mirror of
https://github.com/overleaf/overleaf.git
synced 2024-12-04 10:26:49 -05:00
405 lines
20 KiB
Text
405 lines
20 KiB
Text
|
% Choose pra, prb, prc, prd, pre, prl, prstab, or rmp for journal
|
||
|
% Add 'draft' option to mark overfull boxes with black boxes
|
||
|
% Add 'showpacs' option to make PACS codes appear
|
||
|
% for review and submission
|
||
|
%\documentclass[aps,preprint,showpacs,superscriptaddress,groupedaddress]{revtex4} % for double-spaced preprint
|
||
|
% needed for figures
|
||
|
% needed for some tables
|
||
|
% for math
|
||
|
% for math
|
||
|
% for crossing out text
|
||
|
% for coloring text
|
||
|
%\input{tcilatex}
|
||
|
|
||
|
|
||
|
\documentclass[aps,prl,twocolumn,showpacs,superscriptaddress,groupedaddress]{revtex4}
|
||
|
|
||
|
\usepackage{graphicx}
|
||
|
\usepackage{dcolumn}
|
||
|
\usepackage{bm}
|
||
|
\usepackage{amssymb}
|
||
|
\usepackage{soul}
|
||
|
\usepackage{color}
|
||
|
|
||
|
%TCIDATA{OutputFilter=LATEX.DLL}
|
||
|
%TCIDATA{Version=5.50.0.2960}
|
||
|
%TCIDATA{<META NAME="SaveForMode" CONTENT="1">}
|
||
|
%TCIDATA{BibliographyScheme=BibTeX}
|
||
|
%TCIDATA{LastRevised=Tuesday, May 20, 2014 03:06:00}
|
||
|
%TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
|
||
|
|
||
|
\hyphenation{ALPGEN}
|
||
|
\hyphenation{EVTGEN}
|
||
|
\hyphenation{PYTHIA}
|
||
|
\def\be{\begin{equation}}
|
||
|
\def\ee{\end{equation}}
|
||
|
\def\bea{\begin{eqnarray}}
|
||
|
\def\eea{\end{eqnarray}}
|
||
|
%\input{tcilatex}
|
||
|
|
||
|
\begin{document}
|
||
|
|
||
|
\title{Transport measurements of the spin wave gap of Mn}
|
||
|
\input author_list.tex
|
||
|
\date{\today}
|
||
|
|
||
|
\begin{abstract}
|
||
|
Temperature dependent transport measurements on ultrathin antiferromagnetic
|
||
|
Mn films reveal a heretofore unknown non-universal weak localization
|
||
|
correction to the conductivity which extends to disorder strengths greater than
|
||
|
100~k$\Omega$ per square. The inelastic scattering of electrons off of
|
||
|
gapped antiferromagnetic spin waves gives rise to an inelastic scattering
|
||
|
length which is short enough to place the system in the 3D regime. The
|
||
|
extracted fitting parameters provide estimates of the energy gap ($\Delta
|
||
|
\approx$~16~K) and exchange energy ($\bar{J} \approx$~320~K). %\st{which are in
|
||
|
%agreement with values obtained with other techniques}.
|
||
|
\end{abstract}
|
||
|
|
||
|
\pacs{75}
|
||
|
|
||
|
\maketitle
|
||
|
|
||
|
Hello world
|
||
|
|
||
|
|
||
|
|
||
|
Thin-film transition metal ferromagnets (Fe, Co, Ni, Gd) and
|
||
|
antiferromagnets (Mn, Cr) and their alloys are not only ubiquitous in
|
||
|
present day technologies but are also expected to play an important role in
|
||
|
future developments~\cite{thompson_2008}. Understanding magnetism in these
|
||
|
materials, especially when the films are thin enough so that disorder plays
|
||
|
an important role, is complicated by the long standing controversy about the
|
||
|
relative importance of itinerant and local moments~\cite%
|
||
|
{slater_1936,van_vleck_1953,aharoni_2000}. For the itinerant transition
|
||
|
metal magnets, a related fundamental issue centers on the question of how
|
||
|
itinerancy is compromised by disorder. Clearly with sufficient disorder the
|
||
|
charge carriers become localized, but questions arise as to what happens to
|
||
|
the spins and associated spin waves and whether the outcome depends on the
|
||
|
ferro/antiferro alignment of spins in the itinerant parent. Ferromagnets
|
||
|
which have magnetization as the order parameter are fundamentally different
|
||
|
than antiferromagnets which have staggered magnetization (i.e., difference
|
||
|
between the magnetization on each sublattice) as the order parameter~\cite%
|
||
|
{blundell_2001}. Ferromagnetism thus distinguishes itself by having soft
|
||
|
modes at zero wave number whereas antiferromagnets have soft modes at finite
|
||
|
wave number~\cite{belitz_2005}. Accordingly, the respective spin wave
|
||
|
spectrums are radically different. These distinctions are particularly
|
||
|
important when comparing quantum corrections to the conductivity near
|
||
|
quantum critical points for ferromagnets~\cite{paul_2005} and
|
||
|
antiferromagnets~\cite{syzranov_2012}.
|
||
|
|
||
|
Surprisingly, although there have been systematic studies of the effect of
|
||
|
disorder on the longitudinal $\sigma_{xx}$ and transverse $\sigma_{xy}$
|
||
|
conductivity of ferromagnetic films~\cite%
|
||
|
{bergmann_1978,bergmann_1991,mitra_2007,misra_2009,kurzweil_2009}, there
|
||
|
have been few if any such studies on antiferromagnetic films. In this paper
|
||
|
we remedy this situation by presenting transport data on systematically
|
||
|
disordered Mn films that are sputter deposited in a custom designed vacuum
|
||
|
chamber and then transferred without exposure to air into an adjacent
|
||
|
cryostat for transport studies to low temperature. The experimental
|
||
|
procedures are similar to those reported previously: disorder, characterized
|
||
|
by the sheet resistance $R_0$ measured at $T=$~5~K, can be changed either by
|
||
|
growing separate samples or by gentle annealing of a given sample through
|
||
|
incremental stages of disorder~\cite{misra_2011}. Using these same procedures our results for
|
||
|
antiferromagnets however are decidedly different. The data are well
|
||
|
described over a large range of disorder strengths by a non-universal three
|
||
|
dimensional (3d) quantum correction that applies only to spin wave gapped
|
||
|
antiferromagnets. This finding implies the presence of strong inelastic
|
||
|
electron scattering off of antiferromagnetic spin waves. The theory is
|
||
|
validated not only by good fits to the data but also by extraction from the
|
||
|
fitting parameters of a value for the spin wave gap $\Delta$ that is in
|
||
|
agreement with the value expected for Mn. On the other hand, the
|
||
|
exchange energy $\bar{J}$ could be sensitive to the high disorder in our
|
||
|
ultra thin films, and it turns out to be much smaller compared to the known values.
|
||
|
|
||
|
In previous work the inelastic scattering of electrons off of spin waves has
|
||
|
been an essential ingredient in understanding disordered ferromagnets. For
|
||
|
example, to explain the occurrence of weak-localization corrections to the
|
||
|
anomalous Hall effect in polycrystalline Fe films~\cite{mitra_2007}, it was
|
||
|
necessary to invoke a contribution to the inelastic phase breaking rate $%
|
||
|
\tau_{\varphi}^{-1}$ due to spin-conserving inelastic scattering off
|
||
|
spin-wave excitations. This phase breaking rate, anticipated by theory~\cite%
|
||
|
{tatara_2004} and seen experimentally in spin polarized electron energy loss
|
||
|
spectroscopy (SPEELS) measurements of ultrathin Fe films~\cite%
|
||
|
{plihal_1999,zhang_2010}, is linear in temperature and significantly larger
|
||
|
than the phase breaking rate due to electron-electron interactions, thus
|
||
|
allowing a wide temperature range to observe weak localization corrections~%
|
||
|
\cite{mitra_2007}. The effect of a high $\tau_{\varphi}^{-1}$ due to
|
||
|
inelastic scattering off spin-wave excitations is also seen in Gd films
|
||
|
where in addition to a localizing log($T$) quantum correction to the
|
||
|
conductance, a localizing linear-in-$T$ quantum correction is present and is
|
||
|
interpreted as a spin-wave mediated Altshuler-Aronov type correction to the
|
||
|
conductivity~\cite{misra_2009}.
|
||
|
|
||
|
Interestingly, this high rate of inelastic spin rate scattering becomes even
|
||
|
more important for the thinnest films as shown in theoretical calculations
|
||
|
on Fe and Ni which point to extremely short spin-dependent inelastic mean
|
||
|
free paths~\cite{hong_2000} and in spin-polarized electron energy-loss
|
||
|
spectroscopy (SPEELS) measurements on few monolayer-thick Fe/W(110) films in
|
||
|
which a strong nonmonotonic enhancement of localized spin wave energies is
|
||
|
found on the thinnest films~\cite{zhang_2010}.
|
||
|
|
||
|
Inelastic spin wave scattering in highly disordered ferromagnetic films can
|
||
|
be strong enough to assure that the associated $T$-dependent dephasing
|
||
|
length $L_{\varphi }(T)=\sqrt{D\tau _{\varphi }}$ (with $D$ the diffusion
|
||
|
constant)~\cite{lee_1985} is less than the film thickness $t$, thus putting
|
||
|
thin films into the 3d limit where a metal-insulator
|
||
|
transition is observed~\cite{misra_2011}. Recognizing that similarly high
|
||
|
inelastic scattering rates must apply to highly disordered antiferromagnetic
|
||
|
films, we first proceed with a theoretical approach that takes into account
|
||
|
the scattering of antiferromagnetic spin waves on the phase relaxation rate
|
||
|
and find a heretofore unrecognized non-universal 3d weak localization
|
||
|
correction to the conductivity that allows an interpretation of our experimental
|
||
|
results.
|
||
|
|
||
|
We mention in passing that the 3d interaction-induced quantum correction
|
||
|
found to be dominant in the case of ferromagnetic Gd
|
||
|
films which undergo a metal-insulator transition\cite{misra_2011} is
|
||
|
found to be much smaller in the present case and will not be considered further (for an estimate of this contribution see \cite{muttalib_unpub}.
|
||
|
|
||
|
As discussed in detail in Ref.~[\onlinecite{wm10}], the phase relaxation
|
||
|
time $\tau _{\varphi }$ limits the phase coherence in a particle-particle
|
||
|
diffusion propagator $C(q,\omega )$ (Cooperon) in the form
|
||
|
\begin{equation}
|
||
|
C(q,\omega _{l})=\frac{1}{2\pi N_{0}\tau ^{2}}\frac{1}{Dq^{2}+|\omega
|
||
|
_{l}|+1/\tau _{\varphi }}.
|
||
|
\end{equation}
|
||
|
where $N_{0}$ is the density of states at the Fermi level, $\tau $ is the
|
||
|
elastic scattering time and $\omega _{l}=2\pi lT$ is the Matsubara
|
||
|
frequency. Labeling the Cooperon propagator in the absence of interactions
|
||
|
as $C_{0}$, we can write
|
||
|
\begin{equation}
|
||
|
\frac{1}{\tau _{\varphi }}=\frac{1}{2\pi N_{0}\tau ^{2}}[C^{-1}-C_{0}^{-1}].
|
||
|
\end{equation}
|
||
|
|
||
|
In general, $C(q,\omega )$ can be evaluated diagrammatically in the presence
|
||
|
of interactions and disorder in a ladder approximation \cite{fa} that can be
|
||
|
symbolically written as $C=C_{0}+C_{0}KC$ where the interaction vertex $K$
|
||
|
contains self energy as well as vertex corrections due to both interactions
|
||
|
and disorder. It then follows that $1/\tau _{\varphi }$ is given by
|
||
|
\begin{equation}
|
||
|
\frac{1}{\tau _{\varphi }}=-\frac{1}{2\pi N_{0}\tau ^{2}}K.
|
||
|
\end{equation}%
|
||
|
In Ref.~[\onlinecite{wm10}], the leading temperature and disorder dependence
|
||
|
of the inelastic diffusion propagator was evaluated diagrammatically, in the
|
||
|
presence of ferromagnetic spin-wave mediated electron-electron interactions.
|
||
|
Here we consider the antiferromagnetic case. We only consider large
|
||
|
spin-wave gap where the damping can be ignored. Using the antiferromagnetic
|
||
|
dispersion relation $\omega _{q}=\Delta +Aq$, where $A$ is the spin
|
||
|
stiffness, the inelastic lifetime is given by
|
||
|
\be
|
||
|
\frac{\hbar }{\tau _{\varphi }}=\frac{4}{\pi \hbar }nJ^{2}\int_{0}^{1/l}%
|
||
|
\frac{q^{d-1}dq}{\sinh \beta \omega _{q}}\frac{Dq^{2}+1/\tau _{\varphi }}{%
|
||
|
(Dq^{2}+1/\tau _{\varphi })^{2}+\omega _{q}^{2}}
|
||
|
\ee%
|
||
|
where $n=k_{F}^{3}/3\pi ^{2}$ is the 3d density, $J$ is the effective
|
||
|
spin-exchange interaction and $\beta =1/k_{B}T$. Here we will consider the
|
||
|
limit $\hbar /\tau _{\varphi }\ll \Delta $, relevant for our experiment on
|
||
|
Mn. In this limit we can neglect the $1/\tau _{\varphi }$ terms inside the
|
||
|
integral. The upper limit should be restricted to $\Delta /A$ in the limit $%
|
||
|
\Delta /A<1/l$. For large disorder, we expect the parameter $x\equiv
|
||
|
\hbar Dk_{F}^{2}\Delta / \bar{J}^{2}\ll 1$, where the spin-exchange energy
|
||
|
is given by $\bar{J}=Ak_{F}$. In this limit, $L_{\varphi }$ can be
|
||
|
simplified as
|
||
|
\be
|
||
|
k_{F}L_{\varphi }\approx \left( \frac{\bar{J}}{\Delta }\right) ^{3/2}\left(
|
||
|
\frac{5\sinh \frac{\Delta }{T}}{12\pi }\right) ^{1/2},\;\;\;x\ll 1
|
||
|
\label{L-phi-3d}
|
||
|
\ee%
|
||
|
which is independent of $x$, and therefore, independent of disorder.
|
||
|
|
||
|
Given the inelastic lifetime, the weak localization correction in 3d is
|
||
|
usually given by \cite{lee_1985} $\delta \sigma _{3d}=\frac{e^{2}}{\hbar \pi
|
||
|
^{3}}\frac{1}{L_{\varphi }},$ where the prefactor to the inverse inelastic
|
||
|
length is a universal number, independent of disorder. However, at large
|
||
|
enough disorder, we show that there exists a disorder dependent correction,
|
||
|
due to the scale dependent diffusion coefficient near the Anderson
|
||
|
metal-insulator transition. In fact, the diffusion coefficient obeys the
|
||
|
self consistent equation \cite{WV}
|
||
|
\begin{equation}
|
||
|
\frac{D_{0}}{D(\omega )}=1+\frac{k_{F}^{2-d}}{\pi m}\int_{0}^{1/l}dQ\frac{%
|
||
|
Q^{d-1}}{-i\omega +D(\omega )Q^{2}}
|
||
|
\end{equation}%
|
||
|
where $D_{0}=v_{F}l/d$ is the diffusion coefficient at weak disorder. While
|
||
|
the significance of the prefactor to the integral is not clear, the above
|
||
|
equation remains qualitatively accurate over a wide range near the Anderson
|
||
|
transition. Setting $\omega =i/\tau _{\varphi }$ and doing the $Q$-integral
|
||
|
in 3d,
|
||
|
\bea
|
||
|
\frac{D_{0}}{D} &\approx & 1+\frac{1}{\pi mk_{F}}\int_{1/L_{\phi }}^{1/l}dQ\frac{%
|
||
|
Q^{2}}{DQ^{2}}\cr
|
||
|
&=& 1+\frac{D_{0}}{D}\frac{3}{\pi k_{F}^{2}l^{2}}-\delta
|
||
|
\left( \frac{D_{0}}{D}\right) ,
|
||
|
\label{delta}
|
||
|
\eea%
|
||
|
where
|
||
|
\bea
|
||
|
\delta \equiv \frac{D_{0}}{D}\frac{3}{\pi k_{F}^{2}l^{2}}\frac{l}{%
|
||
|
L_{\varphi }}
|
||
|
\eea
|
||
|
is assumed to be a small correction, and Eq.~(\ref{delta})
|
||
|
should not be solved self-consistently. This follows from the fact that the
|
||
|
diffusion coefficient of electrons at fixed energy entering the Cooperon
|
||
|
expression is that of non-interacting electrons, and is given by the limit $%
|
||
|
T\rightarrow 0$, $L_{\varphi }\rightarrow \infty $ and therefore $\delta
|
||
|
\rightarrow 0$. Then the correction at finite $T$ is given by
|
||
|
\bea
|
||
|
\frac{D}{D_{0}} &=& \frac{1}{\left( \frac{D_{0}}{D}\right) _{0}-\delta \left(
|
||
|
\frac{D_{0}}{D}\right) }\cr
|
||
|
&\approx & \left( \frac{D}{D_{0}}\right) _{0}+\left( \frac{D}{D_{0}}\right) _{0}
|
||
|
\frac{3}{\pi k_{F}^{2}l^{2}}\frac{l}{L_{\varphi }}%
|
||
|
\eea%
|
||
|
where
|
||
|
\be
|
||
|
\lim_{T\rightarrow 0}\frac{D}{D_{0}}\equiv \left( \frac{D}{D_{0}}\right)
|
||
|
_{0}.
|
||
|
\ee%
|
||
|
Using the relation $\sigma _{3d}=(e^{2}/\hbar )nD$ where the longitudinal
|
||
|
sheet conductance $\sigma _{\square }=\sigma _{3d}t$, with $t$ being the
|
||
|
film thickness, we finally get the temperature dependent weak localization
|
||
|
correction term
|
||
|
\bea
|
||
|
\frac{\delta \sigma _{\square }}{L_{00}} &=& \left( \frac{D}{D_{0}}\right) _{0}%
|
||
|
\frac{2}{\pi }\frac{t}{L_{\varphi }}\cr
|
||
|
\left( \frac{D}{D_{0}}\right)_{0} &\approx &\frac{2}{1+\sqrt{1+\frac{4R_{0}^{2}}{a^{2}}}}
|
||
|
\label{WL}
|
||
|
\eea%
|
||
|
where $R_{0}=L_{00}/\sigma _{\square }(T$=$0)$, $L_{00}=e^{2}/\pi h$, $%
|
||
|
a=3\pi/2k_{F}tb_{0}$, $b_{0}$ is a number of order unity and we
|
||
|
have solved the self-consistent equation for $D$ in order to express $D_{0%
|
||
|
\text{ }}$in terms of $D$ and finally $R_{0}$. Thus in this case, the weak
|
||
|
localization correction has a prefactor which is not universal. While this
|
||
|
reduces to the well-known universal result at weak disorder $R_{0}\ll a$, it
|
||
|
becomes dependent on disorder characterized by the sheet resistance $R_{0}$
|
||
|
at strong disorder and at the same time substantially extends the 3d regime
|
||
|
near the transition.
|
||
|
|
||
|
Using the expression for $L_{\varphi }$ (Eq.~(\ref{L-phi-3d})) into Eq.~(\ref%
|
||
|
{WL}), we finally obtain the total conductivity, including the quantum
|
||
|
correction to the conductivity due to weak localization in 3d arising from
|
||
|
scattering of electrons off antiferromagnetic spin waves in Mn,
|
||
|
\begin{equation}
|
||
|
\frac{\sigma _{\square }}{L_{00}}=A+\frac{B}{\sqrt{\sinh [\Delta /T]}},
|
||
|
\label{sigmaWL}
|
||
|
\end{equation}%
|
||
|
\textbf{\textbf{}}where the parameter $A$ is temperature independent and the parameter
|
||
|
\bea
|
||
|
B &\equiv & \left( \frac{D}{D_{0}}\right) _{0}\frac{2}{\pi ^{2}}\left( \frac{%
|
||
|
12\pi }{5}\right) ^{1/2}\left( \frac{\Delta }{\bar{J}}\right) ^{3/2}tk_{F}\cr%
|
||
|
&=&\frac{2c}{1+\sqrt{1+\frac{4R_{0}^{2}}{a^{2}}}},
|
||
|
\label{BFit}
|
||
|
\eea%
|
||
|
where
|
||
|
\be
|
||
|
c\equiv \left( \frac{\Delta }{\bar{J}}\right) ^{3/2}\left( \frac{%
|
||
|
48t^{2}k_{F}^{2}}{5\pi}\right) ^{1/2}.
|
||
|
\label{cFit}
|
||
|
\ee
|
||
|
|
||
|
The data presented here is for a single film prepared with an initial $R_0
|
||
|
\approx$~6~k$\Omega$. Disorder was consequently increased in incremental
|
||
|
stages up to 180~k$\Omega$ by annealing at approximately 280~K~\cite%
|
||
|
{misra_2011}. Additional samples were grown at intermediate disorder and
|
||
|
measured to check reproducibility.
|
||
|
|
||
|
Figure~\ref{fig:cond} shows the conductivity data for two samples with
|
||
|
disorder $R_{0}=$~17573~$\Omega $ and 63903~$\Omega $ with corresponding
|
||
|
fittings to the expression (\ref{sigmaWL}) where $A$ and $B$ are taken as
|
||
|
fitting parameters and $\Delta =$~16~K is the spin wave gap. The fits are
|
||
|
sensitive to the parameters $A$ and $B$ but relatively insensitive to $%
|
||
|
\Delta $. We find that $\Delta =$~16~$\pm $~4~K provides good fittings in
|
||
|
the whole range of disorder (from 6 to 180~k$\Omega $).
|
||
|
|
||
|
\begin{figure}[tbp]
|
||
|
\begin{center}
|
||
|
\includegraphics[width=9cm]{fig_1_16.eps}
|
||
|
\end{center}
|
||
|
\caption{The temperature-dependent normalized conductivity (open squares)
|
||
|
for two samples with the indicated disorder strengths of $R_0 =$~17573~$%
|
||
|
\Omega$ and 63903~$\Omega$ show good agreement with theory (solid lines).
|
||
|
The fitting parameters $A$ and $B$ are indicated for each curve with the
|
||
|
error in the least significant digit indicated in parentheses.}
|
||
|
\label{fig:cond}
|
||
|
\end{figure}
|
||
|
|
||
|
Figure~\ref{fig:parb} shows the dependence of the parameter $B$ on the
|
||
|
disorder strength $R_0$ (open squares) and a theoretical fit (solid line)
|
||
|
using Eq.~(\ref{BFit}), where $c$ and $a$ are fitting parameters. The solid
|
||
|
line for this two-paramener fit is drawn for the best-fit values $c=0.67 \pm
|
||
|
0.04$ and $a= 28 \pm 3$~k$\Omega$. We note that the fit is of reasonable
|
||
|
quality over most of the disorder range except for the film with the least
|
||
|
disorder ($R_0 = 6$~k$\Omega$) where $B = 0.77$,
|
||
|
somewhat below the saturated value
|
||
|
$B = c = 0.67$ evaluated from Eq.~(\ref{BFit}) at $R_0 = 0$. Using higher
|
||
|
values of $c$ (e.g., $c=0.8$) and lower values of $a$ (eg., $a = 22$~k$\Omega$)
|
||
|
improves the fit at low disorder strengths but
|
||
|
increases the discrepancy at higher disorder strengths.
|
||
|
|
||
|
%L_phi/t = 2/pi*2/(1+sqrt(1+16))/0.5, 2/pi*2/(1+sqrt(1+1))/0.25
|
||
|
|
||
|
%http://hyperphysics.phy-astr.gsu.edu/hbase/tables/fermi.html , k_F = sqrt(2*m_e*(10.9 eV))/(hbar) = 1.7E10 1/m
|
||
|
|
||
|
% (bar(J) / \Delta) ^ 3/2 = (48*(2e-9)^2*(2.7e9)^2/5/pi/(0.65)^2) ^0.5 = 8360 = 20 ^ 3
|
||
|
%A = \bar{J} / k_F , \bar{J} = nJ
|
||
|
|
||
|
Substituting the Fermi energy for bulk Mn~\cite{ashcroft_1976},
|
||
|
a thickness $t=2$~nm known to 20\% accuracy, together with the best-fit
|
||
|
value for $c$ into Eq.~(\ref{cFit}), we calculate the value $\bar{J} =$~320~$%
|
||
|
\pm$~93~K. Gao et al.~\cite{gao_2008} performed inelastic scanning tunneling
|
||
|
spectroscopy (ISTS) on thin Mn films and reported $\Delta$ in the range from
|
||
|
30 to 60~K and $\bar{J}=vk_F=$~3150~$\pm$~200~K. The agreement of energy gaps is
|
||
|
good; however our significantly lower value of $\bar{J}$ is probably due to the
|
||
|
high disorder in our ultra thin films.
|
||
|
|
||
|
Since the temperature-dependent correction $B/\sqrt{\sinh (\Delta /T)}$ of
|
||
|
Eq.~\ref{sigmaWL} is small compared to the parameter $A$, we can write
|
||
|
$\sigma_{\square} \approx 1/R_0$ so that Eq.~\ref{sigmaWL} reduces to the
|
||
|
expression $A \approx 1/L_{00}R_0$. The logarithmic plot derived by taking the
|
||
|
logarithm of both sides of this approximation is shown in the inset of
|
||
|
Fig.~\ref{fig:parb}. The slope of -1 confirms the linear dependence of $A$ on
|
||
|
$1/R_0$ and the intercept of 5.01 (10$^{5.01}\approx $~102~k$\Omega$) is
|
||
|
within 20\% of the expected theoretical value $L_{00}=$~81~k$\Omega $,
|
||
|
for the normalization constant. Accordingly, the conductivity corrections in
|
||
|
Eq.~\ref{sigmaWL} are small compared to the zero temperature conductivity and
|
||
|
the normalization constant $L_{00}$ for the conductivity is close to the
|
||
|
expected theoretical value.
|
||
|
|
||
|
Using Eq.~(\ref{WL}) and the obtained value for $a\approx $~28~k$\Omega $ we can
|
||
|
compare the dephasing length ($L_{\varphi }$) with the thickness ($t\approx $%
|
||
|
~2~nm) at 16~K. For the sample with $R_{0}=$~63903~$\Omega $ the ratio $%
|
||
|
L_{\varphi }/t\approx $~0.5 and for the sample with $R_{0}=$~17573~$\Omega $
|
||
|
$L_{\varphi }/t\approx $~2. The latter estimate assumes no spin
|
||
|
polarization, while a full polarization would imply $L_{\varphi }/t\approx $%
|
||
|
~1. Thus $L_{\varphi }$ is smaller than or close to the thickness of the
|
||
|
film, which keeps the film in the three-dimensional regime for almost all
|
||
|
temperatures and disorder strengths considered.
|
||
|
|
||
|
\begin{figure}[tbp]
|
||
|
\begin{center}
|
||
|
\includegraphics[width=9cm]{fig_2_16.eps}
|
||
|
\end{center}
|
||
|
\caption{Dependence of the fitting parameters $B$ and $A$ (inset) on
|
||
|
disorder $R_0$ for $\Delta=$~16~K. The fitting parameters are indicated for
|
||
|
each curve with the error in the least significant digit indicated in
|
||
|
parentheses.}
|
||
|
\label{fig:parb}
|
||
|
\end{figure}
|
||
|
|
||
|
In conclusion, we have performed \textit{in situ} transport measurements on
|
||
|
ultra thin Mn films, systematically varying the disorder ($R_{0}=R_{xx}$($T=$%
|
||
|
~5~K)). The obtained data were analyzed within a weak localization theory in
|
||
|
3d generalized to strong disorder. In the temperature range considered
|
||
|
inelastic scattering off spin waves is found to be strong giving rise to a
|
||
|
dephasing length shorter than the film thickness, which places these systems
|
||
|
into the 3d regime. The obtained value for the spin wave gap was close to
|
||
|
the one measured by Gao et al.~\cite{gao_2008} using ISTS, while the
|
||
|
exchange energy was much smaller.
|
||
|
|
||
|
This work has been supported by the NSF under Grant No 1305783 (AFH).
|
||
|
PW thanks A.\ M.\ \ Finkel'stein for useful discussions and acknowledges
|
||
|
partial support through the DFG research unit "Quantum phase transitions".
|
||
|
|
||
|
\bibliographystyle{apsrev}
|
||
|
\bibliography{bibl}
|
||
|
|
||
|
\end{document}
|