Refine
Has Fulltext
- yes (5)
Is part of the Bibliography
- yes (5)
Document Type
- Doctoral Thesis (5) (remove)
Language
- English (5) (remove)
Keywords
- MBE (5) (remove)
Institute
In the present thesis the MBE growth and sample characterization of HgTe structures is investigated
and discussed. Due to the first experimental discovery of the quantum Spin Hall effect
(QSHE) in HgTe quantum wells, this material system attains a huge interest in the spintronics
society. Because of the long history of growing Hg-based heterostructures here at the Experimentelle
Physik III in Würzburg, there are very good requirements to analyze this material
system more precisely and in new directions. Since in former days only doped HgTe quantum
wells were grown, this thesis deals with the MBE growth in the (001) direction of undoped
HgTe quantum wells, surface located quantum wells and three dimensional bulk layers. All
Hg-based layers were grown on CdTe substrates which generate strain in the layer stack and
provide therefore new physical effects. In the same time, the (001) CdTe growth was investigated
on n-doped (001) GaAs:Si because the Japanese supplier of CdTe substrates had a
supply bottleneck due to the Tohoku earthquake and its aftermath in 2011.
After a short introduction of the material system, the experimental techniques were demonstrated
and explained explicitly. After that, the experimental part of this thesis is displayed.
So, the investigation of the (001) CdTe growth on (001) GaAs:Si is discussed in chapter 4.
Firstly, the surface preparation of GaAs:Si by oxide desorption is explored and analyzed.
Here, rapid thermal desorption of the GaAs oxide with following cool down in Zn atmosphere
provides the best results for the CdTe due to small holes at the surface, while e.g. an atomic
flat GaAs buffer deteriorates the CdTe growth quality. The following ZnTe layer supplies the
(001) growth direction of the CdTe and exhibits best end results of the CdTe for 30 seconds
growth time at a flux ratio of Zn/Te ~ 1/1.2. Without this ZnTe layer, CdTe will grow in the
(111) direction. However, the main investigation is here the optimization of the MBE growth
of CdTe. The substrate temperature, Cd/Te flux ratio and the growth time has to be adjusted
systematically. Therefore, a complex growth process is developed and established. This optimized
CdTe growth process results in a RMS roughness of around 2.5 nm and a FWHM value
of the HRXRD w-scan of 150 arcsec. Compared to the literature, there is no lower FWHM
value traceable for this growth direction. Furthermore, etch pit density measurements show
that the surface crystallinity is matchable with the commercial CdTe substrates (around 1x10^4
cm^(-2)). However, this whole process is not completely perfect and offers still room for improvements.
The growth of undoped HgTe quantum wells was also a new direction in research in contrast
to the previous n-doped grown HgTe quantum wells. Here in chapter 5, the goal of very low
carrier densities was achieved and therefore it is now possible to do transport experiments in
the n - and p - region by tuning the gate voltage. To achieve this high sample quality, very precise
growth of symmetric HgTe QWs and their HRXRD characterization is examined. Here,
the quantum well thickness can now determined accurate to under 0.3 nm. Furthermore, the transport analysis of different quantum well thicknesses shows that the carrier density and
mobility increase with rising HgTe layer thickness. However, it is found out that the band
gap of the HgTe QW closes indirectly at a thickness of 11.6 nm. This is caused by the tensile
strained growth on CdTe substrates. Moreover, surface quantum wells are studied. These
quantum wells exhibit no or a very thin HgCdTe cap. Though, oxidization and contamination
of the surface reduces here the carrier mobility immensely and a HgCdTe layer of around 5 nm
provides the pleasing results for transport experiments with superconductors connected to the
topological insulator [119]. A completely new achievement is the realization of MBE growth
of HgTe quantum wells on CdTe/GaAs:Si substrates. This is attended by the optimization of
the CdTe growth on GaAs:Si. It exposes that HgTe quantum wells grown in-situ on optimized
CdTe/GaAs:Si show very nice transport data with clear Hall plateaus, SdH oscillations, low
carrier densities and carrier mobilities up to 500 000 cm^2/Vs. Furthermore, a new oxide etching
process is developed and analyzed which should serve as an alternative to the standard
HCl process which generates volcano defects at some time. However, during the testing time
the result does not differ in Nomarski, HRXRD, AFM and transport measurements. Here,
long-time tests or etching and mounting in nitrogen atmosphere may provide new elaborate
results.
The main focus of this thesis is on the MBE growth and standard characterization of HgTe bulk
layers and is discussed in chapter 6. Due to the tensile strained growth on lattice mismatched
CdTe, HgTe bulk opens up a band gap of around 22 meV at the G-point and exhibits therefore
its topological surface states. The analysis of surface condition, roughness, crystalline quality,
carrier density and mobility via Nomarski, AFM, XPS, HRXRD and transport measurements
is therefore included in this work. Layer thickness dependence of carrier density and mobility
is identified for bulk layer grown directly on CdTe substrates. So, there is no clear correlation
visible between HgTe layer thickness and carrier density or mobility. So, the carrier density is
almost constant around 1x10^11 cm^(-2) at 0 V gate voltage. The carrier mobility of these bulk
samples however scatters between 5 000 and 60 000 cm^2/Vs almost randomly. Further experiments
should be made for a clearer understanding and therefore the avoidance of unusable
bad samples.But, other topological insulator materials show much higher carrier densities and
lower mobility values. For example, Bi2Se3 exhibits just density values around 1019 cm^(-2)
and mobility values clearly below 5000 cm2/Vs. The carrier density however depends much
on lithography and surface treatment after growth. Furthermore, the relaxation behavior and
critical thickness of HgTe grown on CdTe is determined and is in very good agreement with
theoretical prediction (d_c = 155 nm). The embedding of the HgTe bulk layer between HgCdTe
layers created a further huge improvement. Similar to the quantum well structures the carrier
mobility increases immensely while the carrier density levels at around 1x10^11 cm^(-2) at 0
V gate voltage as well. Additionally, the relaxation behavior and critical thickness of these
barrier layers has to be determined. HgCdTe grown on commercial CdTe shows a behavior as
predicted except the critical thickness which is slightly higher than expected (d_c = 850 nm).
Otherwise, the relaxation of HgCdTe grown on CdTe/GaAs:Si occurs in two parts. The layer
is fully strained up to 250 nm. Between 250 nm and 725 nm the HgCdTe film starts to relax
randomly up to 10 %. The relaxation behavior for thicknesses larger than 725 nm occurs than
linearly to the inverse layer thickness. A explanation is given due to rough interface conditions
and crystalline defects of the CdTe/GaAs:Si compared to the commercial CdTe substrate. HRXRD and AFM data support this statement. Another point is that the HgCdTe barriers protect the active HgTe layer and because of the high carrier mobilities the Hall measurements provide new transport data which have to be interpreted more in detail in the future. In addition, HgTe bulk samples show very interesting transport data by gating the sample from the top and the back. It is now possible to manipulate the carrier densities of the top and bottom surface states almost separately. The back gate consisting of the n-doped GaAs substrate and the thick insulating CdTe buffer can tune the carrier density for Delta(n) ~ 3x10^11 cm^(-2). This is sufficient to tune the Fermi energy from the p-type into the n-type region [138].
In this thesis it is shown that strained HgTe bulk layers exhibit superior transport data by embedding between HgCdTe barrier layers. The n-doped GaAs can here serve as a back gate.
Furthermore, MBE growth of high crystalline, undoped HgTe quantum wells shows also new
and extended transport output. Finally, it is notable that due to the investigated CdTe growth
on GaAs the Hg-based heterostructure MBE growth is partially independent from commercial
suppliers.
A novel growth method has been developed, allowing for the growth of strained HgTe shells on CdTe nanowires (NWs). The growth of CdTe-HgTe core-shell NWs required high attention in controlling basic parameters like substrate temperature and the intensity of supplied material fluxes. The difficulties in finding optimized growth conditions have been successfully overcome in this work.
We found the lateral redistribution of liquid growth seeds with a ZnTe growth start to be crucial to trigger vertical CdTe NW growth. Single crystalline zinc blende CdTe NWs grew, oriented along [111]B. The substrate temperature was the most critical parameter to achieve straight and long wires. In order to adjust it, the growth was monitored by reflection high-energy electron diffraction, which was used for fine tuning of the temperature over time in each growth run individually. For optimized growth conditions, a periodic diffraction pattern allowed for the detailed analysis of atomic arrangement on the surfaces and in the bulk. The ability to do so reflected the high crystal quality and ensemble uniformity of our CdTe NWs. The NW sides were formed by twelve stable, low-index crystalline facets. We observed two types stepped and polar sides, separated by in total six flat and non-polar facets.
The high crystalline quality of the cores allowed to grow epitaxial HgTe shells around. We reported on two different heterostructure geometries. In the first one, the CdTe NWs exhibit a closed HgTe shell, while for the second one, the CdTe NWs are overgrown mainly on one side. Scanning electron microscopy and scanning transmission electron microscopy confirmed, that many of the core-shell NWs are single crystalline zinc blende and have a high uniformity. The symmetry of the zinc blende unit cell was reduced by residual lattice strain. We used high-resolution X-ray diffraction to reveal the strain level caused by the small lattice mismatch in the heterostructures. Shear strain has been induced by the stepped hetero-interface, thereby stretching the lattice of the HgTe shell by 0.06 % along a direction oriented with an angle of 35 ° to the interface.
The different heterostructures obtained, were the base for further investigation of quasi-one-dimensional crystallites of HgTe. We therefore developed methods to reliably manipulate, align, localize and contact individual NWs, in order to characterize the charge transport in our samples. Bare CdTe cores were insulating, while the HgTe shells were conducting. At low temperature we found the mean free path of charge carriers to be smaller, but the phase coherence length to be larger than the sample size of several hundred nanometers. We observed universal conductance fluctuations and therefore drew the conclusion, that the trajectories of charge carriers are defined by elastic backscattering at randomly distributed scattering sites. When contacted with superconducting leads, we saw induced superconductivity, multiple Andreev reflections and the associated excess current. Thus, we achieved HgTe/superconductor interfaces with high interfacial transparency.
In addition, we reported on the appearance of peaks in differential resistance at Delta/e for HgTe-NW/superconductor and 2*Delta/e for superconductor/HgTe-NW/superconductor junctions, which is possibly related to unconventional pairing at the HgTe/superconductor interface. We noticed that the great advantage of our self-organized growth is the possibility to employ the metallic droplet, formerly seeding the NW growth, as a superconducting contact. The insulating wire cores with a metallic droplet at the tip have been overgrown with HgTe in a fully in-situ process. A very high interface quality was achieved in this case.
Oxide heterostructures attract a lot of attention as they display a vast range of physical phenomena like conductivity, magnetism, or even superconductivity. In most cases, these effects are caused by electron correlations and are therefore interesting for studying fundamental physics, but also in view of future applications. This thesis deals with the growth and characterization of several prototypical oxide heterostructures. Fe3O4 is highly ranked as a possible spin electrode in the field of spintronics. A suitable semiconductor for spin injection in combination with Fe3O4 is ZnO due to its oxide character and a sufficiently long spin coherence length. Fe3O4 has been grown successfully on ZnO using pulsed laser deposition and molecular beam epitaxy by choosing the oxygen partial pressure adequately. Here, a pressure variation during growth reduces an FeO-like interface layer. Fe3O4 films grow in an island-like growth mode and are structurally nearly fully relaxed, exhibiting the same lattice constants as the bulk materials. Despite the presence of a slight oxygen off-stoichiometry, indications of the Verwey transition hint at high-quality film properties. The overall magnetization of the films is reduced compared to bulk Fe3O4 and a slow magnetization behavior is observed, most probably due to defects like anti-phase boundaries originating from the initial island growth. LaAlO3/SrTiO3 heterostructures exhibit a conducting interface above a critical film thickness, which is most likely explained by an electronic reconstruction. In the corresponding model, the potential built-up owing to the polar LaAlO3 overlayer is compensated by a charge transfer from the film surface to the interface. The properties of these heterostructures strongly depend on the growth parameters. It is shown for the first time, that it is mainly the total pressure which determines the macroscopic sample properties, while it is the oxygen partial pressure which controls the amount of charge carriers near the interface. Oxygen-vacancy-mediated conductivity is found for too low oxygen pressures. A too high total pressure, however, destroys interface conductivity, most probably due to a change of the growth kinetics. Post-oxidation leads to a metastable state removing the arbitrariness in controlling the electronic interface properties by the oxygen pressure during growth. LaVO3/SrTiO3 heterostructures exhibit similar behavior compared to LaAlO3/SrTiO3 when it comes to a thickness-dependent metal-insulator transition. But in contrast to LaAlO3, LaVO3 is a Mott insulator exhibiting strong electron correlations. Films have been grown by pulsed laser deposition. Layer-by-layer growth and a phase-pure pervoskite lattice structure is observed, indicating good structural quality of the film and the interface. An electron-rich layer is found near the interface on the LaVO3 side for conducting LaVO3/SrTiO3. This could be explained by an electronic reconstruction within the film. The electrostatic doping results in a band-filling-controlled metal-insulator transition without suffering from chemical impurities, which is unavoidable in conventional doping experiments.
In this work the epitaxial growth and characterization of the half-metallic ferromagnet NiMnSb was presented. NiMnSb crystallizes in the C1b structure which is similar to the zinc blende structure from widely used III-V semiconductors. One special property of NiMnSb is the theoretical 100% spin-polarization at the Fermi edge. This makes it a perfect candidate for spintronic experiments and the material of choice for building novel spintronic devices. Another important topic in this work were the magnetic properties of NiMnSb, especially the low magnetic damping of the grown thin films. All grown layers were fabricated with the technique of MBE. The layer stacks for all different experiments and devices were grown on InP substrate in (001) or (111)B orientation. Before the NiMnSb layer a buffer layer of undoped (In,Ga)As was grown. Additional for some samples on InP(111)B, a Si doped (In,Ga)As layer was grown on top of the undoped (In,Ga)As layer. The dopant concentration of this n-doped layer was determined by ETCH-CV. All layers were investigated by structural and the NiMnSb layer additional by magnetic properties. For the structural investigation the in-situ technique RHEED and ex-situ tool HRXRD were used. RHEED observations showed a good quality of the grown buffer and half-metallic ferromagnet layers on both orientations. These results were strengthened by the HRXRD measurement. The vertical lattice constant could be determined. The received value of a(NiMnSb_vertical) = 5.925 Å for NiMnSb on InP(001) is in good agreement to the value a(NiMnSb_Lit) = 5.903 Å found in literature [Cas55]. For NiMnSb on InP(111)B a vertical lattice constant of a(NiMnSb_vertikal) = 6.017 Å could be determined. The horizontal lattice constant of the buffer and the half-metallic ferromagnet layer could be determined as the same of the substrate. For NiMnSb this conclusion is only valid up to a thickness of ≈40nm. To increase this maximum thickness, NiMnSb samples were grown on InP(001) substrates and capped with Ti/Au layers. Afterwards a reciprocal space map of the (533) reflex was drawn with GIXRD at the synchrotron beamline BW2 of HASYLAB [Kum07]. It has been shown that the critical thickness is more than doubled by depositing a Ti/Au capping directly after growth of NiMnSb without breaking the ultrahigh vacuum (UHV). The magnetic properties were determined with FMR experiments and SQUID measurements. The received magnetic damping parameter α from a 40nm thick NiMnSb layer on InP(001) could be determined to 3.19e−3 along [1-10]. The resulting line width of our NiMnSb layers on InP(001) is more than 4.88 times smaller than measured before [Hei04]. Another result is the direction dependence of the damping. It has been measured that the difference of the damping is changed by more than 42% when rotating the applied field by 45° from [1-10] to [100].With SQUID we measured a saturation magnetization of a 40nm thick NiMnSb layer as 4µB. NiMnSb layers on InP(111)B substrate where also measured with FMR with a surprising result. These layers not only showed a decreasing in the anisotropy field with increasing thickness but also an uniaxial anisotropy. This behaviour can be explained with defects on these samples. With an AFM triangle-like defects were measured. These defects originated from the buffer layer and influenced the magnetic properties. Another part of this work is dedicated to the behaviour of NiMnSb at temperatures around 80K. With our samples, no phase transition can be observed in the data of the Hall, anomalous Hall term and resistivity. The last part of this work discusses different spintronic devices build with our NiMnSb layers. In a first device the magnetization acts on the current. This Giant Magneto Resistance (GMR) device consisted of InP:S(001) - 180nm undoped (In,Ga)As - 40nm NiMnSb - 10nm Cu - 6nm NiFe - 10nm Ru in current perpendicular to plane (CPP) geometry. We received a Magneto-Resistance-Ratio of 3.4%. In a second device the current acts on the magnetization and makes use of the spin torque phenomena. This so called Spin Torque Oscillator (STO) emitted frequencies in the GHz range (13.94GHz - 14.1GHz). The last fabricated device is based on the magnetic vortex phenomena. For switching the core polarity the gyrotropic frequencies f + = 254MHz f − = 217MHz and a total static magnetic field of only mµ0H = 65mT were necessary. The reversal efficiency has been determined as better than 99% [Lou09].
In this work heterostructures based on the half-Heusler alloy NiMnSb have been fabricated and characterized. NiMnSb is a member of the half-metallic ferromagnets, which exhibit an electron spin-polarization of 100% at the Fermi-level. For fabrication of these structures InP substrates with surface orientations of (001),(111)A and (111)B have been used. The small lattice mismatch of NiMnSb to InP allows for pseudomorphic layers, the (111) orientation additionally makes the formation of a half-metallic interface possible. For the growth on InP(001), procedures for the substrate preparation, growth of the lattice matched (In,Ga)As buffer layer and of the NiMnSb layer have been developed. The effect of flux-ratios and substrate temperatures on the MBE growth of the buffer as well as of the NiMnSb layer have been investigated and the optimum conditions have been pointed out. NiMnSb grows in the layer-by-layer Frank-van der Merwe growth mode, which can be seen by the intensity oscillations of the RHEED specular spot during growth. RHEED and LEED measurements show a flat surface and a well-defined surface reconstruction. High resolution x-ray measurements support this statement, additionally they show a high crystalline quality. Measurements of the lateral and the vertical lattice constant of NiMnSb films on (001) oriented substrates show that layers above a thickness of 20nm exhibit a pseudomorphic as well as a relaxed part in the same layer. Whereas layers around 40nm show partly relaxed partitions, these partitions are totally relaxed for layers above 100nm. However, even these layers still have a pseudomorphic part. Depth-dependent x-ray diffraction experiments prove that the relaxed part of the samples is always on top of the pseudomorphic part. The formation and propagation of defects in these layers has been investigated by TEM. The defects nucleate early during growth and spread until they form a defect network at a thickness of about 40nm. These defects are not typical misfit dislocations but rather antiphase boundaries which evolve in the Mn/Sb sublattice of the NiMnSb system. Dependent on the thickness of the NiMnSb films different magnetic anisotropies can be found. For layers up to 15nm and above 25nm a clear uniaxial anisotropy can be determined, while the layers with thicknesses in between show a fourfold anisotropy. Notably the easy axis for the thin layers is perpendicular to the easy axis observed for the thick layers. Thin NiMnSb layers show a very good magnetic homogeneity, as can be seen by the very small FMR linewidth of 20Oe at 24GHz. However, the increase of the linewidth with increasing thickness shows that the extrinsic damping gets larger for thicker samples which is a clear indication for magnetic inhomogeneities introduced by crystalline defects. Also, the magnetic moment of thick NiMnSb is reduced compared to the theoretically expected value. If a antiferromagnetic material is deposited on top of the NiMnSb, a clear exchange biasing of the NiMnSb layer can be observed. In a further step the epitaxial layers of the semiconductor ZnTe have been grown on these NiMnSb layers, which enables the fabrication of NiMnSb/ZnTe/NiMnSb TMR structures. These heterostructures are single crystalline and exhibit a low surface and interface roughness as measured by x-ray reflectivity. Magnetic measurements of the hysteresis curves prove that both NiMnSb layers in these heterostructures can switch separately, which is a necessary requirement for TMR applications. If a NiMn antiferromagnet is deposited on top of this structure, the upper NiMnSb layer is exchange biased by the antiferromagnet, while the lower one is left unaffected. Furthermore the growth of NiMnSb on (111) oriented substrates has been investigated. For these experiments, InP substrates with a surface orientation of (111)A and (111)B were used, which were miscut by 1 to 2° from the exact orientation to allow for smoother surfaces during growth. Both the (In, Ga)As buffer as well as the NiMnSb layer show well defined surface reconstructions during growth. X-ray diffraction experiments prove the single crystalline structure of the samples. However, neither for the growth on (111)A nor on (111)B a perfectly smooth surface could be obtained during growth, which can be attributed to the formation of pyramid-like facets evolving as a result of the atomic configuration at the surface. A similar relaxation behavior as NiMnSb layers on (001) oriented InP could not be observed. RHEED and x-ray diffraction measurements show that above a thickness of about 10nm the NiMnSb layer begins to relax, but remnants of pseudomorphic parts could not be found. Magnetic measurements show that the misorientation of the substrate crystal has a strong influence on the magnetic anisotropies of NiMnSb(111) samples. In all cases a uniaxial anisotropy could be observed. The easy axis is always aligned parallel to the direction of the miscut of the substrate.