0810.2303/ms.tex
1: \documentclass[revtex]{emulateapj}
2: \journalinfo{Accepted for publication in The Astrophysical Journal}
3: %\usepackage{graphicx}
4: 
5: \newcommand{\HI}{{\sc H\,i}}
6: \newcommand{\HII}{{\sc H\,ii}}
7: \newcommand{\kms}{\rm km\ s^{-1}}
8: \newcommand{\thetaB}{\theta_{\rm B}}
9: \newcommand{\thetaopt}{\theta_{\rm opt}}
10: 
11: 
12: \shorttitle{Integrated Polarization of Spiral Galaxies}
13: \shortauthors{Stil et\ al.}
14: 
15: \begin{document}
16: 
17: \title{The integrated polarization of spiral galaxy disks}
18: 
19: 
20: \author{Jeroen M. Stil}
21: \affil{Centre for Radio Astronomy, University of Calgary}
22: \author{Marita Krause and Rainer Beck} 
23: \affil{Max Planck Institut f\"ur Radio Astronomie, Auf dem H\"ugel, Bonn, Germany}
24: \author{A. Russell Taylor}
25: \affil{Centre for Radio Astronomy, University of Calgary}
26: 
27: 
28: \begin{abstract}
29: We present integrated polarization properties of nearby spiral
30: galaxies at 4.8 GHz, and models for the integrated polarization of 
31: spiral galaxy disks as a function of inclination.  Spiral galaxies in
32: our sample have observed integrated fractional polarization in the
33: range $\lesssim 1\%$ to $17.6\%$. At inclinations less than 50
34: degrees, the fractional polarization depends mostly on the ratio of
35: random to regular magnetic field strength. At higher inclinations,
36: Faraday depolarization associated with the regular magnetic field
37: becomes more important.  The observed degree of polarization is lower
38: ($< 4\%$) for more luminous galaxies, in particular those with
39: $L_{4.8} > 2 \times 10^{21}$ W Hz$^{-1}$. The polarization angle of
40: the integrated emission is aligned with the apparent minor axis of the
41: disk for galaxies without a bar. In our axially symmetric models, the
42: polarization angle of the integrated emission is independent of
43: wavelength. Simulated distributions of fractional polarization for
44: randomly oriented spiral galaxies at 4.8 GHz and 1.4 GHz are
45: presented. We conclude that polarization measurements, e.g. with the
46: SKA, of unresolved spiral galaxies allow statistical studies of the
47: magnetic field in disk galaxies using large samples in the local
48: universe and at high redshift. As these galaxies behave as idealized
49: background sources without internal Faraday rotation, they can be used
50: to detect large-scale magnetic fields in the intergalactic medium.
51: \end{abstract}
52: 
53: \keywords{galaxies: magnetic fields --- galaxies: spiral --- radio continuum: galaxies --- polarization }
54: 
55: 
56: \section{Introduction}
57: 
58: 
59: Large-scale regular magnetic fields have been detected by Faraday
60: rotation of radiation of polarized background sources in the Milky Way
61: \citep{han1999,brown2003,brown2007}, and the nearest galaxies
62: \citep{stepanov2008,arshakian2007}, such as M31 \citep{han1998}, and
63: the Large Magellanic Cloud \citep{gaensler2005}. The limiting factor
64: for this work is the number density of bright polarized background
65: sources. A few galaxies in the Local Group (M31, LMC) have angular
66: sizes large enough to probe their magnetic field structure by rotation
67: measures of polarized background sources with current radio
68: telescopes. Magnetic fields in galaxies outside the Local Group have
69: been studied mainly with imaging polarimetry at radio wavelengths
70: (Beck 2005 for a review). These observations indicate a large-scale
71: azimuthal magnetic field in the plane of the disk of spiral galaxies,
72: normally oriented in a spiral pattern with a pitch angle similar to
73: the optical spiral arms. High degrees of polarization have been
74: observed in inter-arm regions of some galaxies, such as NGC~6946
75: \citep{beck2007}. If a magnetic field is observed in a barred galaxy,
76: it is aligned with the bar inside the region of the bar, with a spiral
77: pattern outside the bar region \citep{beck2002}. However, some barred
78: galaxies show no detectable polarized emission or polarized emission
79: along the bar.
80: 
81: Statistical information on magnetic fields in galaxies is required to
82: study the cosmic evolution of these magnetic fields by observing
83: galaxies over a range of redshift. Polarimetric imaging of a large
84: sample of spiral galaxies {\small in the local universe} is impractical with
85: current instrumentation because of the high sensitivity and angular
86: resolution required. On the other hand, deep polarization surveys of
87: the sky are underway or being planned that will have the sensitivity
88: to detect spiral galaxies out to large distances, but not resolve
89: these galaxies.  The first deep wide-area polarimetry survey with the
90: sensitivity to detect spiral galaxies is
91: GALFACTS\footnote{http://www.ucalgary.ca/ras/GALFACTS} at 1.4 GHz,
92: 5$\sigma$ point source sensitivity of 400$\mu$Jy, and resolution
93: $3\arcmin$. SKA pathfinder telescopes such as ASKAP
94: \citep{johnston2007} aim to do all-sky polarimetry surveys at 1.4 GHz
95: with an instantaneous band width of 300 MHz with thousands of spectral
96: channels and a $5\sigma$ point source sensitivity around $50\ \rm \mu
97: Jy$ \citep{johnston2007}. ASKAP will be able to detect a few thousand
98: spiral galaxies at sufficient signal to noise ratio to detect
99: polarization at the few percent level, but the expected angular
100: resolution of $8\arcsec$ to $\sim 1\arcmin$ will not resolve these
101: galaxies. This raises the question what can be learned about magnetic
102: fields from polarimetry of large samples of unresolved spiral
103: galaxies.
104: 
105: Observations of polarization from spiral galaxies at redshifts up to 1
106: or higher will require the Square Kilometre
107: Array\footnote{http://www.skatelescope.org} (SKA), and even then it
108: will be difficult to make resolved polarization maps of these
109: high-redshift galaxies. The origin and evolution of cosmic magnetism
110: is one of five key science drivers for the SKA
111: \citep{gaensler2004}. For the evolution of magnetic fields in spiral
112: galaxies as a function of redshift, polarization of the integrated
113: emission is most readily observable, and a comparison with unresolved
114: galaxies at low redshift is relatively straightforward.
115: 
116: The number density of faint polarized radio sources in the sky
117: determines the unknown confusion limit in polarization, and is an
118: important parameter for rotation measure studies. Below 1 mJy total
119: flux density, star forming galaxies become an increasingly important
120: fraction of all radio sources. Knowledge of the polarization
121: properties of these galaxies is required to make realistic predictions
122: of the number of polarized sources at $\rm \mu Jy$ levels in polarized
123: flux density. Spiral galaxies may become an important fraction of the
124: polarized 1.4 GHz radio source population at $\mu$Jy flux densities
125: \citep{stil2007a}.
126: 
127: This paper presents observations and models that show that unresolved
128: spiral galaxies have detectable polarized emission, sometimes more
129: than 10\% polarized at 4.8 GHz, and that statistical information on
130: magnetic fields and internal Faraday rotation can be derived from
131: integrated polarization measurements of spiral galaxies.
132: 
133: \section{Observations}
134: \label{data-sec}
135: 
136: \begin{deluxetable*}{lrccrcrrrc}
137: \tablecolumns{10}
138: \tablewidth{0pc} 
139: \tabletypesize{\scriptsize}
140: \tablecaption{ Integrated polarization of nearby spiral galaxies }
141: \tablehead{
142: \colhead{Source} & \colhead{Type} & \colhead{$i$} & \colhead{$\nu$} & \colhead{$S_{\nu}$} & \colhead{$S_{\nu}(10\ \rm Mpc)^{a}$} & \colhead{$\Pi_0$} & \colhead{$\thetaB$} & \colhead{$\thetaopt$}  & \colhead{Tel.$^{b}$} \\
143: \colhead{\ } & \colhead{\ }  &\colhead{($\degr$)}& \colhead{(GHz)} &\colhead{(mJy)} & \colhead{(mJy)} & \colhead{(\%)}  & \colhead{($\degr$)} & \colhead{($\degr$)} &   
144: }
145: \startdata
146: IC 342    & SAB(rs)cd   &   25 &  4.8  & 856   &   82 & 2.2  $\pm$ 0.1 &     50 $\pm$ 1 &      39   &    E        \\
147: M 31      & SA(s)b      &   76 &  4.8  &1863   &    9 & 17.6 $\pm$ 3.5 &     68 $\pm$ 4 &      38   &    E        \\
148: M 33      & SA(s)cd     &   56 &  4.8  &1539   &   11 & 1.4  $\pm$ 0.4 &     77 $\pm$ 6 &      23   &    E        \\
149: M 51      & SAbc        &   20 &  4.8  & 380   &  358 & 2.8  $\pm$ 0.3 &  $-$46 $\pm$ 2 &   $-$10   &    E+V      \\
150: M 81      & SA(s)ab     &   59 &  4.8  & 385   &   41 & 7.3  $\pm$ 0.3 &  $-$36 $\pm$ 1 &   $-$28   &    E        \\
151: M 83      & SAB(s)c     &   24 &  4.8  & 809   &  111 & 2.5  $\pm$ 0.4 &      9 $\pm$ 4 &      45   &    E+V      \\
152: NGC 253   & SAB(s)c     &   78 &  4.8  &2707   &   82 & 1.9  $\pm$ 0.2 &     51 $\pm$ 2 &      52   &    E+V      \\
153: NGC 891   & SA(s)b? sp  &   88 &  4.8  & 286   &  264 & 1.6  $\pm$ 0.1 &      6 $\pm$ 1 &      23   &    E        \\
154: NGC 3628  & SAb pec sp  &   89 &  4.8  & 247   &  111 & 4.0  $\pm$ 0.1 &    105 $\pm$ 1 &     104   &    E        \\
155: NGC 4565  & SA(s)b? sp  &   86 &  4.8  &  54   &   85 & 10.5 $\pm$ 0.3 &  $-$43 $\pm$ 1 &   $-$44   &    E        \\ 
156: NGC 4736  & (R)SAB(rs)ab &   35 &  4.8  & 125   &   27 & 2.4  $\pm$ 0.4 &  $-$86 $\pm$ 3 &   $-$65   &   E+V      \\ 
157: NGC 5775  & SAbc        &   81 &  4.8  &  94   &  670 & 0.5  $\pm$ 0.2 &  $-$31 $\pm$ 9 &   $-$35   &    V        \\
158: NGC 5907  & Sc          &   87 &  4.8  &  72   &   87 & 2.2  $\pm$ 0.3 &   $-$6 $\pm$ 3 &   $-$25   &    E        \\
159: NGC 6946  & SAB(rs)cd   &   38 &  4.8  & 457   &  224 & 1.5  $\pm$ 0.5 &  $-$82 $\pm$ 7 &      60   &    E+V      \\
160: \enddata
161: \tablenotetext{a}{Flux density scaled to a distance of 10 Mpc.}
162: \tablenotetext{b}{Imaging telescope: E = Effelsberg 100m; E+V = Effelsberg 100m combined with the Very Large Array(VLA).}
163: \tablenotetext{\ }{References: IC~342: \citet{grave1988}, M~31: \citet{berkhuijsen2003}, M~33: \citet{tabatabaei2007}, 
164: M~51: \citet{fletcher2004}, M~81: \citet{beck1985}, M~83: \citet{beck2007a} from Beck, Ehle \& Sukumar (unpubl.), 
165: NGC~253: \citet{heesen2008}, NGC~891, NGC~3628, and NGC~4565: \citet{dumke1998}, NGC~4736: \citet{chyzy_buta2008}, 
166: NGC~5775: \citet{tullmann2000}, NGC~5907: \citet{dumke2000}, NGC~6946: \citet{beck1996,beck2007}}
167: \label{nearby-tab}
168: \end{deluxetable*}
169: 
170: 
171: \begin{deluxetable*}{lrccrrrrc}
172: \tablecolumns{9}
173: \tablewidth{0pc} 
174: \tabletypesize{\scriptsize}
175: \tablecaption{ Integrated polarization of Virgo galaxies }
176: \tablehead{
177: \colhead{Source} & \colhead{Type} & \colhead{$i$} & \colhead{$\nu$} & \colhead{$S_{\nu}$} & \colhead{$\Pi_0$} & 
178: \colhead{$\thetaB$} & \colhead{$\thetaopt$}  & \colhead{Tel.$^{a}$} \\
179: \colhead{\ } & \colhead{\ }  &\colhead{($\degr$)}& \colhead{(GHz)} &\colhead{(mJy)} & \colhead{(\%)}  & \colhead{($\degr$)} & \colhead{($\degr$)} &   
180: }
181: \startdata
182: NGC 4192  & SABb        &  78    &  8.4  &  16 & 14.5 $\pm$ 1.0 &  $-$28 $\pm$ 1  &     $-$27  &   E      \\
183:           &             &        &  4.8  &  38 & 11.9 $\pm$ 0.5 &  $-$21 $\pm$ 1  &            &   E      \\
184: NGC 4254  & SA(s)c      &  42    &  4.8  & 167 &  1.4 $\pm$ 0.1 &  $-46$ $\pm$ 1  &      68    &   E+V    \\
185: NGC 4302  & SAc         &  90    &  8.4  &  22 &  5.8 $\pm$ 1.0 &  $-$13 $\pm$ 4  &     $-$2   &   E      \\
186:           &             &        &  4.8  &  36 &  8.4 $\pm$ 0.7 &   $-$8 $\pm$ 1  &            &   E      \\ 
187: NGC 4303  & SABbc       &  18    &  8.4  & 107 &  1.5 $\pm$ 0.3 &  $-$13 $\pm$ 4  &      $-$28 &   E      \\
188:           &             &        &  4.8  & 189 &  0.7 $\pm$ 0.1 &  $-$65 $\pm$ 4  &            &   E      \\
189: NGC 4321  & SAB         &  30    &  8.4  &  66 &  2.3 $\pm$ 0.4 &  $-$58 $\pm$ 4  &      $-$60 &   E      \\
190:           &             &        &  4.8  & 109 &  1.0 $\pm$ 0.2 &  $-$52 $\pm$ 4  &            &   E      \\
191: NGC 4388  & SAb         &  82    &  4.8  &  74 &  1.1 $\pm$ 0.1 &  $-$60 $\pm$ 3  &      $-$89 &   E      \\
192: NGC 4501  & SAb         &  61    &  4.8  & 115 &  1.0 $\pm$ 0.1 &  $-$27 $\pm$ 3  &      $-$42 &   E      \\
193: NGC 4535  & SABc        &  44    &  8.4  &  20 &  6.7 $\pm$ 1.1 &     17 $\pm$ 3  &      0     &   E      \\
194:           &             &        &  4.8  &  38 &  8.0 $\pm$ 0.4 &     15 $\pm$ 1  &            &   E      \\
195: NGC 4654  & SABc        &  56    &  4.8  &  51 &  3.3 $\pm$ 0.2 &  $-$47 $\pm$ 1  &     $-$59  &   E      \\
196: \enddata
197: \tablenotetext{a}{Imaging telescope: E = Effelsberg 100m; E+V = Effelsberg 100m combined with the Very Large Array(VLA).}
198: \tablenotetext{\ }{References: NGC~4192, NGC4302, NGC4303:  unpublished images courtesy M. We\.zgowiec and M. Urbanik,
199: NGC~4254: \citet{chyzy2007}, NGC~4388, NGC~4501, NGC~4535, and NGC~4654: \citet{wezgowiec2007}.}
200: \label{virgo-tab}
201: \end{deluxetable*}
202: 
203: \begin{deluxetable*}{lrccrcrrrc}
204: \tablecolumns{10}
205: \tablewidth{0pc} 
206: \tabletypesize{\scriptsize}
207: \tablecaption{ Integrated polarization of barred galaxies }
208: \tablehead{
209: \colhead{Source} & \colhead{Type} & \colhead{$i$} & \colhead{$\nu$} & \colhead{$S_{\nu}$} & \colhead{$S_{\nu}(10\ \rm Mpc)$} & \colhead{$\Pi_0^{a}$} & \colhead{$\thetaB$} & \colhead{$\thetaopt$}  & \colhead{Tel.$^{b}$} \\
210: \colhead{\ } & \colhead{\ }  &\colhead{($\degr$)}& \colhead{(GHz)} &\colhead{(mJy)} & \colhead{(mJy)} & \colhead{(\%)}  & \colhead{($\degr$)} & \colhead{($\degr$)} &  
211: }
212: \startdata
213: NGC 1097  & SBbc(rs)    &   45 &  4.8  & 155 & 397 & (0.7 $\pm$ 0.8) &  \ldots &      $-$45  &  V       \\
214: NGC 1300  & SBb(s)      &   35 &  4.8  &  10 &  38 & 7.6 $\pm$ 2.0\phantom{)} &     89 $\pm$ 6 &      86   &    V       \\
215: NGC 1313  & SBc(s)      &   38 &  4.8  &  22 &   4 & (0.7 $\pm$ 1.3) &   \ldots &        $-$10  &  A      \\
216: NGC 1365  & SBb(s)      &   40 &  4.8  & 205 & 740 & (0.3 $\pm$ 0.2) &    \ldots &        40   &    V       \\ 
217: NGC 1433  & SBb(s)      &   27 &  4.8  &   2 &   3 & (4.7 $\pm$ 3.1) &    \ldots &      17   &    A      \\
218: NGC 1493  & SBc(rs)     &   30 &  4.8  &   2 &   3 & 13.6 $\pm$ 3.5\phantom{)} &     29 $\pm$ 5 &       \ldots$^{c}$   &    A      \\
219: NGC 1559  & SBc(s)      &   55 &  4.8  & 105 & 235 &  1.9 $\pm$ 0.1\phantom{)} &  $-$50 $\pm$ 1 &      $-$115 &  A      \\
220: NGC 1672  & SBb(rs)     &   39 &  4.8  & 105 & 236 &  1.5 $\pm$ 0.2\phantom{)} &  $-$55 $\pm$ 3 &   $-$10  &  A      \\
221: NGC 2336  & SBbc(r)     &   59 &  4.8  &   4 &  35 &  (3.6 $\pm$ 3.5) &  \ldots &      $-$2   &  V       \\
222: NGC 2442  & SBbc(rs)    &   24 &  4.8  &  71 & 182 &  1.1 $\pm$ 0.5\phantom{)} &     51 $\pm$ 9 &      40   &    A      \\
223: NGC 3059  & SBc(s)      &   27 &  4.8  &  29 &  56 &  (0.8 $\pm$ 0.5) &     \ldots &     \ldots$^{c}$   &    A      \\
224: NGC 3359  & SBc(s)      &   55 &  4.8  &  13 &  29 &  (0.7 $\pm$ 1.4) &     \ldots &      $-$10   & V       \\
225: NGC 3953  & SBbc(r)     &   61 &  4.8  &   6 &  14 &  8.5 $\pm$ 2.7\phantom{)} &     50 $\pm$ 6 &      13   &    V       \\
226: NGC 3992  & SBb(rs)     &   59 &  4.8  &   4 &   8 & 14.9 $\pm$ 4.8\phantom{)} &     25 $\pm$ 7 &      67   &    V       \\
227: NGC 4535  & SBc(s)      &   26 &  4.8  &  12 &  30 &  6.4 $\pm$ 1.4\phantom{)} &      8 $\pm$ 4 &      28   &    V       \\
228: NGC 4631  & SB(s)d      &   85 &  4.8  & 476 &  268 & 1.1 $\pm$ 0.1\phantom{)} &     16 $\pm$ 2 &      86   &    E        \\  % From nearby sample
229: NGC 5068  & SBc(s)      &   29 &  4.8  &  14 &   7 &  (1.8 $\pm$ 1.6) &  \ldots &      $-$20&    V       \\
230: NGC 5643  & SBc(s)      &   30 &  4.8  &  59 & 116 &  1.0 $\pm$ 0.2\phantom{)} &  $-$68 $\pm$ 4 &      \ldots $^{c}$   &    A      \\
231: NGC 7479  & SBbc(s)     &   45 &  4.8  &  32 & 365 &  3.9 $\pm$ 0.6\phantom{)} &   $-$9 $\pm$ 3 &      25   &    V       \\
232: NGC 7552  & SBbc(s)     &   31 &  4.8  & 110 & 485 &  1.2 $\pm$ 0.2\phantom{)} &  $-$87 $\pm$ 4 &       1    &    A      \\
233: \enddata
234: \tablenotetext{a}{Galaxies with $\Pi_0$ less than 2 times the formal error listed in parentheses; these galaxies are not included in figures}
235: \tablenotetext{b}{Imaging telescope: A = Autralia Telescope Compact Array (ATCA), E =  Effelsberg 100 m, V = VLA.}
236: \tablenotetext{c}{Position angle of major axis undetermined for this low-inclination galaxy.}
237: \tablenotetext{\ } {References: NGC~4631: Krause et al. (unpubl.) but see \citet{krause2003} and \citet{golla1994}; all other galaxies from \citet{beck2002}.}
238: \label{barred-tab}
239: \end{deluxetable*}
240: 
241: 
242: We use a compilation of archival data for nearby galaxies imaged in
243: polarized radio emission.  Observations at 4.8 GHz are most widely
244: available, so this paper focuses at this frequency to maximize the
245: sample size for a single frequency. Additional observations at 8.4 GHz
246: are available for five galaxies. These are listed alongside the 4.8
247: GHz data.  Three sub samples are distinguished in this paper. The first
248: sub sample is a set of 14 nearby ``field'' galaxies listed in
249: Table~\ref{nearby-tab}. The second sub sample is a set of 9 spiral
250: galaxies in the Virgo cluster from \citet{wezgowiec2007}, We\.zgowiec
251: et~al. (in prep.), and \citet{chyzy2007} listed in
252: Table~\ref{virgo-tab}. These galaxies exist in a cluster environment,
253: with a higher probability of interaction with other cluster members
254: and the intracluster gas. The third sub sample of 20 barred spiral
255: galaxies listed in Table~\ref{barred-tab} allows us to explore the
256: integrated polarization of barred galaxies. One nearby galaxy
257: (NGC~4631, classified as SB) was included in Table~\ref{barred-tab}.
258: 
259: The nearest galaxies in the sample and the galaxies in the Virgo
260: cluster were observed with the 100m-Effelsberg telescope, sometimes in
261: combination with the VLA. These galaxies are not affected by missing
262: short spacings in the interferometer. The barred galaxies were not
263: observed with the Effelsberg telescope, and the images could in
264: principle lack large scale structure resolved by the shortest
265: baselines of the interferometer. A comparison of 6 cm flux densities
266: derived by \citet{beck2002} with published integrated flux densities
267: showed that NGC~1313 and NGC~3992 in Table~\ref{barred-tab} may be
268: affected by missing short spacings, because published flux densities
269: listed in the NASA Extragalactic Database (NED) are substantially
270: higher than the 6-cm flux density in \citet{beck2002}. If no 6 cm flux
271: density was published, we compared the 21 cm flux density corrected
272: for a spectral index 0.7 - 1 with the flux density in
273: \citet{beck2002}. Of the two barred galaxies with suspected missing
274: flux in the interferometer image, NGC~1313 showed no significant
275: integrated polarization, while NGC~3992 has a high degree of
276: polarization of $14.9 \pm 4.8\ \%$. Most of the emission of NGC~3992
277: detected in the interferometer image originates outside the bar region
278: \citep{beck2002}.
279: 
280: 
281: \begin{figure*}
282: \resizebox{\textwidth}{!}{\includegraphics[angle=0]{f1.eps}}
283: \caption{ (a) Integrated fractional polarization at 4.8 GHz of nearby
284: spiral galaxies from Table~\ref{nearby-tab} (filled circles) and Virgo
285: cluster spirals from Table~\ref{virgo-tab} (triangles), with model
286: curves described in Section~\ref{model-sec}. The model curves from top
287: to bottom correspond with models 1--9 in Table~\ref{model_par-tab}, in
288: that order. (b) The same as (a) for the detected barred galaxies from
289: Table~\ref{barred-tab} (squares). The model curves in (b) are shown
290: for reference only.
291: \label{diskpol_model-fig}
292: }  
293: \end{figure*}
294: 
295: 
296: \begin{figure*}
297: \resizebox{8.5cm}{!}{\includegraphics[angle=0]{f2a.eps}}
298: \resizebox{8.5cm}{!}{\includegraphics[angle=0]{f2b.eps}}
299: \caption{ (a) Correlation between $\thetaopt$ and $\thetaB$ for
300: galaxies in Table~\ref{nearby-tab} (filled circles) and the Virgo
301: galaxies in Table~\ref{virgo-tab} (triangles).  The symbol size
302: increases linearly with $\sin(i)$. (b) The same as (a) for the barred
303: galaxies in Table~\ref{barred-tab} (squares).
304: \label{PA-fig}
305: }  
306: \end{figure*}
307: 
308: \begin{figure*}
309: \resizebox{8.5cm}{!}{\includegraphics[angle=0]{f3a.eps}}
310: \resizebox{8.5cm}{!}{\includegraphics[angle=0]{f3b.eps}}
311: \caption{ Integrated fractional polarization as a function of distance
312: (a) and monochromatic 4.8 GHz luminosity (b) for the entire
313: sample. The symbols are the same as in Figure~\ref{PA-fig}. The
314: symbols for the Virgo galaxies NGC~4321 and NGC~4501 overlap
315: completely in (a), and partially in (b).
316: \label{PI_L-fig}
317: }  
318: \end{figure*}
319: 
320: 
321: 
322: 
323: In order to simulate unresolved observations from our maps of
324: nearby galaxies we integrated the Stokes I, Q, and U images of these
325: galaxies separately in ellipses. Their axial ratio was defined by the
326: outer isophotes in the Stokes $I$ image and the ellipses include the
327: entire galaxy up to the noise level of the Stokes I map. The
328: integrated $Q$ and $U$ flux densities were used to calculate the
329: integrated polarized flux density $S_p=\sqrt{(Q^2+U^2)}$, that was
330: corrected for polarization bias following \citet{simmons1985} to
331: obtain the intrinsic polarized flux density $S_{p,0}$. The correction
332: for polarization bias did not have a measurable effect for most
333: galaxies because of the high signal to noise ratio of the integrated
334: flux densities.  The integrated fractional polarization of each galaxy
335: was calculated as $\Pi_0 = S_{p,0}/S$, where $S$ is the total flux
336: density, including thermal and non-thermal emission. The polarization
337: angle $\theta_{\rm pol} = \onehalf \arctan(U/Q)$ of the integrated
338: emission is expected to be aligned with the apparent minor axis of the
339: optical galaxy.  The sensitivity of the observations is included in
340: the errors for $\Pi_0$ through the noise in the maps. The error
341: estimates also include the uncertainty in the zero level, that
342: dominates the uncertainty for galaxies with a large angular size such
343: as M~31. It is customary to list the position angle of the optical
344: major axis, and also to rotate the observed E-vectors by $90\degr$ to
345: obtain B-vectors in the plane of the sky.  We therefore define
346: $\thetaB = \theta_{\rm pol}+90\degr$ for comparison with published
347: position angles of the optical major axis, $\thetaopt$.  The results
348: are listed in Tables~\ref{nearby-tab}, \ref{virgo-tab}, and
349: \ref{barred-tab}.
350: 
351: Bandwidth depolarization is negligible in these data. For a
352: rotation measure as high as 300 rad m$^{-2}$, the polarized intensity
353: is decreased by less than 1\%. The integrated polarized flux densities
354: are not affected by significant residual instrumental
355: polarization. Small values for the integrated fractional polarization
356: are the result of integrating significant flux in Stokes Q and U, with
357: nearly equal positive and negative contributions. The uncertainty in
358: the polarization angle of the integrated emission is typically less
359: than 5 degrees.
360: 
361: Some galaxies (M~31, M~81, NGC~4565 in Table~\ref{nearby-tab},
362: NGC~4192, NGC~4302, NGC~4535 in Table~\ref{virgo-tab}, NGC~1300,
363: NGC~1493, NGC~3953, NGC~3992 in Table~\ref{barred-tab}) have
364: integrated fractional polarization $\Pi_0 > 7\%$.  Five of these
365: galaxies (M~31, NGC~4565, NGC~4192, NGC~1493, and NGC~3992) are more
366: than $10\%$ polarized after integration over solid angle.
367: 
368: Five of the Virgo galaxies have additional observations at 8.4 GHz.
369: The fractional polarization at both frequencies agrees to within a few
370: percent for these galaxies. Small differences can be attributed to
371: statistical errors, mainly at 8.4 GHz where the signal to noise ratio
372: of the emission is lower. The differences in $\Pi_0$ at 4.8 GHz and at
373: 8.4 GHz per galaxy are much smaller than the differences between
374: galaxies. This confirms that the variation in $\Pi_0$ between galaxies
375: represents intrinsic differences between the galaxies. The
376: polarization angles at 4.8 GHz and 8.4 GHz are also mutually
377: consistent. In low-inclination galaxies, the polarization angle of the
378: integrated emission is more affected by asymmetries because the effect
379: of an axially symmetric component on the polarization vanishes mostly.
380: This can explain the difference in polarization angle for NGC~4303 ($i
381: = 18\degr$). The position angle of the major axis can also be affected
382: by ellipticity of the disk.
383: 
384: The fractional polarization of any particular galaxy depends on the
385: direction from which the galaxy is observed, as the projection of the
386: magnetic field on the plane of the sky and the path length of the line
387: of sight through the disk change. Any particular galaxy can only be
388: observed from one direction, but the variation of integrated
389: polarization properties with inclination can be revealed for a
390: sufficiently large sample of galaxies. The wide range in inclination
391: ($20\degr < i < 90\degr$) in the sample allows an effective analysis
392: of the integrated polarization properties of spiral galaxies as a
393: function of inclination. It should be noted that the sample is not
394: randomly distributed in inclination.  The galaxies in
395: Tables~\ref{nearby-tab}, \ref{virgo-tab}, and \ref{barred-tab} were
396: observed to study their magnetic field configuration through imaging
397: polarimetry.  The sample is biased towards low inclination (face-on)
398: and high-inclination (edge-on) galaxies because these were preferred
399: for imaging of the magnetic field. In particular, the sub sample of
400: barred spirals in Table~\ref{barred-tab} contains only galaxies with
401: inclination $i < 60\degr$, with the exception of the added nearby
402: galaxy NGC~4631.  Any sample of spiral galaxies with inclination
403: derived from the apparent axial ratio appears deficient in galaxies
404: with $i \lesssim 20\degr$ because the disks of most spiral galaxies
405: are not exactly circular
406: \citep{binney1981,grosbol1985,lambas1992}. This effect is also
407: apparent in the present sample, but it does not affect our analysis.
408: 
409: Figure~\ref{diskpol_model-fig} shows $\Pi_0$ as a function of
410: inclination. The curves in Figure~\ref{diskpol_model-fig} represent
411: models that will be discussed in Section~\ref{model-sec}.  The formal
412: errors on $\Pi_0$ are generally small because of the high signal to
413: noise ratio in the integrated flux densities.  Some galaxies with
414: large angular size have larger errors because of uncertainties in the
415: zero level of the maps.  The large spread in $\Pi_0$, $1\% \lesssim
416: \Pi_0 \lesssim 15\%$, seen for each subsample reflects intrinsic
417: differences. The highest values of $\Pi_0$ in
418: Figure~\ref{diskpol_model-fig}a are found in the inclination range
419: from $\sim 50\degr$ to $\sim 80\degr$.  Some low-inclination galaxies
420: with high $\Pi_0$ are found among the barred galaxies in
421: Figure~\ref{diskpol_model-fig}b. The model curves in
422: Figure~\ref{diskpol_model-fig}b are shown for reference only, as they
423: do not include a bar.
424: 
425: Figure~\ref{PA-fig}a shows the correlation between the position angle
426: of the major axis, $\thetaopt$, and $\thetaB$ for galaxies in
427: Table~\ref{nearby-tab} and Table~\ref{virgo-tab}. If polarization of
428: the integrated emission is the result of projection of the azimuthal
429: magnetic field in the disk on the plane of the sky, a correlation is
430: expected between $\thetaB$ and the line of nodes of the disk
431: intersecting the plane of the sky. The latter is determined by the
432: position angle of the major axis, $\thetaB$.  The most deviating
433: galaxies in Figure~\ref{PA-fig}a are NGC~6946 and NGC~4254, with
434: inclination $i = 38\degr$ and $i = 42\degr$ respectively. Large-scale
435: departures from axial symmetry are most likely to be noticed for
436: galaxies with a small inclination. The position angles used in
437: Figure~\ref{PA-fig} are also less well defined.  At lower
438: inclinations, the position angle of the major axis is more difficult
439: to measure because the isophotes are nearly circular.  Besides
440: measurement uncertainties, the position angle of the major axis is not
441: a good estimator for the position angle of the line of nodes of the
442: disk if the inclination approaches $26\degr$, the inclination for
443: which a circular disk would have an apparent axial ratio of 0.9,
444: similar to the mean intrinsic ellipticity of spiral galaxy disks found by
445: \citet{lambas1992}. Consequently, the uncertainty associated
446: $\thetaopt$ is substantial but hard to quantify for galaxies with
447: inclination less than $\sim 40\degr$. For galaxies with higher
448: inclinations, $\thetaopt$ is estimated to be accurate to within $\sim
449: 5\degr$. The uncertainty in $\thetaB$ is typically a few degrees, but
450: some additional scatter is expected because of differences in
451: structure between galaxies. The total scatter in Figure~\ref{PA-fig}a
452: for galaxies with inclination $\gtrsim 60\degr$ suggests that the
453: scatter introduced by structural differences is limited to $\pm
454: 15\degr$.
455: 
456: Figure~\ref{PA-fig}b shows the same relation for the detected barred
457: galaxies in Table~\ref{barred-tab}. The scatter between $\thetaopt$
458: and $\thetaB$ appears substantially larger than in
459: Figure~\ref{PA-fig}a.  This may be a result of magnetic fields aligned
460: with the bar, where the orientation of the bar adds a parameter to the
461: problem. In this paper we focus mainly on galaxies without (strong)
462: bars, while noting that unresolved barred galaxies can in fact also be
463: polarized radio sources.
464: 
465: 
466: Figure~\ref{PI_L-fig} shows the integrated fractional polarization
467: $\Pi_0$ as a function of distance, and monochromatic luminosity
468: $L_{4.8}$. There should be no correlation between integrated
469: polarization and distance, and none is found in the present sample.
470: Figure~\ref{PI_L-fig}b shows a correlation between $\Pi_0$ and
471: $L_{4.8}$.  There are no galaxies in the sample with $L_{4.8}> 2
472: \times 10^{21}$ W Hz$^{-1}$ that have $\Pi_0 > 4 \%$.  Most galaxies
473: with $L_{4.8}< 5 \times 10^{20}$ W Hz$^{-1}$ do have $\Pi_0 > 4
474: \%$. As shown in Figure~\ref{diskpol_model-fig}, a galaxy can have low
475: $\Pi_0$ because of strong Faraday depolarization, or because of its
476: inclination. There is no correlation between inclination and
477: luminosity in the sample. The lower integrated polarization of
478: luminous galaxies in Figure~\ref{PI_L-fig}b suggests stronger Faraday
479: depolarization in these galaxies, or a lower degree of uniformity of
480: the magnetic field as will be discussed in Section~\ref{discussion1}.
481: 
482: 
483: \section{Models of the integrated polarization}
484: \label{model-sec}
485: 
486: \subsection{Polarized intensity in the disk}
487: 
488: The principal cause of net polarization of unresolved spiral galaxies
489: is the regular azimuthal magnetic field in the disk projected on the
490: sky along the apparent major axis. Faraday rotation and depolarization
491: are affected by the path length through the disk, the line-of-sight
492: component of the regular magnetic field, and fluctuations in thermal
493: electron density and magnetic field strength.  A first analysis of the
494: integrated polarization of spiral galaxies is best served with the
495: simplest possible model that includes these factors.  We consider
496: axially symmetric models with a regular azimuthal magnetic field in
497: the disk, and an isotropic random component of the magnetic field.
498: 
499: Using this simple geometry, the integrated polarization of a spiral
500: galaxy is modeled by considering each solid angle element of the disk
501: as a slab with a specified line-of-sight component of the
502: regular magnetic field that depends on inclination and location in the
503: disk. \citet{sokoloff1998} provided solutions of polarized intensity
504: from a slab of Faraday-rotating medium mixed with relativistic
505: electrons that produce synchrotron emission. The polarized intensity
506: is expressed as a complex quantity $\mathcal{P} = Q + j U$
507: \citep{spangler1982}, where $Q$ and $U$ are Stokes parameters that
508: describe the linear polarization. Following the conventions adopted by
509: \citet{sokoloff1998}, $\mathcal{P}$, $Q$, and $U$ are normalized by
510: the total intensity of the synchrotron radiation, so that the modulus
511: of $\mathcal{P}$ is the fractional polarization. The polarization
512: angle of the complex polarization is $\Psi = \onehalf \arctan (U/Q)$.
513: 
514: At higher frequencies, a contribution from thermal emission to the
515: total intensity reduces the observed fractional polarization by a
516: factor ($1-f_{th}$), where $f_{th}$ is the fraction of the total
517: intensity generated by thermal emission. We include thermal emission
518: by setting $f_{th} = 0.23$ at 4.8 GHz
519: \citep{condon1990,condon1992,niklas1997}.  Determination of $f_{th}$
520: in individual galaxies is difficult because of the need to separate
521: thermal and non-thermal emission in the radio spectrum.  The rms
522: scatter in $f_{th}$ from galaxy to galaxy is believed to be at most a
523: factor of 2 \citep{condon1992}. This translates into a spread of $\sim
524: 25\%$ in ($1-f_{th}$) from the adopted value of 0.77.
525: 
526: \begin{deluxetable}{lrcccccccc}
527: \tablecolumns{10}
528: \tablewidth{0pc} 
529: \tabletypesize{\scriptsize}
530: \tablecaption{ Sensitivity to model parameters$^a$}
531: \tablehead{
532: \colhead{Inclination} & \colhead{$B$} & \colhead{$f_{\rm B}$}  & \colhead{$n_e$} & \colhead{$f_{i}$} & \colhead{$l_{\rm turb}$} & \colhead{$2h$}
533: } \startdata $i = 40\degr$ & 0.007 & 1.01 & 0.006 & 0.034 & 0.042 &
534: 0.035 \\ $i = 60\degr$ & 0.17 & 1.09 & 0.17 & 0.12 & 0.057 & 0.006 \\
535: $i = 75\degr$ & 0.50 & 0.83 & 0.51 & 0.45 & 0.055 & 0.45 \\ \enddata
536: \tablenotetext{a}{Numbers listed are normalized contributions to the
537: total scatter in a Monte Carlo simulation with 10000 realizations per
538: inclination per parameter, where all parameters were drawn independently from a
539: uniform distribution. Parameters with a small contribution to the
540: total scatter are relatively unimportant. Parameter ranges adopted in
541: this simulation: $1 < B < 10$ $\mu$G, $0.7 < f_{\rm B} < 4.0$, $0.01 <
542: n_e < 0.1$ $\rm cm^{-3}$, $0.1 < f_i < 1.0$, $10 < l_{turb} < 100$ pc,
543: $200 < 2h < 2000$ pc.}
544: \label{model_var-tab}
545: \end{deluxetable}
546: 
547: 
548: In this paper we adopt a solution for the polarized intensity from a
549: uniform slab of finite thickness along the line of sight from
550: \citet{sokoloff1998}. For a given inclination $i$, the component of
551: the regular magnetic field along the line of sight $B_\|$ and
552: perpendicular to the line of sight $B_\bot$ can be evaluated. The
553: uniformity of the magnetic field is expressed as the ratio of the
554: isotropic random magnetic field strength to the regular magnetic field
555: strength, defined as $f_B=\sigma_{\rm B}\sqrt{3}/B$, where
556: $\sigma_{\rm B}$ is the rms strength of the random component of the
557: magnetic field in one direction.
558: 
559: The complex polarization can be written as the product of
560: a wavelength-independent part that describes the polarization if
561: Faraday rotation effects can be neglected, and a wavelength-dependent
562: part that includes Faraday rotation effects. The
563: wavelength-independent part of the complex polarization is
564: \begin{equation}
565: \langle {\mathcal P_0} \rangle = p_i {B_\bot^2 \over B_\bot^2 + 2\sigma_{\rm B}^2 } \exp\bigl [ 2j ( \onehalf \pi + \arctan(B_y/B_x) ) \bigr ],
566: \label{P0-eq}
567: \end{equation}
568: where $p_i \approx 0.75$ is the intrinsic maximum polarization for
569: synchrotron emission in a regular magnetic field, and $B_x$ and $B_y$
570: are the components of the regular magnetic field along two orthogonal
571: axes in the plane of the sky. Equation~\ref{P0-eq} assumes a constant
572: spectral index of $-1$ for the synchrotron emission of a galaxy.  We
573: define the $x-axis$ along the apparent major axis of the disk. The
574: brackets $\langle \cdot \rangle$ indicate that the emission is
575: averaged over a volume in the disk, that is sufficiently narrow to keep
576: the regular magnetic field constant inside the volume, but extends
577: along the line of sight through the disk.  This is a reasonable
578: approximation except for very high inclinations $i \gtrsim 80\degr$,
579: where the azimuthal field must change along the line of sight, but
580: only if the line of sight is near the major axis.
581: 
582: Equation~\ref{P0-eq} includes the effect of decreased polarization
583: because of the presence of a random component of the magnetic field in
584: the plane of the sky. \citet{sokoloff1998} referred to this as
585: wavelength-independent depolarization. The term beam depolarization is
586: sometimes used in this situation, but this term is used also in case
587: of wavelength-dependent Faraday effects.  The plane of polarization is
588: perpendicular to the component of the regular magnetic field projected
589: on the plane of the sky.
590: 
591: 
592: \begin{figure}
593: \resizebox{\columnwidth}{!}{\includegraphics[angle=0]{f4.eps}}
594: \caption{ Representation of polarized intensity in an annulus of a
595: spiral galaxy disk for $i=40\degr$ (a), and $i=60\degr$ (b). Thin
596: lines represent the component of the magnetic field perpendicular to
597: the line of sight, $B_\bot$.  The pitch angle of the field is
598: $15\degr$ in this example. Thick lines represent the polarization
599: E-vectors, with length proportional to the polarized intensity. The
600: line-of-sight component of the magnetic field $B_\|$ is indicated by
601: open circles (towards the observer) and filled circles (away from the
602: observer) with size proportional to $B_\|$. The integrated fractional
603: polarization in (a) is $\Pi_0 = 6.1\%$, and in (b) $\Pi_0 = 13.1\%$,
604: including a thermal fraction $f_{th}=0.23$ at 4.8 GHz. The separation
605: of the positions shown in this figure does not represent the step size
606: of the integration.
607: \label{spiral_proj-fig}
608: }  
609: \end{figure}
610: 
611: 
612: 
613: 
614: The observed polarization $\langle {\mathcal P} \rangle$ differs from
615: $\langle {\mathcal P_0} \rangle$ because the polarized synchrotron
616: emission generated in a small volume element in the disk is Faraday
617: rotated as it travels through the disk to the observer. The effects of
618: Faraday rotation can be written as a wavelength-dependent factor in
619: the complex polarization, that includes an integral along the line of
620: sight.  \citet{burn1966} and \citet{sokoloff1998} evaluated this
621: integral analytically for a uniform slab.  The observed polarized
622: intensity from a small part of the disk is
623: \begin{equation}
624: \langle{\mathcal P} \rangle = \langle {\mathcal P_0} \rangle {1 - \exp(-S) \over S},
625: \label{P-eq}
626: \end{equation}
627: where $S = 2 \lambda^4 \sigma_{\rm RM}^2 -2 j \lambda^2 {\mathcal R}$
628: is a complex number that includes rotation measure fluctuations
629: $\sigma_{\rm RM}$ in the real part and the Faraday depth $\mathcal R$
630: in the imaginary part.
631: 
632: The real part of $S$ contains fluctuations in rotation measure along
633: the line of sight through the factor $\sigma_{\rm RM}$, which
634: represents the rms amplitude of rotation measure fluctuations after
635: integration along the line of sight. The plane of polarization makes a
636: random walk because of fluctuations in density and magnetic
637: field. When polarized emission is integrated over different lines of
638: sight, or different depths inside the source, depolarization of the
639: emission results.  This process is called internal Faraday dispersion.
640: Evaluation of $\sigma_{\rm RM}$ requires assumptions about the
641: correlation scale of fluctuations in density and magnetic field
642: strength. \citet{sokoloff1998} discuss the difficulties estimating a
643: correlation scale length, and assume that the scale length is the same
644: for fluctuations in density, magnetic field, and rotation measure. The
645: strength of the magnetic field in a turbulent cell defined by the
646: correlation length is $\sigma_{\rm B}$.  Rotation measure fluctuations
647: also depend on the path length $L$ through the disk with thickness $2
648: h$, observed at inclination $i$, given by $L = 2h / \cos i$.  The
649: line-of-sight filling factor of the Faraday-rotating medium is $f_i$.
650: Rotation measure fluctuations are then calculated following
651: \citet{sokoloff1998}
652: \begin{equation}
653: \sigma_{\rm RM} = 0.81 n_e \sigma_{\rm B} l_{\rm turb} \Bigl({ 2h f_i \over l_{\rm turb} \cos i } \Bigr)^{\onehalf}
654: \end{equation}
655: 
656: The imaginary part of $S$ describes rotation of the plane of
657: polarization by the line-of-sight component of the regular magnetic
658: field. The Faraday depth $\mathcal{R} = n_e f_i B_\| 2h/\cos i $ is
659: the product of the mean electron density ($n_e f_i$), where $n_e$ is
660: the electron density in a turbulent cell, $B_\|$ the line-of-sight
661: component of the regular magnetic field, and $h$ the path length
662: through the disk. Rotation of the plane of polarization also
663: depolarizes emission integrated along the line of sight as emission
664: from different depths of the source is rotated by different amounts.
665: This mechanism for depolarization is called differential Faraday
666: rotation.  It is sometimes called depth depolarization, but it should
667: be understood that there is a different depolarizing effect in the
668: form of Faraday dispersion that is also the result of integrating
669: emission along the line of sight.
670: 
671: 
672: \subsection{Integrated polarized emission}
673: 
674: \begin{deluxetable*}{lrrccccccc}
675: \tablecolumns{10}
676: \tablewidth{0pc} 
677: \tabletypesize{\scriptsize}
678: \tablecaption{ Model parameters for Figure~\ref{diskpol_model-fig}}
679: \tablehead{
680: \colhead{Model} & \colhead{$B$} & \colhead{$f_{\rm B}$}  & \colhead{$n_e$} & \colhead{$f_{i}$} & \colhead{$l_{\rm turb}$} & \colhead{$2h$} & \colhead{$f_{th}$
681: } & \colhead{$\Pi_{\rm max}$} & \colhead{$\Pi_{\rm med}$}  \\
682: \colhead{\ } & \colhead{($\rm \mu G$)} & \colhead{\ }  & \colhead{($\rm cm^{-3}$)} & \colhead{\ } & \colhead{(pc)} & \colhead{(pc)} & \colhead{\ }  & \colhead
683: {(\%)} & \colhead{(\%)}
684: }
685: \startdata
686: 1$^{a}$      & 5.0  & 1.0   &  0.03   &  0.5  & 50  &  1000  & 0.23  & 16.4 & 10.2 \\
687: 2$^{b}$      & 10.0  & 1.0   &  0.03   &  0.5  & 50  &  1000 & 0.23  & 13.7 &  8.2 \\
688: 3$^{c}$      & 5.0  & 1.0   &  0.09   &  0.5  & 50  &  1000  & 0.23  & 12.0 &  7.0 \\ \\
689: 4$^{a}$      & 5.0  & 2.0   &  0.03   &  0.5  & 50  &  1000  & 0.23  & 6.8  &  4.7 \\ 
690: 5$^{b}$      & 10.0  & 2.0   &  0.03   &  0.5  & 50  &  1000 & 0.23  & 5.6  &  3.7 \\
691: 6$^{c}$      & 5.0  & 2.0   &  0.09   &  0.5  & 50  &  1000  & 0.23  & 4.8  &  3.1 \\ \\
692: 7$^{a}$      & 5.0  & 3.0   &  0.03   &  0.5  & 50  &  1000  & 0.23  & 3.4  &  2.4 \\
693: 8$^{b}$      & 10.0  & 3.0   &  0.03   &  0.5  & 50  &  1000 & 0.23  & 2.7  &  1.8 \\
694: 9$^{c}$      & 5.0  & 3.0   &  0.09   &  0.5  & 50  &  1000  & 0.23  & 2.2  &  1.4 \\
695: \enddata
696: \tablenotetext{a}{Models indicated by a solid curve in Figures~\ref{diskpol_model-fig} and \ref{Pi_PDF-fig}}
697: \tablenotetext{b}{Models indicated by a dashed curve in Figures~\ref{diskpol_model-fig} and \ref{Pi_PDF-fig}}
698: \tablenotetext{c}{Models indicated by a dotted curve in Figures~\ref{diskpol_model-fig} and \ref{Pi_PDF-fig}}
699: \label{model_par-tab}
700: \end{deluxetable*}
701: 
702: Equation~\ref{P-eq} provides the complex polarized intensity
703: $\langle{\mathcal P} \rangle$ for every line of sight, from which we
704: obtain the local Stokes $Q$ and $U$ intensities, normalized by the
705: local total intensity that is presumed uniform in the current
706: models. The model Stokes $Q$ and $U$ intensities can be integrated
707: over the disk to predict the model polarized flux density, fractional
708: polarization $\Pi_0$, and the angle $\thetaB$ of an unresolved spiral
709: galaxy. The integration over the disk reduces to a one-dimensional
710: numerical integration over azimuthal angle for the symmetric models
711: under consideration.
712: 
713: Assuming values for the various model parameters, the integration over
714: the disk can be performed as a function of inclination.  The
715: integrated polarization is more sensitive to certain model parameters,
716: and the influence of model parameters can change with inclination. The
717: relative importance of the parameters in our model was investigated
718: through a Monte Carlo experiment. Each parameter was varied
719: separately, and the resulting range of the integrated polarization was
720: divided by the range obtained when all parameters were varied
721: independently at the same time. The spread in integrated fractional
722: polarization was taken as the difference between the upper quartile
723: and the lower quartile. As the underlying distributions are strongly
724: skewed, the quartiles provide a better indication of the spread in the
725: distribution than the standard deviation. The normalization with the
726: ``total'' scatter allows us to compare the importance of parameters
727: between the three inclinations. It should be noted that the non-linear
728: character of the model, and the opposing effects of certain parameters
729: can make the ratio of the normalized quartile ranges larger than
730: unity.
731: 
732: Table~\ref{model_var-tab} shows for three values of the inclination
733: the result of 10000 realizations of each parameter, drawn from a
734: uniform distribution. The magnetic field strength, electron density,
735: filling factor, turbulent length scale, and the vertical thickness of
736: the disk were varied by an order of magnitude, as indicated in the
737: footnote to Table~\ref{model_var-tab}. The ratio of random to regular
738: magnetic field strength ($f_{\rm B}$) was varied over a smaller range
739: (0.7 to 4.0). At low inclinations, the normalized quartile range for
740: variation of $f_{\rm B}$ is 1.01, while the normalized quartile ranges
741: of all other parameters are less than 0.1. Clearly, the ratio of
742: random to regular magnetic field dominates the fractional polarization
743: of the model for small inclinations. At $i=60\degr$, $f_{\rm B}$ is
744: still the most important parameter, but other parameters such as the
745: regular magnetic field strength and the electron density become more
746: important. At $i=75\degr$, $f_{\rm B}$ is still important, but the
747: parameters that define $\mathcal{R} = n_e f_i B_\| 2h/\cos i$ are
748: almost equally important. The relative importance of $n_e$, $f_i$,
749: $B$, and $2h$ is similar at $i=75\degr$, because they contribute
750: equally to $\mathcal R$, although subtle differences occur because of
751: the dependence on $\sigma_{\rm RM}$. The turbulent scale length
752: $l_{turb}$ is the parameter in our model that is least constrained by
753: observations.  Table~\ref{model_var-tab} shows that its effect on our
754: results is insubstantial.
755: 
756: The simplest model is a disk with uniform parameters throughout that
757: can be represented by a set of concentric annuli, all with the same
758: integrated polarization properties. The model makes no assumptions
759: about the magnetic field structure in the center of the disk. This is
760: not a significant simplification for most spiral galaxies, where most
761: of the polarized emission originates in the disk.
762: Figure~\ref{spiral_proj-fig} shows a graphic representation of
763: polarized intensity at a representative subset of locations (a ring)
764: in the disk for model 1 in Table~\ref{model_par-tab} at inclinations
765: $i=40\degr$ and $i=60\degr$. Thin lines indicate the magnitude and
766: direction of the magnetic field component in the plane of the sky,
767: $B_\bot$. Thick lines represent polarization vectors. Open and filled
768: circles represent the component of the regular magnetic field along
769: the line of sight, $B_\|$.
770: 
771: At low inclination (Figure~\ref{spiral_proj-fig}a), the path length
772: through the disk is relatively small and the line-of-sight component
773: of the regular field is also small.  The polarized intensity in
774: Figure~\ref{spiral_proj-fig}a is fairly uniform across the disk and
775: relatively high. At higher inclination, the path length through the
776: disk increases, and the line-of-sight component of the regular
777: magnetic field increases, in particular near the major axis. The
778: polarized intensity is smaller everywhere in the disk, but most
779: strongly near the major axis. Here, the line-of-sight component of the
780: azimuthal regular field is largest, so Faraday rotation is strongest
781: and the component $B_\bot$ (Equation~\ref{P0-eq}) is smallest.
782: Faraday depolarization is therefore strongest near the major axis, and
783: the intrinsic polarization of the emission is smaller than elsewhere
784: in the disk.  Faraday rotation of the plane of polarization is also
785: stronger near the major axis in Figure~\ref{spiral_proj-fig}.
786: 
787: The model Stokes $Q$ and $U$ intensities, calculated by solving
788: Equation~\ref{P-eq} for every line of sight through the disk, are
789: integrated over the disk to predict the polarized intensity of an
790: unresolved spiral galaxy. The integration over the disk reduces to a
791: one-dimensional numerical integration over azimuthal angle for the
792: symmetric models under consideration.  The model in
793: Figure~\ref{spiral_proj-fig}a has integrated fractional polarization
794: $\Pi_0 = 7.9\%$, while the model in Figure~\ref{spiral_proj-fig}b has
795: integrated fractional polarization $\Pi_0 = 17.0\%$. Despite the
796: stronger depolarization anywhere in the disk at inclination $i =
797: 60\degr$, the {\it integrated } polarization is stronger at $i =
798: 60\degr$ than at $i = 40\degr$, because of the smaller degree of
799: symmetry of the polarized emission.
800: 
801: The model predictions for $\Pi_0$ and $\thetaB$ are independent of
802: the spiral pitch angle of the magnetic field. For a disk with constant
803: thickness and uniform properties, the path length through the disk
804: depends only on inclination, and the polarized intensity depends only
805: on the angle of the regular component of the magnetic field with the
806: line of sight. The effect of a pitch angle of the magnetic field, is
807: to change the azimuthal angle in the disk where the line of sight
808: makes a certain angle with the regular magnetic field. The integral
809: over all azimuthal angles in the disk is not affected by a translation
810: in azimuthal angle resulting from the pitch angle of the magnetic
811: field.
812: 
813: 
814: Table~\ref{model_par-tab} gives values for the parameters in the
815: models that are compared with the data. Model 7 corresponds with the
816: values adopted by \citet{beck2007} for NGC~6946. The filling factor of
817: ionized gas $f_{i}$ and the thickness of the disk $2h$ were kept
818: constant, because their effect on the models is similar to that of the
819: electron density and the magnetic field.  Model curves of fractional
820: polarization as a function of inclination are shown in
821: Figure~\ref{diskpol_model-fig}. Three families of models are shown
822: (Table~\ref{model_par-tab}), with $f_{\rm B} = $ 1, 2, and 3. For each
823: value of $f_{\rm B}$, model curves are shown for $n_e = 0.03$
824: cm$^{-3}$, $B = 5\ \mu$G (solid curves), $n_e = 0.03$ cm$^{-3}$, $B =
825: 10 \mu$G (dashed curves), and $n_e = 0.09$ cm$^{-3}$, $B = 5\ \mu$G
826: (dotted curves).
827: 
828: At low inclination, the models are mainly distinguished by the value
829: of $f_{\rm B}$. The line-of-sight component of the regular magnetic
830: field is small at low inclination, and the main depolarization
831: mechanisms are wavelength independent depolarization and Faraday
832: dispersion. At inclination $i \gtrsim 50\degr$, differential Faraday
833: rotation along the line of sight becomes increasingly important, and
834: significant effects of the strength of the regular magnetic field and
835: the density of thermal electrons are found.
836: 
837: This effect already occurs at low inclination, where the line-of-sight
838: component of the magnetic field near the major axis increases
839: proportionally to $i$, but the path length through the disk $\sim
840: 1/\cos i$ is constant to first order in $i$. The result is that models
841: that include Faraday depolarization predict a {\it higher} degree of
842: polarization at low inclinations than models that do not include
843: Faraday depolarization \citep{stil2007a}. The present models also
844: predict a higher fractional polarization at low inclination if the
845: electron density $n_e$ or the strength of the regular magnetic field
846: is increased, if the ratio of random magnetic field to regular
847: magnetic field strength is not too high.
848: 
849: The models more or less cover the same area of
850: Figure~\ref{diskpol_model-fig}a as the data. This result supports the
851: idea that the observed integrated polarization of spiral galaxies is
852: related to the large-scale field in the disk.  The most notable
853: difference between the data and the models is the observed low $\Pi_0
854: \lesssim 2\%$ of some galaxies with intermediate inclination, $40\degr
855: \lesssim i \lesssim 60\degr$. It is difficult to change any single
856: parameter in the model to obtain a fractional polarization $\Pi
857: \lesssim 2\%$ in this inclination range, unless higher values of
858: $f_{\rm B}$ are assumed.
859: 
860: 
861: 
862: \subsection{Probability distribution of $\Pi_0$}
863: \label{Pi_PDF-sec}
864: 
865: The results in the previous sections suggest that the integrated 4.8
866: GHz emission of spiral galaxies can show substantial
867: polarization. Deep polarization surveys can detect distant unresolved
868: spiral galaxies. How many spiral galaxies can be detected, and the
869: interpretation of unresolved galaxies detected in polarization, is
870: determined by the probability density function (PDF) of the fractional
871: polarization of spiral galaxies. The wavelength dependence of
872: integrated polarization is also best summarized by means of the PDF of
873: fractional polarization.  In this section we discuss the expected
874: shape of the PDF of fractional polarization for unresolved spiral
875: galaxies at 4.8 GHz and 1.4 GHz. We also discuss the PDF of $\Pi_0$ in
876: the special case of low-inclination galaxies ($i \lesssim 30\degr$),
877: where the integrated polarization depends mostly on Faraday
878: dispersion, but the actual value of the inclination cannot be
879: determined reliably from the optical axial ratio.
880:  
881: \begin{figure}
882: \resizebox{\columnwidth}{!}{\includegraphics[angle=0]{f5.eps}}
883: \caption{ Probability density functions (PDF) of $\Pi_0$ for
884: randomly-oriented spiral galaxies derived from the models in
885: Figure~\ref{diskpol_model-fig}. The continuous, dashed, and dotted
886: curves correspond with models 1 -- 9 in Table~\ref{model_par-tab}, in
887: order of decreasing maximum $\Pi_{\rm max}$. The curves are normalized
888: such that the integral of the PDF is 1.  Median values $\Pi_{\rm med}$
889: of these distributions are listed in Table~\ref{model_par-tab}.
890: \label{Pi_PDF-fig}
891: }  
892: \end{figure}
893: 
894: \begin{figure*}
895: \resizebox{\textwidth}{!}{\includegraphics[angle=-90]{f6.eps}}
896: \caption{ (a) Simulated histogram of fractional polarization of 20,000
897: randomly oriented model spiral galaxies at 4.8 GHz. Ranges of model
898: parameters used for this simulation are given in
899: Table~\ref{model_var-tab}.  The fractional polarization for the same
900: galaxies at 1.4 GHz is shown in (b).  The vertical axis in (b) is
901: truncated (lowest bin only).
902: \label{Pi_PDF-sim-fig}
903: }  
904: \end{figure*}
905: 
906: 
907: 
908: The models for the integrated fractional polarization as a function of
909: inclination can be used to derive a probability density function (PDF)
910: for $\Pi_0$ of unresolved, randomly oriented spiral galaxies. This PDF
911: can be used to predict the number of spiral galaxies detectable in
912: deep polarization surveys based on population models of the total
913: intensity source counts \citep{stil2007a}. The shape of the PDF can be
914: modeled as a function of rest-frame wavelength and fitted to
915: observations. \citet{taylor2007} used Monte-Carlo simulations of noise
916: statistics and observational selection effects, and derived the shape
917: of the PDF for $\Pi_0$ for faint polarized sources from a maximum
918: likelihood fit to the data. A similar approach can be used in the
919: future to compare the PDF of $\Pi_0$ for galaxies at high redshift
920: with the PDF for local galaxies.
921: 
922: The PDF for $\Pi_0$ follows from the probability $P(i)$ to observe a galaxy at
923: inclination between $i$ and $i+di$,
924: \begin{equation}
925: P(i)di = \sin(i) di \ \ \ \ \  (0\degr \le i \le 90\degr).
926: \end{equation}
927: The probability to find a fractional polarization between $\Pi_0$ and
928: $\Pi_0+\Delta\Pi_0$ is equal to the probability to find a galaxy with
929: inclination in one of up to two inclination ranges $i_{{\rm min}}$ to
930: $i_{{\rm max}}$ that yield the same fractional polarization. Therefore,
931: \begin{equation}
932: P(\Pi_0) \Delta\Pi_0 = \sum_{k = 1}^N ( \cos i_{{\rm min},k} - \cos i_{{\rm max},k}) 
933: \label{PIdist-eq}
934: \end{equation}
935: In most cases, $N=2$, except at the maximum of $\Pi_0$, where
936: $N=1$.
937: 
938: The PDFs of $\Pi_0$ for models 1 -- 9 in Table~\ref{model_par-tab} are
939: shown in Figure~\ref{Pi_PDF-fig}.  The median inclination for
940: randomly-oriented galaxies is $i_{\rm med}=60\degr$. Nearly half
941: (47\%) of the galaxies in a randomly oriented set have inclination
942: between $50\degr$ and $80\degr$, for which $\Pi_0$ is near the maximum
943: of the curves in Figure~\ref{diskpol_model-fig}a.  Consequently, all
944: models predict a PDF that is sharply peaked toward the maximum
945: possible value of $\Pi_0$.
946: 
947: Qualitatively, the shapes of the PDFs in Figure~\ref{Pi_PDF-fig} do
948: not depend strongly on the details of the models. The PDFs are derived
949: from the curves in Figure~\ref{diskpol_model-fig}. The shapes of these
950: curves can be understood qualitatively in terms of two competing
951: effects.  If the inclination is not too large, the integrated
952: polarization increases with inclination because of a larger net
953: magnetic field in the plane of the sky.  At higher inclinations,
954: Faraday depolarization dominates and the integrated polarization
955: decreases. A broad maximum in $\Pi_0$ must occur at intermediate
956: inclination, that gives rise to the peak of the PDF near the maximum
957: of $\Pi_0$.
958: 
959: Another feature of the PDFs in Figure~\ref{Pi_PDF-fig} is that they
960: converge to a constant value for small $\Pi_0$. The distribution is
961: dominated here by galaxies at low inclination, with a smaller
962: contribution of highly inclined galaxies with strong Faraday
963: depolarization.  Small fluctuations in the curves at small $\Pi_0$ are
964: the result of occurrences where the Faraday rotation of the disk at
965: the major axis is $\mathcal{R}\lambda^2 = n \pi$ ($n = 1, 2, 3,
966: \ldots$) at high inclination. These fluctuations are small, and would
967: not be noticed in observations.
968: 
969: The plateau in the distributions in Figure~\ref{Pi_PDF-fig} can be
970: understood based on a simple analytic argument.  For inclination $i
971: \lesssim 30\degr$, the model curves in Figure~\ref{diskpol_model-fig}
972: can be approximated with the relation $\Pi_0 = a i^2$, with the
973: inclination $i$ in radians and $a$ a constant that depends on the
974: model. The value of $a$ may depend on complicated physics, but the
975: only assumption here is that the increase of $\Pi_0$ for small $i$ can
976: approximated by a parabola, for fixed parameters of the model. The
977: probability to find a fractional polarization between $\Pi_0$ and
978: $\Pi_0+d\Pi_0$ is equal to the probability to find a galaxy with
979: inclination in the interval $i_1 \leqq i \leqq i_2 < 30\degr$, with
980: $\Pi_0 = a i_{{\rm min},1}^2$, and $(\Pi_0+d\Pi_0) = a i_{{\rm
981: max},1}^2$.  Equation~\ref{PIdist-eq} reduces to
982: \begin{equation}
983: P(\Pi_0)d\Pi_0 = \cos i_{{\rm min},1} - \cos i_{{\rm max},1} = {1 \over 2a} \Bigl( {\sin i \over i}\Bigr) d\Pi_0.
984: \label{low_i-eq}
985: \end{equation}
986: The last step used that $(i_{{\rm max},1}-i_{{\rm min},1})$ is the
987: infinitesimal increment $di = d\Pi_0/(2 a i)$, and that
988: $\onehalf(i_{{\rm min},1}+i_{{\rm max},1})=i$.  For $i < 30\degr$ the
989: factor $\sin i / i \approx 1$ to an accuracy better than 5\%, so
990: $P(\Pi_0)$ is a constant that depends only on $a$.  The models
991: presented in Section~\ref{model-sec} suggest that $a$ depends mainly
992: on the ratio of the random magnetic field to the regular magnetic
993: field. This is why models 1, 2, and 3 converge to the same value for
994: small $\Pi_0$, and models 4, 5, and 6 converge to a larger value. The
995: differences between models 7, 8, and 9 arise from the more significant
996: contributions from galaxies with a high inclination. These models also
997: converge to a common value if only galaxies with low inclination are
998: considered.
999: 
1000: The distribution of $\Pi_0$ of an observed sample of spiral galaxies
1001: would be a superposition of a continuous set of model distributions
1002: similar to the nine curves in Figure~\ref{Pi_PDF-fig}. A simulated PDF
1003: for 20,000 galaxies with model parameters varied over the ranges given
1004: in Table~\ref{model_var-tab} is shown in Figure~\ref{Pi_PDF-sim-fig}.
1005: At 4.8 GHz (Figure~\ref{Pi_PDF-sim-fig}a), the distribution of $\Pi_0$
1006: is quite broad, with a peak between $1\%$ and $2\%$ polarization.  The
1007: location of the peak and the depth of the minimum near $\Pi_0 = 0\%$
1008: depends on the number of galaxies with strong Faraday depolarization.
1009: At 1.4 GHz (Figure~\ref{Pi_PDF-sim-fig}b), Faraday depolarization is
1010: much stronger and the distribution is sharply peaked towards $\Pi_0 =
1011: 0\%$.  For the distributions in Figure~\ref{Pi_PDF-sim-fig} we find
1012: that 78\% of galaxies has $\Pi_0 > 1\%$ at 4.8 GHz, while at 1.4
1013: GHz 17\% has $\Pi_0 > 1\%$. The broad tail of the distribution in
1014: Figure~\ref{model_var-tab}b contains galaxies with low to moderate
1015: inclination. The median inclination for galaxies with $\Pi_0 > 5\%$ at
1016: 1.4 GHz is $41\degr$.
1017: 
1018: Integrated polarimetry of an unbiased sample of a few hundred spiral
1019: galaxies would allow a statistical analysis for the $\Pi_0$
1020: distributions of spiral galaxies, including subsets by inclination,
1021: morphological type and star formation rate. The $\Pi_0$ distribution
1022: can be retrieved from the data in the presence of noise through
1023: maximum likelihood fits presented by \citet{taylor2007}.  The skewed
1024: shape on the PDF in Figure~\ref{Pi_PDF-sim-fig}a, suggests that the
1025: best theoretical function to represent the PDF of spiral galaxies
1026: would be the lowest order {\it anti}-symmetric deviation from a
1027: Gaussian that shifts the peak of the distribution to a finite value of
1028: $\Pi_0$. In terms of the Gauss-Hermite series used in
1029: \citet{taylor2007}, the lowest order term would include coefficient
1030: $h_3$ \citep{vdmarel1993}. 
1031: 
1032: 
1033: 
1034: \section{Discussion}
1035: \label{discussion-sec}
1036: 
1037: \subsection{Diagnostics from integrated polarization}
1038: \label{discussion1}
1039: 
1040: We presented observations of the integrated polarization of nearby
1041: spiral galaxies to study the polarization properties of spiral
1042: galaxies as unresolved radio sources. It is shown that unresolved
1043: spiral galaxies are polarized radio sources with fractional
1044: polarization up to $\sim 20\%$ at 4.8 GHz.  This high degree of
1045: fractional polarization arises from the regular azimuthal component of
1046: the magnetic field in the spiral galaxies observed, that is projected
1047: into a component in the plane of the sky that is predominantly
1048: oriented along the apparent major axis of the disk seen at inclination
1049: $i$.  The projected magnetic field direction as implied by the plane
1050: of polarization, rotated by $90\degr$, is well correlated with the
1051: position angle of the major axis for galaxies observed at higher
1052: inclination.  The highest integrated polarization is expected for
1053: galaxies at intermediate inclination. For a sample of randomly
1054: oriented spiral galaxies, 47\% has an inclination between $50\degr$
1055: and $80\degr$.  Galaxies in this inclination range are
1056: under-represented in the present sample because the data used in this
1057: paper were obtained for polarization studies that preferred
1058: low-inclination or high-inclination galaxies. The number of galaxies
1059: with a high fractional polarization in Figure~\ref{diskpol_model-fig}
1060: may therefore be under represented compared with an unbiased sample.
1061: 
1062: 
1063: Three of the eight Virgo spirals in our data set have integrated
1064: fractional polarization $\Pi_0 \geqq 8\%$. The range of $\Pi_0$ of the
1065: Virgo spirals is similar to that of the nearby ``field'' spirals.  We
1066: do not find evidence in the present data that the integrated
1067: polarization of spiral galaxies in a cluster environment would be
1068: different from field galaxies. Although cluster galaxies are distorted
1069: by interactions with other galaxies and the intracluster medium
1070: \citep{wezgowiec2007,chyzy2007,vollmer2007}, the Virgo galaxies in our
1071: sample cannot be distinguished from field galaxies in their integrated
1072: polarization. This result should be verified with a larger,
1073: complete sample of cluster galaxies.
1074: 
1075: The barred galaxies in Table~\ref{barred-tab} show that a high
1076: integrated polarization also occurs for barred galaxies. In some of
1077: these barred galaxies, polarized emission from the bar is negligible
1078: compared to polarized emission from the disk. In other barred spirals,
1079: polarized emission from the bar region dominates, and the integrated
1080: polarization depends on the orientation of the bar, as well as the
1081: inclination of the galaxy. The effect of a bar is not included in the
1082: present models, so the existence of significant departures from the
1083: models is not unexpected. Non-axially symmetric models of the
1084: integrated polarization of spiral galaxies are deferred to a later
1085: paper.
1086: 
1087: Figure~\ref{PI_L-fig} suggests a difference between luminous and less
1088: luminous spiral galaxies. The three sub samples separately also show a
1089: smaller integrated fractional polarization for luminous spiral
1090: galaxies.  Face-on galaxies and depolarized edge-on galaxies should
1091: exist at every luminosity, although they form a minority in a set of
1092: randomly oriented spiral galaxies. The presence of some galaxies with
1093: low $\Pi_0$ at every luminosity in Figure~\ref{PI_L-fig} is therefore
1094: not surprising.  The luminous galaxies in the sample cover the same
1095: range in inclination as the galaxies with lower luminosity. This
1096: excludes selection effects in inclination as a possible origin of the
1097: correlation.  No selection effect could be identified that would
1098: exclude luminous galaxies with a high $\Pi_0$ from the sample.
1099: 
1100: The lack of luminous galaxies with a high $\Pi_0$ gains significance
1101: in view of the results from Section~\ref{Pi_PDF-sec}. Given certain
1102: physical conditions, the inclination dependence of $\Pi_0$ favors high
1103: values of $\Pi_0$. The lower polarization of luminous spiral galaxies
1104: therefore indicates different conditions in these galaxies. A higher
1105: rate of star formation increases the random component of the magnetic
1106: field through the effect of supernova explosions and stellar winds. A
1107: stronger total magnetic field and lower polarization are found locally
1108: in regions of galaxies where the star formation rate is high
1109: \citep{beck2005,chyzy2008}. The correlation in Figure~\ref{PI_L-fig}b
1110: may be explained similarly in terms of a generally higher star
1111: formation rate.  The correlation of polarization with luminosity is an
1112: important factor for predictions of the number of spiral galaxies that
1113: will be detected in deep polarization surveys.  Larger samples of
1114: spiral galaxies are required to confirm this result, and to allow more
1115: detailed modeling.
1116: 
1117: Simple axially symmetric models for the integrated polarization of
1118: spiral galaxies successfully describe the distribution of observed
1119: galaxies in the $\Pi_0$ - $i$ diagram
1120: (Figure~\ref{diskpol_model-fig}a). The results in
1121: Table~\ref{model_var-tab} show that for $i \lesssim 50\degr$ the model
1122: is most sensitive to the parameter $f_{\rm B}$. This is so because
1123: wavelength-dependent depolarization such as Faraday dispersion and
1124: differential Faraday rotation are small compared with
1125: wavelength-independent depolarization in the models at 4.8 GHz with a
1126: small inclination. At higher inclination and at lower frequencies,
1127: wavelength-dependent depolarization is more important.  A complete
1128: sample of spiral galaxies observed at 4.8 GHz may be divided into
1129: sub samples by inclination to separate the effects of different physical
1130: parameters. We discuss three inclination ranges.
1131: 
1132: Galaxies with inclination $i \lesssim 30\degr$ (13.4\% of an unbiased
1133: sample) would display low fractional polarization, provided the galaxy
1134: is axially symmetric. Large-scale departures from axial symmetry
1135: likely result in a higher degree of polarization for these
1136: low-inclination galaxies. A statistical study of the integrated
1137: polarization of face-on spiral galaxies will therefore provide
1138: constraints on the large-scale asymmetry of polarized emission and
1139: therefore magnetic fields in spiral galaxies.  For a sample of face-on
1140: barred galaxies, the orientation of the plane of polarization may be
1141: correlated with the position angle of the bar derived from optical
1142: images.
1143: 
1144: At inclinations $30\degr \lesssim i \lesssim 50\degr$ (22.3\% of an
1145: unbiased sample), the axially symmetric models are most sensitive to
1146: the ratio of the random magnetic field to the regular magnetic
1147: field.  This inclination range seems most promising to provide
1148: constraints of this parameter for a large sample of
1149: galaxies. Integrated polarimetry at more than one frequency may be
1150: combined to separate wavelength-dependent depolarization from
1151: wavelength-independent depolarization.  More detailed models of the
1152: effect of non-axially symmetric structure such as spiral arms are
1153: required.
1154: 
1155: At inclinations $i \gtrsim 50\degr$ (64.3\% of an unbiased sample),
1156: the longer path length and larger line-of-sight components of the
1157: regular magnetic field greatly enhance wavelength-dependent Faraday
1158: depolarization and Faraday rotation. This is confirmed by the results
1159: in Table~\ref{model_var-tab}, where the contribution of $B$, $n_e$,
1160: $f_i$, and $2 h$ to the total scatter all rise to approximately equal
1161: amounts at inclination $75\degr$.  The ratio of random to regular
1162: magnetic field is still an important factor, but this can be
1163: constrained by results obtained from galaxies at lower inclination.
1164: Multi-frequency observations would be valuable because the models
1165: suggest that the strongest frequency dependence of depolarization of
1166: the integrated emission occurs for inclinations more than $\sim
1167: 60\degr$.
1168: 
1169: 
1170: \subsection{Spiral galaxies as polarized background sources}
1171: 
1172: Are spiral galaxies suitable background sources for rotation measure
1173: studies? Internal Faraday rotation in distant spiral galaxies likely
1174: amounts to rotation measures of a few hundred rad m$^{-2}$.  However,
1175: the correlation between $\thetaB$ and $\thetaopt$ in
1176: Figure~\ref{PA-fig}a, and the models presented in
1177: Section~\ref{model-sec} suggest that the plane of polarization is
1178: related to the orientation of the major axis of the optical disk.  The
1179: polarization angle of an unresolved spiral galaxy is determined by the
1180: orientation of its major axis through the projection of the azimuthal
1181: magnetic field in the galactic disk on the plane of the sky. This is
1182: in part a result of the symmetry of the galaxies, as polarized
1183: emission from areas of the disk opposite to each other with respect to
1184: the center of the galaxy is Faraday rotated by equal but opposite
1185: angles.  Depolarization resulting from differential Faraday rotation
1186: also mitigates the effect of Faraday rotation on the position angle of
1187: the integrated emission.  The correlation in Figure~\ref{PA-fig}a
1188: confirms that the plane of polarization is related to the orientation
1189: of the major axis in real galaxies.
1190: 
1191: The important implication is that the plane of polarization of a
1192: spiral galaxy that is unresolved by a radio telescope, is independent
1193: of wavelength. The orientation of the plane of polarization can be
1194: determined from the position angle of the optical major axis, although
1195: with a $180\degr$ ambiguity.  As the plane of polarization does not
1196: depend on wavelength, we find the surprising result that {\it
1197: unresolved spiral galaxies behave as idealized polarized background
1198: sources without apparent internal Faraday rotation.} This result
1199: follows from axial symmetry.  The 8.4 GHz polarization data in
1200: Table~\ref{virgo-tab} seem to confirm that the polarization angle of
1201: the integrated emission is independent of wavelength.  More
1202: observations are required to confirm this mainly theoretical result
1203: for more galaxies and at lower frequencies.
1204: 
1205: The intrinsic position angle of the plane of polarization of an
1206: unresolved spiral galaxy can be estimated from an optical image. It
1207: does not depend on internal Faraday rotation or on the spiral pitch
1208: angle of the magnetic field (Section~\ref{model-sec}).  In principle,
1209: this property allows the detection of possible Faraday rotation by an
1210: intergalactic medium, using a large sample of spiral galaxies.  The
1211: correlation between $\thetaB$ and $\thetaopt$ (Figure~\ref{PA-fig})
1212: can be made for a large number of galaxies, subdivided by distance and
1213: location in the sky. If large-scale magnetic fields exist in the
1214: nearby intergalactic medium, the correlation between $\thetaB$ and
1215: $\thetaopt$ would show an offset for more distant galaxies in a
1216: certain direction. The rotation measure sensitivity of such an
1217: experiment would depend primarily on the sample size of spiral
1218: galaxies with integrated polarimetry, and on the longest wavelength at
1219: which unresolved spiral galaxies remain polarized sources.
1220: 
1221: 
1222: \section{Summary and Conclusions}
1223: 
1224: This paper discusses the integrated polarization properties of nearby
1225: galaxies from archival data at 4.8 GHz, and models for the integrated
1226: polarized emission of a spiral galaxy. We find that
1227: 
1228: 1. Spiral galaxies can show substantially polarized (up to $\sim$20\%
1229: at 4.8 GHz) integrated radio emission. The polarization angle of the
1230: integrated emission of spirals without a (strong) bar is aligned with
1231: the apparent minor axis of the disk.  Barred spirals can also be
1232: polarized sources, but do not show a clear correlation of polarization
1233: angle with the orientation of the major axis.
1234: 
1235: 2. The highest degree of polarization, and the largest spread in
1236: fractional polarization, are found for spiral galaxies with
1237: intermediate to high inclination ($i \gtrsim 50\degr$). 
1238: 
1239: 3. A sub-sample of nine spiral galaxies in the Virgo cluster does not
1240: show a significant difference in integrated polarization properties
1241: from the field galaxies in our sample. This suggests that physical
1242: distortions of these galaxies are not so strong that they have a
1243: noticeable effect on the integrated polarization. Confirmation by a
1244: larger, complete sample is required.
1245: 
1246: 4. The fractional polarization of integrated emission is smaller for
1247: galaxies with a higher radio luminosity. This result suggests a
1248: difference in properties of the magnetic field and/or Faraday
1249: depolarization in luminous spiral galaxies.
1250: 
1251: 5. Our models for the integrated polarization of axially symmetric
1252: spiral galaxies cover the same area in the $\Pi_0$-$i$ diagram
1253: (Figure~\ref{diskpol_model-fig}) as the observed galaxies
1254: without a strong bar. The models indicate that the main cause of
1255: variation in the fractional polarization of galaxies with inclination
1256: less than $50\degr$ is the ratio of the random to the regular magnetic
1257: field component in the disk.
1258: 
1259: 6. Our models provide a probability density function of fractional
1260: polarization of spiral galaxies as a function of wavelength that can
1261: be used to predict the number of spiral galaxies in deep polarization
1262: surveys, and to study the statistical properties of spiral galaxies in
1263: such surveys.
1264: 
1265: 7. The integrated polarization of spiral galaxies opens the
1266: possibility to study magnetic field properties and Faraday rotation in
1267: large samples of spiral galaxies, and in galaxies at high redshift.  A
1268: deep 1 - 2 GHz survey with the SKA could detect normal spiral galaxies
1269: at redshift 1 or higher \citep{wilman2008} with a total intensity flux
1270: $\sim 100\ \mu \rm Jy$. These can be compared in a
1271: resolution-independent way with integrated polarization measurements
1272: of low-redshift galaxies at the same restframe frequency.
1273: 
1274: 8. Unresolved symmetric spiral galaxies behave as idealized background
1275: sources without internal Faraday rotation. They can be used to detect
1276: large-scale magnetic fields in the intergalactic medium, e.g. with the
1277: SKA \citep{gaensler2004, stepanov2008}.
1278: 
1279: 
1280: \acknowledgements We thank Marek We\.zgowiec and Marek Urbanik for
1281: giving us the maps of the Virgo galaxies, some of them prior to
1282: publication. The authors thank K. T. Chy\.zy for the use of data for
1283: NGC~4254 and NGC~4736. We also acknowledge Maike Lensing for her help
1284: to prepare the radio maps for the flux integration. The authors thank
1285: the referee for comments that helped to improve the manuscript.  This
1286: research has made use of the NASA/IPAC Extragalactic Database (NED)
1287: which is operated by the Jet Propulsion Laboratory, California
1288: Institute of Technology, under contract with the National Aeronautics
1289: and Space Administration.
1290: 
1291: 
1292: 
1293: \begin{thebibliography}{}
1294: 
1295: \bibitem[Arshakian et~al.(2007)]{arshakian2007} Arshakian, T., Stepanov, R., Beck, R., Frick, P., \& Krause, M. 2007, arXiv:0711.4979
1296: 
1297: % Reference for M81 data:
1298: \bibitem[Beck et~al.(1985)]{beck1985} Beck, R., Krause, M., \& Klein, U. 1985, \aap, 152, 237
1299: 
1300: % Reference for NGC6946 data:
1301: \bibitem[Beck \& Hoernes(1996)]{beck1996} Beck, R., \& Hoernes, P. 1996, \nat, 379, 47
1302: 
1303: % Magnetic fields in barred galaxies I. The atlas
1304: \bibitem[Beck et~al.(2002)]{beck2002} Beck, R., Shoutenkov, V., Ehle, M., Harnett, J. I., Haynes, R. F., Shukurov, A., Sokoloff, D. D., Thierbach, M. 2002, \aap, 391, 83
1305: 
1306: \bibitem[Beck(2005)]{beck2005} Beck, R. 2005, in Cosmic Magnetic Fields, ed. R. Wielebinski \& R. Beck (Heidelberg:Springer), 41
1307: 
1308: \bibitem[Beck(2007a)]{beck2007a} Beck, R. 2007, EAS Publ. Ser., 23, 19
1309: 
1310: \bibitem[Beck(2007b)]{beck2007} Beck, R. 2007, \aap, 470, 539
1311: 
1312: % Reference for M31 data:
1313: \bibitem[Berkhuijsen et~al.(2003)]{berkhuijsen2003} Berkhuijsen, E. M., Beck, R., \& Hoernes, P 2003, \aap, 398, 937
1314: 
1315: \bibitem[Binney \& de Vaucouleurs (1981)]{binney1981} Binney, J. \& de Vaucouleurs, G. 1981, \mnras, 194, 679
1316: 
1317: \bibitem[Brown et~al.(2003)]{brown2003} Brown, J. C., Taylor, A. R., Wielebinski, R., \& Mueller, P. 2003, \apj 592, L29
1318: 
1319: \bibitem[Brown et~al.(2007)]{brown2007} Brown, J. C., Haverkorn, M., Gaensler, B. M., Taylor, A. R., Bizunok, N. S., McClure-Griffiths, N. M., Dickey, J. M., \& Green, A. J. 2007, \apj 663, 258
1320: 
1321: \bibitem[Burn(1966)]{burn1966} Burn, B. J. 1966, \mnras, 133, 67
1322: 
1323: \bibitem[Chy\.zy et~al.(2007)]{chyzy2007} Chy\.zy, K. T., Ehle, M., \& Beck, R. 2007, \aap, 474, 415
1324: 
1325: \bibitem[Chy\.zy(2008)]{chyzy2008} Chy\.zy, K. T. 2008, \aap, 482, 755
1326: 
1327: \bibitem[Chy\.zy \& Buta(2008)]{chyzy_buta2008} Chy\.zy, K. T., \& Buta, R.J. 2008, \apj, 677, L17 
1328: 
1329: \bibitem[Condon \& Yin(1990)]{condon1990} Condon, J. J. \& Yin, Q. F. 1990, \apj, 357, 97
1330: 
1331: \bibitem[Condon(1992)]{condon1992} Condon, J. J. 1992, \araa, 30, 575
1332: 
1333: % Reference for NGC891, NGC3628, NGC4565
1334: \bibitem[Dumke \& Krause(1998)]{dumke1998} Dumke, M., \& Krause, M., 1998, Lecture Notes in Physics, 506, 555
1335: 
1336: % Reference for NGC 5907:
1337: \bibitem[Dumke et~al.(2000)]{dumke2000} Dumke, M., Krause, M., \& Wielebinski, R. 2000, \aap, 355, 512
1338: 
1339: % Reference for M51 data:
1340: \bibitem[Fletcher et al.(2004)]{fletcher2004} Fletcher, A., Beck, R., Berkhuijsen, E. M., Horellou, C., Shukurov, A. 2004, in How does the Galaxy work?, ed. E. J. Alfaro et al. (Dordrecht: Kluwer), 299
1341: 
1342: \bibitem[Gaensler et~al.(2004)]{gaensler2004} Gaensler, B., Beck, R., \& Feretti, L. 2004, New Ast. Rev., 48, 1003
1343: 
1344: \bibitem[Gaensler et~al.(2005)]{gaensler2005} Gaensler, B., Haverkorn, M., Staveley-Smith, L., Dickey, J. M., McClure-Griffith, N. M., Dickel, J. R., \& Wolleben, M. 2005, Science, 307, 1610
1345: 
1346: % Reference for VLA image of NGC4631 (data not used here)
1347: \bibitem[Golla \& Hummel(1994)]{golla1994} Golla, G., \& Hummel, E. 1994, \aap, 284, 777
1348: 
1349: % Reference for IC342 Effelsberg map
1350: \bibitem[Gr\"ave \& Beck(1988)]{grave1988} Gr\"ave, R., \& Beck, R. 1988, \aap, 192, 66
1351: 
1352: \bibitem[Grosb{\o}l(1985)]{grosbol1985} Grosb{\o}l, P. J. 1985, \aaps, 60, 261
1353: 
1354: \bibitem[Han et~al.(1998)]{han1998} Han, J. L., Beck, R., Berkhuijsen, E. M. 1998, \aap, 335, 1117
1355: 
1356: \bibitem[Han et~al.(1999)]{han1999} Han, J. L., Manchester, R. N., Qiao, G. J. 1999, \mnras, 306, 371
1357: 
1358: \bibitem[Heesen et~al.(2008)]{heesen2008} Heesen, V., Dettmar, R.-J., Kause, M., \& Beck, R. 2008, PoS(MRU)089 {\tt arxiv:0801.3542}
1359: 
1360: \bibitem[Johnston et~al.(2007)]{johnston2007} Johnston, S., et~al. 2007, PASA, 24, 174
1361: 
1362: \bibitem[Krause(2003)]{krause2003} Krause, M. 2003, in The Magnetized Interstellar Medium, Proceedings of the conference, held in Antalya, Turkey, September 8 - 12, 2003, Ed. B. Uyaniker, W. Reich, and R. Wielebinski, Copernicus GmbH, Katlenburg-Lindau., 173
1363: 
1364: \bibitem[Lambas et~al.(1992)]{lambas1992} Lambas, D. G., Maddox, S. J., \& Loveday, J. 1992, \mnras, 258, 404
1365: 
1366: \bibitem[Niklas et~al.(1997)]{niklas1997} Niklas, S., Klein, U., Wielebinski, R. 1997, \aap, 322, 19
1367: 
1368: \bibitem[Simmons \& Stewart(1985)]{simmons1985} Simmons, J. F. L., \& Stewart, B. G. 1985, \aap, 142, 100
1369: 
1370: \bibitem[Sokoloff et~al.(1998)]{sokoloff1998} Sokoloff, D. D., Bykov, A. A., Shukurov, A., Berkhuijsen, E. M., Beck, R., \& Poezd, A. D. 1998, \mnras, 299, 189
1371: 
1372: \bibitem[Spangler(1982)]{spangler1982} Spangler, S. R. 1982, \apj, 261, 310
1373: 
1374: \bibitem[Stil et~al.(2007)]{stil2007a} Stil, J. M., Taylor, A. R., Krause, M., \& Beck, R. 2007, PoS(MRU)069 {\tt arxiv:0802.1374}
1375: 
1376: \bibitem[Stepanov et~al.(2008)]{stepanov2008} Stepanov, R., Arshakian, T. G., Beck, R., Frick, P., \& Krause, M. 2008, \aap, 480, 45
1377: 
1378: % Reference for M33 data:
1379: \bibitem[Tabatabaei et~al.(2007)]{tabatabaei2007} Tabatabaei, F. S., Krause, M., \& Beck, R. 2007, \aap, 472, 785
1380: 
1381: % Reference for NGC 5775 data:
1382: \bibitem[T\"ullmann et~al.(2000)]{tullmann2000} T\"ullmann, R., Dettmar, R.-J., Soida, M., Urbanik, M., Rossa, J. 2000, \aap, 364, L36
1383: 
1384: \bibitem[Taylor et~al.(2007)]{taylor2007} Taylor, A. R., et~al. 2007, \apj, 666, 201
1385: 
1386: \bibitem[Van der Marel \& Franx(1993)]{vdmarel1993} Van der Marel, R. P., \& Franx, M. 1993, \apj, 407, 525
1387: 
1388: \bibitem[Vollmer et~al.(2007)]{vollmer2007} Vollmer, B., Soida, M., Beck, R., Urbanik, M., Chy\.zy, K. T., Otmianowska-Mazur, K., Kenney, J. D. P., Van Gorkom, J. H. 2007, \aap, 464, L37
1389: 
1390: \bibitem[We\.zgowiec et~al.(2007)]{wezgowiec2007} We\.zgowiec, M., Urbanik, M., Vollmer, B., Beck, R., Chy\.zy, K. T., Soida, M., \& Balkowski, Ch. 2007, \aap, 471, 93
1391: 
1392: \bibitem[Wilman et~al.(2008)]{wilman2008} Wilman, R. J., Miller, L., Jarvis, M. J., Mauch, T., Levrier, F., Abdala, F. B., Rawlings, S., Kl\"ockner, H.-R.,  Obreschkow, D., Olteanu, D. \& Young, S. 2008, \mnras, 388, 1335
1393: 
1394: \end{thebibliography}
1395: 
1396: 
1397: \end{document}
1398: 
1399: 
1400: