Sm1-xGdxAl2, a zero-magnetization ferromagnet
The first strategy to generate a magnetic compensation in the SmAl2 compound was to try to modify the contribution of conduction electrons in changing the s-f exchange or the density of states at the Fermi level. Those changes could theoretically be achieved via modification of the lattice constant. The substitution of Sm atoms in SmAl2 by non-magnetic Sc, Y or La (effect of chemical pressure) resulted in the reduction of ordering temperature but no compensated state could be observed.
The second attempt was to modify the 4f contribution, i.e. to increase the 4f spin contribution. Given the magnetic coupling between lanthanides elements (ferromagnetic coupling between spin moments), replacing a small quantity of Sm atoms by other Lanthanide atoms would yield a total moment that is the average of the 4f spin and orbital moments of each RE element. The strategy developed by Adachi et al. was to substitute Gd atoms to Sm ones, to synthetize the Sm1-xGdxAl2 (SGA) compound. This choice has been strongly motivated by the S character of Gd ions that will only contribute to the spin moments via both 4f and conduction electrons contributions. The goal of this section is to give an overview of the magnetic properties investigated in bulk SGA and more especially of the various analysis devoted to the original magnetic compensated state, coexisting with long-range parallel (i.e. ferromagnetic) spin and orbital orders.
First evidence of the zero-magnetization state
The magnetic compensation via the tuning of the 4f contribution in bulk SmAl2 has been demonstrated by H. Adachi et al. and achieved by substituting a small quantity of Gd atoms by Samarium atoms. Given the Gd3+ electronic configuration ( [ ]4 75 16 2), the Hund’s rules lead to = 0 in order to maximize and its total angular momentum is thus a pure spin moment. Previous results of hyperfine studies on SGA alloys have reported the parallel coupling between the Gd and Sm spin contributions. Gd is thus expected to contribute positively to the spin part of the net moment providing two new spin contribution 4 and . The net moment defined by equation (I-2) can be rewritten as: = 4 − 4 − − 4 − (II-1).
Figure 5 presents the temperature dependence of the magnetization in bulk SmAl2 and in two SGA compounds corresponding to different Gd concentrations. While a monotonously decreasing M(T) curve is measured for SmAl2 (decrease of L and S with L > S in the whole temperature range), the substitution of Gd atoms in the SmAl2 matrix yields a magnetic compensation, occurring at 81K and 64K for respectively 1.8% and 2.6% Gd contents. The authors suggest that the ferromagnetic spin order should be kept in the compensated state. No definitive proof is however provided by these macroscopic measurements.
Ferromagnetic order at Tcomp
Following the observation of a zero-magnetization state in bulk SGA by Adachi et al., several groups have tried to clearly identify the underlying mechanism and confirm the co-existing ferromagnetic spin order. A variety of techniques has been used, which enabled to explore the system from different angles and to separately probe the different contributions to the magnetic moment.
The Magnetic Compton scattering, based on the interaction between the spin moment ( + 4 ) and the helicity of light, has brought the first proof of the spin ferromagnetic order via the observation of a finite signal at Tcomp. By changing the thermal history (measurements at Tcomp coming from high or low temperature side), the authors report on opposite spin polarizations, implying a positive (negative) spin contribution to the total moment above (below) Tcomp.
Non-resonant X-ray Magnetic Scattering, based on the interaction between the photon and the spin and orbital 4f moments, has been used to investigate separately the thermal dependences of the 4f spin and 4f orbital moments under an applied field of 1T. These measurements have especially confirmed the dominant character of orbital contribution below compensation and the dominant character of spin contributions above compensation. In the same study, measurements of specific heat show no anomaly at the compensation, implying no magnetic transition and suggesting the persistence of the ferromagnetic order.
Six years after their first report on the SGA magnetic compensation, Adachi et al. unambiguously prove the spin and orbital long-range orders at compensation. Using X-Ray Magnetic Circular Dichroism (XMCD), based on the difference in absorption by a magnetic material when it is excited by a left or a right circularly polarized light, they could separately probe the Gd and Sm 4f contributions (M4,5 edge). Taking advantage of the sum rules, they could extract the spin and orbital contributions, as it is reported in figure 6 (a). The net resulting magnetizations for the Sm, Gd and conduction electrons are compared to the net magnetization of SmAl2 in figure 6 (b).
These results prove the monotonic decrease of the Sm 4f spin moment and its finite value at compensation, in agreement with a remaining long-range ferromagnetic order. They confirm the anti-parallel coupling between Sm orbital and spin contributions and the dominant L contribution over the entire temperature range, as expected from the calculation presented in figure 3. Consequently, the compensated state cannot be achieved by only the Sm 4f electrons; the Gd spin moment and the conduction electron polarization (ferromagnetically coupled to the Sm 4f spin contribution) must be taken into account to cancel the surplus of orbital moment of the Sm3+ ions. Figure 6: (a) Experimental values of (⃝), (+), (Δ) and (x) when the sample was magnetized at 110K (open symbols) and at 43K (filled symbols). The dashed and dotted lines correspond to the theoretical value of the 4f spin and orbital moments of Gd and Sm (respectively). (b) Net magnetization (+) for the Sm (⃝), Gd (□) and conduction electrons (x) compared to the net magnetization of SmAl2 (Δ) (from).
The different studies mentioned in this section have contributed to the definition of bulk SGA compound as a zero-magnetization ferromagnet. These unusual properties make it very attractive for specific fundamental studies as well as for various types of applications.
Because of its zero-magnetization, such a compound doesn’t indeed generate any dipolar field and is insensitive to an external applied field: these characteristics should permit to achieve a uniform magnetic state, without formation of magnetic domains, that doesn’t induce magnetic perturbation for neighboring elements and is magnetically very stable. However, as a ferromagnet, it must be able to spin-polarize or spin-analyze an electrical current.
The combination of these usually antagonist properties would be particularly interesting for applications dealing with the control and/or manipulation of the electron spin, such as magnetic tips for spin-polarized tunnel microscopy or as a component in some spintronic devices. These potential developments have strongly motivated the studies undertaken by our group in Institute Jean Lamour to achieve the growth of SGA epitaxial films, to investigate their structural, magnetic and electronic properties and to further integrate them in more complex hybrid systems. Indeed, these properties should be conserved in thin films used in spintronics devices as spin-valves or magnetic tunnel junctions.
Sm1-xGdxAl2 as epitaxial films
The first paragraph briefly summarizes the growth conditions and structural characteristics of SGA epitaxial films. The second paragraph presents their magnetic properties, investigated by the combination of SQUID magnetometry and XMCD experiments.
Epitaxial growth and structural properties
Our group in IJL has been the first to achieve the growth of single crystalline epitaxial SGA films by Molecular Beam Epitaxy (MBE),. The experimental process is based on the co- ̅deposition at 510°C of Sm, Gd and Al atoms on a Nb buffer (50 nm thick) covering a (1120) sapphire substrate. The different fluxes are previously calibrated by a quartz microbalance at the substrate position in order to get the desired stoichiometry.
Figure 7 presents typical (a) Reflection High Energy Electron Diffraction (RHEED) patterns along two azimuthal directions collected during the SGA growth and (b) large angle X-ray diffraction patterns (specular analysis along the growth direction and rocking curve across the (111)SGA Bragg peak) measured with a PanAnalytical X’Pert Pro diffractometer working at the Cu wavelength.
Perpendicular anisotropy and giant coercivity
The hysteresis loops measured at 95K and 85K (i.e. above the compensation temperature Tcomp=64K) for an external field applied perpendicular to the sample surface (figure 9 (a)) are characteristics of a perpendicular easy magnetization axis, as previously reported by Avisou et al.,. It has been suggested that the perpendicular anisotropy in Sm based compound would result from magneto-elastic effects related to the compressive strain observed along the growth direction.
The XMCD signal recorded below Tcomp is almost constant in the ±17T field range. The coercive field can be properly determined only down to 85K (i.e. approximately 20K above Tcomp). Its temperature dependence is given in figure 9 (b). The strong divergence when approaching the compensation point is expected from the vanishing magnetization: the magnetization becoming small, the Zeeman energy becomes also small and as a result the magnetic field required to reverse the magnetization increases drastically. It is a common feature in usual ferrimagnetic compounds,. However, in ferrimagnetic compounds, the coercive field decreases when further decreasing the temperature away from the compensation. This is obviously not the case in SGA epitaxial films below Tcomp despite the increase of magnetization at low temperature. The giant coercivity is thus not related to the small magnetization value but to other features becoming dominant at low temperatures. The persistence of a huge coercivity below the compensation temperature in SGA epitaxial films also differs from the behavior reported in bulk SGA (for example 0.2T at 10K from). The reasons for such a large coercive field and the specific role of structural characteristics (antiphase boundaries, microstructure) in SGA epitaxial films still remain to be clarified.
5d spin polarization and long-range ferromagnetic order
The negative XMCD signal measured at the Sm L2 edge for the maximum +17T field (figure 9(a)) reveals a positive projection of the 5d spin polarization, i.e. the spin contribution point towards the positive field direction. This configuration (L-down and S-up), sketched by the respective purple and green arrows in figure 9 (c), is favored by the Zeeman interaction above Tcomp where S contributions are dominant. Above Tcomp, the magnetization reversal in SGA epitaxial film is driven by the flipping of the dominant S contribution towards the negative field direction (L-up configuration), thus yielding a positive signal at the Sm L2 edge as observed in figure 9 (a). Below Tcomp, the signal measured in the SGA epitaxial film remains negative over the entire ±17T field range. Despite the dominant L contributions in this temperature range, the smaller S component still point towards the positive field direction: the S-up configuration is stabilized during the cooling process from above Tcomp and persists down to low temperature due to the diverging coercive field, as previously discussed.
The temperature dependence of the XMCD signal recorded at 17T is reported in figure 9 (c), in comparison with SQUID measurements (black squares). It reflects the variation of the 5d spin polarization that monotonically decreases when increasing the temperature. This variation is in good agreement with the one for 4f Sm and Gd spin contributions measured in bulk SGA at the M4,5 edge and confirms the close correlation between the 4f and 5d spin contributions. Two main points have to be noted in figure 9 (c): (i) the Curie temperature can be extrapolated to 126K, in good agreement with the Curie temperature obtained from SQUID measurements and with the bulk SmAl2 Curie temperature, (ii) the 5d spin polarization exhibits a finite value at Tcomp, confirming the persisting long-range ferromagnetic order in the compensated state, as reported previously by Avisou et al..
In order to analyze the possible existence of domains after the cooling process towards below compensation, a specific cooling procedure has been used to specifically prevent the formation of L-up domains: the cooling field has been reversed from +17T to -17T at Tcomp thus “following” the dominant magnetic contribution. The resulting 5d polarization at 20K (star symbol in figure 9 (c)) is very close to the one measured after the +17T usual cooling, which suggests a homogeneous S-up magnetic configuration after the +17T field cooling. A high external field up to 17T does not permit inducing any “Zeeman-favored” L-up domains.
Sm1-xGdxAl2 epitaxial films in magnetic heterostructures
Beyond the epitaxial growth of (111)SGA epitaxial films, the group in Nancy undertook a few years ago the investigation of magnetic SGA-based heterostructures:
(i) Exchange-coupled bilayers, where the highly coercive SGA compound acts as a magnetically hard pinning layer.The investigation of exchange coupling in SGA-based bilayers is first a powerful way to probe the interface magnetic properties of SGA. At the SGA compensation point, SGA/SA bilayers can also be interestingly compared to conventional AFM/FM exchange-bias systems, where the zero-magnetization AFM is responsible for the biased reversal of the FM layer. Despite its zero-magnetization state, as it is the case in AFM, the remaining ferromagnetic order in SGA has however the strong advantage to enable the exploration of its magnetic behavior by specific techniques such as XMCD, as it has been shown for single SGA layers in section II.B.2.
(ii) Magnetic tunnel junctions, where SGA could constitute an original electrode capable to spin polarize a current in its zero magnetization state. The investigation of SGA-based magnetic tunnel junctions permits to directly probe the possibility for zero-magnetization SGA to spin-polarize a current and is also a good way to probe interface magnetic and electronic properties. Moreover, the general context of this study is the exploration of the tunnel properties in magnetic systems where the electrodes are not conventional ferromagnets, as it is the case in the large majority of magnetic tunnel junctions. An intense research activity namely aims at designing materials that would permit to get a perfectly stable spin-polarized electrode. The idea is to eliminate interactions with other magnetic elements and external field and for this purpose, ideal compounds would be half-metallic antiferromagnets or ferrimagnets with 100% spin polarization and very small or zero magnetization. Up to now, these materials however still remain theoretical objects. Conventional ferrimagnetic alloys (RE-TM) have been already used as electrode in MTJ’s and a finite tunnel spin polarization could be measured at the magnetic compensation point. The use of SGA permits to go a step further with a material for which the zero magnetization state coexists with a “true” ferromagnetic state where all spin contributions point towards the same direction.
Table of contents :
Chapter 1: Sm1-xGdxAl2, a recent story
I. From Sm3+ to SmAl2
I.A. Localized magnetism in Lanthanide metals
I.B. Sm multiplets mixing in a metallic matrix
I.C. SmAl2, a self-ferrimagnet
II. Sm1-xGdxAl2, a zero-magnetization ferromagnet
II.A. First evidence of the zero-magnetization state
II.B. Ferromagnetic order at Tcomp
III. SGA as epitaxial films
III.A. Epitaxial growth and structural properties
III.B. Magnetic properties
III.B.1. Perpendicular anisotropy and giant coercivity
III.B.2. 5d spin polarization and long-range ferromagnetic order
IV. SGA epitaxial films in magnetic heterostructures
IV.A. Exchange-coupled bilayers
IV.B. Magnetic Tunnel Junctions
V. Summary and purpose of this thesis
Chapter 2: Photoemission spectroscopy
I. Generalities on photoemission spectroscopy
I.A. Photoemission and the photoelectric effect
I.B. The three-steps model
I.B.1. Optical excitation of one electron in the solid
I.B.2. Propagation of the photoelectron to the surface
I.B.3. Escape of the photoelectron from the solid to the vacuum
I.C. Photoemission and dipole selection rules
II. Angle Resolved PhotoEmission Spectroscopy (ARPES)
II.A. Geometry of the ARPES experiment
II.B. Information provided by ARPES experiments
III. Spin-Resolved PhotoEmission Spectroscopy
IV. CASSIOPEE beamline
Chapter 3: Electronic properties of (111)Sm1-xGdxAl2 surface and interface
I. Literature review
I.A. Electronic band structure in Rare Earth dialuminides
I.B. Valence stability in Sm metal and SmAl2
I.C. Electronic surface state in Lanthanide elements
II. (111) SGA surface preparation and characterization
II.A. Chemical analysis
II.B. Structural analysis of the Nb-free surface
II.C. Electronic analysis
III. Electronic structure of (111) SGA surface
III.A. Samarium 4f multiplets
III.B. Valence band analysis (ARPES)
III.B.1 In-plane wave vector measurements
III.B.2 Out-of-plane wave vector measurements
III.B.3 Discussion on the nature of observed electronic states
III.C. Spin Resolved analysis
IV. Electronic structure of (111)SGA/MgO
IV.A. Interface oxidation
IV.B. Samarium 4f multiplets
IV.C. Valence band analysis
Chapter 4: Overview of the magnetic anisotropy in [Co/Pt] multilayers
I. Introduction to the magnetic anisotropy
I.A. Origin of magnetic anisotropy in thin magnetic films
I.A.1. Shape anisotropy
I.A.2. Magneto-crystalline anisotropy
I.A.3. Surface or interface anisotropy
I.A.4. Magneto-elastic anisotropy
I.B. Effect of the roughness and interface alloy
I.C. Experimental determination of Keff
I.D. Stoner-Wohlfarth model
II. Perpendicular Magnetic Anisotropy in [Co/Pt] multilayers
II.A. Influence of the crystallographic orientation and preparations
II.B. Deviation from the linear behavior
Chapter 5: Magnetic and structural properties in [Co/Pt] multilayers
I. Description of the [Co/Pt] multilayers
II. Structural properties of the [Co/Pt] multilayers
II.A. Generalities on X-ray diffraction
II.B. Experimental setup
II.C. Growth direction in the sputtered [Co/Pt] multilayers
II.D. Reflectometry analysis on [Co/Pt] multilayers
II.D.1. Co and Pt thicknesses
II.D.2. Co and Pt roughnesses
III. Magnetic properties in the [Co/Pt] multilayers
III.A. Experimental setup
III.B. Experimental results
III.B.1 Effective anisotropy
III.B.2. Saturation magnetization
III.B.3. Influence of temperature
Chapter 6: Magnetic Tunnel Junctions with [Co/Pt]-based electrodes
I.A. MTJ’s structure
I.B. Micromagnetic simulations and magnetic parameters of the materials
II. Magnetic properties in MTJs full stack
II.A. Magnetization reversal
II.B. Magnetic configurations
II.B.1. Influence of the applied magnetic field
II.B.2. Influence of the temperature
II.B.3. Calculated (H,T) phase diagram for the hard electrode
III. Transport properties in nano-patterned MTJs
III.A. Common features in perpendicular MTJs
III.A.1. Tunnel characteristics
III.A.2. R(H) characteristics
III.B. Phase diagrams in nano-patterned MTJs
III.B.1. Duplication of domains from the hard to the soft electrode
III.B.2. Spring magnet and exchange bias behavior
III.C. Coupling through the MgO barrier
Chapter 7: Spin-polarization in [Co/Pt]- and SGA- based magnetic tunnel junctions
I. TMR and effective polarization in [Co/Pt] based-MTJs
I.A. Effective polarization of [Co/Pt]: more than an interface phenomenon?
I.B. Bloch low and Curie temperature
II. TMR and spin-polarization in SGA/AlOx/[Co/Pt] MTJs
II.A. Tunnel characteristic
II.B. Finite magneto-resistance tunnel at the magnetic compensation
III. Preliminary results on SGA/MgO/[Co/Pt] MTJs
III.A. SGA surface preparation
III.B. Common features in perpendicular SGA-based MTJs
III.B.1. Tunnel conductance versus voltage
III.B.2. Tunnel conductance versus temperature
III.B.3. Resistance versus applied field characteristics