Durham Research Online

Deposited in DRO:

22 May 2012

Version of attached �le:

Accepted Version

Peer-review status of attached �le:

Peer-reviewed

Citation for published item:

Hindmarch, A.T. (2011) 'Interface magnetism in ferromagnetic metal-compound semiconductor hybridstructures.', Spin., 1 (1). pp. 45-69.

Further information on publisher's website:

http://dx.doi.org/10.1142/S2010324711000069

Publisher's copyright statement:

Electronic version of an article published as Spin, 1,1, 2011, 45-69, 10.1142/S2010324711000069 c© Copyright WorldScienti�c Publishing Company.

Additional information:

Use policy

The full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, forpersonal research or study, educational, or not-for-pro�t purposes provided that:

• a full bibliographic reference is made to the original source

• a link is made to the metadata record in DRO

• the full-text is not changed in any way

The full-text must not be sold in any format or medium without the formal permission of the copyright holders.

Please consult the full DRO policy for further details.

Durham University Library, Stockton Road, Durham DH1 3LY, United KingdomTel : +44 (0)191 334 3042 | Fax : +44 (0)191 334 2971

https://dro.dur.ac.uk

Interface Magnetism in Ferromagnetic Metal – CompoundSemiconductor Hybrid Structures

Aidan T. HindmarchSchool of Physics & Astronomy, University of Nottingham,

Nottingham, NG7 2RD, United Kingdomaidan.hindmarch@nottingham.ac.uk

Interfaces between dissimilar materials present a wide range of fascinating physical phenomena.When a nanoscale thin-film of a ferromagnetic metal is deposited in intimate contact with acompound semiconductor, the properties of the interface exhibit a wealth of novel behaviour,having immense potential for technological application, and being of great interest from theperspective of fundamental physics. This article presents a review of recent advances in the fieldof interface magnetism in (001)-oriented ferromagnetic metal/III-V compound semiconductorhybrid structures. Until relatively recently, the majority of research in this area continued toconcentrate almost exclusively on the prototypical epitaxial Fe/GaAs(001) system: now, a sig-nificant proportion of work has branched out from this theme, including ferromagnetic metalalloys, and other III-V compound semiconductors. After a general overview of the topic, and areview of the more recent literature, we discuss recent results where advances have been madein our understanding of the physics underpinning magnetic anisotropy in these systems: tai-loring the terms contributing to the angular-dependent free-energy density by employing novelfabrication methods and ferromagnetic metal electrodes.

Keywords: Magnetic anisotropy, thin films and interfaces, III-V semiconductors

1. Introduction

Epitaxial growth of Fe on GaAs(001) was first re-ported, seemingly as an aside, in a paper on in-terface chemistry in Schottky contacts by Waldropand Grant in 19791. However, the curious and fas-cinating magnetic properties exhibited by epitax-ial Fe/GaAs(001) systems were not revealed until1983, when Jantz et al. mention, again almost inpassing, a novel uniaxial magnetic anisotropy inthin Fe films on GaAs(001)2. This novel uniaxialanisotropy was finally placed firmly center-stage in1987 by Krebs et al.3: this slow and steady build-upprecipitated decades of intensive research into theproperties of ferromagnetic films on the (001) sur-faces of compound semiconductors — the topic ofthis review.

The Fe/GaAs(001) system is considered one ofthe original prototype systems in the field whichhas now come to be known as ‘spintronics’4: Fe

and GaAs are closely lattice matched, having lat-tice constants d = 2.86 A and a = 5.653 A respec-tively, allowing relatively easy cube-on-cube epitax-ial growth, i.e., Fe(001)<100> || GaAs(001)<100>,with lattice mismatch of ∼ 1.4 %. The interfacesbetween epitaxial bcc ferromagnetic metals andzincblende-structure semiconductors have been in-tensely studied not only because of their applicabil-ity to spintronics and unique magnetic anisotropybehavior, but also as they provide the ability tosynthesize and characterize metastable crystallinephases of ferromagnetic metals. Here, we will coverrecent advances in the field — an extremely com-prehensive overview of the topic, as the state-of-the-art stood in 2005, has previously been providedby Wastlbauer and Bland5: we aim to focus on themore recent work, particularly with respect to mag-netic anisotropies in these systems.

Whilst the microelectronics industry remainsprimarily focussed on silicon, compound semicon-

1

2 Aidan T. Hindmarch

ductors are extensively used as test-bed and proto-type systems in spintronics, in addition to havinga great many applications in their own right. Fer-romagnetic thin-films deposited on the III-V com-pound semiconductors attract the vast majority ofinterest: however, similar effects to those discussedhere may also be observed in the zinc-blende II-VI compound semiconductors, one of the most fre-quently studied cases being ZnSe4. In recent years,the primary driving forces behind the continued in-terest in Fe/GaAs systems has been the quest forefficient injection of a spin-polarized current intoa semiconductor. Currently, one of the more com-mon device structures for spin-injection employs theSchottky barrier formed at the interface between ann-type semiconductor and ferromagnetic metal con-tact6 in order to provide a ‘tunnel’ contact, negat-ing the problem of conductivity mismatch7 on thespin-injection efficiency. Although the principles ofspin-injection apply equally to any semiconductor,GaAs and its related III-V compound semiconduc-tors are particularly useful in this regard as a directband-gap in the near-infra-red energy range, whichmay be easily modulated with III-V alloy composi-tion, allows optical techniques to be employed. Thespin-light-emitting-diode (spin-LED) structure8 —consisting of a p-i-n quantum-well LED beneath aferromagnetic contact, shown in figure 1a) — al-lows a model-independent, quantitative, determina-tion of the carrier spin-polarization in the semicon-ductor from the circular polarization of the emittedelectroluminescence9.

Although planar devices such as the spin-LEDhave attracted significant research effort, in recentyears attention has become increasingly focussed onlateral structures, aimed toward spintronic deviceimplementations such as the now-infamous Datta-Das spin-field-effect-transistor (spin-FET)10, shownin figure 1b). In the Datta-Das device, the spin-polarized current which flows in the semiconductorchannel, often supposed as a 2-dimensional elec-tron gas (2DEG), is manipulated by electrostaticgate-action, hence providing electrical control ofthe source-drain current in addition to that pro-vided by the relative alignment of the ferromag-netic contacts. To this end, imaging of the in-jected spin-polarization from Fe into n-GaAs hasbeen performed in a lateral geometry by Crooker etal.11 and in a cross-sectional geometry by Kotis-sek et al.12, whilst Lou et al.13 and Lou et al.14

have demonstrated all-electrical measurements of

spin-injection, -transport and -detection in a singleFe/GaAs(001) device. Garlid et al.15 have also usedFe/InGaAs(001) contacts to measure the transversespin-current generated via the spin-Hall effect.

An interesting topic at present is magnetic tun-nel junction structures, planar devices where a spin-polarized current tunnels between two ferromag-netic electrodes separated by a nanoscale insulat-ing barrier layer, where the barrier is a compoundsemiconductor. Some theories predict very largetunnel magnetoresistance: the electronic-structureof the zincblende crystal makes ‘coherent tun-nelling’ across the Fe/GaAs(001) interface possi-ble16. Moser et al. fabricated Fe/GaAs/Fe(001) de-vices and find magnetoresistance around 6 % at4.2 K, which reverses sign with increasing bias17:a result somewhat short of the 1000s of % changepredicted. However, more recent theoretical workby Autes et al. has demonstrated that, due tothe strong spin-orbit coupling in GaAs, saturationof the magnetoresistance is reached with increas-ing GaAs thickness, and a limited magnetoresis-tance (in the range as low as 30-50 %) results18.In the related magnetic tunnel transistor, a variantof the metal-base transistor structure with ferro-magnetic metal emitter and base contacts, the fer-romagnetic base acts as a hot-electron spin-filter,resulting in a highly spin-polarized current injectedinto the semiconductor collector19. Since the ad-vent of crystalline ferromagnet/oxide/ferromagnetdevices20,21 it is now possible to fabricate fullyepitaxial Fe/MgO/Fe/GaAs(001) transistor struc-tures. Calculations by Autes et al. have demon-strated that the coherent crystal structure of thesedevices provides potentially very large magnetore-sistive functionality22, and the possibility of suchspin-polarized resonant tunnelling devices has beenshown by Nagahama et al.23. These results suggestpotential for room-temperature operation of spin-transistor and spin-polarized resonant-tunnellingdevices consisting solely of ferromagnetic metalsand compound semiconductors.

One of the critical factors in enabling the tech-nological application of nanomagnetism is that themagnetization direction in, for example, ferromag-netic spintronic storage and processing elementsmust be both stable and controllable. In order tobring this about, a complete understanding of themagnetic anisotropies in ferromagnetic thin films isof vital importance for future nanoscale spintron-ics applications: without such an understanding, a

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 3

(a) (b)

Fig. 1. Cartoon diagrams of (a) spin-LED and (b) Datta-Das spin-FET hybrid ferromagnet-semiconductor device structures.

fundamental limit is placed on the degree to whichminiaturization of device elements may proceed.Thus, although the study of magnetic anisotropyhas been reported for over a century24, it still re-mains an important, pertinent, and interesting topicto this day.

2. Overview of magneticanisotropies

The magnetic anisotropy energy is the differencein the Helmholtz free-energy for magnetization ori-ented along the ‘easy’- and ‘hard’-axes of the mag-netic anisotropy: it provides the means by whichthe magnetization may be confined to lie along agiven spatial direction and, along with the presenceof a spontaneous magnetic moment, is crucial inproducing ferromagnetism. In the absence of mag-netic anisotropy, ordered moments are able only toaccess a superparamagnetic phase.

The magnetocrystalline anisotropy may bethought of as being related, via the spin-orbit in-teraction, to the anisotropy in the orbital magneticmoment associated with each lattice site. In Fe, Co,Ni, and their alloys, the 3d band is over half-filled;the spin-orbit coupling should restrict the maxi-mal component of the orbital moment vector ml

to be parallel to the spin moment vector ms (andhence magnetization). However, the crystal field in-teraction causes the orbital magnetic moment vec-tor to be confined to a given high-symmetry di-rection25 depending, in a non-trivial fashion, onfactors such as the lattice spacing, orbital popula-tion etc.26. Provided the magnetization is alignedalong such a high-symmetry direction of the cubiccrystal, the free-energy of the system is minimizedvia the (leading) spin-orbit coupling term in themagnetic anisotropy energy, ∆ESO ∝ −ml · ms

27.Thus the magnetocrystalline anisotropy takes thesame symmetry as the crystal lattice, and so bcc-ferromagnetic metals possess cubic magnetocrys-

talline anisotropy. One recalls that, in a bulk cubiccrystal, the free-energy density ε is often expressedin terms of an expansion in powers of the direc-tion cosines α1, α2 and α3 which the magnetizationmakes with the crystal axes28, giving

ε = K1

(α21α

22 + α2

2α23 + α2

3α21

)+K2α

21α

22α

23 + ...,

(1)where K1 and K2 are the first- and second-ordercubic magnetocrystalline anisotropy constants, re-spectively. (Typically K1 ≫ K2, and K2 is oftentaken to be zero: it is common also to neglect theisotropic term K0 in the free-energy as it makes nooverall contribution to the magnetic anisotropy en-ergies).

2.1. Cubic anisotropy in aferromagnetic metal

For the bcc crystal-structure, the three high-symmetry directions of the crystal ⟨100⟩, ⟨110⟩, and⟨111⟩, correspond to easy-, intermediate-, and hard-axes of the cubic magnetocrystalline anisotropyin Fe28, corresponding to positive K1 (K2 = 0)in equation 1. In the bulk bcc-CoxFe(100−x) al-loys (having stable bcc crystal structure for x .75 at.%29) the sign of K1 changes, corresponding tothe cubic easy-axes shifting from the ⟨100⟩ to the⟨111⟩ directions — and hence the cubic hard-axesfrom ⟨111⟩ to ⟨100⟩ — at compositions correspond-ing to around x ∼ 40 at.%28.

When the topic of discussion turns to thin-films, one must consider also the demagnetizingfactor30 due to shape-anisotropy, which causes themagnetization to preferentially lie in the plane ofthe layer: thus we need consider the ⟨111⟩ directionsno further as they do not lie within the plane of the(001)-oriented thin-film. The high-symmetry direc-tions in the plane of the film are then the ⟨100⟩ and⟨110⟩ directions, which for Fe are thus the in-planecubic easy- and hard-axes, respectively.

4 Aidan T. Hindmarch

Table 1. Summary of the in-plane easy- and hard-axis directions of the cubic magne-tocrystalline anisotropy for epitaxial body-centered cubic thin-films of the ferromagnetic3d transition-metals and their alloys, deposited on GaAs(001).

Material Composition range Cubic easy-axes Cubic hard-axes Reference

Fe — [100], [010] [110], [110] 5CoxFe(100−x) x . 30 % [100], [010] [110], [110] 31

CoxFe(100−x) 30 . x . 70 % [110], [110] [100], [010] 32, 31, 33, 34

Co — [100], [010] [110], [110] 35, 36Ni — [100], [010] [110], [110] 37

NixFe(100−x) all x [100], [010] [110], [110] 38

Table 1 shows the easy- and hard-axis direc-tions of the cubic magnetocrystalline anisotropyfor thin-films of various ferromagnetic metals onGaAs(001): in each case the ferromagnet has a bcccrystal structure, which is a metastable phase formany of these materials. An interesting, and some-what surprising, result is found: in almost all cases,the exception being the ‘Co-rich’ CoFe alloys, thecubic magnetocrystalline anisotropy has easy axesalong the in-plane ⟨100⟩ directions31,35–38 — as isexpected for Fe. This is despite the fact that in theirstable fcc crystal phases, both Co and Ni have cubiceasy-axes along the ⟨110⟩, i.e., K1 is expected to benegative28. In CoxFe(100−x), the cubic anisotropyconstant indeed changes sign, becoming negativefor x & 30 at.%31,32: however, studies in the com-position range 70 . x < 100 at.%, where it appearsa further sign-reversal must occur, have not yetbeen performed. In the case of the NixFe(100−x) al-loys the cubic anisotropy constant is again positive,with in-plane cubic easy-axes along ⟨100⟩ direc-tions, over the entire composition range: in thesealloys, though, the magnetocrystalline anisotropybecomes vanishingly small over a wide composi-tion range, 20 < x < 80 at.%38, despite bothbcc-Fe and Ni having large, positive K1 in ele-mental form. The supposedly well-understood mag-netocrystalline anisotropy reveals some surprisingresults in these materials.

The first experimental work focussing solelyon magnetism in epitaxial Fe films on GaAs(001)was reported in 1987 by Krebs et al.3. Their fer-romagnetic resonance measurements indicated thatwhilst the free-energy density for magnetizing thefilm along the in-plane ⟨100⟩ directions were equiv-alent, due to the anticipated cubic magnetocrys-talline anisotropy in their Fe films, the in-plane⟨110⟩ directions were inequivalent — indicative of

an additional uniaxial magnetic anisotropy termKU with easy-axis oriented along the [110] andhard-axis along the [110] direction. The origin ofthis uniaxial magnetic anisotropy which is foundin these, and related, films has escaped explana-tion: it is toward gaining an understanding of thisanisotropy we will focus the remainder of this re-view.

2.2. Mixed cubic and uniaxialmagnetic anisotropy

In thin-film magnetism, the ‘effective’ anisotropyconstants are frequently considered to be composedof a combination of volume and interface terms,which may be explicitly separated in the form

Keffa = Kvol

a +K inta /t, (2)

where a = U, 1, 2, etc., and t is the thicknessof the ferromagnetic film. For thin epitaxial bcc-ferromagnet/III-V(001) one typically findsKvol

U ≈ 0and |Kvol

1 | ≫ |K int1 |. Thus the uniaxial magnetic

anisotropy is purely interfacial in origin, related insome way to the interface between the ferromag-netic metal film and the compound semiconductor.

The cubic magnetocrystalline anisotropy arisespredominantly due to the film volume, and has onlya weak thickness dependence due to truncation ofthe crystal lattice symmetry at the interface. Thistruncation of the lattice results in an interfacialcomponent to the cubic anisotropy of opposite signto the volume component: at a critical thickness,typically found to be around 6 monolayers (ML)39,the interface term dominates and the effective cu-bic anisotropy Keff

1 changes sign — the easy axesof the cubic magnetocrystalline anisotropy rotatefrom, e.g., ⟨100⟩ to ⟨110⟩ directions for Fe. Notethat this effect is purely related to loss of transla-tional symmetry in the bcc lattice of the ferromag-

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 5

netic film: the use of different substrate materialshas negligible further influence on the cubic magne-tocrystalline anisotropy.

In the prototypical epitaxial bcc-Fe/GaAs(001)system, within the Stoner-Wohlfarth model40 of co-herent magnetization reversal by rotation of a singledomain, one may rewrite equation 1 to express theangular-dependence of the free-energy density in-cluding terms describing magnetocrystalline, mag-netoelastic, interface, and Zeeman energy terms forsaturation magnetization MS making an angle θwith the in-plane [100] direction as41

ε (θ) ∼ Keff14 sin2 (2θ) + B2ϵ

2 sin2 (2θ)

+Kint

Ut sin2

(θ − π

4

)−HMS cos (θ) , (3)

where B2 is the second order magnetoelastic cou-pling constant, and ϵ the shear strain. Althoughthere is a lattice mismatch of around 1.4 % betweenFe and GaAs, it is common for the magnetoelasticterm to be omitted from the analysis of anisotropiesin Fe/GaAs(001) contacts. For CoxFe(100−x) alloysthe magnetocrystalline anisotropy constant changessign at x ∼ 30− 45 %28,31: hence whilst equation 3remains appropriate for Fe-rich alloys, a more ap-propriate expression for the free-energy density inCo-rich alloys on GaAs(001) is commonly written(omitting the magnetoelastic term) as32

ε (ϕ) ∼ −Keff14 sin2 (2ϕ) +

KintUt sin2 (ϕ)

−HMS cos (ϕ− α) , (4)

where now ϕ is the angle between the magnetizationand the [110] direction, and α is the angle betweenH and the [110] direction. It is obvious that caremust be taken in applying the correct expression:further difficulties may arise when the interfacialuniaxial easy axis is no-longer oriented along the[110] direction, but rather along [110], as is the casein films deposited on, for example, InAs(001).

When considering such an admixture of cu-bic magnetocrystalline and interfacial uniaxial mag-netic anisotropies, three distinct situations may oc-cur: i) when the ferromagnetic film is very thick, inwhich case |Keff

U | ≪ |Keff1 |, the uniaxial anisotropy

is negligible and the film behaves as if it has onlycubic in-plane magnetic anisotropy; ii) in the casewhen the ferromagnetic film is very thin, in whichcase |Keff

U | ≫ |Keff1 |, the cubic anisotropy is neg-

ligible and the film behaves as if it has only uni-axial in-plane magnetic anisotropy; and, iii) when

the film is of a thickness where |KeffU | ≈ |Keff

1 |, bothterms contribute, which results in two-stage magne-tization reversal for fields applied along the uniaxialhard-axis (UHA), depicted schematically in the in-set to the lower right frame of figure 2. Straightarrow represents a coherent rotation of the mag-netization, whilst curved arrows represent abruptchanges in magnetization direction. The strengthof the uniaxial magnetic anisotropy determines thepoints at which the magnetization jumps betweenthe uniaxial hard-axis and cubic easy axis (CEA).

2.3. Evaluation and quantificationof the anisotropy

Where the uniaxial magnetic anisotropy dominates,M(H) is single-valued (figure 2, lower left): we mayobtain an expression for H(m), where the normal-ized magnetization component along the uniaxialhard-axis is m = M/MS = sin θ, from minimizingequation 4 for α = π/2, i.e., field applied along theuniaxial hard-axis. In this case the anisotropy con-stants may simply be evaluated by fitting

H(m) = 2Keff1

(2m3 −m

)/MS + 2Keff

U m/MS. (5)

In cases where a two-stage magnetization-reversalis observed, following the method outlined byDumm et al.32 we are able to extract the effec-tive uniaxial anisotropy constant Keff

U and first or-der effective cubic anisotropy constant Keff

1 fromthe hysteresis loops measured along the UHA, fig-ure 2 lower right, assuming that the magnetizationreversal around zero applied field takes place by acoherent rotation. It may be shown that the effec-tive anisotropy constants are given by

KeffU = −1

2

MSHS

(H2

Ss2 +HSs+ 2

)H3

Ss3 +H2

Ss2 +HSs+ 1

(6)

and

Keff1 = −1

2

MS (HSs− 1)(H3

Ss3 +H2

Ss2 +HSs+ 1

)s, (7)

where HS and s are respectively the shift-field andzero-field slope of the normalized hysteresis loops,as indicated in figure 233,34. Thus extraction of boththe uniaxial and cubic anisotropy constants may bereadily achieved across the whole thickness rangeusing simple hysteresis loop measurements. Morecomplicated experimental techniques, including fer-romagnetic resonance and Brillouin light scatter-ing, have also been extensively applied to these sys-tems in order to extract the magnetic anisotropy

6 Aidan T. Hindmarch

Fig. 2. In-plane magneto-optic Kerr effect hysteresis loops for Co70Fe30/GaAs(001) along (upper) the uniaxial easy-axis and(lower) the uniaxial hard-axis. Left frames are an ultra-thin film, where the uniaxial magnetic anisotropy dominates, and rightframes are a slightly thicker film, where uniaxial and cubic magnetic anisotropies are of similar strength. Shown inset on theright are schematics showing an example of the in-plane magnetic anisotropy and magnetization reversal. The lower left frameshows a fit to the hysteresis data using equation 5, and in the lower-right frame the zero-field slope s and split-field Hs features,which may be used in determining the magnetic anisotropies from equations 6 and 7, are indicated. (After Hindmarch et al.34.)

Table 2. Summary of lattice mismatch between Fe and various III-V compound semi-conductors (and the II-VI compound ZnSe), valence spin-orbit parameter in the semicon-ductor, and the direction of the interface-induced uniaxial easy-axis for an Fe thin-filmdeposited onto the (001) surface.

Material Lattice const. Lattice mismatch S-O coupling UEA References

a (A) η (%) ∆0 (eV) a,b

GaAs 5.65 1.41 0.34 [110] 42, 43, 44Al8Ga92As — — ∼ 0.33 [110] 45

AlAs 5.66 1.25 0.3 [110] 46In20Ga80As — 0 ∼ 0.33 [110] 42In50Ga50As — −2.13 ∼ 0.33 [110] 44, 47 (IBD)Al48In52As — (−)2.2 — [110] 48

InP 5.87 −2.32 0.11 [110] 49InAs 6.06 −5.37 0.38 [110] 50, 42GaSb 6.09 −5.95 0.76 — 46

ZnSe 5.67 1.14 0.43 c [110] 51

Note: a Values for the valence spin-orbit parameter in the binary III-Vs taken from ref-erence 52.b Values for the valence spin-orbit parameter in the ternary III-Vs after reference 53.c Value for the valence spin-orbit parameter in ZnSe taken from reference 54.

constants: very good agreement is found betweenthe results for different techniques. As an exper-imentalist there is now a collection of tools whichare suitable for studying magnetism in ferromagnet-semiconductor hybrid structures.

3. Influence of the underlyingcompound semiconductor

As we have seen, the cubic magnetocrystallineanisotropy is strongly dependent on details of theferromagnetic metal film, particularly its elemen-

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 7

tal composition and thickness: the uniaxial mag-netic anisotropy, on the other hand, seems relativelyrobust against changes to the ferromagnetic metalfilm. The crystallographic direction along which theuniaxial easy-axis lies is also dependent on the ma-terial constituents of the contact: however, in thiscase it is the semiconductor which determines theuniaxial anisotropy, rather than the ferromagnet.

Values for the lattice constant a, lattice mis-match η = (a − aFe)/a, valence spin-orbit cou-pling strength ∆0, and the interface induced uniax-ial easy-axis direction for various epitaxial Fe/III-V(001) systems (and Fe/ZnSe(001)) are given intable 2. Entries for the III-Vs are ordered by in-creasing lattice parameter, and we see that the lat-tice mismatch for the epitaxial Fe overlayer passesthrough a minimum, being roughly zero for theIn20Ga80As ternary compound. One very interest-ing point to note is the dependence of the uniaxialeasy-axis direction on the lattice mismatch: if onewere to assume that the interface anisotropy arisesdue to epitaxial strain or anisotropic stress relax-ation, as have previously been suggested in the lit-erature, one may think that a positive lattice mis-match (2aFe > a) results in interface anisotropyalong [110]. From this, we would then anticipatethe interface anisotropy vanishing when the latticemismatch is zero — leaving only the cubic mag-netocrystalline anisotropy: however, this is not thecase42. Indeed, for both In50Ga50As and Al48In52Asternary compounds, the lattice mismatch is negative(2aFe < a) but the easy-axis of the uniaxial mag-netic anisotropy remains along the [110] direction.This shows that it is unlikely that epitaxial strainis the origin of the uniaxial interface anisotropy, asis now the generally-held consensus33,41,55,56.

An additional point to note is the variation ofthe valence-band spin-orbit splitting parameter, ameasure of the spin-orbit interaction strength in thesemiconductor. In most of the III-V materials it isin the range 0.3–0.4 eV, with the specific exceptionsof InP and GaSb. In GaSb the spin-orbit interac-tion is far stronger, mainly due to the ‘heavy’ anti-mony, but, due to the large lattice mismatch withFe, defect-free epitaxial growth is difficult and theuniaxial magnetic anisotropy has not been observedin this system. However, in InP the spin-orbit inter-action strength is much smaller than in the otherIII-Vs: in reference 49, Zakeri et al. found that theuniaxial magnetic anisotropy in Fe/InP(001) notonly is rotated 90◦ relative to that in Fe/GaAs(001),

but also is significantly weaker.Given the dependence of the uniaxial magnetic

anisotropy on the underlying semiconductor mate-rial, the influence of the capping layer material hasalso been investigated; particularly the influence onthe in-plane uniaxial magnetic anisotropy. Shaw etal. have found that an Al capping can induce avolume component to the uniaxial anisotropy inultra-thin Fe films on GaAs(001), and suggest thatthis may be due to an additional anisotropic strainmechanism which is absent for Au or Cu capping.Morley et al. observe slight differences between Crand Au capping of Fe films on GaAs(001) and In-GaAs(001), concluding that the primary mecha-nism for the effect is additional interdiffusion at theCr/Fe interface over the Au/Fe interface44.

A further mechanism which has been shownto cause a modification of the uniaxial magneticanisotropy in epitaxial films are surface step-edges. Epitaxial CoFe films deposited on vicinalGaAs(001) substrates, miscut at various angles to-wards the two inequivalent [111] planes, have beenstudied by Isakovic et al. and Wolf et al.. In bothcases a rock-salt structured Sc30Er70As buffer layerhas been deposited between GaAs(001) and CoFe:the buffer-layer acts to weaken the interfacial uni-axial magnetic anisotropy, making the the influenceof the substrate miscut more clearly apparent. Thesubstrate miscut is found to influence the rever-sal path taken by the magnetization, allowing thevector magnetization response of the film to anapplied magnetic field to be tailored57,58. Yoo etal. have studied Fe films deposited onto a rangeof vicinal GaAs(001) surfaces59, and shown thatan asymmetric planar Hall effect arises due to themagnetization reversal occurring in the (001) planerather than the (vicinal) plane of the film.

In a similar fashion to, for example,the exchange-bias effect, the unique magneticanisotropies which arise if hybrid ferromagnet-semiconductor systems are now beginning to be uti-lized as a means to control the magnetic anisotropyin nanoscale systems, in addition to being a stand-alone topic of study. Kipferl et al. compared the‘Bloch-3/2 law’ spin-wave parameters30 for Fe filmsepitaxially deposited on GaAs(001) and Au(001),finding a smaller depolarization due to thermalspin-waves in Fe/GaAs(001) as a consequence ofthe additional magnetic anisotropy contributionfrom the interfacial uniaxial magnetic anisotropy60.

8 Aidan T. Hindmarch

Kipferl et al. and Niu et al. have both investigatedmagnetic properties of arrays of sub-micron epitax-ial Fe dots patterned on GaAs(001)61,62. Niu et al.find that using a focussed ion-beam milling tech-nique to pattern the dot-array appears to causemodifications to the magnetic anisotropies. Meng etal. have demonstrated the interplay between cubicmagnetocrystalline, interface-induced uniaxial, andshape anisotropies in a series of arrays of micron-scale rectangular elements63. Despite extensiveuse being made of the ferromagnet/semiconductorinterface-induced uniaxial magnetic anisotropy, thedetails of the origin of the effect are still not wellunderstood.

3.1. What may determine theuniaxial magnetic anisotropy?

Since the initial work of Jantz et al. and Krebs etal. on magnetism in Fe/GaAs(001), the field hasexpanded to an unprecedented extent: hybridferromagnet-semiconductor structures and devicesare now key topics in nanomagnetism and spintron-ics. However, despite extensive study extending overmore than a quarter of a century, the physical ori-gin of the interfacial uniaxial magnetic anisotropyin these systems has continued to retain somethingof an air of mystery. From early work in the field,several ideas as to how the uniaxial anisotropyarises have been put forward, each of which seemsto have found favour for at least a short period oftime. Below we briefly summarize the areas whichhave been studied previously, and how these havebeen shown to contribute (or not) to the interfacialinduced uniaxial magnetic anisotropy.

Direction of the uniaxial magnetic anisotropy : Inthe early literature on the subject there is signif-icant disagreement over exactly which crystallo-graphic direction corresponds to the uniaxial easyaxis: although in some of the first reported casesthe uniaxial easy-axis was not well defined (mostprobably as a result of poorer interface quality inthe earlier works), later reports appeared to dis-agree on whether the uniaxial easy-axis was alignedalong [110] or [110]. Wastlbauer and Bland reana-lyzed the prior literature on the subject and con-cluded that, after correcting for accidentally misre-ported crystallographic axes, the uniaxial easy-axisin Fe/GaAs(001) is oriented along the [110] direc-tion, with the uniaxial hard-axis along [110]5.

Anisotropic island-growth: Numerous studies onthe initial stages of growth of Fe on GaAs(001)have found that the growth proceeds initally as a3-D island growth mode, with islands nucleatingalong the As dimer rows ([110] direction). These Feislands then gradually coalesce as the coverage in-creases, with islands elongating along [110], and av-erage height in the range 1–2 ML64,65. Coalescencegenerally occurs at nominal thickness of around 2–3 ML66; while well-separated islands are observed,the surrounding areas of the GaAs(001) surface arefound typically to retain their initial surface recon-struction. Although the interface structure beneaththe Fe islands is unknown, Ionescu et al. have pro-posed a model of Fe2As cluster formation during theinitial stages of growth65, and Godde et al. foundclusters of both Fe2As and Fe3Ga(2−x)Asx follow-

ing thermal processing67. Gillingham et al. showedthat Fe islands may relax by anisotropic diffusion,favouring [110], over a period of ∼ 30 hours66.However, Thomas et al. demonstrated that theshape-anisotropy due to elongated grain-growthin Fe/GaAs(001) does not influence the uniaxialmagnetic anisotropy41.

Magnetoelastic coupling : Related to the above dis-cussion of island-growth during early stages of filmdeposition, it has been suggested that anisotropicepitaxial strain, and/or anisotropic stress relax-ation may determine the interfacial uniaxial mag-netic anisotropy, e.g., from first-principles calcula-tions by Mirbt et al.68. Thomas et al. also foundthat anisotropic shear-strain has no influence on theuniaxial magnetic anisotropy in their Fe/GaAs(001)films41. It is now considered, in-part also due to ob-servations of strong uniaxial magnetic anisotropyin CoFe/GaAs(001) systems with far lower latticemismatch than Fe/GaAs31–34, that whilst magne-toelastic coupling may contribute to the interfacialuniaxial magnetic anisotropy, it must play only arelatively minor role.

Interface bonding : The idea that the uniaxial mag-netic anisotropy in Fe/GaAs(001) was in some wayrelated to interfacial bonding was first suggestedby Jantz et al., who supposed that the lack offour-fold symmetry could perhaps be reconciled byconsidering that the GaAs(001) surface presenteda reduced symmetry, possibly due to surface recon-struction2. Several years later, Krebs et al. astutelycommented that the dangling bonds at an unrecon-

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 9

structed GaAs(001) surface reduced the in-planesymmetry. If the surface is Ga-terminated thenthe two unsatisfied tetrahedral bonds per surfaceatom are oriented along [110], whilst for an As-terminated surface they are oriented along [110]:this breaking of the in-plane four-fold symmetryat the surface may provide an indirect means bywhich the uniaxial magnetic anisotropy may form3.Freeland et al. demonstrated modifications to theinterfacial electronic structure in Fe/GaAs(001):they found a significant charge-transfer from Fe toGaAs(001) due to interfacial bonding, and, basedon the relative electronegativities of Fe, Ga, andAs, concluded that an Fe-As bonding configurationis dominant at the interface69. These Fe-As bondswill preferentially form along the [110] direction, assuggested by Krebs et al., in order to preserve thetetrahedral bonding symmetry: Freeland et al. sug-gested that this may be the origin of the interfacialuniaxial magnetic anisotropy in Fe/GaAs(001), anidea which is accepted by many.

Semiconductor surface reconstruction: A rangeof studies using various reconstructions of theGaAs(001) surface have demonstrated the pres-ence of a uniaxial magnetic anisotropy5. Whilst theidea that As-dimers running along the [110] direc-tion of the reconstructed GaAs surface break thein-plane cubic symmetry and provide anisotropicinterfacial bonding coordination has been exten-sively put forward as an explanation for the uniax-ial magnetic anisotropy, Kneedler et al.64, and laterMoosbuler et al.70, have clearly demonstrated thatthe reconstruction of the GaAs(001) surface priorto depositing Fe makes no discernible difference tothe uniaxial magnetic anisotropy.

It is concluded that, whilst the initial growth offerromagnetic films on GaAs(001) is an interestingand important topic in itself, it seems that thesestudies are unable to shed light on the origins ofthe uniaxial magnetic anisotropy. Similarly those in-vestigations into the reconstruction of the semicon-ductor surface prior to ferromagnetic film growth,which again show little influence on the uniaxialanisotropy. The influence of epitaxial strain in sys-tems with large lattice mismatch appears to be toreduce the uniaxial anisotropy: however, epitaxialstrain does not seem to be the mechanism whichcauses the anisotropy to appear in the first-place.Finally, the anisotropic interface bonding, for ex-

ample between Fe and As, which was initially pro-posed as a potential mechanism for the anisotropy,still holds the strongest claim to be the ‘cause’ ofthe magnetic anisotropy. It seems very likely thatthe Fe-As interfacial bonding is responsible for thedirection of the uniaxial interface anisotropy: whatexactly determines the strength of the uniaxial mag-netic anisotropy is, at this point in time, still anopen question.

3.2. Theory and modelling ofhybrid structures

The interfacial magnetic anisotropy inferromagnet/III-V(001) systems is one of the fewareas of modern nanoscience where experimentalfindings are only very loosely guided by theory (atleast from the perspective of an experimentalist!).Whilst the ability to fabricate and characterizethese types of structure has rapidly emerged, thetheoretical tools required for truly ab-initio theoret-ical investigations are still lacking at present. Mag-netocrystalline anisotropy energies are usually cal-culated by applying the ‘torque theorem’, wherebythe electronic structure and total energy is calcu-lated for unconstrained magnetization, producingthe result for magnetization along the easy-axis,then the magnetization is constrained to lie alongthe hard axis and the total energy re-minimized.The magnetic anisotropy energy is the (small) dif-ference between the (large) total energies for mag-netization along hard- and easy-axes. Due to thefact that one must calculate the total energies tohigh precision in order to obtain an accurate valueof the anisotropy energy, until the advent of fairlyrecent computational advances it was not possi-ble to accurately predict the sign, never mind themagnitude, of the magnetic anisotropy energy fromfirst-principles — even in an elemental metal.

To further complicate matters, as we are in-terested in a hybrid structure, one requires amethod which is able, at the same time, to ac-curately model the electronic structure of both aferromagnetic metal and a compound semiconduc-tor. Success has been achieved by using parame-terized models, for example tight-binding, wherethe parameters for both metals and semiconduc-tors are well-known. Kosuth et al. obtained rea-sonably good agreement with experimental resultsfor Fe/GaAs(001) using a tight-binding Korringa-Kohn-Rostoker method with spin-orbit interaction,assuming an ideal Fe/GaAs(001) interface55, and

10 Aidan T. Hindmarch

attribute the anisotropy to interfacial Fe-As bond-ing. Sjostedt et al. studied Fe/ZnSe(001) using anab-initio full-potential linearized augmented-plane-wave method, finding that epitaxial strain is notrequired for uniaxial magnetic anisotropy to bepresent, and that the spin-orbit coupling at the in-terfacial Fe sites is important in realizing uniaxialmagnetic anisotropy71. Until fairly recently, theo-retical calculations have ‘only’ been able to suggestmechanisms which may (or may not) be importantin determining the uniaxial magnetic anisotropy:the increased computing power available means thatmodelling of this type of hybrid system is nowrapidly approaching the point where quantitativepredictions may be made.

4. Advances in fabrication andcharacterization

Having given an overview of the topic and discussedthe magnetic anisotropies which arise in ferromag-netic films on III-V semiconductors, we now go onto discuss some of the more important recent ad-vances in the field, with the particular aim of clar-ifying the physical mechanisms which underpin theinterface-induced uniaxial magnetic anisotropy.

4.1. Fabrication of hybridstructures

Until fairly recently the only method which hadbeen employed extensively in fabricating ferromag-net/semiconductor hybrid contacts was molecularbeam epitaxy (MBE): although this technique re-sults in extremely high-quality device structures, inrecent years there has been increasing interest informing epitaxial ferromagnetic films onto semicon-ductor surfaces by other means, which may be lesscostly and allow higher throughput. However, be-fore discussing film-deposition techniques, it seemspertinent to describe the preparation of a smooth,ordered, semiconductor surface on which to deposit.

4.1.1. Semiconductor surface preparation

As discussed above, there is a great deal of infor-mation suggesting that the exact nature of the sur-face reconstruction plays little role in determiningthe magnetic anisotropies of the ferromagnetic film:however, preparation of a smooth, clean, semicon-ductor surface onto which the ferromagnetic filmsmay be grown is of critical importance. Here we

will summarise some of the wide range of surface-preparation protocols which have been employed:further details of the surface reconstructions ofGaAs(001) may be found in reference 5.

The ideal method of fabricating a ferromag-net/semiconductor contact would involve growinga semiconductor epilayer in a dedicated III-V MBEsystem, with subsequent ferromagnetic metal filmgrowth either in the same vacuum chamber or fol-lowing a short in-vacuum transfer to a separatemetal-growth chamber. Using III-V MBE, forma-tion of a given termination and surface reconstruc-tion may be readily achieved through regulationof the substrate temperature and As flux afterdepositing the semiconductor epilayer64,72. Theseprocedures have been used successfully by somegroups9,72,73: unfortunately, such combinations ofcoupled III-V and metal deposition facilities are notalways available, and so other surface preparationschemes are often employed.

One alternate method is to grow a III-V epi-layer by MBE and then passivate the surface withan amorphous As capping layer, which protects thesemiconductor surface from oxidation during trans-fer through atmosphere. Once back under vacuum,the wafer is heated in order to thermally desorb theAs passivation, leaving a clean semiconductor sur-face on which to deposit metal48,66,74: desorption ofthe As passivation typically occurs at temperaturesin the region of 450 ◦C. This method of obtaininga clean III-V surface is particularly common wheresputter-deposition of the ferromagnetic metal is em-ployed34,56,75,76: as these types of deposition systemgenerally lack surface analysis tools, little is knownregarding the phase of the surface reconstructionsachieved. Further surface treatment may also be ap-plied, typically in the form of an in-situ thermal an-neal of the As-desorbed wafer. For example, heat-ing to 560 ◦C for 60 mins has been shown to resultin a GaAs(001)-(2× 4) reconstructed surface42. Wenote that there are, in-fact, several distinct surfacereconstructions that are described as (2 × 4), seereference 5 for further details.

By far the most popular technique for achievinga clean semiconductor surface involves using in-situetch and anneal processes on commercial epi-readyIII-V wafers. This has the obvious benefit withregard to cost, as a commercial epi-ready wafersare significantly cheaper than MBE-grown epilay-ers. Using such an etch-anneal procedure, by far themost commonly reported surface is the GaAs(001)-

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 11

(4× 6) reconstruction. This is formed by processesinvolving annealing in the region of 600 ◦C, whilstbombarding the surface with 0.5-1 keV Ar+-ionsfor a period of between 30–80 mins32,37,43,77 — asomewhat broad range of process parameters. Insome cases multiple flash-anneal steps are also in-cluded78,79. Reconstructions termed ‘(4 × 6)’ are,more correctly, in fact a combination of regions of(4×2) and (2×6) reconstructed surface phases66,80,i.e., the surface reconstruction is properly termedthe GaAs(001)-(4 × 2)/(2 × 6)5. This perhaps ex-plains the relative ease in achieving such a surfacereconstruction.

Single-phase surface reconstructions may beachieved using similar etch-anneal parameters.Godde et al. have shown that, following an ini-tial surface etch with 0.5 keV Ar+-ions for 45 minsat room-temperature, annealing at 500 ◦C for afurther 45 mins results in the GaAs(001)-(2 × 6)reconstructed surface, whilst annealing at 600 ◦Cresults in the GaAs(001)-(4 × 2) surface phase67.Lee et al. showed that, following an unspecified ini-tial etch-anneal cycle, a final anneal at 570 ◦C for30 mins also produces the GaAs(001)-(4×2) recon-struction81. In an earlier study by Moosbuhler etal., an initial etch-anneal for 30 mins at 600 ◦Cwith 1 keV Ar+-ion bombardment, then followedby 5 mins at 600 ◦C, with 0.5 keV Ar+-ion bom-bardment resulted in the GaAs(001)-(2×4) surface,whilst 5 mins at 500 ◦C, 0.5 keV Ar+ producedGaAs(001)-(2× 6)70.

We note in passing that the GaAs(110) sur-face does not have the rich variety of interface ter-minations and reconstructions which may be ob-served in the case of GaAs(001) — in-fact, due toits non-polar nature, the GaAs(110) has only thetrivial (1×1) reconstruction (with slight relaxationof the surface-layer bond angles) and forms a natu-ral cleavage-plane of the zinc-blende crystal lattice.Whilst formation of a clean, oxide-free, semiconduc-tor surface is critical both to the epitaxial growthof the FM metal overlayer and overall quality ofthe deposited structure, as discussed by Kneedler etal. and Moosbuhler et al., the magnetic anisotropyin an Fe film is independent of the GaAs(001) sur-face reconstruction onto which it is deposited. Thus,knowledge of the exact nature of the semiconductorsurface reconstruction/composition is of less signif-icance than had initially been suggested. However,details of the film growth are often of importancein determining the interface properties.

4.1.2. Ferromagnet film deposition

Since the original inception of ferromag-net/semiconductor hybrid contacts, the growthmethod of choice for the ferromagnetic metal hasbeen MBE. Deposition of the ferromagnetic metalfrom either an electron-beam hearth or thermalevaporation (Knudsen) cell typically results ingrowth rates of the order of 1 ML/min (∼ 0.05 A/s)and produced extremely high-quality epitaxial sin-gle crystalline films.

In recent years, alternative methods for deposit-ing epitaxial ferromagnet contacts onto semicon-ductors have begun to be explored. One major driv-ing factor behind this is the fact that MBE growth isboth expensive and time-consuming; consequentlymaking this technique unattractive for industrialapplication. Ion-beam sputter deposition has beenused for epitaxial growth of Fe on GaAs(001),by Damm et al.82, and on InGaAs(001) by Ri-chomme et al.47, and in both cases the structuralquality of the films, in addition to the magneticanisotropies, are found to be comparable to thoseobtained in the respective MBE-grown systems.

Formation of epitaxial Fe contacts toGaAs(001), GaAs(110), and GaAs(111) from aferrous ammonium sulphate (Fe(NH4)2SO4) elec-trolytic solution by Bao et al.83, and a compar-ison between contacts to GaAs(001) using ironchloride (FeCl2) and ferrous ammonium sulphateelectrolytes by Svedberg et al.84, have demon-strated that electrodeposition may also be a use-ful technique for ferromagnet/semiconductor con-tact formation. As electrodeposition is a very low-energy method, the formation of very abrupt in-terfaces should be possible by this technique84.Low-pressure metal organic chemical vapour de-position (LP-MOCVD) has also been successfullyused by Liu et al. to fabricate epitaxial Fe films onGaAs(001)85.

CoFe

epi-GaAs

Taglue

epi-GaAs

CoFe

2 nm2 nm

Fig. 3. Cross-sectional transition electron micrographs ofTa[25 A]/Co70Fe30[35 A] and Ta[25 A]/Co70Fe30[1100 A]

12 Aidan T. Hindmarch

films on GaAs(110), viewed along the GaAs⟨111⟩ zone axis:the scale bars are 2 nm. The Co70Fe30 film grows coherentlyon GaAs(110), despite the large surface features, and the co-herent crystal structure extends through the film thickness.(After Hindmarch et al.86.)

Magnetron sputter deposition of epitaxialCo70Fe30 films on GaAs(001)34,87 and GaAs(110)86

has been demonstrated by Hindmarch et al.: thevery small lattice mismatch between Co70Fe30 andGaAs allows a relatively high deposition rate of∼ 2 A/s to be used, with no formation of secondarycrystal phases in Co70Fe30 films over ∼ 100 nmthick: figure 3 shows a high-resolution transmis-sion electron micrograph of a Co70Fe30 film onGaAs(110), in the region of a large surface feature,demonstrating that epitaxy is maintained even inthe absence of a smooth interface. Hindmarch etal. have also demonstrated deposition of amorphousCoFeB films onto GaAs(001)56. One significant ad-vantage of sputter deposition over MBE growth isthe ability to easily apply a magnetic field to thestructure during film deposition in order to ‘set’ afurther magnetic anisotropy: in MBE the electron-beam surface analysis tools typically employed tostudy the surface structure during growth are in-trinsically incompatible with such a magnetic ‘form-ing’ field. Conversely, in sputter deposition, the lackof these surface analysis tools results in reduced in-formation regarding the semiconductor surface re-construction prior to ferromagnet film growth.

Kardasz et al.80 have performed a comparativestudy of Fe deposited onto GaAs(001) by thermalevaporation and pulsed-laser deposition (PLD). Byusing 2 ML thick 57Fe marker-layers and perform-ing conversion electron Mossbauer spectroscopy,they demonstrated that the highly-energetic PLDgrowth causes significant interdiffusion of Fe intoGaAs, with the effect of reducing the interfacialuniaxial magnetic anisotropy by roughly a factorof four in comparison to films deposited by thermalevaporation: highlighting the importance of creat-ing an abrupt, clean ferromagnet/semiconductor in-terface.

In MBE growth it was, until fairly recently,common for ferromagnetic contacts to semiconduc-tors to be grown at slightly elevated temperatures,usually in the range 50–300 ◦C. However, low-temperature MBE growth of Fe on GaAs(001) hasmore recently been used in an attempt to enhancethe interface cleanliness by Lee et al.88. They foundthat by reducing the GaAs substrate temperature

to around 130 K they could suppress both out-diffusion of As and Ga from the substrate, and in-terfacial intermixing, during Fe film growth. Lee etal.81 also show that similarly low-temperature de-posited ultra-thin (2.1-2.8 ML) Fe films exhib-ited ferromagnetism with a perpendicular magneticanisotropy. It is now far more common for low-temperature (room-temperature and below) MBE-growth to be used for ferromagnet/semiconductorcontact deposition.

Another route which has been suggested forproviding enhanced ferromagnet/semiconductor in-terface cleanliness is to use post-growth thermalprocessing: care must be taken in choosing suit-able thermal anneal conditions due to the chemi-cal reactivity and tendency for intermixing betweenFe and GaAs. As the cleanliness of the ferromag-net/semiconductor interface is not only of criti-cal importance for the interfacial uniaxial magneticanisotropy80,87, but also for spin-coherent electrontransport across the ferromagnet/semiconductor in-terface89, in recent years significant research efforthas been directed toward determining the structureand properties of the buried interface in these sys-tems, and the influence of thermal treatments.

4.2. Characterization of the buriedinterface

In the earlier literature on ferromag-net/semiconductor interface magnetism, significanteffort has been expended on investigating the de-pendence of the growth of thin Fe films, and their in-herent magnetic anisotropies, when deposited ontoGaAs(001) surfaces which have undergone a rangeof treatments in order to prepare different surfacereconstructions (see reference 5). We recall that, de-spite surface preparations which result in either As-or Ga-rich reconstructions, the consensus reachedwas that the surface reconstruction of the III-Vsemiconductor prior to ferromagnet film growthhas little, if any, influence of the resultant magneticanisotropy in the ferromagnetic film64,70: despitehaving significant influence on the growth of theferromagnetic film. It is perhaps not so outrageousa supposition, then, to suggest that the structureof the buried Fe/GaAs(001) interface may be of fargreater importance than the initial GaAs(001) sur-face prior to ferromagnet film deposition in deter-mining the physical properties of the ferromagneticoverlayer.

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 13

4.2.1. Atomic arrangement at theferromagnet/semiconductorinterface

One must consider that the deposition of the fer-romagnetic film also has an influence on the un-derlying semiconductor surface: promoting atomicoutdiffusion primarily of As, but also Ga to a lesserextent. From a theoretical point of view, Erwin etal. suggested that Fe adsorption causes the charac-teristic As-dimers of the reconstructed GaAs(001)surface to become unstable90. With this in mind,density functional theory calculations, carried outby Demchenko and Liu, assuming an interface struc-ture based on the GaAs(001)-(1 × 1) reconstruc-tion, suggest that, due to the preference for Fe-As over Fe-Ga interfacial bonding, As-terminatedFe/GaAs(001) favours an atomically abrupt inter-face whilst the Ga-terminated structure favoursan intermixed interface region over several atomicplanes91. Driven by these theoretical predictions,efforts to unambiguously determine the detailedatomic structure of the buried Fe/GaAs(001) inter-face experimentally have begun.

From a combination of x-ray photoelectronspectroscopy, reflection high-energy electron diffrac-tion, and photoelectron diffraction, Schieffer et al.have suggest the formation of a bcc-Fe-based substi-tutional alloy during the first 4 ML of Fe growth onthe GaAs(001)-(4× 2) surface, followed by pure-Fegrowth thereafter92. Whilst spectroscopy and scat-tering techniques prove useful, it is clear that micro-scopic imaging techniques are preferable for precisedetermination of the interface structure.

Cross-sectional transmission electron mi-croscopy, using both the high-resolution (HRTEM)and high-angle annular-dark-field (HAADF) modesof operation, has been used to measure the inter-face structure in both the Fe/AlGaAs(001)93 andFe/GaAs(001) systems72. For Fe/AlGaAs(001),Zega et al. found that a post-growth anneal for10 minutes at 200 ◦C both reduced the intermixedinterface width and increased the injected spin-polarization. Using HAADF (Z-contrast) imaging,a technique which is chemically sensitive at highspatial resolution, a model interface structure con-sisting of As-terminated AlGaAs-(1×1) with a sin-gle intermixed ML of Fe and As at the interface wasproposed93. However, this study consisted only ofcross-sectional images of the interface viewed alongthe in-plane AlGaAs[110] direction: Lebeau et al.extended this study by taking HAADF images along

both the GaAs[110] and [110] directions in order tounambiguously determine the interface structurein a Fe/GaAs(001) contact, after annealing for 1hour at 200 ◦C. They found a subtly different,novel interface structure, shown in figure 4, con-sisting of As-terminated GaAs(001)-(1 × 1) and apartially-occupied, non-intermixed, interfacial Felayer, terminating at the interface-plane.

Further studies on the influence of ther-mal anneal processing on the structure of theFe/GaAs(001) interface by Schultz et al. haveshown that temperatures in the region of 400-450 ◦Cpromote the diffusion of both Fe into GaAs (form-ing Fe2As clusters) and Ga into Fe (forming an ad-ditional Fe3Ga interface layer). The formation ofthese two distinct stable binary phases due to in-terdiffusion occurs preferentially to the anticipatedstable Fe3Ga2−xAsx ternary alloy phase due to Febeing more weakly diffusive that Ga or As in thissystem94. Schultz et al. have also demonstrated theinfluence of such an interfacial Fe3Ga alloy, in thiscase formed during elevated-temperature growth,on the spin-injection from Fe into GaAs; showinga reversal in the sign of the injected spin-polarizedcurrent from majority- to minority-spin at a growthtemperature of 175 ◦C. Further post-depositionthermal treatment, for 1 hour at 250 ◦C results inthe sign of the injected spin-polarization revertingto majority-spin89, demonstrative of atomic rear-rangement at the Fe/GaAs(001) interface.

The influence of thermal annealing on the in-terfacial magnetic anisotropy has been studied byBianco et al.33, who demonstrated an enhancementin the interfacial uniaxial magnetic anisotropy inCo69Fe31/GaAs(001) by a factor of three after an-nealing at 200 ◦C for 10 mins. They show that theanneal treatment does not result in a significantchange in strain, and attribute the enhanced uniax-ial magnetic anisotropy to improvements in inter-facial bonding, i.e. atomic rearrangement, broughtabout by the thermal treatment. One may antici-pate different interfacial alloys to form due to in-terdiffusion at the CoFe/GaAs(001) interface, dueto the increased diffusivity of Co compared to Fein GaAs. Given that the magnetic anisotropy andspin-injection efficiency depend critically on theatomic structure at the interface, it seems perti-nent also to consider interface magnetism in thesestructures.

14 Aidan T. Hindmarch

Fig. 4. Left: high-angle annular-dark-field images of epitaxial Fe/GaAs(001), viewed along (a) GaAs[110], and (b) GaAs[110]directions. Right: model structure for the Fe/GaAs(001) interface, showing a partially-occupied, non-intermixed, Fe interfacelayer. (Reprinted with permission from LeBeau et al.72. Copyright (2008), American Institute of Physics.)

4.2.2. Ferromagnet/semiconductorinterface magnetism

Early studies of Fe on GaAs(001) suggested that,for ultra-thin films, the onset of room-temperatureferromagnetism occurs at higher Fe film thicknessthan may be anticipated from Fe growth on met-als. For example, Kneedler et al. find ferromag-netism only for Fe films thicker that 6 ML (grownon GaAs(001)-(2 × 4) and GaAs(001)-c(4 × 4))64,whilst Xu et al. showed that their Fe films (grownon GaAs(001)-(4 × 6)) were non-magnetic below3.5 ML, consisted of superparamagnetic islands be-tween 3.5 and 5 ML, and were ferromagnetic above5 ML95. Such a magnetically ‘dead’ layer at theinterface is quite apparently detrimental for spin-injection applications. Indeed, for many years, sug-gestions were made that the interfacial magneticdead-layer found in Fe/GaAs(001) structures maynot, in-fact, be an intrinsic feature of the interfacebetween Fe and GaAs(001).

The first demonstration that it is possible foran Fe film to possess its full magnetic moment rightup to the interface with GaAs(001) was providedby Claydon et al.96 using x-ray magnetic circulardichroism (XMCD) spectroscopy97. They depositedthin Fe films (0.25 – 18 ML) where the thinnestfilms were buried below Co in order to remainferromagnetic. Utilizing the element specificity ofXMCD, the orbital- and spin-components of theFe magnetic moments were extracted, demonstrat-

ing a bulk-like value for the Fe spin magnetic mo-ment and a large enhancement in the orbital mag-netic moment due to the reduced symmetry at theinterface. (Bulk-like Fe magnetic moments at theFe/GaAs(001) interface were later theoretically pre-dicted by calculations of Demchenko and Liu91, andTobin et al. have used spin-resolved photoelectronspectroscopy to also reveal bulk-like magnetism atthe Fe/GaAs(001) interface, despite significant in-termixing74.)

A similar experimental method, again employ-ing the element-specificity of XMCD, has been usedby Giovanelli et al.78,79, who have studied the spinand orbital magnetic-moments in a 0.5 ML Comarker layer, as its position is moved through a 6ML Fe film on GaAs(001). They find that the spinmagnetic moment of the Co marker layer is highestin the center of the film, and drops at both theGaAs and vacuum interfaces. A large enhancementin the orbital magnetic moment is again found closeto the interface with GaAs(001), as was found byClaydon et al.96. As the films employed by bothgroups were deposited onto the GaAs(001)-(4×6)surface, the reason for the discrepancy in behaviourof the interfacial spin magnetic moment is unknown,and may again be related to a different behaviourof Co and Fe atomic species at the interface withGaAs(001).

Whilst photoelectron spectroscopies are pow-

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 15

erful techniques, e.g., for determining element-specific magnetic properties, they are less suitablefor obtaining depth-selective magnetic information.In order to obtain such information, scatteringtechniques are commonly employed. The applica-bility of soft-x-ray resonant magnetic (exchange)scattering (XRMS) to study interface magnetismin ferromagnet/semiconductor contacts has beendemonstrated, for example, by Hindmarch et al.87.They used off-specular XRMS to record element-specific rocking curves, and compared the mag-netic disorder with the structural disorder: show-ing that, even under a saturating magnetic field,interfacial magnetic disorder was present in aCo70Fe30/Al10Ga90As(001) contact. The presenceof interfacial magnetic disorder in these structureswas correlated with a reduction in the strength ofthe interfacial uniaxial magnetic anisotropy. Reso-nant x-ray techniques are useful due to their ele-ment specificity. However, quantitative analysis ofXRMS is difficult, due in no small part to the in-direct nature of the spin-photon interaction uponwhich it relies. The far more direct interaction ofthe neutron with the local magnetization makes po-larized neutron reflectivity (PNR)98 an attractivetechnique from which to obtain quantitative, depth-selective information on the magnetization of thin-film structures99.

Magnetic depth profiling studies of epitax-ial CoFe/GaAs(001) structures has been reportedby Park et al. using PNR100. They found that,for an epitaxial CoFe[20 nm] film grown on theGaAs(001)-(2 × 4) reconstructed surface at 95 ◦C,the resultant ‘magnetic thickness’ was reduced fromthe structural thickness of the CoFe film by around6 A; indicative again of a magnetic dead-layer at theCoFe/GaAs(001) interface — the dead-layer thick-ness in CoFe is somewhat thinner than those typi-cally found in Fe on GaAs(001), again suggesting adifferent behaviour of Co and Fe elemental speciesat the GaAs(001) interface. Park et al. have fur-ther used the PNR technique to investigate the in-fluences of both growth- and anneal-temperaturesin epitaxial CoFe, again deposited onto GaAs(001)-(2 × 4)101. They found that reducing the temper-ature of the CoFe growth resulted in improved in-terface sharpness and, for CoFe growth at −15 ◦C,resulted in an interface free of any degraded mag-netically dead layer. They also found that, for sam-ples grown at any temperature (within the studiedrange of −15–175 ◦C), anneal processing at 250 ◦C

for 1 hour resulted in an 11 A thick interfacial dead-layer forming: showing that annealing at elevatedtemperatures and long times can be highly detri-mental to the interfacial magnetic properties of thecontact.

In an interesting twist on the discussion of theusual quasistatic magnetization reversal, Zhao etal. have demonstrated that the magnetization inthe interfacial Fe layer appears decoupled from thebulk of the film: it appears to reverse as if pos-sessing only uniaxial magnetic anisotropy, whilstthe remainder of the film appears to reverse asif acted on by a admixture of uniaxial and cu-bic anisotropies45. They used a combination of(bulk-sensitive) magneto-optical Kerr effect and(interface-sensitive) magnetization-induced second-harmonic generation magnetometry techniques onFe/AlGaAs(001) structures to show that, despitebeing coupled together via the strong exchangefield in Fe, in-plane angles in the range 40◦– 85◦

may open between interfacial and bulk magneticmoments. Using time-resolved variants of thesetechniques, Zhao et al. have further determinedthe angular-dependence of the damping parame-ter in Fe/AlGaAs(001) from time-resolved Kerr ef-fect magnetometry102, and investigated the preces-sional dynamics of Fe moments at the interfacewith AlGaAs(001) using the time-resolved second-harmonic generation103. These surprising and unan-ticipated demonstrations that the interfacial andbulk magnetization, and their respective dynamicalbehaviours, are decoupled to such an extent, clearlyshows that there is much still to be learned aboutinterface magnetism in ferromagnet/semiconductorhybrid structures.

5. Tailoring the free-energy density

In order to gain further insight into themagnetic anisotropy arising at the ferromag-net/semiconductor interface, it is instructive tofind ways in which one may controllably mod-ify the angular-dependence of the free-energy den-sity, equation 3, in a ferromagnetic thin-filmon GaAs(001). The full expression for the free-energy density ε for a (Co-rich) CoFe film con-sisting of terms relating to the cubic magne-tocrystalline anisotropy, magnetoelastic anisotropy,interfacial-induced uniaxial magnetic anisotropy,volume magnetization-induced uniaxial magnetic

16 Aidan T. Hindmarch

anisotropy104, and Zeeman energy, may be written

ε (ϕ) ∼ −Keff14 sin2 (2ϕ)− B2ϵ

2 sin2 (2ϕ)

+Kint

Ut sin2 (ϕ) +Kvol

U sin2 (β)

−HMS cos (ϕ− α) , (8)

where α is the angle between the applied magneticfield and the GaAs[110] during measurement, and βis the angle between the applied (saturating) mag-netic field and the GaAs[110] during film deposition.

Equation 8 contains numerous terms of differ-ing symmetry and origin, making quantitative anal-ysis difficult: a problem which has previously hin-dered advance in unravelling the details of the ferro-magnet/semiconductor interface anisotropy. How-ever, by judicious choice of ferromagnetic thin-filmmaterial, one is able to significantly simplify equa-tion 8 by ‘removing’ various terms in the free-energy density; thereby making the physical ori-gins of the various magnetic anisotropy contribu-tions more easily experimentally accessible.

5.1. Removal of the magnetoelasticanisotropy term

As discussed in section 3.1, one mechanism whichhas been extensively suggested as underpinning theuniaxial interfacial magnetic anisotropy is epitax-ial strain in the ferromagnetic film. Both in ex-perimental work reported by Thomas et al.41 andby Bianco et al.33, and in previous theoreticalwork presented by Sjostedt et al.71, evidence tothe contrary has been provided: uniaxial magneticanisotropy arises in the absence of epitaxial strain.

We would thus like to find a ferromagneticmetal where the lattice mismatch with GaAs, andhence epitaxial strain, is negligible. Although it iscommon in the literature to disregard the mag-netoelastic term, for Fe the lattice-constant ofd = 2.86 A results in epitaxial strain of around1.4 % when deposited on GaAs (table 2). Epitax-ial Co70Fe30 films have negligible lattice mismatchwith GaAs: for Co70Fe30 d ≈ 2.835 A29, hence whenusing Co70Fe30, the magnetoelastic term is indeednegligible. For this CoFe composition the second-order cubic anisotropy constant K2 is negligible andthe omission of this term also is justified28.

5.1.1. Interfacial magnetic anisotropy inepitaxial films

One obvious question to ask is whether we see aneffect by changing the III-V semiconductor mate-

rial, keeping all other factors constant. In order thatthe magnetoelastic term in equation 8 remain neg-ligible, we have compared Co70Fe30 films depositedon GaAs(001) and Al30Ga70As(001) surfaces. In theabsence of an applied magnetic field during film de-position, equation 8 reduces to equation 4, the ex-pression for the free-energy density which is com-monly applied to epitaxial CoFe on GaAs(001),

ε (ϕ) ∼ −Keff14 sin2 (2ϕ) +

KintUt sin2 (ϕ)

−HMS cos (ϕ− α) .

Figure 5 shows magneto-optical Kerr-effecthysteresis loops along the uniaxial hard-axisfor Co70Fe30 films on GaAs(001) (a) andAl30Ga70As(001) (b): we are clearly able to ob-serve interfacial uniaxial magnetic anisotropy inepitaxial sputtered Co70Fe30 films. The effectiveanisotropy constants are Keff

1 ≈ −2.8 × 104 J/m3

and KeffU = K int

U /t ≈ 1.5 × 104 J/m3 for Co70Fe30on both GaAs(001) and Al30Ga70As(001).

-4

-3

-2

-1

0-6

-5

-4

-3

-2

-1

0

-150 -100 -50 0 50 100 150

-1.0

-0.5

0.0

0.5

1.0

780 790 800 810

-1.5

-1.0

-0.5

0.0

0.5

XM

CD

(Arb

.)

Photon energy (eV)

d) ml / ms = 0.21

Inte

grat

ed X

MC

D (A

rb.)

-2.0

-1.5

-1.0

-0.5

0.0

0.5

c) ml / ms = 0.21-1.0

-0.5

0.0

0.5

1.0

Nor

mal

ized

mag

netiz

atio

n

CoFe/GaAs(001)

a)

b)

Applied field (Oe)

CoFe/AlGaAs(001)

Fig. 5. MOKE hysteresis loops along the uniaxial hard axisfor Co70Fe30[3.5 nm] on (a) GaAs and (b) Al30Ga70As, andCo-edge XMCD spectra for Co70Fe30 on (c) GaAs and (d)Al30Ga70As. Similar magnetic anisotropies are obtained forboth films and the orbital:spin magnetic moment ratio ex-tracted from sum-rule analysis of the XMCD spectra areidentical.

XMCD spectra around the Co LII and LIII ab-sorption edges are also shown in figure 5 for filmson GaAs (c) and Al30Ga70As (d) with magnetiza-tion along the uniaxial easy-axis. For both films theratio of orbital to spin magnetic moments are ex-tracted using sum-rule analysis105, and are foundto be ml/ms = 0.21 for Co, and ml/ms = 0.11for Fe (not shown), apparently independent of theunderlying semiconductor material. For a ferromag-net with only uniaxial anisotropy, ml/ms measured

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 17

along the uniaxial easy axis should be proportionalto the degree of anisotropy in ml, and hence to theeffective uniaxial magnetic anisotropy25,27. How-ever, in this case there is also a volume cubic mag-netocrystalline anisotropy with similar strength tothe interface induced uniaxial magnetic anisotropy:the measured anisotropy inml is due predominantlyto the magnetocrystalline anisotropy, owing to theshort penetration-depth of electron-yield XMCD.

One could, at this point, conclude that the un-derlying compound semiconductor material playsno role in the uniaxial magnetic anisotropy in theabsence of epitaxial strain. Rather, we suggest thatit is difficult to accurately determine the effect ofthe substrate material on the uniaxial magneticanisotropy in epitaxial films due to the interplaywith the cubic magnetocrystalline anisotropy: asthe ml/ms are biased toward the volume ratherthan interfacial moments, one may again obtain lit-tle information on the origin of the interfacial uni-axial magnetic anisotropy from XMCD on epitaxialferromagnetic films on GaAs(001), and a subtly dif-ferent approach is necessary.

5.1.2. Modification of the volume uniaxialanisotropy term

Before continuing to further simplify the expres-sion for the free-energy density, we first consider theeffects of reintroducing the volume magnetization-induced uniaxial magnetic anisotropy term to equa-tion 4, giving

ε (ϕ) ∼ −Keff14 sin2 (2ϕ) +

KintUt sin2 (ϕ)

+KvolU sin2 (β)−HMS cos (ϕ− α) . (9)

It then appears a trivial matter to separate thecontributions from magnetization- and interface-induced uniaxial magnetic anisotropies from eitherthe thickness, t, or angular, ϕ and/or α, depen-dence.

Magneto-optical Kerr-effect hysteresis loops fora series of thin Co70Fe30 films on GaAs(001) areshown in figure 6a), which have been sputter-deposited in either zero applied magnetic field, orwith a saturating magnetic field, HDep, appliedalong either the easy- or hard-axis of the interfacialuniaxial magnetic anisotropy ([110] or [110]): the ef-fective magnetic anisotropy constants are given intable 3. It is found that applying a magnetic fieldalong the [110] direction during film growth has anegligible effect on the magnetic anisotropy, whilst

applying a magnetic field along [110], the hard-axisof the interface-induced anisotropy, results in an in-crease in the effective uniaxial magnetic anisotropy.

The reason for this may be revealed fromXMCD spectroscopy: figures 6b)-d) show XMCDspectra around the Co and Fe LII and LIII absorp-tion edges for these films. Significant differences inthe ratio of orbital-to-spin magnetic moments be-tween films deposited with and without an appliedmagnetic field are observed. Applying the deposi-tion field along the interfacial uniaxial easy-axisproduces an increase in the component of the ra-tio of the atomic species resolved orbital and spinmagnetic moments, ml/ms, directed along the uni-axial easy axis of magnetization, whilst applying thedeposition field along the uniaxial hard-axis causesa decrease, as summarized in table 3. The ratioml/ms along the easy-axis may be taken as rep-resentative of the degree of anisotropy in the or-bital magnetic moment, and hence as a measureof the magnetic anisotropy itself27. The spin-orbitcontribution to the uniaxial magnetic anisotropy is∆ESO ∝ −ml ·ms ∼ −ml

34; hence the depositionfield changes the maximal projection of the atomicorbital magnetic moments onto the easy-axis, whichcorresponds to a deposition-field-induced shift inthe free-energy landscape via a modification of thecontribution of the spin-orbit interaction to the to-tal energy.

5.2. Removal of themagnetocrystalline anisotropyterm

It is clear that the spin-orbit interaction plays apivotal role in determining the magnetic anisotropyin epitaxial ferromagnetic films: in a crystallineferromagnetic film the orbital moment contributesboth to the uniaxial and cubic magnetocrystallineanisotropies. Due to this fact, the origin of theinterface-induced uniaxial magnetic anisotropy ispartially obscured — in order to disentangle theseinfluences, crystalline ferromagnets must be aban-doned.

Using an amorphous ferromagnet, i.e. one hav-ing no long-range crystal symmetry, it is possible toisolate the interfacial magnetic anisotropy, by en-tirely removing the magnetocrystalline anisotropycontribution. The expression for the free-energydensity then reduces to the trivial form

ε (ϕ) ∼K int

U

tsin2 (ϕ)−HMS cos (ϕ− α) : (10)

18 Aidan T. Hindmarch

-150 -100 -50 0 50 100 150-3

-2

-1

0

1

-1.5

-1.0

-0.5

0.0

0.5

1.0

ml / m

s = 0.11

ml / m

s = 0.21

(b) Hdep

= 0 L2 edge

L3 edge

775 780 785 790 795 800 805-2.0

-1.5

-1.0

-0.5

0.0

0.5

Fe

Co ml / m

s = 0.10

Energy (eV)

ml / m

s = 0.14

(d) Hdep

UHA

-1.5

-1.0

-0.5

0.0

0.5

ml / m

s = 0.18

SXM

CD

ml / m

s = 0.27

(c) Hdep

// UEA

-0.75-0.50-0.250.000.250.50

705 710 715 720 725

-0.75-0.50-0.250.000.250.50

705 710 715 720 725

-0.75-0.50-0.250.000.250.50

705 710 715 720 725 HDep = 0

HDep // UEA HDep // UHA

Applied field (Oe)

Nor

mal

ized

mag

netiz

atio

n

(a)

Fig. 6. (Color online) Room temperature magnetic hysteresis loops (a) for epitaxial Co70Fe30[3.5 nm] films on GaAs(001)deposited with no applied field and magnetic fields parallel to the interfacial uniaxial easy- and hard-axes. XMCD spectraaround the Co (main) and Fe (inset) LII and LIII edges for (b) zero magnetic field, (c) magnetic field applied along the uniaxialeasy axis, and (d) magnetic field applied along the uniaxial hard axis. The orbital-to-spin magnetic-moment ratios ml/ms

extracted from sum-rule analysis are indicated. (After Hindmarch et al.34.)

Table 3. Summary of the effective uniaxial and cubic magneticanisotropy constants (equation 2), and XMCD ml/ms ratios alongthe uniaxial easy-axis, for epitaxial Co70Fe30[3.5 nm] films onGaAs(001). (After Hindmarch et al.34.)

HDep = 0 HDep // UEA HDep // UHA

KeffU (×105 erg/cc) 1.5± 0.1 1.4± 0.1 2.1± 0.1

Keff1 (×105 erg/cc) −2.8± 0.1 −2.8± 0.1 −2.2± 0.1

Keff1 /Keff

U −1.9± 0.1 −2.0± 0.1 −1.05± 0.1

ml/ms (Co) 0.21± 0.02 0.27± 0.02 0.14± 0.02ml/ms (Fe) 0.11± 0.02 0.18± 0.02 0.10± 0.02

no cubic magnetocrystalline anisotropy is present,and the origin of the uniaxial magnetic anisotropyshould be more easily accessible experimentally56.

An amorphous ferromagnet may be producedby alloying dilute metalloid into the ferromag-netic metal matrix: addition of ∼20% metalloid(for example, boron) in the ferromagnetic transi-tion metal alloy, CoFe, slightly reduces the Curietemperature and saturation magnetization rela-tive to CoFe, whilst at the same time destroyinglong-range crystalline order. HRTEM micrographsof Co40Fe40B20[3.5 nm] films on GaAs(001) and

Al30Ga70As(001) surfaces are shown in figure 7:the clear crystallinity of the underlying semicon-ductor terminates abruptly at the interface with theCoFeB layer, which is amorphous. For films on theGaAs(001) substrate, a surface corrugation is ob-served when viewed along the [110] direction, fig-ure 7b), with amplitude ∆ and period ξ indicated.A longer period corrugation, ξ ∼ 100s nm, is foundat the surface of Al30Ga70As(001).

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 19

5.2.1. Interfacial magnetic anisotropy inamorphous films

To demonstrate that the interfacial uniaxialmagnetic anisotropy is still present for amor-phous ferromagnetic films on III-V(001) surfaces,magneto-optical Kerr-effect hysteresis loops forCo40Fe40B20[3.5 nm] films on GaAs(001) andAl30Ga70As(001) are shown in figure 8. In this casethe films were deposited in an applied magneticfield along the GaAs[110] direction (adding a vol-ume uniaxial magnetic anisotropy term to the free-energy density for each film): the magnetic fieldduring measurement is applied either parallel orperpendicular to the deposition field. Strong uni-axial magnetic anisotropy is observed in both films,and the uniaxial easy-axes are perpendicular to theapplied deposition field direction: indicating thatthe uniaxial magnetic anisotropy is not simply ofa magnetization-induced origin.

50 Å 50 Å

50 Å 50 ÅGaAs

AlGaAs

CoFeB

CoFeB

Ta

Ta

TaOx

TaOx

Hole / glue

gluea) b)

c) d)

ξ ∆UHA

UHA

UEA

UEA

HDep

HDep

HDep

HDep

X

X

Fig. 7. Cross-sectional HRTEM mi-crographs of Co40Fe40B20/GaAs/Al30Ga70As(001), a) & b),and Co40Fe40B20/Al30Ga70As(001), c) & d), contacts. Thegrowth field direction and uniaxial magnetic anisotropy axesare indicated in each frame. The corrugation period ξ andamplitude ∆ are indicated in frame b). (After Hindmarch etal.56.)

-1.0

-0.5

0.0

0.5

1.0

Nor

mal

ized

mag

netiz

atio

n

Al0.3Ga0.7As/GaAs/Co0.4Fe0.4B0.2

H // HDep

a) b)

H HDep

-75 -50 -25 0 25 50

-1.0

-0.5

0.0

0.5

1.0

d)c)

Al0.3Ga0.7As/Co0.4Fe0.4B0.2

H // HDep

Applied field (Oe)-50 -25 0 25 50 75

H HDep

Fig. 8. Hysteresis loops of Co40Fe40B20[3.5 nm] films onGaAs(001), a) & b), and Al30Ga70As(001), c) & d). The fieldis applied parallel or perpendicular to the deposition field, asindicated. The solid red line in frame a) is a fit using theStoner-Wohlfarth model for the free-energy density, based onequation 10. (After Hindmarch et al.56.)

Co40Fe40B20 films on GaAs(001) exhibit anentirely reversible magnetization reversal by co-herent rotation for fields applied along the hard-axis; whereas for Al30Ga70As(001), hard-axis re-versal takes place partially by domain wall mo-tion. The uniaxial anisotropy field is also reducedfor Al30Ga70As(001), indicating a stronger uniaxialmagnetic anisotropy in structures with GaAs(001).The reason for the difference in reversal-mechanismmay be explained due to the surface corrugation:one may calculate a Neel-type domain-wall width30

of D ∼ 150 nm for these Co40Fe40B20 films: it seemsreasonable that the substrate surface corrugation isable to influence domain-walls for Co40Fe40B20 onAl30Ga70As(001), where D ∼ ξ. Fortuitously, theorientation of the surface corrugation relative thethe easy-axis of the uniaxial magnetic anisotropyallows step-edge anisotropies also to be ruled outas contributory factors in determining the magneticanisotropy56. Indeed, the only mechanism whichmay plausibly explain both the direction and mag-nitude of the uniaxial magnetic anisotropy is theinterface-interaction between ferromagnetic metalsand III-V(001) semiconductors; despite the lack ofcrystal symmetry in the Co40Fe40B20 films.

The important question which must be an-swered is, what determines the difference instrength of the uniaxial magnetic anisotropy inAl30Ga70As compared to GaAs? It is well knownthat orbital angular momentum plays a dominantrole in determining the strength of magnetocrys-

20 Aidan T. Hindmarch

talline anisotropy; applying XMCD spectroscopy tothese films allows determination of the orbital- andspin-resolved components of the atomic magneticmoments in both Co and Fe atomic species, demon-strating that the different semiconductor alloy sur-faces influence the uniaxial magnetic anisotropythrough the orbital component of the atomic mag-netic moments.

Interfacial bonding, long thought to be im-portant in determining the interfacial magneticanisotropy, is found not, in-fact, to be the wholestory: no significant difference in interface bond-ing charge transfer between Co40Fe40B20 films onAl30Ga70As(001) and GaAs(001) at either Co orFe sites, from the integrated absorption spectra69.This may be anticipated as the fraction of interfacialAs sites is unchanged — whilst interfacial bondingmay be important in determining the directionality,another factor must be at play in determining thestrength of the uniaxial magnetic anisotropy.

Table 4. Ratios of the orbital–to-spin magnetic moments for Feand Co atomic species in amorphousCo40Fe40B20[3.5 nm] films on GaAs(001)and Al30Ga70As(001) as measured byXMCD. (After Hindmarch et al.56.)

Species GaAs Al30Ga70As

Fe ml/ms = 0.45 ml/ms = 0.34Co ml/ms = 0.38 ml/ms = 0.19

The results of sum-rule analysis of the recordedXMCD spectra for Co40Fe40B20 films on GaAs(001)and Al30Ga70As(001) are shown in table 4. Inall cases moments are enhanced over those forCoFe alloys (table 3) and elemental metals (forbulk elemental Fe ml/ms = 0.043, and for Coml/ms = 0.099105) due the the reduced orbitalmoment quenching arising from the lack of crystalsymmetry in CoFeB. The observed differences be-tween the ml/ms ratios for films on GaAs(001) andAl30Ga70As(001) are unrelated to interfacial bond-ing, and, due to the electronic structure of CoFe(B)alloys, cannot arise due to charge transfer via inter-nal bonding and back-donation interactions56,106.

Additional enhancement in the atomic orbitalmoment can occur in systems which exhibit a spin-reorientation transition, where the enhancement isrelated to a non-local spin-orbit interaction whichmodifies the anisotropy in the magnetic layer due

to the strongly spin-orbit coupled adjacent ma-terial, e.g., Co/Au or Co/Pt107. The criteria forthis enhancement to occur are that there shouldbe strong spin-orbit coupling in the adjacent ma-terial, and, strong wavefunction overlap across theinterface,i.e., strong interfacial bonding107. As dis-cussed previously, the bonding across the ferro-magnet/GaAs(001) interface is well-known to bestrong, and to be capable of influencing the elec-tronic structure in the ferromagnetic metal69: and,as the spin-orbit interaction parameters in Au andPt are ∼ 0.005 eV, compound semiconductors havefar stronger valence spin-orbit coupling (Table 2)than these elemental metals.

Comparing the spin-orbit coupling strength inGaAs (∆0 = 0.34 eV) and the series of AlGaAs al-loys (0.3 ≤ ∆0 ≤ 0.34 eV), we see that adding Al toGaAs causes a reduction in the valence spin-orbitcoupling. Were the uniaxial magnetic anisotropy inthe ferromagnetic overlayer to be related to the non-local spin-orbit interaction, we would anticipate aweaker uniaxial magnetic anisotropy for films onAl30Ga70As(001) than for films on GaAs(001): asis found in these CoFeB films, and also by Za-keri et al. for epitaxial Fe/InP(001) (which has∆0 = 0.11 eV, table 2). Whilst interfacial-bondingis crucially important, both in determining the di-rection of the uniaxial magnetic anisotropy andin allowing the non-local spin-orbit interaction tocouple to the ferromagnetic film, the observed dif-ferences in the strength of the uniaxial magneticanisotropy, figure 8, and species-resolved orbitalmagnetic moments, table 4, seems to point towardthe spin-orbit coupling strength in the semiconduc-tor playing a role in producing the interface-induceduniaxial magnetic anisotropy in ferromagnetic filmson the (001) surface of III-V compound semicon-ductors.

In this section we have discussed several meansby which the in-plane magnetic anisotropies inferromagnet-III-V(001) hybrid contacts may bemodified, and suggest a possible explanation forthe origin of the interfacial uniaxial magneticanisotropy. Despite these recent advances, there ismuch work still to be done in fully understandingthe intriguing magnetic behavour exhibited by thisclass of structures.

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 21

6. Perspectives

One of the so-called ‘grand challenges’ in nanomag-netism today is to harness the ability to manipu-late the magnetization direction using an electricfield. The electric-field strength predicted to be re-quired in order to significantly modify the magneticanisotropy in Fe monolayer has been calculated tobe up to 1 V/A (108 V/cm)108 — a very strongelectric field in comparison to the dielectric break-down threshold of GaAs. However, Ohta et al. usedmagneto-optical measurements under an AC back-gate voltage applied across the Schottky barrier ina thin Fe film on n-GaAs(001) with in-plane magne-tization109, and showed a change in Kerr-ellipticitywith application of a comparatively weak electricfield: they applied a voltage of 1 V across a Schot-tky barrier with n = 5×1017 cm−3 (depletion width∼ 24 nm), i.e., an electric field of order 0.005 V/A— comparable to the dielectric breakdown thresh-old6. However, other groups have been unable toverify this result using other magnetometry tech-niques, suggesting that ‘parasitic’ electro-optic orpiezoelectric effects may play a more significant rolethan originally suggested in reference 109. A suc-cessful method for obtaining voltage-control of aferromagnet is to indirectly couple the voltage tothe magnetization via mechanical strain: applyinga voltage to a piezo-stressor attached to the film re-sults in a macroscopic strain of ∼ 10−5 V−1: signif-icant changes in anisotropy may be achieved withrelatively modest voltages110 using the Villari ef-fect111. Recently, Nolting et al. have demonstratedthat the x-ray magnetic linear dichroism techniquemay be used to attempt to distinguish spin-orbitand strain effects in Fe/GaAs(001) structures112.

An area not touched on so-far in this article hasbeen the incorporation of a ferromagnetic semicon-ductor into the hybrid structure, the archetypalexample being GaMnAs. The major stumbling-block to application of these materials is thelow Curie-temperatures, below room-temperature,which make spintronic devices based on dilute mag-netic semiconductors impractical for most applica-tions. However, Maccherozzi et al. found that inepitaxial Fe/GaMnAs/GaAs(001) structures, aninterfacial proximity effect allows the exchange in-teraction in the Fe film to stabilize interfacial fer-romagnetism in GaMnAs to room temperature.Olejnik et al. found that, in a similar structure,the Fe layer causes an exchange-bias effect on theGaMnAs layer when cooled through its Curie tem-

perature: the direction of the exchange-bias is con-trolled by the direction in which the Fe film ismagnetized during cooling73. Mark et al. depositedNi80Fe20/GaMnAs/GaAs(001) films, and found anabsence of exchange coupling across the interfacein these materials; in effect producing an abrupt,and tuneable, magnetization gradient at the inter-face between Ni80Fe20 and GaMnAs. These resultsshow great potential for enabling room-temperatureoperation of spintronic devices based on dilute mag-netic semiconductors.

The topic of ferromagnetic metal contacts tocompound semiconductors has come a long waysince its humble beginnings: in 1983 it wouldhave been impossible to augur the impact and ex-tent of the field as it stands at the present day.Whilst a great deal of progress has been made,there are many aspects of the behaviour of hy-brid ferromagnet-semiconductor structures whichare not well understood, and many exciting discov-eries still to be made — although, of course, onecannot predict the nature and potential impact ofthese discoveries, even in the short-term.

Acknowledgements

The author wishes to thank the many peo-ple who have contributed to the work pre-sented here; including C.H. Marrows, D.A. Arena,B.J. Hickey, C.J. Kinane, M. Henini, A.K. Suszka,K.J. Dempsey, D. Taylor, J. Dvorak, M. MacKen-zie, and J.N. Chapman. Interesting and fruit-ful discussions with B.L Gallagher, A.W. Rush-forth, K.W. Edmonds, R.P. Campion, C.T. Foxon,T. Jungwirth, J.A. Haigh, J.S. Claydon, M. Ali,D. Ciudad, and G. Burnell are also gratefully ac-knowledged, as is financial support under EU GrantNo. 214499 NAMASTE.

References

1. J. R. Waldrop and R. W. Grant, Interface chem-istry of metal-GaAs Schottky-barrier contacts,Appl. Phys. Lett. 34, 630 (1979).

2. W. Jantz, G. Rupp, R. Smith, W. Wettling and G.Bayreuther, Investigation of single crystal Fe filmsgrown by MBE on GaAs substrates, IEEE Trans.Mag. 19, 1859 (1983).

3. J. J. Krebs, B. T. Jonker and G. A. Prinz, Proper-ties of Fe single-crystal films grown on (100)GaAsby molecular-beam epitaxy, J. Appl. Phys. 61, 2596(1987).

22 Aidan T. Hindmarch

4. G. A. Prinz, Hybrid ferromagnetic-semiconductorstructures, Science 250, 1092 (1990).

5. G. Wastlbauer and J. A. C. Bland, Structuraland magnetic properties of ultrathin epitaxial Fefilms on GaAs(001) and related semiconductor sub-strates, Adv. Phys. 54, 137 (2005).

6. S. M. Sze, Physics of semiconductor devices, 2ndedn. (J. Wiley and sons, 1981).

7. E. I. Rashba, Theory of electrical spin injection:tunnel contacts as a solution of the conductiv-ity mismatch problem, Phys. Rev. B 62, R16267(2000).

8. Y. Ohno, D. K. Young, B. Beschoten, F. Mat-sukura, H. Ohno and D. D. Awschalom, Electri-cal spin injection in a ferromagnetic semiconductorheterostructure, Nature 402, 790 (1999).

9. B. T. Jonker, Progress toward electrical injection ofspin-polarized electrons into semiconductors, Proc.IEEE 91, 727 (2003).

10. S. Datta and B. Das, Electronic analog of theelectro-optic modulator, Appl. Phys. Lett. 56, 665(1990).

11. S. A. Crooker, M. Furis, X. Lou, C. Adelmann,D. L. Smith, C. J. Palmstrøm and P. A. Crow-ell, Imaging spin transport in lateral ferromag-net/semiconductor structures, Science 309, 2191(2005).

12. P. Kotissek, M. Bailleul, M. Sperl, A. Spitzer,D. Schuh, W. Wegscheider, C. H. Back and G.Bayreuther, Cross-sectional imaging of spin injec-tion into a semiconductor, Nature Phys. 3, 872(2007).

13. X. Lou, C. Adelmann, M. Furis, S. A. Crooker,C. J. Palmstrøm and P. A. Crowell, Electricaldetection of spin accumulation at a ferromagnet-semiconductor interface, Phys. Rev. Lett. 96,176603 (2006).

14. X. Lou, C. Adelmann, S. A. Crooker, E. S. Gar-lid, J. Zhang, K. S. M. Reddy, S. D. Flexner,C. J. Palmstrom and P. A. Crowell, Electrical de-tection of spin transport in lateral ferromagnet-semiconductor devices, Nature Phys. 3, 197 (2007).

15. E. S. Garlid, Q. O. Hu, M. K. Chan, C. J. Palm-strøm and P. A. Crowell, Electrical measurement ofthe direct spin Hall effect in Fe/InxGa1−xAs het-erostructures, Phys. Rev. Lett. 105, 156602 (2010).

16. X. G. Zhang and W. H. Butler, Large magnetore-sistance in bcc Co/MgO/Co and FeCo/MgO/FeCotunnel junctions, Phys. Rev. B 70, 172407 (2004).

17. J. Moser, M. Zenger, C. Gerl, D. Schuh, R. Meier,P. Chen, G. Bayreuther, W. Wegscheider, D. Weiss,C.-H. Lai, R.-T. Huang, M. Kosuth and H. Ebert,Bias dependent inversion of tunneling magnetore-sistance in Fe/GaAs/Fe tunnel junctions, Appl.Phys. Lett. 89, 162106 (2006).

18. G. Autes, J. Mathon and A. Umerski, Theory of

tunneling magnetoresistance of Fe/GaAs/Fe(001)junctions, Phys. Rev. B 82, 115212 (2010).

19. X. Jiang, R. Wang, S. van Dijken, R. Shelby, R.Macfarlane, G. S. Solomon, J. Harris and S. S. P.Parkin, Optical detection of hot-electron spin injec-tion into GaAs from a magnetic tunnel transistorsource, Physical Review Letters 90, 256603 (2003).

20. S. S. P. Parkin, C. Kaiser, A. Panchula, P. M.Rice, B. Hughes, M. Samant and S. H. Yang, Gianttunnelling magnetoresistance at room temperaturewith MgO (100) tunnel barriers, Nature Mat. 3, 862(2004).

21. S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzukiand K. Ando, Giant room-temperature magnetore-sistance in single-crystal Fe/MgO/Fe magnetic tun-nel junctions, Nature Mat. 3, 868 (2004).

22. G. Autes, J. Mathon and A. Umerski, Strongenhancement of the tunneling magnetoresistanceby electron filtering in an Fe/Mgo/Fe/GaAs(001)junction, Phys. Rev. Lett. 104, 217202 (2010).

23. T. Nagahama, H. Saito and S. Yuasa, Hot electrontransport in magnetic tunnel transistors with anepitaxial MgO tunnel barrier, Appl. Phys. Lett. 96,112509 (2010).

24. P. Weiss, Magnetization of crystalline Magnetite,J. Phys. 5, 453 (1896).

25. H. A. Durr, G. Y. Guo, G. van der Laan, J. Lee, G.Lauhoff and J. A. C. Bland, Element-specific mag-netic anisotropy determined by transverse mag-netic circular X-ray dichroism, Science 277, 213–215 (1997).

26. S. Chikazumi, Physics of ferromagnetism (OxfordUniversity Press, 1997).

27. P. Bruno, Tight-binding approach to the or-bital magnetic moment and magnetocrystallineanisotropy of transition-metal monolayers, Phys.Rev. B 39, 865 (1989).

28. R. M. Bozorth, Ferromagnetism (Van Nostrand,1951).

29. W. Ellis and E. Greiner, Equilibrium relations inthe solid state of the iron-cobalt system, Trans.Am. Soc. Met. 29, 415 (1941).

30. S. Blundell, Magnetism in condensed matter (Ox-ford University Press, 2001).

31. M. Dumm, B. Uhl, M. Zolfl, W. Kipferl andG. Bayreuther, Volume and interface magneticanisotropy of Fe1−xCox thin films on GaAs(001),J. Appl. Phys. 91, 8763 (2002).

32. M. Dumm, M. Zolfl, R. Moosbuhler, M. Brock-mann, T. Schmidt and G. Bayreuther, Magnetismof ultrathin FeCo (001) films on GaAs(001), J.Appl. Phys. 87, 5457 (2000).

33. F. Bianco, P. Bouchon, M. Sousa, G. Salis and S. F.Alvarado, Enhanced uniaxial magnetic anisotropyin Fe31]Co69 thin films on GaAs(001), J. Appl.Phys. 104, 083901 (2008).

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 23

34. A. T. Hindmarch, D. A. Arena, K. J. Dempsey,M. Henini and C. H. Marrows, Influence of depo-sition field on the magnetic anisotropy in epitax-ial Co70Fe30 films on GaAs(001), Phys. Rev. B 81,100407 (2010).

35. S. J. Blundell, M. Gester, J. A. C. Bland, C.Daboo, E. Gu, M. J. Baird and A. J. R. Ives,Structure induced magnetic anisotropy behaviorin Co/GaAs(001) films, J. Appl. Phys. 73, 5948(1993).

36. Y. Z. Wu, H. F. Ding, C. Jing, D. Wu, G. L. Liu,V. Gordon, G. S. Dong, X. F. Jin, S. Zhu and K.Sun, In-plane magnetic anisotropy of bcc Co onGaAs(001), Phys. Rev. B 57, 11935 (1998).

37. C. S. Tian, D. Qian, D. Wu, R. H. He, Y. Z. Wu,W. X. Tang, L. F. Yin, Y. S. Shi, G. S. Dong,X. F. Jin, X. M. Jiang, F. Q. Liu, H. J. Qian, K.Sun, L. M. Wang, G. Rossi, Z. Q. Qiu and J. Shi,Body-centered-cubic Ni and its magnetic proper-ties, Phys. Rev. Lett. 94, 137210 (2005).

38. L. F. Yin, D. H. Wei, N. Lei, L. H. Zhou, C. S. Tian,G. S. Dong, X. F. Jin, L. P. Guo, Q. J. Jia and R. Q.Wu, Magnetocrystalline anisotropy in permalloy re-visited, Phys. Rev. Lett. 97, 067203 (2006).

39. G. Bayreuther, M. Dumm, B. Uhl, R. Meier andW. Kipferl, Magnetocrystalline volume and inter-face anisotropies in epitaxial films: universal re-lation and Neel’s model, J. Appl. Phys. 93, 8230(2003).

40. E. C. Stoner and E. P. Wohlfarth, A mechanism ofmagnetic hysteresis in heterogeneous alloys, Trans.Roy. Soc. 240, 599 (1948).

41. O. Thomas, Q. Shen, P. Schieffer, N. Tournerie andB. Lepine, Interplay between anisotropic strain re-laxation and uniaxial interface magnetic anisotropyin epitaxial Fe films on (001) GaAs, Phys. Rev. Lett90, 017205 (2003).

42. J. S. Claydon, D. X. Niu, Y. B. Xu, N. D. Telling,I. W. Kirkman and G. van der Laan, Spin andorbital moments of ultra-thin Fe films on vari-ous semiconductor surfaces, IEEE Trans. Mag. 41,3325 (2005).

43. K. Zakeri, T. Kebe, J. Lindner and M. Farle, Mag-netic anisotropy of Fe/GaAs(001) ultrathin filmsinvestigated by in situ ferromagnetic resonance, J.Magn. Magn. Mater. 299, L1 (2006).

44. N. A. Morley, M. R. J. Gibbs, E. Ahmad, I. G. Willand Y. B. Xu, Comparison between the in-planeanisotropies and magnetostriction constants of thinepitaxial Fe films grown on GaAs and Ga0.8In0.2Assubstrates, with Cr overlayers, J. Appl. Phys. 99,08N508 (2006).

45. H. B. Zhao, D. Talbayev, G. Lupke, A. T. Han-bicki, C. H. Li, M. J. van’t Erve, G. Kioseoglouand B. T. Jonker, Interface magnetization reversaland anisotropy in Fe/AlGaAs(001), Phys. Rev. Lett

95, 137202 (2005).46. B. Lepine, C. Lallaizon, S. Ababou, A. Guivarc’h,

S. Deputier, A. Filipe, F. Nguyen Van Dau, A.Schuhl, F. Abel and C. Cohen, Fe/GaAs(001) andFe/GaSb(001) heterostructures: epitaxial growthand magnetic properties, J. Crystal Growth 201-202, 702 (1999).

47. F. Richomme, A. Fnidiki and J. P. Eymery, Mag-netic study of epitaxial Fe/InGaAs/InP(100) de-posited by ion-beam sputtering, J. Appl. Phys. 97,123902 (2005).

48. N. Tournerie, P. Schieffer, B. Lepine, C. Lal-laizon and G. Jezequel, Structural and magneticanisotropy properties in epitaxial Fe films onAl0.48In0.52As(001), IEEE Trans. Mag. 41, 3322(2005).

49. K. Zakeri, J. Lindner and M. Farle, Thickness-dependent reorientation of the magnetization of Femonolayers on InP(001), Europhys. Lett. 83, 17006(2008).

50. Y. B. Xu, D. J. Freeland, M. Tselepi and J. A. C.Bland, Anisotropic lattice relaxation and uniaxialmagnetic anisotropy in Fe/InAs(100)-4 × 2, Phys.Rev. B 62, 1167 (2000).

51. J. J. Krebs, B. T. Jonker and G. A. Prinz, Mag-netic properties of single-crystal Fe films grown onZnSe epilayers by molecular-beam epitaxy, J. Appl.Phys. 61, 3744 (1987).

52. J.-M. Jancu, R. Scholz, E. A. de Andrada e Silvaand G. C. La Rocca, Atomistic spin-orbit couplingand k.p parameters in III-V semiconductors, Phys.Rev. B 72, 193201 (2005).

53. O. Stier and D. Bimberg, Modeling of strainedquantum wires using eight-band k·p theory, Phys.Rev. B 55, 7726 (1997).

54. Y. Rajakarunanayake, R. H. Miles, G. Y. Wu andT. C. McGill, Band structure of ZnSe-ZnTe super-lattices, Phys. Rev. B 37, 10212 (1988).

55. M. Kosuth, V. Popescu, H. Ebert and G.Bayreuther, Magnetic anisotropy of thin Fe filmson GaAs, Europhys. Lett. 72, 816 (2005).

56. A. T. Hindmarch, C. J. Kinane, M. MacKenzie,J. N. Chapman, M. Henini, D. Taylor, D. A. Arena,J. Dvorak, B. J. Hickey and C. H. Marrows, Inter-face induced uniaxial magnetic anisotropy in amor-phous CoFeB films on AlGaAs(001), Phys. Rev.Lett. 100, 117201 (2008).

57. A. F. Isakovic, J. Berezovsky, P. A. Crowell, L. C.Chen, D. M. Carr, B. D. Schultz and C. J. Palm-strom, Control of magnetic anisotropy in Fe1−xCoxfilms on vicinal GaAs and Sc1−yEryAs surfaces, J.Appl. Phys. 89, 6674 (2001).

58. J. A. Wolf, K. K. Anderson, E. D. Dahlberg, P. A.Crowell, L. C. Chen and C. J. Palmstrom, Tunablemagnetization reversal in epitaxial bcc Fe1−xCoxfilms on vicinal surfaces, J. Appl. Phys. 93, 8256

24 Aidan T. Hindmarch

(2003).59. T. Yoo, S. Khym, H. Lee, S. Chung, S. Lee, X. Liu

and J. K. Furdyna, Asymmetry in the planar Hallresistance of Fe films grown on vicinal GaAs sub-strates, J. Appl. Phys. 107, 09C505 (2010).

60. W. Kipferl, M. Sperl, T. Hagler, R. Meier andG. Bayreuther, Stabilization of ferromagnetic or-der in epitaxial ultrathin Fe films, J. Appl. Phys.97, 10B313 (2005).

61. W. Kipferl, M. Dumm, P. Kotissek, F. Steinbauerand G. Bayreuther, Bloch’s law for epitaxial ultra-thin dot arrays with uniaxial magnetic anisotropy,J. Appl. Phys. 95, 7417 (2004).

62. D. Niu, X. Zou, Y. Zhai, Z. Huang, I. Will, P. Wong,J. Wu and Y. Xu, Reduction of in-plane uniaxialmagnetic anisotropy in patterned single-crystal Fedot arrays, IEEE Trans. Mag. 45, 3507 (2009).

63. K. K. Meng, J. Lu, S. L. Wang, H. J. Meng, J. H.Zhao, J. Misuraca, P. Xiong and S. von Molnar,Magnetic anisotropies of laterally confined struc-tures of epitaxial Fe films on GaAs(001), Appl.Phys. Lett. 97, 072503 (2010).

64. E. M. Kneedler, B. T. Jonker, P. M. Thibado,R. J. Wagner, B. V. Shanabrook and L. J. Whit-man, Influence of substrate surface reconstructionon the growth and magnetic properties of Fe onGaAs(001), Phys. Rev. B 56, 8163 (1997).

65. A. Ionescu, M. Tselepi, D. M. Gillingham, G.Wastlbauer, S. J. Steinm&uuml;ller, H. E. Beere,D. A. Ritchie and J. A. C. Bland, Submonolayergrowth of Fe on a GaAs(100)-2 × 6 reconstructedsurface, Phys. Rev. B 72, 125404 (2005).

66. D. M. Gillingham, M. Tselepi, A. Ionescu, S. J.Steinmuller, H. E. Beere, D. A. Ritchie and J. A. C.Bland, Smoothening of ultrathin Fe films grownon GaAs(100) observed by scanning tunneling mi-croscopy and Brillouin light scattering, Phys. Rev.B 76, 214412 (2007).

67. C. Godde, S. Noor, C. Urban and U. Kohler, Struc-tural changes and alloying of annealed iron layerson GaAs(001) and GaAs(110), Surf. Sci. 602, 3343(2008).

68. S. Mirbt, B. Sanyal, C. Isheden and B. Johansson,First-principles calculations of Fe on GaAs(100),Phys. Rev. B 67, 155421 (2003).

69. J. W. Freeland, I. Coulthard, W. Antel and A. P. J.Stampfl, Interface bonding for Fe thin films onGaAs surfaces of differing morphology, Phys. Rev.B 63, 193301 (2001).

70. R. Moosbuhler, F. Bensch, M. Dumm and G.Bayreuther, Epitaxial Fe films on GaAs(001): doesthe substrate surface reconstruction affect the uni-axial magnetic anisotropy?, J. Appl. Phys. 91, 8757(2002).

71. E. Sjostedt, L. Nordstrom, F. Gustavsson and O.Eriksson, Uniaxial magnetocrystalline anisotropy

of metal/semiconductor interfaces: Fe/ZnSe(001),Phys. Rev. Lett. 89, 267203 (2002).

72. J. M. LeBeau, Q. O. Hu, C. J. Palmstrom andS. Stemmer, Atomic structure of postgrowth an-nealed epitaxial Fe/(001)GaAs interfaces, Appl.Phys. Lett. 93, 121909 (2008).

73. K. Olejnik, P. Wadley, J. A. Haigh, K. W. Ed-monds, R. P. Campion, A. W. Rushforth, B. L.Gallagher, C. T. Foxon, T. Jungwirth, J. Wun-derlich, S. S. Dhesi, S. A. Cavill, G. van derLaan and E. Arenholz, Exchange bias in a ferro-magnetic semiconductor induced by a ferromag-netic metal: Fe/(Ga,Mn)As bilayer films studied byXMCD measurements and SQUID magnetometry,Phys. Rev. B 81, 104402 (2010).

74. J. Tobin, S.-W. Yu, S. Morton, G. Waddill, J.Thompson, J. Neal, M. Spangenberg and T. Shen,Highly polarized emission in spin resolved pho-toelectron spectroscopy of α-Fe(001)/GaAs(001),Surf. Sci. 604, 1342 (2010).

75. S. van Dijken, X. Jiang and S. S. P. Parkin, Roomtemperature operation of a high output currentmagnetic tunnel transistor, Appl. Phys. Lett. 80,3364 (2002).

76. X. Jiang, R. Wang, R. M. Shelby, R. M. Mac-farlane, S. R. Bank, J. S. Harris and S. S. P.Parkin, Highly spin-polarized room-temperaturetunnel injector for semiconductor spintronics usingMgO(100), Phys. Rev. Lett. 94, 056601 (2005).

77. J. M. Shaw, S. Lee and C. M. Falco, Overlayer-induced magnetic uniaxial anisotropy in nanoscaleepitaxial Fe, Phys. Rev. B 73, 094417 (2006).

78. L. Giovanelli, G. Panaccione, G. Rossi, M. Fab-rizioli, C. S. Tian, P. L. Gastelois, J. Fujiiand C. H. Back, Layer-selective spectroscopy ofFe/GaAs(001): influence of the interface on themagnetic properties, Phys. Rev. B 72, 045221(2005).

79. L. Giovanelli, G. Panaccione, G. Rossi, M. Fab-rizioli, C. S. Tian, P. L. Gastelois, J. Fujii andC. H. Back, Interface magnetization profiling byX-ray magnetometry of marker impurities onFe/GaAs(001)-(4×6), Appl. Phys. Lett. 87, 042506(2005).

80. B. Kardasz, J. Zukrowski, O. Mosendz, M. Przybyl-ski, B. Heinrich and J. Kirschner, Interface atomicstructure and magnetic anisotropy in ultrathinFe films grown by thermal deposition and pulsedlaser deposition on GaAs(001), J. Appl. Phys. 101,09D110 (2007).

81. J.-M. Lee, J.-Y. Kim, S.-U. Yang, B.-G. Park, J.-H.Park, S.-J. Oh and J.-S. Kim, Magnetism of pristineFe films on GaAs(100), Phys. Rev. B 76, 052406(2007).

82. T. Damm, M. Buchmeier, A. Schindler, D. E. Bur-gler, P. Grunberg and C. M. Schneider, Magnetic

Interface Magnetism in Ferromagnetic Metal – Compound Semiconductor Hybrid Structures 25

properties of Fe films and Fe/Si/Fe trilayers grownon GaAs(001) and MgO(001) by ion-beam sputterepitaxy, J. Appl. Phys. 99, 093905 (2006).

83. Z. L. Bao and K. L. Kavanagh, Epitaxial Fe/GaAsvia electrochemistry, J. Appl. Phys. 98, 066103(2005).

84. E. B. Svedberg, J. J. Mallett, L. A. Bendersky,A. G. Roy, W. F. Egelhoff, Jr. and T. P. Mof-fat, A structural study of electrodeposited Fe on n-GaAs(001), J. Electrochem. Soc. 153, C807 (2006).

85. K. Liu, D. Shen, J. Zhang, X. Wu, B. Li, B. Li,Y. Lu and X. Fan, MOCVD growth, structure andmagnetic properties of Fe films grown on GaAs(001) substrates, Sol. Stat. Comm. 140, 33 (2006).

86. A. T. Hindmarch, A. K. Suszka, M. MacKenzie,J. N. Chapman, M. Henini, D. Taylor, B. J. Hickeyand C. H. Marrows, Structural and magnetic prop-erties of magnetron sputtered Co70Fe30 films onGaAs(110), J. Appl. Phys. 105, 073907 (2009).

87. A. T. Hindmarch, C. J. Kinane, C. H. Marrows,B. J. Hickey, M. Henini, D. Taylor, D. A. Arena andJ. Dvorak, In-plane magnetic anisotropies of sput-tered Co0.7Fe0.3 films on AlGaAs(001) spin lightemitting diode heterostructures, J. Appl. Phys.101, 09D106 (2007).

88. J.-M. Lee, S.-J. Oh, K. J. Kim, S.-U. Yang, J.-H.Kim and J.-S. Kim, Kinetic stabilization of a pris-tine Fe film on (4×2)-GaAs(100), Phys. Rev. B 75,125421 (2007).

89. B. D. Schultz, N. Marom, D. Naveh, X. Lou, C.Adelmann, J. Strand, P. A. Crowell, L. Kronik andC. J. Palmstrøm, Spin injection across the Fe/GaAsinterface: role of interfacial ordering, Phys. Rev. B80, 201309 (2009).

90. S. C. Erwin, S.-H. Lee and M. Scheffler, First-principles study of nucleation, growth, and inter-face structure of Fe/GaAs, Phys. Rev. B 65, 205422(2002).

91. D. O. Demchenko and A. Y. Liu, Influence of inter-face structure on electronic properties and Schottkybarriers in Fe/GaAs magnetic junctions, Phys. Rev.B 73, 115332 (2006).

92. P. Schieffer, A. Guivarc’h, C. Lallaizon, B. Lep-ine, D. Sebilleau, P. Turban and G. Jezequel, For-mation of a body-centered-cubic Fe-based alloy atthe Fe/GaAs(001) interface, Appl. Phys. Lett. 89,161923 (2006).

93. T. J. Zega, A. T. Hanbicki, S. C. Erwin, I. Zutic,G. Kioseoglou, C. H. Li, B. T. Jonker and S. R. M.,Determination of interface atomic structure andits impact on spin transport using Z-contrast mi-croscopy and density-functional theory, Phys. Rev.Lett 96, 196101 (2006).

94. B. D. Schultz, C. Adelmann, X. Y. Dong, S. McKer-nan and C. J. Palmstrom, Phase formation in thethin film Fe/GaAs system, Appl. Phys. Lett. 92,

091914 (2008).95. Y. B. Xu, E. T. M. Kernohan, D. J. Freeland, A.

Ercole, M. Tselepi and J. A. C. Bland, Evolution ofthe ferromagnetic phase of ultrathin Fe films grownon GaAs(100)-4× 6, Phys. Rev. B 58, 890 (1998).

96. J. S. Claydon, Y. B. Xu, M. Tselepi, J. A. C. Blandand G. van der Laan, Direct observation of a bulk-like spin moment at the Fe/GaAs(100)-4× 6 inter-face, Phys. Rev. Lett 93, 037206 (2004).

97. J. Stohr, Exploring the microscopic origin of mag-netic anisotropies with X-ray magnetic circulardichroism (XMCD) spectroscopy, J. Magn. Magn.Mater. 200, 470 (1999).

98. J. F. Ankner and G. P. Felcher, Polarized-neutronreflectometry, J. Magn. Magn. Mater. 200, 741(1999).

99. P. Steadman, M. Ali, A. T. Hindmarch, C. H. Mar-rows, B. J. Hickey, S. Langridge, R. M. Dalglieshand S. Foster, Exchange bias in spin-engineereddouble superlattices, Phys. Rev. Lett. 89, 077201(2002).

100. S. Park, M. R. Fitzsimmons, X. Y. Dong, B. D.Schultz and C. J. Palmstrøm, Magnetic degrada-tion of an FeCo/GaAs interface, Phys. Rev. B 70,104406 (2004).

101. S. Park, M. R. Fitzsimmons, C. F. Majkrzak,B. D. Schultz and C. J. Palmstrøm, The in-fluence of growth temperature and annealing onthe magnetization depth profiles across ferromag-netic/semiconductor interfaces, J. Appl. Phys. 104,083905 (2008).

102. H. B. Zhao, D. Talbayev, Q. G. Yang, G. Lupke,A. T. Hanbicki, C. H. Li, O. M. J. van ’t Erve,G. Kioseoglou and B. T. Jonker, Ultrafast magne-tization dynamics of epitaxial Fe films on AlGaAs(001), Appl. Phys. Lett. 86, 152512 (2005).

103. H. B. Zhao, D. Talbayev, G. Lupke, A. T. Han-bicki, C. H. Li and B. T. Jonker, Ultrafast interfacemagnetization dynamics in Fe/AlGaAs(001) het-erostructure, Appl. Phys. Lett. 91, 052111 (2007).

104. M. S. Blois Jr., Preparation of thin magnetic filmsand their properties, J. Appl. Phys. 26, 975 (1955).

105. C. T. Chen, Y. U. Idzerda, H. J. Lin, N. V. Smith,G. Meigs, E. Chaban, G. H. Ho, E. Pellegrin andF. Sette, Experimental confirmation of the x-raymagnetic circular-dichroism sum-rules for iron andcobalt, Phys. Rev. Lett 75, 152–155 (1995).

106. R. Richter and H. Eschrig, Spin-polarised elec-tronic structure of disordered bcc FeCo alloys fromLCAO-CPA, J. Phys. F: Met. Phys. 18, 1813(1988).

107. C. Andersson, B. Sanyal, O. Eriksson, L. Nord-strom, O. Karis, D. Arvanitis, T. Konishi, E.Holub-Krappe and J. H. Dunn, Influence of ligandstates on the relationship between orbital momentand magnetocrystalline anisotropy, Phys. Rev. Lett.

26 Aidan T. Hindmarch

99, 177207 (2007).108. K. Nakamura, R. Shimabukuro, Y. Fujiwara, T.

Akiyama, T. Ito and A. J. Freeman, Giant mod-ification of the magnetocrystalline anisotropy intransition-metal monolayers by an external electricfield, Phys. Rev. Lett. 102, 187201 (2009).

109. K. Ohta, T. Maruyama, T. Nozaki, M. Shiraishi,T. Shinjo, Y. Suzuki, S.-S. Ha, C.-Y. You andW. Van Roy, Voltage control of in-plane magneticanisotropy in ultrathin Fe/n-GaAs(001) Schottkyjunctions, Appl. Phys. Lett. 94, 032501 (2009).

110. A. W. Rushforth, E. De Ranieri, J. Zemen, J.Wunderlich, K. W. Edmonds, C. S. King, E. Ah-

mad, R. P. Campion, C. T. Foxon, B. L. Gal-lagher, K. Vyborny, J. Kucera and T. Jungwirth,Voltage control of magnetocrystalline anisotropyin ferromagnetic-semiconductor-piezoelectric hy-brid structures, Phys. Rev. B 78, 085314 (2008).

111. E. Villari, Change of magnetization by tension andby electric current, Ann. Phys. Chem. 126, 87(1865).

112. F. Nolting, D. Legut, J. Rusz, P. M. Oppeneer,G. Woltersdorf and C. H. Back, Anisotropy of theL2,3 X-ray magnetic linear dichroism of Fe films onGaAs: Experiment and ab initio theory, Phys. Rev.B 82, 184415 (2010).