Characterization of Fe Silicide Growth on Si(111) Surface by Weissenberg RHEED

Fe silicide growth on a Si(111) surface has been characterized by Weissenberg RHEED. The silicides were grown in the present study via two fundamental methods, reactive deposition epitaxy (RDE) and solid phase epitaxy (SPE), with various deposition amounts of Fe. The silicide species and its epitaxial orientation were determined from three-dimensional diffraction patterns obtained using Weissenberg RHEED. It was found that only α-FeSi2 islands grew by the RDE method. Both α-FeSi2 and β-FeSi2 islands grew by the SPE method. The proportion of β-FeSi2 increased with increasing Fe deposition amount, until only β-FeSi2 was finally observed. The present results could be explained by the surface or interface reaction model wherein the reaction at the Si surface and at the Fe/Si solid interface caused α-FeSi2 and β-FeSi2 growth, respectively. [DOI: 10.1380/ejssnt.2009.866]


I. INTRODUCTION
The integration of magnetic devices with silicon technology is one of the hottest current subjects in advanced electronics.Since Fe is the most common ferromagnetic element, it is crucial to understand how Fe interacts with Si, and how Fe grows on a Si substrate.It is known that when Fe is deposited on a Si substrate, several types of Fe silicides, such as α-FeSi 2 , β-FeSi 2 , γ-FeSi 2 , c-FeSi and -FeSi are formed on the substrate after thermal reactions [1][2][3][4][5][6][7][8][9][10][11][12].Although most of these silicides are non-magnetic, they are important from another point-of-view.In particular, β-FeSi 2 is a semiconductor with a low environmental load, whose band gap is suitable for optical communication devices.However, most previous studies are contradictory in terms of the growth conditions or the resulting silicide species on Si [1][2][3][4][5][6][7][8][9][10][11][12].
It is known that Fe silicides with a cubic unit cell grow on Si(111) with an epitaxial orientation of silicide(111)//Si(111) and silicide [11 2]//Si [ 11 2] under certain conditions [2,3,6,10].A schematic phase diagram of the surface phase and Fe silicide on a Si(111) surface grown by solid phase epitaxy (SPE) was reported by Kataoka et al. [10].The phase diagram was mainly drawn using LEED and STM measurements, and was based on the surface structure.It is generally difficult to distinguish between three-dimensional (3D) form silicide species which have similar cubic unit cells, i.e., γ-FeSi 2 (CaF 2 type), FeSi (CsCl type) and α-FeSi 2 .Although α-FeSi 2 has a tetragonal structure, four tetragonal unit cells can be considered as one quasi-cubic cell with the lattice constant close to that of the substrate [3].In order to exactly determine the silicide species, one needs to investigate the reciprocal space in three dimensions.In addition to silicides with cubic unit cells, it is also known that β-FeSi 2 (orthogonal unit cell) grows with the epitaxial orientation of (110)//Si(111) and [001]//Si[1 10], or (101)//Si(111) and [010]//Si [1 10] [7-12].
The objective of the present work was to determine the Fe silicide species grown on Si(111) surfaces using two fundamental methods, reactive deposition epitaxy (RDE) and solid phase epitaxy (SPE), with various deposition amounts of Fe.We then compared the results of the two different methods to better understand the mechanism of Fe silicide formation on the Si(111) surface.Crystal structure and morphology of the epitaxial films were investigated using SEM and Weissenberg RHEED [13,14] wherein the principle governing the Weissenberg camera for x-ray crystallography [15] were incorporated into RHEED in order to obtain a 3D reciprocal image.The crystal structure could be immediately determined from the 3D reciprocal image.

II. EXPERIMENTAL
Experiments were performed using an ultra high vacuum (UHV) azimuthal-scan RHEED system, shown in Fig. 1, consisting of an electron gun, an energy filter, a phosphor screen, a five-axis sample manipulator, and an ebeam evaporator for the Fe deposition.Base pressure was < 1 × 10 −8 Pa.Polar and azimuth rotations of the sample manipulator are motorized.An energy filter was used to remove inelastically scattered electrons that generally disturb observations of diffraction patterns.A high-pass energy filter consisting of three spherical retarding grids, in accordance with ref. [16], with an energy resolution of ∼4 eV at 10 keV of electron beam energy was used.The primary beam energy for RHEED measurements was 10 keV in the present work.Energy-filtered RHEED patterns displayed on the phosphor screen were captured by a CCD video camera.
A mirror-polished Si(111) wafer (0.004 Ωcm, As-doped) with dimensions of 25 × 3 × 0.6 mm 3 was used as the substrate.The sample was resistively heated, and the sample temperature was measured by infrared and optical pyrometers.A clean Si(111) 7×7 surface was prepared by flashing at 1500 K for ∼5 s followed by annealing at 1250 K for ∼5 min under UHV.
Fe was deposited on the clean Si(111) surfaces with deposition amounts ranging from 0.5 to 50 ML.ML corresponds to the surface atomic density of the bulk truncated Si(111), and the deposition amount was measured using a quartz thickness monitor.Fe was deposited on the substrate heated to 920 K in the RDE method.In the case of the SPE method, Fe was deposited on a substrate at RT and the sample was then heated to 920 K and maintained at the temperature for ∼60 min.Both Fe deposition and heating were carried out under the vacuum of < 5 × 10 −7 Pa.
In order to characterize the crystallinity, the samples were investigated using Weissenberg RHEED at RT.The Weissenberg RHEED measurement was conducted by rotating the sample azimuthal angle by 60 • from the [2 11 ] incidence to the [1 21] incidence, taking advantage of the C 3V symmetry.The rotation interval was 0.1 • , which is sufficiently fine for a continuous survey of the reciprocal space.Thus, more than 600 RHEED patterns were obtained in total in a single azimuthal scan of Weissenberg RHEED.After the RHEED measurements, we observed the surface morphology of select samples by SEM.The samples were then transferred in air to an SEM chamber.

III. RESULTS AND DISCUSSION
A 3D image of the reciprocal space, wherein the diffraction intensities are mapped with the scattering vector s = (s x , s y , s z ), can be reconstructed from hundreds of RHEED patterns obtained by a single azimuthal scan in accordance with the Weissenberg method [13,14].The s x and s z axes are parallel to the [ 11 2] and [111] directions of the substrate, respectively.Sections of the reciprocal pattern for the sample prepared by RDE with 0.5 ML deposition are shown in Fig. 2. The vertical section at s y = 1.6 Å−1 is shown in Fig. 2(a) [17].The intensity was normalized by the maximum intensity of the 3D reciprocal pattern.We clearly see vertical rods arranged with even spacing as well as spots.The rods correspond to the reciprocal rods of the 1 × 1 surface unit of Si(111).These rods indicate that the bare Si(111) surface was largely exposed in the RDE 0.5 ML deposition sample.This is consistent with the SEM image shown in Fig. 3(a).Although very small and sparsely scattered islands are observed, most of the surface is flat in the SEM image.
In addition to the rods in Fig. 2(a), spots corresponding to the transmitted patterns of conventional RHEED measurement are also observed.These indicate the existence of 3D crystals, i.e., islands, on the surface.As demonstrated here, it is easy to identify the transmitted patterns from the 3D reciprocal data by the Weissenberg RHEED.Furthermore, it is easy to understand the 3D arrangements of these spots and rods from the reciprocal pattern in 3D form, and any section can be obtained in the 3D reciprocal pattern.The section parallel to both the A-A' line in Fig. 2(a) and the y axis is shown in Fig. 2(b).Spots are arranged in a square lattice with each side having a length of ∼2.3 Å−1 in Fig. 2(b).It is obvious that this section is parallel to one of the low index planes of the island crystal.Spots also exist at every center point on each side of the square, but not at the center of the square.From many sections of the 3D reciprocal image (not shown), we obtained a 3D unit of the reciprocal space, shown in Fig. 4(a), which is consistent with that of tetragonal α-FeSi 2 .As shown in Fig. 4 We measured the azimuthal-scan RHEED for the samples prepared by RDE up to 50 ML.Although the 1 × 1 rods disappeared above 2 ML, spots from the α-FeSi 2 islands were always observed in the 3D reciprocal images from the RDE samples.An SEM image of the RDE 50-ML sample is shown in Fig. 3(b).Thick α-FeSi 2 islands were observed on the surface.They did not fully cover the substrate surface, and some open flat surface areas remained between the islands.Since we could not see any rod structure in the reciprocal images, the remaining surface might be disordered, or the electron beam with grazing incidence could not reach the surface, blocked by the thick and dense islands.
For SPE growth, Fe was deposited on a RT substrate before annealing to form silicides.Sections of 3D recip-  [11,12].In addition to the diffuse spots from Fe(111) islands, 1 × 1 surface rods were also observed below ∼10 ML as shown in Fig. 5(a).Thus the Fe islands did not fully cover the substrate surface, which was partially exposed below ∼10 ML Fe deposition.
The Fe deposited sample was then annealed at 920 K for ∼60 min for SPE.Vertical sections at s y = 1.6 Å−1 of the 3D reciprocal pattern for the SPE sample with 10 and 50 ML depositions are shown in Figs.6(a) and (b), respectively.Comparing the section for the 10 ML SPE with that for the 0.5 ML RDE in Fig. 2(a), it is obvious that the spots from α-FeSi 2 islands also exist in Fig. 6(a).We also see some rods vertically crossing over α-FeSi 2 spots in Fig. 6(a).The rods become the main structure above 20 ML as shown in Fig. 6(b), while the spots from α-FeSi 2 disappear above 20 ML.The spacing between the rods is about half of that of the 1 × 1 rods (see Fig. 2(a)).We can see spot-like modulations on these rods, especially at higher s z in Fig. 6 Flat islands, which are very different from the thick α-FeSi 2 islands in Fig. 3(b), cover nearly half the surface.There are smaller (∼20 nm) but 3D islands between the flat islands in Fig. 3(c).The rod pattern of β-FeSi 2 and the spotty pattern of α-FeSi 2 suggest that the flat islands and the small 3D islands are islands of β-FeSi 2 and α-FeSi 2 , respectively.It was observed by SEM (not shown) that the flat islands were aggregated and covered most of the surface above 20 ML.
The 3D reciprocal patterns clearly show the crystal structures and the epitaxial orientations of the Fe silicides grown on Si(111) by the RDE and SPE methods.Only α-FeSi 2 islands grew by the RDE method up to 50 ML of Fe deposition.On the SPE sample, the α-FeSi 2 islands and the flat β-FeSi 2 islands co-exist at lower deposition amounts of < 20 ML.However, the intensity ratio of β-FeSi 2 pattern over α-FeSi 2 pattern increased with increasing Fe deposition, and α-FeSi 2 islands disappeared above 20 ML.It is interesting that the different silicides are formed depending on the deposition amount, while other growth conditions remain the same.Here, we consider what determines the species of Fe silicides formed.First, we carefully observed and compared the SEM images to the reciprocal images.It was found that flat surfaces were always exposed between islands in the SEM images when α-FeSi 2 islands are present on the sample.For RDE, the substrate or flat surface would always be exposed to Fe flux during the reaction.In the case of SPE, bcc Fe islands did not fully cover the substrate before the post-anneal when the deposition amount of Fe was less than 10 ML.This is roughly coincident with the existence of α-FeSi 2 islands on the SPE sample.The exposure of the bare substrate might contribute to the growth mechanism of α-FeSi 2 islands.We assume that diffusion and reaction on the surface of the Si substrate leads to growth of α-FeSi 2 on the Si(111) surface.For the RDE sample, Fe atoms from the deposition source might diffuse on the surface and then arrive at α-FeSi 2 islands.The Fe atoms then react with Si atoms and are captured by the α-FeSi 2 islands.When the SPE sample was annealed after RT deposition, Fe atoms might be released from the bcc Fe islands to the Si surface when the surface was not fully covered.Then, α-FeSi 2 islands formed and grew on the substrate surface.
On the other hand, the present β-FeSi 2 growth can be understood if β-FeSi 2 formed at the solid interface between the Si substrate and the bcc Fe islands.There was no Fe/Si solid interface during the RDE preparation when the deposition speed was not extremely fast.Thus, β-FeSi 2 was not formed by RDE in the present study.When the Fe deposited sample was annealed in the SPE method, β-FeSi 2 started to grow at the solid interface between the Fe islands and the Si substrate.If the Fe islands fully cover the substrate, there is no surface for α-FeSi 2 formation, and then only β-FeSi 2 grew from the interface.Our growth model for β-FeSi 2 does not contradict the growth model proposed by Hattori et al. in refs. [11] and [12].They found an Fe-rich silicide layer near the interface.β-FeSi 2 might be formed on the surface through such an Fe-rich interface region.Although specific reactions or the exact mechanism could not be elucidated from the present study, the surface or interface reaction model explains the present α-or β-phase selection by the RDE and SPE methods very well.

IV. CONCLUSION
Fe silicide growth on an Si(111) surface by RDE and SPE methods for various deposition amounts of Fe has been investigated using Weissenberg RHEED and SEM.The exact silicide species and epitaxial orientation were determined from the 3D reciprocal images obtained by azimuthal-scan RHEED.Only α-FeSi 2 islands grew by the RDE method, up to 50 ML Fe.Both α-FeSi 2 and β-FeSi 2 islands grew by the SPE method when the deposition amount of Fe was less than 20 ML.The ratio of β-FeSi 2 increased with increasing amount of Fe, and only β-FeSi 2 was observed above 20 ML Fe.We have proposed a surface or interface reaction model in which the reaction at the Si surface and the reaction at the Fe/Si solid interface results in α-and β-FeSi 2 growth, respectively.

FIG. 2 :
FIG. 2: Sections of the reciprocal map obtained by Weissenberg RHEED for the RDE 0.5 ML sample.(a) Vertical section at sy = 1.6 Å−1 , which is parallel to the [ 11 2] direction of the substrate.(b) Section perpendicular to (a) along A-A'.
FIG. 3: SEM images of (a) the RDE 0.5 ML sample, (b) the RDE 50 ML sample, and (c) the SPE 10 ML sample.

FIG. 4 :
FIG. 4: (a) Reciprocal unit obtained from the present study for the RDE 0.5 ML sample.Crystalline directions are indicated by cubic indexes.(b) Unit cell of tetragonal α-FeSi2.(c) Reciprocal unit cell of α-FeSi2.
FIG. 5: Vertical sections at sy = 1.6 Å−1 of the reciprocal map obtained by Weissenberg RHEED for the as-deposited surface for (a) 5 ML Fe and (b) 50 ML Fe.