1: \documentclass[twocolumn,showpacs,aps,amsmath,amssymb,superscriptaddress]{revtex4}
2: \usepackage{graphicx}
3: \begin{document}
4: \title{Shot Noise in Magnetic Tunnel Junctions: Evidence for Sequential Tunneling}
5: \author{R. Guerrero}
6: \author{F. G. Aliev}
7: \affiliation{Departamento de Fisica de la Materia Condensada, C-III,
8: Universidad Autonoma de Madrid, 28049, Madrid, Spain}
9: \author{Y. Tserkovnyak}
10: \affiliation{Lyman Laboratory of Physics, Harvard University,
11: Cambridge, MA 02138 and
12: Department of Physics and Astronomy, University of California, Los Angeles, California 90095, USA }
13: \author{T. S. Santos}
14: \author{J. S. Moodera}
15: \affiliation{Francis Bitter Magnet Laboratory, Massachusetts Institute of Technology, Cambridge, MA 02139, USA}
16: \date{\today}
17: \begin{abstract}
18: We report the experimental observation of sub-Poissonian shot noise in single magnetic tunnel junctions,
19: indicating the importance of tunneling via impurity levels inside the tunnel barrier. For junctions with weak
20: zero-bias anomaly in conductance, the Fano factor (normalized shot noise) depends on the magnetic configuration
21: being enhanced for antiparallel alignment of the ferromagnetic electrodes. We propose a model of sequential
22: tunneling through nonmagnetic and paramagnetic impurity levels inside the tunnel barrier to qualitatively
23: explain the observations.
24: \end{abstract}
25: \pacs{72.25.-b; 72.25.Mk; 73.40.Gk}
26:
27: \maketitle
28:
29:
30: The discovery of the giant magnetoresistance \cite{Fert88} followed by the observation of a large tunneling
31: magnetoresistance (TMR) at room temperature in magnetic tunnel junctions (MTJ's) \cite{Moodera95} have boosted
32: interest in spin-dependent electron transport in magnetic nanostructures, especially in the spin-dependent
33: tunneling phenomena \cite{RevMoodera99,RevTsymbal03}. During recent years, there has been a growing interest in
34: controlling the TMR and also the statistics of tunneling events in MTJ's by nanostructuring of the insulating
35: barrier \cite{Barnas98}. A variety of new electron-correlation mechanisms have been proposed, typically based on
36: transport through double MTJ's with either an open or Coulomb-blockaded quantum dot (QD) contacted by
37: ferromagnetic electrodes. Electric shot noise (SN) is a powerful tool for studying correlations of tunneling
38: processes in nanostructures beyond the capabilities of dc measurements \cite{BB00}. The growing list of
39: theoretically investigated topics regarding spin-dependent shot noise includes the noise asymmetry between
40: parallel (P) and antiparallel (AP) ferromagnetic (FM) alignment \cite{Bulka99} as well as continuous variation
41: of the SN over the relative angle between FM electrodes \cite{Tserkovnyak01}, SN through an artificial (QD)
42: Kondo impurity \cite{Lopez03} contacted by magnetic leads, shot-noise enhancement by dynamic spin blockade in
43: tunneling through a small QD \cite{Cottet04}, and shot noise for spin-polarized and entangled electrons with
44: spin-orbit interaction in the leads \cite{Egues02}. The scope of experimental efforts
45: \cite{Nowak99,Nowak04,Nowak04II} has however so far been much more limited and inconclusive with regard to the
46: nature of tunneling electron correlations even in the conceptually simplest spintronic devices, viz. MTJ's, as
47: manifested by shot-noise measurements.
48:
49:
50: Current fluctuations due to discreteness of electron charge flowing through the structure out of equilibrium,
51: which provide the shot noise, contain information not accessible by time-independent conductance. Sensitivity to
52: quantum statistics, interference, and interactions between electrons passing through the device has made SN an
53: effective tool for investigating quantum transport in meso- and nano-structures \cite{BB00}. In the absence of
54: any correlations, Poissonian shot noise is practically frequency independent at low frequencies with the noise
55: power given by $S=2eI$, in terms of the average current $I$. The Fano factor $F=S/2eI$ representing normalized
56: shot noise is in general lowered below 1 for noninteracting electrons due to fermionic statistics.
57: Electron-electron interactions can either further suppress or enhance the Fano factor (even beyond the
58: Poissonian value).
59:
60: Despite the theoretical excitement about perspectives of using the shot noise for investigation of
61: spin-polarized electrons, behavior of the SN even in simple nonstructured MTJ's remains unclear.
62: Jiang~\textit{et al.} \cite{Nowak04} reported an observation of the ``full" SN (i.e., $F\sim1$) in MTJ's with AP
63: alignment of electrodes. Later the same group \cite{Nowak04II} measured a strong suppression (down to
64: $F\approx0.45$) of the SN in magnetic tunnel junctions, which was not understood. Our Letter reports the first
65: systematic investigation of the tunneling statistics in a magnetic tunneling device by measuring shot noise in
66: Co(80~\AA )$\mid$Al$_{2}$O$_{3}$(14~\AA )$\mid$Py(100~\AA) MTJ's with and without Cr doping of the insulating
67: barrier. We demonstrate a decrease of the Fano factor and its dependence on the alignment of the ferromagnetic
68: electrodes for certain barrier conditions.
69: %Version 24/10/06
70: %Experimental observations suggest sequential tunneling through intrabarrier impurity levels. Cr doping and
71: %magnetic-configuration control by an applied magnetic field allow us to engineer statistics of spin-polarized
72: %hopping via the tunnel barrier, opening new venues for studying physics of spin-dependent transport and quantum
73: %information.
74:
75: Details of sample preparation have been published previously \cite{Jansen00}. For Cr-doped samples, the tunnel
76: barriers were deposited in two steps. After deposition of the underlying Co electrode, a first tunnel barrier
77: was formed by deposition and subsequent oxidation of 7-9~\AA\ of Al. Subsequently, sub-monolayer amounts of Cr
78: were deposited on the Al$_2$O$_3$ surface, followed by a second Al layer deposition (5-7~\AA) and oxidation,
79: resulting in a ``$\delta$-doped" Al$_2$O$_3$$\mid$Cr$\mid$Al$_2$O$_3$ tunnel barrier. The noise measurements use
80: a setup described in Ref.~\cite{Guerrero05}, which employs the cross-correlation method. This technique removes
81: uncorrelated noise from the amplifiers and the noise of the leads
82: %CUTs Oct06
83: %Computer control of the current through the
84: %sample permits measurements of the noise and dynamic tunneling
85: %resistance as a function of bias current at fixed magnetic field, in
86: %this way
87: and takes into account nonlinearity of the dynamic resistance while converting the obtained voltage noise into
88: current noise. Out of 13 samples investigated, the shot noise was measured for 11 MTJ's: 5 without and 6 with
89: $\delta$-layer of Cr in the middle of the barrier, ranging between 0.2 and 1.2~\AA\ in thickness.
90: %Cuts Oct06
91: %The current flowing through the sample is converted into the voltage, which is analyzed either by a dc voltmeter
92: %or by a lock-in amplifier to measure the bias voltage and the dynamic conductance, respectively. The voltage
93: %from the MTJ's is measured by two identical homemade dc-coupled ultra-low--noise amplifiers placed at the top of
94: %the cryostat. The pre-amplified signals are further amplified by additional low-noise amplifiers (Stanford
95: %Research SR560). A spectrum analyzer SR780 calculates the cross-correlation spectrum in general containing
96: %thermal, shot, and $1/f$ contributions to the noise. The influence of the capacitance of the line (about 400~pF) and of
97: %the MTJ's (about 10~nF) as well as their resistance ($10-600$~k$\Omega$) are taken into account during noise
98: %analysis without using fitting parameters.
99: \begin{figure}
100: \includegraphics[width=\linewidth,clip=]{fig1}
101: \caption{\label{fig1}(Color online) Typical dynamic resistance
102: obtained in P state for the Cr-free (a) and Cr-doped (b) junctions with
103: 0.8~\AA\ $\delta$-layer measured at 300, 77, and 2~K. (c) Reduction of the TMR with applied
104: voltage for Cr-free and Cr-doped MTJ's at
105: $T=2$~K. (d) Dependence of the ZBA(\%) $=100\times [R(0~{\rm m}V)-R(100~{\rm m}V)]/R(100~{\rm m}V)$ determined for the P alignment
106: on Cr (at 2~K).}
107: \end{figure}
108:
109:
110: Figure~\ref{fig1} shows typical electron transport characteristics of the studied MTJ's. The dynamic tunneling
111: resistance vs bias $V$ [Figs.~\ref{fig1}(a),(b)] measured at three temperatures for P alignment proves
112: pinhole-free MTJ's \cite{RevMoodera99}. For all MTJ's studied, an asymmetric parabolic conductance background
113: \cite{brinkman} plus a zero-bias anomaly (ZBA) below $T\sim77$~K, appeared in the resistance of the junction
114: ($R_{J}$) [Figs.~\ref{fig1}(a),(b)].
115: %change to editor III
116: %The general view of the ZBA is that finite temperature and bias allow for some
117: %inelastic-scattering processes coupled to tunneling events, opening additional transport channels.
118: Presently, there exists several possible explanations of the ZBA's in MTJ's
119: \cite{ZangLevyZBA,JAP98}, which consider magnon- or phonon-assisted tunneling or two-step
120: tunneling through impurities inside the tunnel barrier which are also coupled to some additional degrees of
121: freedom.
122: %Such processes require hot electrons, leading to the low-bias/temperature suppression of the
123: %conductance.
124: Simultaneous ZBA and SN measurements on our samples suggest the ZBA is provided by sequential
125: tunneling through impurities accompanied with spin flips.
126: \begin{figure}
127: \includegraphics[width=\linewidth,clip=]{fig2}
128: \caption{\label{fig2} (Color online) (a) Typical voltage noise
129: measured for a Cr-free MTJ at $T=2$~K
130: with the applied currents (from bottom to up) of 3.4, 5 and 6~$\mu$A.
131: (b) Voltage dependence of $F$ on bias, for the Cr-free (filled) and 0.4~\AA\ Cr-doped
132: (open stars), also measured at $T=2$~K. The error bars show
133: standard deviations.}
134: \end{figure}
135: %Doping of the barrier with Cr generally suppresses both the TMR and
136: %the conductivity, but the relations between these parameters and the
137: %nominal Cr concentration are not strictly monotonic [see, e.g.,
138: %Fig.~\ref{fig1}(d)].
139:
140: Doping of the barrier with Cr enhances the normalized ZBA, although this trend presents rather large dispersion
141: [Fig.~\ref{fig1}(d)]. Conductivity and TMR are generally suppressed when Cr thickness is increased, but the
142: relations between these parameter and the nominal Cr concentration are not strictly monotonic. We have found,
143: however, that the changes in the TMR are correlated with those of the tunneling resistance (see below). This can
144: be understood as follows: As the barrier width and the resistance increase, the relative role of two-step
145: tunneling increases, which generally reduces the TMR. The TMR is also monotonically reduced with the applied
146: voltage both for the Cr-free and Cr-doped MTJ's [see Fig.~\ref{fig1}(c)], in accordance with the previous
147: reports \cite{ZangLevyZBA}.
148: %The monotonic suppression of the
149: %TMR and TR with bias indicates we do not have Kondo impurities in the strong-coupling regime (at least down to
150: %the lowest temperature $T=2$~K), where tunneling resistance decreases at $T, V\to0$ due to the Kondo resonance \cite{KondoIV}.
151:
152:
153: \begin{figure}[b]
154: \includegraphics[width=\linewidth,clip=]{fig3}
155: \caption{\label{fig3} (Color online) (a) Dependence of the TMR and $F$
156: (for the P state) on the resistance area product ($R\times A$).
157: Solid horizontal line marks $F=1$. Vertical dashed
158: line separates the Cr-free and Cr-doped regions. (b) Dependence of the
159: relative variation of $F$ with alignment
160: $\Delta F/F_{\rm P}(\%)=100\times(F_{\rm AP}-F_{\rm P})/F_{\rm P}$ on the relative strength of the
161: ZBA. Solid symbols point the undoped samples. The lines are guides for the eye.}
162: \end{figure}
163:
164:
165: The measured low-frequency noise has a typical form for MTJ's, with the $1/f$ noise dominating at $f<100$~Hz and
166: the ``white" noise dominating at $f\gtrsim 100$~Hz. Figure~\ref{fig2}(a) shows a typical voltage noise for the
167: frequency and bias range where the $1/f$ noise does not affect the data and the applied bias ($eV\gg k_{B}T$)
168: ensures that SN presents the dominant contribution to the total noise. Fig.~\ref{fig2}(b) shows a typical
169: dependence of $F$ on bias. For most of the undoped MTJ's, the Fano factor was reduced below the Poissonian value
170: ($F<1$), while for the Cr-doped MTJ's $F$ was always close to one.
171:
172: Figure~\ref{fig3}(a) shows the TMR and the Fano factor for the P alignment as a function of the resistance by
173: area product ($R\times A$) at $T=2$~K. The Fano factor was averaged over the range $40-120$~mV where it is
174: nearly bias independent. For the undoped MTJ's in the range where TMR is only weakly reduced with the product
175: $R\times A$ ($<10^{4}$~M$\Omega\mu $m$^{2}$), we observed a gradual suppression of the Fano factor down to
176: $F\sim0.65$. Doping of the barrier with Cr further increases the tunneling resistance and restores the
177: Poissonian SN ($F\sim1$). The suppression of $F$ in a certain tunneling resistance range is not accompanied by
178: the appearance of random telegraph noise as in Ref.~\cite{Nowak99}, reduced TMR \cite{shotnoise02}, or by
179: metallic temperature dependence $R(T)$, clearly ruling out pin-holes/hot spots across the barrier.
180: Figure~\ref{fig3}(b) shows the normalized AP-P $F$ asymmetry as a function of the normalized ZBA for the P
181: alignment. Surprisingly, we find that $F$ depends on the alignment of the electrodes with $F_{\rm AP}/F_{\rm P}>1$ only in
182: the MTJ's with a weak zero-bias anomaly and becomes nearly independent of the alignment above some threshold
183: value of the ZBA. We stress that the observed Fano factor asymmetry reflects only alignment of the FM
184: electrodes, but not orientation of the magnetic field.
185: \begin{figure}[b]
186: \includegraphics[width=0.9\linewidth,clip=]{fig4}
187: \caption{Two models: (a) a large normal region $R_M$ such that electrons can be treated as noninteracting,
188: coupled to biased emitter and collector reservoirs in P or AP configuration, (b) $R_M$ is a single
189: spin-polarized impurity level that can hold only one electron with spin at a (random) angle $\theta$ with
190: respect to the collector magnetization.} \label{sc}
191: \end{figure}
192: %It was reported previously that for nonmagnetic tunnel junctions
193: %tunneling through localized states within the barrier could indeed
194: %account for the measured reduced Fano factor $F<1$ \cite{Iann03}.
195: %To our best knowledge there have not been
196:
197:
198: Previous studies of the shot noise in non magnetic TJ's with Al$_2$O$_3$ barrier have observed Poissonian value
199: $F\simeq 1$ \cite{SNthermometer}. It was reported, however, that for non-magnetic TJ's with SiO$_2$ barrier,
200: tunneling through localized states within the barrier, could indeed account for the measured reduced $F$
201: \cite{Iann03}. In the following, we consider two simple models for sequential tunneling via an island inside the
202: tunnel barrier (see Fig.~\ref{sc}), which capture some qualitative aspects of our measurements. First, consider
203: tunneling through a normal region ($R_M$) inside the tunnel barrier [Fig.~\ref{sc}(a)]. Neglecting charging
204: effects, we can simply sum the contributions to the (averaged) current and noise for the two spin species. To
205: this end, suppose $R_M$ is coupled asymmetrically to the left and right reservoirs ($R_L$ and $R_R$) with the
206: respective spin-dependent conductances given by
207: \begin{align}
208: g_{L\uparrow}=g/\sqrt{\beta}\,\,\,\,\,\,&{\rm and}\,\,\,\,\,\,g_{R\uparrow}=g\sqrt{\beta}\,,\\
209: g_{L\downarrow}=\alpha g/\sqrt{\beta}\,\,\,\,\,\,&{\rm
210: and}\,\,\,\,\,\,g_{R\downarrow}=\alpha g\sqrt{\beta}\,
211: \end{align}
212: $\beta$ is a dimensionless left-right asymmetry parameter and $\alpha$ characterizes spin polarization. The
213: charge current at the voltage bias $V$ is given by $I=igV$, parametrized by a dimensionless current $i$ that
214: depends on $\alpha$ and $\beta$ only. Let us furthermore write the zero-frequency shot noise as $S=2esgV$. The
215: Fano factor thus becomes $F=S/(2eI)=s/i$.
216: \begin{figure}
217: \includegraphics[width=\linewidth,clip=]{fig5}
218: \caption{Fano factor as a function of the left-right asymmetry parameter $\beta$ setting $\alpha=1/5$ for the two models
219: sketched in Fig.~\ref{sc}. Solid lines are for the P and dashed for the AP magnetic configurations. Insets show
220: the dimensionless current $i$ and noise $s=Fi$ defined in the text. Note the logarithmic scale for $\beta$:
221: Assuming tunnel rates depend exponentially on the barrier thickness, this corresponds to a linear scale for the
222: $R_M$ position inside the tunnel junction (small $\beta$ corresponding to the proximity to $R_L$ and large
223: $\beta$ to $R_R$).} \label{theory}
224: \end{figure}
225:
226: Let us recall first that, in general, noninteracting spinless electrons in double-barrier structures have the
227: series conductance and $F$ (for spinless electrons)
228: %equation 3 in the text
229: %\begin{equation}
230: %G=\frac{g_Lg_R}{g_L+g_R}\,\,\,\,\,{\rm
231: %and}\,\,\,\,\,F=\frac{g_L^2+g_R^2}{(g_L+g_R)^2}\,, \label{GF}
232: %\end{equation}
233: \begin{equation}
234: G=g_{L}g_{R}/(g_{L}+g_{R})\,,\,\,F=(g_{L}^{2}+g_{R}^{2})/(g_{L}+g_{R})^{2}\,, \label{GF}
235: \end{equation}
236: which are valid not only for large semiclassical $R_M$ but also for sequential tunneling through a small $R_M$
237: described by master-equation approach, in which case $g$'s become respective transition rates instead of the
238: tunnel-barrier conductances \cite{BB00}. Summing corresponding current and noise for the two spin channels in
239: the P configuration (neglecting correlations between two spin species), one trivially obtains for the current
240: and $F$ \cite{BB00}
241: %equation 4 in the text
242: \begin{equation}
243: i_{\rm P}=(1+\alpha)\sqrt{\beta}/(1+\beta)\,,\,\,F_{\rm P}=(1+\beta^2)/(1+\beta)^2\,.
244: \end{equation}
245: In the AP configuration,
246: %equation 5 and 6 in the text
247: \begin{align}
248: i_{\rm AP}&=\alpha(1+\alpha)(1+\beta)\sqrt{\beta}/[(\alpha+\beta)(1+\alpha\beta)]\,,\\
249: F_{\rm AP}&=\frac{\alpha^2(1+2\beta-2\beta^2+2\beta^3+\beta^4)}{(\alpha+\beta)^2(1+\alpha\beta)^2}+(\alpha\leftrightarrow\beta)\,,
250: \end{align}
251: where $(\alpha\leftrightarrow\beta)$ is the same as the first summand but with $\alpha$ and $\beta$
252: interchanged. These results are plotted in Fig.~\ref{theory}(a) for $\alpha=1/5$. Note that $F_{\rm AP}-F_{\rm
253: P}>0$ for $\beta\sim1$, roughly corresponding to the center of the junction for the $R_M$ location, which is the
254: region contributing the largest current, see Fig.~\ref{theory}(a). The Fano factor asymmetry is reversed closer
255: to the junction interfaces where the tunneling is asymmetric.
256:
257: Consider now hopping through a single level that can hold only one extra electron, see Fig.~\ref{sc}(b). If
258: there is a large exchange-energy splitting along certain direction $\theta$, one could imagine a situation when
259: only spins polarized along $\theta$ are energetically allowed to tunnel through. We can then calculate the
260: current and noise using Eqs.~(\ref{GF}) where the rates $g_L$ and $g_R$ now depend on $\theta$ and the relative
261: magnetic orientation in the leads \cite{Sloncz05}:
262: %equation 7 and 8in the text
263: %\begin{align}
264: %\label{gL}
265: %g_L&=g_{L\uparrow}\frac{1+\cos\theta}{2}+g_{L\downarrow}\frac{1-\cos\theta}{2}\,,\\
266: %g_R&=g_{R\uparrow}\frac{1\pm\cos\theta}{2}+g_{R\downarrow}\frac{1\mp\cos\theta}{2}
267: %\label{gR}
268: %\end{align}
269: \begin{align}
270: \label{gL}
271: g_L&=g_{L\uparrow}(1+\cos\theta)/2+g_{L\downarrow}(1-\cos\theta)/2\,,\\g_R&=g_{R\uparrow}(1\pm\cos\theta)/2+g_{R\downarrow}(1\mp\cos\theta)/2
272: \label{gR}
273: \end{align}
274: for the P (AP) configuration. Assuming $\theta$ is random, we average the current and noise:
275: $\langle...\rangle_\theta=(1/2)\int_{-1}^1d(\cos\theta)...$. This results in simple expressions for the P MTJ:
276: %equation 9 in the text
277: \begin{equation}
278: i_{\rm P}=(1+\alpha)\sqrt{\beta}/[2(1+\beta)]\,,\,\,F_{\rm P}=(1+\beta^2)/(1+\beta)^2\,,
279: \end{equation}
280: %\begin{equation}
281: %i_{\rm
282: %P}=\frac{1}{2}\frac{(1+\alpha)\sqrt{\beta}}{1+\beta}\,\,\,\,\,{\rm
283: %and}\,\,\,\,\,F_{\rm P}=\frac{1+\beta^2}{(1+\beta)^2}\,,
284: %\end{equation}
285: which are the same as just averaging over $\theta=0$ and $\pi$. There is no simple analytic form for the current
286: and noise in the AP case. We plot the results in Fig.~\ref{theory}(b). Notice that the AP-P asymmetry is
287: significantly reduced in comparison to Fig.~\ref{theory}(a).
288:
289: In undoped MTJ's, we measured typically $F_{\rm AP}>F_{\rm P}$ and both are significantly suppressed below 1,
290: apart from the thinnest tunnel barrier. Both of these findings are consistent with the results in
291: Fig.~\ref{theory} for tunneling predominantly through impurities in the middle of the barrier.
292: %As seen from Fig.~\ref{theory}(a), $F$ can be considerably reduced
293: %below unity even for randomly-distributed impurities within the barrier, as the current is dominated by hopping
294: %through the center.
295: Since $F$ for tunneling through uniformly-distributed point-like
296: localized states is in general $3/4$ (in the absence of hopping
297: correlations between the two spin species) \cite{Nazarov}, which in
298: particular applies to both models in Fig.~\ref{sc}, the AP-P
299: asymmetry would require some structural preference towards tunneling
300: through the middle of the barrier. The Fano factor is reduced to
301: $F\sim 3/4$, as shown in Fig.~\ref{fig3}(a), as the tunnel barrier
302: becomes wider and the role of the two-step tunneling processes
303: become relatively more important. The tunneling resistance does not
304: indicate variable-range hopping involving multi-step tunneling,
305: which was observed for wider tunnel barriers \cite{Xu}. Observation
306: of Poissonian noise after Cr doping could be due to an offset in Cr
307: deposited nominally in the center of the junction, which leads to
308: systematically asymmetric hopping. Finally, the observed correlation
309: in the AP-P Fano factor asymmetry and the ZBA [Fig.~\ref{fig3}(b)]
310: suggest that an inelastic spin-flip mechanism in the barrier is
311: responsible for concurrent reduction of the former and enhancement
312: of the latter.
313:
314: In summary, first systematic shot noise measurements in magnetic tunnel junction show
315: an evidence for sequential tunneling mediated by defects. We demonstrate for the first time that electron
316: tunneling statistics can be manipulated by an applied magnetic field due to their dependence on the relative
317: orientation of ferromagnetic electrodes and also by deliberately doping the tunnel barrier with impurities.
318: Control over the sequential tunneling could find applications in optimizing signal-to-noise ratio in
319: magnetoelectronic devices and provide a new tool for investigating spin-dependent transport of electrons
320: injected by ferromagnetic electrodes.
321: %Finally, presence of sequential tunneling in epitaxial(MgO) barriers through Oxygen vacancies would resolve
322: %disagreement between theory of coherent tunneling and experiment
323: %\cite{Tsymbal}
324:
325: Authors acknowledge P.~LeClair and R.~Villar for critical reading of
326: the manuscript. The work at UAM was supported in parts by Spanish
327: MEC (MAT2003-02600, MAT2006-07196, 05-FONE-FP-010-SPINTRA) and CM
328: (NANOMAGNET). The work at MIT was supported by NSF and KIST-MIT
329: project grants.
330: \begin{thebibliography}{}
331: %\bibitem{Fert88} M.~N.Baibich, J.M.Broto, A.Fert, F.N.Van Dau, F.Petroff, P.Eitenne, G. Creuzet, A. Friederich, and J. Chazelas, Phys. Rev. Lett. {\bf 61}, 2472 (1988).
332: \bibitem{Fert88} M.~N. Baibich \textit{et al.}, Phys. Rev. Lett. {\bf 61}, 2472 (1988).
333: \bibitem{Moodera95} J.~S. Moodera \textit{et al.}, Phys. Rev. Lett. {\bf 74}, 3273 (1995);
334: %\bibitem{Moodera95} J.~S. Moodera, L. R. Kinder, T. M. Wong, and R. Meservey, Phys. Rev. Lett. {\bf 74}, 3273 (1995).
335: T.~Miyazaki and N.~Tezuka, J. Magn. Magn. Mater. {\bf 139}, L231 (1995).
336: \bibitem{RevMoodera99} J.~S. Moodera, J.~Nassar, and G.~Mathon, Ann. Rev. Mater. Science {\bf 29}, 381 (1999).
337: \bibitem{RevTsymbal03} E.~Y. Tsymbal, O.~N. Mryasov, and P.~R LeClair, J. Phys.: Condens. Matter {\bf 15}, R109 (2003);
338: %\bibitem{Yuasa04} S.~Yuasa, T.Nagahama, A.Fukushima, Y.Suzuki, and K.Ando, Nature Materials {\bf 3}, 868 (2004).
339: S.~Yuasa \textit{et al.}, Nature Materials {\bf 3}, 868 (2004);
340: %\bibitem{Parkin04} S.S.~P. Parkin, C.Kaiser, A.Panchula, P.M.Rice, B.Hughes, M.Samant, and S.H.Yang, Nature Materials {\bf 3}, 862 (2004).
341: S.~S.~P. Parkin \textit{et al.}, \textit{ibid.} {\bf 3}, 862 (2004).
342: \bibitem{Barnas98} J.~Barna\'{s} and A.~Fert, Phys. Rev. Lett. {\bf 80}, 1058 (1998); S.~Takahashi and S.~Maekawa, \textit{ibid.} {\bf 80}, 1758 (1998).
343: \bibitem{BB00} Ya.~M. Blanter and M.~B{\"{u}}ttiker, Phys. Rep. {\bf 336}, 1 (2000).
344: %\bibitem{Russek} N.A.~Stutzke, S.~L.~Burkett, S.~E.~Russek, Appl.Phys.Lett. {\bf 82}, 91 (2003).
345: %\bibitem{Bulka99} B.R. Bulka, J. Martinek, G. Michalek, J. Barnas, Phys. Rev. B {\bf 60}, 12246 (1999).
346: \bibitem{Bulka99} B.~R. Bulka \textit{et al.}, Phys. Rev. B {\bf 60}, 12246 (1999).
347: \bibitem{Tserkovnyak01} Y.~Tserkovnyak and A.~Brataas, Phys. Rev. B {\bf 64}, 214402 (2001).
348: \bibitem{Lopez03} R.~Lopez and D.~Sanchez, Phys. Rev. Lett. {\bf 90}, 116602 (2003).
349: \bibitem{Cottet04} A.~Cottet, W.~Belzig, and C.~Bruder, Phys. Rev. Lett. {\bf 92}, 206801 (2004).
350: \bibitem{Egues02} J.~C. Egues, G.~Burkard, and D.~Loss, Phys. Rev. Lett. {\bf 89}, 176401 (2002).
351: %\bibitem{Nowak04} L.~Jiang, E.R. Nowak, P.E. Scott, J. Johnson, J.M. Slaughter, J.J. Sun, R. W. Dave, Phys. Rev. B {\bf 69}, 054407 (2004).
352: \bibitem{Nowak04} L.~Jiang \textit{et al.}, Phys. Rev. B {\bf 69}, 054407 (2004).
353: %\bibitem{Nowak04II}L. Jiang, J.F. Skovholt, E.R.Nowak, J.M Slaughter, Proceedings of SPIE, {\bf 5469}, 13 (2004)
354: \bibitem{Nowak04II}L.~Jiang \textit{et al.}, Proceedings of SPIE {\bf 5469}, 13 (2004)
355: \bibitem{Nowak99} E.~R. Nowak, M.~B. Weissman, and S.~S.~P. Parkin, Appl. Phys. Lett. {\bf 74}, 600 (1999)
356: %\bibitem{shotnoise02} P.K. George, Y. Wu, R.M. White, E. Murdock, M. Tondra, Appl. Phys. Lett. {\bf 80}, 682 (2002)
357: \bibitem{Jansen00} R.~Jansen and J.~S. Moodera, Phys. Rev. B {\bf 61}, 9047 (2000).
358: %\bibitem{Guerrero05} R.~Guerrero, F. G. Aliev, R. Villar, J. Hauch, M. Fraune, G. G{\"{u}}ntherodt, K. Rott, H. Br{\"{u}}ckl, and G. Reiss, Appl. Phys. Lett. {\bf 87}, 042501 (2005).
359: \bibitem{Guerrero05} R.~Guerrero \textit{et al.}, Appl. Phys. Lett. {\bf 87}, 042501 (2005).
360: \bibitem{brinkman} W.~F. Brinkman, R.~C. Dynes, and J.~M. Rowell, J. Appl. Phys. {\bf 41}, 1915 (1970).
361: %\bibitem{Guntherodt01} U.~May, K.Samm, H.Kittur, J.Hauch, R.Calarco, U.R{\"{u}}digier, G.G{\"{u}}ntherodt, Appl. Phys. Lett. {\bf 78}, 2026 (2001).
362: \bibitem{ZangLevyZBA} S.~Zhang \textit{et al.}, Phys. Rev. Lett. {\bf 79}, 3744 (1997);
363: %\bibitem{MooderaZBA} J.~S. Moodera, J.~Nowak, R.~J.~M. van~de~Veerdonck, Phys. Rev. Lett. {\bf 80}, 2941 (1998).
364: J.~S. Moodera \textit{et al.}, \textit{ibid.} {\bf 80}, 2941 (1998).
365: \bibitem{JAP98} J.~Zhang and R.~White, J. Appl. Phys. {\bf 83}, 6512 (1998); L.~Sheng, D.~Y. Xing, and D.~N. Sheng, Phys. Rev. B {\bf 70}, 094416 (2004).
366: %\bibitem{KondoIV} S.~Bermon, D.~E. Paraskevopoulos, and P.~M. Tedrow, Phys. Rev. B {\bf 17}, 2120 ( 1978); J.~Martinek {\em at al.}, Phys. Rev. Lett. {\bf 91}, 127203 (2003); S.-Y. Bae and S.~X. Wang, IEEE Trans. Mag. {\bf 38}, 2721 (2002).
367: %\bibitem{Egues94} J.~C. Egues, S.~Hershfield, and J.~W. Wilkins, Phys. Rev. B {\bf 49}, 13517 (1994).
368: \bibitem{shotnoise02} P.~K. George \textit{et al.}, Appl. Phys. Lett. {\bf 80}, 682 (2002).
369: \bibitem{SNthermometer} L.~Spietz \textit{et al.}, Science, {\bf 300}, 5627 (2003).
370: %\bibitem{Iann03} G.~Iannaccone, F. Crupi, B. Neri, and S. Lombardo, IEEE Trans. El. Dev. {\bf 50}, 1363 (2003).
371: \bibitem{Iann03} G.~Iannaccone \textit{et al.}, IEEE Trans. El. Dev. {\bf 50}, 1363 (2003).
372: \bibitem{Sloncz05} J.~C. Slonczewski, Phys. Rev. B {\bf 71}, 024411 (2005).
373: \bibitem{Nazarov} Y.~V. Nazarov and J.~J.~R. Struben, Phys. Rev. B {\bf 53}, 15466 (1996).
374: \bibitem{Xu} Y.~Xu, D.~Ephron, and M.~R. Beasley, Phys. Rev. B {\bf 52} 2843 (1995).
375: %\bibitem{Tsymbal} Private communication.
376: \end{thebibliography}
377: \end{document}
378: