Table Of Content

Mon.Not.R.Astron.Soc.000,000–000 (0000) Printed1February2008 (MNLATEXstylefilev1.4) Red synchrotron jets in Parkes quasars M.T.Whiting1, R.L.Webster1 and P.J.Francis2,3 1 Astrophysics group, School of Physics, Universityof Melbourne, Victoria 3010, Australia 2 Research School of Astronomy and Astrophysics, Australian National University,Canberra, ACT 0200, Australia 3 Joint Appointment withthe Department of Physicsand Theoretical Physics, Faculty of Science 1February2008 1 0 0 ABSTRACT 2 WepresentmodelfitstospectralenergydistributionsintheopticalandNIRof>100 n flat-spectrum radio quasars from the Parkes Half-Jansky Flat-spectrum Sample. We a find that ∼ 40% of the sources have power-law SEDs, while a similar number show J evidencefortwoprimarycomponents:abluepowerlawandopticalsynchrotronemis- 9 sion.Thebluepowerlawissimilartothedominantcomponentobservedinthespectra 2 of optically-selected QSOs. There is strong evidence that the synchrotron component hasaturnoverinthe UV-opticalrestframeofthe spectrum.Inthe remainingsources 1 it is likely that the synchrotron peaks at longer wavelengths. This mixture of two v components is supported by optical polarisation measurements in a subgroup of the 2 sources.The sourceswith power lawSEDs showevidence for an excessnumber of red 0 powerlawslopescomparedtooptically-selectedquasars.Thereareadditionalspectral 5 1 componentsinsomeofthesources,suchasdustandtheunderlyinggalaxy,whichhave 0 not been considered here. 1 Keywords: quasars:general–BLLacertaeobjects:general–radiationmechanisms: 0 non-thermal / h p - o r 1 INTRODUCTION NIRwavelengthscomesfromboththevariabilityintheNIR, t s andfrom thestrongpolarisation of afew sources.Stickelet a The spectral energy distributions of Active Galactic Nuclei al.(1996)foundthatvariabilitywasageneralcharacteristic v: provideasum of theemission processes contributingto the ofoptically faint flat-spectrumradiosources, althoughthey i energyoutputoftheAGN.Clearly,severalmechanismscon- did not specifically discuss synchrotron emission. Instead, X tributeatmostfrequencies.Ifwecandeterminethedifferent they suggest that reddening due to intervening galaxies or r components as a function of frequency, then we can deter- thehost galaxy itself was thelikely cause of thered optical a mine the energy generation mechanisms which are impor- colours. tant. While radio jets are quite common, and have been Flat-spectrumradioquasarsareknowntobedominated widelyobserved,opticalcounterpartstothesejetsarequite intheradiospectrumbysynchrotronemission.Itislessclear rare. In fact, only 14 optical jets are currently known (see how far these synchrotron spectra extend towards higher O’Deaetal.(1999)andreferencestherein).Thefeatures(i.e. frequencies,orthemagnitudeofthecontributiontheymake emissionknots)intheopticaljetsmatchthepositionsoffea- to the optical and near-infrared (NIR) spectrum – a region tures in the radio jets, indicating that the emission regions dominated bythe Big Blue Bump. andmechanismsforthetwospectralregimesareassociated. Thefar-IRemissionofflat-spectrumquasarsappearsto Optical jets can be used to put strong constraints on the be dominated by synchrotron emission (Haas et al. 1998), particleenergetics,byprovidinglimitsonthemaximumen- consistent with aDoppler-boosted component that swamps ergies,andontheaccelerationmechanisms,duetotheshort the expected emission from dust components. Synchrotron lifetimes of particles at these energies (Meisenheimer et al. emission in the optical has been considered for many years 1996). tobenecessarytoexplaintheopticaltoNIRphotometryof The quasars considered in this paper are selected opticallyfaintandredflat-spectrumradioquasars.Thishas fromtheParkesHalf-JanskyFlat-spectrumSample(PHFS) mostlybeenonthebasisofthesteepnessoftheoptical–NIR (Drinkwater et al. 1997), which consists of 323 objects se- spectra(Riekeetal.1979; Beichmann 1981;Bregman 1981; lected to be radio-loud (S >0.5 Jy), and have a flat 2.7 GHz Riekeetal.1982).Thesteepnessofthesespectrasuggesteda radiospectrum(α > 0.5,S(ν) να).Thesequasars 2.7/5.0 − ∝ cutoff,orsharpbreak,intheelectronenergyspectrum.Sup- have been shown (Webster et al. 1995; Francis et al. 2000) portingevidenceforthepresenceofsynchrotronemissionat to have a large spread in B K colours, with the reddest − (cid:13)c 0000RAS 2 M.T.Whiting, R.L.Webster and P.J.Francis objectshavingB K >7.Masci,Webster&Francis(1998) redshift galaxies, which had a prominent 4000˚A break be- − showed that this spread could not be accounted for by the tween the B and V bands. These sources show strong ev- emission from the host-galaxy. idence for flux from the underlying galaxy in their spectra In this paper we test the idea that optical synchrotron (Masci et al. 1998), and thus need an additional galactic emission causes the red optical-NIR colours of the PHFS component to be modelled accurately. Sources with z > 3 quasars. Models representing emission from both an opti- were also excluded,as these had strong Lyαbreaks present cal synchrotron component and a blue optical power law betweenB andV.Thisreducedthenumberofsourcesfrom (representing the continuum emission from an unreddened 157 to117. quasar) arefittedtobroad-bandoptical-NIR spectra.Stan- Twenty-one of the reddest sources (B K >5) in the − dard goodness-of-fit techniquesare used todetermine mod- subsample were amongst those excluded, as they have only els which are consistent with the observations. upper limits on B, V and possibly R, or no redshift. The Thedataset wascompiled byFrancis,Whiting&Web- sources with upper limits are both red and optically faint. ster(2000)(FWWhereafter),andcomprisesbroadbandop- Sincethefluxofthesesourcestypicallydecreasesrapidlyin ticalandNIRphotometryinthebandsB,V,R,I,J,H and the blue (i.e. from V to B), there are three possible expla- K. This photometry is quasi-simultaneous, meaning all ob- nations for them. They are either dust-reddened (causing servations for agiven source weremade within several days thebluedecreaseinflux),highredshift(andhenceLyαab- (six at most) of one another. This minimises the effects of sorbed), or they are dominated by synchrotron that turns source variability. The source selection used for this paper overrapidly.Futureanalysisofthespectralenergydistribu- is explained in Section 2. The simultaneity, as well as the tions (SEDs) of these sources will only strengthen our final breadthand densityof thespectral coverage of thisdataset conclusions. providesanexcellentbasistomodelthebroadbandemission Those sources excluded purely because of their lack of from a large numberof flat-spectrum quasars. measured redshift generally have high B K colours (B − − The dataset is described in Section 2. In Section 3 we K > 5), since the most likely reason they have no redshift test thehypothesisthattheemission in theoptical-NIR re- isthattheyarefaintintheoptical(andinparticularinthe gion is well-fitted by a single power law. In Section 4 we blue). Further discussion of the sources excluded from the describethemoresophisticatedmodelthatwasfittedtothe sample is given in Section 10. data–representingtheaccretiondiskandsynchrotronemis- sion–andthemethodthatwasusedtoperformthefitting. The results of this fitting are described in Sections 5 & 6, 2.2 Errors on photometry while further tests of the model fits are examined in Sec- ThephotometrygiveninFWWquotederrorbars,wherethe tion 7, using polarisation and emission line measurements. estimated error comprised two parts: a random error com- Apossiblerivalmodel–thatofblack-bodyemission byhot ponent and an assumed 5% error in the photometric zero dust–isconsideredinSection8.Theeffectofemissionlines points, which were added together in quadrature. The pho- on the photometry and the resulting fits is investigated in tometric zero point errors were estimated from the scatter Section 9. Finally, Section 10 contains discussion of the re- inzeropointsbetweendifferentstandardstarmeasurements sults and their implications. in an individual night: Francis et al. adopted a value of 5% to account for this error. However, thiszero point error ignores a numberof fac- 2 DATA tors that we believe may be important for the analysis in thispaper. These factors are as follows: The data which have been fitted by these models are described in detail in FWW. A total of 157 sources from the The optical zero magnitude fluxes were taken from PHFS were observed, with quasi-simultaneous broadband • Bessell,Castelli&Plez(1998).Thesefluxeswerederivedfor photometryobservationsinthebandsB,V,R,I,J,H and anA0star,andsowillbeslightlyincorrectforaquasarspec- K.Thesemagnitudeswereconvertedintobroadbandfluxes trum. This has the effect of introducing small colour terms using thezero points given in the same paper. into the photometry, the size of which will depend on the spectralindexof theobject beingobserved.Asimilar effect will of course be present in the NIR.Bersanelli, Bouchet & 2.1 Excluded sources Falomo (1991) found that spectral shape differences could Not all sources in FWW were used. Firstly, we only con- produce systematic errors of at least a few percent in the sidered those sources with complete contemporaneous pho- NIRflux. tometry in all bands, or those with only one observation The zero point fluxes in the optical are taken from • missing (either not observed or an upper limit). This was a different reference (Bessell et al. 1998) to those in the to ensure that the number of degrees of freedom in the NIR,whichwerecalculated byP.McGregor (CASPIRman- modelfittingwasgreaterthanzero.Thosesourceswithmore ual,MSSSO,ANU),assumingthatVegaiswellrepresented than one band missing were not included in the analysis. in the NIR by a black body of temperature 11200 K, and Sources without a measured redshift were also excluded, normalisation F (555nm) = 3.44 10−12W cm−2 µm−1 λ × since the redshift is needed to obtain the correct shape (Bersanelli etal.1991). Thesedifferentzeropointsmaynot of the observed synchrotron spectrum. One of the sources be exactly equivalent, which will produce small offsets be- that had an unknown redshift (z = 0) in Drinkwater et al. tween the optical and NIRparts of the SED. (1997),0829+046,hasapublishedredshiftvalueofz=0.18 IftheskyconditionsatSidingSpringObservatorywere • (Falomo 1991), which is used here. Also excluded were low not completely photometric for all observations for a given (cid:13)c 0000RAS,MNRAS000,000–000 Red synchrotron jets in Parkes quasars 3 source(particularlyifthetransparencychangedbetweendif- ferentbands),thenthemeasuredphotometrywillhavesmall band-to-banderrors present. The observations for each source were taken quasi- • simultaneously (meaningall observationsweremadewithin atmostasix-dayperiod),tominimisetheeffectsofvariabil- ity.However,anumberofthesourcesinthePHFShavebeen foundtoexhibitintra-dayvariabilityintheoptical(Romero et al. 1999; Heidt & Wagner1996), and a larger numberno doubthavesimilarpropertiestothese.Therefore,variability on timescales of the order of those separating our observations is likely for some of the sources. Such variability can beup to 0.1 mag over theperiod of a night. Finally, the presence of strong emission lines in the • quasars’ spectra could boost the flux of a band above the levelofthecontinuum.Wehavetriedtoquantifythiseffect in Section 9, although this is not possible to do for many sources,duetothelackofbothphotometryandaspectrum. Some additional effects may be due to line blends (such as Feiiblends), but this is probably not as strong an effect as you would find in a radio-quiet sample. Figure 1. Histogram of fitted power law indices. Hatched histogram indicates good fits, while open histogram indicates all Hence, to take account of all these factors, we have in- fits. creasedthesystematicerrorinthephotometryto10%.The random error is kept at the same level as that presented in FWW, and is added in quadratureto thesystematic error. distributionofresultingpowerlawindices,bothforthegood fitsand for all sources, is shown in Fig. 1. Thespectralindicesofthesegoodfitsspanawiderange of values. At one extreme there are the sources with rela- 3 POWER LAW MODEL tively blue SEDs (α . 1.4). These sources are charac- − The object of our analysis is to find physical models to ex- terised by their blue continuum, the presence of moderate plaintheopticalandNIRemission.FWWfoundthatabout tostrongemission lines,andgenerally lowX-rayflux(most 90% of thePHFS haveapproximately powerlaw SEDs.We were not detected byROSAT (Siebert et al. 1998)). firstwishtotestthismorerigorously.Asastartingpoint,we AttheotherextremearethesourceswithredderSEDs choose to fit (naively) a simple power law, with an uncon- (that is, flatter in f ), with 1.α. 0. These sources are λ − strained spectral index. This will separate out the sources blazar typeobjects, with high optical polarisation (Wills et that have power law SEDs from those that show some cur- al.1992)aswellasrelativelyweak(orevenabsent)emission vaturein theirspectrum. lines – in fact, all the BL Lac objects in our sample are in thisregion. 3.1 Fits to data The model we choose to fit is f (λ) = cλα, so that the 3.2 Interpretation of power laws PL normalisation c and the spectral index α vary. (Note that this model implies f ν−2−α.) So,wehavefitapowerlawtoalargemajorityofthesources ν ∝ in our sample, spanning a wide range of spectral indices. This model is fit to the data using a least-squares Arethephysicalprocesses that generatethispower law the method. This generates a χ2 value: sameforallsources?Thatis,doesthepowerlawintheblue 7 (y f (λ ))2 sources havethe same origin as that in thered sources? χ2 = i− σP2L i , Thefirst class of sources –thebluesources –are being Xi=1 i fit by a blue power law, which has similar colours to the which indicatesthegoodness of fit.A fittoasourcewill be blue power law emission seen in optically-selected quasars deemed to be “good” when the value of χ2 is less than the (Francis 1996). This is likely to be the optical part of the cut-off value corresponding to the 99% confidence level of continuum emission from the accretion disk (often termed the χ2 distribution. For a source with 5 degrees of freedom thebig bluebump). (asisthecaseformost sourceswiththispowerlaw model), However,thepowerlawbeingfittothereddersourcesis this cutoff level is 15.09. (Note that increasing the value of mostlikelyofdifferentorigintothatseeninthebluesources. this cutoff is equivalent to increasing the confidence level These objects exhibit characteristics commonly associated – for example, the 99.5% cutoff is 16.75 for 5 degrees of with optical synchrotron emission (such as high optical po- freedom.) If the χ2 value is greater than this cutoff level, larisation and lack of prominent emission lines), and so we thenwerejectthenullhypothesisthatthepowerlawmodel postulatethatthisemissionis,atleastinpart,someformof fitsthe data. synchrotron emission. The slope of the power law can then When we fit this power law model to thedata, we find beusedtodeterminep,thepowerlaw indexoftheelectron that 83 sources (or 71% of the total) have good fits. The energydistribution(i.e.definedsuchthatN(E) E−p:see ∝ (cid:13)c 0000RAS,MNRAS000,000–000 4 M.T.Whiting, R.L.Webster and P.J.Francis Section 4.2). Using the values shown in Fig. 1, we obtain Over the wavelength of our observations, the BBB is 2<p<6 (using 1.5<α<0.5 and p=2α+5). modelledasasimplepowerlaw, f λαB orf ν−2−αB. λ ν − ∝ ∝ If synchrotron emission is present in the spectra of at Francis (1996) found the median slope for a subsample of least some of these quasars, then we can ask the question quasarsfromtheLargeBrightQuasarSurvey(LBQS),taken “Isthesynchrotroncomponentbestmodelledbythispower from optical/NIR photometry, was f ν−0.35±0.3, and ν ∝ law?” Thesynchrotron component will bepresent in one of notedthattheobservationswereconsistentwithanintrinsic twoforms: apower law caused byan unbroken(power law) continuum slope of f ν−0.3 that is reddened by various ν ∝ electron energy distribution, or a turning-over component amounts of dust. We therefore take our value of α to be B caused by a break or a cut-off in the electron energy distri- 1.7. The effects of allowing the value of α to vary are B − bution. (Note that the power law can also be produced by considered in Section 10. a synchrotron spectrum turning over at higher frequencies than those observed.) Bothoftheseformscanbetested.Thepowerlawmodel 4.2 Synchrotron emission must produce power law indices that are consistent with slopes of plausible energy distributions. The presence or All the PHFS sources are radio-loud flat-spectrum sources, otherwise of a turnover can be evaluated by examining the and thus very likely have relativistic jets that emit syn- sharpest possible turnover (caused by an abrupt cut-off in chrotron radiation, at least at radio frequencies. Could this theenergydistributionatsomemaximumenergy).Thiswill synchrotron emission extend up to the optical/NIR part of provide the maximum contrast with the power law, and is thespectrum?Ourpowerlaw fittingfrom theprevioussec- consistent with modeling done by other authors (Meisen- tion provides circumstantial evidence for this: the redder heimer et al. 1996, for example). sources in the optical tend to be the ones with higher po- Many of the sources, while they have power law fits larisation (a good sign of synchrotron emission) and less that we can not reject at the 99% confidence level, show prominent emission lines (possibly a sign that the emission evidenceforcurvatureintheirSEDs.Thiscurvaturecanbe lines are being swamped by the presence of a synchrotron ‘n’-shaped (higher in the middle than at either end), ‘u’- component).Thesepiecesofevidenceareinvestigated more shaped (lower in the middle) or perhaps take the form of deeply in Section 7. aninflection(e.g.thefluxdecreases,levelsoffanddecreases As discussed above, the synchrotron spectrum could again). (See Fig. 3 for illustrations of the different types.) take the form of either a power law, from a power law dis- MostofthesourcesclassedasBLLacobjectsare‘n’-shaped, tribution of electron energies, or a power law with a break and so we postulate that these sources are dominated by a orturn-over,duetoanelectronenergydistributionthatex- synchrotron component that is turningover in the optical. hibitsabreakorevenacut-off.Thislattertypeofspectrum Othersources,however,arebluerintheopticalthanin hasbeenseeninopticalsynchrotronjets(O’Deaetal.1999; the NIR (i.e. show an inflection, without the turn-up seen Scarpaetal.1999),wheretheopticalspectrumislikeν−1.2– in ‘u’-shaped sources). This is a possible indication of the ν−3.0,comparedtoaradio–opticalspectrumofν−0.6–ν−1.0. presenceofexcessemissionintheNIR,inadditiontoablue Additionally, a synchrotron spectrum that has a turn- power law. We propose that this excess emission is due to over will, when combined with the blue power law, be able a synchrotron component that has turned over in the NIR, to reproduce an inflection-like SED. Such a spectrum, par- andsodoesnotdominateintheoptical,wherethedominant ticularlyintheregionoftheturn-over,willalsobequitered, emissionisinsteadabluepowerlawsimilartothatobserved thusaccounting for the red colours of many of theSEDs. in thebluest sources. 4.2.1 Analytic modelling 4 PHYSICAL MODELS We consider here synchrotron emission from a population of electrons with an energy distribution with the form of a In response to this phenomenological classification, we pro- power law up to some maximum energy and zero beyond pose the following physical model. There are two compo- this (i.e. an energy spectrum with an abrupt cutoff). This nents in this model: one is a blue power law, representing canbeexpressedasadistributionoftheLorentzfactorγ of continuumemission from theaccretion disk region; and the theradiating electrons: second is synchrotron emission, representing emission from the relativistic jet (that we know to be present due to the κγ−pdγ 1 γ γ flat-spectrum radio emission seen in these objects). N(γ)dγ ≤ ≤ c ∝(0 γ >γc ThisisconsistentwithmodellingdonebyMeisenheimer 4.1 Accretion Disk emission et al. (1996) on thejet ofM87. Theyfound that theoverall We find, from the simple power law fitting, that the bluest synchrotron spectrum of the brightest parts of the jet was sourceshavepowerlawcontinua.Theslopesofthesepower best described by a spectrum that had a sharp cutoff at lawsareconsistentwiththembeingthesamecomponentas ν 1015Hz, with an energy distribution of the form of a c that seen in optically-selected quasars, commonly termed stra≈ightpowerlawN(γ) γ−2.31,witharatherabrupthigh ∝ the “Big Blue Bump” (BBB). We take this component energy cutoff. We consider the effect of using a power law to represent the underlying quasar continuum in the UV- synchrotron spectrum instead in Section 10. optical part of the rest-frame spectrum – that is, the “un- Such a synchrotron spectrum is straightforward to reddened”quasar continuum. model analytically. We use the “classical” synchrotron (cid:13)c 0000RAS,MNRAS000,000–000 Red synchrotron jets in Parkes quasars 5 Figure2.Thesynchrotronmodelusedintheanalysis,inbothfλ andνfν units.Bothplotshavebeennormalisedsothatthepeakflux infλ is1.Notethattheverticalscaleisdifferentinthetwoplots. model, as first calculated by Schwinger (1949), and as de- We consider here a range of p values from p = 2.0 to rivedbyanumberofauthors,particularlyPacholczyk(1970) p = 3.0, which gives a range of long-wavelength power law and Longair (1994), whose derivation we follow. slopesofα = 0.5toα = 1.0.Thisrangecoversthedis- S S − − The single particle luminosity for a radiating electron tribution of radio-to-optical slopes observed in optical syn- (mass m , charge e, Lorentz factor γ = E/m c2, and pitch chrotronjets(Scarpa&Urry1999).Allowingptovarydoes e e angle,oranglebetweentrajectory andmagnetic fielddirec- not significantly alter the results of our analysis – see Sec- tion, θ) in a uniform magnetic field B is given by tion 10 for further discussion. We also note here that a value of p > 3 means that √3e3Bsinθ L1p(ω)= 8π2ǫ0cme F(x) FtheνFνν−fl(up−x1w)/i2llainncdresaoseνtFowardνs−lo(pn−g3e)r/2w)aavneldentghtihssr(essiunlctes ν ν ∝ ∝ where ω=2πν is the angular frequency, in the radio flux being severely overestimated by the fitted synchrotroncomponent,sincetheradioemissionalwayshas 2ωβm x= e , a lower νFν flux than the optical. (This assumes that the 3γ2eBsinθ samesynchrotroncomponentisresponsibleforboththeop- and F(x)is defined in terms of theBessel function K5/3(z) tical andradio emission, which is anassumption commonly by made, particularly for the modelling of optical synchrotron ∞ jets (Meisenheimer et al. 1996).) F(x)=x K (z)dz. 5/3 Zx We are interested in the luminosity of a population of 4.3 Model fitting particles,soweneedtointegrateL (ω)oversuitabledistri- 1p butionsofenergiesandpitchangles.Theenergydistribution These two components (the blue power law and the syn- is that given above, while the pitch angle distribution that chrotroncomponent)arecombinedlinearlytoformamodel weuseisanisotropicone,wheretheprobabilitydistribution f (λ) = aλ−1.7+bf (λ) that is fit to the data in the C synch is p(θ)dθ= 1sinθdθ. Thus,theintegrated luminosity from samewayasthepowerlawmodel(thatis,usingχ2 minimi- 2 such a population is sation).Thereducedχ2 value(thatis,χ2/ν)foreachofthe two models (combined and power law) are compared, and L(ω)= √3e3Bκ πsin2θ γcγ−pF(x)dγ dθ the model with the lowest χ2/ν is chosen to be the best fit 16π2ǫ cm 0 e Z0 (cid:18)Z1 (cid:19) model. An example of such a spectrum is shown in Fig. 2, for Infittingthecombinedmodel, thelocation ofthepeak γ = 104 and B = 10−4T = 1G (the value of κ has been wavelengthofthesynchrotronspectrum,λ ,wasallowedto c p takentobe1).Thepeakfrequencyν dependsonthesetwo vary. This variation was allowed to occur over a range of c values, and can be shown by simple arguments (Blandford rest frame wavelengths such thatthecurvatureofthespec- 1990) to be approximated by ν γ2B MHz (where B is trum caused by the turn-over affected the synchrotron flux c ∼ c measured in Gauss). in theregion of thedata points (in otherwords, wedid not Theslope of thepower law tail(at frequencies ν ν ) wanttojustbefittingthepowerlawpartofthesynchrotron c ≪ is related to the energy power law index by α =(p 5)/2 spectrum).Quantitatively,wetooktheminimumpeakwave- S (where f λαS). The energy distribution for the spe−ctrum lengthtobehalfadecadeshorterthantheBband(0.44µm) inFig.2i∝stakentobeN(γ) γ−2.5 (i.e.p=2.5),givinga shifted to the rest frame. We then considered 20 λ values p power law of f λ−1.25. ∝ per decade (evenly spaced in log λ ), up to a maximum λ ∝ 10 p (cid:13)c 0000RAS,MNRAS000,000–000 6 M.T.Whiting, R.L.Webster and P.J.Francis Figure 5. Distributions of the fitted power index α. The open histogramshowstheresultsofthepowerlawfittoall117sources, thehatched histogram showswhichsourceshave apower lawfit accepted at 99% confidence, while the cross-hatched histogram showswhichofthosesourceshavepowerlawfitsbetterthanthe combinedfit. While many sources have been fitted better with the Figure 4. Histograms of the reduced χ2 values, for all sources combined model, a large number are still preferentially fit andforeach ofthe best-fit models.Thedotted histogram shows with the power law. If we plot a histogram (Fig. 5) of the thosesourceswhosefitsarerejectedatthe99%confidence level. powerlawindicesofthosestillfitbythepowerlawmodel,we can see that those sources that are preferentially fit by the peak of 10µm. For each of these synchrotron functions, a power law model are the bluer sources, while the majority best fit to the data was found, and then the best of these of the sources with indices α > 1 are fit better by the − was chosen, giving thebest fit λ valuefor that source. combined model. p 5 RESULTS FOR PHYSICAL MODELS 6 FITTED SYNCHROTRON COMPONENTS Thiscombinedmodelwasfittedtothephotometry,andcom- Thepropertiesofthesynchrotroncomponentsthatarefitted paredtothepowerlawfits.Thebestfittingmodelwascho- aspartofthecombinedmodelareofparticularinterest.As senonthebasisofthelowestreducedχ2 value,asdescribed canbeseeninFig6,thepeakwavelengthsarerestrictedtoa above.Forthedefaultvaluesoftheparameters(p=2.5and relativelynarrowrangeofvalues(approximatelyadecadein α = 1.7), we find that 93 sources (or 79% of the total) wavelength).However,thisislikelytobelargelyareflection B − are well fit by one of the models. Of these, 48 are best fit of the distribution of the wavelengths of the photometric by the power law model, and 45 by the combined model. points. How these numbers change with different parameter values Thestrengthofthefittedsynchrotroncomponentvaries isdiscussedinSection10.4.Aselectionofthefitsareshown considerably from source to source. In Fig. 7, the ratio of in Fig. 3, for a range of power law slopes and synchrotron the synchrotron and power law components at a rest-frame peak wavelengths. wavelength of 0.5µm is shown for all sources best fit by the We note here that although there are 48 sources best combined model. The main bulk of this distribution spans fit by the power law model, many of these have combined nearly four ordersof magnitude. This large range of values, fitsthatareonlyslightlyworsethanthepowerlawfit.This which is also seen in the normalisations of the individual indicatesthatthereisnotagreatdealofdifferencebetween components, indicates that we are seeing a continuum of the fits of the two models. This is not the case with the variationsofthesecomponents,probablyduetovariationsin combined model sources, as for most of thesethe combined thestrengthsoftheinnerjetandemissionfromtheaccretion model fit is a lot betterthan the power law fit. disk and/or surrounding regions. Ahistogramofχ2/ν valuesisshowninFig.4,withthe Wealso notethat asmall numberof thesources at the distributions for thetwo different models shown separately. high-ratioendofthedistributionarefaint,redsources,that The distribution for the sources best fit by the combined arelikelytobesignificantlydust-reddened.Theyarethusfit model is noticeably broader than that for the power law withadominantsynchrotroncomponent,asthesynchrotron sources, with more sources havingvery low χ2 values. spectrumhas theapproximateform of apowerlaw with an (cid:13)c 0000RAS,MNRAS000,000–000 Red synchrotron jets in Parkes quasars 7 Figure3.Examplesofthevarioustypesofphotometry,andthefitstothem.Thehorizontalaxisoneachplotiswavelength,inµm,and theverticalaxisisfλ inWm−2µm−1.Thelocationsofnotableemissionlinesareshown,basedontheredshiftofeachquasar.Forthe combinedfits,theindividualcomponents–thepowerlaw(αB =−1.7)andthesynchrotroncomponent(p=2.5)–areshownasdotted lines.The λp value, where given, is for the quasar restframe. A fullcompilation of the fits to all the quasars inthe sample isgiven in Whiting(2000). exponential cut-off (which is the same as a power law with ationisitshighdegreeofpolarisation. Ifthereissignificant dust extinction). synchrotron emission at optical and NIRwavelengths, then one would expect to be able to detect a corresponding polarisation. Indeed, this has been used as a way to confirm the presence of synchrotron emission in optical jets (Baade 7 TESTING THE FITS (1956)providedthefirstexampleofthisforthejetofM87.) 7.1 Polarisation In our combined model, the synchrotron component is We have shown that a large fraction of the sources in the the only polarised component, as we assume that the BBB PHFSshowevidenceforthepresenceofopticalsynchrotron component,whichisessentially emission fromtheaccretion emission,wheretheamountofsynchrotronemissionpresent disk,isunpolarised(P <1%fortheBBB(Antonucci1988)). in the spectrum changes with wavelength. How else can we Thus,theamountofpolarisationwilldependonthepropor- testthismodel?Oneofthekeyfeaturesofsynchrotronradi- tionofthetotalfluxthatisduetothesynchrotronemission. (cid:13)c 0000RAS,MNRAS000,000–000 8 M.T.Whiting, R.L.Webster and P.J.Francis Figure 8. Polarisation from Wills et al. (1992) as a function of the proportion of the total flux made up by synchrotron. The sources are given different symbols according to the nature of theirbestfitmodel. Furthermore,iftherelativeamountofsynchrotronemission changeswithwavelength,asitdoeswiththemodelswehave fitted, then the amount of polarisation should also change with wavelength. Suchdependencieshavebeeninvestigatedpreviouslyby anumberofdifferentauthors,forsamplesthatincludesome Figure6.Histogramofpeakwavelengths,showingallsourcesfit sources considered here. Wills et al. (1992) studied a large best with the combined model, in a) the rest frame, and b) the sampleofbright,flat-spectrumcore-dominantquasars,mea- observed frame. The solid histograms show all sources accepted at 99% confidence, while the hatched histograms show the BL suringtheiropticalpolarisation.Aninterestingresultisthat Lacobjects.Thelocationsofthreeprominentemissionlinesand they found that the fraction of quasars with P > 3% in a photometricbandsaregivenforreference. fixed observed passband decreased with increasing z, possibly indicating that the percentage polarisation decreases towardsshorterrestframewavelengths.Thiswould becon- sistent with thepresenceof asynchrotron component turning overin theoptical region. Impey&Tapia(1990)presentradioandopticaldatafor a slightly larger sample of radio-selected quasars, including optical polarisation measurements. They find strong statis- ticallinksbetweenstrongopticalpolarisationandproperties such as compact radio structure, superluminal motion and weak emission lines. They explain this by requiring the optical emission, as well as the compact radio emission, to be relativistically beamed. Smithetal.(1988)obtainedmulticolour(UBVRI)polarisation measurementsof11highlypolarised quasars,and found that three of these exhibited decreasing polarisation towardshorterwavelengths,whichtheymodelled asacom- bination of polarised synchrotron emission, and two unpo- larised components, from the broad-line region and the ac- cretiondisk.Noneofthesesources,however,arepartofour sample. Figure 7. Histogram of ratios of the synchrotron component 7.1.1 Optical polarisation to the power law component of the combined model, for those sourcesbestfitbythecombinedmodel.Theratioiscalculatedat Firstly,wewishtocompareourmodelpredictionswithpub- arest-framewavelength of0.5µm.Thehatched histogramshows lished optical polarisation measurements. We use the large theBLLacobjects. catalogue of measurements compiled by Wills et al. (1992). Wewanttocomparethesemeasurementswiththefractional (cid:13)c 0000RAS,MNRAS000,000–000 Red synchrotron jets in Parkes quasars 9 amount of flux due to the fitted synchrotron component. ness,wealsoshowalineofconstantpolarisation,normalised However, since the observations of Wills et al. were made totheK-band point.All thesesources are best fit with the withoutafilter,wehavedeterminedtheaveragesynchrotron combined model, with the exception of 1101 325, which is − fraction byintegrating over therange 0.3µm – 1µm. fit best by the power law model (its SED taking the form WehavethenplottedinFig.8thepercentagepolarisa- of a blue power law, λ−1.81). We have used the combined tion as a function of this average synchrotron fraction. For model fit to it for the purposesof Fig. 9. those sources that were best fit by the power law model, Wehavealso included polarisation measurementsfrom we have calculated the fraction from the combined model Wills et al. (1992) where they exist. Again, these are put fit, and indicated these sources by a different symbol. All at an observed wavelength of 0.5µm. Note, of course, that the power law sources with polarisation measurements had thesepoints are not simultaneous with the NIRpoints. relatively red power law indices (i.e. α> 1.3). The fitted synchrotron component generally replicates − The spread in polarisation measurements at large syn- well the wavelength dependence of the polarisation, al- chrotronfractions ismuchgreaterthanat low fractions, in- though for some sources, such as 1313 333, the points are − dicating that the high polarisation sources generally have equally well given by a constant polarisation component. large amounts of synchrotron fitted to them (at the wave- Thisiswhatyouwouldexpectfromapuresynchrotroncom- lengthsatwhichthepolarisationismeasured).Additionally, ponent,andthesesourcesaretypicallyBLLacobjects,from allbutoneofthesourcesatzero(ornear-zero)synchrotron which you would expect to see a synchrotron-dominated fraction havelow optical polarisation. SED.Anotableexceptiontothisis0537 441,whosepolar- − The two exceptions to this picture are 0202 172 and isation isnotfitwellbyaconstantsynchrotroncomponent, − 1020 103. Firstly, 1020 103 has a very high synchrotron but is fit better by a combination of a synchrotron and a − − fraction ( 95%), but has very little optical polarisation significant power law component. However, it is apparent ∼ (P =0.58%). In this case, the source has a power law con- from these plots that having simultaneous optical polarisa- tinuumwithaslightcurvature,whichisfitwellbyanalmost tion measurements would better help discriminate between puresynchrotroncurve.Itmaybethatthiscurvatureisdue thecombined model and a constant polarisation model. toothereffectsthansynchrotron,whichexplainsthelackof polarisation.Apossiblecandidateiscontaminationfromthe 7.2 Emission lines very strong Hαline, which would boost thecontinuum level in thecentreof theSED. Synchrotron flux emitted from the PHFS quasars is most Secondly,0202 172hasameasuredpolarisationof5.15, likelytocomefromathin,relativisticjet,andtheradiation − butalso hasaverybluepowerlaw continuum(α= 1.85). will not be isotropic. This means that there will be very − This continuum slope is possibly too steep to attribute to little synchrotron radiation directed towards the emission synchrotron(itimpliesavalueofp=1.3–inturnimplying lineclouds, which havea muchlarger covering angle. Thus, a rather flat energy distribution ). However, the location of the ionisation of these clouds will be due to the continuum the synchrotron peak may have shifted in the period since emission from the central accretion disk region – in other the polarisation measurements were made (thepolarisation words, the Big Blue Bump. measurements were taken 8 years prior to our photometry One might expect that adding a non-ionising syn- observations). chrotron component to the continuum will have the effect of reducing the equivalent width of the emission lines from theBLR, dueto the fact that theflux in theemission lines 7.1.2 Near-infrared polarisation is not changed but thecontinuum fluxis increased. Totrytoavoidtheproblemofnon-simultaneityofthepho- To test this prediction, we compared the equivalent tometry and polarisation observations, we obtained polari- widths of five emission lines (Civ 1549, Ciii] 1909, Mgii sationmeasurementsof8quasarsintheNIR,usingtheIRIS 2798, Hβ, and the doublet [Oiii] 4959,5007) with the ratio instrument on the Anglo Australian Telescope. The details of synchrotron to continuum fluxat theline wavelength, to and results of the observations will be presented elsewhere see if some form of an anti-correlation is present. The de- (Whiting et al., in preparation). These measurements are tailsoftheobservationswillpresentedelsewhere(Franciset nearly simultaneous with the photometry measurements (a al. 2001). Objects that had spectra taken were essentially differenceof 40days),andsocanbedirectlyrelatedtothe a random sample of the PHFS (subject to visibility during ∼ fittedcomponents.Thewavelengthdependenceofthepolar- theobserving run). isationcanthenprovideanimportanttestonthemodelswe InFig.10,weshowtheresultsofthiscomparisonforthe are fitting. Mgii and Hβ lines, which are the two broad lines with the Since the synchrotron component is the only polarised longest wavelength (and hence the two lines most likely to component, one would expect the percentage polarisation showareductioninequivalentwidth).Wehavealsoplotted to be directly related to the amount of synchrotron flux those sources best fit with the power law model. The value present,and,in fact, thepercentagepolarisation will bedi- oftheratiousedforthesesourceswastakenfromfittingthe rectly proportional to the ratio of synchrotron flux to total combined model, and so are upperlimits to theratio. flux. In Fig. 9, we have plotted the polarisation of each of The Mgii line does not show much relationship to the thesequasarsasafunctionofwavelength,and,onthesame synchrotron ratio, while the Hβ line does show a reduction plot, the synchrotron ratio normalised (arbitrarily, as it is in equivalent width with increasing amount of synchrotron. thewavelengthdependenceweareinterestedin,notthepre- This lends some support to the hypothesis that excess syn- cise normalisation) to the longest wavelength polarisation chrotron light is present. The difference in the two plots is point (which is usually the K-band point). For complete- likelyduetothepresenceoftheturnoverinthesynchrotron (cid:13)c 0000RAS,MNRAS000,000–000 10 M.T.Whiting, R.L.Webster and P.J.Francis Figure 9. Near-infraredpolarisation of a selection of quasars as a function of wavelength. Alsoshown are optical measurements from Wills et al. (1992). The solidlines show the fraction of the total flux made up by the synchrotron component (right-hand axis), while thedashedlineshowsthisnormalised(seetext) totheK-banddatapoint,andisinpolarisationunits(left-handaxis).Thedottedline showsconstantpolarisation,normalisedtotheK-banddatapointaswell.A∗ denotes theobjectisaBLLac. flux, so that is has less effect at the shorter wavelength of other model be used instead of synchrotron? One possible Mgii. There are, however, several sources that have high alternative is blackbody emission resulting from hot dust. synchrotronratiostogetherwithhighequivalentwidths.The Totestthismodel,weusedablackbodyemission spec- implications of theseresults are discussed in Section 10. trumduetodust at a temperatureof 1750 K(thesublima- tion temperature characteristic of dust grains consisting of graphite and silicates, e.g. Laor & Draine (1993)) emitted in the quasars’ rest frame. A blackbody curve at this tem- 8 HOT DUST: AN ALTERNATIVE TO perature would have its peak, in the quasar rest-frame, at SYNCHROTRON? 1.66µm. This blackbody spectrum was combined with the So far, we have shown that, for a number of sources, the same λ−1.7 power law used in the combined model to pro- optical–NIR photometry is well fit with a power law plus ducea model that was fit to thedata. a curved component, which we have assumed to be the The fits generated by this model were almost always turnover of a synchrotron component. However could an- worsethanthoseofthesynchrotronmodel.Thiswasdueto (cid:13)c 0000RAS,MNRAS000,000–000