Interface Conductivity in Oxide Perovskite Heterostructures

(1)

.

(2)

Interface Conductivity in Oxide Perovskite Heterostructures

Master’s Thesis

Mart Salverda Student number: s1916432

Daily supervisor:

dr. Saeedeh Farokhipoor Supervisor:

prof. dr. Beatriz Noheda Solid State Materials for Electronids Zernike Institute for Advanced Materials

University of Groningen

December 16, 2015

(3)

Introduction

The heterostructure of LaAlO3/SrTiO3, two band insulators, where the LaAlO3 film is deposited epitaxially on a TiO2-terminated (001)-oriented SrTiO3substrate, has a (quasi) 2-dimensional electron gas (2DEG) at its interface. In this chapter, the basics of this oxide perovskite heterostructure will be discussed, together with the possible origins of this 2DEG. After that, a short outline of this thesis is presented.

1.1 Perovskite Crystal Structure

Perovskites are a group of materials that have the perovskite crystal structure. The general chemical formula is ABX3, where A and B are two different cations and X is an anion. The ideal perovskite structure has a cubic unit cell with the A-cations in the center, the B-cations on the corners and the X-anions on all the mid-edge positions¹ (figure 1.1a) forming an octahedron around the B-cation (figure 1.1b).

(a) (b) (c)

Figure 1.1: Perovskite crystal structure (with X=O): (a) Perovskite unit cell; (b) multiple unit cells show the octahedra (of oxygen, in this case); (c) The crystal structure can also be regarded as alternating planes of AO and BO₂. Taken from [1].

In the case of oxide perovskites the X-anions are oxygen ions. Although the ideal perovskite structure is cubic, many perovskites are actually quasi-cubic, with a distorted unit cell [2]. Even the original Perovskite CaTiO3, of which the name of the structure is derived, is not cubic. Many interesting physical properties are related to these distortions, such as piezoelectricity and ferroelectricity [2]. However, in this thesis, these properties are not the main interest.

1The unit cell of ABX3 can also be described by placing the A-cations on the corners, the B-cations in the center and the X-anions on the faces. This is equivalent to the unit cell shown here.

(6)

1.2 Thin films

When a crystalline material is deposited epitaxially² onto another crystal, it can become strained, i.e.

the film’s crystal lattice is deformed with respect to its bulk lattice. This happens due to a stress in the lattice that comes from the mismatch of the bulk lattice parameters of the film material and the substrate material. If the two lattice parameters are similar in value (i.e. the mismatch is small) and the substrate surface is of sufficient quality, it is possible for the film to form a coherent interface with the substrate and therefore grow in a strained form[3], see figure 1.2b. In contrast, if the mismatch or the film thickness is large, it is unlikely that the film makes a coherent interface. More likely, defects (figure 1.2c) or domains are formed that allow for a relaxation of the film.

Figure 1.2: Epitaxy: (a) Regarding the substrate and film as bulk, before being in contact with each other; (b) Epitaxial and strained film, with a coherent (defect and domainless) interface; (c) Epitaxial, but partially relaxed film, with incoherent interface. Taken from [4].

Strain in a material can lead to enhanced properties[5][6]. For example, it can modify the phase transition temperature of a superconducting, ferromagnetic or ferroelectric material [3], such as in the perovskite PbTiO3[7].

Strain is not the only aspect of thin films that allows for new functionality: the mere stacking of a crystalline film on a crystalline and atomically flat substrate can already introduce new physics at the interface. For example, a ferromagnetic interface is formed at the interface of two antiferromagnetic materials[8] and a 2-dimensional electron gas (2DEG) is formed at the interface of two insulators[9].

1.3 The LaAlO

3

/SrTiO

3

2DEG

One of such systems in which a film on a substrate brings out a new physical property is the heterostructure of LaAlO3grown on a substrate of SrTiO3. In 2004, A. Ohtomo and H. Y. Hwang [9] published their results on this system, in which they reported a 2-dimensional electron gas (2DEG) and the interface of the two materials. Since the two separate bulk materials are insulating, this was a remarkable observation. Since this discovery, over a thousand articles have been published on this system or similar systems on possible explanations and potential applications. Now, eleven years later, much more is known about the origin of the 2DEG, but still no perfect and conclusive mechanism has been reported that can explain all observations.

1.3.1 Material Properties

SrTiO3

SrTiO₃ (STO) has a perovskite crystal structure, which is cubic at room temperature[10]. Its lattice parameter is 3.905˚A [11]. STO is a wide band gap semiconductor with an indirect band gap energy of 3.2 eV [12]. It is a popular substrate material, due to its lattice parameter being very similar to that

2If there is a clear relation between the atomic planes of a crystalline film and a crystalline substrate, the film is called epitaxial (in this thesis).

(7)

of some interesting materials, such as multiferroic BiFeO₃[13] and ferroelectric PbTiO₃ [6][14], but also because there is a well understood and reliable chemical treatment, reported in section 4.9, that provides a single terminated crystal surface, which is important for epitaxial growth of numerous films [15].

STO is also a material with many interesting properties: it can be driven ferroelectric by strain or doping, it can be turned metallic by doping (with lanthanum, niobium or oxygen vacancies[16]) and even superconducting at low temperatures for sufficient doping levels[17][16] or electric field strength[18].

LaAlO3

LaAlO₃(LAO) has a perovskite cyrstal structure, which is distorted to a rhombohedral unit cell at room temperature. Above 435^◦C it has a cubic unit cell[19]. The lattice parameter of the pseudo-cubic unit cell is 3.778˚A [11]. The rhombohedral unit cell leads to twin formation in single crystal substrates (see section 4.11). LAO is a wide band gap insulator with a gap energy of 5.6 eV [20]. LAO is used as a substrate for epitaxial growth of perovskite films, most notably for the growth of superconducting cuprate films[21].

1.3.2 The Origin of the 2DEG

When LAO is deposited on (001)-oriented STO epitaxially by Pulsed Laser Deposition (PLD, see section 2.1), a 2D-electron gas³ is formed at their interface [9]. This 2DEG is only formed if the substrate is terminated with TiO₂ (i.e. when the interface is TiO₂-LaO, figure 1.3b) not when it is SrO-terminated (SrO-AlO₂, figure 1.3c)[9]. And only after a layer of 4 unit cells has been deposited (so, above a critical thickness)[23][24]. Several potential explanations have been investigated, which will be discussed in this section.

Polar Catastrophe and Electronic Reconstruction

The first mechanism discussed here is the electronic reconstruction model. To avoid a scenario called Polar Catastrophe an electronic reconstruction takes place at the interface, generating a 2DEG. Already in 2004 this was the mechanism that Ohtomo and Hwang reported as a possibility for the existence of a 2DEG [9].

The (001) direction of a perovskite crystal can be regarded as alternating planes of AO and BO₂. In the ionic limit, STO consists of planes of SrO and TiO₂, which are both charge neutral (Sr²⁺, O²⁻, Ti⁴⁺).

In contrast, LAO is built up of planes of LaO and AlO₂, which have a +1 and -1 charge, respectively.

Although the ionic limit is not fully applicable in these systems[25], they do grant an easier understanding of the electronic reconstruction model.

When an LAO film is deposited on STO, the stacking goes from neutral layers (in the STO) to alternating +1 and -1 layers (in the LAO), see figure 1.1c. The electric potential for such a system would diverge, as shown in figure 1.4a. This diverging potential can be avoided by bringing half a charge per unit cell to the interface to screen the potential build-up⁴[25]. The potential then merely alternates and stays finite. So, if the first plane is LaO (figure 1.3a), which is positively charged, half an electron should go to the interface (per unit cell) to prevent the divergence, and the interface would contain free electrons (n-2DEG). On the other hand, if the first plane would be AlO2(figure 1.3b), which is negatively charged, it should be half a hole and the conduction would be carried out by holes (p-2DEG).

As mentioned before, only the n-2DEG is observed in experiments, i.e. only the TiO₂-LaO interface is conducting. The difference between the TiO₂-LaO and SrO-AlO₂interface can be explained in this model as follows: the former interface has to take up half an electron per unit cell. The valence of titanium can be 3+ and 4+, and since it is 4+ in STO, it can easily take up an extra electron and accomodate the electronic reconstruction. However, strontium and aluminium are not multivalent and cannot take up a hole or electron easily. Therefore, the polar catastrophe has to be avoided in a different way than by electronic reconstruction, for example by a structural reconstruction[26][25] .

3Actually, unlike in semiconductors like GaAs and Si, the electrons are confined to the Ti3d orbital bands and are therefore much stronger correlated. Therefore, one could also speak of an 2D electron liquid (2DEL)[22]

4This is not only true for interfaces, but also for polar surfaces.

(8)

(a) (b) Conducting (c) Insulating

Figure 1.3: Sheet resistance of LAO/STO grown at different oxygen growth pressures [9]: square= 10⁻⁴ torr, triangle= 10⁻⁵ torr. Values are in Ω/, not mΩ/ as is stated in the graph. This error was corrected by the authors after publication; Stacking of planes for a (b) TiO₂-terminated substrate and a (c) SrO-terminated substrate. Interface is marked (adapted from [1]).

(a) (b) (c)

Figure 1.4: (a) Electronic potential with and without reconstruction (taken from [27]); (b) Band diagram explaning formation of 2DEG and critical thickness; (c) adsorbates on the surface of the LAO film could lower the potential inside the film, while the 2DEG remains (taken from [28]).

The critical thickness is easily explained by this model: the potential build-up has to be great enough for electrons to tunnel to the interface. After four unit cells of LAO this is the case. Graphically this is easily understood in figure 1.4b, where the electrons can tunnel to the interface as soon as the valence band of LAO has more energy than the conduction band of STO at the interface.

What is special about this mechanism, is that it describes an intrinsic property of the interface: it does not depend on extrinsic factors such as extrinsic dopants, vacancies or other defects, which is true for silicon-based electronics.

However, this model does not seem to explain all observations that were made on this system. For example:

• LAO/STO also forms a 2DEG in the (011) direction. The planes in this direction are [ABO]⁴⁺

and [O₂]⁴⁻, regardless of the composition of the perovskite. Therefore, no polar discontinuity is occurring at the interface and no 2DEG is expected. Yet, this is observed [29][30];

• amorphous films on a neutral STO(001) substrate have no polar discontinuity. However, also in these systems, a 2DEG is observed[29];

• the 2DEG at LAO/STO(001) is only observed if the films are grown at a pressure below 10⁻²

(9)

mbar[31];

• according to this model, an electric field is what drives the electrons to the interface. Such fields can be observed by X-ray photoelectron spectroscopy (XPS). However, this field appeared to be absent[32]. However, this could be explained by a lowering of the potential by adsorbates, as reported by Xie et al.[28], see figure 1.4c.

To explain these other observations, alternative mechanisms have been investigated, discussed in the following sections.

Lanthanum doping

The possibility that STO is being doped with lanthanum ions during the deposition is the first alternative⁵. It has been shown that STO becomes conducting when it is sufficiently doped with lanthanum [16]. In that case, La³⁺ replaces Sr²⁺, which results in free electrons in the material. A similar way to turn STO conducting is by sustituting Ti⁴⁺ with Nb⁵⁺ [17].

When an LAO film is deposited by PLD at 850^◦C, it is possible for the lanthanum ions to diffuse into the STO substrate. This is shown to be true by TEM [33][34]. However, when looking in detail at the LAO/STO system, it appears that only films grown onto a TiO2-terminated surface generate a 2DEG, while the interface remains insulating with a SrO-terminated substrate. Intermixing would take place in both cases, so this mechanism cannot explain the 2DEG.

Also, intermixing would not result in a universal critical thickness. Perhaps it is true that only after depositing enough layers of LAO, sufficient lanthanum ions can dope the STO to become conducting.

However, the number of layers required for this threshold doping level would also depend on the growth conditions, which are not equal in different PLD systems around the world. Yet, the critical thickness of 4 unit cells is universally found. Therefore, La doping is an unlikely candidate for the origin of the 2DEG.

Another strong argument against lanthanum doping as cause of the 2DEG is given by Warusawithana et al.[33]. Increasing the amount of lanthanum in their LAO films⁶ resulted in insulating interfaces, even though this should have increased conduction levels if lanthanum doping was the origin of the 2DEG.

Although intermixing is not expected to cause the 2DEG, it can influence the mobility, since the intermixed ions can act as trapping acceptor states[35].

Oxygen Vacancies

Another alternative mechanism is the doping of STO with oxygen vacancies. STO becomes conducting (and even superconducting at low T) when oxygen vacancies are present [16]. Oxygen vacancies n-dope the STO by the following mechanism[32][35]:

O⁰_O↔ 2e⁻+ V²⁺_O +¹₂O₂

where O⁰_O is an occupied neutral oxygen site and V²⁺_O a positively charged oxygen vacancy.

During growth in the PLD system, the STO substrate is at high temperature in an often oxygen poor/vacuum environment, in which oxygen is easily removed from the STO. In chapter 3 (figure 3.1) the sheet resistance is shown for a reduced STO substrate, which shows metallic behaviour. If an LAO layer is deposited in vacuum, such a measurement results in a very similar graph. Also, no distinction is found between a TiO2-LaO and a SrO-AlO2 interface in this case. These observations lead to the conclusion that the measured conduction of vacuum-grown films originates at the substrate, not at the LAO/STO interface.

However, when the LAO film is deposited at an oxygen pressure of 10⁻³mbar, or when a vacuum- grown sample is annealed in an oxygen-rich environment (P_O₂ ≈ 300mbar), the behaviour seems to fulfill the criteria of interface termination selectivity and the existence of a critical thickness, as described in the beginning of this section (section 1.3.2)⁷.

5In literature this is often called intermixing: also part of the strontium enters the LAO film.

6Deposited by molecular beam epitaxy (MBE)

7This is verified in Chapter 3

(10)

(a) (b) (c)

Figure 1.5: Cross-sectional conductive AFM images of (a) non-annealed LAO/STO sample and (b) annealed LAO/STO sample. The low resistance line (red) is the interface between the top LAO and bottom STO. Taken from [36]; (c) Oxygen could diffuse from the STO substrate into an oxygen deficient film, resulting in conduction by oxygen vacancies at the interface (taken from [32]).

Figure 1.5 shows cross-sectional conductive AFM images of the LAO/STO interface. The non- annealed sample (figure 1.5a) shows that the entire substrate is somewhat conducting, with the highest conductivity near the interface, while the annealed LAO/STO sample (figure 1.5b) shows only conduction at the interface. As the bulk-conduction is clearly oxygen-vacancy induced, the conduction mechanism at the annealed interface is still unclear.

Not only the growth conditions influence oxygen vacancy concentrations in STO, also the deposited film material can do so. If a material such as LAO, which has a higher electron affinity than STO, is deposited on the STO, oxygen can be drawn to the LAO layer, resulting in oxygen vacancies in the STO [37][32], see figure 1.5c.

On the other hand, although it is clear that oxygen vacancies would be induced under any circum- stances, the critical thickness is not explained by holding only oxygen vacancies responsible. Also, films deposited under the same conditions can show both conduction and insulation, depending on other parameters (e.g. stoichiometry[33]).

Redox screening

Another possibility combines the oxygen vacancies with the electronic reconstruction model. Oxygen vacancies at the surface of LAO could donate their carriers to the interface[35] or a redox hydroxilation reaction with water and an oxygen vacancy could take place, also leading to tunneling electrons as a result of the potential inside the film. This model is supported by the observation that the interface of a 3 u.c. film of LAO and STO can be switched from insulating to conducting by applying a bias voltage to the film with an SPM tip [38], but only in the presence of water[39].

(11)

1.4 Thesis Outline

Chapter 2 explains the experimental techniques that were used throughout this thesis.

In Chapter 3, more evidence is found for oxygen vacancies being the cause of conduction in systems where STO is the substrate. The work of this chapter is a continuation of the BSc thesis of A. Kingma.

Chapter 4 invesitages the possibility to obtain the same interface (LaO-TiO2) as the one with the 2DEG, but in reverse order, where STO is the film and LAO the substrate (see figure 1.6b). According to the electronic reconstruction model, also this interface should generate a 2DEG. The many experimental difficulties are discussed and overcome step-by-step.

Chapter 5 takes a different approach. In the electronic reconstruction model, the doped electrons are taken up by the titanium ions. If not STO, but a different oxide perovskite with TiO2-termination is used as substrate, also a 2DEG should be formed at the interface of this substrate and an LAO film.

This has been done by growing LAO/CTO/STO heterostructures as shown in figure 1.6c.

(a) (b) (c)

Figure 1.6: (a) Famous LAO/STO heterostructure with 2DEG at the interface; (b) Inverted system:

STO on top of LAO with a possible 2DEG; (c) LAO/CTO/STO with potential 2DEG at LAO/CTO interface.

(12)

Chapter 2

Experimental Techniques

2.1 Pulsed Laser Deposition (PLD)

One of the techniques to produce thin films is Pulsed Laser Deposition (PLD). It is based on laser ablation of a material using a pulsed laser with a high intensity beam. Inside a vacuum chamber this beam is focused onto a target, which ablates due to the high energy density of the focused beam. The ablated material forms a plume of atoms and ions which deposites on a substrate which is held in the path of the plume. The plume maintains the stoichiometry of the target. Figure 2.1a shows a basic PLD setup.

(a)

(b) (c)

(d) (e)

Figure 2.1: (a) Setup of PLD and RHEED (image taken from [40]);Possible growth modes during PLD:

(b) Frank-Van der Merwe (layer-by-layer growth), (c) Stranski-Krastanov (layer+island growth), (d) Volmer-Weber (island growth) and (e) step-flow growth. Taken from [41].

Depending on several parameters, such as chamber pressure, substrate-to-target distance, heater temperature and deposition frequency, one can obtain films that can be crystalline and epitaxial to the substrate, sometimes even strained. The values of these parameters depend on the properties of the substrate surface, the target material and the desired film composition and structure.

In PLD growth, 4 growth modes can be distinguished: layer-by-layer growth, layer+island growth, island growth and step-flow growth mode, see figures 2.1b-e[41]. The first and last growth mode give flat surfaces, while the second and especially the third can give (very) rough surfaces.

The laser that was used for this research was a Lambda PhysiK COMPex Pro 205 KrF excimer laser with λ=248 nm and τ =25 ns.

(13)

(a) (b)

Figure 2.2: (a) Graphical representation of Bragg’s Law: X-rays will only be reflected when Bragg’s law is satisfied (dots are scatter sites, such as atoms); (b) In reciprocal space, Bragg’s law is fulfilled when the Ewald sphere intersects with two lattice points, which are the dots in this image. Taken from [7].

2.2 Diffraction Techniques

Waves reflect off a layered surface if the conditions for constructive interference are met. This is described by Bragg’s Law :

nλ = 2d sin θ (2.1)

where n is the order of reflection, λ the wavelength of the radiation source, d the distance between reflecting planes (atoms, surfaces and interfaces). In the case of a crystal lattice, this can be graphically represented as in figure 2.2a. To make it insightful which planes in which orientations would give rise to constructive interference, it can be shown that Bragg’s Law is equal to saying that Q, which is defined as:

−

→Q =−→ K₀−−→

K_d (2.2)

connects two reciprocal lattice points. Here,−→

K₀ is the scattering vector of the incident beam and−→

K_d the scattering vector of the diffracted beam. In turn, this is equal to saying that constructive interference will take place if the Ewald’s sphere, which is constructed with the two vectors, intersects two lattice points. This is shown graphically in figure 2.2b.

In this thesis, two techniques make use of this: Reflection High Energy Electron Diffraction and X-ray Diffraction. These are described in the following paragraphs.

2.2.1 Reflection High-Energy Electron Diffraction (RHEED)

The PLD system is equipped with a Reflection High Energy Electron Diffraction (RHEED) cannon and camera to provide real-time in-situ monitoring of the film’s growth. In RHEED an electron beam is aimed at the substrate under a very small angle. This direct beam will diffract at the crystalline substrate surface and its diffractions will end up at a phosphor screen, which emits light which is, in turn, captured by a camera.

The pattern that is observed with the camera is only sensitive to the surface, since the electrons do not penetrate very deeply into the material. This lack of periodicity in the out-of-plane direction, while in-plane the periodicity is bulk-like, gives rise to reciprocal rods instead of reciprocal lattice points, see figure 2.3a&b. Careful analysis of the RHEED pattern can therefore give information about the in-plane crystal structure of the surface. To illustrate how a RHEED pattern looks, figure 2.3c shows the RHEED pattern of an STO substrate at 850^◦C.

If the surface of the sample is very flat, the specular reflection of the beam is very good, but when the surface is less than perfect, more electrons will scatter diffusely and the intensity of the specular reflection will be less. This concept is shown in figure 2.4.

By using this technique during PLD growth, one can monitor in-situ and real-time the quality of the surface of the sample. If a film grows layer-by-layer, the surface is alternating between poor quality (when

(14)

(a) (b) (c)

Figure 2.3: (a) Side view of Ewald’s sphere in RHEED. Reciprocal rods are perpendicular to the sample surface; (b) Top view of Ewald’s sphere. In this orientation, the rods appear as points; (c) Typical RHEED pattern where both the direct and specular beams are visible, together with diffractions corre- sponding to the lines drawn in (b). Images (a) and (b) taken from [7].

the surface is not yet covered with enough material to form a full layer) and good quality (when the new layer is complete), which shows up as oscillations of the RHEED intensity. As in PLD, under the right deposition conditions, the material that is deposited is stoichiometric, a complete layer corresponds to a layer of full unit cells, not the half-unit cell layers used to explain polar catastrophe in chapter 1.

2.2.2 X-Ray reflectivity (XRR)

X-ray diffraction experiments were performed using a PANalytical X’Pert thin film diffractometer. A monochromator selects the CuKα from the X-ray source and suppresses the others. Since the wavelength of CuKα is in the order of (but smaller than) the structures that are of interest in this thesis (crystal unit cells and thin films), the X-rays can be used to study these structures.

In X-ray reflectivity this condition can be used to determine the thickness of a film. Here, also amorphous films can be measured, as long as the surface and the interface between film and substrate is flat (so that eq. 2.1 holds for enough surface area to obtain a decent reflection. In a θ-2θ scan with small values for θ (to about 8^◦) oscillations will appear, called Kiessig fringes (see figure 2.5). The period of these fringes is related to the thickness of the film[43]. And if, for example, two layers are present, two periodicities will be present. This is especially useful when no oscillations were observed in the RHEED signal during PLD from which the layer thickness could otherwise be obtained.

The general shape of the reflectivity measurement (i.e. when excluding the fringes) gives information about the electron densities of the substrate and film material and about the surface and interface roughness[43].

2.2.3 X-Ray Diffraction (XRD)

In X-ray diffraction (XRD), Bragg’s Law (eq. 2.1) is used to determine the structure of crystals. Since X-ray penetrate much deeper than electrons, this technique is sensitive to bulk as well. Due to their longe-range order, X-rays scattering off electrons in the crystal can constructively interfere (see figure 2.2a). Crystal planes can therefore be regarded as mirrors for X-rays.

The information on all these crystal planes is used to construct the unit cell of the crystal. Also, other periodic phenomena occuring in the crystal, such as periodic domains or superlattices, can be observed by X-ray diffraction.

Figures 2.7a and b show the reciprocal lattice points of a relaxed and strained film on a substrate, respectively. If the film is epitaxial, the lattice points in the [00l]-direction (specular) are aligned, but only if the film is fully strained will the off-specular point align, since then the in-plane lattice parameters of film and substrate are equal.

(15)

Figure 2.4: Intensity of specular RHEED reflection is determined by the surface roughness.

When the surface is complete, the intensity is high, but if the coverage of the surface with new material from PLD is low, the intensity is low. The process repeats itself in layer-by-layer growth. (Image from www.pascal-co-ltd.co.jp)

Figure 2.5: Simulation of typical X-ray reflectivity measurement with clear Kiessig fringes[42].

θ-2θ-Scans

To check for epitaxy, a scan along the [00l]-direction is taken of the sample. As explained in figures 2.7a and b, the lattice points of both the film and substrate are aligned if the film is epitaxial. Also, it gives information about the out-of-plane lattice parameter. If the value deviates from the film’s bulk value, the film is likely to be strained.

For a scan along the [00l] direction, the angle of both the source and the detector with the surface has to be increased at the same rate to maintain the specular reflection conditions. The angle between the source and the sample surface is often called ω, while the angle between the direction beam and the detector is called 2θ. In specular conditions, ω equals precisely half of 2θ, which is θ. Therefore, a scan along [00l] is often called a θ-2θ scan.

In such scans, the film thickness plays a similar role as in XRR. Near a peak, fringes appear due to the thickness of the film. Often these are also called Kiessig fringes, yet sometimes Laue fringes.

Reciprocal Space Maps (RSMs)

The reciprocal space of a crystal can be further investigated by doing 2-dimensional scans, in which both ω and 2θ are scanned, see figure 2.7b. Such a 2D scan (or set of scans) maps the reciprocal space and is therefore called a reciprocal space map (RSM). From these maps, one can quickly see graphically if a film is epitaxial (see figure 2.7a), strained to the substrate or if there are any periodicities in the film, for example. Since strain is an important property in this thesis, most X-ray related data is presented in the form of RSMs.

Powder diffraction

In contrast with single crystals, where an X-ray beam is diffracted only in certain directions, polycrys- talline materials, such as crystalline powders, diffract in cones¹, due to their random crystal orientation,

1So called Debye-Scherrer cones.

(16)

(a) (b) (c)

Figure 2.6: (a) A relaxed but epitaxial cubic film will have its [00l] lattice points aligned with the substrate, but not the off-specular points; (b) A strained film will have all hkl peaks aligned, parallel to [00l], with the substrate peaks. Images a and b taken from [44]; (c) A θ-2θ scan is performed by scanning the angle of the source and detector with the sample crystal surface while keeping ω=θ.

(a) (b)

Figure 2.7: (a) XRD Reciprocal Space Mapping (RSM). By varying ω and 2θ (i.e. the source/sample and detector orientation), a 2-dimensional image of the reciprocal space is obtained. If one takes ω such that ω=θ, a θ-2θ scan is obtained; (b) Diffraction of X-rays by single crystal. Instead of spots, a patern of circles is observed. Therefore, no reciprocal space mapping is required. Rather, a θ-2θ scan already provides all the information about the unit cell (red line). Adapted from [44]

see figure 2.7b. A θ-2θ scan, while the powder is rotated, will give all available diffraction peaks in one measurement, from which the unit cell of the material can be reconstructed.

2.3 Atomic Force Microscopy (AFM)

Atomic Force Microscopy is a technique that can be used to obtain topography images of a surface, which allows a far greater magnification than optical microscopes, since it is not limited by diffraction.

In contact mode, similar to a needle on a record player, a cantilever with a tip is moved across the surface.

A laser is aimed at the reflecting back side of the cantilever, which reflects the laser to the center of a detector. When the cantilever moves across the surface, the laser reflection moves across the detector, which is translated into a topography image. Also similar to the needle and the record player, this mode can damage the sample surface if it is soft and also the tip will erode quite fast.

In the more advanced Tapping Mode, the cantiliver is oscillated near a resonance. The tip then only touches the surface of the sample for short moments, damping the oscillation slightly. If the tip encounters a higher feature of the sample, the oscillation is damped even more. A feedback loop keeps the amplitude

(17)

of the oscillation constant by moving the average cantiliver position up and down². Since this is done by a piezoelectric crystal, the voltage on the crystal is a measure for the height of the measured surface and its features [45], see figure 2.8b. Since the tip only touches the surface for short times, the tip erodes much slower and the sample surface is hardly damaged, compared to the previously mentioned contact mode.

(a) (b)

Figure 2.8: AFM example: (a) Topography image of doubly terminated LaAlO₃ substrate; (b) Phase image of the same substrate. In the phase image, the two terminations can be clearly distinguished by their contrast.

Making use of the lock-in amplifier of the instrument, also phase images can be taken of the sample, see figure 2.8b. Differences in adhesion or elasticity of the sample surface result in a phase shift between the oscillation of the cantilever and the excitation of the piezo crystal. Although these phase shifts are highly non-linear in hard materials (such as the ones treated in this thesis) and therefore cannot be used for quantitative analysis, they can be used to make the distinction between two different materials[45].

Figures 5.5c and 4.9a show that the phase image can provide information about the presence of different materials on the surface. For example, when a LaAlO₃substrate is cut in the (001) direction, the surface is alternatingly composed of LaO and AlO23. This shows up very clearly in figure 4.9a. Which phase shift corresponds to which material cannot be extracted from this measurement however.

2.4 Van der Pauw Measurements

To compare the resistance of different sized and geometries of a material, people often use resistivity. For systems that are quasi-2D, such as the 2DEG in LAO/STO, a different quantity can be used, namely sheet resistance Rs, which is defined by:

ρ = RS· t (2.1)

where t is the thickness of the material and ρ the resistivity. From this it follows that the unit of R_S is Ω. However, as its meaning is very different from bulk resistance, the unit is usually represented as Ω/ (Ohms per square).

The sheet resistance of a system can be obtained by a measurement technique called the Van der Pauw method [46][47]. In order to perform Van der Pauw measurements, one has to make 4 contacts to 4 corners of the sample (which is, in our case, usually square), see figure 2.9. By sourcing a current through two contacts on one side and measuring a voltage between the two on the other side, 4 resistances can be defined.

R12,34= V₁₂ I34

R14,23=V₁₄ I23

R_34,12= V₃₄ I12

R_23,14=V₂₃ I14

2This technique is referred to as Amplitude Modulation. Tapping mode is therefore sometimes called AM-AFM.

3Double (single) termination can also be observed in the height image, when the step height amounts to half a unit cell (a full unit cell) step. This is actually the way to prove single termination. The phase image is mainly a graphical indicator of double termination for a large area, which can then be checked locally by taking a cross-section of the height image.

(18)

Figure 2.9: Van der Pauw sheet resistance geometry (Taken from www.nist.gov.)

Two resistances would already be sufficient in the ideal case, if the contacts were in a perfect geometry, the sample was homogeneous and the measurement set-ups did not introduce errors. Since this is difficult to accomplish in the real world, a work-around is to measure the same property in different orientations, resulting in 4 resistances instead of two. Reversing the polarity of the current source and voltage measurement device, 8 values are measured. Averaging these improves the accuracy of this method.

In the end, two resistances are defined:

RA=R_12,34+ R_34,12+ R_21,43+ R_43,21 4

RB =R14,23+ R23,14+ R41,32+ R32,41

4

The sheet resistance RS can subsequently be calculated (numerically) by the following formula:

exp −πRA

R_S

+ exp −πRB

R_S

= 1 (2.2)

RS can then be compared with other samples, as it does not depend on geometry or size.

Van der Pauw measurements are time consuming. Therefore a LabVIEW program was written, which is discussed in appendix B. The measurements of sheet resistance versus temperature were taken inside a Physical Property Measurement System (PPMS) by Quantum Design

2.5 STO substrate treatment

To obtain single termination of SrTiO3(001) substrates, the surface had to undergo a chemical treatment.

First, the substrate is cleaned with acetone and ethanol, both for about 15 minutes, in an ultrasonic bath.

Afterwards, the substrates are submerged in demineralized water for at least 30 minutes in the same bath, to let the following reaction occur[48]:

SrO + H2O → Sr(OH)₂ (2.1)

Subsequently, the substrate is held in hydrofluoric acid (HF) for about 30 seconds, during which the following reaction takes place[48]:

Sr(OH)₂+ 4HF → Sr²⁺+ 4F⁻+ 2H₃O⁺ (2.2) The strontium is etched away and the surface is left with a single termination of TiO₂. Afterwards, the substrate is annealed at 955^◦C. How long the annealing should last is influenced by the terrace length.

This is determined by AFM before etching. A flow of oxygen is present during the annealing⁴. After this treatment, one can see clear unit cell steps in the AFM topography image, see figure 2.10a&b. More proof is given by the phase images (figures 2.10c&d), in which only one termination is present at the terraces. Since the contrast is reversed for the trace and retrace images, it can be concluded that the observed contrast at the step edges is merely an effect of the terrace edges themselves, not from any double termination present at those edges.

4This is done to keep the surface clean during annealing. At high temperature, particles from the tube, crucible and atmosphere can stick to the surface. A flow of oxygen gas takes these particles away.

(19)

(a) (b)

(c) (d)

Figure 2.10: AFM images: (a) Topography image of singly terminated SrTiO3substrate; (b) Line scan across the line in (a). The step height is approximately one STO unit cell height; (c) Trace and (d) retrace of phase image of the same substrate. Only one termination is observed in between the clear contrast of the step edges.

(20)

Chapter 3

Oxygen Vacancy Induced Conduction

3.1 Introduction

To support the claim from chapter 1 that oxygen vacancies cause conduction in oxide perovskite heterostructures on SrTiO₃ substrates, several films of different material (CaTiO₃, PbTiO₃ and BiFeO₃) have been deposited onto STO. In some of them, a 2DEG is predicted, while others are predicted to be insulating and act as a control sample. First, the samples are grown in reducing environments (low pressure: PO₂ = 10⁻⁶mbar) and their conducting properties are then compared to the same samples after annealing in an oxygen-rich environment (PO₂= 300mbar).

PbTiO

₃

/SrTiO

₃

Lead titanate (PbTiO₃/PTO) is an oxide perovskite with a tetragonal crystal structure at room temperature. Its lattice parameters are a=3.902˚A and c=4.16˚A[49] and it has a band gap of 3.4eV[50]. Similar to SrTiO₃, which is a A²⁺B⁴⁺O₃perovskite as well, the PTO has neutral planes in the (001) direction (in the ionic limit). Growing PTO on STO would therefore not lead to a 2DEG based on electronic reconstruction in the simple charged planes picture. However, since PTO is ferroelectric[7], the polarization of the film already leads to a build-up of potential which would also lead to a polar catastrophe if no reconstruction would take place[25][51]. Therefore, one might also find a 2DEG in this system.

If this would work, the ferroelectric PTO film would already have a back electrode (the 2DEG) and would not need an additional process step to fabricate an additional electrode layer, for which usually SrRuO3is used[7]. Reducing the number of process steps for devices is favourable in industry applications.

Another possible application one could imagine is that the level of conductivity is switched between a high and low state by switching the ferroelectric PTO layer, since this changes the polarization at the interface, similar to what is mentioned in [52][53]. This could then be used in memory applications.

BiFeO

₃

/SrTiO

₃

Bismuth ferrite (BiFeO3/BFO) is a rhombohedrally distorted oxide perovskite with a pseaudo-cubic lattice parameter of 3.965˚A[13] and a band gap of 2.7eV[54]. It is of the same type as LaAlO3, namely A³⁺B³⁺O3. Therefore, also BFO has charged planes in the (001) direction (BiO⁺ and FeO⁻). Hence, the interface of BFO grown on STO should give rise to a 2DEG. Calculations have indeed shown that a 2DEG is expected[55]. BFO could be an even more interesting film material than ’ordinary’ LAO, since it has multiferroic properties at room temperature [13][56]. Also here, having a 2DEG as intrinsic back electrode could reduce the number of process steps for devices.

(21)

CaTiO

₃

/SrTiO

₃

The original Perovskite calcium titanate (CaTiO3/CTO), which is similar to STO and PTO (A²⁺B⁴⁺O3), actually does not have the perfect perovskite crystal structure: it has an orthorhombic structure with a pseudo-cubic lattice parameter of 3.82˚A[57] and a band gap of 3.5eV[58]. It has neutral planes in the (001) direction (CaO and TiO2). Therefore, in the ionic limit, the interface of a CTO on STO would not result in a 2DEG. As seen in the case of PTO, the ionic limit is not always the full story, since a polarization could also incude a 2DEG, theoretically. Although calculations have shown that CTO could be ferroelectric when strained, this has not been shown experimentally on STO yet. It is therefore not believed that this would occur in a system without taking any effort. Also, it is not reported that a 2DEG would form in this system. Therefore, this system is used as a reference for the insulating interface.

Part of the results in this chapter are the work of former Bachelor student A. Kingma[59], who measured electrical properties of the films.

(a) (b)

Figure 3.1: (a) Sheet resistance of perovskite films right after deposition (all films PO₂ = 10⁻⁶mbar except PTO/STO (PO₂ = 2 · 10⁻⁵mbar); (b) Sheet resistance of LAO/STO samples grown at low (PO₂ = 10⁻⁶mbar) and high (PO₂ = 10⁻³mbar) pressure and after post-annealing in PO₂ = 300mbar.

All measurements on the low-resistive samples (∼ 10Ω/square) performed by A. Kingma[59] (except bare STO).

3.2 Thin film growth and characterisation

Thin films of PTO, BFO and CTO have been deposited on TiO2-terminated STO(001) substrates at the growth conditions mentioned in Table 3.1. These conditions were optimized for the growth of LAO on STO, which resulted in the 2DEG. LAO/STO samples were grown as reference, with the same growth conditions as the other films. The pressure (10⁻⁶mbar) is similar to the pressure used by A. Ohtomo to obtain the highly conductive heterostructures. The film thickness was estimated to be around 20 u.c.

LAO/STO

The sheet resistance of the LAO reference samples is shown in figure 3.1b. The order of magnitude is the same as the values measured by A. Ohtomo[9], shown in figure 1.3a. As mentioned before, the sample that was grown at low oxygen pressure was likely to be oxygen deficient and therefore have bulk conduction of the STO substrate. Evidence for this is given in figure 3.1b as well: a bare STO substrate has been exposed to the same growth conditions as the low-pressure sample (T =880^◦C, PO₂=10⁻⁶mbar), but without any deposition. This substrate showed the same sheet resistance behaviour and values as the STO substrate with the LAO film grown on top. In this as-grown sample, apparently the conduction is through the bulk of the substrate, not only along the interface of the film and substrate.

(22)

Heater temperature 880^◦C

Oxygen pressure 10⁻⁶ mbar

Deposition frequency 1 Hz (LAO) - 0.5 Hz (PTO,BFO,CTO)

Heater-target distance 49 mm

Laser fluence 1.2 J/cm²

Table 3.1: Growth conditions for growth of LAO, PTO, BFO and CTO on STO.

After annealing in PO₂=300mbar at T =700^◦C, the sheet resistance was measured again. The sample had become about 10⁴times more resistive than before annealing. This increase in resistivity is generally ascribed to the amount of oxygen vacancies being completely removed from the bulk, leaving only the interface conducting. Whether there are still oxygen vacancies at the interface or whether this conductivity is caused by electronic reconstruction is still a point of debate in the scientific community, as is already mentioned in chapter 1.

The Other Films

The sheet resistance of the other as-grown films (BFO, PTO and CTO) all show similar behaviour as the low-pressure LAO/STO, see figure 3.1a. Since they are in the same order of magnitude as the oxygen-vacancy dominated samples of LAO/STO and reduced STO substrate, it is likely that this is also mainly caused by oxygen vacancies in the STO bulk. Therefore, like the LAO/STO film, these films were annealed in an oxygen-rich environment.

After annealing, the conduction had disappeared in all cases (i.e. no sheet resistance could be measured due to measurement limits of the measurement setup)(GΩ), except for the LAO/STO system, see figure 3.1b. From this observation one can conclude that, also here, the low resistivity of the as-grown samples was caused by oxygen vacancies in the bulk of the STO substrate, which were removed during the annealing process. The interface of these heterostructures is different from the LAO/STO system from the previous section, apparently.

3.3 Characterisation

The BFO/STO interface was expected to harbour a 2DEG, but did not. But before drawing any conclusions about the validity of the electronic reconstruction model, one should prove that the interfaces are equivalent, i.e. that also the BFO/STO interface is crystalline. If the interface has a lot of defects that already screen the polar discontinuity, there is no need for an electronic reconstruction and nothing can be said about electronic reconstruction. The same holds for PTO/STO. To investigate the crystallinity of the interface, thicker films were grown (approximately 50 u.c.) to grant a stronger signal in XRD.

RHEED

The growth was monitored in-situ by RHEED. Information about the growth mode already gives an indication on the quality of the films. In figure 3.2a the RHEED intensities are plotted for the different films. After growth of BFO, the RHEED diffraction pattern shows a clear signature of 3D (island) growth (figure 3.2b). On the other hand, the RHEED pattern of the PTO film has decreased strongly in intensity, indicating a rough surface. AFM topography images of both films show that the surface is indeed very rough/3D, see figure 3.2c.

In contrast with BFO and PTO, CTO appears to grow in layer-by-layer mode for a long time (over 20 oscillations can be counted in figure 3.4a). After the deposition, RHEED shows a pattern that is typical for a flat surface(figure3.4b). AFM indeed shows that after deposition of about 50 u.c., the step structure from the substrate is still visible, see figure 3.4c.

(23)

(a) (b) (c)

Figure 3.2: BFO/STO; (a) RHEED intensity during deposition. Inset shows close-up around 5 minutes of deposition. (b) RHEED pattern after deposition. This pattern is typical for films which have 3D growth (islands); (c) AFM topography image shows a relatively rough surface, in agreement with the RHEED pattern.

(a) (b) (c)

Figure 3.3: PTO/STO; (a) RHEED intensity during deposition. Intensity decreases, indicating a rough- ening of the surface. (b) RHEED pattern shows streaks that indicate a thin film, but the intensity if very low; (c) AFM topography shows that the surface is indeed rougher, but the signature terrace structure of the substrate is still observable.

(a) (b) (c)

Figure 3.4: CTO/STO; (a) RHEED intensity shows oscillations, which indicate a layer-by-layer growth, followed by step-flow growth. Inset shows recovery after each laser pulse while maximum intensity is stable; (b) RHEED pattern is typical for 2D growth; (c) AFM topography shows that the film is flat.

(24)

XRD

Finally, X-ray diffraction was performed on the films. Firstly, X-ray reflectivity measurements were taken of the four samples. These are shown in figures 3.5a,d,h,j. Both the LAO and CTO films give rise to Kiessig fringes in the intensity, which indicates that the film has a low surface roughness, which was already concluded from AFM as well. On the other hand, both PTO and BFO films show no fringes, which indicates that the the films have a high surface roughness. This was also observed in AFM.

Secondly, θ-2θ scans were performed around the (002) reflection of the STO substrate. Any epitaxial film should give rise to a peak in this scan, since the film’s crystal planes are aligned with those of the substrate. What’s more: in thin films, the main Bragg peaks are accompanied by Laue fringes¹, as mentioned in section 2.2. Since the incident angle is much larger for (002)-reflections than for XRR, the measurement is much less sensitive to surface roughness. In this case, one might see fringes that were not observable in XRR.

Figures 3.5b,e,h,k show the θ-2θ scans. Both the LAO and CTO films have a clear peak and also clear fringes. But this time, also the PTO and BFO films show some fringes near the STO substrate peak.

Since the pseudo-cubic lattice parameters for these materials are very close to that of STO, the film peak is partly obscured by the substrate peak. But the presence of the fringes indicates that some crystalline phase is present.

Lastly, reciprocal space maps were made for the four films, which are shown in figures 3.5c,f,i,l. A crystalline epitaxial film would show a film peak in this scan and if it is fully strained, the peak should be on the same vertical line as the substrate peak in this image. This is only seen in the CTO/STO and LAO/STO system. Since the pseudo-cubic lattice parameters of PTO and BFO are close to STO, the peak should be close to the substrate peak. However, from the θ-2θ scans it already became evident that only a small fraction of the film is crystalline, resulting in a very low intensity peak, compared to the crystalline CTO and LAO films. Any signature of epitaxy is therefore expected to also be very weak in the RSM.

Both the LAO and CTO films show a peak that is aligned with the substrate peak. This shows that the film is strained and that the interface is likely to have a low defect concentration. The other two films (PTO and BFO) do not show such a clear peak. Perhaps the small vertical streak is related to the films, similar to the fringes in the θ-2θ scans. From this absence of a strong peak, as in the LAO and CTO heterostructures, it can be concluded that the crystal quality in the BFO and PTO films is very poor compared to the LAO and CTO films.

3.4 Discussion

From the AFM and XRD measurements, it can be concluded that both PTO and BFO do not grow in crystalline epitaxial films on STO under the growth conditions that were used. Although all heterostructures were conducting after growth in vacuum, all of them lost their conductivity and became highly-insulating, except for LAO/STO. In this heterostructure, the LAO film is strained to the STO substrate, therefore the interface must have few defects. The heterostructure that is also strained, is CTO/STO. However, as this system does not have any polar discontinuity or well-pronounced polarization, no 2DEG is predicted here. The fact that no conduction is measured, does not tell us anything about the electronic reconstruction model, however, only that it is not in disagreement with it. The absence of the 2DEG in the other two systems (PTO/STO and BFO/STO) is likely to be caused by a poor film quality.

Usually, PTO and BFO films are not deposited under the same conditions as those that were used in this thesis. Lead is volatile at high temperature. Deposition of a PTO film at 850^◦C in vacuum surely result in evaporation of lead into the vacuum chamber, causing a non-stoichiometric film which is less likely to become crystalline. Therefore, PTO films are usually deposited at lower temperatures (e.g.

≈ 570^◦C[7]).

Also for BFO films, usually the growth temperature is much lower than 850^◦C, namely around 670^◦C[60][61]. The parameters that were used for this thesis were so different that no crystalline film was formed. Also, in literature, generally a back electrode of SrRuO3(SRO) is deposited before the BFO

1Also called Kiessig fringes, but generally that name is used for the fringes in X-ray reflectivity

(25)

films, as mentioned in section 3.1. With the optimized growth conditions, strained films can be grown[60]

(see also appendix A), but these conditions do not apply for growth without the SRO layer.

Very recently (July 2015), Chen et al. reported on a 2DEG at the interface of BFO and STO[61].

With TEM they observe an atomically flat interface between the two crystalline materials. Similar to the measurements in figure 1.5, they use cross-sectional conductive AFM to show that the interface is conducting. Unfortunately, no sheet resistance has been measured. Therefore, its value cannot be compared to LAO/STO yet.

3.5 Conclusions

In this chapter, oxide perovskite films on (001) oriented STO substrates have been investigated for their 2DEG properties. Results by A. Kingma [59] showed that the vacuum-grown samples were conducting, and in this thesis it was shown that these films had values of conduction that were similar to those of reduced STO substrates. During the work for this thesis it was found that the conducting properties were lost upon annealing in an oxygen rich environment, except for the LAO/STO heterostructure. Also, XRD analysis showed that PbTiO3 and BiFeO3do not grow in a crystalline films under the used growth conditions.

Hence, the conclusion can be drawn that the measured conductive properties of the films, with the exception of LAO/STO, were a result of oxygen vacancies in the bulk of STO instead of a 2DEG at a crystalline interface of a polar perovskite film and non-polar substrate. An implication of this conclusion is that, for conducting oxide samples, one should always try to rule out oxygen vacancies first, for example by annealing in an oxygen-rich environment. Whether oxygen vacancies can really be ruled out in the case of LAO/STO is still unclear.

These results do not exclude a 2DEG at crystalline interfaces of PTO/STO and BFO/STO, however.

The latter has recently been observed, even[61]. It is therefore still interesting to study these systems.

(26)

(a) LAO/STO (b) LAO/STO (c) LAO/STO

(d) PTO/STO (e) PTO/STO (f) PTO/STO

(g) BFO/STO (h) BFO/STO (i) BFO/STO

(j) CTO/STO (k) CTO/ST (l) CTO/ST

Figure 3.5: (a,d,g,j) X-ray reflectivity; (b,e,h,k) θ-2θ scans; (c,f,i,l) RSMs around (103) reflection.

(27)

Chapter 4

The Interface of STO/LAO

4.1 Introduction

In this Thesis, the main hypothesis being tested is the following: the 2DEG at the interface of (among others) LAO films on STO substrates is a result of electronic reconstruction to prevent polar catastrophe, in the fashion as discussed previously in section 1.3.2. During the investigation of this phenomenon, researchers most often use STO as substrate, while growing many different films. LAO is not used much for this kind of heterostructures because the surface of LAO reconstructs due to its polarity, because of twinning due to its rhombohedral crystal structure at room temperature and because of the absence of a reliable way to obtain a single termination of the (001) surface. Any attempt to grow an epitaxial film with a coherent interface that generates a 2DEG should at least fix the first and last of these points.

In this chapter, we have tested if restoring a structural reconstruction of LAO is possible and, if so, if that would allow us to show coherent growth of STO on LAO, unlike the work reported by Liu et al.[62].

Also, several ways to obtain single termination have been examined. Finally, a way to achieve a twin-less LAO substrate is investigated. In the end, the goal of these three improvements was to obtain an interface that is as similar to the LAO/STO interface as possible, while not having STO as the substrate but the film.

4.2 The STO/LAO Interface

STO was first deposited by PLD on an as-delivered LAO substrate. The growth conditions are in table 4.1. RHEED oscillations (figure 4.1) show that the first few monolayers have a poor crystallinity, as their intensity is very low. After the growth of one monolayer, the film crystallinity improves and the intensity recovers. During the first 12 monolayers, the growth mode is layer-by-layer, since oscillations can be observed, after which the films grows in step-flow mode. This can be conluded from the absence of oscillations but presence of a recovery after each laser pulse and high intensity (figure 4.1a). Also, AFM images show that the surface is still flat (see figure 4.1b).

Heater temperature 850^◦C Oxygen pressure 10⁻⁶ mbar Deposition frequency 1 Hz Heater-target distance 49 mm

Laser fluence 1.2 J/cm²

Table 4.1: Growth conditions for growth of STO on LAO.

(28)

(a) (b)

Figure 4.1: (a) RHEED oscillations during growth of STO on LAO. Inset shows signal after 500 seconds, where after each laser pulse, a recovery is observed, which is characteristic of step-flow mode; (b) AFM image of STO/LAO after growth.

The low intensity of the first monolayer can be linked to growth of a film with lots of defects. After some time, the intensity increases again, which indicates that the film’s crystal quality (or at least the surface quality) is improved.

If a lot of defects are present near the interface, it is unlikely that the film has grown fully strained onto the substrate. Therefore, the expectation is that this film is at least partially relaxed. The film appears not to be fully relaxed; in the XRD θ-2θ scan in figure 4.2a the angle at which the (002) reflection is expected for bulk STO is 46.45^◦; instead, it is found at 45.95^◦. To rule out misalignment: the LAO substrate peak is found at 47.96^◦, which is almost equal to the expected 48.02^◦.

(a) (b)

Figure 4.2: (a) θ-2θ scan around (002) reflection. Lines show where the peaks would be for bulk or fully relaxed materials; (b) RSM around (103) reflection of LAO.

To determine if the film has really been grown epitaxially, an RSM was taken around the (103) reflection of STO. The result is shown in figure 4.2b. It is clear that the film does not have the same in plane lattice parameter as the substrate since the peaks are not found at the same value for qx. Therefore, the film is not fully strained, yet it is not fully relaxed at the same time.

(29)

When performing Van der Pauw measurements on this sample, no conduction was observed. This is in agreement with the measurements by Liu et al.[62].

4.3 Buffer Layer STO/LAO/LAO

Liu et al.[62] investigated the interface of STO grown on LAO as well. They obtained a singly terminated LAO substrate by annealing it at 1000^◦C for 2 hours. After depositing a film of STO, they do not obtain a 2DEG but a highly insulating material. They attribute this to a non-polar top-layer of the LAO substrate, which prevents the formation of a clean polar-to-non-polar interface between the LAO and STO.

In a similar system, namely CaTiO3 grown on LAO(111), such a layer exists as well [63][64]. In the absence of a film, the top layers of LAO undergo a certain structural reconstruction to prevent a charged surface, since a polar-to-vacuum surface is quite similar to a polar-to-non-polar interface and, thus, highly unstable[25]. When the substrate is heated to deposition condictions (700^◦C), this reconstruction remains. If another perovskite with a different lattice parameter is deposited onto this substrate, the top- layer prevents strained/coherent growth, as the top layer has a different crystal structure than perovskite.

However, the film can still be epitaxial as shown in figure 4.3a.

(a) (b)

Figure 4.3: TEM images of CaTiO₃ film on LaAlO₃(111) substrates (a) without and (b) with buffer layer of LaAlO₃(taken from [64]).

This problem was resolved by Blok et al. by depositing 3 monolayers of LAO onto the LAO substrate before growing the film[64]. Apparently the surface reconstruction can be restored to bulk termination by doing so. According to them, the LAO becomes sufficiently conducting at high temperature to screen the charge from the surface, preventing the otherwise resulting surface reconstruction. After growing the CTO on the LAO with this so-called buffer layer in between (without cooling down in between the two depositions), the film proved to be coherent and the interface sharp, shown by TEM in figure 4.3[64]. So, a buffer layer might resolve the issue of incoherence at the interface STO/LAO which leads to a (partial) relaxation of the film and which could be related to the non-polar top-layer mentioned by Liu et al.

Hence, a buffer layer of LAO was deposited before growing the STO film. The same growth conditions were used for this buffer layer as for the STO film (see table 4.1). RHEED shows layer by layer growth (figure 4.4a). At the third maximum in the RHEED signal, the deposition of the buffer layer was stopped and changed to growth of STO. The growth of STO starts off as layer-by-layer, but the oscillations soon disappear. Only after a minute clear step-flow growth is observed.

AFM images (figure 4.4b) show that the surface of the sample is flat, supporting the observation of step-flow growth in the RHEED intensity.

An XRD θ-2θ scan (figure 4.5a) shows that the STO film peak is found at the same position as when no buffer layer was introduced (45.99^◦). The LAO substrate is at 47.99^◦again. Figure 4.5b shows a RSM

(30)

(a) (b)

Figure 4.4: (a) RHEED oscillations during deposition of LAO buffer layer and STO film. Inset shows recovery after each laser during step-flow mode; (b) AFM image of STO/LAO/LAO.

around the (103) reflection, where it can be seen that also in this sample, the film has a different in-plane lattice parameter than the substrate and is therefore not fully strained.

Like the sample without buffer layer, this sample also shows insulating behaviour when attempting electrical transport measurements. Possible explanations are:

• the non-polar top-layer was not removed by introducing the buffer layer. Therefore, there is no difference with the unbuffered sample;

• even if the non-polar top-layer was removed by the buffer layer, the substrate still has a double termination. This would also result in an insulating surface, since two points on the surface are unlikely to be connected by LaO-TiO2 interfaces: there are AlO2-SrO interfaces blocking the carriers.

• not electronic reconstruction but a different mechanism (that does not result in conduction) fixes polar catastrophe, as mentioned in the introduction (section 1.3).

(a) (b) (c)

Figure 4.5: STO/LAO/LAO: (a) XRR; (b) θ-2θ scan around (002) reflection; (c) RSM around (103) reflection.

Interface Conductivity in Oxide Perovskite Heterostructures