Influence of oxygen in the sputter gas
on creating a conducting interface in
LaAlO
3
/SrTiO
3
THESIS
submitted in partial fulfillment of the requirements for the degree of
MASTER OF SCIENCE
in PHYSICS
Author : C.S. Remeijer BSc
Student ID : 0778745
Supervisor : Prof. dr. J. Aarts
C. Yin MSc
2ndcorrector : Prof. dr. ir. T.H. Oosterkamp
Influence of oxygen in the sputter gas
on creating a conducting interface in
LaAlO
3
/SrTiO
3
C.S. Remeijer BSc
Huygens-Kamerlingh Onnes Laboratory, Leiden University P.O. Box 9500, 2300 RA Leiden, The Netherlands
22 February 2019
Abstract
In this study we investigate if we can change the interfacial conductivity in
LaAlO3/SrTiO3heterostructures by adding some oxygen to the argon sputter
gas.
We vary the deposition parameters to minimise the effect of the oxygen. We find that an increase in oxygen partial pressure increases the surface roughness. We also find that an increase in argon flow increases surface roughness. The optimised parameters, which minimise surface roughness, are in agreement with previous results of growing a conducting interface in LaAlO3/SrTiO3by
sputter-ing. However all our samples grown with oxygen are found to be insulatsputter-ing. A possible explanation for the effect of oxygen in the sputter gas is that the species oxidise before reaching the sample.
We conclude that even a very small amount of oxygen in the sputter gas gives an insulating interface and that we can not make a conducting interface with oxygen in the sputter gas with the experimental set-up used.
Contents
1 Introduction . . . . 5
2 Theory . . . . 7
2.1 LaAlO3/SrTiO3heterointerface . . . 7
2.2 Sputtering . . . 14
3 Experimental methods . . . . 17
3.1 Substrate termination . . . 17
3.2 Thin film deposition . . . 21
3.2.1 90° off-axis sputter machine . . . 21
3.2.2 Sputter deposition procedure . . . 22
3.3 Sample characterization . . . 26
4 Results . . . . 29
4.1 Total pressure . . . 29
4.2 Oxygen partial pressure . . . 33
4.3 Sputter power and temperature . . . 37
4.4 Control measurements . . . 40
5 Discussion . . . . 43
6 Conclusion . . . . 47
References . . . . 49
Chapter
1
Introduction
In 2004 Ohtomo & Hwang discovered a conducting two-dimensional layer with a high electron mobility at the interface of LaAlO3(LAO) and SrTiO3(STO) [1].
Its discovery stimulated many investigations into oxide heterostructures. The LAO/STO interface exhibits a wide range of intriguing physical phenomena des-pite STO and LAO being simple band insulators. LAO/STO has been reported to display interfacial superconductivity [2], magnetism [3, 4], a gate tunable insu-lator to superconductor transition [5] and gate tunable spin-orbit interactions [6]. Oxide heterostructures are of interest as electronic devices because they exhibit a high degree of tunability and a wide variety of physical phenomena. However the mechanism behind the interfacial conductance and other phenomena is still not clear despite the many investigations.
Ohtomo & Hwang used pulsed laser deposition (PLD) to make LAO/STO and ever since it is the most common growth method for LAO/STO heterostructures. In PLD laser pulses ablate particles from a target, which then transfer to a sub-strate as a plasma plume and form a thin film. PLD can be performed in vacuum, but oxides are usually grown with an appropriate oxygen background to deposit crystalline films with the correct oxygen stoichiometry [7]. A disadvantage of PLD is that the high energy incident particles cause oxygen vacancies and cation intermixing at the interface [8].
Other growth techniques shed some light on the mechanism behind the forma-tion of the conducting interface. Molecular-beam epitaxy (MBE) showed the im-portance of stoichiometry when Warusawithana et al. [9] found that only Al-rich LAO gives a conducting interface and stoichiometric and La-rich samples have insulating interfaces. Sputter deposition confirmed the role of stoichiometry [10].
6 Introduction
The insulating nature of stoichiometric samples calls into question the electronic reconstruction model for the conducting interface. The results from MBE and sputtering support the oxygen vacancy model instead.
Sputtering has the advantage over PLD that it is a scalable process. The ability to produce larger samples is an important aspect for industrial applications. Sputtering is also cheaper than PLD.
Initial attempt to make a conducting interface in LAO/STO by sputtering gave insulating samples with a high La/Al ratio of 1.1 [11, 12]. The samples were grown in high oxygen pressure with an on-axis geometry. Later Podkaminer et al. [13] demonstrated that sputtering can give interfacial conductance by using a 90° off-axis sputter geometry and sputtering in argon.
The gas pressure is an important growth parameter in both PLD and sputter-ing. The switch of sputter geometry lowered the necessary gas pressure, but Podkaminer et al. also changed the sputter gas from oxygen to argon. They also tried a mixture of oxygen and argon in a 3:4 ratio and found only insulating interfaces [13]. These oxygen partial pressures are quite high though.
Here we investigate the effect of oxygen in the sputter gas on creating a conduct-ing interface in LAO/STO. We introduce a small amount of oxygen in the argon sputter gas and optimise the growth parameters. We find that oxygen in the sputter gas does not give a smooth surface and results in an insulating interface and that higher oxygen partial pressures give rougher surfaces.
Chapter
2
Theory
2.1
LaAlO
3/SrTiO
3heterointerface
In this section we describe the materials studied in this research project, i.e.
SrTiO3(STO) and LaAlO3(LAO), and how a two-dimensional conducting layer
can form at the interface of the LAO/STO heterostructure.
LAO and STO
Strontium titanate and lanthanum aluminate are both band insulators with a perovskite structure. Table 2.1 lists their most important material properties.
Table 2.1:Material properties of STO and LAO at room temperature.
Material Band gap Crystal structure Lattice parameter
SrTiO3 3.2 eV Cubic 3.905 ˚A
LaAlO3 5.6 eV Rhombohedral 3.789 ˚A
Perovskite oxides have a chemical formula of the form ABO3. Figure 2.1 shows
the perovskite structure with a larger A and a smaller B cation and three oxygen atoms per unit cell. The A atom is located at the cube corner (0, 0, 0), the B atom is located at the body-centred position (1/2, 1/2, 1/2) and the oxygen atoms sit at face-centred positions, e.g. (1/2, 1/2, 0). The A and B cations can take several charge values to make the ABO3structure neutral, e.g. A4+B2+or A3+B3+
8 Theory
Figure 2.1: Ideal cubic perovskite structure. The green atoms are the larger A cations
located at the cube corners, the blue atoms the smaller B cations located at the body-centred positions and the red atoms are the oxygen atoms located at the face-body-centred positions.
Sr has valence state Sr2+and Ti has several valance states of which Ti4+and Ti3+ are the most common so neutral STO has Sr2+Ti4+O3 electronic configuration.
The valance state of La is La3+and of Al it is Al3+so neutral LAO has La3+Al3+O3
electronic configuration. The (001) planes of these ABO3materials are
alternat-ing AO and BO2 planes. In the (001) direction STO can be seen as a stack of
alternating neutral (001) planes (SrO)0and (TiO2)0. LAO can be seen as a stack
of alternating polar (001) planes of (LaO)+and (AlO2)−.
The LAO/STO heterostructure we discuss here is epitaxially grown LAO on STO cleaved in the (001) direction. The STO crystal has a small miscut angle giving rise to a mixed termination of TiO2and SrO. It is possible to make the surface
singly terminated. LAO grown on TiO2-terminated STO always starts with
LaO and LAO grown on SrO-terminated STO starts with AlO2[1]. Figure 2.2a
shows LAO grown on TiO2-terminated STO resulting in the (LaO)+/(TiO2)0
interface and figure 2.2b shows LAO grown on SrO-terminated STO resulting in the (AlO2)−/(SrO)0interface. The lattice mismatch between LAO and STO is
3 %, which means that under the right conditions LAO can be grown strained to the STO.
2.1 LaAlO3/SrTiO3heterointerface 9
a b
Figure 2.2: Schematic models of the two possible interfaces between LaAlO3 and
SrTiO3in the (001) orientation. aSchematic of the resulting (LaO)+/(TiO2)0interface,
showing the composition of each layer and the ionic charge state of each layer. b Schem-atic of the resulting (AlO2)−/(SrO)0interface. Figure reprinted from [1].
Interfacial conductance
Ohtomo & Hwang [1] showed in 2004 that the n-type (LaO)+/(TiO2)0interface
is conducting. It has a high carrier density (1.05×1013cm−2 at 250 mK) and a very high carrier mobility (2.9×103cm2·V−1·s−1 at 250 mK) [14]. The two-dimensional nature of the interfacial conductance was proven several years later by Shubnikov–de Haas experiments [14]. The conducting interface is also called a two-dimensional electron gas (2DEG), a two-dimensional electron liquid (2DEL) or a two-dimensional electron system (2DES).
Conducting interfaces have also been found in LAO/STO for the (110) and (111) orientations of STO [15, 16] and in STO based heterostructures with other polar layers such as LaVO3[17], LaTiO3[18] and NdGaO3[19]. Several models have
been proposed to explain the interfacial conductance: electronic reconstruction, oxygen vacancies and cation intermixing.
Electronic reconstruction
If the interfaces are as the perfect model interfaces in figure 2.3a,b then the polar discontinuity, the transition from non-polar planes to polar planes, gives a poten-tial in the LAO that increases with LAO layer thickness: the polar catastrophe. Figure 2.3a,b also shows the diverging potentials for both LAO/STO interfaces.
10 Theory
Nakagawa et al. [20] proposed an electronic reconstruction in order to prevent the potential from diverging. Unlike in semi-conductors where only the ions can move, in oxide structures the electrons can move so that the ions do not have to. If a mixed valance state is available and the cost of redistributing electrons is lower than redistributing ions then a mixed valance charge compensation can occur. Figure 2.3a,c shows the electronic reconstruction of the (LaO)+/(TiO2)0
interface. Half an electron per unit cell is transferred from the LAO surface to the (LaO)+/(TiO2)0interface, where the Ti4+ion becomes Ti3.5+. This electronic
reconstruction makes the (LaO)+/(TiO2)0interface a n-type interface with half
an electron per two-dimensional unit cell available at the interface. Figure 2.3 does not show the surface of the LAO, but at the surface oxygen vacancies avert the polar catastrophe leading to an insulating surface [21].
Figure 2.3: The polar catastrophe illustrated for atomically abrupt (001) interfaces
between LaAlO3and SrTiO3. aThe unreconstructed interface has neutral (001) planes
in SrTiO3, but the (001) planes in LaAlO3have alternating net charges (ρ). If the interface plane is (LaO)+/(TiO2)0, this produces a non-negative electric field (E), leading in turn to an electric potential (V) that diverges with thickness. b If the interface is instead placed at the (AlO2)−/(SrO)0plane, the potential diverges negatively. c The divergence catastrophe at the (LaO)+/(TiO2)0interface can be avoided if half an electron is added to the last Ti layer. This produces an interface dipole that causes the electric field to oscillate about 0 and the potential remains finite. d The divergence for the (AlO2)−/(SrO)0 interface can also be avoided by removing half an electron from the SrO plane in the form of oxygen vacancies. Figure reprinted from [20].
At the (AlO2)−/(SrO)0interface the polar catastrophe can be averted by
2.1 LaAlO3/SrTiO3heterointerface 11
Figure 2.3b,d shows the atomic reconstruction of the (AlO2)−/(SrO)0interface.
This interface was expected to be p-type, but until recently experiments showed that this interface is insulating [1]. No mixed valance states are available (Sr3+ and Ti5+ are energetically inaccessible) so an atomic reconstruction was thought to be required to avert the polar catastrophe leading to an insulating interface, introducing oxygen vacancies in the top SrO layer [20].
In 2018 Lee et al. [21] found simultaneous conducting interfaces, both a 2DEG and a two-dimensional hole gas (2DHG), in a STO/LAO/STO heterostructure. Growing this heterostructure in the correct conditions prevents positive ionic charges in the form of oxygen vacancies from averting the polar catastrophe at the p-type interface. They found p-type conduction with only holes as carriers. Thiel et al. [22] showed that a critical thickness of 4 unit cells of LAO is needed for the conducting interface to form. The internal electric field in LAO due to the polar nature of the (001) layers builds up a potential that bends the electronic bands. At some critical thickness tc, see figure 2.4a, the valance band of LAO
reaches the energy of the conduction band of STO and a Zener breakdown occurs. Above the critical thickness electrons are transferred from the surface of LAO to the interface making it conducting. Interfacial conductivity of crystalline LAO on STO shows little dependence on LAO thickness beyond the critical thickness [23]. The simple electrostatic model in figure 2.4 by Reinle-Schmitt et al. [24] provided further evidence for the electronic reconstruction model. They calculated the critical thickness tc of LAO according to equation 2.1 with e0the electric constant,
eLAOthe relative permittivity of LAO,∆E the difference in energy between the
valance band of LAO and the conduction band of STO, e the electron charge and PLAO0 the formal polarization of LAO. This gives a critical thickness of 3.5 unit cells, in good agreement with experiment.
tc = e0eLAO∆E
e PLAO0 (2.1)
They also made LAO/STO heterostructures where the LAO was replaced by various mixtures of LAO and STO and found that the critical thickness increases as more LAO is replaced by STO, providing a strong argument in favour of the electronic reconstruction model. Figure 2.4b shows the difference in potential build-up and critical thickness for pure LAO and a layer consisting of 50 % LAO and 50 % STO (LASTO:0.5) [24].
12 Theory
a b
Figure 2.4: Electrical potential build-up at polar interfaces. aThe band-level scheme
shows band bending in the pure LAO layer of the valence band (VB) and conduction band (CB), and the critical thickness tcand potential build-up eVcrequired to induce the electronic reconstruction. φnis the valence-band offset between STO and LAO. b Schem-atic of the potential build-up as a function of its thickness t for LAO and LASTO:0.5, assuming the same relative permittivity but different formal polarizations P0induced by the charge of the successive A-site and B-site sublayers. The critical thicknesses for the electronic reconstruction are labelled t(c1)and t(c2)for LAO and LASTO:0.5, respectively. Figure reprinted from [24].
Oxygen vacancies
The second proposed explanation is that oxygen vacancies in STO cause the interfacial conductance [25]. Oxygen vacancies are point defects in the lattice, an empty lattice site where normally an oxygen atom sits. Oxygen vacancies can be induced in STO by growing a film on top in reducing conditions [26].
Growing amorphous LAO, La2O3 or Al2O3 films on STO can also result in a
conducting interface [27]. At low enough oxygen pressure during deposition a redox reaction at the interface oxidises the deposited films and reduces the STO. These samples turn insulating after annealing in oxygen showing that the charge carriers of the conducting interface came from oxygen vacancies [27, 28].
Where the interfacial conductance vanishes in amorphous LAO/STO, it decreases in crystalline LAO/STO after oxygen annealing. This shows that both an
elec-2.1 LaAlO3/SrTiO3heterointerface 13
tronic reconstruction and oxygen vacancies give interfacial conductance [29, 30]. These results also show that crystallinity is a requirement for the electronic re-construction to occur. A combined experimental and theoretical research by Slooten et al. [31] found that oxygen vacancies in the topmost AlO2layer of the
LAO can provide half an electron per unit cell to the interface thereby creating a conducting interface.
Cation intermixing
The third proposed explanation of the interfacial conductance is cation intermix-ing. Willmott et al. [32] performed surface x-ray diffraction and phase-retrieval measurements to show that the interface is not a smooth transition, but that there is cation intermixing in the interface. Sr and La intermix at greater depth than Al and Ti atoms leading to the formation of one or two monolayers of metallic La1-xSrxTiO3. Density functional theory calculations showed that the Fermi level
near the interface rises above the bottom of conduction band giving a conducting interface. The high energy of the incident particles in PLD leads to a strong tendency to intermix at the interface rather than form an abrupt transition [8].
Stoichiometry
Another factor that is important to the formation of the conducting interface is the stoichiometry. Warusawithana et al. [9] made LAO/STO heterostructures with molecular-beam epitaxy (MBE) and found that the LAO layer needs to be Al-rich to give interfacial conductance. Stoichiometric and La-rich samples are insulating. The maximum La/Al ratio for interfacial conductance to occur is 0.97. The stoichiometry can be controlled with PLD by tuning the laser fluence [23] and with sputtering by tuning the pressure [10].
Warusawithana et al. [9] argues that La vacancies can be filled by Al, but not the other way around, so in Al-rich samples an electronic reconstruction takes place. Breckenfeld et al. [23] on the other hand argues that Al-rich films induce oxygen reduction in the STO. In La-rich samples Al2O3-vacancy complexes form,
which leads to an atomic reconstruction and insulating samples [9]. However the cation vacancies argument for electronic reconstruction does not explain why stoichiometric samples made by MBE and sputtering are insulating. This result seems to support oxygen vacancies as the source of the interfacial conductance. Even after fifteen years of research no consensus exists over the origin of the interfacial conductance in LAO/STO. Considering the role of stoichiometry both the electronic reconstruction and oxygen vacancies seem to play a role in the formation of the conducting interface.
14 Theory
2.2
Sputtering
In this section we describe how the thin film deposition technique sputtering works and why we use RF magnetron sputtering in a 90° off-axis geometry. Sputtering is a physical vapour deposition technique suitable to deposit (very) thin films. In sputtering atoms are ejected from a target by high energy ionized particles and then fly to a substrate. The ejected particles from the target, the species, arrive at the substrate and a thin film grows in a continuous way until sputtering is stopped.
In order to grow a pure film the target needs to be a high purity material and the sputtering needs to take place in a vacuum chamber. The chamber is evacuated to remove most contaminants and then a sputter gas, often argon as that is an inert gas, is let into the chamber for the sputtering process. Next a plasma needs to be created. Paschen’s law states that you need a minimum breakdown voltage to ignite a plasma depending on the pressure, distance and type of gas, which is why sputtering requires high voltages and a minimum gas pressure.
DC sputtering
Sputtering is most easily understood by considering DC sputtering. In DC sputtering a negative DC voltage is applied to the target, while the rest of the chamber is grounded, to ignite a plasma. The potential causes free electrons to flow away from the target material. The electrons collide with gas atoms, say argon for simplicity, knocking an electron from the argon atoms making them positively charged argon ions. The result of this collision is an argon ion and two electrons. The electrons can cause further collisions with argon atoms sustaining the plasma. Argon ions can also react with electrons becoming neutral again. This recombination releases energy in the form of a photon causing the plasma to glow.
Figure 2.5 shows a schematic representation of DC sputtering in an on-axis geometry. The argon ions are accelerated to the target by the negative potential and the collisions eject atoms from the target material. Atoms are only ejected if the incoming particle has sufficient energy to overcome the binding energy of the atoms in the target. Hence sputtering always needs high applied potentials. The number of collisions is inversely related to the gas pressure. The sputter power, gas pressure and target-sample distance determine the energy with which the species arrive at the sample. The energy of the arriving species and the sample temperature determine whether the species form an amorphous film or a crystalline one. A high sample temperature gives the arrived species the energy
2.2 Sputtering 15
Figure 2.5:Schematic representation of DC sputtering in an on-axis geometry.
to perform a random walk on the surface and find a place with low energy to sit. With the correct combination of sputter power, gas pressure, target-sample distance and sample temperature a thin films can be grown epitaxially.
RF magnetron sputtering
If the target material is conducting sputtering can be performed with a DC voltage, but if the target material is insulating then an applied negative DC voltage causes a charge build-up in the target. This causes electrical discharges. Sputtering of insulating materials is commonly done by RF sputtering. In RF sputtering an AC voltage is applied to the target and during the part of the cycle with negative voltage the gas ions continue to impact on the target. During the positive part of the cycle the charge build-up is removed preventing arcing from happening. The applied AC voltage usually has a radio frequency of 13.56 MHz, which is an international standard for RF power supply equipment.
In magnetron sputtering magnets are used behind the target to trap the charged particles close to the target. The electrons follow helical trajectories around the magnetic field lines causing more ionizing collisions with neutral gas atoms. This lowers the minimum gas pressure needed to sustain the plasma. It also prevents the electrons from being repelled by the negatively charged target and bombarding the sample. The ejected species are neutral and thus are not affected by the electric and magnetic field.
16 Theory
90° off-axis sputtering
The most used sputter geometry is on-axis, where the sample is located in front of the target. This is the simplest geometry and gives the highest deposition rate as a larger part of the ejected particles from the target reaches the sample compared to other geometries. However the sputter machine used in this research has an off-axis sputter geometry. The sample is not located in front of the target, but at an angle. Here an angle of 90° is employed.
The off-axis geometry means that the deposition rate is considerably lower than in on-axis sputtering because a large part of the particles ejected from the target goes in a different direction. The off-axis geometry also allows stoichiometry control. The gas pressure can be used to control the stoichiometry of the grown LAO film [10]. The weight of argon is 40 u so the light aluminium (27 u) scatters much more than the heavy lanthanum (139 u). A higher gas pressure increases the scattering and causes mostly the light aluminium to scatter a lot.
Chapter
3
Experimental methods
3.1
Substrate termination
In this section we describe the materials, equipment, machines and methods we use to process the as-received STO substrate. The goal of processing the STO is to make a TiO2-terminated STO substrate, which enables us to make a conducting
interface by depositing a thin layer of a few nanometre of LAO on top of it. The used STO wafer has a small miscut angle and thus has a mixed SrO- and TiO2-termination. To obtain TiO2-terminated STO we etch the STO in a buffered
HF solution [33]. We clean the STO, immerse it in water, etch the STO in a buffered HF solution, clean it again and anneal it at high temperature in oxygen.
The STO substrate we use is a 5 mm×5 mm×0.5 mm piece of SrTiO3 (100)K
with a miscut angle of 0.2° to 0.3° that is polished on one side. The STO is bought from CrysTec.
We use several solvents to clean glassware, tweezers and substrates: deionized Type I water from Millipore (18.2 MΩ·cm at 25◦C), ethanol (C2H5OH) and
isopropanol (C3H7OH or 2-propanol).
Before use we clean tweezers, tubes and the microscope glass with isopropanol applied to a wipe from Kimtech and blow them dry with nitrogen gas. Beakers and the tweezers for HF etching we clean by rinsing them with demiwater from a water gun five times and by blowing them dry with nitrogen gas afterwards. Before use we rinse all beakers with the solvent we are going to put in them twice and blow them dry with nitrogen gas afterwards.
18 Experimental methods
We use a digital optical microscope in a cleanroom to look for contaminants and large structures on our samples and to save images of this. This microscope has a magnification range from 50 to 1000 times.
Cleaning the STO (before etching)
The STO is cleaned by sliding Salm en Kipp lens paper with a little bit of isopro-panol applied to it over the substrate followed by sliding the substrate upside down over lens paper soaked in isopropanol while applying light pressure. We examine the substrate with an optical microscope and repeat if necessary. The next step is removing smaller contaminants with an ultrasonic bath. The STO is sonicated in ethanol for 5 min, in isopropanol for 10 min and in Type I water for 30 min. During transfer from one solvent to the next we spray the next solvent on the substrate to keep it wet at all times.
The last step of sonicating in Type I water is not just for cleaning purposes, but also a crucial step to make nearly perfect TiO2-terminated STO. The SrO reacts
with H2O and CO2 at room temperature, which causes the topmost layer of
SrO-terminated domains to form a Sr-hydroxide complex, while the chemically very stable TiO2does not react with water. This Sr-hydroxide complex dissolves
much faster in an acidic solution than the TiO2[33].
HF etching
After the sonication in Type I water we transfer the STO to a beaker with a stand-ard commercially available buffered HF (BHF) solution. For the BHF we use a beaker made of polytetrafluoroethylene (PTFE) and we transfer the substrate into and out of the BHF with tweezers made of Ethylene
ChloroTriFluoroEthyl-ene (ECTFE). The BHF is a buffered HF solution (12.5 % HF and 87.5 % NH4F)
bought from J.T.Baker.
The BHF etches the Sr-hydroxide complex much faster than the TiO2allowing
us to obtain single TiO2-termination with the correct etching duration. After 30 s
we take the substrate out, dip it and move it around in two consecutive beakers with Type I water to wash away remaining BHF and put it in a third beaker with Type I water.
Cleaning the STO (after etching)
After the HF etching remnants of the BHF remain on the substrate. We clean the substrate again by rinsing it with isopropanol, sliding it upside down over lens paper soaked in isopropanol while applying light pressure, rinsing it with
3.1 Substrate termination 19
isopropanol again and blowing it dry with nitrogen gas. This removes large contaminants. Note that the substrate has been immersed in liquid or been wet since it is put in ethanol for sonication. We examine the substrate with an optical microscope and repeat if necessary.
Figure 3.1:Optical microscope image with the contrast increased of a STO substrate after
HF etching and cleaning. The substrate is very clean with only very few contaminants sitting near the edges of the substrate.
When the substrate is sufficiently clean the surface ideally is completely free from contaminants visible with the optical microscope, but a handful of small contaminants near the edges of the substrate is acceptable. At this moment we save optical microscope images for reference. Figure 3.1 shows an example of a substrate that has only a handful of very small contaminants of the surface after HF etching and cleaning.
20 Experimental methods
Annealing STO
To facilitate recrystallization and further remove remnants of the HF etching the STO is annealed at 980◦C for 70 min. The substrates are placed on a quartz boat in a quartz tube through the oven and we let 100 sccm to 150 sccm of 5N oxygen flow through this tube. The temperature is raised at a rate of 16◦C·min−1and after the annealing the temperature setpoint is lowered at the same rate. The
temperature drops to≈50◦C in roughly two hours by opening the oven when
the temperature reaches 400◦C. Afterwards the oxygen flow is stopped and the substrates taken out of the oven.
We process three substrates simultaneously and we have checked the cleanliness of one batch of three substrates after annealing with the optical microscope. The substrates were very clean before the annealing and were also very clean after annealing.
AFM STO
The last step of the substrate processing is to image the surface of the substrate with an Atomic Force Microscope (AFM). If the terraces are clearly visible, (nearly) straight, clean and have unit cell step height then the substrate is good and is used for sputter deposition. A description of imaging surfaces with the AFM and an example of AFM images of a substrate and the terrace step height can be found in section 3.3.
3.2 Thin film deposition 21
3.2
Thin film deposition
In this section we describe the materials, equipment, machines and methods we use to deposit LAO on top of the TiO2-terminated STO. The goal of sputtering
the LAO on the STO is to determine if some oxygen in the sputter gas affects the properties of the conducting interface.
3.2.1
90° off-axis sputter machine
We start with a description of the 90° off-axis sputter machine (hereafter just called sputter machine) we use and its components. A schematic view of the sputter machine and peripherals can be seen in figure 3.2.
Figure 3.2:Schematic view of the 90° off-axis sputter machine including pumps, gauges
and computer controlled operation. Image reprinted with permission from C. Yin.
The sputter machine has one vacuum chamber with a top cover that can be hoisted up to allow a sample to be inserted or taken out. To create the vacuum a TPU 261 PC turbomolecular pump from Pfeiffer Vacuum with a pumping speed of 170 L·s−1for N2is used backed by a Duo 10 M rotary vane pump from Pfeiffer
Vacuum with a maximum pumping speed of 12 m3·h−1to create a fore-vacuum. Between the turbo pump and the vacuum chamber a butterfly valve (BFV) from VAT is situated that can be opened in steps of 0.1%.
The sputter machine is equipped with three pressure gauges: a Pirani with a detection range from 5×10−4Torr to atmospheric pressure, a Baratron ab-solute pressure transducer from MKS Instruments with a detection range of
22 Experimental methods
5×10−3mbar to 10 mbar and a hot-cathode Bayard–Alpert ionization gauge
(hereafter called UHV gauge) from Kurt J. Lesker Company with a detection range of 10−9mbar to 10−4mbar. The Baratron is an absolute pressure gauge, whereas the Pirani and UHV gauge are gas-sensitive gauges.
The substrate is glued to a heater with silver paste, which is heated by resistive heating. The temperature of the heater is measured with a type K thermocouple. A molybdenum shutter, which can be rotated away or in front of the heater, is placed underneath the heater to shield the substrate from the plasma.
Mass flow controllers from Brooks Instrument are used to control the flow of gas into the sputter chamber. Three gases can be input: argon, oxygen and nitrogen. The minimum of these gases is 0.5 sccm (standard cubic centimetre per minute), a lower value can’t be set in a controllable way. The maximum flow of argon is 200 sccm and of oxygen 100 sccm.
The sputtering target is a 2 inch single crystal LAO wafer from MTI Corporation. The distance between the surface of the target and the axis of the heater is 45 mm and the distance between the surface of the heater and the axis of the target is 75 mm. Note that the correct total pressure for deposition depends on the working distance. The sputter machine has four targets in total and all targets have a molybdenum shutter, which can be rotated away or in front of the target, to shield the target from the plasma of another target.
To create the plasma a radio frequency (RF) sputter controller and impedance matching unit from Advanced Energy is used.
3.2.2
Sputter deposition procedure
In this subsection we describe the parameters and methods we use to sputter LAO on top of the TiO2-terminated STO.
Preparations for sputtering
Before inserting a substrate into the sputter machine we check the cleanliness of the substrate with the optical microscope and blow away any dirt with nitrogen gas.
To insert a substrate we take out the heater, clean the heater and utensils with isopropanol on a wipe and blow them dry with nitrogen gas. Then the substrate is glued to the heater with a drop of silver paste on a wooden stick with a flat end. The silver paste provides good heat conductance from the heater to the
3.2 Thin film deposition 23
substrate. Finally the substrate is pressed on the heater by applying some force to opposite corners of the substrate with tweezers tips.
After inserting the heater back in the sputter machine and pumping it down, the silver paste is baked to harden it. The silver paste bake-out consists of ramping to 200◦C with 5◦C·min−1, dwelling at 200◦C for 1 h and ramping back to 20◦C with 5◦C·min−1.
Setting temperature, gas flow and pressure
For sputtering in oxygen and argon the heating program consists of ramping up to the desired temperature with 25◦C·min−1, dwelling at that temperature
for the pre-sputtering and sputtering and then ramping back to 20◦C with
10◦C·min−1.
For sputtering in argon the heating program includes annealing. The program is as follows: ramp up to 800◦C with 25◦C·min−1, dwell at 800◦C for the pre-sputtering and pre-sputtering, ramp down to 600◦C with 10◦C·min−1, dwell for 1 h and then ramp back to 20◦C with 10◦C·min−1.
When the temperature has almost reached the desired value, we close the BFV or set it to a low setpoint and start the gas flow into the sputter chamber. The mass flow controller adjusts the gas flow quickly and the new flow is reached within a second. If oxygen is used, nitrogen gas is let into the turbo pump directly to protect it from oxidation.
We use two methods to control the total pressure: pressure control and position control. For pressure control a calibration has been made of the correct combin-ation of flow and CPA setpoint for a number of different pressures. After the gas flow(s) have been started the desired setpoint is set in the CPA and then the software controls the setpoint of the BFV with a feedback mechanism to attain the desired pressure. We use this method at first and for the control samples. Position control does not have a feedback mechanism. It has a wider range of possible combinations of flow and pressure because the feedback mechanism leads to a heavily oscillating BFV if the flow is too high. With position control the gas flow(s) are started and the BFV is set to a fixed position. After a couple of minutes the pressure is stabilized. This method requires a separate calibration to find out which BFV positions give the desired pressures for different total flows. Note that the BFV setpoint depends on the temperature of the heater and also can be slightly different between samples. Also note that the minimum step of the BFV is 0.1 % and at pressures of about 5.0×10−2mbar this BFV step gives a
24 Experimental methods
pressure step of≈0.4×10−2mbar. In order to exactly attain the desired pressure the flow might need to be adjusted.
Oxygen partial pressure
The Pirani gauge is only used to measure rough vacuum when we vent or pump down the sputter machine. The Baratron gauge is used to measure the total pressure for sputtering. The UHV gauge is used to measure the base pressure after a night (or longer) of pumping down before sputtering a sample. The base pressure is a measure of how clean the sputter chamber and the substrate are. The only pressure gauge that is an absolute pressure gauge is the Baratron. The oxygen partial pressures we are use are in the 10−4mbar to 10−3mbar regime, which is below the detection range of the Baratron. We calculate the oxygen partial pressure by multiplying the total pressure measured with the Baratron by the fraction of the oxygen flow of the total flow.
The pressure is a result of the continuous flow of gas into the sputter chamber and the continuous pumping of gas out of the chamber. However the pumping speed of a turbo pump depends on the mass of the molecule. The mass of argon is 40 u and the mass of oxygen is 32 u, which means that the pumping speed for oxygen is≈5 % higher than for argon*. We ignore this small difference in
pumping speed and assume the pressure ratio is the same as the flow ratio.
Pre-sputtering and sputtering
When the temperature, flow(s) and pressure are all set, we close the substrate shutter and make sure the LAO shutter is open. Then we turn on the RF sputter controller and the impedance matching unit, turn on the sputtering, look at the plasma to see if it looks normal, and pre-sputter for at least 15 min. We pre-sputter to make sure that any contaminants on the target are removed and that the plasma is stable.
After the pre-sputtering we open the substrate shutter, start a timer and set the remaining dwelling time for the heater to the remaining sputter time. After the desired sputter time we turn the sputtering off and the heating program starts to ramp down immediately.
When we sputter in argon and oxygen and use a fixed BFV setpoint we stop the argon flow after the sputtering is finished, but keep the oxygen flow and BVF
setpoint constant. When the temperature has dropped below 70◦C we stop the
*Working with Turbopumps, Introduction to high and ultra high vacuum production, Pfeiffer
3.2 Thin film deposition 25
oxygen flow and open the BFV completely. The sample is cooled down in the same oxygen partial pressure as it is sputtered in until the temperature is low enough that the oxygen is not mobile anymore.
When sputtering in pure argon we stop the argon flow after the sputtering is finished and open the BFV to pump away the argon. When the temperature has reached 600◦C we start the annealing.
Annealing LAO
To anneal a sample we close the BFV when the temperature has reached 600◦C, input 40 sccm oxygen and set the pressure in the CPA at 1 mbar. Then we anneal the sample for 1 h. Afterwards the heating program starts to ramp down and we stop the oxygen flow and open the BFV completely when the temperature has
dropped below 70◦C. The sample is cooled down in the same oxygen partial
pressure as it is sputtered in until the temperature is low enough that the oxygen is not mobile anymore.
26 Experimental methods
3.3
Sample characterization
In this section we describe the equipment, machines and methods we use to characterize the LAO/STO heterostructures by Atomic Force Microscope (AFM) imaging and resistance measurement.
Atomic Force Microscopy
To assess the structure of the surface of substrates and samples and to check for dirt too small to see with an optical microscope we use the Bruker MM-NS5 AFM. We use tapping mode in air and we use an OMCL-AC160TS cantilever bought from Olympus. It is made from silicon and has a nominal resonance frequency of 300 KHz and a nominal spring constant of 26 N·m−1.
An AFM has a cantilever hanging over a sample with a very sharp tip at the free end with a typical tip radius of 10 nm to 20 nm. The cantilever is driven to oscillate with a frequency near its resonance frequency. The tip oscillates between actually touching the surface and no contact at a distance of 50 nm to 100 nm from the surface. Proximity to the surface of the sample leads to interaction forces like Van der Waals forces and dipole-dipole interactions. The interaction forces between sample and tip changes the amplitude of the oscillation, which is kept constant with a feedback mechanism. The interaction forces also depend on the material the surface is made of. Different materials give a difference in phase, but height differences also give phase differences. The amplitude and the phase of the tip with respect to the driving signal are measured in addition to the height of the surface.
We use Gwyddion to process AFM images. AFM images can be levelled using different algorithms. The first algorithm we use is aligning rows, which can be done with e.g. a polynomial function, and the second algorithm is fitting a plane through three points. These points are circles with a certain radius and the averages of the circles are set to zero height, which levels the height of the entire image by the plane they define. If necessary data is tilted manually to make the terraces horizontal. We also make line scans and save the height profiles. The surface roughness of samples was determined with Gwyddion. We define
the average surface roughness Sa of a 1 µm×1 µm area levelled by aligning
rows as the roughness of a sample.
AFM on STO
We image the STO substrates with the AFM to determine their quality and usability for thin film deposition. An area of 1 µm×1 µm is imaged to determine
3.3 Sample characterization 27
whether the terraces are flat and straight and a larger area of 4 µm×4 µm is imaged to look at larger features and to more easily see surface defects and anomalies in the terraces on a larger scale. The larger area is scanned at a different spot than the smaller image, shifted by 3 µm.
400 nm 913 pm 0 200 300 400 500 600 700 800 Height a 400 nm 6.0 nm 0.0 1.0 2.0 3.0 4.0 5.0 Height b 400 nm 1 6.0 nm 0.0 1.0 2.0 3.0 4.0 5.0 Height c 0 , 0 0 , 2 0 , 4 0 , 6 0 , 8 0 1 2 3 4 5
H e i g h t p r o f i l e
H e ig h t (n m ) P o s i t i o n ( µ m ) dFigure 3.3:AFM images of TiO2-terminated STO, where the colour represents the height
according to the scale bar. a,b Height images levelled by (a) aligning rows with a polynomial function and (b) setting the plane defined by the indicated points to equal height. c The line scan from right to left with d the height profile along this line.
Figure 3.3 gives an example of good quality TiO2-terminated STO. Figure 3.3a is
levelled by aligning rows with a polynomial function. Figure 3.3b is levelled by fitting a plane through three points making the terraces flat. The AFM images are 512 x 512 pixels and the indicated points are circles with a 10 pixels radius. Figure 3.3c shows the 10 pixels wide line scan that is used to determine terrace
28 Experimental methods
smoothness and the step height. Figure 3.3d is the height profile of the line in figure 3.3c, which is an average over the width of 10 pixels.
The right half of the plateau around 1 nm in figure 3.3d has an average height of 0.992 nm and the left half of the next plateau around 1.4 nm has an average height of 1.395 nm. The step height is 0.403 nm. The same procedure gives a step height of 0.384 nm between the terraces at 4.3 nm and 4.7 nm. Both values are close to the theoretical value for STO of 0.3905 nm. We look at only one step at a time and do not average over multiple steps and also take only half of a terrace to average over in order to look only at the step height and not at any residual tilt in the data.
AFM on LAO
After depositing the LAO the sample is taken out of the sputter machine and checked for cleanliness with the optical microscope. Then the sample is im-aged with the AFM to determine the surface morphology. A smaller area of 1 µm×1 µm is imaged to look closely at the surface and the terraces and a larger area of 4 µm×4 µm is imaged to look at larger features and to see whether the features of the small area are representative for the sample. Subsequent AFM images of the same sample are shifted by 3 µm.
Resistance measurement
We use a Fluke 115 True RMS multimeter to test whether the sample is conducting or insulating. We test all six combinations of corners for conductivity by pressing the sharp multimeter probe tips against the corner of the sample from above. The LAO itself is an insulator, but the metal tips of the multimeter probes penetrate the LAO layer that is only a few nanometre thick thereby contacting the interface and measuring its resistance. The highest resistance the multimeter can measure, the measurement limit, is 40 MΩ.
If the resistance of a sample is above the measurement limit of the multimeter and the surface of the sample is of good quality, a quality that in principle allows the formation of a conducting interface, then we wirebond the sample, i.e. we locally melt an aluminium wire with an ultrasonic pulse while pressing one end against the top surface of the sample and repeat this for the other end while pressing it against a contact pad of a device the sample is glued on. The wire and contact pad give a better connection to the sample, where the wire penetrates the top LAO layer to connect to the interface because the wire is pressed against the sample with great force.
Chapter
4
Results
4.1
Total pressure
In this section we show the results of our first attempts to make a conducting interface by 90° off-axis sputtering with oxygen in the sputter gas and the results of determining the total pressure that gives the best surface morphology.
Parameters that are relevant for sputter deposition are the surface roughness (Sa), the total pressure (P), the oxygen partial pressure (PO2), the oxygen pressure as
a percentage of the total pressure (%O2), the argon flow (QAr), the oxygen flow
(QO2), the RF sputter power (FP), the sample temperature (T) and the growth
time (t). Each section contains a table with the deposition parameters of the samples discussed in that section, while appendix A gives an overview of the deposition parameters of all samples.
Yin et al. [10] have shown that a conducting interface can be made in a smooth film of LAO/STO by 90° off-axis sputtering in an argon environment. This was done with a pressure of 0.04 mbar to 0.08 mbar, an argon flow of 16 sccm
to 20 sccm, 50 W RF sputter power and a sample temperature of 800◦C. After
sputtering the samples were annealed in 1 mbar oxygen for 1 h to remove the oxygen vacancies induced in the STO substrate by sputtering at high temperature in a reducing atmosphere.
The stoichiometry is controlled by the pressure: Al-rich samples grown at lower pressures give a conducting interface and increasing the pressure increases the La/Al ratio up to an insulating stoichiometric sample at 0.10 mbar, which is in agreement with refs. [9, 34].
30 Results
We use the same sputter machine as in ref. [10] to make our samples. The deposition rate depends on the total pressure: 4.27 ˚A·min−1 at 0.04 mbar and 3.20 ˚A·min−1 at 0.10 mbar total pressure [10]. The samples we grow at higher total pressures are sputtered for longer in order to give all samples approximately the same thickness of 15 unit cells (≈5.6 nm).
Table 4.1:LAO sputter deposition parameters.
Sample Sa P PO2 %O2 QAr QO2 FP T t
nm mbar mbar sccm sccm W ◦C min
LAO03 0.21 4.0×10−2 1.3×10−3 3.1 15.5 0.5 50 800 13.5 LAO05 0.42 8.0×10−2 2.0×10−3 2.5 19.5 0.5 50 800 15 LAO04 0.23 1.0×10−2 4.2×10−4 4.2 11.5 0.5 50 800 15 LAO06 0.29 1.8×10−1 4.5×10−3 2.5 19.5 0.5 50 800 25 LAO07 0.80 1.4×10−1 1.8×10−3 1.3 39.5 0.5 50 800 30 400 nm 2.03 nm 0.24 0.60 0.80 1.00 1.20 1.40 1.60 1.80 Height a 400 nm 2.82 nm 0.29 1.00 1.50 2.00 2.50 Height b
Figure 4.1: Effect of a small amount of oxygen. AFM images of samples LAO03 (a) and
LAO05 (b), where the colour represents the height according to the scale bar next to the image.
We start by replicating the method in ref. [10] and replacing a bit of argon flow by
oxygen flow. We use the lower (4.0×10−2mbar, sample LAO03) and upper end
(8.0×10−2mbar, sample LAO05) of the pressure regime that gives a conducting interface. The smallest flow of oxygen possible in the experimental set-up, which is 0.5 sccm, is used to substitute for the same amount of argon flow. The deposition parameters of these samples can be found in table 4.1. Figure 4.1 shows the AFM images of samples LAO03 and LAO05. Both samples have a grainy surface and are insulating. A small amount of oxygen in the sputter
4.1 Total pressure 31
gas already has a significant effect on the surface morphology and transport properties of the LAO/STO samples.
400 nm 2.13 nm 0.00 0.50 1.00 1.50 Height a 400 nm 0.3 deg −46.1 −35.0 −30.0 −25.0 −20.0 −15.0 −10.0 −5.0 Phase b 400 nm 3.19 nm 0.00 0.50 1.00 1.50 2.00 2.50 Height c 400 nm 4.93 nm 0.61 1.50 2.00 2.50 3.00 3.50 4.00 4.50 Height d
Figure 4.2: Effect of lower and higher total pressure.AFM images of
samples LAO04 (a,b, 1.0×10−2mbar), LAO06 (c, 1.8×10−1mbar) and LAO07 (d, 1.4×10−1mbar).
In order to find out whether the total pressure has the same effect on the smooth-ness of the film as when LAO is sputtered in pure argon we make samples with lower and higher total pressure. Reducing the total pressure also decreases the oxygen partial pressure, but increases the oxygen percentage. Higher total pressures also allow us to lower the oxygen percentage to see if a lower oxygen percentage gives smoother films. Sample LAO07 is grown using a higher argon flow lowering the oxygen partial pressure and oxygen percentage even further.
32 Results
Samples LAO04, LAO06 and LAO07 are grown with total pressures of
1.0×10−2mbar, 1.8×10−1mbar and 1.4×10−1mbar respectively. Their depos-ition parameters can be found in table 4.1. Figure 4.2 shows the AFM images of samples LAO04, LAO06 and LAO07. Sample LAO04 made with a lower total pressure is not smooth, but has holes in the surface. Both samples grown with a higher total pressure have a grainy surface. All three samples are insulating.
4.2 Oxygen partial pressure 33
4.2
Oxygen partial pressure
In this section we describe the results of our efforts to get the oxygen partial pressure as low as possible to reduce the effect the oxygen in the sputter gas has on the surface morphology and transport properties of the samples. This is accomplished by increasing the argon flow far beyond the flow used in ref. [10] and setting the pressure by position control instead of pressure control (see subsection 3.2.2). The oxygen flow used is 0.5 sccm for all samples (except for the control samples grown in pure argon).
Table 4.2:LAO sputter deposition parameters.
Sample Sa P PO2 %O2 QAr QO2 FP T t
nm mbar mbar sccm sccm W ◦C min
LAO08 0.24 7.7×10−2 4.8×10−4 0.63 79.5 0.5 50 800 15 LAO13 0.10 4.9×10−2 2.5×10−4 0.50 99.5 0.5 50 800 15 LAO14 0.14 4.9×10−2 2.0×10−4 0.42 119.5 0.5 50 800 15 LAO15 0.13 5.0×10−2 1.6×10−4 0.31 159.5 0.5 50 800 15 LAO24 0.29 5.0×10−2 2.5×10−4 0.50 99.5 0.5 50 800 15 400 nm 2.76 nm 0.00 0.50 1.00 1.50 2.00 2.50 Height a 400 nm 1.20 nm 0.00 0.20 0.40 0.60 0.80 1.00 Height b
Figure 4.3: Effect of a higher argon flow.AFM images of samples LAO08 (a, 80 sccm)
and LAO13 (b, 100 sccm).
The oxygen flow can not be lowered in the current experimental set-up. In order to decrease the oxygen partial pressure the argon flow is increased while the total pressure is kept constant. The total pressure of sample LAO08 is in the range that gives a conducting interface when LAO is sputtered in pure argon, but only after this sample we settled at 5×10−2mbar as a good standard for the
34 Results
total pressure. 5×10−2mbar is closer to the lower end of the pressure regime that gives a conducting interface and allows the oxygen partial pressure to be lower compared to when using a higher total pressure.
Samples LAO08, the first sample made with position control, and LAO13 are grown with oxygen partial pressures of 4.8×10−4mbar and 2.5×10−4mbar and total flows of 80 sccm and 100 sccm respectively. Their deposition parameters can be found in table 4.2. Figure 4.3 shows the AFM images of samples LAO08 and LAO13. Sample LAO08 has a grainy surface with a roughness of 0.24 nm and sample LAO13 has terraces with what could be small islands on top and a roughness of 0.10 nm. Both samples are insulating.
400 nm 3.14 nm 2.07 2.40 2.60 2.80 3.00 Height a 400 nm 1.21 nm 0.15 0.40 0.60 0.80 1.00 Height b
Figure 4.4: Effect of an even higher argon flow.AFM images of samples
LAO14 (a, 120 sccm) and LAO15 (b, 160 sccm).
After the first sample that was not grainy, the argon flow is increased even more. Samples LAO14 and LAO15 are grown with oxygen partial pressures of 2.0×10−4mbar and 1.6×10−4mbar and total flows of 120 sccm and 160 sccm respectively. Their deposition parameters can be found in table 4.2. Figure 4.4 shows the AFM images of samples LAO14 and LAO15. Sample LAO14 has a grainy surface and sample LAO15 looks to have a lot of holes in its surface. In-creasing the flow beyond 100 sccm makes the surface morphology and roughness worse. Both samples are insulating.
Figure 4.5 shows the relation between roughness and oxygen partial pressure for two different total pressures while keeping all other parameters constant. The black squares of samples made with 0.05 mbar total pressure show a minimum in the roughness. At both total pressures the roughness increases when the oxygen partial pressure increases beyond the minimum.
4.2 Oxygen partial pressure 35 0 , 0 5 , 0 x 1 0 - 4 1 , 0 x 1 0 - 3 1 , 5 x 1 0 - 3 2 , 0 x 1 0 - 3 0 , 0 0 , 1 0 , 2 0 , 3 0 , 4 R o u g h n e s s v s . o x y g e n p a r t i a l p r e s s u r e R o u g h n e s s ( n m ) O x y g e n p a r t i a l p r e s s u r e ( m b a r ) 0 . 0 5 m b a r 0 . 0 8 m b a r
Figure 4.5: Effect of the oxygen partial pressure on surface roughness.Surface
rough-ness as a function of oxygen partial pressure for two different total pressures.
A second sample is grown with the oxygen partial pressure and total flow that produced the sample with the best surface, but this sample turns out to be different than the one grown in the same conditions. Sample LAO24 is grown with an oxygen partial pressure of 2.5×10−4mbar and a total flow of 100 sccm. Its deposition parameters can be found in table 4.2. Figure 4.6 shows the AFM images of sample LAO24. The terraces are covered with small islands and the sample is insulating. Sputtering samples with oxygen is not very reproducible.
36 Results 400 nm 2.46 nm 0.00 0.50 1.00 1.50 2.00 Height a 400 nm 36.8 deg 22.5 26.0 28.0 30.0 32.0 34.0 Phase b
4.3 Sputter power and temperature 37
4.3
Sputter power and temperature
In this section we describe the results of varying two important parameters in the sputtering process: the sputter power that attracts the gas ions to the target and the temperature of the sample, which gives the species arriving at the sample the needed energy and mobility on the surface to form a smooth thin film.
Table 4.3:LAO sputter deposition parameters.
Sample Sa P PO2 %O2 QAr QO2 FP T t
nm mbar mbar sccm sccm W ◦C min
LAO10 0.26 6.2×10−2 3.9×10−4 0.63 79.5 0.5 50 850 15
LAO18 0.14 4.9×10−2 3.1×10−4 0.63 79.5 0.5 50 750 15
LAO16 0.38 4.9×10−2 2.5×10−4 0.50 99.5 0.5 30 800 25
LAO17 0.33 5.1×10−2 2.6×10−4 0.50 99.5 0.5 80 800 9.5
In finding the best total pressure and oxygen partial pressure we used the same sputter power and deposition temperature as in ref. [10]. Having found the best pressures we vary the sample temperature: a sample with higher temperature and one with lower temperature is made.
Samples LAO10 and LAO18 are grown with sample temperatures of 850◦C
and 750◦C respectively. Their deposition parameters can be found in table 4.3. Figure 4.7 shows the AFM images of samples LAO10 and LAO18. Sample LAO10 has a grainy surface with a strange pattern, which makes it hard to say something about surface composition. Sample LAO18 has a lot of very small islands. The height difference between the islands and the terraces is quite small, below 1 nm, and the islands are also visible in the phase image with a small positive phase difference of about 2°. The phase difference is quite small as well, which means that the islands are probably made of LAO. Changing the sample temperature makes the surface morphology worse. Both samples are insulating.
Next the sputter power is varied. We make a sample with higher power and one with lower power. The sputter time is decreased and increased accordingly to give these samples the same approximate thickness as the other samples. Samples LAO16 and LAO17 are grown with sputter powers of 30 W and 80 W respectively. Their deposition parameters can be found in table 4.3. Figure 4.8 shows the AFM images of samples LAO16 and LAO17. Sample LAO16 has a lot of islands on the terraces that are significantly larger than the islands of the sample grown at lower temperature. The terraces are visible in the height image mostly because there are no islands at the steps between terraces. These islands
38 Results 400 nm 3.35 nm 0.00 1.00 1.50 2.00 2.50 3.00 Height a 400 nm 17.1 deg 12.0 13.0 14.0 15.0 16.0 Phase b 400 nm 1.34 nm 0.22 0.40 0.60 0.80 1.00 1.20 Height c 400 nm 9.49 deg 3.22 5.00 6.00 7.00 8.00 Phase d
Figure 4.7: Effect of higher and lower sample temperature. AFM images of samples
LAO10 (a,b, 850◦C) and LAO18 (c,d, 750◦C).
are also visible in the phase image with a larger phase difference of 5° to 8°. The height and phase differences of these islands are larger, which makes it difficult to draw a conclusion about the nature of the islands. It is a possibility that the islands consist of a material other than LAO. The surface of sample LAO17 is completely covered by islands or grains and at some places the terraces are not visible because they are covered by larger features. The homogeneity of the phase image indicates that the surface probably consists of LAO. Changing the sputter power makes the surface morphology worse. Both samples are insulating.
4.3 Sputter power and temperature 39 400 nm 2.89 nm 0.00 0.50 1.00 1.50 2.00 2.50 Height a 400 nm 143 deg 128 132 134 136 138 140 Phase b 400 nm 3.03 nm 0.00 0.50 1.00 1.50 2.00 2.50 Height c 400 nm 5.6 deg −2.5 −1.0 0.0 1.0 2.0 3.0 4.0 Phase d
Figure 4.8: Effect of lower and higher sputter power.AFM images of samples LAO16
40 Results
4.4
Control measurements
In this section we describe the results of two different control measurements. In order to compare the insulating samples grown with some oxygen in the sputter gas to samples grown in only argon we make a control sample with a low argon flow and a control sample with an argon flow of 100 sccm.
Table 4.4:LAO sputter deposition parameters.
Sample Sa P PO2 %O2 QAr QO2 FP T t
nm mbar mbar sccm sccm W ◦C min
LAO22 0.07 5.0×10−2 16 50 800 14 LAO23 0.11 4.9×10−2 100 50 800 14 400 nm 5.1 nm 0.0 1.0 2.0 3.0 4.0 Height a 0 , 0 0 , 2 0 , 4 0 , 6 0 , 8 1 2 3 4 5
H e i g h t p r o f i l e
H e ig h t (n m ) P o s i t i o n ( µ m ) bFigure 4.9: aAFM image of sample LAO22. b Height profile of the line in a from left to
right.
Sample LAO22 is the first control sample. It is grown in only argon using pressure control with the same flow as in ref. [10]. Its deposition parameters can be found in table 4.4. Note that the control samples are annealed in oxygen after sputter deposition to remove the oxygen vacancies induced in the STO by sputtering at high temperature in a reducing atmosphere. Figure 4.9a shows an AFM image of sample LAO22 with a 10 pixels wide line scan perpendicular to the terraces and figure 4.9b shows the height profile of the line in figure 4.9a, which is an average over the width of 10 pixels. The sample has flat terraces, a roughness of 0.07 nm and is conducting.
Sample LAO23 is the second control sample. It is grown in only argon using position control with the same total flow as sample LAO13, the sample that has
4.4 Control measurements 41 400 nm 7.5 nm 0.0 2.0 3.0 4.0 5.0 6.0 Height a b
Figure 4.10: aAFM image of sample LAO23. b Multimeter resistance measurement.
the lowest surface roughness. Its deposition parameters can be found in table 4.4. Figure 4.10a shows an AFM image of sample LAO23 and figure 4.10b an image of measuring the resistance with the multimeter. The surface is slightly rougher than that of the first control sample, but is almost smooth with a roughness of 0.11 nm, comparable to the roughness of sample LAO13. Sample LAO23 is conducting. Pressing the multimeter contacts against the sample gave a resistance of the order of 9 MΩ at room temperature.
At a certain moment we switched from the pressure control method to the posi-tion control method. This enabled us to reduce the oxygen partial pressure while keeping the total pressure constant by increasing the total flow. Figure 4.5 shows that for 2.5×10−4mbar to 2×10−3mbar decreasing the oxygen partial pressure decreases the roughness. This effect is visible at two different total pressures while keeping the other growth parameters constant. The samples grown at
0.05 mbar total pressure show a minimum in roughness at 2.5×10−4mbar. We
believe this is due to the effect of the increased flow necessary to reduce the oxygen partial pressure. Beyond the optimum the roughness increases again as the effect of the increase in flow is larger than the effect of the decrease in oxygen partial pressure.
The control samples also indicate the effect the higher flow has. With sputtering in argon an increase of flow from 16 sccm to 100 sccm increases surface roughness and makes the terraces be not completely flat. Figure 4.11 shows that increasing the flow increases the roughness for the control samples (red circles) and that increasing the flow beyond 100 sccm increases the roughness for samples with oxygen in the sputter gas and a total pressure of 0.05 mbar (black squares). It is
42 Results 0 2 0 4 0 6 0 8 0 1 0 0 1 2 0 1 4 0 1 6 0 0 , 0 0 , 1 0 , 2 R o u g h n e s s v s . f l o w R o u g h n e s s ( n m ) F l o w ( s c c m ) O x y g e n A r g o n
Figure 4.11: Effect of the flow on surface roughness. Surface roughness as a function
of gas flow with and without oxygen in the sputter gas.
possible that too high a flow affects the plasma or the flow of species from the target to the sample.
Chapter
5
Discussion
In this chapter we discuss the results of our research and the experimental details and limitations of our research. We also present a possible explanation of the effect the oxygen in the sputter gas has.
We have used the calibration of ref. [10] for the deposition rate. We have not measured the LAO thickness of our samples. The LAO layer needs to be at least four unit cells thick in order to create a conducting interface in LAO/STO [22]. All samples have a LAO thickness between 10 and 20 unit cells, a regime in which the interfacial conductivity shows little dependence on thickness [34]. AFM measurements measure height, but other information like the phase and amplitude of the tip relative to the drive signal are also recorded. The exact composition of the surface can not be determined with AFM measurements, but AFM images can be used for qualitative statements about surface composition. The phase of the tip relative to the drive signal is different for different materials, but a height difference also leads to a phase difference. A significant phase difference with a very small (< 1 nm) height difference is a strong indication that the surface consists of regions of different materials.
No quantitative measurements have been performed to assess the composition of the sample surface as the goal of our research was to determine whether a conducting interface could be created in LAO/STO by introducing some oxygen in the sputter gas and not to determine what happens precisely when sputtering in oxygen and argon. Instruments to measure the sample composition such as X-ray photoelectron spectroscopy and Rutherford backscattering spectroscopy are not available at Leiden University and using such instruments at other universities did not fit in the scope or allotted time for this research project.
44 Discussion
Growth parameters
The first objective was to make samples with a smooth surface grown in an epi-taxial manner. The parameters that produce the sample with the lowest surface roughness and no grains are a total pressure of 5.0×10−2mbar, an oxygen par-tial pressure of 2.5×10−4mbar, a sputter power of 50 W, a sample temperature of 800◦C and a growth time of approximately 15 min. These parameters are the same as when samples are grown in a pure argon environment. It is sufficient to show that a total pressure of 5.0×10−2mbar, which falls in the pressure range of 4×10−2mbar to 8×10−2mbar of ref. [10], is the total pressure that produces the best surface. The exact range of total pressures that give smooth-like terraces without grains is not determined.
The higher flow with respect to ref. [10] not only increases surface roughness, but also causes an experimental problem. Inputting this much gas and keeping the BFV opened at a constant position puts a severe load on the turbo pump. The current the turbo pump uses comes close to its maximum when using a total flow of 160 sccm, but already with a total flow of 100 sccm the turbo pump becomes warm after some time. This strain on the turbo pump and its heating up reduces the life expectancy of the turbo pump. Decreasing the oxygen partial pressure below the found optimum of 2.5×10−4mbar cannot be achieved in the current experimental set-up by increasing the argon to oxygen flow ratio.
Comparison with pure argon sputter gas
The optimised parameters for oxygen in the sputter gas have the same total pres-sure, temperature, sputter power and growth time as used to make a conducting interface in pure argon. The two differences are oxygen added to the sputter gas and the higher flow needed to keep the oxygen partial pressure low. So we made a control sample just the same as in ref. [10] to see if our sample fabrication process was done correctly and can create a conducting interface to begin with. Control sample LAO22 shows that our substrate preparation and sample growth can generate a conducting interface.
The second control sample is to eliminate the higher flow as a difference. Sample LAO23 grown with an argon flow of 100 sccm is conducting even though the surface is rougher and the terraces not completely sharp. So now the conducting sample with high argon flow can be compared to the insulating sample grown with the same deposition parameters save that a small part of that flow is oxygen instead of argon. This shows the effect of introducing oxygen in the sputter gas.
45
The effect of oxygen
The minimum flow the mass flow controllers can input in a controllable way is 0.5 sccm. As discussed before decreasing the oxygen partial pressure by further increasing the argon flow does not work. If the effect of a lower oxygen partial pressure than used in this research is to be studied, one has to either change something in the experimental set-up or find a clever way to do so in the current set-up. One method that might work is using a mixture of argon and oxygen instead of pure oxygen. The oxygen partial pressure can be decreased with e.g. a factor of 10 by using a mixture of 10 % oxygen and 90 % argon. The issue with making such a mixture however is that it is very hard, maybe even impossible, to do this in a controllable manner.
A conducting interface in LAO/STO is most often made with PLD and can also be made with MBE and sputtering. In PLD pure oxygen is always used as gas. Early attempt to make a conducting interface in LAO/STO by (on-axis) sputtering used oxygen as sputter gas, but this resulted in insulating interfaces with the high La/Al ratio of 1.1. The parameter window necessary for a smooth surface gave a stoichiometry that did not produce a conducting interface. Most notably the pressure was an order of magnitude higher than with off-axis sputtering [11, 12]. Podkaminer et al. [13] were the first to make a conducting interface in LAO/STO by sputtering. They used 90° off-axis sputtering with argon as sputter gas. They claim that the sputtered single crystal LAO target provides a large enough
background pressure of O2and atomic oxygen, which creates an environment
similar to PLD. This raises the question what the partial pressures of O2 and
atomic oxygen are. We also have the same background gas coming from the target and add to that a small oxygen partial pressure. However it is unclear how much oxygen comes from the target and what oxygen partial pressure this effectively gives. Oxygen must go from the target to the sample or else only an alloy of lanthanum and aluminium would be grown with argon as sputter gas. The effect we find from a low oxygen partial pressures is an indication that the oxygen from the target gives a partial pressure that can not be very high.
We discussed the effect of oxygen on the surface morphology and transport properties, but the question remains how oxygen in the sputter gas affects the sample. A possible explanation for the effect of oxygen in the sputter gas is that the species oxidise before reaching the sample. Oxygen can react with the species before they arrive at the sample, whereas argon is a noble gas. La can react with oxygen to form LaO [35]. Orsel et al. [36] deposited LAO with PLD and measured the Al and AlO content of the plasma plume during propagation to a sample. They found that with an oxygen background Al oxidized at the
