Polarized synchrotron radiation from the Andromeda Galaxy M31 and background sources at 350 MHz
aa r X i v : . [ a s t r o - ph . C O ] S e p Astronomy & Astrophysicsmanuscript no. M31_350MHz c (cid:13)
ESO 2018October 19, 2018
Polarized synchrotron radiation from the Andromeda Galaxy M31and background sources at 350 MHz
R. Gießübel , G. Heald , R. Beck , and T.G. Arshakian Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, 53121 Bonn, Germany ASTRON, PO Box 2, 7990 AA Dwingeloo, The Netherlands Kapteyn Astronomical Institute, University of Groningen, PO Box 800, 9700 AV Groningen, The Netherlands I. Physikalisches Institut, Universität zu Köln, Zülpicher Straße 77, 50937 Köln, Germany Byurakan Astrophysical Observatory, Byurakan 378433, Armenia and Isaac Newton Institute of Chile, Armenian BranchReceived 25 April 2013 / Accepted 8 September 2013
ABSTRACT
Context.
Low-frequency radio continuum observations are best suited to search for radio halos of inclined galaxies. Polarizationmeasurements at low frequencies allow the detection of small Faraday rotation measures caused by regular magnetic fields in galaxiesand in the foreground of the Milky Way.
Aims.
The detection of low-frequency polarized emission from a spiral galaxy such as M31 allows us to assess the degree of Faradaydepolarization, which can be compared with models of the magnetized interstellar medium.
Methods.
The nearby spiral galaxy M31 was observed in two overlapping pointings with the Westerbork Synthesis Radio Telescope(WSRT), resulting in about 4 ′ resolution in total intensity and linearly polarized emission. The frequency range 310–376 MHz wascovered by 1024 channels, which allowed the application of rotation measure (RM) synthesis on the polarization data. We derived adata cube in Faraday depth and compared two symmetric ranges of negative and positive Faraday depths. This new method avoidsthe range of high instrumental polarization and allows the detection of very low degrees of polarization. Results.
For the first time, di ff use polarized emission from a nearby galaxy is detected below 1 GHz. The degree of polarization isonly 0 . ± . DP (90 , = . ± . ∼ Conclusions.
As expected, polarized emission from M31 and extragalactic background sources is much weaker at low frequenciesthan in the GHz range. Future observations with LOFAR, with high sensitivity and high angular resolution to reduce depolarization,may reveal di ff use polarization from the outer disks and halos of galaxies. Key words.
Instrumentation: interferometers – Techniques: polarimetric – Galaxies: individual: M31 – Galaxies: magnetic fields –Radio continuum: galaxies
1. Introduction
The Andromeda Galaxy (M31) at a distance of ∼
750 kpc was one of the first external spiral galaxies from which polar-ized radio emission was detected (Beck et al. 1978). Most ofthe radio continuum emission is synchrotron emission, originat-ing from cosmic-ray electrons that spiral around the interstellarmagnetic field lines. An early analysis of the polarized emis-sion showed that the turbulent and ordered components of themagnetic field are concentrated in a ring-like structure at about10 kpc radius (Beck 1982). This ring is a superposition of severalspiral arms with small pitch angles, seen under the inclinationof 75 ◦ (Chemin et al. 2009). Faraday rotation measures (RM)showed that the ordered field of M31 is coherent, meaning thatit preserves its direction around 360 ◦ in azimuth and across sev-eral kiloparsecs in radius (Beck 1982; Berkhuijsen et al. 2003).This large-scale field can be traced out to about 20 kpc distance For example, 752 ±
27 kpc from the luminosity of the Cepheids (Riesset al. 2012) or 744 ±
33 kpc using eclipsing binaries (Vilardell et al.2010). from the centre of M31, using polarized background sources inthe M31 field as an RM grid (Han et al. 1998).By now we know that spiral (and even some irregular) galax-ies exhibit ordered magnetic fields (Beck 2005) with averagefield strengths of 5 ± µ G, while the average random field istypically three times stronger (assuming energy equipartition be-tween magnetic fields and cosmic rays) (Fletcher 2010). In M31the strengths of the ordered and random fields are about equal(5 ± µ G) (Fletcher et al. 2004), which is unique among thegalaxies observed so far.These ordered fields on scales of the entire galaxy can be bestexplained by dynamo theory (Beck et al. 1996). It requires weakmagnetic seed fields and an interplay of turbulence and shearto generate and maintain such a field. The turbulence can beprovided by supernova explosions, while shear is a consequenceof the di ff erential rotation of the galactic disk. The exception-ally well ordered large-scale magnetic field in M31 can be welldescribed by an axisymmetric spiral pattern, the basic dynamomode m =
0, disturbed by a weaker m = Article number, page 1 of 19 o our knowledge, there has been no detection of di ff use po-larization from spiral galaxies below 1 GHz. All observationsof polarization from nearby galaxies were restricted to the GHzrange so far. Lower frequencies are advantageous in severalaspects: (1) synchrotron emission is stronger, especially fromsteep-spectrum radio halos, and (2) Faraday rotation is stronger,allowing the measurement of small RM even with low signal-to-noise ratios. Multichannel polarimetry allows the applicationof RM synthesis (Brentjens & de Bruyn 2005), which generatesFaraday spectra for each map pixel. The achieved resolution inFaraday spectra increases with coverage in λ space ( ∆ λ ) andhence is higher at low frequencies. On the other hand, Faradaydepolarization increases towards lower frequencies at a rate de-pending on the depolarization mechanism (Burn 1966; Sokolo ff et al. 1998; Tribble 1991). As a result, polarized emission de-creases below some characteristic frequency that depends on theproperties of the Faraday-rotating medium (Arshakian & Beck2011). Depolarization in M31 is relatively low because of itsweak turbulent magnetic field (Sect. 5.3). This makes M31 anexcellent candidate for low-frequency polarization studies.The properties of nearby galaxies as observed in radio con-tinuum below 250 MHz will soon be explored with the LOwFrequency ARray (LOFAR, van Haarlem et al. 2013), to searchfor extended synchrotron emission far away from the galacticdisks and in the galactic halos. The observations at 350 MHzpresented in this paper are a crucial observational link betweenobservations at gigahertz frequencies and the upcoming observa-tions with LOFAR. To this point, the polarization properties ofgalactic disks and the feasibility to use polarized point sources asa background grid to explore magnetic fields in the foregroundare untested at low frequencies.
2. Observations
M31 was observed on four di ff erent days using the WSRT inDecember 2008. Two pointings centred at R a = ec = a = ec = × λ
3. Data reduction
Calibration was made in
CASA 3.3.0 (Common AstronomySoftware Applications) after correcting for the system temper- http: // casa.nrao.edu / index.shtml ature in AIPS (Astronomical Image Processing System) . Forlow-frequency observations with the WSRT there are some limi-tations of the standard CASA tasks, but since the software is basedon the script language python , the user has total control overthe data and can also run tasks in batch mode. This is important,since calibration and imaging has to be performed for each chan-nel separately. Due to the large number of channels, this has tobe automated.
With this amount of data, manual flagging of every single chan-nel is no longer possible. The software package rficonsole by O ff ringa et al. (2010) was used, which was specifically de-veloped for low-frequency data of LOFAR and the WSRT. Thealgorithm is described in O ff ringa et al. (2010). It features ageneral user interface, with which one can inspect the data man-ually. Here one develops a so-called strategy (essentially a pa-rameter file used by rficonsole ), by randomly checking forsingle baselines how well the algorithm detects any RFI.Since the bandpass response across a spectral window is fil-tered and drops smoothly to zero at the edges to reduce aliasinge ff ects, a bandpass calibration is performed beforehand to fa-cilitate the operation of the algorithm. Rficonsole is appliedto the bandpass-corrected data in the
CORRECTED_DATA column,but the flags are stored in a separate table and are also appliedto the
DATA column, which holds the raw data and is used forthe following steps. The final bandpass calibration is made afterthe flagging. The first 3 and last 17 channels of all IFs are unus-able due to the anti-aliasing filter and are flagged as well. Sincethe bands overlap, one does in general not lose any frequencychannels.However, here two of the bands, IF2 and IF3 (358.75 MHzand 350.0 MHz) were unusable due to RFI and had to be re-moved entirely for all four days.Because the maxi-short configuration was used for the ob-servation, antennas RT9 and RTA were subject to shadowing .RT9 was manually flagged for hour angles ≦ − ≧ + The system temperature ( T s y s ) calibration had to be performedin AIPS , since
CASA 3.3.0 was unable to read the T s y s infor-mation table provided from the telescope. However, for thecalibration in AIPS , the polarization products have to be trans-formed from linear polarization (XX, XY, YX, YY) to circularpolarization (RR, LL, RL, LR). Detailed instructions are givenin the CookBook for WSRT data reduction using classic
AIPS by R. Braun .The observation of 3C295 and 3C303 failed on the secondday. For consistency only 3C147 and DA240 were used forcalibration. Where possible, the gain solutions were applied to3C295 and 3C303 to check their validity.The flux of 3C147 was derived using the analytic functiongiven in the VLA Calibrator Manual (Perley & Taylor 2003),log( S ν ) = A + B log( ν ) + C log ( ν ) + D log ( ν ) , (1) http: // / see the WSRT shadowing calculator athttp: // / ~heald / tools / wsrtshadow.php http: // astron.nl / radio-observatory / astronomers / analysis-wsrt-data / analysis-wsrt-dzb-data-classic-aips / analysis-wsrt-d http: // / astro / calib / manual / baars.htmlArticle number, page 2 of 19. Gießübel et al.: Polarized synchrotron radiation from the Andromeda Galaxy M31 and background sources at 350 MHz with A = . , B = − . , C = − . , D =+ . RM = + . ± .
14 rad m − and a polarization angleat λ = pa = ◦ ± ◦ .At these wavelengths 3C303 is expected to be 5% polarizedwith an RM of +
15 rad m − (Ger de Bruyn, private communica-tion). This value di ff ers slightly from (but still agrees with) thepublished value of RM = + ± − at GHz frequencies(Simard-Normandin et al. 1981).The results of calibrating 3C303 confirm that the polariza-tion calibration remains constant over the the 12 h of each obser-vation, which means that the ionosphere was reasonably stableduring this time span. In December 2008, solar activity was neara minimum, ionospheric Faraday rotation is thus expected to beonly a few rad m − and stable, which means that any e ff ects canbe handled by the calibration.The calibration has to be made for each channel individually,solving for all elements in the Jones matrices (Hamaker et al.1996), to circumvent problems caused by the so-called 17 MHzripple. This is a variation of the gains with frequency (at a periodof 17 MHz), caused by a standing wave between the dish andthe primary focus of the antennas. It sometimes results in strongfrequency-dependent variations in the spectra of o ff -axis sourcesacross the primary beam and polarization leakages (Popping &Braun 2008; Brentjens 2008).The properties of the calibrators (namely the expected valuesfor the four Stokes parameters) were calculated for each channelindependently and were written into the MODEL_DATA columnof the measurement set. Afterwards the
CASA tasks gaincal , polcal , and applycal were used to calculate and apply thegains to the other calibrators and the M31 fields. This was doneseparately for each channel and each of the four observations. Several rounds of self-calibration (selfcal) were performed be-fore final imaging of the visibility (uv) data of M31. A singleselfcal step consists of imaging and cleaning the uv data to ob-tain a clean model of the field and using that clean model forcalculatin and applying gain solutions to the field. But first Cas-siopeia A and Cygnus A had to be removed from the data, sinceduring imaging side-lobes from both sources were visible withinthe M31 field. The two sources are far away from the point-ing centre, but they are the brightest sources in the sky at thesefrequencies.Sources can be subtracted from the uv data using the so-called peeling method. This is a general method for removingo ff -axis sources from the field, for which good gain solutionscannot be derived with normal selfcal. It requires the followingsteps:1. Obtain a good clean model of the observed field. Usuallya selfcal step (excluding the source(s) to be subtracted) isperformed to be able to produce a better clean-model.2. Subtract that model from the uv data. This step is performedso that no side-lobes from the field interfere with the sourcethat is to be peeled. If a selfcal step was made, the model hasto be subtracted from the CORRECTED_DATA column. Since the gain-solutions from the selfcal step did not include thesource that is being peeled, they will be entirely wrong forthis source. Thus, after subtracting the calibration with thefield sources has to be reversed by inverting the gain tableand applying the inverted gains.3. Use the field-subtracted uv data to obtain a good clean modelof the source that is to be peeled. Usually this involves set-ting the phase centre to the position of the source, and againperforming selfcal on the source in question.4. Subtract the new model from the original uv data. If selfcalwas used, first apply the gain solutions, then subtract andreverse the calibration by again applying the inverted gains.For better results, this can be repeated in an iterative process,since the model for the field will improve after the disturbinginfluences of the peeled source are reduced, leading to a bettersubtraction and thus a better model for the source to be peeled,and so on.Here it was su ffi cient to run selfcal on the field using onlythe brightest sources, subtract that model from the field, ap-ply the inverted gains from the selfcal and then subtract Cas-siopeia A and afterwards subtract Cygnus A without any selfcalsteps. (Both sources are too far away from the original phase-centre to achieve any good solutions.)After peeling, three selfcal runs were performed on eachM31 field (again separately for each channel and each of thefour days). For the initial model, the uv range was restricted tobaselines > . λ to exclude the extended emission and startwith a simple model. Only the brightest sources and uniformweighting were used. The resulting image was cleaned to threetimes the noise-level measured in the Stokes V image for eachchannel individually. In the second run the uv range was stillrestricted, and more sources as well as the bright centre of M31were included. The threshold for cleaning was reduced to twicethe noise. In the final step the uv range was no longer restrictedand the entire M31 was included in the model and cleaned downto the noise-level.This procedure was tested for di ff erent channels across theentire bandwidth and then automated with python to run with-out user-interaction for each channel separately. For the final images the uv range was restricted from 34.5 λ to2801.0 λ . This is the maximum uv range that all channels havein common, and ensures that for all frequencies the spatial res-olution is the same. The images for all four Stokes parameterswere cleaned automatically using Briggs-weighting and a uv ta-per and afterwards were slightly smoothed to a final resolutionof 230 ′′ × ′′ . The Stokes Q and U images are then used forRM synthesis (Sect. 5).For the total power map, the channels were once again in-spected using the images from the automatic imaging run, andobviously bad channels were excluded. Then the uv data wasconcatenated per field and per IF, keeping the individual chan-nels, and imaged manually, resulting in one image per field andper IF. As a last imaging step the two fields had to be mosaickedand primary-beam corrected. Article number, page 3 of 19 ig. 1.
Final total power image of M31 at 90 cm. Contours are from 4 to 128 mJy / beam in steps of 10 mJy / beam. HPBW: 230 ′′ × ′′ ;rms = / beam. The black crosses mark the pointing centres of the two fields. Note: The tick marks may be misleading since the map isrotated with respect to the celestial coordinates, see the coordinate grid in Fig. 2 or Fig.4! Fig. 2.
Spectral index (SI) map between 90 cm and 20 cm overlaid with the same contours as Fig. 1. The SI map was calculated for intensities > σ in Stokes I. According to the ASTRON webpage, the primary beam responsecan be described by the function A ( r ) = cos ( c ν r ) , (2)where r is the distance from the pointing centre in degrees and ν the observing frequency in GHz. The automatic primary beamcorrection in CASA uses a di ff erent function for the WSRT. Theconstant c is only constant for GHz frequencies and is said todecline to c =
66 at 325 MHz and c =
63 at 4995 MHz. It isrelated to the light crossing time across the the aperture (i.e. thee ff ective diameter of the dish, Brentjens 2008). A good primarybeam correction is crucial in our case, because M31 spans twopointings. Since there is no exact value given and recommenda-tions vary (e.g. c =
64, Ger de Bruyn, private communication),we manually determined the best value for our observation. http: // astron.nl / radio-observatory / astronomers / wsrt-guide-observations / wsrt-guide-observations For four arbitrary chosen channels across all bands (IF 0,channel 15; IF 1, channel 83; IF 5, channel 26; and IF 7 channel57), the primary-beam correction was applied for di ff erent val-ues of c . Then the di ff erence between the peak fluxes of each ofthe ten sources in the overlap area of the primary beams betweenthe northern and southern field at the same frequency was calcu-lated and normalized to the flux density in the northern field, f i ( c ) = S i , north ν − S i , south ν S i , north ν . (3)Hence, for each constant c there are ten values f i ( c ) (per chosenchannel) between -1 and +
1. For a perfect primary-beam correc-tion, each value would be equal to 0. The error in the mean ofthese ten values is thus a measure for the quality of the primarybeam correction, σ ( c ) = vt n ( n − n X i (cid:16) f ( c ) − f i ( c ) (cid:17) . (4)For c =
65, all the systematic errors are at their lowest com-mon point and are also all below 10%, which is deemed an ac-
Article number, page 4 of 19. Gießübel et al.: Polarized synchrotron radiation from the Andromeda Galaxy M31 and background sources at 350 MHz ceptable level. Using this value for the primary beam correction,the resulting systematic flux error was estimated to be 7 − I LM ( l ) = P p A ( l − l p ) I p ( l ) P p A ( l − l p ) , (5)where A ( l − l p ) is the primary beam attenuation at distance l fromthe pointing centre l p (eq. 2) and I p is the cleaned image of therespective field.
4. Total emission from M31 at 90 cm
The final total power image is shown in Figure 1. It is auniformly weighted average of the single IF images (see Sec-tion 3.4). Note that even at these low frequencies, no excessradio continuum emission (i.e. a radio halo) is detected aroundM31. This has already been inferred by Gräve et al. (1981).After subtracting all sources < . R = − . of10 . ± . . ◦ . Thiswould still enable us to detect a halo along the minor axis.Figure 2 shows a spectral index (SI) map between our 90 cmmap and the VLA + E ff elsberg 20 cm map by Beck et al. (1998).The spectral index is very similar to that presented by Berkhui-jsen et al. (2003) between 20 cm and 6 cm. Like at GHz fre-quencies, we found on average a slightly steeper spectral indextowards the southern major axis ( α , ≈ − .
7) compared withthe northern major axis ( α , ≈ − . ii regions andshows a flat spectral index of only α , ≈ − .
4. This is evenflatter than the average value found in the SI map by Berkhuijsenet al. (2003) at GHz frequencies. With a constant thermal frac-tion and a constant nonthermal spectral index one would expectthe spectral index to steepen at low frequencies, so this indicatesthat thermal absorption is considerable. We note that missingspacings are no problem for the spectral index here, since theyonly become significant in the faint outermost regions. The over-all spectral index, dominated by the bright regions, is thereforenot a ff ected.A more thorough analysis would require a higher resolutionmap at 90 cm to allow proper subtraction of all point sources andis beyond the scope of this paper.
5. Polarized emission and RM synthesis
The Faraday depth (FD) is proportional to the integral alongthe line of sight over the cosmic-ray electron density n e and thestrength of the line of sight component of the regular magneticfield B (Burn 1966), φ ∝ Z observersource n e B · d l , (6) Previous measurements used the radius interval R = −
16 kpc basedon the old distance estimate of 690 kpc by de Vaucouleurs & de Vau-couleurs (1964), which corresponds to R = − . while the classical RM is an observable quantity that describesthe di ff erence of polarization angles ∆ χ observed at two (ormore) di ff erent wavelengths λ i ,RM = ∆ χλ − λ . (7)The classical RM is equivalent to the FD φ only if there is just abackground source and a dispersive Faraday screen in the fore-ground along the line of sight. If there are several emitting androtating components along the line of sight, the linear relation-ship between ∆ χ and ∆ ( λ ) in eq. 7 does not hold. In addition,there is an ambiguity because the polarization vector could haverotated by n π ( n a natural number) without being noticed.RM synthesis, on the other hand, yields a spectrum F ( φ ) inFD φ , in which each polarization-emitting component along theline of sight will produce a separate signal. Its position in FDcorresponds to the total rotation of the rotating components be-tween the emitting component and the observer along the line ofsight (see Fig. 3). For more details on RM synthesis we refer toBrentjens & de Bruyn (2005) and Heald (2009). (a) Sketch of di ff erent components along the line of sight. Some ofthem are emitting ( Ei ), are Faraday-rotating ( Ri ), or both.(b) Resulting Faraday spectrum of the components depicted in 3a. Fig. 3.
Example for a set of di ff erent polarization-emitting and rotatingcomponents along the line of sight and the resulting Faraday spectrum. E E E , R R R
3, and R ff ects are neglected here for simplicity. RM synthesis was performed on the automatically cleaned Q andU images after mosaicking, using the software by Michiel Bren-tjens (Brentjens & de Bruyn 2005; Brentjens 2008). The result-ing Faraday cube was then cleaned using the RM Clean codefor
MIRIAD (Sault et al. 1995) implemented by one of us (GH)(Heald et al. 2009), using 1 σ as cuto ff level and nmax = ∼
975 iterations.The resolution in FD φ of our observation is given by themeasured full width at half maximum of the RM spread function(RMSF) ϕ = .
12 rad m − . The largest detectable structure inFaraday space is φ max = π/λ min ≈ .
95 rad m − . Article number, page 5 of 19 ote that the terms RM clean and RMSF are misleading be-cause only in exceptional cases RM and FD φ are the same (seeabove). Fig. 4.
Position and Faraday depth of the detected sources on contoursof the total power map.
We compiled a catalogue of all (33) polarized sources de-tected in the Faraday cube in Table A.1. The listed positionscorrespond to the brightest, central pixel of that source in theimage plane at the FD it was detected. The polarized intensi-ties (PI) are the peak intensities in the Faraday cube, and eachpolarization degree ( p ) was calculated with the respective in-tensity at that position in the Stokes I map. The name of thecorresponding source from the Walterbos et al. (1985) catalogueand / or from Taylor et al. (2009) is given if the source is listed. Ifmarked with a ‘?’ the assignment is not entirely certain becausethe position of the peak and that in the catalogue di ff er slightly.This can occur because of uncertainties due to the low resolu-tion in the WSRT data, sources consisting of several unresolvedcomponents, or in fact an erroneous assignment. For compari-son the polarization properties measured at 1.4 GHz (21 cm) byHan et al. (1998) and Taylor et al. (2009) are listed where avail-able. We note that only nine of the 21 sources listed by Han et al.(1998) were detected at 90 cm. The FD derived from RM syn-thesis at 90 cm is denoted by φ , whereas RM denotes Faradayrotation measures obtained from two frequency bands.The average value of φ of all sources in the catalogue is − ±
40 rad m − (where ±
40 rad m − is the standard deviation). Theaverage degree of polarization is 1.36%. Most values are below3%; only five sources are more strongly polarized.An important selection criterion for a source to be includedin the catalogue was a clear recognition as a point-like source inthe cube. Plots of the FD along slices through the source posi-tions in R a / D ec are a good tool for determining the spatial extentof peaks in the Faraday cube. In Appendix B the Faraday spectraof all detected sources are shown. The polarization data (Figures B.1 to B.6) is available in FITS format at the CDS . Some haveseveral components, but these peaks belong to extended struc-tures in the cube that are unrelated to M31 (Sect. 5.3). The errorin FD is estimated by ∆ φ = ϕ S / N , (8)where ϕ is the full width at half maximum of the RMSF and S / N the signal-to-noise ratio of the peak. This is a similar error esti-mation as is used for source detection in imaging (e.g. Fomalont1999). The error in polarized intensity is the noise level of theFaraday cube, which is about 1 / √ N chan times lower than the rmsin the single Q and U images ( N chan is the number of frequencychannels). The lowest S / N >
8, therefore polarization-bias cor-rection can be neglected.Figure 4 shows the position of the detected sources on con-tours of the total power map. The colour corresponds to the mea-sured FD. There are clearly not enough sources to define an RMgrid, which could be used to trace M31 and its magnetic fieldas a foreground screen (Han et al. 1998; Stepanov et al. 2008).However, 37W175b in the north, 37W207B, 37W115 / T1237?and 37W158C in the middle, and 37W058 in the south seemto follow the RM distribution seen in M31 at GHz frequencies(Berkhuijsen et al. 2003).The average RM value for the foreground determined at GHzfrequencies is -93 rad m − (Fletcher et al. 2004), consistent withthe average φ of the 90 cm sample. However, towards the westa number of sources seem to increasingly show a φ of about-50 rad m − . This may be an indication that the foregroundscreen is in fact not constant, but has an FD gradient. However,the western part of the cube is heavily a ff ected by foregroundemission, as can be seen from the Faraday spectra of the sources(Sect. B).Moreover the three sources marked as (triplet)(J003901 + + + ff erent FDs of that feature. The feature emerges at the locationof 37W021 and extends star-shaped to the denoted positions. Itseems unlikely that there is a physical connection to 37W021because of their relatively large angular separation and the lackof counterparts in total intensity. If these features are trulyconnected, they are probably part of the Galactic foreground.There are two ubiquitous features, detected at most pixelsin R a / D ec : a peak at φ = − , which is caused bythe frequency-independent instrumental polarization, and a peakaround φ = −
12 rad m − (sometimes accompanied by one or twomore peaks within the range -30 to 0 rad m − ). This could bepolarized emission from the near side of the foreground mediumof our Milky Way. Since this hampers the detection of signalsfrom background sources or M31 in the Faraday cube, the range ±
30 rad m − is marked grey in all plots. An emitting fore-ground region is expected to be recognizable as an extendedfeature in the Faraday spectrum, but is not entirely visible inour observations. With a shortest wavelength of λ min ≈
80 cm,the widest detectable feature in Faraday spectra has an extentFD max = π/λ min ≈ − . A pair of Faraday componentswith similar heights would indicate an extended FD structure.The quality of our data is not su ffi cient to rule out such struc-tures. A large coverage in λ is needed to recognize componentswith a range of scales in the Faraday spectrum (e.g. Brentjens &de Bruyn 2005; Beck et al. 2012). http://cdsweb.u-strasbg.fr/cgi-bin/qcat?J/A+A/ Article number, page 6 of 19. Gießübel et al.: Polarized synchrotron radiation from the Andromeda Galaxy M31 and background sources at 350 MHz(a) φ = −
44 rad m − (b) φ = −
40 rad m − (c) φ = −
36 rad m − (d) φ = −
32 rad m − Fig. 5.
Frames from the Faraday cube of the triplet feature at di ff erentFDs. Not all sources detected in the Faraday cube have been listed byHan et al. (1998) or Taylor et al. (2009) and vice versa. For the subset of sources appearing in multiple catalogues, the measuredrotation measures and polarized intensities can be compared. InFigure 6 the FD detected in our catalogue is plotted against theRM values found by Han et al. (1998) (green) and Taylor et al.(2009) (red). The solid line indicates where both values coin-cide. Although most sources agree well, several sources deviate.Here they are identified with numbers from 1 to 8, which arealso given behind the names of the corresponding source in Ta-ble A.1. Sources 1 to 6 are systematically o ff set from the solidline towards lower absolute FDs, which suggests a systematice ff ect caused by Faraday depolarization. (cid:0) (cid:1) (cid:2) (cid:3)
50 0RM [rad m (cid:4) ] Han / Taylor (cid:5) (cid:6) (cid:7) (cid:8) (cid:9) [ r a d m (cid:10) ] Han et al. (1998)Taylor et al. (2009)
Fig. 6.
Measured FD from the Faraday cube at 90 cm versus RMvalues measured by Han et al. (1998) and Taylor et al. (2009) at 20 cm.The grey area marks the range of FD that probably is a ff ected by fore-ground emission and instrumental polarization. Dashed lines indicatethe expected rotation measure of the Galactic foreground, dotted linesshow the range spanned by the FWHM of the RMSF. The RM values by Han et al. (1998) and Taylor et al. (2009)were determined in the traditional way, using only two frequen-cies (1.365 GHz and 1.652 GHz by Han et al. 1998, 1.365 GHzand 1.435 GHz by Taylor et al. 2009). Although the frequencypairs are very close, some of their values could still be a ff ectedby the n π ambiguity. However, this would yield an o ff set of ∼ ±
650 rad m − , which is too high to explain the o ff sets seenin Fig. 6. (Moreover Han et al. (1998) and Taylor et al. (2009)do not always in agree (see e.g. source 6 (37W115 / T1237? or37W045 / T1126) where the PI values disagree by an order ofmagnitude.)Inspection of the Faraday spectra of the sources (Figs. B.1–B.6) suggests that most of the spectra of the deviating sourceshave a more complex structure than those where the φ valuematches the RM from previous works. The simple linear ap-proximation of the variation of polarization angle with λ isonly valid for the simplest case of a background source with-out rotation and a constant foreground screen without emission(see above). For example, the spectra of 37W074A / B (Fig-ure B.3, bottom right) and T1061 (Fig. B.6, top left) suggestthe presence of several possibly extended RM structures alongthe line of sight, while the spectra of 37W175b (Fig. B.2) or37W045 / T1126 (Fig. B.4), for instance, are much more simple.Multiple components along the line of sight can easily lead towrong results when measuring the RM between only two fre-quencies, which probably explains the deviations in Fig. 6.
Article number, page 7 of 19 here may also be systematic errors in the output of RMsynthesis (e.g. Farnsworth et al. 2011). However, such ambi-guities are within the FWHM of the RMSF. As can be seen inFigure 6, the o ff set sources are well outside of this range, sothey are unlikely to be caused by such systematic uncertaintiesresulting from RM synthesis of complicated sources. It might be possible that the deviating sources in Fig. 6 are ofa di ff erent type from those near the solid line. If these sourcessu ff er from internal Faraday depolarization, the intrinsic FD atlow frequencies can be entirely di ff erent from that at high fre-quencies, since the visible region of polarized emission becomessmaller (because all the emission from deeper layers is depolar-ized and will no longer contribute to the emission and Faradayrotation). In that case these sources should be more stronglydepolarized than others. In Figure 7 the polarization degreemeasured at 90 cm is plotted against that measured at 20 cmby Han et al. (1998) and Taylor et al. (2009). The deviatingsources are denoted with the same numbers as given in Fig 6.The solid grey line is a linear fit through all points; its slopecorresponds the mean depolarization between 90 cm and 20 cm( DP (90 , = . ± . χ = . p [ % ] Han et al. (1998)Taylor et al. (2009)
Fig. 7.
Measured polarization degree from the Faraday cube at 90 cmversus the values measured by Han et al. (1998) and Taylor et al. (2009)at 20 cm. The solid grey line is a linear fit through all points ( DP = . ± . χ = . The strong depolarization of the background sources couldoriginate (1) in the Galactic foreground, (2) in M31, (3) in inter-vening galaxies along the line of sight, or (4) in the sources them-selves. The condition for model (1) is that the angular extent ofthe source is larger than that of a typical turbulence cell. Noneof our sources shows significant extension in total intensity onthe scale of our telescope beam (4 ′ ), which is much smaller thanthe angular size of a turbulence cell in the Galactic foregroundof typically 50 pc linear size out to several kiloparsecs withinthe Galaxy. Model (2) can be excluded because the amount ofdepolarization does not increase towards the inner part of M31. We conclude that the depolarization occurs in distant interveninggalaxies or within the sources.Similarly, the multiple components seen in the Faraday spec-tra of Figs. B.1–B.6 can hardly originate in the Galactic fore-ground, because they would have to emit in polarization on sim-ilar levels as the background sources themselves (Fig. 3). Theyare either intrinsic features of these sources, for instance, thelobes of radio galaxies, or occur in the turbulent medium of in-tervening galaxies (Bernet et al. 2012). If the source is coveredby a discrete number of turbulent cells in the intervenor or inthe lobe, a corresponding number of components appears in theFaraday spectrum (see Fig. 5 in Bernet et al. (2012)). If unre-solved, these components lead to depolarization. If the numberof components is large, depolarization can be described by dis-persion in Faraday rotation (see equation (11) below).This scenario is supported by the fact that the two sourceswith least depolarization (T1379 and 37W014 / T1119) have sim-ple Faraday spectra without multiple components outside therange of ±
30 rad m − , which is a ff ected by polarized emissionfrom the nearby foreground and by instrumental polarization.Furthermore, the three sources with the highest degrees of polar-ization (T1379, 37W154, and 37W019) also have simple Fara-day spectra. A more detailed analysis is hampered by the smallnumber of sources and by the fact that for most sources no data(distance, optical classification, and spectrum) are available fromthe literature.37W115 / T1237 is a known AGN, consisting of three com-ponents (core and two lobes, Morgan et al. 2013) that are un-resolved at the resolution of our 90 cm data. It is possible thatat low frequencies one of the lobes becomes depolarized, andin turn the polarized emission of the other lobe dominates (theLaing–Garrington e ff ect, Garrington et al. 1988). In this case acomparison with 20 cm data is not practical, since one essen-tially observes two di ff erent source patterns. The same may alsobe the case for radio galaxies at larger distances.Finally, none of our sources fits into the depolarization modelfor compact, steep-spectrum (CSS) sources by Rossetti et al.(2008). This model predicts that the degree of polarizationshould be constant at long wavelengths. The strong depolariza-tion seen in Fig. 7 indicates that our sources are not of type CSS. No extended polarized emission of M31 is visible by eye in theFaraday cube. Internal Faraday dispersion is the dominating de-polarization mechanism towards low frequencies. Changes ofthe polarization angle increase with decreasing frequency. Dueto turbulent cells in the magnetized plasma, the polarized emis-sion becomes increasingly patchy and the polarization angle ran-domized, which is the true cause of the depolarization.Thus it is unclear whether any coherent polarized emissionon large scales can be expected at all (either in the image-plane,or in FD), therefore we propose a new method for uncoveringweak di ff use polarized emission at low frequencies.From observations at GHz frequencies it is known that thepolarized emission in M31 is strongest around the 10-kpc ring,showing an RM range of roughly -200 rad / m to 0 rad / m . Theintrinsic RM in M31 is thus -100 rad / m to +
100 rad / m , shiftedby the RM of the Galactic foreground by ∼ −
93 rad / m . RM syn-thesis allows us to compare the positive and negative FD ranges.An excess signal in the negative FD range, which is confined tothe position of M31, would be a detection of M31 in polariza-tion. Article number, page 8 of 19. Gießübel et al.: Polarized synchrotron radiation from the Andromeda Galaxy M31 and background sources at 350 MHz
Ellipse resol.elem.
I [mJy] PI [mJy] p [%]inner ∼
35 2567 ±
36 3 ± . ± . ∼
45 4100 ±
19 9 ± . ± . ∼
54 2325 ±
22 7 ± . ± . ±
77 19 ± . ± . ∼
63 876 ± − ± ∼
72 201 ± − ± Table 1.
Number of independent resolution elements, integrated totalflux density, residual integrated polarized flux density over the selectedFD range (see text) and degree of polarization. Line 5 gives the totalof inner-, ring- and outer ellipse and the resulting polarization degree.Integration is made over residual histograms (see text), which explainsthe occurrence of negative values.
We used the range of ( − ±
50) rad / m and compared itwith the complementary range of ( + ±
50) rad / m . The se-lected range is thus centred on the FD where we expect most ofthe signals. Since we are in the Faraday-thick regime ( opaquelayer approximation , see e.g. Sokolo ff et al. 1998), the totalrange in rotation measure will be smaller. The restriction to( − ±
50) rad / m also excludes the range in the Faraday cubethat is a ff ected by foreground emission and instrumental polar-ization (see end of Section 5.2.1).For each pixel in the map, we integrated along the Faradayspectrum over the absolute values in the selected positive FDrange and subtracted that result from the integral along the Fara-day spectrum in the selected negative FD range. Since we in-tegrated over the amplitude of a complex-valued function, theresult is by definition positive and contains a positive non-zerocomponent from the noise. The second component in the spec-trum consists of signals from M31 itself plus a few weak, polar-ized point sources.Owing to the rotation measure of the Galactic foreground,the positive FD range does not contain any significant signals ofthe second component. By subtracting the complementary posi-tive FD range, we subtracted the contribution of the noise, whichis equal over the entire spectrum. The residual is therefore an es-timate for the intrinsic polarized signal from M31. The positivebias from the noise component is removed with the subtraction,but due to the nonlinear addition of noise and signals in polar-ized intensity, weak signals are suppressed by our subtractionmethod, and our result is an underestimate.Figure 8 shows the resulting residual map. Since there isstill no clear detection of polarized emission, we integrated overthe residual map in the image plane. After masking the detectedpoint sources, we defined five ellipses with a width of 6 ′ : one onthe exact position where the 10-kpc ring is seen in total power(and where we expect the strongest polarization signal), one in-side the ring, and three outside the ring. The outlines of theellipses are overlaid in Figure 8 and the results are listed in Ta-ble 1. The error in each ellipse was estimated by the standarddeviation of the value in each ellipse multiplied by the sqare-rootof independent resolution elements.We repeated the analysis for the complementary FD ranges ± ±
50 rad m − (not shown here). As expected, no excess ofpolarized emission was found for these FD ranges.Since the excess is confined to the position of polarized emis-sion at GHz frequencies and to the expected FD range, we iden-tified it as polarized emission originating from M31. However,the weakness of the signals does not allow us to measure thedistribution of the FD along the ring and compare this to previ- h m s m m m R.A. (2000.0) (cid:11) (cid:12) (cid:13)(cid:14) (cid:15) (cid:16) D e c . ( . ) Fig. 8.
Residual polarization map after integrating each pixel over theFD ranges ( − ±
50) rad / m and ( + ±
50) rad / m and subtractingthe results. Overlaid are the outlines of the ellipses used to integrate inthe image plane. The outline of the ellipse on the 10-kpc ring is markedin boldface. Detected point sources have been masked (seen as whitepoints). inner ring outer outer2 outer3Ellipse (cid:17) (cid:18) r e s i d u a l P I [ m J y ] Fig. 9.
Integrated residual polarized flux densities for the FD range ± ±
50 rad m − for the di ff erent ellipses (see Table 1). ous results. It must be somewhat uniform or it would be visibleat particular azimuthal ranges in the residual map in Figure 8.The layer of the regular field known to exist from observationsat higher frequencies is broken up by Faraday dispersion intosmall regions emitting in polarization. This is clearly visible inthe PI map at 20 cm by Beck et al. (1998). Instead of one broadstructure in the Faraday spectrum, we expect many components.Our integration collects all these features.Now the depolarization between 6 cm and 90 cm can beestimated. The 6 cm polarized intensity map (Gießübel 2012,Gießübel et al. in prep.) was smoothed to the resolution of the90 cm map and the total polarized flux density within the inner Article number, page 9 of 19 hree ellipses was calculated ( PI ≈ .
296 Jy). The depolariza-tion can then be calculated as in Beck (2007), DP (90 , = ( PI / PI ) ∗ ( ν /ν ) α n , (9)where α n = − . ν i therespective frequency, leading to DP (90 , = . ± . DP (20 , ≈ .
1. Referring to Burn (1966), we can calcu-late the expected depolarization at 90 cm. At these low frequen-cies depolarization is dominated by internal Faraday dispersion. DP int = − exp( − S ) S (10)with S = σ RM λ . (11)The factor σ RM describes the dispersion of the mediumin rad m − . In a simplified model of a magneto-ionic mediumit can be written as (Arshakian & Beck 2011) σ RM = (0 . n e B turb d ) f Ld ≃ (0 . h n e ih B turb i ) Ldf , (12)where n e = h n e i / f is the thermal electron density in cm − withinthe turbulent cells of size d in pc, h n e i is the average electrondensity in the volume along the pathlength traced by the tele-scope beam in pc, f is the filling factor of the cells, h B turb i thestrength of the turbulent field in µ G and L the pathlength alongthe line of sight through the Faraday-active emitting layer. Theterm in brackets describes the rotation measure of a single tur-bulent cell, while f Ld gives the number of cells along the line ofsight.The following quantities were used: h n e i = .
015 cm − (inconcordance with the depolarization caused by internal Faradaydispersion of 0.1, 0.3 and 0.2 measured by Fletcher et al. (2004)for di ff erent rings at 8 −
10 kpc, 10 −
12 kpc and 12 −
14 kpc, re-spectively), a filling factor of f = . B turb = µ G and d =
50 pc (Fletcher et al. 2004). Accord-ing to Fletcher et al. (2004), the scale height of the synchrotron-emitting layer is h s y n =
300 pc. The scale height was measuredfrom the mid-plane. The pathlength through the entire layeralong the line of sight is thus L = h s y n cos( i ) ≈ , (13)where i = ◦ (Chemin et al. 2009) is the inclination angle ofthe disk.The classical formula by Burn (1966) would yield DP (90 , ≈ . p ≈ . λ to λ for longer wavelengths and equa-tion (11) becomes S = σ RM λ (14)(see also Sokolo ff et al. 1998; Arshakian & Beck 2011). This isbecause the dispersion causes the (spatial) correlation length ofthe polarized emission to decrease with increasing wavelength,until it drops below the size of the turbulent cells. In this picture, no extended structures would be visible from M31 in the Faradaycube, which is consistent with our observations.Since S ≫
1, eq. 10 becomes DP = σ RM λ , (15)which (using the values given above) results in DP = . p = . DP (90 , = . ± . p = . ± .
05% are thus between the predictions of Burn and Trib-ble (but closer to the latter, which is just outside our 3 σ uncer-tainty bound). Our detection is a lower limit for the true polar-ized signal. If some polarized emission from M31 is extendedin Faraday space, we underestimate the detected polarized fluxeven more because of missing zero frequencies (see Brentjens& de Bruyn 2005). We are therefore unable to comment on thevalidity of the Tribble formula, but we can confidently rule out a λ dependence of the depolarization for low frequencies.
6. Conclusions
We have presented the first detection of polarized emission froma nearby galaxy at 90 cm. The polarized emission is mainly con-fined to the 10-kpc ring. At these low frequencies the emissioncomes from cosmic-ray electrons with lower energies. They suf-fer less from energy losses and therefore are able to propagatefarther out from the disk or into a halo than at higher frequen-cies. However, no signs of a radio halo can be seen in our totalpower map (Fig. 1). The reason may be the preferred propa-gation of cosmic-ray electrons along the highly ordered field inthe ring and suppression of di ff usion perpendicular to the ring(Fletcher et al. 2004).Depolarization by internal Faraday dispersion becomesstrong at long wavelengths. Our observations show that thestrong λ wavelength dependence predicted by Burn (1966) un-derestimates the remaining polarization at 325 MHz. We wereunable to conclude whether a λ depolarization law as predictedby Tribble (1991) is correct, but we note that the correspondingDP does fall within our 3 σ error margin. The amount of de-polarization furthermore depends on the magnetic field strength,thermal electron density, and the pathlength along the line ofsight and can accordingly be very di ff erent in other galaxies,depending on the conditions. A detection of di ff use polarizedemission from nearby galaxies with LOFAR will be di ffi cult andrequires very high sensitivity, achievable with the RM synthe-sis technique in combination with a broad bandwidth. With thesame parameters in eq. (15) as used in Sect. 5.3, depolariza-tion in the LOFAR high band around 150 MHz will be aboutten times stronger. Searches for di ff use polarized emission ingalaxies with LOFAR are most likely to yield detections in outerdisks and halos, where σ RM is expected to be much smaller thanin the ring of M31.Using polarized background sources to probe the magneticfields of nearby galaxies as a foreground screen seems a morepromising prospect. Hence, we compiled a catalogue of all po-larized point-like sources. The RM synthesis analysis resulted in33 detections, but fewer than 50% of the sources listed by Hanet al. (1998) at 20 cm were detected at 90 cm. Fig. 4 shows thatthe number of detected sources is not su ffi cient to provide a gridof sources for probing the magnetic field of M31 as a foregroundscreen to these sources. According to Stepanov et al. (2008), fora galaxy such as M31, about 20 polarized sources on a cut alongthe projected minor axis would be needed to detect the domi-nant field structure (the number depends on the inclination of Article number, page 10 of 19. Gießübel et al.: Polarized synchrotron radiation from the Andromeda Galaxy M31 and background sources at 350 MHz the galaxy, for an inclination of 45 ◦ a total number of ten wouldbe su ffi cient), but the 33 sources detected here are scattered overthe entire field of view. Furthermore, the estimate by Stepanovet al. (2008) assumed a constant contribution of the Milky Wayforeground, which is not the case here (see Section 5.2.1). Inconclusion, the number of polarized sources from our 90 cm ob-servations is insu ffi cient.A higher angular resolution will help with the detection ofmore sources, since fewer features extended in RA / DEC (blend-ing with the emission of sources) will be present in the Faradaycube. LOFAR can provide the necessary resolution, both angularand in FD.The analysis in Section 5.2.2 showed that a more detailedknowledge about the structure of the sources is necessary to in-terpret their Faraday spectra. About 30% of the sources from theliterature showed a systematically di ff erent FD at 90 cm than at20 cm. Additional studies are needed to understand the cause ofthis deviation. A possible explanation is the presence of radiolobes, as discussed in Sect. 5.2.2. The systematics of the devia-tion (see Fig. 6 in Sect. 5.2.2) indicates that it may be possible todefine distinct classes of sources and select those suitable for abackground grid for future observations with LOFAR. The depo-larization of the sources is on average DP (90 , = . ± . Acknowledgements.
The authors would like to thank Björn Adebahr and CarlosSotomayor (Ruhr-Universität Bochum) for their initial help with calibration andpeeling in
CASA , and Elly M. Berkhuijsen for valuable comments. We are alsograteful for the comments of the anonymous referee, which helped to improveour paper. RB acknowledges support from DFG FOR1254. TGA acknowledgessupport by DFG project number Os 177 / References
Arshakian, T. G. & Beck, R. 2011, MNRAS, 418, 2336Beck, R. 1982, A&A, 106, 121Beck, R. 2005, in American Institute of Physics Conference Series, Vol. 784,Magnetic Fields in the Universe: From Laboratory and Stars to PrimordialStructures., ed. E. M. de Gouveia dal Pino, G. Lugones, & A. Lazarian, 343–353Beck, R. 2007, A&A, 470, 539Beck, R., Berkhuijsen, E. M., & Hoernes, P. 1998, A&AS, 129, 329Beck, R., Berkhuijsen, E. M., & Wielebinski, R. 1978, A&A, 68, L27Beck, R., Brandenburg, A., Moss, D., Shukurov, A., & Sokolo ff , D. 1996,ARA&A, 34, 155Beck, R., Frick, P., Stepanov, R., & Sokolo ff , D. 2012, A&A, 543, A113Berkhuijsen, E. M., Beck, R., & Hoernes, P. 2003, A&A, 398, 937Bernet, M. L., Miniati, F., & Lilly, S. J. 2012, ApJ, 761, 144Brentjens, M. A. 2008, A&A, 489, 69Brentjens, M. A. & de Bruyn, A. G. 2005, A&A, 441, 1217Burn, B. J. 1966, MNRAS, 133, 67Chemin, L., Carignan, C., & Foster, T. 2009, ApJ, 705, 1395Cornwell, T. J., Holdaway, M. A., & Uson, J. M. 1993, A&A, 271, 697de Vaucouleurs, G. & de Vaucouleurs, A. 1964, Ref. Catalogue of Bright Galax-ies (Austin: Univ. of Texas Press) Farnsworth, D., Rudnick, L., & Brown, S. 2011, AJ, 141, 191Fletcher, A. 2010, in Astronomical Society of the Pacific Conference Series, Vol.438, Astronomical Society of the Pacific Conference Series, ed. R. Kothes,T. L. Landecker, & A. G. Willis, 197Fletcher, A., Berkhuijsen, E. M., Beck, R., & Shukurov, A. 2004, A&A, 414, 53Fomalont, E. B. 1999, in Astronomical Society of the Pacific Conference Series,Vol. 180, Synthesis Imaging in Radio Astronomy II, ed. G. B. Taylor, C. L.Carilli, & R. A. Perley, 301Garrington, S. T., Leahy, J. P., Conway, R. G., & Laing, R. A. 1988, Nature, 331,147Gießübel, R. 2012, PhD dissertation, Universität zu Köln, Cuvillier Verlag Göt-tingen, ISBN 978-3-95404-308-8Gräve, R., Emerson, D. T., & Wielebinski, R. 1981, A&A, 98, 260Hamaker, J. P., Bregman, J. D., & Sault, R. J. 1996, A&AS, 117, 137Han, J. L., Beck, R., & Berkhuijsen, E. M. 1998, A&A, 335, 1117Heald, G. 2009, in IAU Symposium, Vol. 259, IAU Symposium, 591–602Heald, G., Braun, R., & Edmonds, R. 2009, A&A, 503, 409Morgan, J. S., Argo, M. K., Trott, C. M., et al. 2013, ArXiv e-printsO ff ringa, A. R., de Bruyn, A. G., Biehl, M., et al. 2010, MNRAS, 405, 155Perley, R. A. & Taylor, G. B. 2003, NRAOPopping, A. & Braun, R. 2008, A&A, 479, 903Riess, A. G., Fliri, J., & Valls-Gabaud, D. 2012, ApJ, 745, 156Rossetti, A., Dallacasa, D., Fanti, C., Fanti, R., & Mack, K.-H. 2008, A&A, 487,865Sault, R. J., Staveley-Smith, L., & Brouw, W. N. 1996, A&AS, 120, 375Sault, R. J., Teuben, P. J., & Wright, M. C. H. 1995, in Astronomical Society ofthe Pacific Conference Series, Vol. 77, Astronomical Data Analysis Softwareand Systems IV, ed. R. A. Shaw, H. E. Payne, & J. J. E. Hayes, 433Scaife, A. M. M. & Heald, G. H. 2012, MNRAS, L434Simard-Normandin, M., Kronberg, P. P., & Button, S. 1981, ApJS, 45, 97Sokolo ff , D. D., Bykov, A. A., Shukurov, A., et al. 1998, MNRAS, 299, 189Stepanov, R., Arshakian, T. G., Beck, R., Frick, P., & Krause, M. 2008, A&A,480, 45Taylor, A. R., Stil, J. M., & Sunstrum, C. 2009, ApJ, 702, 1230Tribble, P. C. 1991, MNRAS, 250, 726van Haarlem, M. P., Wise, M. W., Gunst, A. W., et al. 2013, ArXiv e-printsVilardell, F., Ribas, I., Jordi, C., Fitzpatrick, E. L., & Guinan, E. F. 2010, A&A,509, A70Walterbos, R. A. M. & Braun, R. 1994, ApJ, 431, 156Walterbos, R. A. M., Brinks, E., & Shane, W. W. 1985, A&AS, 61, 451 Article number, page 11 of 19 pp e nd i x A : C a t a l ogu e o f d e t ec t e d s ou r ces Name R a D ec φ [rad / m ] PI [mJy] p [%] RM [rad / m ] PI [mJy] p [%] RM [rad / m ] PI [mJy] p [%](Han et al. 1998) (Taylor et al. 2009)J005058 + ± ± + ± ± ± ± ± ± ± (5) ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± (1) ± ± ↓ ±
119 4.56 ± ± ± ± ± ± / T1332 00h45m12.3s 41d12m20.8s -86.0 ± ± ± ± ± ± ± (2) ± ± ± ± + ± ± ↓ ± ± ↓ ± ± (8) ± ± ↓ ± ± ± ± ± ± ↓ ± ±
006 337W115 / T1237? (6) ± ± ↓ ± ± ± ± ± ± ± ↓ / B (4) ± ± ± ± ± / T1126 00h39m55.9s 41d11m48.8s -100.0 ± ± ± ± ± ± ± ± ± ↓ + ± ± + ± ± ± ± ↓ + ± ± + ± ± ± ± / T1119 00h38m25.1s 41d37m53.3s -110.0 ± ± ↓ ± ± ± ± ± ± ± ± ± ± (3) ± ± ± ± ± ± ± ↓ ± ± ± (7) ± ± ± ± ± Table A.1.
Catalogue of the detected Faraday sources. Numbers in brackets behind the names ( (1)(2)(3)(4)(5)(6)(7)(8) ) are sources with deviating values with regard to the literature, seeSection 5.2.2 and Figures 6 and 7. J003901 + + + p given). Values of p marked with ↓ denote upper limits, assuming a reasonable calibration limit of p = A r ti c l e nu m b e r , p a g e f . Gießübel et al.: Polarized synchrotron radiation from the Andromeda Galaxy M31 and background sources at 350 MHz Appendix B: Faraday spectra of the sources detected at λ
90 cm
The following pages show the Faraday spectra of the sources detected at 90 cm. We refer to Section 5.2.1 for details. The peakassociated with the detected source is marked with a blue arrow and a blue error bar. Green error bars and dashed lines showliterature values from Han et al. (1998), the red errorbars and dashed lines from Taylor et al. (2009) where available.An important selection criterion for a source to be included in the catalogue was a clear recognition as a point-like source in thecube (in R a / D ec ). This is why sometimes higher peaks than the one marked as detection are present in the spectra. Any other peaksthat possibly appear in the spectra are partly extended features.The grey area marks the range of FD ( ±
30 rad m − ) that probably is a ff ected by emission from the nearby foreground of theMilky Way and by instrumental polarization.Like in Table A.1, the sources are ordered from east to west. Article number, page 13 of 19 (cid:20) (cid:21)
100 0 100 200 300 (cid:22) [rad m (cid:23) F ( (cid:24) ) [ J y / b e a m / (cid:25) ] J005058+420637RA 00h50m58.2s DEC 42d 6m37.1s (cid:26) (cid:27) (cid:28)
100 0 100 200 300 (cid:29) [rad m (cid:30) F ( (cid:31) ) [ J y / b e a m / ] J005000+414836RA 00h50m00.8s DEC 41d48m36.0s ! "
100 0 100 200 300 $ [rad m % F ( & ) [ J y / b e a m / ’ ] T1421RA 00h48m43.9s DEC 40d45m36.7s ( ) *
100 0 100 200 300 + [rad m , F ( - ) [ J y / b e a m / . ] T1407RA 00h48m13.0s DEC 40d22m57.2s /
100 0 100 200 300 [rad m F ( ) [ J y / b e a m / ] T1375?RA 00h46m56.1s DEC 40d32m54.5s
100 0 100 200 300 [rad m : F ( ; ) [ J y / b e a m / < ] T1379?RA 00h46m55.1s DEC 40d37m24.6s F i g . B . . F a r a d a y S p ec t r a o f t h e d e t ec t e d s ou r ce s . S ee t e x t onp r e v i ou s p a g e f o r d e t a il s . A r ti c l e nu m b e r , p a g e f . G i e ßüb e l e t a l . : P o l a r i ze d s yn c h r o t r on r a d i a ti on fr o m t h e A nd r o m e d a G a l a xy M a ndb ac kg r ound s ou r ce s a t M H z = > ?
100 0 100 200 300 @ [rad m A F ( B ) [ J y / b e a m / C ] D E F
100 0 100 200 300 G [rad m H F ( I ) [ J y / b e a m / J ] K L M
100 0 100 200 300 N [rad m O F ( P ) [ J y / b e a m / Q ] R S T
100 0 100 200 300 U [rad m V F ( W ) [ J y / b e a m / X ] Y Z [
100 0 100 200 300 \ [rad m ] F ( ^ ) [ J y / b e a m / _ ] J004440+424938RA 00h44m40.6s DEC 42d49m38.9s ‘ a b
100 0 100 200 300 c [rad m d F ( e ) [ J y / b e a m / f ] F i g . B . . F a r a d a y S p ec t r a o f t h e d e t ec t e d s ou r ce s ( c on ti nu e d ) . A r ti c l e nu m b e r , p a g e ff
100 0 100 200 300 c [rad m d F ( e ) [ J y / b e a m / f ] F i g . B . . F a r a d a y S p ec t r a o f t h e d e t ec t e d s ou r ce s ( c on ti nu e d ) . A r ti c l e nu m b e r , p a g e ff h i
100 0 100 200 300 j [rad m k F ( l ) [ J y / b e a m / m ] n o p
100 0 100 200 300 q [rad m r F ( s ) [ J y / b e a m / t ] u v w
100 0 100 200 300 x [rad m y F ( z ) [ J y / b e a m / { ] | } ~
100 0 100 200 300 (cid:127) [rad m (cid:128) F ( (cid:129) ) [ J y / b e a m / (cid:130) ] (cid:131) (cid:132) (cid:133)
100 0 100 200 300 (cid:134) [rad m (cid:135) F ( (cid:136) ) [ J y / b e a m / (cid:137) ] (cid:138) (cid:139) (cid:140)
100 0 100 200 300 (cid:141) [rad m (cid:142) F ( (cid:143) ) [ J y / b e a m / (cid:144) ] F i g . B . . F a r a d a y S p ec t r a o f t h e d e t ec t e d s ou r ce s ( c on ti nu e d ) . A r ti c l e nu m b e r , p a g e f . G i e ßüb e l e t a l . : P o l a r i ze d s yn c h r o t r on r a d i a ti on fr o m t h e A nd r o m e d a G a l a xy M a ndb ac kg r ound s ou r ce s a t M H z (cid:145) (cid:146) (cid:147)
100 0 100 200 300 (cid:148) [rad m (cid:149) F ( (cid:150) ) [ J y / b e a m / (cid:151) ] (cid:152) (cid:153) (cid:154)
100 0 100 200 300 (cid:155) [rad m (cid:156) F ( (cid:157) ) [ J y / b e a m / (cid:158) ] (cid:159) (cid:160) ¡
100 0 100 200 300 ¢ [rad m £ F ( ⁄ ) [ J y / b e a m / ¥ ] ƒ § ¤
100 0 100 200 300 ' [rad m “ F ( « ) [ J y / b e a m / ‹ ] J003901+410928 (triplet)RA 00h39m01.6s DEC 41d 9m28.0s › fi fl
100 0 100 200 300 (cid:176) [rad m – F ( † ) [ J y / b e a m / ‡ ] J003847+412056 (triplet)RA 00h38m47.6s DEC 41d20m56.3s · (cid:181) ¶
100 0 100 200 300 • [rad m ‚ F ( „ ) [ J y / b e a m / ” ] F i g . B . . F a r a d a y S p ec t r a o f t h e d e t ec t e d s ou r ce s ( c on ti nu e d ) . A r ti c l e nu m b e r , p a g e ff
100 0 100 200 300 • [rad m ‚ F ( „ ) [ J y / b e a m / ” ] F i g . B . . F a r a d a y S p ec t r a o f t h e d e t ec t e d s ou r ce s ( c on ti nu e d ) . A r ti c l e nu m b e r , p a g e ff … ‰
100 0 100 200 300 (cid:190) [rad m ¿ F ( (cid:192) ) [ J y / b e a m / ` ] J003825+411538 (triplet)RA 00h38m25.3s DEC 41d15m38.2s ´ ˆ ˜
100 0 100 200 300 ¯ [rad m ˘ F ( ˙ ) [ J y / b e a m / ¨ ] J003830+415409RA 00h38m30.8s DEC 41d54m09.2s (cid:201) ˚ ¸
100 0 100 200 300 (cid:204) [rad m ˝ F ( ˛ ) [ J y / b e a m / ˇ ] — (cid:209) (cid:210)
100 0 100 200 300 (cid:211) [rad m (cid:212) F ( (cid:213) ) [ J y / b e a m / (cid:214) ] (cid:215) (cid:216) (cid:217)
100 0 100 200 300 (cid:218) [rad m (cid:219) F ( (cid:220) ) [ J y / b e a m / (cid:221) ] (cid:222) (cid:223) (cid:224)
100 0 100 200 300 Æ [rad m (cid:226) F ( ª ) [ J y / b e a m / (cid:228) ] T1067RA 00h37m09.1s DEC 41d34m56.1s F i g . B . . F a r a d a y S p ec t r a o f t h e d e t ec t e d s ou r ce s ( c on ti nu e d ) . A r ti c l e nu m b e r , p a g e f . G i e ßüb e l e t a l . : P o l a r i ze d s yn c h r o t r on r a d i a ti on fr o m t h e A nd r o m e d a G a l a xy M a ndb ac kg r ound s ou r ce s a t M H z (cid:229) (cid:230) (cid:231)
100 0 100 200 300 Ł [rad m Ø F ( Œ ) [ J y / b e a m / º ] T1061RA 00h37m05.0s DEC 41d36m10.3s (cid:236) (cid:237) (cid:238)
100 0 100 200 300 (cid:239) [rad m (cid:240) F ( æ ) [ J y / b e a m / (cid:242) ] T1006RA 00h35m25.2s DEC 39d43m32.2s (cid:243) (cid:244) ı
100 0 100 200 300 (cid:246) [rad m (cid:247) F ( ł ) [ J y / b e a m / ø ] T974RA 00h34m29.2s DEC 40d37m34.2s F i g . B . . F a r a d a y S p ec t r a o f t h e d e t ec t e d s ou r ce s ( c on ti nu e d ) . A r ti c l e nu m b e r , p a g e ff