Advanced Functional Materials: A Perspective from Theory and Experiment

Advanced Functional Materials

Advanced Functional Materials

Tara Prasad Das

Biplab Sanyal

Olle Eriksson

ISSN  1875-4023

Volume 2 • Number Suppl. (C) • 2012

Table of Contents

Cover image

Title page

Copyright

Foreword

Preface

Contributors

Theory of Diluted Magnetic Semiconductors

Molecular Beam Epitaxial Growth and Magnetotransport Properties of GaMnAs

1 Introduction

2 Molecular beam epitaxy growth of GaMnAs

3 Magnetic and transport properties of GaMnAs

4 Summary

Spectroscopic Investigations of Electronic Structure

1.1 X-ray absorption spectroscopy and X-ray magnetic circular dichroism

1.2 X-ray photoelectron spectroscopy and resonant photoelectron spectroscopy

1.3 X-ray photoelectron spectroscopy and resonant photoelectron spectroscopy

1.4 X-ray emission spectroscopy

1.5 Extended X-ray absorption fine structure (EXAFS)

Theory of Magnetism in Graphene

1 Introduction

2 Experimental evidence

3 Theoretical framework

4 Concluding remarks

Synthesis of Graphene and Derivatives

1 Introduction

2 Synthesis by scalable exfoliation from graphite

3 Molecular synthesis

4 Graphene by CVD

5 Transfer/deposition

6 Further modifications by functionalization

Multiferroics: Theory, Mechanisms, and Materials

1 Introduction

2 Theory of proper multiferroics

3 Theory of improper multiferroics

Multiferroic Complex Metal Oxides: Main Features of Preparation, Structure, and Properties

1 Multiferroic materials

2 Incompatibility between ferroelectricity and magnetism

3 The classification of multiferroic compounds

4 Perovskites: introduction and scope

5 Coexistence of ferroelectricity and magnetism in Pb-based perovskites

6 Ternary Fe-containing perovskite oxides

12 Experimental techniques

13 Sample preparation

14 Sample characterization

15 Crystal chemistry search of multiferroics

16 Crystal chemistry criteria for the search of multiferroic complex metal oxides

17 Single phase multiferroic materials

18 Predictions for new multiferroics

19 Other approaches

20 Ferroelectricity induced by A-site disorder

21 Concluding remarks

Index

Copyright

Foreword

Preface

Contributors

Theory of diluted magnetic semiconductors

Olle Erikssona, Diana M. Iusanb, Patrik Thunstroma, Johan Hellsvikb, Igor di Marcoa, Lars Bergqvistc, Peter Dederichsd, Biplab Sanyala

a Division of Materials Theory, Department of Physics and Astronomy, Uppsala University, Box 516, 75120 Uppsala, Sweden

b Consiglio Nazionale delle Ricerche Istituto Nazionale per la Fisica della Materia (CNR-INFM), Laboratorio Regionale CASTI, LAquila, Italy

c Tillampad Materialfysik, KTH, Stockholm, Sweden

d IFF, Forschungszentrum Juelich, Germany

Abstract

In order to understand and manipulate the properties of diluted magnetic semiconductors, a precise knowledge of interactions in these systems should be available. In this chapter, we review ab initio studies on DMS, in particular, calculations of ordering temperatures by Monte Carlo simulations. We argue why we do not expect large values of ordering temperatures from the point of view of dilute spin system, short ranged magnetic exchange interactions and magnetic percolation. Also, we show results of atomistic spin dynamics using Landau–Lifshitz–Gilbert equations for Mn-doped GaAs. The effects of strong electron correlation on the electronic and magnetic structure are demonstrated by dynamical mean field theory.

Keywords

• Diluted magnetic semiconductors • Density functional theory • Exchange interaction • Electron correlation • Dynamical mean field theory • Atomistic spin dynamics • Monte Carlo simulations

1 Introduction

A very promising way to obtain novel devices is to connect different physical properties of materials, thus potentially enriching the functionality of these devices. This line of thought is pursued in the field of spintronics, where one goal is to find materials whose magnetic properties can be tuned by electrical fields and vice versa. Among the most promising candidates are a group of materials called diluted magnetic semiconductors (DMS), materials with a low concentration (typically 5–10%) of magnetic elements embedded into a semiconducting host. Their potential in spintronic applications has led to very intensive investigations, both experimentally [1], as well as theoretically [2–5]. These materials show complexity regarding the crystal chemistry, the magnetic properties, as well as transport behavior (for a short summary of these materials see Ref. [6]).

2 Experimental background of selected materials

alloys, attributed to an RKKY-like interaction for low values of xas high as 110 K [1] (which in subsequent works has been pushed up to 170 K for a Mn concentration of 7–8%[10]). This was soon followed by the theoretical predictions by Dietl et al. [2] of high transition temperatures of a series of Mn-doped semiconductors. The theory was built on Zener’s model of p–d exchange for ferromagnetism [11], where the exchange splitting of localized d-electrons of magnetic impurity atoms (e.g., Mn) was argued to induce a hole in the valence band (e.g., of GaAs), with the related valence band moment coupling anti-ferromagnetically to the impurity d-moment. The exchange coupling between the localized moments was assumed to be mediated by the conduction carriers of the host, and a long ranged character of the exchange interaction was proposed. This implies that even for dilute doping concentrations, with Mn atoms distributed far apart, a significant ferromagnetic coupling between transition metal moments might be obtained. Based on a mean field analysis, ordering temperatures above room temperature were predicted for several semiconducting hosts with only a few percent of transition metals. The picture of long-ranged interactions has been challenged from calculations based on first principles theory, that suggest a much more short ranged interaction in general, with magnetic percolation as a key ingredient [4,5].

In practical applications, an ordering temperature above room temperature is necessary, and two materials were predicted [2] to be very promising in this regard, namely Mn-doped GaN and Mn-doped ZnO. The consequence of this prediction was the exponential increase in the interest in the field, and numerous groups have been involved in the growth and characterization of these materials by a variety of techniques. Unfortunately, it has been very difficult experimentally to unambiguously reach sufficiently high values of the ordering temperature. As a matter of fact no unanimous ordering temperature well above room temperature has been established.

As Mn-doped GaN and Mn-doped ZnO have been predicted to be room-temperature ferromagnets [2], much of the attention has been focusing on these systems. Some early studies of Mn-doped GaN reported ferromagnetic properties [12,13]. More recently, Yoshii et al. reported ferromagnetism above room temperature, which was suppressed below 10 K [14]. In contrast, ferromagnetism with a low Tphase [17].

The magnetic properties of Mn-doped ZnO also continue to be lively debated, and the scientific community is also here divided into two groups, either claiming intrinsic room temperature ferromagnetism or the lack of it. Ferromagnetism at room temperature or above was for instance reported in Refs. atmosphere were paramagnetic [21]. Similarly, Hou et al. showed that the ferromagnetism of Mn:ZnO thin films is destroyed by annealing in oxygen [22]. Sharma and Varma found that hydrogenation induces week ferromagnetism in otherwise paramagnetic Mn:ZnO samples [23]. In these cases the observed ferromagnetism was attributed to the presence of defects. Also, in a study by Yan et al. the ferromagnetism of highly resistive samples was attributed to zinc vacancies [24]. Ferromagnetic properties were also observed for Mn:ZnO co-doped with N [25], or P [26]. Yang et al. showed a clear dependence of the ferromagnetic properties on the electron concentration in n-type Mn:ZnO co-doped with Ga, for which the magnetic moment decreases with decreasing electron concentration [27]. Paramagnetism of Mn:ZnO has also been reported by several groups [28–33]. Studies by Blasco et al. of 4% TM-doped ZnO (TM = Mn, Fe, Co) point out that the ferromagnetism at room temperature is not intrinsic, but due to the presence of secondary phases, often below the detection limits of XRD.

ferromagnetism [34,35] and paramagnetism [36] were made. In addition, antiferromagnetic Co–Co interactions for n-type Co:ZnO were reported [36]. Recently, Zhou et al. have shown that ferromagnetism of Co:ZnO is due to the formation of magnetic precipitates and they emphasized the difficulty of using XRD for their detection [37].

From the description above it seems that the experimental situation in DMS materials seems unclear, but it may be argued on general grounds that DMS materials with a homogeneous distribution of magnetic impurity atoms (e.g., Mn or Co) should have substantial magnetic moments of these impurities (of order μB/atom). This is indeed what first principles theory results in for the presently considered materials (see, e.g., Ref. [50–52]). Hence one may suspect that reports of large ordering temperatures for materials with a tiny magnetic moment for the magnetic impurity atoms are caused by antiferromagnetic clusters with uncompensated spins at the surface of the cluster.

There are several unsolved issues of DMS materials, e.g., the nature and strength of the interatomic exchange interaction, the dynamical properties of the magnetization, as well as the possible influence of electron correlations on the electronic structure and magnetic properties. Several reviews with various focus have been published for diluted magnetic semiconductors [38–41]. Here we summarize some of the more recent aspects of diluted magnetic semiconductors, with emphasis of electron correlations, magnetization dynamics and possible maximum ordering temperatures. We start however with a short description of the general features of magnetism and exchange in these materials, based on density functional theory in the local spin density approximation or parametrized Hartree–Fock like theory as given by the LDA + U method.

3 First principles theory

First principles theory of DMS materials relies on the theorems of Ref. [42], which turns the many-body problem of the electronic structure into a one-electron problem with an effective potential. The latter is often approximated by the so-called local density approximation (LDA) [43]. By now several numerical methods for finding a solution to this effective one-electron problem have been suggested and found to give reliable results, both as concerns the magnetic moments and inter-atomic exchange interaction between magnetic atoms of a solid. Hence first principles calculations based on, e.g., linear muffin-tin orbitals [44–46] are found to agree with results of the KKR method [47]. For close packed solids it is also by now well known that the atomic sphere approximation gives results in agreement with full-potential calculations [48]. Since these calculations can be made in a spin-polarized fashion, it is possible to calculate magnetic moments of a material, and if electron correlation is not a too dominant feature, experiment and theory agree with an error often not exceeding a few percent.

First principles theory also offers the possibility to evaluate the pair exchange interaction between atomic spins, and hence it can be concluded that the parameters of the Heisenberg Hamiltonian are possible to calculate, without adjustable parameters. The most common approach is the use of Liechtenstein’s theory [49], where the exchange interaction between two spins is calculated using a classical Heisenberg Hamiltonian and making use of the magnetic force theorem. The expression for the pair-exchange parameter reads

     (1)

, t being the on-site scattering matrix. T is the trace over the orbital indices of the scattering matrices. The first principles calculations presented in this chapter made use of a Green’s function Korringa–Kohn–Rostoker (KKR) method within the atomic sphere approximation (ASA) [46]. The local density approximation was used for the exchange–correlation functional, together with an spdf .

The values of calculated exchange parameters among the Mn spins embedded in all semiconducting hosts considered in this work are given in . This seems to be a value most first principles calculations reach for the majority of the DMS materials studied so far. Table 1 also lists the exchange interaction between Mn atoms, as a function of distance, or rather as a function of coordination shell. It is clear from the table that the interaction is dominated by the nearest neighbor contribution, which is ferromagnetic for all studied materials. The exchange interaction between atoms further away is quickly decaying; after four shells it is in general less than one percent of the nearest neighbor interaction. One may note that for some distances the interaction is actually negative and hence of an antiferromagnetic nature. However, the overall ordering of all materials listed in Table 1 is ferromagnetic. The short ranged nature of the inter-atomic exchange of DMS systems has led to the conclusion that magnetism in these materials is heavily influenced by percolation [4,5].

Table 1 Calculated Mn projected magnetic moments and exchange parameters for various diluted magnetic semiconductors incorporating a Mn concentration of 5%.

4 Magnetization dynamics

Recently the dynamical properties of spin-systems have attracted intense focus. This is spurred both by technical developments, like pump–probe measurements using intense lasers, as well as a technological drive, where the speed of reading and writing information in a magnetic medium is of technological importance. However, the dynamics of magnetism offers also a way to interpret the nature of the exchange interactions of a material, e.g., if it is magnetically ordered or a spin-glass. The latter aspect is of particular importance for DMS materials, since for some cases spin-glass behavior has been demonstrated experimentally, whereas for some materials speculations about a spin-glass material have not been confirmed. In this sub-section we outline one approach with which to undertake theoretical studies of magnetization dynamics of DMS materials, namely the use of atomistic spin dynamics (ASD) simulations.

Observing that the dominant microscopic magnetic moments are large and stable in many of these materials, a Langevin atomistic spin dynamics (ASD) approach can be used for simulations of the magnetization dynamics [50–52]. This technique relies on the adiabatic approximation and coarse graining to atomic magnetic moments. The central entity of the technique is the stochastic Landau–Lifshitz–Gilbert equation

     (2)

is calculated from a parametrized magnetic Hamiltonian that can contain contributions from isotropic and anisotropic exchange, single-ion magnetocrystalline anisotropy (MAE), and Zeeman energy. When focusing on shorter length scales, the influence of magnetostatic interactions can typically be neglected. For a comprehensive description of the ASD approach, see, e.g., Refs. [50–52].

As described in the previous section, the Heisenberg exchange interactions in a DMS can be calculated using DFT methods. Recently Hals et al. used scattering matrix theory to calculate the Gilbert damping tensor for GaMnAs [53]. The formalism takes into account the sample shape and treats chemical disorder in a non-perturbative way. It is well known from experiments that the sample quality can strongly influence the damping, with dirtier samples typically exhibiting a larger damping than cleaner and well-annealed samples. This is captured by the scattering matrix technique. Furthermore it was proposed that for smaller samples it is the sample shape that is the main cause for anisotropic damping, rather than the anisotropies of the crystal structure.

Insights into the microscopic interactions can alternatively be inferred from various experimental techniques. Of importance are here not least magnetic resonance (MR) techniques which can give insight into exchange stiffness, the effective damping as well as MAE and shape anisotropy. To disentangle the contributions to the MR signal can be a challenging task, which can be further complicated by surface effects and sample impurities. Bihler et al. have recently analyzed the magnetic resonances in GaMnAs taking into account the sample geometry, and also effects from surface spin pinning [54].

[55]. It was established that the oscillatory part of the Kerr signal, probed with a linear polarized probe pulse is connected to the ferromagnetic order and shows only a weak dependence on the polarization of the pump pulse. The pump pulse is exciting a comparably small number of carriers of the GaMnAs sample and consequently the ferromagnetism is only weakly perturbed [56]. It was proposed that the phonons emitted by photo-injected carriers cause a change in the MAE, thereby initiating magnetic precession which reveals itself in the oscillating Kerr signal. Using more intense pump pulses, with a fluence in the range of 1–10 mJ/cm², Zahn et al. investigated the carrier and magnetization dynamics in GaMnAs [56]. The laser pulse this time achieves a strong, but not complete demagnetization of the sample. The carrier dynamics was studied through the reflectivity signal and it was observed that the recombination of carriers proceeds over several nanoseconds. As demonstrated by Scherbakov et al., coherent magnetic precession in GaMnAs can be initiated by picosecond acoustic pulses [57]. The MAE in GaMnAs is highly sensitive to strain. In contrast to the optical pump methods, the change of the anisotropy is here achieved without affecting the concentration of charge carriers.

In simulations based on Eq. (2) one has the freedom to choose freely the initial spin configuration and to include or exclude the various terms in the Hamiltonian. Using exchange parameters calculated for different As antisite concentrations, Langevin spin dynamics simulations have been performed on GaMnAs, focusing on how the magnetization responds to changes in the temperature [58]. The pair correlation function for equal times is defined as

     (3)

is calculated. In Figures 1 and 2 the pair correlation function is plotted at different times as a function of distance between atoms for two different starting configurations of the atomic moments; a random distribution (denoted rDLM in Figs. 1 and 2) and a ferromagnetic distribution (denoted FM in Figs 2.a and 2). The simulations are performed for As antisite concentrations of 0.75% and 1.25% at a temperature of 100 K. For the lower antisite concentration, this temperature is below the Curie temperature, whereas for the higher antisite concentration it is above. In these figures the pair correlation function approaches its equilibrated value after a sufficiently long time of the simulations. For the FM configuration in Fig.1 this is seen to take place after approximately 5 ps. It is also clear from Fig.1 that the FM starting configuration reaches equilibrium faster than a rDLM configuration. It can be observed in Figure 2 that substantial short range correlation occurs also above the Curie temperature. The rate of magnetic relaxation was in Ref. [58] also studied by following the spin autocorrelation function and comparisons were performed with other materials, e.g., typical spin-glass materials like Mn-doped Cu. It was concluded [58] that in the dilute limit Mn-doped GaAs is not a spin-glass, a situation which most likely applies also to other DMS materials. It was also found [58] that the relaxation, as a consequence of the comparably weak exchange field associated with the dilution, is a slower process in a DMS material than in ordinary ferromagnetic Fe.

Figure 1 . The equilibrium pair correlation is shown by the black thick line. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this book.)

Figure 2 Same as Figure 1 but for an As antisite concentration of 1.25%.

5 Dynamical mean field theory

So far we have discussed theoretical results based on the local spin density approximation (LSDA), in which correlation effects are included to some extent in the exchange and correlation part of the function. However, it is clear that this way to treat correlations has limitations, in particular for transition metals forming compounds with pnictides and chalcogenides (oxides are a typical example). Hence one may suspect that also for DMS materials, e.g., Mn-doped GaAs, a more direct method to include electron correlation is needed. So far, the LDA + U method has been tried and one can conclude that, e.g., for Mn-doped GaAs the magnetic properties are not dramatically changed, when compared to calculations based on LSDA. This holds both for the magnetic moment as well as the ordering temperature. However, neither LSDA or LDA + U seems to reproduce the measured valence band spectrum of, e.g., Mn-doped GaAs, suggesting a complex form of correlations in DMS materials. A recent experimental finding of Mn-doped GaAs shows [59] that the Fermi level is situated essentially in the band gap, in a fashion similar to what LSDA or LDA + U calculations give [5]. The assumptions of the model of Ref. [2], result in a situation where the Fermi level cuts through the top of the valence band, and with the Mn d-states situated at significantly lower energy, seemingly in contrast to the experiment of Ref. [59].

The discussion above point to that a thorough analysis of the electronic structure of Mn-doped GaAs is highly needed, in which the effects of electron–electron are included at a level beyond a static Hartree–Fock level, as provided by the LDA + U method. We present here the electronic structure and magnetic properties using dynamical mean field theory and LDA + U, in combination with a full-potential electronic structure method. We have for this calculation used a full-potential linear-muffin tin orbital method [60–64], in