The Parkes Galactic Meridian Survey (PGMS): observations and CMB polarization foreground analysis
E. Carretti, M. Haverkorn, D. McConnell, G. Bernardi, N.M. McClure-Griffiths, S. Cortiglioni, S. Poppi
aa r X i v : . [ a s t r o - ph . C O ] J u l Mon. Not. R. Astron. Soc. , 1–23 (2010) Printed 8 June 2018 (MN L A TEX style file v2.2)
The Parkes Galactic Meridian Survey (PGMS):observations and CMB polarization foreground analysis
E. Carretti, , ⋆ M. Haverkorn, , , D. McConnell, G. Bernardi, N.M. McClure-Griffiths, S. Cortiglioni, and S. Poppi, ATNF, CSIRO Astronomy and Space Science, P.O. Box 276, Parkes, NSW 2870, Australia INAF, Istituto di Radioastronomia, Via Gobetti 101, I–40129 Bologna, Italy Jansky Fellow, National Radio Astronomy Observatory Astronomy Department, University of California, Berkeley, 601 Campbell Hall, Berkeley, CA 94720 ASTRON, Oude Hoogeveensedijk 4, 7991 PD Dwingeloo, The Netherlands ATNF, CSIRO Astronomy and Space Science, P.O. Box 76, Epping, NSW 1710, Australia Kapteyn Astronomical Institute, University of Groningen, P.O. Box 800, 9700 AV Groningen, the Netherlands INAF, Istituto di Astrofisica Spaziale e Fisica Cosmica Bologna, Via Gobetti 101, 40129 Bologna, Italy INAF, Osservatorio Astronomico di Cagliari, Loc. Poggio dei Pini, Strada 54, 09012 Capoterra, Italy
Accepted xxxx. Received yyyy; in original form zzzz
ABSTRACT
We present observations and CMB foreground analysis of the Parkes Galactic MeridianSurvey (PGMS), an investigation of the Galactic latitude behaviour of the polarizedsynchrotron emission at 2.3 GHz with the Parkes Radio Telescope. The survey consistsof a 5 ◦ wide strip along the Galactic meridian l = 254 ◦ extending from Galactic planeto South Galactic pole. We identify three zones distinguished by polarized emissionproperties: the disc, the halo, and a transition region connecting them. The halo sectionlies at latitudes | b | > ◦ and has weak and smooth polarized emission mostly at largescale with steep angular power spectra of median slope β med ∼ − .
6. The disc regioncovers the latitudes | b | < ◦ and has a brighter, more complex emission dominatedby the small scales with flatter spectra of median slope β med = − .
8. The transitionregion has steep spectra as in the halo, but the emission increases toward the Galacticplane from halo to disc levels. The change of slope and emission structure at b ∼ − ◦ is sudden, indicating a sharp disc-halo transition. The whole halo section is just oneenvironment extended over 50 ◦ with very low emission which, once scaled to 70 GHz,is equivalent to the CMB B –Mode emission for a tensor–to–scalar perturbation powerratio r halo = (3 . ± . × − . Applying a conservative cleaning procedure, we estimatean r detection limit of δr ∼ × − at 70 GHz (3-sigma C.L.) and, assuming a dustpolariztion fraction < δr ∼ × − at 150 GHz. The 150 GHz limit matchesthe goals of planned sub-orbital experiments, which can therefore be conducted atthis high frequency. The 70 GHz limit is close to the goal of proposed next generationspace missions, which thus might not strictly require space-based platforms. Key words:
Cosmology: CMB – Galaxy: disk – Galaxy: halo – polarization
The study of the Galactic polarized synchrotron emission isessential for two cutting-edge fields of current astrophysicsresearch: the detection of the B –Mode of the Cosmic Mi-crowave Background (CMB), for which the Galactic syn-chrotron is foreground emission, and the investigation of themagnetic field of the Galaxy.The CMB B –mode is a direct signature of the ⋆ E-mail: [email protected] (EC) primordial gravitational wave background (GWB) leftby inflation (e.g., Kamionkowski & Kosowsky 1998;Boyle, Steinhardt, & Turok 2006). The amplitude of itsangular power spectrum is proportional to the GWB power,which is conveniently expressed relative to the amplitude ofdensity fluctuations—the so-called “tensor-to-scalar pertur-bation power ratio” r . Still undetected, the current upperlimit is set to r < .
20 (95% C.L., Komatsu et al. 2009) by Refer to Peiris et al. (2003) Eq. 10 for a full definition of r .c (cid:13) E. Carretti et al. the results of the Wilkinson Microwave Anisotropy Probe(WMAP). A detection of the B –Mode would be evidencefor primordial gravitational waves, and a measurement of r would help distinguish among several inflation models andinvestigate the physics of the early stages of the Universe.Reaching this spectacular science goal will be difficultbecause of the tiny size of the CMB B -Mode signal, fainterthan the current upper limit of 0.1 µ K, and perhaps as faintas the 1 nK corresponding to the smallest r accessible byCMB ( r ∼ − , Amarie, Hirata & Seljak 2005). At suchlow levels the cosmic signal is easily obscured by the Galacticforeground of the synchrotron and dust emissions.Investigations of the synchrotron contribution havebeen conducted over recent years, but data are still insuf-ficient to give a comprehensive view (Fig. 1). Page et al.(2007) analysed the 23 GHz WMAP polarized maps and findthat the typical emission at high Galactic latitude is strong:at 70 GHz it is equivalent to r ∼ .
3, even higher than thecurrent upper limit. An analysis of the same WMAP databy Carretti et al. (2006b) identified regions covering about15 per cent of the sky with much lower emission levels, offer-ing a better chance for B –Mode detection. In these regionsthe polarized foreground is fainter, equivalent to a B –Modesignal corresponding to r in the range [1 × − , × − ]. Abetter characerization of the polarized foreground is crucialespecially for sub-orbital experiments (ground-based andballoon-borne), which will observe small sky areas.The design of future experiments is dependent on thefrequency of minimum foreground emission. WMAP findsthat this is in the range 60–70 GHz for high Galactic lat-itudes (Page et al. 2007), but it remains unknown for thelowest emission part of the sky. It has been suggested thatthe dust emission might have deeper minima than the syn-chrotron in the areas of lowest emission (e.g., Lange 2008),shifting the best window for B –Mode detection to higherfrequencies.Synchrotron emission from the Milky Way is not onlya foreground for CMB polarization measurements, but canalso be used to study the Galactic magnetic field. The totalintensity of synchrotron emission can be used to estimate thetotal magnetic field strength, while the polarized intensitygives the strength of the regular component. This analysis inexternal galaxies has shown that the spiral arms are usuallydominated by a small-scale, tangled, magnetic field with aweaker coherent large-scale field aligned with the arms. Inthe inter-arm regions the regular component dominates andin some spirals, magnetic arms with coherent scales up tothe size of the disc have been detected in between the gasarms (e.g. see Beck 2008 for a review).The synchrotron emissivity of our own Galaxy is harderto understand because of our location inside it, but has theadvantage that it can be studied in detail. Frequency de-pendent synchrotron depolarization can be used to deter-mine typical scale and strength of small-scale magnetic fields(e.g.Gaensler et al. 2001), and all-sky synchrotron emis-sivity maps can characterise the synchrotron scale height By equivalent to r we signify the strength of a foreground whosespectrum would match the spectrum of the CMB B -mode emis-sion at the ℓ = ∼
90 peak arising from conditions characterised bythe given value of r . Figure 1.
Summary of the current knowledge of the B –Modepower spectra of the Galactic synchrotron emission. Spectra areestimated at 70 GHz. A brightness temperature frequency spec-tral index of α = − . C B WMAP is the general contamination at high Galactic latitude, asestimated by the WMAP team using a ∼
75% sky fraction at 22.8-GHz (Page et al. 2007). Spectra measured in small areas selectedfor their low emission are also shown: the target fields of the exper-iments BOOMERanG ( C B BOOM , Carretti et al. 2005b) and BaR-SPOrt ( C B BaR , Carretti et al. 2006a), and the upper limit foundin the two fields of the experiment DASI ( C B DASI , Bernardi et al.2006). Finally, C S/N< shows the estimate of the emission in thebest 15% of the sky (Carretti et al. 2006b, the shaded area indi-cates the uncertainties), while C DRAO the values found in someintermediate regions using 1.4 GHz data (La Porta et al. 2006).For comparison, CMB spectra for three values of r are also shown. (Beuermann et al. 1985), or can be used for large-scale mod-eling of the Galactic magnetic field, especially in the halo .The relative parity of the toroidal magnetic field compo-nent is still under discussion (see e.g., Han 1997; Frick et al.2001; Sun et al. 2008). Jansson et al. (2009) use WMAPsynchrotron maps at 23 GHz to show that the magneticfield behaviour in the Galactic disc and halo may differ con-siderably.Data from external galaxies does not help in constrain-ing the Milky Way magnetic halo, as there is a wide varietyof magnetic field configurations: from galaxies without ev-ident halo field, to X-shaped fields centred at the galaxycentre, to large almost spherical magnetic halos (see Beck2008 for a review).Recent maps of polarized Galactic synchrotron radia-tion at 1.4 and 22.8 GHz (Wolleben et al. 2006; Testori et al.2008; Page et al. 2007) show polarized emission across theentire sky, and can be used to study the Galactic mag-netic field. However, the 1.4 GHz maps show that the discemission is strongly depolarized up to latitudes | b | ≈ ◦ (Wolleben et al. 2006; Testori et al. 2008), while Faradaydepolarization effects are still present up to | b | ≈ − ◦ (Carretti et al. 2005a). Furthermore, those data consist of asingle frequency band and do not enable rotation measure Galactic halo is used in this context as the gaseous and mag-netic field distributions out of the Galactic disc, and is not nec-essarily connected to the stellar halo.c (cid:13) , 1–23
GMS: observations and CMB foreground analysis computations. The WMAP data at 22.8 GHz are virtuallyunaffected by Faraday rotation (FR) effects, but the sensi-tivity is not sufficient since, once binned in 2 ◦ pixels, about55% of the sky has a signal to noise ratio S/N <
3. Thisarea corresponds to all the high Galactic latitudes with theexception of large local structures, which is most of the skyuseful both for CMB studies and to investigate the Galacticmagnetic field.Therefore, synchrotron maps at intermediate frequen-cies over all Galactic latitudes are needed to explore thebehaviour of the contamination of the CMB with latitudeas well as to study the Galactic magnetic field in the disc,the halo, and the disc-halo transition.In this work we present the Parkes Galactic Merid-ian Survey (PGMS), a survey conducted with the ParkesRadio Telescope to cover a strip along an entire south-ern Galactic meridian at 2.3 GHz. The area is freefrom large local structures, making it ideal for investi-gating both the CMB foregrounds and the Galactic mag-netic field. The PGMS overlaps the target area of severalCMB experiments like BOOMERanG (Masi et al. 2006),QUaD (Brown et al. 2009), BICEP (Chiang et al. 2009),and EBEX (Grainger et al. 2008). Our results may have di-rect implications for all these experiments.In this paper we present the survey, observations, and acharacterisation of the polarized emission. We also presentan analysis of the measured emission as a contaminatingforeground to CMB B –mode studies. Analysis and impli-cations for the Galactic magnetic field will be subject of aforthcoming paper (Paper II, Haverkorn et al. 2010 in prepa-ration). A third paper will deal with the polarized extra-galactic sources (paper III, Bernardi et al. 2010 in prepara-tion).Survey and observations are presented in Section 2, theground emission analysis in Section 3, and the maps in Sec-tion 4. The analysis of both the angular power spectrum andemission behaviour is presented in Section 5, while the dustcontribution is investigated in Section 6. The detection lim-its of r are discussed in Section 7 and, finally, our summaryand conclusions are reported in Section 8. The available data and the properties of the synchrotronemission discussed in Section 1 lead to the following mainrequirements for a survey. Observations must(i) be conducted at a low enough radio frequency for thesynchrotron emission to dominate the other diffuse emissioncomponents, but at a frequency higher than 1.4 GHz to avoidsignificant Faraday Rotation effects;(ii) cover all latitudes from the Galactic plane to the pole,to explore the behaviour with the Galactic latitude b ;(iii) cover regions free from large local structures, suchas the big radio loops, that would distort the estimates oftypical conditions at high latitudes.The Parkes Galactic Meridian Survey (PGMS) is aproject to survey the diffuse polarized emission along aGalactic meridian designed to satisfy these requirements. It surveys a 5 ◦ × ◦ strip along the entire southern merid-ian l = 254 ◦ from the Galactic plane to the south Galacticpole (Fig. 2). The observations have been made at 2.3 GHzwith the Parkes Radio Telescope (NSW, Australia), a facil-ity operated by the ATNF - CSIRO Astronomy and SpaceScience a division of CSIRO . It also includes an 10 ◦ × ◦ extension centred at l = 251 ◦ and b = − ◦ .The selected meridian goes through one of the low emis-sion regions of the sky identified using the WMAP data(Fig. 2, see also Carretti et al. 2006b) and is free of large lo-cal emission structures. The meridian also goes through thearea of deep polarization observations of the BOOMERanGexperiment (Masi et al. 2006); the 10 ◦ extension near b = − ◦ is positioned to best cover that field.At long wavelengths, measurements of Galactic polar-ized emission in regions of high rotation measure are cor-rupted by Faraday depolarization. At 1.4 GHz Faraday de-polarization is significant up to Galactic latitudes | b | < ◦ (Carretti et al. 2005a) where RM >
20 rad/m . At2.3 GHz this RM limit increases to 60 rad/m , allowinga clear view of polarized emission over all high Galactic lat-itudes and well into the upper part of the disc.The observations were made in four sessions from Jan-uary 2006 to September 2007 with the Parkes S-band Galileoreceiver, named after NASA’s Jupiter exploration probe forwhich the Parkes telescope and this receiver were used fordown-link support (Thomas et al. 1997). The receiver re-sponds to left- and right-handed circular polarization, whosecross-correlation gives Stokes parameters Q and U (e.g.Kraus 1986). This scheme provides more protection againsttotal-power (gain) fluctuations than the alternative: a re-ceiver responding directly to the linearly polarized signals.The original feed used for the Galileo mission has beenreplaced by a wide-band corrugated horn, highly tapered toreduce sidelobes and the response to ground emission. Thefeed illuminates the dish with a 20 dB edge taper and thefirst side-lobe is 30 dB below the main beam.The ATNF’s Digital Filter Bank 1 (DFB1) was used toproduce all four Stokes parameters, I , Q , U , and V . DFB1was equipped with an 8-bit ADC and configured to give256 MHz spectrum with 128 2-MHz channels. Spectra aregenerated using polyphase filters that provided high spectralchannel isolation. The isolation between adjacent channelsis 72 dB, an enormous improvement over the 13 dB isola-tion of Fourier-based correlators. This, in combination withthe high sample precision, gives excellent protection againstRFI leaking from its intrinsic frequency to other parts of themeasured spectrum. This is valuable in the 13-cm band asstrong RFI can be present. Recording spectra with spectralresolution greater than required for the polarimetry analysishas allowed efficient removal of RFI-effected channels, max-imising the effective useful bandwidth. Data were reducedto 30 8-MHz channels. The RFI removal typically yieldedan effective total bandwidth of 160 MHz.The source B1934-638 was used for flux calibration as-suming the polynomial model by Reynolds (1994) for an The first observing session, in September 2005, used a Fourier-based correlator, and spectra were strongly contaminated by RFI.Subsequent use of DFB1 greatly improved the measurements.c (cid:13) , 1–23
E. Carretti et al.
Figure 2.
The PGMS strip (dark gray) plotted on the WMAP polarized intensity map ( L = p Q + U ) at 22.8 GHz (Page et al. 2007)binned in 2 ◦ pixels (HEALPIX pixelation with Nside=32). Pixels with S/N < l ∼ ◦ , b ∼ ◦ ). The big radio loops are structures large several tens of degrees, like loop-I which extends fromthe north Galactic pole down to the Galactic plane, with a possible continuation into the southern hemisphere. The map is in Galacticcoordinates with longitude l = 0 ◦ at centre and increasing leftward. accuracy of 5%. The polarization response was calibratedusing the sources 3C 138 and PKS 0637-752, whose polar-ization states were determined using the Australia TelescopeCompact Array (ATCA) with an absolute error of 1 ◦ . Thestatistical error of our polarization angle calibration is 0 . ◦ .The astronomical IAU convention for polarization angles isused: angles are measured from the local northern meridian,increasing towards the east. It is worth noting that this dif-fers from the convention used in the WMAP data, for whichthe polarization angle increases westwards. The unpolarizedsource B1934-638 was also used to measure the polarizationleakage, for which we measure a value of 0.4%. The off-axisinstrumental polarization due to the optics response is about1%. The use of a system with both Q and U as corre-lated outputs mitigates gain fluctuation effects. To check thelevel of a 1/f noise component in the data we observed theSouth Celestial Pole (SCP), thereby avoiding azimuth (AZ)and elevation (EL) dependent variations of atmospheric andground emissions. Remaining variations in the signal arisefrom intrinsic atmospheric changes and receiver fluctuations.Power spectra of the Q and U time-series are almost flat withno evidence of a 1/f component down to 3 mHz (see Fig. 3).This confirms that the system is stable and characterised by white noise up to 7-min time scales, sufficient for theduration of our scans.The Galactic meridian was observed in 16 5 ◦ × ◦ fieldsand one 10 ◦ × ◦ field. The fields are named PGMS-XX,where XX is the Galactic latitude of the field centre. Eachfield except PGMS-02 includes a 1 ◦ extension along b at thenorth edge for an actual size of 5 ◦ × ◦ and an overlap of5 ◦ × ◦ with the next northern field.The fields were observed with sets of orthogonal scansto give l – and b –maps (scans along Galactic longitude andlatitude, respectively). Each field was observed with 101latitude scans ( b –maps) and 121 longitude scans ( l –maps)spaced by 3 arcmin to ensure full Nyquist sampling of thebeam (FWHM = 8.9 arcmin). The same sample spacing wasused along each scan by scanning the telescope at 3 ◦ /minwith a 1-second integration time.One full set of l – and b –maps were observed for the6 disc fields at latitude | b | < ◦ and the 10 ◦ × ◦ field,giving final a sensitivity of ∼ . | b | > ◦ ), where a weakersignal was expected, two full passes were made to give asensitivity of ∼ . c (cid:13) , 1–23 GMS: observations and CMB foreground analysis Figure 3.
Power spectra of Q (top) and U (bottom) time seriesfor a south Celestical pole observation. Both spectra are mostlyflat with no evidence of a 1/f component up to 3 mHz. Table 1.
Main features of the PGMS observations.Central frequency 2300 MHzEffective bandwidth 240 MHzUseful bandwidth
160 MHzFWHM 8 . l = 254 ◦ Latitude coverage b = [ − ◦ , ◦ ]Area size 5 ◦ × ◦ Pixel size 3 ′ × ′ Observation runs Jan 2006Sep 2006Jan 2007Sep 2007 Q , U beam-size pixel rms sensitivity (halo fields) 0.3 mK Q , U beam-size pixel rms sensitivity (disc fields) 0.5 mK After RFI channel flagging. estimated and cleaned up by the procedure described in Sec-tion 3.The map-making procedure is based on the algorithmby Emerson & Gr¨ave (1988), which combines l – and b –mapsin Fourier space and recovers the power along the directionorthogonal to the scan, otherwise lost through the baselineremoval. The algorithm is highly efficient and effectively re-moves residual stripes.Table 1 summarises the main features of the PGMSobservations. Ground emission can seriously affect continuum observa-tions, especially in our halo fields where the emission hasa brightness of only a few mK. Our tests have shown thatthe highest ground emission occur in the Zenith cap at eleva-tions above 60 ◦ where large fluctuations are observed in thedata. Even though not yet fully understood, the most likelyreason is the loss of ground shielding by the upper rim of thedish at large elevations. The receiver is located at the primefocus and is shielded by the dish up to this elevation, receiv-ing only atmospheric contributions from beyond the upperrim. Above this limit the ground becomes visible, contribut-ing a ground component that rapidly varies as the telescopescans. To avoid this contamination all PGMS observationswere limited to the elevation range EL = [30 ◦ , ◦ ], betweenthe lower limit of the telescope’s motion and this region ofground sensitivity.Even though these precautions have significantly re-duced the effect, some contamination is still present in thehalo fields, requiring us to develop a procedure to estimateand clean the ground contribution. The procedure is based on making a map of the groundemission in the AZ–EL reference frame. Any AZ–EL bingathers the contributions of data taken at different Galacticcoordinates and the weak sky emission is efficiently averagedout. Since the PGMS meridian goes through low emissionregions, its high latitude data are ideal for such an aim.The smooth behaviour of the ground emission enables theuse of large bins, which further helps average out the skycomponent. We therefore use a bin size of ∆EL = 1deg inEL and average over 8 degrees in Azimuth. The binningis performed in the instrument reference frame before thecorrection for parallactic angle.For a given 8 degree AZ bin we smooth the map alongthe EL direction. This reduces the residual local devia-tions that are mainly due to strong point sources. We usea quadratic running fit: for each EL bin, we fit the 7 binscentred at it with a 2-degree polynomial, and the bin valueis then replaced with the fit result at the bin position.In the Azimuthal direction the data are sufficientlysmooth that no fit is necessary in that dimension. We there-fore shift our 8 degree azimuth averages by 1 degree incre-ments, performing the elevation fit for each 1 degree bin.This results in a map of the ground emission in the AZ-EL frame with a bin size of 1 degree in both Azimuth andElevation.The ground emission contamination mainly comes fromthe far lobes, which are frequency dependent. We generateground emission maps for each frequency channel. A set ofthese maps is generated at each observing session. They arechecked for constancy of ground emission before a grandaverage set is formed from all observations.Fig. 4 shows example AZ cuts in the 2300 MHz spec-tral channel. Over a range of about 30 ◦ in both AZ andEL the ground emission varies by less than 50 mK, smallerby an order of magnitude than ground emission variationsreported for other polarization surveys (e.g. Wolleben et al.2006). We attribute this low response to ground emission to c (cid:13) , 1–23 E. Carretti et al.
Figure 4.
Examples of azimuth cuts of the ground emission mapsof both Q (top) and U (bottom) from the 2300 MHz channel ofthe September 2007 observations. the high edge tapering of the S-band feed and its consequentsmall sidelobes.Finally, the sky emission measurements, observed ona 3-arcminute grid, were cleaned of ground emission usingthe model just described. For each sky measurement, theground emission at its actual (AZ, EL) was obtained bylinear interpolation through a standard Cloud-In-Cell tech-nique (Hockney & Eastwood 1981). The low emission fields in the halo have been observed twicegiving two independent maps taken in different runs and atdifferent AZ and EL. That way, they are contaminated bydifferent ground emission enabling us to test the cleaningprocedure and estimate the residual contamination. The er-ror map can be estimated as half the difference of the twomaps, in which the sky is cancelled, leaving the noise andany residual (unmodelled) ground emission.An example is given in Fig. 5 in which one of the fieldswith lowest emission is shown: PGMS-87. The images showStokes Q (left) and U (right) of both the observed (top) anddifference maps (bottom).The difference maps are clearly dominated by whitenoise, indicating that most of the ground emission has been removed at the level needed to measure the sky signal, andthat the residual ground emission does not contribute sig-nificantly to the error budget. That residual can be seen asfaint shadows of large scale structure in the difference maps.These visual impressions of the ground emission removalwe know quantify by measuring the angular power spectra of both the sky and the difference map. Fig. 6 reports themean of the E – and B –mode power spectra – ( E + B )/2– which is the most complete description of the polarizedemission.The spectrum of the sky signal is dominated by thediffuse emission at low multipole ℓ , where it follows a powerlaw C ℓ ∝ ℓ β with a steep slope ( β < − . ℓ end due to both noise and a point sourcecontribution. A white noise spectrum would be flat ( C ℓ =constant).The difference spectrum also follows a power law, butis much flatter than the sky signal. The best fit slope is β noise = − . ± .
06, which, although not pure white noise,is close to the ideal β noise = 0. Furthermore, the differencebetween sky signal and noise increases at large angular scale,giving a rapidly increasing S/N.The ground emission has a smooth behaviour thatmakes the largest scales the most susceptible to contami-nation. On a scale of 2 ◦ the rms fluctuation of the differencemap is N ◦ = 60 µ K, a factor 2.5 larger than that expectedfrom white noise (24 µ K), but much smaller than the skysignal (few mK).This field (PGMS-87) is the worst with regard to groundemission residuals; over all fields, angular spectral slopes β noise fall in the range [-0.7, 0.0] and the rms noise onthe 2-degree scale N ◦ lie in [24, 60] µ K. The mean val-ues over all fields are ¯ β noise = − . N ◦ = 43 µ K. Oncethe pure white noise component is subtracted, the effectivecontribution by only ground emission can be estimated in¯ N grnd , ◦ = 36 µ K. With such results the impact of theground emission may be considered marginal on the finalmapping.
Maps of the Stokes parameters Q , U and the polarized inten-sity L of the PGMS meridian are shown in Fig. 7, 8, and 9,while Fig. 10 displays the whole 10 ◦ × ◦ field (PGMS-34). All images are smoothed with a Gaussian filter ofFWHM = 6 ′ to give a better idea of the sensitivity on beam-size scale, for an effective resolution of FWHM = 10 . ′ . Alldata at latitude | b | > ◦ are plotted with the same in-tensity range to show clearly the power and morphologicaldifferences. The disc fields ( | b | < ◦ ) require a more ex-tended scale. The two strongest sources present in our data(Pic A and NGC 612 in field PGMS-34 and PGMS-77, re-spectively) have been blanked before the map generationof their fields. Without blanking, the high brightness rangecauses the map-making procedure to generate artefacts.The disc has the strongest emission, extending to lat-itude | b | ∼ ◦ with little variation of emission power. At See Section 5 for the description of the power spectrum com-putation. c (cid:13) , 1–23
GMS: observations and CMB foreground analysis Figure 5. Q (left) and U images (right) of the observed (top) and difference maps (bottom) of the field PGMS-87 ( b =[-84 ◦ , -90 ◦ ]).No smoothing is applied for a resolution of FWHM=8.9’. Position coordinates are Galactic latitude and longitude; the brightness unit ismK, the intensity colour scales are linear. Figure 6.
Power spectrum of the mean ( E + B )/2 of E – and B –Modes measured for the PGMS-87 polarized emission: skyemission (diamonds), difference map (stars) and best fit curves(dashed). higher latitudes, the emission starts to decrease up to thehalo where it settles on levels one order of magnitude lower.The clear visibility of the bright polarized disc emission and its contrast with the fainter halo is a new result, notapparent from previous observations carried out at lowerfrequencies where the disc is strongly depolarized up to | b | ∼ ◦ . This allows us to locate the disc-halo boundaryin polarization, which has not been visible so far because ofeither strong depolarization (at 1.4 GHz) or insufficient sen-sitivity (at 23 GHz). A more quantitative analysis is givenin Section 5, but the visual inspection of the maps clearlyshows that it starts at | b | ∼ ◦ .The emission of the halo has a smooth behaviour withthe power mostly residing on large angular scales. The discemission is also smooth, at least at latitudes higher than | b | = 6 ◦ –7 ◦ . Closer to the Galactic plane the pattern has amore patchy appearance, suggestive of Faraday depolariza-tion effects being significant at 2.3 GHz.This supports the view that Faraday depolarization ef-fects are marginally significant in the disc, and are relevantonly in a narrow belt a few degrees wide across the Galacticplane.Several polarized point sources are visible, especially inthe halo where the diffuse emission fluctuations are smaller.To enable a cleaner analysis of the diffuse component thesources are identified, fitted and subtracted from the maps.Each source is located by a 2D-Gaussian fit of the stronger c (cid:13) , 1–23 E. Carretti et al.
Figure 7. Q (left), U (mid), and polarized intensity L = p Q + U images (right) of the six PGMS fields in the latitude range b = [ − ◦ , ◦ ] (PGMS-27 through PGMS-02). Position coordinates are Galactic latitude; the brightness unit is mK.c (cid:13) , 1–23 GMS: observations and CMB foreground analysis Figure 8.
As for Fig. 7 but in the range b = [ − ◦ , − ◦ ] (PGMS-57 through PGMS-34: of the latter only the 5 ◦ across the meridian l = 254 ◦ are imaged). Position coordinates are Galactic latitude; the brightness unit is mK.c (cid:13) , 1–23 E. Carretti et al.
Figure 9.
As for Fig. 7 but in the range b = [ − ◦ , − ◦ ] (PGMS-87 through PGMS-62). Position coordinates are Galactic latitudeand longitude; the brightness unit is mK. c (cid:13) , 1–23 GMS: observations and CMB foreground analysis Figure 10. Q (top), U (mid), and L images (bottom) of the whole 10 ◦ × ◦ fields (PGMS-34). Position coordinates are Galacticlatitude and longitude; the brightness unit is mK.c (cid:13) , 1–23 E. Carretti et al.
Table 2.
Polarization flux limits S lim p used to selected the polar-ized sources. The chosen limits are latitude dependent. b -range S lim p [mJy][ − ◦ , ◦ ] 40[ − ◦ , − ◦ ] 30[ − ◦ , − ◦ ] 20[ − ◦ , − ◦ ] 15[ − ◦ , − ◦ ] 10 component, either Q or U . Its position is then fixed in the fitof the second and weaker component to improve the fit ro-bustness. A polarization flux limited selection is applied withthreshold set to ensure S/N ratios of at least 5. The ampli-tude of fluctuations in the maps is dominated by sky emis-sion (rather than by the instrument sensitivity), which variesalong the PGMS meridian. The threshold we use is thereforea function of Galactic latitude, running from 10 mJy at highlatitudes up to 40 mJy near the Galactic plane (Table 2).In this work, the point source identification is carriedout only for cleaning purposes. The catalogue and a detailedanalysis of their properties are subject of a forthcoming pa-per (Bernardi et al. 2010, in preparation).
The angular power spectra (APS) of E – and B –Mode ofthe polarized emission have been computed for each field.They account for the 2-spin tensor nature of the polarizationand give a full description of the polarized signal and itsbehaviour across the range of angular scales. In addition,the E – and B –Modes are the quantities predicted by thecosmological models enabling a direct comparison with theCMB.To cope with both the non-square geometry and theblanked pixels at the locations of the two brightest sources,we use a method based on the two-point correlation func-tions of the Stokes parameters Q and U described bySbarra et al. (2003). The correlation functions are estimatedon the Q and U maps of the regions as˜ C X ( θ ) = X i X j X = Q, U (1)where X i is the emission in pixel i of map X , and i and j identify pixel pairs at distance θ . Data are binned withpixel-size resolution. Power spectra C E,Bℓ are obtained byintegration C Eℓ = W Pℓ Z π [ ˜ C Q ( θ ) F ,ℓ ( θ ) + ˜ C U ( θ ) F ,ℓ ( θ )] sin( θ ) dθ (2) C Bℓ = W Pℓ Z π [ ˜ C U ( θ ) F ,ℓ ( θ ) + ˜ C Q ( θ ) F ,ℓ ( θ )] sin( θ ) dθ (3)where F ,ℓm and F ,ℓm are functions of Legendre polynomi-als (see Zaldarriaga 1998 for their definition), and W Pℓ is thepixel window function accounting for pixel smearing effects.Since the emission power is best described by the quan-tity ℓ ( ℓ + 1) C ℓ / (2 π ), hereafter we will denote an angularspectrum following a power law behaviour C ℓ ∝ ℓ β as • flat , if β = − .
0: power equally distributed across theangular scales; • steep , if β < − .
0: large scales dominate the power bud-get; • inverted , if β > − .
0: small scales dominates.We tested the procedure using simulated maps gener-ated from a known input power spectrum by the proce-dure synfast of the software package HEALPix (G´orski et al.2005). The input spectra are power laws with differentslopes; for each slope we generated 100 simulated maps andcompute their APS. The mean of the 100 APS reproducedthe input spectrum and its slope correctly, with the excep-tion of an excess at the largest scales, mainly at the firsttwo multipole bands. E – and B –Mode are related to thepolarization angle pattern and this excess is likely due tothe discontinuity of the pattern abruptly interrupted at thearea borders. To account for this we corrected our spectrafor the fractional excess estimated from the simulation as theratio between the mean of the computed and input spectra.For a cleaner measure of the diffuse component, thepoint sources are subtracted from the polarization maps.The E – and B –mode spectra C Eℓ and C Bℓ havebeen computed for the 17 fields along with their mean C ( E + B ) / ℓ = ( C Eℓ + C Bℓ ) /
2. Artificial fluctuations are gen-erated on E and B spectra because of the limited sky cover-age of the individual areas, but their mean suffers less fromthat effect and is a more accurate estimator if the power isdistributed equally between the two modes, as is the casefor Galactic emission. In addition, C ( E + B ) / ℓ gives a full de-scription of the polarized emission which the two individualspectra cannot give separately. Therefore we mostly use themean spectrum ( E + B )/2 to investigate emission behaviourand properties in the following analysis.Fig. 11 shows C ( E + B ) / ℓ for all the fields. As an ex-ample of all three spectra, Fig. 12 shows those of the twofields PGMS-52, which is from the low emission halo, andPGMS-34, our biggest field and the area observed by theBOOMERanG experiment. All spectra are shown withoutcorrection for the window functions. In most fields the spec-tra follow a power law behaviour that flattens at the highmultipole end because of the noise contribution. Exceptionsare the four fields closest to the Galactic plane (PGMS-17to PGMS-02) where a power law modulated by the beamwindow function dominates everywhere. .We fit the angular power spectra to a power law modu-lated by the beam window function W Bℓ for the synchrotroncomponent and a constant term N for the noise: C Xℓ = (cid:20) C X (cid:16) ℓ (cid:17) β X W Bℓ + N (cid:21) W Pℓ , (4)where C X is the spectrum at ℓ = 200 and X = E, B, ( E + B ) / E –Mode, B –Mode, and their mean( E + B )/2. Possible residual contributions by point sourcesare accounted for by the constant term.Plots of the best fits are shown in Fig. 11 and 12, whilethe parameters of the synchrotron component are reportedin Table 3.To analyse the behaviour of the synchrotron compo-nent, the power law component of the best fit spectra areplotted together in Fig. 15 where they are also extrapo-lated to 70 GHz for comparison with the CMB signal. Todetermine the spectral index for the frequency extrapola-tion we computed a map of spectral index of the polarized c (cid:13) , 1–23 GMS: observations and CMB foreground analysis Figure 11.
Angular power spectra C ( E + B ) / ℓ of the 17 PGMS fields. Both the measured spectra (diamonds) and the best fit curve(solid) are plotted.c (cid:13)000
Angular power spectra C ( E + B ) / ℓ of the 17 PGMS fields. Both the measured spectra (diamonds) and the best fit curve(solid) are plotted.c (cid:13)000 , 1–23 E. Carretti et al.
Figure 12.
Angular power spectrum of E –Mode (top), B –Mode (mid), and their mean ( E + B )/2 of the two fields PGMS-52 (left) andPGMS-34 (right). Both the measured spectra (diamonds) and the best fit curve (solid) are plotted. Table 3.
Best fit amplitude C X (referenced to ℓ = 200) and angular spectral slope β X of the PGMS fields ( X = E, B, ( E + B ) / E –, B –Mode, and ( E + B )/2, respectively).Field C E [ µ K ] β E C B [ µ K ] β B C ( E + B ) / [ µ K ] β ( E + B ) / PGMS-02 4500 ± − . ± .
04 4860 ± − . ± .
13 4610 ± − . ± . ± − . ± .
07 4120 ± − . ± .
06 4030 ± − . ± . ± − . ± .
05 3160 ± − . ± .
09 3070 ± − . ± . ± − . ± .
10 1270 ± − . ± .
08 1840 ± − . ± . ± − . ± .
17 246 ± − . ± .
16 376 ± − . ± . ± − . ± .
24 131 ± − . ± .
19 263 ± − . ± . . ± . − . ± .
07 171 . ± . − . ± .
07 180 . ± . − . ± . ± − . ± .
19 61 . ± . − . ± .
14 74 . ± . − . ± . . ± . − . ± .
20 49 . ± . − . ± .
13 56 . ± . − . ± . . ± . − . ± .
13 42 . ± . − . ± .
17 51 . ± . − . ± . . ± . − . ± .
10 16 . ± . − . ± .
21 32 . ± . − . ± . . ± . − . ± .
22 19 . ± . − . ± .
40 24 . ± . − . ± . . ± . − . ± .
19 5 . ± . − . ± .
34 12 . ± . − . ± . . ± . − . ± .
16 14 . ± . − . ± .
33 29 . ± . − . ± . . ± . − . ± .
27 33 . ± . − . ± .
16 38 . ± . − . ± . . ± . − . ± .
35 18 . ± . − . ± .
27 21 . ± . − . ± . . ± . − . ± .
07 11 . ± . − . ± .
07 19 . ± . − . ± . (cid:13) , 1–23 GMS: observations and CMB foreground analysis Figure 13.
Map of frequency spectral index of the polarized syn-chrotron emission computed using the all-sky polarization mapsat 1.4 GHz (Wolleben et al. 2006; Testori et al. 2008) and at22.8 GHz (WMAP-5yr, Hinshaw et al. 2009)
Figure 14.
Distribution of the spectral indexes reported inFig. 13. synchrotron emission (Fig. 13) using the all-sky polariza-tion surveys at 1.4 GHz (Wolleben et al. 2006; Testori et al.2008) and 22.8 GHz (WMAP, Hinshaw et al. 2009). The in-dex distribution at high Galactic latitudes ( | b | > ◦ ) peaksat α pol = − .
21 with dispersion ∆ α pol = 0 .
15 (Fig. 14).This is consistent with the analysis of the WMAP-5yr databy Gold et al. (2009), who find a polarized synchrotron spec-tral index of α WMAP = − . α I = − . α synch = − . • The high and mid latitude fields ( b = [-90 ◦ , -20 ◦ ]), withsteep spectra ( β < − . β ∼ [-3.0, -2.0] (except for a few outliers). Thereis no clear trend with latitude and the slopes are rather Figure 15.
Best fits to the spectra of E –Mode (top), B –Mode(mid), and their mean ( E + B )/2 (bottom) of all PGMS fields.The plot reports only the power law component which describesthe synchrotron emission. All PGMS fields are plotted togetherfor a direct comparison of the behaviour with the Galactic lati-tude. The colour code goes from blue throughout red by a rain-bow palette for the areas from the south Galactic pole (PGMS-87) throughout the Galactic plane (PGMS-02), respectively. Thespectra are scaled to 70 GHz for a comparison with the CMBsignal (frequency spectral index α = − . B –Mode spec-tra for different values of tensor-to-scalar power ratio r are shownfor comparison. The quantity plotted here is ℓ ( ℓ + 1) / (2 π ) ∗ C Xℓ ,which provides a direct estimate of the power distribution throughthe angular scales: a flat spectrum means the power is evenlydistributed; a decreasing spectrum (steep) is dominated by thelargest scales; a rising spectrum (inverted) is dominated by thesmallest scales.c (cid:13) , 1–23 E. Carretti et al.
Figure 16.
Angular spectral slopes of the PGMS fields plot-ted against the field’s centre latitude for E –Mode (top), B –Mode(mid), and ( E + B )/2. uniformly distributed. The median is β ( E + B ) / = − . σ β ( E + B ) / = 0 .
24 is significantly largerthan the individual measurement errors, meaning that thiswide spread is an intrinsic property of the synchrotron emis-sion at these latitudes. We prefer to use the median to estimate the typical angularslope in this region because of the possible significant deviationsby the outliers.
Table 4.
Mean ( ¯ β X ), median ( β X med ), and dispersion ( σ β ) of theangular slopes at mid-high Galactic latitudes for all the threespectra X = E, B, ( E + B ) / β X β X med σ β X E -2.67 -2.62 0.34 B -2.39 -2.46 0.27( E + B ) / Table 5.
As for Table 5 but for low Galactic latitudes.LOW latitudesX ¯ β X β X med σ β X E -1.83 -1.81 0.14 B -1.78 -1.78 0.10( E + B ) / • The low latitude fields ( b = [ − ◦ , ◦ ]), which showinverted spectra ( β > − . β ( E + B ) / = [ − . , − . σ β ( E + B ) / = 0 .
08 (Table 5). All spectra have mean and me-dian slope ¯ β X = β X med = − .
8, which can be considered thetypical value of this region.This change from steep to inverted spectra is quite sud-den and clearly separates two different environments: themid-high latitudes, characterised by a smooth emission withmost of the power on large angular scales, and the disc fields,whose power is more evenly distributed with a slight pre-dominance of the small scales.Does this change indicate an intrinsic feature of the po-larized emission of the disc, or is it the effect of Faradaymodulation, which transfers power from large to small an-gular scales in the low latitude fields? The answer is un-clear with the information available, but some points can benoted. In the disc the ISM is more turbulent than in the haloand the intrinsic emission might have more power on smallangular scales. In addition, the low-latitude lines of sight gothrough much more ISM including more distant structures;these are expected to give more power to the small angularscales. Also the smooth emission of the two highest latitudedisc fields (PGMS-12 and PGMS-17) make the presence ofsignificant Faraday depolarization unlikely. Finally, we havecomputed the power spectrum of the individual frequencychannels to search for a possible variation of the angularslope with frequency. Since the lowest frequencies would bemore affected, steeper spectra at highest frequencies wouldsupport the presence of FR effects. We find that all the fourdisc fields have non-significant slope variation compatiblewith zero within 1.0–1.5 sigma, with the only exception ofPGMS-02 which approaches 2-sigma. All these points sup-port the view that the structure of the low-latitude polarizedemission derives from the intrinsic nature of the synchrotronemitting regions close to the plane, and is not imposed byFaraday depolarization along the propagation path.Considering the amplitude distribution of the PGMSfields, we further divide the mid-high latitude region iden- c (cid:13) , 1–23 GMS: observations and CMB foreground analysis Figure 17.
As bottom panel of Fig. 15, but with spectragrouped according to latitude region. The disc fields (solid, red: b = [ − ◦ , ◦ ]) are clearly distinguished by their spectral slopesand higher amplitudes, and the fields in the transition region (dot-ted, black: b = [ − ◦ , − ◦ ]) have amplitudes quite distinct fromboth the disc and halo fields (solid, blue: b = [ − ◦ , − ◦ ]). tified above into an halo ( b = [ − ◦ , − ◦ ]) and transitionregion ( b = [ − ◦ , − ◦ ]). Thus we identify three distinctlatitude sections: two main regions (disc and high latitudes)well distinguished by both emission power and structure ofthe emission, and an extended transition about 20 ◦ wideconnecting them. Fig. 17 reports the spectra for all field,showing how they belong to the three regions, which: • Halo ( b = [ − ◦ , − ◦ ]): the emission is weak here and,scaled to 70 GHz, is between the peaks of CMB modelswith r = 10 − and r = 10 − . The weakest fields (PGMS-87and PGMS-67) match models with r = 10 − . The fluctua-tions from field-to-field dominate with no clear trend withlatitude. A weak trend might be present with the emissionpower increasing toward lower latitudes, but the effect is aminor in comparison to the dominant field-to-field fluctua-tions. • Galactic disc ( b = [ − ◦ , ◦ ]): the emission is stronger,about two orders of magnitude brighter than that of thehalo. Within the area there is no large variation of the emis-sion power, but slight increase toward the Galactic plane isevident. • Transition strip ( b = [ − ◦ , − ◦ ]): here a transitionbetween the faint high latitudes and the bright disc occurs.This is clearer at large scales, where the northernmost field(PGMS-22) is almost as bright as the weakest disc field(PGMS-17) and the southermost field approaches the up-per end of the halo brightness range.An important consequence is the identification of a cleartransition between disc and halo. The sudden change in theangular spectral slope at | b | ∼ ◦ and the approximatelyconstant emission power from the Galactic plane up to thattransition clearly separate the 20 ◦ equatorial zone from thehigher latitudes. Characterised by a more complex structureof the ISM this area can be associated with the Galactic disc.A second environment characterised by steep spectraand low emission is clearly present for | b | > ◦ . Both an-gular slope and amplitude exhibit wide fluctuations without Figure 18.
Mean of the power spectra of the halo fields ( b =[ − ◦ , − ◦ ]) for E –Mode (top), B –Mode (mid), and their mean( E + B )/2 (bottom). Both the mean of the measured spectra (di-amonds) and its best fit curve (solid) are plotted. any clear trend with latitude. We consider this high Galacticlatitude section as a single environment with regard to itspolarized synchrotron emission properties. Characterised bya smoother emission and simpler ISM structure, this areawe associate with the Galactic halo.The emission of the halo section is very weak. In spiteof large fluctuations, once scaled to 70 GHz the synchrotroncomponent is equivalent to r values between 1 × − and7 × − , which matches the weakest areas observed so far inpolarization. It is worth noticing that PGMS fields PGMS-87 and PGMS-67 have the weakest polarized synchrotronemission observed so far.The high Galactic latitudes above 40 ◦ are thus just oneenvironment, at least in a low emission region not contami-nated by local anomalies like the area intersected by PGMS.This is very important for CMB investigations, since it tellsthat, in principle, it is possible to find large areas with op-timal conditions (extended over 50 ◦ , in the PGMS case).It is thus important to measure the mean emission prop-erties of the entire halo section ( | b | > ◦ ). The mean spec-trum of the 10 halo fields and its best fit are plotted in c (cid:13) , 1–23 E. Carretti et al.
Table 6.
Best fit amplitude C Xℓ and angular spectral slope β X of the mean spectrum of the PGMS halo section ( X = E, B, ( E + B ) / E –, B –Mode, and ( E + B )/2, respectively).Field C E [ µ K ] β E C B [ µ K ] β B C ( E + B ) / [ µ K ] β ( E + B ) / PGMS HALO 40 . ± . − . ± .
10 28 . ± . − . ± .
12 35 . ± . − . ± . Figure 19.
Best fit of the mean halo spectrum C ( E + B ) / ℓ scaledto 70 GHz (solid) alongside CMB B –Mode spectra for different r values (dashed). The mean synchrotron contamination at allhigh Galactic latitudes in also shown for comparison ( C B WMAP )together with the previous estimates in other low emission re-gions: C BS/N< (shaded area), C DRAO , C B BOOM , and C B BaR (seeFig. 15 for details).
Fig. 18; Table 6 gives the best fit parameters; the extrapo-lation to 70 GHz is shown in Fig. 19.The angular slope is β X ∼ − . E , B , and ( E + B )/2 and is thus be considered the typ-ical slope of the halo section. Note that this matches theslope measured at high latitudes by WMAP at 22.8 GHz(Page et al. 2007), indicating that the power distributionthrough the angular scales is the same at 2.3 and 22.8 GHz.This further argues against the significance of Faraday de-polarization, which would have transferred power from largeto small scales and made the angular spectra frequency de-pendent.Once scaled to 70 GHz, the amplitude is equivalent to r halo = (3 . ± . × − , (5)roughly in the middle of the range covered by the individualfields. As mentioned earlier, this is a very low level and corre-sponds to the faintest areas observed previously, which thusseem to be more the normal condition of the low emissionregions rather than lucky exceptions.Finally, the field PGMS-34 deserves mention as the areaobserved by the experiment BOOMERanG. It lies in thetransition region, and although at the high-latitude end ofthis zone, has emission about five times greater than fieldsin the halo. It is suitable for measuring the stronger E -mode(the aim of the 2003 BOOMERanG flight), but our resultsidentify more suitable fields for detecting the B-mode. Figure 20.
Temperature power spectrum of the dust emis-sion model at 94 GHz in the halo section of the PGMS ( b =[ − ◦ , −
40 deg]). The spectrum is the mean of those computedin the individual PGMS fields. Both the mean of the computedspectra (diamonds) and its best fit curve (solid) are plotted. Thelatter is a power law with slope β Td = − . ± .
14 (see text andTable 7).
The dust emission is the other most significant foregroundfor CMB observations. It has a positive frequency spectralindex and dominates the foreground budget at high fre-quency.An estimate of the local dust contribution in the sameportion of halo covered by the PGMS is thus important tounderstand the overall limits of a CMB B –Mode detectionin that area.However, no polarized dust emission has been detectedover the PGMS region, and even the total intensity dustmap of WMAP is noise dominated in that area. We there-fore use the Finkbeiner, Davis & Schlegel (1999) model ofthe total intensity dust emission applying an assumed po-larization fraction. First, we generate maps at 94 GHz usingtheir model-8 for each of the 10 PGMS fields at | b | > ◦ .The temperature power spectra of each are then computedand averaged together to estimate the mean conditions ofthe whole 5 ◦ × ◦ section. Fig. 20 shows both the meanspectrum and its best fit, whose parameters are reported inTable 7. Within the errors, the angular slope matches wellthat of the synchrotron.The total polarized spectrum is estimated from thistemperature spectrum assuming a polarization fraction f pol = 0 .
10, as inferred for high Galactic latitudesfrom the Archeops experiment data (Benoˆıt et al. 2004;Ponthieu et al. 2005). The spectrum is further divided bytwo to account for an even distribution of power between c (cid:13) , 1–23 GMS: observations and CMB foreground analysis Table 7.
Best fit amplitude C T and angular spectral slope β Td of the mean Temperature dust spectrum at 94 GHz in the halosection of the PGMS ( | b | > ◦ ).Field C T [ µ K ] β Td PGMS halo Dust (12 . ± . × − − . ± . Figure 21.
Estimates of the dust polarized emission in the halosection of the PGMS at 70 GHz (solid). The best fit to the tem-perature spectrum is used assuming a dust polarization fraction f pol = 0 .
10 and then scaled to 70 GHz. The synchrotron emissionspectrum estimated in Section 5 is shown for comparison (dot-ted), as well as the CMB B –Mode spectra for three different r values. E – and B –Mode, a reasonable assumption for the Galacticemission.For frequency extrapolations, we use a single Planckfunction modulated by a power law T d ∝ ν α d / ( e hνkT − α d = 2 .
67 and temperature T = 16 . , α PL d = 1 .
5, consistent with the 5-yr WMAP result of α WMAP d = 1 . ± . ± . B –Mode polarized dust spectrum at70 GHz is given in Fig. 21. At this frequency the two com-ponents of Galactic polarized emission are approximatelyequal, and so the total polarized foregound is at a mini-mum. This frequency is mildly dependent on assumed po-larized fraction of the dust component: alternate assump-tions of five percent or twenty percent shift the frequency ofminimum to 80 GHz and 60 GHz respectively.This result is similar to that of the general high Galacticlatitudes and suggests that the synchrotron-to-dust powerratio is only marginally dependent on the strength of theGalactic emission. r To estimate the detection limits of r in the presence of theforeground contamination in the PGMS halo section, we consider an experiment with resolution θ cmb = 1 ◦ at theCMB frequency channel to have adequate sensitivity at the ℓ ∼
90 peak.We also account for the cleaning provided by foregroundseparation techniques. We consider the two cases discussedby Tucci et al. (2005):(i) cleaning by subtracting the foreground map scaled tohigher frequencies using just one frequency spectral indexfor all pixels. It is a coarse method and represents a worstcase. The residual contamination depends on the spread ofspectral indices in the area. We use ∆ α s = 0 .
15 for thesynchrotron component (Section 5; see also Bernardi et al.2004; Gold et al. 2009), and assume the same dispersion forthe dust emission (∆ α d = 0 . clean .(ii) cleaning by foreground subtraction, but assumingknowledge of the frequency spectral index for any individualpixel. In this case the amplitude of the residual contamina-tion depends on the measurement error of the frequencyslopes. Here we assume a combination of the PGMS datawith a synchrotron channel at 22 GHz onboard the CMBexperiment with resolution θ s = 1 ◦ and sufficient sensitivityto give S/N = 5. For the dust emission, we assume that theCMB experiment includes a dust channel at 350 GHz witha sensitivity to allow a
S/N = 5 and resolution scaled fromthe CMB channel ( θ d = θ cmb ν cmb /ν d , where ν cmb and ν d are the frequencies of the CMB and dust channel). We referto this method as clean .These two methods are at the two ends of cleaning capabil-ities ( ℓ ∼
90 CMB peak, but has a flat white–noise–like spectrumwhich performs much better on large angular scales. Whilethe simpler clean ℓ ∼
90, the clean r is the only parameter to be measured,(other cosmological parameters being provided by Temper-ature and E –Mode spectrum from other experiments suchas WMAP or PLANCK), the Fisher matrix reduces to thescalar c (cid:13) , 1–23 E. Carretti et al.
Figure 22.
Residual contaminations after foreground cleaning in the PGMS halo section at 70 (left) and 150 GHz (right) for synchrotron(red) and dust emission (blue). Results for both the two cleaning methods described in text are shown. The gravitational lensingcontribution is also shown assuming that cleaning has reduced its contribution by a factor 10 in the power spectrum (green).
Table 8.
Detection limits of r at 3-sigma C.L. ( δr ) at 70 and 150 GHz in the PGMS halo area and in a 2500-deg region assumed tohave equivalent foreground contamination (see text). Results for both cleaning types described in the text are reported.Area clean type δr (70 GHz) δr (150 GHz)PGMS clean . × − . × − PGMS clean . × − . × − clean . × − . × − clean . × − . × − F rr = X ℓ C Bℓ ) (cid:18) ∂C Bℓ ∂r (cid:19) (6)and the rms error on r is σ r = F − / rr . (7)The uncertainty ∆ C Bℓ of the B –mode spectrum is a functionof the CMB spectrum C B, cmb ℓ and the cleaning residuals ofsynchrotron, dust, and gravitational lensing:∆ C Bℓ = r ℓ + 1)∆ ℓ f sky C B, cmb ℓ + (cid:0) C B, synch − res ℓ + C B, dust − res ℓ + C B, lens − res ℓ (cid:1) , (8)where ∆ ℓ is the width of the multipole bins and f sky isthe sky coverage fraction. As a set of cosmological param-eters we use the best fits of the 5-yr WMAP data release(Komatsu et al. 2009).At 70 GHz the detection limits of r in the PGMS haloregion is δr ∼ × − (3-sigma C.L.) for both cleaningmethods (Table 8). This is a very low level which makes thePGMS strip an excellent target for CMB experiments andenables accessing levels of r much lower than previously es-timated for areas of comparable size (e.g. Tucci et al. 2005;Verde et al. 2006, who used higher foreground levels esti-mated from total intensity data). An important point is thatthere is only a marginal benefit from using the more sophis-ticated cleaning method. For a 250-deg area most of thesensitivity resides in the ℓ ∼
90 peak, where the dominant residual is the gravitational lensing and a better cleaning ofthe other contaminants is not critical.This result is therefore quite robust since it is based onactual measurements of the foreground contamination in aspecific area and is marginally dependent on the cleaningmethod. Moreover, the leading residual term is the gravi-tational lensing, giving a good margin against errors in ourdust polarization fraction assumption.It is also important to estimate δr for 150 GHz, a fre-quency that, although far from the foreground minimum,is preferred by experiments based on bolometric detectors.However, as shown in Fig. 22, right panel, the major residualat this frequency is dust emission, making the result depen-dent on the assumed dust polarization fraction f pol . A valueof f pol = 0 .
10 gives the limit δr = 6–8 × − (Table 8),while the goal of the most advanced sub-orbital experimentsplanned for the next years ( r ∼ × − ) is still achievableunder the reasonable assumption that f pol < . ◦ long and itis unlikely that its optimal conditions are confined to its 5 ◦ width. Therefore we consider that a larger area, say of 50 ◦ × ◦ extent, could be identified with properties comparableto those of the PGMS halo region. Such an area, about 6per cent of the sky, matches in size the southern portionof the low emission region identified in the WMAP data(Carretti et al. 2006b).The detection limit achievable over such an area dropsto δr = 5 × − (3-sigma C.L.) if method δr = 1 . × − under the coarser method c (cid:13) , 1–23 GMS: observations and CMB foreground analysis Note that in this case there is a significant differencebetween the two cleaning methods, justifying the use of themore sophisticated method
The PGMS has mapped the radio polarized emission at allGalactic latitudes in a 5 degree strip at a frequency suffi-ciently high not to be affected by Faraday depolarizationand with sufficient sensitivity to detect the signal in lowemission regions. It is the largest area observed so far athigh Galactic latitude uncontaminated by large local struc-tures.This has enabled us to investigate the behaviour of thepolarized emission with latitude by computing the polarizedangular power spectrum in 17 fields from the Galactic planeto the South Galactic pole. We can distinguish three lati-tude sections: two main regions well distinguished by bothbrightness and structure of the emission (disc and halo), andan extended transition connecting them. In detail they are:1. The halo at high Galactic latitudes ( b = [ − ◦ , − ◦ ])characterised by low emission fields with steep spectra (an-gular slope β = [ − . , − . β med = − . b = [ − ◦ , − ◦ ])whose angular spectra are steep like those of the halo, butshows an increase of the emission power with decreasinglatitude.3. The disc at low latitudes ( b = [ − ◦ , ◦ ]) characterizedby inverted spectra with slopes in a narrow range of median β med = − .
8. The amplitudes are two orders of magnitudebrighter than in the halo and the power gradually increasestowards the Galactic plane.The change in the angular slope around b = − ◦ isabrupt and identifies a sharp disc-halo transition from thesmooth emission of the mid-high latitudes to the more com-plex behaviour of the disc; this is likely related to the moreturbulent and complex structure of the ISM in the disc.The halo section has no clear trend with latitude ofeither emission power or angular slope, and, at least alongthe meridian sampled by PGMS, can be considered a singleenvironment. This is very important for CMB investigations,as it indicates that it is possible to find large areas withoptimal conditions for seeking the B –Mode.The synchrotron emission of the whole halo section isvery weak. Once scaled to 70 GHz it is equivalent to r =3 . × − , so that an experiment aiming for a detection limitof δr = 0 . r detection limit of this area account-ing for the use of foreground cleaning procedures. We applyboth a coarse and a more refined method. The Galactic emis-sion is so low that the dominant residual contamination isfrom gravitational lensing, even assuming a 10-fold reduc-tion of the lensing foreground from cleaning. For both thetwo cleaning methods the detection limit is δr ∼ × − (3-sigma C.L.) if the CMB B –Mode search is conducted at70 GHz. This result provides a sound basis for investigat-ing the B –Mode. The detection limit we have found here iseven better than the goals of the most advanced sub-orbitalexperiments ( r ∼ .
01, e.g.: SPIDER, EBEX, and QUIET,Crill et al. 2008; Grainger et al. 2008; Samtleben 2008) andproves that there exists at least one area of the sky where itis realistic to carry out investigations of the B –mode downto very low limits of r .At 150 GHz the detection limit rises, but is still bet-ter than δr = 0 .
01 assuming a reasonable dust polarizationfraction ( f pol < ◦ along one dimension and it is unlikely that its ex-cellent conditions are confined to its 5 ◦ width. We have ex-plored the likely results from a larger 50 ◦ × ◦ area, havingproperties similar to those of the PGMS halo section. In sucha region the r detection limit would drop to δr = 5 × − (3-sigma) at 70 GHz.It is worth noticing that the gravitational lensing needsto be cleaned to take advantage of the low Galactic emissionof the PGMS halo section. This can be effectively carried outonly with high resolution data (10 ′ or finer, Seljak & Hirata2004) and the design of CMB experiments should complywith that rather than be limited to 1-degree to fit the peakat ℓ ∼ B –Mode and associated inves-tigations of the inflationary scenarios. While the detectionlimit is limited to δr = 1 × − by the detector array sizeand sensitivity, observations at 150 GHz might be sufficientif conducted in an area like the PGMS with clear advantagesof using the currently best detectors (bolometers) and of anexperiment more compact than at 70 GHz. Moreover, thesynchrotron emission is sufficiently weak at 150 GHz not torequire any cleaning, which removes the need for low fre-quency channels to measure it. Experiments like EBEX andBICEP already match such conditions, not only because ofthe design choices, but also because their target areas inter-sects the PGMS strip (Grainger et al. 2008; Chiang et al.2009).A deeper detection limit, down to δr = 2 × − , couldbe reached by a sub-orbital experiment observing the samearea but with the CMB channel shifted to 70 GHz. Theinclusion of a channel at a lower frequency would be requiredto measure the synchrotron component. Finally, detectionslimits down to δr = 5 × − coiuld be achieved by observingat 70 GHz in a large area of 50 ◦ × ◦ having the PGMSforeground levels. The location of the most suitable regionmust be determined, a task that can be accomplished bythe forthcoming large foreground surveys like the S-band c (cid:13) , 1–23 E. Carretti et al.
Polarization All Sky Survey (S-PASS), or the C-band AllSky Survey (C-BASS).These limits are comparable to the goals of the spacemissions currently under study such as B-POL and CMBPol(de Bernardis et al. 2008; Baumann et al. 2008), but withthe significant advantage that such an area is still sufficientlycompact to be observable by a sub-orbital experiment. Thelimit r ∼ × − is an important threshold for the in-flationary physics since it is about the lower limit of theimportant class of inflationary models with low degree offine tuning (Boyle et al. 2006). Our study shows that thisthreshold may be reached with an easier and cheaper sub-orbital experiment rather than a more complex space mis-sion, making this goal more realistically achievable with asmaller budget and in a shorter time than that required todevelop space-borne equipment.The PGMS data will be made available at the site ACKNOWLEDGMENTS
This work has been partly supported by the project SPOrtfunded by the Italian Space Agency (ASI) and by the ASIcontract I/016/07/0
COFIS . M.H. acknowledges supportfrom the National Radio Astronomy Observatory (NRAO),which is operated by Associated Universities Inc., under co-operative agreement with the National Science Foundation.We wish to thank Warwick Wilson for his support in theDFB1 set-up, John Reynolds for the observations set-up,and an anonymous referee for useful comments. Part of thiswork is based on observations taken with the Parkes RadioTelescope, which is part of the Australia Telescope, fundedby the Commonwealth of Australia for operation as a Na-tional Facility managed by CSIRO. We acknowledge the useof the CMBFAST and HEALPix packages.
REFERENCES
Amarie M., Hirata C., Seljak U., 2005, PRD, 72, 123006Baumann D., et al., 2008, arXiv:0811.3911 [astro-ph]Beck R., 2008, in ”The UV Window to the Universe”, eds.A.I. Gomez de Castro and M. Castellanos, Ap&SS, inpress, arXiv:0711.4700 [astro-ph]Benoˆıt A., et al., 2004, A&A, 424, 571Bernardi G., Carretti E., Fabbri R., Sbarra C., Poppi S.,Cortiglioni S., Jonas J.L., 2004, MNRAS, 351, 436Bernardi G., Carretti E., Sault R.J., Cortiglioni S.,Poppi S., 2006, MNRAS, 370, 2064Boyle L.A., Steinhardt P.J., & Turok N., 2006, Phys. Rev.Lett. 96, 111301Brown J.C., Haverkorn M., Gaensler B.M., Taylor A.R.,Bizunok N.S., McClure-Griffiths N.M., Dickey J.M.,Green A.J., 2007, ApJ, 663, 258Brown M.L., et al, 2009, ApJ, 705, 978Beuermann K., Kanbach G., & Berkhuijsen E.M., 1985,A&A, 153, 17Carretti E., Bernardi G., Sault R.J., Cortiglioni S., & PoppiS., 2005a, MNRAS, 358, 1 Carretti E., McConnell D., McClure-Griffiths N.M.,Bernardi G., Cortiglioni S., & Poppi S., 2005b, MNRAS,360, L10Carretti E., Poppi S., Reich W., Reich P., F¨urst E.,Bernardi G., Cortiglioni S., Sbarra C., 2006a, MNRAS,367, 132Carretti E., Bernardi G., Cortiglioni S., 2006b, MNRAS,373, L93Chiang H.C., et al, 2009, submitted to ApJ,arXiv:0906.1181 [astro-ph.CO]Crill B.P., et al., 2008, in Space Telescopes and In-strumentation 2008: Optical, Infrared, and Millime-ter, Eds. J.M. Oschmann Jr., M.W.M. de Graauw &H.A. MacEwens, Proc. of SPIE, 7010, 70102Pde Bernardis P., Bucher M., Burigana C., Piccirillo L.,2008, Exp. Astron., in press, arXiv:0808.1881 [astro-ph]Emerson D.T., Gr¨ave R., 1988, A&A,190, 353Finkbeiner D.P., Davis M., Schlegel D.J., 1999, ApJ, 524,867Frick P., Stepanov R., Shukurov A., & Sokoloff D., 2001,MNRAS, 325, 649Gaensler B.M., Dickey J.M., McClure-Griffiths N.M.,Green A.J., Wieringa M.H., & Haynes R.F., 2001, ApJ,549, 959Gold B., et al., 2009, ApJS, in press, arXiv:0803.0715[astro-ph]G´orski K.M., et al., 2005, ApJ, 622, 759Grainger W., et al., 2008, in Millimeter and SubmillimeterDetectors and Instrumentation for Astronomy IV, Eds.W.D. Duncan, W.S. Holland, S. Withington, J. Zmuidzi-nas, Proc. of SPIE Vol. 7020 70202N-2Han J.L., 2002, in Astrophysical Polarized Backgrounds,eds. S. Cecchini, S. Cortiglioni, R. J. Sault, C. Sbarra,AIP Conf. Ser., 609, 96Han J.L., Manchester R.N., Berkhuijsen E.M., & Beck R.,1997, A&A, 322, 98Han J.L., Manchester R.N., Lyne A.G., Qiao G.J., vanStraten W., 2006, ApJ, 642, 868Haverkorn, M., Gaensler B.M., McClure-Griffiths N.M.,Dickey, J.M., Green A.J., 2006, ApJS, 167, 230Hinshaw G., et al., 2009, ApJS, 180, 225Hockney R.W., Eastwood J.W., 1981, Computer Simula-tion Using Particles (New York: McGraw-Hill)Jansson R., Farrar G.R., Waelkens A.H., & Ensslin T.A.2009, JCAP, 7, 21Kamionkowski M., Kosowsky A., 1998, PRD, 57, 685Komatsu E., et al., 2009, ApJS, 180, 330Kraus J.D., 1986. Radio Astronomy (Ed. CygnusQuasarBooks: Powell, OH)Lange A., 2008, in CMB component sep-aration and the physics of foregrounds,http://planck.ipac.caltech.edu/content/ForegroundsConference/presentationsForWEB/58 andrewLange.pdfLa Porta L., Burigana C., Reich W., Reich P., 2006, A&A,455, L9Masi S., et al., 2006, A&A, 458, 687Page L., et al, 2007, ApJS, 170, 335Peiris H.V., et al, 2003, ApJS, 148, 213Ponthieu N., et al., 2005, A&A, 444, 327Reynolds J.E. 1994, ATNF Tech. Doc. Ser. 39.3040Samtleben D., in Cosmology 2008, Proc. of the 43rd Ren-contres de Moriond, in press, arXiv:0806.4334 [astro-ph]Sbarra C., Carretti E., Cortiglioni S., Zannoni M., Fab- c (cid:13) , 1–23 GMS: observations and CMB foreground analysis bri R., Macculi C., Tucci M., 2003, A&A, 401, 1215Seljak U., Hirata C.M., 2004, PRD, 69, 043005Sun X.H., Reich W., Waelkens A., Enßlin T.A., 2008, A&A,477, 573Tegmark M., 1997, PRD, 56, 4514Tegmark M., Taylor A.N., Heavens A.F., 1997, ApJ, 480,22Testori J.C., Reich P., Reich W., 2008, A&A, 484, 733Thomas B.M., Schafer J.T., Sinclair M.W., Kesteven M.J.,Hall P.J., 1997, IEEE Antennas & Propagation Magazine,39, 54Tucci M., Mart´ınez-Gonz´alez E., Vielva P., Delabrouille J.,2005, MNRAS, 360, 935Verde L., Peiris H.V., Jimenez R., 2006, JCAP, 1, 19Wolleben M., Landecker T.L., Reich W., Wielebinski R.,2006, A&A, 448, 411Zaldarriaga M., 1998, Ph.D. Thesis, M.I.T.,astro-ph/9806122This paper has been typeset from a TEX/ L A TEX file preparedby the author. c (cid:13)000