) using ultrathin PAA templates as shadow templates. An Fe3O4 film was also prepared using LMBE. The AFM and SEM images showed that Fe3O4 nanodot arrays had large areas with well-ordered hexagonal arrays with D of 40, 70, and 140 nm; height of around 20 nm; and Dint of 67, 115, and 160 nm. The nanodot density can be calculated to be as high as 0.18 Tb in.−2 when D = 40 nm. The film or nanodot array exhibited magnetic anisotropy. The anisotropy of the arrays and the S weakened with decreasing dot size. Magnetization reversal may be dominated by the domain wall movement of the film and of the sample with D = 140 nm, while the sample with D = 40 nm may reverse their magnetization by coherent rotation because of its single domain size of 38 nm. On the other hand, the sample with D = 70 nm exhibited smaller S, large Hc, and unsaturation at the external field of 2,000 Oe. It demonstrated that the sample with D = 70 nm may undergo a magnetization reversal transition, which should be further investigated. The Tv values of the film and the samples with large dots (D = 70 and 140 nm) were close to those of the bulk, whereas the Tv of the small dot sample with D = 40 nm was undetected. Based on the above analysis, magnetic anisotropy, S, and Tv of Fe3O4 could be tailored by the morphology of nanodots. To the best of our knowledge, no study has shown epitaxial Fe3O4 nanodots to date. These results facilitate a deeper understanding of the micromagnetization inside nanostructures of Fe3O4, and the output could be a candidate for ultrahigh-density data storage and spintronic devices.