Revisiting Rotation Measures from the Canadian Galactic Plane Survey: the Magnetic Field in the Disk of the Outer Galaxy
C. L. Van Eck, J. C. Brown, A. Ordog, R. Kothes, T. L. Landecker, B. Cooper, K. M. Rae, D. A. Del Rizzo, A. D. Gray, R. Ransom, R. I. Reid, B. Uyaniker
DDraft version February 8, 2021
Typeset using L A TEX twocolumn style in AASTeX63
Revisiting Rotation Measures from the Canadian Galactic Plane Survey: the Magnetic Field in theDisk of the Outer Galaxy
C.L. Van Eck , J.C. Brown , A. Ordog ,
2, 3, 4
R. Kothes , T.L. Landecker , B. Cooper, K.M. Rae, D.A. Del Rizzo, A.D. Gray , R. Ransom ,
5, 4
R.I Reid , and B. Uyaniker Dunlap Institute for Astronomy and Astrophysics, University of Toronto, 50 St. George Street, Toronto, ON M5S 3H4, Canada Department of Physics and Astronomy, University of Calgary, Calgary, Alberta, Canada T2N 1N4 Department of Computer Science, Math, Physics, & Statistics, Irving K. Barber Faculty of Science, University of British Columbia,Okanagan Campus, Kelowna, BC V1V 1V7 Canada National Research Council Canada, Herzberg Research Centre for Astronomy and Astrophysics, Dominion Radio AstrophysicalObservatory, PO Box 248, Penticton, BC, Canada V2A 6J9 Department of Physics and Astronomy, Okanagan College, 583 Duncan Avenue West, Penticton, BC V2A 8E1, Canada Department of Information Technology, Mayo Clinic and Foundation, 200 1st Street SW, Rochester, MN, USA 55905 DataSpeckle Scientific Inc., Kelowna, BC, Canada
AbstractFaraday rotation provides a valuable tracer of magnetic fields in the interstellar medium; catalogsof Faraday rotation measures provide key observations for studies of the Galactic magnetic field. Wepresent a new catalog of rotation measures derived from the Canadian Galactic Plane Survey, coveringa large region of the Galactic plane spanning 52 ◦ < l < 192 ◦ , -3 ◦ < b < 5 ◦ , along with northern andsouthern latitude extensions around l ≈ ◦ . We have derived rotation measures for 2234 sources (4 ofwhich are known pulsars), 75% of which have no previous measurements, over an area of approximately1300 square degrees. These new rotation measures increase the measurement density for this region ofthe Galactic plane by a factor of two. INTRODUCTIONMagnetic fields are an important component of the in-terstellar medium (ISM) in terms of dynamics and evo-lution, with typical energy densities comparable to theturbulence and significantly exceeding the thermal en-ergy (Beck 2007). As a result, magnetic fields have beenfound to play a role in many ISM processes, includingthe acceleration and confinement of cosmic rays (Aha-ronian et al. 2012), cloud collapse in the early stagesof star formation (Padoan & Nordlund 2011), and thevertical structure of the Galactic disk (Boulares & Cox1990).Several different methods of observing magnetic fieldsin the ISM have been used to explore different proper-ties of these magnetic fields. Among these is the mea-surement of Faraday rotation, which has been used tostudy large scale structure in Galactic magnetic fields(e.g., Han et al. 2006; Brown et al. 2007; Sun et al.2008). Faraday rotation measures (RMs) of polarizedradio sources give the strength of the Faraday rotation
Corresponding author: C.L. Van [email protected] between a source and the observer, which in turn givesinformation on the magnetic field along the line of sight.RMs are determined by observing how the polariza-tion angle changes with frequency; for a Faraday-simplesource , the relationship between polarization angle ( ψ )and wavelength ( λ ) is ψ = ψ + λ (cid:34) .
812 rad m − (cid:90) (cid:126)d (cid:16) n e cm − (cid:17) (cid:32) (cid:126)B µ G (cid:33) · (cid:32) (cid:126)dl pc (cid:33)(cid:35) = ψ + λ RM , where ψ is the polarization angle before Faraday rota-tion, n e is the density of free electrons, (cid:126)B is the magneticfield, (cid:126)dl is a differential element of the radiation path,and the integral is evaluated along the path from thesource at a location (cid:126)d to the observer.RMs of pulsars and extragalactic sources thus provideinformation on the magnetic field and free electron den-sity along the line of sight to each source. Combined Faraday-simple refers to a line of sight with a single dominantsource of polarized emission, with all of that emission undergo-ing the same amount of Faraday rotation. Lines of sight withmultiple polarized components with different Faraday rotationare called Faraday-complex. a r X i v : . [ a s t r o - ph . GA ] F e b Van Eck et al. with models of the free electron density (determinedthrough other tracers such as dispersion measures), RMscan be used to construct models of the magnetic field.Since each RM gives information for only its specificline of sight, and includes contributions from the small-scale fluctuations in the magnetic field and free electrondensity, many RMs on adjacent lines of sight (and atdifferent distances, when pulsars are used) are neededto disentangle structures on different scales and to sta-tistically remove the effects of small scale fluctuations.For this reason having a high sky density of RMs isvery important in maximizing the value and reliabilityof analysis using RMs, and surveys that produce largenumbers of RMs are needed.Faraday rotation measure surveys can be loosely di-vided into three generations. The first generation RMsurveys, encompassing most of those surveys conductedprior to the 1990’s, are characterized by small numbersof objects (typically log( N ) ≈ − ), very few frequencychannels (often observed asynchronously and with differ-ent receivers) with a correspondingly high vulnerabilityto nπ ambiguities, and low sky-surface density of RMs.Examples of such surveys are Vallee & Kronberg (1975),Rudnick & Jones (1983), and Clegg et al. (1992). Thesecond generation RM surveys improved on this by hav-ing larger source counts ( log( N ) ≈ − ), greater band-width divided into more frequency channels to overcome nπ ambiguities and in some cases to enable the use ofRM synthesis (Brentjens & de Bruyn 2005), as well ashigher sky density and/or larger sky coverage. Exam-ples include Brown et al. (2003), Brown et al. (2007),Taylor et al. (2009), Mao et al. (2010), and Van Ecket al. (2011).The third generation RM surveys, which have just be-gun, take full advantage of advances in receiver and com-puter technology to have hundreds to thousands of fre-quency channels with high total fractional bandwidth,allowing the use of RM synthesis and other advancedtechniques to explore additional polarization propertieslike Faraday complexity (e.g., Brown et al. 2019). Thisgeneration will include large all-sky surveys, in partic-ular the Very Large Array Sky Survey (VLASS, Maoet al. 2014), the POlarization Sky Survey of the Uni-verse’s Magnetism (POSSUM, Gaensler et al. 2010),and ultimately surveys with the Square Kilometre Ar-ray (SKA, Heald et al. 2020). Each of those surveyswill massively increase the number of measured RMs( log( N ) ≈ − . ). The third generation will also in-clude high-precision RM surveys using low frequency ra-dio telescopes (e.g., Van Eck et al. 2018; Riseley et al.2018), and surveys with very high fractional bandwidths (e.g., Shanahan et al. 2019; Schnitzeler et al. 2019; Maet al. 2020).The Canadian Galactic Plane Survey (CGPS) was amulti-wavelength observing campaign to study the inter-stellar medium in the region of the Galactic plane visiblefrom the northern hemisphere (Taylor et al. 2003). Thisincluded a radio polarization survey performed using theSynthesis Telescope at the Dominion Radio Astrophys-ical Observatory (DRAO-ST; Landecker et al. 2010).Observations for the survey began in 1995 and continuedin several phases until 2009, and covered the Galacticdisk from 52 ◦ < l < 192 ◦ , -3 ◦ < b < 5 ◦ as well as an ex-tension to higher latitudes above the plane covering 101 ◦ < l < 116 ◦ , 5 ◦ < b < 17.5 ◦ . Later observations addeda southern latitude extension (SLE) covering 100 ◦ < l < 111 ◦ , -10 ◦ < b < -3 ◦ . A rotation measure catalog ofpart of the first phase of the CGPS (82 ◦ < l < 96 ◦ , 115 ◦ < l < 147 ◦ ) was produced by Brown et al. (2003).In this paper we report on the search for compact po-larized sources with reliable RMs within the full CGPSand SLE data, and present a catalog of the resultingmeasurements; this may be the last large catalog of thesecond generation RM surveys. In Sect. 2 we describethe CGPS polarization data in more detail, and describeour method of determining which sources had significantpolarization and measuring their RMs. In Sect. 3 wepresent the resulting catalog of 2234 polarized sources,and compare this catalog with previously reported RMdata. Sect. 4 presents some analysis of the Galacticmagnetic field using our catalog and demonstrates theunique value of this catalog. Finally we summarize andpresent our conclusions in Sect. 5. DATA AND RM DETERMINATION2.1.
CGPS data
The details of the CGPS 1.4 GHz observations andsubsequent processing through to the final images arereported in full detail by Landecker et al. (2010); wehighlight a few key parameters in Table 2.1. For ouranalysis, we have used the data from the DRAO-ST only,without the single-dish data from the DRAO 26-m JohnGalt telescope and Effelsberg 100-m telescope. The datafor the SLE region were processed in the same way asthe CGPS data. The data were supplied in the form ofmosaic images, each 5.1 ◦ square, with a single channel-averaged Stokes I image and Stokes Q and U imagesfor each channel. The SLE region was combined into asingle 10 ◦ square mosaic.2.2. Polarized source identification
The identification of polarized sources and determina-tion of their RMs followed the method of Brown et al.
GPS Rotation measure Catalogue Parameter Value (cid:48)(cid:48) × (cid:48)(cid:48) cosec δ Nominal sensitivity 0.23-0.30 mJy/beam rms
Table 1.
Key Observational Parameters for the CGPS Data. (2003), with some improvements. For completeness, themethod is described in full below, and a flowchart of themethod is shown in Fig. 1.The first step was the identification of polarized sourcecandidates. We began with a Stokes I source list, andsearching each source location for statistically signifi-cant polarized intensity. Within the main CGPS region,we used the CGPS Stokes I source catalog from Tayloret al. (2017) as the input source list; for the SLE regionwe used the Aegean source-finder (Hancock et al. 2012,2018) with the default parameters to generate an inputsource list.Each Stokes I source in the input lists was assessed forstatistically significant polarized emission on the sourcelocation. This was made difficult by the combinationof two related challenges: the presence of diffuse polar-ized emission in many regions, as well as the position-dependent noise level across the survey. To account forthese factors, we determined the local noise and fore-ground around each source. The local off-source regionaround each source was defined as an ellipse with theshape of the synthesized beam (calculated at the sourcelocation, as the beam shape changes significantly withdeclination) and size equal to 4 times the full width athalf maximum (FWHM) of the beam in each dimen-sion (producing a beam-shaped region with an area of16 beams). Pixels within this region with Stokes I lev-els above 1.2 mJy/beam (approximately 5 sigma acrossmost of the survey) were classified as ‘source pixels’and those below this threshold as ‘off-source/foregroundpixels’; this prevented any polarization from the targetsource or neighbouring sources from being included inthe foreground calculations. Before calculating the noiseand foreground in the off-source pixels, we required thatthere be at least 5 beam-areas worth of pixels in theoff-source region; if there were insufficient pixels we in-creased the size of the local-region ellipse by 1 beamFWHM in each dimension until sufficient off-source pix-els were present, or until the ellipse passed 10 FWHM ineach dimension in which case the source was discardedas being part of an extended Stokes I object. Within the off-source pixels, the outlier-resistant mean and standard deviation were calculated for eachchannel’s Stokes Q and U . The means were taken as theforeground polarization around the source ( Q fg , U fg ),and the root-mean-square of the standard deviations(averaging over channels and Stokes Q and U ) was takenas the local value for the noise combined with the un-certainty in the foreground estimate.Within the on-source pixels, the noise ( σ QU ) was cal-culated pixel-wise as the quadrature sum of the localnoise determined from the off-source pixels and an in-strumental leakage term, which was set equal to 0.3% ofthe Stokes I map (Brown 2002): σ QU = (cid:113) σ + (0 . I ) . The foreground-subtracted and de-biased polarized in-tensity was calculated per channel, using a modifiedform of the equation of Wardle & Kronberg (1974), as P = (cid:113) ( Q − Q fg ) + ( U − U fg ) − σ QU . This value was averaged over the 4 channels, and thendivided by σ QU / √ (where the factor of √ accountsfor the decreased noise in the channel-averaged map) toproduce a polarized intensity signal-to-noise ratio map.If any of the on-source pixels were found to have a po-larized signal-to-noise ratio greater than 5, the sourcewas classified as a candidate polarized source, and wasprocessed through the RM determination and additionaltesting described in the next section.This procedure produced a large number of false pos-itives. In many cases, polarized pixels belonging toother neighboring sources would be included as on-source pixels, causing sources with polarized neighborsto be labeled as candidates. Another common causewas sources associated with extended Stokes I emission,which would cause many pixels to be above the thresh-old and thus count as on-source, increasing the oddsof encountering a high-polarization outlier in the noisedistribution. Since polarized intensity does not followGaussian statistics, especially if interferometric imageartifacts are present, the S:N>5 threshold is not as re-strictive as might be ordinarily expected (George et al.2012). We considered this high number of false positivesas tolerable, since subsequent quality control tests done Points more than twice the median absolute deviation away fromthe median were rejected before computing the mean and stan-dard deviation. This is approximately equivalent to sigma clip-ping with a 3-sigma cutoff, but is more robust against extremeoutliers which removes the need for iterating the clipping proce-dure.
Van Eck et al.
Per Stokes I Source:Per Candidate:
Extract region around Stokes I source.Identify ‘source’ pixels: Stokes I > 1.2 mJy/beamFit a Gaussian to (foreground-subtracted, debiased) polarized intensity mapIdentify foreground pixels:Draw annulus 4x larger in diameter than beam.‘Foreground’ pixels:Stokes I < 1.2 mJy/beam ≥ No Yes
Calculate outlier-resistant mean and σ of foreground pixels Stokes Q and U (per channel); take as foreground level and noiseSubtract foreground from source pixels, calculate bias-corrected polarized intensity and signal-to-noiseAny source pixels with S:N > 5? NoYes: valid candidate
Using foreground subtracted Q and U , calculate RMs per pixel (unwrapping polarization angles as necessary)Calculate weighted mean RM over valid pixels (inside fitted FWHM, S:N>5) Yes YesYesYes NoNoNoNoYesYesTests: NoNoNo
Reject sourceAccept sourceSource within >20% sensitivity region?Manual inspection: everything looks good?Mean probability of fractional polarization variation<5%?Mean probability-of-fit (RM linearity test)>10%?RM across pixels consistent with >95% confidence?Fractional polarization: < 100%?> 2%? ≥ Figure 1.
A flowchart showing the key steps in identifying polarized source candidates and evaluating the reliability of theircalculated rotation measure. Details of each step are given in the text.
GPS Rotation measure Catalogue
RM determination
Candidate polarized sources were processed very simi-larly to the method of Brown et al. (2003). To accuratelyconstrain the on-source pixels, a 2D Gaussian was fit tothe location of the source in the foreground-subtractedand debiased polarized intensity map calculated in theprevious step. For a small number of candidates ( ∼ Q and U parameters were subtracted and the po-larization angles of the remaining on-source polarizationwere calculated. These angles were subject to the ‘ nπ ’ambiguity that occurs for Faraday-rotated polarizationangles, where a change in the measured polarization an-gle by 180 ◦ would ‘wrap’ back into the [0 ◦ ,180 ◦ ) range.Given our channel spacing, a change in polarization an-gle of ± ◦ (half a wrap) would require an RM of ± − , which is much larger than we expect in thisregion of the Galactic disk. We assumed that the changein polarization angle between adjacent channels will al-ways be less than 90 ◦ , and applied corrections of ± ◦ as appropriate to the polarization angles to satisfy thiscondition.Using the unwrapped polarization angles, a linearfit to polarization angle as a function of λ was per-formed giving per pixel the fitted RM, error in RM, andprobability-of-fit. The pixel-wise RM was averaged overthe on-source pixels, weighted by the inverse square ofthe error in RM, to give the source RM. To assess theconsistency of the RM across all the source pixels, a re- duced χ value was calculated from the pixel-wise differ-ences from the mean. The maximum (foreground sub-tracted, debiased) polarized intensity and Stokes I fromthe on-source pixels were also determined, as well as thepixel-averaged fractional polarization.After all of these values were computed, a series ofquality control tests was applied to check for problemswith the source. First, sources with fewer than 5 on-source pixels were rejected, where on-source was definedas being within the fitted FWHM and having polarizedintensity signal-to-noise ratio greater than 5. Sourcesfor which the fractional polarization was greater than100%, which could happen when the Gaussian fit lockedonto a diffuse polarization feature near an unpolarizedStokes I source were rejected. Sources with fractionalpolarization below 2% were likewise rejected, to avoidany residual instrumental polarization leakage from be-ing identified as a polarized source.We observed that there was a population of partiallyresolved sources with strong RM gradients, for which asingle value could not be assigned; the χ value calcu-lated from the pixel RM averaging procedure was usedto confirm that the source had a single well-defined RMand did not have any significant RM gradients acrossthe on-source pixels. We defined a threshold for eachsource (dependent on the number of pixels, and thusthe number of free parameters) corresponding to a 95%confidence level in the χ distribution, meaning that asource with a single RM would fall below that thresh-old in 95% of cases. Sources with χ values above thisthreshold were considered to have too much RM varia-tion across the pixels and were rejected.To test for Faraday complexity, we evaluated the lin-earity of the relationship between polarization angle and λ using the probability-of-fit metric returned by the lin-ear fitting routine. The probability-of-fit was averagedover the on-source pixels, and if the average probabilitywas below 10% the source was rejected for not havingclear one-component Faraday-thin behavior.An additional test for Faraday complexity was per-formed using the fractional polarization, which is ex-pected to not vary with frequency for a Faraday-simplesource (Le Roux 1961). We determined the channel-averaged fractional polarization, then performed a χ test on the residuals (sum of the squares of the channel-wise differences from the mean). We rejected sourceswith χ values above the 95% confidence level as beingpossibly Faraday-complex.Sources that passed all of these criteria were then sub-jected to a manual inspection. This inspection verified afew conditions that were difficult to test in an automatedway. First, it was confirmed that the fitted FWHM re- Van Eck et al. gion of the Gaussian fit was inside the boundaries of theStokes I source; sources for which a significant portionof the polarization was outside the Stokes I counterpartwere rejected. Second, it was manually verified thatthe source’s polarized intensity was statistically signif-icant, by confirming that the polarized intensity of thesource stood out from the surrounding off-source pix-els; sources where the on-source polarized intensity wasindistinguishable from the surroundings were rejected.The initial processing concluded with this step, butinspection of the resulting catalog showed that sourcesat the edges of the mosaics were less reliable, i.e. consid-erably more likely to have an RM significantly differentfrom neighbouring sources. To maximize the reliabilityof our catalog, we removed the sources at the edges ofthe mosaics where the primary beam sensitivity was low,even if they passed all other tests. We used a thresholdof 20% of peak sensitivity, as given by the mosaic weightmaps; sources below this threshold were discarded fromthe catalog.At this stage all sources that were not rejected forone or more reason were considered as valid polarizedsources with well-defined RMs. These sources wentthrough a second manual inspection step, in which theirpolarized and Stokes I morphology were inspected andrecorded for source classification purposes. We identifiedeach source as either resolved or unresolved in polarizedintensity and Stokes I , as well as whether the source hadany neighbouring sources within approximately 1 (cid:48) , andwhether those neighbouring sources were also polarized.For resolved sources, we evaluated if the polarized inten-sity morphology matched the Stokes I morphology. Forunresolved sources, we also noted if there was an offsetof at least 1 pixel (0.3 (cid:48) ) between the polarized intensitypeak and the Stokes I peak, as this was seen in severalsources. The source classifications were included in thefinal catalog. FINAL CATALOGBefore assembling the final catalog, it was necessaryto remove duplicate detections. Many sources were pro-cessed multiple times, due to the overlap between ad-jacent CGPS mosaics. Since the overlap regions wereproduced from the same observations, they were not in-dependent measurements and it was not appropriate tocombine the multiple detections together. In each casewe chose to keep the detection that was farthest fromthe edge of the mosaic.Finally, we checked our list of RMs against the knownpulsars in this region of the sky, using the ATNF pul-sar catalog (Manchester et al. 2005). Using a cross-matching radius of 0.9 (cid:48) we found 4 pulsars in our sam- ple: B0355+54, J2007+2722, B2111+46, J2229+6114.For the remainder of this paper we assume that all thesources remaining in the main catalog are extragalactic.The final catalog contains 2234 RMs (including the4 pulsars), distributed over approximately 1300 squaredegrees. Selected columns of the catalog are shown inTable 2. The structure of this table is based on a in-development version of a new standardized format forreporting RMs (Van Eck et al, in prep). This tablereports the following quantities which vary by source:• A source ID number, running from 1 to 2230 forthe non-pulsar sources and giving the pulsar namefor each of the 4 pulsars.• Positions in Galactic and equatorial coordinates.The positions are defined as the coordinates of thepixel closest to the Stokes I source location, as re-ported by Taylor et al. (2017) or AEGEAN. Sincethese positions are quantized in units of the pixelsize (0.005 ◦ ), the uncertainty in position is half thepixel size.• Rotation measure and associated error, calculatedas described above.• Polarized intensity, determined as the highest(foreground subtracted and debiased) polarized in-tensity of the on-source pixels• Stokes I intensity, determined as the highestStokes I value of the on-source pixels.• The fractional polarization, determined as the av-erage over the on-source pixels.• The beam major axis, which depends on declina-tion, determined as 58 (cid:48)(cid:48) cosec δ (the beam mi-nor axis and position angle were constant for allsources).• Source classification: the 4 pulsars are labelled assuch, all other sources remain unclassified• Morphology flags from the manual inspection, de-fined as follows. These are stored as ‘Flag A value’in the table. – C = Compact – R = Resolved (in Stokes I and polarization,unless otherwise flagged) Details of this format can be found athttps://github.com/Cameron-Van-Eck/RMTable
GPS Rotation measure Catalogue – S = Subset (polarized intensity morphologyis smaller than the Stokes I extent) – N = Neighbouring source within approxi-mately 1 (cid:48) – P = Additional polarized component(s) seenin source or in neighbour – O = Offset (Polarized peak location does notmatch Stokes I peak)• The CGPS mosaic in which the source was found,stored in the ‘Flag B value’ column• The observation date, defined as the center of theapproximately two-month period over which eachmosaic was observed. This should be treated asapproximate, as some mosaics had a few fields re-observed at a later time, and also the observationdates were supplied to us quantized to the nearestmonth.Table 3 reports values that apply to all rows in thecatalog, which are included in the catalog for conformitywith the standard format. Columns that are part of thestandard that are not included in either of Tables 2 or3 are not supplied and have their default (blank) valuesas defined in the standard.Figure 2 shows the positions and RMs of all the non-pulsar sources. The density of sources is relatively uni-form with two exceptions: a region around l =80 ◦ wherethe bright Cygnus X region dominates Stokes I andimpacted source finding, and a smaller region around l =112 ◦ , b =-2 ◦ where the bright supernova remnant CasA significantly increases the noise levels.3.1. Comparison with previous catalogs
To assess the quality of our catalog, we compared theRMs against a new master catalog of previously pub-lished RMs. The previously published CGPS RMs(Brown et al. 2003) were deliberately excluded and willbe considered separately. Using a cross-matching radiusof 30 (cid:48)(cid:48) , 601 matches were found: 564 from the catalog ofTaylor et al. (2009, TSS09 in Fig 3), 16 from Van Ecket al. (2011, VE11 in Fig 3), 14 from Mao et al. (2012,Mao12 in Fig 3), 4 from Costa & Spangler (2018, CS18in Fig 3), 2 from Tabara & Inoue (1980, TI80 in Fig3), and 1 from Law et al. (2011, Law11 in Fig 3). Thecomparison of their RM values to ours is shown in the This master catalog is being assembled as part of an effort to stan-dardize the reporting of RMs between different projects (Van Ecket al, in prep). We used version 0.1.8 of this catalog, which canbe found at https://github.com/Cameron-Van-Eck/RMTable. top panel of Figure 3. No sources were found that hadcounterparts within two or more previous catalogs.In general there is qualitative agreement between thenew RMs and previous measurements. There is a smallpopulation of sources from Taylor et al. (2009) thatare offset by approximately ± rad m − because ofa known problem with the angle unwrapping ambiguityin their algorithm. These sources appear near the twodotted lines in the top panel of Fig. 3.A few sources show conspicuously large deviations be-tween new and old RM measurements. Both of thesources that were also present in the Tabara & Inoue(1980) catalog have very different RMs (>100 rad m − difference), as do two of the sources from Costa & Span-gler (2018). However, all 4 of these sources show signs ofFaraday complexity in the older measurements: signif-icant changes in polarized fraction at different frequen-cies for the Tabara & Inoue (1980) RMs, and large lin-ear fit residuals for the Costa & Spangler (2018) RMs.This can explain how different RMs can be observedover different frequency ranges. We interpret this asfurther evidence that individual source RMs should beinterpreted carefully whenever the presence of Faradaycomplexity cannot be reliably ruled out. In addition tothese sources, one of the Mao et al. (2012) RMs is alsosignificantly different from ours, but no reason for thisdifference can be easily identified.We further investigated the differences between ourcatalog and the matching sources in the Taylor et al.(2009) catalog to search for possible systematic effects.When testing for correlations between the absolute valueof the RM difference (between the two catalogs) andsource properties, we found a significant anti-correlationwith the signal-to-noise ratio (Spearman ρ = -0.36,p ≈ − ), as well as with related quantities such aspolarized intensity. However, this correlation was nolonger present ( ρ = -0.027, p=0.51) after the differenceswere normalized by the uncertainty in the difference (thequadrature sum of the uncertainties in the individualRMs). We interpret this to indicate that the RM uncer-tainties in both catalogs have the correct dependence onsignal-to-noise ratio. We find no significant correlationswith the uncertainty-normalized RM differences.3.2. Comparison with the 2003 CGPS RM catalog
In addition to comparing against other observations,we also compared this new CGPS catalog against theinitial CGPS catalog (Brown et al. 2003) to determinehow significantly the change in processing affected theresulting RMs. Of the 380 sources in the 2003 cata-log, all were present in the Taylor et al. (2017) catalog,and all but four were identified as polarized source can-
Van Eck et al. T a b l e . S o u r c e - D e p e nd e n t C o l u m n s f r o m t h e C a t a l og T a b l e C a t a l og I D R A D ec l b R M R M E rr o r P I [ ◦ ][ ◦ ][ ◦ ][ ◦ ][ r a d m − ][ r a d m − ][ m J y / b e a m ] . . . . - . . .
87 227 . . . - . - . . .
27 324 . . . - . - . . .
33 428 . . . . - . . .
50 521 . . . - . - . . .
04 621 . . . - . - . . . ........................ B + . . . . . . . J + . . . - . - . . . B + . . . - . - . . . J + . . . . - . . . C a t a l og I D S t o k e s I F r a c t i o n a l B e a mm a j o r S o u r ce M o r ph o l og y M e d i a n o b s e r v i n g [ m J y / b e a m ] p o l a r i z a t i o n a x i s [ ◦ ] t y p ee p o c h [ M J D ] . . . C . . . C . . . C . . . C . . . C . . . C ..................... B + . . . P u l s a r N J + . . . P u l s a r C B + . . . P u l s a r N J + . . . P u l s a r R S P N o t e — T h e f u ll c a t a l og i s a v a il a b l e i n F I T S f o r m a tt h r o u g h t h e a r X i v a s a n c ill a r y d a t a , a nd w ill b e a l s oo n li n e a tt h e A s t r oph y s i c a l J o u r n a l S u pp l e m e n t s a nd V i z i e r a f t e r pub li c a t i o n . T h e s t r u c t u r e o f t h i s t a b l e i s b a s e d o n t h e p r o p o s e d s t a nd a r d i ze d f o r m a t f o rr e p o r t i n g R M s . T h e t a b l e s h o w nh e r e i s j u s t a p o r t i o n f o r g u i d a n ce r e ga r d i n g i t s f o r m a nd c o n t e n t . I n a dd i t i o n t o t h e s o u r ce - d e p e nd e n t v a l u e ss h o w nh e r e , t h e . fi t s v e r s i o nh a s t h e s o u r ce - i nd e p e nd e n t v a l u e s g i v e n i n T a b l e . GPS Rotation measure Catalogue Column Value
RM determination method ‘EVPA-linear fit’Ionospheric correction method ‘None’Stokes I reference frequency 1.4207809 GHzPolarization reference frequency 1.4207809 GHzBeam minor axis 0.013611 ◦ (58 (cid:48)(cid:48) )Beam position angle 0 ◦ Polarization bias correction method ‘1974ApJ...194..249W’(Wardle & Kronberg 1974)Peak or integrated flux? ‘Peak’Minimum frequency 1.407194 GHzMaximum frequency 1.43463 GHzChannel width 7.5 MHzNumber of channels 4Telescope ‘DRAO-ST’Interval of observation a
60 daysCatalog ‘New CGPS (Van Eck et al 2020 in prep)’ b Flag A name ‘Morphology’Flag B name ‘Mosaic name’ a Defined as the interval between the first and last observations used. This value is approximate; most fields were observed repeatedlyover a period of two months with different baseline configurations to produce the final images. b This will be changed to the bibcode of this paper after publication.
Table 3.
Source-independent Columns in the Catalog Table Van Eck et al.
Figure 2.
The locations and RMs of polarized sources found in the CGPS region. Circle diameter is proportional to themagnitude of the RM (capped at 500 rad m − ), with positive and negative RMs colored in red and blue respectively. RMs withmagnitude smaller than 25 rad m − are dispayed as a filled black circle. GPS Rotation measure Catalogue Figure 3.
Top:
Comparison of reported RMs for sourcespresent in our catalog and previously published catalogs.The solid line marks 1:1 agreement between the catalogs,while dashed lines show a ±
650 rad m − offset correspond-ing to the nπ ambiguity in the Taylor et al. (2009) catalog.The catalog abbreviations are given in the text. Bottom:
Asthe top plot, but only comparing against the Brown et al.(2003) RM catalog. didates. Those 4 were found to be located in regionsof strong, position-dependent diffuse polarized emission.This caused our foreground subtraction algorithm to as-sign them high uncertainties. Thus, even though clearon-source polarization could be seen by a visual inspec-tion, our algorithm classified them as below the signalto noise threshold and they were discarded. Of the remaining 376 sources, 95 did not pass the qual-ity control tests of the new pipeline. Four were foundto fail Gaussian fitting in polarized intensity (probablyas a result of the improved foreground subtraction), 15were found to not have enough pixels above the signal-to-noise threshold (as a result of the new noise calcu-lations), 11 failed the pixel-averaging χ test and 15failed the linearity probability-of-fit test (for both tests,as a result of the new noise and error estimates). Thesesources, and their reported results from both the newand original pipelines, were inspected, and we found thatnearly all of these were close to the pass/fail thresholdsin the original pipeline and were pushed across one ormore of these thresholds by the changes to the noiseand foreground calculations. The χ test for constantfractional polarization was not used in the previous cat-alog; this test caused 43 previous sources to fail. Sevensources were found to be outside of the 20% sensitiv-ity threshold we used to remove edge sources with lowerreliability.The remaining 281 passed all tests in the improvedpipeline and are included in the new catalog with up-dated RMs; a comparison of the updated RMs to thoseof the original appears in the lower panel of Fig. 3. Westrongly caution users of this catalog that these RMswere derived from the same observations as the 2003catalog; they are not independent measurements andshould not be combined with the previous catalog forstatistical analyses. We recommend that our catalogcompletely replace the original CGPS catalog for all fu-ture analyses. ANALYSISIn this section we look at a few examples of analy-ses that can be done with the new RMs, including ashort description of some work that has already beenpublished using preliminary versions of these data.4.1.
Large-scale trends
In Figure 4 we show the statistical properties of therotation measures as a function of longitude. The meanRM (RM, top panel) generally shows smooth trendswith longitude: the outer Galaxy region 130 ◦ < (cid:96) <180 ◦ has a smooth and steady trend in RM from nega-tive towards zero with increasing longitude. This indi-cates that the large-scale magnetic field is embedded ina ubiquitous phase of the ISM. This was confirmed byFoster et al. (2013), who used a preliminary version ofthis catalog and found a strong relation between RMsof extragalactic sources and the optically thin hydro-gen column density in the same direction, indicatingthat the warm neutral medium is the main carrier of2 Van Eck et al. the large-scale magnetic field in the Galaxy. RMs crosszero near the anti-center, indicating that the magneticfield is nearly azimuthal in the outer Galaxy (Van Ecket al. 2011). This is supported by the study of the su-pernova remnant G182.4+4.3, close to the anti-center,in which the ambient field is almost perpendicular tothe line of sight (Kothes & Brown 2009). In the in-ner Galaxy the magnetic field closely follows the spiralarms, and, starting near (cid:96) < 70 ◦ the RMs show a strongswing from negative to positive with decreasing longi-tude. This is the reversal of the field between the localand Sagittarius arms (Van Eck et al. 2011).The third panel of Figure 4 shows the longitude-dependence of the ratio between the standard devia-tion ( σ RM ) and the absolute value of the mean of theRMs within the 2 degree bins we used to compute thestatistics. We note that over a substantial extent ofthe longitude range, from l ≈ ◦ to l ≈ ◦ , theratio σ RM / | RM | is well constrained, with a correlationcoefficient of 0.57 (p=0.003) between σ RM and | RM | .This is consistent with the similar result presented inBrown & Taylor (2001) using RM data from the firstphase of the CGPS (Taylor et al. 2003). This correlationwas interpreted by Brown & Taylor (2001) to indicatea preferential alignment between the small- and large-scale magnetic field components, a precursor to the con-cept of an ordered-random GMF component (e.g., Jaffeet al. 2010). At longitudes where the RMs approachzero (near the anti-center and the large-scale reversalregion), the ratios σ RM / | RM | increase significantly as aconsequence of dividing by very small values of | RM | .Enhanced variability in the Cygnus X region (see §4.2)also contributes deviations from the near-constant ra-tio seen in the mid-longitude range. The mid-longituderange of the CGPS, where the magnetic field has a suf-ficient line-of-sight component to yield substantial RMvalues, is an ideal testbed for statistical studies of theconnection between small- and large-scale GMF compo-nents.The variation of RM with Galactic latitude is shownin Figure 5 for the two latitude extensions. The abso-lute value of RM falls significantly away from the mid-plane, and drops almost to zero at high latitudes, as ex-pected. However, RM variation is not symmetric aroundthe mid-plane. This asymmetry appears to be caused bythe anomaly known as Region A (Simard-Normandin &Kronberg 1980), which is primarily affecting RMs southof the Galactic disk. A more thorough analysis of thelatitude extensions, using an earlier version of the cata-log, was done by Cooper (2014).4.2. Localized anomalies
Several regions or individual points that deviate fromthe large-scale trends are also visible, and are usuallyreflected by larger variations in the RM within each bin(as shown by the second panel of Fig. 4). Most of thesedeviations are caused by smaller scale structures in theISM, such as H II regions. Several examples can beeasily identified: the enhanced RM in the highest lon-gitude bin (191.5–193.5 ◦ ) is due to a cluster of sourcesbehind an H II enhancement at l ≈ ◦ , b ≈ +2.8 ◦ with much larger positive RMs than the surroundingsources; the anomalously low mean RM, large scatter,and lower source density around l = 172 ◦ is associatedwith a large H II region complex centred on Sh2-230; thesimilar anomaly around l = 135 ◦ is due to the W3/4/5complex. Figure 6 shows the smoothed distribution ofRMs with these regions highlighted.Figure 4 reveals a large scatter in RMs towards anarea 15 ◦ wide centered on (cid:96) = 80 ◦ . This is the CygnusX region where we are looking along our own spiral arm,the Orion spur. In Cygnus X we are looking throughseveral layers of star forming regions and H II regions(Gottschalk et al. 2012), and the heavy concentrationsof ionized material explain the scatter in RM. RM sourcedensity is low in this area (see Figure 4) because of thevery strong extended emission, and the area is left blankin Figure 6. Pulsar RMs towards Cygnus X are almostall positive above latitude − ◦ < (cid:96) <82 ◦ (Figure 9 of Kothes et al. 2020). Theregion of positive RMs around (cid:96) = 75 ◦ , b = +4 ◦ maybe associated with the low-longitude side of Cygnus Xor may be associated with the positive RMs found atlower longitudes. A study of compact source RMs anda comparison with the RM of extended emission (as inOrdog et al. 2019) would contribute to understandingthis region.Other departures from the overall trend can be seenin Figures 2 and 6, but are less obvious in Figure 4. Wecan identify these features with H II regions or supernovaremnants through comparison with CGPS images of thetotal-intensity along the Galactic plane; we use the 408-MHz maps of Tung et al. (2017), especially their Figures6 to 12:• A region of strong negative RM near (cid:96) = 143 ◦ ,0 ◦ < b < 2 ◦ is associated with the W3/4/5 H II region complex.• A region of strong negative RM near (cid:96) = 172 ◦ , b = − ◦ is associated with the H II complex Sh 2-230. This complex spans almost the whole latituderange of the survey, accounting for a low densityof RMs in this area. GPS Rotation measure Catalogue
13• At (cid:96) = 93 ◦ , b = 0 ◦ we see an area of positive RMwhere the surrounding RMs are negative. Thisfeature was reported by Clegg et al. (1992) andwas noted in the 2003 CGPS RM catalog (Brownet al. 2003). This anomaly is related to the super-nova remnant CTB104A and/or its environment(Uyaniker et al. 2002).There is also an area with very low RM magnitudesembedded within large RMs slightly below the mid-plane at approximately ◦ < (cid:96) < ◦ (see Figure 2).The reason for this is unclear. This region is largelyempty in Stokes I, from the edge of Cygnus X to approx-imately where Sagittarius arm begins, with no obviousStokes I counterpart to these low RMs.4.3. Comparison with diffuse emission
One weakness in a latitude-averaged analysis likeFig. 4 is that transitions that are not parallel to theGalactic plane can be lost or mis-represented. Ordoget al. (2017) used an earlier version of this catalog toidentify that the transition in the sign of RM around l = 60 ◦ is in fact not aligned with the Galactic plane butis clearly along a diagonal line (their Figure 2, repro-duced in Figure 6). They also found that this transitionwas reflected in the diffuse polarized emission present inthe CGPS data.Ordog et al. (2019) used a pre-final version of this RMcatalog (which was made without the 20% sensitivitythreshold and fractional polarization variation tests) tocompare with rotation measures derived from the diffusepolarized emission that is also present in the CGPS data.They found a strong correspondence between the diffuseemission RMs and the extragalactic source RMs, exceptin a few regions (most of the regions that are discussed inSects. 4.1 and 4.2) where smaller local features stronglyinfluence the RMs. They found that through most of thelines of sight the extragalactic RMs (which probe the fullline of sight through the Milky Way) were approximatelytwice as large as the RMs of the diffuse emission (whichis distributed through the Milky Way in a complex way).The details of this result and their interpretation are notrepeated here and can be found in their paper.Stutz et al. (2014) performed a power spectrum anal-ysis of the CGPS diffuse polarized emission at a resolu-tion of 2.67 ◦ . In the high-latitude extension, they founda sharp transition to a steeper power law index at a lat-itude of about +9 ◦ , which indicates a transition fromsmall-scale to large-scale structures. This is interpretedas the disk-halo interface and agrees with the locationwhere the RMs reduce to approximately zero (Figure 5). SUMMARY AND CONCLUSIONS R M [ r a d m ] R M [ r a d m ] R M /| R M | Galactic Longitude [ ] D e n s i t y o f R M s [ p e r s q . d e g . ] Figure 4.
Trends in the RM statistics as a function of longi-tude (computed over 2 degree bins), in the disk ( | b | < . ◦ ). Top: mean of rotation measure; second: scatter (standarddeviation) of rotation measure; third: ratio of the top twopanels; bottom: source density of RMs. The error bars forthe mean RM and the standard deviation were determinedas the standard errors of the mean and standard deviationrespectively. The error bars on the ratios take both of theseinto account using error propagation. The error bars on thesource density were calculated assuming a Poisson distribu-tion in the source counts.
The Canadian Galactic Plane Survey (CGPS) covers alarge area of the Galactic plane, from l =52 ◦ to l =192 ◦ ,and provides full polarization data in 4 closely spacedfrequency channels, enabling measurements of Faradayrotation of background sources. We have expanded onthe work by Brown et al. (2003) determining rotationmeasures from the CGPS data, with an improved pro-cessing pipeline that identifies more polarized sourcesand improves the foreground subtraction and error anal-ysis. Applying this pipeline to the full CGPS region,including north and south latitude extensions, we haveproduced a new catalog of 2234 RMs covering approxi-mately 1300 square degrees. Of these, 4 were identifiedas known pulsars. We have compiled this catalog follow-4 Van Eck et al.
10 5 0 5 10 15200150100500 R M [ r a d m ]
10 5 0 5 10 15
Galactic Latitude [ ] R M [ r a d m ] Figure 5.
Trends in the RM statistics as a function of lat-itude (computed over 2 degree bins), in latitude extensionregions ( ◦ < l < ◦ ). Top: mean of rotation measure; bottom: scatter (standard deviation) of rotation measure.The Galactic mid-plane has been marked by vertical dashedlines. ing a forthcoming standard format to try to maximizethe future value of the RMs.As a verification of the RM values, we identified 564sources in our catalog with previously observed (inde-pendent) RMs, and found good agreement between ourvalues and the previous measurements, with exceptionsmostly identified as problems in the previous observa-tions or as sources that showed signs of Faraday com-plex behaviour. We also compared the RMs we ob-tained against those from the original CGPS RM catalog(Brown et al. 2003), and found generally good agreementthere as well. Of the original 380 sources, 95 were foundto no longer pass the quality control tests in the newpipeline; most of these were cases where a source transi-tioned from marginally passing a test to marginally fail-ing or as the result of failing the fractional polarizationvariation test that was implemented in the new pipeline.In addition, several sources present in both catalogs werefound to have different RMs of order several tens of radm − . Since the underlying data for both sets of RMs arethe same, we interpreted both the change in RM and thechange in the quality control test outcomes to be dueto differences in the foreground subtraction. While weare confident that our improved pipeline performs fore-ground subtraction as effectively as possible, we cautionusers that individual RMs may be subject to unquanti-fied systematic errors of the order of a few tens of rad m − . We also warn users that these RMs are not in-dependent measurements from the Brown et al. (2003)catalog. We recommend that this new catalog replacethe old one in future analysis.With a typical source density of about 2 RMs persquare degree, our catalog significantly improves on thetwo previous large surveys of this area, Brown et al.(2003) and Taylor et al. (2009), which have source den-sities of approximately 1 RM per square degree. Ourcatalog is now the highest density large RM survey of theGalactic plane, and will be very useful for future studiesof magnetic fields inside the Galactic disk. This includesboth studies of large scale structure in the Galactic mag-netic field (Jaffe 2019) as well as studies of magnetismin smaller scale objects (e.g., Tahani et al. 2018). Ourcatalog has already contributed significantly to severalsuch studies (Ordog et al. 2019; Ma et al. 2020).The next generation of rotation measure surveys is al-ready underway. While POSSUM (Gaensler et al. 2010)will not have the declination coverage to overlap with theCGPS region, VLASS (Mao et al. 2014) will cover theCGPS region at higher frequencies. While the VLASSRM catalog is expected to have a higher source densitythan the CGPS, our catalog will be complementary withits coverage of lower frequencies. This will be useful as aprobe of Faraday complexity, for example by searchingfor frequency dependence in the RM. We expect thatthis catalog will be a unique and useful resource formany years. ACKNOWLEDGMENTSThe authors would like to thank Diane Parchomchuk,Jack Dawson, and Ev Sheehan, who operated the tele-scope and prepared the data for analysis through thelong period of the CGPS observations. We are in-debted to them for their skill, and for their thoroughand painstaking work.This work has made use of the following software pack-ages: NumPy (Harris et al. 2020), Matplotlib (Hunter2007), Astropy (Astropy Collaboration et al. 2013;Price-Whelan et al. 2018), and the Karma visualizationtools (Gooch 1996).The Canadian Galactic Plane Survey is a Canadianproject with international partners. The Dominion Ra-dio Astrophysical Observatory is operated as a nationalfacility by the National Research Council Canada. Thisresearch has been supported by grants from the NaturalSciences and Engineering Research Council.The Dunlap Institute is funded through an endow-ment established by the David Dunlap family and theUniversity of Toronto. GPS Rotation measure Catalogue Cygnus X W3/4/5Sh2-230 CTB104A Ordog et al. (2017)H II enhancementCas A Figure 6.
Smoothed distribution of RM, computed per-position as the weighted mean of RMs within 2 degrees. Positions withless than 5 sources within 2 degrees are not shown. Individual regions described in the text are marked.
REFERENCES