1: %% This document created by Scientific Word (R) Version 3.0
2:
3: \documentclass[12pt]{article}
4: \usepackage{amsmath}
5: \usepackage{graphicx}
6: \usepackage{amsfonts}
7: \usepackage{amssymb}
8: %TCIDATA{OutputFilter=latex2.dll}
9: %TCIDATA{CSTFile=LaTeX article (bright).cst}
10: %TCIDATA{Created=Sunday, May 16, 2004 09:34:43}
11: %TCIDATA{LastRevised=Wed Jan 12 12:58:41 2005}
12: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
13: %TCIDATA{<META NAME="DocumentShell" CONTENT="Standard LaTeX\Blank - Standard LaTeX Article">}
14: %TCIDATA{Language=American English}
15: \setcounter{MaxMatrixCols}{30}
16: \newtheorem{theorem}{Theorem}
17: \newtheorem{acknowledgement}[theorem]{Acknowledgement}
18: \newtheorem{algorithm}[theorem]{Algorithm}
19: \newtheorem{axiom}[theorem]{Axiom}
20: \newtheorem{case}[theorem]{Case}
21: \newtheorem{claim}[theorem]{Claim}
22: \newtheorem{conclusion}[theorem]{Conclusion}
23: \newtheorem{condition}[theorem]{Condition}
24: \newtheorem{conjecture}[theorem]{Conjecture}
25: \newtheorem{corollary}[theorem]{Corollary}
26: \newtheorem{criterion}[theorem]{Criterion}
27: \newtheorem{definition}[theorem]{Definition}
28: \newtheorem{example}[theorem]{Example}
29: \newtheorem{exercise}[theorem]{Exercise}
30: \newtheorem{lemma}[theorem]{Lemma}
31: \newtheorem{notation}[theorem]{Notation}
32: \newtheorem{problem}[theorem]{Problem}
33: \newtheorem{proposition}[theorem]{Proposition}
34: \newtheorem{remark}[theorem]{Remark}
35: \newtheorem{solution}[theorem]{Solution}
36: \newtheorem{summary}[theorem]{Summary}
37: \newenvironment{proof}[1][Proof]{\noindent\textbf{#1.} }{\ \rule{0.5em}{0.5em}}
38:
39: \begin{document}
40:
41:
42: \title{The anomalous Hall effect in ferromagnetic Fe: Skew scattering or side jump?}
43: \author{Gerd Bergmann and Manjiang Zhang\\Physics Department, University of Southern California,\\Los Angeles, CA 90089-0484, USA}
44: \date{\today}
45: \maketitle
46: \begin{abstract}
47: The question is investigated whether the anomalous Hall effect (AHE) in Fe
48: films is due to skew scattering or side jump. For this purpose sandwiches of
49: FeIn are investigated in which the conduction electrons carry their drift
50: velocity across the interface. This yields an additional AHE conductance
51: $\Delta G_{xy}$ whose dependence on the In mean free path is used to determine
52: the mechanism of the AHE in the Fe film. The structure of the Fe film is kept constant.
53:
54: PACS: 72.10.-d, 72.25.Mk, 73.40.Jn72.25.Ba, B145
55: \end{abstract}
56:
57:
58:
59: In ferromagnetic metals and metals with magnetic impurities one observes two
60: contributions to the Hall effect, (i) the normal Hall effect and (ii) the
61: anomalous Hall effect (AHE). The AHE is caused by spin-orbit scattering
62: through the interaction of the conduction electron spin with the magnetic
63: moments of the sample. The anomalous Hall effect was already observed by Hall
64: a century ago \cite{H28}. However, theoretically it is a rather complicated
65: problem. There are two main mechanisms discussed in the literature, (a) skew
66: scattering and (b) side jump. Both require a scattering mechanism for the
67: conduction electrons and vanish in pure samples. The first models of skew
68: scattering were developed by Karpulus and Luttinger \cite{L25} and Smit
69: \cite{S30}, while the side jump was proposed by Berger \cite{B138}. Due to its
70: importance in spintronics, the anomalous Hall effect has been intensively
71: studied in recent years \cite{N10}, \cite{S45}, \cite{K48}, \cite{O27} (for
72: further references see \cite{C19}). Recently an additional mechanism has been
73: under discussion which is connected with the Berry phase and believed to occur
74: even in the absence of any scattering (see for example \cite{K49}). Here we
75: restrict the discussion to the skew scattering and the side jump.
76:
77: It is often stated in theoretical papers that for skew scattering the
78: anomalous Hall resistivity $\rho_{yx}$ is proportional to the resistivity
79: $\rho_{xx}$ while for the side jump $\rho_{yx}$ is proportional to the square
80: of the resistivity $\rho_{xx}^{2}$. A number of experimental investigation
81: tried to identify the mechanism of the AHE by changing the resistivity of
82: their sample and analyzing the dependence of $\rho_{yx}$ on $\rho_{xx}$.
83:
84: In this paper we will first show that the power law $\rho_{yx}\varpropto
85: \rho_{xx}^{p}$ with $p=1$ for skew scattering and $p=2$ for the side jump is
86: rather poorly justified. In the second part of the paper we use a very
87: different experimental approach to identify the origin of the AHE in thin
88: amorphous Fe films.
89:
90: We briefly review the AHE resistivity for skew scattering and the side-sump in
91: a ferromagnetic sample where the magnetic moments are aligned in the
92: z-direction. The smple is disordered and has a finite $\rho_{xx}$. Part of the
93: scattering will be potential scattering, i.e. spin-independent, and another
94: part will be magnetic or spin-dependent.
95:
96: We begin with \textbf{skew scattering}. First we consider conduction electrons
97: with spin $\sigma$. In Fig.1 an electron propagates in the x-direction. A part
98: of the wave is skew-scattered by a magnetic moment. We describe the potential
99: scatterers by their concentration $n_{i}$ and their total scattering cross
100: section $a_{i}$ (The index $i$ for impurity). Similarly the magnetic
101: scatterers have the concentration $n_{m}$ with the scattering cross section
102: $a_{m\sigma}$, i.e. the strength of the scattered wave is given by
103: $a_{m\sigma}$ ($a_{m\sigma}$ depends on the spin of the conduction electron).
104: The integrated momentum of the skew scattered wave possesses an electron
105: momentum in the y-direction with the weight $a_{AH,\sigma}.$Here
106: $a_{AH,\sigma}$ is the AHE cross section. It is defined so that $\hbar
107: k_{F}a_{AH,\sigma}$ is equal to the y-component of the integrated momentum of
108: the scattered wave. (A possible forward scattering should be incorporated into
109: the scattering cross section $a_{m\sigma}$.) With the definition of a
110: relaxation time $\tau_{\sigma}$ and the corresponding mean free path
111: $l_{\sigma}$%
112: \[
113: \frac{1}{\tau_{\sigma}}=\frac{v_{F,\sigma}}{l_{\sigma}}=v_{F,\sigma}\left(
114: n_{i}a_{i}+n_{m}a_{m\sigma}\right)
115: \]
116: one obtains for the longitudinal and transverse resistivities $\rho
117: _{xx,\sigma}$ and $\rho_{yx,\sigma}$ for each spin direction $\sigma$%
118: \begin{align}
119: \rho_{xx,\sigma} & =\frac{m}{n_{\sigma}e^{2}\tau_{\sigma}}\label{royx_skew}\\
120: \rho_{yx,\sigma} & =\rho_{xx,\sigma}l_{\sigma}n_{m}a_{AH,\sigma}%
121: =\rho_{xx,\sigma}\frac{n_{m}a_{AH,\sigma}}{n_{i}a_{i}+n_{m}a_{m\sigma}%
122: }\nonumber
123: \end{align}
124:
125: The ratio $\rho_{yx,\sigma}/\rho_{xx,\sigma}$ (and therefore $\rho_{yx,\sigma
126: }$) is rather intuitive. According to Fig.1 the electrons propagate the
127: distance of the mean free path $l_{\sigma}$ (MFP) before they lose their drift
128: velocity. After the skew scattering the fraction $n_{m}a_{AH,\sigma}/\left(
129: n_{i}a_{i}+n_{m}a_{m\sigma}\right) $ of the electron propagates the same MFP
130: $l_{\sigma}$ in y-direction. This yields the ratio $l_{y,\sigma}/l_{x,\sigma
131: }=\rho_{yx,\sigma}/\rho_{xx,\sigma}=n_{m}a_{AH,\sigma}/\left( n_{i}%
132: a_{i}+n_{m}a_{m\sigma}\right) $.%
133: \[%
134: \begin{tabular}
135: [c]{l}%
136: %TCIMACRO{\FRAME{itbpF}{3.3059in}{3.8032in}{0in}{}{}{ahe1.eps}%
137: %{\special{ language "Scientific Word"; type "GRAPHIC";
138: %maintain-aspect-ratio TRUE; display "USEDEF"; valid_file "F";
139: %width 3.3059in; height 3.8032in; depth 0in; original-width 6.4508in;
140: %original-height 7.4271in; cropleft "0"; croptop "1"; cropright "1";
141: %cropbottom "0"; filename 'AHE1.eps';file-properties "NPEU";}}}%
142: %BeginExpansion
143: {\includegraphics[
144: height=3.8032in,
145: width=3.3059in
146: ]%
147: {AHE1.eps}%
148: }%
149: %EndExpansion
150: \\
151: Fig.1a: An (spin up) electron wave with momentum $\mathbf{k=}\left(
152: k_{x},0,0\right) $\\
153: propagates in the x-direction. A part $a_{m\sigma}$ of the wave is
154: skew-scattered\\
155: by a magnetic moment and carries a momentum in the y-direction.\\
156: b) This time the side jump displaces the scattered electrons wave\\
157: by $\Delta y_{\sigma}$ in the y-direction.
158: \end{tabular}
159: \ \ \ \
160: \]
161:
162: Equation (\ref{royx_skew}) yields the well known statement that the AHE
163: resistivity $\rho_{yx,\sigma}$ is proportional to the resistivity
164: $\rho_{xx,\sigma}$, \textbf{for a single spin direction! }Since the electron
165: has two spins we have to (i) invert the resistivity tensors $\left( \rho
166: _{ij}\right) _{\uparrow,\downarrow}$ for each spin to obtain the conductivity
167: tensors $\left( \sigma_{ij}\right) _{\uparrow,\downarrow}$, (ii) add the two
168: conductivity tensors and (iii) invert the resulting tensor.
169: \begin{equation}
170: \left( \rho\right) =\left( \left( \rho\right) _{\uparrow}^{-1}+\left(
171: \rho\right) _{\downarrow}^{-1}\right) ^{-1} \label{invro}%
172: \end{equation}
173: Since the MFPs of spin up and down electrons in the ferromagnet are generally
174: quite different the original linearity between $\rho_{yx,\sigma}$ and
175: $\rho_{xx,\sigma}$ for the individual spin is replaced by a complicated dependence.
176:
177: For the \textbf{side jump} the electron does not propagate in the y-direction
178: after the scattering but the whole scattered electron is displaced by the
179: distance $\Delta y_{\sigma}$ in the y-direction. For the event sketched in
180: Fig.1 the propagation in the x-direction is again $l_{x,\sigma}=l_{\sigma}$
181: while the "propagation" in the y-direction is equal to $l_{y,\sigma}=$ $\Delta
182: y_{\sigma}n_{m}a_{m\sigma}/\left( n_{i}a_{i}+n_{m}a_{m\sigma}\right) $. This
183: yields for $\rho_{yx,\sigma}$ (for each spin) the following value for the side
184: jump
185: \begin{equation}
186: \rho_{yx,\sigma}=\rho_{xx,\sigma}\frac{l_{y,\sigma}}{l_{\sigma}}%
187: =\rho_{xx,\sigma}\Delta y_{\sigma}n_{m}a_{m\sigma} \label{royx_side}%
188: \end{equation}
189: If $\rho_{xx,\sigma}$ is proportional to density of magnetic scattering
190: centers $n_{m}$ then $\rho_{yx,\sigma}$ is proportional to the square of
191: $\rho_{xx,\sigma}$ if the scattering cross section $a_{m\sigma}$ and the side
192: jump are independent of the resistivity, \textbf{for a single spin direction!}
193: Using equ. (\ref{invro}) for the total AHE resistivity destroys the quadratic
194: relation. Furthermore the parameters $a_{i}$, $a_{m\sigma},a_{AH,\sigma}$
195: $\Delta y_{\sigma}$ are not independent of the disorder. The scattering
196: potential is generally not the atomic potential but the deviation from the
197: periodic potential. This potential is generally not spherically symmetric but
198: is rather the gradient of a spherical potential. All of the scattering
199: parameters $a_{i}$, $a_{m\sigma}$, $a_{AH,\sigma}$ and $\Delta y_{\sigma}$
200: will change in complicated ways with the disorder or alloying of the
201: ferromagnet. In particular the behavior of \ $\Delta y_{\sigma}$ as a function
202: of disorder is very critical. As Berger showed the side jump, which is caused
203: by the spin-orbit interaction, only becomes significant because the spin-orbit
204: interaction in the magnetic atoms can be enhanced by a factor of
205: $3\times10^{4}$. Even the smallest change in the local environment of the
206: magnetic atom could change this enhancement factor.
207:
208: We summarize: The simple power law dependence of $\rho_{yx}$ on the
209: resistivity $\rho_{xx}$ for skew scattering and side jump may not be reliable
210: for the following reasons:
211:
212: \begin{itemize}
213: \item The contribution of two kinds of electrical carriers in ferromagnets,
214: spin up and down electrons, destroys the simple relation between $\rho_{yx}$
215: and $\rho_{xx},$
216:
217: \item The scattering parameters $a_{i}$, $a_{m\sigma}$, $a_{AH,\sigma}$ and
218: $\Delta y_{\sigma}$ will change in complicated ways with the disorder or
219: alloying of the ferromagnet.
220:
221: \item The enhancement of the spin-orbit interaction (which determines the
222: side-jump parameter) by a factor of $3\times10^{4}$ will be very sensitive to
223: the disorder.
224: \end{itemize}
225:
226: In the present investigation we use a new approach to investigate the AHE of a
227: ferromagnetic film. We use a thin ferromagnetic film as the target of a
228: scattering experiment by exposing it to incident electrons. The momentum of
229: the incident electrons is varied and the electrons are scattered by the
230: target. Their (integrated) angular scattering is measured. This appears to be
231: a conventional scattering experiment but there is an important difference. The
232: probing electrons are the conduction electrons of a normal metal film which is
233: condensed on top of the ferromagnetic film. The sandwich is shown in Fig.2. In
234: the experiment we use amorphous Fe with a very short MFP for the ferromagnet
235: and In with a much larger MFP for the normal metal. This simplifies the
236: underlying physics and the evaluation of the experiment.
237: \[%
238: \begin{tabular}
239: [c]{l}%
240: %TCIMACRO{\FRAME{itbpF}{3.7592in}{1.6364in}{0in}{}{}{ahe2.eps}%
241: %{\special{ language "Scientific Word"; type "GRAPHIC";
242: %maintain-aspect-ratio TRUE; display "USEDEF"; valid_file "F";
243: %width 3.7592in; height 1.6364in; depth 0in; original-width 8.7339in;
244: %original-height 3.7725in; cropleft "0"; croptop "1"; cropright "1";
245: %cropbottom "0"; filename 'AHE2.eps';file-properties "NPEU";}}}%
246: %BeginExpansion
247: {\includegraphics[
248: height=1.6364in,
249: width=3.7592in
250: ]%
251: {AHE2.eps}%
252: }%
253: %EndExpansion
254: \\
255: Fig.2: A sandwich consisting of ferromagnetic amorphous Fe and\\
256: the normal metal In is quench condensed. In the presence of an\\
257: electric field the conduction electrons in the In carry their larger\\
258: drift velocity into the upper layer of the Fe and create a large\\
259: anomalous Hall effect (AHE) in the Fe. Its dependence on the mean free\\
260: path of the conduction electrons in the In identifies the origin\\
261: of the AHE.
262: \end{tabular}
263: \ \ \ \ \
264: \]
265:
266: In the presence of an electric field $E$ (in the x-direction) the electrons
267: accumulate finite drift velocities: $v_{n}$ in the normal metal and
268: $v_{f\uparrow},v_{f\downarrow}$ in the amorphous ferromagnet. The electrons of
269: both metals cross the interface. The electrons which cross from the normal
270: metal into the ferromagnet increase the current density in the upper layers of
271: the Fe dramatically because they carry a much larger drift velocity. This
272: injected high current density in the ferromagnet is proportional to the MFP in
273: the normal metal. It creates an additional AHE in the Fe. If the AHE is due to
274: the side-jump mechanism then the injected current yields an AHE conductance
275: which is proportional to the MFP $l_{n}$ \ in the normal metal. If the AHE is
276: due to skew scattering then a large fraction of the scattered electrons
277: returns into the normal metal and propagates there the distance $l_{n}$.
278: Therefore their contribution to the AHE conductance is proportional to the
279: square of the MFP in the normal metal. By changing the MFP in the normal metal
280: we can analyse the origin of the AHE in the ferromagnet without changing the
281: structure of the ferro-magnet.
282:
283: Our FeIn sandwiches are prepared at liquid helium temperatures. To obtain very
284: flat and homogeneous Fe films we first condense 10 atomic layers of insulating
285: amorphous Sb. On this fresh substrate the Fe film shows conductance already
286: for one mono layer. The thickness of the Fe films lies in the range of 5 to 10
287: atomic layers. On top of the Fe film the In is condensed in several steps up
288: to a thickness of $25nm$. The MFP of the In lies in the range of $5-20nm$
289: while the MFP of the Fe is of the order of a few Angstroms. Fig.3 shows the
290: anomalous Hall curves for a sandwich of 5 atomic layers of amorphous Fe
291: covered with increasing layers of In. The normal Hall conductance is
292: subtracted. From these curves we obtain the AHE conductance $L_{xy}^{AHE}$ by
293: back extrapolation of the high field part of the curve to zero magnetic field.
294: We observe an increase of the AHE conductance with the In thickness. We denote
295: the additional AHE conductance as \emph{interface} AHE conductance.%
296:
297: \begin{align*}
298: &
299: %TCIMACRO{\FRAME{itbpF}{4.3462in}{2.9805in}{0pt}{}{}{f_{a}rb81_{1}%
300: %.eps}{\special{ language "Scientific Word"; type "GRAPHIC";
301: %maintain-aspect-ratio TRUE; display "USEDEF"; valid_file "F";
302: %width 4.3462in; height 2.9805in; depth 0pt; original-width 6.2391in;
303: %original-height 4.2706in; cropleft "0"; croptop "1"; cropright "1";
304: %cropbottom "0"; filename 'F_B145_1.eps';file-properties "NPEU";}}}%
305: %BeginExpansion
306: \raisebox{-0pt}{\includegraphics[
307: height=2.9805in,
308: width=4.3462in
309: ]%
310: {F_B145_1.eps}%
311: }%
312: %EndExpansion
313: \\
314: &
315: \begin{tabular}
316: [c]{l}%
317: Fig.3: The anomlous Hall conductance curves for a thin Fe film\\
318: covered with In of increasing thickness.
319: \end{tabular}
320: \end{align*}
321:
322: In Fig.4 we have plotted the interface AHE conductance $\Delta L_{xy}$ as a
323: function of the In thickness for two different amorphous Fe film thicknesses
324: of 5 and 8.6 atomic layers. The fact that the two curves lie very close to
325: each other demonstrates that the interface AHE conductance does not depend on
326: the thickness of the ferromagnet (as long as the thickness is larger than the
327: MFP).
328: \begin{align*}
329: &
330: %TCIMACRO{\FRAME{itbpF}{4.0598in}{3.2594in}{0in}{}{}{f_{a}rb81_{2}%
331: %.eps}{\special{ language "Scientific Word"; type "GRAPHIC";
332: %maintain-aspect-ratio TRUE; display "USEDEF"; valid_file "F";
333: %width 4.0598in; height 3.2594in; depth 0in; original-width 3.8497in;
334: %original-height 3.0843in; cropleft "0"; croptop "1"; cropright "1";
335: %cropbottom "0"; filename 'F_B145_2.eps';file-properties "NPEU";}}}%
336: %BeginExpansion
337: {\includegraphics[
338: height=3.2594in,
339: width=4.0598in
340: ]%
341: {F_B145_2.eps}%
342: }%
343: %EndExpansion
344: \\
345: &
346: \begin{tabular}
347: [c]{l}%
348: Fig.4: The interface anomalous Hall conductance of two FeIn\\
349: sandwiches as a function of the In thickness. The thicknesses\\
350: of the amorphous Fe films are 5 and 8.6 atomic layers.
351: \end{tabular}
352: \end{align*}
353:
354: For the evaluation and interpretation of the experiment we calculated the AHE
355: conductance of an FN (ferromagnet/non-magnetic metal) sandwich. (The details
356: will be published elsewhere). We applied the (linearized) Boltzmann equation
357: using Chamber's method of the vector mean free path \cite{C17}. The fact that
358: the Fermi energies differ in the two metals complicates the calculation
359: considerably. Therefore we follow here the examples in the theory of giant
360: magneto-resistance and the superconducting proximity effect where the first
361: theoretical approaches simplified the problem by assuming identical electronic
362: properties in both metals. Furthermore we take the densities of spin up and
363: down electrons in the ferromagnet as identical and equal to $n/2$.
364:
365: The thickness and MFPs of the Fe film are denoted as $d_{f},l_{f\uparrow},$
366: $l_{f\downarrow}$for spin up and down and $d_{n},l_{n}$ for the In. Because
367: the Fe and In film are in parallel their conductances would simply add if
368: there would be no interface crossing between the films. Without the crossing
369: the In would not contribute to the AHE.
370:
371: The conduction electrons in the In with $k_{z}<0$ cross through the interface
372: into the Fe. In the following we call them the injected electrons. In the Fe
373: they carry an injected current $I_{in}$ for each spin which for $l_{n}%
374: >>l_{f\uparrow},l_{f\downarrow}$ is
375: \[
376: I_{in}=\frac{1}{16}e^{2}N_{0}v_{F}l_{n}\left( l_{f\uparrow}+l_{f\downarrow
377: }\right) E
378: \]
379: where $N_{0}$ is the density of electron states per spin. This current flows
380: in a thin layer of the Fe whose thickness is half the MFP, i.e.,
381: $l_{f\uparrow}/2$ or $l_{f\downarrow}/2$ for spin up and down electrons and is
382: proportional to the MFP in the normal metal. The result for the longitudinal
383: part of the conductance $G_{xx}=I_{in}/E$ is similar to Fuchs \cite{F31} and
384: Sondheimer \cite{S36} for thin films but extended to sandwiches. The injected
385: current yields an additional large AHE. The resulting contribution to the
386: anomalous Hall conductance depends on the mechanism of the AHE.
387:
388: \textbf{side jump:} The electrons which cross from the normal metal into the
389: ferromagnet contribute to the side jump. They yield an additional AHE
390: conductance%
391: \[
392: \Delta G_{xy}^{\left( sj\right) }=\frac{1}{16}e^{2}N_{0}v_{F}l_{n}\left(
393: l_{f\uparrow}\Delta y_{\uparrow}n_{m}a_{0\uparrow,m}+l_{f\downarrow}\Delta
394: y_{\downarrow}n_{m}a_{0\downarrow,m}\right)
395: \]
396: The electrons which cross from the ferromagnet to the normal metal do not
397: contribute to the AHE.
398:
399: \textbf{skew scattering:} In contrast to the side jump, here part of the
400: important physics happens after the scattering because half of the skew
401: scattered electrons propagate back towards the normal film. Therefore one
402: obtains two additional contributions due to the normal metal film: (i) The
403: conduction electrons which are accelerated in the ferromagnet and cross into
404: the normal metal after the scattering and (ii) the conduction electrons which
405: are accelerated in the normal metal, cross the interface into the ferromagnet,
406: experience skew scattering and then cross back into the normal metal. The
407: second effect is proportional to $l_{n}^{2}$ and is dominant. It yields an
408: additional anomalous Hall conductance
409: \[
410: \Delta G_{xy}^{(ss)}=\frac{\beta}{32}e^{2}N_{0}v_{F}l_{n}^{2}n_{m}\left(
411: a_{ah\uparrow}l_{f\uparrow}+a_{ah\downarrow}l_{f\downarrow}\right)
412: \]
413: The additional factor $\beta/2$ (with respect to the current) is due to the
414: fact that only half of the scattered electrons move back towards the normal
415: metal. Since they are roughly the distance $l_{f\uparrow}/2,l_{f\downarrow}/2$
416: from the interface only the fraction $\beta$ reaches the normal metal without
417: being scattered in the ferromagnet. The factor $\beta$ is less than one and of
418: the order of $1/2$. (If the scattering in the ferromagnet would be isotropic
419: then $\beta$ would have the value 1/2)
420:
421: For skew scattering the additional anomalous Hall conductance is proportional
422: to the square of the MFP in the normal metal. This result is quite physical.
423: The interface AHE conductance is (i) proportional to the drift velocity, i.e.
424: $l_{n}$, and (ii) proportional to the distance the electrons travel after the
425: scattering, which yields another factor of $l_{n}$.
426:
427: For the analysis of the experiment we plot in Fig.5 the total anomalous Hall
428: conductance of the two sandwiches versus the MFP of the conduction electrons
429: in the In film. Together with the experimental points is shown a linear fit.
430: Obviously the interface AHE conductance is linear in the MFP $l_{n}$ of the
431: In. This proves clearly that the anomalous Hall effect in the amorphous Fe
432: film is due to the side jump. Originally Berger's argument for observing the
433: side jump in metals with small MFP was that there the small MFP reduces the
434: magnitude of the skew scattering so that one could observe the side jump. Our
435: conclusion goes beyond that suggestion. Since the propagation in the normal
436: metal yields a contribution which is proportional to the square of the MFP in
437: the normal metal our experiment should detect a contribution of the skew
438: scattering even if its contribution in the amorphous Fe would be very small.
439: So the conclusion of our experiment is that there is practically no skew
440: scattering in the amorphous Fe.%
441: \begin{align*}
442: &
443: %TCIMACRO{\FRAME{itbpF}{4.0598in}{3.1631in}{0in}{}{}{f_{a}rb81_{3}%
444: %.eps}{\special{ language "Scientific Word"; type "GRAPHIC";
445: %maintain-aspect-ratio TRUE; display "USEDEF"; valid_file "F";
446: %width 4.0598in; height 3.1631in; depth 0in; original-width 3.8497in;
447: %original-height 2.9929in; cropleft "0"; croptop "1"; cropright "1";
448: %cropbottom "0"; filename 'F_B145_3.eps';file-properties "NPEU";}}}%
449: %BeginExpansion
450: {\includegraphics[
451: height=3.1631in,
452: width=4.0598in
453: ]%
454: {F_B145_3.eps}%
455: }%
456: %EndExpansion
457: \\
458: &
459: \begin{tabular}
460: [c]{l}%
461: Fig.5: The total anomalous Hall conductance of the two FeIn\\
462: sandwiches shown in Fig.5 as a function of the In mean free path.
463: \end{tabular}
464: \end{align*}
465:
466: In this paper we have investigated the mechanism of the AHE in amorphous Fe.
467: The standard approach assumes that $\rho_{yx}$ depends on $\rho_{xx}$ linearly
468: for skew scattering and quadratically for side jump. First \ we pointed out
469: that this assumption is not well justified. Instead we introduced a method for
470: which the structure of the ferromagnet is kept constant. By preparing a
471: sandwich of amorphous Fe with the non-magnetic metal In we observed an
472: increased AHE conductance because conduction electrons with a larger drift
473: velocity cross from the In into the Fe and cause an additional "interface" AHE
474: within the MFP of the Fe. For the side jump this interface AHE conductance
475: $\Delta G_{xy}$ is proportional the MFP $l_{In}$ of the In. If the AHE is due
476: to skew scattering then about $1/4$ of the skew-scattered electrons cross back
477: into the In and propagate the distance $l_{In}$. This yields a quadratic
478: dependence of $\Delta G_{xy}$ on $l_{In}$ for skew scattering. The great
479: advantage of the interface AHE is that the structure and scattering potentials
480: of \ the ferromagnet are kept constant.
481:
482: Our experimental results yield a linear dependence of $\Delta G_{xy}$ on the
483: In MFP $l_{In}$. This not only shows that the side jump is the dominant
484: mechanism for the AHE in amorphous Fe, but the experiment did not detect any
485: skew scattering in the amorphous Fe film.
486:
487: Abbreviations: AHE=anomalous Hall effect, MFP=mean free path.
488:
489: Acknowledgment: The research was supported by NSF Grant No. DMR-0124422.\newpage
490:
491: \begin{thebibliography}{99}
492:
493: \bibitem{H28}E.H.Hall, Philos.Mag. 12, 157 (1881)\newline
494:
495: \bibitem{L25}R.Karpulus and J.M.Luttinger, Phys.Rev. 95,1154 (1954)\newline
496:
497: \bibitem{S30}J.Smit, Physica 16, 612 (1951)\newline
498:
499: \bibitem{B138}L.Berger, Phys.Rev. B2, 4559 (1970)\newline
500:
501: \bibitem{N10}K.Rhie, D.G.Naugle, O.Beom--hoan and J.T.Markert, Phys.Rev. B48,
502: 5973--5982 (1993) \newline
503:
504: \bibitem{S45}J.Stankiewicz, L.Morellon, P.A.Algarabel, and M.R.Ibarra,
505: Phys.Rev. B61, 12651 (2000) \newline
506:
507: \bibitem{K48}P.Khatua, A.K.Majumdar, A.F.Hebard, and D.Temple, Phys.Rev. B68,
508: 144405 (2003) \newline
509:
510: \bibitem{O27}Wei-Li Lee, S.Watauchi, V.L.Miller, R.J.Cava, N.P.Ong, Science
511: 303, 1647 (2004) \newline
512:
513: \bibitem{C19}A.Cr\'{e}pieux and P.Bruno, Phys.Rev. B64, 014416 (2001) \newline
514:
515: \bibitem{K49}Y. Yao , L.Kleinman, A. H. MacDonald, J.Sinova, T. Jungwirth,
516: D.Wang, E.Wang and Q.Niu, Phys. Rev. Lett. 92, 037204 (2004) \newline
517:
518: \bibitem{C17}R.G.Chambers, The physics of metals, p. 175, ed. J.M.Ziman,
519: Cambridge University Press, 1969\newline
520:
521: \bibitem{F31}K.Fuchs, Proc.Camb.Phil.Soc. 34, 100 (1938)\newline
522:
523: \bibitem{S36}E.H.Sondheimer, Adv.Phys. 50, 499 (2001)\newline
524: \end{thebibliography}
525: \end{document}