Mon.Not.R.Astron.Soc.000,1–16(2014) Printed27January2015 (MNLATEXstylefilev2.2) C-Band All-Sky Survey: A First Look at the Galaxy M.O.Irfan,1∗ C.Dickinson,1† R.D.Davies,1 C.Copley,2,3,4 R.J.Davis,1 P.G.Ferreira,4 C.M.Holler,4,5 J.L.Jonas,2,3 MichaelE.Jones,4 O.G.King,4,6 J.P.Leahy,1 J.Leech,4 E.M.Leitch,7 S.J.C.Muchovej,6 T.J.Pearson,6 M.W.Peel,1 A.C.S.Readhead,6 M.A.Stevenson,6 D.Sutton,4,8,9 AngelaC.Taylor,4 J.Zuntz1,4 5 1 1Jodrell Bank Centre for Astrophysics, School of Physics and Astronomy, The University of Manchester, Oxford Road 0 Manchester, M13 9PL, Manchester, U.K. 2 2Department of Physics and Electronics, Rhodes University,Drostdy Road, Grahamstown, 6139, South Africa 3SKA SA, 3rd Floor, The Park, Park Road, Pinelands, 7405, South Africa n 4Sub-department of Astrophysics, University of Oxford, DenysWilkinson Building, Keble Road, Oxford OX1 3RH, U.K. a 5Munich Universityof Applied Sciences, Lothstr. 34, 80335 Munich, Germany J 6California Institute of Technology, Pasadena, CA 91125, USA 4 7Universityof Chicago, Chicago, Illinois 60637, USA 2 8Institute of Astronomy, University of Cambridge, Madingley Road, Cambridge CB3 0HA, U.K. 9Kavli Institute for Cosmology, Cambridge, Madingley Road, Cambridge, CB3 0HA, U.K. ] A G 27January2015 . h p - ABSTRACT o We present an analysis of the diffuse emission at 5GHz in the first quadrant of the r Galacticplaneusing twomonthsofpreliminaryintensity datatakenwiththe C-Band t s All Sky Survey (C-BASS) northern instrument at the Owens Valley Radio Observa- a tory, California. [ Combining C-BASS maps with ancillary data to make temperature-temperature 1 plotswefindsynchrotronspectralindicesofβ =−2.65±0.05between0.408GHzand v 5GHz and β = −2.72±0.09 between 1.420GHz and 5GHz for −10◦ < |b| < −4◦, 9 20◦ <l<40◦. Through the subtraction of a radio recombination line (RRL) free-free 6 template we determine the synchrotronspectralindex inthe Galactic plane (|b|<4◦) 0 to be β =−2.56±0.07 between 0.408GHz and 5GHz, with a contribution of 53±8 6 per cent from free-free emission at 5GHz. These results are consistent with previous 0 low frequency measurements in the Galactic plane. . 1 ByincludingC-BASSdatainspectralfitswedemonstratethepresenceofanoma- 0 lous microwave emission (AME) associated with the Hii complexes W43, W44 and 5 W47 near 30GHz, at 4.4σ, 3.1σ and 2.5σ respectively. The CORNISH VLA 5GHz 1 source catalogue rules out the possibility that the excess emission detected around : v 30GHz may be due to ultra-compact Hii regions. Diffuse AME was also identified at i a 4σ level within 30◦ <l<40◦, −2◦ <b<2◦ between 5GHz and 22.8GHz. X r Key words: radiation mechanisms: non-thermal – radiation mechanism: thermal – a diffuse radiation – radio continuum: ISM. 1 INTRODUCTION At frequencies > 100GHz, the total Galactic emission is dominated by thermal emission from dust (Finkbeiner & DiffuseGalacticradioemissionisacombinationoffree-free, Schlegel 1999). Measurements of the diffuse Galactic emis- synchrotron, anomalous microwave and thermal dust emis- sionareusedforboththeinterpretationofcosmologicaldata sion.Below10GHz,free-freeandsynchrotroncontributions andtheunderstandingofourGalaxy.Synchrotronemission account for the majority of the emission, while anomalous can reveal information on the Galactic magnetic field and microwaveemission(AME)reachesitspeakbetween10and cosmic ray physics (Jaffe et al. 2011) while free-free emis- 30GHz(Goldetal.2009;PlanckCollaborationetal.2014e). sion measurements can be used to explore Hii regions and the warm ionised medium (Davies et al. 2006). The cosmo- logical demand for increasingly sensitive, large-scale maps E-mail:[email protected] of Galactic emission is driven by the need to remove these ∗ E-mail:[email protected] † c 2014RAS (cid:13) 2 M.O.Irfan et al. ‘foreground’ emissions to obtain an accurate measurement Table1.C-BASSNorthobservationalparameters.Thefull-beam of the cosmic microwave background (CMB) signal. This area is defined as being within 3◦.5 of the main beam peak and is particularly the case for imaging the CMB polarization thefirstmain-beamnullisat1◦ fromthemain-beampeak. signal and identifying B-modes, where the intrinsic cosmo- logicalsignalisasmallfractionoftheforegrounds(Dunkley Parameter C-BASSNorth et al. 2009; Betoule et al. 2009; Bennett et al. 2013; Er- rard & Stompor 2012). This issue has been exemplified by Latitude 37◦.2N Antennaoptics symmetricGregorian the recent claim of a detection of intrinsic CMB B-modes Antennageometry Az/El (Ade et al. 2014), which now appears to be contaminated Diameter 6.1m by foregrounds (Planck Collaboration et al. 2014a). Primarybeamwidth(FWHM) 44arcmin As different foregrounds dominate over different spec- FirstNull 1◦.5 tral and spatial ranges, it is important to have all-sky data Main-beamefficiency(<1◦.0) 72.8% available covering arange of frequencies and with 1◦ res- Full-beamefficiency(<3◦.5) 89.0% ∼ olution if CMB polarization experiments are to live up to Intensitycentrefrequency 4.76GHz their full potential in determining cosmological parameters Intensitynoise-equivalentbandwidth 0.489GHz (Planck Collaboration et al. 2014f). The low frequency all- Noiseequivalenttemperature 2mK√s sky intensity surveys currently most used for this purpose are at 0.408GHz (Haslam et al. 1982), 1.420GHz (Reich & Reich1986;Testorietal.2001;Reich,Testori&Reich2001), of preliminary C-BASS North intensity data (January and 22.8GHz(Bennettetal.2013)and28.4GHz(PlanckCollab- February 2012), which has a sufficient signal-to-noise ratio oration et al. 2014b). Jonas, Baart & Nicolson (1998) and to constrain the properties of the bright emission near the Carretti et al. (2013) provide southern Hemisphere maps Galactic plane. The main observational parameters for C- at 2.3GHz in intensity and polarization, respectively. Sun BASS North are shown in Table 1. et al. (2007) present a polarization survey of the Galactic This paper is organized as follows: Section 2 describes plane at 5GHz with a FWHM of 9.5 arcmin so as to cap- the C-BASS and ancillary data, and presents a derivation ture the small-scale structures. However the lack of all-sky, of the free-free template used in subsequent analysis. The degreescalesurveysbetween1.420GHzand22.8GHzintro- quality of the C-BASS data calibration is verified in Sec- duces great uncertainties in the spectral behaviour of free- tion 3. In Section 4 we investigate the synchrotron spectral free,synchrotronandanomalousmicrowaveemission across indexoftheGalacticplane(b <4deg),determiningthein- thefull sky (Peel et al. 2012). termediatelatitude(4◦ < b|<| 10◦)synchrotron-dominated | | C-BASS, the C-Band All Sky Survey, is a project cur- spectralindexbetween0.408/1.420GHzand5GHz,aswell ◦ ◦ rently mapping Galactic intensity and linear polarization as the in-plane (0 < b < 4 ) synchrotron spectral index | | at a frequency of 5GHz (Jones et al. in prep, King et al. onceweaccountforfree-freeemission inthisregion.InSec- 2010). At this frequency the polarized emission is negligi- tion 5 we use the C-BASS data alongside higher-frequency blyaffectedbyFaradayrotationathighlatitudes(∆ψ.1◦ data from the WMAP and Planck satellites to constrain across most of the b > 10deg latitude sky). C-BASS con- thespectral behaviour of anomalous microwave emission in | | ◦ ◦ ◦ ◦ sistsoftwoindependenttelescopes:C-BASSNorthobserves 2 < b < 2, 30 < l < 40 . Conclusions are presented in − from theOwensValley Radio Observatory(OVRO)in Cal- Section 6. ◦ ifornia, USA (latitude 37.2 N) whilst C-BASS South is lo- cated at theKlerefontein support basefortheKaroo Radio ◦ Astronomy Observatory (latitude 30.7 S). C-BASS North achievedfirstlightin2010andentereditsfinalsurveymode 2 DATA in late 2012 after a period of commissioning and upgrades 2.1 C-BASS (Muchovejet al. in prep).C-BASSSouthiscurrentlybeing commissioned. The final all-sky maps will have a FWHM C-BASS North observes by repeatedly scanning over 360◦ ′ resolution of 43.8 and a nominal sensitivity in polariza- in azimuth at a constant elevation. Several different scan tion of 0.1mK per beam. On completion the C-BASS speeds close to 4◦ per second are used, to ensure that a ≈ data will be used to probe the Galactic magnetic field us- given feature in the time-ordered data does not always cor- ing synchrotron radiation, constrain the spectral behaviour respond to the same angular frequency on the sky, in case of AME, and provide a polarized foreground template for of data corruption by effects such as microphonics. The CMBpolarization experiments.Thelargeangularscaleand C-BASS North data are subject to 1.2Hz (and harmon- high sensitivity of the final survey will be of particular use ics) microphonic oscillations but these effects are reduced for confirming B-mode detections at low multipoles. to a negligible level within the intensity data, especially In this paper we present some of the Galactic science withinthehigh signal-to-noise regions investigatedhere,by that can be achieved using preliminary C-BASS intensity our data reduction pipeline (Muchovej et al., in prep.). We ◦ results. We focus on the composition of the Galactic plane scan at an elevation of 37.2 (through the North Celestial at 5GHz, determining the spectral index of synchrotron Pole) for about two-thirds of the time, and at higher ele- ◦ emission between 0.408 and 5GHz and the fractional com- vations (mostly 47.2) for the remainder of the time, in or- position of free-free and synchrotron emission present. We der to even out the sky coverage. The northern receiver is extend this analysis to include higher frequency data to a continuous-comparison receiver (King et al. 2014), which constrain the AME amplitude and spectrum towards a few measuresthedifferencebetweenskybrightnesstemperature compact regions. For this analysis we use only two months and a temperature-stabilized resistive load. This architec- c 2014RAS,MNRAS000,1–16 (cid:13) C-Band All-Sky Survey: A First Look at the Galaxy 3 turereducesreceiver1/f noise,whichwouldotherwise con- dependingon whatisbeingobserved.Itisthegenerally ac- taminate thesky signal. ceptedpracticetoconvertfromantennatemperaturetothe The northern receiver has a nominal bandpass of full-beam scale, as described above, in order to fully repre- 4.5 – 5.5GHz,butin-bandfilterstoremoveterrestrialradio sentthelarge-scale structuresofthemap.WhileusingTFB frequencyinterference(RFI)reducetheeffectivebandwidth is correct for diffuse emission, as demonstrated in Section to 0.489GHz, with a central frequency of 4.76GHz (King 3, it underestimates point source emission by 5 per cent ≈ et al. 2014). The finite and relatively large (20 per cent) andsoTA isusedfortheanalysisconductedinSection 5.2. bandwidth, means that adopting a single central frequency The conversion between flux density and full-beam is a simplification. This is because the effective frequency brightnesstemperaturerequiredforthiscalibrationscheme, is a convolution of the instrument bandpass with the sky alongside the opacity (calculated from skydips) and colour signal across this bandpass: corrections,carrywiththemafewpercentuncertainty.The νβ = f(ν)Gm(ν)νβdν, (1) mainsourceofuncertaintyinthecalibrationcomesfromthe eff R f(ν)Gm(ν)dν telescopeapertureefficiency,whichisrequiredtoconvertbe- R tween flux density and antenna temperature. The aperture where f(ν) is the bandpass response and Gm(ν) is the for- efficiency of the northerntelescope has been determined by ward gain (e.g., Jarosik et al. 2011). Any source that has a simulations, to an accuracy that we estimate to be 5 per spectralshapedifferenttothatofthecalibratorsource,will cent, dominated by uncertainties in the modeling of stand- resultinaslightlydifferenteffectivefrequency.Onecancor- ing waves between the feed and the subreflector. Further rectforthisbymaking‘colourcorrections’(e.g.,Leahyetal. work is being doneon both simulation and measurement of 2010) assuming the bandpass shape and the spectral shape theaperture efficiency for the calibration of the full survey. ofthesourcebeingobserved.ForC-BASS,weestimatethat In the meantime a conservative 5 per cent calibration error thesecolourcorrections are 1percentfortypicalspectra has been assigned to thesepreliminary data. ≈ (Irfan 2014); only spectral shapes significantly different to thatofourcalibrators(primarilyTauAandCasA)require The C-BASS maps are constructed from the pipeline- corrections. For this analysis, we do not make any colour processedtime-ordereddatausingtheDEStripingCARTog- corrections and assume an effective frequency of 4.76GHz. rapher,Descart(Suttonetal.2010).Descartperformsa We include an additional 1 per cent uncertainty, which is maximumlikelihoodfittothedata,bymodellingthecontri- added in quadraturewith theother uncertainties. bution of 1/f noise in the timestreams as a series of offsets Therawtelescopetime-ordereddataareprocessedbya of a given length. The multiple crossings of each pixel over datareductionpipelinethatidentifiesRFIevents,performs a range of parallactic angles in long sets of observations al- amplitude and polarization calibration and applies atmo- low these offsets to be well determined. The characteristic spheric opacity corrections (Muchovej et al. in prep). The timescale (‘knee frequency’) of typical 1/f fluctuations in northern receiver includes a noise diode which, when acti- the C-BASS intensity data is tens of mHz (corresponding vated, injects a signal of constant noise temperature. The to tens of seconds), depending on the atmospheric condi- noise diode excess noise ratio (ENR) is stable to within tions. We adopt an offset length of 5seconds, although we 1 per cent over time periods of several months. The data achieve similar results with 10seconds. We use power spec- reduction pipeline calibrates the intensity signal onto the traofthetime-ordereddatatoestimatethewhite-noisevari- noise diode scale, and the noise diode is subsequently cali- anceandkneefrequency,providingamodelofinstrumental brated against the astronomical sources Cas A, Tau A and noise which is used in the mapping process. The data are CygA.The calibrator point-source fluxdensities arecalcu- inverse-variance weighted in each pixel to achieve the op- latedfromthespectralformsgiveninWeilandetal.(2011), timal weighting for signal-to-noise ratio. No striations are Baars et al. (1977) and Vinyaikin (2007) and converted to visible in our finalmap. Thedata are gridded into HEALPix anequivalentantennatemperature.Next,themodelsofthe (G´orski et al. 2005) maps with Nside = 256, corresponding beam are used to convert the antenna temperature scale to 13.7 arcmin pixels. (TA) to a ‘full beam’ temperature scale (TFB): The completed intensity survey made using all avail- ΩA able data will be confusion limited, but the maps used in TFB =TOFF + ΩFBTA, (2) this analysis were made with just two monthsdata and are noise limited. The C-BASS confusion noise limit in inten- whereΩFB isthefullbeamsolidangle,ΩAisthetotalbeam sityisestimatedtobe 130mJybeam−1 (Condon2002)or solid angle and TOFF is the survey zero level (e.g., Jonas, 0.8mK,whilethetherm≈alnoiselimitforthetwomonthpre- Baart & Nicolson 1998). Following Reich (1982) and Jonas, liminary dataused in thisanalysis is 6mK.TheGalactic Baart & Nicolson (1998), we define the extent of the full- ≈ ◦ signals analysed in thispaper are > 500mK. beam to be 3.5 from boresight. Using GRASPphysical op- ticssimulations(Holleretal.2013)wecalculatetheC-BASS ThefourregionslistedinTable2andoutlinedinFig.1. Northantennatemperaturetofull-beambrightnesstemper- The four regions can be considered as off-plane (Area 1, atureconversionfactortobe1.124. Theinitialtemperature which is split into two) and in-plane (Areas 2, 3 and 3). scaleoftheC-BASSmapsisreferencedtotheantennatem- Fig.1showstheGalacticplaneandhigherlatitudesasseen perature of the noise diode, and this is subsequently con- at 4.76GHzbyC-BASS Northas well as thenorthernpart ◦ verted antenna temperature and then the more useful full- of the Gould Belt, a star forming disc tilted at 18 to the beam brightness temperature to account for power lost in Galacticplane.TheGouldBeltisknownforitsincreasedsu- the far sidelobes. The level of power lost is dependent on pernovaactivity(Grenier2000)andincludestheOphiuchus the observed source morphology and so this factor varies cloud complex. c 2014RAS,MNRAS000,1–16 (cid:13) 4 M.O.Irfan et al. 2.326GHz radio maps to provide additional low frequency information on synchrotron/free-free emission. The Haslam data are used in their unfiltered form so point sources and striations arestill present.Thesedata, alongside theJonas, Baart & Nicolson (1998) data are also used to determine the spectral index between 0.408 and 4.76GHz outside of the Galactic plane. The Jonas, Baart & Nicolson (1998) ◦ ◦ data only cover the southern sky ( 83 < δ < 13 ) but − fortunately the survey extendsto theGalactic plane region underinvestigation in thiswork. Haslam et al. (1982) estimate an uncertainty of 3K in their global zero level. Although the true random noise is farlessthanthis,striationspresentinthedataaretypically 0.5K and worsen to3Kin some regions, so wetake3K as a conservativelocal uncertainty in the pixelvalues. Thetemperaturescaleofthe1.420GHzmapisproblem- atic,as extensivelydiscussed byReich&Reich (1988). The mapisnominallyonthefull-beamscale.Becauseofthenear- sideloberesponsewithin3◦.5,fluxesofpointsourcesobtained byfittingaGaussiantothecentrallobeofthebeamareex- pected to be underestimates, and this may be corrected by convertingtoa‘main-beam’ brightnesstemperature.Based on the observed beam profile, this correction was expected tobeabout1.25 0.15; however,Reich&Reich(1988)find Figure 1. C-BASS4.76GHz intensity map using data taken in ± 1.55 0.08 from direct observationsofcalibrators, soeither JanuaryandFebruary2012showingaregioninthefirstGalactic ± themeasuredbeamareaortheoriginaltemperaturescaleis quadrant.ThedataareatNside=256andFWHMof44arcmin. inerrorbyaround24percent.Inourapplication wemulti- The white dashed rectangles highlight the four regions, used in thisanalysisandthewhitesolidcurvedlinesdelineatethenorth- plied the data by the empirically deduced correction factor ernpartoftheGouldBelt.Theemptymapregion,bottomright, of 1.55. We too find this factor to be consistent with our isduetoanabsenceofC-BASSNorthdatabelow declination= calibration observations of Barnard’s Loop (Section 3). We 30◦;thisareawillbemappedbyC-BASSSouthobservationsin assigna10percentuncertainty,mainlyduetotheunquanti- − thefuture.Thecolourbarscaleisinkelvinandtheblackcontours fiedvariationofthetemperaturescalewiththeangularsize arespacedat0.2Kintervals. of the source. We note that Reich & Reich (1988) estimate a full-to-main beam factor of 1.04 0.05 for the 0.408GHz ± survey,so it suffers much less from this ‘pedestal’ effect. Table 2. The four Galactic plane regions investigated in this To allow for scanning artefacts in the 1.42GHz maps work. weuseaconservativelocaluncertaintyof0.5Kequaltothe original estimate of the zero-level uncertainty. Area Long.(◦) Lat.(◦) Motivation 1 20–40 10 4&4 10 Synchrotrondominated − →− → 2.2.2 High frequency data 2 20–40 4 4 TheRRLdataregion − → 3 21–26 4 4 Typicaldiffuseregion We use the Planck and WMAP 9-year data to help con- − → 4 30–39 2 2 AMEregion strainthespectralform ofAME.TheWMAP K-banddata − → are specifically used to determine the ratio of AME to to- tal emission between 5 and 23GHz. The WMAP thermal 2.2 Ancillary Data noiseineach HEALPixpixel(σ)was calculated usingthere- WehavecomparedtheC-BASSmapswiththeall-skymaps lation (Bennett et al. 2013) σ = σ0/√Nobs, where Nobs is at other frequencies listed in Table 3. We will now briefly the map hit count and σ0 is the thermal noise per sample. discusseachdatasetinturn.AllmapswereusedinHEALPix Therefore the r.m.s. thermal noise values listed in Table 3 for WMAP are average values across the sky. The Planck format.Thosemapsthatwereoriginallyinadifferentformat 100GHzand217GHzmapswerenotusedasthesedatain- were converted to HEALPix using a nearest-neighbour pixel clude unwanted and significant contributions from the CO interpolation. The maps were first smoothed to a common ◦ lines (Planck Collaboration et al. 2014c). 1 resolution,assumingthattheoriginalsurveybeamswere An additional 3 per cent uncertainty was assigned to Gaussian with the FWHM specified in Table 3. Depending both theWMAP and Planck data, as in Planck Collabora- on theapplication, themaps were resampled as required to HEALPixN =256(13.′7pixels)orN =64(55′ pixels). tionetal.(2011),toaccountforresidualbeamasymmetries side side after smoothing. Both WMAP and Planck have unblocked andseverelyunder-illuminatedaperturestoreducesidelobes to extremely low levels, and so the difference between the 2.2.1 Low frequency data antennatemperatureandfull-beam brightnesstemperature We use the Haslam et al. (1982) 408MHz, Reich & Re- scale is less than 0.5 per cent (in the case of WMAP, the ich (1986) 1.42GHz, and Jonas, Baart & Nicolson (1998) data are corrected for far sidelobe contamination and the c 2014RAS,MNRAS000,1–16 (cid:13) C-Band All-Sky Survey: A First Look at the Galaxy 5 Table 3.Thedata usedinthis paper alongsidetheir frequency, FWHM,HEALPix Nside,calibrationuncertainty, average thermal noise (inmKRayleigh-Jeans)unlessstatedotherwise)andreference.Thetwobottom rowsarederivedcomponent separationproducts. Survey ν(GHz) Res.(arcmin) Nside Calibration(%) σ Reference Haslam 0.408 51 512 10 700 Haslametal.(1982) RRLs 1.4 14.8 512 15 5 Alvesetal.(2012) Reich 1.420 35 512 10 140 Reich&Reich(1986) Jonas 2.3 20.0 256 5 30 Jonas,Baart&Nicolson(1998) C-BASS 4.76 43.8 256 5 6 Kingetal.(2010) WMAP 22.8 49 512 0.2 0.04 Bennett etal.(2013) WMAP 33.0 40 512 0.2 0.04 Bennett etal.(2013) WMAP 40.7 31 512 0.2 0.04 Bennett etal.(2013) WMAP 60.7 21 512 0.2 0.04 Bennett etal.(2013) WMAP 93.5 13 512 0.2 0.03 Bennett etal.(2013) Planck 28.4 32.65 1024 0.4 0.009 PlanckCollaborationetal.(2014b) Planck 44.1 27.92 1024 0.4 0.009 PlanckCollaborationetal.(2014b) Planck 70.4 13.01 1024 0.4 0.008 PlanckCollaborationetal.(2014b) Planck 143 7.04 2048 0.4 0.0006 PlanckCollaborationetal.(2014b) Planck 353 4.43 2048 0.4 0.0003 PlanckCollaborationetal.(2014b) Planck 545 3.80 2048 7 0.0001 PlanckCollaborationetal.(2014b) Planck 857 3.67 2048 7 0.00006 PlanckCollaborationetal.(2014b) COBE-DIRBE 1249 37.1 1024 13.5 0.5MJysr−1 Hauseretal.(1998) COBE-DIRBE 2141 38.0 1024 10.6 32.8MJysr−1 Hauseretal.(1998) COBE-DIRBE 2997 38.6 1024 11.6 10.7MJysr−1 Hauseretal.(1998) Derived MEM 22.8 60 128 - - Bennett etal.(2013) MCMC 22.8 60 64 - - Bennett etal.(2013) corrections from the WMAP ‘main-beam’ to the C-BASS the Galactic plane: dust absorption along the line-of-sight full-beam are less than 0.6 per cent (Jarosik et al. 2007), of the Hα lines (which needs to be corrected for through- and we havenot made any correction in this analysis. outtheGalaxy)reachesamaximumintheplaneduetothe WeusetheCOBE-DIRBEband8,9and10datatohelp increased density of dusty star forming regions. RRLsfrom constrain the spectral form of thermal dust emission. The ionised hydrogen provide an absorption-free alternative for COBE-DIRBE data used are the Zodi-Subtracted Mission tracingfree-free emission in theGalactic plane(Alvesetal. Average (ZSMA) data(Hauser et al. 1998). 2012). We derive the free-free template used in this analy- sis from RRL data constructed from the HI Parkes All-Sky Survey(HIPASS;Barnes et al. 2001). TheRRLdatahaveabeamFWHMof14.8arcmin,and 2.2.3 Free-Free templates have thermal noise and calibration uncertainties of 5mK At low Galactic latitudes (b < 4◦), the intensities of the and 15 per cent respectively. The RRL data (Alves et al. | | ◦ ◦ ◦ ◦ diffusefree-freeandsynchrotronemissionarecomparableat 2012) cover 20 < l < 40 , 4 < b < 4 and were used − gigahertzfrequencies.Weseparatethesynchrotronandfree- by Alves et al. to produced a free-free emission template free contributionstototal emission in theGalactic planeat from the data by using an average electron temperature of 5GHz;todothisafree-freeemissiontemplatewasrequired 6000Kacrossthewholeregion.Thisaveragevaluecarriesa aswellastheC-BASSdata.Free-freeemissionisunpolarized largeuncertainty( 1000K),asitdoesnotaccountforthe ± and characterised by a spectral index of β 2.1 (Gold presence of warmer and cooler regions. This uncertainty is ≈ − et al. 2009; Dickinson, Davies & Davis 2003), for optically thedominantfactorin the15percentuncertaintyassigned thin regions. At high frequencies ( 90GHz), the free-free to thefree-free template. ≈ spectral index will steepen (Draine 2011) to 2.14 due to For this analysis the RRL data provide our free-free − the Gaunt factor, which accounts for quantum mechanical template of choice. As this is a novel approach to deriving corrections, which become important at higher frequencies. free-free templates, we find it illuminating to compare the The source of the diffuse Galactic free-free emission is RRL-derived result with those from WMAP. All-sky free- the Warm Ionized Medium (WIM), with its intensity pro- freemodelsareavailablefromtheWMAP9-yearproductre- portional to the emission measure (EM), given by EM = leaseconstructedusingamaximumentropymethod(MEM) (ne)2 dl, where ne is the electron density. The WIM andMarkovChain MonteCarlo simulation (MCMC)1.The Ralso radiates Hα recombination lines, including the optical MCMC free-free model (Gold et al. 2009) at 22.8GHzused Balmer-alphaline(656.28nm)andhigh-nradiorecombina- inthiswork(‘modelf’)fitsthetotalemissionmodelforeach tion lines (RRLs) whose strength also depends on the EM. pixelasasum ofapower-law withafreespectralindexpa- Asaresult,Hαlinescanbeusedtocalculatefree-freeemis- sionintensitiesprovidedtheelectrontemperatureisknown. However, Hα lines are not satisfactory free-free tracers in 1 http://lambda.gsfc.nasa.gov/product/map/dr5 c 2014RAS,MNRAS000,1–16 (cid:13) 6 M.O.Irfan et al. rameter for synchrotron emission, a fixed-index power-law for the free-free emission, a CMB term, a power-law with freespectralindexforthethermaldustemission,andathe- oreticalspinningdustcurvewithamplitude.TheMEMfree- freemodelat22.8GHzisasimilarspectralpixel-by-pixelfit, which usesBayesian priorsfor pixelswherethedatadonot constrain theparameters verywell (see Gold et al. 2009 for details). 3 DATA VALIDATION We have considered the effect of pointing errors, absolute intensity calibration, colour (bandpass) corrections and at- mospheric opacity on the C-BASS temperature scale, but the dominant error is the uncertainty on the aperture effi- ciency, which we estimate at 5 per cent. To assess the con- sistency of the C-BASS temperature scale, we compare C- BASSdatawithotherradiosurveys.WefocusonBarnard’s Loop, an Hii shell within the Orion complex (Heiles et al. 2000). This region is known to be dominated by optically thin free-free emission at these frequencies, which allows us to use the well-defined spectral index to compare data at a range of frequencies. Fig. 2 shows Barnard’s Loop in Hα Balmer line emission and in 4.76GHz radio continuum as seenbyC-BASSNorth.Theboxedregion showsthespecific area selected for analysis in Fig. 3. The close morpholog- ical similarity indicates that free-free emission from warm ionized gas dominates. The temperature spectrum of a diffuse Galactic emis- sion is generally approximated in theform of a power-law: T(ν) νβ, (3) ∝ where T(ν) represents the brightness temperature at fre- quency ν and β is the spectral index of the temperature distribution. Given two intensity data sets taken at dif- ferent frequencies (ν1 and ν2) a temperature-temperature Figure 2.Top:Barnard’sLoopasseenat4.76GHzbyC-BASS (T-T)plotofthesetwosetswillrevealalinearrelationship, the gradient of which relates to the emission spectral index North. Nside = 256 and FWHM resolution of 0◦.73. Orion A, OrionB and NVSS J060746-062303 are masked out. The colour (Turtle et al. 1962) via: scaleininkelvinandthecontoursarespacedat0.01Kintervals. T(ν1)= ν1 βT(ν2) + baselineoffsets. (4) aBnodttormes:olBuatironnar6d.’1saLrocompinin. THhαeecmoilsosuiornscaatleHEiAsLiPnixraNylseidigeh=s a5n1d2 (cid:18)ν2(cid:19) the contours spaced at 100 rayleigh intervals between 100 and 500(Finkbeiner2003).Themorphologicalsimilaritybetweenthe An advantage of theT-T plot method is its insensitivity to maps suggests that free-free emission is dominant at 4.76GHz. zero-level uncertainties, though if several emission mecha- The white box encapsulates the region selected for temperature nismsarepresentseverallinearrelationshipswillbeseenso analysis. the region for analysis must be chosen with care. Optically thinfree-freeemissionhasawell-documentedspectralindex of 2.12 0.02 (Draine 2011), therefore asimple and effec- − ± tivevalidation of the C-BASS data quality can be achieved using T-T plots of thisarea. The compact sources Orion A (M42), Orion B and at Nside =64 toreducecorrelations between thepixels– at ◦ NVSSJ060746-062303havebeenmaskedoutoftheC-BASS a FWHMof 1 , this gives <2 pixels per beam. imageusingadiameter3.5 theC-BASSFWHMforOrion The error bars in Fig. 3 represent the pixel noise for × A and B and 2 the C-BASS FWHM for NVSS J060746- the respective radio maps calculated using 1000 iteration × 062303 (see Fig. 2). A smaller mask size could be used for Monte Carlo simulations. These include the effects of con- NVSSJ060746-062303 asthesourceisfainterthanOrionA fusion noise (0.8mK for the C-BASS data), zero-level un- or Orion B. certainties (3K for the Haslam et al. (1982) and 0.5K for Fig. 3 displays T-T plots using the C-BASS data for theReich & Reich (1986) data), CMB anisotropies (for the Barnard’sLoopagainstHaslametal.(1982),Reich&Reich WMAP data),HEALPixsmoothinganddegradingaswellas (1986) and WMAP K-band data. The T-T plots are made ther.m.s. thermal noise. The CMB anisotropies were simu- c 2014RAS,MNRAS000,1–16 (cid:13) C-Band All-Sky Survey: A First Look at the Galaxy 7 latedusingtheCAMBwebinterface2andthestandard,pre-set cosmologicalparameters.Itwasdeterminedthatcalibration uncertainty,andnotpixelnoise,wasthedominantsourceof uncertainty. The uncertainties quoted for the spectral in- dices are a quadrature combination of the pixel errors and calibration uncertainties associated with each dataset. The0.408–4.76GHz,1.42–4.76GHzand4.76–22.8GHz T-T plotsinFig.3showlinearfits;thereducedχ2valuesare 1.9, 0.5 and 1.3 for the0.408–4.76GHz, 1.42–4.76GHzand 4.76–22.8GHzdata,respectively.Theaveragespectralindex ofβ = 2.15 0.03confirmsafree-freedominatedspectrum − ± forBarnard’sLoop.Thisspectralindexisconsistentwithin 1σ of theexpected value (β 2.12). ≈− The multiplicative factor required to bring C-BASS datatobeinperfectagreementwithWMAP 22.8GHzdata (assuming all the emission is due to free-free emission) is 0.95 0.05. This validation gives us confidence that there ± are no significant systematic calibration errors in excess of the 5 per cent calibration error assigned to the preliminary C-BASS data. 4 DIFFUSE EMISSION BETWEEN 0.408 AND 4.76GHz The total diffuse Galactic plane radio continuum emission seen at 4.76GHz is primarily a mixture of free-free and synchrotron emission. Free-free emission is known to have a narrow latitude distribution while synchrotron emission is broader (Alves et al. 2012; Planck Collaboration et al. 2014d).Thenarrowwidthofthefree-freedistributionshown inSection 4.2suggeststhatthefree-freecontributionisneg- ligiblecomparedtothesynchrotroncontributionatlatitudes higher than four degrees. This suggestion is supported by theHαfree-freetemplate(Dickinson,Davies&Davis2003), whichcoversoff-planeregions.Itisthereforepossibletode- terminethespectral indexofsynchrotronemission between 0.408 and 4.76GHz at intermediate and high Galactic lati- tudes without having to perform any free-free/synchrotron emission separation. 4.1 Intermediate latitude total emission spectral indices The spectral index of the Galactic synchrotron emission varies significantly, both spatially and with frequency. Strong,Orlando&Jaffe(2011)reviewresultsfromskymaps between22MHzand23GHzatintermediatelatitudes,find- Figure 3. T-T plots of the Barnard’s Loop region between ingasteepeninginthesynchrotronspectrumfromβ 2.5 ≈− 4.76GHzand(a)0.408GHz,(b)1.42GHzand(c)22.8GHz.The at 0.1GHz to β 3 at 5GHz. Between 0.408GHz and ≈ − spectralindicesconfirmafree-freeemissiondominatedregion. 3.8GHzanindexof 2.7hasbeenestimatedintheGalac- ≈− tic plane whereas between 1.42 and 7.5 GHz the spectrum appears to steepen to 3.0 (Platania et al. 1998). Jaffe etal.(2011),Lawsonet≈al−.(1987)andBennettetal.(2003) index β = (δ +3)/2, this spectral index steepening re- − alsodiscussthespectralsteepeningofsynchrotronemission sultsfromasteepeningofthecosmicrayelectronspectrum. in the Galactic plane and it is clear that between 2.3 and It resembles that expected from synchrotron losses, i.e. a 22.8GHz the spectral index undergoes a steepening from ‘break’ (actually a rather smooth steepening) of 0.5–1 in β 2.7to 3.1.Duetotherelationshipbetweenelectron assumingasteadystatebetweeninjectionoffreshparticles, e≈ne−rgy dist≈rib−ution N(E) Eδ, and synchrotron spectral radiativeloss,andtransportoutoftheGalaxy(e.g.Bulanov ∝ &Dogel1974; Strong,Orlando&Jaffe2011). However,the timescale for electrons to leave the Galaxy implied by this 2 http://lambda.gsfc.nasa.gov/toolbox interpretation is far shorter than the cosmic ray residence c 2014RAS,MNRAS000,1–16 (cid:13) 8 M.O.Irfan et al. timescales inferred from the presence of spallation product Table4.Theoff-plane(4◦ < b <10◦)spectralindicesbetween nucleiinthecosmicrays,soincurrentmodels(e.g.Orlando 0.408/1.420GHzand4.76GHz|.| &Strong2013)thesteepeningisexplainedbyabreakinthe electron injection energy spectrum; the break due to radia- Lon.(◦) Lat.(◦) ν (GHz) β tive losses is predicted to occur at a much lower frequency than is actually observed. 20–40 10 4 0.408–4.76 2.65 0.05 − →− − ± The steepening of the synchrotron spectrum at a few 20–40 4 10 0.408–4.76 2.69 0.05 → − ± GHzisconsistentwiththemeasuredsteepeningofthelocal 20–40 −10→−4 1.420–4.76 −2.72±0.09 20–40 4 10 1.420–4.76 2.64 0.09 cosmic ray electron spectrum (Abdo et al. 2009), with δ → − ± varying from < 2.3 to 3 between 1 and 10 GeV (solar ≈ modulationpreventsaccuratemeasurementofδmuchbelow The larger uncertainties on the 1.420–4.76 spectral indices 1 GeV). are due to the larger uncertainties in the 1.42GHz survey. Reich & Reich (1988) observed a steepening along We expect the full C-BASS survey to have a significantly the Galactic plane of the synchrotron β (between 0.408 and 1.420GHz) with β = 2.86 at (l,b) = (20◦,0◦) and reduced uncertainty in its calibration, and in conjunction ◦ −◦ with other modern surveys such as S-PASS (Carretti et al. β= 2.48 at (l,b) = (85 ,0 ). Peel et al. (2012) comment − 2013) weshouldbeabletoconstrainthespectralindexand on thelack of flatter-spectrum (β 2.7) emission at high ≈− possible curvature (steepening/flattening) of the spectrum latitudes between 2.3GHz and 30GHz, in contrast to the even more accurately. Galactic plane. Conversely, the most sophisticated models of cosmic ray propagation in the Galaxy to date (Orlando & Strong 2013) predict variations in β of no more than 4.2 Free-free cleaned maps at 0.408/1.420/4.76 0.1 between different directions, e.g. between theplane and GHz higher latitudes. These models are explicitly of the large- scalediffuseemissiononly,andsowillomitanyspectralvari- At low Galactic latitudes (b < 4◦) the free-free emis- | | ations caused by features on sub-kpc scales, including the sion cannot be assumed to be negligible; the synchrotron radio “loops” that dominate the high-latitude synchrotron and free-free components need to be separated. Assuming emission. These are expected to cause variations in β be- that below 5GHz only synchrotron and free-free emission cause they may be sites of particle injection, hence with are present at a detectable level, synchrotron maps can be flatter spectra (as for most supernova remnants) and be- formedfromtheHaslametal.(1982),Reich&Reich(1986) cause they have stronger magnetic fields, so emission at a and C-BASS data using a pixel-by-pixel subtraction of a given frequency is from lower energy electrons (which will free-free template of choice. More sophisticated component also flatten the spectrum). Spectra steeper than the diffuse separation methods, such as parametric fitting, could be emission requirethattheycontainoldelectronsconfinedby used to acquire synchrotron maps but this would require a very low effective diffusion coefficient in magnetic fields an accurate estimate of the C-BASS zero-level. Therefore, muchstrongerthantypicalinterstellarfields,sothatsignifi- until the full survey is complete we use simpler methods of cantinsituradiativelossescanoccurwithoutcompensating component separation. particle acceleration. It is also worth noting that the low- In Fig. 5, we present two latitudeprofiles for theRRL, latituderegionsclaimed toshowflattersynchrotronspectra MEM, and MCMC free-free templates over-plotted on the are the ones most contaminated by free-free emission, e.g. C-BASS data for the same region. The RRL, MCMC and theCygnus X region at l=75◦–86◦. MEMtemplateshavebeenscaledto4.76GHzusingthefree- Fig.4showsT-T plotsofsemi-correlatedpixels(6pixels free spectral index of β = 2.1. The free-free profiles all − perbeam),which were madeintheoff-planelatituderange have their baselines subtracted so that the in-plane emis- ◦ ◦ of 4 < b < 10 with no masking of specific regions. We sionscanbecomparedwithoutbiasingfromdifferentsurvey | | expectthisregiontobedominatedbysynchrotronemission zero-levels. The baseline level was established using a cose- between 0.408/1.420GHz and 4.76GHz. The positive and cant(1/sin(b))plusaslopetothe b <10◦ pointsforeach | | | | negativelatituderangesareplottedseparately todifferenti- of thecurves. Wethen subtract theslope (offset and gradi- ate between Galactic plane emission and emission from the ent)toleavetheplaneemission. The1◦ FWHMmapswere NorthPolarSpurandtheGouldBelt.TheregioninFigs.4b resampledatHEALPixN =64tomatchtheMCMCtem- side and4dencompasspartoftheGouldBeltandtheNorthPo- plate.Thetwolatitudeprofilesaverageovera)21◦ <l<26◦ larSpur.Thevariation insynchrotronspectralindexacross andb)thefullRRLlongituderange20◦ <l<40◦.The21◦ the positive latitude range results in a bulged point distri- to 26◦ longitude range was selected to be a typical bright bution in Fig. 4b between 0.12 and 0.16K on thex-axis. diffuseregion asitcontained thefewest bright,compact ar- The mean spectral indices shown in Fig. 4 are sum- eas in the C-BASS map. Both profiles span the full RRL ◦ ◦ marised in Table 4 and range between β = 2.64 and latitude range of 4 <b<4 . − − β = 2.72. The weighted mean spectral indices between The C-BASS data were used as a ‘testbed’ for these − 0.408–4.76GHz and 1.420–4.76GHz are 2.67 0.04 and templatesandhelptoidentifyregionsoverwhichtheyfailto − ± 2.68 0.06,respectively.Theseresultsareconsistentwith, describethemeasuredemission.TheRRLfree-freetemplate − ± but more accurate than, those in the literature, which typ- is systematically lower in temperature than the MEM and ically give an average synchrotron spectral index of 2.7 MCMC templates because the electron temperature used ≈− both within the galactic plane and away from the radio is 1000K lower than the 7000K assumed by WMAP. The “loops”. The C-BASS data provide a longer lever-arm for RRLs are a more direct free-free measure as they are not constraining spectral indices with uncertainties ∆β < 0.1. subject to the same parameter degeneracies faced by the c 2014RAS,MNRAS000,1–16 (cid:13) C-Band All-Sky Survey: A First Look at the Galaxy 9 Figure 4. Top row: T-T plots between 0.408GHzand 4.76GHz for(a) 10◦ <b< 4◦ (Below Plane) and (b) 10◦ >b>4◦ (Above Plane).Bottom row:T-T plotfor1.420GHzand4.76GHzfor(c) 10◦<−b< 4◦ and−(d)10◦>b>4◦.Theoff-planespectralindices − − between0.408and4.76GHzcanbeseentorangebetween 2.67and 2.72. − − MCMC and MEM templates but of course are so far only full analysis of the intensity and polarization data sets will ◦ available for b <4 and a limited range of longitudes. reveal which of these differences are physically interesting. | | Three synchrotron maps at 0.408GHz, 1.420GHz and 4.76GHzweremade bysubtractingtheRRLfree-free tem- These in- and off-plane results confirm a synchrotron plate from the Haslam et al. (1982), Reich & Reich (1986) spectral index of 2.72 < β < 2.64 between 0.408 and ◦− ◦ − and C-BASS maps. The RRL free-free template was scaled 4.76GHz for 10 < b < 10 . Once again no significant − to the different frequencies using a single power-law with spectral steepening of the synchrotron spectral index is ob- β= 2.1.Thesynchrotronmaps,scaled to4.76GHz(using served and the values are in agreement with the expected β va−lues given in Table 5), are shown as latitude profiles index of 2.7. − in Fig. 6. The best-fit synchrotron spectral indices required to match the 0.408/1.420GHz latitude profiles with their corresponding 4.76GHz values are shown in Table 5. The At 1.5GHz, diffuse Galactic emission has been shown data are shown at HEALPix N = 256 so as to fully sam- to be roughly 30 per cent free-free and 70 per cent syn- side ple the latitude profile, resulting in correlated pixel errors. chrotron emission (Platania et al. 1998). The ratio between The thermal noise is shown as error bars while the average free-free and synchrotron emission at 4.76GHz can now be calibrationerrors(basedona0.4Ksignal)areshowninthe determined using the derived latitude distributions of the ◦ ◦ legend.Fig.6aaveragesoverthe21 to26 longituderange 4.76GHz pure synchrotron map. Fig. 7 shows latitude pro- ◦ ◦ ◦ ◦ while 6baveragesovertheentire20 to40 longituderange. files across 4 < b < 4 for the total 4.76GHz emission − AsinFig.5,thesynchrotroncurvesareshownafterthesub- as measured by C-BASS, the RRL free-free map scaled to traction of their zero-levels. Both Fig. 6a and Fig. 6b dis- 4.76GHzusingβ= 2.1andthe4.76GHzpuresynchrotron − play the broad peaks typically associated with synchrotron map. Extrapolating our results to 1.5GHz, the fraction of emission. Although one average synchrotron spectral index synchrotron emission present is 70 10 per cent while at ◦ ± ◦ ◦ is clearly a good approximation across the b =4 latitude 4.76GHzitwasfoundtobe53 8percentfor20 <l<40 , | | ◦ ◦ ± range,smallscaledifferencesareseenbetweenthecurves.A 4 <b<4 . − c 2014RAS,MNRAS000,1–16 (cid:13) 10 M.O.Irfan et al. Figure 5.Free-freelatitude distributionfor b <4◦ usingaver- Figure 6.LatitudeprofilesofGalacticsynchrotronemissionfor agedlongitudedatafromtheRRL,MCMC,M| |EMfree-freetem- b < 4◦ using averaged longitude data from the RRL free-free plates all scaled to 4.76 GHz for the longitude range of a) 21◦ s|u|btracted Haslam et al. (1982), Reich & Reich (1986) and C- – 26◦ and b) 20◦– 40◦. The maps have been smoothed to 1◦ BASS data, all scaled to 4.76 GHz for the longitude range of resolutionanddowngradedtoHEALPix Nside=64. a) 21◦ – 26◦ and b) 20◦ – 40◦ (see Table 5). The maps have been downgraded to HEALPix Nside = 256 and are smoothed to 1◦ resolution. The thermal noise is shown as error bars on the data points while the calibration uncertainty (based on a 0.4K 5 IDENTIFICATION OF ANOMALOUS signal)areshownatthebottom ofeachpanel.Thebroadprofile EMISSION AT 22.8GHz expected forsynchrotronemissionisclearlyvisible. AttheWMAP K-bandfrequencyof22.8GHz,diffuseemis- sion intheGalactic planeisacombinationoffree-free,syn- Table 5.TheGalactic planesynchrotron spectral indicesasde- chrotron and anomalous microwave emission (AME). The terminedusingsynchrotronmapsformedusingtheRRLfree-free favouredemission mechanism forAMEiselectricdipolera- templates. diation from spinning dust grains (e.g. Draine & Lazarian 1998; Planck Collaboration et al. 2011; Davies et al. 2006). Long.(◦) ν range β AME has been readily identified in dark clouds and molec- ularclouds (Vidalet al. 2011; Watson et al. 2005; Casassus 21–26 0.408–4.76 −2.59±0.08 21–26 1.420–4.76 2.69 0.17 etal.2008;AMIConsortiumetal.2011). Howeveritisalso − ± 20–40 0.408–4.76 2.56 0.07 presentthroughouttheGalacticplaneandaddstothecom- − ± 20–40 1.420–4.76 2.62 0.14 plexityof component separation of CMB data (Planck Col- − ± laboration et al. 2011, 2014e). For example, with only full- sky data at 1.4 and 22.8GHz, the presence of AME, which usedincombinationwith22.8GHzdatatoseparatethedif- peaks in flux density at 30GHz (Planck Collaboration ∼ ferent spectral components. et al. 2011), will appear only as a flattening of the total- emission power-law. Here we use the preliminary C-BASS data to help characterise the synchrotron signal and so de- 5.1 Diffuse AME tectthepresenceofAME.The4.76GHzC-BASSdatapro- vides a higher-frequency measure of the synchrotron emis- Forthisanalysis thefollowing threeregionswerechosenfor sion, which is uncontaminated by AME, and can thus be their combination of bright diffuse and compact regions: c 2014RAS,MNRAS000,1–16 (cid:13)