- Email: [email protected]
Atomic arrangement and magnetism of iron silicide on Fe(1 0 0) surface
Applied Surface Science 259 (2012) 166–171
Contents lists available at SciVerse ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Atomic arrangement and magnetism of iron silicide on Fe(1 0 0) surface T.T. Suzuki ∗ , S. Hishita National Institute for Materials Science, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan
a r t i c l e
i n f o
Article history: Received 30 March 2012 Received in revised form 2 July 2012 Accepted 3 July 2012 Available online 20 July 2012 Keywords: Spin-polarized ion scattering spectroscopy Surface magnetic structure Si/Fe interfaces
a b s t r a c t We investigated surface magnetic structure in the initial stage of Si deposition on an Fe(1 0 0) surface by spin-polarized ion scattering spectroscopy (SP-ISS). We found silicidation at the Si/Fe interface after Si deposition followed by annealing at 823 K. The silicidation occurs by the incorporation of silicon into the Fe substrate via the substitutional site of bcc Fe. After annealing, the incorporated Si atoms are distributed in surface layers several nanometers thick. The SP-ISS analysis revealed that the average magnetic moment of Fe in the silicide surface layer is about 70 % of that of Fe in the Fe(1 0 0) surface layer, whereas that of Si is almost zero. These surface magnetic moments are discussed in terms of the local magnetic environment. It is likely that the outermost surface of the silicide layer has an atomic arrangement similar to that of Fe3 Si(1 0 0) with surface termination by the Fe–Si plane. © 2012 Elsevier B.V. All rights reserved.
1. Introduction An Fe/Si interfacial reaction, i.e., silicidation, causes the formation of many stable and metastable Si–Fe compounds. These compounds exhibit a wide variety of electrical properties from semiconductor to metal and magnetic properties such as diamagnetism, paramagnetism, and ferromagnetism [1,2]. Therefore, tailoring the Si–Fe interfacial reaction has the potential to realize silicon-based optoelectronics and spintronics. These future devices are expected to be compatible with established and emerging Si technologies. For example, semiconducting iron disilicide ˇ-FeSi2 is a promising candidate for optoelectronics and photovoltaics because it acts as a Si-based light emitter as well as a photodetector. Photoluminescence and electroluminescence at a wavelength (∼1.5 m) relevant to optical fiber communications have already been demonstrated using ˇ-FeSi2 [3–5]. On the other hand, ferromagnetic Fe3 Si is a potential material for use in spintronics. A Schottky tunnel barrier source/drain made of Fe3 Si has been proposed to construct metal-oxide-semiconductor field effect transistors [6]. Spin injection in silicon through the Fe3 Si/Si Schottky barrier has already been demonstrated [7]. Silicidation at Fe/Si interfaces has been employed to grow iron silicides on Si substrates. Therefore, it is of central importance to elucidate the atomic arrangement and magnetic properties at the interfaces to fabricate device components of the abovementioned future technologies. However, although many studies have investigated silicidation at the interface, complex methods to produce a certain Si–Fe compound at the interface are still
∗ Corresponding author. E-mail address: [email protected] (T.T. Suzuki). 0169-4332/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.apsusc.2012.07.012
controversial [2,4,8]. For example, the identification of threedimensional iron-silicide islands on Si surfaces is still a matter of intensive discussion [9]. In particular, very little is known about the magnetic structure at the Fe/Si interfaces. Since spins at the interfaces typically characterize the principle of a spintronics device, such as the tunneling magnetoresistance effect [10], it is essential for the development of spintronics to analyze the magnetic structure of a ferromagnetic atomic layer in contact with a nonmagnetic layer at the interface. In this context, spin polarization of a single Si (–Fe) layer induced by the magnetism of the Fe layer at the Si/Fe interface is informative. To the best of our knowledge, however, no study on the magnetic structure of the single atomic layer has been reported for Fe/Si or Si/Fe interfaces. In the present study, we investigated the magnetic structure of an Fe(1 0 0) outermost surface deposited with silicon by spinpolarized ion scattering spectroscopy (SP-ISS). We concentrated on a certain silicide phase formed at the Fe surface. The silicide phase was grown with good reproducibility after Si deposition followed by annealing in an ultra-high vacuum (UHV), i.e., solid phase epitaxy. We investigated the atomic arrangement, spin arrangement, and spin polarization at the Fermi level of this silicide phase. Most of studies on the silicide at the Fe/Si interface have prepared the sample in the opposite way: Fe deposition on Si surfaces followed by annealing. The magnetic structure analysis of the Si/Fe outermost surface in the present study should provide important clues of the magnetic structure and its asymmetric nature at the Fe/Si and Si/Fe interfaces [11,12]. Details of SP-ISS have been previously described [13]. Briefly, electron spin-polarized 4 He+ ions are projected on a sample surface, and the kinetic energy of the scattered ions is analyzed in SP-ISS measurements. The kinetic energy of the incident He+ ions is of the order of keV, and therefore, most incident He+ ions are
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171
neutralized at the sample surface, typically via Auger neutralization (AN). Because SP-ISS detects the scattered He+ ions that survive efficient AN, it has extreme surface sensitivity [14]. According to the Pauli exclusion principle, the scattered He+ ion intensity should differ depending on whether the He+ spin is parallel to majority or minority spins of the magnetic surface in the AN process. Thus, SP-ISS detects the spin polarization of the target atom on the outermost surface with element selectivity. Moreover, spin polarization at the Fermi level is estimated from the spin dependence of electron emission with the He+ ion impact on the surface (spinpolarized ion neutralization spectroscopy (SP-INS)) [13]. In SP-ISS and SP-INS studies, the spin dependence of the scattered ions and emitted electrons are discussed in terms of spin asymmetry. The spin asymmetry is defined as (I↑ − I↓ )/[PHe+ · (I↑ + I↓ )], where PHe+ expresses the spin polarization of the incident He+ ion beam, and I↑ and I↓ denote the scattered ion intensity for SP-ISS and the emitted electron intensity for SP-INS with the projectile He+ ions polarized parallel and anti-parallel to the majority spins, respectively.
167
Fig. 1. A series of ISS spectra obtained on the Fe(1 0 0) surface with the Si deposition. The spectra are normalized by the incident ion beam current at the sample. The ˚ by annealing bottom spectrum was obtained after the Si deposition of 22 Afollowed at 823 K for 5 min. The incident energy was 1.77 keV, and ˛ and were 0◦ and 150◦ , respectively. Arrows indicate the binary collision energies for some elements.
2. Experimental method and setup Experiments were performed in an UHV chamber (base pressure of 5 × 10−11 Torr) equipped for SP-ISS, reflection high energy electron diffraction (RHEED), and deposition of iron and silicon. We grew Fe(1 0 0) films on MgO(1 0 0) single-crystalline substrates. We prepared the MgO(1 0 0) clean surface by annealing the mirror polished MgO substrate (10 mm × 10 mm × 1 mm) above at 1300 K by electron bombardment in UHV. Then, we deposited iron at room temperature, followed by annealing at about 700 K for 10 min. Under these growth conditions, bcc Fe films grew epitaxially with the orientation relationships Fe(1 0 0)//MgO(1 0 0) and Fe[1 1 0]//MgO[1 0 0] [15]. The thickness of the iron film was about ˚ We deposited silicon on the Fe(1 0 0) film by using an elec300 A. tron beam evaporator (Omicron EFM3) at room temperature in UHV, and subsequently, we annealed the Si/Fe(1 0 0) sample at elevated temperatures. The sample was pulse-magnetized in-plane by a retractable pulse magnet placed in the UHV chamber before SPISS measurements. The magnetization direction was parallel to the Fe[0 0 1] easy axis of the bcc-Fe films. Electron spin-polarized 4 He+ ions were generated from Penning ionization of spin-polarized metastable He 23 S1 atoms (He*) [16]: He∗ + He∗ → He+ + He0 + e− . We employed an optical pumping (OP) technique, using the D0 line of 1083 nm radiation (He* 23 S1 → 23 P0 ), for spin polarization of He*. The spin angular momentum is conserved in Penning ionization; thus, if He* is spin polarized by OP, spin-polarized He+ ions are generated [17]. The OP radiation was provided by an ytterbium-doped fiber laser (Keopsys KPS-BT2-YFL-1083-FA) via an ytterbium-doped fiber amplifier (Keopsys KPS-BT2-YFA-1083200-COL). The wavelength of the 1083 nm seed light from the fiber laser was precisely tuned to the D0 line. The spin directions of He+ ions (up or down) were controlled by the polarization of the OP radiation. The spin polarization of the He+ ion beam PHe+ was about 0.2. The entire apparatus was surrounded by a three-axis coil to compensate for the Earths magnetic field. An additional coil produced a weak guiding field (∼0.3 Oe), which was perpendicular to the scattering plane. Thus, the spin direction of the incident He+ ion beam was defined by the guiding field, which was parallel to Fe[0 0 1]. The scattering plane was perpendicular to the Fe(1 0 0) surface (the inset of Fig. 1). Both the scattered He+ ions and the emitted electrons were measured using a rotatable hemispherical sector analyzer (Omicron SHA50). The measurements were conducted in a constant pass
energy mode with a pass energy of 318 eV for SP-ISS and 40 eV for SP-INS. 3. Results and discussion Fig. 1 shows ISS spectra obtained at the Fe(1 0 0) surface deposited with silicon. The bottom spectrum was obtained after ˚ the Si deposition of 22 Afollowed by annealing at 823 K for 5 min. The incident energy was 1.77 keV, and the incident angle measured from the surface normal (˛) and the scattering angle () were 0◦ and 150◦ , respectively. The binary collision energies of the projectile He+ ion with iron (1354 eV), silicon (1036 eV), and oxygen (682 eV) are indicated by arrows at the bottom of the spectra. In Fig. 1, the iron peak decreases with the Si deposition at room ˚ It is temperature; it almost vanishes with the Si thickness of 22 A. noted that the Fe peak position shifts to the low energy side with the Si deposition. This is due to the increase of inelastic scattering components, such as electronic-states excitation and reionization [18]. Such inelastic scattering processes occur in the Si overlayers on iron. Accordingly, the ISS peak reflects the composition of several surface layers. The Fe peak reappears after annealing; hence, the outermost surface is composed of both iron and silicon. The reappearance of the Fe atoms at the outermost surface is consistent with the shift of the Fe peak to the high energy side with annealing. In our RHEED observation, the 1×1 streak on Fe(1 0 0) changed into the diffusive 1×1 pattern with the Si deposition, indicating the disordered structure of √ the Si thin-film (not shown). We also √ observed that the ( 2 × 2) R45◦ pattern appears after annealing at 823 K, although it was not a sharp pattern. We observed no transmission pattern through our experiments. This, together with the ISS results in Fig. 1, shows possibility of the silicidation on the surface after annealing. Further increase of the annealing temperature did not essentially change the RHEED pattern but substantially reduced the Si peak intensity in the ISS spectra. The Si peak in the ISS spectra almost vanishes after annealing at 1173 K for 5 min, and accordingly, the 1×1 pattern reappeared in RHEED. Since we are interested in the silicide phase on the Fe substrate, we will, hereafter, concentrate on the abovementioned silicide formed by the Si deposition followed by annealing at 823 K. Fig. 2 shows (a) ˛ and (b) azimuthal angle scans of the scattered He+ intensity with the incident energy of 1.77 keV and of 150◦ . The peak intensity of iron corresponds to integrated counts between 1240 and 1400 eV, while that of silicon corresponds to the counts between 920 and 1100 eV obtained by subtracting the background of the Fe peak estimated by a linear function. In the ˛
168
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171
Fig. 3. A series of ISS spectra obtained on the Si/Fe(1 0 0)/MgO(1 0 0) surface with 2 keV Ar+ ion beam irradiation. The sample was prepared by the Si deposition of ˚ Fe(1 0 0)/MgO(1 0 0) followed by annealing at 823 K for 5 min. In ISS, the 20 Aon incident energy was 1.71 keV, and ˛ and were 0◦ and 150◦ , respectively.
Fig. 2. ˛ scan (a) and azimuthal angle scan (b) of scattered He+ ion intensity with the incident energy of 1.77 keV and of 150◦ . The samples were the Fe(1 0 0) clean sur˚ deposited face (red open circles) and the Fe(1 0 0) surface on which silicon of 20 Awas followed by annealing at 823 K for 5 min (black solid squares). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
scan measurement, ˛ was varied along the Fe[0 1 0] azimuth and the scattering plane was parallel to Fe(0 0 1) as shown in the inset of Fig. 2 (a). The Fe intensity variation shown in Fig. 2 (a) exhibits a sharp increase at 68◦ . This is due to the focusing effect on the Fe atoms located at the atomic layer of the outermost surface. The critical angle of this focusing effect is explained from the bulk termination structure of the Fe(1 0 0) outermost surface. This is consistent with recent calculations of the Fe(1 0 0) surface structure [19]. Some humps at smaller ˛ occur because of the focusing effect on the iron atom at the subsurface atomic layers [13]. On the other hand, the intensity variation of silicon also shows an increase at an angle similar to that of iron. This shows that the Si atoms at the outermost surface are situated in the same atomic plane as the Fe atoms. The critical angle of the focusing effect is indistinct for silicon compared with that of iron. This is partly due to the imperfect atomic arrangement around the Si atom. No clear humps are observed for silicon at smaller ˛, unlike for iron. This is attributed to intensity dispersion, which principally arises from the estimation procedure of the Fe peak background at the Si peak position. In the azimuthal angle scan in Fig. 2 (b), ˛ was fixed to 77◦ . In this scattering geometry, the scattered ions only come from the outermost surface because of the shadowing effect. Therefore, the intensity variation in Fig. 2 (b) is attributed to the atomic arrangement
of the outermost surface. The four-fold symmetry of the Fe intensity variation reflects the surface symmetry of Fe(1 0 0). The similar behaviors of silicon and iron show that these two elements crystallographically occupy equivalent sites. This indicates that the Si atoms are located at the substitutional site of bcc Fe. This is consistent with the ˛ and azimuthal angle dependence of the Fe intensity on Si/Fe(1 0 0) which is essentially the same with those on the Fe(1 0 0) clean surface (not shown). The abovementioned structure analysis manifests that some deposited Si atoms remain at the outermost surface as a result of silicide formation after annealing. Because the Fe surface was entirely covered by silicon before annealing as observed in Fig. 1, the excess Si atoms that are not located at the outermost surface after annealing should desorb from the surface or diffuse into the Fe substrate. It is important to clarify the incorporation of silicon into the Fe subsurface layers to analyze silicidation at the Si/Fe interface. Therefore, we briefly discuss below elemental composition at deep layers. Fig. 3 shows a series of ISS spectra obtained after 2 keV Ar+ ion sputtering with a fluence of 2×1013 ions·cm−2 ·s−1 . In the sputtering, the incident angle measured from the surface normal was 30◦ and the incident direction was parallel to Fe(0 1 0). Thus, it shows ˚ ˚ 0 0) (300 A)/MgO(1 0 0) subthe depth profiling of the Si (20 A)/Fe(1 ˚ strate. The starting surface was prepared by Si deposition (20 A) followed by annealing at 823 K for 5 min, which is identical to conditions used in Fig. 2 (Si/Fe(1 0 0)). The outermost surface constituents are iron and silicon in addition to oxygen, which is a minor contaminant. The Si peak intensity decreases with Ar+ ion beam irradiation. It is noted that the Fe peak slightly shifts to the lower energy side after the initial step of Ar+ ion beam irradiation. This is due to the disordering of the surface atomic arrangement, which enhances multiple scattering components in the ISS surface peak. The ion beam irradiation removes the minor surface contamint, which is mainly composed of light molecules. This explains the decrease of the secondary ion yield below 200 eV with the ion beam irradiation. In addition, the Si peak position shifts to the low energy side with the ion beam irradiation. This is also attributed to the disordering of the atomic arrangement. The Si peak survives with the ion beam irradiation for 20 min; this manifests the incorporation of silicon into the Fe substrate. The decrease of the Si peak intensity with the ion beam irradiation indicates a larger concentration of silicon in the shallower atomic layer. Almost no Si peak is observed after the ion beam irradiation for 30 min. Further increase of the irradiation dosage removes the Fe film. Finally, we observed the appearance of the Mg and O peaks of the MgO substrate and substantial decrease of the Fe peak intensity after the irradiation
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171
169
Fig. 4. ISS spectrum (a black curve) and SP-ISS spin asymmetries on the Si/Fe(1 0 0) surface (solid black squares), and a SP-ISS spin asymmetry on the Fe(1 0 0) surface (a red open circle). The dashed blue line shows the level of zero spin asymmetry. The incident energy was 860 eV, and ˛ and were 0◦ and 150◦ , respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
for 100 min (top spectrum). After the Ar+ ion beam irradiation for 110 min, the ISS measurements became impossible because of the charging effect of the MgO substrate. Thus, no intermixing occurs at the Fe/MgO interface, and the Fe film is removed by the ion beam irradiation for about 110 min. Judging from the Fe film thickness ˚ several equivalent layers of Fe(1 0 0) are removed by the (∼ 300 A), ion beam irradiation for 1 min. Thus, together with the element dependent sputtering speed, the disappearance of the Si peak by the irradiation for 20 min indicates that the incorporated Si atoms are distributed in surface layers several nm thick. The magnetic moments of iron and iron silicides mostly originate from spin moment. This is because of the freezing of orbital angular momentum owing to the strong ligand field, lifting the degeneracy of the angular momentum. Therefore, the magnetic moments are the quantity to be compared with the SP-ISS spin asymmetry, which reflects the total spin polarization of the target element [13]. Fig. 4 shows the SP-ISS result of the Fe(1 0 0) and the Si/Fe(1 0 0) surfaces. The incidence and scattering angles were 0◦ and 150◦ , respectively. In this scattering geometry, the SP-ISS signal is limited to the first and second layers due to the shadowing effect. The contribution from deeper layers through the reionization is negligible because the reionization is a spin independent charge exchange process [13]. The spin asymmetry is proportional to the spin polarization of the electron neutralizing the projectile He+ ions. Hence, it is approximately proportional to the total spin polarization of the target atom [20]. Therefore, the SP-ISS result in Fig. 4 reflects the average magnetic moment of the target elements in the two atomic layers of the outermost surface. Because the spin asymmetry appears only at the ISS peak position [20], the SP-ISS measurements were made for the He+ ion-target element binary collision energy. The SP-ISS data shows that the magnetic moment of iron in the silicide layer is substantially smaller than that of Fe(1 0 0). It is also revealed that the moment of the Si atoms is negligibly small. This magnetic moment of Si/Fe(1 0 0) is intuitively understood in terms of the local magnetic environment as discussed below. Neglecting the chemical species, silicon-iron alloys form the bcc substitutional structure up to 25 at% Si [21]. The ordering takes place when the composition approaches that of Fe3 Si, which forms the D03 structure. The D03 structure can be considered to comprise four fcc sublattices arranged regularly along the body diagonal, as shown in Fig. 5. A characteristic of the structure is that it has almost the same Fe-nearest-neighbor interatomic distance as bcc Fe. From
Fig. 5. D03 structure of ordered Fe3 Si.
the crystallographic viewpoint, two different Fe atoms exist in a unit cell of the D03 lattice, which are labeled as Fe I and Fe II in Fig. 5. The magnetism of ordered Fe3 Si has been well investigated because it has attracted attention as the Heusler magnetoresistance material. Both experiment and theory have consistently reported that the magnetic moment of Fe II is close to that of bcc Fe (2.2 B ) [22–27]. As shown in Fig. 5, Fe II is eight-fold coordinated to Fe I; this gives the similar magnetic moment of Fe II to that of bcc Fe. On the other hand, the moment of Fe I is induced through the interaction with the nearest-neighbor Fe II. Thus, it is reasonable to infer that replacing Si with Fe II reduces the moment of Fe I. This is the essence of the phenomenological model by Niculescu et al., which indicates that the moment of Fe I changes linearly with the average number of iron atoms in their nearest-neighbor sites, whereas Fe II maintains the moment of bcc Fe [28]. This model, known as the local magnetic environment model, is consistent with past studies on iron silicides with the D03 structure [23,26,27]. The moment of Si is reported to be very small compared with those of Fe I and Fe II. These experimental reports are quantitatively consistent with the theoretical calculations [22,24–26]. The abovementioned local magnetic environment model indicates that the average magnetic moment of Fe in Fe3 Si should be 75 % of that of bcc Fe. The SP-ISS data in Fig. 5 shows the similar ratio (∼ 0.7) of the Fe spin asymmetry on Si/Fe to that on Fe. Therefore, we indicate the iron silicide (Fe3+x Si1−x ) formation with x ∼ 0 at the outermost surface. Because our ISS and RHEED analyses show that the Si atoms replace the Fe atoms at the outermost surface while maintaining the bcc lattice, it is most likely that the atomic arrangement of the outermost surface is similar to that of Fe-Si plane terminated Fe3 Si(1 0 0) with the D03 structure. Our magnetic structure analysis is consistent with a density functional calculation for Fe3 Si(1 0 0) surface [29]. It is also consistent with our RHEED observation previously mentioned in the present paper. The Fe–Si bonding reduces the average magnetic moment of the Fe atom in the local environment. It is reflected as the change in the band structure mainly near the Fermi level [26]. Thus, the Fe–Si magnetic interaction, which describes the transportation of spin polarization through the Fe–Si bonding, is analyzed from electron spin polarization at the Fermi level.
170
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171
4. Conclusion
Fig. 6. SP-INS spin asymmetry at the kinetic energy of 15 eV, which corresponds to the Fermi level, as a function of the Si thickness. The measurements were obtained without annealing. The dashed blue line shows the level of zero spin asymmetry. The ˚ 0 0) inset shows INS spectra on the Fe(1 0 0) surface (black curve) and the Si 5 A/Fe(1 surface (red curve). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
Fig. 6 shows the SP-INS spin asymmetry at the Fermi level as a function of the Si thin-film thickness. The measurements were made without annealing. The incident and emitted angles from the surface normal were 0◦ and 30◦ , respectively, and the electrons emitted at the kinetic energy of 15 eV were measured. To reduce the kinetic electron emission, which acts as a background in the SPINS spin asymmetry, the incident He+ energy was lowered to about 50 eV. The kinetic energy of 15 eV corresponds to the high-energy limit of the emitted electron as shown in the INS spectra (the inset of Fig. 6). Under such conditions, the origin of the emitted electrons is attributed to the Fermi level in AN. Therefore, the SP-INS spin asymmetry at 15 eV reflects spin polarization at the Fermi level of the outermost surface. The Si deposition on the Fe surface reduces the emitted electron intensity as observed in the inset of Fig. 6. This is because of the semiconducting electronic structure of the Si surface. In other words, the reduced electron density at the Fermi level lowers the AN rate. The spin asymmetry linearly decreases with the Si thick˚ Thus, the Si thin-film acts as ness, and it vanishes at about 15 A. a magnetically dead layer, and Fe–Si hybridization is the mechanism that causes charge transfer, which reduces spin polarization at the Fermi level. Thus, the behavior of the SP-INS spin asymmetry in Fig. 6 is consistent with that of the local magnetic environment model. There seems to be a general agreement for the formation of a ferromagnetic Fe–Si layer at room temperature at Fe/Si interfaces. The typical thickness range of the Fe film for observing the solid phase reaction at the Fe/Si interface is below 1 nm [30–32]. It is likely that further Fe deposition causes the pure Fe film formation on the topmost surface atomic layer. The thickness and the magnetic property of the Fe–Si layer at the interface are still matters of intensive discussion. Some reports claim the formation of a ferromagnetic Fe–Si layer whose thickness reaches up to about 10 monolayers even at room temperature [30,33]. On the other hand, our (SP-) ISS results show that the ferromagnetic Fe3 Si layer is formed at the outermost surface, but the Si concentration has a strong gradient along the surface normal as observed in our ISS measurements for depth profiling. Thus, it seems that the tendency of Si segregating to the Fe surface affects the solid phase reaction at the Si/Fe and Fe/Si interfaces. This may be the origin of the discrepancy between the Si/Fe and Fe/Si interfaces showing asymmetric nature of these interfaces.
The atomic arrangement and spin polarization at the Fe(1 0 0) outermost surface deposited with Si were investigated by an electron spin-polarized 4 He+ ion beam to elucidate the magnetic structure of the Si/Fe interface. From the structure analysis by He+ ISS and RHEED, we found that the deposited Si layer√forms √ an iron silicide on the Fe surface with the periodicity of ( 2 × 2) R45◦ after annealing at 823 K. We observed the incorporation of silicon into the Fe substrate via the substitutional site by annealing. By SPISS, we observed that the average magnetic moment of Fe in the silicide layer is substantially smaller than that in the Fe(1 0 0) surface. Furthermore, the moment of Si is under the detection limit of SP-ISS. This magnetism of the silicide surface is understood from the bulk magnetic structure of Fe3 Si. By taking the structure analyses of ISS and RHEED, it is indicated that the outermost surface of the iron silicide has a similar atomic arrangement as that of ordered Fe3 Si(1 0 0) whose surface is terminated by the Fe–Si plane. We also discussed the change in the surface magnetism of Fe(1 0 0) by the Si deposition in terms of the Si–Fe bonding as observed in the spin polarization at the Fermi level by SP-INS. Considering past studies which report that the thickness of the ferromagnetic iron silicide is more than several monolayers at room temperature for Fe/Si, the asymmetric nature of the interfacial reaction between Si/Fe and Fe/Si is suggested. Acknowledgments This work was partially supported by SENTAN-JST and KAKENHI 22760032. References [1] V.V. Balashev, V.V. Korobtsov, T.A. Pisarenko, E.A. Chusovitin, K.N. Galkin, Physics of the Solid State 52 (2010) 397. [2] J.C. Gonzalez, D.R. Miquita, M.I.N. da Silva, R. Magalhaes-Paniago, M.V.B. Moreira, A.G. de Oliveira, Physical Review B 81 (2010) 113403. [3] D. Leong, M. Harry, K.J. Reeson, K.P. Homewood, Nature 387 (1997) 686. [4] M. Shaban, K. Nomoto, S. Izumi, T. Yoshitake, Applied Physics Letters 94 (2009) 222113. [5] M. Suzuno, S. Murase, T. Koizumi, T. Suemasu, Applied Physics Express 1 (2008) 021403. [6] S. Sugahara, M. Tanaka, Applied Physics Letters 84 (2004) 2307. [7] Y. Ando, K. Hamaya, K. Kasahara, Y. Kishi, K. Ueda, K. Sawano, T. Sadoh, M. Miyao, Applied Physics Letters 94 (2009) 182105. [8] S.P. Dash, H.D. Carstanjen, Physical Status Solidi B 248 (2011) 2300. [9] M. Someta, K. Maetani, K. Hattori, H. Daimon, Surface Science 604 (2010) 21. [10] M. Cinchetti, J.-P. Wustenberg, M.S. Albaneda, F. Steeb, A. Conca, M. Jourdan, M. Aeschlimann, Journal of Physics D 40 (2007) 1544. [11] S.R. Naik, S. Rai, M.K. Tiwari, G.S. Lodha, Journal of Physics D 41 (2008) 115307. [12] A. Gupta, D. Kumar, V. Phatak, Physical Review B 81 (2010) 155402. [13] T.T. Suzuki, H. Kuwahara, Y. Yamauchi, Surface Science 604 (2010) 1767. [14] H.H. Brongersma, M. Draxler, M. de Ridder, P. Bauer, Surface Science Reports 62 (2007) 63. [15] Y. Yamauchi, M. Kurahashi, Applied Surface Science 169 (2001) 236. [16] T. Suzuki, Y. Yamauchi, Nuclear Instruments and Methods in Physics Research Section A 575 (2007) 343. [17] D.L. Bixler, J.C. Lancaster, F.J. Kontur, R.A. Popple, F.B. Dunning, G.K. Walters, Review of Scientific Instruments 70 (1999) 240. [18] R. Souda, T. Aizawa, C. Oshima, S. Otani, Y. Ishizawa, Physical Review B 40 (1989) 4119. [19] J. Yu, X. Lin, J. Wang, J. Chen, W. Huang, Applied Surface Science 255 (2009) 9032. [20] T. Suzuki, Y. Yamauchi, Surface Science 602 (2008) 579. [21] M. Hansen, Constitution of Binary Alloys, McGraw-Hill, New York, 1958. [22] J. Kudrnovsky, N.E. Christensen, O.K. Andersen, Physical Review B 43 (1991) 5924. [23] W.A. Hines, A.H. Menotti, J.I. Budnick, T.J. Burch, T. Litrenta, V. Niculescu, K. Raj, Physical Review B 13 (1976) 4060. [24] N.I. Kulikov, D. Fristot, J. Hugel, A.V. Postnikov, Physical Review B 66 (2002) 014206. [25] A. Bansil, S. Kaprzyk, P.E. Mijnarends, J. Tobola, Physical Review B 60 (1999) 13396. [26] K. Zakeri, S.J. Hashemifar, J. Lindner, I. Barsukov, R. Meckenstock, P. Kratzer, Z. Frait, M. Farle, Physical Review B 77 (2008) 104430.
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171 [27] J. Moss, P.J. Brown, Journal of Physics F 2 (1972) 358. [28] V.A. Niculescu, T.J. Burch, J.I. Budnick, Journal of Magnetism and Magnetic Materials 39 (1983) 223. [29] J. Hafner, D. Spisak, Physical Review B 75 (2007) 195411. [30] F. Zavaliche, W. Wulfhekel, H. Xu, J. Kirschner, Journal of Applied Physics 88 (2000) 5289.
171
[31] I.I. Pronin, M.V. Gomoyunova, D.E. Malygin, D.V. Vyalikh, Y.S. Dedkov, S.L. Molodtsov, Applied Physics A 94 (2009) 467. [32] N.G. Gheorghe, M.A. Husanu, G.A. Lungu, R.M. Costescu, D. Macovei, C.M. Teodorescu, Journal of Materials Science 47 (2012) 1614. [33] R. Klasges, C. Carbone, W. Eberhardt, C. Pampuch, O. Rader, T. Kachel, W. Gudat, Physical Review B 56 (1997) 10801.
Contents lists available at SciVerse ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Atomic arrangement and magnetism of iron silicide on Fe(1 0 0) surface T.T. Suzuki ∗ , S. Hishita National Institute for Materials Science, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan
a r t i c l e
i n f o
Article history: Received 30 March 2012 Received in revised form 2 July 2012 Accepted 3 July 2012 Available online 20 July 2012 Keywords: Spin-polarized ion scattering spectroscopy Surface magnetic structure Si/Fe interfaces
a b s t r a c t We investigated surface magnetic structure in the initial stage of Si deposition on an Fe(1 0 0) surface by spin-polarized ion scattering spectroscopy (SP-ISS). We found silicidation at the Si/Fe interface after Si deposition followed by annealing at 823 K. The silicidation occurs by the incorporation of silicon into the Fe substrate via the substitutional site of bcc Fe. After annealing, the incorporated Si atoms are distributed in surface layers several nanometers thick. The SP-ISS analysis revealed that the average magnetic moment of Fe in the silicide surface layer is about 70 % of that of Fe in the Fe(1 0 0) surface layer, whereas that of Si is almost zero. These surface magnetic moments are discussed in terms of the local magnetic environment. It is likely that the outermost surface of the silicide layer has an atomic arrangement similar to that of Fe3 Si(1 0 0) with surface termination by the Fe–Si plane. © 2012 Elsevier B.V. All rights reserved.
1. Introduction An Fe/Si interfacial reaction, i.e., silicidation, causes the formation of many stable and metastable Si–Fe compounds. These compounds exhibit a wide variety of electrical properties from semiconductor to metal and magnetic properties such as diamagnetism, paramagnetism, and ferromagnetism [1,2]. Therefore, tailoring the Si–Fe interfacial reaction has the potential to realize silicon-based optoelectronics and spintronics. These future devices are expected to be compatible with established and emerging Si technologies. For example, semiconducting iron disilicide ˇ-FeSi2 is a promising candidate for optoelectronics and photovoltaics because it acts as a Si-based light emitter as well as a photodetector. Photoluminescence and electroluminescence at a wavelength (∼1.5 m) relevant to optical fiber communications have already been demonstrated using ˇ-FeSi2 [3–5]. On the other hand, ferromagnetic Fe3 Si is a potential material for use in spintronics. A Schottky tunnel barrier source/drain made of Fe3 Si has been proposed to construct metal-oxide-semiconductor field effect transistors [6]. Spin injection in silicon through the Fe3 Si/Si Schottky barrier has already been demonstrated [7]. Silicidation at Fe/Si interfaces has been employed to grow iron silicides on Si substrates. Therefore, it is of central importance to elucidate the atomic arrangement and magnetic properties at the interfaces to fabricate device components of the abovementioned future technologies. However, although many studies have investigated silicidation at the interface, complex methods to produce a certain Si–Fe compound at the interface are still
∗ Corresponding author. E-mail address: [email protected] (T.T. Suzuki). 0169-4332/$ – see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.apsusc.2012.07.012
controversial [2,4,8]. For example, the identification of threedimensional iron-silicide islands on Si surfaces is still a matter of intensive discussion [9]. In particular, very little is known about the magnetic structure at the Fe/Si interfaces. Since spins at the interfaces typically characterize the principle of a spintronics device, such as the tunneling magnetoresistance effect [10], it is essential for the development of spintronics to analyze the magnetic structure of a ferromagnetic atomic layer in contact with a nonmagnetic layer at the interface. In this context, spin polarization of a single Si (–Fe) layer induced by the magnetism of the Fe layer at the Si/Fe interface is informative. To the best of our knowledge, however, no study on the magnetic structure of the single atomic layer has been reported for Fe/Si or Si/Fe interfaces. In the present study, we investigated the magnetic structure of an Fe(1 0 0) outermost surface deposited with silicon by spinpolarized ion scattering spectroscopy (SP-ISS). We concentrated on a certain silicide phase formed at the Fe surface. The silicide phase was grown with good reproducibility after Si deposition followed by annealing in an ultra-high vacuum (UHV), i.e., solid phase epitaxy. We investigated the atomic arrangement, spin arrangement, and spin polarization at the Fermi level of this silicide phase. Most of studies on the silicide at the Fe/Si interface have prepared the sample in the opposite way: Fe deposition on Si surfaces followed by annealing. The magnetic structure analysis of the Si/Fe outermost surface in the present study should provide important clues of the magnetic structure and its asymmetric nature at the Fe/Si and Si/Fe interfaces [11,12]. Details of SP-ISS have been previously described [13]. Briefly, electron spin-polarized 4 He+ ions are projected on a sample surface, and the kinetic energy of the scattered ions is analyzed in SP-ISS measurements. The kinetic energy of the incident He+ ions is of the order of keV, and therefore, most incident He+ ions are
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171
neutralized at the sample surface, typically via Auger neutralization (AN). Because SP-ISS detects the scattered He+ ions that survive efficient AN, it has extreme surface sensitivity [14]. According to the Pauli exclusion principle, the scattered He+ ion intensity should differ depending on whether the He+ spin is parallel to majority or minority spins of the magnetic surface in the AN process. Thus, SP-ISS detects the spin polarization of the target atom on the outermost surface with element selectivity. Moreover, spin polarization at the Fermi level is estimated from the spin dependence of electron emission with the He+ ion impact on the surface (spinpolarized ion neutralization spectroscopy (SP-INS)) [13]. In SP-ISS and SP-INS studies, the spin dependence of the scattered ions and emitted electrons are discussed in terms of spin asymmetry. The spin asymmetry is defined as (I↑ − I↓ )/[PHe+ · (I↑ + I↓ )], where PHe+ expresses the spin polarization of the incident He+ ion beam, and I↑ and I↓ denote the scattered ion intensity for SP-ISS and the emitted electron intensity for SP-INS with the projectile He+ ions polarized parallel and anti-parallel to the majority spins, respectively.
167
Fig. 1. A series of ISS spectra obtained on the Fe(1 0 0) surface with the Si deposition. The spectra are normalized by the incident ion beam current at the sample. The ˚ by annealing bottom spectrum was obtained after the Si deposition of 22 Afollowed at 823 K for 5 min. The incident energy was 1.77 keV, and ˛ and were 0◦ and 150◦ , respectively. Arrows indicate the binary collision energies for some elements.
2. Experimental method and setup Experiments were performed in an UHV chamber (base pressure of 5 × 10−11 Torr) equipped for SP-ISS, reflection high energy electron diffraction (RHEED), and deposition of iron and silicon. We grew Fe(1 0 0) films on MgO(1 0 0) single-crystalline substrates. We prepared the MgO(1 0 0) clean surface by annealing the mirror polished MgO substrate (10 mm × 10 mm × 1 mm) above at 1300 K by electron bombardment in UHV. Then, we deposited iron at room temperature, followed by annealing at about 700 K for 10 min. Under these growth conditions, bcc Fe films grew epitaxially with the orientation relationships Fe(1 0 0)//MgO(1 0 0) and Fe[1 1 0]//MgO[1 0 0] [15]. The thickness of the iron film was about ˚ We deposited silicon on the Fe(1 0 0) film by using an elec300 A. tron beam evaporator (Omicron EFM3) at room temperature in UHV, and subsequently, we annealed the Si/Fe(1 0 0) sample at elevated temperatures. The sample was pulse-magnetized in-plane by a retractable pulse magnet placed in the UHV chamber before SPISS measurements. The magnetization direction was parallel to the Fe[0 0 1] easy axis of the bcc-Fe films. Electron spin-polarized 4 He+ ions were generated from Penning ionization of spin-polarized metastable He 23 S1 atoms (He*) [16]: He∗ + He∗ → He+ + He0 + e− . We employed an optical pumping (OP) technique, using the D0 line of 1083 nm radiation (He* 23 S1 → 23 P0 ), for spin polarization of He*. The spin angular momentum is conserved in Penning ionization; thus, if He* is spin polarized by OP, spin-polarized He+ ions are generated [17]. The OP radiation was provided by an ytterbium-doped fiber laser (Keopsys KPS-BT2-YFL-1083-FA) via an ytterbium-doped fiber amplifier (Keopsys KPS-BT2-YFA-1083200-COL). The wavelength of the 1083 nm seed light from the fiber laser was precisely tuned to the D0 line. The spin directions of He+ ions (up or down) were controlled by the polarization of the OP radiation. The spin polarization of the He+ ion beam PHe+ was about 0.2. The entire apparatus was surrounded by a three-axis coil to compensate for the Earths magnetic field. An additional coil produced a weak guiding field (∼0.3 Oe), which was perpendicular to the scattering plane. Thus, the spin direction of the incident He+ ion beam was defined by the guiding field, which was parallel to Fe[0 0 1]. The scattering plane was perpendicular to the Fe(1 0 0) surface (the inset of Fig. 1). Both the scattered He+ ions and the emitted electrons were measured using a rotatable hemispherical sector analyzer (Omicron SHA50). The measurements were conducted in a constant pass
energy mode with a pass energy of 318 eV for SP-ISS and 40 eV for SP-INS. 3. Results and discussion Fig. 1 shows ISS spectra obtained at the Fe(1 0 0) surface deposited with silicon. The bottom spectrum was obtained after ˚ the Si deposition of 22 Afollowed by annealing at 823 K for 5 min. The incident energy was 1.77 keV, and the incident angle measured from the surface normal (˛) and the scattering angle () were 0◦ and 150◦ , respectively. The binary collision energies of the projectile He+ ion with iron (1354 eV), silicon (1036 eV), and oxygen (682 eV) are indicated by arrows at the bottom of the spectra. In Fig. 1, the iron peak decreases with the Si deposition at room ˚ It is temperature; it almost vanishes with the Si thickness of 22 A. noted that the Fe peak position shifts to the low energy side with the Si deposition. This is due to the increase of inelastic scattering components, such as electronic-states excitation and reionization [18]. Such inelastic scattering processes occur in the Si overlayers on iron. Accordingly, the ISS peak reflects the composition of several surface layers. The Fe peak reappears after annealing; hence, the outermost surface is composed of both iron and silicon. The reappearance of the Fe atoms at the outermost surface is consistent with the shift of the Fe peak to the high energy side with annealing. In our RHEED observation, the 1×1 streak on Fe(1 0 0) changed into the diffusive 1×1 pattern with the Si deposition, indicating the disordered structure of √ the Si thin-film (not shown). We also √ observed that the ( 2 × 2) R45◦ pattern appears after annealing at 823 K, although it was not a sharp pattern. We observed no transmission pattern through our experiments. This, together with the ISS results in Fig. 1, shows possibility of the silicidation on the surface after annealing. Further increase of the annealing temperature did not essentially change the RHEED pattern but substantially reduced the Si peak intensity in the ISS spectra. The Si peak in the ISS spectra almost vanishes after annealing at 1173 K for 5 min, and accordingly, the 1×1 pattern reappeared in RHEED. Since we are interested in the silicide phase on the Fe substrate, we will, hereafter, concentrate on the abovementioned silicide formed by the Si deposition followed by annealing at 823 K. Fig. 2 shows (a) ˛ and (b) azimuthal angle scans of the scattered He+ intensity with the incident energy of 1.77 keV and of 150◦ . The peak intensity of iron corresponds to integrated counts between 1240 and 1400 eV, while that of silicon corresponds to the counts between 920 and 1100 eV obtained by subtracting the background of the Fe peak estimated by a linear function. In the ˛
168
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171
Fig. 3. A series of ISS spectra obtained on the Si/Fe(1 0 0)/MgO(1 0 0) surface with 2 keV Ar+ ion beam irradiation. The sample was prepared by the Si deposition of ˚ Fe(1 0 0)/MgO(1 0 0) followed by annealing at 823 K for 5 min. In ISS, the 20 Aon incident energy was 1.71 keV, and ˛ and were 0◦ and 150◦ , respectively.
Fig. 2. ˛ scan (a) and azimuthal angle scan (b) of scattered He+ ion intensity with the incident energy of 1.77 keV and of 150◦ . The samples were the Fe(1 0 0) clean sur˚ deposited face (red open circles) and the Fe(1 0 0) surface on which silicon of 20 Awas followed by annealing at 823 K for 5 min (black solid squares). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
scan measurement, ˛ was varied along the Fe[0 1 0] azimuth and the scattering plane was parallel to Fe(0 0 1) as shown in the inset of Fig. 2 (a). The Fe intensity variation shown in Fig. 2 (a) exhibits a sharp increase at 68◦ . This is due to the focusing effect on the Fe atoms located at the atomic layer of the outermost surface. The critical angle of this focusing effect is explained from the bulk termination structure of the Fe(1 0 0) outermost surface. This is consistent with recent calculations of the Fe(1 0 0) surface structure [19]. Some humps at smaller ˛ occur because of the focusing effect on the iron atom at the subsurface atomic layers [13]. On the other hand, the intensity variation of silicon also shows an increase at an angle similar to that of iron. This shows that the Si atoms at the outermost surface are situated in the same atomic plane as the Fe atoms. The critical angle of the focusing effect is indistinct for silicon compared with that of iron. This is partly due to the imperfect atomic arrangement around the Si atom. No clear humps are observed for silicon at smaller ˛, unlike for iron. This is attributed to intensity dispersion, which principally arises from the estimation procedure of the Fe peak background at the Si peak position. In the azimuthal angle scan in Fig. 2 (b), ˛ was fixed to 77◦ . In this scattering geometry, the scattered ions only come from the outermost surface because of the shadowing effect. Therefore, the intensity variation in Fig. 2 (b) is attributed to the atomic arrangement
of the outermost surface. The four-fold symmetry of the Fe intensity variation reflects the surface symmetry of Fe(1 0 0). The similar behaviors of silicon and iron show that these two elements crystallographically occupy equivalent sites. This indicates that the Si atoms are located at the substitutional site of bcc Fe. This is consistent with the ˛ and azimuthal angle dependence of the Fe intensity on Si/Fe(1 0 0) which is essentially the same with those on the Fe(1 0 0) clean surface (not shown). The abovementioned structure analysis manifests that some deposited Si atoms remain at the outermost surface as a result of silicide formation after annealing. Because the Fe surface was entirely covered by silicon before annealing as observed in Fig. 1, the excess Si atoms that are not located at the outermost surface after annealing should desorb from the surface or diffuse into the Fe substrate. It is important to clarify the incorporation of silicon into the Fe subsurface layers to analyze silicidation at the Si/Fe interface. Therefore, we briefly discuss below elemental composition at deep layers. Fig. 3 shows a series of ISS spectra obtained after 2 keV Ar+ ion sputtering with a fluence of 2×1013 ions·cm−2 ·s−1 . In the sputtering, the incident angle measured from the surface normal was 30◦ and the incident direction was parallel to Fe(0 1 0). Thus, it shows ˚ ˚ 0 0) (300 A)/MgO(1 0 0) subthe depth profiling of the Si (20 A)/Fe(1 ˚ strate. The starting surface was prepared by Si deposition (20 A) followed by annealing at 823 K for 5 min, which is identical to conditions used in Fig. 2 (Si/Fe(1 0 0)). The outermost surface constituents are iron and silicon in addition to oxygen, which is a minor contaminant. The Si peak intensity decreases with Ar+ ion beam irradiation. It is noted that the Fe peak slightly shifts to the lower energy side after the initial step of Ar+ ion beam irradiation. This is due to the disordering of the surface atomic arrangement, which enhances multiple scattering components in the ISS surface peak. The ion beam irradiation removes the minor surface contamint, which is mainly composed of light molecules. This explains the decrease of the secondary ion yield below 200 eV with the ion beam irradiation. In addition, the Si peak position shifts to the low energy side with the ion beam irradiation. This is also attributed to the disordering of the atomic arrangement. The Si peak survives with the ion beam irradiation for 20 min; this manifests the incorporation of silicon into the Fe substrate. The decrease of the Si peak intensity with the ion beam irradiation indicates a larger concentration of silicon in the shallower atomic layer. Almost no Si peak is observed after the ion beam irradiation for 30 min. Further increase of the irradiation dosage removes the Fe film. Finally, we observed the appearance of the Mg and O peaks of the MgO substrate and substantial decrease of the Fe peak intensity after the irradiation
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171
169
Fig. 4. ISS spectrum (a black curve) and SP-ISS spin asymmetries on the Si/Fe(1 0 0) surface (solid black squares), and a SP-ISS spin asymmetry on the Fe(1 0 0) surface (a red open circle). The dashed blue line shows the level of zero spin asymmetry. The incident energy was 860 eV, and ˛ and were 0◦ and 150◦ , respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
for 100 min (top spectrum). After the Ar+ ion beam irradiation for 110 min, the ISS measurements became impossible because of the charging effect of the MgO substrate. Thus, no intermixing occurs at the Fe/MgO interface, and the Fe film is removed by the ion beam irradiation for about 110 min. Judging from the Fe film thickness ˚ several equivalent layers of Fe(1 0 0) are removed by the (∼ 300 A), ion beam irradiation for 1 min. Thus, together with the element dependent sputtering speed, the disappearance of the Si peak by the irradiation for 20 min indicates that the incorporated Si atoms are distributed in surface layers several nm thick. The magnetic moments of iron and iron silicides mostly originate from spin moment. This is because of the freezing of orbital angular momentum owing to the strong ligand field, lifting the degeneracy of the angular momentum. Therefore, the magnetic moments are the quantity to be compared with the SP-ISS spin asymmetry, which reflects the total spin polarization of the target element [13]. Fig. 4 shows the SP-ISS result of the Fe(1 0 0) and the Si/Fe(1 0 0) surfaces. The incidence and scattering angles were 0◦ and 150◦ , respectively. In this scattering geometry, the SP-ISS signal is limited to the first and second layers due to the shadowing effect. The contribution from deeper layers through the reionization is negligible because the reionization is a spin independent charge exchange process [13]. The spin asymmetry is proportional to the spin polarization of the electron neutralizing the projectile He+ ions. Hence, it is approximately proportional to the total spin polarization of the target atom [20]. Therefore, the SP-ISS result in Fig. 4 reflects the average magnetic moment of the target elements in the two atomic layers of the outermost surface. Because the spin asymmetry appears only at the ISS peak position [20], the SP-ISS measurements were made for the He+ ion-target element binary collision energy. The SP-ISS data shows that the magnetic moment of iron in the silicide layer is substantially smaller than that of Fe(1 0 0). It is also revealed that the moment of the Si atoms is negligibly small. This magnetic moment of Si/Fe(1 0 0) is intuitively understood in terms of the local magnetic environment as discussed below. Neglecting the chemical species, silicon-iron alloys form the bcc substitutional structure up to 25 at% Si [21]. The ordering takes place when the composition approaches that of Fe3 Si, which forms the D03 structure. The D03 structure can be considered to comprise four fcc sublattices arranged regularly along the body diagonal, as shown in Fig. 5. A characteristic of the structure is that it has almost the same Fe-nearest-neighbor interatomic distance as bcc Fe. From
Fig. 5. D03 structure of ordered Fe3 Si.
the crystallographic viewpoint, two different Fe atoms exist in a unit cell of the D03 lattice, which are labeled as Fe I and Fe II in Fig. 5. The magnetism of ordered Fe3 Si has been well investigated because it has attracted attention as the Heusler magnetoresistance material. Both experiment and theory have consistently reported that the magnetic moment of Fe II is close to that of bcc Fe (2.2 B ) [22–27]. As shown in Fig. 5, Fe II is eight-fold coordinated to Fe I; this gives the similar magnetic moment of Fe II to that of bcc Fe. On the other hand, the moment of Fe I is induced through the interaction with the nearest-neighbor Fe II. Thus, it is reasonable to infer that replacing Si with Fe II reduces the moment of Fe I. This is the essence of the phenomenological model by Niculescu et al., which indicates that the moment of Fe I changes linearly with the average number of iron atoms in their nearest-neighbor sites, whereas Fe II maintains the moment of bcc Fe [28]. This model, known as the local magnetic environment model, is consistent with past studies on iron silicides with the D03 structure [23,26,27]. The moment of Si is reported to be very small compared with those of Fe I and Fe II. These experimental reports are quantitatively consistent with the theoretical calculations [22,24–26]. The abovementioned local magnetic environment model indicates that the average magnetic moment of Fe in Fe3 Si should be 75 % of that of bcc Fe. The SP-ISS data in Fig. 5 shows the similar ratio (∼ 0.7) of the Fe spin asymmetry on Si/Fe to that on Fe. Therefore, we indicate the iron silicide (Fe3+x Si1−x ) formation with x ∼ 0 at the outermost surface. Because our ISS and RHEED analyses show that the Si atoms replace the Fe atoms at the outermost surface while maintaining the bcc lattice, it is most likely that the atomic arrangement of the outermost surface is similar to that of Fe-Si plane terminated Fe3 Si(1 0 0) with the D03 structure. Our magnetic structure analysis is consistent with a density functional calculation for Fe3 Si(1 0 0) surface [29]. It is also consistent with our RHEED observation previously mentioned in the present paper. The Fe–Si bonding reduces the average magnetic moment of the Fe atom in the local environment. It is reflected as the change in the band structure mainly near the Fermi level [26]. Thus, the Fe–Si magnetic interaction, which describes the transportation of spin polarization through the Fe–Si bonding, is analyzed from electron spin polarization at the Fermi level.
170
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171
4. Conclusion
Fig. 6. SP-INS spin asymmetry at the kinetic energy of 15 eV, which corresponds to the Fermi level, as a function of the Si thickness. The measurements were obtained without annealing. The dashed blue line shows the level of zero spin asymmetry. The ˚ 0 0) inset shows INS spectra on the Fe(1 0 0) surface (black curve) and the Si 5 A/Fe(1 surface (red curve). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.)
Fig. 6 shows the SP-INS spin asymmetry at the Fermi level as a function of the Si thin-film thickness. The measurements were made without annealing. The incident and emitted angles from the surface normal were 0◦ and 30◦ , respectively, and the electrons emitted at the kinetic energy of 15 eV were measured. To reduce the kinetic electron emission, which acts as a background in the SPINS spin asymmetry, the incident He+ energy was lowered to about 50 eV. The kinetic energy of 15 eV corresponds to the high-energy limit of the emitted electron as shown in the INS spectra (the inset of Fig. 6). Under such conditions, the origin of the emitted electrons is attributed to the Fermi level in AN. Therefore, the SP-INS spin asymmetry at 15 eV reflects spin polarization at the Fermi level of the outermost surface. The Si deposition on the Fe surface reduces the emitted electron intensity as observed in the inset of Fig. 6. This is because of the semiconducting electronic structure of the Si surface. In other words, the reduced electron density at the Fermi level lowers the AN rate. The spin asymmetry linearly decreases with the Si thick˚ Thus, the Si thin-film acts as ness, and it vanishes at about 15 A. a magnetically dead layer, and Fe–Si hybridization is the mechanism that causes charge transfer, which reduces spin polarization at the Fermi level. Thus, the behavior of the SP-INS spin asymmetry in Fig. 6 is consistent with that of the local magnetic environment model. There seems to be a general agreement for the formation of a ferromagnetic Fe–Si layer at room temperature at Fe/Si interfaces. The typical thickness range of the Fe film for observing the solid phase reaction at the Fe/Si interface is below 1 nm [30–32]. It is likely that further Fe deposition causes the pure Fe film formation on the topmost surface atomic layer. The thickness and the magnetic property of the Fe–Si layer at the interface are still matters of intensive discussion. Some reports claim the formation of a ferromagnetic Fe–Si layer whose thickness reaches up to about 10 monolayers even at room temperature [30,33]. On the other hand, our (SP-) ISS results show that the ferromagnetic Fe3 Si layer is formed at the outermost surface, but the Si concentration has a strong gradient along the surface normal as observed in our ISS measurements for depth profiling. Thus, it seems that the tendency of Si segregating to the Fe surface affects the solid phase reaction at the Si/Fe and Fe/Si interfaces. This may be the origin of the discrepancy between the Si/Fe and Fe/Si interfaces showing asymmetric nature of these interfaces.
The atomic arrangement and spin polarization at the Fe(1 0 0) outermost surface deposited with Si were investigated by an electron spin-polarized 4 He+ ion beam to elucidate the magnetic structure of the Si/Fe interface. From the structure analysis by He+ ISS and RHEED, we found that the deposited Si layer√forms √ an iron silicide on the Fe surface with the periodicity of ( 2 × 2) R45◦ after annealing at 823 K. We observed the incorporation of silicon into the Fe substrate via the substitutional site by annealing. By SPISS, we observed that the average magnetic moment of Fe in the silicide layer is substantially smaller than that in the Fe(1 0 0) surface. Furthermore, the moment of Si is under the detection limit of SP-ISS. This magnetism of the silicide surface is understood from the bulk magnetic structure of Fe3 Si. By taking the structure analyses of ISS and RHEED, it is indicated that the outermost surface of the iron silicide has a similar atomic arrangement as that of ordered Fe3 Si(1 0 0) whose surface is terminated by the Fe–Si plane. We also discussed the change in the surface magnetism of Fe(1 0 0) by the Si deposition in terms of the Si–Fe bonding as observed in the spin polarization at the Fermi level by SP-INS. Considering past studies which report that the thickness of the ferromagnetic iron silicide is more than several monolayers at room temperature for Fe/Si, the asymmetric nature of the interfacial reaction between Si/Fe and Fe/Si is suggested. Acknowledgments This work was partially supported by SENTAN-JST and KAKENHI 22760032. References [1] V.V. Balashev, V.V. Korobtsov, T.A. Pisarenko, E.A. Chusovitin, K.N. Galkin, Physics of the Solid State 52 (2010) 397. [2] J.C. Gonzalez, D.R. Miquita, M.I.N. da Silva, R. Magalhaes-Paniago, M.V.B. Moreira, A.G. de Oliveira, Physical Review B 81 (2010) 113403. [3] D. Leong, M. Harry, K.J. Reeson, K.P. Homewood, Nature 387 (1997) 686. [4] M. Shaban, K. Nomoto, S. Izumi, T. Yoshitake, Applied Physics Letters 94 (2009) 222113. [5] M. Suzuno, S. Murase, T. Koizumi, T. Suemasu, Applied Physics Express 1 (2008) 021403. [6] S. Sugahara, M. Tanaka, Applied Physics Letters 84 (2004) 2307. [7] Y. Ando, K. Hamaya, K. Kasahara, Y. Kishi, K. Ueda, K. Sawano, T. Sadoh, M. Miyao, Applied Physics Letters 94 (2009) 182105. [8] S.P. Dash, H.D. Carstanjen, Physical Status Solidi B 248 (2011) 2300. [9] M. Someta, K. Maetani, K. Hattori, H. Daimon, Surface Science 604 (2010) 21. [10] M. Cinchetti, J.-P. Wustenberg, M.S. Albaneda, F. Steeb, A. Conca, M. Jourdan, M. Aeschlimann, Journal of Physics D 40 (2007) 1544. [11] S.R. Naik, S. Rai, M.K. Tiwari, G.S. Lodha, Journal of Physics D 41 (2008) 115307. [12] A. Gupta, D. Kumar, V. Phatak, Physical Review B 81 (2010) 155402. [13] T.T. Suzuki, H. Kuwahara, Y. Yamauchi, Surface Science 604 (2010) 1767. [14] H.H. Brongersma, M. Draxler, M. de Ridder, P. Bauer, Surface Science Reports 62 (2007) 63. [15] Y. Yamauchi, M. Kurahashi, Applied Surface Science 169 (2001) 236. [16] T. Suzuki, Y. Yamauchi, Nuclear Instruments and Methods in Physics Research Section A 575 (2007) 343. [17] D.L. Bixler, J.C. Lancaster, F.J. Kontur, R.A. Popple, F.B. Dunning, G.K. Walters, Review of Scientific Instruments 70 (1999) 240. [18] R. Souda, T. Aizawa, C. Oshima, S. Otani, Y. Ishizawa, Physical Review B 40 (1989) 4119. [19] J. Yu, X. Lin, J. Wang, J. Chen, W. Huang, Applied Surface Science 255 (2009) 9032. [20] T. Suzuki, Y. Yamauchi, Surface Science 602 (2008) 579. [21] M. Hansen, Constitution of Binary Alloys, McGraw-Hill, New York, 1958. [22] J. Kudrnovsky, N.E. Christensen, O.K. Andersen, Physical Review B 43 (1991) 5924. [23] W.A. Hines, A.H. Menotti, J.I. Budnick, T.J. Burch, T. Litrenta, V. Niculescu, K. Raj, Physical Review B 13 (1976) 4060. [24] N.I. Kulikov, D. Fristot, J. Hugel, A.V. Postnikov, Physical Review B 66 (2002) 014206. [25] A. Bansil, S. Kaprzyk, P.E. Mijnarends, J. Tobola, Physical Review B 60 (1999) 13396. [26] K. Zakeri, S.J. Hashemifar, J. Lindner, I. Barsukov, R. Meckenstock, P. Kratzer, Z. Frait, M. Farle, Physical Review B 77 (2008) 104430.
T.T. Suzuki, S. Hishita / Applied Surface Science 259 (2012) 166–171 [27] J. Moss, P.J. Brown, Journal of Physics F 2 (1972) 358. [28] V.A. Niculescu, T.J. Burch, J.I. Budnick, Journal of Magnetism and Magnetic Materials 39 (1983) 223. [29] J. Hafner, D. Spisak, Physical Review B 75 (2007) 195411. [30] F. Zavaliche, W. Wulfhekel, H. Xu, J. Kirschner, Journal of Applied Physics 88 (2000) 5289.
171
[31] I.I. Pronin, M.V. Gomoyunova, D.E. Malygin, D.V. Vyalikh, Y.S. Dedkov, S.L. Molodtsov, Applied Physics A 94 (2009) 467. [32] N.G. Gheorghe, M.A. Husanu, G.A. Lungu, R.M. Costescu, D. Macovei, C.M. Teodorescu, Journal of Materials Science 47 (2012) 1614. [33] R. Klasges, C. Carbone, W. Eberhardt, C. Pampuch, O. Rader, T. Kachel, W. Gudat, Physical Review B 56 (1997) 10801.