Selection of AGN candidates in the GOODS-South Field through SPITZER/MIPS 24 microns variability
Judit García-González, Almudena Alonso-Herrero, Pablo G. Pérez-González, Antonio Hernán-Caballero, Vicki L. Sarajedini, Víctor Villar
Highlights of Spanish Astrophysics VIII, Proceedings of the XI Scientific Meeting of the Spanish Astronomical Society held on September 8 – 12, 2014, in Teruel, Spain. A. J. Cenarro, F. Figueras, C. Hernández-‐Monteagudo, J. Trujillo, and L. Valdivielso (eds.)
Selection of AGN candidates in the GOODS-SouthField through SPITZER/MIPS 24 micronsvariability
J. Garc´ıa-Gonz´alez , A. Alonso-Herrero , , P. G. P´erez-Gonz´alez , A.Hern´an-Caballero , V. L. Sarajedini , and V. Villar Instituto de F´ısica de Cantabria, CSIC-UC, 39005 Santander, Spain Augusto Gonz´alez Linares Senior Research Fellow Departamento de Astrof´ısica, UCM, 28040 Madrid, Spain Departament of Astronomy, University of Florida, Gainesville, FL 32611, USA
Abstract
We present a study of galaxies showing mid-infrared variability in the deepest
Spitzer/MIPS µ m surveys in the GOODS-South field. We divide the dataset in epochs and subepochsto study the long-term (months-years) and the short-term (days) variability. We use a χ -statistics method to select AGN candidates with a probability ≤
1% that the observedvariability is due to statistical errors alone. We find 39 (1.7% of the parent sample) sourcesthat show long-term variability and 55 (2.2% of the parent sample) showing short-termvariability. We compare our candidates with AGN selected in the X-ray and radio bands,and AGN candidates selected by their IR emission. Approximately, 50% of the MIPS 24 µ mvariable sources would be identified as AGN with these other methods. Therefore, MIPS24 µ m variability is a new method to identify AGN candidates, possibly dust obscured andlow luminosity AGN that might be missed by other methods. However, the contribution ofthe MIPS 24 µ m variable identified AGN to the general AGN population is small ( ≤ Variability can be used to select active galactic nucleus (AGN). Practically all AGN vary ontime-scales from hours to millions of years ([19]; [7]). Any variability detected in galaxies onhuman time-scales must originate in the nuclear region, because the typical timescale for starformation variability is ≥
100 Myr ([7]). In particular low-luminosity AGN are expected toshow stronger variability than the luminous ones ([18]). Therefore, variability is likely to bean effective method to select low-luminosity AGN. Although the mechanisms that produce a r X i v : . [ a s t r o - ph . GA ] N ov AGN candidates selected by MIPS 24 µ m variability variability are not well understood, the main explanations involve disk instabilities ([13]) orchanges in the amount of accreting material ([9]).The aim of this work is to identify AGN through mid-IR variability in the GOODS-South field using 24 µ m observations taken with the Multiband Imaging Photometer for Spitzer (MIPS, [16]) on board the
Spitzer
Space Telescope ([20]). The near and mid-IRnuclear emission of AGN, once the stellar component is subtracted, is believed to be due tohot and warm dust (200 − µ m and found that 1.1% of the sources satisfied their variability criteria. We compiled all the data taken around the GOODS-South field with the MIPS instrumentat 24 µ m by querying the Spitzer
Heritage Archive. This field was observed by
Spitzer duringseveral campaigns from January 2004 to March 2007. We focused our study on a regionaround RA=3 h m s (J2000) and DEC= − o (cid:48) (cid:48)(cid:48) (J2000). We divided these data setsinto 7 different epochs in order to detect variable sources. For this study we decided toexclude Epochs 2, 4, and 5 because their FoV is small when compared to the other epochs.The common area for the epochs 1, 3, 6, and 7 is ∼ . They probe time scales ofmonths up to three years, and henceforth are used to study the long-term variability coveringa period of over three years. We also subdivided Epoch 7 in three epochs, namely Epochs7a, 7b, and 7c to study the short-term variability. The short-term variability epochs have acommon area of ∼ and probe time scales of days, covering a period of 7 days.To study the temporal variability of MIPS 24 µ m sources detected in the commonregions we built a source catalog for each epoch and subepoch. We used SExtractor (Source-Extractor, [4]) to detect sources and the Image Reduction and Analysis Facility (IRAF) toperform the photometry following the procedure explained in [14] and [15]. We obtained a24 µ m source catalog for each epoch. In this work we restrict the analysis to sources abovethe 5 σ detection limit in the shallowest data in the mosaics. This corresponds to MIPS 24 µ mfluxes of 80 µ Jy and 100 µ Jy for the long-term and the short-term epochs, respectively. Wealso discarded sources with neighbours at distances of less than 10” to minimize crowdingeffects in the photometry that could affect the flux measurements and produce false variabilitypositives. To identify the common sources in all the epochs we cross-matched the catalogsusing a 2” radius, imposing additionally that the 2” criterion was fulfilled in each pair ofepochs. Our final catalogs contain 2277 MIPS 24 µ m sources detected in Epochs 1, 3, 6, and7 and 2452 MIPS 24 µ m sources in Epochs 7a, 7b, and 7c, covering an area of 1360 and 1960arcmin , respectively. arc´ıa-Gonz´alez et al. µ m variable sources in GOODS-South.The left panel corresponds to a long-term variable source (four epochs) and the right panelto a short-term variable candidate (three epochs). The flux for each epoch is plotted withits corresponding photometric error. The solid line is the 24 µ m mean flux of the source andthe gray shaded area is the average of the errors of the source. Each plot lists the name ofthe source, the χ value, and Var. µ m variable sources To select the 24 µ m variable sources we used a χ -statistics method to account for thevariations of intrinsic flux uncertainties of each epoch (related to differences in depth). Thisis the case for our study as different epochs have different depths and within a given mosaicthere are some variations in depth. The latter effect is most prominent in epoch 7, which isdeeper in the center. This method associates each flux with its error. The χ -statistics isdefined as follows: χ = (cid:80) ni =1 ( F i − ¯ F ) σ i , where n is the number of epochs, F i is the flux in agiven epoch, σ i is the associated error in the i th epoch, and ¯ F is the mean flux.We calculated the χ value for each source without neighbours. We selected as variablecandidates those sources above the 99 th percentile of the χ distribution expected from pho-tometric errors alone. That is, only 1% of non-variable sources satisfy the selection criteria.This value corresponds to χ ≥ .
34 for the 4 epochs sample (3 degrees of freedom) and χ ≥ .
21 for the 3 epochs one (2 degrees of freedom).Every object with a χ value higher than the threshold was visually inspected to removeartefacts. We also discarded objects that fell close to the edge of the mosaic. After discardingproblematic objects, our final sample contains 39 MIPS 24 µ m long-term variable sources(0.03 sources × arcmin − ) and 55 MIPS 24 µ m short-term variable sources (0.03 sources × arcmin − ). The selected MIPS 24 µ m long-term and short-term variable sources represent1.7 and 2.2% of the original parent samples, respectively. Only two sources are identified ashaving both, long and short-term variability. 28 MIPS 24 µ m long-term and 33 MIPS 24 µ mshort-term are located in the Extended Chandra Deep Field South (E-CDFS). µ m variable sources The 24 µ m fluxes of the variable sources are dominated by sources with mean fluxes below300 µ Jy. The median 24 µ m flux is 168 µ Jy for the long-term variable sources and 209 µ Jy AGN candidates selected by MIPS 24 µ m variability Figure 2: IRAC colour-colour plot of MIPS 24 µ m variable sources in GOODS-South fromthe Rainbow database (left panel) and plotted according to their redshift (right panel). Thedifferent AGN wedges are shown as blue solid line for [11] and black solid line for [5]. Themulticoloured lines (right panel) are the predicted IRAC colours of the star-forming templatesfor four different templates from [5] with a 20% AGN contribution.for the short-term variable sources. This slight difference in the median values of the 24 µ mfluxes for long and short-term variability is likely reflecting the different depths (i.e., 5 σ detection limits) of the epochs rather than different intrinsic properties of the sources).An estimate of the variability is the ratio between the maximum and minimum valuesand the mean flux ¯ f measured as a %: V ar = f max − f min ¯ f × µ m V ar valuesof the long-term and short-term variable sources are 37-43%, with typical errors of 12-13%.In Figure 1 we show two example light curves, one of long-term and the other of short-termvariable sources. Each plot shows the name of the source, the χ value and the measure ofthe variability V ar ).We also studied the Spitzer-IRAC mid-IR (3.6, 4.5, 5.8, and 8.0 µ m) properties ofthe MIPS 24 µ m variable sources as the IRAC emission has also been used to select AGNcandidates (e.g., [11]; [17]; [1]; [5]; [12]). To obtain the IRAC data for our sources, we usedthe Rainbow
Cosmological Surveys Database, which contains multi-wavelength photometricdata as well as spectroscopic information for sources in different cosmological fields, includingGOODS-South [14, 15]. In Figure 2 we show the IRAC colour-colour plot. 44% of the variablesources fall in the [11] AGN wedge. The majority of the variable sources are compatible witha low AGN contribution in the IR.From the
Rainbow database we also obtained the photometric redshifts of the variablesources (average redshift ∼ µ m luminosity. Themean value of rest-frame log( νL µm /L (cid:12) ) is 10.5 for both the long-term and the short-term variable sources. For those candidates satisfying the [11] AGN selection criteria themean values are log( νL µm /L (cid:12) ) = 10 . arc´ıa-Gonz´alez et al. µ m variable sources selected as AGN by othercriteria. No. X-ray radio other AGN IR Combined [11] Combined variable excess catalogs power law criteria criteriasources No. (%) No. (%) No. (%) No. (%) No. (%) No. (%) No.(%)Long-term variable sourcesIn the E-CDFS 28 7 (25) 2 (7) 4 (14) 1 (4) 8 (29) 12 (43) 17 (61)Outside the E-CDFS 11 0 (0) 0 (0) 0 (0) 2 (18) 2 (18) 5 (45) 5 (45)All 39 7 (18) 2 (5) 4 (10) 3 (8) 10 (26) 17 (44) 22 (56)Short-term variable sourcesIn the E-CDFS 33 4 (12) 1 (3) 3 (9) 0 (0) 5 (15) 14 (42) 17 (52)Outside the E-CDFS 22 1 (5) 1 (5) 1 (5) 1 (5) 2 (9) 5 (23) 5 (23)All 55 5 (9) 2 (4) 4 (7) 1 (2) 7 (13) 19 (35) 22 (40) Variable MIPS 24 µ m sources detected in X-rays. Variable MIPS 24 µ m sources with radio excess. Variable MIPS 24 µ msources in other AGN catalogs. Variable MIPS 24 µ m sources detected as IR power-law AGN. Combined 1 st , 2 nd , 3 rd , and 4 th criteria. Variable MIPS 24 µ m sources satisfying the [11] criteria. All the criteria combined. candidates. Conversely, the candidates not satisfying the [11] criteria have mean values oflog( νL µm /L sun ) = 10 . q = log ( f µ m /f . ) which might be anindication of AGN activity (see [3]).We also studied the X-ray properties and obtained that the 30% of the 24 µ m variablesourced are detected in X-rays in the central part of the E-CDFS (covered by [21]). The X-ray0 . − ∼ × erg s − to ∼ × erg s − . 4% of theX-ray sources satisfying the properties of our parent MIPS 24 µ m catalogs are variable at 24 µ m on the timescales probed by our study. Assuming that deep X-ray exposures provide themajority of the AGN, the 24 µ m variable sources not detected in X-ray would only account ∼
13% of the total AGN population.Finally, the compared our variable sources with sources selected as AGN by othercriteria. We found ∼
56% of the variable 24 µ sources in the E-CDFS would be identified asAGN by other methods. Table 1 summarizes this comparison. We used a χ method to select long-term (years) and short-term (days) variable sources at 24 µ m using deep Spitzer /MIPS imaging data from 7 epochs (2004-2007) in the GOODS-Southfield. We found 39 long-term and 55 short-term variable sources. Of them, 28 long-term and33 short-term variable sources are located in the E-CDFS. The average redshift is ∼ µ m rest-frame. The contribution ofthe AGN to the 24 µ emission is low, which probably implies that they are low-luminosity AGN candidates selected by MIPS 24 µ m variability AGN. 30% of the variable sources are detected in X-ray in the central part of the E-CDFS.Sources without X-ray detection are ≤
13% of the total AGN population in the central part ofthe E-CDFS. Approximately 56% of the variable sources in the E-CDFS would be identifiedas AGN by other methods. Therefore, MIPS 24 µ m variability provides a new method toidentify AGN in cosmological fields. See [6] for more details. Acknowledgments
J.G.-G., A.A.-H., and A.H.-C. acknowledge support from the Augusto G. Linares research programof the Universidad de Cantabria and from the Spanish Plan Nacional through grant AYA2012-31447.P.G.P.-G. acknowledges support from MINECO grant AYA2012-31277.