Magnon-Paramagnon Effective Theory of Itinerant Ferromagnets
Extracted source text ยท arXiv cond-mat/9908024
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%% This is RevTex 3.0 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\documentstyle[preprint,aps]{revtex}
\tightenlines
\begin{document}
\draft
\title{
Magnon-Paramagnon Effective Theory of
Itinerant Ferromagnets
}
%
\author{Naoum Karchev\cite{byline}}
%
\address{
Department of Physics, University of Sofia, 1126 Sofia, Bulgaria
}
%\date (Received \today)
\maketitle
%
\begin{abstract}
The present work is devoted to the derivation of an effective magnon-paramagnon
theory starting from a microscopic lattice model of ferromagnetic metals.
For some values of the microscopic
parameters it reproduces the Heisenberg theory of localized spins.
For small magnetization the effective model describes the physics
of weak ferromagnets in accordance with the experimental results.
It is written in a way which keeps $O(3)$ symmetry manifest,and describes both
the order and disordered phases of the system.
Analytical expression for the Curie temperature,which takes the magnon
fluctuations into account exactly, is obtained.
For weak ferromagnets $T_c$ is well below the Stoner's critical temperature
and the critical temperature obtained within
Moriya's theory.
\end{abstract}
%
\pacs{75.10.Lp,71.27.+a,75.40.Gb}
{\bf I Introduction}
The Heisenberg model of ferromagnetism, based on the exchange interaction
of localized electrons gives an explanation of many properties of
non-conducting magnetic systems at both low and high temperatures. However the
development of a satisfactory theory of ferromagnetic metals has ran into
difficulties.
Theories of weak ferromagnetic metals based on the Landau Fermi liquid theory
have been developed by several theorists \cite{mp1,mp2,mp3,mp4}. The
spectrum of the spin excitations has been found. It consists of spin
fluctuations of paramagnon type and a transverse spin-wave branch.
Murata and Doniach \cite{mp6} have proposed a phenomenological mode-mode
coupling theory to describe the temperature dependence of the quantities for
a weak ferromagnet. However, they started from a classical Hamiltonian and
ignored the quantum effects, which are important because of the low
temperature nature of the weak ferromagnet.
An alternative approach to magnetic phase transition in Fermi systems
has been developed by Moriya and Kawabata \cite{mp10,mp10a}. It is a
self-consistent one loop approximation which accounts for the spin
fluctuations.
The nonlinear effects of spin fluctuations are treated and in \cite{mp10b},
using a self-consistent
rotationally invariant Hartree approximation.
New results concerning magnetic phase diagram in Hubbard-type
model have been obtained within the dynamical mean-field theory
\cite{mp11,mp12}.
Perhaps the most striking feature of the itinerant ferromagnets is
the quantum transition to paramagnet. In the paper \cite{mp7},
Hertz derives an effective paramagnon theory of a
paramagnet to ferromagnet quantum phase transition. He analyzed
the effective model by means of a renormalization group method
that generalizes Wilson's treatment of classical phase transition
and concludes that the critical behavior of itinerant ferromagnets
in dimensions larger then one are described by a mean-field fixed
point. Hertz's work has been reexamined in Ref.\cite{mp8}.
The results of this paper support those of Hertz in 3D case,
but in many aspects differ from
them in two dimensions.
More realistic model of ferromagnetic metals is considered in Refs
\cite{mp9,mp17}. It contains particle-hole spin-triplet
interaction that causes ferromagnetism, as well as particle-hole
spin-singlet and particle-particle interactions. The spin-triplet
interaction is decoupled by introducing a vector field whose
average is proportional to the magnetization, and performing a
Hubbard-Stratanovich transformation. Then the fermionic degrees of
freedom are integrated out accounting for the rest part of
interaction by means of perturbation theory. The resulting
effective field theory is nonlocal. It contains an effective
long-range interaction between the order parameter fluctuations.
The analysis in Refs \cite{mp9,mp17} is restricted to power
counting arguments at tree level. It shows that the critical
behavior is governed by Gaussian fixed point and that all
non-Gaussian terms are irrelevant in renormalization group sense.
Logarithmic corrections to power-low scaling are obtained in $D=3$
case. For $D<3$ the deviations from the mean-field theory results
depend on $D$.
There is a considerable interest in the finite temperature properties of weak
itinerant ferromagnets. It was obtained in \cite{mp4}, that in the region below
but close to the Curie temperature, the paramagnon contribution is dominant.
Experimentally, the transition in the weak ferromagnet $MnSi$ has been
investigated at different Curie temperatures by applying hydrostatic pressure
\cite{mp16}. The transition at high temperatures was found to be of second
order, while at lower transition temperatures it is of first order. In the
first order regime the transition temperature was found to scale with pressure
as $T_c\sim (p_c\,-\,p)^{\frac 12}$. The scaling law is explained by a
mean-field analysis, assuming a dynamical exponent $z=3$. But in Hertz's
theory \cite{mp7}, the dynamical critical exponent is equal to three
due to paramagnon excitations. Hence, in order to
correctly explain the characteristic features of itinerant ferromagnets it
should be important to work with an effective theory that keeps explicitly
the magnon as well as paramagnon excitations.
The present work is devoted to the derivation of an effective
magnon-paramagnon theory starting from a microscopic lattice model
of ferromagnetic metals. This is an effective theory of
ferromagnets, and for some values of microscopic parameters it
reproduces the Heisenberg theory of localized spins. On the other
hand, for small magnetization the effective model describes the
physics of weak ferromagnets in accordance with the experimental
results. Thus, the effective magnon-paramagnon theory interpolates
between theories of localized and itinerant electrons. The zero
temperature dimensionless magnetization per lattice site $m$ is
introduced to describe this interpolation. It parameterizes the
ground state of the system, and moving $m$ we change the ground
state and respectively the fluctuations above it. When $m$ is
maximal ($m=\frac 12$) all the lattice sites are occupied by one
electron with spin up in the ground state, and the only relevant
excitations are magnons. The effective theory is the Heisenberg
theory of localized spins. When the magnetization is smaller, i.e.
some of the lattice sites are doubly occupied or empty in the
ground state, the spectrum consists of paramagnon and magnon
excitations and the effective theory is a "spin $m$" Heisenberg
theory coupled to paramagnon fluctuations. Decreasing the
magnetization results in the changing the parameters of the spin
fluctuations, and close to quantum critical point $m=0$, the
paramagnon becomes important due to the singularity in the
paramagnon propagator. In the quantum paramagnetic phase $(m=0)$,
the magnon excitations disappear from the spin spectrum and one
obtains Hertz's effective model.
The ferromagnetic order parameter is a vector $\vec M$ field.
The transverse spin fluctuations (magnons) are described by $M_1+iM_2$
($M_1-iM_2$) and the longitudinal fluctuations (paramagnons) by
$M_3-<M_3>$. Alternatively the vector field can be written as a product of
its amplitude $\rho=\sqrt{M_1^2+M_2^2+M_3^2}$ and an unit vector $\vec n$,
$\vec M=\rho\vec n$. In the ferromagnetic phase one sets $M_3=<M_3>+\varphi$
and in linear (spin-wave) approximation obtains $\rho=<M_3>+\varphi$.
It is evident now, that the fluctuations of the $\rho$ field,
$\rho-<M_3>$ are exactly the paramagnon excitations in a formalism
which keeps $0(3)$ symmetry manifest. One can write the
effective theory in terms of $\vec M$-vector components, or in terms of $\rho$
and an unit vector $\vec n$. In ferromagnetic phase, in spin-wave
approximation the effective theories coincide. I use the parameterization in
terms of unite vector and spin singlet amplitude. Then, the effective action
keeps the $O(3)$ symmetry manifest, and describes both the order and disorder
phases of the system.
Above Curie temperature the spectrum consists of spin
singlet fluctuations of paramagnon type and spin-$\frac 12$ spinon
fluctuations. The spinon has a gap, but near the critical temperature it
approaches zero and spin-$\frac 12$ fluctuations as well as paramagnon are
essential describing the thermal phase transition of itinerant ferromagnets.
Analytical expression for the Curie temperature,which takes the magnon
fluctuations into account exactly, is obtained. For weak
ferromagnets the critical temperature scales like $T_c\sim m^{\frac 53}$.
It is well below the Stoner's critical temperature $T_c\sim m$
and the critical temperature obtained within
Moriya's theory $T_c\sim m^{\frac 32}$ \cite{mp19}.
Scaling arguments based on the Hert's renormalization group technique allow to
obtain a relation between spin stiffness and the Stoner's enhancement factor.
The relation is a consequence of the fact that quantum critical exponent is
equal to three. Making use of the relation one can obtain the contribution of
magnon scattering to the Stoner's enhancement factor.
To derive the effective model a Schwinger-bosons-slave-fermions
representation of the operators is used. The salient point is that
within this approach the local Coulomb repulsion is treated
exactly. As a result, the constants in the effective theory are
finite and well defined for all values of the magnetization $m$.
The paper is organized as follows. In Sec.II an effective magnon paramagnon
theory of ferromagnetic metals is obtained. The Curie temperature is
calculated and written in terms of magnetization and spin stiffness constant.
In Sec III, Hertz's renormalization group in tree approximation is extended
to include the magnon fluctuations. Section IV is devoted to the concluding
remarks.
\ \\
{\bf II Magnon-paramagnon effective model}
The simplest model of ferromagnetic metals is determined by the
Hamiltonian \cite{mp18}
\FL
\begin{equation}
\hat H = -t \sum\limits_{<i,j>,\sigma}
(\hat{c}^+_{i\sigma} \hat{c}_{j\sigma} + {\rm h.c.})
-J\sum\limits_{<i,j>}
{\hat{\vec {S}}}_i\cdot {\hat{\vec {S}}}_j
+ U \sum\limits_{i} \hat{n}_{i\uparrow} \hat{n}_{i\downarrow} -
\mu\sum\limits_i \hat n_i
\label{mp1}
\end{equation}
Here $\hat{c}^+_{i\sigma}$ and $\hat{c}^{\phantom +}_{j\sigma}$ are creation
and annihilation operators for electrons,
$\hat{n}_{i \sigma} = \hat{c}^+_{i \sigma}\hat{c}^{\phantom +}_{i \sigma}$
and $\hat{n}_{i} = \sum_{\sigma} \hat{n}_{i\sigma}$ are density operators,
and ${\hat{\vec{S}}}_{i}=1/2 \sum\limits_{\sigma\sigma'} \hat{c}^+_{i\sigma}
{\vec {\tau}}_{\sigma\sigma'} \hat{c}^{\phantom +}_{i\sigma'}$,
where ${\vec {\tau}}$ denotes the vector of Pauli matrices, are spin operators.
The sums are over all sites of a three-dimensional cubic lattice,
$<i,j>$ denotes the sum over the nearest neighbors, and $\mu$ is the chemical
potential.
In (\ref{mp1}) the $J$-term corresponds to direct Heisenberg exchange which is
generically ferromagnetic ($J>0$) in nature.
In terms of Schwinger-bosons ($\hat \varphi_{i,\sigma},\hat
\varphi_{i,\sigma}^{\dagger}$) and slave-fermions ($\hat h_i,\hat
h_i^{\dagger},\hat d_i,\hat d_i^{\dagger}$) the operators have the
following representation:
\FL
\begin{eqnarray}
& & \hat {c}_{i\uparrow} = \hat h_i^{\dagger}\hat \varphi _{i1}+
\hat \varphi_{i2}^{\dagger}\hat d_i,
\qquad
\hat {c}_{i\downarrow} = \hat h_i^{\dagger}\hat \varphi _{i2}-
\hat \varphi_{i1}^{\dagger}\hat d_i,
\qquad
\hat n_i = 1 - \hat h^+_i\hat h_i + \hat d^+_i\hat d_i,
\nonumber \\
& & \hat {c}_{i\uparrow}^{\dagger}\hat {c}_{i\uparrow}
\hat {c}_{i\downarrow}^{\dagger}\hat {c}_{i\downarrow} =
\hat d_i^{\dagger}\hat d_i,
\qquad
\hat{\vec{S}}_{i}=\frac 12 \sum\limits_{\sigma\sigma'} \hat\varphi^+_{i\sigma}
{\vec {\tau}}_{\sigma\sigma'} \hat\varphi_{i\sigma'},\qquad
\hat \varphi_{i\sigma}^{\dagger}\hat \varphi_{i\sigma}\,
+\,\hat d_i^{\dagger}\hat d_i\,+\,\hat h_i^{\dagger}\hat h_i\,=\,1
\label{mp2}
\end{eqnarray}
The partition function can be written as a path integral over the complex
functions of the Matsubara time $\tau$\,\, $\varphi_{i\sigma}(\tau)\,
\left(\bar\varphi_{i\sigma}(\tau)\right)$ and Grassmann functions
$h_i(\tau)\,\left(\bar h_i(\tau)\right)$ and $d_i(\tau)\,
\left(\bar d_i(\tau)\right)$.
\FL
\begin{equation}
{\cal Z}(\beta)\,=\,\int\,D\mu\left(\bar\varphi,\varphi,\bar h,h,\bar d,d,\right)
e^{-S}.
\label{mp4}
\end{equation}
The action is given by the expression
\FL
\begin{equation}
S\,=\,\int\limits^{\beta}_0 d\tau\left[\sum\limits_i\left(\bar\varphi_{i\sigma}
(\tau)
\dot\varphi_{i\sigma}(\tau)\,+\,\bar h_i(\tau)\dot h_i(\tau)\,+
\,\bar d_i(\tau)\dot d_i(\tau)\right)\,+\,
H\left(\bar\varphi,\varphi,\bar h,h,\bar d,d\right)\,\right],
\label{mp5}
\end{equation}
where $\beta$ is the inverse temperature and the Hamiltonian is obtained
from Eqs.(\ref{mp1}) and (\ref{mp2})
replacing the operators with the functions. In terms of Schwinger
bosons and slave-fermions the theory is $U(1)$ gauge invariant, and the measure
includes $\delta$ functions that enforce the constraint and the
gauge-fixing condition
\FL
\begin{eqnarray}
D\mu\left(\bar\varphi,\varphi,\bar h,h,\bar d,d\right) & = &
\prod\limits_{i,\tau,\sigma}\frac {D\bar\varphi_{i\sigma}(\tau)
D\varphi_{i\sigma}(\tau)}{2\pi i}\prod\limits_{i\tau}D\bar h_i(\tau)
D h_i(\tau)D\bar d_i(\tau)D d_i(\tau) \nonumber \\
& & \prod\limits_{i\tau}\delta\left(\bar\varphi_{i\sigma}(\tau)
\varphi_{i\sigma}(\tau)\,+\,\bar h_i(\tau) h_i(\tau)\,+\,
\bar d_i(\tau) d_i(\tau)\,-\,1\right)
\prod\limits_{i\tau}\delta\left(g.f\right).
\label{mp6}
\end{eqnarray}
I make a change of variables, introducing new Bose fields $f_{i\sigma}(\tau)
\left(\bar f_{i\sigma}(\tau)\right)$ Ref.\cite{mp13}
\FL
\begin{eqnarray}
f_{i\sigma}(\tau) & = & \varphi_{i\sigma}(\tau)
\left(1-\bar h_i(\tau)h_i(\tau)-\bar d_i(\tau)d_i(\tau)\right)^
{-\frac 12},\nonumber \\
\bar f_{i\sigma}(\tau) & = &\bar \varphi_{i\sigma}(\tau)
\left(1-\bar h_i(\tau)h_i(\tau)-\bar d_i(\tau)d_i(\tau)\right)^
{-\frac 12},
\label{mp7}
\end{eqnarray}
where the new fields satisfy the constraint
\FL
\begin{equation}
\bar f_{i\sigma}(\tau)f_{i\sigma}(\tau)\,=\,1.
\label{mp9}
\end{equation}
In terms of the new fields the spin vector and the
action have the form
\FL
\begin{equation}
\vec{S}_{i}(\tau)=\frac 12 \sum\limits_{\sigma\sigma'} f^+_{i\sigma}(\tau)
{\vec {\tau}}_{\sigma\sigma'} f_{i\sigma'}(\tau)
\left(1-\bar h_i(\tau)h_i(\tau)-\bar d_i(\tau)d_i(\tau)\right)
\label{mp10a}
\end{equation}
\FL
\begin{eqnarray}
S & = & \int\limits^{\beta}_0 d\tau \left\{\sum\limits_i
\left[\bar f_{i\sigma}(\tau)\dot f_{i\sigma}(\tau)\,+\,
\bar h_i(\tau)\left(\frac {\partial}{\partial\tau}-\bar f_{i\sigma}(\tau)
\dot f_{i\sigma}(\tau)
\right)h_i(\tau)
\right.\right. \nonumber \\
& & \left.\left.+\,\bar d_i(\tau)\left(\frac {\partial}{\partial\tau}-
\bar f_{i\sigma}(\tau)\dot f_{i\sigma}(\tau)
\right)d_i(\tau)\right]\,+\,H\left(\bar f,f,\bar h,h,\bar d,d\right)\right\},
\label{mp10}
\end{eqnarray}
where $H\left(\bar f,f,\bar h,h,\bar d,d\right)$ is the Hamiltonian
\FL
\begin{eqnarray}
H & = & -t\sum\limits_{<i,j>}\left[
\bar d_i d_j \bar f_{j\sigma} f_{i\sigma}\,+
\,\bar d_j d_i \bar f_{i\sigma} f_{j\sigma}\,-
\,\bar h_i h_j \bar f_{j\sigma} f_{i\sigma}\,-
\,\bar h_j h_i \bar f_{i\sigma} f_{j\sigma}\right. \nonumber \\
& &\left.+\,\left(h_id_j\,-\,h_jd_i\right)\left(\bar f_{i1}\bar f_{j2}\,-\,
\bar f_{i2}\bar f_{j1}\right)
\,+\,\left(\bar d_j\bar h_i\,-\,\bar d_i\bar h_j\right)\left(f_{i1} f_{j2}\,-\,
f_{i2} f_{j1}\right)\right] \nonumber \\
& & \times \left(1\,-\,\bar h_i h_i\,-\,\bar d_i d_i \right)^{\frac 12}
\left(1-\bar h_j h_j\,-\,\bar d_j d_j\right)^{\frac 12} \nonumber \\
& & +\,\frac J2 \sum\limits_{<i,j>}\left[\bar h_i h_i\,+\,\bar d_i d_i\,+\,
\bar h_j h_j\,+\,\bar d_j d_j\right]\,
-\,\frac J2 \sum\limits_{<i,j>}\left(\bar h_i h_i\,+\bar d_i d_i\right)
\left(\bar h_j h_j\,+\bar d_j d_j\right) \nonumber \\
& & +\frac J8\sum\limits_{<i,j>}
\left(\vec n_j\,-\,\vec n_i\right)^2
\left(1-\bar h_i h_i\,-\,\bar d_i d_i \right)^{\frac 12}
\left(1-\bar h_j h_j\,-\,\bar d_j d_j \right)^{\frac 12} \nonumber \\
& & +U\sum\limits_{i}\bar d_i d_i\,-\,\mu\sum\limits_{i}\left(1\,-
\,\bar h_i h_i\,+\,\bar d_i d_i\right)
\label{mp11}
\end{eqnarray}
In Eq.(\ref{mp11})
$\vec n_i\,=\,\sum\limits_{\sigma\sigma'}\bar
f_{i\sigma}\vec\tau_{\sigma,\sigma'}f_{i\sigma'}$ is an unit vector.
The Eq.({\ref{mp10a}}) describes in an $O(3)$ covariant way the spin.
When the lattice site is empty or doubly occupied the spin
vector is zero. When the lattice site is occupied by one electron
the unit vector $\vec n_i$ identifies the local orientation.
One can consider
the first two components $n_{i1}$ and $n_{i2}$ as
independent, and then $n_{i3}\,=\,\sqrt{1\,-\,n_{i1}^2\,-\,n_{i2}^2}$.
In the leading order of the fields, the spin vector has the form
\FL
\begin{equation}
S_{i1}\simeq \frac 12 n_{i1},
\qquad
S_{i2}\simeq \frac 12 n_{i2},
\qquad
S_{i3}\,-\,\frac 12 \simeq -\frac 12 \left(\bar h_i h_i\,+\,\bar d_i
d_i\right).
\label{mp11a}
\end{equation}
The last equation shows that the longitudinal spin fluctuations are associated with
the collective fields $(\bar h_i h_i\,+\,\bar d_i d_i)$.
To avoid
misunderstandings, it is important to point out that the charge-waves are
associated with the collective field $(\bar d_i d_i\,-\,\bar h_i h_i)$ (see
the representation of the electron number operator (\ref{mp2})).
To formulate a mean-field theory I drop the terms of order equal or higher
then six in the Hamiltonian Eq.(\ref{mp11}).
It is convenient to replace the term
$\frac J2 \sum\limits_{<i,j>}\left(\bar h_i h_i\,+\bar d_i d_i\right)
\left(\bar h_j h_j\,+\bar d_j d_j\right)$ in Eq.(\ref{mp11})
with the local one
$\frac {3J}{2} \sum\limits_{i}\left(\bar h_i h_i\,+\bar d_i d_i\right)^2$.
The difference is of higher order of derivatives, and I'll drop it.
Then one can decouple this term, by means of the Hubbard-Stratanovich
transformation,
introducing a real, spin-singlet and gauge invariant field.
\FL
\begin{eqnarray}
& & e^{\frac {3J}{2}\int\limits^{\beta}_0 d\tau\sum\limits_i
\left(\bar h_i(\tau) h_i(\tau)\,+\,\bar d_i(\tau) d_i(\tau)\right)^2}\nonumber \\
& & =\int\prod\limits_{i\tau} DS_i(\tau)
e^{-\int\limits^{\beta}_0 d\tau\sum\limits_i
\left[\frac {3J}{2}S_i(\tau) S_i(\tau)\,-\,3J \left(\bar h_i(\tau) h_i(\tau)\,+
\bar d_i(\tau) d_i(\tau)\right)S_i(\tau)\right]}
\label{mp13}
\end{eqnarray}
Now, the action is quadratic
with respect to the fermions and one can integrate them out.
The resulting action depends on the spinons and the real field $S_i$. It has
a minimum at the point $S_i=s_0, f_{i\sigma}=f_{\sigma}$, and the stationary
condition is
\FL
\begin{equation}
s_0\,=\,<\bar h_i h_i\,+\,\bar d_i d_i>
\label{mp14}
\end{equation}
Expanding the effective action around this point one obtains the effective
model.
To improve the calculations I account for the terms of order six and higher,
replacing the collective field $\bar h_i h_i\,+\,\bar d_i d_i$ in these
terms by its mean-field value from Eq.({\ref{mp14}). The new Hamiltonian
depends on the fields $\bar f_{i\sigma}(\tau)\,,\,f_{i\sigma}(\tau)\,,\,
2\varphi_i(\tau)=s_0-S_i(\tau)$, and is quadratic with respect to the
fermions
\FL
\begin{eqnarray}
H & = & -t\sum\limits_{<i,j>}\left[\bar d_i d_j\,+ \,\bar d_j d_i\,
- \,\bar h_i h_j \,- \,\bar h_j h_i\right]\,+\,
\left(6mJ+U\,-\mu\right)\sum\limits_{i}\bar d_i d_i\,+\,
\left(6mJ+\mu\right)\sum\limits_{i}\bar h_i h_i \nonumber \\
& & -2mt\sum\limits_{<i,j>}\left[
\left(\bar d_i d_j\,-\,\bar h_i h_j\right)\left(\bar f_{j\sigma}
f_{i\sigma}\,-\,1\right)\,+\,\
\left(\bar d_j d_i\,-\,\bar h_j h_i\right)\left(\bar f_{i\sigma}
f_{j\sigma}\,-\,1\right)\right. \nonumber \\
& & \left.+\,\left(h_id_j\,-\,h_jd_i\right)\left(\bar f_{i1}\bar f_{j2}\,-\,
\bar f_{i2}\bar f_{j1}\right)
\,+\,\left(\bar d_j\bar h_i\,-\,\bar d_i\bar h_j\right)\left(f_{i1} f_{j2}\,-\,
f_{i2} f_{j1}\right)\right] \nonumber \\
& & +6J\sum\limits_i\varphi_i \left(\bar h_i h_i\,+ \bar d_i d_i\right)
+\frac {(2m)^2J}8\sum\limits_{<i,j>}
\left(\vec n_j\,-\,\vec n_i\right)^2
\label{mp13a}
\end{eqnarray}
where $m=\frac 12(1-s_0)$.
Integrating out the fermions one obtains the action of the effective theory.
To get intuition how the effective action looks like, it is important to stress
that the spinon fields contribute the action through the fields
$\bar f_{i\sigma}(\tau)\dot f_{i\sigma}(\tau)$,\,\,
$\bar f_{i\sigma}(\tau)\left(
f_{j\sigma}(\tau)-f_{i\sigma}(\tau)\right)$,\,\,
$\left(f_{i1}(\tau)f_{j2}(\tau)\,-\,f_{j1}(\tau)f_{i2}(\tau)\right)$\,\,and
$\left(\vec n_j\,-\,\vec n_i\right)^2$ (see Eq.\ref{mp13a})
In the continuum limit they have the form
$\bar f_{\sigma}\partial_{\mu}f_{\sigma}$\,$(\mu=\tau,x,y,z)$,\,\,
$\left(f_1\partial_{\nu}f_2-f_2\partial_{\nu}f_1\right)$\,\,
$(\nu=x,y,z)$ and
$\partial_{\nu}\vec n\cdot\partial_{\nu} \vec n$.
The four-vector\, $A_{\mu}=i\bar f_{\sigma}\partial_{\mu}f_{\sigma}$\, transforms
as an\, $U(1)$\, gauge field,\,
$\left(f_1\partial_{\nu}f_2-f_2\partial_{\nu}f_1\right)$\,
as a charge-two complex field and
\,$\partial_{\nu}\vec n\cdot\partial_{\nu} \vec n$ is a gauge invariant.
Hence, the simplest gauge invariant and spin-
singlet contributions of the first two fields have the form
$\left(\bar f_1\partial_{\nu}\bar f_2-\bar f_2\partial_{\nu}\bar f_1
\right)\left(f_1\partial_{\nu}f_2-f_2\partial_{\nu}f_1\right)$ and
$\left(\partial_{{\mu}_1}A_{{\mu}_2}-\partial_{{\mu}_2}A_{{\mu}_1}\right)
\left(\partial_{{\mu}_1}A_{{\mu}_2}-\partial_{{\mu}_2}A_{{\mu}_1}\right)$.
It is not difficult to check that
\FL
\begin{equation}
\left(\bar f_1\partial_{\nu}\bar f_2-\bar f_2\partial_{\nu}\bar f_1
\right)\left(f_1\partial_{\nu}f_2-f_2\partial_{\nu}f_1\right)\,=\,
\frac 14\partial_{\nu}\vec n\cdot\partial_{\nu} \vec n.
\label{mp15}
\end{equation}
The term $\left(\partial_{{\mu}_1}A_{{\mu}_2}-\partial_{{\mu}_2}
A_{{\mu}_1}\right)
\left(\partial_{{\mu}_1}A_{{\mu}_2}-\partial_{{\mu}_2}A_{{\mu}_1}\right)$
is of the same order as $\left(\partial_{\nu}\vec
n\cdot\partial_{\nu} \vec n\right)^2$ and I'll ignore it. To
obtain the effective theory it is convenient to set the gauge
field $A_{\mu}=i\bar f_{\sigma}\partial_{\mu}f_{\sigma}$ equal to
zero, and to account for the contribution of the complex field
$\left(f_1\partial_{\nu}f_2-f_2\partial_{\nu}f_1\right)$, the real
field $\varphi$, and $\partial_{\nu}\vec n\cdot\partial_{\nu} \vec
n$. An important exclusion is the $\bar
f_{\sigma}\partial_{\tau}f_{\sigma}$ field which contributes the
action linearly.
Expanding the effective functional around the mean-field point and keeping
only the first three terms, one can write the effective
action in the form
\FL
\begin{equation}
S_{\text eff}\,=\,S_{\text H}\,+\,S_{\text p}\,+\,S_{\text int},
\label{mp16}
\end{equation}
$S_{\text H}$ is the action of the Heisenberg theory of localized spins.
In continuum limit it has the form
\FL
\begin{equation}
S_{H}\,=\,\int d\tau d^3\vec r\left[2m\bar f_{\sigma}(\tau,\vec r)\dot
f_{\sigma}(\tau,\vec r)\,+\,\frac {m^2J_{\text r}}{2}\sum\limits_{\nu=1}^3
\partial_{\nu}\vec n(\tau,\vec r)\cdot\partial_{\nu}\vec n(\tau,\vec r)
\right].
\label{mp17}
\end{equation}
In Eq.(\ref{mp17}),\, $m=\frac 12(1-s_0)$, and $s_0$ comes from
"tadpole" diagrams with one $h$ or $d$ line.
The renormalized exchange coupling constant has the following representation in
terms of
microscopic parameters and at zero temperature \cite{mp14}
\FL
\begin{eqnarray}
J_r & = & J\,-\,\frac {4t^2}{12Jm+U}\,
+ \,\frac 83 \frac {t^2}{12Jm+U}\frac 1N\sum\limits_k\left(\sum
\limits_{\nu=1}^{3} \sin^2 k_{\nu}\right)\left(n_k^h+n_k^d\right).
\label{mp18}
\end{eqnarray}
In Eq.(\ref{mp18}),\, $n_k^d$ and $n_k^h$ are the
occupation numbers for $d$ and $h$ fermions respectively. The first term
is due to the direct Heisenberg exchange term in Eq.(\ref{mp13a}),
and the other terms are due to Anderson's superexchange. The last two
terms are obtained calculating the one-loop self-energy diagrams
of $h$ and
$d$ fermions. The superexchange contribution to the exchange coupling
constant goes to zero for small magnetization.Hence, near the quantum phase
transition one can replace the renormalized coupling constant $J_r$ by bare
one $J$.
$S_p$ is the contribution to the effective
action of the paramagnon excitations
\FL
\begin{equation}
S_p\,=\,\frac 12 \int\frac {d\omega}{2\pi}\frac
{d^3p}{(2\pi)^3}\varphi(\omega,\vec p)
\left( r\,+\,a\frac
{|\omega|}{p}\,+\,b p^2\right)\varphi(-\omega,-\vec p)
\label{mp23}
\end{equation}
where
\FL
\begin{equation}
r\,=\,12J\left[1\,-\,3J\left(N(\epsilon_F^h)\,+\,N(\epsilon_F^d)
\right)\right]
\label{mp24}
\end{equation}
and the constants $a$ and $b$ are calculated in continuum limit.
\FL
\begin{equation}
a\,=\,18\pi J^2\left(\frac {N(\epsilon_F^h)}{v_F^h}\,+
\,\frac {N(\epsilon_F^d)}{v_F^d}\right)\quad
b\,=\,3 J^2\left (\frac {N(\epsilon_F^h)}{\left(k_F^h\right)^2}\,+
\,\frac {N(\epsilon_F^d)}{\left(k_F^d\right)^2}\right).
\label{mp25}
\end{equation}
It is obtained from the Lindhard functions for $h$ and $d$ fermions in
the limit when $p$ and $\frac {\omega}{p}$ are small.
Finally, the spinon-paramagnon interaction has the form
\FL
\begin{equation}
S_{\text int}\,=\,m^2\lambda\int d\tau d^3\vec r\, \varphi(\tau,\vec r)
\left[\sum\limits_{\nu =1}^{3}\partial_{\nu}\vec n(\tau,\vec r)\cdot
\partial_{\nu}\vec n(\tau,\vec r)\right]
\label{mp26}
\end{equation}
where
\FL
\begin{equation}
\lambda\,=\,
\frac {16Jt^2}{(12Jm+U)^2}\frac
1N\sum\limits_k\left(\sum\limits_{\nu=1}^{3}
\sin^2
k_{\nu}\right)\left(1\,-\,n_k^h-n_k^d\right).
\label{mp27}
\end{equation}
The effective magnon-paramagnon coupling is obtained from triangular
diagrams with two $h$ and one $d$
lines or with two $d$ and one $h$ lines.
To analyze the effective model, it is more convenient to rewrite it in terms of
rescaled spinon fields
\FL
\begin{equation}
\bar \zeta_{i\sigma}\,=\,\sqrt{2m}\bar f_{i\sigma},\qquad
\zeta_{i\sigma}\,=\,\sqrt{2m}f_{i\sigma}.
\label{mp90a}
\end{equation}
The new fields satisfy the constraint
\FL
\begin{equation}
\bar \zeta_{i\sigma}\zeta_{i\sigma}\,=\,2m ,
\label{mp90b}
\end{equation}
and the action of the effective theory has the form
\FL
\begin{eqnarray}
S_{\text eff} & = & \int d\tau d^3\vec r\left[\bar \zeta_{\sigma}(\tau,\vec
r)\dot \zeta_{\sigma}(\tau,\vec r)\,+\,\frac {J_{\text
r}}{2}\sum\limits_{\nu=1}^3 \partial_{\nu}\vec M(\tau,\vec
r)\cdot\partial_{\nu}\vec M(\tau,\vec r)\right. \nonumber \\
& & \left.+\,\frac {\lambda}{4}
\varphi(\tau,\vec r) \left[\sum\limits_{\nu
=1}^{3}\partial_{\nu}\vec M(\tau,\vec r)\cdot \partial_{\nu}\vec M(\tau,\vec
r)\right]\right]\,+\,S_p,
\label{mp90c}
\end{eqnarray}
where $\vec M$ is the spin vector
\FL
\begin{equation}
\vec M\,=\,\frac 12 \bar \zeta_{\sigma}
\vec\tau_{\sigma,\sigma'}\zeta_{\sigma'}, \qquad
\vec M^2\,=\,m^2
\label{mp90d}
\end{equation}
and $S_p$ is given by Eq.(\ref{mp23}).
It follows from Eq.(\ref{mp10a}) that the dimensionless magnetization
of the system, per
lattice site is defined by the equation,
\FL
\begin{equation}
<S_i^3>=\frac 12 <n_i^3>\left(1-<\bar h_i h_i+ \bar d_i d_i>\right).
\label{mp90e}
\end{equation}
At zero temperature $<n_i^3>=1$ and
using the Eq.(\ref{mp14}) one obtains that $m$ is zero temperature
dimensionless magnetization of the system per lattice site, $m=<S_i^3>$. The
parameter
$m$ depends on the microscopic parameters of the theory and
characterizes
the vacuum. If, in the vacuum state, every lattice site is
occupied by one
electron with spin up, then $m=\frac 12\quad (s_0=0)$, the
parameters $a$ and
$b$
from Eq.(\ref{mp23}) are equal to zero and $r=\frac
{3J}{2}$. In this case one
can integrate over the paramagnons and the
resulting theory is the spin $\frac 12$ Heisenberg
theory of the localized
spins. When, in the vacuum state, some of the sites are
doubly occupied
($<\bar d_i d_i>\neq 0$) or empty $(<\bar h_i h_i>\neq 0)$,
then $m<\frac
12$, the relevant excitations are the spinon and paramagnon
excitations and
the effective theory is a "spin $m$" Heisenberg
theory coupled to paramagnon fluctuations defined by
Eqs.(\ref{mp90b},\ref{mp90c},\ref{mp90d}). The system approaches the
quantum critical
point when $m\rightarrow 0\quad (s_0\rightarrow 1)$. One can
see directly, from
the stationary condition (\ref{mp14}), that $r(m)$
approaches zero when
$m\rightarrow 0$. Hence, the parameter $r$ measures the
distance from the quantum critical point. In quantum paramagnetic phase
$(m=0)$, the spinon excitations disappear from the spin spectrum (see
Eqs.(\ref{mp90b},\ref{mp90d})) and one obtains Hertz's effective model. One
can add a four-paramagnon term , calculating one-loop diagrams with four $h$
or $d$ fermion lines, but I have dropped it motivated by the Hertz's result.
The effective theory is $U(1)$ gauge invariant. Below the Curie temperature
it is convenient to introduce explicitly the magnon excitations. To this end,
I impose the gauge-fixing condition in the form
$arg \zeta_{i1}=0$. Then the
constraint (\ref{mp90b}) can be solved by means of
the complex field
$a_i(\tau)=\zeta_{i2}$ and $\zeta_{i1}=\sqrt {2m-\bar a_i(\tau)
a_i(\tau)}$.
For the components of the spin vector
$M^+\,=\,M_1\,+\,iM_2,\,\,M^-\,=\,M_1\,-\,iM_2,$ and $M_3$ one obtains the
Holstein-
Primakoff representation:
\FL
\begin{eqnarray}
M_i^+(\tau) & = & \sqrt {2m - \bar a_i(\tau)\,a_i(\tau)}\,\,a_i(\tau), \,\,
M_i^-(\tau)\,=\,\bar a_i(\tau)\sqrt {2m - \bar a_i(\tau)\,a_i(\tau)},
\nonumber \\
M_i^3(\tau) & = & m\,-\,\bar a_i(\tau)\,a_i(\tau)
\label{mp91}
\end{eqnarray}
The kinetic term in the
action
and the measure are the same as the kinetic term and the measure in
the theory
of Bose field. The only difference is that the complex fields are
subject to
the condition $\bar a_i(\tau) a_i(\tau)\leq 1$.
In the spin-wave theory one approximates $\sqrt {2m-\bar a_i(\tau)a_i(\tau)}$
and integrates over the whole complex plane. Then, the model is simplified and
can be written in terms of magnon $a_i(\tau)\,\,(\bar a_i(\tau))$ and
paramagnon $\varphi_i(\tau)$ fields
\FL
\begin{eqnarray}
S_{\text eff} & = & \int \frac {d\omega}{2\pi}\frac {d^3p}{(2\pi)^3}\left[
\bar a(\omega,\vec p)\left (i\omega\,+\,\rho p^2\right) a(\omega,\vec p)
\,+
\,\frac 12 \varphi(\omega,\vec p)\left(r\,+\,a\frac {|\omega|}{p}\,+\,b
p^2\right)
\varphi(-\omega,-\vec p)\right] \nonumber \\
& + & \frac {m\lambda}{2} \int\prod\limits_{l=1}^{2}\frac {d{\omega}_l}{2\pi}
\frac {d^3p_{l}}
{(2\pi)^3}\left(\vec p_1\cdot\vec p_2\right)\bar a({\omega}_1,\vec p_1)
a({\omega}_2,\vec p_2)\varphi({\omega}_1-{\omega}_2,\vec p_1-\vec p_2)
\label{mp30}
\end{eqnarray}
where
\FL
\begin{equation}
\rho\,=\,m\,J_r
\label{mp30n}
\end{equation}
is the spin stiffness constant.
Let's rewrite the spin vector Eq.(\ref{mp10a}) in terms of the vector $\vec M$
Eq.(\ref{mp90d}) and use the Holstein- Primakoff representation
Eq.(\ref{mp91}). Then the magnetization of the system per lattice site is given
by the expression
\FL
\begin{equation}
<S_i^3>\,=\,m\,-\,<\bar a_i a_i>
\label{mp92}
\end{equation}
The equation for the critical temperature is $<S_i^3>\,=\,0$. Having in
mind that the magnon dispersion is $\epsilon_a(p)\,=\,\rho p^2$ one
obtains for the Curie temperature
\FL
\begin{equation}
T_c\,=\,\kappa\,m^{\frac 23}\,\rho(m)
\label{mp94}
\end{equation}
where the constant $\kappa$ can be written in terms of gamma $\Gamma(z)$ and
Riemann $\zeta(z,q)$ functions $\kappa\,=\,(\Gamma(\frac 32)\zeta(\frac
32,1)/4\pi^2)^{-\frac 23}$. In the spin-wave approximation the spin stiffness
constant is given by Eq.(\ref{mp30n}), hence, when the system approaches
quantum critical point ($m\rightarrow 0$), the critical temperature scales
with magnetization like $T_c\sim m^{\frac 53}$. Recently, it was
experimentally shown \cite{mp16}, that for weak ferromagnets the transition
temperature scales with pressure like $T_c\sim (p_c\,-\,p)^{\frac 12}$. To
explain the experimental result one has to assume that magnetization scales
with pressure like $m\sim (p_c\,-\,p)^{0,3}$.
One can improve the equation for the Curie temperature replacing the zero
temperature dimensionless magnetization $m$ by the finite temperature solution
$m(T)$ of the mean-field equation Eq.(\ref{mp14}).
\FL
\begin{equation}
T_c\,=\,\kappa\,m^{\frac 23}(T_c)\,\rho(m(T_c))
\label{mp94a}
\end{equation}
For conventional weak ferromagnets $m(T_c)\sim m$ and Eq.(\ref{mp94}) is an
appropriate expression for Curie temperature. But for high $T_c$ weak
ferromagnets $m(T_c)>m$ and the correct equation for the critical temperature
is Eq.(\ref{mp94a}).
It is important to stress that the equations for the critical temperature
follow from the exact representations for the third component of the spin (\ref
{mp10a},\ref {mp91},\ref {mp92}).Hence they take the magnon fluctuations into
account exactly.
In the spin-wave approximation the transverse components of the spin
fields are
proportional to the magnon fields
\FL
\begin{equation}
S_i^+(\tau)=\sqrt {2m} a_i(\tau),\,\,\,\,S_i^-(\tau)=\sqrt {2m} \bar
a_i(\tau)
\label{mp30a}
\end{equation}
and the field $\varphi_i(\tau)$ is exactly the paramagnon (longitudinal
spin fluctuation)
\FL
\begin{equation}
S_i^3(\tau)-<S_i^3>\,=\,\varphi_i(\tau).
\label{mp30b}
\end{equation}
Hence, in Gaussian approximation, spin-spin
correlation functions have the form
\FL
\begin{equation}
D^{\text tr}(\omega,\vec p)\,=\,\frac {2m}{i\omega\,+\,\rho p^2}\,\,\,,
\qquad\qquad
D^{\text long}(\omega,\vec p)\,=\,\frac {1}{r\,+
a\frac {|\omega|}{p}\,
+\,bp^2}
\label{mp31}
\end{equation}
%By definition the magnetic susceptibility is
%\FL
%\begin{equation}
%\chi\,=\,D^{\text long}(0,0)\,=\,\frac {1}{2r}
%\label{mp94}
%\end{equation}
\ \\
{\bf III Scaling behavior}
Scaling arguments, based on the Hertz's
renormalization group technique
allow to obtain a relation
between the spin stiffness and the parameter $r$-the distance
from the quantum critical point.
To begin with, I introduce a cutoff and redefine the momenta
introducing dimensionless one. RG construction starts with a definition of
"soft" and "fast" modes,
\FL
\begin{eqnarray}
\Psi(\omega,\vec p) & = & \Psi_{<}(\omega, \vec p)\,+\,\Psi_{>}(\omega,
\vec p) \\
\Psi_{<}(\omega, \vec p) & = & \Psi(\omega, \vec p) \qquad {\text for}
\qquad |p|<e^{-l} \nonumber \\
\Psi_{>}(\omega, \vec p) & = & \Psi(\omega, \vec p) \qquad {\text for}
\qquad e^{-l}\leq |p| \leq 1 .\nonumber
\label{mp95}
\end{eqnarray}
where $\Psi(\omega,\vec p)$ stands for magnons, $a(\omega,\vec
p),\bar a(\omega,\vec p)$ and paramagnon $\varphi(\omega,\vec p)$ fields.
The action Eq.(\ref{mp30}) depends on the "slow" and the "fast" modes. In the
tree approximation one keeps only the "slow" modes
\FL
\begin{equation}
S^{\text tree}(\bar a,a,\varphi)\,=\,S(\bar a_{<},a_{<},\varphi_{<}).
\label{mp96}
\end{equation}
The next step is to change the variables, letting
\FL
\begin{equation}
\vec p'\,=\,e^l\vec p, \qquad \omega'\,=\,e^{zl}\omega
\label{mp97}
\end{equation}
The new momenta runs over the whole interval $0\leq |p'|\leq 1$, and in terms
of the new variables the action has the form
\FL
\begin{eqnarray}
S^{\text tree} & = & e^{-(z+3)l}\int \frac {d\omega'}{2\pi}\frac
{d^3p'}{(2\pi)^3}\left[\bar a_{<}(e^{-zl}\omega',e^{-l}\vec p')\left
(ie^{-zl}\omega'\,+\,mJ_r e^{-2l}p'^2\right) a_{<}(e^{-zl}\omega',e^{-l}\vec
p')\right. \nonumber \\
& + &\left. \frac 12 \varphi_{<}(e^{-zl}\omega',e^{-l}\vec
p')\left(r\,+\,e^{(-z+1)l}a\frac {|\omega'|}{p'}\,+\,e^{-2l}b
p'^2\right)\varphi_{<}(-e^{-zl}\omega',-e^{-l}\vec p')\right] \nonumber \\
& + & \Lambda e^{-2(z+4)}\int\prod\limits_{l=1}^{2}\frac {d{\omega'}_l}{2\pi}
\frac {d^3p'_{l}} {(2\pi)^3}\left(\vec p'_1\cdot\vec p'_2\right)\bar
a_{<}({\omega'}_1,\vec p'_1)a_{<}({\omega'}_2,\vec
p'_2)\varphi_{<}({\omega'}_1-{\omega'}_2,\vec p'_1-\vec p'_2),
\label{mp98}
\end{eqnarray}
where I have used a short notation for the magnon-paramagnon coupling constant
$\Lambda\,=\,\frac {m\lambda}{2}$.
It is apparent that if one chooses $z\,=\,3$, the coefficient of $a\frac
{|\omega'|}{p'}$, in the paramagnon quadratic term, is the same as the
coefficient of $bp'^2$. Then rescaling the paramagnon $\varphi$, one can make
the total coefficient of both of them unity. \FL
\begin{equation}
\varphi'(\omega',\vec p')\,=\,e^{-\frac
{z+5}{2}l}\varphi_{<}(e^{-zl}\omega',e^{-l}\vec p')
\label{mp99}
\end{equation}
To complete
the RG transformation, one has to rescale the magnon
fields too, to make the coefficient of $i\omega'$ unity,
\FL
\begin{equation}
a_{<}(\omega',\vec p')\,=\,e^{-\frac
{2z+3}{2}} a_{<}(e^{-zl}\omega',e^{-l}\vec p'), \qquad
\bar a_{<}(\omega',\vec p')\,=
\,e^{-\frac {2z+3}{2}}\bar a_{<}(e^{-zl}\omega',e^{-l}\vec p')
\label{mp100}
\end{equation}
Then, for the transformed action one obtains
\FL
\begin{eqnarray}
& S' & (\bar a',a',\varphi')\,=\,\int \frac {d\omega'}{2\pi}\frac
{d^3p'}{(2\pi)^3}\left[ \bar a'(\omega',\vec p')\left (i\omega'\,+\,\rho'
p'^2\right) a'(\omega',\vec p')\right. \nonumber \\
& + &\left. \frac 12 \varphi'(\omega',\vec p')\left(r'\,+\,a\frac
{|\omega'|}{p'}\,+\,b p'^2\right) \varphi'(-\omega',-\vec p',)\right]
\nonumber \\ & + & \Lambda' \int\prod\limits_{l=1}^{2}\frac
{d{\omega'}_l}{2\pi} \frac {d^3p'_{l}}
{(2\pi)^3}\left(\vec p'_1\cdot\vec p'_2\right)\bar a'({\omega'}_1,\vec p'_1)
a'({\omega'}_2,\vec p'_2)\varphi'({\omega'}_1-{\omega'}_2,\vec p'_1-\vec p'_2)
\label{mp101}
\end{eqnarray}
where $r',\rho'$ and $\Lambda'$ are the transformed parameters
\FL
\begin{eqnarray}
r' & = & r(l)\,=\,re^{2l} \nonumber \\
\rho' & = & \rho(l)\,=\,\rho e^{(z-2)l}\,=\,\rho e^{l} \\
\Lambda' & = & \Lambda(l)\,=\,\Lambda e^{\frac {z-5}{2}l}\,=\,\Lambda e^{-l}
\nonumber
\label{mp102}
\end{eqnarray}
Excluding the scaling parameter $l$ one obtains the relations (RG invariants)
\FL
\begin{equation}
\frac {r'}{\rho'^2}\,=\,\frac {r}{\rho^2}, \qquad
r'\Lambda'^2\,=\,r\Lambda^2
\label{mp103}
\end{equation}
The bare parameter $r$ scales with magnetization like $r\,=\,r_0m^2$, and the
bare spin stiffness constant is proportional to the magnetization (see
Eq.\ref{mp30}). Hence, for weak ferromagnets, one obtains
\FL
\begin{equation}
r(m)\sim \rho^2(m)
\label{mp104}
\end{equation}
%Using the spin-wave expression for the spin stiffness and the assumed scaling
%of the magnetization with the pressure one obtains the scaling low of the
%magnetic susceptibility with the pressure
%\FL
%\begin{equation}
%\chi^{-1}\sim \left(p_c\,-\,p\right)^{\frac 35}
%\label{mp105}
%\end{equation}
The relation Eq.(\ref{mp104}) is a consequence of the fact that quantum
critical exponent is equal to three ($z=3$). It holds even if the
magnon-magnon interaction is introduced. Calculating the magnon contribution
to the spin stiffness, one can use it to obtain the the magnon scattering
corrections to the parameter $r$ (Stoner's enhancement factor)
Making use of Eq.(\ref{mp104}) one can obtain that
\FL
\begin{equation}
T_c\sim m^{\frac 23}\,T_s
\label{mp106}
\end{equation}
where, $T_s\,=\,const\sqrt r\sim m$ is the Stoner's expression for the
critical temperature (see Eq.\ref{mp24}). For weak ferromagnets $m<\frac 12$,
and $T_c$ is well below the Stoner's critical temperature, as should
be.Curie temperature obtained within Moriya's theory scales wit magnetization
like $\sim m^{\frac32}$ \cite {mp19}.
The result indicates that the spin-$m$
Heisenberg model coupled to paramagnon provides a good description of the
itinerant ferromagnets.
\ \\
{\bf IV Conclusions}
An effective magnon-paramagnon theory of
itinerant ferromagnets has been obtained. It is written in a way
which keeps $O(3)$ symmetry manifest.
The effective model differs from the models discussed in
Refs.\cite{mp7,mp9,mp17} in two ways. First, it describes in unified way both
the order
and disordered phases of the system. Altering the parameters, it
interpolates between the Heisenberg theory of localized spins and
Hertz's
theory of nearly ferromagnetic metals. In ferromagnetic phase the important
spin fluctuations are transversal magnon fluctuations and longitudinal
paramagnon fluctuations. In thermal paramagnetic phase (above Curie
temperature) the spectrum consists of spin singlet fluctuations of paramagnon
type and spin-$\frac 12$ spinon fluctuations. Well above the
critical temperature the spinon has a large gap, and the physics of
ferromagnetic metals is dominated by the paramagnon fluctuations. But just
above $T_c$ the spinon's gap approaches zero \cite{mp20}, and the
contribution of the spin-$\frac 12$ fluctuations is essential. Crossing the
quantum critical point ($m=0$), the spinon excitations disappear from the
spectrum and only the paramagnon survives in the quantum paramagnetic phase.
Second, I have
accounted for the on site Coulomb repulsion exactly. The Hubbard interaction
is the strongest, and to drop it, as in Ref.\cite{mp7}, or to treat it
perturbatively, as in Refs. \cite{mp9,mp17} is not adequate. As a result, I
obtain that the vertices in the effective functional exist in the limit of zero
frequencies and wavenumber, and that the constants are well defined for all
values of the magnetization.
The effective model Eqs.(\ref{mp90c},\ref{mp30}) enables to estimate
the correctness of the Green's functions calculations \cite{mp4}, and
the Moriya Kawabata approximation \cite{mp10}.Within these approaches,
the spin stiffness constant is proportional to the magnetization. The same
result follows from the effective model Eq.(\ref{mp30}) where the nonlinear
magnon-magnon interaction is not considered. From spin-wave expansion follows
that the magnon vertices are of order $(\frac 1m)^n$, and on the verge of
ferromagnetism are relevant. One way to obtain the effects of magnon-magnon
interaction is to use a renormalization group approach in the spin-wave
theory, which allows for the analysis of systems with small spins \cite{mp21}.
\ \\
\acknowledgments
The author would like to thank K.Bedell and K.Blagoev for valuable
discussions. This research is supported by the USA National
Science Foundation Grant-DINT\#9876873.
\begin{references}
%
\frenchspacing
%
\bibitem[*]{byline} Electronic address: naoum@phys.uni-sofia.bg
%
\bibitem{mp1} A. A. Abrikosov and I. E. Dzyloshinskii, Zh. Eksp. Teor. Fiz.
{\bf 35}, 771 (1959) [Sov. Phys. JETP. {\bf8}, 535 (1959)].
%
\bibitem{mp2} T. V. Ramakrishnan, Solid State Commune. {\bf 14}, 449 (1974);
Phys. Rev. B {\bf 10}, 4014 (1974).
%
\bibitem{mp3} A. Kawabata, J. Phys. F {\bf 4}, 1477 (1974).
%
\bibitem{mp4} I. E. Dzyaloshinskii and P. S. Kondratenko, Zh. Eksp. Teor.
Fiz.
{\bf 70}, 1987 (1974) [Sov. Phys. JETP, {\bf 43}, 1036 (1975)].
%
\bibitem{mp6} K. Murata and S. Doniach, Phys. Rev. Lett. {\bf 29}, 285 (1972)
%
\bibitem{mp10} T. Moriya and A. Kawabata, J. Phys. Soc. Japan {\bf34},639
%(1973); {\bf 35}, 669 (1973).
%
\bibitem{mp10a} The works are summarized in T. Moriya, {\it Spin
Fluctuations in Itinerant Electron Magnetism.} (Springer-Verlag, Berlin
1985).
%
\bibitem{mp10b} G. Lonzarich and L. Taillefer, J. Phys. C {\bf 18}, 4339
(1985).
%
\bibitem{mp11} W. Metzner and D. Vollhardt, Phys. Rev. Lett. {\bf
62}, 324
(1989).
%
\bibitem{mp12} N. Bl\"umer, J. Wahle, J. Schlipf, K. Held, and D. Vollhardt,
in preparation.
%
\bibitem{mp7} J. Hertz, Phys. Rev. B {\bf 14}, 1165 (1976).
%
\bibitem{mp8} A. J. Millis, Phys.Rev. B {\bf 48}, 7183 (1993).
%
\bibitem{mp9} T. Vojta, D. Belitz, R. Narayanan, and T. R. Kirkpatrick,
Europhys.Lett. {\bf 36}, 191 (1996)
%
\bibitem{mp17} T. Vojta, D. Belitz, R. Narayanan, and T. R. Kirkpatrick,
Z. Phys. B {\bf 103}, 451 (1997).
%
%\bibitem{mp15} D. Belitz, T. R. Kirkpatrick, A. J. Millis, and T. Vojta,
%Phys. Rev. B {\bf 58}, 14155 (1998).
%
\bibitem{mp16} C. Pfleiderer, G. J. McMullan, S. R. Julian, and
G. G. Lonzarich, Phys. Rev. B {\bf 55}, 8330 (1997).
%
\bibitem{mp19} G. Gumbs and A. Griffin Phys. Rev. B {\bf 13}, 5054 (1976)
%
\bibitem{mp18} D. Vollhard, N. Bl\"umer, K. Held, M. Kollar, J.
Schlipf, and M. Ulmke, Z. Phys. B {\bf 103}, 283 (1997).
%
\bibitem{mp13} D. Schmeltzer, Phys. Rev. B {\bf 43}, 8650 (1991).
%
\bibitem{mp14} The present discussion is correct if
the renormalized exchange coupling constant is
positive. To this end, it is
sufficient the microscopic parameters
to satisfy the inequality $\frac
{4t^2}{U} < J$.
%
\bibitem{mp20} D. P. Arovas, and A. Auerbach, Phys. Rev. B {\bf 38}, 1316
(1988)
%
\bibitem{mp21} N.Karchev, Phys. Rev B {\bf 55}, 6372 (1997)
\end{references}
\end{document}