Noncollinear Ferromagnetism in (III,Mn)V Semiconductors
Extracted source text ยท arXiv cond-mat/0107573
\documentstyle[aps,prl,epsfig,float,twocolumn]{revtex}
\begin{document}
\twocolumn[\hsize\textwidth\columnwidth\hsize\csname
@twocolumnfalse\endcsname
\title{Noncollinear Ferromagnetism in (III,Mn)V Semiconductors}
\author{John Schliemann and A. H. MacDonald}
\address{Department of Physics, The University of Texas, Austin, TX 78712}
\address{Department of Physics, Indiana University, Bloomington, IN 47405}
\date{\today}
\maketitle
\begin{abstract}
We investigate the stability of the collinear ferromagnetic state
in kinetic exchange models for (III,Mn)V semiconductors with randomly
distributed Mn ions . Our results
suggest that {\em noncollinear ferromagnetism} is commom to these
semiconductor systems. The instability of the collinear state
is due to long-ranged fluctuations involving a large fraction of
the localized magnetic moments.
We address conditions that favor the occurrence
of noncollinear groundstates and discuss unusual behavior that we
predict for the temperature and field dependence of its saturation
magnetization.
\end{abstract}
\vskip2pc]
In systems with local moments, interactions that favor parallel spin
alignment ({\em ferromagnetic} interactions) yield ferromagnetic ground states.
Interactions that favor antiparallel spin alignment ({\em antiferromagnetic}
interactions), on the other hand, yield antiferromagnetism only in the special
case of short-range interactions on bipartite lattices. More generally,
longer range antiferromagnetic interactions often lead to
complex order or, when spatial disorder is also present, to
spin-glass ground states \cite{Yosida}.
In this context, the occurrence of
ferromagnetism\cite{expgen} in III-V semiconductors with
Mn substituted on a relatively small fraction of the cation sites is unusual,
since Mn provides $S=5/2$ local moments in this system and the distribution
of Mn ions among the cation sites is thought to be random.
A ground state with parallel alignment
can be understood in a system with randomly located local moments only if
interactions are dominantly ferromagnetic, a property not normally
associated with the carrier-mediated interactions thought\cite{meanfield} to
be responsible for ferromagnetism in
these systems. Indeed there is ample evidence that randomness can play an
important role in these ferromagnets, since transport and magnetic
properties depend strongly on the MBE growth conditions used for materials
synthesis, with reproducible results achieved only if growth protocols are
followed carefully \cite{expgen}.
An important step toward the systematic
analysis of such effects was taken very recently by Potashnik {\it et al.}
\cite{Schiffer} who studied the effect of annealing procedures on magnetic
and transport properties for samples with a Mn fraction close to that at
which the maximum Curie temperatures ($T_c$) occurs, $x \sim 0.05$.
They find that $T_c$, the ground
state saturation magnetization $M(T=0)$, and the shape of the $M(T)$ curve
all depend on the temperature and duration of annealing.
Since the Mn content of the system is not believed to be changed by
annealing, changes in the $T=0$ magnetization can be explained only
if Mn moments are not always, or perhaps never, completely aligned
in the ground state. Moreover, drastic changes in the shape of the $M(T)$
curve are evidence of changes in the magnetic excitation spectrum.
Both changes most likely reflect the sensitivity of carrier-mediated
interactions between the Mn local moments to the defect distribution that is
varied by annealing.
In this Letter we study the stability of states with parallel
Mn local moments using the kinetic-exchange model of carrier-induced
ferromagnetism. We find that fully aligned ferromagnetic ground states do
not occur
when the band electrons wavefunctions are strongly concentrated around Mn ion
locations, and propose that this circumstance must be avoided to achieve
high ferromagnetic transition temperatures.
On the other hand we
find, consistent with the observations of Potashnik {\it et al.},
that noncollinear ferromagnetism is common in (III,Mn)V semiconductors.
On the basis of calculations of external-field dependent magnetic excitation
spectra, we discuss unusual temperature and field dependence
of the magnetization that we predict will be characteristic of non
collinear states in these ferromagnets.
Ferromagnetism in (III,Mn)V semiconductors has been addressed using
first-principles local-spin-density approximation calculations
\cite{theo-lsda}, phenomenological impurity-band models\cite{bhatt} and,
most-commonly\cite{meanfield,spinwave,mc,dassarma}, using a
semi-phenomenological approach that sidesteps atomic scale physics by
using an envelope function description of the bands. The last approach is
strongly supported by the large literature on the closely related
(II,Mn)VI semiconductors \cite{dmsreviews} and we follow it
here. The model Hamiltonian we study,
\begin{equation}
{\cal H}=\sum_{i}\frac{\vec p_{i}\,^{2}}{2m}
+\sum_{I}\int\,d^{3}r\,J(\vec r-\vec R_{I})\vec s(\vec r\, ) \cdot \vec S_{I}
\, .
\label{defmod1}
\end{equation}
describes noninteracting carriers in a parabolic band
characterized by an effective mass $m$, whose spin density
$\vec s(\vec r\, )$ is coupled to ionized Mn acceptor spins $\vec S_{I}$
at locations $\vec R_{I}$ by an spatially extended antiferromagnetic
exchange coupling $J(\vec r\, )$, which we take to be of the form
\begin{equation}
J(\vec r \,)=\frac{J_{pd}}{(2\pi a_{0}^{2})^{\frac{3}{2}}}
e^{-\frac{r^{2}}{2 a_{0}^{2}}}\,,
\label{excpl}
\end{equation}
where $J_{pd}$ parametrizes the strength and $a_{0}$ the range of the
interaction.
Similar models have been studied some time ago\cite{Abrikosov} in connection
with classical spin glass systems; in that case however the density of
magnetic impurities $N$ is
taken to be much smaller than the density of charge carriers $n$,
$n/N\gg 1$ \cite{Abrikosov}. (III,Mn)V semiconductors are in the opposite,
$n/N\ll 1$ because of strong compensation of Mn acceptors by antisite defects
$N$.
The model (\ref{defmod1})
is unrealistic in its use of a single parabolic band to
represent the semiconductor valence bands, in its neglect of Coulomb
interactions between holes in the valence band
and other holes and Mn acceptors, and in the absence of any representation
of the defects (other than the Mn ions) whose
distributions presumably vary during the annealing process.
We return to these limitations of the model later. In addition, the
range of the exchange interaction in this
model accounts \cite{isolatedMn} for the atomic scale physics only crudely.
We believe, nevertheless, that this relatively simple model
captures much of the physics associated with the
approach of and the occurrence of reduced saturation magnetization states
for non-optimal defect configurations.
Recently a theory of magnetic fluctuations around a fully aligned state
of magnetic ions was developed using the {\em virtual crystal}
approximation \cite{spinwave}.
In contrast to that work
we do not adopt this simplifying approximation in
which the Mn magnetic distribution is replaced by a continuum,
retaining individual spins of length $S=5/2$ placed at
arbitrary locations instead.
The virtual {\em crystal approximation},
which has a long and successful history
in analyzing the properties of (II,Mn)VI ferromagnets, completely
eliminates disorder and, as we discuss further below, noncollinear
ferromagnetism.
We first address the stability
of the collinear state in which all Mn moments have parallel orientation
taken to be in the $\hat z$ direction,
parameterizing changes in the orientation of spin $I$ by a complex number,
$z_I=(S^{x}_{I}+iS^{y}_{I})/\sqrt{2S}$.
The zeroth-order band electron Hamiltonian is given by Eq.~(\ref{defmod1}) with
${\vec S}_I = S \hat z$. For small changes in spin orientation
the change in the band electron Hamiltonian is
\begin{equation}
{\cal H}_{1}=\frac{1}{2}\sum_{I}\left[J(\vec r-\vec R_{I})
\left(\begin{array}{cc}
-z_{I}\bar z_{I} & \sqrt{2S}\bar z_{I} \\
\sqrt{2S}z_{I} & z_{I}\bar z_{I}.
\end{array}\right)
\right]
\label{perturb}
\end{equation}
The contribution to the carrier groundstate energy of first order in
the $z_{I}$ vanishes. At second order one obtains
\begin{equation}
E^{(2)}=\sum_{I,J}\bar z_{I}M_{IJ}z_{J}
\label{flucaction}
\end{equation}
where $M_{IJ}=L_{IJ}+K_{IJ}$ and
\begin{eqnarray}
L_{IJ} & = &-\delta_{IJ}
\int d^{3}rJ(\vec r-\vec R_{I})\langle s^{z}(\vec r)\rangle
\label{L}\\
K_{IJ} & = & \frac{S}{2}\sum_{\alpha,\beta}\Biggl[
\frac{n_{F}(\xi_{\alpha})-n_{F}(\xi_{\beta})}
{\xi_{\alpha}-\xi_{\beta}}
F^{\alpha\downarrow,\beta\uparrow}_{I}
F^{\beta\uparrow,\alpha\downarrow}_{J}\Biggr]
\label{K}
\end{eqnarray}
Here $\langle\vec s(\vec r)\rangle$ is the band-electron spin density in the
collinear state, $n_{F}$ is the Fermi function at zero temperature
In these equations we have defined
\begin{equation}
F^{\alpha\sigma,\beta\mu}_{I}=\int d^{3}rJ(\vec r-\vec R_{I})
\bar\psi_{\alpha\sigma}(\vec r)\psi_{\beta\mu}(\vec r)
\label{F}
\end{equation}
with $\psi_{\alpha\sigma}(\vec r)$ being the spin $\sigma$ component of the
the carrier wave function with energy $\varepsilon_{\alpha}=
\xi_{\alpha}+\mu$, where $\mu$ is the Fermi energy.
All eigenvalues of $M$ must be positive for the collinear state to be stable.
The diagonal contribution to $M$ from the positive definite matrix (\ref{L})
gives the energy change
associated with reorienting each localized spin in the mean-field created by
the polarized
band electrons, while the contributions from the matrix (\ref{K})
represents the energy reduction due to the response of band electrons to
reorientation of the localized spins.
When the Mn spins are treated quantum mechanically using a
Holstein-Primakoff boson representation \cite{Auerbach} and
coherent state path integrals the $z_I$ are boson coherent state labels,
eqivalent to replacing $z_I$ and $\bar z_{I}$ by bosonic creation and
annihilation operators $a_{I}$ and $a^{+}_{I}$. Then Eq.(~\ref{perturb}) is the
spin-wave Hamiltonian,
and the eigenvalues\cite{retarded} of $M_{IJ}$ are the system's
spin-wave energies.
%
\begin{figure}
\centerline{\includegraphics[width=8cm]{fig1.eps}}
\caption{Upper panel:
Disorder-averaged density of states for spin-wave excitations.
These results were obtained with a simulation cube of volume
$V=L^{3}=400{\rm nm}^{3}$ with a Mn density of $N=1.0{\rm nm}^{-3}$,
a carrier density $n=0.15{\rm nm}^{-3}$, and $m=0.5 m_{el}$.
The strength of the exchange interaction
between ions and carriers is $J_{pd}=0.05{\rm eVnm}^{3}$ with a spatial
range of $a_{0}=0.40{\rm nm}$.
Lower panel: Disorder-averaged participation ratios for the same system.
In both graphs the value at zero energy is enhanced
due to the contribution of the uniform rotation mode which occurs for any
disorder realization.
\label{fig1}}
\end{figure}
%
We have evaluated the matrix $M_{IJ}$ and its spectrum by solving
for the band electron mean-field eigenstates, applying periodic boundary
conditions to cubic simulation cells.
The single-particle wavefunctions $\psi_{\alpha\sigma}(\vec r)$ are computed
in a plane-wave basis taking into account wavevectors $\vec q$ with length up
to an appropriate cutoff $q_{c}$.
The upper panel of Fig.~\ref{fig1} shows typical results
for the eigenvalue distribution of $L_{IJ}$ averaged over
a large number of randomly chosen Mn distributions.
We have established by further calculations that the effects of the wavevector
cutoff on the low-lying excitations have already
saturated for the value of $q_{c}$ used here.
We note that for any arrangement of the Mn positions $\vec R_{I}$ the
matrix $M_{IJ}$ contains a zero eigenvalue corresponding to a uniform
rotation of all spins where all components $z_{I}$ are the same. This implies
summing over all columns of $M$ gives the zero vector, a sum rule that
is perfectly fulfilled in our numerics independently of
$q_{c}$ since this plane wave cutoff does not influence the rotational
invariance in spin space.
In the above calculations the Mn positions were chosen
completely at random with a uniform distribution, while in a real
(III,Mn)V semiconductor the Mn ions are located on
the fcc lattice of cation sites. Our disorder-averaged
data for the two cases is are indistinquishable, however, establishing that
atomic-scale correlations in Mn locations do not alter our results.
Our most important finding, illustrated in Fig.~\ref{fig1} and discussed
at length below, is that that for a wide range of parameter values
{\em there are always a few negative eigenvalues, i.e. that the collinear
ferromagnetic state is not stable.}
To analyze the nature of this instability we have evaluated
participation ratios for these elementary excitations which we define by
\begin{equation}
p_{j} =\left[NV\sum_{I}|\alpha^{j}_{I}|^{4}\right]^{-1}
\label{partrat}
\end{equation}
where $\alpha^{j}_{I}$ is the $I$-th component of the $j$-th normalized
eigenvector of $M_{IJ}$. $p_{j}$ is an estimate for the fraction of
Mn sites that have important involvement in
the $j$-th spin wave. For instance, if
a vector contains exactly a fraction of $p$ nonzero components of equal
modulus and all others being zero, its participation ratio is $p$.
The largest participation ratio of unity is achieved for the zero-energy
uniform rotation mode where all components of the corresponding
eigenvector are equal.
The lower panel of
Fig.~\ref{fig1} shows the disorder-averaged participation ratio as a
function of spin-wave energy for the same situation as in the top panel.
The property that negative energy excitations have large
participation ratios shows that the instabilities of the collinear state
involve correlated reorientations of many spins, rather than lone loosely
coupled moments. It is the high energy excitations, for which the
mean-field term dominates, that are single-ion in character
with participation ratios comparable to $1/400$ where $400$ is the number of
Mn ions in these simulations cells.
The shape of the spin-wave density-of-states is, in the model we have studied,
sensitive to the Mn density $N$, the carrier density $n$, and to the
Hamiltonian parameters $m$, $J_{pd}$, $a_{0}$. Situations in which the
collinear ferromagnetic state is, for certain disorder realizations,
stable can be approached most simply (for technical reasons)
by letting $a_{0}$ be larger, but also we believe for larger $n/N$.
For the value of $n/N$ illustrated in Fig.~\ref{fig1},
which corresponds to a carrier density somewhat smaller than
that measured in the highest $T_c$ samples,
negative eigenvalues occur for nearly any Mn ion distribution.
We note that negative eigenvalues increase in number as the
wavevector cutoff is increased toward its converged value.
These results suggest that noncollinear ferromagnetic states are common,
and that they are sensitive to the distribution of Mn ions and defects,
especially those that trap carriers. We find that the collinear ferromagnetic
state tends to become unstable as mean-field band eigenfunctions become
more strongly localized around Mn ion sites.
The effect of a weak external magnetic field on the eigenvalue spectrum
of $M_{IJ}$ is particularly simple. The field
couples to the local moment through its Land\'e g-factor,
adding $2 \mu_B H$ to the energy of a spin-wave
excitation for $S=5/2$, and to the spin and orbital degrees of freedom
of the band electrons. The orbital coupling leads to Landau levels
that do not play an important role at weak fields in these highly disordered
samples. Zeeman coupling is
also unimportant, because it gives a contribution to the
spin-splitting that is
negligible compared to the mean-field exchange splitting
$\Delta=J_{pd}NS\sim 0.1 {\rm eV}$.
Thus, provided that the modulus of the smallest eigenvalue
(i.e. the onset of the eigenvalue distribution of $M_{IJ}$ at negative
energies) is small compared to $\Delta$ (as it is the case in data shown
above), this modulus is the
value of $g_L \mu_B H$ necessary to force full spin alignment of a
noncollinear state. The experiments of Potashnik {\it et al.} demonstrate
that the maximum value of $M(T=0)$ is achieved over a certain range of
annealing histories.
We associate this maximum value with
the fully aligned collinear Mn configuration state; indeed the maximum
moment per Mn is consistent with full alignment partially compensated by
band electrons. In this circumstance, the {\em virtual crystal approximation}
used in earlier spin-wave calculations which neglected
disorder\cite{spinwave} can be accurate.
For other annealing histories, $M(T=0)$ is reduced
corresponding, we propose, to noncollinear order of the Mn spins.
Calculations like those described above show that these spins gradually
align as an external field is added to the Hamiltonian.
Since full alignment is achieved at a finite external field, we expect full
alignment to be indicated experimentally by a kink in the
$M(T=0,H)$ curve. At this point, we predict that
the system will still have gapless excitations and power-law temperature
dependence of the magnetization, in sharp contrast to the gapful excitations
and exponentially suppressed temperature dependence of
conventional ferromagnets in an external
magnetic field. At the same time, spin-resonance experiments will see a
gapped spin-wave spectrum since they couple only to the
uniform-rotation mode which has finite energy $g_L \mu_B B$
over the noncollinear groundstate.
To address the noncollinear state, we generalize our formalism by expanding
around and
defining Holstein-Primakoff bosons \cite{Auerbach}
with respect to a general noncollinear
orientation configuration, of each Mn spin,
${\vec S}_{I}=S\hat\Omega_{I}=S(\sin\vartheta_{I}\cos\varphi_{I},
\sin\vartheta_{I}\sin\varphi_{I},\cos\vartheta_{I})$.
If this configuration is not an extremum, there will be an energy correction
that is {\em linear} in the Holstein-Primakoff variables:
\begin{equation}
E^{(1)}=\frac{1}{2}\sum_{I}
\left[\bar g_{I}z_{I}+g_{I}\bar z_{I}\right]
\label{gradientfluc}
\end{equation}
with $g_{I}=g^{1}_{I}+ig^{2}_{I}$, and
\begin{eqnarray}
g^{1}_{I} & = & \sqrt{2S}\left(\vec e_{\varphi_{I}}\times\vec e_{z}\right)
\cdot\int d^{3}r\Bigl[J(\vec r-\vec R_{I})\nonumber\\
& & \Bigl(\left(\langle\vec s(\vec r)\rangle\cdot\vec e_{\varphi_{I}}\right)
\vec e_{\varphi_{I}}+
\left(\langle\vec s(\vec r)\rangle\cdot\vec e_{z}\right)\vec e_{z}
\Bigr)\times\vec \Omega_{I}\Bigr]\\
g^{2}_{I} & = &\sqrt{2S}\vec e_{z}\cdot
\left(\vec e_{\varphi_{I}}\times\int d^{3}rJ(\vec r-\vec R_{I})
\langle\vec s(\vec r)\rangle\right)
\end{eqnarray}
where $\vec e_{\varphi_{I}}=(\cos\varphi_{I},\sin\varphi_{I},0)$ and
$\vec e_{z}=(0,0,1)$.
The contributions to the carrier ground state energy quadratic in the
Holstein-Primakoff variables can also be obtained via perturbative
calculations similar to those for the the collinear case.
The components $g_{I}$ represent the gradient of the
energy with respect to deviations, parameterized by the $z_{I}$, from the
reference
orientations. A given orientation of Mn spins is stationary with
respect to fluctuations
if all complex coefficients $g_{I}$ vanish. This is the case if and only if
$\int d^{3}rJ(\vec r-\vec R_{I})\langle\vec s(\vec r)\rangle$
is parallel with the direction $\hat\Omega_{I}$ of the local ion spin.
(The collinear ferromagnetic state is always stationary but not necessarily
stable.)
We have employed the energy gradient expression (\ref{gradientfluc}) in a
numerical steepest descent procedure to search for true energy minima.
Briefly, our results are as follows. In cases where the energy minimum
found by this method is close to that of the collinear state
(with $M(T=0)$ about
90 or more percent of the maximum value), this minimum appears to be
unique for each disorder realization.
We can therefore be confident that we have located
the absolute ground state of our model. In situations
where the magnetisation is reduced more substantially, however, by about 20
percent or more typically, we converge to different
energy minima from different starting points.
In these case the model has substantial
spin-glass character with a complex energy landscape.
For the system shown in Fig.~\ref{fig1}, for
instance, magnetisation values at local energy minima are typically
30 to 40 percent of the collinear state value.
We thank T. Dietl, J. K\"onig, B. Lee, F. von Oppen, N. Samarth, P Schiffer,
C. Timm, and S.-R.~E. Yang for useful discussions.
This work was supported by the
Deutsche Forschungsgemeinschaft, the Indiana 21st Century Fund, DARPA/ONR
Award No. N00014-00-1-0951, and the Welch foundation.
\begin{thebibliography}{50}
\bibitem{Yosida}
For a general overview see e.g. K. Yosida, {\it Theory of Magnetism},
Springer, (1996).
\bibitem{expgen}
For reviews see
H. Ohno, Science {\bf 281}, 951 (1998);
H. Ohno, J. Magn. Magn. Mater. {\bf 200}, 110 (1999);
H. Ohno and F. Matsukura, Solid State Commun. {\bf 117}, 179 (2001).
\bibitem{meanfield}
T. Dietl, A. Haury, and Y.~M. d'Aubign{\'e},
Phys. Rev. B {\bf 55}, R3347 (1997);
T. Dietl, H. Ohno, F. Matsukura, J. Cibert, D. Ferrand,
Science {\bf 287}, 1019 (2000);
T. Dietl, H. Ohno, and F. Matsukura,
Phys. Rev. B {\bf 63}, 195205 (2001);
M. Abolfath, J. Brum, T. Jungwirth, and A.~H. MacDonald,
Phys. Rev. B {\bf 63}, 054418 (2001).
\bibitem{Schiffer}
S.~J. Potashnik, K.~C. Ku, S.~H. Chun, J.~J. Berry, N. Samarth,
and P. Schiffer, Appl. Phys. Lett. {\bf 19}, 1495 (2001).
The dependence of the crystal properties of these ferromagnets
on the growth conditions has also been
studied recently by G.~M. Schott, W. Faschinger, and L.~W. Molenkamp,
cond-mat/0105562.
\bibitem{theo-lsda}
H. Akai, Phys. Rev. Lett. {\bf 81}, 3002 (1998);
S. Sanvito, P. Ordejon, and N.~A. Hill, Phys. Rev. B {\bf 63}, 165206 (2001);
S. Sanvito and N.~A. Hill, Appl. Phys. Lett. {\bf 78}, 3493 (2001);
M. van Schilfgaarde and O.~N. Mryasov, Phys. Rev. B {\bf 63}, 233205 (2001).
\bibitem{bhatt}
X. Wan and R.~N. Bhatt, cond-mat/0009161;
M. Berciu and R.~N. Bhatt, Phys. Rev. Lett. {\bf 87}, 107203 (2001);
M.~P. Kennett, M. Berciu, R.~N. Bhatt, cond-mat/0102315.
\bibitem{spinwave}
J. K\"onig, H.~H. Lin, and A.~H. MacDonald,
Phys. Rev. Lett. {\bf 84}, 5628 (2000);
cond-mat/0010471, published in {\it Interacting Electrons in
Nanostructures}, edited by R. Haug and H. Schoeller, Springer (2001);
J. K\"onig, T. Jungwirth, and A.~H. MacDonald,
Phys. Rev. B {\bf 64}, 184423 (2001).
\bibitem{mc}
J. Schliemann, J. K\"onig, H.~H. Lin, and A.~H. MacDonald,
Appl. Phys. Lett. {\bf 78}, 1550 (2001);
J. Schliemann, J. K\"onig, and A.~H. MacDonald,
Phys. Rev. B {\bf 64}, 165201 (2001).
\bibitem{dassarma} A. Chattopadhyay, S. Das Sarma, and A.~J. Millis,
cond-mat/0106455.
\bibitem{dmsreviews} J.~K. Furdyna, J. Kossut, {\it Diluted Magnetic
Semiconductors}, in {\it Semiconductors and Semimetals}, volume 25,
Academic Press (1988);
T. Dietl, {\it Diluted Magnetic Semiconductors}, in {\it Handbook of
Semiconductors}, volume 3B, North-Holland (1994).
\bibitem{Abrikosov}
For a review see A.~A. Abrikosov, Adv. Phys. {\bf 29}, 869, (1980).
\bibitem{isolatedMn} A.~K. Bhattacharjee and C. Benoit $\grave { \rm a}$
la Guillaume, Solid State Commun. {\bf 113}, 17 (2000);
V.~I. Litvinov and V.~K. Dugaev, Phys. Rev. Lett.
{\bf 86}, 5593 (2001).
\bibitem{Auerbach}
See e.g. A. Auerbach, {\it Interacting Electrons and Quantum Magnetism}
Springer, (1994).
\bibitem{retarded} Strictly speaking $K_{IJ}$ is frequency dependent in a
quantum formulation, reflecting the retarded nature of the carrier-mediated
interaction between
local moments. The frequency dependence is weak however, because of the
small ratio between spin-wave excitation energies and the mean-field
spin-splitting of the itinerant electron bands.
It can be shown that, for high participation ratio
low frequency excitations, the only role
of the frequency dependence is to incorporate the band electron
contribution to the magnetization and to rescale the collective mode
energies by a factor of $1/(1-n/2SN)$.
Terms at second and higher order in the low frequency expansion of $K_{IJ}$
reflect correlated fluctuations between local moment and itinerant electron
spins that supress the total magnetization by a further small factor.
\end{thebibliography}
\end{document}