Probing Accretion Disk Winds in AGN I. Asymmetric broad Balmer emission lines 4 1 Cosmos DUMBA 0 2 Department of Physics, Faculty of Science p e Mbarara University of Science & Technology S 9 P. O. Box 1410, Mbarara, Uganda 1 ] September 23, 2014 A G . h p The Broad Line Region of Active Galactic Nuclei is characterized by broad - o Balmer emission lines in their optical spectra. The broad Balmer emission lines r t are found to be asymmetric, some blue sided and others red sided in their asym- s a metry. One of the components behind the asymmetry is thought to be an accretion [ disk wind. We probe the accretion disk wind using the broad balmer emission line 1 profiles. v 4 7 1 This asymmetry of the broad balma emission line profiles is measured in velocity 6 space after a measurement of the line shift at percentiles from 0, in increaments of . 9 10, up to 90. In addition, the Kurtosis Index is obtained at appropriate points of the 0 emission lines’ profiles. This study is based on many hundreds of SDSS spectra, 4 1 starting with low redshift high signal to noise ratio spectra. We also consider a def- : v inite number in each bin of their FWHM, in bins of 1000 km/s (atleast 40 per bin), i X startingfrom1000km/stotheverybroademissionlines. r a We present how strong the asymmetry (by plotting Asymmetry Index as a function of percentile) of the broad/narrow lines (in percent) is, what the Kurtosis (R20,80) is. WealsopresentwhattheAsymmetryIndexasafunctionoflinewidth(FWHM), luminosity(V-band),core-radiofluxandIonizationDegree. AGN: Accretion Disk, Accretion Disk Wind – Line: Asymmetry, Balmer, Emission, Profiles 1 1 Introduction the nucleus). This limit defines a charac- teristic luminosity, the Eddington luminosity Active Galaxies have been widely studied (L )throughtherelations; Edd by many authors revealing many fascinating F = F (5) properties which among all include; compact grav rad nuclear emission (Clavel et al. 1990), non- GM(m +m ) GMm σ L p e p T Edd thermalcontinuumemission ≈ = (6) r2 r2 4πcr2 4πGm c F ∼ ν−α (1) L = p M (7) ν Edd σ T (Bregman1990). M L = 1.26×1031 [W] (8) Edd In addition, authors notice that; the con- M sun tinuum stretches from the radio to X- TheimportantparameterinAGNis L . ray/Gamma rays (Mehdipour et al. 2011), LEdd One of the consequences of high L are the luminosity of the nucleus exceeds that LEdd winds,AccretionDiskWinds. JustlikeinSo- of the host (Osterbrock 1989), they have larWinds....onceaparticleexceedstheescape strongemissionlines(DeBreucketal.2000), velocity, we see this as a wind from the ac- are highly variable (Ulrich et al. 1997), and cretion disk. Accretion disk winds have been also have X-ray emission (Turner & Pounds confirmed by some authors by studying BAL 1989). Quasars(Hamann1998). Hamann(1998)ob- The energy source in the AGN is believed served strong absorption trouphs in the rest- to be accretion (Blandford & Znajek 1977). frame UV spectrum and estimated outflows Thiscanbedemonstratedthroughtherelation with10,000km/s. : In this work, we study the accretion disk E = ηmc2 (2) whereη istheefficiency. Theluminositycanthenbere-writtenas: dE dM L = = η c2 = ηM˙ c2 (3) dt dt whereM˙ = dM istheaccretionrate. dt ForatypicalAGN; L L M˙ = ≈ 1.8×10−3 37M yr−1 (4) ηc2 η sun Figure1:The rest frame of the UV spectrum where L37 is the luminosity in units of 1037 ofaBALQuasarshowingasignifi- W. cantoutflowintheabsorptionlines This energy, at one point, through the princi- ofupto10,000km/s pleofhydrostaticequilibrium,reachesapoint Hamann(1998) at which gravitational forces (causing the ac- cretion) balance with radiative forces (from winds by analysing broad Balmer line profile 2 asymmetryandsteepnessvalues. of obtaining quality spectra in all ranges of We define the two parameters in line shape FWHM from 1000 km/s to above 10,000 analysisas: Km/s. This is so because as we are selecting thespectrawiththehighestsignaltonoisera- S A.I = (9) tio, there is a tendency to obtain only spectra IPV thatarebothfromsourcesnear(lowredshift), IPV thusneglectingthosewithhigherredshift,and U K.I = (10) IPV also biasing the data primarily on signal to L noise ratio. This is avoided when we split the where S is the line shift, IPV is the inter- sampleintwosamples,oneforthosebetween percentile velocity, IPV and IPV are the U L 1000 km/s FWHM and 3000 km/s FWHM, interpercentile velocities for the upper and and another for those having FWHM greater lower parts of the emission line profile. The than 3000 km/s. We end up having spectra following sections will outline what we did with very high signal to noise ratio across inmoredetailandexplainthemostimportant the whole spectrum of broad Balmer emis- stepswecarriedoutduringthestudy. sionlines. RememberbroadBalmeremission linesarethosewithFWHMgreaterthan1000 2 Data km/s. In this section, we explain how we download 2.1 Introduction to Data the data, how we treat the data, describing the software we use and all the tasks used. Analysis We also explain all the steps we carry out, In order to study any sample of galaxies and giving details of the output in each step and quasars,itisimportantnottoforgetafewuse- why it is carried out. In addition we show ful conditions that make the sample produce samplesofthevaluesextractedfromtheplots reliableinformation: (thewholelistbeingfoundintheappendices). Lastlywedescribethesecondarytreatmentof • thesizeofthesampleand the extracted data from the plots, explaining how and why we carried out the treatment in • the quality of the spectra, both in terms such ways and show samples of the results of resolution and Signal to Noise ra- in tables (the actual results being shown in tio.(Whittle1985a) the next chapter). But then, it is necessary to Our data meets both criteria thanks to dedi- first briefly describe the database (SDSS) and cated surveys like the SDSS that have made thereafter the telescope and the surveys it has such homogeneous data sets available to the carried out so that one gets a feeling of the public. whole process from observing, collection of Wehaveobtained,ineachsample,thetop600 data,treatmentofdataandanalysis. highsignaltonoisegalaxiesandquasarsfrom the seventh data release (DR7) (York et al. 2.1.1 SDSS Database 2000). The samples have the highest spectro- scopic quality in the release (Abazajian et al. The Sloan Digital Sky Survey consists of 2009),andareuniformintermsofcalibration three major surveys that together provide sci- letalonebeingcomplete. entists with immense volumes of data ob- We have two samples for the same reasons tained from a dedicated 2.5m telescope lo- 3 cated at Apache Point Observatory in South- Encodedwithinthespectraldataarethecom- ern New Mexico. This survey has been col- position and temperature of these stars, vi- lecting data since the year 2000 (Abazajian tal clues for determining the age and origin et al. 2009; York et al. 2000). The three sur- of different populations of stars within the veysare; Galaxy(Yannetal.2009). • Legacy 2.1.4 The SDSS Supernova Survey • SEGUE The SDSS Supernova Survey was one of threecomponents(alongwiththeLegacyand • Supernova SEGUE surveys) of SDSS-II, a 3-year ex- tension of the original SDSS that operated 2.1.2 SDSS Legacy Survey from July 2005 to July 2008. The Super- The SDSS Legacy Survey provided a uni- nova Survey was a time-domain survey, in- form, well-calibrated map in ugriz of more volving repeat imaging of the same region than7,500squaredegreesoftheNorthGalac- of sky every other night, weather permitting. tic Cap, and three stripes in the South Galac- The primary scientific motivation was to de- tic Cap totaling 740 square degrees. The cen- tectandmeasurelightcurvesforseveralhun- tral stripe in the South Galactic Gap, Stripe dred supernovae through repeat scans of the 82, was scanned multiple times to enable a SDSS Southern equatorial stripe 82 (about deep co-addition of the data and to enable 2.5◦ wide by 120◦ long) (Frieman et al. discovery of variable objects. Legacy data 2008). Theabovethreesurveyshaveprovided supported studies ranging from asteroids and scientists with a catalog derived from the im- nearby stars to the large-scale structure of the ages obtained by the 2.5m telescope. These universe. Almost all of these data were ob- images include more than 350 million celes- tainedinSDSS-I,butasmallpartofthefoot- tial objects, and spectra of 930,000 galaxies, printwasfinishedinSDSS-II. 120,000quasars,and460,000stars. Thisdata is not only fully calibrated and reduced, care- fully checked for quality, and publicly acces- 2.1.3 SEGUE - Sloan Extension for sible through efficient databases, but has also Galactic Understanding and been been publicly released in a series of an- Exploration nual data releases. It is through this effort SEGUE was designed to explore the struc- that we are able to carry out this study on the ture; formation history; kinematics; dynam- asymmetry of the broad emission lines of ac- ical evolution; chemical evolution; dark mat- tivegalacticnuclei. terdistributionoftheMilkyWay. Theimages In this Data release, this study exploits the and spectra obtained by SEGUE allowed as- immense volume of spectra of galaxies and tronomers to map the positions and velocities quasars (Abazajian et al. 2009; York et al. of hundreds of thousands of stars, from faint, 2000). The sample obtained from the release relativelynear-by(withinabout100parsecor includes around 600 spectra of both galaxies roughly 300 light-years) ancient stellar em- and quasars in the redshift range of 0 and 1, bers known as white dwarfs to bright stellar withbroadHβ linesfrom 1500km/senabling giants located in the outer reaches of the stel- us contain all groups of broad line emission lar halo, more than 100,000 light-years away. objects.SDSS(2013) 4 Figure2:The2.5mtelescopelocatedatApachePointObservatoryinSouthernNewMexico BBC(2013) 2.1.5 Obtaining the Data Inaddition,thisdataconsistsofaround283 spectra with measured Hβ profiles around In order to obtain data from the SDSS 165 measured Hα profiles. This is because database with your own constraints on the not all spectra could have both the Hβ emis- sample,oneneedstowriteanSQLQuerythat sion line and Hα emission line due to red- generates a list of the sources that meets your shifting at both ends of the optical window. requests. However, most of the spectra with Hβ lines In this study, we focus on the parameters also have Hα lines but not vice versa. Of of the Hα & Hβ emission line, looking for course spectra that are at the extreme end of asymmetryintheselines. our redshift selection are the ones affected with not having the Hα line visible, but this 2.1.6 Data Properties and Constraints is not an issue for us to worry about since we Invoked weremoreinterestedintheHβ lineprofiles. Figures 3 show the distribution of the sig- The data in our samples combined consists nal to noise ratio of our sample and the red- of around 300 objects, galaxies and quasars, shift. It is seen here in the signal to noise restricted to a redshift range between 0 and distribution that we have a fairly good sig- 1. The samples consist of reduced spectra of nal to noise ratio, the minimum being 37 and these objects in fits files that we renamed in maximum 79. The distribution is binned in 5 ascending order from those with the highest starting from 35. The figure therefore clearly signal to noise ratio. This helps us analyze shows that most of our sample spectra have those with the highest signal to noise ratio a signal to noise ratio between 40 and 45, to first. Thisisimportantbecausetheresultsob- bespecificaround120outof300(40%),with tained from high signal to noise ratio sources a few having more than that. This is not a are more reliable for the derivation empiri- problem since, given the nature of our study, cal relations. The redshift limit was chosen asignaltonoiseratioofeven30wouldbesuf- so that we have a good coverage of the Hβ ficient for good measurements (Thorne et al. line and its adjacent spectral regions. Figure 1999). presents the general properties of the sample in terms of redshift, signal to noise ratio and In the same way, the redshift distribution thebroadeningoftheHβline. is shown. Because we chose spectra with the 5 Figure3:SignaltoNoiseratio&RedshiftdistributionofSample highest signal to noise ratio, it seemed obvi- 9000 km/s, there isn’t any increase in counts ous that most of the sample spectra will be as there are very few AGN with such ex- in the near end of the redshift window we tremely broad lines. Our study will obtain in- chose with those below redshift 0.2 domi- formationaboutthebroadlinesbasingalmost nating. However, since the distribution does entirelyonthefirstsixorsevenbinswithsuf- not fall rapidly as we move to higher red- ficientnumbers. shift, the data will provide a sufficient study The distribution of the FWHM in our sam- of AGN within this redshift bin chosen. In ple is in agreement with the AGN statistics a later study, it would be good to also study obtainedbyHaoetal.(2005)whentheyplot- other redshift ranges, possibly using Ultravi- ted the distribution of the FWHM values of olet emission lines which can be obtained in the Hα emission line for over 40,000 emis- the optical spectra after they have been red sion line galaxies. Their distribution was bi- shifted. modal since they included all AGN, narrow Figures 4 show the distribution of the and broad line AGN. Their boarder line of FWHM of all the Hβ emission lines mea- broadlineAGNwasnaturallyplacedatthose sured. The distribution shows that there is with a FWHM of over 1200 km/s. To com- a peak between 3000 km/s and 4000 km/s. pare with our distribution, we only consider But still it is not a big range from the bins thoseover1000km/s,anditgivesusthesame at both ends which are at 50 each. Our inten- distribution. Intheirstudy,definingbroadline tion is to have a good distribution across the AGN as objects with a FWHM greater than whole range from 1000 km/s to over 10,000 1200 km/s, they obtained 1,317 objects out km/s, and since we obtain enough counts for of 42,435 emission line galaxies. This makes the first five bins, then the sample will pro- our sample of 300 objects not bad since it is vide a statistically sufficient analysis for our a quarter of this value, let alone having been findings. restrictedtoaredshiftbetween0and1. The other figure here is a cumulative dis- Following this simple look at the gen- tribution, which shows that from FWHM of eral properties of the downloaded data, it is 6 Figure4:FullWidthatHalfMaximumdistributionofSample now necessary to explain the treatment of the information i needed from the broad Balmer data in order to obtain measurements that are lines in my sample. This was principally the needed for our later analysis. The detailed partofthethesisthatoccupiedmemostsince spectral properties of the sample will follow itinvolvedcarefulvisualanalysisofthespec- laterintheforthcomingchapters. trum first and accurate execution of the re- quired commands for which a mistake with oneofthemmeansredoingthewholeprocess 2.2 Image Plotting and from the beginning. Explained below are the Analysis with IRAF stepsitookduringmydataextraction. IRAF (an acronym for Image Reduction and Analysis Facility) is a collection of software 2.2.1 Spectrum Visual Analysis written at the National Optical Astronomy After downloading the data, (the fits files), Observatory (NOAO) geared towards the re- and having renamed them, starting from the duction of astronomical images in pixel array one having the highest signal to noise ra- form. In this thesis, i used IRAF to plot spec- tio, i plotted each spectrum and analyzed the trafromoursampleAGNandanalyzedafew Balmer emission lines. The reason why i spectral lines needed to derive more infor- needed to do this is because each spectrum is mation for analysis later (Tody 1986; Valdes unique and the tasks to perform on each var- 1986). To be specific, we plotted the Balmer ied from one spectrum to another. Some of lines, Hα and Hβ emission lines, although thethingsilookedoutforwere; not all spectra contained both emission lines. Most of them had the Hβ emission line, a • The availability of both Hβ and Hα good number had both Hβ and Hα emission emissionlines. line, while very few(∼2%) had only the Hα emissionline. • Iron emission around the Hβ emission In the following subsections, i will explain line,theFeEmission. brieflythestepsandtasksiusedtoextractthe This was important to note because a lot 7 of Fe emission could provide a poor es- 1000 km/s and 3000 km/s, and sample 2 be- timate for the continuum measurements. ingthatofAGNwithFWHMfrom3000km/s Iftheemissionwasonlyoneithersideof and above, this is what we observed from the theBalmerline,iwouldusetheendwith visualanalysis; lessemissionasmyproxyforthecontin- uum. Incaseswheretherewastoomuch • TheAGNinsample1havebothHαand Fe emission, i excluded such a spectrum Hβ,meaningtheyarefoundinthelower from my sample since it meant an extra part of the redshift range from 0 to 1, process of using a template to subtract preferably less than z = 0.6. There are theFeemissionfirst. less than 10 out of 81 that had only the Hβ emissionline. • Availability of neighboring narrow lines ([OIII]forHβ &[SII]forHα). • Sample 2 has at least 40% of the AGN having both Hα and Hβ. This means Thiswasforuseinestimatinganequiva- 60% ofthem are foundin the higherend lent FWHM of the narrow component to of the redshift bin, preferably above z = subtract from the Balmer line. However, for the case of the Hα emission line, 0.7. Recallthatmyredshiftrangeisfrom z=0toz=1. the [SII] emission doublet was not accu- rate enough because the two lines were blended together in most cases. It meant 2.2.2 Continuum Subtraction usingthe[NII]linesthatwerewithinthe broadened Hα emission. This meant a This is the task done after visual analysis of visualestimationonceboth[SII]&[NII] the spectrum. Spectra with both the Hα and lines were unavailable. On a positive Hβ emission lines look similar to the ones in note, there were very few cases where i figure 5 although some with significant red- couldnotuseanyofthetwo. shiftwillonlyhavetheHβ emissionlinestill available. The subtraction of the continuum • Thefunctiontousewhilesubtractingthe isdoneinsplot,withthekeys”t”followedby continuum. a ”-”, followed with the appropriate function This was important because the unique- necessaryfortheselectedregion. Thecontin- ness of each spectrum, and the quest to uum is subtracted for each emission line sep- apply a good estimate to a flat spectrum, arately and separate images are saved out of entails using different functions for each theoriginalimagehavingbothemissionlines. spectrum in order to obtain a flat spec- Thiscanbeseeninthesampleplotsinfigure6 trum from the power law spectra down- on page 9. Continuum subtraction is the sec- loaded. ond step to spectrum analysis for the Balmer emission line features and it sets a vantage All the above mentioned reasons vary from point for line normalization, as we shall dis- spectrumtospectrumandeachofthemisim- cussinthenextsubsection. portant in order for me to obtain the best re- sults from the spectrum, and for minimizing 2.2.3 Line Normalization errorsasmuchasican. In our data, which has two samples, sample To normalize the emission line simply means 1 being that of AGN with FWHM between to have its base at a zero point and its peak 8 Figure5:Seenabovearetwosamplesofthespectraduringvisualanalysis. Itshouldbenoted thatnotallspectracontaintheHαandHβ emissionlinesasseenhere,reasonbeing that those red-shifted significantly will eliminate the Hα since it will be Doppler shifted to wavelengths outside the optical range. In addition, not all spectra are relativelyflatliketheonesshownduetothevaryingamountsofcontinuumemission atdifferentwavelengthsformanyAGN Figure6:The images show an example of extracted Balmer emission lines after continuum subtraction on each of them. One should note here that all the extracted emission lines have their continuum at zero, making it the starting point of the percentile rangesweneedexaminelater. 9 at unity. It should be noted that we are nor- threat to any measurements needed from the malizing the broad Balmer lines. This means Hα emission line. This can be seen in the thatweeithersubtractthenarrowcomponents examplesbelow. or simply leave them visible, but ignore their additional height as we simply place a unity However, for the Hβ emission line, there valueatthepeakofthebroadline. is an undeniable problem. The neighboring Inthefigure7,examplesonthediversityof emission lines are significant enough to pose broad and narrow components in our sample athreattomeasurementstakenfromit. These with some not having narrow components at emission lines are; the Oxygen III lines and all and others dominated by the narrow com- the Iron emission at both sides of the Hβ ponentareshown. emission line. The Iron emission will in- To perform the normalization of the broad crease on the error in the continuum subtrac- Balmer emission lines, we use the arithmetic tion since it creates a pseudo continuum and task,”imarith”,whichcanbeusedanytimein thus making it difficult to estimate a contin- IRAF. However, after continuum subtraction, uumlevel. Toreduceontheerrorinthismea- thecountsatthepeakofthebroadBalmerline surement, one has to subtract the iron emis- arerecordedinatable. Alsointhesametable sion, preferably using some already devel- isrecordedthevelocityatwhichtheemission oped templates before the actual continuum line is centered. This velocity will be of use levelcanbeestimated. while transforming to velocity space, center- In our case, since it was a small fraction ingtheemissionlinepeakatthatvelocity. of spectra that had significant iron emission Table 1 shows part of the table. The whole on both sides, we neglected this since it af- tablecanbeviewedintheappendicessection. fects less than 5% of the data. The Oxy- It shows the Spectrum Number, the Balmer gen III emission lines also increase on the line heights used to normalize the profile to uncertainty of the FWHM values measured unity and the corresponding wavelengths at since they are embedded in the broad Balmer which they are identified. These wavelengths Hβ emission lines. The most reliable way to are the wavelengths at which we center the deal with this uncertainty is to subtract them correspondinglineswhiletransformingtove- before any values are measured. However, locityspace. since they are also broadened at their bases, their subtraction, using a Gaussian model, in- creases absorption features in the residue Hβ 2.2.4 Line Correction : Filtering out emission line left. Thus, since the emission other lines and Smoothening lines are already broad, we decided to simply out the noise cutthemoutinordertoleaveasmootherpro- Inanyspectrum,thereisalwaysaprobability file for the residue for better measurements that the emission line you are interested in is without introducing a smoothing factor. It is surrounded by other emission or absorption noticed that subtracting the OIII lines intro- lines. For some emission lines, the neigh- duces absorption features which in turn adds boring emission lines or absorption lines are anuncertaintytothemeasurements,sosimply not a threat to any measurements pertaining cutting them off and preventing this uncer- the emission line under study, for example, tainty seemed better because the uncertainty for the Hα emission line, the neighboring we deal with by not doing the subtraction it- Sulphur II emission lines do not pose a selfismuchless. 10
Description: