抑制矽化物生成的低溫鐵薄膜之成長與磁性研究

全文

(1)國立臺灣師範大學物理研究所碩士論文. 抑制矽化物生成的低溫鐵薄膜之成長與磁性研究 Growth and Magnetism of Low-temperature Deposited Fe/Si(111) Films as an Intermediate Layer for Suppression of Silicide Formation. 研究生：涂文廷指導教授：林文欽博士. 中華民國一百年一月.

(2) ABSTRACT. Low temperature (LT: 100 K) deposition of Fe on Si(111)7x7 surface effectively reduces Fe-silicide formation at the Fe/Si interface, as compared with conventional room temperature (RT) growth. The interface condition of 5-15 monolayer (ML) LT-Fe/Si(111) remains stable at least up to 350 K. Si segregation was observed after annealing at 400 K. LT-grown Fe films also reveal a relatively flat surface morphology with a roughness of 0.4-0.6 nm. Thus, LT-Fe films were suggested as an intermediate layer for the subsequent RT-growth of Fe. We use a single domain model of magnetic anisotropy to fit the magnetic coercivity evolution of n ML RT-Fe on 5 ML LT-Fe/Si(111). Accordingly, we deduce the surface and volume-contributed magnetic anisotropy for discussion.. i.

(3) Content Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. i. Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. ii. List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . .. iv. List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . .. v. 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1. 1.1. Si(111)7x7 . . . . . . . . . . . . . . . . . . . . . . . . .. 2. 2 Basic Concept . . . . . . . . . . . . . . . . . . . . . . . . . .. 8. 2.1. Adsorption of Molecules . . . . . . . . . . . . . . . . .. 8. 2.1.1. Modes of Adsorption . . . . . . . . . . . . . . .. 8. 2.1.2. Rate of Adsorption . . . . . . . . . . . . . . . .. 8. 2.1.3. Crystal . . . . . . . . . . . . . . . . . . . . . . .. 11. 2.1.4. Wood’s Notation . . . . . . . . . . . . . . . . .. 13. 2.2. Film growth . . . . . . . . . . . . . . . . . . . . . . . .. 16. 2.3. Ferromagnetism . . . . . . . . . . . . . . . . . . . . . .. 18. 2.3.1. Origin of magnetism . . . . . . . . . . . . . . .. 19. 2.3.2. Anisotropy energy . . . . . . . . . . . . . . . .. 20. 2.3.3. Magnetic hysteresis loop . . . . . . . . . . . . .. 21. ii.

(4) 2.3.4. Stoner-wahlfarth model . . . . . . . . . . . . . .. 22. 3 Experimental Apparatus . . . . . . . . . . . . . . . . . . . .. 27. 3.1. Multi-functional UHV systems . . . . . . . . . . . . . .. 27. 3.2. Scanning Tunneling Microscopy (STM) . . . . . . . . .. 32. 3.3. AES and LEED . . . . . . . . . . . . . . . . . . . . . .. 39. 3.4. Magneto-Optical Kerr effect(MOKE) . . . . . . . . . .. 42. 4 Experiment and Results . . . . . . . . . . . . . . . . . . . . .. 48. 4.1. LT(100K)-growth of Fe/Si(111) . . . . . . . . . . . . .. 51. 4.2. MOKE measurement . . . . . . . . . . . . . . . . . . .. 57. 5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 64. iii.

(5) List of Tables. 2.1. Typical characteristics of adsorption processes [30] . . .. 9. 3.1. Description of UHV chamber in C207, NTNU . . . . .. 31. 3.2. Description of MOKE chamber in C207, NTNU . . . .. 45. iv.

(6) List of Figures. 1.1. Reconstructed 7x7 structure of the silicon (111) face. .. 1.2. Schematic of various silicide phase diagram in different coverage and annealing temperature [23] . . . . . . . .. 1.3. 4. 5. Thickness dependencies of the remanent Kerr intensities for ultrathin Co/6 ML Ag/Ge(111)(squares) and Co/Ge(111)(circles) films grown at 300 K on the longitudinal configuration. The nonferromagnetic layer is efficiently reduced to zero by using 6 ML Ag as a buffer layer. [5] . . . . . . . . . . . . . . . . . . . . . . . . . .. 5. 2.1. Planes with different directions . . . . . . . . . . . . .. 14. 2.2. fcc(100) surface . . . . . . . . . . . . . . . . . . . . . .. 15. 2.3. fcc(111) surface . . . . . . . . . . . . . . . . . . . . . .. 15. 2.4. fcc(100) surface; (2x2) structure . . . . . . . . . . . . .. 15. 2.5. √ √ fcc(100) surface; ( 2x 2)R45 . . . . . . . . . . . . . .. 15. 2.6. Schematic display of the ideal growth mode. (a) Layerby-layer growth. (b) Layer plus island growth. (c) Is-. 2.7. land growth. . . . . . . . . . . . . . . . . . . . . . . . .. 16. Schematic illustration of dynamic processes at a surface. 17. v.

(7) 2.8. An example of hysteresis loop. The magnetization at zero external field is the remenance. The external field at zero net magnetization is called the coercive field and. 2.9. coercivity. . . . . . . . . . . . . . . . . . . . . . . . . .. 22. Illustration of the Stoner-Wolhfarth model. . . . . . . .. 23. 2.10 An example solution of the Stoner-Wolhfarth model. The solid A and B curves are energy minimum, the dashed A and B lines are energy maxima. . . . . . . .. 25. 2.11 Some hysteresis loops predicted by the Stoner-Wolhfarth model for different angles between the field and easy axis. 26 3.1. The UHV system in lab. C207 in NTNU . . . . . . . .. 29. 3.2. The diagram of the pumping process. . . . . . . . . . .. 30. 3.3. Schematic diagram of electron tunneling . . . . . . . .. 36. 3.4. The loop of a working STM(Text and graphics by Michael Schmid, TU Wien) . . . . . . . . . . . . . . . . . . . .. 38. 3.5. STM constant current mode . . . . . . . . . . . . . . .. 39. 3.6. STM constant height mode . . . . . . . . . . . . . . . .. 39. 3.7. Process for demagnetization of atomic core holes. (a) Emission of X-ray radiation. (b) Emission of an AES. 3.8. electron. . . . . . . . . . . . . . . . . . . . . . . . . . .. 40. Schematic display of a LEED structure . . . . . . . . .. 41. vi.

(8) 3.9. Schematic illustration of magneto optical Kerr effect. After reflected from the ferromagnetic sample, the linear polarized laser beam becomes elliptical polarized. .. 42. 3.10 Different geometry for MOKE measurement . . . . . .. 43. 3.11 Schematic illustration of AC MOKE . . . . . . . . . .. 44. 3.12 MOKE chamber in NTNU C207 Lab . . . . . . . . . .. 46. 3.13 Specific gasket for LT-MOKE measurement. . . . . . .. 47. 3.14 Cu wire for cooling sample . . . . . . . . . . . . . . . .. 47. 3.15 Schematic display of a DC-MOKE loop in C207, NNU. 47. 4.1. The whole experiment process . . . . . . . . . . . . . .. 48. 4.2. 12 ML RT-growth Fe/Si(111). The signal of carbon is much smaller than iron and silicon after 12 ML Fe deposition. . . . . . . . . . . . . . . . . . . . . . . . . .. 49. 4.3. The process of preparation Si(111)7x7. . . . . . . . . .. 49. 4.4. Si(111) 7x7 image obtained in our laboratory . . . . . .. 50. vii.

(9) 4.5. Auger ratios of Fe47eV /Si92eV (left axis) and Si92eV /Fe47eV (right axis) measured as a function of Fe thickness. The thickness means the total thickness of Fe deposition on Si(111). The solid circles and solid squares indicate the Fe/Si AES ratio of RT-grown Fe films and RT-Fe/5 ML LT-Fe films on Si(111), respectively. The open circles and open squares indicate the Si/Fe AES ratio of RT-Fe and RT-Fe/5 ML LT-Fe films on Si(111), respectively. The solid and dotted curves are the fitting results by Eq. (4.1) and (4.2). . . . . . . . . . . . . . . . . . . . .. 52. 4.6. Schematic illustration of AES electron emission. . . . .. 53. 4.7. The relation of the current passing through the sample and the temperature. . . . . . . . . . . . . . . . . . . .. 4.8. 56. Auger ratios of Si/Fe measured as a function of annealing temperature. All the LT-grown Fe films are annealing to RT at first and then annealing to different temperature. The circles, squares, and triangles indicate the series data of 5, 15 and 25 ML LT-grown Fe/Si(111) films. The solid lines are guides to the eye. .. viii. 56.

(10) 4.9. Surface morphology of (a) 24 ML and (b) 40 ML LTgrown Fe films, investigated by scanning tunneling microscopy (STM). The figures are all of 120x120 nm2 . The insets (30x30nm2 ) reveal the detailed surface structures. The line profiles indicated in the STM images are plotted at the bottom. . . . . . . . . . . . . . . . . . .. 58. 4.10 Polar MOKE hysteresis loops measured at room temperature (RT) for n ML of RT-grown Fe on a intermediate layer of 5 ML LT-grown Fe on Si(111). . . . . . .. 59. 4.11 Longitudinal MOKE hysteresis loops measured at room temperature (RT) for n ML of RT-grown Fe on a intermediate layer of 5 ML LT-grown Fe on Si(111). . . . .. 59. 4.12 Kerr saturation signal (Ms : left-axis) and coercivity field (Hc : right axis) plotted as function of Fe coverage. The data points were analyzed from the hysteresis loops of n ML RT-Fe/5 ML LT-Fe/Si(111) in Fig. 4.11. The solid line and curve are the fitting results of Ms and Hc using a linear function and Eq.4.6, respectively.. ix. 60.

(11) Chapter 1. Introduction. The science of combining metallic thin films with semiconductor gets development continually. Especially the growth of magnetic materials, such as Fe and Co [1–7], on semiconductor surfaces has been widely studied because of the potential applications combining magnetism and semiconductor techniques, so called spintronics. It uses the quantum mechanic characteristics of electron spin. For example, many spin in non-magnetic semiconductor has much longer coherent time than electronic polarization, and it can be controlled by electric field. Therefore it is easy to realize the spin controlling in a quantum mechanical system. The characteristic can promote the processor of quantum information development, like modulator. Compared with the traditional semiconductor components, it has the advantage of good stability, faster in data processing, reducing the power loss. And according to theory prediction and experiment results, it can satisfy the operating condition that the Currie temperature reaches to room temperature. So ferromagnet/semiconductor compound materials are suitable to used on electric components in new generation. In our case, we try to deposit Fe thick films on Si(111) 7x7 substrate. The materials with ferromagnetism includes Fe, Co and Ni. In the paper review, we find many studies about Co, and the magnetism of Ni is much smaller than Fe. Therefore we choose to study the ability of Fe/Si(111)7x7. In the process of Fe deposition, it will result in many silicide. The silicide is hard to exist at the state of 1.

(12) single crystalline and it has no magnetism. So we do not wish to find the formation of silicide and try to solve the problem.. 1.1. Si(111)7x7. Semiconductors, especially silicon, being the basic material for electronic devices, really change our life a lot. In recent years, crystalline samples can be manufactured in ways, thanks to silicon-based computers, as a matter of fact that their surface has a high degree of perfection [8]. Moreover, surface observation with atomic resolution in tunneling microscopy is ever easier on semiconductors than that on most metals. The crystal structure of silicon is diamond cubic. In this structure, due to the hybridization of s and p electrons, stems from the fact that each atom trend to have four neighbours arranged in regular tetrahedron. Cleaving a highly covalent body in two parts to create a surface, one breaks bonds which are now ”dangling”. They recombine to lower their energy resulting in a surface reconstruction, a typical phenomenon for all semiconductors. The (001) and (111) surface orientations are especially used for growth of microelectronic components. In (001) surface, the diamond lattice has two atoms per unit cell, thus it is not a Bravais lattice. In practice, (001) surfaces are most often used in molecular beam epitaxy growth, and some steps are always present. The (111) surface is characterized by a very complicated recon2.

(13) struction. To understand its nature, we consider a 1x1 non-reconstruction structure first. In a 1x1 structure, each surface atom has a dangling bond. Compared to(001) one can easily reduce this number by placing one adatom on the head of each group of three atoms. This adatom possesses one unsaturated bond, but the number of dangling bonds per unit surface reduced by a factor three. However, such adatoms deeply perturb the other atoms. On silicon, this perturbation favours the appearance of dimer bonds, stacking faults, and even surface vacancies. The outcome is the celebrated DAS (dimer-adatom-stacking-fault) 7x7 model, which is sketched in figure. It includes 12 adatoms, 9 dimer bonds and a stacking fault in each unit cell. The order of magnitude for surface energies is the electron volt. But energy difference may be much smaller. For instance, take the cleavage (111) plane of Silicon. This surface reconstructs after cleavage in metastable 2x1 structure, which transforms into the 7x7 upon annealing. ab-initio calculations showed that the energy difference between the two structures is 60 mV pre 1x1 cell [9,10]. For a long time, the serious silicide formation at the transition metal/Si interface has been always a key issue in this field [11–18]. The silicide formation not only randomizes the crystalline structure and electronic structure near interface, but also leads to thick magnetic dead layer and unstable magnetic anisotropy [7]. From figure1.2 we see it is easily for iron grown on silicon surface to form iron silicides, we try to use MBE method to grow iron on silicon surface under lower temperature to reduce the formation of ”alloy”, therefore we 3.

(14) Figure 1.1: Reconstructed 7x7 structure of the silicon (111) face.. may possibly study the reaction of real iron epitaxial growth on silicon surface. By experiment, we know the way that iron deposition at lower temperature can only reduce the formation of iron silicide but cannot prevent silicide production completely. Here we define a unity monolayer (ML) as the surface atomic density of an unreconstructed Si(111) plane: 7.83×1014 atoms/cm2 about 0.8˚ A thick Fe coverage. [23]. So, how to suppress the silicide formation is really important for us. Some researchers gets an idea to put something between the metal/semiconductor to achieve this effect. Up to now, many buffer layers were proposed and demonstrated for the suppression of silicide formation [5,17–22]. In some cases, the buffer layer needs to be thick. 4.

(15) Figure 1.2: Schematic of various silicide phase diagram in different coverage and annealing temperature [23]. Figure 1.3: Thickness dependencies of the remanent Kerr intensities for ultrathin Co/6 ML Ag/Ge(111)(squares) and Co/Ge(111)(circles) films grown at 300 K on the longitudinal configuration. The nonferromagnetic layer is efficiently reduced to zero by using 6 ML Ag as a buffer layer. [5]. 5.

(16) enough in order to block the chemical activity of semiconductor surface [5,17,18,22]. For example, a 6 monolayer (ML) Ag buffer layer was demonstrated to block the alloy formation in Co/Ge(111) [5]. From Fig1.3, we can see that the remanent Kerr intensity of Co/6 ML Ag/Ge(111) and Co/Ge(111) are zero up to 2ML and 8ML respectively. And subsequent Co deposition causes the linear increase of the magnetism on both the two systems. It’s successful to reduce the formation of silicide and the dead layer for magnetism. In other cases, deposition of sub-monolayer buffer atoms with high temperature annealing can result in a very stable superstructure on the semiconductor surface, which provides a sharp interface for the subsequent deposition of magnetic materials [19]. Nevertheless, inserting an intermediate layer of different materials might increase more complexity in changing the original properties of semiconducting substrate and magnetic deposit. Therefore we propose the idea of using a low temperature(LT:100K)-deposited Fe layer, instead of the buffer with different materials, as an intermediate layer for suppression of silicide formation. The Fe/Si interface might be frozen and fixed because of the low mobility and small thermal activation energy at LT. Such interface is expected to be stable not only at LT, but also after annealing to room temperature(RT:300K), owning to the binding from subsequent deposit layers. With this idea of using LT-Fe as the buffer layer, not only the planar Si surface, but also many recently developed Si-based nanostructures [24–29] can be used as suitable templates for magnetic material deposition without introducing new materials in between or destroying the nano pattering on the substrate. 6.

(17) In this study, the Fe/Si interface, thermal stability, and surface morphology of LT:100 K-deposited Fe/Si(111) were investigated. For demonstration, the 5ML LT-Fe film was used as an intermediate layer for the subsequent RT-growth of Fe. In the coverage-dependent magnetic measurement, the evolution of magnetic anisotropy waa also deduced.. 7.

(18) Chapter 2 2.1. Basic Concept. Adsorption of Molecules. 2.1.1. Modes of Adsorption. The adsorption of molecules on surfaces has two principal modes :. 1. Physical Adsorption : the bonding is contributed from the Van der Waals force. It’s a quite weak force and there is no difference of electronic structure in the molecule and substrate. 2. Chemical Adsorption : a chemical bond formed by molecule and substrate. The electronic structure in molecule and substrate surface has an obviously variation.. From table2.1, we can see the typical characteristics of adsorption process. 2.1.2. Rate of Adsorption. The rate of the molecules adsorbed on the surface can be written as [30] :. 8.

(19) chemical adsorption. physical adsorption. Material Specificity (variation. Substantial variation. Slight. between substrates of different. between materials. upon substrate com-. chemical composition) Crystallographic. dependence. position. Specificity. (variation between different. Marked variation be-. Virtually independent. tween crystal planes. of surface atomic ge-. surface planes of the same. ometry. crystal) Temperature. Range. which adsorption occurs). (over. Virtually. unlimited. Near or below the con-. (but a given molecule. densation point of the. may effectively adsorb. gas (e.q. Xe < 100K,. only. CO2 < 200K). over. a. small. range) Adsorption Enthalpy. Wide range (related. Related to factors like. to the chemical bond. molecular mass and. strength)-typically 40-. polarity - typically 5-. 800 kJ mol−1. 40 kJ mol−1 (similar to heat of liquefaction). Nature of Adsorption. Saturation Uptake. Kinetics of Adsorption. Often dissociative. Non-dissociative. May be irreversible. Reversible. Limited to one mono-. Multilayer. layer. possible. Very variable - often. Fast - since it is a non-. an activated process. activated process. Table 2.1: Typical characteristics of adsorption processes [30]. 9. uptake.

(20) 0. Rads = k P χ. where. χ : kinetic order 0. k : rate constant P : partial pressure if the rate constant is then expressed in Arrhenius form, then we obtain the equation of the form :. Rads = Aexp(−Ea /RT )P χ. (2.1). where Ea is the activation energy for adsorption, and A the preexponential (frequency) factor. As the molecular level, the rate of the molecules arrive at the surface and the proportion of incident molecules which undergo adsorption will govern the rate of adsorption. So we can rewrite the rate by incident molecular flux,F ,and the sticking probability,S . Rads = SxF [molecules · m−2 s−1 ]. (2.2). F = P/(2ΠmkT )1/2 [molecules · m−2 s−1 ]. (2.3). and. p : gaspressure[N m−2 ] where. m : massof onemoleculekg. T : temperature[K] The sticking probability,S, must lie in the range 0 < S < 1 and it clearly depends on the character of adsorbate and substrate : existing coverage of adsorbed species (θ) and the presence of activation barrier 10.

(21) to adsorption. Therefore : S = f (θ)·exp(−Ea /RT ). (2.4). where, f (θ) is a function of the existing surface coverage of adsorbed species. Combining the equations for S and F : f (θ)·P R=√ exp(−Ea /RT ) 2πmkT 2.1.3. (2.5). Crystal. The materials deposited on single crystal substrate surface are often found that they have a long-range ordering and a well-defined overlayer structure. In order to obtain the result, the lattice match plays an important role in the process. At the same time, the ordering arrangement of structure exists over a limited coverage range of the adsorbate because the adsorbate and substrate have their own lattice constant respectively. At first, the falling deposited atoms may pile on the substrate surface along the arrangement of the surface crystal by the weak Van der Waals force. When the amount of adsorbate is more and more, the atoms deposited will order according to their original characteristic getting rid of the weak force. Of course, there are still many other factors in the process. We just make preliminary understanding here.. Unit cell. The crystal structure of a material or the arrangement of atoms with a given type of crystal structure can be described in terms of 11.

(22) its unit cell. The unit cell is a tiny box containing one or more atoms, a spatial arrangement of atoms. The unit cells stacked in three-dimensional space describe the bulk arrangement of atoms of the crystal. The crystal structure has a three dimensional shape. The unit cell is given by its lattice parameters, the length of the cell edges and the angles between them, while the positions of the atoms inside the unit cell are described by the set of atomic positions (xi , yi , zi ) measured from a lattice point.. Planes and directions. The crystallographic directions and planes are fictitious linking some nodes like atoms, ions or molecules. The surface can be understood as the direction of its normal vector(Fig. 2.1). Some directions or planes have the higher density of the atoms and they will influence the behavior of the crystal as follows : 1. Optical properties: Refractive index is directly related to density (or periodic density fluctuations). 2. Adsorption and reactivity: Physical adsorption and chemical reactions occur at or near surface atoms or molecules. These phenomena are thus sensitive to the density of nodes. 3. Surface tension: The condensation of a material means that the atoms, ions or molecules are more stable if they are surrounded by other similar species. The surface tension of an interface thus varies according to the density on the surface. 12.

(23) 4. Microstructural defects: Pores and crystallites tend to have straight grain boundaries following higher density planes. 5. Cleavage: This typically occurs preferentially parallel to higher density planes.. 2.1.4. Wood’s Notation. When we deposit the atoms on the surface of a single crystal substrate, the surface is composed of repeated translation of unit cell, the adsorbate may also form the ordering arrangement. So we need to make sure how to describe the structure of the substrate and adsorbate adequately at first. Now we use the clean surface structures of the surface planes of fcc metals for examples. From figure2.22.3 ,we can select anyone of the two possible choices of unit cell(marked). It is obvious that the size of a unit cell needs to be the simplest and repeatable unit on this surface. They have the same shape, size and orientation but differing in their translational position. You can choose the unit cell what you prefer. Wood’s notation is the simplest and frequently used method for describing a surface structure. But it has some limits for using that the unit cell of adsorbate must be the same or closely-related symmetry to the substrate, i.e. so called lattice match. For example: as figure2.4, when |b~1 | = 2|a~1 | and |b~2 | = 2|a~2 |, we can call it (2x2) overlayer. As if 13.

(24) Figure 2.1: Planes with different directions. 14.

(25) Figure 2.3: fcc(111) surface Figure 2.2: fcc(100) surface. the structure like the figure2.5, the unit cell of the adsorbate rotates √ √ 450 with the substrate. We can use the ”( 2 × 2)R45” to express the overlayer.. Figure 2.4: fcc(100) surface; (2x2) struc-. √ √ Figure 2.5: fcc(100) surface; ( 2x 2)R45 structure. ture. 15.

(26) 2.2. Film growth. In a phenomenological analysis, three modes of film growth can be distinguished, as shown in Fig. 2.6. In (a), each new layer starts to grow only when the last one has been complete, and thus called layer by layer growth. The opposite case in (c) shows that only 3D island deposit exists. In (b) after formation of the first or several complete monolayers, island growth starts. These modes is related to γS the free surface energy of the substrate/vacuum interface, γF that of film/vacuum, and γF/S that of film/substrate as the following : 1. γS γF + γF/S ⇒ islandgrowth 2. γS ≤ γF + γF/S ⇒ layergrowth. Figure 2.6: Schematic display of the ideal growth mode. (a) Layer-by-layer growth. (b) Layer plus island growth. (c) Island growth.. Many factors might be responsible for the layer-plus-island growth, such as the lattice mismatch between the substrate and the deposited. 16.

(27) film and alternatively, the symmetry or orientation of the overlayers with respect to the substrate. Additionally, if the microscopic kinetic processes taking place at the surface is taken into account, the substrate topology can be found to have much influence on the growth mode. The possible kinetic processes taking place at the surface is shown in Fig. 2.7. Figure 2.7: Schematic illustration of dynamic processes at a surface. 17.

(28) 2.3. Ferromagnetism. Ferromagnetism is a property referring to the magnetic condition of a material which has the spontaneous magnetization phenomenon and iron is the most widely known in all of the materials. The attraction between a magnet and ferromagnetic material is the quality of the magnetism first apparent to the ancient world, and to us today. Some materials have been magnetized by an external magnetic field, and remain their magnetization with the absence of the additional field, so called spontaneous magnetization. The term ferromagnet was used for any material that could exhibit spontaneous magnetization : a net magnetic moment in the absence of an external magnetic field. And the definition is still common used now. By the investigation of magnetism, we can make a more stricter definition now. A material which magnetic domains add a positive contribution to the net magnetization is called ”ferromagnet”. If the moment of aligned and anti-aligned domains balance completely that have zero net magnetization, despite the magnetic ordering, it is an antiferromagnet. All of these alignment effects only occur at temperatures below a certain critical temperature, called Curie temperature (for ferromagnets) or Néel temperature (for antiferromagnets).. 18.

(29) 2.3.1. Origin of magnetism. The spin of an electron, combined with its electric charge, results in a magnetic dipole moment and creates a small magnetic field. Although an electron can be visualized classically as a spinning ball of charge, spin is actually a quantum mechanical property with differences from the classical picture, such as the fact that it is quantized into discrete up/down states. The spin of the electrons in atoms is the main source of ferromagetism, although there is also some contribution from the orbital angular moment of the electron about the nucleus, whose classical analogy is a current loop. When these tiny magnetic dipoles are aligned in the same direction, their individual magnetic fields add together to create a measurable macroscopic field. However in many materials, the total dipole moment of all the electrons is zero because the spins are in up/down pairs. Only atoms with partially filled shells(i.e. unpaired spins) can have a net magnetic moment, so ferromagnetism only occurs in materials with partially filled shells. Because of Hund’s rules, the first few electrons in a shell tend to have the same spin, thereby increasing the total dipole moment. These unpaired dipoles (often called simply ”spins” even though they also generally include angular momentum) tend to align in parallel to an external magnetic field, an effect called paramagnetism. Ferromagnetism involves an additional phenomenon, however: the dipoles tend to align spontaneously, without any applied field.. 19.

(30) 2.3.2. Anisotropy energy. According to classical electromagnetism, two nearby magnetic dipoles tend to align in opposite directions. However in some ferromagnetic materials, they tend to align in the same direction because of the Pauli exclusion principle : two electrons with the same spin can’t have the same position. But it does not against the electrons have opposite spin. So, the electrostatic energy of system will decide by the related direction of the spin. The difference in energy is called the ”exchange energy”. Although the exchange energy let the spins aligned, it doesn’t align them in any particular direction. Without magnetic anisotropy, the spins will change directions randomly. There are several kinds of magnetic anisotropy.. 1. magnetocrystalline anisotropy : This is a dependence of the energy on the direction of magnetization relative to the crystallographic lattice. 2. shape anisotropy : When a magnetic material has been magnetized, it will form the surface magnetic pole in the both sides. And the correlation will produce a equivalent magnetic field with opposite direction to the magnetized direction. The intensity of the field is dependent with the shape of the sample and the intensity of the magnetization. 3. exchange anisotropy : It’s a phenomenon which produced by the 20.

(31) exchange coupling formed by interface between ferromagnetic and antiferromagnetic materials.. 2.3.3. Magnetic hysteresis loop. Hysteresis loop is one of the most distinctive experimental facts of ferromagnetism. An example is shown in Fig2.8. Loops like this are obtained by applying to the sample a cyclic magnetic field H and by recording the ensuing change of the magnetization M along the field, where M is defined as the average magnetic moment per unit volume. Hysteresis lops may be of many different shapes, thus it’s important to get some parameters to characterize the loop properties. Two quantities of particular importance are the remanent magnetization or remanence, Mr , and the coercive field or coercivity, Hc . As indicated in Fig2.8, remanence represents the magnetization obtained by applying a magnetic field and then moving it, and coercivity is the field needed to bring the remanence to zero. Unlike remanence, the coercive field spans an very wide interval, from less than 1 A/m to more than 106 A/m, in different specimens. The shape of a hysteresis loop strongly depends on not only the intrinsic properties of the specimen but also the external factors such as the measurement method, the specimen geometry etc.. Therefore, we have to make sure that the loops are measured in the same process, specimen geometry etc before comparing their results. The variety of the hysteresis loop shape is the direct consequence of the variety of possible magnetic domain structure. The primary mechanisms of the hysteresis loop are magne21.

(32) tization rotation and wall motion. In general, wall motion can occur in low field, and magnetization rotation needs strong applied field to overcome the energy barrier.. Figure 2.8: An example of hysteresis loop. The magnetization at zero external field is the remenance. The external field at zero net magnetization is called the coercive field and coercivity.. 2.3.4. Stoner-wahlfarth model. The Stoner-Wohlfarth model [31–36] is a widely used model for the magnetization of single domain ferromagnets. It is a simple example of magnetic hysteresis and is useful for simulating small magnetic particles in magnetic storage. In the Stoner-Wohlfarth model, the magnetization does not vary within the ferromagnet and it is represented by a vector M . This 22.

(33) vector rotates as the magnetic field H changes. The magnetization changes only in one axis. It is positive in one direction and negative in the opposite direction. The ferromagnet is assumed to have a uniaxial magnetic anisotropy with anisotropy parameter Ku . As the magnetic field varies, the magnetization varies in the plane containing the magnetic field direction and the easy axis. It can therefore be represented by an angle φ between the magnetization and the field and the angle θ between the field and the easy axis.(Fig. 2.9). Figure 2.9: Illustration of the Stoner-Wolhfarth model.. The energy of the system can be written as. E = Ku sin2 (φ − θ) − Ms Hcosφ. (2.6). where Ms is the saturation magnetization. The first term is the magnetic anisotropy and the second term is the Zeeman energy. A given magnetization direction is in mechanical equilibrium if the forces on it are zero. This occurs when the first derivative of the energy with 23.

(34) respect to the magnetization direction is zero :. ∂E = Ku sin(2(φ − θ)) + Ms Hsinφ = 0 ∂φ. (2.7). This direction is stable against perturbations when it is at energy minimum, having a positive second derivative :. ∂ 2E = 2Ku cos(2(φ − θ)) + Ms Hcosφ > 0 ∂φ2. (2.8). In zero field the magnetic anisotropy term is minimized when the magnetization is aligned with the easy axis. In a large field, the magnetization is pointed towards the field.. hysteresis loop. For each angle θ between easy axis and field, equation (2.7) has a solution which includes two solution curves. One of the curves is for φ between 0 and π and another one is for φ between π and 2π, the solutions at φ = 0andπ correspond to h = ±∞. Because the magnetization in the direction of the field is Ms cosφ, we usually plot the curves in the normalized form mh vs. h, where mh = cosφ is the component of magnetization in the direction of the field. An example is shown in Fig 2.10. The solid curves connect all the stable magnetization directions. In the magnetic hysteresis measurement, h starts from a large positive value to a large negative value. The magnetization direction starts on the A curve. At h = 0.5 24.

(35) the B curve appears, but for h > 0 the A state has a lower energy (equation4.3)because it is closer to the direction of the magnetic field. When the field becomes negative, the B state has the lower energy, but the magnetization cannot immediately jump to this new direction because there is an energy barrier. At h = −0.5, the energy barrier disappears, and in more negative fields the A state no longer exists. It must therefore jump to the B state. After this jump, the magnetization remains on the B curve until the field increases past h = 0.5, where it jumps to the A curve.. Figure 2.10: An example solution of the Stoner-Wolhfarth model. The solid A and B curves are energy minimum, the dashed A and B lines are energy maxima.. dependence of field direction. The shape of the hysteresis loop has a strong dependence on the angle between the magnetic field and the easy axis (Fig. 2.11). If the two are parallel (θ = 0), the hysteresis loop is at its biggest (with mh = hs = 1 in normalized units). The magnetization starts parallel 25.

(36) to the field and does not rotate at all until it becomes unstable and jumps to the opposite direction. In general, the larger the angle, the more reversible rotation occurs. At the other extreme of θ = 900 , with the field perpendicular to the easy axis, no jump occurs. The magnetization rotates continuously from one direction to the other (it has two choices of rotation direction, though). For a given angle θ, the switching field is the point where the 2. solution switches from an energy minimum ( ∂∂φE2 > 0) to an energy 2. maximum ( ∂∂φE2 < 0). Thus, it can be calculated directly by solving equation (2.7) along with. Figure 2.11:. ∂2E ∂φ2. = 0.. Some hysteresis loops predicted by the Stoner-Wolhfarth model for. different angles between the field and easy axis.. 26.

(37) Chapter 3 3.1. Experimental Apparatus. Multi-functional UHV systems. The thickness of the deposited layer we made is about only few mono layer, that any nano-scale unexpected deposition, or, pollution will contribute unknown effect to our measurement, or kill the weak micro-phenomenon we want to observe. For example, Lieberman discovered the pollution will result in magnetic dead layer in 1969 [37,38], and Hope found that only 0.08 mono layer CO falls on Co/Cu(110) will also change the easy axis [39]. For the purpose to avoid the sample surface being polluted, an ultra high vacuum (< 10−10 ) is basically recommend. The other way, we can also sketch the pollution rate with respect to the base pressure by the dynamics of gas. We know the frequency of the gas atoms collide the sample surface is nva 4. r=. (3.1). where the root mean square velocity of a gas molecular of mass m under Kelvin degree T is 2 υrms =. 3kB T m. where and the mean velocity of gas is r υa = υrms r =. 8 3π. 8kB T mπ. 27. (3.2). (3.3) (3.4).

(38) Combine these into ideal gas formulation P = nkB T. (3.5). We can get r=√. P 2πkB mT. (3.6). Substituting, making this formula more readily useful by expressing P in torr and merging all convention factors into a constant. The result is. P (torr) r = 3.52 × 1022 p /cm2 s m(a.m.u)T (K). (3.7). Using nitrogen of mass 28, room temperature is 300K and the nitrogen pressure is 1 × 10−6 torr to demonstrate the frequency of gas collide on the sample surface, we obtain r = 3.84×1014 bombardments /cm2 s. Assume that 100% nitrogen molecules will stick on the sample surface, after 0.1 seconds, will have about 2 ML nitrogen atoms on sample surface of 0.5cm2 . In our UHV system, we can divide it into several areas roughly including the load lock, main chamber, MOKE chamber and the ”beer barrel” which is an important and interesting design that will be illustrated later. In the pumping system, two sides of the main chamber link the load lock and MOKE chamber with valves respectively. The load lock and MOKE chamber also link the ”beer barrel” with valves respectively. There are two paths to discharge the gas out of the chamber. One of the paths is going from main chamber through the load 28.

(39) Figure 3.1: The UHV system in lab. C207 in NTNU. lock to the barrel and then into the air; and the other one is through the MOKE chamber. By these valves, we can control the pumping paths and lock the vacuum of the chamber optionally. The ”beer barrel” is used for the buffer between the UHV system and the air because of the difficulty for turbo to discharge the air into air directly. In usually, we will open the valve and turn on the mechanic pump to let the pressure of the ”buffer” be about 10−2 10−3 torr four times a day to reduce the consumption of the turbo. Fig3.2 is the diagram of pumping process. To reach UHV condition, the pumping system and process play an important roles. Usually we use a mechanical pump to fore-pumping. 29.

(40) Figure 3.2: The diagram of the pumping process.. down to the pressure about 10−2 to 10−3 torr, then a turbo pump is capably on to help further pumping to 10−7 torr while the main chamber is also baked by about 120 0 C for 24 hours and reach about 10−8 to 10−10 torr after baking stopped and cool down to room temperature. When the base pressure is near 10−7 torr, we on and off the ion pomp to out gas, the ion pump will continuously on until it is clean enough. During the process with the base pressure is under 10−6 torr, TSP (titanium sublimation pump) is applied to help pumping and keep UHV getting better.. 30.

(41) description UHV chamber. Clear environment. (Base pressure ∼ 1 × 10−10 torr) Scanning Tunneling Microscopy. Morphology analysis. (STM) e-beam evaporator. Fe-deposition. (home-made) Auger electron spectroscopy. Chemical element analysis. (AES) Low energy electron diffraction. Structure analysis. (LEED) Magneto-Optical Kerr effect. Magnetic property analysis. (MOKE). (in-plane & perpendicular). Ion sputtering (Ar+ 2KeV ). Sample cleaning. Table 3.1: Description of UHV chamber in C207, NTNU. 31.

(42) 3.2. Scanning Tunneling Microscopy (STM). Basic concept of STM. The basic idea of scanning tunneling microscopy (STM) is to measure the tunneling current between STM tip and the conductive sample. By applying a voltage between the tip and the sample, although the tip is indirect contact with the sample, a small electric current (0.01 nA - 50 nA) still flows from the sample to the tip or vice versa (depending on the polarity of the applied voltage). This phenomenon is called electrical tunneling, which can be explained by quantum mechanics. The exponential dependence of the tunneling current on the tip to sample distance results in a high vertical resolution. By scanning the tip across the surface and detecting the tunneling current (one can also use the current as a vertical positioning signal for the tip) a map of the surface morphology can be generated with a resolution in the order of ˚ A(depending on the wave function overlapping of atoms). It has to be mentioned that the image cannot just be interpreted as a topographic map because the tunneling current is influenced by the lateral (one can consider the shape limit of tip) and vertical variation of the electronic state density at the surface. The lateral resolution is about 1 ˚ Awhereas a vertical resolution up to 0.01 ˚ Acan be achieved. The STM can be used in ultra high vacuum, air or other environments.. 32.

(43) Theory. In classical physics, an electron cannot penetrate into or across a potential barrier if its energy E is smaller than the potential within the barrier. A quantum mechanical treatment predicts an exponential decaying solution for the electron wave function in the barrier. For a rectangular barrier we get. Ψ(d) = Ψ(0)e−k·d p 2m(Φ − E) k= ~. (3.8) (3.9). Scanning Tunneling Microscope (STM) is a potent tool for us to investigate the physics of surface at atomic level including the action of particle and surface morphology. For the STM, its resolution can reach to be 0.1 nm lateral resolution and 0.01 nm depth resolution. With the resolution, we can routinely catch the image of the atoms. The STM can be used not only in ultra high vacuum but also in air and other liquid or gas ambient, and at temperatures ranging from near 0 K to a few hundred degree Celsius. The basic concept of STM is based on the quantum tunneling effect. When we apply a bias (a voltage differential value) between the conducting tip and the surface, which the tip is very close to the surface, the electrons will tunnel through the vacuum between them. The resulting tunneling current will be the function of tip position, applied voltage (bias) and the local density of states (LDOS) of the. 33.

(44) sample. How we get the information is monitoring the current as the tip scans through the surface and it will form an image. Since STM is such a micro-probing machine, a clean surface, a sharp tip and a stable ambient are highly required. Classically, an object hitting an impenetrable barrier will not pass through. In contrast, objects with a very small mass, such as electrons, have wavelike characteristics which permit an event, referred to as tunneling. Assuming the 1-D case, the electrons behave as beams of energy and in the presence of potential U (z). Then the energy levels ψn (z) are given by the solutions to Schrödinger’s equation,. ~2 ∂ 2 ψn (z) + U (z)ψn (z) = Eψn (z) 2m ∂z 2. (3.10). where ~ is the reduced Plank’s constant, z is the position, and m is the mass of an electron. When E > U (z), the wave function is :. ψn (z) = ψn (0)e±ikz. (3.11). p 2m(E − U (z)) k= ~. (3.12). where. For E < U (z), inside the barrier, the wave function is :. ψn (z) = ψn (0)e±κz where 34. (3.13).

(45) p 2m(U − E) κ= ~. (3.14). In the case of tunneling, the wave functions of the tip and sample will overlap when we apply a bias and it will have the probability to find the electron in the barrier region and even on the other side of the barrier. The probability P is :. P ∝ |ψn (0)|2 e−2κW. (3.15). where W is the barrier width. If the bias is small, we can let U − E ≈ ϕM in the expression for κ, where ϕM , the work function, gives the minimum energy needed to bring an electron from an occupied level, the highest of which is at the Fermi level, to vacuum level. When a small bias V is applied to the system, only electronic states very near the Fermi level are excited and these excited electrons can tunnel across the barrier. In other words, the tunneling current is from the electrons near the Fermi level, and is proportional to the probability of the decaying wave function.. I∝. Ef X. |ψn (0)|2 e−2κW. (3.16). Ef −eV. One can sum the probability over energies between Ef − eV and eV to get the number of states available in this energy range per unit volume, thereby finding the local density of states (LDOS) near the 35.

(46) Figure 3.3: Schematic diagram of electron tunneling. Fermi level. The LDOS near some energy E in an interval ε is given by. E. 1X ρs (z, E) = |ψn (z)|2 . (3.17). E−. The tunneling current applied a small bias is proportional to the LDOS near the Fermi level. So. I ∝ V ρs (0, Ef )e−2κW. (3.18). where ρs (0, Ef ) is the LDOS near the Fermi level of the sample surface. Now, Fermi’s Golden Rule gives the rate for electron transfer across the barrier :. 36.

(47) w=. 2π |M |2 δ(Eψ − Eχ ) ~. (3.19). where Eψ − Eχ restricts tunneling to occur only between electron levels with the same energy. The tunnel matrix element, given by. ~ M= 2π. Z (χ ∗ z=z0. ∂χ∗ ∂ψ −ψ )dS ∂z ∂z. (3.20). is a description of the lower energy associated with the interaction of wave functions at the overlap, also called the resonance energy. Summing over all the states gives the tunneling current as. 4πe I= ~. Z. +∞. [f (Ef −eV +)−f (Ef +)]ρs (Ef −eV +)ρT (Ef +)|M |2 d. −∞. (3.21) where f is the Fermi function, ρs and ρT are the density of states in the sample and tip, respectively.. Operation mode. As I know, the operation of STM can divide into two modes roughly. 1. Constant current mode(Fig3.5): By using a feedback loop, the tip is vertically adjusted in such a way that always stays constant. The software will record the vertical position of the tip during scanning. Because of the current is proportional to the 37.

(48) Figure 3.4: The loop of a working STM(Text and graphics by Michael Schmid, TU Wien). local density of states, it can transfer to the image of surface morphology. 2. Constant height mode(Fig3.6): By contrast with the constant current mode, this mode will fix the vertical height of the tip and record the variation of tunneling current to get the surface morphology.. 38.

(49) Figure 3.5: STM constant current mode. 3.3. Figure 3.6: STM constant height mode. AES and LEED. An atom which has been ionized in a core level may come back to its ground state by the following way. One is that an electron in higher state jumps into the core level and simultaneously releases the energy by x-ray radiation (Fig. 3.7(a)). The other one is that an electron in higher state jumps into the core level and simultaneously the energy is transmitted to another electron, which may leaves the atom with a characterizing kinetic energy (Fig. 3.7(b)). The second process is called Auger process and the emitted electron is Auger electron. Therefore, the Auger electron has a characterizing kinetic energy denoting the kind of atom where it comes from. By detecting the Auger electron, the composition of the sample can be characterized. In general, the kinetic energy of the Auger electrons are less than 1000 eV, so their mean free path are almost less than 20 ˚ A. That is why the Auger electron spectroscopy (AES) which can only detect the region within the mean free path of Auger electron is used as a standard tool 39.

(50) in surface science.. Figure 3.7: Process for demagnetization of atomic core holes. (a) Emission of X-ray radiation. (b) Emission of an AES electron.. Because many properties of ultrathin film, such as mechanical, electronic, and magnetic properties are strongly correlated to the arrangement of atoms, i.e. the geometric structure of the surface, the exact knowledge of this structure is an important information for the quantitative understanding. Well-defined and periodic structures like the surfaces of single crystals are most suitable for structural investigations. In most cases the bulk structure is known through X-ray diffraction. However, the surface structure within only a few atomic layers, usually is not identical to that of the bulk. From the de Broglie relation λ = h/p, with electron energy ranging from 25 eV to 600 eV, the wavelength varies from 2.5 ˚ A to 0.5 ˚ A which is comparable to the lattice constant. Therefore the most ideal and reliable tool we can utilize to investigate those structures is the low energy electron diffraction (LEED). By the analysis of LEED patterns, the intensities, 40.

(51) the size and the shape of the surface unit cell, the degree of order and the atomic structure can be determined with high precision. Furthermore, LEED measurement with a CCD camera allows fast and reliable acquisition of data.. Figure 3.8: Schematic display of a LEED structure. 41.

(52) 3.4. Magneto-Optical Kerr effect(MOKE). If a linear polarized light is incident into a ferromagnetic sample, since of the different reflection coefficients of right and left circular polarization components, the reflection beam will become elliptical polarized. This phenomenon is so called magneto-optical Kerr effect. The angle between the primary axis of the elliptical polarization and the linear polarization is called Kerr rotation, and the ellipticity of the elliptical polarization is called Kerr elliptical, as shown in Fig. 3.9.. Figure 3.9: Schematic illustration of magneto optical Kerr effect. After reflected from the ferromagnetic sample, the linear polarized laser beam becomes elliptical polarized.. Let r+ eiθ+ and r− eiθ− stand for the reflection coefficients of right and left circular polarization, respectively. The Kerr rotation and − and εk = Kerr ellipticity can be illustrated as ϕk = − θ+ −θ 2. r+ −r− r+ +r− ,. b a. =. respectively. Both of them are proven to be proportional to the. magnetization of sample. Thus by measuring ϕk and εk with cyclic 42.

(53) applied magnetic field, we can get the hysteresis loop. In general, there are three types of MOKE measurement. Each of them has different geometry of the magnetization and the light path, as shown in Fig. 3.10. In the polar Kerr effect, the magnetization lies in the plane of incidence and is perpendicular to the surface. In the longitudinal Kerr effect, the magnetization lies in the plane of incidence and is parallel to the surface. In the transverse geometry, the magnetization is perpendicular to the plane of incidence and on the surface.. Figure 3.10: Different geometry for MOKE measurement. In magnetic ultrathin films, the Kerr signal is so small that the noise may result in significant effect. Therefore, in our experiment, a modulator is added between the polarizer and the sample such that the modulated signal can be taken by lock-in technique with a larger ratio of signal to noise. The schematic illustration is shown in Fig. 3.11. Due to the difference with the DC MOKE shown in Fig. 3.9, this method is called AC MOKE. Fig. 3.12 is a MOKE chamber in our system. Part A is a combination of ”tilt” and ”rotator”. It can adjust the direction and angle of the sample to let the angle between incident laser and sample be. 43.

(54) Figure 3.11: Schematic illustration of AC MOKE. 450 . Part B has four electromagnets. We can control the direction of magnetism by changing the direction of current passed through the electromagnets. As a result of the above two equipments, we can measure the magnetism in longitudinal and polar direction at the same time without moving the sample. In our system, we can measure the magnetism at low temperature. It’a s simple concept that let the cold trap be full of liquid nitrogen. The bottom of cold trap sticks to sample holder tightly by using a specific gasket which is done with copper. Because of the heat transmitted by the contact between sample, sample holder and clod trap, the gap of screw reduces the efficiency of thermal conduction. So we add some cooper wire to help the sample cooling down to 100 K in ten minutes, just like Fig. 3.14. And the process of measurement is shown in Fig. 3.15.. 44.

(55) characteristics Magnetic field. We have four electromagnets to form the magnetic field which is varied as the direction of passing current being changed. As a result, we can measure the magnetism in in-plane and perpendicular directions at the same time without moving the sample.. Temperature. We can do the MOKE measurement at room temperature(RT) & low temperature(LT:100K). Modulator. In our system, it’s used in a different way with the general AC MOKE. A modulator is added on the laser source and modulated signal can be taken by lock-in technique with a larger ratio of signal to noise.. Parameter of SR810 DSP Lock-in. frequency : 10KHz. Amplifier. Amplitude : 0.4V Phase : −88.020 Table 3.2: Description of MOKE chamber in C207, NTNU. 45.

(56) Figure 3.12: MOKE chamber in NTNU C207 Lab. 46.

(57) Figure 3.14: Cu wire for cooling sample Figure 3.13: Specific gasket for LT-MOKE measurement. Figure 3.15: Schematic display of a DC-MOKE loop in C207, NNU. 47.

(58) Chapter 4. Experiment and Results. The whole experimental process, including sample preparation, transferring, and characterization were performed in an ultra high vacuum (UHV) system with a base pressure better than 3 × 10−10 torr. The Si(111) 7x7 template was prepared by repeated flashing at 12000 C [3,40,41]. For LT and RT-growth, the Fe atoms were deposited while the substrate was at 100 K and 300 K, respectively. The film thickness was calibrated by using Auger electron spectroscopy (AES) and scanning tunneling microscope (STM). Both the detections of AES and STM were performed while the sample was at room temperature. The magnetic behavior was investigated by using magneto optical Kerr effect (MOKE) in both longitudinal and polar directions with the lock-in technique. The MOKE measurements shown in this report were performed at RT.. Figure 4.1: The whole experiment process. 48.

(59) The whole process is shown in Fig. 4.1. Because of the different resistance of all Si slabs we need to adjust the current pass through Si carefully to avoid the sample overheated. We can know the temperature of sample by the thermal couple or the power we add on it. We have to deal properly with the Si slab by heating at 6000 C for more than 6 hours to completely get rid of the oxidation. Then we measure the AES to check if the sample is clean enough.. Figure 4.2: 12 ML RT-growth Fe/Si(111). The signal of carbon is much smaller than iron and silicon after 12 ML Fe deposition.. Figure 4.3: The process of preparation Si(111)7x7. The last process is the key to obtain large Si(111)7x7 terrace. We 49.

(60) need to reduce the current very slowly, or said, cooling down slowly, to get the large terrace. By STM image, we can confirm if the surface is Si(111) 7x7 and the terrace is large enough. In Fig. 4.4, we can see a terrace with the width about 30nm and its reconstruction of Si(111)7x7.. Figure 4.4: Si(111) 7x7 image obtained in our laboratory. Up to now, we have obtain the substrate for following experiment. We deposit the Fe on Si(111) at RT and measure its AES data as the control group. In our system, we use LT-grown films as the buffer layer and then deposit Fe at RT several times. We measure AES, MOKE and STM data after Fe deposited on Si substrate every time.. 50.

(61) 4.1. LT(100K)-growth of Fe/Si(111). In order to investigate the different interface condition of LT and RT-grown Fe on Si(111), the coverage-dependent AES detection was performed. Fig. 4.5 shows the AES signal ratio of F e47eV /Si92eV (left axis) as a function of Fe thickness for RT-Fe films on Si(111) and on 5 ML LT-Fe/Si(111). The AES measurements were carried out at RT. This means for LT-grown films, the sample underwent RT annealing before AES detection. In RT-grown Fe/Si(111) films, the Fe signal becomes observable after 3-4 ML, and then reaches the comparative signal intensity of Si at ∼9 ML. After 10 ML, the Fe signal dominates Si. The question of when the Si really disappears from the surface can be answered by seeing from the right axis of Fig. 4.5, the AES ratio of Si92eV /F e47eV . The Si/Fe ratio gradually approaches 0.06 ± 0.03 at 15 ML. This is the minimum ratio of Si/Fe we could observe. The small and broad feature of Fe at 87 eV partially overlaps with the possible small Si-92 eV peak. The overlapping actually prevents us from checking the presence of Si more precisely. Thus, we can conclude, at least to the limit of our AES measurement, the Si(111) substrate is fully covered after deposition of 15 ML RT-grown Fe. For comparison, the solid squares in Fig. 4.5 indicate the Fe/Si AES ratio of RT-Fe/5 ML LT-Fe/Si(111). Apparently, the Fe/Si ratio of 5 ML LT-Fe/Si(111) is much larger than 5 ML RT-Fe/Si(111). Sequential deposition on 5 ML LT-Fe/Si(111) gradually increases the Fe/Si ratio in the similar trend of RT-Fe films. The horizontal shift. 51.

(62) Figure 4.5: Auger ratios of Fe47eV /Si92eV (left axis) and Si92eV /Fe47eV (right axis) measured as a function of Fe thickness. The thickness means the total thickness of Fe deposition on Si(111). The solid circles and solid squares indicate the Fe/Si AES ratio of RT-grown Fe films and RT-Fe/5 ML LT-Fe films on Si(111), respectively. The open circles and open squares indicate the Si/Fe AES ratio of RT-Fe and RTFe/5 ML LT-Fe films on Si(111), respectively. The solid and dotted curves are the fitting results by Eq. (4.1) and (4.2).. between the two series of data is 3.5 ± 0.5 ML. A recent study of Kataoka et al. [23] showed, for RT-grown Fe/Si(111), metastable FeSi with the CsCl-type structure persists at least to 6 ML of Fe. Afterward, the bcc-Fe(111) film starts to grow. Our AES data supports their conclusion of serious silicide formation in RT-growth. Based on Kataoka et al.’s observation [23], the equation of Fe/Si AES ratio is written down by assuming the first 6 ML of Fe was mixed with Si, forming 12 ML of FeSi. 52.

(63) Figure 4.6: Schematic illustration of AES electron emission.. Fig. 4.6 is the schematic illustration of AES electron emission. The quantitative AES analysis was depicted by assuming that the possibility of AES electron to emit out of the sample without losing its characterizing kinetic energy decays as e−1 of the original one after penetrating the depth of a mean free path (λ). Because of the FeSi formation, the intensity of Auger signal is contributed to the Fe films and half of the FeSi films.. −(t−6). −(t−6). −(t+6). Fe IF e (1 − e λF e ) + 0.5 × (e λF e − e λF e ) ( )RT = × −(t+6) −(t−6) −(t+6) Si ISi (e λSi ) + 0.5 × (e λSi − e λSi ). (4.1). t is the deposited Fe thickness. λF e and λSi are the mean free path of Fe47eV and Si92eV Auger electrons.. IF e ISi. is the relative sensitivity. between Fe and Si Auger intensity. Note, Eq. (4.1) is for RT-grown 53.

(64) Fe/Si(111) with t ≥ 6 ML. Besides, for RT-Fe/5 ML LT-Fe/Si(111), the Fe/Si AES ratio is written down by assuming no silicide formation at the LT-Fe/Si interface. −t. Fe IF e (1 − e λF e ) ( )LT = × −t Si ISi (e λSi ). (4.2). First, the RT-Fe/Si(111) data was fitted by Eq.(4.1), using the relative sensitivity. IF e ISi =0.66±0.06,. which is consistent with the value. in the Auger handbook. [42] The fitting curve matches the experimental RT data well, as shown in Fig. 4.5 with λF e = 3.6 ± 0.3 ML, and λSi = 4.0±0.3 ML. With these parameters, the curve of Eq. (4.2) also matches the experimental data of RT-Fe/5 ML LT-Fe/Si(111). Actually, within the error bars of AES data, we cannot totally exclude the possible presence of 1-2 ML Fe-silicide formation at the interface. However, when considering the detailed chemical composition and AES signal of the Fe-silicide interface, the suppression of silicide formation by LT-Fe is even larger than the shift of 3.5±0.5 ML in Fig. 4.5. Our experimental and fitting results all indicate the LT-Fe intermediate layer indeed effectively reduces the Fe-silicide thickness by at least 4 ML Fe (∼8 ML FeSi) compared with RT-growth. Even after RT-annealing, the LT-grown intermediate layer can sustain the Fe/Si interface and prohibit further silicide formation. Since LT deposition of Fe on Si(111) can effectively reduce the alloy formation at the interface even after RT annealing, we were curious about the thermal stability of the LT-grown films. Fig. 4.8 54.

(65) shows the AES ratio of Si/Fe as a function of annealing temperature for 5-25 ML LT-Fe films. All the AES detections were performed after keeping the sample at the annealing temperature for 10 mins by passing through the current directly and then cooling it to RT naturally. Fig. refc shows the relation of current and temperature. For 5 and 15 ML LT-Fe, the AES ratio remains invariant during 300-350 K and then we observe the increased Si signal at ∼400 K. For 25 ML LT-film, the Si/Fe ratio remains stable up to a higher temperature of 400 K. For a thicker film, like 25 ML LT-Fe, it is hard to judge if the higher stability is truly because of the higher thickness, or only because the coverage is too thick to detect the slight changes at the Fe/Si interface. Nevertheless, we can still conclude 5-15 ML LT-grown Fe films annealing to RT are stable at least up to 350 K. It means that the composition does not vary with temperature before 350 K. It also supports our suggestion of using LT-grown layer as an intermediate layer for subsequent RT-growth or processing. The previous studies of Gallego et al. and Tsushima et al. concluded the Fe(111) films grows epitaxially up to a thick region, maintaining the bcc structure. [1,16,23,41] Besides the crystalline structure, the study of surface morphology is still lacking, especially for LTFe/Si(111) films. Usually the LT-deposition of thin film is expected to result in statistically flat morphology, due to the low mobility and random landing distribution of deposited atoms. Recent studies of quantum well systems also report the 2-step growth, meaning LTdeposition with post-annealing may lead to atomically flat surface. 55.

(66) Figure 4.8:. Auger ratios of Si/Fe mea-. Figure 4.7: The relation of the current. sured as a function of annealing temper-. passing through the sample and the tem-. ature. All the LT-grown Fe films are an-. perature.. nealing to RT at first and then annealing to different temperature.. The cir-. cles, squares, and triangles indicate the series data of 5, 15 and 25 ML LT-grown Fe/Si(111) films.. The solid lines are. guides to the eye.. morphology in some cases of the thin film system. Thus, we try to probe the growth result of LT-deposited Fe films on the Si(111)7x7 surface. For example, the STM images of 24 ML and 40 ML LTFe/Si(111) are shown in Fig. 4.9 with the line profiles. For 24 ML LT-Fe, the surface was composed of elongated and winding islands. One can see it more clearly from the inset. The width of the islands is 4 ± 1 nm and the length ranges from 5 to 15 nm. The surface corrugation observed from the line profile is only within 0.4 ± 0.2 nm. For the higher coverage of 40 ML LT-Fe/Si(111), the morphology 56.

(67) changes to be composed of larger islands, which are no longer elongated, with the diameter = 8 ± 2 nm. The surface corrugation also increases to 0.6±0.2 nm. Actually, our STM investigation of 24-40 ML 100K-deposited Fe/Si(111) reveals similar results as 2.5 and 20.4 ML 150K-deposited Fe/Au/Si(001), which were previously reported by F. Zavaliche et al.. [21] In their study of 2.5 ML Fe, the surface was also composed of small coalesced islands of 0.4 nm height and ∼ 6−8 nm in size. Increasing the thickness to 20.4 ML Fe only enhances the roughness to ∼0.6 nm. In contrast to the LT-growth, previous RT-growth results of Fe/Si(111) and (001) all reveal a relatively rough surface. For example, the morphology of 20.4 ML RT-Fe/Si(001) consists of elongated islands of 1.0-1.5 nm height. [20]. 4.2. MOKE measurement. In the above sections, compared with RT-grown films, LT-growth can effectively reduce the Fe-silicide formation at the interface, and lead to a flat surface morphology. Thus, for the magnetic study, we used a 5 ML LT-grown Fe film as an intermediate layer for the sequential RT-deposition, since RT-deposition is easier to perform, especially in future applications. The MOKE hysteresis loops for n ML RT-Fe/5 ML LT-Fe/Si(111) measured at RT are shown in Fig. 4.11. The hysteresis loops are close to the square shape. Both the Kerr saturation signal (Ms ) and coercivity field (Hc ) increase with Fe coverage, but in different ways. The quantitative analysis is shown in Fig. 4.12. Ms. 57.

(68) Figure 4.9: Surface morphology of (a) 24 ML and (b) 40 ML LT-grown Fe films, investigated by scanning tunneling microscopy (STM). The figures are all of 120x120 nm2 . The insets (30x30nm2 ) reveal the detailed surface structures. The line profiles indicated in the STM images are plotted at the bottom.. 58.

(69) increases linearly with Fe coverage. The extrapolation of the linear fitting line indicates that the magnetic dead layer could be 3.5 ± 0.5 ML in our RT-MOKE measurement. Thus, we expect, at low temperature, the magnetic dead layer might be even smaller than 3 ML, which can be due to a slight Fe-silicide formation or the extensive hybridization of Si electronic states into Fe.. Figure 4.10: Polar MOKE hysteresis loops measured at room temperature (RT) for n ML of RT-grown Fe on a intermediate. Figure 4.11: Longitudinal MOKE hysteresis loops measured at room temperature (RT) for n ML of RT-grown Fe on a in-. layer of 5 ML LT-grown Fe on Si(111).. termediate layer of 5 ML LT-grown Fe on Si(111).. In Fig. 4.12, Hc increases monotonically and gradually becomes saturate around 70 Oe. To deduce more detailed insights from the coverage dependent evolution of Hc , we use the Stoner-Wohlfarth single domain model to simulate the Hc evolution behavior with Fe film thickness. The energy (E) correlated to magnetism in our thin film system can be written as followings. [43,44] E = −H×M ×cos(θ − φ) + K ef f ×sin2 θ 59. (4.3).

(70) Figure 4.12: Kerr saturation signal (Ms : left-axis) and coercivity field (Hc : right axis) plotted as function of Fe coverage. The data points were analyzed from the hysteresis loops of n ML RT-Fe/5 ML LT-Fe/Si(111) in Fig. 4.11. The solid line and curve are the fitting results of Ms and Hc using a linear function and Eq.4.6, respectively.. θ and φ are the rotation angles of magnetic moment and magnetic field, respectively, relative to the surface normal direction. The first term is Zeeman energy. M is the averaged magnetic moment per unit volume and H is the applied magnetic field. In the longitudinal MOKE measurement, as shown in Fig. 4.11, we apply the magnetic field in the in-plane directions, and thus φ only can be ±π/2. Kef f in the second term is the uniaxial magnetic anisotropy. Since the easy axis is in the surface plane, one should expect Kef f < 0, which we can check later from the fitting results. By minimizing the magnetic energy in Eq.4.3, the magnetization direction θ can be determined as a function of the applied field H. Since we have φ = ±π/2, the. 60.

(71) minima of E(θ) are always at ±π/2. This can be easily checked by solving the equation of dE(θ)/dθ=0 with φ = ±π/2. Therefore, in this model, the magnetization only switches between θ = ±π/2, resulting in the square magnetic hysteresis loops, which are actually consistent with our measured loops. Further, from Eq.4.3 we can deduce the coercivity field Hc at which the magnetization direction switches. For example, if the magnetic moment is at θ = π/2 initially, we apply the field in the φ = −π/2 direction to switch the magnetic moment. The coercivity field can be solved from the following equation. d2 E(θ, φ = −π/2) |θ=π/2 = 0 dθ2. (4.4). Hc = −2×Kef f /M. (4.5). In a thin film system, we have both the surface and the volumecontributed magnetic anisotropy. The magnetic anisotropy Kef f can be decomposed as Kef f =kv +ks /t, [44] where kv and ks are volume and surface anisotropy terms, respectively. Since Kef f is the anisotropy energy per volume, the surface contribution ks will gradually decay with 1/t, where t is the thickness of thin film. Then we rewrite the Hc as a function of film thickness t. Hc (t) = −. 2 ks ×(kv + ) M t. (4.6). We take the magnetic moment of Fe: M = 2.5 µB /atom from the literature, [45] and then set kv and ks as the free parameters. The solid curve in Fig.. 4.12 is the best fitting by using Eq.4.6 with. kv = −0.5 ± 0.1µeV /atom and ks = 1.9 ± 0.1µeV /atom. The fitting curve qualitatively reproduces the monotonic increase and saturation 61.

(72) of Hc with the Fe film thickness. With these fitted kv and ks values, Kef f =kv +ks /t is always negative after 5 ML. This is consistent with the in-plane magnetization of the measured Fe films. Besides, Kef f changes signs from positive to negative at the critical thickness of tc = −ks /kv = 4 ± 1M L, indicating the spin reorientation transition (SRT) from surface normal to in-plane. The SRT and tc is actually consistent with the previous study of Nazir et al.. [3] They reported the perpendicular to in-plane SRT at 3.6-5.5 ML for 100Kgrown Fe/Si(111) films. Compared with other SRT systems, our fitted Kv and Ks are relatively small, but actually close to the crystalline anisotropy of bodycentered-cubic (BCC) Fe. [43,44] One possible reason could be the domain wall motion, which is not considered in the single domain model. Another possibility is the anisotropy model could be interpreted in another way. During the magnetization switching, instead of going through the surface normal direction, the magnetic moment may undergo 180◦ rotation just on the surface plane. In this case, we can redefine the θ and φ relative to another direction, which is perpendicular to the applied magnetic field, but on the surface plane. Then we can still keep the same calculation and fitting. The only difference is the magnetic anisotropy is no more the energy difference between surface normal and in-plane directions, but between the different directions on the surface plane. Even after modifying the anisotropy from uniaxial to 6-fold symmetry, [17] we will still get similar results of kv and ks . We have to agree, from our experiment data, it is hard 62.

(73) to judge which is the real situation in our system. Further work will be needed for a more advanced conclusion. However, we can at least conclude the volume and surface contributions, kv and ks , are of opposite signs and compete with each other while the film thickness is increased. Such kind of situation eventually results in an Hc increase and saturation in our measured n ML RT-Fe/5 ML LT-Fe/Si(111) system.. 63.

(74) Chapter 5. Conclusion. LT-deposition of Fe/Si(111) films was performed in this experiment.. 1. LT-deposition can effectively reduce the Fe-silicide formation at the Fe/Si interface compared with conventional RT growth. 2. The LT-grown Fe films reveal relatively flat surface morphology and its interface condition remains stable around room temperature. 3. The roughness of LT-grown Fe films is within 2-3 atomic monolayers even for 24-40 ML LT-Fe films. 4. A single domain model of magnetic anisotropy is proposed to fit the magnetic coercivity evolution with film thickness. We deduce the values of surface and volume-contributed magnetic anisotropy, which are inverse signs and their competition is the main origin for the variation in coercivity.. 64.

(75) Bibliography [1] J.M. Gallego, J. M. García, J. Alvarez, and R. Miranda, Phys. Rev. B 46, 13339 (1992). [2] J. S. Tsay and Y. D. Yao, Appl. Phys. Lett. 74, 1311 (1999). [3] Z.H. Nazir, C.-K.Lo, M. Hardiman, J. Magn. Magn. Mater. 156, 435 (1996). [4] B. Ghebouli, S.-M. Ch´ erif, A. Layadi, B. Helifa, M. Boudissa, J. Magn. Magn. Mater. 312, 194 (2007). [5] J. S. Tsay, H. Y. Nieh, C. S. Yang, Y. D. Yao, T. S. Chin, J. Appl. Phys. 93, 8728 (2003). [6] J. S. Tsay, T. Y. Fu, M. H. Lin, C. S. Yang, and Y. D. Yao, Appl. Phys. Lett. 88, 102506 (2006). [7] J. Alvarez, A. L. V´ azquez de Parga, J. J. Hinarejos, J. de la Figuera, E. G. Michel, C. Ocal. and R. Miranda, Phys. Rev. B 47, 16048 (1993). [8] A.Pimpinelli,and. J.. Villain. in. Physics. of. crystal. growth,(Cambridge University press, United Kingdom, 1998),p. 144. [9] Brommer, J. Appl. Phys. 89, 7139 (2001). [10] Stich, Phys. Rev. B 55, 5886 (1997).. 65.

(76) [11] S. Banerjee, G. Raghavan, M. K. Sanyal, J. Appl. Phys. 85, 7135 (1999). [12] A. Mascaraque, J. Avila, C. Teodorescu, M. C. Asensio, E. G. Michel, Phys. Rev. B 55, R7315 (1997). [13] R. Kl¨ asges, C. Carbone, W. Eberhardt, C. Pampuch, O. Rader, T. Kachel, W. Gudat, Phys. Rev. B 56, 10801 (1997). [14] D. Sander, A. Enders, and J. Kirschner, Appl. Phys. Lett. 67, 1833 (1995). [15] A. Wawro, S. Suto, R. Czajka, and A. Kasuya, Phys. Rev. B 67, 195401 (2003). [16] Le Thanh Vinh, J. Chevrier, and J. Derrien, Phys. Rev. B 46, 15946 (1992). [17] G. Garreau, S. Hajjar, J. L. Bubendorff, and C. Pirri, D. Berling, A. Mehdaoui, R. Stephan, P. Wetzel, S. Zabrocki, G. Gewinner, S. Boukari, E. Beaurepaire, Phys. Rev. B 71, 094430 (2005). [18] D. L. Goroshko, N. G. Galkin, D. V. Fomin, A. S. Gouralnik, and S. V. Vavanova, J. Phys.: Condens. Matter 21, 435801 (2009). [19] D. L. Goroshko, N. G. Galkin, and A. S. Gouralnik, J. Appl. Phys. 107, 063709 (2010). [20] F. Zavaliche, W. Wulfhekel, M. Przybylski, S. Bodea, J. Grabowski and J. Kirschner, J. Phys. D: Appl. Phys. 36, 779 (2003).. 66.