Atomistic spin dynamics : foundations and applications 1st Edition Bergman 2024 Scribd Download
Atomistic spin dynamics : foundations and applications 1st Edition Bergman 2024 Scribd Download
com
https://textbookfull.com/product/atomistic-spin-dynamics-
foundations-and-applications-1st-edition-bergman/
OR CLICK BUTTON
DOWNLOAD NOW
https://textbookfull.com/product/foundations-of-space-dynamics-ashish-
tewari/
textboxfull.com
https://textbookfull.com/product/decision-diagrams-for-
optimization-1st-edition-david-bergman/
textboxfull.com
https://textbookfull.com/product/explosion-dynamics-fundamentals-and-
practical-applications-1st-edition-rangwala/
textboxfull.com
Mathematical Foundations and Applications of Graph Entropy
1st Edition Matthias Dehmer
https://textbookfull.com/product/mathematical-foundations-and-
applications-of-graph-entropy-1st-edition-matthias-dehmer/
textboxfull.com
https://textbookfull.com/product/mathematical-foundations-and-
applications-of-graph-entropy-1st-edition-matthias-dehmer-2/
textboxfull.com
https://textbookfull.com/product/spin-with-me-1st-edition-ami-
polonsky/
textboxfull.com
https://textbookfull.com/product/spin-current-sadamichi-maekawa/
textboxfull.com
ATOMISTIC SPIN DYNAMICS
Atomistic Spin Dynamics
Foundations and Applications
Olle Eriksson
Department of Physics and Astronomy, Uppsala University, Box 516, SE-751 20 Uppsala,
Sweden
Anders Bergman
Department of Physics and Astronomy, Uppsala University, Box 516, SE-751 20 Uppsala,
Sweden
Lars Bergqvist
Department of Materials and Nano Physics, KTH Royal Institute of Technology,
Electrum 229, SE-164 40 Kista, Sweden
Johan Hellsvik
Department of Materials and Nano Physics, KTH Royal Institute of Technology,
Electrum 229, SE-164 40 Kista, Sweden
3
3
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Olle Eriksson, Anders Bergman, Lars Bergqvist, Johan Hellsvik 2017
The moral rights of the authors have been asserted
First Edition published in 2017
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2016938429
ISBN 978–0–19–878866–9
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Preface
We would like to acknowledge fruitful discussions and a great collaboration with the
following colleagues (in alphabetical order):
The authors also acknowledge support from the Swedish Research Council (VR),
eSSENCE, the Swedish e-Science Research Centre (SeRC), the Knut and Alice
Wallenberg (KAW) Foundation (grants 2012.0031 and 2013.0020), and the Göran
Gustafsson Foundation (GGS).
Contents
Part 3 Applications
8 Ferromagnetic Resonance 139
8.1 Experimental set-up and demagnetization field 139
8.2 Kittel equations 141
8.2.1 Damping and anisotropy ignored 141
8.2.2 Including anisotropy 143
8.2.3 Full treatment including damping 143
8.3 The Smit–Suhl equation 151
8.4 Spin wave resonance 153
9 Magnons 154
9.1 Spin excitations in solids 154
9.2 Experimental methods 158
9.3 Adiabatic magnon spectra and the frozen magnon method 160
9.4 Dynamical magnetic susceptibility 167
9.5 Surface magnons from atomistic spin dynamics 169
9.5.1 Thin films of Co on Cu substrates 170
9.5.2 A comparison of approaches 175
9.5.3 Fe on Cu(001) 175
9.5.4 Fe on Ir(001) 177
9.5.5 Fe on W(110) 178
9.6 Relativistic effects 181
9.7 Magnon lifetimes 182
10 Skyrmions 185
10.1 Background 185
10.2 Magnetism and topology 186
10.3 Magnetic skyrmions 187
10.4 Theoretical prediction and experimental identification 188
10.5 Dimensionality and stability 190
x Contents
References 231
Index 253
Part 1
Density Functional Theory
and its Applications
to Magnetism
Density functional theory (DFT) has been an invaluable tool for understanding and
analysing the magnetism of materials. In the three introductory chapters forming Part
1, we review the most central and important features of this theory and give examples
of magnetic properties that are accessible with this theory. We start this description with
the well-established Hartree–Fock theory, to demonstrate the principal mechanism of
intra-atomic spin pairing. The master equation of DFT, the so-called Kohn–Sham equa-
tion, is derived, and we show how it can be used for spin polarized calculations of spin
and orbital moments, for magnetic form factors, for magnetic anisotropy, and for inter
atomic (Heisenberg) exchange parameters. We also describe special considerations of
this equation when applied to the solid state. We cover both translational and rotational
symmetries, as well as complications that emerge when spin–orbit coupling is included
in the calculations. Also outlined in Part 1 of this book is how concepts from DFT
can be used for a multiscale approach to atomistic spin dynamics simulations. Hence,
Part 1 shows how DFT calculations of only a few atoms can be used to enable simula-
tions of billions of atomic spins. Throughout these three chapters, theoretical results are
compared to existing experimental data, and the level of agreement between theory and
observation is discussed.
1
Density Functional Theory
Density functional theory has been established as a very practical platform for model-
ling, from first principles, the electronic, optical, mechanical, cohesive, magnetic, and
structural properties of materials. Starting from the Schrödinger equation for the many-
body system of electrons and nuclei, an effective theory has been developed allowing for
material-specific and parameter-free simulations of non-magnetic as well as magnetic
materials. In this chapter, an introduction will be given to density functional theory, the
Hohenberg–Kohn theorems, the Kohn–Sham equation, and the formalism for how to
deal with spin polarization and non-collinear magnetism.
1 1 e 2 1 ZI e 2
+ – , (1.1)
2 i=j
4π0 |ri – rj | i,I
4π0 |ri – RI |
where the indices i, j denote electrons; I, J are for atomic nuclei; and the masses are
denoted MI for nuclei, and m for electrons. Furthermore, RI and ri stand for nucleus and
electron coordinates, respectively, whereas ZI denotes atomic number. In the following,
we will adopt Hartree atomic units, that is, e = m = h̄ = 4π0 = 1. Since nuclei are
much heavier than electrons, one may adopt the Born–Oppenheimer approximation and
assume that the nuclei are fixed, while the electrons are dynamic objects. This allows,
one to deal with electron states separately from the atomic nuclei. Thus, we are left ’only’
Atomistic Spin Dynamics. Olle Eriksson, Anders Bergman, Lars Bergqvist, Johan Hellsvik. First Edition.
© Olle Eriksson, Anders Bergman, Lars Bergqvist, Johan Hellsvik 2017. First published in 2017 by Oxford University Press.
4 Density Functional Theory
with the description of the electron system, and the Hamiltonian acting on the electrons
is written as
1 2 1 1 ZI
Ĥ = – ∇i + – = T̂ + Ŵ + V̂ ext . (1.2)
2 2 |ri – rj | |ri – RI |
i i=j i,I
Here, T̂ is the kinetic energy operator of the electrons, Ŵ is the operator determining
the Coulomb energy of electron–electron interactions, and V̂ ext is the external poten-
tial accounting for the Coulomb interactions between the electrons and the nuclei. The
corresponding total energy E is the expectation value of Ĥ , that is
E = Ĥ = T + W + d 3r Vext (r)n(r), (1.3)
with T and W denoting the expectation values of the kinetic energy and electron–
electron interaction operators, respectively, and n(r) denoting the electron charge
density.
1
(x1 , x2 , . . . , xi , . . . , xN ) = √ det[ψi (xj )], (1.4)
N
where x is a composite coordinate of space and spin of an electron, N is the num-
ber of electrons, and we have used a compact expression for the Slater determinant.
Forming the expectation value of the electron Hamiltonian with this approximate many-
electron wave function yields an expression of the total energy. Minimization of this
energy expression with respect to the single-particle wave functions that compose the
Slater determinant gives the Hartree–Fock equation:
⎛ ⎞
1 ψ ∗
(x )ψ j (x
)
⎝– ∇i2 + Vext (r) + dr
j ⎠ ψi (x)
2 |ri – rj |
i =j
ψ∗
(x )ψ i (x
)
dr
j
– ψj (x)δsi sj = i ψi (x), (1.5)
|ri – rj |
i=j
where Vext represents the electron–nucleus interaction. This equation is similar to the
Schrödinger equation for electrons that move in a potential and, in fact, the first two
The Hartree–Fock theory 5
terms that follow the kinetic energy in Eqn (1.5) represent the effective potential that
an electron experiences because of its attractive interaction with the nucleus and its re-
pulsive interaction with all other electrons in the system (this interaction is referred to
as the Hartree term; Atkins and Friedman, 2005). The last term on the left-hand side
of Eqn (1.5) is referred to as the exchange interaction; it is active only if the spins of
electrons i and j are the same, and is hence a term that lowers the energy if as many
spins as possible have the same direction. This spin pairing (Jorgensen, 1962) is respon-
sible for the fact that most atoms of the Periodic Table have a ground-state electronic
configuration with a maximized number of parallel spins, provided that, the Pauli exclu-
sion principle is obeyed. This microscopic mechanism behind Hund’s first rule follows,
interestingly, from a Hamiltonian that has no spin-dependent terms in it, Eqn (1.2).
It is entirely quantum mechanical in nature and follows from the requirement that the
wave function should be antisymmetric with respect to permutation of the coordin-
ates of any two electrons. Note that, in Eqn (1.5), we have explicitly kept electron i from
10 Lanthanides
Actinides
9
7
Spin-pairing energy (eV)
interacting with itself, thus adopting a physically meaningful approach. If, however, we
allow for the summations in Eqn (1.5) to also include terms where i = j, we would not
introduce an error, since then the extra interaction in the Hartree term would be can-
celled exactly by the extra interaction of the exchange, or Fock, term. This is easily seen
by setting i = j in the subscript of the wave functions of Eqn (1.5). This is normally
described as the lack of self-interaction in the Hartree–Fock equation, and is relevant
when discussing approximations to density functional theory, as will be discussed in
Section 1.4. It should also be noted that the interactions in the Hartree–Fock equation
are non-local, meaning that they involve interactions over distances described by |ri – rj |.
In order to illustrate the strength of the spin-pairing energy as defined by Jorgensen
(1962), Fig. 1.1 shows the experimental values of the spin-pairing energy of lanthan-
ides and actinides for the f -shell (Nugent, 1970). Note that the energies involved can
be rather large, of the order of 10 eV. Hence, discussions of, for example, ultrafast de-
magnetization using intense laser pulses must take these strong intra-shell couplings into
consideration. It is also clear from the figure that spin pairing is slightly larger for the
lanthanide series than for the actinide series, because the 4f wave function is less ex-
tended in the lathanide series than in the actinide series and thus forces the electrons of
the 4f shell to occupy a smaller volume than those in the 5f shell do. For this reason, the
exchange energy is larger for the lanthanides than for the actinides, as may be seen from
Eqn (1.5).
Theorem 1.1 The total energy of a system is a unique functional of the ground–state electron
density.
To demonstrate this, we consider the expectation value of Eqn (1.3). We evaluate the
ground-state many-body wave function gs (r1 , r2 , . . . , rN ) for N electrons and their
energy from the equation
N
ngs (r) = d 3 ri | gs (r1 , r2 , . . . .rN ) |2 δ(r – ri ). (1.7)
i=1
We will now demonstrate that using two different external potentials in Eqn (1.3), for
example, Vext and Vext , gives rise to two different ground-state electron densities: ngs
and ngs , respectively. For simplicity, we consider first a non-spin polarized system, with
m = 0. To show this, we note first that, for the system with external potential Vext ,
we have
Ĥ gs = Egs gs . (1.8)
To estimate the energy of the rightmost term of Eqn (1.9), we add and subtract Vext to
the expectation value so that
gs Ĥ gs = gs Ĥ + Vext – Vext gs = gs Ĥ + Vext – Vext gs . (1.10)
Note that, in this expression, we have, for simplicity, omitted to write out the r-
dependence of the external potential and density. Combining Eqns (1.9) and (1.11)
yields the relationship
Egs < Egs + ngs (Vext – Vext
) d 3 r. (1.12)
We could start all over from Eqn (1.9) and, going to Eqn (1.11), arrive at an expression
similar to Eqn (1.12) but with all primed and unprimed symbols being interchanged,
that is, the following expression would also follow from the variational principle:
Egs < Egs + ngs (Vext – Vext ) d 3 r. (1.13)
If we now assume that ngs = ngs , an absurd relationship follows, since we can replace ngs
in Eqn (1.12) with ngs and then add Eqns (1.12) and (1.13) together to obtain
Egs + Egs < Egs + Egs + ngs (Vext – Vext ) d3r + ngs (Vext – Vext ) d 3 r, (1.14)
8 Density Functional Theory
V n
V´ n´
V-space n-space
from which it would follow that Egs + Egs < Egs + Egs . This is clearly a result that is
incorrect. Since deriving it follows from the rules of quantum mechanics, plus one as-
sumption, we must conclude that this initial assumption was wrong, that is, that ngs and
ngs cannot be equal. Figure 1.2 shows a schematic of the relationship between external
potential and electron density, following the above analysis of density functional the-
ory. According to this figure, two different external potentials can never point to the
same ground-state density; they exclusively will identify unique and separate ground-
state densities. In addition, there are no points in n-space that cannot be reached from a
point in V -space. Mathematically, such mappings are called bijective.
So far in this discussion, we have made the derivation following the direction of the
arrows in Fig. 1.2, namely, starting from different potentials in V -space always results in
unique positions in n-space. In principle, it is possible to follow these arrows backwards,
starting from the ground state density and ending up in a unique spot of V -space, with
this spot uniquely establishing the external potential where this density came from. Since
the expressions for kinetic energy and electron–electron interaction are the same for any
system, it is the form of the external potential which makes the Hamiltonian unique and
hence specifies it. The arguments of density functional theory thus make the following
claim: knowing ngs implies we know which external potential was used in the Hamilton-
ian. This, in turn, specifies the full Hamiltonian and, from it, all states, even the excited
ones. Thus we establish electron density as a property which describes a system, and we
formally allow the use of electron density in place of a many-electron wave function. In
particular, we are now able to express the following functional relationship between the
ground-state energy of the Hamiltonian, and the ground-state electron density:
The three terms in Eqn (1.15) are the same as those defined in Eqn (1.3). Using similar,
straightforward arguments, one can arrive at the second important theorem underlying
density functional theory:
Theorem 1.2 The exact ground-state density minimizes E[n(r)] in Eqn (1.15).
The Kohn–Sham equation 9
If we had an explicit form for E[n(r)], we could go ahead and minimize it with respect
to the electron density and in this way calculate the ground-state energy. The expression
for Vext [n(r)] is straightforward, but those for T [n(r)] and W [n(r)] are more difficult.
Attempts have been made at formulating such expressions, for example, Thomas–Fermi
theory, but this theory comes short when compared to the topic of the next section, the
Kohn–Sham approach, when results are compared to, for example, measured magnetic
and cohesive properties.
1
– ∇ 2 + Veff (r) ψi (r) = i ψi (r). (1.16)
2
From this equation, which in the density functional theory community is also referred
to as the Kohn–Sham equation, one can calculate an electron density from the occupied
one-particle (op) states. Since no direct electron–electron interactions are considered
when evaluating this density, it is referred to as a one-particle density:
nop (r) = |ψi (r)|2 , (1.17)
occ
where the sum is over occupied states. In this case, the energy functional which describes
the total energy may be written as
and the electron density which minimizes this functional is obtained from the require-
ment that the energy functional is stationary for small variations of the electron density
around the ground-state density, that is Eop [nop + δn] – Eop [nop ] = 0, which also can be
written as
δTop [nop (r)] + δn(r)Veff (r) d 3 r = 0. (1.19)
Carrying out this minimization, using Eqn (1.18) for the kinetic energy, leads to
Eqn (1.16), and we have shown that non-interacting electrons which are the solutions
10 Density Functional Theory
to Eqn (1.16) result in an electron density which minimizes the total energy of this sys-
tem via the functional in Eqn (1.18). The basic principle of the Kohn–Sham approach
is the assumption that one can find an effective potential Veff such that its density nop is
the same as the ground-state density of the fully interacting system, ngs . The assumption
is proven to hold for a homogenous electron gas and small deviations from it but no
practical experience shows that the assumption holds in the general case. Nevertheless,
since we know the coupling between potential and density, as described, for example, in
Fig. 1.2, it seems like an efficient route to get the ground-state density of the interacting
system, by careful selection of an effective potential, even if it is from a one-particle
system.
The question now is, how do we determine Veff so that nop becomes equal to ngs ?
In order to find a way to do this, we first recast the energy functional in Eqn (1.15) to
the form
E[nop (r)] = Top [nop (r)] + nop (r)Vext (r) d 3 r
nop (r) · nop (r ) 3 3
+ d r d r + Exc [nop (r)]. (1.20)
r – r
In Eqn (1.20), we have introduced the one-particle kinetic energy functional Top instead
of the true kinetic energy functional, and we have introduced the Hartree electrostatic
interaction instead of the true electron–electron interaction. Hence in order to make
Eqn (1.20) equal to Eqn (1.15) we must introduce a term that corrects for this, and this
is what the exchange and correlation energy Exc [nop (r)] does. Since the first three terms
on the right-hand side of Eqn (1.20) can be calculated numerically, we have moved
the complexity of the fully interacting system to finding the exchange and correlation
functional. For a uniform electron gas, one can, however, calculate Exc [nop (r)] for all
values of the electron density, and parameterized forms of Exc [nop (r)] as a function of
nop (r) are available. For uniform densities, it is hence possible to evaluate Eqn (1.20)
with excellent accuracy, and obtain the total energy using the electron density as the
decisive variable of the system.
The local density approximation (LDA; Hedin, 1965; Hedin and Lundqvist, 1971;
Barth and Hedin, 1972; Ceperley and Alder, 1980) assumes that the parameterizations
used for the uniform electron gas work even in cases where the electron gas is not uni-
form, and is applicable to molecules, solids, surfaces, and interfaces. This is done by
assuming that locally, at a given point in space of, for example, a solid or molecule,
one may consider the density as uniform and hence use the parameterized version from
the uniform electron gas. This means that one uses the following expression for the
exchange–correlation energy:
Exc [nop (r)] = xc [nop (r)]nop (r) d 3 r, (1.21)
where xc [nop (r)] is the exchange–correlation energy density; in a parameterized form,
its dependence on nop (r) is relatively simple. We now have an expression for the
The Kohn–Sham equation 11
energy functional that can be minimized with respect to the electron density, using
an expression similar to Eqn (1.19) but now including electron–electron interaction
via the Hartree term and the exchange–correction functional. This minimization re-
sults in a one-electron Schrödinger-like equation similar to Eqn (1.16). However, the
minimization procedure leads to an explicit form of the effective potential Veff from
Eqn (1.16):
nop (r ) 3
Veff (r) = Vext (r) + d r + μxc [nop (r)], (1.22)
| r – r |
where
∂{xc [nop (r)]nop (r)} ∂{xc [nop (r)]}
μxc [nop (r)] = = xc [nop (r)] + nop (r) . (1.23)
∂nop (r) ∂nop (r)
We now can evaluate the total energy of a system, using electron density as the key vari-
able that determines things. In practice, this means solving Eqn (1.16) with the effective
potential specified by Eqn (1.22). Since the effective potential to be used in Eqn (1.22)
depends on electron density, which is the property we want to calculate, one has to per-
form a self-consistent field calculation where an initial electron density is guessed and
an effective potential is calculated from Eqn (1.22). This potential is then used to solve
Eqn (1.16), and a new electron density is calculated from Eqn (1.17), which is then
put back into Eqn (1.22). This procedure is repeated until convergence is obtained,
that is, until the density does not change appreciably with successive iterations. Once a
self-consistent electron density has been found, one can calculate the ground-state en-
ergy of the Kohn–Sham LDA energy functional, via Eqn (1.20). We comment here on
the evaluation of the kinetic energy in Eqn (1.20). Since an accurate expression of it in
terms of the electron density is missing, one evaluates it from occ ψ| – 12 ∇ 2 |ψ by using
Eqn (1.16), moving Veff to the right-hand side of the equation and multiplying from the
left with ψ| on both sides, yielding
1
Top = ψi | – ∇ 2 |ψ = i – d 3r Veff (r)nop (r). (1.24)
occ
2 occ
The last term on the right-hand side of this equation is sometimes referred to as the
double-counting term, which is not to be confused with the double counting used in
the (local density approximation + Hubbard parameter) LDA + U extension of density
functional theory.
One of the effects that the exchange interaction Eqn (1.5) or the exchange–correlation
interaction Eqn (1.21) has is that one electron can dig out a hole in the surrounding
density provided by all the other electrons. This hole is normally referred to as the
exchange–correlation hole (Gunnarsson and Lundqvist, 1976), in the density functional
12 Density Functional Theory
theory literature and, in fact, the exchange–correlation energy can be expressed as the
energy due to interaction of an electron with its exchange–correlation hole.
It has been shown that only the spherical average of the exchange–correlation hole
contributes to Exc (Gunnarsson and Lundqvist, 1976). Hence, it may be argued that,
even if the exact exchange–correlation hole may in general be strongly aspherical and
non-local, it is not necessary for approximate functionals, for example, in LDA, to
describe the non-spherical parts. Consequently, approximate functionals often do not
experience major obstacles in evaluating materials properties with good accuracy, as will
be described in Chapter 3.
Other differences between approximate functionals, like LDA, and exact functionals
or even the Hartree–Fock theory are well known, in particular, that the self-interaction
of the Hartree term cancels exactly the self-interaction of the Fock term. This means
that one may include terms i = j in the summations of Eqn (1.5) without causing
an error. This could, for example, be done in order to make the Hartree–Fock the-
ory more comparable to density functional theory, in which the contributions from all
electrons are used to evaluate the density. Including all electrons in the terms of the
Hartree–Fock theory therefore does not introduce an error, and one may say that the
Hartree–Fock theory does not have a self-interaction error. However, most approxima-
tions of the exchange–correlation term of density functional theory do not generate a
self-interaction-free functional and, in many cases, this deficiency has been known to
cause unacceptable errors. Attempts to overcome this self-interaction error have been
made; the most popular form is given by Perdew and Zunger (1981). It has been pointed
out that this correction only partially removes the self-interaction (Lundin and Eriks-
son, 2001), and truly self-interaction-free functional forms of density functional theory
have been suggested. However, the functionals suggested so far have not been shown to
drastically improve the results of, for example, Perdew and Zunger (1981).
Parametrizations of μxc (nop (r)) in Eqn (1.22), in terms of electron density, are avail-
able, where the exchange part is proportional to n1/3 op . This was done by, for example,
Hedin and Lundqvist (1971) and is discussed in several textbooks (e.g. in Ashcroft and
Mermin, 1976; Marder, 2010). Extensions of this analysis of the exchange–correlation
properties of the uniform electron gas to spin polarized situations allow majority spin-up
(α) and minority spin-down (β) densities to not be the same, thus enabling calcu-
lations of finite magnetic moments. The exchange–correlation potential for electrons
of a specific spin density nα is shown from the work of Barth and Hedin (1972)
to be
1
nα (r) 3
μαxc [nα (r), nβ (r)] = A[nop (r)] + B[nop (r)], (1.25)
nop (r)
spin polarization, one must treat the Kohn–Sham equation, Eqn (1.16), separately for
the majority spin-up and minority spin-down states.
1 α
– ∇ 2 + Veff (r) ψiα (r) = i ψiα (r),
2
1 β
– ∇ 2 + Veff (r) ψiβ (r) = i ψiβ (r), (1.26)
2
respectively. The effective potential for the majority spin channel becomes a little deeper
than that for the minority spin channel, which is why an imbalance between majority
and minority spin states emerges in the first place. The difference between the effect-
ive potential of majority spin states and that of minority spin states often amounts to a
constant shift between the electron states of the two spin channels. This shift is referred
to as exchange splitting. It can be seen from Eqn (1.25) that the exchange contribution
to the effective potential is proportional to the ‘local density’ of spin-up and spin-down
states, and again a spin polarized calculation has to be done self-consistently, as de-
scribed above, but now for each spin channel. Alternatively, one may express this as
self-consistency which has to be achieved both for the charge density n = nα + nβ and for
the magnetization density m = nα – nβ . The latter description is to be preferred since, in
the self-consistent cycle, a larger mixing of densities can be used for m than for n. Spin-
polarized parameterizations of the LDA, normally referred to as the local spin density
approximation (LSDA), have been established for some time (Barth and Hedin, 1972).
The LSDA is known to give accurate total energies for many bulk systems. This theory is
also known to reproduce with great accuracy the magnetic spin moments of most tran-
sition metals and their alloys. As an example, we show in Fig. 1.3 the measured and
Magnetic moment per atom (μB)
2 bcc
fcc
hcp
1
Fe alloy Co
calculated (using LSDA) spin moments of Fe-Co alloys (James et al., 1999). Note that
we have calculated spin moments by multiplying the integrated spin density with the
electron g-factor. This corresponds to the standard formulation in quantum mechan-
ics, where the size of the spin moment of one electron is calculated from ms = gs sz ,
where gs = 2.002319, to get moments in units of magnetic moment per atom (μB ; used
throughout this book). The phase diagram of the alloys in Fig. 1.3 shows that the crystal
structure is body-centred cubic (bcc), for almost all concentrations. Only for the very
Co-rich alloys does the hexagonal closed packed (hcp) crystal structure become stable.
Also, for some concentrations, the face-centred cubic (fcc) phase has been stabilized,
as precipitates in an fcc matrix of Cu. In Fig. 1.3 we see that, for the bcc phase, the
measured and calculated magnetic moments (per ‘average’ atom of the alloys) agree
with good accuracy. Theory is even found to reproduce the maximum value of the spin
moment, the Slater–Pauling maximum, at ∼25 % Co. For the fcc and hcp phases, theory
also reproduces observation, for cases where a comparison can be made.
where the wave functions are two-component, time-dependent Pauli spinors; ψi (r, t) =
[ψ∗iα (r, t), ψ∗iβ (r, t)]T ; and the summation is over the occupied Kohn–Sham orbitals. The
charge density and the magnetization density are calculated as traces of the density
matrix:
where σ̂ = (σ̂x , σ̂y , σ̂z ) is a vector of Pauli matrices. Reciprocally the density matrix can
be expressed in terms of n(r, t) and m(r, t) as
1
ρ(r, t) = n(r, t)1 + m(r, t) · σ̂ , (1.29)
2
where 1 is the 2 × 2 identity matrix. Note that in relationship to Eqn (1.26), we have
now made a generalization in that we consider non-collinear arrangements of magnetic
moments, and hence the Kohn–Sham equation must be written in a somewhat more
general form. However, before doing this, we note that Eqn (1.28) also contains the time
dependence of the charge and magnetization density, and hence we should consider the
time-dependent Kohn–Sham equation, normally written as a Pauli–Schrödinger-type
equation,
∂ψiα (r, t)
i = Hαβ ψiβ (r, t), (1.30)
∂t
for single-particle orbitals ψiα (r, t). Note that, in Eqn (1.30), we have used the Einstein
convention of implicit summation over repeated indices. If an external magnetic field
Bext is included together with the spin–orbit coupling, the Hamiltonian for the electron
system is written as
∇2 1
Ĥαβ = – δαβ + Vαβ (r, t) +
eff
σ̂ · B (r, t)
eff
2 2c αβ
1
+ σ̂ · ∇V eff (r, t) × i∇ , (1.31)
4c2 αβ
eff
Vαβ (r, t) = V ext (r, t)δαβ + V H (r, t)δαβ + μxc
αβ (r, t)δαβ (1.32)
consists of the external potential V ext (r, t), the Hartree potential V H (r, t), and the non-
magnetic part of the exchange–correlation potential μxc αβ (r, t). The effective magnetic
field
can be decomposed into the weaker external magnetic field Bext and the stronger
exchange–correlation magnetic field Bxc (r, t). The scalar and magnetic exchange–
correlation potentials are calculated, following Eqn (1.23), as the functional derivatives
of the exchange–correlation energy
16 Density Functional Theory
δE xc [n, m]
μxc (r, t) = , (1.34)
δn(r)
δE xc [n, m]
Bxc (r, t) = – . (1.35)
δm(r)
We have in this section provided the most general form of a practical scheme to calcu-
late magnetic properties from density functional theory, in the sense that no restriction
on the shape of the charge or on the magnetization direction is imposed and that time
dependence is included. From this formalism, one can evaluate the ‘direction’ of the
magnetization density, by diagonalization of the density matrix Eqn (1.27) (Nordström
and Singh, 1996; Sandratskii, 1998; Kübler, 2000). This is described in Chapter 4,
where we connect Eqn (1.31) to the concepts of atomistic spin dynamics. For most ma-
terials, the ‘direction’ of the magnetization density does not vary over regions where it
is large; typically, this region is centred on each atomic nucleus, within a radius that is
significantly smaller than interatomic distances. Examples of this are given in Chapter 2,
and we note that this property implies that a collinear atomic description is meaningful
in such a region (Sandratskii, 1998; Kübler, 2000), although examples of the oppos-
ite property have been discussed (Nordström and Singh, 1996). Before entering the
discussion of how to derive from Eqn (1.31) the relevant equations of atomistic spin
dynamics, we describe in Chapter 2 those aspects of a solid that are important when one
solves the Kohn–Sham equation. We also give some examples of the magnetic properties
of materials and how theory compares with observations (Chapter 3).
2
Aspects of the Solid State
Symmetries play an important role in the theory of the solid state. As will be de-
veloped in this chapter, density functional theory (DFT) calculations for crystalline
materials are commonly performed for k-points in the irreducible part of the first
Brillouin zone (BZ), an approach which relies on the use of translational and point-
group symmetries. Two central properties that result from a calculation in reciprocal
space are the wavevector-resolved energy spectra, that is, the so-called band structure
or band dispersion, and the energy-resolved density of states. For magnetic materials,
atomic magnetic moments can be defined and calculated, as well as effective inter-
atomic exchange interactions. In this chapter, the essential aspects of such calculations
are described.
Atomistic Spin Dynamics. Olle Eriksson, Anders Bergman, Lars Bergqvist, Johan Hellsvik. First Edition.
© Olle Eriksson, Anders Bergman, Lars Bergqvist, Johan Hellsvik 2017. First published in 2017 by Oxford University Press.
18 Aspects of the Solid State
where N is a (large) integer. It is then possible to show that electrons moving through
an infinite, periodic crystal must obey Bloch’s theorem (Ashcroft and Mermin, 1976;
Marder, 2010), that is, that the one-electron wave function (e.g. the solution to the
Kohn–Sham equation) must obey
A vector k has been introduced in this expression. This is a vector of reciprocal space.1
In self-consistent DFT-based calculations one has only to consider k-vectors which lie
inside the first BZ for the calculation of charge and magnetization density. The BZ is
defined as the Wigner–Seitz primitive cell of the reciprocal lattice and, as an example, we
show in Fig. 2.1 the BZ for the bcc crystal structure. Note that a smaller region is marked
that describes the so-called irreducible wedge of the BZ. Most crystal structures can
undergo a rotation along one or several axis, such that the rotated geometry is identical to
the starting geometry. These rotations define the point group of the crystal, and we give
one example of a point group below. The point group operations imply, importantly,
1
Reciprocal space is spanned by the vectors Gi , defined as Gi · Rj = 2πδij .
Exploring the Variety of Random
Documents with Different Content
tuomittu kolmeksi vuodeksi linnavankeuteen», sanoi hän tietäen
tämän uutisen koskevan minuun yhtä kipeästi kuin häneenkin.
Kirjeen sävy oli hyvin rauhallinen. Hän koetti lohduttaa omaisiaan ja
sanoi tietäneensä, mikä häntä odotti tuon artikkelin jälkeen.
Senpävuoksi hän ei myöskään surrut, sitä vähemmän kun hänellä ei
enää, kuuroille korville kaikuneen kehoitushuutonsa jälkeen, ollut
mitään sanottavaa.
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
textbookfull.com