Theory of Magneto-Optic Kerr Effect in Ferromagnetic Compounds

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/272065058
Theory of the Magneto-Optical Kerr Effect in Ferromagnetic Compounds
Book · March 1999

DOI: 10.13140/2.1.3171.4083
CITATIONS READS
2 7,164
1 author:
Peter M. Oppeneer
Uppsala University
560 PUBLICATIONS 17,090 CITATIONS
SEE PROFILE
All content following this page was uploaded by Peter M. Oppeneer on 10 February 2015.
The user has requested enhancement of the downloaded file.

! " # $&%'()
* (# + ,- . / + 0-1 3254
6 798;:=<>:@?A7B?C:=DFEHGIGIJLKNMO:>PQ?
TR S VM U MO<=7WENX S ENXZYN[G7B\B7WYN[T]:=GIJLKN[E_^MC7WYN[G
`MabMI[Ma E;7c?dabK;798;:=<ea
f DWMIXF[T<@[X9? f DWE
`VMTa MI[.MabE;7B?Ta#ghajikamlnHno[EN[.[M
XF[8oDWMI[TEp:=Eq6Dr[\Hs9t:@[TYN[.MO<u7WEHYN[
vwEHGx?C:@? S ?yPAS z My{yKN[DWMI[.?C:=GIJLKN[(g#Kr|}GI:>\

{[JLKHEH:>GOJAKN[~EH: f [MOGI:@?c7Bz ?`MI[TGIYN[E
i7Bz M R.FF
HE Y^"DY_G nmDW\F[(?CDFDW8
`[TJ.<u7BMO[WN:>P|rD S \EHDd&:?V7W<=<W
KN[MI[:=G)?CKN[(7d|<=:>XFKr?Y})[<=<>Gm
FDW8F)HWI 9 HJ7Na9FW-
i
¡¡¢,£5!¢¤
¥§¦©¨Lª¬«®°¯±T²wªe³Á¨ª¬#µ¶I·A¨T¸¹ªeIº©»m¼Tª¬³½²¿¾ ¥
1.1 Historical overview of magneto-optics . . . . . . . . . . . . . . . . . . . ÀÀ
1.1.1 Early history . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Á
1.2
1.1.2 Recent history . . . . . . . . . . . . .
Outline of the Kerr effect . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. Â
1.3 Recent applications and extensions of MO
Ã
ÃÄ
spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.1 Applications of MO effects . . . . . . . . . . . . . . . . . . . . .
Ä
1.3.2 MO effects in the X-ray energy regime . . . . . . . . . . . . . .
ÅÇÆ;ÈT¸xO«ÊÉ Ë
1.3.3 Non-linear MOKE . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 Classical optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ÌÌ

ÀeÀ¬ÍÁ
2.1.1 The Maxwell equations . . . . . . . . . . . . . . . . . . . . . .
2.1.2 The Fresnel equation . . . . . . . . . . . . . . . . . . . . . . . .
ÀeÀ¬ÍÃ
2.2 The Kerr and Faraday effects . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 The polar MO Kerr effect . . . . . . . . . . . . . . . . . . . . .
ÀeÀeÀ¬ÎÌÌ
2.2.2 The longitudinal and transversal MO Kerr effects . . . . . . . . .
2.2.3 The Faraday effect . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Linear-response theory . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 General formulation . . . . . . . . . . . . . . .
2.3.2 The conductivity tensor . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. ÁIÁwÍÀ
2.3.3 Single-particle formulation . . . . . . . . . . . . . . . . . . . . .
Á®Â
ÍwÍ¿ÂÂ
2.3.4 Properties of the conductivity tensor . . . . . . . . . . . . . . . .
2.4 Density-functional theory . . . . . . . . . . . . . . . . . . . . . . . . .
ÍxÍ¹ÎÄ
2.4.1 Basic formulation . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.2 LSDA approach to the excitation spectrum . . . . . . . . . . . .
Í¹Â¹ÎÏ
2.5 Numerical implementation . . . . . . . . . . . . . . . . . . . . . . . . .
2.5.1 The relativistic band structure . . . . . . . . . . . . . . . . . . .
Â¿Â
2.5.2 The matrix elements . . . . . . . . . . . . . . . . . . . . . . . .
2.5.3 k-space integration and numerical convergence . . . . . . . . . .
ii CONTENTS
ÐÑ»Ò«Ê¯T¸Ó«®¸Ó¯yÔ¸xÕH¯BÖB×jCÕH¯9ÖdÔ.¸xÕNª¿Öd¶I¨T¯×jCÕNª ØLÙ
ÂÄ
ÂÓÎ
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ÚIÀ
3.2 Magnetic moments and anisotropy . . . . . . . . . . . . . . . . . . . . .
ÚÀ
3.3 MO spectra of FePt, FePd, CoPd, and CoPt . . . . . . . . . . . . . . . .
ÚwÍ
3.3.1 FePt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ú¹Ä
3.3.2 FePd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ú¿Ì
3.3.3 CoPd and CoPt . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ø&ÛÕNª¬ÜÝÛ(Þß_ª¬à×jâáo×jAãä¼BA±¨T¯T¾ å°¥
æ ÃOÀ
Ã¿Í
4.1 Physical aspects of XPt compounds . . . . . . . . . . . . . . . . . . .
ÃxÄ
4.2 Calculated MOKE spectra . . . . . . . . . . . . . . . . . . . . . . . . .
ÃÓÄ
4.3 Comparison of calculated quantities . . . . . . . . . . . . . . . . . . . .
Ä®Ï
4.3.1 Comparison to other first-principles calculations . . . . . . . . .
ÄÀ
4.3.2 Comparison to the experimental magnetic moments . . . . . . .
Ä®Â
4.4 Origin of the large Kerr effect . . . . . . . . . . . . . . . . . . . . . . .
4.5 Discussion and conclusions . . . . . . . . . . . . . . . . . . . . . . . .
çèµ¨Téo³9¶I¨T¯yê¸xë½¶ ª¬¸x¯ì×jCãä¼9C±T¨¯T¾ ÙOË
ÄwÌ
ÎIÀ
5.1 Introduction to MnBi . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Î¹Í
5.2 Calculated MOKE spectra of MnBi . . . . . . . . . . . . . . . . . . . .
Î¿Í
5.3 Origin of the record Kerr effect in MnBi . . . . . . . . . . . . . . . . . .
ÎwÂ
5.3.1 MnBi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ÎÓÄ
5.3.2 MnAs, MnSb, CrTe, and CrBi . . . . . . . . . . . . . . . . . . .
Î¿Î
5.3.3 Al-doped MnBi . . . . . . . . . . . . . . . . . . . . . . . . . .
Î¹Ì
5.4 Temperature dependence . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
åÇÕNªeµ¨dí°îì¶I¨T¯yê¸xë´¶Iª¬¸x¯V×jCãä¼9C±T¨¯T¾ Ë°¥
ÌOÀ
Ì¿Í
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Ì¹Í
6.2 Calculated MOKE spectra . . . . . . . . . . . . . . . . . . . . . . . . .
Ì¿Í
6.2.1 Outline of ternary compounds . . . . . . . . . . . . . . . . . . .
Ì¹Ú
6.2.2 Computational aspects . . . . . . . . . . . . . . . . . . . . . . .
À¬Ï¹Ú
6.2.3 PtMnSb, PdMnSb, and NiMnSb . . . . . . . . . . . . . . . . . .
ÀeÏÓÄ
6.2.4 PtMnSn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ï ï À¬Ï¹Ì
6.2.5 PtCrSb and PtFeSb . . . . . . . . . . . . . . . . . . . . . . . . .
À¹À¬Ï
6.2.6 Pt MnSb and Co HfSn . . . . . . . . . . . . . . . . . . . . . .
À¹ÀðÂ
6.2.7 PtMnBi, BiMnPt, and PtGdBi . . . . . . . . . . . . . . . . . . .
À¿À®Ä
6.2.8 NiMnAs, PdMnAs, PtMnAs, and RuMnAs . . . . . . . . . . . .
6.3 Conclusions Heusler compounds . . . . . . . . . . . . . . . . . . . . . .
CONTENTS iii
ÙÇñ«®¶I¨³½±ã'×jCãä¼9C±T¨¯T¾ ¥C¥ Ë
7.1 Introduction to the physics of uranium
compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . À¹ÀeÁIÀ¬ÌÀ
ÀÊÀeÀÊÁ¿Á¿ÁwÎÁÚ
7.2 Results for selected uranium compounds . . . . . . . . . . . . . . . . .
7.2.1 UAsSe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ò
7.2.2 URhAl . . . . . . . . . . . . .
7.2.3 UFe . . . . . . . . . . . . . .
. .
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
À¬Í¹Ï
éo³óîdë´³½C·C«®¶ ¼TÈLÉ ¥xÐ.¥
7.3 Summary uranium compounds . . . . . . . . . . . . . . . . . . . . . . .
ô ²¬õ°¨TxöÒë´¸x¯T·Cã¸x¨Lª ¥ çC÷
ø ³´¾¬ªmAù ô îTîT«Ê¸Óú°³½¶ ª¬³½C¨T¾ ¥ çAÅ
iv CONTENTS
1
/ ( û
ü !¢þý&ÿ - ¢¡ ¢,
£q¢, Ñ¢ q¤
"!#$!%&#' (*)+-,/.0#12-345%

6879687:6 0; <2=4>@?BA$C9DFEHGI=J?
The field of magneto-optics (MO) has a long history. The discovery of the Faraday effect
by Michael Faraday as early as 1845 marks its birth. Faraday observed the rotation of the
K
polarization plane of linearly polarized light upon transmission through a piece of lead-
borosilicate glass in a magnetic field (Faraday 1846). Other researchers at that time had
been thinking too if light could possibly interact with a magnetic material, but it took the
outstanding experimental expertise of Faraday to make the discovery (e.g., Busch 1977).
Faraday’s report on his discovery drew immediately a wide attention. The effect was soon
confirmed by a number of other researchers (see, e.g., the bibliography of magneto-optics
by Palik and Henvis 1967).
The next highlight of magneto-optics was the discovery of the polar MO Kerr effect
in 1876 by the Scotch scientist Rev. John Kerr (1876, 1877). Kerr observed the rotation
of the polarization plane of linearly polarized light upon reflection from the surface of a
magnetic piece of iron. Kerr used the pole of a magnet to reflect the incident light; the
MO phenomenon in this particular geometry is therefore called the polar magneto-optical
Kerr effect (MOKE). Two years later, Kerr discovered a similar MO phenomenon in the
reflection of linearly polarized light, but from an in-plane magnetized piece of iron (Kerr
1878). This phenomenon is nowadays known as the longitudinal MO Kerr effect, where
L
the plane of incidence is parallel to the magnetic field.
There were other researchers who claimed to have observed an interaction of light and magnetism,
see the discussion by Busch (1977). The scientist and author G. C. Lichtenberg wrote 1796 in his
diary: “If one would make a mirror of steel, would it show a difference in the reflection, when one
would bring it close to a magnet?” The author is indebted to F. U. Hillebrecht for bringing this to his
attention.
2 CHAPTER 1. INTRODUCTION TO MAGNETO-OPTICS
The discoveries of Faraday and Kerr created a lively interest in MO effects. One of
the prime fundamental consequences of these discoveries was that they prompted to think
of light as an electromagnetic entity, something that had not been conceived before (see,
e.g., Faraday’s Diary 1933). A first application of the Faraday effect was proposed by
Faraday himself, who suggested to apply it as a tool to measure the magnetic field strength,
since it appeared to be linear in the magnetic field (see Voigt 1908). Later measurements
of the Faraday effect in specially prepared thin films of Fe, Co, and Ni showed that the
linear dependence did not hold for these ferromagnetic materials (Kundt 1884, du Bois
1887). A further spin-off was that studies of MO phenomena initiated other fundamental
discoveries. Zeeman was one of the scientists, who was at the end of the last century
working on the Kerr effect in Fe, Co, and Ni (e.g., Zeeman 1894, 1895). In the sequel of
his studies, Zeeman discovered in 1896 the splitting of emitted spectral lines of Na atoms
in a magnetic field (Zeeman 1897). Other subsequent discoveries where the occurrence of
magnetic double refraction in Na vapor (Voigt 1899), in a colloidal suspension of metal
particles (Majorana 1902), and in paramagnetic liquids (Cotton and Mouton 1907). An
overview of the early work on MO phenomena has been collected by Voigt (1908).
At the turn of the century MO phenomena hat become an important research topic.
Quantum mechanics had not yet emerged, therefore the theoretical understanding of the
phenomena was completely lacking. An early theoretical model of the Faraday effect was
proposed by Lorentz (1884), based on the idea that left- and right-circularly polarized
light coupled differently to classical electron oscillators in the solid. Further theoretical
extensions were made by Drude (1900a,b). The basic understanding of MO effects grew
with the development of quantum mechanics. Hulme (1932) and Halpern (1932) were
the first to proposed that the Faraday effect arises from spin-polarized electron motion in
the presence of spin-orbit (SO) coupling. Hulme used for his considerations the Kramers-
Heisenberg dispersion equation, which gives the index of refraction in terms of the energy
eigenvalues and matrix elements of the electric dipole operator (Kramers and Heisenberg
1925). He accounted for the difference between the indices of refraction of the left- and
right-circularly polarized light beams by considering the SO induced splitting of the energy
eigenvalues, but neglected the effect of SO coupling on the wave functions. Kittel (1951)
argued that the influence of SO coupling on the wave functions could give rise to an equally
large contribution. A more complete formulation, in which both SO interaction and spin
polarization were treated, was developed by Argyres (1955). The basic origin of magneto-
optics was thereby in the fifties understood as an interplay of SO coupling and exchange
splitting. In the same period from the turn of the century to the fifties, there had been
continuing improvements of the experimental techniques, but no new discoveries could be
reported. The only exception was the discovery of the MO Kerr effect in paramagnetic
metals in an applied field by Majorana (1944).
67:687NM OQPSRTPSUVE A$C9DFEHGI=J?

Two rather different discoveries stimulated the interest in MO research again in the fifties.
The first one was the discovery that magnetic domains could be observed using the MO
Kerr effect (Williams et al. 1951, Fowler and Fryer 1952). As a result, a new approach
CHAPTER 1. INTRODUCTION TO MAGNETO-OPTICS 3
to visualizing surface and subsurface magnetic domains developed. Furthermore, it was

realized that the MOKE could be employed to read-out suitably stored magnetic information
(Williams et al. 1957, Conger and Tomlinson 1962, Supernowicz 1963). This initiated
the interest in MO recording, which since then has developed into a leading technological
application of MOKE (see, e.g., the recent overviews by Mansuripur 1995, and Carey et al.
1995). The other discovery was that of the microwave magneto-absorption and Faraday
effect in semiconductors (Dresselhaus et al. 1955, Zwerdling and Lax 1957, Lax et al.
1959). MO spectroscopy thereby became a technique to investigate the band structure of
semiconductors. A substantial amount of experimental and theoretical work on the Faraday
effect in semiconductors followed (see, e.g., Lax and Nishina 1961, Boswarva et al. 1962,
Roth 1964, Bennett and Stern 1965). At that time, it started to become customary to relate
MO phenomena to the materials band structure.
Progress in experimental techniques made it possible to perform measurements of MO
spectra in a wide energy range of about 0.5 to 5 eV. Particularly detailed studies of the
various MOKE spectra of the elementary ferromagnets Fe, Co, and Ni were carried out by
Krinchik and coworkers (Krinchik 1964, Krinchik and Artem’ev 1968a). Magneto-optical
measurements continued to mature into an appealing and widely-used spectroscopic tool
in solid-state research. As a consequence, numerous publications on measured MO spectra
of many materials appeared in the last three decades. The amount of experimental data
has thereby become too large to be covered in a single review. Most of the accumulated
experimental MO data have recently been surveyed by Buschow (1988) and by Schoenes
(1992), to whom the reader is referred. Two major contributions to the field of MOKE
spectroscopy deserve to be mentioned here. The first one is the large scale investigation of
the polar MOKE that was undertaken by Buschow and van Engen. With the aim to search
for materials which could be applied in MO recording devices, Buschow and van Engen
investigated the MOKE spectra of several hundred materials (see, e.g., Buschow and van
Engen 1981, Buschow et al. 1983a, Buschow 1988). Another goal of MO research which
became intensively pursued in the eighties was to extracted information on the electronic
structure of, in particular, lanthanide and actinide compounds. This research directive
was undertaken and perfected by Schoenes and coworkers (see Schoenes 1984, Reim and
Schoenes 1990).
While the experimental research of MO spectra was booming, its theoretical description
hardly progressed over the same period. The microscopic origin of MO phenomena had
been attributed by Argyres (1955) to an interplay of SO coupling and spin polarization, but
there was no direct comparison of theoretical and measured MO spectra possible. Most
experimental data were therefore still being analyzed using the classical models dating back
to Lorentz and Drude. A first interband theory of the Kerr effect in Ni, that allowed for a
comparison to experiment, was proposed by Cooper and Ehrenreich (1964), and Cooper
(1965), who took the Ni band structure into account, which at that time had just become
known. The required optical transition matrix elements were, however, not yet calculated,
but only estimated. This situation changed with the advent of density-functional theory and
the local-spin-density approximation (LSDA) (Hohenberg and Kohn 1964, Kohn and Sham
1965), whereby accurate band-structure calculations became feasible. On this basis, and
employing also an expression for the optical conductivity stemming from linear-response
theory (Kubo 1957), the next decisive step was taken by Callaway and coworkers, who
WXIY Z\[][\^`_Vacb d@efY Z\[]gh^`_Vacb

computed the absorptive parts of both the diagonal and the off-diagonal optical conductivity
(i.e., and ) of Ni and Fe (Wang and Callaway 1974, Singh et al.
1975). As the MO Kerr and Faraday effects are related directly to the off-diagonal
optical conductivity (see Section 2.2), this was the first band-theoretical calculation of a
MO spectrum. The agreement between theory and experiment was not overwhelming.
Callaway and coworkers did then not proceed to try to compute the dispersive parts
of the diagonal and off-diagonal conductivities, from which they could have calculated
the Faraday and Kerr spectra. At the end of the eighties several groups picked up the
i
subject of computing MO spectra again. Ebert (1989, 1990) and Uspenskii and Khalilov
(1989) computed the absorptive parts of the diagonal and off-diagonal conductivities of 3
ferromagnets, which compared similarly – or even somewhat less good – to experiment as
the spectra computed by Callaway and coworkers. A further step was taken by Daalderop et
al. (1988a, 1988b), who computed the absorptive components of the optical conductivities
of UNiSn, from which they computed the dispersive components through Kramers-Kronig
j
transformation and subsequently the first theoretical MO Kerr spectrum. Daalderop et
i
al. (1988a) predicted a large Kerr rotation of 5 for UNiSn, but did, unfortunately, not
publish a test of the computational method on simpler systems like the elemental 3 metals.
Such tests appear to be mandatory, in view of the difficulties inherent to MO calculations.
Moreover, UNiSn is an antiferromagnetic material, for which so far no measurements of
the Kerr effect in the ferromagnetic phase could be carried out. No conclusions regarding
the applicability of the method could therefore be drawn. The first explicit calculations of
the Kerr spectra of Fe and Ni are due to Oppeneer et al. (1991, 1992a), who developed a
precise computational technique to solve the difficulties inherent to the evaluation of MO
spectra. The thus-computed MO Kerr effect of Fe provides a minute description of the
experimental MOKE, and the computed Kerr rotation of Ni compares reasonably well to
low-temperature data (Di and Uchiyama 1994). These results proved that MO spectra can
accurately be calculated on the basis of relativistic band-structure theory. Since then ab
initio calculations of the Kerr spectra of various compounds have been reported (see, e.g.,
Oppeneer and Antonov 1996). Altogether, these investigations have supplied a wealth of
evidence that precise MO Kerr spectra can be calculated from first-principles energy-band
theory, and even beyond, that it is feasible to make ab initio predictions of MOKE spectra.
lk mon%p.0#q(Qr0#qst#u8v#wx#uy
Linearly polarized light that is reflected from a magnetized solid is modifided in two ways
upon reflection: First, the polarization plane of the reflected light is rotated over a small
z|{
angle with respect to that of the incident light, and second, the reflected light has become
}F{
elliptically polarized. The first phenomenon is called the Kerr rotation, denoted , and
the latter is called the Kerr ellipticity, denoted . These modifications in the reflected
light occur in the polar geometry, where the magnetic field and incident wave vector are
perpendicular to the surface, and also in the longitudinal geometry, where the magnetization
is parallel to the plane of incidence and to the surface plane. A schematic illustration of
y y' y
a b
x' qK
x x
~IJ&xH`HSchematic illustration of the polar MO Kerr effect. Linearly polarized
polarization plane is tilted over a small angle

light is, after reflection from a magnetized material, elliptically polarized and the main
with respect to that of the incident light.
the polar Kerr effect is shown in Fig. 1.1. The Kerr ellipticity is determined from and ,
} {&V /
the long and short axis, respectively, of the ellipse in Fig. 1.1. It is commonly defined as:
^ 1.1)
Both z{ and }{ are thereby expressed as an angle. There exists a third type of geometry
for MOKE, which is the transversal geometry, where the magnetization lies in the surface
plane, and is perpendicular to the plane of incidence. In this geometry the MO effect is
measured as a magnetization modulated intensity difference of the reflected light (Zeeman
1896, Ingersoll 1912, Krinchik and Nuralieva 1959).
Various experimental techniques for detecting the Kerr rotation and ellipticity have been
discussed in the literature (see, e.g., Robinson 1963, Jasperson and Schnatterly 1969, Suits
1971, Sato 1981). Schoenes (1992) has given a classification of the available techniques
z{ }{
and a discussion of their respective advantages and disadvantages.
For most transition metal (TM) compounds, and are small quantities, which
Àj Àj
are usually smaller than . A few TM compounds are known in which the Kerr rotation
exceeds . In such cases it has become customary to speak of a ‘giant Kerr rotation’.
Famous examples of such materials are PtMnSb (van Engen et al. 1983a) and MnBi (Di
j
et al. 1992b). The situation differs much when one considers the lanthanide and actinide
compounds. In these -electron systems, much larger Kerr rotations up to some 10 have
j
been detected at low temperatures (see, e.g., Reim and Schoenes 1990). Very recently, the
exciting new discovery of a record Kerr rotation of 90 in CeSb was reported (Pittini et al.
1996). The stupendous Kerr rotation of CeSb is thus two orders of magnitude larger than
j
that maximally found in TM compounds, and about one order of magnitude larger than that
found in other -electron materials. Actually, 90 is the absolute maximum Kerr rotation
that can be measured. The singular MO behavior of CeSb and the underlying mechanism
attracted naturally a wide attention.
v#u#$./v55 &/.0¡.0¢£#1¤¥2#".02.0¡(¦ m

25§#uy825©¨
67Nª17:6 «Q¬$¬ >:C:RT<2E\CGVU$D§G®%¯±°²P´³PSRE\D
All MO spectroscopies are rooted in the fundamental property of polarized light to couple
to the orbital motion of individual, spin-polarized electron states, which is in turn coupled
to the spin through the spin-orbit interaction. Because spin and orbital polarization are at
the heart of all magnetic phenomena, MO spectroscopy techniques provide highly sensitive
tools for probing, on a nearly atomic level, those parts of the electronic structure which are
responsible for the magnetism. For this reason, MO spectroscopy in its various forms has
become widely used in fundamental as well as applied research.
If one considers which type of MO effect is used mostly, one finds that the vast majority
of MO investigations nowadays utilize the MO Kerr effect, the Faraday effect is employed
much less. The obvious reason is that for metallic materials the intensity of the transmitted
light drops off exponentially with the thickness of the material. The Faraday effect has
remained to be employed in investigations of semiconductors and insulators doped with, for
example, rare earth ions (e.g., Shen 1964, Shen and Bloembergen 1964). We mention that
the situation is different for X-ray MO effects. Here the magnetic circular dichroism, which
is equivalent to the Faraday ellipticity, is a commonly used technique for the investigation
of metals (see the following Section 1.3.2).
As already mentioned, the main technological application of MOKE is MO recording
(Mansuripur 1995). Besides this application, the Kerr effect has continually met with
new applications. For example, MO Kerr spectroscopy has recently been used to examine
surface magnetism (Liu et al. 1988, Liu and Bader 1991), and even to visualize the motion
of surface magnetic domains and domain walls (Schäfer and Hubert 1990, Hubert and
Schäfer 1998). MO Kerr spectroscopy has been applied to study properties of magnetic
multilayers, as oscillatory interlayer magnetic coupling (Bennett et al. 1990, Carl and
Weller 1995), MO enhancement effects due to plasma resonances (Katayama et al. 1988),
and recently also quantum-confinement effects in ultrathin ferromagnetic films (Suzuki et
al. 1992, 1993, 1998). Other recent fields in which MO Kerr spectroscopy was employed,
are the discovery of the formation of a new crystallographic phase in the Co-Pt phase
diagram (Harp et al. 1993, Weller 1996). In addition, the magnetocrystalline anisotropy of
single crystals, i.e., the dependence of magnetic properties on the magnetization orientation
with regard to the crystallographic axes, has explicitly been observed using MO Kerr
spectroscopy (Weller et al. 1994a, Weller 1996, Iwata et al. 1997, Katayama et al.
1998). The theoretical prediction dated back to Donovan and Medcalf (1965). Another
recent application is to use MOKE to record subpicosecond spin dynamics and magnetic
relaxation processes in thin magnetic films (Beaurepaire et al. 1996, Ju et al. 1998),
and also to visualize the temporal and spatial response to a magnetic pulse (Hiebert et al.
1997). It can be conceived that other, new, applications of the Kerr effect will be reported
in the future.
687µª17NM ¯¶°·P´³P¸REJD C:U¹E\AuPfº¼»=\<?¹PSU$P=\½8?o=\P½IC9¾¿P

The MO effects that were known till the seventies all involved optical transitions in the
valence-band energy regime, i.e., photon energies up to about 12 eV. Erskine and Stern
(1975) proposed that MO effects should also occur in X-ray excitations from core levels to
valence states. Ten years later, the first X-ray magnetic dichroism effects were discovered
by van der Laan et al. (1986) and by Schütz et al. (1987). Magnetic dichroism is,
for historical reasons, the term which became used instead of Faraday ellipticity. After
the initial discovery of X-ray MO effects, many other such effects have been discovered,
as, for example, MO phenomena in resonant X-ray scattering (Gibbs et al. 1988), and
X-ray transversal MOKE (Kao et al. 1990). A newly discovered phenomenon, which
had no counterpart in the valence-band energy regime, is that of MO effects in core-level
photoemission, using either circularly or linearly polarized incident light (Baumgarten et
al. 1991, Roth et al. 1993, Hillebrecht et al. 1995). In addition to the observation of new
effects, theoretical progress on sum rules stimulated the advance of X-ray magneto-optics.
Particularly, the theoretically derived sum rules for X-ray magnetic circular dichroism
(XMCD) (Thole et al. 1992, Carra et al. 1993) proved to be extremely useful in examining
magnetic properties of solids on an atomic scale. These sum rules relate the integrated
X-ray absorption spectra for left- and right-circularly polarized light to the spin and orbital
moments on a particular element in the material. Therefore element specific information can
be derived, which is a great advantage beyond valence-band MO spectroscopy. Although
a large number of approximations is involved in deriving the sum rules (see, e.g., Ebert
1996a,b), they work convincingly in practice. The spin or orbital moments are obtained
to an accuracy of about 10% to 20% (Ebert 1996a,b), but sometimes only 50% (Wu and
Freeman 1994). A recent discussion of various X-ray MO effects is due to Altarelli (1997).
687µª17Nª À*GVU1»>9C:U$P<= ¯¶°ÂÁ ;

The frequency of the reflected light detected in a standard MO Kerr experiment is the
same as that of the incident light. There can exist, however, a tiny fraction of light that is
reflected with a doubled frequency. This is known as second harmonic generation, or, more
generally, as non-linear optics (see, e.g., Shen 1984). Second harmonic generation appears
when the inversion symmetry is broken, which occurs for a centrosymmetric medium at its
surface. Pan et al. (1989) predicted that in the case of a magnetic surface layer there would
ò
be a MO Kerr effect in the second harmonic reflection. Experimental evidence for what
has become known as the non-linear MO Kerr effect (NOLI-MOKE) was first observed
from an Fe surface by Reif et al. (1991). NOLI-MOKE has then gained popularity as a
probe to surface magnetism and magnetic interfaces (see, e.g., Rasing et al. 1996 for a
recent outline). One particular feature of NOLI-MOKE is that the measured non-linear
Kerr rotations are in general an order of magnitude larger than the ordinary Kerr rotation
j
of the same material (Reif et al. 1993). The resolution of the non-linear Kerr rotation
is, however, with a root mean square error of about 1 much less than that of the normal
j
Kerr rotation (Reif et al. 1993). The latter can be measured to a resolution better than
0.001 , depending on the measurement technique (Schoenes 1992). A recent account of
the theoretical understanding of NOLI-MOKE has been summarized by Pustagowa et al.
(1996).
Ã
Å Ä Ã
The adjectives ‘linear’ and ‘non-linear’ appear to be used in a confusing manner. Linear X-ray
dichroism refers to linearly polarized incident light, but the effect is proportional to . Non-linear
Æ
MOKE, on the other hand, means that the polarization is not linear, but quadratic, in the electrical
field vector . There are other MO effects as Voigt or Cotton-Mouton effects, that are similar to linear
dichroism, which are called ‘non-linear’, because these are not antisymmetric in the magnetization
as the Kerr or Faraday effect, but symmetric, i.e., in lowest order quadratic in the magnetic field.
9
/ ( Ç
È É £3 ýËÊ
Ìk Í &-22¼5%
M179687:6 ÏÎ A$Pf¯<¸Ð1ÑxPS>:>/PSÒ1Ó$<¸E\C:GIU$D
The Kerr and Faraday effects appear because left- and right-circularly polarized light
waves propagate differently in a magnetic solid. The propagation in a magnetic medium
of any electromagnetic wave with wave length large compared to the interatomic distance
is governed by the Maxwell equations. These equations are macroscopic equations, in the
sense that they are obtained by taking suitable spatial averages over volume elements that
are large compared to the interatomic distances (see, e.g., Jackson 1975). The Maxwell
ÔÖÕH× ÂFØ´Ù$Ú
equations for the electromagnetic fields in a medium are
^ 2.1)
ÔÖÕJÛ ÏÚ ^ 2.2)
ÔÝÜÞ ÂFß©Ø à©á ß%À â × Ú ^ 2.3)
Ô+Üåä æ ßÀ â Ú âã Û
^ 2.4)
× the electric displacement, ä the electric âfield
ã averaged over microscopic regions
with
Û
on the atomic scale,
microscopic regions, and
Þ the phenomenological macroscopic magnetic field. Ù and à
the magnetic induction, that is, the magnetic field averaged over
× Þ
are the macroscopic (free) charge and current densities. Apart from the Maxwell equations,
ä
there are the material relationships, which relate the Û derived fields , , in the medium
× Þ¡ç ÕJä Þ ä á ÂFØSèéÚ

to the respective averaged microscopic fields and :
^ 2.5)
Û Þ¡ê ÕJÞ Þ Þ á ÂØSëìÚ ^ 2.6)
10 CHAPTER 2. THEORY
à Þ¡í HÕ ä
and also Ohm’s law
^ 2.7)
è is the electric polarization, ë ç ê í
the magnetization, and , , , the dielectric tensor,
îlï À ð´ï À î]ï

magnetic permeability tensor, and conductivity tensor, respectively. In the above notation,
ðï
the convention , , for the vacuum dielectric constant , and vacuum
permeability , respectively, has been adopted.
The material equations Eqs. (2.5)–(2.7) are, in the above given form, valid for homo-
geneous media. For inhomogeneous media, these have to be replaced by
× ò^ ñ¸Ú a óô ñ4õ óô õ ç ^òñ0öéñhõµ÷ æ õøa Õ ä ò^ ñõNÚ õøa´Ú ^ 2.8)
ã ã ã ã ã
and similar expressions for Eqs. (2.6), (2.7) (see, e.g., Callaway 1974). Here it has been as-
sumed that the origin of the time scale is irrelevant, and that in a periodic medium the choice
of the coordinate origin is irrelevant also. An electric wave field in an inhomogeneous
medium can, most generally, be written as a Fourier integral
ä ò^ ñ¸Ú a À úó ôHûóüô _ ä ^ û Ú_VaJXpýøþÿ ý Ú ^ 2.9)
ã ^½ÁØSaù
where û is the wave vector and _ the frequency. For monochromatic light in vacuum, the
above expression for the electric field reduces to the standard plane wave form
ä ^ ñSÚ a ä ï X ýøþÿ ý ^ 2.10)
ã
The negative sign to _ in Eq. (2.10) is the common convention, but sometimes the opposite
convention is used, which eventually gives rise to differently looking expressions for
magneto-optical quantities, or a different sign convention. A Fourier transform of Eq. (2.8)
leads to the material relationship in reciprocal space
× ^ û Ú:_Va ç ^ û Ú_Va Õä ^ û Ú:_Va ^ 2.11)

à ^ û Ú:_Va ;í Ý û Ú_Va Õä ^ û Ú:_Va
Analogously, Ohm’s law becomes
^ 2.12)
At this point it is appropriate to examine the relevance of the expressions with regard

to the physical application we have in mind. The materials of which the MO spectra are
investigated consist of unit cells with a lattice parameter of 2 – 10 . These are
certainly inhomogeneous on a length scale of less than 10 nm. Visible light, however,
ÁØ
has a wave length of about 600 – 800 nm. Consequently, there is no detectable resolution
ûæ Ï
on a unit-cell scale; the material appears homogeneous. Since , one can make
the approximation . In the following, the dielectric, magnetic permeability, and
_
conductivity tensors, and also the field quantities, are therefore considered to be dependent
only on the frequency .
CHAPTER 2. THEORY 11
For any electromagnetic wave in a medium, the Maxwell equations together with the
material relationships determine its propagation. Before we derive the equations describing
ê ^`_Va ¥
normal modes in a medium, an important point with respect to the Maxwell equations has
to be mentioned. It has been noted that even for metallic ferromagnets, one has to set
Lifshitz consider in their derivation of

ê ò^ _Va ¥
at optical frequencies (Landau and Lifshitz 1960, Pershan 1967). Landau and
the contributions to the averaged induced
current density
à ^`_Va â è ^ò_Va á ß Ô±Ü ë¶^`_VaTÚ ^ 2.13)

âã
side. The term ß
+
Ô Ü ë ^ò_Va becomes only non-negligible when averaging over volumes
and estimate the size of the first term against that of the second term on the right hand
ÔÝÜ Ï . The consequence is, with Eq. (2.6), that Û ^`_Va Þ ^`_Va and thus
ê ^`_Va ë± ^ò¥ _V. a Another
much less than an atomic volume. The second term can therefore simply be neglected,
consequence is that è becomes the only time-varying source term
current density is only a direct current. The fact that ^ò_Va

ê ¥ does of course not mean
to the induced current, under the common assumption that the macroscopic, extrinsic
that ferromagnetic materials cannot be described, but rather that the electric response
× ^`_Va
dominates the electromagnetic wave propagation. The reason for this situation is that there
Þ ^`_Va
è ë
exist ambiguities in the definitions of the material fields and at non-zero
ê ^ò_Va ¥
frequencies. The splitting into an electric polarization and a magnetic part is rather
arbitrary (see, e.g., Pershan 1967, Dolgov and Maksimov 1989). In a manner consistent
with the observation that at optical frequencies, it is always possible at all
_ æ Ï
frequencies to introduce solely an effective dielectric tensor, without the neglect of any
magnetic property (Pershan 1967). Only, for it does not converge to the commonly
used static quantity. Consequently, it turns out that it is appropriate to develop classical
optics on the basis of a single, intrinsic material quantity, which is the dielectric tensor.
Equivalently to the dielectric tensor, the optical conductivity tensor can be used to
describe classical optics. The relationship between the dielectric and conductivity tensor
_
follows from the material equations Eqs. (2.11), (2.12), through taking the time derivative
of Maxwell’s Eq. (2.1) (and assuming that there is no -component of the macroscopic
charge density)
ç ^`_Va ¥ á Â Ø í `^ _Va ^ 2.14)
_
ç
The dielectric and optical conductivity tensor are thus equivalent quantities. Another often
tensor K . The latter quantity is defined as the linear dependence of è on , i.e.,
used relationship is that between the dielectric tensor and the linear electric susceptibility
ä
èÏ^òñ¸Ú ã a ó ô ñõó ô ã õ K ^ ñ ö ñhõµ÷ ã æ ã õ a Õä ^ ñ4õµÚ ã õ a ^ 2.15)
æ Ï,
ç ^ò_Va ¥ á ÂFØ K ^ò_Va
Substitution of the Fourier transform of this expression in Eq. (2.5) yields, for
^ 2.16)
Note, that in Eq. (2.15) those contributions to è

that are non-linear in are neglected.
ä
These are the ones that are responsible for the second-harmonic generation and the NOLI-
MOKE. An extension of the theory to include such contributions has been outlined by
Shen (1984).
M7:687NM ÎÏA$P!=\PSDHUuP¸>P¸ÒyÓ$<2EJC:GIU
The propagation of normal light modes in a medium is described by the Fresnel equation. To
derive the Fresnel equation a monochromatic wave traveling in the medium is considered.
The phase velocity of the light wave in the material is less than the vacuum velocity. It is
convenient to define the complex refractory index "
" ß_ û Ú ^ 2.17)
_ _
which is the quotient of the velocity of light in vacuum and the phase velocity of light (i.e.,
# ) in the medium. Obviously, " depends on the frequency . The considered electric
ä ò^ ñSÚ a ä ï X ý ÿ ý Ú
wave field in the medium is thus
ã
$
%'&)(

2.18) ^
where " -^`_Va
is to be determined. Taking the time derivative of Maxwell’s Eq. (2.3) and
Ô ò ä á Ô ^ ÔÂÕä a æ ß À ò â ò × ò
combining with Eq. (2.4), one obtains the differential equation
^ 2.19)
× âã
K *
Using the material relationship Eq. (2.11) for , and inserting the above given wave form,
, ò ¥ æ ç æ ".-"#/ Õä Ï
the Fresnel equation results +
^ 2.20)
¥ ç æ ".-"2/ Ï . A
*
quantity, therefore the normal mode solution follow from 1JX , ò æ
ý
Here ".-" is the dyadic product, i.e., , ,0 . The dielectric tensor is a given, fixed material
peculiarity of the form of the Fresnel equation is, that there exist in general not 3, but at
ä ò^ ñ¸Ú a
the most only 2 independent vector solutions for " . Next to the solutions to the refractive
indices, also the corresponding eigenmodes for
eigenvectors of Eq. (2.20). ã
are required, that follow from the
medium, i.e.,
ç îï¥
, for which there is only one solution ,
ò îï
The simplest solution to the Fresnel equation is that for an isotropic, non-magnetic
. The form of the
dielectric tensor is, of course, often more complicated. Solutions to the Fresnel equation
for several types of dielectric tensors have been discussed in the literature (Landau and
Lifshitz 1960, Hunt 1967, Kahn et al. 1969).
All optical experiments relevant to MO phenomena involve either the transmission
L
through the material of a light beam coming from vacuum or its reflection back into the
Fresnel derived the equation in the early 19 354 century before Maxwell theory was developed, on
the basis of analogies to classical mechanics.
vacuum. Therefore, the solutions to the Fresnel equation in the two media have to be
supplemented with the appropriate boundary conditions at the interface. These follow
from the continuity requirements for the Maxwell’s equations and for the phase factor of
the wave, and give rise to the familiar Fresnel formulas for diffraction and reflection, and
also to Snell van Royen’s law (see, e.g., Jackson 1975). The complete set of equations
allows the treatment of all (linear) optical and MO effects.
kÌlk úr0#st#$8v.0¢87 -8-¢u¨ #wx#uy2

6
In the following, expressions are derived for the Faraday effect and for the three kinds
of Kerr effect, the polar MOKE (P-MOKE), the longitudinal MOKE (L-MOKE), and the
transversal MOKE (T-MOKE). As the current focus is particularly on the polar Kerr effect,
the case of P-MOKE is considered first in more detail.
M17µM17:6 ÎÏA$P ¬-GV><=¼¯±° Á P=\=¼P´³PSRE

The Faraday and Kerr effects are observed in the transmission or reflection of linearly
polarized light incident on the magnetized material. Linearly polarized incident light
is composed of equal-amplitude left- and right-circularly polarized waves. The familiar
notation for such waves is
ä:9 ò^ ñ¸Ú a ; À < ¥ï ^>= [@? A= g aHXpýøþCB)D E ý Ú ^ 2.21)

ã Á
where the wave propagates along the F -axis. Traditionally, the wave having positive
helicity, i.e., the electric field vector turns counter-clockwise when the observer is facing
in the on-coming wave, is called left-circularly polarized. This terminology is opposite to
what one would expect from the standard concept of ‘handedness’. Under time inversion
left-circularly polarized light becomes right-circularly polarized, and vice versa. Also,
upon reflection, an incident left- (right-) circularly polarized wave becomes reflected as a
right- (left-) circularly polarized wave.
To derive expressions for MO phenomena, the incident left- and right-circularly po-
larized waves are to be augmented to the corresponding modes in the medium, which
follow from the Fresnel equation (2.20). The dielectric tensor of a magnetized medium
adopts a particular form, which is related to the crystal symmetry and the direction of the
magnetization with respect to the crystal axes (see, e.g., Birss 1961, 1964, Kleiner 1966).
The general tensorial form can be complex, but for most practical situations it suffices to
ç
consider special symmetry arrangements. For such cases, a consequence of the magnetism
is the occurrence of non-diagonal elements in . A special symmetry arrangement is the
polar geometry, which is the geometry relevant to the Faraday effect and to P-MOKE. The
polar geometry has the magnetization parallel to the F -axis, and, if in addition the F -axis is
a symmetry axis of at least 3-fold symmetry, the related dielectric tensor has the following
form
î
GH
]
[ [
ç æ î:[]g 9î [][ Ïî ]
[ g Ï KL
^ 2.22)
Ï Ï îJIAI
Here it has been used that due to symmetry î g][¥üæ î []g (Boswarva et al. 1962, Kleiner
that the light is incident on the material parallel to the F -axis (i.e., " , = I )
1966). The normal mode solutions to the Fresnel equation (2.20) are, under the assumption
, ò9 î []@ [ ? @î []g ^ 2.23)

The corresponding electric field modes are
ä 9 ò^ ñSÚ a ; À < ï ^>=J[ ? A=|gFaHX ý $M D EN&)( ý ^ 2.24)
ã Á
the reflected right- (left-) circularly polarized mode corresponds to õO
Choosing the geometry such that the reflected waves travel in positive F -direction, then
ä ä (ä Põ äRQ ,
account of the non-zero î []g in Eq. (2.23), which is the quantity ultimately giving rise
respectively). The amplitudes and phases of the two reflected modes are unequal on
diagonal element vanishes, î []g Ï .

to MO phenomena. Only for a non-ferromagnetic material in this arrangement, the off-
The reflected light is elliptically polarized and its polarization plane is tilted over a
ä ä
amplitudes, the second to unequal phases of õO , and Põ . The quotient of the two complex
small angle with respect to that of the incident light. The first quantity is due to unequal
amplitudes can be written as
õ õ Xý ÀÀ áæ ÀÀ áæ
< <
O
õ õ
O
< P TS < P S VUW UX
, P
, P
, O
, O Ú 2.25) ^
S S
with Y the phase factor. Here the Fresnel expression for the reflection coefficients Z for
perpendicular reflection has been used (Born and Wolf 1980):
< 9
õ
, 9 æ À ^
á À
9
Z < 9 , 9 2.26)
If one considers the positions of the two reflected electric field vectors in the F Ï
z{
plane,
then the Kerr rotation is the angle over which the main axis of the polarization ellipse
is rotated with respect to that of the incident light. The Kerr rotation is therefore
z{ ÁÀ [^ Y P æ Y O a ^ 2.27)
At this point a sign convention enters the expression: zF{
is usually chosen to be positive
if the rotation vector of the polarization plane is parallel to the magnetization vector (van
Engen 1983, Fumagalli 1996). The latter is in turn chosen along the propagation vector of
æ
the incident light, i.e., F -direction. The Kerr ellipticity }|{
follows from the quotient of
the short and long axes of the ellipse, i.e.,
}{ õõ áæ õõ Ú
< < P
O
O
\S < S S < S
P 2.28) ^
S S S S
which can be rewritten as
<
õ À á l } { ^
õ À æ l } {
O
S< S 2.29)
P
S S
The sign of the Kerr ellipticity is said to be positive if the rotation associated with it has
its rotation vector parallel to the magnetization vector (see, e.g., van Engen 1983). From
Eq. (2.25), together with Eqs. (2.27), (2.29), one obtains for the polar MO Kerr rotation
and ellipticity
À á }{ X ò Vý ]A^ À á ,, PP À æ ,, OO ^ 2.30)
À æÏ } { Àæ Àá
ò
The above equation is the exact expression for polar MOKE, which is identical to the
expression given by van Engen (1983) .
It should be noted that in the literature on MO phenomena different sign conventions
have been adopted. The MOKE spectra reported by different groups can therefore differ up
to a minus sign. The sign convention in magneto-optics is, in fact, a long-lasting problem.
Faraday (see Faraday’s Diary, 1933) took an empirical stand and simply introduced the
positive sign of the Faraday rotation by defining the rotation of lead glass as positive. The
Kerr rotation of light reflected through lead glass should then also be positive, irrespective
of concepts of handedness that became developed later. The problem of the sign convention
has recently been discussed by Fumagalli (1996). It should be emphasized, though, that in
any case no physical relevance is related to the sign.
}{ z{
The Kerr rotation and ellipticity are for most materials less than . The above exact ÀFj
æ
expression can be approximated for small , , by an expansion to the leading order of
, O , P
z { á µ} {`_ æ
æ ÀÚ ^ 2.31)
, O , P
, , P
O
which once more shows that the difference in the propagation of the two normal modes in
Schoenes (1992) for the P-MOKE. Substituting , Oba P ? ò î9[][ cî9[]g

the material leads to the MO effect. Expression (2.31) corresponds to the one derived by
leads to
z { á µ } {`_ î []K [& ò ^@æ À î æ[]g î:[][a Z [][dc À Z Jù [] e g ý Z [][f K & ò ^ 2.32)
Ã á
In the expression given by van Engen (1983) the gh and j g i are interchanged with respect to
Eq. (2.30), but also the magnetic field vector is taken in k#l -direction, so that effectually the
expressions are the same.
Û Z4[]gh^ò_uÚ æ Û &a
I , while normally it is more
Zh[]g
Z\[]g4^ò_uÚ æ Û &a á æ JZ []gh^`_uÚ Û &a
Here one has to keep in mind that this holds for
appropriate to relate to I as, e.g., in the Hall effect. On account of the Onsager
I I one may rewrite Eq. (2.32) as
Û
relationship
z{ á µ}{ _ ZJ[][dc>æÀ Z []g ù) e^ ý ZJI []a [f K & ò Ú ^ 2.33)
á
which is the expression for P-MOKE mostly used in calculations. From the Onsager
Û
relationship together with Eq. (2.32) one recovers the familiar property that polar MOKE
Z []g ^`_Va
is antisymmetric in . MO measurements thus primarily determine the off-diagonal
conductivity , which is the quantity containing the magnetic information.
î []g
In the case that the incident light is not perpendicular to the material surface, but
incident at an angle m , the solution to the Fresnel equation is, to first order in

, 9 ò î [][ cî []g
n? oqp m r
9
Ú
2.34) ^
9
9
where m r are the angles of refraction of the two beams. To first order in one can replace î []g
m r _ m r in Eq. (2.34), where m r is the angle of refraction for the non-magnetic material.
Using the angular dependent analog of the Fresnel formula Eq. (2.26) (Rasing 1998), the
expression for the polar Kerr effect at an angle of incidence becomes, for s -polarized light
z{ á µ}{ _ Z [][ c æ À Z\[]g Jù e oqý pZ m[][ f K & ò ^ 2.35)

á
From this expression it is obvious that P-MOKE experiments carried out at a small angle m
differ only in second order in m from Kerr effect measurements performed at perpendicular
incidence.
M7NM17NM ÎÏA$Pé>GIU$½IC@E\Óut"C:U$<><8UutoEJ=\<8U$DqvIP=4DF<8>Ì¯¶°£Á*PS=\=ÂP´³P¸REJD
In the longitudinal Kerr effect (L-MOKE) and transversal Kerr effect (T-MOKE) the
magnetization direction is in the surface plane. These two effects are normally measured
for a fixed, non-zero, angle of incidence m . Their exact derivation is somewhat lengthy,

but an expansion to first order in the small quantity , O æ
, P is convenient and sufficiently
accurate. Approximate expressions for these MO quantities have been derived by Hunt
(1967).
For L-MOKE one can choose the surface normal parallel to the F -axis, the plane of
æ
incidence as the F `w –plane, and the magnetization parallel to the w -axis. The dielectric
tensor then adopts the form
î9[][ Ï î:[
Ïæ î î K ò î ÏÏ K îî Ï ò
GH KLzy GH KL
ç Ïæ î9[ î9g]Ï g î Ï
xI
{ ^
2.36)
xI JIAI K
Since î9[]g is equivalent to îx[ I through an elementary basis transformation, it is customary
to denote the off-diagonal component by î ò and the diagonal ones by î , î { . The solutions
K K
ò î K ? c îuò pb| m r9
of the Fresnel equation are
, 9 _ ^ 2.37)
z{ }{
A derivation similar to the one given above for P-MOKE leads to the approximate
expression for the longitudinal Kerr rotation } and Kerr ellipticity ~}
z { á N} { ^
, O xop m rO æ aò
, P oqp m rP oqp m
^
} } _ ,
xæ
O xop m rO , P xoqp m r
P
xop m

2.38)
9
After substituting the ,
îò
from Eq. (2.37), the expression for the longitudinal Kerr effect
becomes, to first order in the small quantity
z {} á µ} }{ _ Áh^À æ î æ a&îu^Nòî p|øæ Ápbm| ò m a K & ò Ú ^ 2.39)

K K
longitudinal Kerr effect disappears to first order in î ò , which is the reason that L-MOKE
again for s -polarized incident light (cf. Rasing 1998). For perpendicular incidence the
is measured only at non-zero incidence. Experimental studies of the L-MOKE spectra are
less frequent than studies of P-MOKE spectra. L-MOKE spectra of Fe, Ni, and permalloy
were studied by Lissberger (1961a, 1961b) and Robinson (1963). More recently, the
longitudinal geometry has been chosen in studies of the non-linear Kerr effect (Reif et al.
1993, Vollmer et al. 1996a, Rasing et al. 1996).
T-MOKE is different from both L-MOKE and P-MOKE, because it is not measured as
an angle, but as a magnetization modulated intensity difference of the reflected light. The
in-plane magnetization, which for T-MOKE is perpendicular to the plane of incidence, can
have two orientations, either in plus or minus -direction. The two reflection coefficients
have, for -wave modes, the form (Voigt 1908, Freiser 1968)
Z [
9
J^ a _ Z ? À î ò ò î ò æ ø î ÁK á ò Ú
p| m
^
K
2.40)
oqp m pb| m

where Z is the reflection coefficient for the non-magnetic material, and it has been assumed
îK îK
that { _
{
. The normalized reflectivity difference for the two magnetization directions
is the T-MOKE quantity , which is thus defined by
{ Á À ò J^ a ò æ J^ a ò
Q

CZ S
Z [
S S
Z [

S
2.41) ^
{
Á
We note, first, that in some investigations of T-MOKE the MO quantity
{
is defined
without the factor in the denominator (e.g., Martin et al. 1965, Ferguson and Romagnoli
1969), and, second, that the sign of depends once again on the choice of the 9positive
{
magnetic field direction. The expression for becomes, after substitution of Z [ from J^ a
Eq. (2.40),
W X î ò oqp ò 8Á m îuò æ p|øî Áá m pb|ø ò m

{ ^ 2.42)
K K
This expression exemplifies that T-MOKE is also directly related to the off-diagonal dielec-
tric tensor, and, like L-MOKE and P-MOKE, antisymmetric in the magnetization. Investi-
gations of the T-MOKE spectrum of various compounds have been reported (Krinchik and
Nuralieva 1959, Krinchik and Gushchin 1969, Martin et al. 1965, Ferguson and Romag-
nolli 1969, Voloshinskaya and Bolotin 1974), yet T-MOKE is much less frequently used in
spectroscopic studies as P-MOKE is. T-MOKE has over the last decades developed into a
ð ò
tool for analyzing surface magnetic structures. Magnetization modulated reflectivities can
ð ò
be detected with a resolution of approximately 0.3 0.3 ( m) on surfaces areas extending
over 30 30 ( m) . T-MOKE is therefore particularly useful as a tool to visualize surface
magnetic domain structures (Hubert and Schäfer 1998). In combination with the surface
magnetic sensitivity of NOLI-MOKE the transversal Kerr configuration has been applied
to obtain a high magnetic contrast from a magnetic Fe surface (Vollmer et al. 1996b).
M7NM17Nª ÎÏA$P!<=\<tu<´?¡P´³PSRE
An expression for the Faraday effect follows easily from the considerations made in the
derivation of the polar MO Kerr effect. For the situation of perpendicular incidence along
, 9 are given by Eq. (2.23), and the corresponding
ä
the F -axis, the indices 9 of refraction
æ
electric field modes by Eq. (2.24). Just like for P-MOKE, it is customary to choose the
magnetization vector along the propagation vector of the incident light, i.e., F -direction.
The transmitted modes in the medium travel in opposite Q ä ä
direction of the reflected modes,
( , respectively). Suppose that the
ô
therefore the right- (left-) circular mode is now
material has a thickness . Then, after the modes have propagated through the material,
both the phases and the amplitudes of the two circularly polarized waves have become
unequal. The quotient of the two electric field vectors is, with Eq. (2.24), given by
<
^ ñSÚ a <
Xý Xý ^
^ ñÚ ã ã a
O O
< P
I

S < P S VUW UX
S S
~W X &)( 2.43)

Note, that after the light waves have left the material, both the phases and amplitudes are not
changed anymore. Multiple reflections, that can occur within a low-absorbing medium, are
z }
neglected in the present formulation (see Donovan and Medcalf 1963, Niess and Kessler
1989). The Faraday rotation ~ and ellipticity ~ are commonly defined by adopting the
same conventions as adopted for the polar Kerr rotation and ellipticity. Keeping in mind
that the propagation direction is now opposite to that used in the polar Kerr configuration,
the expression for the Faraday effect becomes
À æ l } X ò ý ]Ì X ý ~W X &)( ^ 2.44)

À á l }C
For historical reasons the Faraday ellipticity is better known as the magnetic circular
dichroism. An expansion to first order in the small quantities yields
z á N } :_ _ Á ßô ^ , O æ , P a _ _ Á ßô @î î9[]K [][& gò Ú ^ 2.45)

which, in terms of the components of the conductivity tensor reads
z á µ} _
ZJ[]gh^ á Û I&a ò Á Ø ß ô ^ 2.46)
c>À á ùJ e ý Z\[][f K &
Here ZJ[]gh^ á
Û I&a qæ ZJ[]g\^ æ Û I&a is the off-diagonal conductivity defined for the mag-
netization in á F -direction. Both the Faraday and Kerr effects are evidently caused by
the non-zero off-diagonal conductivity and are antisymmetric in the magnetization. The
Faraday effect has mostly been used to investigate MO properties of transparent semicon-
ductors. A survey of the experimental data accumulated until 1960 has been given by Jaggi
et al. (1962), while Schoenes (1992) has reviewed data collected from 1960 till 1992.
The Faraday and the Kerr effects were for a long time the only known MO effects
that are antisymmetric in the magnetization. Very recently, a new MO quantity has been
z y^ò_Va
introduced, the optical Hall angle (Spielman et al. 1994, Kaplan et al. 1996). The optical
Hall angle C is defined by
z å_ z WX ZZ\[][][g Ú ^ 2.47)

Û
where the magnetic field I is chosen in á F -direction. z ^`_Va is the equivalent at op-
*
tical frequencies of the so-called Hall angle, z ^´Ï|a _ WX Z\[]gh^´ÏFaJZ\[][J^½ÏFaA/ (see, e.g.,
Campbell and Fert 1982). The optical Hall angle can be measured in microwave magneto-
z z z{
transmission experiments (Spielman et al. 1994, Kaplan et al. 1996, Lihn et al. 1996).
From equation (2.47) for ~ and the respective expressions for and it is clear that the
optical Hall angle and the Kerr and Faraday angles are closely related quantities. The re-
lationship between the zero-frequency Hall angle and the Faraday rotation at a wavelength
of 633 nm has been studied experimentally for a large group of alloys by Hartmann and
McGuire (1983), who find a nearly linear dependence.
kÌ .0#u-¸348#u5§.02#t8r0#u¨
ç í
To evaluate any of the MO spectra from the expressions derived in the preceding Section,
either the dielectric tensor , or equivalently, the conductivity tensor has to be calculated.
The theory which enables such a calculation is the linear-response formulation for the
conductivity tensor, originally due to Kubo (1956, 1957). Its derivation is given in the
following.
M17µª17:6 Q P¸UuP=\<8>/®lG=h¾ Ó"><2E\CGIU

In linear-response theory one considers a many-particle Hamiltonian , which consists of
two parts
ï á K ^ ã aTÚ ^ 2.48)
where ï
is the unperturbed Hamiltonian, which is time independent, and defines
K
the interaction of the system with a time-dependent external field. The density matrix
corresponding to is
Ù^ ã a X q¡ ¢§Ú ^ 2.49)

with £ À¤¥ and ¢ is the partition function, ¢ § ¦S©¨ Xx«ª^ æ £© )a ¬ . A density matrix
Ù|ï is similarly defined for the unperturbed Ìï .
The aim of the linear-response approach is to approximate the expectation value of an
operator to first order in the perturbing field. The expectation value of an operator is
defined by ®
^ ã aN¯ §¦S©¨ ÙT^ ã ab°¬¥Ú

^ 2.50)
which is time dependent through Ù^ a . The time dependence of Ù is described by the
Liouville-von Neumann equation, i.e.,ã
ô ÙT^ ã a À ^ a]Ú9Ù^ a²/

*
ô ã A± ã ã ^ 2.51)
It is assumed that for ¶ æ ³ there is no perturbing field, so that ÙT^ æ³ a Ùhï . With this
initial value Eq. (2.51)ã can straightforwardly be integrated. To first order in
becomes
K the result
Ù ^ ã a _ |Ù ï á AÀ ± ó ô ã õ K ^ ã õà]Ú9ÙFï/8Ú
*
^ 2.52)
´
where the superscript µ denotes that the observables are in the interaction picture, i.e.,
Ù ^ ã a X ^ /ï ã Ja@ÙT^ ã aX ^ æ /ï ã Ja
x«ª ¶A ~± x«ª A ± . As a direct result, the expectation value of
can be written as
® ® ® *
ï á À ± ó ´ ô ã õ

¥^ ã a)¯
_ °¯ ^ ã a]ÚJ K ^ ã õà / ¯ ï ^ 2.53)
Here ·
ø ï
¸
ÙJï
is the expectation value with respect to . Eq. (2.53) is the general linear-
ï Ï
response expression. Two further assumptions are commonly made: First, it is often
assumed that there is no effect in the unperturbed equilibrium state, i.e., ·²
K ^ãa ^ãa
. Second,
¸
it is assumed that the perturbing Hamiltonian is of the form º¹¼» , where ¹
is a time independent operator and » ^ãa
a scalar field, containing the time dependence.
Eq. (2.53) can then be written in the standard linear-response form

·½ ^ ã a ¸ _ ó ô ã õV¾©^ ã æ ã `õ a » ^ ã õ aTÚ ^ 2.54)
´
where ¾ is the response function, ® *
©^ ha À ± z4^¶¿ha
¾ ¶¿ ^ ha&Ú ¹ ½^ ÏFa / ¯ ï Ú
¶¿ ^ 2.55)
with ¿ ã æ ãõ
ãõ
. An important property of the response function is its causality: its value
ã
at stems from contributions from the field » at an earlier time .
M17µª17NM ÎÏA$PfRTGVUut$Ó"RE\CAvuC@E?EJPSU"D|G=
To derive the linear-response expression for the conductivity tensor, it is convenient to
ä â ä X ý
evaluate the current induced by a perturbing time-varying electric field,
From Ohm’s law it can be recalled that the conductivity is the sought response function to
.
ã
the electric field, i.e., À

U ^ ã a ó ô ã õøZ ^ ã æ ã õøa ^ ã õ a
U
<
2.56) ^
´
y
The interaction of the electrons with the electric field is given by
Õ ä
K § ÁÂ ý ñ ý ^ ã a Xy ý «ÃÄ
¹ ^ 2.57)
Q
Here ñ are the positions of the electrons, and _ _ á A . The small imaginary term
ý
AÅ has been added to assure convergence properties. The expression for the total current
à Æ , with Æ the volume of the system,® thus becomes
À *
Ç
^ a ± ó ô ã õ Ç U ^ ã a]Ú ¹ ^ ã õ a / ¯ ï X ý «ÃÄ¶È
U ã ^ 2.58)
´
Partial integration leads to

® * ® *
^ãa À ^ ã &a Ú ^ ã a ï X æ Ný _ æ À ± ó ô ã õ

Ç Ç
Ã
Q Ç X
^ a]Úd¹ É ^ ã õ a²/ ¯ ï æ µý _ Q ^ 2.59)
Ã È
U ã
¹ ²/ ¯
U ± U
´
¹ É is the time derivative of ¹ , which is again related to the current through
¹É ^ ã a ÊÁ Â ñ É ý Õä æ Å ^ ã a Õä ^ 2.60)

ý
The first
® * term of Eq. (2.59) can be rewritten
® * using the expression for
ò K
Ç
^ ã a]Ú ^ ã a ï æ ò
U ¹ ²/ ¯ @Á Â ZÉ aU
ý ^´Ï|a]Ú ^½ÏFa ï
bZ 0 a A/ ¯
< Á ±
ËÍÌ U
<
^ 2.61)
ý
0
ã Ï
*
ñ ý Ú9ñ æ ý
Note, that at the operators in the interaction picture are equal to those of the
Schrödinger picture, and that the commutator value É 0 / ± 0 Ë has been in-
serted. The full expression for the induced current thus becomes
^ãa
ò ® *
X ý æ À ó ô ãõ

^ ]
a Ú ^ `
õ a X ý
< «Ä
Ã ¶È
ï µ_ ^
Ç Á Ì Q U < Ä

Ã Ç Ç
U _Ë A±
´ ã ã
U / ¯ À
Q 2.62)
U
Ç
Comparing this expression to Eq. (2.56) for the current density
U Æ , one directly
ò
finds for the conductivity tensor
® *
Z U ^ ã æ ã õ a ÁÆ Ë Ì _ UQ J^ ã æ ã õ a á Æ ±FÀ _ Q Ç
U ^ ã æ ã `õ a]Ú Ç ´^ Ï|a / ¯ ï ^ 2.63)
As is done in most treatises of the linear-response formulation the indices µ denoting the
í `^ _Va
interaction picture are suppressed. A Fourier transformation yields the linear-response
expression for
ò
Á Ì UQ Q
´
¿ ¶¿
® *
Z ^ò_Va _ á FÀ _ ó ô zh^ ha ^ ha]Ú ^´Ï|a ï XSý ^

Ç
>¿
Ç
²/ ¯
ÄÃÄÎ
U Ë U 2.64)
Æ Æ ± ´
Here the substitution ¿ æã ã õ has been made. An alternative derivation for í ^ò_Va which
leads to the same expression has been given by Callaway (1974).
The original expression for the conductivity tensor derived by Kubo (1956, 1957) looks
Z
different from the expression (2.64) obtained in the above derivation. Kubo wrote a double
integration in his equation for , i.e.,
K
´ ®
^ á Ja ´^ ÏFa ï XSý Ú
&)ÏNÐ
b¥Z ^`_uÚ "a ó ô ó ô
¿
Ç
U ¶¿ A±
Ç
N¯
«Î
^
ï ï
2.65)
U
but, except for the volume not included by Kubo, the two expressions are identical (see, e.g.,
Verboven 1960). Eqs. (2.64) and (2.65) express the conductivity in terms of current-current
correlations.
Z
Eqs. (2.64), (2.65) can be rewritten in various, equivalent forms (e.g., Zubarev 1961,
Wallis and Balkanski 1986). A valuable, short notation for
U , which is often found, is
ò
obtained when matrix elements of the current operator are exploited:
Z U ^`_Va À X ÒÓ &)ÏNÐ æ X Ò Ó È &)ÏNÐ ·½Ô S Q Ç U S Ô õ < ¸ ·[Ñ Ô õ S Ç < S Ô Ñ ¸ Ú

_ á Æ_
Á Ì UQ Q
Æ
Ë Â ÑÑ
È ¢ ±F_ æ ^ Èæ a
^ 2.66)
where Ô and
< Ñ
are the many-electron state and its energy, i.e., Ô ï < Ñ
Ô , and
·½Ô õ
Ç S ¸
Ô are the current matrix elements (Wallis and Balkanski 1986).
S S ¸
S ¸ S ¸
M17µª17Nª 1Õ C9Uu½I>:PT»¬"<2=\EJC9R´>P®lG=4¾ Ó$>:<¸E\C:GIU

The conductivity is, to first order in the perturbing field, given by an exact formal many-
particle expression. However, for practical applications it is undesirable and even im-
possible to evaluate the Kubo equation on the basis of many-particle states. The only
feasible and workable approach is to rewrite the Kubo equation in terms of single-particle
states. Single-particle formulations of the linear-response conductivity expression were
derived soon after Kubo’s initial work (Chester and Thellung 1959, Verboven 1960). Such
expressions can be obtained by representing the many-particle wave function by a Slater
determinant of single-particle states. Alternatively, one can integrate out the time integral
in Eq. (2.64) using second quantization. The current operator in second quantization is
(see, e.g., Gross and Runge 1986)
õ ^ ã a ^ ã aTÚ
Q
Ç
U ^ãa æ
Ë
Á
Â · , ,
È S U S ¸
È 2.67) ^
~
where ,
S ¸
and ,
S ¸
õ Q
label the one-particle states, and , , are the creation and annihilation

respectively, for a particle in state , . The many-body field operator is given by

operator,
Ö
Ø× ,
S ¸
, while the Hamilton operator ©ïS ¸
is expressed by
õ Ú
Q
ï
Â ·, ,
È SÙ S ¸
È 2.68) ^

with
Ù
æ ò á
ÚÜÛ Ý , the Hamiltonian operating on a single-particle, which is to be specified
later on. The time dependence of the creation and annihilator operator follows from the
Heisenberg equation (in the interaction picture)
ô ^ a ^ a&Úb/ï / ØÂ · , , õ È ^ a î ^ aTÚ

*
±
ôã ã ã È SÙ S ¸ ã ã ^ 2.69)
Q

states of . Performing a time integration of Eq. (2.69) gives ^ a ^½ÏFaHXª ¨Fæ @î ~±¬ .
and the complex conjugate expression for . The one-particle states are chosen as eigen-

® ã ýã
U ^ ã a and ^ ã a in Eq. (2.64), and carrying out the
Ç
Ù
í
integration, the equation for ^ò_Va becomes
After substituting the expressions for
*

Á ò Ì UQ Á ò Q
Z U ^`_Va y Æ Ë _ á Ë ò Æ ±F_ Â È Â È y _ Q æ _ Ï È Ï È / ¯ ï ~U È ÏxÏ È Ú ^ 2.70)

Q Q
È Ú
~ ÏÏ Þ Þ
with ±F_ È
î È æ î , and the definition U È · , S U S , õ ¸ . Both terms still contain the
many-body states in® the Þ
* expectation values. Evaluating these terms yields
Q

È
Q
Ú ï
Ï Ï
È/ ¯ ^Nî a æ ^Nî a Ú
® Ï È È Ï ß ß È 2.71) ^
^ î a ï
Q
and also Ì
×

^ î a
ß , where ß
í ^`_Va
¯ is the Fermi-Dirac distribution function
(see Gross and Runge 1986). The resulting expression for in terms of single-particle
ò ò
states is
Z U ^`_Va ^ î Q a æ ß^ î È a U È È ^ 2.72)

_ á Ë ò Æ ±F_ Q
Á Ì Q Á ß
Ë U
È _ æ _
Â
Æ È
Þ Þ
~
í
This is the desired single-particle expression for ^ò_Va . Note, that its form is similar to that
of the many-electron expression (2.66). Physical implications of the formula
* (2.72) are
Ú u
discussed in the subsequent Section. ® There is an important way in which Eq. (2.72) can be
À
rewritten, which is often quoted in the literature. Using the commutator Z [ / ± U
, ¯ , the first term becomes U
and the closure relation à×
S S
^ î a« U Â ß^Nî a· , S Z U Ú> / S , ¸ À ±

*
Ì U Â ß
ß^ î a æ ß^ î È a

Â È A± Z U È
~
È
æ Ë À ± Â È ß^Nî a _ æ ßÈ ^Nî È a ~U È È

~ Þ
^ 2.73)
~ ~
Þ Þ
a shorter single-particle expression for ^`_Va

í
The first term in Eq. (2.72) can now be joined together with the second term, which yields
ò
Z U ^ò_Va æ Ë ò Á ± Æ Â È ß^ î a _ æ ßÈ ^ î È a _ Þ Q U È æ Þ _ ~ È È ^ 2.74)
~
~

^ î a X S^ æ a
Once again, the many-electron expression (2.66) can be coined in the very< same
form, but using the many-electron quantities, and with ß
replaced by «ª £ J¢
(e.g., Sols 1991). Another approach that also leads to Eq. (2.74) is the random-phase
approximation (RPA) to the dielectric tensor of a many-electron system (Nozi ères and
Pines 1958, Ehrenreich and Cohen 1959). A remark would be in due course that this
î
formulation works under the condition that the true excitation spectrum of the many-body
system is well described by the , that is, the quasi-particle picture holds with quasi-

particle state , .
S ¸
M7Nª17
¶á â =\GI¬-P=JE\CPSDxGI®¥E\AuP RTGIU "Ó$RE\C uCE?EJPSU$D|G=
ut Av
Ô ³½¨³´ª¬¸hë½³´ù ¸Óª¬³´ãà¸¸ãB¸x²wªe¾
One of the customary extensions of the expression for
í ^`_Va
is that to finite lifetime effects.
Optically excited states decay after a certain relaxation time, and emit again the absorbed
photon, withouty which nothing can be observed in any detector. Lifetime effects can be
Q
_ _ á _ á
introduced in a phenomenological manner by replacing in the above given expressions
by A C¿ , where ¿ is the lifetime, which may be state dependent. This
is a simplified way for introducing relaxation time effects, which gives the correct limit
expressions for ¿ ä³ (Ehrenreich and Philipp 1962). However, one has to be cautious
how precisely the replacement _ _á C¿ is carried through, because in some cases this
ç ^ò_Va ¥ á ÂØ í ^`_Va _

has been done erroneously for the RPA dielectric function, which lead to a violation of the
-sum rule (cf. Mermin 1970, Garik and Ashcroft 1980). Such ambiguity can already be
recognized from J . The proper manner to include lifetime effects is
the number-conserving relaxation-time approximation, which has been derived within the
density-matrix formalism (Mermin 1970, Garik and Ashcroft 1980). For the single-particle
ò
expression (2.74) this leads simply to
Z U ^`_Va æ Ë ò Á ± Æ ^ î a æ ß^ î È a U È ~ È

ß
^ 2.75)
_ _ æ Þ _ Þ È á ¿
Â
È È
~ ~

is convoluted with a Lorentzian, of which ¿ K is the half width at half maximum. In

The physical effect of the lifetime parameter is that each infinitely sharp optical transition

í
of the finite lifetime is that the interband contribution, i.e., ,§ å , õ in Eq. (2.75), does not
just the same manner the many-body expression for can be modified. A consequence
vanish at _ Ï (see Garik and Ashcroft 1980).

Also the intraband contribution, i.e., , , õ , in Eq. (2.75) is accordingly modified
ò
through the finite relaxation time. This part becomes
Z U r[æ¶ç ^ò_Va æ Ë Á ò Æ
â ß î è[é Þ _ ~U á Þ ¿~ Ú
Â ^ 2.76)
â
conductivity, i.e., Z8^`_Va ZTï~J^@À æ µ_ê¿ha . For
_ Ï the intraband conductivity Z r>æ¶ç approaches a finite value, which is proportional
which adopts the Drude form for the
to ¿ . Eq. (2.76) is then identical with the Boltzmann equation in constant relaxation-time
_ Ï
approximation (Jones and March 1973). An exact reduction of the many-particle expression
ï
for the limit has been given by Chester and Thellung (1959) and Verboven (1960),
who introduce an additional scattering potential to the Hamiltonian . The scattering of
Z
the electrons from this potential is responsible
r>æ¶ç for the relaxation time. This leads to lowest
order in the scattering strength to the of Eq. (2.76), but higher order terms also exist.
U
More general, it can be shown that the the Kubo formula for DC transport is equivalent
to the extended quantum Boltzmann equation with correction terms (Hansch and Mahan
1983, Chen and Su 1989).
It is known that the intraband relaxation time may be different from the interband
relaxation time. The intraband relaxation time is the combined result from various electron-
scattering processes of electrons in the vicinity of the Fermi energy (Wooten 1972). It can
in addition be sample dependent, e.g., through the amount of defects in the material.
Consequently, the intraband Drude contribution is not so easily captured theoretically.
The constant intraband relaxation time is the simplest, lowest order approximation (cf.
Verboven 1960). Yet, even in this approximation one generally doesn’t know its sample
dependent value. Only for a number of materials the average Drude relaxation times have
been measured. In actual calculations estimated values are therefore mostly used. The
interband relaxation time, on the other hand, depends rather on the frequency: Excited
states with energies high above the Fermi energy decay more rapidly than low-energetic
states (e.g., Santoni and Himpsel 1991). While such a frequency dependence can be taken
into account in calculations, already the constant relaxation-time approximation suffices

for limited frequency ranges. Approximate values for the constant interband relaxation
time have been deduced from a comparison of experimental and calculated MOKE spectra
(Oppeneer et al. 1992a). For TM compounds, the value of = 0.03 Ry has been found to
be on average a good estimate of this phenomenological parameter.
í°É°ãã¸Óª¬«eÉä¼T«®O¼9¸¹«Êªe³´¸x¾
í ^`_Va
Kubo (1956) already analyzed the symmetry properties of the many-body linear-response
expression for . The physical current is defined as the real part of Eq. (2.56), and
since the conductivity tensor follows from a Fourier integral, one obtains for
í ^`_Va
Z©Uë ^`_Va Z U ^ æ _VaTÚ ^ 2.77)
which immediately implies
uW XSZ U ^`_Va WuXSZ U ^ æ _VaTÚ

d@eQZ U ^`_Va æ d@eQZ U ^ æ _Va ^ 2.78)
U ½^ ÏFa
A consequence is that d@eQZ Ï , and WuXSZ U ^½ÏFa constant. Furthermore, the
Onsager relation can be shown to hold
Z U ^`_uÚ Û a Z U ^`_uÚ æ Û a´Ú ^ 2.79)

which follows from time reversal accompanied by a reversion of the magnetic field (Jones
Z []g ^ò_Va éæ Z g][ ^ò_Va

and March 1973). For the off-diagonal conductivity this leads – with the crystal symmetry
property – to
ZJ[]g4^ò_uÚ æ Û &I a éæ Z\[]gh^`_uÚ Û I]a ^ 2.80)

These symmetry properties are fulfilled also in the finite lifetime approximation.
Æ È¸h¼dë´¶O¾¬ã¶ ù½«®¸ìâ±T¸Ó¨T²wÉ
An important quantity in the analysis of optical spectra is the plasma frequency. The
dielectric response of an isotropic gas of classically moving, free electrons to an oscillating
_ò Ú
electric field can be expressed by
ÂØ ò
y
î^`_Va À æ
Ë
Æ
Ì Á
y ²¿
_$^ò_ á K a
í
À æ
½¿ $_ ^`_ á K a 2.81) ^
_
where í is the plasma frequency, _ í ò ÂØ Ì Á ò Ë Æ ¿
_ å_
, and the relaxation time. The
physical meaning of the plasma frequency is that for ïî í an electromagnetic wave is
damped in the plasma, while for .ð í _ _
_ _
the wave propagates through the plasma. The
reflectivity of the plasma vanishes therefore above í , but below í the plasma is highly
reflecting. Such a behavior of the reflectivity is observed for alkali metals, for which the
itinerant electrons can be well treated within the free-electron approximation. Already for
the noble metals this approximation becomes doubtful; however, the reflectivities of many
metals and compounds do display a steep decrease in the spectrum, which is a signature of
a plasma edge. The conductivity corresponding to the classical electron motion induced
by the time-varying electric field is just identical to the conductivity obtained within the
ò N_ ò
Drude model (Drude 1900a,b), i.e.:
y
Äñ æ>òZó
Ë^ò_Va
Æ
Á Ì
²¿
í
^`_ á ²¿ K a ÂØ^ò_ á K a
2.82) ^
The Drude model lacks of course all the ingredients that are essential in the energy-
band theory of electronic states in metals. The latter is undoubtedly the better, quantum-
mechanical approach to the conductivity. Comparing the single-particle, linear-response
ò
expressions (2.72), (2.73), to that for the Drude conductivity, leads to the identification
_ ò æ FÂ Øò
y
í Ë ±
Á
Â
ß ^ î a æ ^ î a
ß È
È _ È
S S
òÚ
2.83) ^
Æ È
Þ
~
í ^`_Va
which can be considered as the quantum-mechanical definition of í . The second term in _
the linear-response expression (2.72) for is neglected in the Drude model. This term,
however, is important in the high-frequency spectrum.
_
There exists another quantity, which is called the intraband plasma frequency, or
Z J^@À æ µ_ a
the í defined
Z
sometimes also plainly plasma frequency, but should not be confusedr>æ¶with ç
jñ2 ê¿xñ ,
_
above. The intraband part of Eq.(2.76) adopts also the Drude form
from which an expression for the intraband plasma frequency ô í can be derived
y
Z ^`_Va ÂØ^ò_ _ á
ò
Ka Ú ^
ô í a
U
²¿
2.84)
UqU
_
with ô í a the components of the intraband plasma frequency, which are given by
U y
_ ô íò a U æ ËÂ Ø ò Á Æ
ò
â ß î Cè[é S ~U S ò
Â ^ 2.85)
â Þ
^ ß î]a è[é qæ J^ î æ a . From
ò ò
<
Eq. (2.83) it can be recognized that _ ô í is precisely the
At zero temperature, one can furthermore exploit y
â â part of _ í . In systems
ò _ ô íò a [ _ ô íò a g _ ô íò a I . The intraband
with cubic symmetry it further holds that _ ô í
intraband
plasma frequency can straightforwardly be computed from the single-particle band states.
However, the corresponding relaxation time ¿ , which may be state dependent, is not known,
and can only be estimated.
Æ;ÈT¸õ«®¶Iã¸Ó«®¾ðºJõ«®C¨T³´·b«®¸xë´¶ ªe³Á¨¾
In the experimental analysis of optical spectra the Kramers-Kronig (KK) relations (Kramers
1926, Kronig 1926) are an important tool to derive either the dispersive part of the optical
conductivity from the absorptive part, or vice versa, even if one part of the spectrum is only
known in a finite frequency range (see Milton et al. 1997). Or, if both the absorptive and
dispersive part of a spectrum have been measured in ellipsometry, the consistency of the
measured data can be tested from the fulfillment of the KK relations (Azzam and Bashara
1986). Furthermore, the KK equations can be applied to extrapolate spectra which are
measured in a finite frequency range, to frequencies outside the experimental boundaries.
The physical reason for the KK equations is solely causality; the response of a system
í ^ò_Va
to an impulse can obviously never precede the incident impulse. The Fourier integral for
can be rewritten as
y
í ^`_Va ó ô í ^ ha@X ý ó ô zh^ ha í ^ haX ý Ú
´
¿ ¶¿
«Î
´
¿ ¶¿ ¶¿
Î
2.86)^
´ ´
í
since for ¿î)Ï one has ^¶¿ha Ï (see Eq. (2.56)) and therefore one may insert a z -function.
Substituting for zh¶^ ¿ha the integral representation (e.g., Zubarev et al. 1997)
zh¶^ ¿ha Í À
÷~öø | e Q ï ÁØ ó ô ù ù X æ ý ú ½û Ú
´ Î
^ 2.87)
´
one obtains for
í
í ^`_Va Í÷~öø | e Q À ó ´ ô _ õ í ^`_ õ a ^ 2.88)
ï ÁØ ´ _ õ æ ^ò_ á ½ûá
After separating on both sides the real and imaginary parts, each part can be expressed as
an integral equation of the other part. An abbreviated notation for the real, imaginary part,
Z U ò d@e¼Z U Ú
of the conductivity which will y be used in the following,
y is
Z U K WXSZ U Ú ^ 2.89)
respectively. The expressions that result from Eq. (2.88) are the traditional KK equations
(Kramers 1926, Kronig 1926), which read
Z U K ^`_Va À ´
À ò ´
Á _õ ò
Ø@ü ó ´ ô _Iõ _ õ æ _ Z U ^`_Iõ a Ø@ü óï ô _Iõ _ õ ò æ _ ò Z U ^`_IõàTÚ
òZ ^`_Va æ À ó ´ ô _Iõ À Z K ^`_Iõ a éæ Á ó ´ ô _Iõ ò _ ò Z K ^ò_Iõøa
U Ø@ü ´ _ õ æ _ U Ø@ü ï _ õ æ _ U
^ 2.90)
denotes the principal value.
ü Since causality is such a fundamental physical concept, in practical applications the
KK equations are assumed to be valid, and measured spectra are not used to verify the KK
equations, but rather the accuracy of spectra is examined from their fulfillment of the KK
formulation of .
í ^`_Va
equations. The KK equations are valid both in the many-particle and in the single-particle
There exist general conditions for complex functions to satisfy KK-type relations,
which rest in the applicability of the Cauchy integral equation to that function (sometimes
called Meiman theorem). As a consequence, KK-type relations can be derived for a broad
æ
variety of response functions (cf. Jahoda 1957, Velick ý 1961). Smith (1976a) derived
Q
KK-type of equations for the MO quantity , , in the polar geometry:
Á ó ô _ õ ò _ õ ò ò d@e ^`_ õ a æ `^ _ õ a 8Ú
* ´ *
WX ^ò_Va æ
, Q , ²/ ^`_Va
ØT_ ï _ õ æ _
ü
, Q , ²/
^ò_Va æ ^`_Va æ Ø Á ó ô _Iõ _ õ ò _ æ õ _ ò WX ò^ _Iõà æ ^ò_Iøõ a ^

* ´ *
d@e , Q , ²/ , Q , A/
ï
2.91)
@ü
z }
The Faraday rotation and ellipticity are approximately proportional to , _$Y æ b
Q ,
z
(cf. Eq. (2.45)), and consequently, can be written as an integral of ~ , and vice versa. }
The exact KK relations for the Faraday effect quantities are
Áz ^`_Va æ ØÁ ó ô _Iõ _ õ ò _ æ õ _ ò ø À æÏ } 2^`_ õ a Ú

´
À á } ^`_ õ a

ö
ï
@ü

ÀÀ áæ ll }} ¸^ò^ò_V_Vaa Á

´
_
Ø óï ô _Iõ _ õ ò æ _ ò Á z ^`_IõàTÚ ^

ö 2.92)
C @ü

}
These can be approximated for small ; the resulting expressions are given below.
Related KK expressions have been derived by Schnatterly (1969) and Smith (1976b) for
9 9
the polar MO Kerr effect. The reflection coefficients Z and the phases Y are connected
through the relations (Smith 1976b)
ò^ _Va $ æ Ø Á ó ô _ õ _ õ ò _ æ õ _ ò ^ò^ò__ õõ aa Ú
* ´ Q
Y
Q
^ò_Va æY / ö
Z
@ü
ï Z

Á ó ô _Iõ ò _ ò ò^ _Iõøa æ ^`_Iõ a ^
*
ø Z
Q
^`_Va ´
Q
ö
Z

^`_Va

Ø ï _õ æ _
@ü
Y Y / 2.93)

z{
}{
Using the definitions (2.27), (2.29) of Section 2.2.1 for the polar Kerr rotation and Kerr
z{ }{
ellipticity , these equations can be rewritten in the form of KK-type relations between
and :
Áz { ^`_Va æ @ØÁ ü ó ô _Iõ _ õ ò _ æ õ _ ò öø ÀÀ áæÏ }}{{u^ò^ò__ õõ aa Ú

´
ï´
À Á

Ï
æ

} { ò
^ V
_ a _
öø
À á } { ^ò_Va

Ø ü óï ô _Iõ _ õ ò æ _ ò Á z{1^ò_Iõøa ^ 2.94)

}{
These are exact, and adopt, interestingly, precisely the same form as the KK relations for
the Faraday effect. For small the approximate KK relations become
z{u^ò_Va _ Á ó
´
ô I
_ õ ò _ õ ò }{u^`_Iõ aTÚ
Øü ï _õ æ _
Á
} { ^ò_Va _ æ Ø@ü ó ô _Iõ _ õ ò _ æ _ ò z { ^`_Iõà
´
^ 2.95)
ï
z }
Again the equations for C and C are precisely the same. It comes as a surprise that
the polar Kerr rotation and ellipticity are connected through approximate KK equations,
because this kind of relationship is not at all obvious from the expression (2.33) for polar
MOKE in terms of the components of the conductivity tensor. Probably for that reason
the KK equations for MOKE are not widely used. Nevertheless, the KK equations for the
polar MOKE quantities have been tested in several occasions, in which they were found to
be reliable (e.g., Kielar 1994, Kučera et al. 1998).
í.±Tãk«®±ë½¸Ó¾
Other powerful spectroscopic tools for analyzing optical spectra are the so-called sum rules
(Wooten 1972, Altarelli et al. 1972). Sum rules state a relationship between an integral
over a spectrum and a certain quantity. Most well-known is the ß -sum rule, which reads
´
ó ô_ Z U KU ^`_Va Ø ÁË ò Ì _ í ò Ú ^ 2.96)
ï
Â
U
Í Á Æ Î
and its off-diagonal counterpart
´
ó ô _¥Z U K ^`_Va ÏÚ ^>Y å £¸a ^ 2.97)
ï
¸ Z K
The sum rule provides the non-trivial information that the integral over the diagonal
conductivity ¦ ¨ Uq U ¬ is related to the total number of particles in the volume, whereas
Z []Kg
the integral over the off-diagonal conductivity vanishes. Like the KK relations, the
sum rule is valid for both the many-particle and the single-particle formulation of
U . Z
The ß -sum rule is also valid in the finite lifetime formulation. It is convenient to start from
the following many-particle expression for Z
U , which is equivalent to Eq. (2.64):
®
Z ^ò_Va À ó ô zh^ ha
´ *
U ± Æ
¿ > ¿
Ç
U ¶ ¿ b ¾ ² / ¯
Q
^ ha&Ú ´^ Ï|a ï XSý ý Ú
ý
Î
2.98) ^
´
y
Z a . Substituting Eq. (2.87) and carrying out the _ -integration leads to
where ¾
ý
Á ×
ý ® *
ó ô _ÌZ U ^`_Va æ ± À Æ ó ô ¿ ó ôù X ù ý æ ú ýÏ ý Q J^¶¿ha Ç U ^>¿ha]Úb¾ ^½ÏFa²/ ¯ ï Ú ^ 2.99)

´ ´ ´ Q Î
´ ´ ´
where the integrals are easily performed. After inserting the commutator value, the ß -sum
rule Eq. (2.96) is obtained.
î
The sum rule Eq. (2.96) can be reformulated in several ways as a sum rule for other
,
î K
optical quantities, such as the dielectric tensor
U , the refractive indices , or the energy-
loss function, (see Wooten 1972, Altarelli et al. 1972, Smith and Shiles 1978).
U
_ Ï
There exists another kind of equation, which is derived from the KK equations, and
which is also named sum rule. In the special case of , the first KK equation (2.90)
yields the sum rule expression
ò
Z U K ^´ÏFa Á üØ ó ô _ Z U _ ^`_Va
´
^ 2.100)
ï
The Hall conductivity Z []Kg ^´ÏFa
and DC diagonal conductivity can thus be obtainedZ []K[ ^½ÏFa
from an integral of the absorptive part over all excitation energies, which again is a useful
test of a measured absorptive spectrum. In a similar manner, other sum rules of this kind
can be derived from the KK equations for the Faraday effect and the polar Kerr effect.
From Eqs. (2.92), (2.94) one obtains
Á ó ´ ô _ À öø À æÏ } { ^ò_Va ÏÚ ^ 2.101)

Ø@ü ï _ À á }{u^ò_Va

because z { ^´ÏFa Ï }
. An identical expression is found for . For the Faraday rotation one
can in addition derive (Smith 1976c)
´
ó ô _©z¸^`_Va ÏÚ ^ 2.102)
ï
which follows from Eq. (2.91). These expressions are apparently rarely used in the analysis
of MO spectra; however, they do hold potential for performing the interpolation of the Kerr
ellipticity or Faraday effect spectrum to either zero or high frequencies, which are the parts
of the spectrum that can normally not easily be measured.
In MO spectroscopy one would of course like to have sum rules that are special for
MO spectra. It turns out that the derivation of such sum rules is not a straightforward
matter. The usefulness of MO sum rules, however, has very recently been underlined
by the newly derived MO sum rules for XMCD spectra (Thole et al. 1992, Carra et al.
1993). These XMCD sum rules provide element specific information about the spin and
orbital moment of an element in a compound from the X-ray absorption spectra of left- and
right-circularly polarized light (see, e.g., van der Laan 1996, Altarelli 1997). The element
specificity of the XMCD sum rules is due to the fact that electrons excited from selected
core levels go dominantly into empty conduction states of the same element, because
the overlap of the core levels with conduction electron states of other elements is much
smaller. This is not the case for MO spectra in the valence-band energy regime, where
there is consequently in general no element-specific part of the MO spectrum. Several
attempts have been undertaken to derive MO sum rules for valence-band MO spectroscopy
(Bennett and Stern 1965, Erskine and Stern 1973, Smith 1976b). The obtained sum rules
were, however, derived under extreme limiting assumptions, namely, for a paramagnetic
material in a small magnetic field, with zero induced magnetic moment, neglect of SO
_IZ []òg ^`_Va

interaction and of crystal-field induced orbital splitting. Under these assumptions, Bennett
and Stern (1965) obtain that the spectral integral of is in first-order proportional
to the (nearly vanishing) applied magnetic field ¹dI (see also Erskine and Stern 1973).
However, since MO effects of ferromagnets depend crucially on the SO interaction and the
spontaneous magnetization, these chosen conditions do not represent the relevant physical
situation. The SO interaction provides the necessary coupling between the electric field
Z []g
perturbed electron motion and the spin moment. Argyres (1955) showed – by applying
perturbation theory with respect to the SO coupling – that is to first-order linear in
the SO coupling, stemming from the product of the current matrix elements. More recent
direct calculations of MO Kerr spectra showed that the Kerr effect itself scales linearly
with the SO interaction (Misemer 1988, Oppeneer et al. 1992b), and, consequently, that
ferromagnetic materials computed without SO interaction yield a zero Kerr effect.
A MO sum rule for oscillator strengths of atomic transitions was proposed by Smith
(1976c):
Â
Q

'ß È ß È _

æ
Á·, I , Ú ^
2.103)
È ~ ~ ± Sþ S ¸

where the oscillator strength is defined by ß È
9
~
Ë ^Nî æ î a õ

È , ? w ,
S · S S ¸ S ~± , and
ò, , Ú õ
label the atomic states. This expression can be considered as the MO analog of the ß -sum
rule for oscillator strengths of Kuhn (1929). It should be noted, though, that Eq. (2.103) is
only valid for atoms with one electron. For atoms with more electrons the approximation
is made that the Hilbert space of the unoccupied states is equal to the Hilbert space spanned
Z []òg
by all states.
Another MO sum rule, which relates the spectral integral of to a part of the orbital
moment, was derived recently by Oppeneer (1998). This sum rule reads, for the polar
geometry,
´

Á ò
ó ô _ÌZ []òg ^`_Va Á Ø ð æ ð ë
^
ï Ë )ÿ ¤ ÿ ¤ 2.104)
y Æ
Here
ÿ
æþ ¸
ð
· I ¤ ± is the F -component of the total orbital moment in the volume Æ ,
whereas
ÿ
ë is a certain, small part of the orbital moment, which is defined below. The
total angular momentum in the volume · I is defined as · I · I , with the :^ ñ a
þ ¸ þ ¸ ¸
relativistic momentum. Making use of the closure relation, and rewriting ñ in terms of
,
via
Ë

Y Ú:ñJbIÚ ^ 2.105)
±
the expression for · I can, in single-particle notation, be rewritten
* as

d@e ~[ È g È /
þ ¸
Á±
· I Ë ^
á îÞ æ îÞ È
Â Â Â Â 2.106)
þ ¸
(A(
a È È È
(A(
Z []òg
The second double sum on the right is the term that defines ë . Comparing the expression
Z []òg ^ò_Va
for · I to Eq. (2.74) for establishes that both quantitiesÿ are closely related. Integration
þ ¸
of
Z []g
yields the MO sum rule Eq. (2.104).
This MO sum rule is consistent with the observation that both and · I are, to first
þ ¸
order, linear in the SO interaction. Experimental investigations of Eq. (2.104) have not yet
been carried out. While it is certainly of interest to know that the orbital moment and the
off-diagonal conductivity are connected, the fact that the integral does not amount up to the
orbital moment only appears a hindrance for practical applications. Test calculations show
that represents the major contribution to the spectral integral, and that ë is small, but
doesÿ not vanish. ÿ
One may compare to the XMCD sum rule for the orbital moment of a -element, which i
reads (Thole et al. 1992, Carra et al. 1993)
Q
ó ô _ ð ^ò_Va æ ð ^`_Va Í ^
Â
· I

Ì
_ þ ¸ 2.107)
Ì ±
9 9 9
ß ð Á_©d@e
ð
Û
, (Born and Wolf
i
Here is the absorption coefficient, which is given by
1980), Ì is the number of unoccupied -states, and Ì is defined by
Ì
Îß Ø ó ô _ ð ^`_Va á ð ^ò_Va

Q

2.108) ^
Û
ò Ü
ò, 9K & ò ò Ü & ò
The part of the spectrum that is considered extends over the and
ô 9
edges only, i.e.,
9

ßð
ð _
excitations from the and core states to the valence bands are considered.
ÂFØ WuXIY Z\[][ ? @ZJ[]gb

Using Eq. (2.23) for , one can rewrite and approximate for large by
. Substituting this expression in the XMCD sum rule (2.107) yields
_
ó ô _¥Z [] òg ^`_Va _ Â Í ÌÌ ÿð¤ Ú ^ 2.109)

Û
which has a resemblance to Eq. (2.104). In practical applications of the XMCD sum rule
ô
(2.107), one still has to make the suitable choice for the spectral range, which may result

in a scaling of both Ì and . Furthermore, the number of holes in the -band has to be
estimated, which introducesÿ a second uncertainty. None the less, these uncertainties have
been mastered for a variety of transition metals and compounds (e.g., Chen et al. 1995,
Arvanitis et al. 1996, Altarelli 1997).
Ìk #$.0lT¨©34(n.0y8&/.08r0#u¨
M7¶áu7:6 x<8D|C9R®lG=h¾ËÓ$><2E\CGVU
The numerical evaluation of the optical conductivity requires knowledge of the single-
particle wave functions and their energies, in the sense of quasi-particles. The density-
functional theory (DFT) in the local spin-density approximation (LSDA) is the appropriate
formalism that provides explicit expressions for single-particle wave functions and en-
ergies. The DFT makes a formally exact mapping of a many-particle problem onto a
single-particle problem (Hohenberg and Kohn 1964, Kohn and Sham 1965), but only for
the ground state, not for the excitation spectrum. Hohenberg and Kohn considered the
many-particle Hamiltonian for Ì identical electrons with charge Á , i.e.
á *á æ
y
À ò á [ ^ñ ýa á À ò Ú ^
Á ý K! ý ý K Áý ñ ý æ ñ
r Á
¥ Æ Â Â Ý Â 0 2.110)
ó 0
S S
where ¥ is the kinetic energy operator, the electron-electron repulsion operator, and Æ
represents the interaction with a given external potential Ý [ r , which is a function only
of the position of the particle. ó
In the present case of interest of electrons in a solid, the
y
^òñ K Ú:ñOòFÚ p Ú9ñ a òÜ
external potential contains the Coulomb interaction of the electrons with the static nuclei.
Ö Ö

To find the ground state of the system, one would have to solve
Ö < Ö

, which, for electron numbers as large as Ì = 10 , is a much too entangled
and complex task to be carried through. Hohenberg and Kohn showed that instead of
ÙT^òñ4a
searching for the ground-state wave function in a complex Hilbert space of 3 Ì variables,
one can equivalently search the ground-state energy using the particle density , which
is defined by
Ù^ ña ·
Ö
Â
K J^ ñ æ ñ ý a Ö
Ú óüô ñ4Ù^ ña Ì ^ 2.111)
ý
S S ¸
Before outlining the theorems of Hohenberg and Kohn, we note that Levy (1982) and Lieb
(1983) gave a formulation in which some restrictions of the initial formulation, e.g., for
systems with a degenerate ground state, werey overcome. As a first step, Hohenberg and
Y Ùb
Kohn showed that the external potential is a unique functional (apart from an additive
Ù
constant) of the ground-state density, Æ Æ . Second, Hohenberg
< and
< Kohn
< showed #" $"&% Y ÙFb
Ù Ù"
for any trial density , it holds for the ground-state energy that , and,
[
for , the ground-state energy is obtained. The total ground-state energy for the
potential Ý r can then be written as
ó
<'" Y [ b e ( ø $) Y ÙFb e (+ * Y ÙFb á óúô ñ [ ^òñTa@ÙT^òña-,Ú
<
^
Ý r | | » Ý r 2.112)
ó ó
under the constraint of particle conservation, Eq. (2.111). » is defined as (Levy 1982)
» ä| ø · ¥
Ö
S
Y Ùb 0 /1. (
Ö
S ¸ á2 T ^ 2.113)
Ù
Y Ùb
The infimum is taken over the manifold of all many-body wave functions that yield . To
calculate »
Y ÙFb
would thus imply the full solution of the many-body problem, which one
wanted to avoid. Consequently, only model guesses of » can be made. In principle
Ù
the second theorem offers the possibility to search the ground-state minimum of the total
Ö
Y Ùb
energy with respect to , which is much easier than with respect to . The Euler-Lagrange
<
variational principle to minimize under the constraint of particle conservation therefore
reads
Ù Y Ùb á î\ æóüô ñÙT^òñTa ÏÚ ^

<
Ì 2.114)
î
where is the Lagrange multiplier.
While thereby in principle a method to find the total-energy minimum is known, the
Y Ùb
practical step to applications needs yet to be made. This step was undertaken by Kohn
and Sham (1965), who devised a way of splitting-off from » well-defined parts. For a
non-interacting system, for which the Hohenberg-Kohn theorems hold likewise, the many-
Ö
3546 ^òñTa
particle wave function can be written as a Slater determinant of single-particle states
. The idea is to rewrite the total-energy functional on this basis, and to separate
those parts that are explicitly known. Eq. (2.111) for the many-particle density exactly
gives
ÙT^òñTa 3 4ë 6 ^òñTa 3546 ^òñTa ^ 2.115)

ý
Â
K a87 9:9
Since now , the » Ï Y ÙFb of Eq. (2.113) reduces to the kinetic energy ¥¸ïY Ùb of the
ò
non-interacting particles
Tï æ 3546 3546 ^ 2.116)

¥
ý
Â
K a;7 9:9
·
S Á S T
¸
1Y Ùb
Also the electron-electron repulsion term in the total-energy functional can be reformulated
<
by splitting-off the known part, which is the Hartree energy . In the case of interacting
1Y ÙFb Ù
particles one can therefore write the electron-electron repulsion term as a sum of the
<
Hartree energy , for which an explicit expression as a functional of is known,
and the remaining part that contains the exchange interaction, and also correlation effects.
ï Y ÙFb
Similarly, the kinetic energy of the interacting particles is not known, but one may separate
the kinetic energy of non-interacting particles ¥ , and group together the remaining
[ 9 Y ÙFb
correlation term with the exchange-correlation< part of the electron-electron repulsion into
the unknown exchange-correlation energy, , i.e.,
» Y ÙFb ï|Y FÙ b á Y ÙFb á [ 9 Y ÙFb
¥
<

<
^ 2.117)
with respect to
3 46 the variation in Eq. (2.114) with respect to Ù , one can make the variation
Instead of making
ë , i.e.,
3 4ë 6
<
Y Ùb á î 46 Ì æ óô ñuÙT^òñ4a ÏÚ ^ 2.118)
with î46
the Lagrange multiplier. This leads to the Kohn-Sham equation for the single-
ò
particle wave function:
y æ Á á Ý ={ < ^òñTa 354 ^ ña î 4-3>4 ^òña´Ú ^ 2.119)

{=< Ý r [ á á [9
Ý , and the superfluous particle label has< been dropped. The
Ý
Ù [9
with Ý
ó potential are defined by Ý ~ , and Ý
[ 9 Ù
Hartree and exchange-correlation
<
~ , respectively.
[9
The Kohn-Sham approach gives an exact reformulation of the many-particle problem
into a much simpler single-particle problem, provided Ý is known. Minimizing the
[ 9 ^òña
ground-state energy is equivalent to solving the Kohn-Sham equation. The latter can
[9
only be solved if a suitable approximation is made for the unknown Ý . Kohn and
Sham (1965) already proposed to use approximate expressions for Ý that are derived
[9
from equivalent expressions for the homogeneous electron gas. This is the local density
<
approximation, in which one writes for :
y
<
[ 9 Y Ù b óô ñ\ÙT^òñTac} [ 9 ^òñ4aTÚ ^
2.120)
where } [ 9 ^òÙT^òñTa:a is a function of ÙT^òña , which depends only on the local value of Ù at
ñ , and not on a part of real space about ñ . This is expectantly valid for slowly varying
densities. For } [ 9 one uses ‘pointwise’ for each value of Ù^ ña the value that } [ 9 ^òÙT^òñTa:a
polarized homogeneous electron gas the exchange-only part of } [ 9 is given by (Kohn and
would have for the case of a homogeneous electron case. For example, for a non-spin
Sham 1965, see also Slater 1951)
Í Á ò ÍÙ K &
Ü
}[\^òÙJa éæ Á ÎØ ^ 2.121)
Soon after the initial LSDA-DFT work extensions to treat spin-polarized systems, to
include correlation in the LSDA exchange-correlation potential, and to treat relativistic
corrections, were formulated (e.g., Hedin and Lundqvist 1971, von Barth and Hedin 1972,
Rajagopal and Callaway 1973, MacDonald and Vosko 1979, Ceperley and Alder 1980).
In addition, numerous applications of the LSDA-DFT in electronic structure calculations
of atoms, molecules, and solids have appeared. Detailed surveys of the fundamentals of
LSDA-DFT and of its applicability to describe the electronic ground state of real systems
can be found in the literature (Williams and von Barth 1983, Jones and Gunnarsson 1989,
Dreizler and Gross 1990, Eschrig 1996). The foremost outcome of the by now extensive
investigations of the LSDA-DFT is that the approach is capable of describing with high
accuracy the ground-state properties of electron systems. Weak points of the LSDA-DFT
approach are: the slight ‘overbinding’, i.e., the calculated lattice constants and bond lengths
come out systematically too small by 2 – 3%; the size of the band gap of semiconductors
is too small; the description of localized electrons is unsatisfactory. The last point is
related to the fact that the LSDA is derived from the homogeneous electron gas, which, for
materials having itinerant electrons, is a suitable approximation, but not anymore for the
inhomogeneous case of localized electrons. Also heavy-fermion systems, for which the
energy bands near the Fermi energy are substantially renormalized due to quasi-particle
interactions, are only modestly explained by the LSDA.
M17>áu7NM ?°ÕA@¼« <¬$¬"=\G<8RhA¡EJG¿E\AuPfP´ÐuRĆ@EJ<2EJC:GIUoDF¬-P¸REH=4Ó$¾

The single-particle eigenenergies that are obtained within DFT are ground-state entities,
which are not equal to the single-particle excitation energies of the many-electron system
(Hedin and Lundqvist 1969). Actually, the Kohn-Sham energies were introduced mathe-
matically as Lagrange multipliers, which had nothing to do with excitation energies. None
the less, the Kohn-Sham single-particle energies have become widely used in the interpre-
tation of excitation spectra, often with obvious success. The appropriate starting point for
ò
the discussion of excitation spectra is the Dyson equation for the quasi-particle excitations,
æ Á á`Ý ó [ r ^ ña á Ý ^ ña 3 ^òña á ó ô ñõCB ^ ñSÚ9ñhõ Úî]a 3 ^ ñ4õ a î 3 ^ ñaTÚ ^ 2.122)
with B ^òñSÚ:ñ õ Úî]a the non-local, and non-hermitian, self-energy operator. Comparing the
Dyson equation to the Kohn-Sham equation, shows that the evident approximation made
if the quasi-particle excitation energies are replaced by the Kohn-Sham energies, is
B ^òñSÚ9ñ õ Ú9î]a _ J ^ ñ æ ñ õ a Ý [ 9 ^ ña ^ 2.123)

Numerical investigations showed that this approximation is surprisingly accurate for the
homogeneous electron liquid at metallic densities, particularly for energies close to the
Fermi energy (Hedin and Lundqvist 1969). Moreover, the Fermi level itself is correctly
given by DFT (e.g., Kohn and Vashista 1983). Therefore, the Kohn-Sham energies con-
stitute a very reasonable approximation to the quasi-particle excitation energies of weakly
correlated systems.
The above already provides arguments for the approximation of excitation energies
by the Kohn-Sham energies, yet, there exist some further arguments that deserve to
be mentioned. Kohn and Vashista (1983) pointed out that in the linear-response for-
mulation of the optical conductivity the occurring expectation value (see Eq. (2.64)) is
Ö
evaluated with respect to the many-body ground-state wave function . Therefore, the
Ù^ ñ4a
linear-response conductivity tensor becomes a fundamentally unknown functional of the
unperturbed ground-state density . To go beyond the LSDA-DFT approximation to
the excitation energies would imply that corresponding corrections to the linear-response
formulation should be taken into account as well. Improvements beyond the approxima-
tion of the Kohn-Sham energies as excitation energies are found in the approach DFE
to the self-energy (Hedin and Lundqvist 1969). The DFE
approach gives, for instance,
better values than the LSDA-DFT for the band gaps of semiconductors (Hybertsen and
Louie 1985). Further improvements of the DFE
, like vertex corrections and dynamical
quasi-particle effects, can be included (e.g., Bechstedt et al. 1997). Yet, in spite of the
attempts to improve sequentially the quasi-particle description of the excitation spectrum,
the LSDA-DFT approximation yields still dielectric functions for diamond-structure C and
Si that are closer to the experimental spectra, except for the gap value (Bechstedt et al.
1997).
Ìk G Hn)+#u p) 5&#")+#$./ .

M78I17:6 ÎÏ$A P =\PS><2E\CAvuC9DE\C9RKJ"<Ut DFEJ=hÓ$RE\Óu=hP
Since magneto-optical effects are of relativistic origin, the relativistic variant of the Kohn-
Sham equation has to be solved, which is called the Kohn-Sham-Dirac equation. It is
instructive to begin with the Dirac Hamiltonian Rñ for a particle in a scalar potential Ý ,
L
without vector potential (see, e.g., Rose 1961)
ñ
3{ ßNM -Õ O Qá P Ë0ß ò Ká RÝ 3 { < 3{ Ú ^ 2.124)
3 M P
with { the four-component wave function. The 4 4 matrices , , and R
í ¥ Ï ¥ ¥ Ï¥
are defined by
M íÏ Ï Ú P Ï æ ÚR Ï Ú ^ 2.125)
í ¥
with the standard Dirac matrices, and the 2 2 unit matrix. In contrast to the non-
relativistic operator, the Dirac operator is not bounded from below, which means that there
are two branches of energy solutions: the upper part formed by the electron states and the
lower part reinterpreted as positron states. Naturally, for electronic structure calculations
only the electron part has to be dealt with.
performed, which has the electronic four-current density

Ç Å TS ^ Ù4Ú a
To arrive at the Kohn-Sham-Dirac equation, a relativistic DFT derivation has to be
as central quantity:
US
The ground state of a spin-polarized, many-electron system is a unique functional of
Ç
(Rajagopal and Callaway 1973). Making the local spin-density and a certain other
approximation, the Kohn-Sham-Dirac equation can be obtained (e.g., MacDonald and
Vosko 1979, Eschrig et al. 1985)
ßNM -Õ O Vá P Ë0ß ò Ká RÝ-áQPÝXW í =ÄI Õ í I 3 { þ cNî þ á Ë0ß ò f 3 { þ ^ 2.126)

where the energy origin has been shifted, î á Ë0ß ò . Here the Kohn-Sham potential is
y <
Ý {< ÝTá¼Ý W í , in which Ý W í ^òñ4a contains the spin-polarized part of the exchange-correlation
Û
potential corresponding to the F quantization axis. All other parts
contained in Ý ^òñá . The abbreviation I Ý W í = I is often used, signifying that the spin-
y of the potential are
Û < Y Y ÙÚZQb> Z , with
í
dependent part of ÝXY causes an internal magnetic field, i.e.,
Z · ¸ the spin density.< The additional* approximation made< inâ Eq.( (2.126) isâ that the
(
functional Y is adopted, [ which implies that in Y the orbital contribution
to the magnetization density Z Á K Á
[non-relativistic
kept (Eschrig et al. 1985).
(
á ]\ / is neglected, and( only the spin density is
The task to be accomplished is to solve the Kohn-Sham-Dirac equation. This has in
the recent years been done in various approximations for most of the widely-used band-
structure schemes, viz. the Korringa-Kohn-Rostoker (KKR) method (Korringa 1947, Kohn
and Rostoker 1954), the linearized muffin-tin orbital (LMTO) method (Andersen 1975),
the augmented spherical wave method (ASW) (Williams et al. 1979), and the augmented
plane wave (APW) method (Slater 1937). A relativistic KKR scheme was developed by
Feder et al. (1983), Strange et al. (1984), and Schadler et al. (1987). A detailed account
is due to Weinberger (1990). Ebert (1988) developed a relativistic version of the LMTO
scheme, and for the ASW scheme this was done by Takeda (1979), and by Krutzen and
Springelkamp (1989), while a relativistic linear combination of atomic orbitals program
was developed by Richter and Eschrig (1989). An early relativistic APW scheme was
made by Loucks (1967).
A particular approximation to solve the Kohn-Sham-Dirac equation is the scalar-
relativistic approach with SO interaction included in a second-variational manner (see,
e.g., Andersen 1975, Koelling and Harmon 1977, Takeda 1978, MacDonald et al. 1980).
ô
It is based on the observation that the SO interaction is for most materials a small quan-
tity, which could be included in a perturbative way. For example, for 3 metals the SO
interaction shifts the band energies only in the mRy range. To obtain the scalar-relativistic
equations, the four-vector is expressed in two spinor components through {
3 _ ^/^bà
3
. The
small component can be eliminated from the equation through
ß í ÕcO 3 ^ î æ Ý æ ÝXW í í I]a 3 Ú ^ 2.127)
which leads to a formal equation
3 î 3 for 3 , with
ß ò O0Õ e^ d á d ò í I&a O á í Õ-O d ÜAO á O I ÕfO d ò ÜgO
K í ÕhO í Õ-O K
æ I dHò æ dHò I á`Ý-á`Ý W í í I Ú ^ 2.128)
î Ë0á ß Á ò Ë0ß ò æ ò Ý ò ÚVdHò
and
d K ^Nî á Á æ Ý a æ Ý Wí ^Nî á Á æ Ý a ò æ Ý Wò í Ú
0
Ë ß ò ÝXW í ^ 2.129)
d;ò
see, e.g., Harmon and Koelling (1977), Sticht (1989). The meaning of the various terms
can more easily be understood by considering the case Ý í = 0, so that W
= 0. This leads
ò ± ò
to
æ ò ß ò ^ Ô Ý a Õ`Ô á ± ò ß ò í Õ c Ô Ý AÜ O f á`Ý Ú ^ 2.130)

qÁ Â Â

y
with Ë
á ^Nî æ Ý ab¿Á ß ò ÀCJ^ Á ß ò d K a . The second term is the Darwin correction,
and the third term the SO coupling. Without 3 the SO coupling, which is the small quantity,
^ þ Ë sa . Thus, in the first step the SO term is left out, and then it is added again to the
the equation can be solved for a function having the non-relativistic quantum numbers
Hamiltonian in the second-variational step, which is carried out in every self-consistent

iteration. It turns out that the second-variational approach leads to results that are in
excellent agreement with those of fully-relativistic calculations (MacDonald et al. 1980).
It should also be mentioned that in spin-polarized calculations based on the Kohn-Sham-

Dirac equation first the non-spin-polarized equation is solved, and then the spin-dependent
potential term is treated in a second-variational manner (Ebert 1988). With regard to our
aim, the evaluation of MO spectra, it appears thus that there are several relativistic band-
structure techniques available that all comprise both the relativistic and the spin polarization
effects.
M78I17NM ÎÏA$Pé¾¿<2EJ=hCÐüPS>PS¾ PSUVE\D

The optical transition matrix element (ME)

is the crucial quantity in numerical inves-
tigations. Much of the success in the computation of an optical spectrum depends on the
Z´[]g
MEs. This holds already for calculations of the non-magnetic optical spectrum, but even
Z [][
more for the MO spectrum. The magnitude of the off-diagonal conductivity, is about
S S
Z [][ Z []g
10 to a 100 times smaller than . Any error in the calculated MEs, that might cause
S S
an error of about 1 – 2% in , can result in much larger errors in . There are several
sources for numerical errors, of which the two major ones are: basis-set incompleteness,
affecting the quality of the final-state wave functions, and (non) hermiticity of the MEs.
The first problem is related to the variational manner in which the Kohn-Sham equation
is commonly solved. A finite, incomplete set of basis functions m is usually chosen, e.g.,
4
3 ^ ña 4ji 4 4 ^òñTa
atomic orbitals, muffin-tin orbitals, from which the Bloch wave functions are constructed:
× m . The wave-function coefficients and energies are solved in
i 4 î

a Rayleigh-Ritz variational approach, from
È
S ¸
· 3 S
È
· S ¸ |
3 æ î 3 3
2.131)
4 Uk e ø Ú ^
which leads to a linear algebraic equation for the ’s. It is well-known that the variational
i
î
energies are accurately obtained thereby, even for a small basis-set, but the variational

wave functions are not. The energies are accurate to second order in deviations from the
minimum, but the wave functions only to first order. Therefore, while most band-structure
methods provide reasonably accurate band energies, the accuracy of the wave functions
can often be quite poor. These deficiencies in the wave functions become dominating in
the higher unoccupied states, where no basis functions are available anymore to precisely
represent those Bloch wave function. These unoccupied states are obviously of direct
importance to the MEs. One solution to the incompleteness problem is to increase the
basis set as much as possible, for example, by adding plane waves to the basis.
The second problem, that of hermiticity, arises in the computation of the MEs when
the customary division of the potential in Wigner-Seitz or atomic spheres centered about
the nuclei is adopted. Obviously, the MEs ought to be hermitian, but in calculations based
l4
on atomic spheres only this is not the case. The integral over the unit cell (UC) can be
subdivided in two parts, one part extending over the atomic spheres , and a part extending
over the interstitial:
3 O 3 ô ñ 3 ë ^ ña O 3 ^ ña á ó ô ñ 3 ë ^ ñ4a O 3
È · È
:mon Â 4 rp ó q È È ^ ña
ý ;s O W
~ S S ¸
^ 2.132)
If the interstitial contribution is neglected, the MEs will not be hermitian. The neglect
of the interstitial contribution can lead to deviations of hermiticity in the MEs of about
20% (Oppeneer et al. 1992a). This can be avoided by performing the integration over the
whole UC, which is computationally more demanding, or by including the interstitial in
an approximative way.
The limited basis-set is the main source of errors in the MEs. The effects of the basis-set
size has been investigated for several of the commonly-used band-structure schemes, the
KKR, LMTO, ASW, and APW methods. In the following, the non-relativistic description
of the MEs is considered, because all problems inherent to the numerical evaluation are
already contained in it. Also, before mentioning explicit numbers, it should be pointed
õ
out, that there exist formally two other, equivalent expressions for the MEs, namely (for
, å , )
3 Ô 3 æ ^ î æ î È a· 3 S ñ 3
îÈ æ î Ú Ú ^ 2.133)
A± Ë
È · Ý S È¸ È È
~ S ~ ± S ¸
which follow from Eq. (2.105) and from the Kohn-Sham equation. Evaluated on a finite
basis-set all three expressions suffer from the same basic deficiency. Chen (1976) investi-
i
gated the accuracy of the MEs using the KKR method. Enlarging the employed basis-set
from
þ utwv Y
= 2 (i.e., x yTz y
basis functions) to
þ
= 4, resulted in changes in the MEs
ctwv Y
of already a factor of two. The MEs are thus not yet converged and all prospects of
î []ò [
precise calculations of optical properties are thereby diminished. KKR-based calculations
of of Cu nevertheless yielded a reasonably good description of the experimental data
(Williams et al. 1972).
Another way of testing the MEs is by evaluating both sides of the equality (Blount
1962)
â± î û þ Ú
^û a 3 O 3
^ 2.134)
Ë
þ S S þ ¸
·
~
â
where the left-hand side is evaluated from the wave functions, and the right-hand side from
the energies. This has been done for the LMTO method, both by using the atomic sphere
approximation (ASA), in which the interstitial is neglected, and by including extensions
to cope approximately with the interstitial part by Uspenskii et al. (1983). In going
from
þ twv Y
= 2 to 3 the MEs fluctuate almost randomly (opposite signs occurring) in the
<
LMTO-ASA, already for energies below . Better results are obtained with LMTO with
combined corrections, leading to changes of a factor 2 – 3 in the MEs when enlarging the
basis from
þ utwv Y
= 2 to 3. The MEs calculated for
þ
= 3 with the LMTO combined-
corrections approach are clearly the more accurate ones, yet, even these are, for energies
ctwv Y
<
higher above , nearly 50% smaller than the more accurate values obtained from the
energies with Eq. (2.134) (see Uspenskii et al. 1983). Although it thus appears that the
î []ò [
LMTO-ASA method is not very suitable for the computation of excitation spectra, Alouani
et al. (1986) employed it and obtained a spectrum for Al and for Fe that modestly
compared to available experimental data. Once again, it has to be kept in mind that for the
MO spectrum a much better precision is required.
A computational scheme that yields sufficiently accurate MEs for MO calculations

has been developed by Oppeneer et al. (1992a), on the basis of the ASW method. The
MEs are evaluated by taking the interstitial into account, which makes them hermitian.
In addition, unaugmented spherical waves are used in the evaluation to achieve better
converged MEs. These unaugmented spherical waves are site-centered, spherical Hankel
and Bessel functions, that are the solutions of the free-electron Schr ödinger equation. In
the ASW scheme, the spherical waves up to angular momentum
þ htwv Y
are augmented in the
atomic spheres, i.e., differentiably continuously matched to exact solutions of the radial
Kohn-Sham equation. The spherical waves of all higher -values are not augmented, but
þ
maintained in the calculation as they are (Williams et al. 1979). These free-electron wave
functions provide the supplementary set of functions that can be invoked to converge better
the MEs. As a result of this approach, it is found, first, that the MEs change very little when
þtwv Yis enlarged from 2 to 3. This exemplifies convergence with respect to
þ <
. Second,
the errors in the MEs are rather small, but increase for energies higher above . Up to
twv Y
<
energies of about 7 eV above , the relative error in the MEs, evaluated with Eq. (2.134),
<
is about 7%. For bands that have energies in between 7 eV and 14 eV< above the errors
are up to 25%. Bands that are again placed more than 14 eV above can exhibit errors
of 50 % (Maurer 1991, Oppeneer et al. 1992a); here the errors are similar to those of
Uspenskii et al. (1983). The MEs do thus become inaccurate for increasing band energies,
but calculations of optical properties for photon energies up to about 15 eV are feasible.
Another method to test the MEs is to analyze the fulfillment of the ß -sum rule, together
with its off-diagonal counter part, Eq. (2.96). This has been done for several TM elements.
For the ASW scheme with
þ utwv Y
= 3 a deviation of 10% of the diagonal ß -sum rule was
calculated for bcc Fe; the off-diagonal sum rule, { ô _ÌZ []Kg ^ò_Va
0, yielded approximately
0.01, a small number, which is to be compared to the value of 8 valence electrons given by
the diagonal sum rule. These numbers appear more favorable than the direct comparison
of MEs through Eq. (2.134) would suggest. Larger deviations of the diagonal ß -sum rule
of about 30% were achieved within the LMTO scheme by Maksimov et al. (1988).
Again a different approach was chosen by Bross and Fehrenbach (1990), who checked
éæ ± ò · , S ò S , ¸T
the fulfillment of the identity
Â
È
Õh È ^ 2.135)
È ~

On the basis of a specially developed modified, spline APW method, Bross and Fehrenbach
(1990) achieved so far the most accurate MEs: the kinetic energy identity is fulfilled to
<
better than 1% even for energies of several Ry above . Tests based on Eq. (2.134)
demonstrated also only deviations of less than 2% for bands with energies up to 5 Ry
<
above (Bross et al. 1990). This highly-precise method has, unfortunately, not been
applied to the computation of MO spectra.
The convergence problems of the MEs are to be mastered irrespective of whether a
non-relativistic or a relativistic formulation is adopted. For MO spectra the SO interaction
is the decisive quantity that must be taken into account. This can either be done in a
fully-relativistic formulation, or by treating the SO interaction in the second-variational
approach (see, e.g., Koelling and Harmon 1977). In the fully-relativistic description, the
MEs are given by
þ S ßNM
^û a 3 3 ^
{ È

S ¸
È 2.136) Ë · {
þ T
Z []òg Z []K[
This expression has been applied in calculations, but Guo and Ebert (1995) found that
no stable could be attained thereby, even though the diagonal spectrum was
sufficiently converged. A much better off-diagonal spectrum could be calculated by Guo
ò
and Ebert using the equivalent expression (see Ebert 1990)
Á ß ò Áá îß á î
® *
3 O M 3
á ß M æ ßP ^
Ë
~
È Ë

È
{
S
Ý
Ý í = I / { È¯
S
2.137) W
Z []K[
The diagonal spectrum , however, proved to be indifferent when calculated with either
Eq. (2.136) or Eq. (2.137), which once more illustrates the crucial sensitivity of the MO
spectrum to the MEs.
In the equivalent approach, where the SO interaction is added to the scalar-relativistic
Kohn-Sham equation in the second-variational manner, the replacement of the four-
3
component wave function by the large and small spinor components can again be carried
through. Making the substitution { ^^ `
3 P
, the following expression for

X , and eliminating the small spinor component
is obtained (see Kraft et al. 1995, Kraft 1997)
0À ß ò 8^ d K á dHò í I a O á O ^8d K á dHò í I a á í Ü Ô d a á

K
Ü ^ O dHòpa
ò í hÕ O á í ÕhO d ò bFÚ
A= I
æ í Yd ò ú á ú d ò b æ = ú Yd
Ë
^ 2.138)
with , d K dHò
as defined by Eq. (2.129). The notation with the canonical momentum and a
O
function only. In the above form, operates on

scalar function enclosed in round brackets means that the operator is supposed to act on this
only and is exact to all orders of C .
Eq. (2.138) is an appropriate expression for the evaluation of MEs in the scalar-relativistic
3 Àß
approach with SO interaction added. To understand better the meaning of individual terms
in expression (2.138), it is elucidating to consider an expansion to lowest orders of ,
which reads (see Kraft et al. 1995)
À ßò
ò
_
O À
á Â Ë0ß ò æ Ë O á ^ O Y Ý-á`ÝXW í í I9bòa á í Ü ^ O Ý a á ²=ÄI Ü ^ O ÝXW í a
æ í YÝ W í I á IÝ í W b æ = I Y Ý W í í hÕ O á í ÕhO Ý W í b ^ 2.139)
If, for clarity sake, one sets Ý í = 0 then W
ò
á Â Ë À ß òu æ Ë O á ^ O Ý a á ± í Ü Ý Ô Ý a / Ú
*
_
O ^ 2.140)
is obtained. From this expansion one can recognize that the second term on the right-hand
side of Eq. (2.139) is related to the mass-velocity term, and that the third term relates to the
Darwin correction. The fourth term contains the SO interaction, which contribution to the
ME was derived in a semi-relativistic formulation already by Wang and Callaway (1974),
who obtained *
_
O
á Â Ë0± ß ò í ÜÔ Ý / ^ 2.141)
À ßò
The other terms in Eq. (2.139) have no special names, but they are all of the same order of
C . There is therefore no justification to include only the SO interaction and neglect the
other contributions. The terms in Eqs. (2.138), (2.139) that contain the Pauli matrix can
í
cause spin-flip optical transitions.
Calculations of the MO spectra of uranium compounds where carried out by Kraft et
al. (1995) and Kraft (1997) using Eq. (2.138), but with the additional approximation that
the dependence of the MEs on the energy is neglected. This is a valid approximation,
<
since only valence states near play a role, whereas the relativistic corrections become
O
noticeable only close to the nucleus. Numerical tests showed that all relativistic correction
terms to do not even add up to corrections of 2% in the spectra (see Kraft et al. 1995,
the MO spectra can be evaluated already from the approximation _

and also Wang and Callaway 1974, Osterloh 1993). This implies that even for actinides
O
, but the initial
and final state must of course be the proper relativistic wave functions.
M78I17Nª V| »D|¬$<8RTP C:UVEJP½=h<¸E\C:GIUB<8UtvUyÓ"¾ P=hC:RT<8>/RTGIU2vIP=\½IPSU$RTP

Once the MEs are computed, the second task is to perform the Brillouin zone (BZ) integrals.
It is sufficient to consider only the integral for the diagonal conductivity, in which all the
}~
ò
salient numerical details are present
óØ T Hô û _ æ S Þ _$^^ ûû aa S á ²

^ 2.142)
y
Here _$^ û a _ ~ È ^ û a , and similar for ^ û a ~[ È ^ û a , but for convenience the super-
for æ Ï , where the denominator vanishes at _ _$^ û a . There exist several standard
fluous indices are dropped. The integration Þ has toÞ performed stably for every , including
techniques to carry the -integration out. Most techniques are based on the subdivision of
-space in tetrahedra, on which a suitable approximation of the integrand is subsequently
made.
æ Ï
Two conceptually different techniques have so far been applied to the BZ integration.
}
The first, often-applied technique starts from considering the limit
ò ò
, which yields
@Ø ü Tó ôHû _ S Þ æ ^ _$û a ^ ûS a á ó ôTS Ô Þ _$^ û ^ ûa S a Ú ^ 2.143)

þ
S S
results from J^`_ æ _$^ û a:a . The latter part can be computed using a standard analytical BZ
where the first term is the dispersive part and the second term is the absorptive part, which
integration scheme (Jepsen and Andersen 1971, Lehmann and Taut 1972). The dispersive
part in Eq. (2.143) can then be computed by a KK transformation of the absorptive part. The
lifetime effects connected to the finite are subsequently incorporated by convoluting both
the absorptive and dispersive spectrum with a Lorentzian. The second technique performs
the BZ integration for the non-zero directly in the complex plane, using a variant of the
analytical tetrahedron method (Coleridge et al. 1982, Oppeneer and Lodder 1987). It has
æ Ï
been shown that this technique has good accuracy and in addition stably yields the proper
limiting integration values for (Maurer 1991, Oppeneer et al. 1992a).
Both techniques have certain advantages and disadvantages. The analytical complex
formulas used in the direct integration technique are longer and require therefore more
_
computing time. In the approach based on the KK transformation, on the other hand, the
expressions are simpler, but a rather dense and extended mesh of -points is required to
_ _
perform the KK transformation. In the direct approach, the integration can be performed
on a small number of -points, even for one -point if desired. Which one of the two
techniques one prefers is a matter of taste. One further drawback of the KK transformation
K õ
technique may be mentioned, which becomes apparent only when a non-constant inverse

lifetime ¿ is adopted. The lifetime ¿ may depend on the specific states , , , , but also
F_
on the excitation energy ± (e.g., Santonini and Himpsel 1991). The KK transformation
and the broadening convolution do not commute for a non-constant (Delin 1998). The
resulting spectrum depends therefore on the order of the lifetime convolution and KK
æ Ï
integrations. One thus has to be cautious about the order of the spectral integrations.
Although the KK equations are strictly only valid in the limit , i.e., for spectra
without finite lifetime effects, the approach suggested by Delin (1998) is to perform first
the lifetime broadening convolution on the calculated absorptive part, and then do the KK
í `^ _ á a
transformation. This is not obvious, as can be recognized from the equivalent of Eq. (2.88)
Ï
for the optical conductivity A that already contains the finite lifetime broadening
( )
í ^`_ áa ÁØÀ ó ô _Iõ _ õ æ À _ í ^ò_Iõ á a

´
A

Q ² ^ 2.144)
´
í
If the argument of on the right-hand side had been _ õ á A È , a KK transformation of the
lifetime convoluted quantities would have been justified, as it is for a constant . The direct
integration scheme is, in contrast, straightforwardly applicable to non-constant lifetime
parameters.
The last numerical convergence that is to be achieved before the machinery of first-
principles calculations of MO spectra can be operated, is the convergence of the BZ
integrals with respect to the number of -points used. This is in itself a rather trivial
task, except if one wants to find the optimal algorithm which has the best performance
zh^ %æ î ý a
on the smallest set of -points. Sufficient care is required for those energy bands that
< <
cross , because the occurring cut-offs of the initial and final energy bands
zh^ î æ a
,
<
and , are properly to be taken into account. These band crossings sometimes
Z []Kg ^`_Va
play a role for the accuracy of the far-infrared range of the spectrum. As an example of
the -space convergence a calculated spectrum of FePd is shown for two sets of
-points in Fig. 2.1. The numerical BZ integration on a subdivision of tetrahedra can thus
0.16
0.12 FePd
σxy (10 s )
-1
0.08
15
0.04
(1)
0.00
-0.04
-0.08
0 1 2 3 4 5 6
Photon energy (eV)
~øJ&h`H
UL8g T
Example of the convergence of the BZ integrals with respect to the
number of tetrahedra used to subdivide the irreducible part of the BZ. Shown is the

spectrum of FePd, calculated for 5200 tetrahedra (dotted curve) and for
r
12300 tetrahedra (full curve), for a lifetime of = 0.03 Ry.
be easily controlled.
The necessary numerical tasks for the ab initio calculation of optical and MO spectra
are thereby summarized, and have been shown to be feasible. In the following Chap-
ters, theoretical polar MOKE spectra of various groups of ferromagnetic compounds are
presented, and where possible, compared to the available experimental data.
47
/ (
ýZÿ £3ý£ ÿ Ñ£ ÿ ÿ Ñ£+ ¢

ÿ ¢
¥ .©28¥¢ n0y8&.

Artificially synthesized transition-metal (TM) multilayers have become a leading edge re-
search activity in the last decade (see, e.g., Bland and Heinrich (1994) for a recent survey).
A series of discoveries lead to the current world-wide research interest in TM multilayer
materials. One of these is the antiferromagnetic exchange coupling of two neighboring
magnetic layers across various kinds of spacer layers, as, e.g., Fe with Cr spacer layer,
Co with noble metal spacer Cu, or Co with a non-magnetic Ru spacer (Gr ünberg et al.
1986, Cebollada et al. 1989, Parkin et al. 1990). Concurrent to the occurrence of an-
tiferromagnetic exchange coupling a large change in the resistivity of such materials in
an applied magnetic field was observed (Baibich et al. 1988, Binash et al. 1989). This
phenomenon has become known as giant magneto-resistance (GMR). GMR materials are
of great technological importance due to their magnetic field tunable transport properties.
Another observation related to the interlayer exchange coupling was the enhancement of
the Kerr rotation concurrent to the appearance of an antiferromagnetic coupling (Bennett
et al. 1990, Carl and Weller 1995). Another, again related, novel discovery was that of
quantum confinement effects in thin Fe/Au layers, which manifest themselves in oscilla-
tions of the MOKE versus layer thickness (Suzuki et al. 1992, 1993).
Other kinds of ferromagnetic TM multilayers, as e.g., Co/Ni, Co/Pd, Co/Pt, and Fe/Pd,
Fe/Pt, and also Fe/Au, have recently attracted considerable interest, because it was dis-
covered that these multilayers may exhibit pronounced perpendicular magnetocrystalline
anisotropies (see, e.g., Garcia et al. 1985, den Broeder et al. 1987, 1988, Garcia 1988,
Zeper et al. 1989, Katayama et al. 1991, Takanashi et al. 1995). The magnetocrystalline
anisotropy energy (MAE) in most of these multilayer systems causes the magnetization to
ô
orient preferably perpendicular to the layers. Through selective variation of the thicknesses
of the non-magnetic spacer layer and of the 3 layer, the magnetic and transport properties
48 CHAPTER 3. ORDERED FEPD, COPD, FEPT, AND COPT
can be “tuned”, so that a materials engineering of desirable materials characteristics be-

comes feasible. This aspect, and the fact that the phenomenon of perpendicular magnetic
anisotropy is of particular importance in various technological applications, especially in
recording technology, caused the tremendous present interest in the magnetic and MO
properties of such multilayer materials.
In this Chapter we consider the ordered compounds CoPd, CoPt, FePd, and FePt. These
ï
are the ordered “parent” compounds for the corresponding multilayer systems. CoPt, FePd,
and FePt crystallize in the AuCu (i.e., L1 ) structure, which is a tetragonal structure con-
sisting of alternating monolayers of Fe (Co) and Pd (Pt, respectively) in direction. CoPd, ß
however, is not known to exist in the AuCu structure, but it has been synthesized through
epitaxial layer disposition (den Broeder et al. 1989). In particular, we study the rela-
tionship between the MAE on the one hand, and the magnetocrystalline anisotropy in the
orbital moment, and the magnetocrystalline anisotropy in the polar MOKE spectra on the
other hand. Since these quantities are all three caused by the simultaneous occurrence of
spin and orbital polarization of individual electron states, it is not surprising that there is a
close relationship between them.
¥lk ¦ -,/.0#1q)+/)+#$./¡.0¢².¸28/5©¨
The magnetocrystalline anisotropy energy is one of the most essential quantities in mag-
netism. Since already many years it is known that SO coupling in the presence of sponta-
neous spin polarization is responsible for the MAE (Van Vleck 1937, Brooks 1940, Fletcher
K [ [ K [ [
1954), so that both the MAE and MO phenomena are rooted in the same microscopic origin.
Experimental investigations on Co Pd and Co Pt alloys demonstrated that large
ô
Kerr rotations and MAEs coincide (Weller et al. 1993b). A model description of the MAE
in 3 metals has been proposed on the basis of a perturbation theory approach (Brooks
po [ Õ \
1940, Fletcher 1954, Bruno 1989). In this approach, the SO term in the Hamiltonian, which

ß ò K ={ <
is given by , with the SO coupling constant, i.e., ·¶Z Ý q Z ¸,
â â
ò
is assumed to be the small perturbation. The MAE is then expressed by the first terms of
a power series in . This formulation does, however, not explain why multilayer systems
for certain layer thicknesses exhibit a perpendicularly oriented magnetization, while for
other thicknesses an in-plane magnetization occurs (see, e.g., den Broeder et al. 1987).
With regard to this rather fundamental problem, it is noteworthy that using perturbation
theory, a relationship between the uniaxial MAE and the anisotropy in the orbital moment
define the uniaxial MAE as

< <
< po æ ¡po¢
was derived (Bruno 1989, Bruno 1993). For a material having a layer structure, we may
p] , where stands for the magnetization £
ô
perpendicular to the layers, and for an in-plane magnetization. The relationship between
SS
ô
the MAE and the anisotropy in the orbital moment for a 3 element with more-than-half
filled 3 shell reads (Bruno 1989, Bruno 1993) ~
< op æ FÂ ð ¤ Y ÿ æ ÿ
¢ bFÚ ^ 3.1)
CHAPTER 3. ORDERED FEPD, COPD, FEPT, AND COPT 49
¤T¥o¦N§ ©¨\`H ª «
Calculated spin (M ) and orbital (M ) moments, and anisotropy energies ( ) for ¬ Æ
U®
Æ ³Æ ¶µ:µ L8_· Æ L@L µ
the ordered XPd, XPt (X = Fe, Co) alloys. All moments are given in Bohr magneton ( ), for
two magnetization directions, perpendicular to the layers ( (001)) and in plane ( ¯±°
(110)). The ¯²°
anisotropy energy is given as ¬ ´ , in units of meV per 3 atom. The experimental ¸
anisotropy energies are quoted from Daalderop et al. (1991).
M ª¹]ºf» ¼½ M «¹]ºf» ¼½ M ª¾-¿u» ¾ 3 M «¾-¿u» ¾ 3

(001) (110) (001) (110) (001) (110) (001) (110) ¬"ÆgÀ;Á:Â À:ÃÄ¬ Æ ½ f Å Ã
FePd 2.917 2.916 0.093 0.077 0.345 0.344 0.030 0.032 -0.55 -0.5
FePt 2.780 2.780 0.083 0.077 0.348 0.347 0.054 0.067 -2.75 -1.2
CoPd 1.814 1.813 0.131 0.116 0.376 0.376 0.036 0.034 0.10 —
CoPt 1.686 1.686 0.097 0.067 0.370 0.370 0.067 0.084 -0.97 -1.0
¢
with M , M , the orbital moment for magnetization perpendicular to the plane, or in the

plane, respectively. Expression (3.1) states that for perpendicular magnetic anisotropy the
perpendicular orbital moment M will be larger than the in-plane one M , whereas for
¢
a preferred in-plane magnetization the reversed situation will occur. Relationship (3.1)
has been investigated experimentally for various Co-based magnetic multilayers using the
XMCD sum rules to determine the Co orbital moment in the multilayers (Weller et al.
1994b, Weller et al. 1995). The anisotropy relationship was found to hold, and on account
of this, it has been stated that the anisotropy in the orbital moment is in fact the microscopic
ô
origin of the MAE (Weller et al. 1995). As a restriction of Eq. (3.1), it should be mentioned
that a perturbative approach is most likely appropriate for 3 elements, but its validity will
be limited for heavier elements as, e.g., Pt. In addition, all aspects related to self-consistent
relaxation of the SO perturbed electron states are left out of the consideration. In spite of
these short comings, it seems that at least as a sort of rule of thumb Eq. (3.1) is applicable
(see also Újfalussy et al. 1996).
In Table 3.1 we list the calculated spin and orbital moments of the FePd, FePt, CoPd,
ï
and CoPt compounds, for two directions of the magnetization, the (001) and (110) direction
in the L1 unit cell. The former corresponds to a magnetization perpendicular to the planes,
ß
whereas for the latter the magnetization is in the plane. We have used for FePd, FePt, and
ß

CoPt, the experimental
ß
lattice spacings, which are: = 3.8550 ,
= 0.9634 for FePd,
= 3.8665 ,
ß
= 0.9570 for FePt, and = 3.8000 , = 0.9737 for CoPt (Villars and
Calvert 1991, Cebollada et al. 1994). For CoPd we used lattice spacings ( = 3.7950 ,

= 0.9721) estimated from those of CoPt. Our calculations were performed on large
Æ
numbers of -points in the Brillouin zone, about 1.4 10 up to 2.9 10 irreducible points ù ù
in the whole Brillouin zone. These large numbers of points were required to achieve
Æ
convergence of -space summations. The von Barth-Hedin (1972) parameterization of the
local spin-density exchange-correlation energy was used in the calculations. The results
listed in Table 3.1 can be summarized as follows: A large spin moment is calculated for Fe
and Co, whereas Pd and Pt carry a relatively large induced spin moment. The spin moment
ð
on each individual atom shows nearly no anisotropy: the differences in the spin moments
for the (001) and (110) directions do not exceed 0.001 ¤ . The spin and orbital moments
ô ô ô
on Fe and Co are for the Pd-based compounds significantly larger than for the Pt-based
ô
compounds. This is a band structure effect, indicating a somewhat different 3 –4 and 3 –
5 bonding. Experimentally, a larger orbital moment has been measured for Co in Co/Pd
multilayers than for Co in Co/Pt multilayers (Weller et al. 1994b). The orbital moments
ô
on Pt, on the other hand, are larger than those on Pd. This appears to be related to the SO
coupling of Pt being larger than that of Pd. For all four compounds the 3 orbital moment is
calculated to be significantly larger for perpendicular orientation. For FePd, FePt, and CoPt
this finding clearly corroborates with the experimentally observed perpendicular anisotropy
in these compounds. Our results support thus for these compounds the perturbation theory
expression (3.1).
The magnetic moments and the anisotropy energies of the XPd, XPt (X = Fe, Co)
compounds were recently investigated by Daalderop et al. (1991) and by Solovyev et
al. (1995), and those of FePt and CoPt by Sakuma (1994). In these calculations lattice
constants that differ from the above cited experimental constants were used. Daalderop et
al. chose identical constants for FePt and for FePd, and also for CoPd and CoPt. In spite
of these differences, trends in the calculated moments very similar to the trends displayed
ð
in Table 3.1 were found. For FePd, where nearly the same lattice parameters were adopted,
the calculated moments of Table 3.1 agree within 0.01 ¤ with those of Daalderop et al.
(1991).
µ µ õ
We also calculated the MAEs from the self-consistently calculated Kohn-Sham total
< r7 r ^µ a æ < r7 r ^µ õa
energies for the two magnetization directions and (cf. Trygg et al. 1995), i.e.,
<
^ 3.2)
This expression is the exact definition of the MAE. Another approach which is commonly
used to compute the MAEs is the so-called force theorem method (see Mackintosh and
î þ
Andersen 1980, Daalderop et al. 1990a). In this approach is the MAE approximated by
Æ-space sums over all occupied Kohn-Sham single particle energies
for two directions
of the magnetic moment, viz.
<
_ Â
7 9:9
î þ ^µ a æ Â
È 7 9:9
î þ
µ
È ^ øõ a ^ 3.3)
þ þ

[ Õ\
po ´
The single-particle energies are computed non-self-consistently, by adding the SO coupling
term to the scalar-relativistic Kohn-Sham equation. Daalderop et al. (1991)
and Solovyev et al. (1995) applied the force theorem method to calculate the MAEs of
ordered XPd and XPt (X = Fe, Co).
Our MAEs computed from the total energies and the experimental values are given
also in Table 3.1. For FePd, FePt, and CoPt we obtain MAE values which are in good
agreement with available experiment data (McCurrie and Gaunt 1966, Sanchez et al. 1989,
ô
Daalderop et al. 1991). The error in our calculated MAEs is, from tiny fluctuations in the
converged self-consistent total energies, estimated to be less than 0.2 meV per 3 atom.
The calculated MAE of FePt is two times larger than the experimental MAE, but this might
be due to the fact that the latter is not a low-temperature value (cf. Daalderop et al. 1991).
ô
Weller (1996) has remeasured the uniaxial magnetic anisotropy of FePt and obtained a
uniaxial <
3.4 meV per 3 atom. This is a high value, which together with the large
^ ÈÇ afÉ [
magnetic moment density categorizes FePt as a hard-magnetic material. Recent studies
obtained for sputtered FePt films an energy product ¹ ç of 35.8 MG Oe, which is
Ê
about 80% more than that of hard-magnetic SmCo (Watanabe et al. 1998), and, for some
films an ultrahigh coercive force of about 48 kOe (Ide et al. 1998).
In the investigations of Daalderop et al. (1991) and Solovyev et al. (1995) different
computational approaches and lattice constants were used, but the same easy axes were
obtained for FePd, FePt, and CoPt (Daalderop et al. 1991, Solovyev et al. 1995). CoPd is
the only compound were the situation is not unambiguously clear. CoPd is not known to
crystallize in the AuCu structure, but attempts have been undertaken to synthesize CoPd
in this structure through epitaxial layer deposition (den Broeder et al. 1989). Our MAE
ô
calculation predicts a very small, but in-plane magnetization for CoPd. The estimated error
ô
is, with 0.15 meV per 3 atom, as large as the MAE value itself. Daalderop et al. (1991)
obtained a perpendicular anisotropy energy of -0.30 meV per 3 atom for CoPd. However,
using the same computational approach, Solovyev et al. found an in-plane magnetization
for CoPd. In a later study, Daalderop et al. (1994), too, computed a small, but in-plane
magnetization for CoPd. Thus, for CoPd most of the available calculations predict an in-
ô
plane magnetization and do therefore not support the perturbation theory expression (3.1).
We furthermore remark, that while the orbital moment of the 3 atom obeys Eq. (3.1),
the orbital moment on Pd and Pt shows a different behavior. For FePd, FePt, and CoPt,
ô
the orbital moment on Pd or Pt is, for magnetization in the plane, larger than that for the
perpendicular-to-the-plane magnetization. This behavior is thus opposite to that of the 3
K ò
element. Interestingly, here again CoPd forms the exception.
ò
The related (Co) /(Pd) multilayers were studied theoretically by Daalderop et al.
magnetic anisotropy, with

< K
(1990b) and Kyuno et al. (1992). Both calculations obtain for (Co) /(Pd) a perpendicular
_ -1 meV per unit cell. This multilayer with two Pd
spacer layers thus behaves already different from CoPd, which illustrates the extreme
K ò
dependence of the MAE on structural parameters. In a detailed investigation of a series
of (Co) /(X) (X = Pd, Pt, Cu, Ag, and Au) multilayers Kyuno et al. (1992) obtained a
larger perpendicular orbital moment on Co for every multilayer which was calculated to
have perpendicular anisotropy, in accordance with Eq. (3.1).
¥ ¦ m 5§#uy28Ý(§70#$ËéwÌ70#$ËË¢ÍÌÍ ËË¢ÍÌ .0¢ ÍúËé

ª17µª17:6 Pâ E
We consider FePt first, because historically FePt was investigated intensively before FePd
was. Several noteworthy discoveries, as, e.g., a huge structural dependence of the polar
MOKE spectra, a pronounced Kerr anisotropy, and an a large MAE were reported for FePt
(Cebollada et al. 1994, Weller 1996). It can be anticipated that the other three compounds
exhibit similar properties. The very first measurements of the MOKE spectra of FePt were
due to Buschow et al. (1983a), who reported that the measured Kerr angle of an arc-melted
FePt
Complex polar Kerr effect (deg) 0.3 θK εK
0.0
-0.3
-0.6 Buschow
Cebollada 1
Cebollada 2
theory
-0.9
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
~øJ&Î¨h`H
Ï Experimental and calculated polar Kerr rotation, , and Kerr ellipticity,
, of FePt. The experimental data are after Buschow et al. (1983a), and are after
Cebollada et al. (1994). The latter MOKE spectra were measured on a highly-ordered,
L
MBE-grown sample (denoted Cebollada 1), and on a disordered sample (denoted Ce-
bollada 2). Note the huge dependence of the MOKE spectra on the degree of ordering.
The theoretical spectra are calculated for a a lifetime broadening of = 0.03 Ry, and XÐ_Ñ
contain an intraband Drude-type conductivity.
j
FePt sample reached up to about -0.5 at 4.5 eV. Other MOKE measurements on sputtered
FePt-alloy samples roughly yielded the same magnitude at 4.5 eV (Katayama et al. 1992,
j
Sato et al. 1992a, 1993a). First-principles LSDA calculations, however, predicted a peak
Kerr angle of about -1 at 4.5 eV (Osterloh et al. 1994). The theoretical MOKE spectra did
thus at first not seem to bear any relevance to the experimental spectra. One piece of this
puzzle was supplied by Lairson and Clemens (1993), and also by Sugimoto et al. (1993),
ï
who both observed changes of the polar MOKE spectra coinciding with the formation of
an ordered L1 AuCu crystal phase. The solution to the disagreement between ab initio
calculated and experimental MOKE spectra of FePt was finally given by Cebollada et al.
(1994), who explicitly showed that the polar MOKE spectra of a MBE-grown, highly-
ordered FePt sample compared excellently to the predicted MOKE spectra. Cebollada et
al. furthermore found the MOKE spectra of a disordered FePt sample to be strikingly
different. In Fig. 3.1 we show these experimental MOKE spectra of two FePt samples as
obtained by Cebollada et al. (1994) as well as the Kerr rotation measured by Buschow
et al. (1983a) and the ab initio calculated spectra (cf. Osterloh et al. 1994, Oppeneer
et al. 1996c). It can be recognized from Fig. 3.1 that the MOKE spectra of FePt are
very sensitive to the crystal structure: The Kerr spectra of the highly ordered FePt sample
differ much from those of a disordered sample. The Kerr angles measured by Cebollada
et al. on the disordered sample nicely agree with those of Buschow et al. (1983a). The
ï
structural dependence appears to be particularly related to the layer structure of FePt in the
L1 structure: The disordered FePt material does not have pure monolayers of Fe and Pt,
and, in addition, it is not found to exhibit perpendicular magnetic anisotropy (Cebollada
et al. 1994). At photon energies less than 1.5 eV there is a small difference visible
Z KÊ K
between the experimental and the calculated Kerr angle. An intraband Drude conductivity

(with ñ = 12 10 s , ± ñ = 0.013 Ry) has been added, but the difference cannot fully
be explained thereby. Also, the theoretical Kerr ellipticity at 5 – 6 eV differs from the
experimental curve, which shows a slight upturn. Earlier measurements by Sugimoto et
}{
al. (1993) as well as recent newer measurements (Weller 1997) confirmed the calculated
behavior of in this energy range. The outstanding experimental finding of Cebollada et
al. (1994) for the Kerr angle of FePt was verified in an independent study by Mitani et al.
(1996). The experimental MOKE spectra of high-quality, single-crystalline FePt compare
thus very well to the first-principles MOKE spectra. We mention in this respect that
the MOKE spectra of FePt have been calculated by several other groups, using different
computational schemes (Halilov and Feder 1993, Ebert et al. 1995, Lim 1997). The
thus-obtained MOKE spectra show some spectral differences, but the overall shape of the
MOKE spectra has been reproduced therein.
The discovery of the magnetocrystalline anisotropy in the MOKE spectra of hcp Co
Ü
(Weller et al. 1994a) prompted that TM multilayer systems could, on account of the
larger MAE, exhibit much larger Kerr anisotropies. The MAE in FePt is with
20 MJ/m more than 40 times larger than that in hcp Co, suggesting thus a much larger
Ò _
Kerr anisotropy (Weller 1996). The occurrence of a significant Kerr anisotropy in FePt
was verified experimentally by Weller (1996). In Fig. 3.2 the measured MOKE anisotropy
^ z { ïï K æ z { 9K K ï a òK ^ z { ïï K á z { 9K K ï a
and the calculated anisotropy are shown. The measured Kerr anisotropy is largest at 2 eV:
b = 0.39, which is more than twice as large as the Kerr
anisotropy observed in hcp Co. Ab initio calculations predicted a notable Kerr anisotropy at
2 eV, in good agreement with experiment (Ebert et al. 1995, Oppeneer and Antonov 1996).
With regard to the experimental MOKE spectra in Fig. 3.2 we note that the fabrication of
single-crystalline FePt (110) was more difficult than that of FePt (001). The FePt (110)
sample was therefore only about 80% chemically ordered (Weller 1996). In view of the
dependence of the MOKE spectra on the degree of ordering an increased Kerr anisotropy
is therefore conceivable for perfect order.
ª17µª17NM Pât
K [ [
The MOKE spectra of FePd were first measured by Buschow et al. (1983a). Reim et
al. (1991) studied various Fe Pd alloys. A detailed investigation of the perpendicu-
lar magnetic anisotropy and the polar MOKE spectra of FePd were recently reported by
Armelles et al. (1997a, 1997b). The theoretical prediction of the MOKE spectra of FePd
is due to Osterloh et al. (1994). In Fig. 3.3 we show the measured polar MOKE spectra
of Armelles et al. (1997a) as well as the calculated spectra. Just as it was previously
discovered for FePt, the measured FePd MOKE spectra depend appreciably on the degree
FePt
0.0
-0.3
-0.6 exp. (001)

exp. (110)
-0.9
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
εK
Complex polar Kerr effect (deg)
0.3
θK
0.0
-0.3
-0.6
-0.9
(001)
-1.2 (110)
-1.5
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
~øJ&Q¨hÓh
Orientation dependence of the polar MOKE in FePt. Top panel: ex-
perimental Kerr anisotropy in (001) and (110) FePt after Weller (1996). Lower panel:
calculated, interband-only, Kerr anisotropy after Oppeneer and Antonov (1996). The
absence of an intraband contribution is responsible for the peak in the Kerr rotation
below 1 eV.
of crystalline ordering of the sample (Cebollada et al. 1994). The measured Kerr spectra
were obtained for two samples which were approximately ordered up to 80% (Armelles et
FePd
θK εK
0.2
0.0
-0.2
-0.4
-0.6
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
~øJ&¨hÓ¨h Ï
XÐ Ñ L
Theoretical and experimental Kerr angle and Kerr ellipticity
spectra of FePd. The theoretical spectra were calculated for the ideally ordered FePd
compound, using a lifetime parameter of = 0.03 Ry, and include an intraband
¬ Ô
Drude contribution. The experimental data ( , ) are for two incompletely ordered
FePd samples (Armelles et al. 1997a).
al. 1997a). The shapes of the experimental Kerr rotation and ellipticity spectra in Fig. 3.3
correspond well to those reported by Buschow et al. (1983a), who obtained a somewhat
smaller overall magnitude. Again, this can be attributed to the degree of crystalline order:
A better ordering is anticipated to lead to a larger Kerr angle and would thus compare better
to the ab initio calculation corresponding to perfectly ordered FePd. It can therefore be
concluded that the experimental polar Kerr spectra of FePd are satisfactorily explained by
the ab initio calculated spectra. The calculated Kerr rotation peak at about 2 eV, however,
is much larger than the experimental peak. The reason of this discrepancy is not yet known.
The MOKE spectra of FePt and of FePd are rather similar in shape, but the magnitude of
the FePt Kerr rotation is larger than that of the FePd. Osterloh et al. (1994) pointed out
that the larger SO coupling of Pt is responsible for the larger Kerr rotation in FePt.
Z [][ ^À á ÂØ @Z [][ _Va9K ò Z []g

The anisotropy in the MOKE spectra arises from a crystalline anisotropy in the de-
Z []òg
nominator & , and a magnetocrystalline anisotropy in (Oppeneer
et al. 1995b, Oppeneer 1998). The magnetocrystalline anisotropy in the spectrum is
Z []òg
connected through the MO sum rule (2.104) to the anisotropy in part of the orbital moment.
In Fig. 3.4 we show the spectrum of FePd, calculated for three inequivalent directions
0.16
FePd
0.12
σxy (10 s )
-1
15
0.08
(2)
0.04 (001)
(111)
(010)
0.00
0 1 2 3 4 5 6
Photon energy (eV)
~øJ&Õ¨h¶Ö
Imaginary part of the off-diagonal conductivity, Ã:
, calculated for the
three non-equivalent directions (001), (111), and (010) of the magnetization in FePd.
of the magnetization, the (001), (111) [i.e., (11 ( )], and (010) direction. The subscript ‘xy’
v
is denoted with respect to the magnetization orientation, which defines the local F -axis,
Z []òg
and can either be the crystallographic (001), (111), or (010) axis. There is a gradual
Z []òg
change in the spectrum when the magnetization is changed from the perpendicular
(001) to the in-plane (010) direction. The magnitude of the first peak in at about
2 eV is gradually reduced in amplitude, but the second maximum at about 4.5 eV becomes
enhanced when the magnetization turns from perpendicular to in-plane. These changes
Z []òg
can be related to the anisotropy in the orbital moments of FePd given in Table 3.1. The
first peak in is due to Fe (see Oppeneer et al. 1992a), while the second peak is mostly
due to Pd (cf. Weller et al. 1993a). Under the assumptions that the anisotropy in part of
×
the orbital moment M is representative for the whole orbital moment, and that the energy
range in Fig. 3.4 is sufficiently large to cover the dominant spectral range, the following
can be noted: The Fe-related spectral peak is smaller for in-plane magnetization, which
corresponds to a reduced orbital moment of Fe, in accordance with the MO sum rule and
the orbital moments given in Table 3.1. The second peak, being mostly due to Pd is larger,
again in accordance with the enhanced in-plane orbital moment on Pd. We emphasize,
however, that in the case of FePd it is possible to identify approximately the elements
Z []òg
from which certain spectral features originate, but this is not always the case. We remark,
lastly, that the relationship between the spectrum and the orbital moment has not yet
received much attention. In the case of the ferromagnetic uranium monochalcogenides
Z []òg
US, USe, and UTe, which crystallize in the cubic rocksalt structure, a larger spectrum
was obtained for the (111) magnetization direction than for the (001) orientation (Kraft
et al. 1995). which corresponds to the larger orbital moment calculated for the (111)
direction, and to total-energy calculations, which predict the (111) orientation to be the
easy-magnetization axis (Kraft 1997).
ª17µª17Nª ÏØ Gêât¹<8UtÙØÏGêâ E
CoPd and CoPt alloys and multilayers initially drew attention when it was discovered that
the multilayers exhibit perpendicular magnetic anisotropy (Garcia et al. 1985, Draaisma
and de Jonge 1987, Draaisma et al. 1987, den Broeder et al. 1987, Garcia 1988).
Subsequently, substantial Kerr rotations in Co/Pd, Co/Pt multilayers and CoPd, CoPt alloys
were reported (Zeper et al. 1989, Weller et al. 1992, Sato et al. 1992a, 1993b, Sato 1993).
After these discoveries, it was soon established that Co/Pd and Co/Pt multilayers fulfill
all the prerequisites for polar MO recording materials: a large perpendicular anisotropy,
NÚ
a suitable ¥ , large MOKE signal in the UV spectral range, good reflectivity, 100%
perpendicular remanence, and a high coercive field (Buschow 1989, Zeper et al. 1989,
Weller et al. 1992).
The structural, magnetic, and MO properties of Co/Pd and Co/Pt multilayers have
been the subject of a number of papers, in which the influences of varying the layer
thicknesses, adding further interlayers of other elements, or doping with selected elements
were studied (see, e.g., Suzuki et al. 1994, Hashimoto 1994, Wang et al. 1995, van Drent
et al. 1996). Before such effects are addressable in first-principles investigations, the
ï
MOKE spectra of ordered CoPd, CoPt compounds must be understood. In the following
ï
we therefore consider the MOKE spectra of CoPd and CoPt in the L1 crystal structure.
As was mentioned before, CoPd is not known to exist in the L1 structure, but has been
synthesized through epitaxial layer disposition (den Broeder et al. 1989). The MOKE
j
spectra of disordered CoPd alloy were measured by Buschow et al. (1983a), who obtained
a maximal Kerr rotation of -0.30 at 3.7 eV. Tosaka et al. (1994) studied several Co–Pd
ï
compositions and obtained MOKE spectra similar to those of Buschow et al. (1983a).
Assuming the ideal L1 crystal structure, we calculated the MOKE spectra of CoPd for
two orientations of the magnetic moment, the (001) and (111) orientation. These MOKE
spectra, and the experimental ones are shown in Fig. 3.5. The measured Kerr spectra are,
compared to the calculated ones, rather featureless, yet a rough correspondence between
[ K [
the calculated and measured spectra can be observed.
The magnetic and MO properties of Co Pt alloys were intensively examined over
the last years (see, e.g., Buschow et al. 1983a, Sch ütz et al. 1990, Sato et al. 1992a, Weller
j ï Ê9ï ï Ê:ï
et al. 1992, 1993b, Lairson and Clemens 1993, Weller 1996). A maximal Kerr angle of
ï
-0.55 for Co Pt alloy was reported by Buschow et al. (1983a). Investigations
of an epitaxially grown CoPt compound with L1 structure, ordered to better than 90%,
were carried out by Weller (1996). In Fig. 3.6 we compare these experimental data to ab
j
initio calculated MOKE spectra of CoPt. The experimental Kerr angle is systematically
smaller than the calculated one by 0.15 . An explanation of this difference has not yet been
found. Weller (1996) observed significant differences between the Kerr spectra of a highly
CoPd
0.2
Complex polar Kerr effect (deg) θK εK
0.0
-0.2
-0.4
Buschow
theory (001)
theory (111)
-0.6
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
~øJ&©¨hÓÛh
L
Experimental and theoretical polar MOKE spectra of CoPd. Theoret-
ical, interband-only, Kerr spectra are given for the (001) and the (111) magnetization
orientation in the L1 crystal structure, for µ rÐ_Ñ
= 0.03 Ry. The experimental MOKE
spectra are after Buschow et al. (1983a).
ordered CoPt sample and of a disordered CoPt alloy. In particular, while the magnitude
of the Kerr angle for disordered CoPt has a minimum at 2.3 eV, in the ordered compound
ï
a small peak develops at 2.2 eV. The development of this small peak coincided with the
formation of the ordered L1 phase (Weller 1996). It can be conferred that this small
experimental peak corresponds to the small peak in the theoretical Kerr angle visible at
2.8 eV. This can be recognized from the MOKE spectra predicted for other magnetization
orientations, which are given in Fig. 3.7. There is a remarkable Kerr anisotropy present
at 2.8 eV: For CoPt (001) there is small maximum Kerr angle, whereas for CoPt (111)
and (110) there is on the contrary a minimum. Assuming that in disordered CoPt alloy
small grain particles exist, which orient in a random fashion, the resulting Kerr signal will
be an average over these random orientations. Therefore, the average Kerr rotation for
the material composed of randomly oriented particles can exhibit a minimum at 2.8 eV,
in accordance with the minimum observed at 2.3 eV for disordered CoPt (Weller 1996).
First-principles calculations do thereby confirm that the small experimental peak at 2.2 eV
is a signature of the ordered CoPt (001) compound, as was proposed by Weller (1996).
j
Anisotropic MO properties of CoPt were investigated recently by Iwata et al. (1997),
who prepared MBE grown CoPt either on MgO (001) or MgO (111). At 500 C CoPt
CoPt
θK εK
0.3
0.0
-0.3
-0.6
exp.
-0.9 theory
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
~øJ&Ü¨hÓÝh
The polar Kerr spectra of CoPt. The calculated MOKE spectra are after
Oppeneer and Antonov (1996), and the experimental data are those of Weller (1996).
ï
grows on MgO in the L1 AuCu structure, either as AuCu (001) or as AuCu (111). The
measured Kerr rotations for CoPt (001) and CoPt (111) are also shown in Fig. 3.7. Iwata
et al. obtained a small maximum in the Kerr angle of ordered CoPt (001) at 2.2 eV,
consistent with the result of Weller (1996). This typifying peak was absent for CoPt (111),
where instead a minimum in the Kerr angle was measured. These observations are in
full agreement with the theoretically predicted Kerr anisotropy. The amplitudes of the
measured Kerr rotations, however, are, like that depicted in Fig. 3.6, smaller than the
calculated Kerr rotations.
¥ Íú.0upn02.0
First-principles LSDA investigations of the magnetocrystalline anisotropy in the orbital
moment, the MAE, and the anisotropy in MO spectra emphasize that these anisotropies are
all closely connected. The orbital moments of FePt, FePd, CoPd, and CoPt, exhibit marked
orientational anisotropies, up to 30% for CoPt. In contrast, the corresponding spin moments
are isotropic. The calculated orientational dependencies of the orbital moments on Fe in
FePd and FePt, and that on Co in CoPt, is found to be consistent with the perturbation
theory expression for the MAE by Bruno (1989, 1993). For CoPd, however, a preferred
CoPt
Complex polar Kerr effect (deg) 0.3 Iwata (001) θK εK
Iwata (111)
0.0
-0.3
-0.6
Theory
(001)
-0.9 (111)
(110)
-1.2
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
~øJ&©¨hÞ\
Theoretical orientational dependence of the polar Kerr effect in CoPt.
The calculated, interband-only, MOKE spectra are for the (001), (111), and (110)
µ
magnetization orientation in the L1 phase. For comparison the measured Kerr rotation
spectra of Iwata et al. (1997) for CoPt (001) and (111) are depicted.
in-plane magnetization is calculated, which does not correspond to the perturbation theory
expression. The MAEs of FePd, FePt, and CoPt, as calculated from LSDA total energies,
are in good agreement with available experimental MAE data.
The MOKE spectra of FePd, FePt, and CoPt, are, as a generic feature, found to depend
sensitively on the degree of crystalline order, and also on the crystallographic orientation
of the magnetic moment. A satisfactory agreement between MOKE spectra calculated for
the perfectly ordered material and the experimental MOKE spectra measured from high-
quality single crystals is achieved for FePd, FePt, and CoPt. Moreover, the orientational
Z []òg
dependence of the polar Kerr effect is properly predicted for FePt and CoPt. The anisotropy
Z []òg
in the spectrum is related through the sum rule to the anisotropy in part of the orbital
moment. In the case of FePd, the respective parts of the spectrum are found to
follow the anisotropy behavior of the orbital moments of Fe and Pd, but such a spectral
decomposition of individual contributions may not be possible in general.
61
/ ß áàâÈã ä
å æ çéè êå ë ì çéí î íðï

î íòñ ó íòô õ ö ÷
øÕùûú ËýüÎþ ÿ ÿ §# ÿ( Ë ÿ
In the early large-scale investigations of the MOKE spectra of TM compounds (van Engen
et al. 1983b, Buschow et al. 1983a) no particular attention was given to the series of
ï ï
XPt compounds (X = V, Cr, Mn, Fe, and Co). Buschow et al. (1983a) reported to have
measured a Co Pt alloy and a MnPt alloy; the latter was found to have a negligible
Kerr rotation. A second investigation of a compound of this series was carried out by
Brändle et al. (1991), who measured the MOKE spectra of MnPt . Brändle et al. (1991),
too, obtained a Kerr rotation for MnPt that was quite small, only -0.12 maximally. A
first indication that the XPt compounds could nevertheless be interesting was obtained
by Weller et al. (1993a), who discovered a reasonable, maximal Kerr rotation of -0.6
at 4.2 eV in a CoPt alloy. The next discovery reported was that of a giant Kerr angle of
-1.18 in MnPt (Kato et al. 1995a,b). For sample-technical reasons this giant Kerr angle
was not measured at the solid–air interface, but through a protective quartz layer. This
implies that the thus measured Kerr spectrum can be enhanced by a factor of about one
and a half over that of the solid–air interface. Since this first discovery of a large-to-giant
Kerr angle in MnPt , several other groups have also measured the Kerr spectra of MnPt
(Vergöhl and Schoenes 1996, Wierman and Kirby 1996, Wierman et al. 1997, Borgschulte
et al. 1999). Wierman and Kirby (1996) obtain a maximal Kerr rotation of -0.9 at
1.5 eV, but could not measure the peak Kerr rotation located at smaller photon energies.
Subsequently Wierman et al. (1997) observed an increase of the Kerr rotation at 95 K,
which, corrected for the glass protective layer, could possibly reach an intrinsic value of up
to -1.5 . Vergöhl and Schoenes (1996) measured the Kerr spectra of both MnPt and CrPt
at room temperature and without protective layer and obtained a smaller maximal Kerr
rotation for MnPt of nearly -0.6 at 1.2 eV. More recently, a maximum room-temperature
Kerr rotation of -0.82 at 1.1 eV was achieved by Borgschulte et al. (1999), whereas
62 CHAPTER 4. XPT (X = V TO CO) COMPOUNDS
Kokuryu et al. (1998) obtained -0.7 at 1.2 eV. Despite of these encouraging results, it
appears that further experimental research will be needed to unambiguously clarify the
Kerr spectra of MnPt . None the less, the Kerr spectra measured so far particularly provide
experimental evidence that the whole XPt series is of quite some interest in MO research.
The XPt series have received very recently attention for a second reason: XMCD
experiments made it possible to study the variation of the spin and orbital magnetic moment
on the Pt site (Maruyama et al. 1995a, 1995b). In these experiments an interesting variation
of the size of the orbital and spin moment on Pt was measured, as well as a variation of
the coupling of the Pt moment to the X moment (Maruyama et al. 1995b). For CrPt ,
Maruyama et al. (1995a) obtained an antiferromagnetic coupling to the X moment, but a
ferromagnetic coupling for FePt and CoPt . These measurements stimulated theoretical
investigations of the magnetic trends in the XPt series. Iwashita et al. (1996) calculated
the magnetic moments on the basis of the LSDA, using a scalar-relativistic FLAPW method
with SO coupling included in the final step. Oppeneer et al. (1996b) performed relativistic
band-structure calculations using the relativistic ASW scheme. Kulatov et al. (1996)
carried out similar calculations, but employed the LMTO method. As we shall discuss in
some detail in Section 4.3.1, these various computational techniques yield the same trends
in the magnetic moments. However, the theoretical moments deviate in some points from
the measured moments. Other groups calculated the spin-only moments of compounds of
the XPt series and compared these to the earlier available experimental moments obtained
from neutron scattering experiments (Kübler 1984, Hasegawa 1985, Podgorny 1991).
Tohyama et al. (1989), and more recently Shirai et al. (1995), investigated theoretically
the trends in the spin-only moments of the whole XPt series.
There are various other reasons for which the magnetic and structural properties of
some of the XPt compounds have drawn attention. VPt , remarkably, was reported to be
ð
ferromagnetic, although its constituents itself are not ferromagnetic (Jesser et al. 1981).
ð
NMR experiments indicated that in VPt there is a moment of 1.0 "! on V, which couples
antiparallel to a moment of -0.3 "! on Pt (Kawakami and Goto 1979). Moreover, VPt ,
and also VPd , crystallize in two structures, the cubic AuCu (L1 # ) structure and the
TiAl (DO#$# ) structure, which are quasi-degenerate (Jesser et al. 1981, Solal et al. 1987,
Cabet et al. 1996). Ferromagnetic ordering in VPt was calculated to occur only in the
L1 # structure (Kübler 1984). A relative structural stability of the L1 # versus the DO #%#
structure occurs in the whole series of XPt and XPd compounds (with X a 3& element).
Total-energy LSDA calculations could basically reproduce the experimental stability trend
within the XPd alloys (Rosengaard and Skriver 1994, Lu et al. 1995). Within the XPt
series (X = V to Co), all compounds order ferromagnetically on the 3& atom, except for
FePt , which orders antiferromagnetically (Bacon and Crangle 1963). Isoelectronic FePd ,
on the other hand, orders ferromagnetically (Cable et al. 1962). The pseudobinary alloy
Fe(Pd ')(+* Pt* ) , consequently, has a rich magnetic phase diagram, exhibiting canted and
non-collinear arrangements of the spin moments, which has been the subject of several
studies (see, e.g., Kouvel and Forsyth 1969, Pebler et al. 1972, Kaburagi and Tonegawa
1987, Kaburagi et al. 1990).
Our aim in investigating the XPt series is twofold. In the first place, the discovery
of a large Kerr rotation in MnPt (Kato et al. 1995a,b) indicates, that the whole group of
CHAPTER 4. XPT (X = V TO CO) COMPOUNDS 63
transition-metal-platinum alloys is exceptionally interesting for MO applications in optical

recording. Even more, it appears that large Kerr effects might still be found in materials
which were previously not regarded for their MO properties. We therefore attempt to obtain
a microscopic understanding of those physical quantities that give rise to large peaks in the
Kerr rotation spectrum. We study to this end the influence on the Kerr effect of the size of
the magnetic moment on the 3 & atom and on Pt, as well as the influence of the SO coupling
on either of the two constituents. In the second place, our aim is to undertaking a systematic
investigation of the trends in transition-metal-platinum alloys. To this end, we consider
the trends in the calculated moments in comparison with the available experiments. Also
the trends in the MOKE spectra are analyzed. All XPt compounds investigated exist in
the AuCu structure (Villars and Calvert 1985) and order ferromagnetically, except for
FePt . For the latter compound we consider only the ferromagnetic state, which should
correspond to FePt in an applied magnetic field.
øÕù-, . #
/0 132 4576 ÿ8Ï# 9:
In our calculations we have adopted the experimental lattice constants for the AuCu unit
cell (see Villars and Calvert 1985). Since we primary want to study trends in the computed
Kerr spectra, we ignore the intraband Drude contribution to the optical conductivity and
consider only the interband spectra. In Fig. 4.1 the calculated polar Kerr spectra of VPt ,
CrPt , and MnPt are shown, and in Fig. 4.2 those of FePt and CoPt (see Oppeneer et al.
1996b). All Kerr spectra given in Figs. 4.1, 4.2 pertain to the (001) magnetization direction
in the cubic L1 # structure. A Lorentzian broadening, furthermore, with a half width at half
maximum of 0.03 Ry, has been applied to all optical conductivity spectra. The recently
measured Kerr spectra of MnPt (Kato et al. 1995a,b) are also given in Fig. 4.1.
The Kerr spectra of VPt , CrPt , and MnPt are found to be very similar, as are those of
FePt and CoPt . This is the reason why we show the spectra in this combination together.
The theoretical Kerr rotations of VPt , CrPt , and MnPt have their minimum at the same
photon energy of 0.8 eV, followed by a zero crossing at 2 eV. This similarity is partially
observed in the Kerr ellipticity, too. The Kerr rotations of FePt and CoPt , however, are
distinctly different, as these have no zero crossing, and exhibit two minima, one smaller
minimum at 1.3 eV, and a larger one at 4.7 eV (see Fig. 4.2). Noticeable also, are the large
Kerr rotations that are predicted by the LSDA-DFT approach for these compounds. The
largest Kerr rotation is found for MnPt , which reaches a value of -1.5 (for interband-only)
at 0.8 eV. But also the Kerr rotation of the CrPt alloy is surprisingly large, being with a
peak value of -0.9 at 0.8 eV larger than that of the transition metals Fe and Co (Krinchik
and Artem’ev 1968a,b, Buschow et al. 1983a). In addition, the Kerr rotations predicted
for FePt and CoPt are substantial too, with peak values of -1.0 to -1.1 at 4.7 eV. With
respect to the magnitudes of the Kerr angles displayed in Figs. 4.1, 4.2, there are three
points to be noted. In the first place, the precise peak rotation magnitude depends on the
applied broadening parameter (Oppeneer et al. 1992a). A larger broadening than 0.03 Ry
would generally lead to slightly smaller, but broader spectral peaks. However, so far it
is has become an empirical experience that the broadening parameter of 0.03 Ry gives a

0.4
θK εK
0.0
-0.4
-0.8
VPt3
-1.2 CrPt3
MnPt3 exp.
MnPt3
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>A@CBDECFHGJI
Calculated polar Kerr rotation KL and Kerr ellipticity L spectra of VPt M , Ï
CrPt M , and MnPt M in the AuCu M crystal structure with (001) magnetization axis. The
theoretical spectra are all calculated with a relaxation-time broadening of 0.03 Ry, and
result from the interband optical conductivity only. The experimental data shown are
those of MnPt M (Kato et al. 1995b).
physically adequate description for TM compounds (Oppeneer et al. 1992a, Osterloh et al.
1994). Secondly, the neglect of the intraband Drude contribution to the optical conductivity
can play a role for VPt , CrPt and MnPt . An intraband Drude contribution to the optical
conductivity can be of importance for the Kerr rotation spectrum at small photon energies.
As the main Kerr rotation peak of the compounds in Fig. 4.1 occurs at a small energy, the
size of this peak will become reduced when a large intraband contribution is present. For
CoPt and FePt , the intraband contribution is less important, because these compounds
already have a relatively small Kerr rotation at low energies (see Fig. 4.2). Thirdly, it should
be noted that the ab initio Kerr spectra are essentially calculated for zero temperature. If
the Kerr spectra are measured at room temperature, where the magnetization is reduced,
then the over-all size of the thus measured Kerr rotation will be reduced, too. This property
is further examined in more detail in Section 4.4.
In Fig. 4.1 the recently measured Kerr spectra of ordered MnPt are shown also (Kato
et al. 1995b). These spectra were measured at room temperature from an annealed thin
film of MnPt on a quartz substrate, but from the substrate side, i.e., through quartz. This
implies that the thus measured spectra can be enhanced over the Kerr spectra measured
in air by a factor of about one and a half. Recently Kato et al. (1995a, 1998) measured
the Kerr rotation of MnPt films at 80 K also and found that NPO becomes enhanced to
-1.30 over the room-temperature result given in Fig. 4.1. A similar enhancement was
0.3 θK εK

0.0
-0.3
-0.6
CoPt3
-0.9 FePt3
-1.2
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;Q<R>A@CBDSECFRT+I
As figure 4.1, but for the theoretical polar Kerr spectra of FePt M and
CoPt M .
reported by Wierman et al. (1997). This is not unexpected, because the Curie temperature
of ordered MnPt is with UWV = 380 K (Auwärter and Kussmann 1950) not so much above
room temperature. After the first report of a ‘giant’ Kerr rotation in MnPt by Kato et al.
(1995a,b), the Kerr spectra of MnPt has been measured by two other groups. Wierman
and Kirby (1996) and Wierman et al. (1997) measured the Kerr spectra of MnPt films
(through quartz protective layer), and obtained reasonable agreement with those of Kato
et al. (1995b). Vergöhl and Schoenes (1996) measured the Kerr spectra of both CrPt
and MnPt on bulk polycrystalline samples at the solid–air interface. The peak in the Kerr
rotation of MnPt comes out smaller by a factor of two, but the shape of the spectrum
corresponds to that of Kato et al. (1995b). Recently, Borgschulte et al. (1999) measured
the room-temperature Kerr rotation of MBE grown MnPt films at the solid–air interface,
and obtained a maximum rotation of -0.82 . Vergöhl and Schoenes (1996) were the first to
measure the Kerr spectra in CrPt . They obtained rotation and ellipticity spectra which are
about two times smaller than the theoretical spectra of Fig. 4.1, yet, their shapes compare
rather good to that of the theoretical counterparts. Thus, within the limitations concerning
the size of the calculated Kerr rotation mentioned above, and the possible influence of the
quartz substrate and temperature on the Kerr spectra, it can as yet only be concluded that
the shape of the theoretical and experimental Kerr rotation and ellipticity spectra are in
good agreement.
The Kerr spectra of Co* Pt ')(+* alloys were measured by several groups on as-cast or
deposited samples near the 1 – 3 composition (Buschow et al. 1983a, Zeper et al. 1989,
Weller et al. 1992, 1993a). The measured Kerr rotation of a Co #$X Pt # alloy was found
to agree very well in shape with an earlier ab initio calculation (using the computational
scheme of Oppeneer et al. 1992b), but the measured rotation was about 30% smaller
than the calculated one (Weller et al. 1993a). This difference was attributed by Weller et
al. (1993a) to a reduction of the sample magnetization at room temperature: the Curie
temperature of CoPt alloys is about 500 K, depending on sample deposition (see Sanchez
et al. 1989, Rooney et al. 1995).
As we have mentioned in Section 4.1, within the XPt series there is a phase stability
competition between the L1 # (i.e., AuCu ) structure and the DO #%# (i.e, TiAl ) structure.
For X = Mn, Fe, and Co, the L1 # structure is experimentally stabilized over the DO #%#
structure, but for VPt the DO #$# and L1 # structures are quasi-degenerate, with the DO #%#
structure being preferred over the L1 # structure at U = 0 (see, e.g., Cabet et al. 1996).
A sort of cross-over occurs also for CrPt : The experimental ground-state structure is
L1 # , which is found also from first-principles total-energy calculations, however, only for
ferromagnetically ordered CrPt (Lu et al. 1995). In the paramagnetic state the DO #%#
structure has the lowest total energy (Lu et al. 1995). In other words, ferromagnetism
stabilizes the L1 # structure over the DO #$# structure. We have calculated the total energies
for VPt and CrPt in both the L1 # and DO #%# structure. Our results, which we do not outline
in detail here, are in accord with those of other calculations: For VPt the paramagnetic
DO#$# structure is just more stable than the ferromagnetic L1 # structure (see, e.g., Kübler
1984), while for CrPt we obtain the same result as Lu et al. (1995). Interestingly, we
ð ð
find a rather large dependence of the Cr magnetic moment on the crystal lattice: the total
ð
moment on Cr is 2.89 ! for the L1 # structure, but only 2.23 ! for the DO #$# structure.
ð
Also, the total moment on Pt is calculated to be -0.14 ! in the DO #$# structure, but only
-0.09 Y! in the L1 # structure. We expect that both the difference in crystal structure and
magnetic moment will lead to different MO Kerr spectra. In Fig. 4.3 we show the calculated
Kerr spectra of CrPt in the two crystal structures. The Kerr rotation of CrPt in the DO #%#
structure is in the energy region of the main peak near 0.8 eV only -0.6 , compared to the
-0.9 of the L1 # structure. The two Kerr rotations and ellipticities are different also in
ð ð
various other respects. The smaller magnitude of the peak Kerr angle can expectantly be
related to the smaller magnetic moment per formula unit (2.62 ! for L1 # versus 1.91 !
for DO #%# ). Of course, the structural differences contribute to the differences in the Kerr
spectra, too. In this respect it is worthwhile to note that recently a new structural phase
of Co Pt was discovered by means of MOKE spectroscopy (Harp et al. 1993). The Kerr
spectra of a hexagonal, ordered Co Pt phase were found to be markedly different from
those of a cubic Co Pt alloy (Harp et al. 1993). Inspired by this discovery, we have
calculated the MOKE spectra of CoPt in both the L1 # and DO #$# structures. Unlike CrPt ,
the magnetic moment on Co is nearly invariant with respect to the crystal structure. In
Fig. 4.4 we show the influence of the crystal structure on the MOKE spectra of CoPt .
We furthermore explore in Fig. 4.4 the anisotropy of the Kerr spectra with respect to the
magnetization axis for CoPt . The Kerr spectra given in Figs. 4.1, 4.2, and 4.3 pertain to the
common magnetization orientation in these compounds, i.e., the (001) magnetization axis.
A non-equivalent magnetization direction in the L1 # structure is, for instance, the (111)
direction. As can be inferred from Fig. 4.4 the polar MO Kerr spectra depend appreciably
CrPt3
0.8

θK εK
0.4
0.0
-0.4
L12
-0.8 DO22
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>A@CBDZECFR[+I
Calculated influence of the crystal structure on the polar Kerr spectra of
CrPt M . Shown are the calculated (interband-only) MOKE spectra of CrPt M in both the
L1 \ and DO \]\ crystal structure.
on the crystal structure. The magnitude of the differences of the L1 # and DO #$# Kerr
spectra is only just a bit smaller than those measured for the Co Pt alloys (Harp et al.
1993). The influence on the polar Kerr spectra of the magnetization direction relative to
the crystallographic axis, on the other hand, is clearly very small. We mention that for
CoPt we found the largest magnetocrystalline Kerr anisotropy within the XPt series.
The comparatively small influence of the magnetization direction on the Kerr spectra is
undoubtedly related to the high degree of isotropy inherent to the cubic L1 # structure.
Recently we found a sizable magnetocrystalline anisotropy effect in the Kerr spectra of
several layered XPt compounds, which consist of alternating X and Pt monolayers, and
thus have a very anisotropic structure (Oppeneer and Antonov 1996). If we compare the
relative differences in the L1 # and DO #$# Kerr spectra of CrPt to those of CoPt , then it
becomes apparent that the larger differences occur for CrPt , where the magnetic moments
depend substantially on the crystal structure.
øÕù$^ . _ `9= ÿ a ( #

/0 1cbd 8)8 #'ÿ
á1egfheji kmlQnporq8s+t%uvlQwyxAlpl:x{z1|Ys~}0s+uxors+t%wrWtjo |uq:%Wr q x{t lwru
The Kerr rotation spectra of CrPt , MnPt , FePt , and CoPt were recently also calculated
by Kulatov et al. (1996) and those of CrPt , MnPt , and CoPt by Köhler (1997). The
CoPt3
0.0
-0.3
-0.6
L12 (001)
L12 (111)
-0.9 DO22 (001)
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>J@CBDECFECI
Dependence of the ab initio polar Kerr spectra of CoPt M on the magne-
tization direction and the crystal structure. Theoretical (interband-only) polar MOKE
spectra are given for both the (001) and the (111) magnetization directions in the L1 \
crystal structure, and the (001) magnetization direction of the DO \]\ structure.
theoretical MOKE spectra of Köhler compare closely to the above given results, which is
understandable, since in both studies a similar computational scheme, based on the ASW
method was employed. Köhler (1997) added the intraband Drude contribution, which
leads to small differences at photon energies below 2 eV. Kulatov et al., however, obtain
Kerr rotation spectra which are in several respects strikingly different from those shown
in Section 4.2. A maximal Kerr angle of -2.5 is computed for MnPt , and of -2.2 for
CrPt . The shape of the Kerr rotation spectra for all four compounds does, however, agree
with the above given results for energies above 2.5 eV. For FePt and CoPt Kulatov et
al. calculate a zero crossing in the Kerr angles already above 1 eV. This is certainly not
supported by the experimental Kerr rotation of CoPt which was measured down to 0.6 eV
by Weller et al. (1993a). Also the fact that the calculated maximal Kerr angle does not
reduce much in going from MnPt to CrPt is not consistent with the measurements of
Vergöhl and Schoenes (1996). The theoretical Kerr rotation spectra shown in Fig. 4.1, on
the other hand, exhibit a clear trend of a reducing maximum Kerr peak from MnPt to
VPt .
The question to be addressed, is thus: How can these differences in two LSDA band-
structure calculations of the Kerr effect arise? The Kerr spectra given in Section 4.2 were
calculated using the ASW method, whereas Kulatov et al. (1996) employed the LMTO
method. It is, however, unlikely that the employed band-structure scheme is responsible
for the differences in the Kerr spectra. In the first place, the early ASW Kerr spectra for Fe
and Ni were confirmed by other MO Kerr effect calculations on the basis of independent
band-structure schemes (Guo and Ebert 1995, Antonov et al. 1995, Mainkar et al. 1996,
MacLaren and Huang 1996). The MO Kerr spectra of MnPt , moreover, were calculated
with both the ASW method and a relativistic version of the LMTO method (Antonov et
al. 1995). Both computational schemes yielded nearly identical MOKE spectra for MnPt
(Oppeneer et al. 1995a). Consequently, one may exclude that the deviations are related
to employing these different band-structure schemes. Second, besides the energy band-
structure, details of the interplay of exchange splitting and SO interaction are of course
responsible for the MOKE spectrum. Relevant measurable quantities for the Kerr effect
are therefore the spin and orbital moments on each of the constituting elements. The spin
and orbital moments on Pt have recently been measured using XMCD by Maruyama et al.
(1995b). In Table 4.1 we give the experimental spin and orbital moments on the Pt site,
as well as the recently reported theoretical moments. The latter have been calculated with
the FLAPW scheme (Iwashita et al. 1996), the ASW scheme (Oppeneer et al. 1996b),
and the LMTO scheme (Kulatov et al. 1996). Table 4.1 illustrates that the three different
band-structure schemes (on the basis of the LSDA) yield essentially the same moments on
Pt throughout the series. Moreover, the calculated LSDA moments compare very well to
the experimental moments for MnPt , FePt , and CoPt . Only for CrPt , the calculated
spin and orbital moment differ each substantially from the experimental values, but the
total moment on Pt compares much better. We can thus safely conclude that the differences
in the published first-principles MO Kerr spectra are definitely not caused by the employed
band-structure schemes. Most likely, these differences originate from the accuracy of the
evaluation of the elements of the optical conductivity tensor (see Section 2.5.2).
8vW DÈCFHGJI
Comparison of the calculated and experimental spin ( Z ) and orbital ( Z ) moments
on Pt in the XPt M (X = V, Cr, Mn, Fe, and Co) series. The theoretical moments given were all
calculated for the ferromagnetically ordered XPt M compounds in the AuCu M crystal structure, using
three different band-structure schemes: the FLAPW method (Iwashita et al. 1996), the ASW method
(Oppeneer et al. 1996b), and the LMTO method (Kulatov et al. 1996). All moments are given in
Bohr magnetons ( ).
VPt M CrPt M MnPt M FePt M CoPt M

Reference
Iwashita et al. (1996) -0.06 -0.03 -0.04 -0.05 0.12 0.00 0.34 0.05 0.29 0.04
Oppeneer et al. (1996b) -0.06 -0.03 -0.04 -0.05 0.12 0.00 0.32 0.06 0.26 0.05
Kulatov et al. (1996) — — -0.04 -0.05 0.12 0.00 0.32 0.05 0.24 0.05
Experiment — — 0.02 -0.12 0.14 -0.01 0.30 0.08 0.30 0.09

Maruyama et al. (1995b)
For a discussion of the energy bands and DOS of the XPt alloys we refer to the
mentioned theoretical papers (Oppeneer et al. 1996b, Iwashita et al. 1996). With respect
to the interesting similarity in the computed Kerr sprectra of VPt , CrPt , and MnPt , on
the one hand, and FePt and CoPt on the other hand, it has been noted that this similarity
cannot simply be understood from the band structures and DOS’s (Oppeneer et al. 1996b).
árefhe kmlQnporq8s+t%uvlQwxJlx{z1||Wo`|Ys+t%n|wQxAq:nq8Qwr|Wx{tj¡nln|wQx{u
The MOKE spectra originate in the calculations from optical transitions between polarized
energy bands of the solid. Besides the energy-band structure, the interplay of exchange
splitting and SO interaction is responsible for the MO Kerr spectrum (Argyres 1955).
Relevant measurable quantities for the Kerr effect are therefore the spin and orbital moment
on the constituting elements. Apart from the recent XMCD results for the spin and orbital
moments on Pt, the magnetic moments on each of the elements of the XPt compounds
(X = Cr to Co) have previously been measured in polarized neutron experiments (in some
cases already more than 30 years ago). The magnetic moments of VPt were obtained in
a NMR experiment (Kawakami and Goto 1979). In analyzing the trends in the theoretical
Kerr spectra, it is of interest to corroborate a relationship of the trend in the spectra to the
magnetic moments. Furthermore, a comparison of experimental and calculated moments
provides insight in the applicability of the LSDA to these materials.
In Table 4.2 we compare the calculated magnetic moments of the XPt compounds to
the experimental moments, as far as these are available. Neutron-scattering experiments
showed that in CoPt and MnPt the Pt moment couples parallel to the 3 & TM moment
(Pickart and Nathans 1962, Menzinger and Paoletti 1966, Antonini et al. 1969). In CrPt
and VPt the Pt moment couples antiparallel to the 3& TM moment (Pickart and Nathans
1963, Burke et al. 1980, Kawakami and Goto 1979). The calculated magnetic moments
of CoPt and MnPt are in good agreement with experiment (see Table 4.2), and for CrPt
and VPt the ferrimagnetic structure is in accordance with experiment, but the values of
'
the moments only agree to a lesser extent . In the case of CrPt , the two neutron-scattering
experiments yield rather different moments on Cr, which could be due to different form
factors used in the evaluation of the moments (Pickart and Nathans 1963, Burke et al.
1980). The total moment on Pt in CrPt , as derived through XMCD (Maruyama et al.
1995b), does in addition not correspond to the moment obtained from polarized neutron
experiments. The moments in CrPt are thus experimentally not very accurately known,
ð
for which reason no definite conclusions can be drawn for this compound. Our calculated
ð
total moment of 2.62 ! for CrPt corresponds, moreover, very well to that of another
recent calculation (Lu et al. 1995), and to the total moment of ¢ 2.5 ! derived from
the bulk magnetization (Besnus and Meyer 1973, Lewis and Williams 1976, Goto 1977).
ð
FePt has an antiferromagnetic ground state, with an experimental total moment on Fe of
ð ð
3.3 Y! (Bacon and Crangle 1963), which is in reasonable agreement with the calculated
total Fe moment of 3.1 "! for ferromagnetic FePt . The total moment of 0.38 "! on Pt
measured for ferromagnetic FePt (Maruyama et al. 1995b), furthermore, agrees well with
£
The magnetic moments have been recalculated using more ¤ -points than in the original work of
Oppeneer et al. (1996b), which lead to minimal differences of less than 0.02 W for VPt M .
8W DE¥FRT+I
Calculated and experimental moments of the XPt M (with X = V, Cr, Mn,
Fe, and Co) compounds. The theoretical moments were calculated for the ferromagnetic
state of the XPt M compounds in the AuCu M crystal structure (see Oppeneer et al. 1996b).
Spin moments are denoted by , orbital moments by . All moments are given in
Bohr magnetons ( ).
X §¦ §¦ §¦¨ª© ÅÃ ¨«© ÅÃ

V 1.47 0.01 -0.06 -0.03 1.0¬ -0.3¬
Cr 2.73 0.16 -0.04 -0.05 2.33 0.10 ® -0.27 0.05 ®
3.37¯ -0.26¯
Mn 3.71 0.03 0.12 0.00 3.60 0.09 0.17 0.04
3.64 0.08 ° 0.26 0.03 °
Fe 3.00 0.11 0.32 0.06 — 0.38 ±
Co 1.66 0.05 0.26 0.05 1.64 0.04 ² 0.26 0.02²
³
´ Kawakami and Goto (1979)
Pickart and Nathans (1963)
µ
Burke et al. (1980)
Pickart and Nathans (1962)
¶
· Antonini et al. (1969)
Maruyama et al. (1995b)
¸
Menzinger and Paoletti (1966)
ð
the calculated Pt moment of 0.38 ! . We may therefore conclude that the LSDA provides
an adequate description of the magnetic moments of particularly MnPt , FePt , and CoPt .
The relative structural stability of the DO #%# and L1 # structures within the XPt series is
also very well explained from LSDA calculations. The relationship between the size of the
magnetic moment and the Kerr rotation will be considered in the next Section.
øÕùø 4 9=¹ a ( 0ü # # t#
/`9:¹ º5 9:9 # x#
»
Density-functional theory in the LSDA predicts a large Kerr effect in the XPt alloys.
Noticeably, the Kerr rotations predicted are much larger than those calculated for, e.g.,
Fe, Co, or Ni, where the same broadening parameter of 0.03 Ry was used (Oppeneer
et al. 1992a). An important issue is therefore to identify the origin of the large Kerr
effect in these compounds. To this end, we examine the dependence of the MO spectra
on the exchange splitting, the SO interaction, and the optical transition matrix elements
(MEs). As it can be expected that the Kerr effect in each of these compounds is of the
same origin, we do this only for one compound, CrPt . The exchange splitting and the
SO coupling can be studied by scaling the corresponding terms in the Kohn-Sham-Dirac
equation artificially with a constant prefactor. This we do in a non-self-consistent way, i.e.,
after self-consistency has been achieved, only one iteration is performed with the modified
equation (a self-consistent calculation would lead to a different band structure). From

the resulting band structure we then compute the optical spectra. These modifications
can in addition be done atom dependent, i.e., within each atomic sphere, so that we can
investigate the separate effects of these quantities on Cr and on Pt. The outcomes of these
model calculations for the Kerr rotation of CrPt are shown in Fig. 4.5. In the upper panel,
the importance of the exchange splitting is illustrated. When the exchange splitting on Pt
is set to zero, the Kerr rotation remains as it is. But when we do the same for the exchange
splitting on Cr, the Kerr rotation totally vanishes. This implies that the exchange splitting
due to Cr is crucial for the sizable Kerr rotation, but that of Pt is unimportant. Furthermore,
an enhancement of the exchange splitting on Cr by a factor of two (dashed line) leads to a
much larger peak in the Kerr rotation. The middle panel of Fig. 4.5 shows the dependence
on the SO coupling. If we set the SO coupling on Cr to zero, the Kerr rotation practically
doesn’t change (dotted line). On the other hand, when the SO coupling on Pt is zero,
the Kerr rotation almost disappears (dashed line). Thus, the SO coupling of Pt is equally
responsible for the large Kerr rotation as is the exchange splitting of Cr. An intermediate
scaling of the SO coupling of Pt by a factor of 0.5 leads to an approximately half as large
Kerr angle, thereby illustrating the almost linear dependence of the Kerr effect on the SO
interaction of Pt in these compounds. The lower panel in Fig. 4.5, finally, displays the
importance of the site-dependent MEs. Within an atomic sphere about either one of the
atomic positions, the optical transition matrix elements are set to zero. If this is done for
the MEs on Cr, the Kerr rotation doesn’t change much. But if the MEs on Pt vanish, a
large impact on the Kerr rotation is found (dashed curve). This indicates that the MEs on
the Pt site are more important for bringing about the large Kerr peak at 1 eV, than are those
of Cr. Making the matrix elements zero gives, of course, only an impression of which site
the main contribution comes from. To obtain information about which kinds of bands are
responsible for the Kerr peak, it is instructive to exclude a particular transition ME. Due to
the selection rules for optical transitions, these transitions can only take place between band
states with an angular momentum difference of ¼½ (see, e.g., Reim and Schoenes 1990).
By excluding, for instance, the ¾¿vÀ transition ME we can investigate the contribution of
this type of transitions. However, as the transition matrix must be hermitian, we have to
exclude also the conjugated transition, i.e., both ¾¿vÀ and ÀP¿)¾ transitions. The results of an
investigation of the importance of the various transitions on Pt are shown in Fig. 4.6 for the
'jÃ
real part of the diagonal optical conductivity, Á=*)Â * , the imaginary part of the off-diagonal
#$Ã
conductivity, Á *)Â Ä , and the Kerr rotation. The upper and middle panel of Fig. 4.6 show
'jÃ #$Ã
that both Á *)Â * and Á *)Â Ä are strongly reduced when the ¾C¿JÀ and ÀP¿)¾ interband transitions
are excluded, more than when the À¿$Å and Å-¿JÀ transitions are excluded. Especially the
off-diagonal conductivity almost disappears in the energy region around 1 eV if the ¾C¿vÀ
and À¿¾ transitions on the Pt sites are excluded. Because this peak in the off-diagonal
conductivity at 1 eV is responsible for the peak in the Kerr rotation spectrum, this shows
that the À¿¾ and ¾C¿JÀ transition MEs on Pt account for most of the Kerr effect in this
frequency region. The other transitions, Æ)¿)¾ and À¿$Å , also have a minor influence, but
excluding these still gives approximately the same Kerr rotation (see Fig. 4.6, lower panel).
Thus, we conclude that the hybridized & orbitals of Pt, being subject to the strong SO
interaction on the Pt site, contribute most to the optical transitions that lead to a large Kerr
0.8
0.4
0.0
CrPt3
-0.4
Cr M=0
-0.8 Pt M=0
Cr M/2
-1.2
Cr M*2
Kerr rotation (deg)
0.4
0.0
CrPt3
-0.4 SOCr=0
SOPt=0
-0.8
SOPt/2
4 CrPt3
MECr=0
2 MEPt=0
-2
0 1 2 3 4 5 6
Energy (eV)
;=<?>A@CBDE¥FRÇ+I
Study of the influence of the exchange splitting (denoted by È ), spin-
orbit (SO) coupling, and optical transition matrix elements (ME) on the Kerr rotation of
CrPt M . The upper panel shows the effect of multiplying the spin-polarized part of the
Kohn-Sham-Dirac equation with a constant factor, either on the Cr site or on the Pt site.
The middle panel shows the effect of multiplying the SO coupling either on Cr or on
Pt with a constant prefactor (see text). The lower panel depicts the effect of setting the
matrix elements on Cr or on Pt to zero.
angle.
80
σxx (10 s ) CrPt3
-1
MEPt(p-d)=0
60 MEPt(d-f)=0
14
MEPt=0
40
(1)
20
0
σxy (10 s )
-1
CrPt3
10 MEPt(p-d)=0
14
MEPt(d-f)=0
MEPt=0
5
(2)
0
Kerr rotation (deg)
2 CrPt3
MEPt(p-d)=0
MEPt(d-f)=0
1
MEPt(s-p)=0
-1
-2
0 1 2 3 4 5 6
Energy (eV)
;=<?>A@CBDECFRÉ{I
Influence of the exclusion of various optical£Ì matrix elements on the Pt
site on the real part of the diagonal optical conductivity, Ê ©jË © , the imaginary part of the
\ Ì
off-diagonal optical conductivity, Ê ©jË Í , and the Kerr rotation. The notation ME (ÎAÏ
Ð Ð Ð
) = 0 means that on the Pt site the ÎAÏ interband transitions and the Ï%Î interband
transitions are excluded from the optical matrix element.
øÕù-Ñ Ò ÿ ÿ!ÿ 1 ÿ ÿ
From these investigations the following picture of the Kerr effect in these compounds
emerges: Pt is the magneto-optically active element, and creates the large Kerr rotation
GdPt3
θK εK
0.4
0.2
0.0
-0.2
-0.4
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>A@CBDE¥F?Ó{I
Calculated, interband-only, Kerr rotation and Kerr ellipticity of GdPt M .
The lifetime parameter used was ÔPÕ = 0.03 Ry.
through its large SO interaction. The important MO transitions are the ¾C¿JÀ and À¿¾
transitions on Pt. The 3& elements are magneto-optically not very active. Their role is to
supply through their exchange splitting enough hybridized, spin-split energy bands. This
understanding suggests the following recipe for finding a material having a sizable Kerr
rotation: such material should contain elements with a large SO coupling, for instance Pt,
Bi, or an actinide. Also should it contain an element having a sufficiently large magnetic
moment, but this element does not need to have a strong SO interaction, like for instance
Mn. Also should there be a substantial hybridization between the valence orbitals of these
two kinds of constituents. Elements which have a large, but localized atomic moment, like
some of the 4 ß elements, are in the latter respect not suitable. The unhybridized, localized
4 ß states do not overlap much with the Pt 5 & orbitals, and therefore they do not contribute
much to the optical spectra, because the corresponding transition matrix elements are quite
small (Oppeneer 1995). Also, due to the lack of hybridization, other band states do not
become sufficiently polarized, and do therefore also not contribute much (Oppeneer 1995).
As an example of such a situation we briefly consider GdPt , which is known to crystallize
ð
in the L1 # structure (Colinet and Buschow 1986). In GdPt , the calculated total magnetic
moment on Gd is large, 6.9 Y! , which is thus much larger than the moments on the 3 &
elements in XPt . In spite of the “rule of thumb” requiring a large moment, the Kerr effect
is nevertheless calculated to be comparatively small. In Fig. 4.7 we show the calculated
Kerr spectra of GdPt . The maximal Kerr angle at 4 eV is just less than -0.4 , i.e., only of
GdPt3
5
Energy (eV)
−5
−10
Γ X M Γ Z R A
;=<?>J@CBDECFRÖ+I
Energy band structure of GdPt M calculated within the LSDA approach.
Note the narrow 4 × spin-minority, spin-majority, subbands at 1.5 eV and -3.5 eV that
are exchange split by about 5 eV.
similar magnitude as the Kerr angle of VPt . We note that the Kerr rotation of GdPt has
a typical shape, with a peak at 4 – 5 eV. It has been outlined for pure Gd by Erskine (1976,
1977) that such a structure can be attributed to the 4 ß bands. Primarily intra-atomic 4 ß
to & transitions on Gd are responsible (Erskine 1976, 1977). We find that for GdPt the
#%Ã #$Ã
calculated Á *)Â Ä0Ø«ÙÚ indeed resembles in shape the measured Á *)Â ÄÛØ«ÙÚ of pure Gd (Erskine
1976), which, however, occurs for the latter at energies that are at least 1 eV higher. The
energy position of the 4 ß ’s in GdPt need not be identical to that in pure Gd, but the
applicability of the LSDA to Gd-based compounds could be questioned as well (see, e.g.,
Bylander and Kleinman 1994). In the present LSDA calculation for GdPt we obtain 4 ß
bands which are exchange split by about 5 eV. This can be recognized from the LSDA
energy bands of GdPt which we show in Fig. 4.8. The occurrence of a large, but localized
moment does evidently not suffice to bring a large Kerr effect about.
For the XPt alloys we furthermore conclude that the behavior of the magnitude of the
peak in the Kerr rotation spectra with respect to the 3& element, as shown in Fig. 4.1 and
4.2, and also the dependence of the magnitude of the Kerr effect on the magnetization, can
fully be understood on the basis of our model calculations. The increase of the peak in the
Kerr angle at about 1 eV when going from VPt to MnPt (see Fig. 4.1) is caused basically
by the corresponding increase of the exchange splitting, as is witnessed by the increasing
magnetic moments in Table 4.2. This is most clearly demonstrated by the scaling of the
peak in the Kerr rotation of CrPt with respect to the scaling of the exchange splitting
(Fig. 4.5, upper panel). Also, it can be understood from this behavior that a reduction of
the magnetization at room temperature leads to a reduction of the Kerr rotation, as was
proposed originally for CoPt (Weller et al. 1993a). This dependence on the magnetization
explains also why the Kerr rotation of FePt is larger than that of CoPt .
In conclusion, the series of XPt compounds is predicted to exhibit rather large Kerr
effects. The origin of the Kerr effects is shown to rest in an interplay between the exchange
splitting of the 3 & element and the SO coupling of Pt in the hybridized bands. The
dependence of the Kerr spectra in the XPt compounds on the crystallographic direction of
the magnetization is found to be very small. This finding corroborates with the high degree
of isotropy of the cubic L1 # crystal structure. The dependence of the Kerr spectra on the
crystal structure (L1 # versus DO #$# ) is larger, and should in principle be measurable. The
agreement between the ab initio calculated Kerr spectra and the experimental results for
MnPt Kato et al. 1995a,b, 1998, Kato 1997, Verg öhl and Schoenes 1996, Wierman and
Kirby 1996, Wierman et al. 1997, Kokuryu et al. 1997, Borgschulte et al. 1999), finally,
looks very promising. Yet, further measurements on these compounds are still desirable,
in order to obtain a complete picture of the correspondence between experimental and
first-principles Kerr spectra.
79
Ü ÝßÞáà7âäãæå ç
õ è é ê õ ö ë ìîí0ê çìö
Ñ ùûú ï 9 Zä : _2 ð
Manganese-bismuth has been studied for two decades because it is considered a promising
material for magneto-optical memories (see, e.g., Buschow 1988, Schoenes 1992). MnBi
crystallizes in the hexagonal NiAs crystal structure, with the ñ -axis being the easy mag-
netization axis (Roberts 1956). In MnBi films deposited on a substrate, the ñ -axis orients
perpendicularly to the substrate, so that the magnetization orients perpendicular to the film
plane also (Chen 1971). In addition, MnBi was found to exhibit a large Faraday effect
(Williams et al. 1957, Chen et al. 1965, 1968, Chen and Aagard 1970), and a giant polar
Kerr effect (Chen and Gondo 1964, Egashira and Yamada 1974, Yoshii and Egashira 1974,
Wang 1990, Huang et al. 1994, Di et al. 1991, 1992a,b, 1994a,b), which can both be
utilized to read-out magnetically stored information. Although these properties are very
favorable for MO recording, MnBi has the drawbacks, first, that the coercive field increases
with temperature, showing a maximum at 550 K, which is unfavorable for thermomagnetic
writing (Guo et al. 1993). And second, at about 630 K MnBi undergoes a structural phase
transition from the low-temperature NiAs phase to a distorted NiAs structure at higher
temperatures, in which some Mn atoms occupy interstitial positions in the hexagonal NiAs
unit cell (Andresen et al. 1974). This phase transition is accompanied with a simultaneous
ferromagnetic to paramagnetic transition (Andresen et al. 1974, Chen 1974). But even
in the quenched high-temperature phase the room-temperature Kerr and Faraday rotation
are sufficiently large to be interesting for optical recording memories (Chen and Aagard
1970, Di et al. 1994b). However, within the process of thermo-magnetic writing the
high-temperature phase is metastable with respect to the low-temperature phase, so that a
continuous transformation to the low-temperature phase takes place (Chen et al. 1970).
This phase transition would limit the practical operation time of a MnBi memory to only a
few years (see Chen et al. 1970), thereby making this technological application of MnBi
80 CHAPTER 5. MNBI AND RELATED COMPOUNDS
impossible. In order to overcome this shortcoming, many attempts have been undertaken
to improve the structural stability of MnBi by selective doping (Unger et al. 1972, Wang
1990, Di et al. 1992a,b, Shang et al. 1995, 1997, Shang 1996). In the present status of this
line of research, it appears that especially doping with Al leads to an improved structural
stability, while the MO quality does not differ much from that of pure MnBi (Wang 1990,
'
Di et al. 1992b, 1994a, Huang et al. 1994). Given the fact that the figure of merit of MnBi
is more than two times larger than that of Co/Pt multilayers (Wang 1990, Di 1992), the
practical application of Al-doped MnBi in MO memories is therefore becoming likely. It
was recently reported that the number of achievable write/erase cycles of Al-doped MnBi
is over ½òó times, which is a significant improvement over the ½ò cycles possible for
undoped MnBi (Xu et al. 1995).
Apart from the technological view point, the MO properties of MnBi are also inter-
esting from a fundamental point of view: The Kerr rotation of -1.6 measured on MnBi
at 85 K and that of -1.25 measured at room temperature are among the record values
obtained for the air–material interface of transition-metal compounds (Di et al. 1992a, Di
1995, Di and Uchiyama 1996). The low-temperature Kerr rotation is of similar size as
that of PtMnSb, which has a maximum Kerr angle reported of -1.27 (van Engen et al.
1983a) up to -2.0 (Takanashi et al. 1988, Sato et al. 1992b, Ikekame et al. 1993). For
this reason, it is important to obtain an understanding of the microscopic mechanism that
gives rise to the record MO properties. As has been shown in the case of PtMnSb, actual
relativistic band-structure calculations are required to address the microscopic mechanism
of the giant Kerr rotation (Oppeneer et al. 1995a, Antonov et al. 1997). The MO Kerr
spectra of MnBi were calculated recently by Sabiryanov and Jaswal (1996), Oppeneer et
al. (1996a), and Köhler and Kübler (1996). A first MO investigation of MnBi was due to
Kübler (1995). The band structure of MnBi (which in itself is obviously insufficient for
explaining the MO properties) was calculated earlier by Coehoorn and de Groot (1985).
In the following we shall first present our theoretical results for the optical conductivity
spectra and MOKE spectra of MnBi (Oppeneer et al. 1996a). We also give calculated
spectra for Mn # Bi to estimate the influence of the Mn–Bi stoichiometry. Subsequently, the
origin of the giant Kerr effect in MnBi will be addressed. This is done in comparison to the
relatively small Kerr effects calculated for the related compounds MnAs and MnSb, but
the large Kerr effect predicted for CrBi, and the moderate Kerr effect predicted for CrTe.
Experimentally it has been observed that the Kerr effect in MnBi exhibits a pronounced
temperature dependence (Di et al. 1992a, Di and Uchiyama 1996). Within the present
theoretical approach it is not yet possible to study temperature effects in an ab initio man-
ner, but we investigate the observed unusual temperature dependence of the MnBi Kerr
spectra in a model approach, in which we investigate the influence of a temperature related
reduction of both the magnetization and average lifetime. This model study together with
the achieved insight in the record Kerr effect allow, as will be outlined, for an almost
complete understanding of the outstanding MO behavior of MnBi.
£ £õ \
The figure of merit that is used in the work of Wang (1990) is defined as ô K L ,ô being the
reflectivity.
CHAPTER 5. MNBI AND RELATED COMPOUNDS 81
ù
Ñ -, . #
/0 132 4576 ÿ8Ï# 9: (ö2 ð
The structural and magnetic properties of MnBi were investigated already several decades
ago (Roberts 1956, Andresen et al. 1967, Chen 1974). In our calculations we used the
experimental lattice constants ÷ø 0.4290 nm, ñø 0.6126 nm (Andresen et al. 1967), but
also made test calculations using low-temperature constants (Roberts 1956). This did not
lead to noticeable differences in the optical spectra. We further mention that in the calcu-
lations we tested the influence of ß states included in the basis functions and performed
electronic structure calculations with additional empty spheres on the interstitial lattice
sites. This also did not change much in the calculated MOKE spectra. Although the MO
Kerr effect in MnBi was measured already some twenty years ago, only very recently a
systematic investigation of the spectral dependence of the Kerr effect under variation of
the Mn–Bi composition, Al-doping, and temperature was carried out (Di 1992, Di et al.
1991, 1992a,b, 1994a,b, Di and Uchiyama 1996). Besides, while most experiments have
been conducted on samples with a protective layer, which can lead to an enhancement
of the measured Kerr effect over that of the material–air interface by about 30 – 50%,
these recent measurements were corrected for the refractive indices of the captive layer
(Di 1995). We shall therefore compare our first-principles results to these measurements.
These experimental Kerr spectra obtained at 85 K as well as the calculated ones are shown
in Fig. 5.1. First-principles theory predicts a very large Kerr rotation in MnBi of about
-1.75 at 1.8 eV, which is even larger than the measured peak value of -1.6 (Di 1995). The
lifetime broadening parameter ùJú used was 0.04 Ry (i.e., ú is the half width at half maxi-
mum of a Lorentzian), which is a moderately large value for transition-metal compounds.
We wish to point out that a smaller (but still reasonable) relaxation time broadening of
0.02 Ry would already result in a theoretical peak value of -2.22 . Therefore, according
to theory, a larger Kerr effect than measured as yet should be possible. Experiment shows
a second maximum in the Kerr angle NO at 3.4 eV, where our calculation gives only a
shoulder. This second peak at about 3.4 eV has received wide attention, because its energy
position is attractive for UV laser-light MO recording applications (420 nm recording tech-
nology) (Huang et al. 1994, Shang et al. 1995, 1997). The origin of this peak cannot fully
be understood on the basis of our calculations. A tentative explanation of the difference
might be the sample composition, which is – in the experimental data shown in Fig. 5.1 –
Mn ' #$# Bi (Di 1992, Di et al. 1992a, Di and Uchiyama 1996). There is thus an excess
of Mn in the sample. To examine the changes caused by the excess of Mn we performed
test calculations for a hypothetical Mn # Bi compound in the Heusler C1 ´ structure where
each Mn atom is tetragonally surrounded by Bi. In the (111) direction this compound has
a trigonal symmetry, like the (0001) NiAs phase. The calculated Kerr spectra of Mn # Bi
are also shown in Fig. 5.1 (dotted curve). In the theoretical Kerr rotation of Mn # Bi there
is a peak at about 4.3 eV and a smaller one at 2 eV, which is at the same position as the
main peak of MnBi. As Mn # Bi in this structure is only a hypothetical compound, we have
used a guessed lattice constant. The position of the peak at about 4 eV is rather sensitive
to the value of the lattice constant. Thus, if the stoichiometry shifts from MnBi to Mn # Bi
there appears to be a tendency to reduce the first peak at 1.8 eV and to enhance the peak at
about 3.5 eV. This corresponds exactly to what is seen in the experimental Kerr spectrum
MnBi
2

1
-1
-2 theory MnBi δ=0.04 Ry

theory MnBi δ=0.02 Ry
exp. Mn1.22Bi 85 K
theory Mn2Bi δ=0.04 Ry
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>A@CBDûÇ+FHGvI
Calculated and experimental polar MOKE spectra of Mn–Bi compounds.
The theoretical spectra are calculated for MnBi in the NiAs structure, for two lifetime
broadenings, and for hypothetical Mn \ Bi in the Heusler C1 structure. The experimental
®
Ã
Kerr spectra were measured at 85 K on a Mn £ \]\ Bi film (Di et al. 1992a, Di 1995).
of composition Mn ' #%# Bi. Other recent experiments on Al-doped MnBi samples with an
almost 2:1 Mn–Bi ratio confirm the trend of an increased Kerr rotation above 3 eV (Shang
et al. 1995, Shang 1996).
Another feature of the experimental Kerr rotation is that it exhibits a sign reversal at
0.9 eV. This sign reversal is actually also given by theory, but only for a smaller broad-
ening parameter. This is consistent with the observation that experimentally it disappears
in the room-temperature Kerr rotation (Di et al. 1992a). Lastly, we mention that there
appears to be a substantial intraband contribution to the conductivity present in the sam-
ple. In the calculations shown in Fig. 5.1 we accounted for the intraband conductivity by
adding a Drude-type conductivity to the calculated interband conductivity (Oppeneer et al.
1992a). For this Drude conductivity we used the calculated intraband plasma frequency
ü
Ù"ý ø 0.26 Ry and an estimated Drude broadening parameter ùJúÿþ¡ø 0.02 Ry. But, since
adding a Drude conductivity shifts especially the Kerr ellipticity below 3 eV upwards,
we would judge that in the sample there is likely a larger intraband contribution to the
conductivity. This can be due to some disorder and the Mn–Bi stoichiometry.
The Kerr spectra depend on the MO conductivity spectra in an entangled way (see,
e.g., Oppeneer and Antonov 1996). Therefore it is difficult to assign features in the Kerr
' Ã
spectra to particular band transitions. The absorptive parts of the optical conductivity, Áh*)Â *
#$Ã
and Á *)Â Ä , however, relate directly to the interband optical transitions, and provide therefore
more physical insight (Reim and Schoenes 1990). These absorptive parts have been dis-
cussed by several authors (Oppeneer et al. 1996a, Sabiryanov and Jaswal 1996), to whom
we refer for details. From the absorptive parts of the conductivity tensor it follows that
#$Ã
the main peak in the Kerr rotation of MnBi at 1.8 eV is due to a maximum in ÁQ*)Â Ä0ØHÙÚ near
2 eV, i.e., the main Kerr peak is a result of magneto-optically active interband transitions.
It has been found that mainly optical transitions from Bi bands are responsible for the
#$Ã
maximum in Á *)Â Ä ØHÙÚ (Oppeneer et al. 1996a).
In view of the fact that for Fe, Co, and Ni independent ab initio band-structure calcu-
lations of the MO Kerr spectra lead to different spectra (see, e.g., Oppeneer et al. 1992a,
Kim et al. 1994, Gasche et al. 1996, Mainkar et al. 1996), it is of interest to compare also
the existing calculated MO spectra of MnBi. The theoretical Kerr rotation as obtained by
Oppeneer et al. (1996a) and Köhler and Kübler (1996) are practically identical. The Kerr
rotation evaluated by Sabiryanov and Jaswal (1996), however, displays a different structure
particularly above 2.5 eV. At about 3 eV a large second peak of -1.2 is present, which
corresponds nicely with the experimental peak at 3.4 eV. It therefore appears that the ab
initio spectra of Sabiryanov and Jaswal (1996) agree better with experiment, whereas the
correspondence of the ab initio Kerr spectra of Oppeneer et al. (1996a) and K öhler and
Kübler (1996) support the absence of a second peak at 3.4 eV in the theoretical Kerr angle
spectrum of pure MnBi. The latter investigations tentatively attribute this second peak to
the non-stoichiometry of the MnBi sample, having a surplus of Mn atoms, or interstitial
oxygen atoms. In a first study on MBE grown, very pure MnBi samples, Harder et al.
(1998) find no peak at 3.4 eV. Further experimental studies on pure and well-characterized
MnBi samples seem thus required to provide a definite answer to the interrelation of the
microscopic structure and the Kerr angle at 3.4 eV. On account of the reasonable agreement
with the experimental Kerr spectra, it can nevertheless be concluded that first-principles
electronic structure calculations give a satisfactory description of the giant Kerr rotation in
MnBi.
ù
Ñ $^ 4 9=¹ a ( 0ü # #
9 9 a5 t# 9:9 # x#
» 2 ð
hegfheji
w t
Owing to the recent progress achieved in fully relativistic band-structure calculations it very
recently became possible to investigate the Kerr effect from first principles (Oppeneer et al.
1992a, Antonov et al. 1995, Guo and Ebert 1995). Such investigations carried out for TM
compounds like MnPt showed unambiguously how a giant Kerr effect microscopically
develops (Oppeneer et al. 1996b). As was outlined in the latter paper, large Kerr effects can
be anticipated in compounds which fulfill the following conditions: one of the constituting
elements must be heavy (i.e., have a strong SO coupling), but this element does not have to
be magnetic, for instance Pt or Bi. One of the other elements must have a large moment, but
this element does not have to be heavy, for instance Mn or Fe. There should be, in addition,
a good hybridization between the electronic states of the two kinds of atoms. Within such
a composition, the large exchange splitting (i.e., good magnetic moment) and the strong
SO coupling give through the hybridized bands rise to a big Kerr rotation (Oppeneer et
ð
al. 1996b). Although this model description is somewhat simplified, we find that it fully
ð
applies to MnBi. Manganese has the required large magnetic moment of 3.71 ! , while Bi
has a small induced moment of -0.10 ! . In an applied field the small moment on Bi shifts
ð
to Mn (with sign reversed). The total calculated moment compares thus reasonably well
with the moment of (3.84 ¼ 0.03) ! obtained from the saturation magnetization (Chen
and Stutius 1974). Thus, in the interplay of exchange splitting and SO coupling which
leads to the Kerr effect (Argyres 1955), Mn brings in the exchange splitting and Bi the SO
coupling. The degree of hybridization between Mn and Bi, furthermore, can be recognized
from the partial densities of states, which are shown in Fig. 5.2. There exists a strong
hybridization between the Bi and Mn -type states as well as Bi and Mn & -type states,
as can be seen from the identical shape of the partial densities (see also Coehoorn and
de Groot 1985). The magneto-optically active transitions take place mainly on Bi, from
occupied to unoccupied & states, in the minority spin states (dotted curves).
efhe
w
u
w ks|q:w kms t
To illustrate the above given explanation of the Kerr effect in MnBi it is instructive to
compare the Kerr effect in MnBi to that in the isoelectronic compounds MnAs and MnSb,
where the SO coupling strengths of As and Sb are much smaller than that of Bi. Other
interesting reference compounds are CrTe and CrBi, which will have a smaller magnetic
moment than MnBi. CrTe is actually also isoelectronic with MnBi, but with a different
number of valence electrons per atom. CrTe also crystallizes in the NiAs structure (Mayer
1958). In contrast to MnAs, MnSb, and CrTe, it must be mentioned that CrBi is not known
to exist in the NiAs structure, although (antiferromagnetic) CrAs and CrSb do exist in
ð ð ð ð
this structure. Our band-structure calculations predict for ferromagnetic CrTe and CrBi
moments that are not so small: 3.49 "! and -0.02 Y! for Cr, Te, and 3.40 Y! , -0.11 W!
ð ð ð ð
for Cr, Bi, respectively. The moments on Cr are even bigger than those calculated for
MnAs and MnSb, where we find 3.14 ! , -0.08 ! , and 3.34 ! , -0.07 ! , for MnAs and
MnSb, respectively. These results for MnAs and MnSb are in accord with earlier band
calculations by Coehoorn et al. (1985) and Motizuki (1987), while for CrTe our calculated
total moment compares to that obtained by Dijkstra et al. (1989). In Fig. 5.3 we show
the calculated interband-only MOKE spectra of these four compounds, for the moderate
spectral broadening of 0.03 Ry. As can be seen from Fig. 5.3, the shapes of the Kerr
rotation spectra of MnAs, MnSb, and CrBi are very similar, with a broad peak around 2 eV,
but that of CrTe is of a different shape. A most surprising result is the theoretical Kerr
rotation of hypothetical CrBi which has a maximal value of even -1.37 . The Kerr angles
of MnAs and MnSb are less than half as large as that calculated for MnBi. These spectra
can be understood within the simplified picture of SO coupling and magnetization: the SO
coupling of As and Sb is much smaller than that of Bi, leading to a smaller Kerr effect. In
the case of CrBi, the magnetic moment on Cr is a bit smaller than the corresponding one
on Mn in MnBi, thereby also resulting in a somewhat smaller Kerr effect in CrBi. In a
;=<?>A@CBDÛÇ+FRT+I

Partial densities of states for MnBi, in units of states/(atom eV). Majority
spin density of states are given by the solid curves, minority spin density of states by
the dotted curves.
way of speaking, MnBi exhibits its record Kerr angle because it fortuitously combines the
features of a maximal moment, SO coupling, and hybridization.
With respect to MnSb, we mention that very recently it has experimentally been found
that MnSb becomes a suitable material for MO data storage when it is doped with Pt
0.8
0.4
0.0
-0.4
-0.8
CrBi
-1.2 CrTe
MnAs
-1.6 MnSb
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>A@CBDÇ+FR[+I
Theoretical, interband only, MOKE spectra of MnAs, MnSb, CrTe, and
CrBi in the NiAs structure. The lifetime broadening used is ÔPÕ 0.03 Ry.
(Takahashi et al. 1994). Within the above given picture this would be understandable, as
Pt contributes through its SO coupling to the Kerr effect (see, e.g., Oppeneer et al. 1996b).
The Kerr spectra of the parent compound MnSb were measured also by Takahashi et al.
(1994) and by Sato et al. (1995), and Ikekame et al. (1996). The experimental Kerr
rotation is similar in shape to our theoretical result, but the amplitude is a bit smaller, only
0.3 to 0.5 at 2.5 eV. Doping MnSb with 6% Pt enormously enhances the Kerr rotation by
a factor of two as compared to the rotation of undoped MnSb (Takahashi et al. 1994). The
magnitude of the Kerr angle of Pt-doped MnSb is then of the same size as that of PtMnSb
C1 ´ Heusler alloy (van Engen et al. 1983a, Takanashi et al. 1988), but the advantages of the
Pt-doped MnSb over PtMnSb Heusler alloy are the lower content of Pt, the technologically
better suited position of the maximum rotation at about 2.5 eV, and a Curie temperature
UWV ø 573 K (Ido 1985), which is not too high for thermo-magnetic writing. It should
be mentioned, however, that other investigation of Pt-doped MnSb showed evidence of
a phase separation into MnSb and PtMnSb (Sato et al. 1995). The enhancement of the
Kerr rotation in Pt-doped MnSb could thus result as an average of spectral contributions of
MnSb and PtMnSb present in the sample (Sato et al. 1995).
Our first-principles calculation of the Kerr spectra of CrTe predict that this compound
could be interesting for UV laser-light MO recording: at 3.6 eV there is a maximum of
-0.76 in the Kerr angle (see Fig. 5.3). In addition to this, CrTe has its easy axis along the
ñ -axis, exhibits perpendicular magnetic anisotropy in evaporated films, while good quality
films are rapidly produced, and has a U V ø 340 K (Mayer 1958, Comstock and Lissberger
1970). This U V is rather small, but might just be appropriate for Curie point writing.
However, the magnetization is probably not sufficiently saturated at room temperature
(Comstock and Lissberger 1970). Experimental Kerr spectra of CrTe have not yet been
reported, only its Faraday rotation was previously measured (Comstock and Lissberger
1970). In view of these attractive features, an experimental investigation of CrTe would
certainly be worth while.
hegfhef 1l=o`|
w t
Although there are numerous experimental Kerr effect data of doped MnBi samples avail-
able, the influence of, e.g., Al or Si doping on the Kerr spectra is not yet understood (Wang
1990, Huang et al. 1994, Di et al. 1991, 1992b, 1994a, Shang et al. 1995, 1997, Sellmyer
et al. 1995, Rüdiger et al. 1995, 1996a,b, Fumagalli et al. 1996). Wang (1990), Huang et
al. (1994) and Xu et al. (1995) reported that they found an enlarged Kerr rotation as well
as an improved thermal stability in Al, Si doped MnBi films. It was proposed by Huang
et al. (1994) that Al atoms might occupy the interstitial positions in the NiAs-type unit
cell. Other groups found that doping with a small Al percentage (5 to 7%) does not affect
the MO, magnetic and structural properties of MnBi (Di et al. 1991, 1992a,b, 1994a,b,
Sellmyer et al. 1995, Rüdiger et al. 1996a). In some cases, however, it was observed that
Al promotes the formation of smaller grain sizes, which are more favorable for low-noise
ï ï :ï
MO recording (Di et al. 1991, 1994a, Sellmyer et al. 1995, Fumagalli 1996, R üdiger et al.
1996a). Other experiments were recently performed on a Mn # Bi/ Al sample where
ó
the Mn–Bi stoichiometry is nearly Mn # Bi (Shang et al. 1995, 1997, Shang 1996). An
enhancement of the Kerr rotation was for this composition attributed to the particular effect
of Al (Shang et al. 1995). The Kerr rotation measured through glass substrate reaches for
parts of the sample values up to -3.2 (Shang et al. 1995, Shang 1996). It was proposed
that these peak Kerr angles are related to a new MnBiAl crystal phase being formed (Shang
1996). Its crystal structure is not yet known.
Theoretical investigations of the effect of doping with Al and other elements were car-
ried out by Sabiryanov and Jaswal (1996), Oppeneer et al. (1996a), and K öhler and Kübler
(1996). Each of these three theoretical studies found, under the assumption that Al occupies
ïÊ
the interstitial positions in the NiAs unit cell, a reduction of the Kerr effect for compounds
with stoichiometry MnBiAl and MnBiAl / with respect to that of undoped MnBi. This
either indicates that the employed Al concentration is too high, or that that Al does not
occupy the interstitial positions. The latter option agrees with the experimental observation
ï ï 9ï
that Al is located on the boundaries of MnBi grains (Di et al. 1991, Di 1992, Sellmyer et
al. 1995, Rüdiger et al. 1996a). However, for samples with composition Mn # Bi/ Al/
ó
it is likely that Bi atoms are at least partially substituted by Mn and Al (Shang et al. 1995,
1997). Obviously, more experiments are needed to resolve unambiguously the function of
Al (and of other types of doping) in MnBi.
Ñ ùø _ 9:0 9
The temperature dependence of the Kerr effect in MnBi has drawn considerable attention.
In earlier, non-polar, measurements it was found that when the temperature was decreased
from room temperature to 85 K, the Kerr rotation increased from 0.9 to 9 ! (Chen and
Gondo 1964). In the more recent, polar Kerr measurements also an unusual temperature
dependence was found, but by no means of such an enormous size (Di 1992, Di et al.
1992a, Di and Uchiyama 1996). The observed temperature dependence needs nevertheless
to be explained quantitatively.
To explain the observed behavior, we investigated the influence of the variation of
two model parameters within our first-principles approach, which is normally valid for
zero temperature. The appropriate parameters that can be expected to play a role in
the temperature dependence of the Kerr effect are a temperature related reduction of the
magnetization and a reduction of the average lifetime. Experimentally, a reduction of
the saturation magnetization of 6% has been observed when the temperature is raised
from 85 to 300 K (Di et al. 1992a, Di and Uchiyama 1996). In our calculations we can
artificially reduce the magnetic moment, which we do by reducing the exchange splitting
(see Oppeneer and Antonov 1996). The magnetic moments remain in this procedure
parallel, but their magnitude is decreased. In reality, there are probably disordered local
magnetic moment fluctuations present, resulting in a smaller projection of the average
moment on the ñ -axis. In Fig. 5.4 we show the influence of reducing the exchange splitting
on the Kerr spectra calculated for a fixed lifetime broadening of 0.04 Ry. The experimental
reduction of the magnetization at 300 K corresponds approximately to an exchange splitting
reduction of 25%. This is close to that of 30% given by the dashed curve ( *0.7) in
Fig. 5.4. The small magnetization reduction does indeed strongly affect the main peak in
the Kerr rotation at 1.8 eV, in quantitative agreement with experiment (Di 1991, 1995, Di
and Uchiyama 1996). If we, in addition to the magnetization reduction, assume a reduction
of the average lifetime (i.e., a larger broadening ú ), then the Kerr rotation spectrum would
furthermore become smoother (see Fig. 5.1). Thus, the calculated rotation agrees well with
experiment. This explanation of course also applies to the Kerr rotation measured at 450 K
(Di 1995).
The only difference we find is the behavior of the second peak, which becomes more
reduced in experiment than in the calculations. It appears that the experimental reduction
actually supports the idea that the second peak is caused by the excess of Mn in the sample.
Namely, a reduction of the magnetic moment on the additional Mn atoms must lead to a
reduction of the corresponding Kerr peak,which happens at about 3.4eV. In our calculations
these additional Mn atoms are not present, so that there is consequently almost no reduction
of the theoretical Kerr rotation at this energy. With regard to the temperature dependence of
the MnBi Kerr spectra, we note that also for Ni film a remarkable temperature dependence
of the Kerr spectra was recently reported by Di and Uchiyama (1994). These authors
observed for Ni that temperature affects one peak in the Kerr rotation more than another.
Theoretically it could be shown that a reduction of the exchange splitting exactly reproduces
the observed behavior (Oppeneer and Antonov 1996). This result also demonstrates that
the temperature dependence of Kerr spectra can be explained quantitatively from first-
MnBi
1.0
exp. 85 K
exp. 300 K
exp. 450 K
0.5 theory
Mn M*0.7
Mn M*0.5
Kerr rotation (deg)
0.0
-0.5
-1.0
-1.5
0 1 2 3 4 5
Photon energy (eV)
;=<?>A@CBDdÇ+FECI
Ã
Temperature dependence of the experimental Kerr rotation of Mn £ \]\ Bi
(Di 1995, Di and Uchiyama 1996), and the calculated dependence of the MnBi Kerr

rotation on the Mn exchange splitting, calculated for a lifetime broadening ÔPÕ 0.04 Ry.
È *0.7 denotes an exchange splitting of 70% of the first-principles value.
principles calculations employing the temperature dependence of the magnetization.
ù
Ñ -Ñ . 1 ÿ ÿ
The experimentally observed record Kerr rotation of -1.6 (for the air–material interface)
of MnBi is straightforwardly explained by band-structure theory (Sabiryanov and Jaswal
1996, Oppeneer et al. 1996a, Köhler and Kübler 1996). These first-principles calculations
even predict a somewhat larger Kerr angle for pure MnBi than was measured (Di 1992,
1995, Di et al. 1992a,b, 1994a,b). The experimental peak at 3.4 eV in the Kerr angle
spectrum, however, comes out smaller in the ab initio calculations (Oppeneer et al. 1996a,
Köhler and Kübler 1996). The ab initio calculation of Sabiryanov and Jaswal (1996),
in contrast, does reproduce a peak of -1.2 at 3.4 eV; a discrepancy which is not yet
understood. As we discussed in Section 5.2, a tentative explanation of this feature can be
ïÊ
sought in the Mn–Bi stoichiometry. Also K öhler and Kübler (1996) found a small peak at
4 eV for the hypothetical compound MnBiMn , but attributed the measured peak at 3.4 eV
to oxygen interstitial atoms. This has also been done recently by Harder et al. (1998).
The record Kerr effect in MnBi, furthermore, can be understood within a simplified model,
based on the maximal 3& magnetic moment on Mn, the nearly maximal SO coupling of
Bi, as well as the strong hybridization between Mn and Bi states. Related compounds like
MnAs, MnSb and CrBi do not match up to MnBi in respect to the magnetic moment or SO
coupling, therefore these compounds exhibit a smaller Kerr effect. Yet, the Kerr spectra of
hypothetical CrBi and also CrTe are noteworthy: for CrBi a maximal Kerr angle of -1.37 is
predicted, whereas for CrTe about -0.76 is predicted. Especially CrTe appears to be suited
for modern MO storage applications, since its maximal Kerr rotation is in the UV energy
range. The role of Al atoms in MnBi has been studied computationally, assuming that the
Al atoms occupy the interstitial positions in the NiAs-type crystal structure (Sabiryanov
and Jaswal 1996, Oppeneer et al. 1996a, Köhler and Kübler 1996). However, the MOKE
spectra calculated for this kind of compound are reduced in magnitude (Sabiryanov and
Jaswal 1996, Oppeneer et al. 1996a, Köhler and Kübler 1996), and do therefore not
correspond to the experimentally observed spectra. Band-structure theory therefore does
not confirm the proposal that Al occupies the interstitial positions in the NiAs structure.
Further investigations of the structural and MO properties of Al-doped MnBi are required.
The unusual temperature dependence of the MnBi Kerr effect, lastly, can be explained
completely by a temperature induced reduction of both the magnetization and the average
lifetime of the excited states.
91
Ü ÝßÞáà7âäãæå
æ ç õ ! ê õ ö ë ì í0ê çìö
" ùûú ï 9 Zä :
The Heusler alloys NiMnSb, PdMnSb, and PtMnSb have been the subject of intensive
experimental and theoretical investigations since the early 1980’s. The interest in these
compounds arose first from the experimental discovery of an extremely large magneto-
optical Kerr rotation of -1.27 in PtMnSb at room temperature (van Engen et al. 1983a).
This value was for many years the record Kerr rotation observed in a transition-metal
compound at room temperature and therefore called “giant” Kerr effect (see, e.g., the re-
cent surveys of Buschow (1988), and Schoenes (1992)). Almost simultaneously with the
experimental discovery, the theoretical finding of the so-called “half-metallic” nature of
PtMnSb was reported by de Groot et al. (1983). Half-metallicity means that according to
(semirelativistic) band-structure theory the material is metallic for majority-spin electrons,
but insulating for minority-spin electrons (de Groot et al. 1983). Such a gap for one
spin type naturally may give rise to unusual magneto-transport and optical properties. The
isoelectronic Heusler alloy NiMnSb also was predicted to be half-metallic, whereas the
isoelectronic compound PdMnSb was predicted not to be half-metallic (de Groot et al.
1983). The Kerr rotations in both NiMnSb and PdMnSb on the other hand, were exper-
imentally found to be much smaller than that of PtMnSb, which resulted in a puzzling
combination of features. Experimental efforts were undertaken to verify the proposed
half-metallic character of NiMnSb and PtMnSb (Bona et al. 1985, Kisker et al. 1987,
Hanssen et al. 1990, Moodera and Mootoo 1994), which was subsequently established in
the case of NiMnSb (Hanssen et al. 1990, Moodera and Mootoo 1994). Very recently,
also experimental evidence in favor of half-metallicity in PtMnSb was reported by Bobo
et al. (1997).
On the theoretical side, several model explanations of the MOKE spectra of the com-
pounds were proposed (de Groot et al. 1984, Feil and Haas 1987, Wijngaard et al. 1989).
92 CHAPTER 6. PTMNSB AND RELATED COMPOUNDS
One of these was based on a possible loss of the half-metallic character due to SO coupling,
which was suggested to lead to a symmetry breaking between the different # states of
the (presumed) Sb bands in the vicinity of the Fermi energy $&% (de Groot et al. 1984).
Another explanation was based on differences of the semirelativistic effects in NiMnSb
and PtMnSb (Wijngaard et al. 1989), and another one on enhancement of the MOKE
spectra near the plasma resonance (Feil and Haas 1987). While the proposed models con-
tain interesting physical mechanisms in themselves, one of the remaining major stumble
blocks was to explain the measured differences in the MOKE spectra of the isoelectronic
NiMnSb, PdMnSb, and PtMnSb.
Owing only to the development of ab initio calculations of MOKE spectra a detailed
quantitative comparison between experiment and first-principles spectra became feasible
(Oppeneer et al. 1991, 1992a, Halilov and Feder 1993, Guo and Ebert 1994, 1995, Antonov
et al. 1995). The Heusler compounds are, of course, most attractive materials for ab initio
calculations of their MOKE spectra on account of the mentioned unusual features. Several
first-principles calculations for these compounds were reported very recently (Wang et al.
1994, Kulatov et al. 1994, Oppeneer et al. 1995a, Uspenskii et al. 1995, Oppeneer and
Antonov 1996, van Ek et al. 1997, van Ek and MacLaren 1997). The various calculated
MOKE spectra, however, spread rather widely. The origin of the differences in the spectra
obtained in the various calculations traces back, first, to the fact that the Kerr effect is a
tiny quantity in calculations, related to the difference in the reflection of left- and right-
hand circularly polarized light (see, e.g., Schoenes 1992). SO coupling in the presence
of spontaneous magnetization is responsible for the symmetry breaking in the reflection
of left- and right-hand circularly polarized light. Second, since the Kerr effect is only a
tiny quantity in first-principles calculations, numerical accuracy and the influence of ap-
proximations made in the evaluation gain an appreciable importance. For this reason, the
evaluation of the MOKE spectra of the ferromagnetic 3& transition metals Fe, Co, and Ni
have become benchmark test cases for MOKE calculation schemes (Oppeneer et al. 1991,
1992a, Gasche 1993, Kim et al. 1994, Guo and Ebert 1995, Antonov et al. 1995, Gasche
et al. 1996, Mainkar et al. 1996, MacLaren and Huang 1996). Numerical accuracy plays
normally not a decisive role if the physical mechanism is to be sought. However, it has
been shown by several groups (Oppeneer et al. 1992a, Guo and Ebert 1995) that for the ab
initio calculation of MOKE spectra an accurate evaluation of the dipole matrix elements
is essential for obtaining numerically reliable MOKE spectra. Moreover, in the particular
case of the Heusler alloys, the half-metallic band gap depends sensitively on technicalities
of the band structure calculation, as, e.g., atomic sphere radii (Oppeneer et al. 1995a).
In this Chapter we investigate in detail the MOKE spectra of NiMnSb, PdMnSb, and
PtMnSb. In addition to examining these well-known compounds, we investigate a group
of twelve other ternary compounds, viz. PtMnSn, PtCrSb, PtFeSb, Pt # MnSb, Co # HfSn,
NiMnAs, PdMnAs, PtMnAs, RuMnAs, PtMnBi, BiMnPt, and PtGdBi. These ternary com-
pounds are related to PtMnSb and can be derived from PtMnSb through selective chemical
substitution (with the exception of Co # HfSn). This study of the compounds derived by
selective substitution can be viewed as a sort of “materials engineering” on the computer.
The aim of investigating the whole group of compounds is twofold: First, it serves as an
examination of the MOKE in PtMnSb on those quantities altered by substitution, as, e.g.,
CHAPTER 6. PTMNSB AND RELATED COMPOUNDS 93
SO coupling strength, half-metallicity, and magnetic moment. In the second place, the
compounds are considered for their potential as MO storage material. In spite of the large
Kerr rotation of PtMnSb, this material has as a drawback for MO recording applications
the lack of a sufficient magnetocrystalline anisotropy which is required for a magnetic
orientation perpendicular to the solids surface (see, e.g., Shiomi et al. 1987, Kautzky and
Clemens 1995). In this respect we mention that in the last few years promising materials
for MO recording applications have been predicted on the basis of ab initio band-structure
calculations (Osterloh et al. 1994, Cebollada et al. 1994, Oppeneer et al. 1996b). The
technologically motivated search, especially for suitable MO recording materials in the
UV laser-light range, is still continuing. For several of the here investigated compounds
we obtain, as will be discussed, promising MOKE spectra.
" ù-, . /0 32 4576 ÿ8 9:
'hegheji ( *)
x{jt%w1|öl xA|Ys+w1q8s ,+ lQnpo`lQrw ru
The key material of interest is PtMnSb. In Fig. 6.1 we show schematically the group of
ternary intermetallics related to PtMnSb that are investigated in this Chapter. The rela-
tionship of all of the displayed compounds to PtMnSb might not be obvious at first sight.
The motivation is given in the following: The subgroup NiMnSb, PdMnSb, and PtMnsb
constitute isoelectronic compounds, which exhibit the already mentioned unusual features.
In PtCrSb and PtFeSb the magnetic element Mn has been substituted by Cr or Fe, its respec-
tive neighbors in the Periodic Table. Similarly, Sb has been substituted by Sn for PtMnSn.
For PtMnBi, Sb has been replaced by Bi, which, again, is an isoelectronic substitution.
The effect of the atomic positions in the crystal structure is studied by interchanging, e.g.,
Pt and Bi, which yields BiMnPt. Other compounds which are isoelectronic with PtMnSb
are given in the subgroup NiMnAs, PdMnAs, and PtMnAs. These compounds, interest-
ingly, are known not to crystallize in the MgAgAs crystal structure, but in the hexagonal
Fe # P structure (Villars and Calvert 1991). Also RuMnAs crystallizes in the Fe # P structure
(Chenevier et al. 1985). The inclusion of these compounds in the present study serves an
illustration of the trends in the subgroup of the ternary arsenic intermetallics. In Pt # MnSb
the Pt content is doubled with respect to PtMnSb. The crystal structure changes corre-
spondingly from the C1 ´ half-Heusler structure, which has one empty site, to the so-called
L2 ' crystal structure (cf. van Engen 1983). Co # HfSn crystallizes also in the L2 ' structure
and can be considered as a reference compound with regard to its measured MOKE spectra
(van Engen 1983, Buschow 1988). PtGdBi, finally, serves to illustrate the impact of an
enlargement of the magnetic moment due to replacement of Mn by Gd.
'heghe kmlQnpo1xAqx{tjlQw1q:q8uvo`|"Cx{u
To compute the MOKE spectra of the ternary compounds we used mainly the fully rel-
ativistic LMTO method (Andersen 1975, Nemoshkalenko et al. 1983), with variational
inclusion of spin polarization as proposed by Ebert (1988), and MEs after Antonov et al.
;=<?>A@CBDdÉ+FHGJI
The group of ternary intermetallic compounds which are investigated in
relationship to the key compound, PtMnSb. The arrows indicate how the compounds
are derived from PtMnSb through chemical substitution.
(1993). The MOKE spectra calculated with this band-structure scheme have been tested
against MOKE spectra obtained with the relativistic ASW scheme (Oppeneer et al. 1995a).
For the LSDA exchange-correlation potential we have used the parameterization proposed
by von Barth and Hedin (1972). We further mention that for evaluating the optical spectra
we used (unless stated otherwise) a constant interband relaxation parameter of 0.03 Ry.
The interband relaxation parameter is always non-zero, because excited states have a finite
lifetime. The intraband, or Drude, relaxation time parameter - þ may be different from
(Y'
the interband relaxation parameter. For very pure materials - þ will approach zero. We
(Y'
adopted therefore here the perfect crystal approximation, i.e., - þ

(W'
. ò . The lattice pa-
rameters which have been used in the calculations are the experimental ones (Villars and
Calvert 1991). In the few cases where the lattice parameter was not known, estimated
values have been used.
'heghef âäx/
w â0 1
w `q:w32t4ºw1
One of the most surprising results reported for PtMnSb and NiMnSb, is that band-structure
calculations proposed these compounds to be half-metallic, but the isoelectronic compound
PdMnSb was predicted not to be half-metallic (de Groot et al. 1983). The prediction of
half-metallicity was based on semirelativistic band-structure calculations, i.e., the effect
of SO coupling was neglected (de Groot et al. 1983). This approximation is often ac-
ceptable for the study of certain material properties, but especially for the investigation of
MO properties it is not, because the Kerr effect is a purely relativistic effect. Recent ab
initio relativistic calculations showed that the MO Kerr effect scales linearly with the SO
coupling strength for monoatomic crystals (Oppeneer et al. 1992b, Ebert et al. 1996). The
inclusion of SO interaction in the energy-band calculation could in principle destroy the
half-metallic behavior, as it was proposed in one model explanation of the MOKE spectra
of PtMnSb (de Groot et al. 1984, Buschow 1988). In this model it is assumed that SO
coupling lifts the degeneracy of the hybridized triplet Sb band at the 5 point in such a
way that the #ºø76½ level is raised above $% , but the #ºøpò and #ºø98`½ levels fall
below $:% (de Groot et al. 1984). If such a situation would actually happen, it would lead
to an imbalance of MO transitions stemming from #ºø;6½ and #7ø98`½ levels, which
could be the reason for the large Kerr effect in PtMnSb (de Groot et al. 1984). However,
in recent relativistic band-structure studies of PtMnSb it was found that the half-metallic
property nearly completely prevails (Ebert and Sch ütz 1991, Oppeneer et al. 1995a, Youn
and Min 1995), and that therefore the supposed loss of half-metallicity due to SO coupling
does not occur. Also, it has been pointed out that the triplet minority spin band at the 5
point just below $ % is actually composed predominantly of Mn 3 & states (Youn and Min
1995, van Ek and MacLaren 1997). To illustrate the half-metallic nature of PtMnSb as ob-
tained from relativistic band-structure calculations, we show in Fig. 6.2 the spin-projected,
relativistic partial densities of states of PtMnSb. As can be recognized from Fig. 6.2, the
partial densities of states for minority spin have evidently a gap at the Fermi level. We
found a similar behavior for NiMnSb, but of course not for PdMnSb. For all three Heusler
compounds we show, in Fig. 6.3, the calculated relativistic energy bands and total densities
of state (DOS). In the case of PdMnSb, three spin-orbit split energy bands are just above the
Fermi level at the 5 point, therefore half-metallic behavior is not supported for PdMnSb by
band-structure theory. In the case of NiMnSb and PtMnSb, these important bands are just
below $ % , rendering the half-metallicity in these compounds. Our band-structure results
are in agreement with recent experiments on NiMnSb and PtMnSb, in which half-metallic
behavior to a degree of nearly 100% was observed (Hanssen et al. 1990, Moodera and
Mootoo 1994, Bobo et al. 1997).
After having verified the half-metallic band-structure property we turn to the MOKE
spectra. In Fig. 6.4 we show the calculated and experimental (van Engen et al. 1983a)
MOKE spectra of the three isoelectronic Heusler compounds. There exists apparently a
rather good agreement between the experimental Kerr spectra and the ab initio calculated
ones. Overall, the experimental features are reasonably well reproduced, except for the
magnitude of the Kerr rotation of PdMnSb, for which theory predicts larger values than
are experimentally observed. The first and important conclusion, which we draw from
;Q<R>A@CBDöÉ+FRT+I
Spin-projected, partial densities of state calculated for PtMnSb. Majority-spin
densities are given by the full curves, minority-spin densities by the dotted curves. The half-metallic
behavior can be seen from the band gap at the Fermi level, which is present for minority spin, but
not for majority spin.
the correspondence between experimental and calculated Kerr spectra, is: the anomalous
behavior of the MOKE spectra in these compounds is well described by normal band-
structure theory.
To investigate the origin of the Kerr spectra, we consider the separate contribu-
tions of both the numerator of Eq. (2.33), i.e., ÁC*)Ä ØHÙÚ and the denominator, < ØHÙÚ>=
'HG$#
Á{*)* Ø ½?6A@CF BED ÁA*)* Ú . In Fig. 6.5 we show how the separate contributions of numerator and
denominator bring about the Kerr angle of NiMnSb. The imaginary part of the inverse de-
nominator (times the photon frequency), Im I Ù <:J , displays a typical resonance structure
(W'
#$Ã
at about 1 eV. The imaginary part of Ù Á+*)Ä , i.e., Ù Á *)Â Ä , displays a double peak structure.
;=<?>A@CBDÉ+FR[+I
Spin-polarized, relativistic energy-band structures and total densities of
state (DOS’s) of NiMnSb, PdMnSb, and PtMnSb.
The double peak structure of the Kerr rotation results roughly as the product of Im I Ù <J
(W'
#%Ã
and Ù Á *)Â Ä : The first peak in the Kerr rotation at 1.5 eV is predominantly caused by a
θK εK
0.2
NiMnSb
0.0
-0.2
0.2 PdMnSb
0.0
-0.2
-0.4
0.5 PtMnSb
0.0
-0.5
-1.0 theory
exp.
-1.5
0 1 2 3 4 0 1 2 3 4 5
Photon energy (eV)
;=<?>A@CBDmÉ+FECI
Calculated and experimental Kerr rotation KÿL and Kerr ellipticity L KÏ
spectra of the Heusler compounds NiMnSb, PdMnSb, and PtMnSb. The experimental
data are those of van Engen et al. (1983a).
minimum of the denominator, whereas the second peak in the Kerr rotation at 4 eV is due
#$Ã
to a maximum in the off-diagonal conductivity, Ù Á *)Â Ä . The nature of the peak in Im I Ù <:J
(W'
can be understand from the top panel in Fig. 6.5, where the complex diagonal dielectric
NiMnSb
40
(2)
εxx
20
(2)
εxx , εxx
0
(1)
(1)
εxx
-20
(2)
ωσxy 6
4
2
0
0
-1
-2
Im[ωD]
-4
-6
-8
Kerr rotation (deg)
theory
exp.
0.0
-0.2
0 1 2 3 4 5
Energy (eV)
;=<?>A@CBD É{FRÇ+I
Decomposition of the Kerr rotation spectrum of NiMnSb in separate
contributions. £Ì Top panel: calculated real and imaginary part of the diagonal dielectric
\ Ì
L L
function, £ ©jË © and ©%Ë © , respectively.
£Ì ThirdÌ panel from the top: The imaginary part
of M N?O&PRQ which results from ©%Ë © and£
\
©%Ë © . Bottom
\ Ì
L L
panel: The Kerr rotation which
results as a combination of Im M NTSUP Q
and Ê ©jË Í (second panel from the top). The N
experimental Kerr angle spectrum is after van Engen et al. (1983a).
î 'jÃ
function is shown: its real part, *)Â * , becomes small at about 1 eV, and its imaginary part,
î #$Ã
*)Â * , has a shallow minimum at 1 eV. The second peak in the Kerr rotation stems from
#$Ã
the maximum in Ù Á *)Â Ä . The latter, in turn, is known to be due to the interplay of SO
PdMnSb
40
(2)
εxx
20
(2)
εxx , εxx
0
(1)
(1)
εxx
-20
3
(2)
ωσxy
0
-3
0
-1
Im[ωD]
-2
-4
Kerr rotation (deg)
0.0
-0.2
theory
-0.4 exp.
0 1 2 3 4 5
Energy (eV)
;=<?>J@CBDÉ{FRÉ+I
As Fig. 6.5, but for PdMnSb.
coupling and spin polarization (Argyres 1955, Oppeneer et al. 1992b). Thus, the two
similar looking peaks in the Kerr rotation arise in fact from quite a different origin.
Next, we consider the spectra for the compound PdMnSb in more detail, which are
#%Ã
shown in Fig. 6.6. In this compound Ù Á:*)Â Ä is larger than that of NiMnSb in the energy
range 1 – 4 eV. This is simply due to the larger SO interaction on Pd as compared to that on
Ni. The inverse denominator Im I Ù <J , however, does not exhibit such a strong resonance
(W'
'jÃ #$Ã
as it does for NiMnSb. The latter is related to the particular shape of the *)Â * and *)Â *
#%Ã
î î
spectra. The Kerr rotation in effect displays the same shape as Ù Á *)Â Ä , being enhanced at
PtMnSb
40
(2)
20 εxx
(2)
εxx , εxx
0
(1)
(1)
-20 εxx
5
(2)
ωσxy
-5
0
-1
Im[ωD]
-5
-10
1
Kerr rotation (deg)
-1 theory δ=0.05
theory δ=0.02
o,∆ exp.
-2
0 1 2 3 4 5
Energy (eV)
;=<?>A@CBD É+F?ÓAI
As Fig. 6.5, but for PtMnSb. The experimental data are after Ikekame
et al. (1993). The data for annealed PtMnSb are denoted by , and those for non- V
W
annealed, polished PtMnSb by . The calculated Kerr rotation spectrum is shown for
two interband lifetime parameters, 0.05 Ry and 0.02 Ry.
1 – 2 eV by the contribution from the denominator.

In Fig. 6.7 we show the spectral quantities for PtMnSb. The inverse denominator
Im I Ù <J
(W'
again displays for PtMnSb a strong resonance at 1 eV, which is even larger
#$Ã
than that for NiMnSb. In addition, the off-diagonal conductivity Ù Á *)Â Ä is for PtMnSb
again larger than that of PdMnSb, in accordance with the larger SO coupling on Pt. The
resulting Kerr rotation has a “giant” peak of -1.2 up to -2.0 depending on the applied
lifetime parameter (see Fig. 6.7). These values are in good agreement with the available
experimental Kerr peak values for PtMnSb, which range from about -1 to -2 depending
on sample preparation and surface quality (see, e.g., van Engen et al. 1983a, Inukai et al.
1986, Shiomi et al. 1987, Naoe et al. 1988, Takanashi et al. 1988, 1990, 1991, Sato et al.
1992b, Ikekame et al. 1993, Kautzky and Clemens 1995). Careful investigations of the
consequences of sample preparation have been performed by Takanashi et al. (1988) and
Sato and co-workers (Sato et al. 1992b, Ikekame et al. 1993). These investigations showed
that annealing of the PtMnSb sample raises the Kerr angle to a maximum value of -2 ,
#$Ã
whereas the non-annealed Kerr angle is only -1.2 . The off-diagonal conductivity Ù Á *)Â Ä
was found to be rather insensitive to annealing (see Fig. 6.7) (Sato et al. 1992b, Ikekame
et al. 1993). The main impact of annealing thus evidently occurs in the denominator. The
reason for the calculated resonance in the inverse denominator lies again in the frequency
î
dependence of the diagonal dielectric function, which is shown in the top panel of Fig. 6.7.
The calculated *)* compares reasonably well with the experimental one of Ikekame et al.
î 'jÃ
(1993), except for the important first root frequency of P*)Â * which is shifted by about 0.5 eV.
This difference leads to a shifted position in the resonance peak of Im I Ù <:J , which in
(Y'
turn results in a shifted main Kerr rotation peak of just the same amount. The position
of the maximum in Im I Ù <J
(W'
thus dominantly determines the position of the main Kerr
#$Ã
rotation peak. We mention, in addition, that the second maximum in the calculated Ù Á *)Â Ä
at 4.4 eV (see Fig. 6.7) is also present in the experimental spectrum, but at a higher energy
of 5.2 eV (Ikekame et al. 1993).
The origin of the giant Kerr angle in PtMnSb as compared to the Kerr angles in NiMnSb
and PdMnSb can completely be understood from our calculations. In these three com-
pounds, first, the off-diagonal conductivities Á *)Ä Ø«ÙÚ are quite different, which is a direct
result of the different relativistic electronic structure. Although both NiMnSb and PtMnSb
#$Ã
are both half-metallic, their Ù Á:*)Â Ä spectra are distinctly different, while on the other hand
#$Ã
the Ù Á *)Â Ä of PdMnSb and PtMnsb have a similar structure, but not a similar magnitude
(see Figs. 6.5, 6.6, and 6.7). In the second place, there is the influence of the denominators,
as exemplified in Im I Ù <J . These are similar in shape and magnitude for NiMnSb and
(W'
PtMnSb, but the magnitude of Im I Ù <J
(Y'
in PdMnSb is about a factor of 2 smaller. We
find that this difference relates to the half-metallic nature of both NiMnSb and PtMnSb,
which is not present for PdMnSb. From Fig. 6.3 it can be seen that for the half-metallic
compounds there are three lesser bands at $% . One consequence is therefore that the intra-
band contribution to Á+*)* will be smaller (see Eq. (2.85)). In Fig. 6.8 we show the impact
of the half-metallic character of the band structure on the Kerr rotation of PtMnSb. The
ü
calculated intraband plasma frequency in PtMnSb is small, ù Ù ý = 4.45 eV. Experimentally
a somewhat larger plasma frequency of (6.1 ¼ 0.4) eV was found for PtMnSb, and a smaller
ù Ù"ý = (4.9 ¼ 0.2) eV for NiMnSb (van der Heide et al. 1985). One should, however, not
forget that the sample purity can affect the plasma frequency through the position of $ % .
To investigate the influence of the half-metallicity, we can artificially shift the Fermi energy
down, and calculate the spectra for this position of $ % , or we can leave $ % as it is, and
ü
model a non-half-metallic band structure by adopting a larger Ù:ý . Both ways to mimic
PtMnSb
(1)
εxx
-10
-20
5
(2)
ωσxy
-5
0
-1
Im[ωD]
-5 ~
ωp= 4.45 eV
~
ω = 6.40 eV
p
EF shifted
-10
Kerr rotation (deg)
0.5
0.0
-0.5
-1.0
-1.5
0 1 2 3 4 5
Energy (eV)
;=<?>A@CBD É{FRÖ+I
Model investigation of the influence of the half-metallic character of
PtMnSb on the optical and MO spectra. A non-half-metallic band structure has been
modeled in two ways: by artificially shifting XY
in PtMnSb down, and by increasing
the calculated intraband plasma frequency of 4.45 eV to 6.40 eV. NZ [
non-half-metallicity have a drastic impact on the resulting Kerr rotation, which becomes
î
reduced by a factor 2. As can be seen from the top panel in Fig. 6.8, in the absence of
'jÃ
half-metallicity the shape of *)Â * changes, and closely resembles that of PdMnSb (see
Fig. 6.6). This is especially so for the model where $ % is shifted, since this leads to the
#$Ã
smaller Im I Ù <J
(Y'
and also to a reduction of Ù Á *)Â Ä at photon energies below 2 eV. The
latter is due to the exclusion of optical transitions from the SO-split bands just below $% .
The consequence of both models for non-half-metallic behavior is that the maximum in
Im I Ù <J
(W'
becomes about two times smaller. The Kerr angles derived in these models
#$Ã
resemble now that of PdMnSb in shape, but are larger, because Ù Á *)Â Ä is larger than that of
PdMnSb. Previously we have shown that if the SO coupling on Pt in PtMnSb is artificially
set to zero, the Kerr rotation peak in PtMnSb becomes reduced by a factor of 3 (Oppeneer
et al. 1995a). We mention with respect to the influence of the denominator on the Kerr
rotation in PtMnSb, that experiments in which the stoichiometry and crystalline sample
quality were varied also concluded that the denominator contributed appreciably to the
giant Kerr rotation (Takanashi et al. 1988, 1990, 1991, Sato et al. 1992b, Ikekame et al.
1993).
In conclusion, we find that the Kerr spectra of NiMnSb, PdMnSb, and PtMnSb can
be fully understood from their electronic structure. The puzzling anomalies in the Kerr
spectra of these compounds arise from an interplay of compound related differences in
the SO interaction, in the half-metallic character, and also in relative positions of energy
bands.
Previously, several model explanations for the giant Kerr effect in PtMnSb were pro-
posed (de Groot et al. 1984, Feil and Haas 1987, Wijngaard et al. 1989). We already
discussed the model which assumed the destruction of half-metallicity caused by SO cou-
pling (de Groot et al. 1984). Another model, by Wijngaard et al. (1989) suggested that
scalar-relativistic effects and not the SO interaction are responsible for the differences in
the Kerr spectra of the three Heusler compounds. However, as we already emphasized, the
î
SO interaction determines largely the Kerr spectra in these intermetallic compounds. The
'jÃ
explanation proposed by Feil and Haas (1987) is based on the assumption that at *)Â * = 1 a
(W'
resonance in the inverse denominator D occurs, and that Á{*)Ä plays no role. We do find
î
that an important resonance in the denominator occurs, however, not necessarily precisely
'jÃ
at those energies where ÿ*)Â * = 1 (see also Schoenes and Reim 1988). For example, in the
î 'jÃ
case of NiMnSb and PtMnSb there are two energies where *)Â * = 1. In addition, as we have
shown above, Á+*)Ä does as much play a role in the giant Kerr rotation as the denominator
does. Attention to the enhancing effect that a small denominator has on the Kerr rota-
tion was drawn previously for TmS and TmSe (Reim et al. 1984), and very recently for
CeSb, for which compound a record Kerr rotation of 90 was detected (Pittini et al. 1996,
Schoenes and Pittini 1996).
Recently, three other ab initio calculations of the Kerr spectra of these Heusler com-
pounds were reported (Wang et al. 1994, Kulatov et al. 1994, Uspenskii et al. 1995,
van Ek and MacLaren 1997). In the study of Wang et al. (1994) the correct experimental
magnitude of the Kerr rotation in NiMnSb was obtained, but the large Kerr rotation in
PtMnSb could not be reproduced. Conversely, in another study by Kulatov and co-workers
a large Kerr rotation was obtained for PtMnSb; however, for NiMnSb a theoretical Kerr
rotation three times larger than the experimental one was calculated (Kulatov et al. 1994,
Uspenskii et al. 1995). Also, in a recent study van Ek and MacLaren (1997) obtain a Kerr
rotation of experimental magnitude for NiMnSb, but a rather small magnitude of -0.7 to
-0.9 for PtMnSb, even though a rather small relaxation time parameter of 0.2 eV was used.
Undoubtedly, this unusually wide spread in the ab initio calculated Kerr spectra reflects the
major influence of numerical accuracy in the calculations. With regard to the wide spread
in the theoretical Kerr rotations we mention that we very recently predicted for CrPt and
MnPt large Kerr rotations with peak (interband) values of -0.9 and -1.5 , respectively
(Oppeneer et al. 1996b), in reasonable agreement with existing experiments for MnPt
(Kato et al. 1995b, Wierman and Kirby 1996). Yet, another first principles calculation of
Kulatov et al. (1996) gave much larger Kerr rotations of even -2.1 and -2.5 for CrPt
and MnPt . Notably, apart from the magnitude of the MOKE peaks, the later calculation
also predicts a different trend with respect to the 3& element. However, the trend and
shape of the Kerr spectra we predicted for CrPt and MnPt was very recently confirmed
experimentally by Vergöhl and Schoenes (1996). Particular for the Heusler compounds,
computational accuracy is imperative, in view of the conclusions concerning the physical
mechanism that are to be derived. As documented by our previous experience in computing
MOKE spectra, we believe that we can say that we have sufficiently taken care to obtain
trustworthy first-principles spectra. (cf. Oppeneer et al. 1992a, 1995a, 1996a, Antonov et
al. 1995).
'heghe¶á âäx/
w hw
One of the compounds that, from the beginning of the discovery of the anomalous properties
of MnPtSb has been considered in relationship to PtMnSb, is PtMnSn (van Engen et al.
1983a, de Groot et al. 1983, de Groot and Buschow 1986, Otto et al. 1987). Although Sb
and Sn are neighbors in the Periodic Table, the MO Kerr rotation in PtMnSn was measured
by van Engen et al. (1983a) to be much smaller than that of PtMnSb. Scalar-relativistic
band-structure calculations predicted that PtMnSn is not half-metallic (de Groot et al.
1983, de Groot and Buschow 1986). We have performed fully relativistic band-structure
calculations which lead to the same conclusion; see Fig. 6.9. If one were to fill the energy
]\
;=<?>A@CBDdÉ+F +I
Spin-polarized, relativistic energy-band structure and DOS of PtMnSn.
PtMnSn
0.6
0.4
0.2
0.0
-0.2
-0.4
exp.
-0.6 theory
PtMnSb EF shifted
-0.8
0 1 2 3 4 0 1 2 3 4 5
Photon energy (eV)
;=<?>A@CBD E^
É+FHG ¥I
Calculated theoretical and experimental (van Engen et al. 1983a)
L
Kerr angle ( K/L ) and Kerr ellipticity ( %L ) of PtMnSn. For comparison the Kerr spectra
that result for PtMnSb when the Fermi level is artificially shifted down are also shown
(dashed curves).
bands of PtMnSn with one more electron, a half-metallic band structure comparable to that
of PtMnSb (see Fig. 6.3) would result. For this reason, we expect the MOKE spectra of
PtMnSn to be similar to those calculated for PtMnSb with a down-shifted $1% . In Fig. 6.10
we show the theoretical and experimental Kerr spectra for PtMnSn. The calculated Kerr
rotation and ellipticity curves of PtMnSn resemble the shape of the experimental curves,
but the theoretical size of the spectra is much larger. At present we do not know what the
reason for this unusual discrepancy between LSDA energy band theory and experiment is.
One possible explanation, which is as yet purely speculative, is that it could be that the
PtMnSn sample had a poor crystalline quality. A poor crystalline ordering of a sample was
previously shown to reduce the magnitude of the Kerr effect (cf. Cebollada et al. 1994,
Weller 1996). Another possibility is that the PtMnSn sample was poorly magnetized.
This is most likely the case, since the Curie temperature of PtMnSn is only just above
room temperature (U V = 330 K), and the magnetic field used by van Engen et al. (1983a)
was only 1.2 T. As a consequence, van Engen et al. (1983a) estimated that the achieved
magnetization is about half of the low-temperature value. In Fig. 6.9 we also show the
effect of a down-shifted Fermi level on the MOKE spectra of PtMnSb. As anticipated,
these spectra closely correspond to the ab initio calculated spectra of PtMnSn. We can
therefore safely conclude that the main reason for the differences in the Kerr effect in these
two compounds is the critical position of the Fermi level.
'heghe_ âäxCkms`q:w âäx/a|?
Cr and Fe are the respective neighbors of Mn in the Periodic Table. Both the ternary
alloys PtCrSb and PtFeSb were studied previously in connection to PtMnSb (Buschow
et al. 1983b). It came as a surprise when it was found that PtCrSb and PtFeSb do not
crystallize in the half-Heusler crystal structure, but in a closely related structure, in which
the Sb and Cr (or Fe) atoms have shifted somewhat from the ideal positions (Buschow et
ð
al. 1983b). Moreover, it was found that in PtFeSb, Fe carries a large magnetic moment
of 2.5 Y! , but PtCrSb was found to be non-magnetic (Buschow et al. 1983b). In addition,
the Kerr spectra of PtFeSb were measured to be totally different from those of PtMnSb:
The maximal Kerr rotation in PtFeSb was about 20 times smaller than that of PtMnSb!
With the aim of understanding the microscopic reason for this unusual materials depen-
dence, we have performed calculations for PtFeSb and PtCrSb in their distorted Heusler
structure, but for comparison we also calculated PtFeSb assuming an ideal half-Heusler
structure. In Fig. 6.11 we show the relativistic band structure of PtFeSb assuming the ideal
C1 ´ structure. In this structure, there is for PtFeSb a quasigap at about 0.4 eV below $ % . A
closer inspection of the spin-projected DOS has revealed that at this energy the band struc-
ture is half-metallic. Thus, PtFeSb can in a simplified picture be considered as an analog
of PtMnSb, in which the bands are filled with one more electron. In Fig. 6.12 we depict the
experimental and theoretical MOKE spectra of PtFeSb. The theoretical spectra are shown
for three cases: the distorted half-Heusler crystal structure, the ideal C1 ´ half-Heusler
structure, and the ideal C1 ´ structure, but with $% shifted such that the material would
be half-metallic. We find that for the half-Heusler structure LSDA band-structure theory
;=<?>A@CBDÉ+FHGJGvI
Spin-polarized, relativistic energy-band structure and total DOS of
PtFeSb in the C1 half-Heusler crystal structure.
®
PtFeSb
0.2
0.0
-0.2
-0.4
exp.
theory
theory (Heusler)
-0.6
theory (Heusler) EF shifted
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>A@CBDmÉ+FHG T{I
Theoretical and experimental (Buschow et al. 1983b) results for the
polar Kerr spectra of PtFeSb. The theoretical Kerr spectra are given for two crystal
structures: the distorted C1 crystal structure, which is the experimentally observed
®
structure (solid curves), and the ideal C1 half-Heusler structure (dashed curves). For
®
the latter structure also the influence of a shift of the Fermi level into the spin-minority
band gap, making the band structure half-metallic, is shown.
predicts for PtFeSb a reasonable maximum Kerr rotation of about -0.4 . If the Fermi level
is shifted, however, so that the material artificially becomes half-metallic, the Kerr angle
peak is predicted to be enhanced to nearly -0.7 . The mechanism for this enhancement
is, of course, just the same as what we explained for PtMnSb. Noticeably, we recognize
from Fig. 6.12 that the crystal structure has an enormous impact on the Kerr spectra: the
calculated Kerr spectra for the true crystal structure are much smaller than obtained for
the C1 ´ structure. The ab initio calculated Kerr rotation and Kerr ellipticity are in shape
and magnitude in good agreement with the experimental data of Buschow et al. (1983b).
Thus, our first-principles calculations fully support the previously as unusual considered
differences between the MOKE spectra of PtFeSb and PtMnSb. The reason for the enor-
mous difference between the calculated Kerr spectra of PtFeSb in the two structures is the
difference in the corresponding electronic structures. We do not show here the electronic
structure of PtFeSb in the distorted half-Heusler structure (space group No. 198). We refer
to Antonov et al. (1997) for the energy-band structure and for a discussion of the resulting
differences in the absorptive components of ÁC*)* and Á{*)Ä . In the distorted Heusler structure
all allowed interband contributions of all bands in PtFeSb cooperate in such a way that
#$Ã 'jÃ
these nearly cancel for Á8*)Â ÄÛØ«ÙÚ . The diagonal conductivity Á:*)Â *0ØHÙÚ , on the other hand,
is additive in the allowed interband optical transitions, and is much larger in the distorted
phase than it is in the ideal C1 ´ phase. In effect, the combination of a small Á *)Ä and a
relatively large Á *)* brings the small Kerr effect in PtFeSb about.
The related ternary compound PtCrSb also crystallizes in the distorted half-Heusler
structure. This compound does not order magnetically (Buschow et al. 1983b). To study
the influence of the Cr substitution on the MOKE spectra, ab initio calculations were per-
formed for PtCrSb using constraint LSDA calculations to obtain a ferromagnetic solution.
For the resulting MOKE spectra of PtMnSb we refer to Antonov et al. (1997). It turns
'jÃ
out that the diagonal conductivity Á:*)Â * is almost the same for PtFeSb and PtCrSb, but in
#%Ã
PtCrSb the contributions to Á=*)Â Ä do not cancel as much as in PtFeSb (Antonov et al. 1997).
In comparison to PtMnSb, the substitution of Fe or Cr for Mn is thus not favorable for the
MO properties. We find that this is primarily an effect of the concomitant lattice structure
changes.
'heghe_' âäx/bc 1
w q:w kml*b,de)fhw
The C1 ´ half-Heusler structure contains one empty lattice site (see, e.g., van Engen 1983).
If this lattice site in the formula composition g0hji is filled with one g atom, the L2 '
structure for the g # hji composition arises. The compound Pt # MnSb is therefore in this
sense the L2 ' extension of PtMnSb. It would be interesting to examine the MO properties
of Pt # MnSb in comparison to those of PtMnSb, yet to our knowledge it appears that
this has not been done experimentally. In several previous theoretical studies it has been
demonstrated that when the strength of the SO interaction is increased in a ferromagnetic
material, the Kerr effect will in general become larger (Oppeneer et al. 1992b, 1996a, Ebert
et al. 1996). One possible means of increasing SO coupling is of course by substitution
of, or alloying with, heavy elements, as, e.g., Pt or Bi. The Kerr effect in FePt will
consequently be larger than that in FePd, which is in turn larger than that in Fe (Osterloh
et al. 1994). One might on these arguments expect a larger Kerr effect in Pt # MnSb than in
PtMnSb, if one of course would forget for a moment that PtMnSb does not at all exhibit
a standard behavior. In addition to the SO coupling, also the local symmetry of Pt atoms
around one Mn atom is different: For the L2 ' crystal structure the local symmetry around
the Mn site is cubic, whereas for the C1 ´ crystal structure it is trigonal.
In Fig. 6.13 we show the calculated Kerr spectra of Pt # MnSb. For comparison, we
also give calculated spectra for Co # HfSn, because for this compound experimental spectra
were measured (van Engen et al. 1983b, van Engen 1983). In spite of the SO coupling
which is added through the extra Pt atom, the Kerr effect in Pt # MnSb is calculated to be
much smaller than that in PtMnSb. From the comparison between the experimental and
calculated Kerr spectra of Co # HfSn, we conclude that the theoretical MOKE description
of this compound is quite good, which we would thus anticipate to hold also for Pt # MnSb.
The reason that the Kerr effect in Pt # MnSb is much smaller than that in PtMnSb again
follows from the band structures. The band structure of Pt # MnSb, which we do not show
here, is quite different from that of PtMnSb, and, moreover, not half-metallic. The Kerr
effect in Pt # MnSb is also small compared to that of MnPt , where a maximum interband
0.4
Complex polar Kerr effect (deg) Co2HfSn exp.
Co2HfSn theory
θK εK
0.3
Pt2MnSb theory
0.2
0.1
0.0
-0.1
-0.2
-0.3
0 1 2 3 4 0 1 2 3 4 5
Photon energy (eV)
;=<?>A@CBDZÉ{FHG [+I
Kerr rotation K L and Kerr ellipticity L of the Heusler alloys Pt \ MnSb Ï
and Co \ HfSn. The experimental Kerr spectra of Co \ HfSn are after van Engen et al.
(1983b).
Kerr angle of -1.5 was calculated (Oppeneer et al. 1995a, 1996b). In Pt # MnSb the
interband transitions contributing to Á *)Ä cancel to a large extent each other, as a particular
consequence of the electronic structure.
'ehelk âäx/ mt Ht

w w@â n
x rq:w â xop t
The basic idea behind the investigation of PtMnBi and BiMnPt is similar to that of Pt # MnSb,
namely, to try to enlarge the SO coupling strength within the hybridized bands by sub-
stitution of a heavier element. An isoelectronic substitution in this manner is PtMnBi.
Previously it was shown in ab initio calculations, that MnBi has a larger Kerr rotation than
MnSb, because of the larger SO coupling of Bi (Oppeneer et al. 1996a). In PtMnBi the
Mn atom has Pt nearest neighbors. There is thus the largest overlap between Pt and Mn
orbitals. In MnBi, which is experimentally known to exhibit a giant Kerr angle up to -2
(Di et al. 1992b, Di and Uchiyama 1996), the Mn & orbitals overlap with Bi orbitals.
Therefore it is also of interest to consider BiMnPt, i.e., Bi and Pt atoms interchanged.
We note, however, that the Kerr spectra of PtMnBi and BiMnPt have not been measured,
perhaps these compounds do not even exist in this crystal structure. An earlier theoreti-
cal study examined both NiMnSb and MnNiSb in this respect, in order to determine the
θK εK

0.4
0.2
0.0
-0.2
-0.4
PtMnBi
-0.6 BiMnPt
PtGdBi
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>A@CBDdÉ+FHGE¥I
Theoretical Kerr angle K L and Kerr ellipticity L of PtMnBi, BiMnPt, Ï
and PtGdBi in C1 half-Heusler structure.
®
ground-state energy composition (Helmholdt et al. 1984), but this is not our present aim.
The third compound, PtGdBi, does exist in the C1 ´ crystal structure (Canfield et al. 1991,
Fisk et al. 1991), but MOKE measurements were not yet undertaken for this compound.
In PtGdBi the Gd atom is expected to carry a much larger magnetic moment than that of
Mn. A larger moment leads often to a larger Kerr rotation (Oppeneer and Antonov 1996,
Oppeneer et al. 1996b).
The theoretical Kerr spectra of these three compounds are given in Fig. 6.14. Let us
first consider PtMnBi in comparison to PtMnSb. We find for PtMnBi a maximal Kerr
rotation of -0.7 , which is much smaller than the one calculated for PtMnSb. The band
structure of PtMnBi, which is shown in Fig. 6.15, gives us the information that PtMnBi, in
contrast to PtMnSb, is not half-metallic. At 0.2 eV above $ % there is a small gap visible
at the 5 point. With the Fermi level at this energy position, the band structure would be
half-metallic. However, compared to PtMnSb there are further differences: at the 5 point,
the triplet bands and the singlet levels, which are of mixed Pt and Mn q character, hybridize
in a different manner from the equivalent bands in PtMnSb (see, e.g., de Groot et al. 1984).
The consequence is that the optical conductivities, and thus the Kerr spectra, differ from
those of PtMnSb. The Kerr effect in BiMnPt is, for photon energies up to 3 – 4 eV, similar
to that in PtMnBi. The Mn & states overlap in BiMnPt mainly with Bi states, yet, due
to the large exchange splitting of nearly 3 eV in the Mn 3 & bands, the energy bands of
;=<?>J@CBDÉ{FHG Ç+I
Spin-polarized, relativistic energy bands and total DOS of PtMnBi.
;=<?>A@CBD É+FHGPÉ+I
As Fig. 6.15, but for PtGdBi. Note the narrow gap at the Fermi energy.
BiMnPt are, close to the Fermi level, comparable to those of PtMnBi.

The calculated Kerr effect in PtGdBi shows a spectral dependence which is rather dif-
ferent from that of PtMnBi and BiMnPt (see Fig. 6.14). Compared to PtMnBi, the absolute
size of the Kerr angle in PtGdBi is particularly small. With respect to our calculations for
this compound, we should mention that it is still being discussed in the literature if LSDA
band-structure theory can properly describe pure Gd (see, e.g., Bylander and Kleinman
1994, Heinemann and Temmerman 1994). Such doubts could be equally valid for Gd
compounds, but for the present consideration we shall ignore these doubts. We calculate
the LSDA exchange splitting of Gd in PtGdBi to be about 5 eV, which is thus larger than
that of Mn. The spin-polarized, relativistic energy bands and DOS of PtGdBi are shown
;=<?>A@CBDäÉ+FHG Ó{I
Spin-projected partial DOS’s of PtGdBi. Majority-spin densities are given by the
full curve, minority-spin densities by the dotted curves.
in Fig. 6.16. Although the exchange-splitting on Gd is large, this does not give rise to a
large Kerr effect. The reason for this lies in the fact that the rather localized Gd 4 ß bands
do not hybridize so much with the Bi and Pt & bands. In Fig. 6.17 the partial DOS’s
of PtGdBi are depicted. In contrast to the partial DOS’s of PtMnSb, which are shown in
Fig. 6.2, the Pt and Bi bands in PtGdBi do nearly not spin polarize, as a result of the small
hybridization with the Gd 4 ß bands, in spite of the large moment being present on the Gd
site. Therefore the contributions of the optical transitions for the two spin directions that
take place within the hybridized Pt and Bi bands, cancel each other to a large extent, and a
small Á{*)Ä is the consequence. We remark furthermore that our calculations predict PtGdBi
to be a narrow gap semiconductor, in agreement with experiment (Canfield et al. 1991).
This property we obtain in the ferromagnetic phase of PtGdBi, which thus would place
PtGdBi theoretically in the exceptional group of ferromagnetic semiconductors, to which
belong, for instance, CrBr , CrI , EuO, and EuS, (see, e.g., Tsubokawa 1960, Matthias
et al. 1961, Morisin and Narath 1964). Experimental investigations do, however, indi-
cate that most likely antiferromagnetic, instead of ferromagnetic, ordering takes place in
PtGdBi (Canfield et al. 1991). Our calculations for the MOKE spectra of PtGdBi would
then correspond to PtGdBi in an applied magnetic field.
'ehe_r 2¡tH
w
u â0
w
u âäx/ su 0q:wt ºw
w u
Next we consider the related group of ternary compounds NiMnAs, PdMnAs, and PtMnAs,
as well as RuMnAs. Similar to the isoelectronic substitution of Bi for Sb, investigated
in the preceding Section, we can substitute As for Sb. This substitution leads thereby
to a reduced SO coupling strength of As as compared to that of Sb. At first sight the
isoelectronic substitution of As might seem straightforward, however, it has experimentally
been discovered that with As instead of Sb, NiMnAs and PdMnAs do not crystallize in
the MgAgAs crystal structure, but in the hexagonal Fe # P structure. (Nylund et al. 1972,
Chaudouet et al. 1982, Harada et al. 1990). The compound PtMnAs has, to our knowledge,
not been investigated experimentally. We shall assume for this compound the Fe # P crystal
structure too.
The MOKE spectra of these ternary arsenic compounds have not been measured as
yet. In Fig. 6.18 we present their ab initio calculated Kerr spectra. The Kerr rotations are
for these compounds calculated to be relatively small, of about -0.2 to -0.5 maximally.
Apart from the smaller SO coupling of As, of course the crystal structure different from
the C1 ´ structure comes into play. Hence, the shape of the Kerr rotation spectra bears no
direct correspondence to that of the Sb-based half-Heusler alloys. For the band structures
and total DOS’s of the three As-based compounds we refer to Antonov et al. (1997).
The three Fe # P structure compounds are not calculated to exhibit large Kerr effects.
Nevertheless, these compounds could be attractive for MO recording applications. One
of the reasons for this is that the magnetic anisotropy in the hexagonal Fe # P phase is
in general much larger than that of the cubic C1 ´ compounds. In cubic compounds
the magnetic anisotropy energy scales approximately with the fourth power of the SO
coupling strength, whereas in a hexagonal or uniaxial symmetry it scales as the square
of the SO coupling (Brooks 1940, Fletcher 1954, Bruno 1989). Therefore either uniaxial
multilayered or hexagonal materials are more favorable for MO recording applications,
since such materials exhibit the sufficiently large magnetic anisotropy which is required
to sustain the local magnetization orientation perpendicular to the MO disk surface (see,
e.g., Buschow 1988, 1989, Carey et al. 1995, Mansuripur 1995). In addition to the cubic
structure, PtMnSb has for MO recording applications the drawback that at the photon
frequency where the peak Kerr rotation occurs, the reflectivity becomes small (Takanashi
et al. 1991, Sato et al. 1992b, Ikekame et al. 1993). This is of relevance for the so-called
θK εK
0.2
0.0
-0.2
-0.4 NiMnAs
PdMnAs
PtMnAs
-0.6
0 1 2 3 4 5 6 0 1 2 3 4 5 6 7
Photon energy (eV)
;=<?>A@CBDÉ+FHG Ö{I
Theoretical Kerr rotation K L and Kerr ellipticity L spectra of the Ï
ternary alloys NiMnAs, PdMnAs, and PtMnAs in the Fe \ P crystal structure.
figure of merit, which is proportional to the signal-to-noise-ratio in MO reading, and which

'HG%# # 'HG$#
is therefore used as a measure for the MO recording quality of a material (Buschow 1989,
Schoenes 1992). The figure of merit is commonly defined as u Ø N O 6wv O Ú
#
, with
u the reflectivity, which can directly be computed from the optical conductivity (Reim
and Schoenes 1990). In the case of PtMnSb the small reflectivity therefore partially
compensates the advantage of the giant Kerr rotation at 1.7 eV. A third drawback that
PtMnSb has, particularly for modern MO recording applications, is the energy position of
the peak rotation. Future second-generation high-density MO storage devices utilize UV
lasers, i.e., having photon energies of about 3 – 4 eV (Weller 1996).
In comparison to PtMnSb, the As-based ternary intermetallics do have the following
advantages for MO recording applications: a reasonable Kerr effect at 3 – 4 eV, a normal
metallic reflectivity, and, expectantly, a sufficiently large magnetocrystalline anisotropy.
The physical properties of these compounds are only poorly investigated. It is therefore at
present unclear if they do not exhibit certain disadvantages for applications in MO devices.
In the case of PdMnAs it is known that the Curie temperature is about 200 K (Harada et al.
1990), which thus excludes this compound for room temperature MO applications.
The question thus remains if we can identify compounds which would make suitable
MO recording materials. Within the group of MnAs-based compounds, it has previously
been found that RuMnAs crystallizes in the Fe # P structure (Chenevier et al. 1985), and
RuMnAs
0.2
0.0
-0.2
-0.4
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>J@CBDÉ{FHG +I x\ Ï
Predicted Kerr rotation K L and Kerr ellipticity L of RuMnAs.
is ferromagnetic with a U"V = 497 K (Harada et al. 1990). In contrast to PdMnAs, this
UWV is appropriate for thermomagnetic writing (cf. Buschow 1988). We have therefore
investigated the Kerr spectra for this compound. These are shown in Fig. 6.19. Our
calculations predict for RuMnAs a nice Kerr rotation peak at 3 eV of about -0.4 . A
comparison of the band structure of RuMnAs, which we do not display here, with that
of, e.g., PdMnAs (Antonov et al. 1997), shows that there are hybridized Ru 4& bands
near $ % . The Kerr angle peak at the appropriate photon-energy position, together with the
appropriate U V and the magnetocrystalline anisotropy, would make RuMnAs an interesting
candidate for UV laser-light MO recording applications. Unfortunately, not much is know
about other physical properties of RuMnAs. In the related compound RhMnAs, which, in
contrast, orders antiferromagnetically, an exotic magnetic behavior (with canted magnetic
moments) has been observed (Chenevier et al. 1984, Bacmann et al. 1986, Kanomata et
al. 1987, Rillo et al. 1992). It is presently unclear if not also RuMnAs exhibits unusual
magnetic behavior (cf. Chenevier et al. 1985, Harada et al. 1990). If this is the case, it
would certainly diminish the prospect of applying RuMnAs in MO recording devices.
" ù$^ . 1 ÿ ÿy ÿ 9 ÿ

The unusually large MOKE angle of PtMnSb, as compared to those of NiMnSb and
PdMnSb, can satisfactorily be explained on the basis of ab initio band-structure calcu-
lations. The differences in the MOKE spectra of these three compounds arise straight-
forwardly as true relativistic band-structure effects. As we have shown explicitly, the
half-metallic property of both NiMnSb and PtMnSb does have an indirect influence on
the MO Kerr angle, since it leads to a rather small (intraband) plasma frequency. A small
plasma frequency in turn helps to create a minimum in the denominator of Eq. (2.33).
Ið
The other quantities that play a role in the “giant” Kerr rotation of PtMnSb, are the SO
coupling strength of Pt, the large magnetic moment of z ! on Mn, as well as the strong
hybridization of Mn, Pt, and Sb orbitals.
In the second place we have considered an additional group of other ternary inter-
metallic compounds in relationship to PtMnSb. We find computationally that none of
these related compounds exhibits a Kerr effect comparable to that in PtMnSb. This result
can be considered to be quite unexpected, because previous computational MO investiga-
tions (Oppeneer et al. 1996b) revealed unambiguous guidelines for predicting compounds
having large Kerr rotations. Although it can be understood how the giant Kerr angle of
PtMnSb arises, this illustrates also that PtMnSb does not follow a standard MO behavior.
Rather, the aforementioned characteristics of PtMnSb combine fortuitously in a manner
which amplifies the Kerr effect.
An interesting off-spin of our investigations is that we predict PtGdBi to be a semicon-
ductor in its ferromagnetic phase. The combination of ferromagnetism and semiconduc-
tivity is rarely realized in nature. In the case of PtGdBi the semiconducting property stems
from the same origin as the half-metallicity in NiMnSb and PtMnSb, namely, a combined
effect of a relatively strong hybridization of the various energy bands and the particular
energy-band filling.
Finally, we predict the As-based ternary intermetallic compounds NiMnAs, PdMnAs,
PtMnAs, and RuMnAs, to have promising MOKE spectra for MO recording applications.
RuMnAs, particularly, could be a suitable material for MO recording applications, as
RuMnAs exhibits a Kerr angle of -0.4 at 3 eV photon energy. The MOKE spectra of
these As-based intermetallics have not been measured as yet. Therefore, experimental MO
investigations of this group of ternary alloys are encouraged.
119
Ü ÝßÞáà7âäãæå {
| } ê õ 1é ô ñ î íòñ ó íòô õ ö ÷
~ ùûú ï 9 Zä : ü üÎþ ÿ ÿ ¡ 9:
ÿ
Actinide compounds display a rich variety of unusual physical properties, for which, nat-
urally, the 5 ß electrons are responsible. Varying from one actinide compound to another,
the 5 ß electrons may cause different anomalous phenomena (see, e.g., Sechovsk ý and
Havela 1988, Nieuwenhuys 1995). For example, in several intermetallics exotic magnetic
@
ordering has been found, as, e.g., in UNi B (Mentink et al. 1994), and U # Rh Si (Becker Ê
et al. 1997). Other compounds display exceptional phase transitions, as, e.g., the inverse
metal-insulator transition in UNiSn (Palstra et al. 1987, Fujii et al. 1989). In other
uranium intermetallic compounds, metamagnetic phase transitions accompanied by a giant
magneto-resistivity were discovered. Typical examples are UNiGa and UNiGe (Havela et
al. 1996). Another important group is undoubtedly that of the heavy-fermion materials.
Some archetypal, intensively investigated, uranium heavy-fermion compounds are UBe '
(Ott et al. 1983), UPt (Stewart et al. 1984), U # Zn ' (Ott et al. 1984), URu # Si # (Palstra
et al. 1985), and UPd # Al (Geibel et al. 1991). UBe ' , UPt , URu # Si # , and UPd # Al
are, moreover, heavy-fermion superconductors, i.e., the superconductivity coexists with
ð
the heavy-fermion state. UPd # Al is even more unusual, since it exhibits in addition pro-
nounced antiferromagnetic ordering with a substantial moment (0.85 ! ). The latter has
been proposed to be carried by localized 5 ß electrons (Feyerherm et al. 1994).
An efficient tool to investigate the electronic structure of lanthanide and actinide com-
pounds in much detail is optical spectroscopy. Recently, reflectivity spectroscopy in the
infra-red energy regime has successfully been applied to examine the absorption in the
heavy-fermion state of UPt (Wachter 1994, Donovan et al. 1997), URu # Si # (Bonn et
al. 1988), UPd # Al (Degiorgi et al. 1994), and UBe ' (Bommeli et al. 1997). Also,
reflectivity spectroscopy was applied recently to intermediate valence and dense Kondo
materials (see Wachter 1994 for a survey). MOKE spectroscopy is, as supplement to reflec-
tivity spectroscopy, ideally suited for studying magnetic materials, since Kerr spectroscopy
120 CHAPTER 7. URANIUM COMPOUNDS
probes both the spin and orbital polarization of the electron states (Reim and Schoenes
1990). In the last decade numerous MOKE spectra of lanthanide and actinide compounds
have been measured (see, e.g., Reim and Schoenes 1990). Recently, some spectacular new
results were achieved with this technique, as, e.g., the discovery of a record Kerr angle
of 90 in CeSb (Pittini et al. 1996, Schoenes and Pittini 1997), and the observation of a
'
magnetically polarized, empty 4 ß level in LaSe (Pittini et al. 1997). These recent discov-
eries, as well as the optical spectra of the above mentioned U compounds, are undoubtedly
topics of forefront importance in solid state physics. The theoretical explanation of the
measured spectra is highly relevant to the understanding of these complex phenomena. In
particular, a proper theoretical description of the measured spectra would certainly provide
a strong argument in debating the underlying physics. Currently the development on the
theoretical side has not proceeded so far that the optical spectra of strongly correlated elec-
tron systems can be evaluated in a fashion similar to the standard-type ab initio calculation
of such spectra in TM compounds. In the following we shall therefore concentrate on
another intriguing aspect of actinide compounds, which can be addressed within ab initio
electronic structure calculations of the optical spectra, namely the localization degree of
the 5 ß electrons.
It is perplexing that the localization behavior of the 5 ß electrons can change, in going
from one actinide compound to another, from being nearly localized to being practically
itinerant. The early actinides tend more towards delocalized 5 ß behavior (Zachariasen
1973). The 5 ß localization tendency and concomitant physical properties have been ex-
tensively investigated for uranium intermetallic compounds (see, e.g., Holland-Moritz and
Lander 1994, Sechovský and Havela 1988). The 5 ß localization tendency may be looked
upon in terms of the 5 ß band width, which is narrower than the 3 & band width, yet broader
than that of 4 ß ’s (Brooks and Johansson 1993). These two enveloping bounds mark the
different approaches to treat the 5 ß electrons that have become customary: Either a model
of localized ß ’s is adopted, which has proven to be applicable to explaining many properties
of lanthanides, or a delocalized band model is adopted. The latter has proven to be the
valid approach for transition metals.
If we next consider the group of U intermetallics in more detail, and look how many
compounds have actually been classified as having localized 5 ß electrons, then it appears
that only a few compounds could be classified as such. The clearest example is UPd ,
where experimentally the 5 ß electrons were detected at 1 eV below the Fermi energy ( $ % )
(Baer et al. 1980), in accordance with de Haas–van Alphen experiments, which excluded
ß states at $ % (Ubachs et al. 1986). Sharp crystal-field (CF) transitions, furthermore,
were observed (Shamir et al. 1978, Murray and Buyers 1980), as well as an intermultiplet
transition (McEwen et al. 1993, Bull et al. 1996). The CF and intermultiplet transitions
are, in analogy to such transitions in lanthanides, a finger-print of localized ß ’s. A sec-
ond example of localized 5 ß behavior was proposed to be UO # , for which CF transitions
were reported by Kern et al. (1985) and Amoretti et al. (1989). It appears that both
in UPd and UO# localized 5 ß behavior occurs because the 5 ß states are positioned in
a kind of gap between the various bands present, and therefore do not hybridize with
other states (Kelly and Brooks 1980, Johansson et al. 1986). Apart from U compounds
having evidently localized ß ’s, there exists another category, containing quite a number of
CHAPTER 7. URANIUM COMPOUNDS 121
compounds, for which it has been found that the 5 ß ’s are not really localized, neither are
they fully delocalized. This sort of intermediate group has been termed ‘semi-localized’ or
‘quasi-localized’. An experimental definition of semi-localized 5 ß ’s is that the occupied
5 ß ’s are detected at 0.5 eV to 0.8 eV below $ % . This is found to be the case for, e.g., USe
(Reihl et al. 1982b), UTe (Reihl et al. 1981), and UNiSn and UPtSn (H öchst et al. 1986).
In the case of USe and UTe it has, e.g., been found that the specific-heat coefficient is
rather small ( 12 mJ mol K ) (Rudigier et al. 1985), while for UTe upon diluting with
(Y' (C#
ï ï Ê ï Êï
Y and La CF effects appear in U / # (La/ ' Y X ) X Te (Schoenes et al. 1996). In various
other U compounds the 5 ß ’s are considered to be delocalized. Well-known examples are
UN (Reihl et al. 1982a) and UFe # (Aldred 1979). We mention that this (de)localization
classification is somewhat simplified, because the origin of 5 ß delocalization (due to direct
ß 8 ß overlap or ß -ligand hybridization) is not taken into consideration.

A direct comparison of first-principles optical and MOKE spectra with experimental
spectra constitutes a powerful mean to address the 5 ß electron localization degree (see,
e.g., Delin et al. 1997, Köhler et al. 1997, Oppeneer et al. 1997, 1998). In the case that
the electrons are sufficiently delocalized, the itinerant LSDA approach provides a minute
description of the optical and MO spectra. This kind of agreement normally holds for
TM compounds (see, e.g., Oppeneer and Antonov 1996). If the electrons are, however,
partially localized, inherent many-electron correlation effects, which can already be im-
portant for the proper description of ground-state properties, should become imperative
for describing optical excitations, for example when a strong on-site Coulomb interaction
between electron and hole quasi-particles plays a dominating role (Kunz and Flynn 1983,
Anisimov et al. 1993). Therefore, if the uranium 5 ß electrons are localized, one would
particularly expect to observe corresponding correlation effects in the optical spectra. We
note in this context that ab initio band-structure calculations on the basis of the itinerant
LSDA approach failed in giving a satisfactory description of the optical and MOKE spectra
of the uranium monochalcogenides (Cooper et al. 1992, Lim et al. 1993, Brooks et al.
1995, Kraft et al. 1995, Kraft 1997). This failure has not yet been fully clarified, but was
supposed to be due to an insufficient treatment of the partially localized 5 ß electrons in
the LSDA (Cooper et al. 1992). Apart from the uranium monochalcogenides, theoretical
studies of the optical and MOKE spectra of actinide intermetallics are still rare, the re-
search field is thus only in its infancy. One important conclusion, which has so far been
reached, is that an appropriate description of the electronic structure (including degree of
ß localization) will provide an accurate explanation of the measured spectra. Precisely this
notion is of great help in deciding if an actinide intermetallic can be regarded as having
delocalized 5 ß ’s or not.
~ ù-, 3 ÿ ) ÿP 9 ÿ 9: ÿ

In the following we consider several U compounds: UAsSe, URhAl, and UFe # . UAsSe is
one of the U compounds which, on account of photoemission experiments, was classified to
be a localized 5 ß material (Brunner et al. 1981). URhAl has been investigated intensively
in the last years (Paixão et al. 1992). Quite recently, an inelastic neutron peak was observed
for URhAl, which could be due to an intermultiplet transition (Hiess et al. 1997). This
observation thus has raised the interesting question if URhAl could possibly be one of the
few localized ß uranium intermetallics. UFe # , on the other hand, is generally considered
as a typical example of a U compound having itinerant 5 ß electrons (Brooks et al. 1988).
k eheji u |
UAsSe is one the ternary U compounds which was proposed to have localized 5 ß elec-
trons (Zygmunt et al. 1972, Bielov et al. 1973, Brunner et al. 1981). Photoemission
experiments revealed final-state effects, and thus indicated localized 5 ß electrons in UAsSe
(Brunner et al. 1981). Magnetic susceptibility and magnetization measurements showed
a strong anisotropy in magnetic properties, which were interpret as localized-moment be-
havior, and explained within a CF model for a 5 ß configuration (i.e., U @C ) (Zygmunt
#
and Duczmal 1972, Bielov et al. 1973). On the other hand, reflectivity and MO Kerr
spectroscopy revealed a pronounced spectral intensity at small photon energies ( eV)
which was attributed to a 5 ß band located at the Fermi energy (Reim et al. 1985, Reim
1986). Also, the specific-heat coefficient = 41 mJ mol K (Blaise et al. 1980) is
(Y' (C#
not particularly small, indicating a tendency to itineracy. These apparently contradictory
observations show that the behavior of the 5 ß electrons and the related magnetic properties
of UAsSe are not yet well understood. A characterization of the 5 ß electrons in accord
with all of these experiments would be one of, in some sense, partially delocalized or
partially localized 5 ß states. In this regard the 5 ß behavior in UAsSe compares to that
in the uranium monochalcogenides which are categorized to have partially localized 5 ß ’s,
with an increasing localization tendency towards UTe (Reihl et al. 1981, Rudigier et al.
1985, Lander et al. 1990, Brooks and Johansson 1993).
UAsSe crystallizes in the tetragonal PbFCl crystal structure (also called ZrSiS structure,
P4/nmm space group) and orders ferromagnetically along the ñ -axis below U V 110 K
(Hulliger 1968, Leciejewicz and Zygmunt 1972). The unit cell of UAsSe in the PbFCl
structure is shown in Fig. 7.1. The U atoms are arranged in planes perpendicular to the
ñ -axis; there is a sequential stacking of As–U–Se–Se–U–As layers in ñ -direction. As in
other uranium compounds with non-cubic structures such as, e.g., the ternary U(TM) # X#
compounds (Sechovský and Havela 1988, Sandratskii and Kübler 1994) there exists –
enforced by the crystal structure – the possibility of partially delocalized electrons in an-
other sense. These compounds with a preferred ñ -axis tend to have the uranium atoms
arranged in layers which leads to both anisotropic bonding and magnetic properties due to
hybridization of 5 ß and or & states. It is therefore quite possible that the 5 ß electrons
are delocalized in the planes but localized along the ñ -axis and it can be suspected that
this may be the case in UAsSe which distinguishes it from UAs and USe, both of which
are cubic rocksalt compounds. In conjuncture to the crystal structure, anomalous transport
properties were reported for both UAsSe and ThAsSe single crystals (Schoenes et al. 1988,
Henkie et al. 1995).
In Fig. 7.2 the calculated polar MOKE spectra of UAsSe are shown together with the
experimental spectra (Reim 1986). The theoretical Kerr spectra are given for two different
broadening parameters, 0.03 Ry (dashed-dotted curve) and 0.04 Ry (solid curve). The
;Q<R>A@CBDäÓ{FHGvI
Crystallographic unit cell of UAsSe in the PbFCl crystal structure. The
magnetic moments on the uranium sites are depicted by the arrows.
spectra in Fig. 7.2 illustrate that there is good agreement between band theory and experi-
ment: Both the position and height of the main peak in the Kerr angle at 3 eV are correct,
with the usual (small) dependence of the theoretical peak height on the broadening. There
is also a calculated smaller peak in the Kerr rotation at 1 eV, where there is a shoulder in
the measured Kerr angle. A Kramers-Kronig related peak structure is visible in the Kerr
ellipticity spectrum at 1.5 eV. In this energy region the influence of intraband conductivity
may be large enough to make the interband and total spectra differ appreciably. The in-
fluence of the intraband contribution may be estimated by adding an intraband Drude-type
conductivity to the computed interband conductivity. The thus-resulting Kerr spectra are
'Ê
given by the dashed curve, where the interband conductivity spectra are those for 0.04 Ry
spectral broadening, and the Drude parameters are Á þ = 3 ½ò s , ù- þ = 0.02 Ry. In
(W' (W'
effect, the phenomenological Drude conductivity reduces the height of the peak in the Kerr
rotation at 1 eV, but it leaves the peak at 3 eV almost unchanged (see Fig. 7.2), and also the
small maximum in the Kerr ellipticity at 1.5 eV becomes much less pronounced. It thus
brings the calculated Kerr spectra in better agreement with the measured spectra at photon
UAsSe
1.5
Complex polar Kerr effect (-deg) θK εK
1.0
0.5
0.0
-0.5
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>J@CBD Ó{FRT+I
Ï
Theoretical and experimental ( , Reim 1986) polar Kerr rotation K L
and Kerr ellipticity L spectra of UAsSe. The theoretical, interband-only, spectra
are given for two different relaxation-time parameters, Ô Õ = 0.03 Ry (solid curve) and
0.04 Ry (dotted curve). The approximate influence of a Drude intraband contribution is
illustrated by the dashed curved, for which the interband part of the spectra are those of
the 0.04 Ry relaxation-time broadening.
energies below 2 eV.

It is quite surprising – in view of the earlier experience with the uranium monochalco-
genides (Cooper et al. 1992, Brooks et al. 1995, Kraft et al. 1995, Kraft 1997) – that the
LSDA band-structure theory gives such a good description of the MOKE spectra. UAsSe
is actually the first U compound where convincing agreement with experimental MOKE
spectra was obtained (Oppeneer et al. 1996d). If the uranium 5 ß states were localized,
then one would expect that the related strong electron correlations would have a substantial
effect upon the optical spectra so that these would differ from the LSDA-itinerant electron
result (Kunz and Flynn 1983, Liechtenstein et al. 1994, Yaresko et al. 1996). This appears
not to be the case for UAsSe. Likely, this is because the 5 ß electrons in UAsSe are indeed
itinerant in the U planes which are selectively probed by the polar MO spectroscopy. The
perpendicular direction is not probed, and with regard to the experiments which support
localized behavior (Zygmunt and Duczmal 1972, Bielov et al. 1973, Brunner et al. 1981),
it may still be possible that the 5 ß electrons are more localized perpendicular to the planes.
If this is the case, then it may even be that the 5 ß electron correlation perpendicular to the
planes is described poorly by the LSDA which – due to its construction from the homoge-
neous electron gas – is a delocalized description in all directions. We note further that the
rather ‘absolute’ distinction between localized and delocalized 5 ß electrons is normally
a classification stemming from experiment. In comparison to the uranium monochalco-
genides, we find that the average extent of the calculated 5 ß wave function in UAsSe is
similar to that of USe. The 5 ß states of US are found to be more extended, and those of
UTe to be more localized than those of USe.
In spite of the close correspondence between experimental and theoretical Kerr spectra,
ð ð ð
not all properties of UAsSe are equally well given. The calculated total magnetic moment
ð
on uranium is only 0.77 ! (with spin moment -1.86 ! and orbital moment 2.63 ! ) which
is smaller than the experimental moment of about 1.36 – 1.50 ! (Zygmunt and Duczmal
ð ð
1972, Bazan and Zygmunt 1972). The calculated moment is practically completely due to
the 5 ß states: the ß components of the spin and orbital moment are -1.76 8! and 2.60 W! ,
respectively. It is a well-known fact, however, that within the LSDA the total magnetic
moment of uranium compounds in general comes out too small (Brooks and Johansson
1993). Corrections that simulate Hund’s second rule in solids, which describes orbital
correlations absent in the homogeneous electron gas, such as the orbital polarization (OP)
are needed to bring the magnetic moment in better agreement with experiment (Brooks
1985, Eriksson et al. 1990a, Severin et al. 1993). It appears therefore that the MOKE
spectra can be properly given by band theory, even if the orbital moment is not. The
calculated linear-temperature specific-heat coefficient = 30 mJ mol K , furthermore,
(W' (#
(W' (C#
corresponds reasonably well to the experimental value of 41 mJ mol K (Blaise et al.
1980).
The anisotropic structure of UAsSe suggests the interesting possibility of measuring
the anisotropy of the polar MOKE spectra in UAsSe. The MO spectra can be expected
to be highly anisotropic if the uranium 5 ß bonding is anisotropic; the MO response being
expectantly dominated by that of the U atoms. The anisotropic Kerr spectra are shown
in Fig. 7.3, for the polar Kerr axis (and magnetic moment) being oriented both along
the ñ -axis and along the ÷ -axis. The latter geometry probes the MO response of the 5 ß
electrons perpendicular to the plane. The intraband Drude conductivity has been added
to the calculated interband conductivities for both orientations. The spectra for the polar
÷ -axis differ in the energy range of 1 – 5 eV substantially from those for the polar ñ -axis.
Currently the polar Kerr spectra have not yet been measured for the magnetic moment
oriented along the ÷ -axis. Such measurements would undoubtedly contribute valuable
information concerning the behavior of the 5 ß ’s perpendicular to the plane. In conclusion,
the measured MO Kerr spectra of UAsSe are satisfactorily reproduced within a band-like
description of the 5 ß electrons. This puts forward unambiguous evidence for the existence
of at least partially itinerant 5 ß electrons in UAsSe.
keghe pt
z
URhAl belongs to the class of ternary uranium intermetallics which crystallize in the
hexagonal Fe # P structure (sometimes called ZrNiAl structure) (Veenhuizen et al. 1988a,
Sechovský and Havela 1988). In this structure, the U atoms (and 33 of the Rh atoms) are
UAsSe
1.5
Complex polar Kerr effect (-deg) θK εK
1.0
0.5
0.0
M || c
M || a exp.
-0.5
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;=<?>A@CBDÓ{FR[+I
Geometrical dependence of the calculated polar Kerr spectra of UAsSe.

The spectra for which the -axis is the axis of incident light are given by the solid curve,

while those for which the -axis is the incident light axis are given by the dashed curve.
In both cases has a Drude conductivity been taken into account and a broadening of

0.03 Ry been applied. Experimental spectra ( , Reim 1986) are shown for comparison.
arranged in planes perpendicular to the ñ -axis. These planes are separated from one another
by a plane consisting of the remaining Rh atoms and the Al atoms (see, e.g., Sechovsk ý
and Havela 1988). In the first investigations of URhAl, it was found that URhAl is a
ferromagnet (UWV = 27 K), exhibiting a high magnetocrystalline anisotropy, with the easy
axis along the ñ -axis (Veenhuizen et al. 1988a).
Very recently, URhAl has drawn particular attention, because inelastic neutron-scatter-
ing experiments revealed a peak at 380 meV, which could be the signature of an intermul-
tiplet transition in URhAl (Hiess et al. 1997). The value of 380 meV is quite close to the
intermultiplet transition energy of 390 meV measured for UPd (McEwen et al. 1993, Bull
ð
et al. 1996). Several other properties of URhAl, however, would rather advocate delocal-
ized 5 ß behavior in URhAl. A small U moment of only 0.94 ! was measured which does
#
corresponds to the magnetic moment of neither a 5 ß nor a 5 ß configuration (Paixão et al.
1992). A significant amount of anisotropic 5 ß ligand hybridization was reported (Paix ão
et al. 1993). Also, the measured specific heat = 60 mJ mol K is not particularly
(W' (C#
small (Veenhuizen et al. 1988b). These contradictory observations demonstrate that the
5 ß behavior in URhAl is not yet understood.
URhAl
εK

0.2 θK
0.0
-0.2
-0.4
-0.6
0 1 2 3 4 5 0 1 2 3 4 5 6
Photon energy (eV)
;Q<R>A@CBDÓ{FECI
Experimental and theoretical MOKE spectra of URhAl. The theoretical
MOKE spectra are calculated with the itinerant LSDA approach. The experimental
spectra are after Beránková et al. (1996).
In order to examine the applicable electronic structure for URhAl, we consider the
MOKE spectra. In Fig. 7.4 we show the experimental Kerr spectra of URhAl measured by
Beránková et al. (1996), together with the theoretical spectra calculated using the itinerant
LSDA approach. The first spectral peak at 1 eV, and the second one at 2 – 3 eV in the Kerr
angle are definitely reproduced in the theoretical spectrum. The theoretical Kerr rotation
drops off between 4 and 5 eV, but it is not yet clear if this also occurs in the experimental
N O spectrum. It can, nevertheless, be concluded that the itinerant 5 ß model explains the
measured Kerr spectra fairly well. Clearly, this supports the picture of delocalized 5 ß elec-
trons in URhAl. The calculated electronic specific-heat coefficient, = 41 mJ mol K ,
(W' (C#
is – compared to the measured = 60 mJ mol K (Veenhuizen et al. 1988b) – quite
(W' (#
ð ð
reasonable since a many-body enhancement of 1.5 can be considered to be normal. The U
ð ð
magnetic moment is calculated to be -1.2 ! , 1.6 ! , for spin, orbital moment, respectively,
whereas the measured counterparts are -1.16 ! and 2.10 Y! (Paixão et al. 1992). These
values indicate that the orbital polarization of URhAl is underestimated within the LSDA.
One of the shortcomings of the LSDA in describing actinides is the underestimation of the
OP (Brooks and Johansson 1993). Interestingly, the fact that the OP is too small within
the LSDA does not prohibit a reasonable explanation of the MO Kerr effect.
With respect to the signified intermultiplet transition, one could speculate that the 5 ß ’s
in URhAl divide into two groups, relatively delocalized, rather hybridized 5 ß ’s in the U–Rh
plane, and more localized 5 ß ’s perpendicular to this plane, in accord with the observation
of anisotropic ß hybridization in URhAl (Paixão et al. 1993). The possible intermultiplet
transition might correspond to the localized 5 ß ’s, whereas Kerr spectroscopy in the polar
geometry probes the MO response in the U–Rh plane. A similar proposal has been put
forward to explain the 5 ß behavior in UAsSe (Oppeneer et al. 1996d), and it has also been
suggested for UPd # Al by Feyerherm et al. (1994). In the current stage, more experiments
are obviously needed to answer the question of localized 5 ß behavior in URhAl. None the
less, the correspondence between the calculated and experimental Kerr spectra speaks in
favor of delocalized 5 ß behavior.
k ehef a|
UFe # is one of the cubic actinide Laves phase compounds that has been intensively studied
for many years (cf. Aldred 1979, Andreev et al. 1979, Paolasini et al. 1996). UFe #
is ferromagnetic with U V 172 K (Lin and Ogilvie 1963). The general picture that has
emerged for UFe # is that of delocalized 5 ß ’s which strongly hybridize with Fe 3 & electrons
(Aldred 1979, Brooks et al. 1988). Photoemission experiments established the existence
of a 5 ß band at $ % , as well as the absence of final state effects (Naegele 1985). As a
ð ð
consequence of the delocalization, the spin moment on U is substantially reduced from its
ð
free ion value to -0.22 ! , and the orbital moment to 0.23 ! (Lebech et al. 1989). The
total moment on Fe is reduced to 0.60 ! (Lebech et al. 1989). Even though the delocalized
5 ß behavior of UFe # is firmly established, and though the LSDA approach does predict the
correct trend in UFe # , it has not been able to explain quantitatively the vanishing moment
ð
on U (Brooks et al. 1988). The inclusion of the OP correction brings the calculated total
ð
U moment to -0.15 Y! (Eriksson et al. 1990b), in better agreement with experiment. The
ð
total moment per formula unit becomes, however, 1.63 ! (Eriksson et al. 1990b), which
ð
is larger than the computed total moment without OP of 1.40 8! and also larger than the
ð ð
experimental total moment of 1.19 "! (Lebech et al. 1989). Also, the calculated spin and
orbital moment (with OP) amount to -1.03 ! and 0.88 ! , respectively, each of which
deviates more from experiment than the corresponding LSDA values.
These findings illustrate that not everything about the electronic structure of UFe # is
yet explained. Another stringent test for the applicability of the LSDA approach to UFe #
would be the description of the MO Kerr spectra. In Fig. 7.5 we show the theoretical,
interband-only, MO Kerr spectra, calculated using the LSDA approach, for both the (001)
and (111) orientations of the magnetic moments. The Kerr rotation of UFe # has so far been
measured by Kirby et al. (1991) on a polycrystalline sample. The measured Kerr angle of
Kirby et al. varies between +0.2 at 1.1 eV and +0.3 at 2.5 eV, which does not correspond
to the theoretical Kerr angle predicted. In the present stage, however, it appears to be too
early to conclude on the correspondence between measured and calculated MOKE spectra.
Rather, we would highly encourage precise optical and MOKE effect measurements on
single-crystalline UFe # samples.
In Fig. 7.5 we have included the interband-only, MOKE spectra of Fe. It turns out that
0.2
θK εK

0.0
-0.2
-0.4
UFe2 (001)
UFe2 (111)
Fe
-0.6
0 2 4 6 0 2 4 6 8
Photon energy (eV)
;=<?>A@CBDÓ{FRÇ{I
Theoretical, intraband only, Kerr rotation K L and Kerr ellipticity L Ï
spectrum of UFe \ for the (001) and (111) orientation of the magnetization. For compar-
ison, the Kerr spectra of bcc Fe are also shown. All spectra have been calculated using
the itinerant LSDA approach.
the Kerr spectra predicted for UFe # are not so different from those of bcc Fe. One of the
reasons for this is that the moment on U is small, i.e., direct optical transitions on U will
not contribute much to the spectrum. Pilot calculations which we performed, have shown
that U contributes indirectly to the Kerr spectra through its large SO coupling contained in
the hybridized, spin-polarized U–Fe bands. The magnetocrystalline anisotropy in the Kerr
spectra of UFe # , furthermore, is calculated to be quite small. This is related to the cubic
crystal symmetry, and possibly, also to the small U moment again. Experimentally the (111)
axis is known to be the easy magnetization axis (Aldred 1979). Our calculations for UFe #
indeed confirm the total energy for the (111) orientation to be lower than that of the (001)
orientation: $ = (-0.28 ¼ 0.14) meV/formula unit, for ÷ = 7.055 . We mention that we
have not taken the small rhombohedral lattice distortion which is known to occur in UFe #
into account in the calculations. We note that this value of the magnetocrystalline anisotropy
energy is very small compared to the value of -80 meV, which has been extrapolated
experimentally for the cubic rocksalt US (Lander et al. 1991). We can thus confirm
that the itinerant 5 ß treatment does correctly explain many physical properties of UFe #
(cf. Brooks et al. 1988). More detailed information about the electronic structure could
expectedly be gained from the comparison of calculated and accurate experimental spectra.
~ ù$^ `9 þ 9: ÿ
The recent progress in first-principles calculations of optical spectra has illustrated that
optical and MO spectra are developing into a powerful tool for tracing in much detail
the electronic structure of both lanthanide and actinide compounds. The optical spectra
depend quite sensitively on the underlying electronic structure, therefore these spectra can
be utilized to assess the degree of localization of the ß electrons. The basic notion which is
extremely helpful in this respect, is that the proper electronic structure model will provide
the correct description of the measured optical and MO spectra (cf. Cooper et al. 1995,
Yaresko et al. 1996, Delin et al. 1997, Oppeneer et al. 1997, 1998). Calculated optical
and MOKE spectra represent, moreover, a new argument in discussing the appropriate
electronic structure model that is applicable to a certain U compound. Because such
calculations were not yet possible a few years ago, no conclusive evidence could be
extracted from considerations of the optical spectra. In this respect, it is quite remarkable
that for two compounds, UAsSe and URhAl, which were discussed in the literature to
have localized 5 ß electrons, explicit calculations of the MOKE spectra supported rather
the picture of delocalized, or partially delocalized ß electrons. Two other U compounds,
UNiGa and U P , that are not directly considered as compounds having localized 5 ß ’s,
@
were investigated recently. For U P it was found that a delocalized 5 ß description
@
provided a good explanation of the measured reflectivity and MOKE spectra (K öhler et al.
1997). Yet, in spite of these first theoretical successes in explaining the MOKE spectra
of U intermetallics, we have to emphasize that still the theoretical development in this
area is only in an early stage. To arrive at a more complete picture of the limitations
and possibilities of the theoretical approach, many more actinide intermetallics are to
be investigated, both experimentally and computationally. In view of the attention that
was given to U compounds exhibiting localized or semi-localized 5 ß ’s, it would be of
particular interest to compare also measured and ab initio calculated MOKE spectra of
materials which are regarded to be definitely itinerant, as, e.g., UFe # , UN, or UGa (see
Kaczorowski et al. 1993).
A recent survey of those U compounds that were studied experimentally was given
by Reim and Schoenes (1990) and Schoenes (1992). Two groups of U compounds which
received considerable attention in the last years, are the uranium monochalcogenides and
monopnictides. USe and UTe are examples of U systems having semi-localized 5 ß ’s
(Reihl et al. 1981, 1982b). US, on the other hand, is regarded to be partially itinerant
(Brooks and Johansson 1993). While the U monochalcogenides deserve surely further
investigations, a first result obtained is that the gross features of the experimental Kerr
spectra are better described using the LSDA+ approach (Cooper et al. 1995, Oppeneer
et al. 1997) than the LSDA approach (Cooper et al. 1992, Brooks et al. 1995, Kraft et al.
1995). Although the physical model applicable to the U monochaclogenides needs still to
be clarified further, it appears in first instance that the LSDA+ approach can be applied
to actinide materials having semi-localized 5 ß electrons. The U monopnictides display a
complicated antiferromagnetic structure and a different 5 ß localization behavior than the U
chalcogenides (see, e.g., Lander and Burlet 1995). First-principles LSDA calculations for
the U monopnictides shows that the normal optical conductivity spectra agrees reasonably
with experimental data (Kraft 1997). A direct comparison of calculated and measured Kerr
spectra is hindered by the fact that experimentally the magnetization could not be saturated
(cf. Reim and Schoenes 1990). This, once more, illustrates that there remains a demand
for well-defined, experimental MOKE spectra of single-crystalline U compounds.
133
è &
é í`é í} ê ó
Aldred, A. T., J. Magn. Magn. Mater. c , 42 (1979).

Alouani, M., J. M. Koch, and `¢£ M.sA.
¤ Khan, J. Phys. F: Metal Phys. , , 473 (1986).
Altarelli, M., Physica B ¡ ¡ , 365 (1997).
Altarelli, M., D. L. Dexter, H. M. Nussenzweig, and D. Y. Smith, Phys. Rev. B , 4502
(1972).
Amoretti, G., A. Blaise, R. Caciuffo, J. M. Fournier, M. T. Hutchings, R. Osborn, and A.
Taylor, Phys. Rev. B ¥` , 1856 (1989).
Andersen, O. K., Phys. Rev. B ,¡ , 3060 (1975).
Andreev, A. V., A. V. Deryagin, R. Z. Levitin, A. S. Markosyan, and M. Zeleny, Phys.
Status Solidi (a) ¦s¡ , K13 (1979).
Andresen, A. F., W. Hälg, P. Fischer, and E. Stoll, Acta Chem. Scand. ¡ , 1543 (1967).
Anisimov, V.¤ I., I. V. Solovyev, M. A. Korotin, M. T. Czyzyk, and G. A. Sawatzky, Phys.
Rev. B ¥ , 16929 (1993). ¤`¢
Antonini, B., F. Lucari, F. Menzinger, and A. Paoletti, Phys. Rev. , 611 (1969).
Antonov, V. N., A. I. Bagljuk, A. Y. Perlov, V. V. Nemoshkalenko, Vl. N. Antonov, O. K.
Andersen, and O. Jepsen, Low Temp. Phys. c§ , 494 (1993).
Antonov, V. N., A. Y. Perlov, A. P. Shpak, and A. N. Yaresko, J. Magn. Magn. Mater. c¥ ,
205 (1995).
Antonov, V. N., P. M. Oppeneer, A. N. Yaresko, A. Y. Perlov, and T. Kraft, Phys. Rev. B
¦s , 13012 (1997). ¢
Argyres, P. N., Phys. Rev. § , 334 (1955).
Armelles, ¤ G., D. Weller, B. Rellinghaus, P. Caro, A. Cebollada, and F. Briones, J. Appl.
Phys. ¡ , 4449 (1997a).
Armelles, G., D. Weller, B. Rellinghaus, R.CF. C. Farrow, M. F. Toney, P. Caro, A. Cebollada,
and M. I. Alonso, IEEE Trans. Magn. , 3220 (1997b).
Arvanitis, D., M. Tischer, J. Hunter Dunn, F. May, N. Mårtensson, and K. Baberschke, in
Spin-Orbit-Influenced Spectroscopies of Magnetic Solids, edited by H. Ebert and G.
Schütz (Springer, Berlin, 1996), p. 145. ¢
Auwärter, M., and A. Kussmann, Ann. Phys. , 169 (1950).
Azzam, R. M. A., and N. M. Bashara, Ellipsometry and Polarized Light (North-Holland,
Amsterdam, 1986).
Bacmann, M., B. Chenevier, ¢ D. Fruchart, O. Laborde, and J. L. Soubeyroux, J. Magn.
Magn. Mater. ¦/¥/¨¦ , 1541 (1986). ¢
Bacon, G. E., and J. Crangle, Proc. Roy. Soc. (London) A ¡ ¡ , 387 (1963).
134 BIBLIOGRAPHY

Baer, Y., H. R. Ott, and K. Andres, Solid State Commun. , 387 (1980).
Baibich, M. N., J. M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P. Etienne, G. Creuzet,
A. Friederich, and J. Chazelas, Phys. Rev. Lett. , 2472 (1988).
Baumgarten, L., C. M. Schneider, H. Petersen, F. Schäfers, and J. Kirschner, Phys. B ¥¥ ,
4406 (1991).
Bazan, C., and A. Zygmunt, Phys. Status Solidi (a) /¡ , 649 (1972). ¢
¢s¤
Beaurepaire, E., J.-C. Merle, A. Daunois, and J.-Y. Bigot, Phys. Rev. Lett. , 4250 (1996).
Bechstedt, F., K. Tenelsen, B. Adolph, R. Del Sole, Phys. Rev. Lett. , 1528 (1997).
¢s¤
Becker, B., S. Ramakrishnan, A. A. Menovsky, G. J. Nieuwenhuys, and J. A. Mydosh,
Phys. Rev. Lett. , 1347 (1997). `¢
Bennett, H. S., and E. A. Stern, Phys. Rev. , A448 (1965).
Bennett, W. R., W. Schwarzacher, and W. F. Egelhoff, Phys. Rev. Lett. `¦ , 3169 (1990).
¶ª©h¶4«
Beránková, P., M. Kučera, M. Matyáš, and A. A. Menovsky, in 26 D Journées des
Actinides, Szklarska Poreba, Poland, p. 70 (1996).
Besnus, M. J., and A. J. P. Meyer, Phys. Status Solidi (b) ¦¦ , 537 (1973).
Bielov, K. P., A.S.
¢ Dmitrievskii, A. Zygmunt, R. Z. Levitin, and W. Trzebiatowski, Sov.
Phys. JETP , 297 (1973).
¢
Binash, G., P. Grünberg, F. Sauernbach, and W. Zinn, Phys. Rev. B § , 4828 (1989).
Birss, R. R., Proc. Phys. Soc. § , 946 (1961).
Birss, R. R., Symmetry and Magnetism (J. Wiley, New York, 1964).
Blaise, A., R. Lagnier, A. Wojakowski, A. Zygmunt, and M. J. Mortimer, J. Low Temp.
Phys. ¥¬ , 61 (1980).
Bland, J. A. C., and B. Heinrich (Eds.), Ultrathin Magnetic Structures (Springer, Berlin,
1994).
Blount, E. I., Solid State Phys. , 305 (1962).
Bobo, J. F., P. R. Johnson, M.¤ Kautzky, F. B. Mancoff, E. Tuncel, R. L. White, and B. M.
Clemens, J. Appl. Phys. , 4164 (1997).
Bommeli, F., L. Degiorgi, P. Wachter, F. B. Anders, and A. V. Mitin, Phys. Rev. B. ¦ ,
R10001 (1997).
Bona, G. L., F. Meier, M. Taborelli, E. Bucher, and P. H. Schmidt, Solid State Commun.
¦s , 391 (1985).
Bonn, D. A., J. D. Garrett, and T. Timusk, Phys. Rev. Lett. , 1305 (1988).
Borgschulte, A., D. Menzel, K. U. Harder, T. Widmer, and J. Schoenes, J. Magn. Soc. Jpn.,
in press (1999).
Born, M., and E. Wolf, Principles of Optics (Pergamon, Oxford, 1980).
Boswarva, I. M., R. E. Howard, and A. B. Lidiard, Proc. Roy. Soc. (London) ®¡s§ , 125
(1962).
Brändle, H., J. Schoenes, F. Hulliger, W. Reim, unpublished results (1991), cited by J.
Schoenes, in Materials Science and Technology, Vol. 3A: Electronic and Magnetic
Properties of Metals and Ceramics (volume editor: K. H. J. Buschow), edited by R. W.
Cahn, P. Haasen and ¤E. J. Kramer (Verlag Chemie, Weinheim, 1992), p. 147.
Brooks, H., Phys. Rev. ¦ , 909 (1940).
Brooks, M. S. S., Physica B , 6 (1985).
Brooks, M. S. S., and B. Johansson, in Handbook of Magnetic Materials, edited by K. H. J.
Buschow (North-Holland, Amsterdam, 1993), Vol. 7, p. 139.
BIBLIOGRAPHY 135
Brooks, M. S. S., O.¤ Eriksson, B. Johansson, J. J. M. Franse, and P. H. Frings, J. Phys. F:

Metal Phys. , L33 (1988).
Brooks, M. S. S., T. Gasche, and B. Johansson, ¤ J. Phys. Chem. Solids ¦s , 1491 (1995).
Bross, H., and G. M. Fehrenbach, Z. Phys. B , 233 (1990).
Bross, H., O. Belhachemi, B. Mekki, and A. Seoud, J. Phys.: Condens. Matter ¡ , 3919
(1990). s¤
Brunner, J., M. Erbudak, and F. Hulliger, Solid State Commun. , 841 (1981).
Bruno, P., Phys. Rev. B § , 865 (1989).
Bruno, P., Physical origins and theoretical models of magnetic anisotropy, in Magnetismus
von Festkörpern und Grenzflächen (Ferienkurse des Forschungszentrums J ülich, Jülich,
1993). s£
Bull, M. J., K. A. McEwen, R. Osborn, and R. S. Eccleston, Physica B ¡s¡ ¡¡/¥ , 175
(1996).
¤
Burke, S. K., B. D. Rainford, D. E. G. Williams, P. J. Brown, and D. A. Hukin, J. Magn.
Magn. Mater. ,¦c¨?¤ , 505 (1980).
Busch, G., Physica B § , 1 (1977).
Buschow, K. H. J., in Ferromagnetic Materials, edited by E. P. Wohlfarth and K. H. J.
Buschow (North-Holland, Amsterdam, 1988), Vol. 4, p. 558.
Buschow, K. H. J., J. Less-Common Metals ,¦¦ , 307 (1989).
Buschow, K. H. J., and P. G. van Engen, J. Magn. Magn. Mater. ¡s¦ , 90 (1981). s¤
Buschow, K. H. J., P. G. van Engen, and R. Jongebreur, J. Magn. Magn. Mater. , 1
(1983a).
¢ P. G. van Engen, D. B. de Mooij, and A. M. van der
Buschow, K. H. J., J. H. N. van Vucht,
Kraan, Phys. Status Solidi (a) ¦ , 617 (1983b).
Bylander, D. M., and L. Kleinman, Phys. Rev. B ¥`§ , 1608 (1994). ¢
Cabet, E., A. Pasturel, F. Ducastelle, and A. Loiseau, Phys. Rev. Lett. , 3140 (1996). C
Cable, J. W., E. O. Wollan, W. C. Koehler, and M. K. Wilkinson, J. Appl. Phys. , S3,
1340 (1962).
Callaway, J., Quantum Theory of the Solid State (Academic Press, New York, 1974).
Campbell, I. A., and A. Fert, in Ferromagnetic Materials, edited by E. P. Wohlfarth (North-
Holland, Amsterdam, 1982), Vol. 3, p. 747.
¢
Canfield, P. C., J. D. Thompson, W. P. Beyermann, A. Lacerda, M. F. Hundley, E. Peterson,
Z. Fisk, and H. R. Ott, J. Appl. Phys. , 5800 (1991). ¤
¢
Carey, R., D. M. Newman, and B. W. J. Thomas, J. Phys. D: Appl. Phys. ¡ , 2207 (1995).
Carl, A., and D. Weller, Phys. Rev. Lett. ¥ , 190 (1995). ¢
Carra, P., B. T. Thole, M. Altarelli, and X. Wang, Phys. Rev. Lett. , 694 (1993).
Cebollada, A., J. L. Martinez, J. M. Gallego, J. J. de Miguel, R. Miranda, S. Ferrer, F.
Batallan, G. Fillion, and J. P. Rebouillat, Phys. Rev. B § , 9726 (1989).
Cebollada, A., D. Weller, J. Sticht, G. R. Harp, R. F. C. Farrow, R. F. Marks, R. Savoy, and
J. C. Scott, Phys. Rev. B ¦ , 3419 (1994).
Ceperley, D. M., and B. J. Alder, Phys. Rev. Lett. ¥n¦ , 566 (1980).
Chaudouet, P., J. Roy Montreuil, J. P. Sénateur, D. Boursier, A. Rouault, and R. Fruchart,
VII Intern. Conference on Solid Compounds of Transition Elements, Grenoble, 1982.
Chen, A. B., Phys. Rev. B ¥ , 2384 (1976).
136 BIBLIOGRAPHY
Chen, C. T., Y. U. Idzerda, H.-J. Lin, N. V.¢ Smith, G. Meigs, E. Chaban, G. H. Ho, E.
Pellegrin, and F. Sette, Phys. Rev. Lett. ¦ , 152 (1995).
Chen, D., J. Appl. Phys. ¥n¡ , 3625 (1971).
Chen, D., and R. L. Aagard, J. Appl. Phys. ¥ , 2530 (1970).
Chen, D., and Y. Gondo, J. Appl. Phys. ¦ , 1024 (1964).
Chen, D., Y. Gondo, and M. D. Blue, J. Appl. Phys. , 1261 (1965).
Chen, D., J. F. Ready, and E. G. Bernal, J. Appl. Phys. § , 3916 (1968).
Chen, D., R. L. Aagard, and T. S. Liu, J. Appl. Phys. ¥ , 1395 (1970).
Chen L.-Y., and Z.-B. Su, Phys. Rev. B ¥` , 9309 (1989).
Chen, T., J. Appl. Phys. ¥`¦ , 2538 (1974).
Chen, T., and W. E. Stutius, IEEE Trans. Magn. c , 581 (1974).
Chenevier, B., D. Fruchart,¤ M. Bacmann, J. P. Senateur, P. Chaudouet, and L. Lundgren,
Phys. Status Solidi (a) ¥ , 199 (1984).
Chenevier, B., M. Anne, M. Bacmann, D. Fruchart, and P. Chaudouet, VIII Intern. Confer-
ence on Solid Compounds of Transition Elements, ¢¥ Vienna, 1985.
Chester, G. V., and A. Thellung, Proc. Phys. Soc. , 745 (1959).
Coehoorn, R., and R. A. de Groot, J. Phys. F: Metal Phys. ,¦ , 2135 (1985).
Coehoorn, R., C. Haas, and R. A. de Groot, Phys. Rev. B , 1980 (1985).
Coleridge, P. T., J. Molenaar, and A. Lodder, J. Phys. C: Solid State Phys. /¦ , 6943 (1982).
Colinet, C., and K. H. J. Buschow, Philips J. Research ¥ , 445 (1986).
Comstock, R. L., and P. H. Lissberger, J. Appl. Phys. C ¥ , 1397 (1970).

Conger, R. L., and J. L. Tomlinson, J. Appl. Phys. , S1059 (1962).
Cooper, B. R., Phys. Rev. § , A1504 (1965).
Cooper, B. R., and H. Ehrenreich, Solid State Commun. ¡ , 171 (1964).
Cooper, B. R., Q. G. Sheng, ¤ S. P. Lim, C. Sanchez-Castro, N. Kioussis, and J. M. Wills, J.
Magn. Magn. Mater. , , 10 (1992).
Cooper, B. R., S. P. Lim, I. Avgin, Q. G. Sheng, and D. L. Price, J. Phys. Chem. Solids ¦s ,
1509 (1995).
Cotton, A., and H. Mouton, Compt. Rend. c¥n¦ , 229 (1907).
Daalderop, G. H. O., F. M. Mueller, R. C. Albers, and A. M. Boring, Appl. Phys. Lett. ¦s¡ ,
1636 (1988a).
¢
Daalderop, G. H. O., F. M. Mueller, R. C. Albers, and A. M. Boring, J. Magn. Magn. Mater.
¥ , 211 (1988b).
Daalderop, G. H. O., P. J. Kelly, and M. F. H. Schuurmans, Phys. Rev. B ¥ , 11919 (1990a).
Daalderop, G. H. O., P. J. Kelly, and M. F. H. Schuurmans, Phys. Rev. B ¥`¡ , 7270 (1990b).
Daalderop, G. H. O., P. J. Kelly, and M. F. H. Schuurmans, Phys. Rev. B ¥s¥ , 12054 (1991).
Daalderop, G. H. O., P. J. Kelly, and M. F. H. Schuurmans, in Ultrathin Magnetic Structures,
edited by J. A. C. Bland and B. Heinrich (Springer, Berlin, 1994), Vol. I, p. 40.
Degiorgi, L., M. Dressel, G. Grüner, N. Sato, and T. Komatsubara, Europhys. Lett. ¡s¦ ,
311 (1994). ¢
de Groot, R. A., and K. H. J. Buschow, J. Magn. Magn. Mater. ¦¥,¨¬¦ , 1377 (1986).
de Groot, R. A., F. M. Mueller, P. G. van Engen, and K. H. J. Buschow, Phys. Rev. Lett. ¦ ,
2024 (1983).
de Groot, R. A., F. M. Mueller, P. G. van Engen, and K. H. J. Buschow, J. Appl. Phys. ¦s¦ ,
2151 (1984).
BIBLIOGRAPHY 137
Delin, A., Ph.D. Thesis, University of Uppsala (1998).

Delin, A., P. M. Oppeneer, M. S. S. Brooks, T. Kraft, B. Johansson, and O. Eriksson, Phys.
Rev. B ¦s¦ , R10173 (1997).
den Broeder, F. J. A., H. C. Donkersloot, H. J. G. Draaisma, and W. J. M. de Jonge, J. Appl.
Phys. , 4317 (1987).
den Broeder, F. J. A., D. Kuiper, A. P. van den Mosselaer, and W. Hoving, Phys. Rev. Lett.
, 2769 (1988).
den Broeder, F. J. A., D. Kuiper, H. C. Donkersloot, and W. Hoving, Appl. Phys. A ¥§ , 507
(1989).
Di, G. Q., Ph.D. Thesis, Nagoya University (1992).
Di, G. Q., private communication (1995). ¢
Di, G. Q., and S. Uchiyama, J. Appl. Phys. ¦ , 4270 (1994).
Di, G. Q., and S. Uchiyama, Phys. Rev. B ¦ , 3327 (1996).
Di, G. Q., S. Iwata, S. Tsunashima, and S. Uchiyama, J. Magn. Soc. Jpn. ,¦ , 191 (1991).
Di, G. Q., S. Iwata, S. Tsunashima, and S. Uchiyama, J. Magn. Soc. Jpn. c , 113 (1992a). ¢
Di, G. Q., S. Iwata, S. Tsunashima, and S. Uchiyama, J. Magn. Magn. Mater. ,¥/¨?, ,
1023 (1992b).
Di, G. Q., S. Iwata, and S. Uchiyama, J. Magn. Magn. Mater. , 242 (1994a). C
Di, G. Q., S. Oikawa, S. Iwata, S. Tsunashima, and S. Uchiyama, Jpn. J. Appl. Phys. ,
L783 (1994b).
Dijkstra, J., H. H. Weitering, C. F. van Bruggen, C. Haas, and R. A. de Groot, J. Phys.:
Condens. Matter , 9141 (1989).
Dolgov, O. V., and E. G. Maksimov, in The Dielectric Function of Crystalline Systems,
edited by L. V. Keldysh, D. A. Kirzhnitz, and A. A. Maradudin (Elsevier, Amsterdam,
1989) p. 221. ¢
Donovan, B., and T. Medcalf, Phys. Lett. , 303¤ (1963).
Donovan, B., and T. Medcalf, Proc. Phys. Soc. , 1179 (1965). ¢
Donovan, S., A. Schwartz, and G. Grüner, Phys. Rev. Lett. § , 1401 (1997).
Draaisma, H. J. G., and W. J. M. de Jonge, J. Appl. Phys. `¡ , 3318 (1987).
Draaisma, H. J. G., W. J. M. de Jonge, and F. J. A. den Broeder, J. Magn. Magn. Mater. ,
351 (1987).
Dreizler, R. M., and E. K. U. Gross, Density Functional Theory (Springer, Heidelberg,
1990). ¤
Dresselhaus, G., A. F. Kip, and C. Kittel, Phys. Rev. § , 368 (1955).
Drude, P., Ann. Physik , 566 (1900a).

Drude, P., Ann. Physik , 369 (1900b).
du Bois, H., Ann. Phys. Chem.
¤ , 941 (1887).
Ebert, H., Phys. Rev. B , 9390 (1988).
Ebert, H., Physica B , , 175 (1989).
Ebert, H., Habilationsschrift, University of Munich (1990).
Ebert, H., in Spin-Orbit-Influenced Spectroscopies of Magnetic Solids, edited by H. Ebert
and G. Schütz (Springer, Berlin, 1996a), p. 159.
Ebert, H., Rep. Prog. Phys. ¦s§ , 1665 (1996b).
Ebert, H., and G. Schütz, J. Appl. Phys. § , 4627 (1991).
Ebert, H., G.-Y. Guo, and G. Schütz, IEEE Trans. Magn. , 3301 (1995).
138 BIBLIOGRAPHY

Ebert, H., H. Freyer, A. Vernes, and G.-Y. Guo, Phys. Rev. B ¦ , 7721 (1996).
Egashira, K., and T. Yamada, J. Appl. Phys. ¥n¦ , 3643 (1974).
Ehrenreich, H., and M. H. Cohen, Phys. Rev. s/¦ ¤ , 786 (1959).
Ehrenreich, H., and H. R. Philipp, Phys. Rev. /¡ , 1622 (1962).
Eriksson, O., M. S. S. Brooks, and B. Johansson, Phys. Rev. B ¥ , 7311 (1990a).
Eriksson, O., M. S. S. Brooks, and `¢ B. Johansson, Phys. Rev. B ¥ , 9087 (1990b).
Erskine, J. L., Phys. Rev. ¤ Lett. , 157 (1976).
Erskine, J. L., Physica B § , 83 (1977). ¤
Erskine, J. L., and E. A. Stern, Phys. Rev. B , 1239 (1973).
Erskine, J. L., and E. A. Stern, Phys. Rev. B /¡ , 5016 (1975).
Eschrig, H., The Fundamentals of Density Functional Theory (Teubner, Stuttgart, 1995).
Eschrig, H., G. Seifert, and P. Ziesche, Solid State Commun. ¦s , 777 (1985).
Faraday, M., Phil. Trans. Roy. Soc. , 1 (1846).
Faraday, M., Faraday’s Diary (Bell and Sons, London, 1933).
Feder, R., F. Rosicky, and B. Ackermann, ¤ Z. Phys. B ¦s¡ , 31 (1983).
Feil, H., and C. Haas, Phys. Rev. Lett. ¦ , 65 (1987).
Ferguson, P. E., and R. J. Romagnoli, J. Appl. Phys. ¥ , 1236 (1969).
Feyerherm, R., A. Amato, F. N. Gygax, ¢¥ A. Schenck, C. Geibel, F. Steglich, N. Sato, and T.
Komatsubara, Phys. Rev. Lett. , 1849 (1994).
Fisk, Z., P. C. Canfield, W. P. Beyermann, J. D. Thompson, M. F. Hundley, H. R. Ott, E.
Felder,¢ M. B. Maple, M. A. Lopez de la Torre, P. Visani, and C. L. Seaman, Phys. Rev.
Lett. , 3310 (1991). ¢
Fletcher, G. C., Proc. Phys. Soc. A , 505¤ (1954).
Fowler, C. A., and E. M. Fryer, Phys. Rev. , 426 (1952).
Freiser, M. J., IEEE Trans. Magn. ¥ , 152 (1968).
Fujii, H., H. Kawanaka, T. Takabatake, M. Kurisu, ¤ Y. Yamaguchi, J. Sakurai, H. Fujiwara,
T. Fujita, and I. Oguro, J. Phys. Soc. Jpn. ¦ , 2495 (1989).
Fumagalli, P., Habilitationsschrift, RWTH Aachen (1996).
Fumagalli, P., U. Rüdiger, P. Dworak, B. Holländer, U. Nowak, and G. Güntherodt, J.
Magn. Soc. Jpn. ¡ , S1, 433 (1996).
Garcia, P. F., J. Appl. Phys. , 5066 (1988). ¢
Garcia, P. F., A. D. Meinhaldt, and A. Suna, Appl. Phys. Lett. ¥ , 178 (1985).
Garik, P., and N. W. Ashcroft, Phys. Rev. B ¡ , 391 (1980).
Gasche, T., Ph.D. Thesis, University of Uppsala (1993).
Gasche, T., M. S. S. Brooks, and B. Johansson, Phys. Rev. B ¦ , 296 (1996).
Geibel, C., S. Thies, D. Kaczorowski, A. Mehner, A. Grauel, B. Seidel, ¤C U. Ahlheim, R.
Helfrich, K. Petersen, C. D. Bredel, and F. Steglich, Z. Phys. B , 305 (1991).
Gibbs, D., D. R. Harshman, E. D. Isaacs, D. B. McWhan, D. Mills, and C. Vettier, Phys.
Rev. lett. , 1241 (1988).
Goto, T., J. Phys. Soc. Jpn. ¥ , 1848 (1977).
Gross, E. K. U., and E. Runge, Vielteilchentheorie (Teubner, Stuttgart, 1986). ¢
Grünberg, P., R. Schreiber, Y. Pang, M. B. Brodsky, and H. Sowers, Phys. Rev. Lett. ¦ ,
2442 (1986).
Guo, G.-Y., and H. Ebert, Phys. Rev. B ¦s , 10377 (1994).
Guo, G.-Y., and H. Ebert, Phys. Rev. B ¦ , 12633 (1995).
BIBLIOGRAPHY 139
¢¥
Guo, X., X. Chen, Z. Altounian, and J. O. Str öm-Olsen,¤s¤ J. Appl. Phys. , 6275 (1993).
Halilov, S. V., and R. Feder, Solid State Commun. , 749 (1993b).
Halpern, O., Ann. Physik ,¡ , 181 (1932). ¤
Hansch, W., and G. D. Mahan, Phys. Rev. B ¡ , 1902 (1983).
Hanssen, K. E. H. M., P. E. Mijnarends, L. P. L. M. Rabou, and K. H. J. Buschow, Phys.
Rev. B ¥`¡ , 1533 (1990). £
Harada, T., T. Kanomata, and T. Kaneko, J. Magn. Magn. Mater. § § ¤ , 169 (1990).
Harder, K.-U., D. Menzel, T. Widmer, and J. Schoenes, J. Appl. Phys. ¥ , 3625 (1998).
¢ G. R., D. Weller, T. A. Rabedeau, R. F. C. Farrow, and M. F. Toney, Phys. Rev. Lett.
Harp,
, 2493 (1993).
Hartmann, M., and T. R. McGuire, Phys. Rev. Lett. ¦n , 1194 (1983).
Hasegawa, A., J. Phys. Soc. Jpn. ¢ ¦¥ , 1477 (1985).
Hashimoto, S., J. Appl. Phys. ¦ , 438 (1994).
Havela,£ L., V. Sechovský, K. Prokeš, H. Nakotte, H. Fujii, and A. Lacerda, Physica B
¡¡ ¡s¡¥ , 245 (1996).
Heinemann, M., and W. M. Temmerman, Phys. Rev. B ¥§ , 4348 (1994).
Hedin, L., and B. I. Lundqvist, J. Phys. C: Solid State Phys. ¥ , 2064 (1971).
Hedin, L., and S. Lundqvist, Solid State Phys. ¡ , 1 (1969).
Helmholdt, R. B., R. A. de Groot, F. M. Mueller, P. G. van Engen, and K. H. J. Buschow, J.
Magn. Magn. Mater. ¥ , 249 (1984).
Henkie, Z., R. Fabrowski, and A. Wojakowski, J. Alloys Compnds. ¡¢ c§ , 248 (1995).
Hiebert, W. K., A. Stankiewicz, and M. R. Freeman, Phys. Rev. Lett. § , 1134 (1997).
Hiess, A., L. Havela, K. Prokeš, R. S. Eccleston, and G. H. Lander, Physica B ¡ c¨¬¡ ¡ ,
89 (1997).
Hillebrecht, F. U., T. Kinoshita,¢ D. Spanke, J. Dresselhaus, C. Roth, H. B. Rose, and E.
Kisker, Phys. Rev. Lett. ¦ , 2224 (1995). ¢
Höchst, H., K. Tan, and K. H. J. Buschow, J. Magn. Magn. Mater. ¦¥,¨¬¦ , 545 (1986).
Hohenberg, P., and W. Kohn, Phys. Rev. , B864 (1964).
Holland-Moritz, E., and G. H. Lander, in Handbook on the Physics and Chemistry of Rare
Earths, edited by K. A. Gschneidner, Jr., L. Eyring, G. H. Lander, and G. R. Choppin
(North-Holland, Amsterdam, 1994), Vol. 19, p. 1. ¢
Huang, D., X. W. Zhang, C. P. Luo, H. S. Yang, and Y. J. Wang, J. Appl. Phys. ¦ , 6351
(1994).
Hubert, A., and R. Schäfer, Magnetic Domains – The Analysis of Magnetic Microstructures
(Springer, Berlin, 1998).
Hulliger, F., J. Less-Common Metals , , 113 (1968).
Hulme, H. R., Proc. Roy. Soc. ¤ (London) A ¦ , 237 (1932).
Hunt, R. P., J. Appl. Phys. , 1652 (1967).
Hybertsen, M. S., and S. G. Louie, Phys. Rev. Lett. ¦¦ , 1418 (1985).
Ide, Y., T. Goto, K. Kikuchi,¢s¢ K. Watanabe,
¤ J. Onegawa, H. Yoshida, and J. M. Cadogan, J.
Magn. Magn. Mater.¢ ¨¯ , 1245 (1998).
Ido, H., J. Appl. Phys. ¦ , 3247 (1985).
Ikekame, H., K. Sato, K. Takanashi, and H. Fujimori, Jpn. J. Appl. Phys. ¡ , Suppl. 32-3,
284 (1993).
140 BIBLIOGRAPHY
Ikekame, H., Y. Yanase, M. Akita, Y. Morishita, and K. Sato, J. Magn. Soc. Jpn. ¡ , S1,
153 (1996).
Ingersoll, L. R., Phys. Rev. ¦ , 312 (1912).
Inukai, T., N. Sugimoto, M. Matsuoka, and K. Ono, Appl. Phys. Lett. ¥§ , 52 (1986).
Iwashita, K., T. Oguchi, and T. Jo, Phys. Rev. B ¦/¥ , 1159 (1996).C
Iwata, S., S. Yamashita, and S. Tsunashima, IEEE Trans. Magn. , 3670 (1997).
Jackson, J. D., Classical Electrodynamics (J. Wiley, New York, 1975).
Jaggi, R., S. Methfessel, and¢ R. Sommerhalder, in Landolt-B örnstein II, Vol. 9, p. 1-180.
Jahoda, F. C., Phys. Rev. , , 1261 (1957).
Jasperson, S. N., and S. E. Schnatterly, Rev. Sci. Instrum. ¥` , 761 (1969).
Jepsen, O., and O. K. Andersen, Solid State Commun. § , 1763 (1971).
Jesser, R., A. Bieber, and R. Kuentzler, J. Phys. (France) ¥`¡ , 1157 (1981).
Johansson, B., O. Eriksson, M. S. S. Brooks, and H. L. Skriver, Phys. Scripta T , 65
(1986).
Jones, R. O., and O. Gunnarsson, Rev. Mod. Phys. , 689 (1989).
Jones, W., and N. H. March, Theoretical Solid State Physics (Dover, New York, 1973)
Vol. 2.
Ju, G., A. Vertikov,¢ A. V. Nurmikko, C. Canady, G. Xiao, R. F. C. Farrow, and A. Cebollada,
Phys. Rev. B ¦ , R700 (1998). ¢
Kaburagi, M., and T. Tonegawa, J. Magn. Magn. Mater. , 408 (1987). £
Kaburagi, M., T. Tonegawa, and K. Ebina, J. Magn. Magn. Mater. § § , 161 (1990).
¤
Kaczorowski, D., R. Troć, D. Badurski, A. Böhm, L. Shlyk, and F. Steglich, Phys. Rev. B
¥ , 16425 (1993). ¤
Kahn, F. J., P. S. Pershan, and J. P. Remeika, Phys. Rev. , 891 (1969). ¤
Kanomata, T., K. Shirakawa, H. Yasui, and T. Kaneko, J. Magn. Magn. Mater. , 286
(1987).
Kao, C., J. B. Hastings, E. D. Johnson, D. P. Siddons, G. C. Smith, and G. A. Prinz, Phys.
Rev. Lett. `¦ , 373 (1990).
Kaplan, S. G., S. Wu, H. T.¢ S. Lihn, H. D. Drew, Q. Li, D. B. Fenner, J. M. Philips, and S. Y.
Hou, Phys. Rev. Lett. , 696 (1996).
Katayama, T., Y. Suzuki, H. Awano, Y. Nishihara, and K. Koshizuka, Phys. Rev. Lett. s ,
1426 (1988).
Katayama, T., Y. Suzuki, Y. Nishihara, T. Sugimoto, and M. Hashimoto, J. Appl. Phys. § ,
5658 (1991).
Katayama, T., T. Sugimoto, Y. Suzuki, ¢ M. Hashimoto, P. de Haan, and J. C. Lodder, J.
Magn. Magn. Mater. ,s¥,¨¯c , 1002 (1992).
Katayama, T., N. Nakajima, N. Okusawa, ¢Y.¢ Miyauchi, ¤ T. Koide, T. Shidara, Y. Suzuki,
and S. Yuasa, J. Magn. Magn. Mater. ¨? , 1251 (1998).
Kato, Y., Ph.D. Thesis, Nagoya University (1997).
Kato, T., S. Iwata, S. Tsunashima, and S. Uchiyama, J. Magn. Soc. Jpn. ,§ , 205 (1995a).
Kato, T., H. Kikuzawa, S. Iwata, S. Tsunashima, and S. Uchiyama, J. Magn. Magn. Mater.
c¥,¨¯¥¥ , 713 (1995b).
Kato,¢sT.,¢ S.¤ Iwata, M. Yasui, K. Fukawa, and S. Tsunashima, J. Magn. Magn. Mater.
¨¯ , 1427 (1998).
Kautzky, M. C., and B. M. Clemens, Appl. Phys. Lett. s , 1279 (1995).
BIBLIOGRAPHY 141
Kawakami, M., and T. Goto, J. Phys. Soc. Jpn. ¥ , 1492 (1979).

Kelly, P. J., and M. S. S. Brooks, Physica B c`¡ , 81 (1980).
Kern, S., C. K. Loong, and G. H. Lander, Phys. Rev. B ¡ , 3051 (1985).

Kerr, J., Rept. Brit. Assoc. Adv. Sci. p. 40 (1876).
Kerr, J., Phil. Mag. , 321 (1877).
Kerr, J., Phil. Mag. ¦ , 161 (1878).
Kielar, P., J. Opt. Soc. Am. B s , 854 (1994).
Kim, K. J., T. C. Leung, B. N. Harmon, and D. W. Lynch, J. Phys.: Condens. Matter ,
5069 (1994).
¢
Kirby, R. D., J. X. Shen, J. A. Woollam, and D. J. Sellmyer, J. Appl. Phys. § , 4574 (1991).
Kisker, E., C. Carbone, C. F. Flipse, and E. F. Wassermann, J. Magn. Magn. Mater. , 21
(1987). ¤C
Kittel, C., Phys. Rev. , 208A (1951).
Kleiner, W. H., Phys. Rev. c¥n¡ , 318 (1966).
Koelling, D. D., and B. N. Harmon, J. Phys. C: Solid State Phys. , , 3107 (1977).
Köhler, J., Ph.D. Thesis, Technical University Darmstadt ¤ (1997).
Köhler, J., and J. Kübler, J. Phys.: Condens. Matter , 8681 (1996).
Köhler, J., L. M. Sandratskii, and J. Kübler, Phys. Rev. B ¦s¦ , R10153 (1997).
Kohn, W., and N. Rostoker, Phys. Rev. §¥ , 1111 (1954).
Kohn, W., and L. Sham, Phys. Rev. ¥` , A1133 (1965).
Kohn, W., and P. Vashista, in Theory of the inhomogeneous electron gas, edited by S.
Lundqvist and N. H. March (Plenum, New York, 1983), p. 79. ¤
Kokuryu, M., T. Kato, S. Iwata, and T. Tsunashima, J. Appl. Phys. , 4779 (1997).
Korringa, J., Physica , 392 (1947).
Kouvel, J. S., and J. B. Forsyth, J. Appl. Phys. ¥` , 1359 (1969).
Kraft, T., Ph.D. Thesis, University of Technology Dresden (1997).
Kraft, T., P. M. Oppeneer, V. N. Antonov,¢ and H. Eschrig, Phys. Rev. B ¦¡ , 3561 (1995).
Kramers, H. A., Nature (London) s , 775 (1926).
Kramers, H. A., and W. Heisenberg, Z. Phys. , 681 (1925).
Krinchik, G. S., J. Appl. Phys. ¦ , 1089 (1964).
Krinchik, G. S., and R. D. Nuralieva, Sov. Phys. JETP , 724 (1959).
Krinchik, G. S., and V. A. Artem’ev, Sov. Phys. JETP ¡s , 1080 (1968a).
Krinchik, G. S., and V. A. Artem’ev, J. Appl. Phys. § , 1276 (1968b).
Krinchik, G. S., and V. S. Gushchin, Sov. Phys. JETP ¡s§ , 984 (1969).
Kronig, R. de L., J. Opt. Soc. Am. ,¡ , 547 (1926).
Krutzen, B. C. H., and F. Springelkamp, J. Phys.: Condens. Matter , 8369 (1989).
Kübler, J., J. Magn. Magn. Mater. ¥`¦ , 415 (1984).
Kübler, J., J. Phys. Chem. Solids ¦s , 1529 (1995).
Kubo, R., Can. J. Phys. ¥ , 1274 (1956).
Kubo, R., J. Phys. Soc. Jpn. /¡ , 570 (1957).
¢ M.,¢¥P. Beránková, M. Matyáš, I. Tichý, and A. A. Menovsky, J. Alloys Compnds.
Kučera,
¡ x¨¬¡ , 467 (1998).
Kuhn, W., Z. Phys. Chem. B ¥ , 14 (1929).
Kulatov, E. T., Y. A. Uspenskii, and S. V. Halilov, Phys. Lett. A c§`¦ , 267 (1994).

Kulatov, E. T., Y. A. Uspenskii, and S. V. Halilov, J. Magn. Magn. Mater. c , 331 (1996).
Kundt, A., Ann. Phys. Chem. ¡ , 228 (1884).
142 BIBLIOGRAPHY
Kunz, A. B., and C. P. Flynn, Phys. Rev. Lett. ¦s , 1524 (1983).
Kyuno, K., R. Yamamoto, and S. Asano, J. Phys. Soc. Jpn. , 2099 (1992).
Lairson, B. M., and B. M. Clemens, Appl. Phys. Lett. , 1438 (1993).
Landau, L. D., and E. M. Lifshitz, Electrodynamics in Continuous Media (Pergamon, New
York, 1960).
Lander, G. H., and P. Burlet, Physica B ¡/¦ , 7 (1995).
Lander, G. H., W. G. Stirling, J. M. Rossat-Mignod, M. Hagen, and O. Vogt, Phys. Rev. B
¥¬ , 6899 (1990).
Lander, G. H., M. S. S. Brooks, B. Lebech, P. J. Brown, O. Vogt, and K. Mattenberger, J.
Appl. Phys. § , 4803 (1991).
Lax, B., and S. Y. Nishina, Phys. Rev. Lett. , 464 (1961). ¤
Lax, B., L. M. Roth, and S. Zwerdling, J. Phys. Chem. Solids , 311 (1959).
Lebech, B., M. Wulff, G. H. Lander, J. Rebizant, J. C. Spirlet, and A. Delapalme, J. Phys.:
Condens. Matter , 10299 (1989).
Leciejewicz, J., and A. Zygmunt, Phys. Status Solidi (a) , 657 (1972).
Lehmann, G., and M. Taut, Phys. Status Solidi (b) ¦¥ , 469 (1972).
Levy, M., Phys. Rev. A ¡s , 1200 (1982).
Lewis, B. G., and D. E. G. Williams, J. Less-Common Metals ¥¥ , 337 (1976).
Lieb, E. H., Int. J. Quantum Chem. ¡¥ , 243 (1983).
Liechtenstein, A. I., V. P. Antropov, and B. N. Harmon, Phys. Rev. B ¥`§ , 10770 (1994).¢
Lihn, H.-T. S., S. Wu, S. Kaplan, H. D. Drew, Q. Li, and D. B. Fenner, Phys. Rev. Lett. ,
3810 (1996). ¤
Lim, S. P., J. Appl. Phys. , 5426 (1997). ¤ ¤s¤
Lim, S. P., B. R. Cooper, Q. G. Sheng, and D. L. Price, Physica B c¨¯ , 56 (1993).
Lin, S. T., and R. E. Ogilvie, J. Appl. Phys. ¥ , 1372 (1963).
Lissberger, P. H, J. Opt. Soc. Am. ¦ , 948 (1961a).
Lissberger, P. H, J. Opt. Soc. Am. ¦ , 957 (1961b).

Liu, C., E. R. Moog, and S. D. Bader, Phys. Rev. Lett. s , 2422 (1988).
Liu, C., and S. D. Bader, J. Magn. Magn. Mater. § , 307 (1991).
Lorentz, H. A., Arch. Néerl. c§ , 123 (1884).
Loucks, T.,L., Augmented Plane Wave Method (Benjamin, ¢ New York, 1967).
Lu, Z. W., B. M. Klein, and A. Zunger, Phys. Rev. Lett. ¦ , 1320 (1995).
MacDonald, A. H., and S. H. Vosko, J. Phys. C: Solid State Phys. /¡ , 2977 (1979).
MacDonald, A. H., W. E. Pickett, and D. D. Koelling, J. Phys. C: Solid State Phys. ,
2675 (1980).
Mackintosh, A. R., and O. K. Andersen, in Electrons at the Fermi Surface, edited by M.
¢
Springford (Cambridge University Press, Cambridge, 1980).
MacLaren, J. M., and W. Huang, J. Appl. Phys. § , 6196 (1996).
Mainkar, N., D. A. Browne, and J. Callaway, Phys. Rev. B ¦ , 3692 (1996).
Majorana, Q., Rend. Acca. Lincei Atti, Roma , 90 (1902).
Majorana, Q., Nuovo Cimento ¡ , 1 (1944).
¤
Maksimov, E. G., I. I. Mazin, S. N. Rashkeev, and Y. A. Uspenski, J. Phys. F: Metal Phys.
, 833 (1988).
Mansuripur, M., The Physical Principles of Magneto-Optical Recording (Cambridge Uni-
versity Press, Cambridge, England, 1995).
BIBLIOGRAPHY 143
¤
Martin, D. H., K. F. Neal, and T. J. Dean, Proc. Phys. Soc. , 605 (1965). ¢
Matthias, B. T., R. M. Bozorth, and J. H. van Vleck, Phys. Rev. Lett. , 160 (1961).
Maurer, T., Masters Thesis, Technical University Darmstadt (1991). ¤s£
Maruyama, H., F. Matsuoka, K. Kobayashi, and H. Yamazaki, Physica B ¡ ¡ss§ , 787
(1995a).
Maruyama, H., F. Matsuoka, K. Kobayashi, and H. Yamazaki, J. Magn. Magn. Mater.
c¥,¨¯¥¥ , 43 (1995b).
Mayer, L., J. Appl. Phys. ¡§ , 1454 (1958).
McCurrie, R. A., and P. Gaunt, Philos. Mag. , 567 (1966). ¤ ¤¤
McEwen, K. A., U. Steigenberger, and J. L. Martinez, Physica B ,¨? , 670 (1993).
Mentink, S. A. M., A. Drost, ¢¥G. J. Nieuwenhuys, E. Frikkee, A. A. Menovsky, and J. A.
Mydosh, Phys. Rev. Lett. , 1031 (1994).
Menzinger, F., and A. Paoletti, Phys. Rev. ¥ , 365 (1966).
Mermin, N. D., Phys. Rev. B , 2362 (1970). ¢
¢ Phys. Rev. Lett. § , 3062 (1997).
Milton, G. W., D. J. Eyre, and J. V. Mantese,
Misemer, D. K., J. Magn. Magn. Mater. ¡ , 267 (1988).
Mitani, S., K. Takanashi, H. Nakajima, K. Sato, R. Schreiber, P. Gr ünberg, and H. Fujimori,
J. Magn. Magn. Mater. ,¦s , 7 (1996). ¢
Moodera, J. S., and D. M. Mootoo, J. Appl. Phys. , 6101 (1994).
¢
Morisin, B., and N. Narath, J. Chem. Phys. ¥ , 1958 (1964).
Motizuki, K., J. Magn. Magn. Mater. , 1 (1987).
Murray, A. F., and W. J. L. Buyers, in Crystalline Electric Fields and Structural Effects in
f-electron Systems, edited by J. E. Crow, R. P. Guertin, and T. W. Mihalisin (Plenum,
New York, 1980), p. 257.
Naegele, J. R., Physica B , 52 (1985).
Naoe, M., N. Kitamura, M. Shoji, and A. Nagai, J. Appl. Phys. , 3636 (1988).
Nemoshkalenko, V. V., A. E. Krasovskii, V. N. Antonov, Vl. N. Antonov, U. Fleck, H.
Wonn, and P. Ziesche, Phys. Status Solidi (b) /¡s , 283 (1983).
Niess, R., and F. R. Kessler, Phys. Status Solidi (a) s , 639 (1989).
Nieuwenhuys, G. J., in Handbook of Magnetic Materials, edited by K. H. J. Buschow
(North-Holland, Amsterdam, 1995), Vol. 9, p. 1.
Nozières P., and D. Pines, Phys. Rev. c§ , 741 (1958).
Nylund, M. A., A. Roger, J. P. Sénateur, and R. Fruchart, J. Solid State Chem. ¥ , 115
(1972).
Oppeneer, P. M., unpublished (1995). ¤¤
Oppeneer, P. M., J. Magn. Magn. Mater. , 275 (1998).
Oppeneer, P. M., and V. N. Antonov, in Spin-Orbit-Influenced Spectroscopies of Magnetic
Solids, edited by H. Ebert and G. Schütz (Springer, Berlin, ¢ 1996), p. 29.
Oppeneer, P. M., and A. Lodder, J. Phys. F: Metal Phys. , 1885 (1987).
Oppeneer, P. M., J. Sticht, and F. Herman, J. Magn. Soc. Jpn. /¦ , S1, 73 (1991).
Oppeneer, P. M., T. Maurer, J. Sticht, and J. Kübler, Phys. Rev. B ¤s¤ ¥n¦ , 10924 (1992a).
Oppeneer, P. M., J. Sticht, T. Maurer, and J. Kübler, Z. Phys. B , 309 (1992b).
Oppeneer, P. M., V. N. Antonov, T. Kraft, H. Eschrig, A. N. Yaresko, and A. Y. Perlov,
Solid State Commun. §s¥ , 255 (1995a).
Oppeneer, P. M., T. Kraft, and H. Eschrig, Phys. Rev. B ¦¡ , 3577 (1995b).
144 BIBLIOGRAPHY
Oppeneer, P. M.,¤ V. N. Antonov, T. Kraft, H. Eschrig, A. N. Yaresko, and A. Y. Perlov, J.

Appl. Phys. , 1099 (1996a).
Oppeneer, P. M., V. N. Antonov,¤ T. Kraft, H. Eschrig, A. N. Yaresko, and A. Y. Perlov, J.
Phys.: Condens. Matter , 5769 (1996b).
Oppeneer, P. M., V. N. Antonov, A. N. Yaresko, A. Y. Perlov, T. Kraft, and H. Eschrig, J.
Magn. Soc. Jpn. ¡ , S1, 47 (1996c).

Oppeneer, P. M., M. S. S. Brooks, V. N. Antonov, T. Kraft, and H. Eschrig, Phys. Rev. B
¦ , R10437 (1996d).
Oppeneer, P. M., V. N. Antonov, A. Y. Perlov, A. N. Yaresko, T. Kraft, and H. Eschrig,
Physica B ¡ c¨¬¡ ¡ , 544 (1997).
¢ V.¢+N.
Oppeneer, P. M., A. Y. Perlov, Antonov, A. N. Yaresko, T. Kraft, and M. S. S. Brooks,
J. Alloys Compnds. ¡ E¨¬¡ , 831 (1998).
Osterloh, I., Masters Thesis, Technical University Darmstadt (1993).
Osterloh, I., P. M. Oppeneer, J. Sticht, and J. K übler, J. Phys.: Condens. Matter , 285
(1994).
Ott, H. R., H. Rudigier, Z. Fisk, and J. L. Smith, Phys. Rev. Lett. ¦s¡ , 1595 (1983).
Ott, H. R., H. Rudigier, P. Delsing, and Z. Fisk, Phys. Rev. Lett. ¦s¡ , 1551 (1984).
Otto, M. J., H. Feil, R. A. M. van Woerden, ¢ J. Wijngaard, P. J. van der Valk, C. F. Bruggen,
and C. Haas, J. Magn. Magn. Mater. , 33 (1987).
Paixão, J. A., G. H. Lander, P. J. Brown, H. Nakotte, F. R. de Boer, and E. Br ück, J. Phys.:
Condens. Matter ¥ , 829 (1992).
Paixão, J. A., G. H. Lander, A. Delapalme, H. Nakotte, F. R. de Boer, and E. Br ück,
Europhys. Lett. ¡/¥ , 607 (1993).
Palik, E. D., and B. W. Henvis, Appl. Optics , 603 (1967).
Palstra, T. T. M., A. A. Menovsky, J. van den Berg, A. J. Dirkmaat, P. H. Kes, G. J. Nieuwen-
huys, and J. A. Mydosh, Phys. Rev. Lett. ¦s¦ , 2727 (1985).
¢
Palstra, T. T. M., G. J. Nieuwenhuys, R. F. M. Vlastuin, J. van den Berg, J. A. Mydosh, and
K. H. J. Buschow, J. Magn. Magn. Mater. , 331 (1987).
Pan, R.-P., H. D. Wei, and Y. R. Shen, Phys. Rev. B § , 1229 (1989).
Paolasini, L., G. H. Lander, S. M. Shapiro, R. Caciuffo, B. Lebech, L.-P. Regnault, B.
Roessli, and J.-M. Fournier, Phys. Rev. B ¦/¥ , 7222 (1996).
Parkin, S. S. P., N. More, and K. P. Roche, Phys. Rev. Lett. s¥ , 2304 (1990).
s¤ Vogt, Phys. Status Solidi (a) , 101 (1972).
Pebler, J., F. W. Richter, and E.
Pershan, P. S., J. Appl. Phys. , 1482 (1967). C
Pickart, S. J., and R. Nathans, J. Appl. Phys. , S-1336 (1962).
Pickart, S. J., and R. Nathans, J. Appl. Phys. ¥ , 1203 (1963). ¢¢
Pittini, R., J. Schoenes, O. Vogt, and P. Wachter, Phys. Rev. Lett. ,¢s944 ¤ (1996).

Pittini, R., J. Schoenes, F. Hulliger, and P. Wachter, Phys. Rev. Lett. , 725 (1997).
Podgorny, M., Phys. Rev. B ¥ , 11300 (1991). ¢
Pustogowa, U., T. A. Luce, W. Hübner, and K. H. Bennemann, J. Appl. Phys. § , 6177
(1996). ¢
Rajagopal, A. K., and J. Callaway, Phys. Rev. B , 1912 (1973).
Rasing, T., in Nonlinear Optics in Metals, edited by K. H. Bennemann (Clarendon, Oxford,
1998), p. 132.
BIBLIOGRAPHY 145
¢
Rasing, T., M. Groot Koerkamp, B. Koopmans, and H. van den Berg, J. Appl. Phys. § ,
6181 (1996). ¢
Reif, J., J. C. Zink, C. M. Schneider, and J. Kirschner, ¢ Phys. Rev. Lett. , 2878 (1991).
Reif, J., C. Rau, and E. Matthias, Phys. Rev. Lett. , 1931 (1993).
Reihl, B., N. Mårtensson, P. Heimann, D. E. Eastman, and O. Vogt, Phys. Rev. Lett. ¥` ,
1490 (1981).
Reihl, B., N. Mårtensson, and O. Vogt, J. Appl. Phys. ¦ , 2008 (1982a).
¤
Reihl, B., G. Hollinger, and F. J. Himpsel, J. Magn. Magn. Mater. ¡s§ , 303 (1982b).
Reim, W., J. Magn. Magn. Mater. ¦ , 1 (1986).
Reim, W., and J. Schoenes, in Ferromagnetic Materials, edited by K. H. J. Buschow and
E. P. Wohlfarth (North-Holland, Amsterdam, 1990), Vol. 5, p. 133.
Reim, W., O. E. Hüsser, J. Schoenes, E. Kaldis, and P. Wachter, J. Appl. Phys. ¦s¦ , 2155
(1984).
Reim, W., J. Schoenes, and F. Hulliger, Physica B , 64 (1985).
Reim, W., H. Brändle, D. Weller, and J. Schoenes, J.¢ Magn. Magn. Mater. § , 220 (1991).
Richter, M., and H. Eschrig, Solid State Commun. ¡ , 263 (1989).
Rillo, C., J. Bartolomé, M. ¢ Bacmann, B. Chenevier, D. Fruchart, and R. Fruchart, J. Magn.
Magn. Mater. cs¥,¨¯, , 1995 (1992).
Roberts, B. W., Phys. Rev. cs¥ , 607 (1956).
Robinson, C. C., J. Opt. Soc. Am. ¦ , 681 (1963). ¢
Rooney, P. W., A. L. Shapiro, M. Q. Tran, and F. Hellman, Phys. Rev. Lett. ¦ , 1843 (1995).
Roosengaard, N. M., and H. L. Skriver, Phys. Rev. B ¦s , 4848 (1994).
Rose, M. E., Relativistic Electron Theory (J. Wiley, New York, 1961). ¢
Roth, C., F. U. Hillebrecht,C H. B. Rose, and E. Kisker, Phys. Rev. Lett. , 3479 (1993).
Roth, L. M., Phys. Rev. , A542 (1964).
¢s¤ U., H. Berndt, A. Schirmeisen, P. Fumagalli, and G. Güntherodt, J. Appl. Phys.
Rüdiger,
, 5391 (1995).
¤
Rüdiger, U., P. Fumagalli, H. Berndt, A. Schirmeisen, G. G üntherodt, and B. Holländer, J.
Appl. Phys. , 196 (1996a).
Rüdiger, U., P. Fumagalli, P. Dworak, and G. G üntherodt, IEEE Trans. Magn. ¡ , 4090
(1996b).
Rudigier, H., H. R. Ott, and O. Vogt, Phys. Rev. B ¦s¡ , 4584 (1985).
Sabiryanov, R. F., and S. S. Jaswal, Phys. Rev. B ¦ , 313 (1996).
Sakuma, A., J. Phys. Soc. Jpn. , 3053 (1994).
Sanchez, J. M., J. L. Morán-López, C. Leroux, and M. C. Cadeville, J. Phys.: Condens.
Matter , 491 (1989).

Sandratskii, L. M., and J. Kübler, Solid State Commun. § , 183 (1994).
Santoni, A., and F. J. Himpsel, Phys. Rev. B ¥ , 1305 (1991).
Sato, K., Jpn. J. Appl. Phys. ¡s¢ , 2403 (1981).
Sato, K., J. Magn. Soc. Jpn. , S1, 11 (1993).
Sato, K., H. Hongu, H. Ikekame, J. Watanabe, K. Tsuzukiyama, Y. Togami, M. Fujisawa,
and T. Fukazawa, Jpn. J. Appl. Phys. , 3603 (1992a).
Sato, K., H. Ikekame, H. Hongu, M. Fujisawa, K. Takanashi, and H. Fujimori, in Pro-
ceedings of the Sixth International Conference on Ferrites (Jpn. Society of Powder and
Powder Metallurgy, Tokyo, 1992b), p. 1647.
146 BIBLIOGRAPHY
Sato, K., H. Hongu, H. Ikekame, Y. Tosaka, M. Watanabe, K. Takanashi, and H. Fujimori,

Jpn. J. Appl. Phys. ¡ , 989 (1993a).
Sato, K., H. Ikekame, Y. Tosaka, and S.-C. Shin, J. Magn. Magn. Mater. ,¡s , 553 (1993b).
Sato, K., Y. Tosaka, and H. Ikekame, J. Magn. Soc. Jpn. ,§ , S1, 255 (1995).
Schadler, G., R. Albers, A. Boring, and P. Weinberger,¤ Phys. Rev. B ¦ , 4324 (1987).
Schäfer, R., and A. Hubert, Phys. ¤C Status Solidi (a) , 271 (1990).
Schnatterly, S. E., Phys. Rev. , 664 (1969).
Schoenes, J., in Handbook on the Physics and Chemistry of the Actinides, edited by A. J.
Freeman and G. H. Lander (North-Holland, Amsterdam, 1984), Vol. 1, p. 341.
Schoenes, J., in Materials Science and Technology, Vol. 3A: Electronic and Magnetic
Properties of Metals and Ceramics (volume editor: K. H. J. Buschow), edited by R. W.
Cahn, P. Haasen and E. J. Kramer (Verlag Chemie, Weinheim, 1992), p. 147.
Schoenes, J., and R. Pittini, J. Magn. Soc. Jpn. ¤ ¡ , S1, 1 (1996).
Schoenes, J., and R. Pittini, J. Appl. Phys. , 4853 (1997).
Schoenes, J., and W. Reim, Phys. Rev. Lett. , 1988 (1988). ¤
Schoenes, J., W. Bacsa, and F. Hulliger, Solid State Commun. , 287 (1988).
Schoenes, J., O. Vogt, J. Löhle, F. Hulliger, and K. Mattenberger, Phys. Rev. B ¦ , 14987
(1996).
¤
Schütz, G., W. Wagner, W. Wilhelm, P. Kienle, R. Zeller, R. Frahm, and G. Materlik, Phys.
Rev. Lett. ¦ , 737 (1987).
¢ G, R. Wienke, W. Wilhelm, W. B. Zeper, H. Ebert, and K. Spörl, J. Appl. Phys.
Schütz,
, 4456 (1990).
Sechovský, V., and L. Havela, in Ferromagnetic Materials, Vol. 4, edited by E. P. Wohlfarth
and K. H. J. Buschow (North-Holland, Amsterdam, 1988), p. 309.
Sellmyer, D. J., R. D. Kirby, J. Chen, K. W. Wierman, J. X. Shen, Y. Liu, B. W. Robertson,
and S. S. Jaswal, J. Phys. Chem. Solids ¦ , 1549 (1995). ¢
Severin, L., M. S. S. Brooks, and B. Johansson, Phys. Rev. Lett. , 3214 (1993).
Sham, L. J., and M. Schlüter, Phys. Rev. Lett. ¦ , 1888 (1983).
Shamir, N., M. Melamud, H. Shaked, and M. Weger, Physica B §¥ , 225 (1978).
Shang, C. H., J. Phys. D: Appl. Phys. ¡§ , 277 (1996).
Shang, C. H., Z. H. Guo, Y. Xu, and Y. J. Wang, in Proceedings of the Third International
Symposium on the Physics of Magnetic Materials, Seoul, Korea, p. 515 (1995). ¤
Shang, C. H., Y. J. Wang, L. Y. Chen, H. Zhang, and J. P. Liu, J. Appl. Phys. , 5662
(1997). C
Shen, Y. R., Phys. Rev. , A511 (1964).
Shen, Y. R., The Principles of Nonlinear Optics¥(J. Wiley, New York, 1984).
Shen, Y. R., and N. Bloembergen, Phys. Rev. , A515 (1964).
Shiomi, S., A. Ito, and M. Masuda, J. Magn. Soc. Jpn. , S1, 221 (1987).
Shirai, M., H. Maeshima, and N. Suzuki, J. Magn. Magn. Mater. c¥`c¨¯¥¥ , 105 (1995).
Singh, M., C. S. Wang, and J. Callaway, Phys. Rev. B , 287 (1975).
Slater, J. C., Phys. Rev. ¤¦n , 846 (1937).
Slater, J. C., Phys. Rev. , 385 (1951).
Smith, D. Y., J. Opt. Soc. Am. s , 454 (1976a).
Smith, D. Y., J. Opt. Soc. Am. s , 547 (1976b).
Smith, D. Y., Phys. Rev. B , 5303 (1976c).
BIBLIOGRAPHY 147
¢
Smith, D. Y., and E. Shiles, Phys. Rev. B , 4689 (1978). ¤
Solal, F., R. Caudron, F. Ducastelle, A. Finel, and A. Loiseau, Phys. Rev. Lett. ¦ , 2245
(1987).
¢
Solovyev, I. V., P. H. Dederichs, and I. Mertig, Phys. Rev. B ¦¡ , 13419 (1995).
Sols, F., Phys. Rev. Lett. , 2874 (1991).
¢¥ D. T. Nemeth, F. Ludwig, J. Clarke, P. Merchant, and
Spielman, S., B. Parks, J. Orenstein,
D. J. Lew, Phys. Rev. Lett. , 1537 (1994).
Stewart, G. R., Z. Fisk, J. O. Willis, and J. L. Smith, Phys. Rev. Lett. ¦¡ , 679 (1984).
Sticht, J., Ph.D. Thesis, Technical University Darmstadt (1989). ¢
Strange, P., J. B. Staunton, and B. L. Gyorffy, J. Phys. C: Solid State Phys. , 3355 (1984).
Sugimoto, T., T. Katayama, Y. ¤ Suzuki, T. Koide, T. Sidara, M. Yuri, A. Itoh, and K.
Kawanishi, Phys. Rev. B ¥ , 16432 (1993).
Suits, J. C., Rev. Sci. Instrum. ¥n¡ , 19 (1971).
Supernowicz, E. J., J. Appl. Phys. ¥ , 1110 (1963).
Suzuki, T., S. Iwata, H. Brändle, and D. Weller, IEEE Trans. Magn. , 4455 (1994). ¤
Suzuki, Y., T. Katayama, S. Yoshida, T. Tanaka, and K. Sato, Phys. Rev. Lett. , 3355
(1992).
Suzuki, Y., T. Katayama, A. Thiaville, K. Sato, M. Taninaka, and S. Yoshida, J. Magn.
Magn. Mater. ,¡ , 539 (1993). ¤
Suzuki, Y., T. Katayama, P. Bruno, S. Yuasa, and E. Tamura, Phys. Rev. Lett. , 5200
(1998).
Takahashi, M., H. Shoji, Y. Hozumi, and T. Wakiyama, J. Magn. Magn. Mater. , 67
(1994). ¢
Takanashi, K., H. Fujimori, J. Watanabe, M. Shoji, and A. Nagai, Jpn. J. Appl. Phys. ¡ ,
L2351 (1988).
Takanashi, K., J. Watanabe, G. Kido, and H. Fujimori, Jpn. J. Appl. Phys. ¡§ , L306 (1990).
Takanashi, K., K. Sato, J. Watanabe, Y. Sato, and H. Fujimori, Jpn. J. Appl. Phys. , 52
(1991).
Takanashi,¢ K, S. Mitani, M. Sano, H. Fujimori, H. Nakajima, and A. Osawa, Appl. Phys.
Lett. , 1016 (1995).
Takeda, T., Z. Phys. B ¡ , 43 (1978).
Takeda, T., J. Phys. F: Metal Phys. § , 815 (1979). ¤
Thole, B. T., P. Carra, F. Sette, and G. van der Laan, Phys. Rev. Lett. , 1943 (1992).
Tohyama, T., Y. Ohta, and M. Shimizu, J. Phys.: Condens. Matter , 1789 (1989).
Tosaka,¤ Y., H. Ikekame, K. Urago, S. Kurosawa, K. Sato, and S. C. Shin, J. Mag. Soc. Jpn.
, 389 (1994). ¢
Trygg, J., B. Johansson, O. Eriksson, and J. M. Wills, Phys. Rev. Lett. ¦ , 2871 (1995).
Tsubokawa, I., J. Phys. Soc. Jpn. ,¦ , 1664 (1960).
Ubachs, W., A. P. J. van Deursen, A. F. de Vroomen, and A. J. Arko, Solid State Commun.
, 7 (1986). ¢¢
Újfalussy, B., L. Szunyogh, P. Bruno, and P. Weinberger, Phys. Rev. Lett. , 1805 (1996).
Unger, W. K., E. Wolfgang, H. Harms, and H. Haudek, ¤J. Appl. Phys. ¥ , 2875 (1972).
Uspenskii, Y. A., and S. V. Khalilov, Sov. Phys. JETP , 588 (1989).
Uspenskii, Y. A., E. G. Maksimov, S. N. Rahkeev, and I. I. Mazin, Z. Phys. B ¦ , 263
(1983).
148 BIBLIOGRAPHY
¤
Uspenskii, Y. A., E. T. Kulatov, and S. V. Khalilov, JETP , 952 (1995).
van der Heide, P. A. M., W. Baelde, R. A. de Groot, A. R. de Vroomen, P. G. van Engen,
and K. H. J. Buschow, J. Phys. F: Metal Phys. /¦ , L75 (1985).
van der Laan, G., in Spin-Orbit-Influenced Spectroscopies of Magnetic Solids, edited by
H. Ebert and G. Schütz (Springer, Berlin, 1996), p. 125.
van der Laan, G., B. T. Thole, G. A. Sawatzky, J. B. Goedkoop, J. C. Fuggle, J. M. Esteva,
R. C. Karnatak, J. P. Remeika, and H. A. Dabkowska, Phys. Rev. B ¥ , 6529 (1986). ¢
Van Drent, W. P., T. Suzuki, Q. Meng, J. C. Lodder, and T. J. A. Popma, J. Appl. Phys. § ,
6190 (1996).
van Engen, P. G., Ph.D. Thesis, Technical University Delft (1983).
van Engen, P. G., K. H. J. Buschow, R. Jongebreur, and M. Erman, Appl. Phys. Lett. ¥`¡ ,
202 (1983a).
van Engen, P. G., K. H. J. Buschow, and M. Erman, J. Magn. Magn. Mater. , 374 (1983b).
van Ek, J., and J. M. MacLaren, Phys. Rev. B ¦ , R2924 (1997).
¤
van Ek, J., W. Huang, and J. M. MacLaren, J. Appl. Phys. , 5429 (1997).
Van Vleck, J. H., Phys. Rev. ¦¡ , 1178 (1937).
Veenhuizen, P. A., F. R. de Boer, V. Menovsky, V. Sechovsk ý, and L. Havela, J. Phys.
(France) ¥`§ , C8-485 (1988a).
Veenhuizen, P. A., J. C. P. Klaasse, F. R. de Boer, V. Sechovsk ý, and L. Havela, J. Appl.
Phys. , 3064 (1988b).
Velický, B., Czech. J. Phys. B s , 541 (1961).
Verboven, E., Physica ¡s , 1091 (1960).
Vergöhl, M., and J. Schoenes, J. Magn. Soc. Jpn. ¡ , S1, 141 (1996).
Villars, P., and L. D. Calvert, Pearson’s Handbook of Crystallographic Data for Inter-
metallic Phases (American Society for Metals International, Materials Park, Ohio,
1991). ¢
Voigt, W., Ann. Phys. Chem. , 345 (1899).
Voigt, W., Magneto- und Elektrooptik (Teubner, Leipzig, 1908).
Vollmer, R., M. Straub, and J. Kirschner, Surf. Science ¦¡¨ ¦¥ , 937 (1996a).
Vollmer, R., M. Straub, and J. Kirschner, J. Magn. Soc. Jpn. ¡ , S1, ¤ 29 (1996b).
Voloshinskaya, N. M., and G. A. Bolotin, Fiz. Metal. Metalloved. , 975 (1974).
von Barth, U., and L. Hedin, J. Phys. C: Solid State Phys. ¦ , 1629 (1972).
Wachter, P., in Handbook on the Physics and Chemistry of Rare Earths, edited by K. A.
Gschneidner, Jr., L. Eyring, G. H. Lander, and G. R. Choppin (North-Holland, Amster-
dam, 1994), Vol. 19, p. 132.
Wallis, R. F., and M. Balkanski, Many-Body Aspects of Solid State Spectroscopy (North-
Holland, Amsterdam, 1986).
Wang, C. S., and J. Callaway, Phys. Rev. B § , 4897 (1974).
Wang, X., V. P. Antropov, and B. N. ¤ Harmon, IEEE Trans. Magn. , 4458 (1994).
Wang, Y. J., J. Magn. Magn. Mater. ¥ , 39 (1990).
Wang, Y. J., Z. H. Guo, D. K. Zhu, C. H. Shang, and J. C. Chen, in Proceedings of the Third
Intern. Symposium on Phys. of Magn. Materials, Seoul, Korea, p. 895 (1995). ¢¢ ¤
Watanabe, M., M. Homma, and T. Masumoto, J. Magn. Magn. Mater. ¨? , 1231
(1998).
BIBLIOGRAPHY 149
Weinberger, P., Electron Scattering Theory for Ordered and Disordered Matter (Clarendon,
Oxford, 1990).
Weller, D., in Spin-Orbit-Influenced Spectroscopies of Magnetic Solids, edited by H. Ebert
and G. Schütz (Springer, Berlin, 1996), p. 1.
Weller, D., private communication (1997).
Weller, D., H. Brändle, G. Gorman, C.-J. Lin, and H. Notarys, Appl. Phys. Lett. , 2726
(1992).
Weller, D., J. Sticht, G. R. Harp, R. F. C. Farrow, and H. Br ändle, in Magnetic Ultra-
thin Films: Multilayers and Surfaces/Interfaces and Characterization, edited by B. T.
Jonker, Mat. Res. Soc. Symp. Proceedings No. (Materials Research Society, Pitts-
burgh, 1993a) p. 501.
Weller, D., H. Brändle, and C. Chappert, J. Magn. Magn. Mater. /¡n , 461 (1993b). ¢
Weller, D., G. R. Harp, R. F. C. Farrow, A. Cebollada, and J. Sticht, Phys. Rev. Lett. ¡ ,
2097 (1994a).
Weller, D., Y. Wu, J. Stöhr, M. G. Samant, B. D. Hermsmeier, and C. Chappert, Phys. Rev.
B ¥`§ , 12888 (1994b).
¢
Weller, D., J. Stöhr, A. Carl, M. G. Samant, C. Chappert, R. Mégy, P. Beauvillain, P. Veillet,
and G. A. Held, Phys. Rev. Lett. ¦ , 3752 (1995).
Wierman, K. W., and R. D. Kirby, J. Magn. Magn. Mater. /¦/¥ , 12 (1996).
Wierman, K. W., J. N. Hilfiker, R. F. Sabiryanov, S. S. Jaswal, R. D. Kirby, and J. A. Wool-
lam, Phys. Rev. B ¦¦ , 3093 (1997).
Wijngaard, J. H., C. Haas, and R. A. de Groot, Phys. Rev. B ¥` , 9318 (1989).
Williams, A. R., and U. von Barth, in Theory of the Inhomogeneous Electron Gas, edited
by S. Lundqvist and N. H. March (Plenum, New York, 1983),Cp.189.
Williams, A. R., J. F. Janak, and V. L. Moruzzi, J. Phys. (France) , C3-131 (1972).
Williams, A. R., J. Kübler, and C. D. Gelatt, Phys. Rev. B¤ c§ , 6094 (1979).
Williams, H. J., F. G. Foster, and E. A. Wood, Phys. Rev. ¡ , 119 (1951). ¤
Williams, H. J., R. C. Sherwood, F. G. Foster, and E. M. Kelley, J. Appl. Phys. ¡ , 1181
(1957).
¢+
Wooten, F., Optical Properties of Solids (Academic Press, New York, 1972).
Wu, R., and A. J. Freeman, Phys. Rev. Lett. , 1994 (1994). ¢¢
Xu, Y., X. W. Zhang, W. K. Zhu, H. S. Yang, and Y. J. Wang, J. Appl. Phys. , 5452
(1995).
Yaresko, A. N., P. M. Oppeneer, A. Y. Perlov, V. N. Antonov, T. Kraft, and H. Eschrig,
Europhys. Lett. , 551 (1996).
Yoshii, S., and K. Egashira, IEEE Trans. Magn. c , 587 (1974).
Youn, S. J., and B. I. Min, Phys. Rev. B ¦ , 10436 (1995).
Zachariasen, W. H., J. Inorg. Nucl. Chem. ¦ , 3487 (1973).
Zeeman, P., Verslag Koningkl. Akad. Wetensch. Amsterdam ¡ , 175 (1894).
Zeeman, P., Verslag Koningkl. Akad. Wetensch. Amsterdam , 221 (1895).
Zeeman, P., Leiden Commun. ¡s§ (1896).
Zeeman, P., Phil. Mag. ¥ , 226 (1897).
Zeper, W. B., F. J. A. M. Greidanus, P. F. Garcia, and C. R. Fincher, J. Appl. Phys. ¦ , 4971
(1989).
Zubarev, D. N., Fortschr. Physik § , 275 (1961).
150 BIBLIOGRAPHY
Zubarev, D. N., V. Morozov, and G. Röpke, Statistical Mechanics of Nonequilibrium

Processes (Akademie Verlag, Berlin, 1997).
Zwerdling, S., and B. Lax, Phys. Rev. , , 51 (1957).
Zygmunt, A., and M. Duczmal, Phys. Status Solidi (a) § , 659 (1972).
151
° ±³² õ íµ´ íìö òñ +õòç

ì
It is a pleasure to thank all the colleagues who have contributed to the accomplishment of
this work. Without their help much of what has been done would not have been possible!
The first-principles investigations of MO Kerr spectra started during my post-doc stay
in the group of Prof. J. Kübler at the Technical University Darmstadt. I would like to thank
Prof. J. Kübler, who initiated, and encouraged, this research in its early stage. Also the
other members of the Darmstadt-group, Thomas Maurer, J ürgen Sticht, and Ines Osterloh,
are acknowledged for our collaboration.
In Dresden, the MO research was continued, and matured, under the supervision of Prof.
Helmut Eschrig, who supported and stimulated this research, and provided all necessary
infrastructure to carry it out. Thank you, Helmut!
During the time of the “Max-Planck-Arbeitsgruppe Theorie komplexer und korrelierter
Elektronensysteme” I worked together with Thomas Kraft, and with visitors from Kiev,
Victor Antonov, Alexander Perlov, and Alexander Yaresko. Thank you, Thomas, Victor,
Sasha, and Sasha, for our intensive cooperation.
Other colleagues, with whom I worked together on the theory of MOKE spectra, are
the members of the (extended) Uppsala-group, Mike Brooks, Anna Delin, Olle Eriksson,
Peter James, Börje Johansson, P. Ravindran, and John Wills. Thank you all, your interest
in MOKE calculations has stimulated the theoretical investigations. I am looking forward
to continuing our collaboration in the future!
With thanks I also acknowledge the discussions and collaboration which I had, in
various stages of this work, with Frank Herman, Hubert Ebert, and Barry Cooper.
A number of experimentalists that perform research in the field of magneto-optics have
contributed to the development of the theoretical research. In particular, I wish to thank
Prof. J. Schoenes, Raniero Pittini, and Dieter Weller for many discussions and the fruitful
cooperation. I have had contact to, or collaborated with, or had discussions with, the
following experimentalists, whom I am grateful to (in alphabetic order): Petra Ber ánková,
Andreas Borgschulte, Guo-Qing Di, Prof. Paul Fumagalli, Prof. Gernot G üntherodt,
Satoshi Iwata, Toshikazu Katayama, Takeshi Kato, Mirek Kučera, Karsten Litfin, Ulrich
Rüdiger, Prof. Katsuaki Sato, Chang He Shang, Koki Takanashi, Prof. S. Tsunashima,
Prof. Štefan Višňovský, Prof. P. Wachter, and Prof. Y. J. Wang.
The members of the Prague-group, Vladimir Sechovsk ý, Ladia Havela, Martin Diviš,
Marek Olšovec, and Karel Prokeš, are acknowledged for keeping me interested in the
physics of actinide compounds, and, in several cases, convincing me of doing electronic-
152
structure calculations.
Thanks are also due to the other members of the former Max-Planck Arbeitsgruppe
“KoKo”: Igor Chaplyguine, Arthur Ernst, J örg Forstreuter, Roland Hayn, Samed Khalilov,
Klaus Koepernik, Clemems Kreß, Zhaosen Liu, Ulrike Nitzsche, Ingo Opahle, Mike Rice,
Manuel Richter, Helge Rosner, Vito Servedio, Lutz Steinbeck, Manfred Taut. Thank you
all, for the pleasant atmosphere, the coffee breaks, and all the discussions, which where,
fortunately, not always about physics. Secretary Mrs. Helga Kotte has in multitudes of
small things provided immediate help and was responsible for creating much of the good
and workable atmosphere. Thank you, Helga!
Last, but not least, I am deeply indebted to my beloved wife Maria, who has given me
her dear support and advice in every situation encountered. Much of what you have done
goes beyond saying!
153
¶ #÷²ç
é í· ° }áì¹¸ é0ê ç é íòõ ÷
APW augmented plane wave

ASA atomic sphere approximation
ASW augmented spherical wave
BZ Brillouin zone
CF crystal field
DC direct current
DFT density-functional theory
DOS density of states
FLAPW full-potential linearized augmented plane wave
GMR giant magneto-resistance
KK Kramers-Kronig
L-MOKE longitudinal magneto-optical Kerr effect
LMTO linearized muffin-tin orbital
LSDA local spin-density approximation
MAE magnetocrystalline anisotropy energy
MBE molecular beam epitaxy
ME matrix elements
MO magneto-optical
MOKE magneto-optical Kerr effect
NMR nuclear magnetic resonance
NOLI-MOKE non-linear magneto-optical Kerr effect
OP orbital polarization
P-MOKE polar magneto-optical Kerr effect
RPA random-phase approximation
SO spin orbit
TM transition metal
T-MOKE transversal magneto-optical Kerr effect
UC unit cell
UV ultra-violet
XMCD X-ray magnetic circular dichroism
View publication stats

Theory of Magneto-Optic Kerr Effect in Ferromagnetic Compounds

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Theory of Magneto-Optic Kerr Effect in Ferromagnetic Compounds

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Theory of the Magneto-Optical Kerr Effect in Ferromagnetic Compounds

Book · March 1999

The user has requested enhancement of the downloaded file.

vwEHGx?C:@? S ?yPAS z My{yKN[DWMI[.?C:=GIJLKN[(g#Kr|}GI:>\

2.1 Classical optics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ÌÌ

    "!#$!%&#' (*)+-,/.0#12-345%

67:687NM OQPSRTPSUVE A$C9DFEHGI=J?

to visualizing surface and subsurface magnetic domains developed. Furthermore, it was

WXIY Z\[][\^`_Vacb d@efY Z\[]gh^`_Vacb

~IJ&xH`HSchematic illustration of the polar MO Kerr effect. Linearly polarized

polarization plane is tilted over a small angle 

 v#u#$./v55 &/.0¡.0¢£#1¤¥2#".02.0¡(¦ m

687µª17NM ¯¶°·P´³P¸REJD C:U¹E\AuPfº¼»=\<?¹PSU$P=\½8?o=\P½IC9¾¿P

687µª17Nª À*GVU1»>9C:U$P<= ¯¶°ÂÁ ;

× Þ¡ç ÕJä Þ ä á ÂFØSèéÚ

îlï  À ð´ï  À î]ï

Lifshitz consider in their derivation of

à  ^`_Va  â è ^ò_Va á ß Ô±Ü ë¶^`_VaTÚ ^ 2.13)

consequence is that è becomes the only time-varying source term

current density is only a direct current. The fact that ^ò_Va 

Note, that in Eq. (2.15) those contributions to è

kÌlk úr0#st#$8v.0¢87 -8-¢u¨ #wx#uy2

M17µM17:6 ÎÏA$P ¬-GV>­<=¼¯±° Á P=\=¼P´³PSRE

ä:9 ò^ ñ¸Ú a  ; À < ¥ï ^>= [@? A= g aHXpýøþCB)D E ý Ú ^ 2.21)

, ò9  î []@ [ ? @î []g  ^ 2.23)

diagonal element vanishes, î []g Ï .

amplitudes can be written as

Schoenes (1992) for the P-MOKE. Substituting , Oba P ?  ò  î9[][ cî9[]g

z{ á µ}{ _ Z [][ c æ À Z\[]g Jù  e oqý pZ m[][  f K & ò  ^ 2.35)

z {} á µ} }{ _ Áh^À æ î æ a&îu^Nòî p|øæ  Ápbm|  ò m a K & ò Ú ^ 2.39)

W X  î ò oqp ò 8Á m îuò æ p|øî Áá m  pb|ø ò m 

À æ l }  X ò ý ]Ì X ý  ~W X  &)(  ^ 2.44)

z  á N } :_ _ Á ßô ^ , O æ , P a _ _ Á ßô @î î9[]K [][& gò Ú ^ 2.45)

which, in terms of the components of the conductivity tensor reads

z  å_  z   WX  ZZ\[][][g  Ú ^ 2.47)

M17µª17:6 Q P¸UuP=\<8>/®lG=h¾ Ó">­<2E\C­GIU

Ù^ ã a  X q¡ ¢§Ú ^ 2.49)

^ ã aN¯ §¦S©¨ ÙT^ ã ab­°¬¥Ú

ô ÙT^ ã a  À  ^ a]Ú9Ù^ a²/

where ¾ is the response function, ® *

Partial integration leads to

^ãa  À ^ ã &a Ú ^ ã a ï X æ Ný _ æ À ± ó ô ã õ

¹ É is the time derivative of ¹ , which is again related to the current through

¹É ^ ã a ÊÁ Â ñ É ý Õä  æ Å  ^ ã a Õä  ^ 2.60)

Z ^ò_Va  _ á FÀ _ ó ô zh^ ha ^ ha]Ú ^´Ï|a ï XSý  ^

Z U ^`_Va  À  X ÒÓ &)ÏNÐ æ X Ò Ó È &)ÏNÐ  ·½Ô S Q Ç U S Ô õ < ¸ ·[Ñ Ô õ S Ç < S Ô Ñ ¸ Ú

M17µª17Nª 1Õ C9Uu½I>:PT»¬"<2=\EJC9R´>­P®lG=4¾ Ó$>:<¸E\C:GIU

ô   ^ a    ^ a&Úb/ï / ØÂ · , , õ   È ^ a  î   ^ aTÚ

Z U ^`_Va  ^ î  Q a æ ß^ î  È a U È È ^ 2.72)

^ î  a« U  Â ß^Nî  a· , S Z U Ú> / S , ¸ À ±

 æ Ë À ± Â È ß^Nî  a _ æ ßÈ ^Nî  È a ~U  È  È  

a shorter single-particle expression for ^`_Va

 ç ^ò_Va  ¥ á ÂØ í ^`_Va _

Z U ^`_Va  æ Ë  ò Á ± Æ ^ î  a æ ß^ î  È a  U È  ~ È

is convoluted with a Lorentzian, of which   ¿ K is the half width at half maximum. In

vanish at _  Ï (see Garik and Ashcroft 1980).

into account in calculations, already the constant relaxation-time approximation suffices

uW XSZ U ^`_Va  WuXSZ U ^ æ _VaTÚ

Z U ^`_uÚ Û a  Z U ^`_uÚ æ Û a´Ú ^ 2.79)

Z []g ^ò_Va éæ Z g][ ^ò_Va

ZJ[]g4^ò_uÚ æ Û &I a éæ Z\[]gh^`_uÚ Û I]a  ^ 2.80)

^ò_Va æ ^`_Va  æ Ø Á ó ô _Iõ _ õ ò _ æ õ _ ò WX ò^ _Iõ`a æ ^ò_Iøõ a  ^

vwEHGx?C:@? S ?yPAS z My{yKN[DWMI[.?C:=GIJLKN[(g#Kr|}GI:>\

"!#$!%&#' (*)+-,/.0#12-345%

67:687NM OQPSRTPSUVE A$C9DFEHGI=J?

WXIY Z\[][\^`_Vacb d@efY Z\[]gh^`_Vacb

~IJ&xH`HSchematic illustration of the polar MO Kerr effect. Linearly polarized

polarization plane is tilted over a small angle

v#u#$./v55 &/.0¡.0¢£#1¤¥2#".02.0¡(¦ m

687µª17NM ¯¶°·P´³P¸REJD C:U¹E\AuPfº¼»=\<?¹PSU$P=\½8?o=\P½IC9¾¿P

687µª17Nª À*GVU1»>9C:U$P<= ¯¶°ÂÁ ;

îlï À ð´ï À î]ï

à ^`_Va â è ^ò_Va á ß Ô±Ü ë¶^`_VaTÚ ^ 2.13)

current density is only a direct current. The fact that ^ò_Va

kÌlk úr0#st#$8v.0¢87 -8-¢u¨ #wx#uy2

M17µM17:6 ÎÏA$P ¬-GV><=¼¯±° Á P=\=¼P´³PSRE

ä:9 ò^ ñ¸Ú a ; À < ¥ï ^>= [@? A= g aHXpýøþCB)D E ý Ú ^ 2.21)

, ò9 î []@ [ ? @î []g ^ 2.23)

diagonal element vanishes, î []g Ï .

Schoenes (1992) for the P-MOKE. Substituting , Oba P ? ò î9[][ cî9[]g

z{ á µ}{ _ Z [][ c æ À Z\[]g Jù e oqý pZ m[][ f K & ò ^ 2.35)

z {} á µ} }{ _ Áh^À æ î æ a&îu^Nòî p|øæ Ápbm| ò m a K & ò Ú ^ 2.39)

W X î ò oqp ò 8Á m îuò æ p|øî Áá m pb|ø ò m

À æ l } X ò ý ]Ì X ý ~W X &)( ^ 2.44)

z á N } :_ _ Á ßô ^ , O æ , P a _ _ Á ßô @î î9[]K [][& gò Ú ^ 2.45)

z å_ z WX ZZ\[][][g Ú ^ 2.47)

M17µª17:6 Q P¸UuP=\<8>/®lG=h¾ Ó"><2E\CGIU

Ù^ ã a X q¡ ¢§Ú ^ 2.49)

^ ã aN¯ §¦S©¨ ÙT^ ã ab°¬¥Ú

ô ÙT^ ã a À ^ a]Ú9Ù^ a²/

^ãa À ^ ã &a Ú ^ ã a ï X æ Ný _ æ À ± ó ô ã õ

¹É ^ ã a ÊÁ Â ñ É ý Õä æ Å ^ ã a Õä ^ 2.60)

Z ^ò_Va _ á FÀ _ ó ô zh^ ha ^ ha]Ú ^´Ï|a ï XSý ^

Z U ^`_Va À X ÒÓ &)ÏNÐ æ X Ò Ó È &)ÏNÐ ·½Ô S Q Ç U S Ô õ < ¸ ·[Ñ Ô õ S Ç < S Ô Ñ ¸ Ú

M17µª17Nª 1Õ C9Uu½I>:PT»¬"<2=\EJC9R´>P®lG=4¾ Ó$>:<¸E\C:GIU

ô ^ a ^ a&Úb/ï / ØÂ · , , õ È ^ a î ^ aTÚ

Z U ^`_Va ^ î Q a æ ß^ î È a U È È ^ 2.72)

^ î a« U Â ß^Nî a· , S Z U Ú> / S , ¸ À ±

æ Ë À ± Â È ß^Nî a _ æ ßÈ ^Nî È a ~U È È

ç ^ò_Va ¥ á ÂØ í ^`_Va _

Z U ^`_Va æ Ë ò Á ± Æ ^ î a æ ß^ î È a U È ~ È

is convoluted with a Lorentzian, of which ¿ K is the half width at half maximum. In

vanish at _ Ï (see Garik and Ashcroft 1980).

uW XSZ U ^`_Va WuXSZ U ^ æ _VaTÚ

Z U ^`_uÚ Û a Z U ^`_uÚ æ Û a´Ú ^ 2.79)

Z []g ^ò_Va éæ Z g][ ^ò_Va

ZJ[]g4^ò_uÚ æ Û &I a éæ Z\[]gh^`_uÚ Û I]a ^ 2.80)

^ò_Va æ ^`_Va æ Ø Á ó ô _Iõ _ õ ò _ æ õ _ ò WX ò^ _Iõ`a æ ^ò_Iøõ a ^

Áz ^`_Va æ ØÁ ó ô _Iõ _ õ ò _ æ õ _ ò ø À æÏ } 2^`_ õ a Ú

Áz { ^`_Va æ @ØÁ ü ó ô _Iõ _ õ ò _ æ õ _ ò öø ÀÀ áæÏ }}{{u^ò^ò__ õõ aa Ú

ó ô _ÌZ U ^`_Va æ ± À Æ ó ô ¿ ó ôù X ù ý æ ú ýÏ ý Q J^¶¿ha Ç U ^>¿ha]Úb¾ ^½ÏFa²/ ¯ ï Ú ^ 2.99)

´ ´ ´

Á ó ´ ô _ À öø À æÏ } { ^ò_Va ÏÚ ^ 2.101)

ÂFØ WuXIY Z\[][ ? @ZJ[]gb

ó ô _¥Z [] òg ^`_Va _ Â Í ÌÌ ÿð¤ Ú ^ 2.109)

Ù Y Ùb á î\ æóüô ñÙT^òñTa ÏÚ ^

ÙT^òñTa 3 4ë 6 ^òñTa 3546 ^òñTa ^ 2.115)

Tï æ 3546 3546 ^ 2.116)

y æ Á á Ý ={ < ^òñTa 354 ^ ña î 4-3>4 ^òña´Ú ^ 2.119)

M17>áu7NM ?°ÕA@¼« <¬$¬"=\G<8RhA¡EJG¿E\AuPfP´ÐuR´C@EJ<2EJC:GIUoDF¬-P¸REH=4Ó$¾

B ^òñSÚ9ñ õ Ú9î]a _ J ^ ñ æ ñ õ a Ý [ 9 ^ ña ^ 2.123)

Ìk G Hn)+#u p) 5&#")+#$./ .

æ ò ß ò ^ Ô Ý a Õ`Ô á ± ò ß ò í Õ c Ô Ý AÜ O f á`Ý Ú ^ 2.130)

M78I17NM ÎÏA$Pé¾¿<2EJ=hCÐüPS>PS¾ PSUVE\D

0À ß ò 8^ d K á dHò í I a O á O ^8d K á dHò í I a á í Ü ^O d a á

M78I17Nª V| »D|¬$<8RTP C:UVEJP½=h<¸E\C:GIUB<8UtvUyÓ"¾ P=hC:RT<8>/RTGIU2vIP=\½IPSU$RTP

óØ T Hô û _ æ S Þ _$^^ ûû aa S á ²

@Ø ü Tó ôHû _ S Þ æ ^ _$û a ^ ûS a á ó ôTS Ô Þ _$^ û ^ ûa S a Ú ^ 2.143)

í ^`_ áa ÁØÀ ó ô _Iõ _ õ æ À _ í ^ò_Iõ á a

ýZÿ £3ý£ ÿ Ñ£ ÿ ÿ Ñ£+ ¢

¥ .©28¥¢ n0y8&.

M ª¹]ºf» ¼½ M «¹]ºf» ¼½ M ª¾-¿u» ¾ 3 M «¾-¿u» ¾ 3

¥ ¦ m 5§#uy28Ý(§70#$ËéwÌ70#$ËË¢ÍÌÍ ËË¢ÍÌ .0¢ ÍúËé

Z [][ ^À á ÂØ @Z [][ _Va9K ò Z []g

øÕùûú ËýüÎþ ÿ ÿ §# ÿ( Ë ÿ

øÕù$^ . _ `9= ÿ a ( #

Experiment — — 0.02 -0.12 0.14 -0.01 0.30 0.08 0.30 0.09

X §¦ §¦ §¦¨ª© ÅÃ ¨«© ÅÃ

øÕù-Ñ Ò ÿ ÿ!ÿ 1 ÿ ÿ

Ñ ùûú ï 9 Zä : _2 ð

Ñ ùø _ 9:0 9

" ùûú ï 9 Zä :