0805.3454/ms.tex
1: %-----------------------------------------------------------------------
2: % 
3: %\documentclass[referee]{aa} % for a referee version
4: %
5: \documentclass[twocolumn]{aa}
6: %\documentclass[onecolumn]{aa}
7: \usepackage{graphicx}
8: \usepackage{txfonts}
9: \usepackage{natbib}
10: \usepackage{slashbox}
11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
12: %
13: \bibliographystyle{aa}
14: \def\etal{et al.}
15: \def\aa{A\&A }                  % A&A
16: \def\aas{A\&AS }                % A&AS
17: \def\apj{ApJ }                  % ApJ
18: \def\apjs{ApJS }                % ApJS
19: \def\aj{AJ }                    % AJ
20: \def\mnras{MNRAS }              % MNRAS
21: \def\aap{AAP }                  % AAP
22: \def\araa{ARA\&A }              % ARA&A
23: \def\jrasc{JRASC }              % JRASC
24: \def\rvmp{RvMP }                % Rev of Modern Phys
25: \def\aar{Astron. Astrophys. rev. }
26: \def\physrep{Phys. Rep. }           % Phys. Rep.
27: 
28: \begin{document}
29: %
30:    \title{Observational constraints on models for the
31: interstellar magnetic field in the Galactic disk}
32: 
33:    \author{H. Men
34: 		  \inst{1, 2}
35: 		  , K. Ferri\`ere\inst{2}
36: 	  \and
37: 	  J. L. Han\inst{1}
38: 		  }
39: 
40:    \offprints{K. Ferri\`ere, ferriere@ast.obs-mip.fr}
41: 
42:    \institute{National Astronomical Observatories, Chinese Academy of Sciences,
43: 			  Jia 20 DaTun Road, Chaoyang District, Beijing, 100012, China\\
44:    \and Observatoire Midi-Pyr\'en\'ees, Universit\'e Paul Sabatier Toulouse 3,
45: 		CNRS, 14 av. Ed. Belin, 31400 Toulouse, France\\
46: 			 }
47: 
48:    \date{Received  / Accepted }
49: 
50: \titlerunning{Observational constraints on models for the Galactic magnetic field}
51: \authorrunning{H. MEN et al.}
52: 
53: \abstract
54: {}
55: {Our purpose is to place firm observational constraints on the three
56:   most widely used theoretical models for the spatial configuration
57:   of the large-scale interstellar magnetic field in the Galactic disk,
58:   namely, the ring, the axisymmetric and the bisymmetric field models.
59: }
60: {We use the rotation measures (RMs) of low-latitude Galactic
61:   pulsars and combine them with their dispersion measures and
62:   estimated distances to map out the line-of-sight component of the
63:   interstellar magnetic field in the near half of the Galactic disk.
64:   We then fit our map of the line-of-sight field to the three aforementioned
65:   theoretical field models and discuss the acceptability of each fit,
66:   in order to determine whether the considered field model is allowed
67:   by the pulsar data or not.
68: }
69: {Strictly speaking, we find that all three field models are ruled out
70: by the pulsar data. Furthermore, none of them appears to perform
71: significantly better than the others.
72: From this we conclude that the large-scale interstellar magnetic field 
73: in the Galactic disk has a more complex pattern than just circular, 
74: axisymmetric or bisymmetric.
75: }
76: {}
77: 
78: \keywords{ISM: magnetic fields -- Galaxy: disk -- Galaxies: magnetic fields
79: }
80: 
81: \maketitle
82: 
83: 
84: \section{Introduction}
85: 
86:   The interstellar magnetic field of our Galaxy has been the object of
87:   intense investigation since the early 1980s. Different observational
88:   methods (e.g., based on synchrotron emission, Faraday rotation,
89:   Zeeman splitting, polarization of starlight, polarization of
90:   dust infrared emission) provide information on the magnetic field in
91:   different interstellar regions. Faraday rotation of Galactic pulsars
92:   and extragalactic linearly polarized radio sources make it possible
93:   to directly trace the magnetic field in ionized regions. In
94:   practice, one measures the so-called rotation measure (RM), defined
95:   by
96:    \begin{equation}
97: 	{\rm RM} = 0.81  \int _{0}^{d} n_{\rm e}\ B_{||}\ {\rm d}s \ \ \ {\rm rad \ m^{-2} }\ ,
98:    \end{equation}
99:    where $n_{\rm e}$ is the free-electron density (in ${\rm cm
100:    ^{-3}}$), $B_{||}$ is the line-of-sight component of the magnetic
101:    field (in $\mu{\rm G}$) and $d$ is the distance to the radio
102:    source (in ${\rm pc}$).  Pulsars present a number of advantages
103:    when used as probes of the interstellar magnetic field. In
104:    particular, they are highly linearly polarized, they have no
105:    intrinsic rotation measure and their distances can be estimated
106:    reasonably well. Moreover, the RM of a pulsar divided by its
107:    dispersion measure (DM),
108:    \begin{equation}
109: 	 {\rm DM} = \int _{0}^{d} n_{\rm e}\ {\rm d}s  \ \ \ {\rm cm ^{-3} \ pc}\ ,
110:    \end{equation}
111:   directly yields the $n_{\rm e}$-weighted average value of $B_{||}$
112:   along its line of sight,
113:    \begin{equation}
114: 	 \overline{B_{||}} = 1.232\ \frac{\rm RM}{\rm DM} \ \ \ \mu{\rm G}\ .
115: 	 \label{Eq_B_par_calc}
116:    \end{equation}
117: 
118:    We now know that the interstellar medium (ISM) is highly inhomogeneous
119:    and that the interstellar magnetic field has an important turbulent
120:    component. For this reason, neighboring pulsars may have significantly
121:    different values of RM and DM, and a plot RM {\it versus} DM will
122:    generally exhibit a large scatter.
123:    However, if one considers a Galactic region larger than the scale of
124:    the turbulent field and containing enough pulsars for statistical
125:    purposes, one can infer the large-scale (or regular) component of
126:    $\overline{B_{\parallel}}$ in that region from the slope of the mean
127:    DM-RM relation \citep{RL94}:
128:    \begin{equation}
129: 	 \left< \overline{B_{||}} \right> = 1.232 \ \left <
130: 			 \frac{\rm d\ RM}{\rm d\ DM } \right> \ \ \ \mu{\rm G} \ .
131: 	 \label{Eq_B_par_obs}
132:    \end{equation}
133: 
134:   Various theoretical models have been proposed to describe the spatial
135:   structure of the large-scale magnetic field in the Galaxy.
136:   First and foremost are the ring model, the axisymmetric or axisymmetric
137:   spiral (ASS) model, and the bisymmetric or bisymmetric spiral (BSS) model.
138:   According to the galactic dynamo theory, ASS fields would be easiest
139:   to amplify under typical galactic conditions \citep[e.g.,][]{RSS85,
140:   Ferriere00}, whereas BSS fields could possibly be excited in the presence
141:   of an external disturbance, such as a companion galaxy \citep{moss95,moss96}.
142:   On the other hand, the primordial field theory naturally leads to BSS
143:   fields \citep{Howard97}.
144: 
145:   In principle, RM studies are ideally suited to establish the overall
146:   structure of the Galactic magnetic field.
147:   However, the different RM studies performed so far yield contradictory
148:   results: some favor a ring field \citep{RK89, RL94, vallee05},
149:   others an axisymmetric or ASS field \citep{vallee91, vallee96},
150:   and others a bisymmetric or BSS field \citep{SK80, HanQiao94, ID99, Han06}.
151: %%The work which carefully studied and compared the three models with the
152: %%pulsar data had been done almost ten years ago \citep{HanQiao94, ID99}.
153:   Moreover, although all these studies conclude with a preferred field model,
154:   none of them has seriously considered the possibility that more than
155:   one model is allowed by the RM data or, alternatively, that none of
156:   the three basic models alone can account for the data.
157:   Hence the question we would like to address in this paper:
158:   which among the ring, axisymmetric and bisymmetric models can clearly 
159: be accepted on the grounds that it is consistent with the RM data, 
160: and which model should clearly be rejected on the grounds that it fails
161: to provide a good fit to the data.
162: 
163:   In recent years, numerous pulsars were discovered in the near half
164:   of the Galactic disk and many of them had their RM measured.
165:   At the present time, among the $\sim 1800$ known pulsars,
166:   690 have measured RMs and, among the latter, 524 are located at low
167:   Galactic latitudes ($|b|<10^\circ$). Pulsars with measured RMs now
168:   provide a reasonably good coverage of the near half of the Galactic
169:   disk. Furthermore, pulsar distances can now be estimated with fairly
170:   good accuracy thanks to the improved free-electron density model of
171:   \citet{NE2001} (known as the NE2001 model). The new measurements
172:   enable one to investigate the configuration of the Galactic magnetic
173:   field over a much larger region and with much more confidence than
174:   previously feasible.
175: 
176:   In Sect.~2, we present the three basic theoretical models for the
177:   interstellar magnetic field in the Galactic disk.
178:   In Sect.~3, we describe the procedure used to bin the pulsar data
179:   and to map out the distribution of $B_{||}$.
180:   In Sect.~4, we fit our map of $B_{||}$ to each of the three field models,
181:   and we discuss how good the fits are at reproducing the pulsar data.
182:   In Sect.~5, we summarize our results and conclude our study.
183: 
184: 
185: \section{Description of the field models}
186:    \label{description and requirment}
187: 
188: Throughout this paper, the Galactocentric cylindrical coordinates 
189: are denoted by $(r, \theta, z)$, and the distance from the Galactic 
190: center (GC) to the Sun is set to $r_\odot = 8.5 \ {\rm kpc}$.
191: 
192: In general, the horizontal position of a given pulsar P can be defined either
193: by its distance from the GC, $r$, and its Galactic azimuthal angle 
194: $\theta$ (which increases clockwise from $\theta=0$ along the line 
195: segment GC-Sun), or by its distance from the Sun, $d$, and its Galactic 
196: longitude, $l$ (which increases counterclockwise from $l=0$ along 
197: the line segment Sun-GC).  Another useful angular coordinate is the angle 
198: $\alpha$ between the azimuthal direction at P and the vector P-Sun, 
199: such that $\alpha=\theta+l+\frac{\pi}{2} $ (see Fig.~\ref{Fig.1}).  
200: 
201: % _______________________ Fig 1 ____________________________
202: 
203: 	\begin{figure*}[!htb]
204: 	  \centering
205:      \includegraphics[angle=-90,width=6cm]{8683fig1.ps}
206: 	   \caption{\label{Fig.1}
207: 		Schematics showing the geometrical variables associated with
208: 		a pulsar P (see main text for the exact definitions).}
209: 	\end{figure*}
210: %____________________________________________________________
211: 
212: 
213: 
214: Because Galactic differential rotation efficiently stretches magnetic field
215: lines in the azimuthal direction, $\left<B_{\theta}\right>$ dominates over 
216: both $\left<B_r \right>$ and $\left<B_z \right>$.
217: Moreover, all the pulsars selected for the present work lie at
218: $\left| b \right| < 10 ^{\circ}$ and reside in the Galactic disk. 
219: There, the large-scale magnetic field is nearly horizontal
220: \citep[e.g.,][]{RSS85, Beck96}, so that $|\left<B_z \right>| \ll
221: |\left<B_r \right>|, \ |\left<B_\theta \right>|$ . 
222: In addition, projecting an already small $\left<B_z \right>$ onto 
223: the line of sight to a pulsar further reduces its contribution 
224: by a factor $|\sin b| \ll 1$.
225: Under these conditions, the line-of-sight component of the large-scale 
226: magnetic field depends only on its radial and azimuthal components,
227: and is related to them through
228:         \begin{equation}
229:           \left< B_{||} \right> = \left<B_r\right> \, \sin \alpha
230:                 + \left<B_{\theta}\right> \, \cos \alpha \ .
231:           \label{Eq_B_para}
232:         \end{equation}
233: 
234: We now present the three theoretical field models.
235: 
236: 
237: \subsection{Ring model}
238: 
239: In the ring model, the large-scale magnetic field points everywhere in
240: the azimuthal direction, so that its radial component vanishes:
241:    \begin{equation}
242:      \left<B_r\right>=0 \ .
243:    \label{Eq_ring_Br}
244:    \end{equation}
245: Its azimuthal component is constant along circles, i.e., independent 
246: of $\theta$, but it can vary with $r$\,:
247:    \begin{equation}
248: 	 \left<B_\theta\right> = \left<B_\theta\right>(r) \ ,
249:    \label{Eq_ring_Btheta}
250:    \end{equation}
251: and it can even change sign along a Galactic radius. 
252: As a matter of fact, all RM studies leading to a ring model have found 
253: reversals in $\left<B_\theta\right>$ \citep{RK89, RL94, vallee05}.
254: It should be noted that the ring model constitutes a particular case
255: of the axisymmetric model.
256: 
257: 
258: \subsection{Axisymmetric model}
259: 
260: In the axisymmetric model, $\left<B_r\right>$ and $\left<B_\theta\right>$ 
261: are both independent of $\theta$ and vary only with $r$\,:
262:    \begin{equation}
263: 	 \left<B_r\right> = \left<B_r\right>(r) \ ,
264:    \label{Eq_AS_Br}
265:    \end{equation}
266:    \begin{equation}
267: 	 \left<B_\theta\right> = \left<B_\theta\right>(r) \ .
268:    \label{Eq_AS_Btheta}
269:    \end{equation}
270: Here, too, $\left<B_\theta\right>$ can reverse sign with $r$.
271: Such sign reversals were found in RM studies favoring an ASS
272: magnetic field \citep{vallee91, vallee96}.
273: Interestingly, reversals in $\left<B_\theta\right>$ are also consistent 
274: with dynamo theory, which can produce them under certain conditions, 
275: e.g., when the magneto-ionic disk has a particular shape and thickness 
276: and the seed field itself has strong reversals \citep{PSS93} 
277: or when the Galactic rotation rate decreases not only with radius 
278: but also with height \citep{Ferriere00}.
279: 
280: 
281: \subsection{Bisymmetric model}
282: 
283: In the bisymmetric model, $\left< B_r \right>$ and $\left<B_\theta\right>$
284: have a simple sinusoidal dependence on $\theta$, which can be written
285: in the form
286: 	\citep{Berkhuijsen97}:
287:    \begin{equation}
288: 	  \left<B_r\right>=b_r(r) \ \sin(\theta-\phi(r)) \ ,
289: 	  \label{Eq_BSS_Br}
290: 	\end{equation}
291: 	\begin{equation}
292: 	  \left<B_{\theta}\right>=b_{\theta}(r) \ \sin(\theta-\phi(r)) \ ,
293: 	  \label{Eq_BSS_Btheta}
294: 	\end{equation}
295: 	where $b_r(r)$ and $b_{\theta}(r)$ are the maximum amplitudes of
296: 	$\left<B_r\right>$ and $\left<B_{\theta}\right>$, respectively, 
297:         and $\phi(r)$ is the azimuthal phase.
298: 	Both $\left<B_r\right>$ and $\left<B_\theta\right>$
299:         can reverse sign with $r$.       
300: 	The magnetic pitch angle is defined as
301: 	\begin{equation}
302: 	  \tan p(r)=\frac{\left<B_r\right>}{\left<B_{\theta}\right>}
303: 			   =\frac{b_r(r)}{b_{\theta}(r)} \ ;
304: 	  \label{Eq_BSS_p}
305: 	\end{equation}
306: 	it is positive (negative) if the magnetic field spirals out clockwise
307: 	(counterclockwise) or spirals in counterclockwise (clockwise).
308: 
309: 
310: \section {Mapping of $\left< \overline{B_{||}} \right>$}
311: 
312:    To date, there are 690 pulsars with measured RMs \citep{Hamilton87,
313:    RL94, Qiao95, van97, HMQ99, Crawford01, Mitra03, Weisberg04, Han06, NJKK08}.
314:    Among these pulsars, we selected those that lie at low Galactic
315:    latitudes ($\left| b \right| < 10 ^{\circ}$) and have reliable RMs
316:    (error on ${\rm RM} < 25\ {\rm rad\ m^{-2}}$).
317:    This left us with 482 pulsars.
318:    For the distances and DMs of our selected pulsars, we used the values
319:    given in the ATNF Pulsar Catalog (\citealt{Manchester05},
320:    see http://www.atnf.csiro.au/research/pulsar/psrcat).
321:    Pulsar distances in this catalog were estimated with the help
322:    of Cordes \& Lazio's \citeyearpar{NE2001} NE2001 model for the
323:    free-electron density; for pulsars located in the inner Galaxy,
324:    individual distances are typically uncertain by $\sim20\%$, but
325:    the relative distances of neighboring pulsars have a much lower
326:    uncertainty.  Pulsar DMs, for their part, are known with good accuracy
327:    (error on ${\rm DM}$ generally $< 1\ {\rm cm ^{-3} \ pc}$).
328: 
329:    In order to map out the large-scale component of
330:    $\overline{B_{||}}$, one needs to divide the Galactic disk into
331:    regions (boxes) having sizes intermediate between the large scales
332:    of the regular field and the small scales of the turbulent field
333:    and containing at least a few pulsars each. In previous studies
334:    \citep{RL94, Weisberg04, Han06}, this division was based on a
335:    heliocentric grid defined by circles of constant $d$ and radial
336:    lines of constant $l$. Such a heliocentric division was justified
337:    by the spatial distribution of the available pulsars, but it is
338:    ill-suited to the present work, whose purpose is to test field
339:    models expressed in terms of Galactic radius, $r$. A much more
340:    appropriate division here is one based on a hybrid grid defined by
341:    circles of constant $r$ and lines of constant $l$ (see
342:    Fig.~\ref{Fig.2}).
343: %
344: To make full use of the pulsar data, we consider two different grids.
345: In the first grid, the circles are located at $r = 4\ {\rm kpc},\
346: 5\ {\rm kpc},\ 6\ {\rm kpc},\ 7\ {\rm kpc} {\rm \ and}\ 8\ {\rm kpc}$
347: [i.e., $r=r_i$, with $r_i \equiv i\ {\rm kpc},\ i=4\, ...\, 8$], and
348: the lines of constant $l$ are the lines emanating from the Sun and
349: tangent to one of the circles $r=r_i,\ i=2\, ...\, 7$, plus the line Sun-GC
350: [i.e., $l=l_0,\ l_{\pm i}$, with $l_0 \equiv 0$ and $l_{\pm i} \equiv
351: \pm {\rm asin} \frac{r_i}{r_\odot},\ i=2\, ...\, 7$]
352: (see Fig.~\ref{Fig.2}a).
353: The second grid is defined in an analogous manner with the circles shifted
354: by 0.5~kpc [i.e., $r=r_i,\ i=3.5,\, 4.5\, ...\, 8.5$,
355: and accordingly, $l=l_0,\ l_{\pm i},\ i=1.5,\, 2.5\, ...\, 7.5$]
356: (see Fig.~\ref{Fig.2}b).
357: To ensure a sufficient number of pulsars per box, some of the boxes
358: defined by these grids are paired together.
359: More specifically, the non-outermost boxes along each ring are paired
360: either with their left or right neighbor along the same ring (thereby
361: leading to a single double-size box) or with both neighbors separately
362: (thereby leading to two overlapping boxes).
363: 
364: %___________________  Fig 2  ________________________________
365: 
366: \begin{figure*}[!htb]
367: \centering
368: \includegraphics[angle=-90,width=12cm]{8683fig2a.ps}
369: \bigskip\bigskip\\
370: \includegraphics[angle=-90,width=12cm]{8683fig2b.ps}
371: \caption{\label{Fig.2}
372:   Grids used in our study overlaid on the face-on distribution of our
373:   482 low-latitude ($|b|<10\degr$) pulsars. Pulsars with a positive
374:   (negative) RM are denoted with crosses (circles). ($x$,$y$) are
375:   Galactocentric cartesian coordinates. The Sun is located at
376:   ($x=8.5 \ {\rm kpc}, y=0$). In the upper panel, the circles underlying the
377:   grids are at $r = 4\ {\rm kpc},\ 5\ {\rm kpc},\ 6\ {\rm kpc},\ 7\
378:   {\rm kpc} {\rm \ and}\ 8\ {\rm kpc}$, while in the lower panel, they
379:   are at $r = 3.5\ {\rm kpc},\ 4.5\ {\rm kpc},\ 5.5\ {\rm kpc},\ 6.5\
380:   {\rm kpc}, \ 7.5\ {\rm kpc} {\rm \ and}\ 8.5\ {\rm kpc}$.
381: }
382: \end{figure*}
383: 
384: %____________________________________________________________
385: 
386:   Altogether, we have 9 rings, which are centered on $r_i \equiv i\ {\rm kpc}$,
387:   with $i= 4.5,\, 5.5\, ...\, 7.5$ in the first grid and $i= 4,\, 5\, ...\, 8$
388:   in the second grid. In the following, the ring centered on $r_i$ (and
389:   extending between $r_{i-0.5}$ and $r_{i+0.5}$) is referred to as ring $i$.
390: 
391: We retain only the boxes containing at least 5 RMs. 
392: This minimum number of RMs, which is smaller than generally recommended 
393: for statistical testing, results from the limited number of pulsars with 
394: know RMs;
395: it was chosen as a trade-off between the need to have enough data points 
396: per box to make use of Eq.~(\ref{Eq_B_par_obs}) and the need to have 
397: enough boxes on the Galactic plane to capture the spatial variations 
398: of the large-scale magnetic field.
399: Even with such a small number of RMs in a given box, it is possible 
400: to trace the dependences of RM on distance and on DM for the pulsars 
401: lying in this box \citep[see][]{RL94, Weisberg04, Han06}.
402: 
403: We estimate the average value of $\overline{B_{||}}$, denoted by 
404: $\left< \overline{B_{||}} \right>$, in each of the retained boxes 
405: in the following way: we plot the points (DM,RM) of all the pulsars 
406: in the box, we fit a straight line through the resulting set of data points,
407: and we take $\left< \overline{B_{||}} \right>$ to be 1.232 times the slope 
408: of this line (see Eq.~(\ref{Eq_B_par_obs}) and preceding comment).
409:   To perform the straight-line fit, we resort to a slightly modified
410:   version of the ordinary least-squares linear regression of $Y$ on
411:   $X$ described by \citet{Isobe90},
412:   which is well suited when the dispersion of the data points about
413:   the linear relation cannot be calculated beforehand.
414:   The modifications brought to the original scheme are designed to exclude
415:   the occasional outliers -- such as those arising from H{\sc ii} regions
416:   \citep{Mitra03}. In practice, we discard all the data points
417:   whose absolute deviation from the mean RM or DM exceeds three times
418:   the mean absolute deviation.
419: To illustrate the procedure, we show two examples in Fig.~\ref{Fig.epl}.
420: 
421: %________________  Fig 3  ___________________________________
422: \begin{figure}[!htb]
423: \centering
424: \includegraphics[angle=-90,width=7cm]{8683fig3a.ps}
425: \bigskip\\
426: \includegraphics[angle=-90,width=7cm]{8683fig3b.ps}
427: \caption{\label{Fig.epl}
428: Plots of RM {\it versus} distance (left panels) and {\it versus} DM 
429: (right panels) for the pulsars lying in two different boxes.
430: The first box (top row) is delimited by the circles $r=r_{5.5}$ and
431: $r=r_{6.5}$ in the radial direction and by the tangential lines 
432: $l=l_{5.5}$ and $l=l_{6.5}$ in the longitudinal direction.
433: The second box (bottom row) is delimited by the circles $r=r_{6.5}$ and
434: $r=r_{7.5}$ and by the tangential lines $l=l_{4.5}$ and $l=l_{6.5}$.
435: For each box, the best-fit straight line through the points (DM,RM) 
436: is drawn in the right panel, and the corresponding value of 
437: $\left< \overline{B_{||}} \right>$ with its statistical uncertainty 
438: are written in the upper right corner.
439: The red points represent outliers.
440: }
441: \end{figure}
442: %____________________________________________________________
443: 
444: 
445: The derived values of $\left< \overline{B_{||}} \right>$ 
446: in all the boxes of our two grids are mapped in Fig.~\ref{Fig.3}.
447: For convenience, these values are converted into vectors oriented along
448: the local line of sight.
449: 
450: %________________  Fig 3  ___________________________________
451: 
452: \begin{figure*}[!htb]
453: \centering
454: \includegraphics[angle=-90,width=12cm]{8683fig4.ps}
455: \caption{\label{Fig.3}
456: Face-on map showing the average line-of-sight component of the magnetic
457: field, $\left<\overline{B_{||}}\right>$, obtained in the different boxes
458: of our two grids.
459: Each $\left<\overline{B_{||}}\right>$ is plotted in the form of a vector
460: centered on the box midpoint (point at middle radius, $r=r_i$,
461: and middle longitude, $l = (l_{\rm min} + l_{\rm max}) /2$),
462: and oriented along the local line of sight.
463: }
464: \end{figure*}
465: 
466: %____________________________________________________________
467: 
468: As explained above Eq.~(\ref{Eq_B_par_obs}), the turbulent component
469: of the magnetic field, $\delta {\bf B}$, causes the RMs to scatter
470: about the mean DM-RM line. This physical scatter due to turbulence
471: is typically one order of magnitude larger than the observational
472: scatter due to measurement errors.  The amplitude of the RM scatter
473: is given by the r.m.s. deviation of the measured RMs from the mean DM-RM
474: line.  Since the RM scatter is of predominantly turbulent origin,
475: its amplitude divided by the mean DM in the considered box directly yields 
476: (to a factor 1.232) an estimate for the r.m.s. value of the turbulent 
477: component of the line-of-sight field, $\delta \overline{B_{||}}$.
478: Finally, the r.m.s. value of $\delta \overline{B_{||}}$ divided by 
479: the square root of the number of pulsars in the box provides an estimate 
480: for the statistical uncertainty in $\left< \overline{B_{||}} \right>$, 
481: $\sigma_{||}$, which, again, is predominantly due to turbulence.
482: The exact expression of $\sigma_{||}$ can be found in \citet{Isobe90}.
483: With our data, the typical values of $\sigma_{||}$ lie between 
484: $\sim 0.2\, \mu{\rm G}$ and $1.3 \, \mu{\rm G}$.
485: 
486: 
487: \section{Data fitting to the field models}
488: 
489: Once we have obtained a set of observational values of 
490: $\left< \overline{B_{||}} \right>$ together with their statistical
491: uncertainties (or error bars), we can put the three theoretical
492: field models presented in Sect.~2 to the test. 
493: As in all other studies based on RMs, we proceed on the notion that
494: the large-scale interstellar magnetic field may be identified with 
495: its $n_{\rm e}$-weighted average value (denoted with an overbar).
496: Implicit here is the assumption that fluctuations in magnetic field 
497: strength and in free-electron density are statistically uncorrelated.
498: In reality, this assumption is certainly not strictly satisfied in the ISM
499: \citep[e.g.,][]{Beck03}, and this will cause our results to be somewhat biased.
500: 
501: With this caveat in mind, we now describe the overall procedure. 
502: For each model, we use all our observational values
503: of $\left< \overline{B_{||}} \right>$ to derive the best-fit
504: parameters of the model. We then examine whether the best fit is
505: consistent with the pulsar data, i.e., whether the theoretical
506: line-of-sight fields predicted by it fall within the error bars
507: of the observational $\left< \overline{B_{||}} \right>$
508: (in a statistical sense).
509: If we find that the best fit is not consistent with the data,
510: we may conclude that the considered model must be rejected.  
511: If, on the other hand, the best fit is found consistent with the data, 
512: we may conclude that the model is acceptable; we then determine 
513: the extent of the so-called "consistency domain", i.e., 
514: the parameter domain around the best fit within which solutions 
515: are consistent with the data.
516: 
517: It is important to realize that the concept of acceptability
518: differs from the concept of detectability.
519: A given field model, say, the ring model, is acceptable
520: only if it is not ruled out by the available pulsar data.
521: This does not necessarily imply that the Galactic magnetic field
522: is really of the ring type, nor that a ring field has truly
523: been detected. Detection of a ring field requires not only that
524: the ring model be acceptable, but also that the zero-field solution
525: do not belong to the consistency domain.
526: 
527: Let us now discuss more specifically what exact criterion should be 
528: used to test consistency with the pulsar data for a given field model.
529: Each of the three models is characterized by a number of independent
530: free functions of Galactic radius [$\left<B_\theta\right>(r)$ 
531: in the ring model; $\left<B_r\right>(r)$ and $\left<B_\theta\right>(r)$ 
532: in the axisymmetric model; $b_r(r)$, $b_{\theta}(r)$ and $\phi(r)$
533: in the bisymmetric model], corresponding to the same number of
534: independent free parameters in every ring $i$ [denoted by 
535: $\left<B_\theta\right>_i$ in the ring model; $\left<B_r\right>_i$ 
536: and $\left<B_\theta\right>_i$ in the axisymmetric model; 
537: $b_{r,i}$, $b_{\theta,i}$ and $\phi_i$ in the bisymmetric model].
538: Therefore, the 9 different rings may be analyzed separately.
539: 
540: For any one of the three field models, consider a given ring $i$ 
541: and suppose that this ring contains $n_i$ boxes.
542: For every box $j$, we have derived an observational value of
543: the average line-of-sight field $\left<\overline{B_{||}}\right>$,
544: denoted by $\left<\overline{B_{||}}\right>_{ij}$, together with
545: its statistical uncertainty, denoted by $(\sigma_{||})_{ij}$.
546: Besides, we can calculate a theoretical expression of the large-scale 
547: line-of-sight field $\left<B_{||}\right>$, denoted by 
548: $\left<B_{||}\right>_{ij}$, in terms of the free parameters of ring $i$.
549: %%for the model under test.
550: The best-fit values of these parameters are obtained by minimizing
551:    \begin{equation}
552: 	 \chi^2 = \sum^{n_i}_{j=1}
553: 	 \left( \frac{ \left<\overline{B_{||}}\right>_{ij} -
554: 				   \left<B_{||}\right>_{ij} }
555: 				 { (\sigma_{||})_{ij} }
556: 	 \right)^2 \ .
557: 	 \label{Eq_chi2_par}
558:    \end{equation}
559: 
560: The best fit of ring $i$ can be considered consistent with the pulsar data 
561: if, on average over ring $i$, the theoretical best-fit 
562: $\left<B_{||}\right>_{ij}$ do not differ from the observational 
563: $\left<\overline{B_{||}}\right>_{ij}$
564: by more than the associated uncertainties $(\sigma_\theta)_{ij}$.
565: In mathematical terms, this condition for consistency can be expressed as 
566: $\chi^2 \le n_i $. 
567: However, when the number of data points, $n_i$, is not much greater than 
568: the number of free parameters, $\nu$ [$\nu=1$ for the ring model;
569: $\nu=2$ for the axisymmetric model; $\nu=3$ for the bisymmetric model], 
570: consistency with the data should be tested with the more exact criterion
571:    \begin{equation}
572: 	 \chi^2 \le n_i - \nu \ ,
573: 	 \label{Eq_consist_par}
574:    \end{equation}
575: where $n_i -\nu$ is the number of degrees of freedom, i.e., 
576: the number of data points that cannot automatically be placed on a curve
577: with $\nu$ adjustable parameters.
578: Eq.~(\ref{Eq_consist_par}) provides a rule of thumb for a reasonably
579: good fit \citep[see Sect.~15.1 in][]{ptvf92}.
580: If $\chi^2 \gg n_i -\nu$, the best-fit curve misses too many data points 
581: to be believable.
582: 
583: It is possible to obtain a more rigorous (and, at the same time, 
584: more flexible) criterion for consistency.
585: Suppose, for the sake of argument, that the model we are testing is correct. 
586: If the data points $\left<\overline{B_{||}}\right>_{ij}$ of ring $i$ 
587: follow a Gaussian distribution, $\chi^2$ has a chi-square distribution 
588: with $n_i -\nu$ degrees of freedom. 
589: One can then calculate the {\it a priori} probability,
590: $P [\chi^2>\chi^2_{\rm crit}]$, that the $\chi^2$ obtained 
591: for a particular set of data points exceeds some critical value 
592: $\chi^2_{\rm crit}$.
593: Conversely, one can calculate the critical $\chi^2_{\rm crit}$
594: for which $P [\chi^2>\chi^2_{\rm crit}]$ equals some imposed 
595: probability $P_0$.
596: For instance, if the model is correct, it is unlikely (only 10\% chance) 
597: that $\chi^2 > \chi^2_{\rm crit} (P_0=0.1)$. Turning the statement around, 
598: if we find $\chi^2 > \chi^2_{\rm crit} (P_0=0.1)$, it is unlikely that 
599: the model is correct -- we will say that the model is inconsistent 
600: with the data.
601: This reasoning directly leads to the following consistency condition:
602:    \begin{equation}
603:          \chi^2 \le \chi^2_{\rm crit} (P_0) \ .
604: 	 \label{Eq_consist_prob}
605:    \end{equation}
606: Here, we will adopt $P_0=0.1$ as our default value, 
607: but we will also discuss the results obtained for $P_0=0.05$.
608: In practice, the values of $P [\chi^2>\chi^2_{\rm crit}]$
609: for given $n_i -\nu$ and $\chi^2_{\rm crit}$ are tabulated in various
610: textbooks \citep[e.g.,][]{Yamane64}. The tables can also be used
611: to determine $\chi^2_{\rm crit} (P_0)$ for given $n_i -\nu$ and $P_0$.
612: For reference, the values of $\chi^2_{\rm crit} (P_0)$ 
613: for $n_i-\nu=1,\, 2,\, 3,\, 4,\, 5$ and for $P_0=0.05,\, 0.1,\, 0.2$ 
614: are listed in Table~\ref{Table_Crit}.
615: An important point emerging from Table~\ref{Table_Crit} is that 
616: $\chi^2_{\rm crit} (P_0=0.1) > n_i -\nu$, 
617: so that Eq.~(\ref{Eq_consist_prob}) with $P_0=0.1$ will always be 
618: easier to satisfy than Eq.~(\ref{Eq_consist_par}) .
619: 
620: 
621: \begin{table}
622: \caption{Critical values of $\chi^2$ for 3 probability levels}
623: \label{Table_Crit}
624: \centering
625: \begin{tabular}{c|c c c }
626: \hline\hline
627: \noalign{\smallskip}
628: %%\backslashbox{$n_i -\nu$}{$P_0$} & 0.05 & 0.1 & 0.2\\
629: $n_i -\nu$ & $\chi^2_{\rm crit} (P_0=0.05)$ & 
630: $\chi^2_{\rm crit} (P_0=0.1)$ & $\chi^2_{\rm crit} (P_0=0.2)$ \\
631: \hline
632: \noalign{\smallskip}
633: 1 & 3.841 & 2.706 & 1.642 \\
634: 2 & 5.991 & 4.605 & 3.219 \\
635: 3 & 7.815 & 6.251 & 4.642 \\
636: 4 & 9.488 & 7.779 & 5.989 \\
637: 5 & 11.070 & 9.236 & 7.289 \\
638: \noalign{\smallskip}
639: \hline
640: \end{tabular}
641: \end{table}
642: 
643: In the next three subsections, we present the results obtained 
644: with the rule of thumb (Eq.~(\ref{Eq_consist_par}))
645: and with the more rigorous consistency condition 
646: (Eq.~(\ref{Eq_consist_prob})), for the three field models.
647:  
648: 
649: \subsection {Ring model}
650: 
651: In the ring model, $\left<B_r\right>$ vanishes and $\left<B_\theta\right>$
652: is constant along circles. Hence, there are 9 free parameters: 
653: $\left<B_\theta\right>_i$, the large-scale azimuthal fields 
654: in the 9 rings $i= 4,\, 4.5,\, 5\, ...\, 8$.\footnote{\label{footnote1}
655: As a reminder, ring $i$ is centered on $r_i \equiv i\ {\rm kpc}$ and 
656: extends between $(i-0.5)\ {\rm kpc}$ and $(i+0.5)\ {\rm kpc}$.}
657: Since all the free parameters are independent, the 9 rings can be treated
658: separately.
659: 
660: For every ring $i$, the large-scale line-of-sight field in any box $j$ 
661: is simply the projection of $\left<B_\theta\right>_i$ onto the line of
662: sight (see Eq.~(\ref{Eq_B_para}) with $\left<B_r\right> = 0$):
663:    \begin{equation}
664: 	\left<{B_{||}}\right>_{ij} =
665: 		\left<{B_\theta}\right>_{i}\ \cos \alpha_{ij} \ ,
666:    \label{Eq_ring_Bpij}
667:    \end{equation}
668: where $\alpha_{ij}$ is the angle between the azimuthal direction and
669: the direction to the Sun at the midpoint\footnote{As in Fig.~\ref{Fig.3},
670: the midpoint of a box is defined as the point at middle radius, $r=r_i$, 
671: and middle longitude, $l = (l_{\rm min} + l_{\rm max}) /2$.} of box $j$ 
672: (see Fig.~\ref{Fig.1}). 
673: The best-fit value of $\left<B_\theta\right>_i$ is obtained 
674: by minimizing $\chi^2$ (given by Eq.~(\ref{Eq_chi2_par})).
675: In terms of $\left<\overline{B_\theta}\right>_{ij} =
676: \left<\overline{B_{||}}\right>_{ij} / \cos \alpha_{ij}$,
677: the observational value of the average azimuthal field in box $j$,
678: and $(\sigma_\theta)_{ij} = (\sigma_{||})_{ij} / \cos \alpha_{ij}$,
679: the associated uncertainty, the minimization procedure turns out to be
680: equivalent to taking an uncertainty-weighted average of the different 
681: $\left<\overline{B_\theta}\right>_{ij}$ along ring $i$:
682:    \begin{equation}
683:      \left<B_\theta\right>_i \ = \
684:      \frac{\displaystyle
685:            \sum^{n_i}_{j=1} \frac{\left<\overline{B_\theta}\right>_{ij}}
686:                                  {(\sigma_\theta)_{ij}^2}
687:           }
688:           {\displaystyle
689:            \sum^{n_i}_{j=1} \frac{1}
690:                                  {(\sigma_\theta)_{ij}^2}
691:           }
692:      \ .
693:      \label{Eq_ring_i}
694:    \end{equation}
695: 
696: The values of $\left<\overline{B_\theta}\right>_{ij}$
697: and their uncertainties $(\sigma_\theta)_{ij}$ in the $n_i$ boxes $j$
698: of the 9 rings $i$ are plotted in Fig.~{\ref{Fig_ring_B_theta}},
699: at the Galactic azimuthal angles of the box midpoints, $\theta_{ij}$.
700: For comparison, the best-fit values of $\left<B_\theta\right>_i$
701: in the 9 rings are indicated by horizontal lines spanning the entire
702: azimuthal range.
703: 
704: %_________________ Fig 4 _________________________________
705: 
706: \begin{figure*}[!htb]
707: \centering
708: \includegraphics[angle=-90,width=10cm]{8683fig5.ps}
709: \caption{\label{Fig_ring_B_theta}
710: Observational values of the average azimuthal fields,
711: $\left<\overline{B_\theta}\right>_{ij}$, {\it versus}
712: Galactic azimuthal angles, $\theta_{ij}$, in all the boxes $j$
713: of the 9 successive rings $i$, for the ring model.
714: The uncertainties $(\sigma_\theta)_{ij}$ in the field values
715: are plotted as standard error bars.
716: For each ring $i$, the best-fit value of the model parameter
717: $\left<B_\theta\right>_i$ is indicated by the horizontal
718: dot-dashed line.
719: }
720: \end{figure*}
721: 
722: %____________________________________________________________
723: 
724: In only one ring ($i=4.5$) does the best-fit value of
725: $\left<B_\theta\right>_i$ satisfy the rule-of-thumb 
726: consistency condition, $\chi^2 \le n_i - 1$
727: (Eq.~(\ref{Eq_consist_par}) with $\nu=1$).
728: For this ring, we compute the consistency range of
729: $\left<B_\theta\right>_i$, which contains all the values of
730: $\left<B_\theta\right>_i$ for which $\chi^2 \le n_i - 1$.
731: The best-fit value of $\left<B_\theta\right>_i$ and its consistency
732: range in the sole "good-fit" ring are plotted against $r_i$,
733: in the upper panel of Fig.~{\ref{Fig_ring_B_r}}.
734: For the other 8 rings, the (inconsistent) best-fit values of
735: $\left<B_\theta\right>_i$ are plotted with crosses.
736: Clearly, these 8 rings do not admit any ring magnetic field 
737: consistent with the data.
738: As an immediate consequence, the ring model must be rejected.
739: 
740: In order to gain some feel for how far the ring model is from being
741: able to reproduce the pulsar data, let us, in thought, extend the
742: error bars of all the observational
743: $\left<\overline{B_{||}}\right>_{ij}$ by a factor of 2 and look into
744: the impact of this extension on our results. With twice the original
745: error bars, the $\chi^2$ parameter would be smaller by a factor of 4,
746: so that, in terms of the original $\chi^2$, the rule-of-thumb
747: consistency condition would become $\chi^2  \le 4 (n_i-1)$.  
748: As it turns out, this less stringent consistency condition would be 
749: fulfilled in  5 rings ($i=4, \, 4.5, \, 5.5, \, 6, \, 7$)
750: out of 9. Thus, with twice the original error bars, the ring model 
751: would remain unacceptable.
752: 
753: If we now resort to the more rigorous consistency condition,
754: $\chi^2 \le \chi^2_{\rm crit} (0.1)$
755: (Eq.~(\ref{Eq_consist_prob}) with $P_0=0.1$),
756: to test the ring model, we find that 3 rings ($i=4, \, 4.5, \, 6$) 
757: have their best-fit $\left<B_\theta\right>_i$ consistent with the data;
758: their consistency ranges are drawn in the lower panel of 
759: Fig.~{\ref{Fig_ring_B_r}}.
760: For the other 6 rings, the (inconsistent) best-fit $\left<B_\theta\right>_i$ 
761: are again plotted with crosses.
762: With $P_0=0.05$, 4 rings ($i=4, \, 4.5, \, 6,\, 7$) would be deemed
763: consistent with the data, but the other 5 rings would still fail
764: the consistency test. 
765: 
766: These results confirm our conclusion that the ring model must be rejected.
767: 
768: %_________________  Fig 5____________________________________
769: %
770: \begin{figure*}[!htb]
771: \centering
772: \includegraphics[angle=-90,width=8cm]{8683fig6.ps}
773: \caption{\label{Fig_ring_B_r}
774: Best-fit values of the model parameters, $\left<B_\theta\right>_i$,
775: {\it versus} ring middle radii, $r_i$, for the ring model.
776: The acceptable values (those consistent with the pulsar data) are shown
777: with their consistency ranges, while the unacceptable values are
778: indicated with crosses.
779: The results obtained with the rule-of-thumb consistency condition,
780: $\chi^2  \le n_i-1$, are plotted in the upper panel.
781: Those obtained with the more rigorous consistency condition,
782: $\chi^2 \le \chi^2_{\rm crit} (0.1)$,
783: are plotted in the lower panel.
784: }
785: \end{figure*}
786: %____________________________________________________________
787: 
788: 
789: \subsection {Axisymmetric model}
790: 
791: In the axisymmetric model, $\left<B_r\right>$ and $\left<B_{\theta}\right>$
792: are both constant along circles. Hence, there are 18 free parameters: 
793: $\left<B_r\right>_i$ and $\left<B_{\theta}\right>_i$,
794: the large-scale radial and azimuthal fields
795: in the 9 rings $i= 4,\, 4.5,\, 5\, ...\, 8$.
796: 
797: For every ring $i$, the large-scale line-of-sight field in any box $j$ 
798: can be written as a linear combination of the two parameters
799: $\left<B_r\right>_i$ and $\left<B_{\theta}\right>_i$:
800:    \begin{equation}
801: 	 \left<B_{||}\right>_{ij}
802: 	 = \left<B_r\right>_i \, \sin \alpha_{ij}
803: 	 + \left<B_{\theta}\right>_i \, \cos \alpha_{ij}
804: 	 \label{Eq_ASS_ij}
805:    \end{equation}
806: (see Eq.~(\ref{Eq_B_para})).
807: Again the best-fit values of $\left<B_r\right>_i$ and 
808: $\left<B_{\theta}\right>_i$ are obtained by minimizing $\chi^2$ 
809: (Eq.~(\ref{Eq_chi2_par})).
810: 
811: Here, we find that 1 ring ($i= 4.5$) has its best fit
812: consistent with the data, according to the rule-of-thumb consistency 
813: condition, $\chi^2 \le n_i - 2$ (Eq.~(\ref{Eq_consist_par}) with $\nu=2$).
814: Its consistency domain in the parameter plane
815: $(\left<B_r\right>_i,\left<B_{\theta}\right>_i)$ 
816: is the area delimited by the ellipse $\chi^2 = n_i -2$
817: (grey contour line in the relevant panel of Fig.~\ref{Fig_ASS_B_B}).
818: As none of the other 8 rings can be properly fit with an axisymmetric
819: magnetic field, the axisymmetric model must be rejected.
820: 
821: With twice the original error bars on the observational
822: $\left<\overline{B_{||}}\right>_{ij}$, the rule-of-thumb consistency 
823: condition would become $\chi^2 \le 4 (n_i-2)$ (in terms of the original 
824: $\chi^2$).
825: This less stringent consistency condition would be satisfied in 5 rings   
826: ($i=4,\, 4.5,\, 6,\, 7,\, 8$), but still not in the other 4 rings.
827: Therefore, the axisymmetric model would remain unacceptable.
828: 
829: According to the more rigorous consistency condition,
830: $\chi^2 \le \chi^2_{\rm crit} (0.1)$ (Eq.~(\ref{Eq_consist_prob}) 
831: with $P_0=0.1$), 5 rings ($i=4,\, 4.5,\, 6,\, 7,\, 8$) have their best fits 
832: consistent with the data.
833: Their consistency domains in the parameter planes
834: $(\left<B_r\right>_i,\left<B_{\theta}\right>_i)$ are the elliptical
835: areas enclosed by the curves $\chi^2 = \chi^2_{\rm crit} (0.1)$
836: (black contour lines in the relevant panels of Fig.~\ref{Fig_ASS_B_B}).
837: For the other 4 rings, the (inconsistent) best-fit pairs
838: $(\left<B_r\right>_i,\left<B_{\theta}\right>_i)$
839: are indicated with crosses. 
840: Relaxing the probability level to $P_0=0.05$ would not raise
841: the number of acceptable rings above 5.
842: 
843: From all the above, we conclude that the axisymmetric model must 
844: be rejected.
845: 
846: %_________________ Fig 7 ____________________________________
847: 
848: \begin{figure*}[!htb]
849: \centering
850: \includegraphics[angle=-90,width=12cm]{8683fig7.ps}
851: \caption{\label{Fig_ASS_B_B}
852: Consistency domains in the parameter planes
853: $(\left<B_r\right>_i,\left<B_{\theta}\right>_i)$ of the 9 different rings,
854: for the axisymmetric model.
855: The grey contour lines define the consistency domains obtained 
856: with the rule-of-thumb consistency condition, $\chi^2 \le n_i-2$. 
857: The black contour lines define those obtained with the more rigorous
858: consistency condition, $\chi^2 \le \chi^2_{\rm crit} (0.1)$.
859: The crosses mark the locations of the unacceptable best fits
860: (those inconsistent with the pulsar data according to both criteria).
861: }
862: \end{figure*}
863: 
864: %______________________________________________________
865: 
866: 
867: \subsection {Bisymmetric model}
868: 
869: In the bisymmetric model, $\left< B_r \right>$ and $\left< B_\theta \right>$
870: vary sinusoidally along circles in the manner described by
871: Eqs.~(\ref{Eq_BSS_Br}) and (\ref{Eq_BSS_Btheta}).
872: Hence, there are 27 free parameters:
873: $b_{r,i}$, $b_{\theta,i}$ and ${\phi}_i$, the maximum amplitudes and
874: the azimuthal phases in the 9 rings $i= 4,\, 4.5,\, 5\, ...\, 8$.
875: 
876: For every ring $i$, the large-scale line-of-sight field in any box $j$ 
877: follows from Eq.~(\ref{Eq_B_para}) together with
878: Eqs.~(\ref{Eq_BSS_Br})--(\ref{Eq_BSS_Btheta}):
879: 	 \begin{equation}
880: 	   \left<{B_{||}} \right>_{ij} =
881: 	   b_{r,i} \, \sin (\theta_{ij}-\phi_{i})\, \sin \alpha_{ij} +
882: 	   b_{\theta,i} \, \sin(\theta_{ij}-\phi_{i}) \, \cos \alpha_{ij} \ ,
883: 	   \label{Eq_BSS_ij}
884: 	 \end{equation}
885: where the angles $\theta_{ij}$ and $\alpha_{ij}$ (see Fig.~\ref{Fig.1})
886: refer to the midpoint of box $j$.
887: Similarly to the previous models, the best-fit values of
888: the three parameters $b_{r,i}$, $b_{\theta,i}$ and ${\phi}_i$
889: are obtained through a minimization of $\chi^2$ (Eq.~(\ref{Eq_chi2_par})).
890: 
891: % It is important to note that, in the outermost ring ($i=8$), 
892: % the number of data points ($n_i=3$) is exactly equal to 
893: % the number of adjustable parameters ($\nu=3$).
894: % This means that we will be able to find a curve that passes exactly
895: % through the three data points, but with zero degree of freedom left
896: % ($n_i -\nu = 0$), we will have no way of testing the goodness of fit.
897: % For this reason, we exclude the outermost ring from our discussion 
898: % of the model's acceptability.
899: 
900: According to the rule-of-thumb consistency condition,
901: $\chi^2 \le n_i - 3$ (Eq.~(\ref{Eq_consist_par}) with $\nu=3$), 
902: 2 rings ($i=4.5,\, 5$) have their best fits consistent with the data. 
903: Their consistency domains in the parameter spaces
904: $(b_{r,i},b_{\theta,i},{\phi}_i)$ are the volumes bounded 
905: by the surfaces $\chi^2 = n_i - 3$.
906: Displayed in Fig.~\ref{Fig_BSS_b_b} are the projections of these
907: consistency domains on the parameter planes $(b_{r,i},b_{\theta,i})$
908: (grey contour lines).
909: Since the other 7 rings fail the consistency test, the bisymmetric model
910: must be rejected.
911: 
912: With twice the original error bars on the observational
913: $\left<\overline{B_{||}}\right>_{ij}$, the rule-of-thumb consistency
914: condition would become $\chi^2 \le 4 (n_i-3)$, which would be satisfied
915: in 7 rings ($i=4,\, 4.5,\, 5, \, 5.5,\, 6,\, 7, \, 8$) out of 9.
916: The bisymmetric model would then be nearly acceptable.
917: 
918: According to the more rigorous consistency condition,
919: $\chi^2 \le \chi^2_{\rm crit} (0.1)$ (Eq.~(\ref{Eq_consist_prob}) 
920: with $P_0=0.1$), 5 rings ($i=4.5,\, 5, \, 6,\, 7, \, 8$) have their best fits
921: consistent with the data.
922: Their consistency domains, bounded by the surfaces 
923: $\chi^2 = \chi^2_{\rm crit} (0.1)$, are also shown in projection
924: on the parameter planes $(b_{r,i},b_{\theta,i})$ in Fig.~\ref{Fig_BSS_b_b}
925: (black contour lines).
926: For the other  4 rings, the (inconsistent) best-fit pairs 
927: $(b_{r,i},b_{\theta,i})$ are indicated with crosses.
928: Relaxing the probability level to $P_0=0.05$ would raise the number 
929: of acceptable rings to 7 ($i=4,\, 4.5,\, 5, \, 5.5,\, 6,\, 7,\, 8$),
930: which would render the bisymmetric model nearly globally acceptable.
931: 
932: Altogether, the bisymmetric model must be rejected, though its rejection 
933: is slightly less severe than for the axisymmetric model.
934: 
935: %___________________ Fig 8 __________________________________
936: 
937: \begin{figure*}[!htb]
938: \centering
939: \includegraphics[angle=-90,width=12cm]{8683fig8.ps}
940: \caption{\label{Fig_BSS_b_b}
941: Projections of the 3D consistency domains of the 9 different rings
942: on their parameter planes $(b_{r,i},b_{\theta,i})$, for the bisymmetric model.
943: The grey contour lines correspond to the consistency domains obtained 
944: with the rule-of-thumb consistency condition, $\chi^2 \le n_i-3$.
945: The black contour lines correspond to those obtained with the more
946: rigorous consistency condition, $\chi^2 \le \chi^2_{\rm crit} (0.1)$.
947: The crosses mark the locations of the unacceptable best fits
948: (those inconsistent with the pulsar data according to both criteria). 
949: %The untestable best fit of ring~8 is not shown because it falls 
950: %  outside the frame.
951: }
952:  \end{figure*}
953: 
954: %____________________________________________________________
955: 
956: 
957: \section {Summary and conclusions}
958: 
959: In this paper, we examined the three most common theoretical models for
960: the large-scale magnetic field in the Galactic disk and confronted each
961: of these models with the pulsar data.
962: For each model, we derived the best-fit parameters, through $\chi^2$
963: minimization, in the 9 Galactocentric rings defined in Fig.~\ref{Fig.2},
964: and we delineated the parameter domains around the best fits (referred to 
965: as the consistency domains) wherein the predicted fields are consistent 
966: with the pulsar data.
967: 
968: Compared to existing studies of the kind, we did not attempt to settle
969: the long-standing (and possibly ill-posed) question of whether
970: the Galactic magnetic field is axisymmetric or bisymmetric.
971: %%Nor did we try to find out which of these two types of azimuthal symmetry
972: %%is favored by the pulsar data.
973: Our sole purpose was to determine whether each of the three basic field
974: models, taken separately, is compatible with the available pulsar data
975: or not.
976: In this regard, we note that many previous studies did find
977: a preference for one of the field models, but omitted to put their
978: preferred model through the crucial "goodness-of-fit" test, which checks
979: whether the model can indeed reproduce the data within the error bars.
980: 
981: Here, we tested the three field models on the basis of two different 
982: criteria: first, a standard rule of thumb for a reasonably good fit 
983: (Eq.~(\ref{Eq_consist_par})), and second, a more rigorous consistency 
984: condition for a chi-square distribution of $\chi^2$ 
985: (Eq.~(\ref{Eq_consist_prob}), with the imposed probability 
986: set to $P_0=0.1$).
987: These two criteria were successively applied to all the rings separately, 
988: such that the best fit of ring $i$ was deemed consistent with the pulsar data
989: if the associated value of $\chi^2$, $\chi^2_{\rm min}$, 
990: was less than $n_i - \nu$ (first criterion) or less than 
991: $\chi^2_{\rm crit} (P_0=0.1)$ (second criterion).
992: A model could then be considered globally acceptable if all the rings 
993: had their best fits consistent with the data.
994: 
995: The results obtained for the three field models, with both criteria,
996: are summarized in Table~\ref{Table_summary1}. 
997: All the rings are listed with their labels, $i$ 
998: (see footnote~\ref{footnote1}), and their numbers of boxes
999: (or numbers of data points), $n_i$, from which it is straightforward 
1000: to deduce the numbers of degrees of freedom, $n_i -\nu$ (for the first 
1001: criterion), and the critical $\chi^2_{\rm crit} (P_0=0.1)$ 
1002: (for the second criterion; see Table~\ref{Table_Crit}).
1003: Also given for all the rings are the minimum values of $\chi^2$, 
1004: $\chi^2_{\rm min}$, i.e., the values associated with the best fits, 
1005: as well as the results of both consistency tests (satisfaction of a test 
1006: is indicated with an asterisk), for the three field models.
1007: 
1008: \begin{table*}[!htb]
1009: \begin{minipage}[t]{\columnwidth}
1010: \caption{Summary of the results obtained for the three field models$\ ^a$
1011: }
1012: \label{Table_summary1}
1013: \centering
1014: \renewcommand{\footnoterule}{}  % to avoid a line before footnotes
1015: \begin{tabular}{l l|r c c|r c c|r c c}
1016: \hline\hline
1017: \noalign{\smallskip}
1018: & & \multicolumn{3}{c|}{Ring model \ ($\nu=1$)} 
1019: & \multicolumn{3}{c|}{Axisymmetric model \ ($\nu=2$)} 
1020: & \multicolumn{3}{c}{Bisymmetric model \ ($\nu=3$)} \\
1021: \noalign{\smallskip}
1022: %%\cline{3-5} \cline{7-9} \cline{11-13} \\
1023: \hline
1024: \noalign{\smallskip}
1025: $i$ & $n_i$ & $\chi^2_{\rm min}$ 
1026: & $\!\!\left[ \chi^2_{\rm min} \le n_i - \nu \right]\!\!$
1027: & $\left[ \chi^2_{\rm min} \le \chi^2_{\rm crit} (0.1) \right]$
1028: & $\chi^2_{\rm min}$ 
1029: & $\!\!\left[ \chi^2_{\rm min} \le n_i - \nu \right]\!\!$
1030: & $\left[ \chi^2_{\rm min} \le \chi^2_{\rm crit} (0.1) \right]$
1031: & $\chi^2_{\rm min}$ 
1032: & $\!\!\left[ \chi^2_{\rm min} \le n_i - \nu \right]\!\!$
1033: & $\left[ \chi^2_{\rm min} \le \chi^2_{\rm crit} (0.1) \right]$ \\
1034: \noalign{\smallskip}
1035: \hline
1036: \noalign{\smallskip}
1037: 4   & 4 & 5.03 &        & $\ast$ &  2.91 &        & $\ast$ & 3.08 &        & \\
1038: 4.5 & 6 & 3.17 & $\ast$ & $\ast$ &  3.11 & $\ast$ & $\ast$ & 1.75 & $\ast$ & $\ast$ \\
1039: 5   & 5 & 25.37&        &        & 18.65 &        &        & 1.79 & $\ast$ & $\ast$ \\
1040: 5.5 & 5 & 14.52&        &        & 14.43 &        &        & 4.95 &        & \\
1041: 6   & 6 & 8.84 &        & $\ast$ &  5.06 &        & $\ast$ & 5.48 &        & $\ast$ \\
1042: 6.5 & 5 & 20.29&        &        & 17.93 &        &        &13.55 &        & \\
1043: 7   & 6 & 9.57 &        &        &  5.81 &        & $\ast$ & 3.37 &        & $\ast$ \\
1044: 7.5 & 4 & 19.31&        &        & 19.30 &        &        & 7.44 &        & \\
1045: 8   & 4 & 52.53&        &        &  3.04 &        & $\ast$ & 0.82 &        & $\ast$ \\
1046: \noalign{\smallskip}
1047: \hline
1048: \end{tabular}
1049: \parbox{14cm}{
1050: \begin{flushleft}
1051: 	\tiny{$^a\ $ When a value of $\chi^2_{\rm min}$ satisfies the first or second 
1052:         consistency condition, an asterisk is plotted in the corresponding column.}
1053: \end{flushleft}
1054: }
1055: \end{minipage}
1056: \end{table*}
1057: 
1058: We found that none of the three field models is acceptable, 
1059: in the sense that none of them can be brought into full agreement 
1060: with the pulsar data.
1061: According to the standard rule-of-thumb consistency condition
1062: (Eq.~(\ref{Eq_consist_par})), all three models must be strongly rejected,
1063: as the ring and axisymmetric models fail to provide a good fit 
1064: (consistent with the data) in all the rings save one, 
1065: while the bisymmetric model fails in all the rings save two.
1066: If the error bars of the observational line-of-sight fields
1067: were enlarged by a factor of 2, the bisymmetric model would not be 
1068: too far from acceptable (7 good-fit rings out of 9),
1069: %the ring model would not be too far off (7 good-fit rings out of 9),
1070: while the ring and axisymmetric models would remain truly unacceptable 
1071: (5 good-fit rings out of 9). 
1072: The conclusions reached with the more rigorous consistency condition
1073: (Eq.~(\ref{Eq_consist_prob})) are intermediate between the two situations 
1074: described above: with the imposed probability set to $P_0=0.1$ ($P_0=0.05$), 
1075: 3 (4) rings can be properly fit with a ring magnetic field,
1076: 5 (5) with an axisymmetric field and 5 (7) with a bisymmetric field.
1077: 
1078: The quantitative differences between both criteria are easily understood.
1079: A comparison between Eqs.~(\ref{Eq_consist_par}) and
1080: (\ref{Eq_consist_prob}) immediately shows that the rule-of-thumb 
1081: consistency condition (Eq.~(\ref{Eq_consist_par})) corresponds to
1082: a probability level $P_1$ such that $\chi^2_{\rm crit} (P_1) = n_i-\nu$,
1083: or equivalently, $P_1 = P [\chi^2 > n_i-\nu]$.
1084: With twice the original error bars on the observational line-of-sight
1085: fields, the rule-of-thumb consistency condition would become 
1086: $\chi^2 \le 4 (n_i-\nu)$, corresponding to a probability level $P_2$ 
1087: such that $\chi^2_{\rm crit} (P_2) = 4 (n_i-\nu)$,
1088: or equivalently, $P_2 = P [\chi^2 > 4 (n_i-\nu)]$.
1089: The values of $P_1$ and $P_2$ as functions of $n_i-\nu$
1090: are tabulated in Table~\ref{Table_P1}.
1091: Clearly, the rule-of-thumb consistency condition with the original 
1092: error bars implies high probability levels ($P_1 \sim 30\%-40\%$),
1093: which make it overly difficult to satisfy; if the model under testing 
1094: is correct, there is nonetheless a $\sim 30\%-40\%$ chance 
1095: that $\chi^2$ exceeds $n_i-\nu$ and that the model will be rejected.
1096: On the other hand, with twice the original error bars, the probability 
1097: levels drop very low ($P_2 \sim 0.1\%-4\%$) and the consistency condition
1098: becomes too easily satisfied; the risk is then to accept a model that 
1099: is in fact incorrect.
1100: 
1101: \begin{table}
1102: \caption{Probability levels for the rule-of-thumb consistency condition}
1103: \label{Table_P1}
1104: \centering
1105: \begin{tabular}{c|c c}
1106: \hline\hline
1107: \noalign{\smallskip}
1108: $n_i -\nu$ & $P_1 = P [\chi^2 > n_i-\nu]$ & 
1109: $P_2 = P [\chi^2 > 4 (n_i-\nu)]$ \\
1110: \noalign{\smallskip}
1111: \hline
1112: \noalign{\smallskip}
1113: 1 &  0.3173 & 0.0455 \\
1114: 2 &  0.3679 & 0.0183 \\
1115: 3 &  0.3916 & 0.0074 \\
1116: 4 &  0.4060 & 0.0030 \\
1117: 5 &  0.4159 & 0.0012 \\
1118: \noalign{\smallskip}
1119: \hline
1120: \end{tabular}
1121: \end{table}
1122: 
1123: The bottom line is that the standard rule of thumb is way too stringent,
1124: while the rule of thumb with twice the original error bars is way too easy.
1125: In contrast, the more rigorous criterion with a probability level set to 
1126: $P_0=0.1$, intermediate between $P_1$ and $P_2$, provides a reasonable 
1127: trade-off.
1128: The latter criterion is also more trustworthy, insofar as all the rings 
1129: for all the models are tested with the same probability level.
1130: 
1131: The results of the present study suggest that the true large-scale 
1132: magnetic field in our Galaxy has a more complex configuration than 
1133: a strictly axisymmetric or bisymmetric field.
1134: One possibility is that it consists of the superposition of axisymmetric,
1135: bisymmetric and probably higher-order azimuthal modes.
1136: Such combinations of modes have been observed in a number of external
1137: galaxies \citep[e.g.,][]{Beck96,Berkhuijsen97,Rohde99,Beck07}.
1138: 
1139: 
1140: %----------  acknowledgements   ----------------------------
1141: \begin{acknowledgements}
1142: We thank the referee for his/her valuable comments as well as 
1143: Pierre Jean and William Gillard for useful discussions.
1144: JLH and MH were supported by the National Natural Science Foundation
1145: (NNSF) of China (10521001 and 10773016) and the National Key Basic
1146: Research Science Foundation of China (2007CB815403). MH was also
1147: supported by the French Embassy in Beijing during her stay in Toulouse
1148: in the framework of a bilateral co-supervised PhD program.
1149: \end{acknowledgements}
1150: 
1151: %----------  thebibliography   ----------------------------
1152: 
1153: \begin{thebibliography}{}
1154: 
1155: 
1156:   \bibitem[Beck(2007)]{Beck07}
1157: 	Beck, R. 2007, \aa~470, 539
1158: 
1159:   \bibitem[Beck et al.(1996)]{Beck96}
1160: 	Beck, R., Brandenburg, A., Moss, D., et al. 1996, ARA\&A~34, 155
1161: 
1162:   \bibitem[Beck et al.(2003)]{Beck03}
1163:         Beck, R., Shukurov, A., Sokoloff, D., \& Wielebinski, R. 2003, \aa~411, 99
1164: 
1165:   \bibitem[Berkhuijsen et al.(1997)]{Berkhuijsen97}
1166: 	Berkhuijsen, E. M., Horellou, C., Krause, M., et al. 1997, \aa~318, 700
1167: 
1168:   \bibitem[Cordes \& Lazio(2002)]{NE2001}
1169: 	Cordes, J. M., \& Lazio, T. J. W. 2002, astro-ph/0207156v3
1170: 
1171:   \bibitem[Crawford et al.(2001)]{Crawford01}
1172: 	Crawford, F., Manchester, R. N., \& Kaspi, V. M. 2001, \aj~122, 2001
1173: 
1174:   \bibitem[Ferri\` ere \& Schmitt(2000)]{Ferriere00}
1175: 	Ferri\` ere, K., \& Schmitt, D. 2000, \aa~358, 125
1176: 
1177:   \bibitem[Hamilton \& Lyne(1987)]{Hamilton87}
1178: 	Hamilton, P. A., \&  Lyne, A. G. 1987, \mnras~224, 1073
1179: 
1180:   \bibitem[Han et al.(2006)]{Han06}
1181: 	Han, J. L., Manchester, R. N., Lyne, A. G., et al.  2006, \apj~642, 868
1182: 
1183:   \bibitem[Han et al.(1999)]{HMQ99}
1184: 	Han, J. L., Manchester, R. N., \& Qiao, G. J.  1999,  \mnras~306, 37
1185: 
1186:   \bibitem[Han \& Qiao(1994)]{HanQiao94}
1187: 	Han, J. L., \& Qiao, G. J. 1994, \aa~288, 759
1188: 
1189:   \bibitem[Howard \& Kulsrud(1997)]{Howard97}
1190: 	Howard, A. M., \& Kulsrud, R. M. 1997, \apj~483, 648
1191: 
1192:   \bibitem[Indrani \& Deshpande(1999)]{ID99}
1193: 	Indrani, C., \& Deshpande, A. A. 1999, NewA~4, 33
1194: 
1195:   \bibitem[Isobe et al.(1990)]{Isobe90}
1196: 	Isobe,  T., Feigelson,  E. D., Akritas,  M. G., et al. 1990, \apj~364, 104
1197: 
1198:   \bibitem[Manchester et al.(2005)]{Manchester05}
1199: 	Manchester, R. N., Hobbs, G. B., Teoh, A., et al. 2005, AJ~129, 1993
1200: 
1201:   \bibitem[Mitra et al.(2003)]{Mitra03}
1202: 	Mitra, D., Wielebinski, R., Kramer, M., et al. 2003, \aa~398, 993
1203: 
1204:   \bibitem[Moss(1995)]{moss95}
1205:    Moss, D. 1995, \mnras~275, 191
1206: 
1207:  \bibitem[Moss(1996)]{moss96}
1208:    Moss, D. 1996, \aa~315, 63
1209: 
1210:   \bibitem[Noutsos et al.(2008)]{NJKK08}
1211:    Noutsos, A., Johnston, S., Kramer, M., \&  Karastergiou A. 2008, astroph-0803.0677
1212: 
1213: 
1214:   \bibitem[Poezd et al.(1993)]{PSS93}
1215: 	Poezd, A., Shukurov, A., \& Sokoloff, D. 1993, \mnras~264, 285
1216: 
1217:   \bibitem[Press et al.(1992)]{ptvf92}Press, W. H., Teukolsky, S. A., 
1218:         Vetterling, W. T., \& Flannery, B. P. 1992,
1219:         {\it Numerical Recipes FORTRAN: The Art of Scientific Computing}, 2nd ed.,
1220:          Cambridge: Cambridge Univ. Press
1221: 
1222:   \bibitem[Qiao et al.(1995)]{Qiao95}
1223: 	Qiao, G. J., Manchester, R. N., Lyne, A. G., et al. 1995, \mnras~274, 572
1224: 
1225:   \bibitem[Rand \& Kulkarni(1989)]{RK89}
1226: 	Rand, R. J., \& Kulkarni, S. R. 1989, \apj~343, 760
1227: 
1228:   \bibitem[Rand \& Lyne(1994)]{RL94}
1229: 	Rand, R. J., \&  Lyne, A. G. 1994, \mnras~268, 497
1230: 
1231:   \bibitem[Rohde et al.(1999)]{Rohde99}
1232: 	Rohde, R., Beck, R., \& Elstner, D. 1999, \aa~350, 423
1233: 
1234:   \bibitem[Ruzmaikin et al.(1985)]{RSS85}
1235: 	Ruzmaikin, A. A., Sokolov, D. D., \& Shukurov, A. M. 1985, \aa~148, 335
1236: 
1237:   \bibitem[Simard-Normandin \& Kronberg(1980)]{SK80}
1238: 	Simard-Normandin, M., \& Kronberg, P. P. 1980, \apj~242, 74
1239: 
1240:   \bibitem[Vall\'ee(1991)]{vallee91}
1241: 	Vall\'ee, J. P. 1991, \apj~366, 450
1242: 
1243:   \bibitem[Vall\'ee(1996)]{vallee96}
1244: 	Vall\'ee, J. P. 1996, \aa~308, 433
1245: 
1246:   \bibitem[Vall\'ee(2005)]{vallee05}
1247: 	Vall\'ee, J. P. 2005, \apj~619, 297
1248: 
1249:   \bibitem[van Ommen et al.(1997)]{van97}
1250: 	van Ommen, T. D., D'Alesssandro, F. D., Hamilton, P. A., et al. 1997, \mnras~287, 307
1251: 
1252:   \bibitem[Weisberg et al.(2004)]{Weisberg04}
1253: 	Weisberg, J. M., Cordes, J. M., Kuan, B., et al. 2004,  \apjs~150, 317
1254: 
1255:   \bibitem[Yamane(1964)]{Yamane64} Yamane, T. 1964, {\it Statistics, an introductory Analysis}, 
1256:         Eds. Harper \& Row, New York
1257: 
1258: \end{thebibliography}
1259: 
1260: \end{document}
1261: