1: %\listfiles
2: \documentclass[twocolumn,amssymb,amsmath,amsfonts,showpacs,prb,floatfix]{revtex4}
3: %\documentclass[preprint,tightenlines,ttamssymb,amsmath,amsfonts,showkeys,showpacs,prb]{revtex4}
4:
5: \voffset2cm % for printer on fourth floor
6:
7: %\usepackage{longtable}%
8: \usepackage{dcolumn}%
9: \usepackage{bm}%
10: \usepackage{graphicx}%
11:
12: \def\Hcr{H_{\rm cr}} % coercive force of remanence
13:
14: \def\dH{\Delta H(M,\Delta M)} % Bergers method
15: \def\Neff{N_{\rm eff}} % effective demagnetization factor
16:
17: \def\Hr{H_{\rm r}} % externally applied reversal field
18: \def\Ha{H_{\rm a}} % externally applied field
19: \def\Hi{H_{\rm i}} % internal effective field
20: \def\Hc{H_{\rm c}} % microcoercivity in forc distribution
21: % and macroscopic coercive force of ensemble
22: \def\Hb{H_{\rm b}} % bias field in forc distribution
23: \def\Heff{H_{\rm eff}} % effective applied field
24: \def\Hk{H_{\rm K}} % microcoercivity (2 K/Ms)
25: \def\Hs{H_{\rm s}} % switching field
26:
27: \def\ha{h_{\rm a}} % externally applied field, normalized to Hk
28: \def\hr{h_{\rm r}} % same with Hr
29: \def\hb{h_{\rm b}} % same with Hb
30: \def\hc{h_{\rm b}} % same with Hc
31: \def\heff{h_{\rm eff}} % same with Heff
32: \def\hex{h_{\rm ex}} % same with Heff
33: \def\hs{h_{\rm s}} % switching field
34: \def\hk{h_{\rm K}} % microcoercivity (2 K/Ms)
35:
36: \def\Ku{K_{\rm u}} % uniaxial anisotropy constant
37: \def\Ms{M_{\rm sat}} % magnetization of saturation
38: \def\m0{\mu_0} %
39: \def\M2s{\Ms^2} % square of magnetization of saturation
40: \def\KuCo{K^{\rm Co}_{\rm u}} % uniaxial anisotropy constant of Cobalt
41: \def\MsCo{M^{\rm Co}_{\rm s}} % magnetization of saturation of Cobalt
42:
43: \def\Masc{M^-} % ascending (lower) branch of hysteresis loop
44: \def\Mdes{M^+} % descending (upper) branch
45: \def\Mrev{M^{\lrcorner}} % recoil curve starting from descending branch
46:
47: \def\Dsi{D (H_{\rm s})} % intrinsic switching field distribution
48: \def\Dse{D^{\ast}_{\rm e} (H_{\rm s}} % extracted switching field distribution
49: %\def\DseFORC{D^{\ast}_{\rm FORC}(\Hc)} % extracted switching field distribution from FORC distributions
50: %\def\DseDCD{D^{\ast}_{\rm DCD}(H_{\rm s})} % extracted switching field distribution from DCD curve
51: %\def\DseDCDa{D^{\ast}_{\rm DCD}(H_{\rm s},\alpha)} % extracted switching field distribution from DCD curve with mean-field parameter alpha
52:
53: \def\DseFORC{D_{\rm FORC}(\Hc)} % extracted switching field distribution from FORC distributions
54: \def\DseDCD{D_{\rm DCD}(H_{\rm s})} % extracted switching field distribution from DCD curve
55: \def\DseDCDa{D_{\rm DCD}(H_{\rm s},\alpha)} % extracted switching field distribution from DCD curve with mean-field parameter alpha
56:
57: \def\sest{\sigma^{\ast}} % extracted sigma
58: \def\sestFORC{\sigma^{\ast}_{\rm FORC}} % extracted sigma from FORC distributions
59: \def\sestDCD{\sigma^{\ast}_{\rm DCD}} % extracted sigma from DCD curve
60: \def\sestNeff{\sigma^{\ast}_{\Neff}} % extracted sigma from DCD curve
61: \def\sestDCDN{\sigma^{\ast}_{\Delta H}} % extracted sigma from DCD curve with mean-field parameter alpha
62:
63: \def\komm#1{} %% erscheint nicht im Dokument.
64:
65: \def\nummer#1{\refstepcounter{equation}\label{#1} \eqno(\theequation)}
66: \def\kref#1{(\ref{#1})}
67:
68: \begin{document}
69: \preprint{APL}
70:
71: \title{Extracting the intrinsic switching field distribution in perpendicular media: a comparative analysis}
72:
73: \author{Michael Winklhofer}%
74: \surname{Winklhofer} \email{MICHAELW@LMU.DE}
75: \altaffiliation[Also at~]{Department of Earth and Environmental Science, Ludwig-Maximilians-Universit\"at, Theresienstr. 41, D-80333 M\"unchen, Germany}
76: %\corauth[cor1]{Tel.: +49-89-2180-4207; Fax: +44-89-2180-4205}
77: \homepage{http://www.geophysik.uni-muenchen.de/michael}
78: \author{Gergely T. Zimanyi}%
79: \surname{Zimanyi}
80: %\email{zimanyi@physics.ucdavis.edu}
81: \affiliation{Department of Physics and Astronomy, UC Davis, One Shields Avenue, Davis, CA 95616, USA}
82:
83: \begin{abstract}
84: We introduce a new method based on the first-order-reversal-curve
85: (FORC) diagram to extract the intrinsic (microscopic)
86: switching-field distribution (SFD) of perpendicular recording
87: media (PRM). To demonstrate the viability of the method, we
88: micromagnetically simulated FORCs for PRM with known SFD and
89: compare the extracted SFD with the SFD obtained by means of two
90: different methods that are based on recoil loops, too, which
91: however rely on mean-field approximations and assumptions on the
92: shape of the SFD. The FORC method turns out to be the most
93: accurate algorithm over a broad range of dipolar interaction
94: strengths, where the other methods overestimate the width of the
95: SFD. \komm{The best match between the extracted and intrinsic
96: $\Dsi$ can be obtained in the case of $\pm45\deg$-tilted PM. Here,
97: there is some power in the 45 degrees lines from the center of the
98: distribution towards the reversible axis, but that can easily be
99: corrected for, for example, by means of SORCs. Also ( dM(Hr,
100: Ha)/Hr) gives an indication where the distribution 'really' is).
101: Here, meanfield correction works because the curvilinearity of the
102: loops is conserved. In the perpendicular plane, however, we cannot
103: undo the curvilinear hysterons into square hysterons ... well,
104: unless we do a vector Preisach model, maybe }
105:
106:
107: \end{abstract}
108:
109: %\keywords{nanostructured materials; perpendicular magnetic recording; micromagnetic simulation; magnetic hysteresis; FORC }
110:
111: % PACS codes here, in the form: \PACS code \sep code
112: \pacs{85.70.Ay,75.50.Ss,75.50.Tt,75.60.-d}
113: %2001
114:
115: \maketitle
116:
117: \section{INTRODUCTION}
118:
119: The quality of recording media depends crucially on the intrinsic
120: (microscopic) switching-field distribution (SFD) of the media
121: particles, which determines both magnetic stability and attainable
122: recording density. It is straightforward to obtain the SFD of a
123: diluted magnetic system by taking the derivative of the DC
124: demagnetization (DCD) curve. In the case of high-density magnetic
125: recording media, where magnetic interactions between the media
126: particles are not negligible, the problem of extracting the SFD
127: from bulk magnetization curves cannot be solved rigorously any
128: more and the shape (and to a lesser extent the location) of the
129: extracted SFD will depend on certain model assumptions.
130:
131: Two conceptually different algorithms \cite{Veerdonk:03,Berger:05}
132: have been suggested recently to extract the intrinsic $\Dsi$ of
133: perpendicular recording media (PRM) from macroscopic magnetization
134: curves using a set of recoil loops. The analysis technique by
135: \citet{Veerdonk:03} assumes a constant effective demagnetization
136: factor $\Neff$ for deshearing recoil loops, from which the DCD
137: curve is extracted. Extraction of the DCD curve and deshearing the
138: recoil loops are performed simultaneously in order to arrive at a
139: self-consistent solution\cite{Veerdonk:03}. Although the algorithm
140: converges after a couple of iterations, it is not clear how
141: reliable the self-consistent solution may be under the assumption
142: of an effective demagnetization factor independent of the
143: magnetization. Another drawback of the method is that it requires
144: assumptions about the shape of the SFD, which of course is not
145: known.
146:
147: The second algorithm, referred to as $\dH$-method
148: \cite{Berger:05}, is a generalization of the $\Delta H$ method
149: originally proposed by \citet{TagawaNakamura:91}. The $\dH$-method
150: overcomes the restriction of a constant value of $\Neff$, assuming
151: an effective field of the form $\Heff(\Ha,M)=\Ha+\Hi(M)$, where
152: $\Ha$ and $\Hi$ denote the applied and internal field,
153: respectively, and $M$ is the magnetization in the direction of
154: $\Ha$. The $\dH$-method approximates interactions on the
155: mean-field level in a sense that all microscopic magnetization
156: configurations $\mathbf{M}(\mathbf{r})$ representing the same
157: macroscopic magnetization value $M$ produce the same
158: volume-averaged internal field $\Hi(M)$. An implicit assumption
159: underlying the $\dH$-method is that each particle acts a square
160: hysteron. Then, the lower branch of the major loop $M^-(\Ha)$ can
161: be represented as:
162: \begin{equation}
163: \Masc(\Ha) = - \Ms + 2 \int_{-\infty}^{\Heff(\Ha,M)} \Dsi d\Hs \,,
164: \label{Masc}
165: \end{equation}
166: where $\Dsi$ is the SFD. Any recoil loop $\Mrev(\Hr, \Ha>\Hr)$
167: originating from the upper branch $\Mdes$ at $\Hr$ can be written
168: as
169: \begin{equation}
170: \Mrev(\Hr, \Ha) = \Mdes(\Hr) + 2 \int_{-\infty}^{\Heff(\Ha,M) }
171: \Dsi d\Hs \,. \label{Mrecoil}
172: \end{equation}
173: In order for Eq.~\ref{Mrecoil} to hold, two conditions must be
174: met: firstly, $\Mrev(\Hr,\Ha>\Hr)$ has to saturate at fields
175: $|\Ha|\leq|\Hr|$ and secondly, recoil loops must not cross each
176: other. Then the coercivity distribution can be assumed to consist
177: of disjunct segments $\Dsi d\Hs$ (non-interacting hysterons). The
178: inverse $I^{-1}$ of the cumulative distribution
179: \begin{equation}
180: I (H) = \int_{-\infty}^{H } \Dsi d\Hs \, \label{cumdist}
181: \end{equation}
182: % Eq. \ref{dHeq} from Eqs. \ref{Masc}-\ref{Mrecoil}
183: for a given magnetization value $M$ can then be obtained by taking
184: the difference in field position, $\Delta H (M)$, between
185: $M^-(\Ha)$ and recoil loop $\Mrev(\Hr, \Ha>\Hr)$,
186: \begin{eqnarray}
187: \dH \equiv \Ha(M^-) &-&\Ha(\Mrev) = \label{dHeq} \\
188: = I^{-1}\left[\frac{\Ms + M }{2}\right]
189: &-& I^{-1}\left[\frac{\Ms + (M -\Delta M) }{2}\right]
190: \,, \nonumber
191: \end{eqnarray}
192: where $\Delta M=\Mdes(\Hr) - (-\Ms) $. By fitting the $\dH$ curves
193: against the inverse of a certain parameterized distribution
194: function, the key features of the SFD can be extracted
195: \cite{Berger:05}. In case $I$ is a normal distribution centered at
196: $\Hc$, Eq.~\ref{dHeq} writes to
197: \begin{equation}
198: \frac{\dH}{\Hc} = \sqrt{2}\sigma\left[{\rm erf}^{-1}(m) - {\rm
199: erf}^{-1}(m-\Delta m) \right] \label{GaussfitdH} \,,
200: \end{equation}
201: where lower case $m$'s are the magnetization values relative to
202: the saturation magnetization $\Ms$.
203:
204:
205: The third method in our comparative analysis is based on the
206: first-order-reversal curve (FORC)
207: distribution\cite{Mayergoyz:85,Pike:99}, defined as
208: \begin{equation}
209: \rho(\Hr,\Ha) = - \frac{1}{2} \frac{\partial^2
210: M(\Hr,\Ha)}{\partial \Hr \partial \Ha} \,. \label{defFORC}
211: \end{equation}
212: Although there is no principal difference between a FORC and a
213: recoil loop, the numerical determination\cite{Pike:99} of $\rho$
214: from a set of FORCs requires recoil loops measured on a grid
215: equidistant in $\Hr$ and $\Ha$.
216:
217: For an assemblage of square hysterons, the FORC distribution is
218: identical to their Preisach distribution and we can obtain their
219: SFD as
220: \begin{equation}
221: \DseFORC = \int_{-\infty}^{\infty} \rho(\Hc,\Hb) d\Hb \,,
222: \label{DHsFORC}
223: \end{equation}
224: where $\Hc=(\Ha-\Hr)/2$ and $\Hb=(\Ha+\Hr)/2$ are the coordinates
225: of the Preisach plane, defined by the microscopic switching field
226: $\Hc$ and the bias field $\Hb$. As with the $\dH$ method, the FORC
227: method is reliable as long as the media particles can be
228: reasonably well described by square hysterons. As opposed to the
229: two algorithms above, however, the method based on
230: Eq.~\ref{DHsFORC} relies neither on a mean-field approximation nor
231: on a certain model form for $\Dsi$ and therefore imposes the
232: fewest constraints on the data.
233:
234:
235: \section{COMPUTATIONAL DETAILS}
236:
237: The three methods presented above are best tested on a system with
238: known SFD. For this purpose, we micromagnetically computed sets of
239: FORCs for a PRM, using the OOMMF code\cite{OOMMF} (v. 1.1b2). The
240: simulated medium typically consisted of $N\sim2\cdot 10^3$
241: particles, arranged on a regular grid (mesh size $s=5$~nm),
242: inscribed in a circle of 250~nm diameter. We chose a circular
243: boundary to minimize the effects of corners. The easy axes of the
244: particles are all oriented the same way, roughly perpendicular to
245: the surface ($\theta_K=89.42\deg$) so as to avoid numerical
246: problems that may arise when the applied magnetic field $H$
247: ($\theta_H=90\deg$) is exactly collinear with the easy direction.
248: We modified the module {\tt maganis.cc} to produce normal
249: distributed values of the uniaxial magnetocrystalline anisotropy
250: constant, $\Ku$ such that $\sigma_{\Hk}=0.1~\langle\Hk\rangle$,
251: with $\Hk=2~\Ku/\mu_0\Ms$. It has been shown
252: \cite{ZhouBertramSchabes:02} that a small amount of intergranular
253: exchange helps to reduce the increase in the recording transition
254: parameter due to a small distribution in $\Hk$. We therefore set
255: the nondimensional intergranular exchange coupling constant
256: $\hex=A/(\Ku\,s^2)$ to $\hex=0.046$ in all our simulations. In
257: order to systematically explore the effects of dipolar
258: interactions, we vary the dipolar interaction strength by using a
259: scaling factor $\phi<1$ for $\Ms$, where $\phi$ represents the
260: volume fraction of magnetic material in each cell. The uniaxial
261: magnetocrystalline anisotropy constant, $\Ku$, is scaled
262: simultaneously to keep the value of the microscopic coercivity
263: $\Hk$ constant as $\phi$ varies.\komm{A less efficient way of
264: decreasing the interaction strength in FFT-based codes would be to
265: increase the spacing between adjacent cells by introducing
266: nonmagnetic cells in between. Such an
267: approach%\cite{MuxworthyWilliams:2005}
268: not only inflates the computation time and memory, but also limits
269: the range of possibles value of $\phi$.}
270: % FFT requires a regular grid.
271: %Sticking to the nn-grid allows us to include exchange, too!
272: We normalize all magnetic fields by $\Hk$ and use a mean-field
273: type equation to define the material-independent dipolar
274: interaction strength as $\phi_Q =\phi/Q$, i.e.,
275: \begin{equation}
276: \heff = \ha + \alpha \frac{\phi}{Q}\frac{M}{\Ms} \equiv \ha +
277: \alpha\, \phi_Q m \,, \label{mfscaled}
278: \end{equation}
279: where $\alpha$ is the mean-field parameter and $Q$ denotes the
280: quality factor $Q=2\Ku/(\mu_0\M2s)$. For hcp Cobalt at room
281: temperature, $Q=0.422$. %with $\MsCo=1400$~A/m and $\KuCo=520~\mathrm{J/m}^3$.
282: \komm{This choice of parameters and anisotropy direction yields
283: $694$~mT as a mean value of the intrinsic switching field
284: (compared to $\m0\Hk=743$~mT for the exactly collinear case).}
285:
286:
287: \section{RESULTS AND DISCUSSION}
288:
289: \begin{figure}
290: %\includegraphics[width=15cm]{Figure1.eps}
291: \includegraphics[width=8.5cm]{Figure1.eps}
292: %\includegraphics[width=8.5cm]{Figure1largefont.eps}
293: \caption{\label{compareEstSigma}%(Color online).
294: Comparison of the three methods in terms of the extracted $\sest$
295: of the SFD as a function of the dipolar interaction strength
296: $\phi_Q$. The relative strength of the exchange coupling decreases
297: from left to right ($\hex=0.046$ for all values of $\phi_Q$). The
298: intrinsic SFD is a Gaussian distribution with $\sigma=0.1$ (thin
299: horizontal line). The coercive force is represented by the grey
300: line without plot symbols ($\Hc$ scale on the right).\komm{For
301: $\phi_Q>0.1$, the $\Neff$ method\cite{Veerdonk:03} yielded a
302: better fit when a lognormal distribution was assumed and therefore
303: the $\sigma$ corresponding to the best fitting lognormal
304: distribution is plotted here. Symbols: squares $\Neff$ method;
305: circles $\dH$ method; triangles FORC.} }\end{figure}
306:
307: Figure~\ref{compareEstSigma} shows the estimated width $\sest$ of
308: the SFD as extracted with the three algorithms tested here. In the
309: hypothetical limit case of a PRM controlled solely by
310: exchange-coupling ($\phi_Q\leq 0$), the $\dH$ method is able to
311: capture the intrinsic width of the SFD with great accuracy. For
312: PRM largely dominated by dipolar interactions, however, the $\dH$
313: method is consistently less accurate than the FORC method, which
314: renders the least deviation of $\sest$ from $\sigma$ for $\phi_Q
315: > 0.15$. In contrast, the $\sest$ extracted with the $\Neff$ method always
316: show the largest deviations from $\sigma$. Importantly, the
317: $\Neff$ method does not yield self-consistent solutions any more
318: for $\phi_Q>0.25$, where the asymmetry of the DCD curve is too
319: pronounced to be properly described by a normal distribution
320: (Fig.~\ref{NeffvsFORC}). More consistent solutions in this regime
321: can be found with a log-normal SFD cut-off at high fields. Caution
322: should therefore be taken to not interpret asymmetric DCD curves
323: prematurely as evidence of an asymmetric intrinsic SFD. The FORC
324: method on the other hand captures the intrinsic shape of the SFD
325: to a very good degree despite the shift in location
326: (Fig.~\ref{NeffvsFORC}).
327:
328:
329: \begin{figure}
330: \includegraphics[width=8.5cm]{./CoSc015A003.vsmForcSFDCol.eps}\\[1mm]
331: %\includegraphics[width=8.5cm]{./CoSc015A003.vsmForcSFDBW.eps}\\[1mm]
332: \caption{\label{NeffvsFORC} (Color online). FORC diagram
333: $\rho(\Hc,\Hb)$ (above) obtained for a PRM with $\phi_Q=0.35,
334: \hex=0.046$ and the corresponding SFD extracted on the basis of
335: Eq.~\ref{DHsFORC} (below). The width $\sest=0.107$ is determined
336: from fitting the extracted SFD (solid line) with a normal
337: distribution centered at $\Hc=0.76\,\Hk$ (dashed). The derivative
338: of the DCD curve (dash-dotted) has a distinct asymmetry. The
339: intrinsic SFD (dotted) with $\sigma=0.1$ is shown for comparison.
340: All curves are normalized to unit area. With the $\dH$ method,
341: $\sest$ was obtained as 0.133 (original recoil loops and $\dH$
342: curves available online).}\end{figure}
343: % Berger mean-fit deviation: 1.8%, FORC: > 1%.
344:
345: All the three methods are primarily concerned with finding the
346: right scale parameter (i.e., $\sest$ or FWHM) of the SFD
347: distribution, while the location parameter is considered to be
348: invariant. As can be seen in Figure~\ref{compareEstSigma}, the
349: observed values of $\Hc$ are progressively shifted to lower values
350: with increasing $\phi_Q$, which is due to dipolar interactions
351: between the media particles, deflecting the effective field away
352: from the easy axes. According to the Stoner-Wohlfarth
353: relationship,
354: \begin{equation}
355: \Hs (\psi) = \Hk \, \left((\cos\psi)^{2/3}
356: +(\sin\psi)^{2/3}\right)^{-3/2}
357: \end{equation}
358: observed coercive forces of $\sim 0.8 \Hk$ suggest that a large
359: fraction of grains experience a local effective field that
360: deviates by some 5\% from the applied field direction, giving rise
361: to curvilinear hysterons, in other words, reversible magnetization
362: processes. In the FORC diagram (Fig.~\ref{NeffvsFORC}),
363: curvilinear hysterons manifest themselves in the form of the small
364: negative region centered at $(\Hc,\Hb)=(1,-0.05)\Hk$. This way,
365: one can determine directly from the FORC diagram if and to what
366: degree the model assumptions are not strictly met.
367:
368: Since all methods tested here rely on square hysterons, their
369: underlying assumption starts breaking down with increasing dipolar
370: coupling $\phi_Q$, albeit at different values of $\phi_Q$. More
371: importantly, the fact that the methods do not all start to fail at
372: the same point also shows that the presence of reversible
373: magnetization processes is not the most crucial limitation to a
374: method's applicability, which rather is restricted by imposing
375: constraints on the internal field distribution. Compared to the
376: $\Neff$ method, the $\dH$ method puts a less tight constraint on
377: the expected magnetization curves and so has a larger range of
378: applicability. The very absence of any such approximations makes
379: the method based on the FORC distribution the most robust
380: algorithm and therefore most suitable for characterizing
381: ultra-high-density PRM.
382:
383: %
384: %\begin{acknowledgments}
385: We enjoyed discussion with Chris Pike, Kai Liu and Rene van der
386: Veerdonk. This work was funded by the Campus-Laboratory Exchange
387: Program (UC).
388: %\end{acknowledgments}
389: %\bibliographystyle{unsrtabnat}
390:
391: \begin{thebibliography}{7}
392: \expandafter\ifx\csname
393: natexlab\endcsname\relax\def\natexlab#1{#1}\fi
394: \expandafter\ifx\csname bibnamefont\endcsname\relax
395: \def\bibnamefont#1{#1}\fi
396: \expandafter\ifx\csname bibfnamefont\endcsname\relax
397: \def\bibfnamefont#1{#1}\fi
398: \expandafter\ifx\csname citenamefont\endcsname\relax
399: \def\citenamefont#1{#1}\fi
400: \expandafter\ifx\csname url\endcsname\relax
401: \def\url#1{\texttt{#1}}\fi
402: \expandafter\ifx\csname
403: urlprefix\endcsname\relax\def\urlprefix{URL }\fi
404: \providecommand{\bibinfo}[2]{#2}
405: \providecommand{\eprint}[2][]{\url{#2}}
406:
407: \bibitem[{\citenamefont{van~de Veerdonk et~al.}(2003)\citenamefont{van~de
408: Veerdonk, Wu, and Weller}}]{Veerdonk:03}
409: \bibinfo{author}{\bibfnamefont{R.~J.~M.} \bibnamefont{van~de Veerdonk}},
410: \bibinfo{author}{\bibfnamefont{X.~W.} \bibnamefont{Wu}}, \bibnamefont{and}
411: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Weller}},
412: \bibinfo{journal}{IEEE Trans. Magn.} \textbf{\bibinfo{volume}{39}},
413: \bibinfo{pages}{590} (\bibinfo{year}{2003}).
414:
415: \bibitem[{\citenamefont{Berger et~al.}(2005 (in print))\citenamefont{Berger,
416: Xu, Lengsfield, Ikeda, and Fullerton}}]{Berger:05}
417: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Berger}},
418: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Xu}},
419: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Lengsfield}},
420: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Ikeda}}, \bibnamefont{and}
421: \bibinfo{author}{\bibfnamefont{E.~E.} \bibnamefont{Fullerton}},
422: \bibinfo{journal}{IEEE Trans. Magn.} (\bibinfo{year}{2005 (in print)}).
423:
424: \bibitem[{\citenamefont{Tagawa and Nakamura}(1991)}]{TagawaNakamura:91}
425: \bibinfo{author}{\bibfnamefont{I.}~\bibnamefont{Tagawa}} \bibnamefont{and}
426: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Nakamura}},
427: \bibinfo{journal}{IEEE Trans. Magn.} \textbf{\bibinfo{volume}{27}},
428: \bibinfo{pages}{4975} (\bibinfo{year}{1991}).
429:
430: \bibitem[{\citenamefont{Pike et~al.}(1999)\citenamefont{Pike, Roberts, and
431: Verosub}}]{Pike:99}
432: \bibinfo{author}{\bibfnamefont{C.~R.} \bibnamefont{Pike}},
433: \bibinfo{author}{\bibfnamefont{A.~P.} \bibnamefont{Roberts}},
434: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{K.~L.}
435: \bibnamefont{Verosub}}, \bibinfo{journal}{J. Appl. Phys.}
436: \textbf{\bibinfo{volume}{85}}, \bibinfo{pages}{6660} (\bibinfo{year}{1999}).
437:
438: \bibitem[{\citenamefont{Mayergoyz}(1985)}]{Mayergoyz:85}
439: \bibinfo{author}{\bibfnamefont{I.~D.} \bibnamefont{Mayergoyz}},
440: \bibinfo{journal}{J. Appl. Phys.} \textbf{\bibinfo{volume}{57}},
441: \bibinfo{pages}{3803} (\bibinfo{year}{1985}).
442:
443: \bibitem[{\citenamefont{Donahue and Porter}(Sept. 1999)}]{OOMMF}
444: \bibinfo{author}{\bibfnamefont{M.~J.} \bibnamefont{Donahue}} \bibnamefont{and}
445: \bibinfo{author}{\bibfnamefont{D.~G.} \bibnamefont{Porter}},
446: \bibinfo{type}{Interagency Report} \bibinfo{number}{NISTIR 6376},
447: \bibinfo{institution}{National Institute of Standards and Technology,
448: Gaithersburg, MD} (\bibinfo{year}{Sept. 1999}).
449:
450: \bibitem[{\citenamefont{Zhou et~al.}(2002)\citenamefont{Zhou, Bertram, and
451: Schabes}}]{ZhouBertramSchabes:02}
452: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Zhou}},
453: \bibinfo{author}{\bibfnamefont{H.~N.} \bibnamefont{Bertram}},
454: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{M.~E.}
455: \bibnamefont{Schabes}}, \bibinfo{journal}{IEEE Trans. Magn.}
456: \textbf{\bibinfo{volume}{38}}, \bibinfo{pages}{1422} (\bibinfo{year}{2002}).
457:
458: \end{thebibliography}
459:
460: \newpage
461:
462: \section*{Online Supplements}
463:
464:
465: \vbox{\includegraphics[width=8.5cm]{./Online_raw.eps}\\[1mm]
466: Micromagnetically computed FORCs for $\phi_Q=0.35$.}
467:
468: \vspace{0.4cm}
469:
470: \vbox{
471: \includegraphics[width=8.5cm]{./Online_dHmisfit.eps}\\[1mm]
472: Average deviation of expected $\dH$ curves (according to
473: Eq.~\ref{GaussfitdH}) from $\dH$ dataset as a function of
474: $\sigma$. The best fit is obtained for $\sigma_{FIT} =0.133$,
475: which defines $\sest$.}
476:
477: \vspace{0.4cm}
478:
479: \vbox{
480: \includegraphics[width=8cm]{./Online_dHplot.eps}\\[1mm]
481: Comparison between expected $\dH$ curves (solid lines) for
482: $\sest=0.133$ and $\dH$ dataset (dots). Not all $\dH$ curves are
483: shown to keep the picture tidy.}
484:
485:
486: \end{document}
487: