The growth of metastable fcc Fe78Ni22 thin films on H-Si(100) substrates suitable for focused ion beam direct magnetic patterning

We have studied the growth of metastable face-centered-cubic, non-magnetic Fe78Ni22 thin films on silicon substrates. These films undergo a magnetic (paramagnetic to ferromagnetic) and structural (fcc to bcc) phase transformation upon ion beam irradiation and thus can serve as a template for direct writing of magnetic nanostructures by the focused ion beam. So far, these films were prepared only on single crystal Cu(100) substrates. We show that transformable Fe78Ni22 thin films can also be prepared on a hydrogen-terminated Si(100) with a 130-nm-thick Cu(100) buffer layer. The H-Si(100) substrates can be prepared by hydrofluoric acid etching or by annealing at 1200{\deg}C followed by adsorption of atomic hydrogen. The Cu(100) buffer layer and Fe78Ni22 fcc metastable thin film were deposited by thermal evaporation in an ultra-high vacuum. The films were consequently transformed in-situ by 4keV Ar+ ion irradiation and ex-situ by a 30 keV Ga+ focused ion beam, and their magnetic properties were studied by magneto-optical Kerr effect magnetometry. The substitution of expensive copper single crystal substrate by standard silicon wafers dramatically expands application possibilities of metastable paramagnetic thin films for focused ion beam direct magnetic patterning.


Introduction
Magnetic nanopatterning plays a central role in the development of novel devices and magnetic metamaterials with properties unattainable in bulk systems. The patterning is conventionally achieved by using optical or electron-beam lithography in combination with lift-off and etching procedures.
However, these methods do not allow, for example, for continuous spatial changes in magnetic properties. The resulting structures usually have sharp magnetic-nonmagnetic transitions leading to, e.g., highly localized demagnetizing fields, or stochastic distribution of pinning sites that cannot be controlled merely by the fabrication process itself. An alternative to the traditional lithography is a fabrication of magnetic elements by ion beam [1]. The ion beam can modify the magnetic properties of a material by alloying [2], intermixing [3] or by a structural change [4]. We have shown that it is possible to embed ferromagnetic body-centered-cubic (bcc) nanostructures in epitaxial paramagnetic facecentered-cubic (fcc) films with a 1 keV Ar + ion beam [4,5]. The system used in these works was an fcc Fe grown on a Cu(100) substrate [6,7]. The ion-beam induces a structural and magnetic phase change by overcoming the potential barrier between the fcc local minimum and the bcc global minimum and we, therefore, call these fcc films metastable. The metastable Fe/Cu(100) spontaneously transforms to bcc phase at 10 ML [8], but this limit can be shifted to 22 ML by dosing CO [9] and even further by alloying Fe with 22% of Ni [10]. We have shown that it is possible to fabricate magnetic nanostructures in these films by ion-beam irradiation using proximity masks [5] or focused ion beams (FIB) and that by suitably chosen FIB patterning procedures it is possible to tune saturation magnetization and also the magnetic anisotropy of the transformed structures [11].
Metastable Fe78Ni22/Cu(100) is an excellent system for a one-step fabrication of magnetic nanostructures, but the copper single crystal is an expensive substrate for use in future applications.
Here we show, that it is possible to use also standard Si(100) wafers as a base substrate for this system.
Silicon has been the most commonly used material in the semiconductor industry and scientific research for decades, not only for its electronic properties but also because of the well-known processes for preparation of very well defined substrates with almost perfect crystallographic properties. Our experiments indicate that the Fe does not grow epitaxially on Si(100), but it is possible to grow a Cu(100) buffer layer on Si(100) [12][13][14]. The Cu buffer layer grows in the desired (100) crystallographic orientation when the Si(100) surface is unreconstructed, hydrogen-terminated (H-Si), which is most commonly achieved by etching in hydrofluoric acid (HF) [12][13][14][15][16][17]. The lattice constants of Cu and unreconstructed Si lead to a significant (33.5%) lattice mismatch; however, rotation by 45° decreases the lattice mismatch to 6%.
Additionally to the known HF etching procedure, we have also used an ultrahigh vacuum (UHV) alternative, which requires annealing of Si at 1200°C followed by adsorption of atomic hydrogen. The growth of the 130 nm thick Cu(100) buffer layer on H-Si substrates prepared by both procedures was tested together with the subsequent deposition of the metastable fcc Fe78Ni22 films. The transformation to the bcc phase was performed both, in-situ by a weakly focused ion beam and ex-situ by the FIB.

Measurement of the magnetic properties of the transformed thin films and nanostructures performed by
Kerr magnetometry show that Fe78Ni22/Cu(100)/H-Si(100) is a viable alternative to systems prepared on Cu(100) single-crystal substrates.

Experimental
The UHV system used for the experiments has a base pressure of 7×10 -11 mbar (measured by a Bayard-Alpert ionization gauge) and is equipped with a three-pocket e-beam evaporator (Focus EFM3T), a single pocket e-beam evaporator (Focus EFM3) and a Knudsen effusion cell (CreaTec). The samples can be cooled by liquid nitrogen to 100 K and heated by e-beam bombardment of the sample holder plate to 1000 K. Auger electron spectroscopy (AES) was used to check the cleanliness of the substrates and the composition of grown films. The AES spectra have been normalized to the minimum of the average peak-to-peak height of the dominant peak. Presented concentrations correspond to quantitative analysis using relative elemental sensitivity factors [18]. The structure of the surfaces was measured by Low Energy Electron Diffraction (LEED). The LEED images were post-processed with a dark-field subtraction, flat-field normalization [19] and inverted.
The experiments were performed on B-doped (p-type) Si with a resistivity of 5-20 Ωcm. The dimensions of the samples were 3×12×0.4 mm 3 . Two procedures were applied to obtaining a Hterminated surface. The chemically treated Si was etched for 2.5 minutes in 10% HF to remove the native oxide and to terminate the Si with hydrogen [17], then rinsed for 1 minute in high-purity (Merck Milli-Q) water, dried with argon gas and transferred into a loadlock connected to the UHV chamber within 10 minutes after etching. The H-terminated Si should be inert to the ambient atmosphere for such a period [17]. Then, the sample was outgassed in the UHV chamber at 100°C for 30 minutes. After this procedure, the sample cleanliness was checked by AES and the surface reconstruction by LEED.
In the UHV procedure, Si samples were heated by direct current (DC) heating in a home-built heating stage. Target holders were made of Mo; their design was based on the Omicron sample plates for DC heating. After introducing the samples into UHV, we outgassed the sample holder by heating it to 600 °C for approx. 20h. We then heated the samples by DC to 600 °C for approx. 20 h until the base pressure in the chamber was restored. After the outgassing phase, we annealed the samples repeatedly at 1200 °C by DC for 5 seconds with a 5-second ramp from the outgassing temperature. The highest pressure during the last annealing step was kept below 5×10 -10 mbar. With this approach, we were able to completely remove both the native oxide and also any organic impurities. The H termination was achieved by a home-built H-cracker based on the design of Bischler [20]. A tungsten capillary with a 0.6 mm inner diameter was heated by 1 keV e-beam bombardment to approx. 1800 °C, which completely dissociated the H2 flowing through it [21]. The end of the capillary was approx. 3 cm from the sample and a liquid-N2-cooled Cu plate between the W tube and the sample limited the sample heating to approx. 1 °C/min. The sample was exposed to atomic hydrogen (1×10 -6 mbar H2 backpressure) for 7 mins to achieve a complete H termination. Again, the sample cleanliness was checked by AES and the surface reconstruction by LEED.
Cu was evaporated from two sources, the EFM3, and the effusion cell. The material of the effusion cell crucible contained a small amount of Ca contamination (confirmed by Secondary Ion Mass Spectrometry) which was detectable also in the deposited layer and which turned out to be an essential surfactant needed to stabilize the growth of the Cu buffer layer in the required (100) orientation on UHV treated samples. The temperature of the sample increased by 10 °C via radiation heating during the deposition from the effusion cell. The pressure during the deposition was 1×10 −10 mbar (with the help of a liquid-N2-cooled cryo baffle and a titanium sublimation pump). Deposition rates were calibrated with a quartz crystal microbalance at the position of the substrate, and the deposition was done at room temperature (RT) unless mentioned otherwise. The deposition rate of both Cu evaporators was 0.06 Ås -1 (approx. 5.8 h for 130 nm). The Fe 78 Ni 22 layers were evaporated by the EFM3T from a rod with a 2 mm diameter (MaTecK). A repelling voltage of +1.5 kV was applied to a cylindrical electrode (flux monitor) in the orifice of the evaporator to suppress high-energy ions, which may modify the growth mode of the films [22]. The base pressure during the deposition was 8×10 −11 mbar, which was artificially increased to 5×10 −10 mbar of CO to stabilize the fcc phase in line with previous observations [10]. The deposition rate of Fe78Ni22 was 0.02 Ås -1 (approx. 1h for 8 nm). After each deposition step, the surface composition and the crystallographic structure were measured by AES and by LEED (respectively).
Large-Area irradiation of one set of samples was performed in-situ by an ion gun (Specs) equipped with a Wien filter by scanning the sample with a time-averaged ion flux of approx. 10 13 cm -2 s -1 . To maximize the transformation rate, we used 4 keV Ar + ions, which penetrate the whole Fe78Ni22 layer at perpendicular incidence and do not cause significant Fe-Cu intermixing [23]. During the ion beam irradiation, we periodically measured magnetic hysteresis loops by a home-built Surface Magneto-Optical Kerr Effect (SMOKE) apparatus. The SMOKE experiment can be performed in longitudinal or polar geometry (angle of incidence 60° or 30°, respectively, with a spot size of approx. 1 mm). The plane of incidence and the direction of the magnetic field were parallel to the (010) plane of the Si substrate.
To study the potential of the metastable films for magnetic micro-and nanopatterning, we performed a local ion-beam-induced transformation in a high vacuum (approx. 10 -7 mbar) chamber of a focused ion beamscanning electron microscope (FIB-SEM) system (Tescan LYRA3). A series of 3×3 μm 2 squares was irradiated with the Ga + FIB at 30 keV with a spot size of 20 nm and a beam current of 40 pA with an increasing ion dose. We varied the irradiation dose between individual areas and then measured the patterns by SEM, Kerr microscopy and micro-focused Kerr magnetometry [24].

Results and discussion a. Surface structure and stoichiometry
The chemically prepared Si(100) samples had a (1×1) diffraction pattern after introduction into the UHV and mild annealing, which is shown by a green square in Figure 1a). The unetched, UHV treated Si(100) showed (2×1)-reconstructed domains after annealing to 1200 °C [blue rectangles in Figure 1d)] which changed into a (1×1) after termination with atomic H [green square in Figure 1e)]. The change in reconstruction confirms that the atomic H had saturated all the Si dangling bonds [25,26]. The (1×1) diffraction pattern, therefore, means that both approaches are potentially suitable for the growth of epitaxial Cu(100). Si with a native SiO2 is still slightly contaminated by C (black lines in Figure 2), and after annealing, we observe a complete removal of the native oxide and carbon contamination (compare black and red lines in Figure 2). The H termination does not introduce any impurities (orange lines in Figure 2).   conclusion that the Cu grows as epitaxial islands with terraces wide only a few nm, as already shown by Mewes et al. [14]. Deposition assisted with ion bombardment from the sputter gun did not bring any significant difference in the as-grown films. Deposition at lower or higher temperatures (-20 °C, > 50 °C) led to a growth of a polycrystalline fcc (111) film, in line with previous observations [29,30].
Post-annealing did not lead to flattening of the film; in fact, we could observe a signal from Si in AES while heating to 150 °C for two hours, in contrast to Lukaszew et al. [31]. The Si signal in AES after post-anneal showed a splitting in energy, which corresponds to the formation of bulk copper silicides, i.e., hybridization between the Si 3p states and the Cu 3d states [16]. that the Fe78Ni22 and the buffer layers beneath have similar morphology and that the structure of the metastable film is also fcc (100). As the films were deposited with a CO background pressure, the surface of the metastable films had 8% C and 5% O measured by AES, as shown by the green line in Figure 2a) and by the detail of the C-peak in Figure 2d). The C and O peaks arise from the dissociation of the CO in which the films are deposited, as described elsewhere [9]. The combination of LEED and AES demonstrates that it is possible to grow epitaxial films of Fe78Ni22(100) on both H-Si(100) prepared by wet chemistry or in UHV.

b. In-situ magnetic transformation by broad beam irradiation
The experiments presented in Figures 3 and 4    where it reaches a maximum and then decreases as the magnetic thin film is sputtered away and intermixes with the Cu layer below. The ion dose at which we reach the maximum transformation is equal to the transformation dose for films grown on a Cu(100) single crystal [10]. The Kerr ellipticity of the transformed film is 10% lower than in the case of the films deposited on a Cu(100) single crystal [10], which we attribute to the corrugation of the metastable film arising from the underlying Cu buffer layer.
The coercivity [blue line in Figure 3b)] starts increasing at the same ion dose as the magnetic saturation (around 2×10 14 cm -2 ) to reach a maximum at 2×10 15 cm -2 . It then falls rapidly to its minimum, which matches the maximum of the magnetic saturation at 6×10 15 cm -2 . From observations on Cu (100) single crystal substrate we know that increasing the ion dose increases the number and size of bcc nuclei randomly dispersed in the fcc layer. The maximum in the coercive field corresponds to the nuclei reaching the maximum size for single magnetic domains. This property is well known from magnetic nanoparticle studies [33]. Further lowering of the coercive field is attributed to multi-domain states of magnetic particles and more efficient interaction in between the nanoparticles through the stray field of individual particles. Figure 4 shows the results of patterning the metastable films with the FIB. The grey level in SEM (inset at the bottom of Figure 4) allows for distinguishing the as-deposited (grey) areas and the irradiated (darker or brighter) areas and serves as a valuable pre-characterization tool. The irradiated areas become darker at a dose of 5×10 13 cm -2 but brighten again at doses above 10 16 cm -2 . The reason for the decrease of the SEM signal is the fcc → bcc structural change, which affects electron channeling and associated secondary electron emission [11]. The last square then shows the signal from the Cu buffer layer after a complete removal of the Fe78Ni22 film. The data from Kerr magnetometry (Figure 4) show the onset of the magnetic transformation at an ion dose of 1×10 14 cm -2 . The maximum of the magnetization is achieved at ion doses 1×10 15 cm -2 -3×10 15 cm -2 . The transformation by 30 keV Ga + ions qualitatively shows the same behavior as the transformation performed in-situ [see Figure 3b)], but the ion dose needed to achieve full transformation is lower for the Ga + . We attribute this difference to the higher cross section for creation of recoils with high enough energy to induce the thermal spike [4]. For example, according to SRIM [23], Ga + create 4 times more 1 keV recoils than Ar + as they travel through the metastable film.

c. Focused ion beam magnetic patterning
The inset in Figure 4 shows the hysteresis loops of the squares irradiated by low ion dose (black line) and at the maximum magnetization (red line). The squares irradiated by low ion dose do not show any FM signal, which is in contrast with the measurement by SMOKE in Figure 3 (black line). We assume this is because the SMOKE collects signal from a large area, and the small fraction of the film which is already in the bcc ferromagnetic (FM) phase is not recognizable by the Kerr magnetometer.

Conclusion
We have shown that it is possible to grow metastable, epitaxial fcc Fe78Ni22 on H-Si(100) with a Cu(100) buffer layer, and that using a focused ion beam, we can create magnetic micro-and nanostructures with tuneable magnetization. The lower size limit of such nanostructures is given by the focus of the ion beam and the size of the bcc nuclei [5]. The magnetic properties of the films are comparable to the properties of Fe78Ni22 films prepared on Cu(100) single crystal substrates [11]. We Metastable films for ion-beam-induced magnetic transformation present a promising system for fabrication of magnetic metamaterials, such as magnonic crystals. They are an alternative to standard lithography approaches, and the possibility to use a standard substrate such as Si(100) is undoubtedly an essential step towards applications in rapid prototyping of magnetic metamaterials (by using FIB) and in suitability for mass production (by ion irradiation through a mask).