Analysis of a Sample of RC Catalog Objects in the Region Overlapping with the Areas Covered by FIRST and SDSS Surveys. II: Optical Identification with the SDSS Survey and USNO-B1 and 2MASS Catalogs
aa r X i v : . [ a s t r o - ph . C O ] J un Analysis of a Sample of RC Catalog Obje ts in the Region Overlapping withthe Areas Covered by FIRST and SDSS Surveys. II: Opti al Identi(cid:28) ation withthe SDSS Survey and USNO-B1 and 2MASS CatalogsO. P. Zhelenkova1 and A. I. Kopylov11Spe ial Astrophysi al Observatory of the Russian AS, Nizhnij Arkhyz 369167, RussiaWe report the results of opti al identi(cid:28) ation of a sample of RC atalog radio sour es withthe FIRST and SDSS surveys. For 320 sour es identi(cid:28)ed with NVSS and FIRST obje ts weperform opti al identi(cid:28) ation with the SDSS survey. When sele ting opti al andidates wemake maximum use of the information about the stru ture of radio sour es as provided bythe FIRST survey images. We (cid:28)nd opti al andidates for about % of all radio sour es.1. INTRODUCTIONPowerful radio galaxies an be observed al-most at any osmologi al distan e, however, theirhost galaxies are often opti ally faint. Mu hobservational time is needed to identify a ra-dio sour e and (cid:28)nd the redshift of its hostgalaxy. The history of the identi(cid:28) ation of the3CR atalog of bright radio sour es ( ν =187 MHz, S lim =5 Jy) [1, 2℄ serves to illustrate this point.The limiting magnitude of the Palomar Obser-vatory Sky Survey (POSS) of about . m for Eplates, proved to be su(cid:30) ient to allow the identi-(cid:28) ation of 65% of all radio sour es of the atalog,whereas it took almost three de ades to identifythe remaining 35% [3(cid:21)8℄. The number of opti- ally identi(cid:28)ed radio sour es rapidly de reasesat lower (cid:29)uxes. For example, the fra tion ofPOSS identi(cid:28) ations for B2 survey ( ν =408 MHz, S lim = 250 mJy) is equal to 38% [9℄, whereasthe orresponding fra tion for deeper the First( ν = 610 MHz, S lim ∼ mJy) and the Se ond ( ν =1415 MHz, S lim ∼ mJy) Westerbork sur-veys is even lower, of about 20% [10℄.The (cid:28)eld of view of a large teles ope doesnot ex eed ′ − ′ , making the task of (cid:28)ndingenough referen e stars for the astrometri ali-bration problemati at the times before the re-lease of the APM [11℄, USNO [12℄, and GSC [13℄ atalogs. For this reason, se ondary astrometri standards had to be used in the (cid:28)elds studied.Extended radio sour es with omplex stru -ture are identi(cid:28)ed using omposite images madeup of opti al and radio frames. A reliable hoi eof the opti al andidate requires a urate oor-dinate alibration (to within ′′ ), high-resolutionradio images (e.g., like those of the FIRST sur-vey with an angular resolution of 5.4 ′′ ), and deepdire t opti al images (with an R-band limitingmagnitude of m (cid:21) . m ) with the astrometri alibration a ura y at least as good as that ofthe radio image. If even after this pro edurethere remain several opti al andidates, addi-tional information, e.g., broad-band photometry, an be used to pi k up the right one. However,only spe tros opi observations allow the hostgalaxy of the radio sour e to be reliably iden-ti(cid:28)ed.In early 1980s a deep survey was arriedout with the RATAN-600 within a ′ -widesky strip entered at the de lination of SS433( δ . = +4 ◦ ′ ). The survey angular res-olution at 3.94 GHz ( λ =7.6 m) was equal to ∆ α ∼ ′ . The RC atalog ompiled based onthe observation data of this survey has a o-ordinate a ura y of ′′ × ′′ in right as en-sion and de lination, respe tively [14, 15℄. Thefaintest sour es dete ted have (cid:29)ux densities ofabout mJy. The ompleteness of the atalogis equal to 0.8 for sour es with (cid:29)ux densities S . GHz >7.5 mJy within the ± ′ strip about the entral de lination of the survey. The RC atalog ontains a total of 1165 sour es [16, 17℄.A pilot program of opti al identi(cid:28) ation ofobje ts of the RC atalog was performed us-ing the photographi data obtained in 1984(cid:21)1989with the 6-m teles ope of the Spe ial Astrophys-i al Observatory of the Russian A ademy of S i-en es [18℄. It was followed by the identi(cid:28) ation ofa 266-obje t sample with the re(cid:28)ned oordinatesbased on the TXS atalog [19℄ via inspe tionof enlarged POSS prints [20, 21℄. Opti al andi-dates have been found for 72 ( ∼ ) of theseobje ts.The next stage of the work involved the iden-ti(cid:28) ation of a sample of steep-spe trum radiosour es of the RC atalog on deep CCD im- ages ( m limR ∼ m ) taken with the 6-m teles opewithin the framework of the (cid:16)Big Trio(cid:17) programof the sear h for distant radio galaxies [22(cid:21)24℄.The astrometri alibration of small regions ofCCD frames (about ′ × ′ ) was based on se -ondary standards whose oordinates were deter-mined using POSS-I images [25℄ and APM andGSC atalogs. Identi(cid:28) ation was performed bymeans of overlapping high-resolution VLA radioimages with opti al images.The release of the deep opti al SDSS sky sur-vey [26℄ and the FIRST radio survey [27℄ madeit possible to ontinue the identi(cid:28) ation of RCobje ts and re(cid:28)ne the results of the identi(cid:28) a-tion of radio sour es performed by Soboleva etal. [20℄ and Flet her et al. [21℄. About ofFIRST radio sour es are estimated to be identi-(cid:28)ed down to the limiting magnitude of the SDSSsurvey ( m r = 22 . m ), and we therefore believedthat at least the same fra tion of RC obje ts anbe identi(cid:28)ed with SDSS.2. OPTICAL IDENTIFICATIONThe RC atalog overlaps with the SDSS andFIRST surveys in the right-as ension intervalfrom h m to h m , whi h in ludes a total of432 obje ts. A omparison of the images of theNVSS survey [28℄ with angular resolution of ′′ with the FIRST survey images with an angularresolution of . ′′ shows that the latter shouldbe used for opti al identi(cid:28) ation, espe ially in ases where the orresponding NVSS obje t isnot a point sour e (angular size ex eeds ′′ ) andbreaks into independent sour es in the FIRSTimages.In omplex ases (multi omponent sour es,groups of independent radio sour es, or severalopti al andidates) detailed information aboutthe stru ture of the sour e is required for opti- al identi(cid:28) ation. To establish the stru ture ofthe radio sour es a ording to the FIRST surveydata in more detail, we not only drew the on-tours using the Aladin appli ation [29℄, but alsoemployed a software tool that draws (cid:29)ux densityisophotes in the areas studied without the loss ofangular resolution [30℄.The stru ture of the radio sour e orrelateswith the position of the host galaxy, and there-fore not only on the oordinate oin iden e, butalso on the morphologi al type of the sour eshould be taken into a ount when hoosing theopti al andidate. We based our morphologi al lassi(cid:28) ation of radio sour es on a slightly modi-(cid:28)ed variant of the s heme proposed by Lawren eet al. [31℄. Our lassi(cid:28) ation in ludes the follow-ing types: • point radio sour es ( ore or C). The posi-tions of their host galaxies most likely o-in ide with the (cid:29)ux density peak; • sour es with one-sided jets ( ore-jet orCJ). For su h sour es, the position of theopti al andidate oin ides with that ofthe bright ompa t omponent; • sour es with a bright ore and omponents ( ore-lobe or CL). In these sour es inten-sity of radio emission de reases from the enter toward the periphery. The opti al andidate oin ides with the peak of theintensity distribution; • double sour es (double or D) and doublesour es with double omponents (double-double or DD). The opti al andidate indouble radio sour es is lo ated betweenthe omponents or oin ides with the min-imum of the intensity distribution betweenthe merging omponents; • double sour es with a ore (double- ore orDC). A double sour e with a faint ore lo- ated between the omponents. The opti- al andidate oin ides with the ore; • triple sour es (T). The radio sour e on-sists of three omponents, where the (cid:29)uxof the entral part is omparable to thoseof the other two omponents. The opti al andidate oin ides with the entral om-ponent; • multiple sour es (M). These are multi om-ponent sour es with a stru ture that is dif-(cid:28) ult to put into any of the above at-egories. In this ase the position of theopti al andidate is not easy to deter-mine and additional photometri or spe -tros opi information is needed for a hoi eof the opti al obje ts lo ated within the er-ror box of the oordinate alibration.We lassi(cid:28)ed 320 sour es in a ordan e withthis s heme. We then determined the positionof the opti al andidate based on the morpho-logi al type of the sour e. In a number of asesits position with respe t to the isophotes of theradio image made it possible to re(cid:28)ne the typeof the radio sour e or identify the omponentswith individual radio sour es. For example, ifea h of the omponents of the presumed doublesour e had the size of about ′′ and oin idedwith a ertain opti al obje t, we regarded it astwo point sour es, be ause a idental oin iden eis highly unlikely due to the low surfa e densityof the radio sour es.We took the following oordinates for a in-tended opti al andidate (hereafter referred toas the enter of the radio sour e): • F(cid:22)the oordinates of the point sour e orof the sour e with a well-de(cid:28)ned ore aslisted in the FIRST atalog (213 sour es); • N(cid:22)if the sour e has two omponents withdi(cid:27)erent (cid:29)ux densities in the FIRST sur-vey, whi h onstitute a single obje t in theNVSS, we adopt the NVSS oordinates,whi h orrespond to the lo ation of the enter of mass the radio sour e, as the po-sition of the opti al obje t (32 sour es); • Fm(cid:22)the oordinates of the enter of thedouble sour e as measured by the ontourmap of the image in the FIRST survey (75sour es) in the following ases: (1) if the image ontains a faint ore with oordinates that are not listed in theFIRST atalog,(2) in the absen e of a ore the presumedposition is determined by the on-tours of the FIRST radio image.We onsider the opti al obje t to be the mostlikely andidate for identi(cid:28) ation ((cid:16)+(cid:17)) if it waslo ated (a ording to the SDSS atalog) within σ of the enter of the radio sour e, where σ isthe oordinate error. We onsider identi(cid:28) ationsto be doubtful ((cid:16)?(cid:17)) in the following ases: • if the obje t is a point sour e and the pos-sible opti al ounterpart, albeit lose, liesbeyond σ from the enter of the radiosour e; • if there are two opti al obje ts in the vi in-ity of the expe ted lo ation of the opti al andidate; • if the sour e is double, the position of its ore is un ertain, and the opti al obje t iso(cid:27)set from the line onne ting the bright-ness maxima of the lobes; • if it is di(cid:30) ult to make any de(cid:28)nitive on- lusions about the stru ture of the radiosour e based on the FIRST radio map(whether it should be viewed as a mul-ti omponent radio sour e or several inde-pendent radio sour es).The average oordinate error of the FIRSTsurvey is . ′′ , amounting to 1 ′′ for S . GHz ∼ mJy sour es. The a ura y of the radio oordi-nates is better than 1 ′′ for NVSS survey sour eswith (cid:29)ux densities S . GHz >15 mJy. We estimatethe oordinate a ura y for the obje ts with the entral part of the radio sour e identi(cid:28)ed usingmethod (cid:16)Fm(cid:17) or (cid:16)N(cid:17) to be 1 ′′ , and that for theobje ts with the entral part of the radio sour eidenti(cid:28)ed using method (cid:16)F(cid:17) to be . ′′ . In the skyarea studied the average density of SDSS obje ts(for m limr = 22 . m ) is about . / (cid:3) ′′ . The a - ura y of the oordinate alibration of SDSS isabout . ′′ .We use these values to ompute the normal-ized distan e D between the opti al andidateand the enter of the radio sour e: D = s ∆ α σ α + ∆ δ σ δ , where σ α = σ α rad + σ α opt and σ δ = σ δ rad + σ δ opt .We use the maximum-likelihood ratio LR LR for point sour es bythe following formula: LR ( D ) = 12 λ e D (2 λ − , where λ = πσ α σ δ ρ and ρ is the number of obje ts per square ar se -ond.The above formula an be used to ompute LR if the position of the host galaxy an be reli-ably found from the stru ture of the radio sour e,1 See [32℄ for a detailed des ription of the pro edure em-ployed to ompute the probability, reliability, and om-pleteness of opti al identi(cid:28) ation for point and doubleradio sour es. We adopted the above formulae fromthat paper. whi h is the ase for the CJ, CL, DC, and T typesdes ribed above (65% of all sour es in our sam-ple). We also use this formula to ompute the LR values for double radio sour es in the aseswhere the supposed position of the ore an be re-liably determined from the stru ture of the om-ponents. This is true for almost 2/3 of all doublesour es in our sample.An empiri al relation between the reliabil-ity and ompleteness of identi(cid:28) ation for radiosour es [32℄ as a fun tion of LR determines the uto(cid:27) value ( LR cutoff ), whi h separates identi-(cid:28) ation from eventual han e oin iden e withba kground obje ts.The radio sour es for whi h the opti al an-didates have been found an be subdivided intothe following three groups by the morphologi altype and enter of the radio sour e: • point sour es (C), • sour es where the peak of radio emission oin ides with the opti al obje t (CL, CJ,DC, and T), and • double radio sour es (D and DD), wherethe supposed position of the opti al obje twas determined based on the shape of theisophotes of the radio image and not by the oordinates listed in the FIRST atalog.The uto(cid:27) levels of the maximum-likelihood ra-tios, LR cutoff , and the orresponding o(cid:27)sets d rad − opt between the opti al and radio oordi-nates proved to be more or less the same for allthe three groups (see Table 1), thereby lendinga ertain support to our proposed pro edure ofsear hing for opti al andidates.We use the omputed LR ratios for the opti al andidates and the LR cutoff values as a testing riterion in the sele tion of the most likely iden-ti(cid:28) ation. Figure 1 shows the distribution of theo(cid:27)sets between the radio and opti al oordinatesfor the (cid:16)+(cid:17) and (cid:16)?(cid:17) opti al andidates.There are gaps in the equatorial area of theSDSS survey. Inside these gaps we sear hedfor opti al andidates of radio sour es in theUSNO-B1 atalog and identi(cid:28)ed one sour e(RC J1623+0446) in su h a way.We found another sour e (RC J1052+0458) tohave an opti al andidate in the SDSS images,but la k the data in the SDSS database. Thisobje t is listed in the USNO-B1 ( R . m )and 2MASS atalogs [33℄ ( K = 14 . m ).Below we give a few examples of the identi(cid:28)- ation of radio sour es and the reasoning used tosele t the opti al andidates.We found two opti al andidates for theRC J1257+0458 sour e: a galaxy (SDSS type(cid:16)GALAXY(cid:17)) and a starlike obje t ((cid:16)STAR(cid:17)). Byits u, g, r, i, and z-band magnitudes the lat-ter looks more like a star rather than a quasar(quasars, too, are lassi(cid:28)ed as (cid:16)STAR(cid:17)). We hosethe galaxy as the opti al andidate.In the (cid:28)rst our paper [34℄ dedi ated to theidenti(cid:28) ation of the RC atalog with the VLSS,TXS, NVSS, FIRST, and GB6 radio atalogs wegive an identi(cid:28) ation for the double radio sour e RC J0815+0453. It is evident from the radioisophotes that RC J0815+0453 has a well-de(cid:28)ned ore, whi h exa tly oin ides with the galaxy andhen e the latter is the true opti al identi(cid:28) ationrather than the fainter obje t lo ated pre iselyon the line onne ting the intensity maxima ofthe radio sour e lobes. There is yet anotherfaint FIRST J081521.3+045339 radio sour e( S peak . GHz =2.67 mJy) near RC J0815+0453. Bothradio sour es are identi(cid:28)ed with ellipti al galax-ies with pra ti ally the same photometri red-shifts. Thus, a ording to SDSS data, thephotometri redshift of the host galaxy of thefainter radio sour e is Z phot = 0 . ± . (based on the approximation by pattern spe -tra [35℄) and Z phot = 0 . ± . (neural-network method [36℄). The orrespondingphotometri redshift estimates for the se ondgalaxy are Z phot = 0 . ± . and Z phot = 0 . ± . , respe tively. It seemsfair to suppose(cid:22)although this requires furtherstudies(cid:22)that the two sour es form a lose, pos-sibly an intera ting, pair of radio galaxies.The identi(cid:28) ation of RC J0916+0441 requireda detailed information on its stru ture. Thus onemight on lude, based on the isophotes of theFIRST survey image produ ed by Aladin appli- ation, that RC J0916+0441 should be a blendof two independent radio sour es(cid:22)the Northernsour e onsisting of two obje ts of the FIRST atalog and the Southern sour e onsisting offour sour es of the FIRST atalog. However,the isophotes drawn with angular resolution pre-served (we use the software servi e [30℄ to thisend) allowed us to understand the stru ture ofthe sour e. It proved to be an obje t of morpho-logi al type (cid:16) ore-lobe(cid:17) with the ore oin identwith a galaxy (see Fig. 2).We show our next example in Fig. 3. HereRC J0952+0453 is lo ated between the ompo-nents of a double NVSS sour e, loser to thebrighter omponent. It is evident from the shapeof their isophotes as shown on the FIRST on-tour map that the sour e omponents are un on-ne ted, their orientation is not oaxial, and ea h omponent an be identi(cid:28)ed with a separate op-ti al obje t. The RC J0952+0453 sour e is thusa blend of two independent radio sour es.In similar ases, where the size of the beamof RATAN-600 does not make it possible to re-solve lose radio sour es, we assumed the bright-est omponent to be the main ontributor to theRC atalog sour e and then pro eeded to its op-ti al identi(cid:28) ation.3. RESULTS OF IDENTIFICATIONIn our (cid:28)rst paper [34℄ we identi(cid:28)ed 320 of atotal of 432 sour es of the RC atalog lo atedinside the area of interse tion of the SDSS andFIRST surveys. A visual inspe tion of SDSS im-ages superimposed on FIRST ontour maps us-ing re(cid:28)ned oordinates allowed us to (cid:28)nd bona(cid:28)de ((cid:16)+(cid:17)) opti al andidates for 227 (71%) ofthe radio sour es; likely ((cid:16)?(cid:16)) opti al andidatesfor 25 (8%) of the radio sour es, with no opti al andidates to be found for 68 (21%) of all radiosour es.The area onsidered ontains 36 obje ts of thesample of RC sour es with steep spe tra (SS). Weearlier identi(cid:28)ed 35 sour es using the 6-m tele-s ope CCD images within the framework of the(cid:16)Big Trio(cid:17) program. For 10 of these obje ts wefound no opti al andidates in the SDSS survey.Opti al andidates ((cid:16)+(cid:17) and (cid:16)?(cid:17)) werefound for 264 (82%) radio sour es out of 320obje ts. These in lude the SS sample andMG J1131+04562. Note that the summary sta-tisti al data below does not in lude the 11 ob-je ts identi(cid:28)ed using deeper images.Of all the radio sour es identi(cid:28)ed with SDSS,89 are stellar obje ts ((cid:16)STAR(cid:17)), 158 are galaxies((cid:16)GALAXY(cid:17)), and six faint obje ts are lassi(cid:28)edas (cid:16)UNKNOWN(cid:17) in the SDSS database. Stellarobje ts are generally brighter (see Fig. 4) andbluer than galaxies.Table 2 lists the results of identi(cid:28) ation ofradio sour es for di(cid:27)erent types of opti al an-didates depending on the α . − . GHz spe tralindex. We found the sour es most di(cid:30) ult toidentify to be those with ultrasteep spe tra andfaint obje ts la king the spe tral-index estimates(about %). The fra tion of unidenti(cid:28)ed obje tswith inverse, (cid:29)at, and steep spe tra is equal to16(cid:21)17%. The fra tion of steep and ultrasteep2 The RC J1131+0455 sour e turned out to be the grav-itational lens MG J1131+0456, whi h was earlier iden-ti(cid:28)ed by Hewitt et al. [37℄, Tonry et al. [38℄, andKo hanek et al. [39℄ using deeper images than thoseprovided by SDSS.Figure 1. Distribution of the o(cid:27)set between the radio and opti al oordinates for RC obje ts. The bla kdots and ir les orrespond to bona (cid:28)de ((cid:16)+(cid:17)) and likely ((cid:16)?(cid:17)) opti al identi(cid:28) ations, respe tively. The sour e oordinate di(cid:27)eren es in right as ension and de lination are plotted along the x- and y-axis, respe tively.The inner and outer ir les are the boundaries of the oordinate-o(cid:27)set domains orresponding to LR cutoff for point sour es and sour es with a ore, respe tively.spe tra is higher among the radio sour es iden-ti(cid:28)ed with galaxies than among those identi(cid:28)edwith starlike obje ts.Table 3 lists the results of identi(cid:28) ation of 318obje ts3 depending on the morphologi al type of3 There are only two faint di(cid:27)use obje ts among the 320radio sour es, and these ome from the NVSS survey.These sour es are absent on the FIRST survey mapsand hen e annot be identi(cid:28)ed. the radio sour e. Note that double sour es (D,DC, DD) were identi(cid:28)ed mostly with galaxies.The highest rate of failed identi(cid:28) ations is amongthe point sour es (C). The fra tion of su ess-ful identi(cid:28) ations for CJ, CL, and DC-type ob-je ts is higher and their opti al ounterparts arebrighter than in ase of point and double radiosour es, and there must be more nearby obje ts D E C . Figure 2. A omposite opti al (SDSS) and radio (FIRST) image of the RC J0916+0441 radio sour e. Theisophotes drawn preserving the angular resolution show that both the Northern and Southern sour es are omponents of a single (cid:16) ore-lobe(cid:17) type radio sour e. The arrow indi ates the position of the host galaxy.among the CJ, CL, and DC-type sour es. Dis-tant obje ts must make up a signi(cid:28) ant fra tionof unidenti(cid:28)ed double (D) and point (C) radiosour es. 3.1. Comparison of TwoFlux-Density-Complete SamplesIn our (cid:28)rst paper [34℄ we ompare the proper-ties of radio sour es in two (cid:29)ux-density- ompletesamples of the RC atalog in the entral partof the (cid:16)Cold(cid:17) survey. One of the samples in- ludes 130 sour es lo ated within the ∆ H ≤ | ′ | strip about the enter of the beam and with (cid:29)uxdensities S . GHz ≥ mJy. The se ond sam- ple ontains 117 sour es with ∆ H ≤ | ′ | and S . GHz ≥ mJy. These samples partially over-lap in terms of obje ts. For the sake of brevity,hereafter we refer to the (cid:28)rst and se ond samplesas (cid:16)1S(cid:17) and (cid:16)2S(cid:17), respe tively.Table 4 lists the omparative results of theiropti al identi(cid:28) ation. The fra tion of unidenti-(cid:28)ed radio sour es is the same in both samples.We subdivided ea h sample into twogroups by their 1.4(cid:21)4.85 GHz spe tral indi es.One group in ludes obje ts with (cid:29)at spe tra( α < . ), and the other, sour es with steep( α ≥ . ) spe tra. We then ompared the resultsof identi(cid:28) ation (see Tables 5 and 6). In bothsamples sour es with (cid:29)at spe tra are more0Table 1. The quantities that hara terize the reliability of opti al identi(cid:28) ation for obje ts of various types: themean, root-mean-square deviation and median of the separation between the opti al andidate and the radiosour e; the uto(cid:27) maximum-likelihood ratio, and the orresponding uto(cid:27) sear h radius. No data are available forfour radio sour es (two multi omponent obje ts and two obje ts without FIRST survey maps)Type N obj φ = N + /N obj ∆ d meanrad − opt , σ ∆ d rad − opt , ∆ d medrad − opt , LR cutoff ∆ d cutoffrad − opt ,ar se ar se ar se ar se C 124 (39%) 0.64 0.94 1.30 0.47 1.79 1.88D, DD 108 (34%) 0.72 1.43 0.87 1.26 1.93 3.49DC, CJ, CL, T 84 (26%) 0.76 1.44 2.16 0.75 1.41 3.58 D E C . Figure 3. The RC J0952+0453 sour e is lo ated between the two omponents of an NVSS sour e(indi ated by thinner ontours). The higher angular resolution of the FIRST survey allows RC J0952+0453to be resolved into two separate radio sour es, be ause the isophote ontours are un onne ted, theorientations of the major axes do not oin ide, and ea h of the omponents of the radio sour e is identi(cid:28)edwith a separate opti al obje t.1Figure 4. Distribution of the r-band magnitudes of opti al obje ts identi(cid:28)ed with the RC atalog radiosour es. The distributions for galaxies and stellar obje ts are shown by gray and bla k olor, respe tively.Table 2. Results of identi(cid:28) ation depending on the α . − . GHz spe tral index of the radio sour e ( S ∼ ν − α ) forgalaxies (GALAXY), stellar obje ts (STAR), obje ts of unknown type (UNKNOWN), and unidenti(cid:28)ed sour es(Empty Field or EF)Type of identi(cid:28) ation I ( α < − . ) F ( − . ≥ α < . ) S ( . ≥ α < ) U ( α ≥ ) Data unavailable(29) (77) (142) (31) (41)GALAXY (157) 10 26 86 14 21STAR (88) 14 36 28 5 6EF (68) 5 12 25 11 14UNKNOWN (6) (cid:21) 2 3 1 (cid:21) ompa t in terms of their angular sizes bothin the opti al domain and at radio frequen ies.The sour es with steep spe tra are most oftenidenti(cid:28)ed with galaxies both in the (cid:28)rst andse ond samples. The fra tion of identi(cid:28) ationswith starlike obje ts is higher for sour es with(cid:29)at spe tra, and, moreover, the number ofsu h obje ts is almost twi e higher than thatof galaxies in the brighter 2S sample. Starlike obje ts with (cid:29)at spe tra form a separate group,be ause they di(cid:27)er from other obje ts in bothsamples by their olor indi es.Our two samples are, a ording to re entradio-sour e ounts, dominated by radio-loudAGNs. Thus most of the sour es with (cid:29)ux den-sities S . GHz ≥ mJy are radio-loud AGNswith radio luminosities above × W Hz − [40℄, i.e., FR II-type sour es [41℄, whereas fainter2Table 3. Results of the opti al identi(cid:28) ation of the obje ts of the RC atalog for radio sour es of di(cid:27)erentmorphologi al types. The ounts do not in lude four sour es: two multi omponent obje ts, whi h we ould notattribute to any of the types mentioned below, and another two sour es absent in the FIRST survey. We give themedian values for LAS and m r Type N obj LAS , + ? EF STELLAR m stellarr , GALAXY m galaxyr , UNKNOWNar se mag magC 123 1.4 78 10 35 39 20.23 45 21.04 4CL 17 11.0 6 (cid:21) 1 6 19.64 9 16.00 (cid:21)CJ 28 6.4 22 3 3 14 20.33 11 20.61 (cid:21)T 19 34.9 17 (cid:21) 2 6 18.75 10 17.78 1D, DC, DD 129 17.5 93 11 25 22 19.75 81 21.30 1Table 4. Results of opti al identi(cid:28) ation of two (cid:29)ux-density- omplete samples (1S and 2S) in the entral part ofthe (cid:16)Cold(cid:17) surveySample + ? EF GALAXY STAR UNKNOWN1S (106) 90 (70%) 16 (12%) 24 (18%) 68 (64%) 36 (34%) 2 (2%)2S (96) 84 (72%) 12 (10%) 21 (18%) 59 (61%) 35 (36%) 2 (2%)obje ts ( < S . GHz ≤ mJy) are dominatedby FR I-type radio sour es with luminosities be-low the given limit [42℄. The subsample of iden-ti(cid:28)ed radio sour es maintains the same galaxy-to-starlike obje t ratio (of about ), despitethe fa t that the 1S sample is deeper in (cid:29)ux den-sity terms than the 2S sample. We assume thatthe overwhelming majority of starlike obje ts arequasars as photometri and spe tros opi SDSSdata suggest.3.2. Redshifts of Radio Sour esSpe tros opi data are available in SDSS for58 radio sour es of our identi(cid:28) ation list. These are 28 quasars and 30 galaxies. We found the ra-dio sour es identi(cid:28)ed with galaxies to be rathernearby obje ts with the median redshift andmagnitude of Z mediangalaxy = 0 . and m medianr =18 . m , respe tively, whereas those identi(cid:28)ed withquasars to be distant sour es ( Z medianqsr = 1 . ).The latter are brighter than galaxies by almostone magnitude ( m medianr = 17 . m ). These veryobje ts are in the group of ompa t radio sour eswith (cid:29)at spe tra, whi h we pointed out above.The most distant of the identi(cid:28)ed radio sour esare quasars with redshifts Z > . ; they have (cid:29)ator inverse radio spe tra. These are point radiosour es (C) with angular sizes LAS ∼ ′′ , whi hare unresolvable in the FIRST survey.3Table 5. Results of opti al identi(cid:28) ation of two (cid:29)ux-density- omplete samples (1S and 2S) for (cid:29)at- andsteep-spe trum sour es (in di(cid:27)erent intervals of α . − . GHz indi es)Sample/ Fra tion, S med . GHz , S med . GHz , EF, STELLAR, GALAXY, UNKNOWN, N obj % mJy mJy % % % %1S ( α <0.5) 37 32.7 18 19 33 46 2481S ( α ≥ . ) 63 82.3 30 18 25 56 1822S ( α <0.5) 28 52.5 46 12 52 33 3332S ( α ≥ . ) 72 146.5 55 20 21 69 184Table 6. Results of opti al identi(cid:28) ation of two (cid:29)ux-density- omplete samples (1S and 2S) depending on the lassi(cid:28) ation of obje ts (G (cid:21) (cid:16)GALAXY(cid:17); S (cid:21) (cid:16)STAR(cid:17)) and spe tral index. The olumns give the median valuesSample Type, S . GHz , LAS, α Size opt , m u , m g , m r , m i , m z ,% mJy ar se ar se mag mag mag mag mag1S G (46) 18 2.98 0.19 1.91 22.42 21.12 20.04 19.48 19.14( α < . ) S (33) 18 2.24 0.24 1.29 21.03 20.69 20.28 19.82 19.831S G (56) 33 16.6 0.80 2.42 22.66 22.45 21.17 20.23 19.31( α ≥ . ) S (24) 30 9.7 0.67 1.27 20.56 20.03 19.51 19.41 19.392S G (33) 51 1.72 0.25 1.72 22.85 19.72 18.49 18.10 17.93( α < . ) S (52) 62 2.24 0.13 1.21 21.14 20.38 19.98 19.75 19.392S G (57) 55 15.0 0.84 2.06 22.73 22.77 21.51 20.34 20.04( α ≥ . ) S (21) 79 6.41 0.78 1.31 22.16 20.24 20.13 19.36 19.20We gathered the data on the redshifts andapparent magnitudes of the RC atalog radiosour es studied in luding the SS sample to an-alyze the redshift(cid:21) m R relation (see Fig. 5). Theplot is based on the data for 151 radio sour esof the RC atalog: 109 galaxies and 42 quasars.These in lude the 58 mentioned above obje ts,72 sour es of the SS sample of the RC ata- log with the redshifts measured at the 6-m tele-s ope [23, 43(cid:21)46℄, and 21 galaxies from amongthe obje ts identi(cid:28)ed as a result of this work. Forthe latter we use the photometri redshifts fromthe SDSS database [35℄, be ause the photomet-ri redshifts in the Z phot < . − . interval forgalaxies with m r < m and with redshift errors err Z phot < . − . agree well with the empiri-4 al Z(cid:21)R distribution for obje ts with known spe -tros opi Z .The redshift(cid:21)magnitude relation4 for radiogalaxies shows up onspi uously out to Z ∼ and persists at greater redshifts, whereas quasarsexhibit no su h relation. The Z(cid:21)R diagram forradio galaxies at Z ≥ . forms two bran hes,whi h are possibly due to two groups of radiogalaxies with di(cid:27)erent luminosities.4. COMPARISON OF TWO-FREQUENCYSPECTRAL INDICES OF RADIOGALAXIESIn the last two de ades deep enough all-skyradio surveys have been ondu ted at various fre-quen ies. It goes without saying that the radioloud obje ts at large redshifts dete ted in thesesurveys remain unidenti(cid:28)ed due to the la k ofe(cid:30) ient methods of sele ting andidate ounter-parts for deep obje ts. The use of multifrequen yradio data may help to re(cid:28)ne the radio-sour esele tion riteria aimed at dete ting distant ob-je ts.Our list of 143 su(cid:30) iently bright sour es5 on-tains the data in the 74(cid:21)4850 MHz frequen y in-terval given from the VLSS, TXS, NVSS, andGB6 atalogs. We estimate the (cid:29)ux densities ofsome of the obje ts by analyzing their images in4 We transform m r into m R c by the formula from Jordiet al. [47℄.5 We give for these sour es the median(cid:29)ux densities at the following frequen- ies: S MHz =1070 mJy, S MHz =450 mJy, S . GHz =75 mJy, S . GHz =41 mJy. the VLSS and GB6 surveys with the allowan efor the oordinate dependent limiting sensitivityof the maps.We ompare the two-frequen y spe tral in-di es of these sour es and those of the radiogalaxies with
Z > known from the literature.We took the redshifts of 33 radio galaxies with Z > and the (cid:29)ux densities at 74 and 365 MHz,1.4 and 4.85 GHz from [45℄, the list of powerfulradio galaxies given by Miley and De Breu k [48℄,and from the NED and Vizier databases.We ompare the α − MHz and α − MHz ; α − MHz and α . − . GHz ,as well as α − MHz and α . − . GHz spe tralindi es for the RC- atalog radio sour es andradio galaxies with redshifts
Z > studied inthis paper. Figure 6 shows by way of an examplea omparison of spe tral indi es. Two-frequen yspe tral indi es for radio galaxies with largeredshifts di(cid:27)er from the orresponding indi esfor most of our sour es. Thus radio galaxieswith Z > fall within the region determined bythe following onstraints on spe tral indi es: • α − MHz > 0.5; • α − MHz > 0.9; • α . − . GHz > 0.7.We sele ted from our list 18 radio sour eswith spe tral indi es fall within the same areaas the spe tral indi es of galaxies with largeredshifts. Thirteen of these sour es appearin the SS sample of the RC atalog. Nine of5these 18 sour es have their redshifts measuredby the spe tros opi data obtained with the6-m teles ope of the Spe ial Astrophysi alObservatory of the Russian A ademy of S i-en es [46℄. Their median magnitudes, angularsizes, and redshifts are equal to m R = 22 . m , LAS = 20 . ′′ , and Z =0.82, respe tively. Noredshifts are known for the remaining nineobje ts, whi h are fainter than m medianR = 23 . m and have smaller angular sizes ( LAS median =7 . ′′ ): RC J0820+0454, RC J0945+0454,RC J1347+0441, RC J1439+0455 (thesesour es are in the SS sample of the RC ata-log) and RC J1251+0446, RC J1357+0507,RC 1434+0445, RC J1456+0456, andRC J1607+0438. The empiri al Z(cid:21)R rela-tion for the RC atalog (Fig. 5) allows theredshifts of these obje ts to be estimated as Z ≥ . . Their small angular sizes are onsistentwith this estimate, however, only spe tros opi observations may (cid:28)nally on(cid:28)rm our hypotheses.5. CONCLUSIONSThe software for the analysis and visualiza-tion of observational data developed in the pastde ade, and free a ess to modern surveys o(cid:27)ernew, hitherto unavailable opportunities for thestudy of elestial obje ts. A the same time, thenew tools impose new requirements both on theformulation of the problems and on the meth-ods of investigation, whi h in te hnologi al solu-tions are in reasingly dependent on information te hnologies and, (cid:28)rst and foremost, on databasemanagement systems and web servi es.Mass identi(cid:28) ation of sour es from varioussurveys and atalogs is of real interest for theastronomers. Automating the identi(cid:28) ation pro- ess by using already operational virtual obser-vatory web servi es, whi h appears to be rathereasy to implement, is still far from perfe t. Thisespe ially on erns the radio data be ause of thewide range of angular resolution, limiting sensi-tivity, oordinate a ura y of the atalogs, andthe nature of radio sour es.The oordinate a ura y of the NVSS andFIRST surveys allows them to be automati ally ross-identi(cid:28)ed with opti al surveys. A num-ber of authors report the results of identi(cid:28) ationwith the APM [49℄ and SDSS [50℄ opti al sur-veys. The resulting rate of su essful identi(cid:28) a-tions is rather low. It was equal to 24% and 27%in the former and latter ases, respe tively. Ifthe oordinate a ura y of the atalog is lowerthan that of the NVSS and SDSS surveys, as,e.g., in GB6, the rate of su essful identi(cid:28) ationsis even lower(cid:22)of about . [51℄. The numberof FIRST sour es identi(cid:28)ed with SDSS is about30(cid:21)40% of all obje ts in the atalog. These aresingle- omponent sour es with the o(cid:27)set betweenthe opti al and radio oordinates no greater than ′′ . Cross-identi(cid:28) ation servi es operate by thesimple algorithm sear hing for the nearest ob-je t to the enter of the given region. Thisalgorithm e(cid:30) iently identi(cid:28)es single- omponent6Figure 5. The m R magnitude(cid:21)redshift relation for the radio sour es of the RC atalog. The diamond signsand ir les orrespond to the galaxies and quasars, respe tively (the (cid:28)lled diamond signs and ir les indi atethe redshifts Z measured at the 6-m teles ope, and the redshifts of the obje ts shown by other symbols areadopted from the SDSS database).radio sour es, but handles poorly sour es withmore omplex stru ture. The e(cid:30) ien y of opti- al identi(cid:28) ation of radio sour es delivered by theexisting ross-identi(cid:28) ation algorithms does notex eed 30%.We de ided to analyze the RC- atalog sam-ple in order to understand to whi h extent auto-mati identi(cid:28) ation of radio sour es an be per-formed in ase where the angular resolution and oordinate a ura y of the atalog are insu(cid:30) ientfor opti al identi(cid:28) ation, what fra tion of sour es an be identi(cid:28)ed, and to (cid:28)nd out the problems and restri tions of su h a pro edure.Due to the insu(cid:30) ient oordinate a ura y ofthe RC atalog the pro edure of identi(cid:28) ation ofthe sample in the area overlapping with the SDSSand FIRST surveys onsists of two stages. We(cid:28)rst re(cid:28)ned the oordinates of the radio sour esof the atalog by ross-identifying them with thesour es in other radio atalogs with better oor-dinate a ura y (mostly NVSS, whi h we usedas our referen e atalog). In the doubtful aseswe used not only the data from the atalogs, butalso the images provided in the radio surveys. We7Figure 6. Comparison of two-frequen y spe tral indi es α − MHz (TN) and α . − . GHz (NG) for143 bright sour es from the list studied ((cid:28)lled ir les) and for known distant radio galaxies with
Z > (open ir les). Also shown is the equal-index line. The arrow indi ates the most distant known radio galaxy( Z = 5 . ).performed opti al identi(cid:28) ation using re(cid:28)ned ra-dio oordinates and ontour radio maps based onthe FIRST survey images. We re(cid:28)ned the las-si(cid:28) ation of sour es for it to in lude seven types,be ause we on luded that the lassi(cid:28) ation in- luding FRI, FRII, and point sour es failed tofully re(cid:29)e t the stru tural variety of the sour esof the FIRST survey.We identi(cid:28)ed 320 radio sour es of the RC at-alog with sour es from other radio atalogs andwith opti al obje ts. We measured the angularsizes of these obje ts and ounted the numberof their omponents listed in the FIRST ata-log. We found the ratio of the radio sour es with one-, two, three-, and four or more omponentsto be of about 10:5:2:1.It follows from these ounts that simple ross-identi(cid:28) ation algorithms (based on the nearestneighbor sear h) are the most e(cid:30) ient ones forsingle- omponent sour es ( of the list) pro-vided the opti al survey is deep enough. The al-gorithms must be modi(cid:28)ed if they are going to beused for identifying double sour es, whi h makeup for about one third of the list. One (cid:28)fth ofthe list an be identi(cid:28)ed by inspe ting opti al andradio images or by applying ross-identi(cid:28) ationalgorithms involving elements of pattern re og-nition. We identi(cid:28)ed our sample with SDSS and8found opti al andidates for about of single- omponent sour es, whi h make up for ofthe entire list.Cross-identi(cid:28) ation depends essentially onthe adopted sear h radius, whi h is a parameter hara terizing the two atalogs ompared. Thusthe optimum sear h radius for FIRST and SDSS atalogs for single- omponent sour es is ′′ . Iden-ti(cid:28) ation is unlikely if the the separation betweenthe radio sour e and opti al obje t is greater thanthis radius.We ompared the threshold values for themaximum-likelihood ratio fun tions for threegroups of sour es in our sample(cid:22)point, double,and non-point sour es with the oin ident po-sitions of the opti al obje t and radio-emissionpeak, to obtain our own estimate for the sear hregion for opti al obje ts in the FIRST survey.The sear h radius for point sour es is of about . ′′ , whi h almost oin ides with the sear h ra-dius for FIRST and SDSS mentioned above. Thesear h radius for non-point and double sour es isof about . ′′ if the sear h region enter is set to oin ide with the enter of the radio sour e as wede(cid:28)ne it in our (cid:28)rst paper [34℄.We found andidate opti al identi(cid:28) ations foralmost of the radio sour es of the list studiedand no andidates for the remaining of thesour es, i.e., opti al obje ts proved to be fainterthan the limiting magnitude of the SDSS survey( r = 22 . m ). The galaxy-to-starlike obje t ratioamong identi(cid:28)ed sour es is of about , andstarlike obje ts are most probably quasars judg- ing by the photometri and spe tros opi SDSSdata.The distribution of the results of identi(cid:28) a-tions by morphologi al type of the radio sour eis su h that most of the opti al andidates havebeen found for CJ, CL, and T-type sour es;smaller amount of opti al andidates ould befound for double radio sour es, and even lessfor point sour es. Double sour es are mostlyidenti(cid:28)ed with galaxies. CL and T-type sour esare identi(cid:28)ed with brighter opti al obje ts thanCJ-type, double, and point radio sour es. Alarge fra tion of radio sour es identi(cid:28)ed withgalaxies have steep and ultrasteep spe tra inthe 1.4(cid:21)4.85 GHz frequen y interval ompared tostarlike obje ts, however, the fra tion of obje tswith steep and ultrasteep spe tra is even higheramong the radio sour es in (cid:16)empty (cid:28)elds(cid:17). Someof the unidenti(cid:28)ed point sour es and obje ts withultrasteep spe tra in the 1.4(cid:21)4.85 GHz frequen yinterval may have large Z .The stru ture of a radio sour e may bear in-formation about re urrent phases of its a tiv-ity in the host galaxy. Su h systems in lude(cid:16)winged(cid:17) or (cid:16)X-shaped(cid:17) radio sour es [52, 53℄,(cid:16)double-double(cid:17) radio galaxies [54℄, and triplesour es (T) with two omponents straddling rel-atively bright, unresolved ores [55℄. A ordingto FIRST survey maps, about − of theradio sour es of the sample studied an be las-si(cid:28)ed as (cid:16)X-shaped(cid:17), (cid:16)double-double(cid:17), or triple,i.e., as sour es with re urrent a tivity phases.We analyzed the physi al parameters of the9identi(cid:28)ed radio sour es by omparing two (cid:29)ux-density- omplete samples from the entral partof the survey [34℄: one with S lim . MHz ≥ mJyand another with S lim . MHz ≥ mJy. We foundthe fra tion of unidenti(cid:28)ed radio sour es to bethe same in both samples, and the galaxy-to-starlike obje t ratio to be the same (of about 2:1)in both subsamples of identi(cid:28)ed sour es. How-ever, the two samples di(cid:27)er in the fra tion ofsour es with (cid:29)at and steep spe tra. The deepersample ontains more obje ts with (cid:29)at and in-verse spe tra and less obje ts with steep and ul-trasteep spe tra.In both samples sour es with (cid:29)at spe tra aremore ompa t in terms of angular size bothon opti al and radio images. Steep-spe trumsour es are more often identi(cid:28)ed with galaxies,and (cid:29)at-spe trum sour es(cid:22)with starlike obje ts.Note that in the 2S sample the number of iden-ti(cid:28) ations of (cid:29)at-spe trum sour es with starlikeobje ts is almost twi e the number of identi(cid:28)- ations with galaxies. In both samples starlikeobje ts with (cid:29)at spe tra di(cid:27)er in olor from thegalaxies with (cid:29)at spe tra, whereas no su h di(cid:27)er-en es are observed for sour es with steep spe tra.Opti al identi(cid:28) ation with SDSS re-vealed a group of nearby radio galaxies with Z mediansp =0.20 and m medianr = 18 . m and agroup of rather distant radio-loud quasars with Z mediansp =1.76, m medianr = 17 . m . Note thatquasars with Z > . are point sour es withangular sizes LAS median ∼ ′′ . Quasars have (cid:29)atand inverse radio spe tra and, possibly, make up the above-mentioned group of radio sour es with(cid:29)at spe tra and with a spe tral distribution thatis di(cid:27)erent from that of other sour es.The steepness of the spe trum of a radiosour e from low to high frequen ies is widely usedas a riterion for sele ting possible distant ob-je ts. We ompared the spe tral indi es for thesour es in our list, for those with the data at 74,365 MHz and 1.4, 4.85 GHz are available, with 33known radio galaxies with Z > . We found thatdistant obje ts o upy a ertain region of indexvalues on two-frequen y indi es plots. We se-le ted from the list studied the radio sour es withspe tral indi es fall within the same region asthe spe tral indi es of galaxies at large redshifts.Two thirds of these sour es belong to the SS sam-ple of the RC atalog and most of them have theirredshifts measured ( Z median =0.82) [46℄, whereasobje ts with unknown Z are fainter and havesmaller angular sizes. The last group in terms ofapparent magnitude o upies the domain of RC atalog sour es with Z ≥ . ˘Z . Ivezi ´c et al., Astronom. J. 124, 2364(2002).51. \re(cid:28)tem{arti le}M. Obri ´c´c