A VLA Survey of Radio-Selected SDSS Broad Absorption Line Quasars
Michael A. DiPompeo, Michael S. Brotherton, Carlos De Breuck, Sally Laurent-Muehleisen
aa r X i v : . [ a s t r o - ph . C O ] N ov A VLA Survey of Radio-Selected SDSS Broad Absorption Line Quasars
M. A. DiPompeo , M. S. Brotherton , C. De Breuck , Sally Laurent-Muehleisen ABSTRACT
We have built a sample of 74 radio-selected broad absorption line quasars fromthe Sloan Digital Sky Survey Data Release 5 (SDSS DR5) and Faint Images of theRadio Sky at Twenty Centimeters (FIRST), along with a well matched sample of 74unabsorbed “normal” quasars. The sources have been observed with the NRAO VeryLarge Array/Expanded Very Large Array at 8.4 GHz (3.5 cm) and 4.9 GHz (6 cm). Allsources have additional archival 1.4 GHz (21 cm) data. Here we present the measuredradio fluxes, spectral indices, and our initial findings. The percentage of BAL quasarswith extended structure (on the order of 10%) in our sample is similar to previous studiesat similar resolutions, suggesting that BAL quasars are indeed generally compact, atleast at arsecond resolutions. The majority of sources do not appear to be significantlyvariable at 1.4 GHz, but we find two previously unidentified BAL quasars that may fitinto the “polar” BAL category. We also identify a significant favoring of steeper radiospectral index for BAL compared to non-BAL quasars. This difference is apparent forseveral different measures of the spectral index, and persists even when restricting thesamples to only include compact objects. Because radio spectral index is a statisticalindicator of viewing angle for large samples, these results suggest that BAL quasarsdo have a range of orientations but are more often observed farther from the jet axiscompared to normal quasars.
Subject headings: quasars: general, quasars: radio
1. INTRODUCTION
Radio observations of broad absorption line (BAL) quasars have been interpreted to indicatethe inability of simple orientation schemes to explain the observed incidence of BALs in quasarspectra. The simplest orientation models suggest that all quasars have outflows, but only approx-imately 20% (the observed fraction of optically selected quasars which exhibit BALs, e.g. Kniggeet al. 2008) are seen from a line of sight that intercept these outflows (Weymann et al 1991). Inthis model BAL winds are radiatively accelerated from the edge of the dusty torus surrounding University of Wyoming, Dept. of Physics and Astronomy 3905, 1000 E. University, Laramie, WY 82071, USA European Southern Observatory, Karl Schwarzschild Strasse 2, 85748 Garching bei M¨unchen, Germany Illinois Institute of Technology, 3101 South Dearborn St., Chicago, IL 60616, USA α , defined as f ∝ ν α ,where f is the radio flux and ν is the frequency), suggesting a full range of orientations (Beckeret al. 2000). A significant difference between the distribution of α for BAL and non-BAL quasarshas not been previously identified, but the sample sizes have typically been small (Fine et al.2011). Montenegro-Montes et al. (2008) did identify a difference in the α distributions of theirsamples- however when they restricted their analysis to include only compact sources the differencedisappeared. Additionally, BAL quasars with polar rather than equatorial outflows have likely beenidentified via short timescale variation in radio flux (Zhou et al. 2006; Ghosh & Punsly 2007).The existence of apparently polar BALs and the fact that BAL quasars are often compact inradio maps suggests a different picture. Becker et al. (2000) find that only about 10% of BALquasars show extended structure at 5 ′′ resolution, compared to about 50% of normal quasars. In asample of 15 BAL quasars, Montenegro-Montes et al. (2008) find that all of them are compact atFIRST resolutions, and the majority remain compact at around 80 mas resolution. Even very longbaseline interferometry (VLBI) observations, which has down to milli-arcsecond resolution, oftenshow BAL quasars as compact objects (Doi et al. 2009, Jiang & Wang 2008, Kunert-Bajraszewskaet al. 2009). The fact that it is required to observe many of these objects on size scales on theorder of a few hundred parsecs or less before seeing any resolved structure (Kunert-Bajraszewskaet al. 2010, Liu et al. 2008) suggests that they are intrinsically quite small. This size scale alongwith a convex spectral shape (which has been seen in many BAL quasars; e.g. Montenegro-Monteset al. 2008) is typical of the compact steep spectrum (CSS) and gigahertz-peaked spectrum (GPS)sources. Objects of this type are thought to be young radio sources (O’Dea 1998, Fanti et al 1995).Instead of having a preferred line of sight, maybe there is a BAL phase in all quasars which lasts forapproximately 20% of the quasar lifetime. A model where a BAL phase evolves into a radio-loudphase, with a relatively short overlap has been proposed (Gregg et al. 2002, 2006).The studies of radio-selected BAL quasars have thus far been with small samples. In order 3 –to find more concrete results a large, statistically significant sample with radio data at multiplefrequencies is needed. This project aims at nailing down the radio properties of BAL quasarsconclusively to finally find a consistent model for these important objects. Here we will present theresults of our NRAO Very Large Array/Expanded Very Large Array (VLA/EVLA) observations.We adopt the cosmology of Spergel et al. (2007) for all calculated properties, with H = 71km s − Mpc − , Ω M = 0 .
27 and Ω Λ = 0 .
2. TARGETS
In order to build a large sample, we started with the BAL quasar catalog of Gibson et al.(2009), which is drawn from the Sloan Digital Sky Survey Data Release 5 (SDSS DR5). Gibson etal. (2009) use both the traditional definition of “balnicity index” ( BI ; Weymann et al. 1991) anda modified BI ( BI , which integrates absorption starting from 0 km s − instead of the traditional − − ) in the lines of C IV , Si IV , Al III , and Mg II to identify BAL quasars. They alsorequire absorption to be continuous over a range of 2000 km s − (see their paper for additionaldetails).In order to pick out radio bright sources, we matched the optical positions from the Gibson etal. (2009) sources to both the FIRST and NVSS catalogs, selecting sources that matched within10 ′′ of FIRST radio positions and had integrated fluxes in either survey of more than 10 mJy. Weindependently matched to both catalogs due to the differences in resolution of each; FIRST has aresolution of 5 ′′ , while NVSS is much lower at 45 ′′ . We wanted the higher sensitivity of FIRST,but only matching with the FIRST catalog may miss objects with faint core emission but brightextended lobe emission. Also matching to NVSS allowed us to identify possible extended sources,investigate the FIRST maps to verify that the sources were indeed extended, and add up the fluxin individual FIRST components to get a total source integrated flux. Thus we put together aBAL quasar sample which had ≥
10 mJy FIRST fluxes, including those with extended, individuallyresolved components. As discussed below, many observations were to be carried out in the VLAD-array, which has the smallest baseline and thus lowest resolution. Therefore we inspected theFIRST maps of all sources by eye to eliminate any in which contamination of nearby sources maybe an issue.We also applied a redshift cut of z ≥ . IV emission line in thespectral window of SDSS. Each spectrum remaining at this point was inspected by eye to ensurethat all objects were indeed BAL quasars. Gibson et al. (2009) also examined their spectra by eye,but we wanted to make sure we agreed that all sources were unambiguously BAL quasars; while nostrict signal-to-noise cut was applied, some of the SDSS spectra were too noisy for us agree with The National Radio Astronomy Observatory is a facility of the National Science Foundation operated undercooperative agreement by Associated Universities, Inc. B magnitude,and (7) is the FIRST integrated 1.4 GHz flux (sum of individual components if necessary). Theremaining columns will be discussed below.In order to make meaningful comparisons with the parent populations of quasars, a wellmatched sample of normal, unabsorbed quasars was developed. For each of the individual BALquasars in our sample, we searched the SDSS database for a normal quasar that matched within20% SDSS i-band magnitude, 20% of 1.4 GHz radio flux, and 10% of redshift. Each quasar spectrumwas examined to ensure they were unabsorbed, and the FIRST maps were inspected for extendedstructure and possible contaminating sources. Thus, a one-to-one matched sample of 74 non-BALquasars was built for use as a comparison. Table 2 lists the properties of the non-BAL sample;columns are the same as Table 1 with the subtype column omitted, since all are normal quasars.The first 3 panels of Figure 1 show a comparison of the properties that the samples werematched on. The final panel of this figure shows the K-corrected rest-frame 4.9 GHz luminositydistributions of the samples, to illustrate that the sample matching in flux and redshift did indeedresult in a matched luminosity. The luminosities shown in this figure used the integrated FIRSTflux at 1.4 GHz and radio spectral index between 1.4 and 4.9 GHz (as discussed below) for theK-correction.
3. OBSERVATIONS & MEASUREMENTS3.1. Observations
Observations were performed over two observing periods, all with dynamically scheduled time,at two frequencies: 8.4 GHz (3.5 cm, X-band) and 4.9 GHz (6 cm, C-band). The ultimate aimwas to measure the radio spectral index for all sources. Individual objects were observed at bothfrequencies in the same observing block (with a few exceptions, see below), so the two measure-ments were essentially simultaneous. Since a significant number of BAL quasars are compact atVLA resolutions, the array configuration was not a critical factor in choosing when to performobservations.The first set of observations were performed with the VLA in the D-array configuration, withthe more southern sources observed in the intermediate DnC configuration, giving resolutions ofabout 15 ′′ (C-band) and 10 ′′ (X-band). The widest available bandwith (50 MHz) was used. Gen-erally exposure times were around 2 minutes for C-band observations and around 3 minutes inX-band, although this varied depending on the brightness of the source at 1.4 GHz. We aimed to 5 –obtain rms values of around 0.5 mJy or less for each source, and this was achieved except in a smallnumber of cases where the rms was between 0.5 and 1 mJy. Observations of a VLA flux calibratorwere performed at least once an hour and phase calibrators located within at least 15 ◦ and usuallywithin 10 ◦ of the science targets were done every half hour. A little more than half of the combinedBAL/normal quasar sample was observed in this configuration.The second set of observations to complete the sample used the new EVLA in the B-arrayconfiguration, observing the southernmost sources in the hybrid BnA configuration. The samefrequencies were used, though the bandwidth on the EVLA is slightly larger at 64 MHz. With thisconfiguration the approximate resolutions are 1.5 ′′ (C-band) and 1 ′′ (X-band). Again we aimed toachieve 0.5 mJy rms noise or better in the radio maps, and again except for a few rare cases thisgoal was reached. The same calibrator strategy was used during this observing period.Data at both frequencies were collected for all the sources in the sample, with a few exceptions:3 sources (1 BAL quasar and 2 normal quasars) were only observed at one frequency either due totechnical problems at the telescope or time constraints, and one BAL quasar was not observed at allfor similar reasons, though it will be included in tables below for completeness. Additionally, fiveBAL sources that overlap with the sample of Montenegro-Montes et al. (2008) were not observedby us, as discussed in the next section.The data were calibrated using CASA , distributed by NRAO, using standard procedures onewould use in AIPS for these straightforward continuum observations. The flux density scale is thatof Perley & Taylor (1999). Radio maps for all sources were inspected and the 4.9 and 8.4 GHz integrated flux measure-ments of the BAL and normal quasar samples are listed in Tables 1 (columns (8) and (9)) and 2(columns (7) and (8)). In the case of resolved/extended sources, the integrated fluxes include thecore and lobe emission. Five BAL sources where the flux values from the Montenegro-Montes etal. (2008) sample were used are noted in the table. Among their observations were simultaneousmeasurements at essentially the same frequencies as ours, so their inclusion as substitutes for ourobservations is valid. Five more of their sources were also observed by us, and the values matchwell. The few 4.9 and 8.4 GHz measurements that were not made simultaneously are also noted inthis table.In order to analyze our data in context with the rest of the radio spectrum, data at otherfrequencies were gathered from the literature where available. In particular, we utilized fluxesfrom Montenegro-Montes et al (2008) (2.6, 4.8, 8.3, 10.5, and 15 GHz), the Green Bank 4.85 http://casa.nrao.edu/ The main goal of these observations was to measure the radio spectral index α of the twosamples for comparison. We did this in several ways, and would like to highlight the index measuredbetween 4.9 and 8.4 GHz ( α . . ) because the flux measurements are simultaneous and variabilitycannot affect the slope of the radio spectrum. These values are given in column (10) of Table 1and column (9) of Table 2. We also calculated the spectral index between 1.4 and 4.9 GHz ( α . . ),and these are listed in column (11) of Table 1 and column (10) of Table 2. Finally, we did a simplelinear fit to all available data points for each object (each source has at least 3 flux measurements),which is quite reasonable in most cases but, as is clear from inspecting the spectra in Figures 2and 3, some are more complex. The spectral index measured in this way ( α fit ) is given in column(12) of Table 1 and column (11) of Table 2. Because the rms values in our radio maps and thereforethe errors in the flux measurements are generally small, the errors on the spectral indices are alsoquite small and thus not reported in the table.
4. RESULTS & DISCUSSION4.1. Radio spectra and spectral indices
Because the majority of our sources only have data at three frequencies (1.4, 4.9, and 8.4GHz), we have not attempted to locate peaks in the radio spectra or apply fits beyond the simplelinear models (where all available data are included). In the left panel of Figure 4 we plot a radio“color-color” plot, comparing the spectral indices α . . and α . . . The line is not a fit, just simply toillustrate where the two spectral indices would be equal. We see that in general there tends to be aflattening of the radio spectrum toward lower frequencies in both samples, as the majority of pointsfall above the line. It is possible that this is due to the presence of synchrotron self absorptionin both samples, though again variability, especially at the higher 4.9 GHz frequency, cannot becompletely ruled out. This trend could also be an indication that BAL quasars do not show peakedspectra of the CSS/GPS type more often than normal quasars, but more data at lower frequenciesfor a larger number of objects is needed to do this properly. To confirm that this flattening occurs 7 –at a similar rate in both samples, we also compared the distributions of the difference in spectralindices ( α . . − α . . ) for all objects in each sample (Figure 4, right). A statistical test on thesedistributions yields P KS = 0 . α . . = 1 .
59. While there are otherobjects with inverted slopes at these frequencies, none are nearly this steep. We note that this isone of the few sources with non-simultaneous 4.9 and 8.4 GHz measurements, and so it is possiblethat it is variable at high frequency even though it does not appear to be variable at 1.4 GHz (see § α . . , α . . , and α fit are given in the top half of Table 3. The bottom halfof the table gives these values when only considering compact (Θ < .
1; see § σ approximately between 0.4 and 0.5, itis clear that the means and medians of all measures of α for BAL quasars are lower (steeper) thanthose of non-BAL sources.Figures 5, 6, and 7 graphically show the distributions of α . . , α . . , and α fit , respectively,for the BAL and non-BAL samples. Just by visual inspection of the distributions of α . . we cansee that there is a favoring of steeper spectra in the BAL quasars, or at least an overabundanceof BALs with α . . between − −
2, though as mentioned previously both samples do have awide range. The odd non-BAL with a highly inverted high frequency spectrum that was discussedabove is also a clear outlier in this plot. The shift between the two samples is apparent, thoughnot as strong, when examining the distributions of the spectral index α . . . As noted variabilityissues cannot be ruled out in this comparison, at 4.9 GHz in particular, and may account for thedifference not appearing as striking in this comparison. Regardless, what we wish to highlight isthat the distributions are measurably different, as shown below. We do acknowledge that usingonly two closely spaced points in the spectrum for this analysis can be problematic. However, thedifference in the distributions using all available data ( α fit ) is also clearly apparent, with BALquasars favoring steeper spectra.Both Kolmogorov-Smirnov (K-S) tests and Wilcoxon Rank-Sum (R-S) tests have been per-formed on the three spectral index measurements made, and the results are shown in Table 4. P ks is the probability that the two samples are from the same parent population from the K-S test, and P rs is the probability that the distribution means are the same from the R-S test. In all cases and byboth tests the differences are significant, though most strongly in the cases of α . . and α fit . Whenrestricted to compact sources (these are the values in the bottom half of the table) the significance 8 –remains, though less strongly for α . . . It is possible that this reduction in significance is simply dueto the decreased sample size. It is worth noting that when this comparison was done after the firstperiod of observations, the difference was apparent but uncertain. As more data came in from thesecond period, the difference grew stronger, suggesting that the scatter in the relationship between α and orientation is large enough that large samples such as this one are required to make meaning-ful comparisons. Finally, since our sample consists of matched pairs of BAL and non-BAL quasars,we also performed a signed rank-sum test on the difference between each measure of spectral in-dex for each pair (matched-pair test). The resulting probabilities that the difference distributionsare symmetric about 0 (and therefore that the samples are indistinguishable) are 3 . × − ( α . . ),0.0022 ( α . . ), and 0.0005 ( α fit ). While we have mentioned issues regarding variability, complexitiesin the spectral shapes, and the spacing between some of the measurements, the fact that all statis-tical tests on all three measurements of α show a significant difference, even when only consideringcompact sources, strongly suggests that the difference between the two samples is real.We have also looked at the spectral index distributions of HiBAL and LoBAL subsamples, andthere is no clear difference between the two. However, the numbers are too small to be conclusive,as there are only 11 Lo/FeLoBALs in the sample.Knowing that spectral index is an orientation indicator, at least in a statistical sense for asample, these results indicate that while BAL quasars can have a range of orientations (as suggestedby the widths of the distribution), they seem to more often be seen “edge-on” and therefore exhibitsomewhat steeper radio spectra. Normal, unabsorbed quasars, then are generally seen more “face-on”, though again they likely span a range of orientations. This result illustrates the complexityof the problem- it is clear that the edge-on only, simple orientation schemes cannot fully explainthese objects, but orientation apparently still plays an important role in the presence or absenceof BAL features. This is an important result that has not yet been directly observed, though somemodeling has suggested it (Shankar et al. 2008); therefore, we will highlight it, provide a morequantitative analysis of its implications, and engage in a more detailed discussion in a forthcomingpaper (DiPompeo et al., in preparation). It has been noted that BAL quasars are generally more compact in radio maps than normalquasars, as discussed in §
1. Using the “morphological parameter” Θ, defined by Montenegro-Monteset al. (2008) as Θ = log( S i /S p ), for the FIRST fluxes in our sample, we do not see the significantdifference seen elsewhere. Assuming Θ > . A check for radio variability was done by comparing the FIRST and NVSS fluxes, since thesurveys were performed at different times. As some of our tests combine data taken at differenttimes, the extent and role of variability may be important. Figure 8 shows the comparison betweenFIRST and NVSS fluxes; the solid line indicates where the two survey fluxes are equal, and thedotted lines show the 3- σ variation of the points around the line of equal flux. From this comparisonwe see that most sources fall close to this line and thus do not appear to be significantly variable.The two non-BAL quasars (SDSS J083629.57+345544.8 and SDSS J105611.77+315616.5) that seemto have much larger NVSS fluxes and thus lie well above the line may have contaminating sourcesabout 35 ′′ away, and so the differences in flux are likely due to resolution issues and not realvariability. At the frequencies we observed, the resolution is sufficient that contamination of thesenearby sources is not an issue.For a quantitative check, we also compute the variability parameter as defined by Torniainenet al. (2005): V ar ∆ S = S max − S min S min . The significance of the variation was computed using a modified method of Zhou et al. (2006) inwhich the integrated fluxes from both surveys are used, instead of the peak flux from FIRST (asthe variable sources are likely to be seen more pole-on and thus appear point-like, this should makelittle difference): σ var = | S − S | p σ + σ where σ and σ are the uncertainties in the integrated fluxes. The error in integrated flux fromFIRST here is assumed to be 5%, since only peak rms values are reported in the survey. We chosea conservative value of σ var > σ var >
4. 10 –Two BAL sources have evidence of variability on short timescales, a property that has beenused to suggest viewing angles within 20 ◦ of the jet axis in order to avoid the inverse Comptoncatastrophe. While using FIRST and NVSS for this sort of analysis is useful because of the largenumber of sources and overlap between the two, we are cautious of this method because it maybe extracting more information from these surveys than they were meant to provide- a few mJydifference between sources could just be noise as the absolute calibrations are simply not thataccurate. Both of the BAL quasars identified here only differ in flux between the surveys byabout 5 mJy. However, for completeness we also calculated the brightness temperature T b of thetwo possibly variable BAL quasars in our sample using equation 4 of Ghosh & Punsly (2007):125243.85+005320.1 has T b = 5 . × K and 135550.30+361627.6 has T b = 2 . × K. Bothof these are greater than the 10 K above which the inverse Compton catastrophe should occur,and suggests that these could also be polar BAL quasars. It should also be noted that the dateof observation for an object in FIRST (an important parameter in T b ) is difficult to know withcertainty, as the fields overlap and the final fluxes presented are a combination of data that couldhave been taken days, weeks, or even years apart. This is not likely to be a large factor in the caseof sources with T b much larger than 10 K, but for borderline objects (as one of ours is), it couldplay an important role.The numbers are small, but it is interesting to note that of the “polar” BALs identified inthe literature that are in our sample, as well as the two just discussed (for 8 total), only two ofthem have what would technically be considered flat radio spectra (using α . . > − .
5. Summary
We have observed a large number of BAL quasars (74) and a well matched sample of 74unabsorbed quasars at 4.9 and 8.4 GHz, simultaneously in all but a few cases. All sources also havedata at 1.4 GHz available from the FIRST survey, and other data from 74 MHz to 15 GHz havebeen collected from the literature when available. The main results are as follows:1. About 13% of the BAL sample appears extended in FIRST maps, similar to previous findingsat arcsecond resolutions. Only 21% of the non-BAL sample show resolved structure, lowerthan previously found- however, there is likely a bias in our selection of unabsorbed quasarsagainst those with individually resolved components.2. We identify two new BAL quasars that may be significantly variable at 1.4 GHz with bright-ness temperatures both above 10 K, suggesting they may be viewed pole-on. 11 –3. Analysis of the spectral index between 4.9 and 8.4 GHz (simultaneous), 1.4 and 4.9 GHz(non-simultaneous), and using a simple linear fit to all available data shows that the twosamples do significantly differ, with the level of significance depending on the frequenciesanalyzed and statistical test performed, but always above 3- σ . BAL quasars do in fact showan overabundance of steep spectrum sources compared to non-BAL quasars, though bothsamples span a wide range of spectral index. This difference persists when only compactobjects are considered. It may be that this difference has not been seen before because thesample sizes were simply too small. The simplest interpretation is that although BAL quasarsmay span a range of orientations, they do have a preference for larger viewing angles. Clearlythe simplest orientation models have been ruled out, but that is not to say that orientationdoes not play a role in these objects.M. DiPompeo would like to thank the European Southern Observatory for providing DGDFfunding to support a visit to collaborate with C. De Breuck in 2011, as well as the excellent supportstaff at NRAO that was extremely helpful during the calibration process using CASA. 12 –Table 1. The BAL quasar sample Source Name Type RA DEC z M B S . S . S . α . . α . . α fit (SDSS J) (mJy) (mJy) (mJy)(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)001408.22 − −
08 52 42.3 1.74 − − − − − − −
00 14 52.6 2.23 − − − − − − − − − −
08 19 36.3 1.68 − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − a a − − − − − − − − − − − − − − − − − − − − −
00 24 18.2 1.65 − − − − − − − − − − − − − −
00 58 52.5 1.50 − c c − − − − − − − − − − − − − − − − − − − − − − − a a − − − − − − − − − − − − − − − − − − − −
01 04 14.2 2.66 − − − − − − − −
13 –Table 1—Continued
Source Name Type RA DEC z M B S . S . S . α . . α . . α fit (SDSS J) (mJy) (mJy) (mJy)(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)123511.59+073330.7 Hi 12 35 11.6 07 33 30.8 3.03 − − − − − − − − − − − − a a − − − − − − − − −
00 37 31.3 1.67 − b b − − − − − − − − − − − −
02 46 29.8 3.06 − − − − − − − − − − − − − c c − − − − − − − − − − − − − · · · · · · − − − − − − − − − − −
00 56 24.6 1.92 − − − − − − − − − − − − − − − − − − − c c − − − − c c − − − − c c − − − − − − − − · · · · · · · · · · · · · · · − − − − a Observations not simultaneous- 8.4 GHz observed with VLA in 2010, 4.9 GHz observed with EVLA in 2011. b Observations not simultaneous- 4.9 GHz observed with VLA in 2010, 8.4 GHz observed with EVLA in 2011. c Values from Montenegro-Montes et al. (2008) at 4.8 GHz and 8.3 GHz. These observations were made essentially simultaneously.
14 –Table 2. The unabsorbed quasar sample
Source Name RA DEC z M B S . S . S . α . . α . . α fit (SDSS J) (mJy) (mJy) (mJy)(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)000050.60 − −
10 21 56.0 2.64 − − − − − − − −
00 08 01.3 1.70 − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − b b − − b b − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − a a − − − − − − − − − − − − − − − − − − − − − · · · · · · · · · − − − − − − − − − − − − − − − −
15 –Table 2—Continued
Source Name RA DEC z M B S . S . S . α . . α . . α fit (SDSS J) (mJy) (mJy) (mJy)(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)130941.51+404757.2 13 09 41.4 40 47 57.3 2.91 − − − − − − − − − − − − − − − − − − − · · · · · · − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − − a Observations not simultaneous- 8.4 GHz observed with VLA in 2010, 4.9 GHz observed with EVLA in 2011 b Observations not simultaneous- 4.9 GHz observed with VLA in 2010, 8.4 GHz observed with EVLA in 2011
16 –Table 3. Radio spectral index statistics
Measurement n Mean Median σ BAL α . . − − α . . − − α fit − − α . . − − α . . − − α fit − − α . . − − α . . − − α fit − − α . . − − α . . − − α fit − − σ ) of the variousspectral index measurements for the full BAL andnormal quasar samples. The number of objects witheach measurement is given in the column labeled n .The second half shows the statistics restricting thesamples to only compact sources, defined as havingΘ < . Table 4. Statistical tests on α distributions Measurement n BAL n non-BAL D ks P ks Z rs P rs α . .
72 72 0.347 0.0002 4.00 3 . × − α . .
73 73 0.287 0.0036 3.18 0.0007 α fit
73 74 0.322 0.0007 3.76 8 . × − c α . .
63 56 0.337 0.0016 3.63 0.0001c α . .
63 57 0.342 0.0012 3.70 0.00011c α fit
63 58 0.394 0.0001 4.19 1 . × − Note. — The top half of the table shows the K-S test statistic ( D ks ) followed bythe corresponding p-value ( P ks ), and the mean rank-sum test statistic ( Z rs ) with itscorresponding p-value ( P rs ) on the BAL and normal quasar distributions for threemeasurements of the spectral index. The bottom half shows the results for theseindices restricted to compact sources.
17 –Table 5. Variable sources at 1.4 GHz
Source (SDSS J) S (FIRST) S (NVSS) V ar ∆ S σ var σ var . The two above the horizontal line are from the BAL sample, and thefour below it are from the unabsorbed sample.
18 –Fig. 1.— Comparison of the properties used to build the matched sample of non-BAL quasars;SDSS i-band ( i ), redshift ( z ), and FIRST integrated flux ( S i ). Because each individual BAL sourcewas matched within a percentage of these properties to a non-BAL source, the histograms are notidentical but similar. The final panel shows the distributions of radio luminosity, with a k-correctionto rest frame 4.9 GHz, to confirm that matching in redshift and flux indeed provided a sample wellmatched in intrinsic luminosity. K-S tests on the distributions do not indicate significant differences. 19 – Fig. 2 a Fig. 2.— Radio spectra of the BAL sample, including literature data (see text for details).
Fig. 2 b
20 –
Fig. 2 c
21 –
Fig. 3 a Fig. 3.— Radio spectra of the normal quasar sample, including literature values (see text fordetails).
Fig. 3 b
22 –
Fig. 3 c Fig. 4.— (Left) Comparison between α . . and α . . . The solid line indicates where the spectral indexis the same in both regions of the spectrum. In general, for both BAL and non-BAL quasars thereis a flattening of the radio spectrum towards lower frequencies. (Right) Histograms of α . . − α . . forboth samples, illustrating that the flattening rate is similar in both samples. A K-S test indicatesthat the distributions are from the same parent population. 23 –Fig. 5.— The radio spectral index α . . distributions for the BAL sample (bold) and non-BAL(diagonal fill) samples. While both samples span a range of spectral indices, there is a significantfavoring of steeper spectra for the BAL sample. As a reminder to the reader, the flux measurementsused here are simultaneous (with very few exceptions), and the redshifts of the samples are wellmatched so there would be very little difference if rest-frame spectral indices were used. A K-S testgives D ks = 0 . Z rs = 4 .
00, with a probability that the meansare the same of 3 . × − . 24 –Fig. 6.— The radio spectral index α . . distributions for the BAL sample (bold) and non-BAL(diagonal fill) samples. The favoring of steeper spectra for BAL quasars is not as significantbut still present. A K-S test gives D ks = 0 . Z rs = 3 . α . . are not simultaneous, and so radio variability may affect the results. 25 –Fig. 7.— The radio spectral index α fit distributions for the BAL sample (bold) and non-BAL(diagonal fill) samples. The fits to the spectra are simple linear fits to all available data foreach source. The favoring of steeper spectra for BAL quasars is still present. A K-S test gives D ks = 0 . Z rs = 3 .
76, with a probability that the means are thesame of 8 . × − . 26 –Fig. 8.— Comparing the FIRST and NVSS fluxes as a check for variability. The solid line showswhere the two survey fluxes are equal, and the dotted lines show the 3- σ variation of the sourcesaround the line of equal flux. The two non-BAL sources that seem to lie well above the line mayboth have a second object contaminating the NVSS fluxes, and are probably not actually variable. 27 – REFERENCES
Becker, R. H., White, R. L., & Helfand, D. J. 1995, ApJ, 450, 559Becker, R. H., White, R. L., Gregg, M.D., Brotherton, M.S., Laurent-Muehleisen, S.A. & Arav, N.2000, ApJ, 538, 72Brotherton, M.S., van Breugel, W., Smith, R.J., Boyle, B.J., Shanks, T., Croom, S.M., Miller, L.& Becker, R.H. 1998, ApJ, 505, L7Brotherton, M.S., Laurent-Muehleisen, S.A., Becker, R.H., Gregg, M.D., Telis, G., While, R.L. &Shang, Z. 2005, AJ, 130, 2006Cohen, A. S., Lane, W. M., Cotton, W. D., Kassim, N. E., Lazio, T. J. W., Perley, R. A., Condon,J. J., & Erickson, W. C. 2007, AJ, 134, 1245Condon, J., Cotton, W.D., Greissen, E.W., Yin, Q.F., Perley, R.A., Taylor, G.B. & Broderick, J.J.1998, ApJ, 115, 1693DiPompeo, M.A., Brotherton, M.S., Becker, R.H., Tran, H.D., Gregg, M.D., White, R.L., &Laurent-Muehleisen, S.A. 2010, ApJS, 189, 83DiPompeo, M.A., Brotherton, M.S., De Breuck, C. 2011, ApJS, 193, 9Doi, A., Kawaguchi, N, Kono, Y. et al. 2009, PASJ, 61, 1389Douglas, J.N., Bash, F.N., Bozyan, F.A., Torrence, G.W., Wolfe, C. 1996, AJ, 111, 1945Fanti, C., Fanti, R., Dallascasa, D., Schilizzi, R.T., Spencer, R.E., & Stanghellini, C. 1995, A&A,302, 317Fine, S., Jarvis, M.J., & Mauch, T. 2011, MNRAS, 412, 213Ghosh, K.K. & Punsly, B. 2007, ApJ, 661, L139Gibson, R.R., Jiang, L., Brandt, W.N., Hall, P.B., Shen, Y., Wu, J., Anderson, S.F., Schneider,D.P., Vanden Berk, D., Gallagher, S.C., Fan, X., & York, D.G. 2009, ApJ, 692, 758Goodrich, R. W. & Miller, J. S. 1995, ApJ, 448, L73Gregg, M.D., Becker, R.H. & de Vries, W. 2006, ApJ, 641, 210Gregg, M.D., Lacy, M., White, R.L., Glikman, E., Helfand, D., Becker, R.H., Brotherton, M.S.2002, ApJ, 564, 133Gregory, P.C., Scott, W.K., Douglas, K., & Condon, J.J. 1996, ApJS, 103, 427Hines, D. C. & Wills, B. J. 1995, ApJ, 448, L69 28 –Hines, D.C. & Schmidt, G.D. 1997, Mass Ejection from Active Galactic Nuclei, 128, 59Jiang, D.R., & Wang, T.G. 2003, A&A, 397, L13Knigge, C., Scaringi, S., Goad, M.R. & Cottis, C.E. 2008, MNRAS386, 1426Kunert-Bajraszewska, M., Siemiginowska, A., Katarzynski, K., & Janiuk, A. 2009, ApJ, 705, 1356Kunert-Bajraszewska, M., Jankiuk, A., Gawronski, M.P., & Siemiginowska, A. 2010, ApJ, 718,1345Liu, Y., Jiang, D.R., Wang, T.G., & Xie, F.G. 2008, MNRAS, 391, 246Miller, B.P., Brandt, W.N., Gibson, R.R., Garmire, G.P. & Shemmer, O. 2009, ApJ, 702, 911Montenegro-Montes, F.M., Mack, K.-H., Vigotti, M., Benn, C.R., Carballo, R., Gonzalez-Serrano,J.I., Holt, J. & Jimenez-Lujan, F. 2008, MNRAS, 388, 1853O’Dea, C.P. 1998, PASP, 110, 493Perley, R.A.. & Taylor, C.B. 1999, VLA Calibrator ManualRengelink, R. B., Tang, Y., de Bruyn, A. G., Miley, G. K., Bremer, M. N., Roettgering, H. J. A.,& Bremer, M. A. R. 1997, A&A, 124, 259Shankar, F., Dai, X., & Sivakoff, G. 2008, ApJ, 687, 859Spergel, D.N., Bean, R., Dore, O., et al. 2007, ApJ, 170, 377Stocke, J.T., Morris, S.L. & Weymann, R.J. 1992, ApJ, 396, 487Torniainen, I., Tornikoski, M., Terastranta, H., Aller, M.F., Aller, H.D. 2005, A&A, 435, 839Weymann, R.J. & Morris, S.L. 1991, ApJ, 373, 23Zhou, H., Want, T., Wang, H., Wang, J., Yuan, W. & Lu, Y. 2006, ApJ, 639, 716