Accepted forpublicationin ApJ PreprinttypesetusingLATEXstyleemulateapjv.11/26/03 COMPLEX STRUCTURE OF GALAXY CLUSTER ABELL 1689: EVIDENCE FOR A MERGER FROM X-RAY DATA? K.E. Andersson1,2 and G.M. Madejski1 Accepted for publication inApJ ABSTRACT Abell 1689 is a galaxy cluster at z = 0.183 where previous measurements of its mass using various techniques gave discrepant results. We present a new detailed measurement of the mass with the data based on X-ray observations with the European Photon Imaging Camera aboard the XMM- Newton Observatory, determined by using an unparameterized deprojection technique. Fitting the 4 total mass profile to a Navarro-Frenk-White model yields halo concentration c = 7.2+1.6 and r = 0 −2.4 200 0 1.13 0.21h−1Mpc,correspondingtoamasswhichislessthanhalfofwhatisfoundfromgravitational ± 2 lensing. Adding to the evidence of substructure from optical observations, X-ray analysis shows a highly asymmetric temperature profile and a non-uniform redshift distribution implying large scale n relative motion of the gas. A lower than expected gas mass fraction f =0.072 0.008 (for a flat a gas ± J ΛCDM cosmology) suggests a complex spatial and/or dynamical structure. We also find no signs of anyadditionalabsorbingcomponentpreviouslyreportedonthebasisoftheChandradata,confirming 9 2 the XMM low energy response using data from ROSAT. Subject headings: dark matter — galaxies: clusters: individual (Abell 1689) — X-rays: galaxies: 1 clusters v 4 0 1. INTRODUCTION initial power spectrum of fluctuations, and cosmologi- 6 cal model. However, observations often disagree with Galaxy clusters are the largest known gravitationally 1 the numerical models. For instance, one disagreement bound systems in the Universe. The detailed analysis of 0 regards the rotation curves of dwarf elliptical galaxies the massdistributionofclustersisthusimportantinthe 4 which appear to be the result of a constant-density core process of understanding the large scale structure, and 0 whereas numerical simulations predict cuspy dark mat- the nature of dark matter. The three main methods of / h measuring galaxy cluster masses: virial masses from ve- ter halo profiles (Moore et al1999a). In addition, obser- p locity dispersions of cluster galaxies, X-ray imaging and vations show fewer Milky Way satellites than predicted - by the models (Kauffman, White, & Guiderdoni 1993; spectroscopy of the intra-cluster medium (ICM) emis- o Moore et al 1999a). Clearly, to understand the nature r sion, and the gravitational lensing of background galax- t ies,havebeenfoundinrecentyearstobeindisagreement ofgalaxyclusters andthe dark matter they consistof,it s is important to measure the matter distribution in clus- a for some clusters. Generally, the X-ray estimates are in ters via all available means. Fortunately, there are two : good agreement with gravitational lensing for clusters v superbX-rayobservatories,ChandraandXMM-Newton, with a high concentrationof centralX-ray emission(the i featuringexcellentangularresolutionandexceptionalef- X so-called “cooling flow” clusters) but seemingly in dis- fective area coupled with good spectral resolution, and r agreementfor other,less centrally peakedobjects (Allen a 1998). To obtain the estimate of the total mass (includ- those are well suited for detailed analysis of the X-ray emitting gas of galaxy clusters ing that due to dark matter) of a galaxy cluster from Abell 1689 is a cluster showing a large discrepancy itsX-rayemission–commonlyassumedtobe fromopti- among various mass determinations, and we chose it for cally thin, hot plasma that subtends the space between a detailed study. It is a rich cluster, R = 4, without galaxies – it is necessary to make the assumption of hy- a pronounced cooling flow but with an approximately drostatic equilibrium. This is appropriate of course only circular surface brightness distribution suggesting a rel- for clusters that have had time to relax into equilibrium andhavenotexperiencedanyrecentmergerevents. Gen- atively relaxed structure. The large mass, 1015 M⊙ ∼ (Tyson & Fischer 1995), and apparentsymmetric distri- erally,it is assumed that clusters with circular isophotes bution ofAbell 1689make it a suitable cluster for gravi- meet this criterion. tationallensing measurementsaswellas X-raymeasure- Cold dark matter (CDM) hierarchal clustering is the ments. However, the type of clusters that are believed leading theory describing the formation of large scale to be the most relaxed have a coolcentral component of structure quite well. In particular, the numerical simu- enhancedsurfacebrightness. Theabsenceofsuchacom- lations such as Navarro, Frenk, & White (1997) (NFW) ponentinAbell1689suggeststhattheclusterisnotfully successfully reproduce the observed dark matter halo relaxed. Also,thegalaxycontentoftheclusterisunusu- profiles, which appear be independent of the halo mass, allyspiralrichforaclusterwithhighsphericalsymmetry 1 Stanford Linear Accelerator Center, 2575 Sand Hill Road, and richness, with a galaxy type ratio E : S0 : Sp of ap- MenloPark,CA94025, USA proximately 22 : 22 : 28 plus 25 unidentified galaxies 2 Royal Institute of Technology (KTH), AlbaNova University (Duc et al. 2002). Teague, Carter, & Gray (1990) mea- Center(SCFAB),S-10691, Stockholm,Sweden sure a velocity dispersion of σ = 2355+238 km s−1 Electronicaddress: [email protected] 1D −183 2 Andersson & Madejski 0.201 0.188 19:30 0.180 -1:20:00 0.193(cid:9) 0.191 0.202 0.192 30 0.174 0.184 0.192 0.181 0.204 21:00 0.191 0.188(cid:9) 30 1 arcmin = 119 h-1 kpc (EdS) 22:00 34 13:11:30 26 Fig. 1.— R magnitude image of Abell 1689 from the STScI Digitized Sky Survey with all magnitude R < 18 cluster members within the central 2′ region plotted over logarithmic X-ray contours from XMM Mos. The bump toward northeast is due to a foreground point source. Circle sizes are proportional to galaxy R magnitudes. for 66 cluster members, unusually high for a cluster of mass interior to 0.24 h−1 Mpc from this method is thistemperature. PositionsandredshiftsfromDuc et al. M2D(< 0.24 h−1 Mpc) = 1.8 0.1 1014 h−1 M⊙ ± × (2002) for all cluster members with red magnitude R < (Taylor et al. 1998). Measurement of the distortion 18andwithin2′ofthebrightestcentralgalaxyareshown of the luminosity function due to gravitational lens in Fig. 1 together with logarithmic X-ray intensity con- magnification of background galaxies gives M (< 2D tours from XMM Mos 1. 0.25 h−1 Mpc)=0.48 0.16 1015 h−1 M⊙ (Dye et al. ± × Thegravitationallensingestimate from6000bluearcs 2001). Finally, the weak gravitational shear of galax- and arclets calibrated by giant arcs at the Einstein ra- ies in a ESO/MPG Wide Field Imager 33′ 33′ im- × dius of Abell 1689 gives a best fit power-law expo- age gives a mass profile with best fit NFW profile with nent of n = 1.4 0.2 for the projected density pro- r = 1.14 h−1 Mpc and c = 4.7 or a best fit SIS 200 file from 200−h−1 k±pc to 1 h−1 Mpc (Tyson & Fischer with σ = 1028+35 km s−1 (Clowe & Schneider 2001; 1D −42 1995). (Unless otherwise stated, we assume an Einstein- King, Clowe, & Schneider 2002). deSitter (EdS) cosmology with ΩM = 1.0,ΩΛ = 0.0 There is a good indication from the optical and H0 = 100 h km s−1 Mpc−1.) This is steeper than data that the cluster consists of substructures. the profile of an isothermal sphere (n = 1). The Miralda-Escud´e& Babul (1995) suggest a strong lens- − strong lensing analysis of two giant arcs directly gives ing model with two clumps in order to reproduce the M2D(< 0.10 h−1 Mpc) = 1.8 0.1 1014 h−1 M⊙ positions of the brightest arcs. A larger mass clump (Tyson & Fischer 1995). ± × (σ = 1450 km s−1) is centered on the brightest clus- r The mass profile derived from the deficit of lensed ter galaxy while a smaller clump (σ = 700 km s−1) is r red galaxies behind the cluster due to magnifica- located1′ northeastofthemainclump. Theyarriveata tion and deflection of background galaxies suggests mass a factor 2 - 2.5 lowerfor their X-ray estimate com- a projected mass profile of M2D(< R) 3.5 paredto theirgravitationallensingmodel. Girardi et al. 1015(R / h−1Mpc)1.3 h−1 M⊙ for R < 0.32 h≈−1 Mp×c, (1997) identify two distinct substructures centered on close to that of an isothermal sphere (M2D R). The redshifts z = 0.175 and z = 0.184 using positional ∝ XMM Newton observation of Abell 1689 3 and redshift data of cluster galaxies from Teague et al. the count rate of approximately 2σ above the quiescent (1990), providingfurther evidencethatthe cluster isnot period in the 0.3 10 keV band. − relaxed. These clumps are also aligned in the southwest –northeastdirectionbutthelocationsdonotagreewith Vignetting correction theonesofMiralda-Escud´e & Babul(1995). Bothstruc- tures are found to have σ 300 400 km s−1 yielding The effective mirror collecting area of XMM-Newton r ∼ − is notconstantacrossthe fieldof view: it decreaseswith virial masses several times smaller than those derived increasing off-axis angle and this decrease is energy de- from lensing and X-ray estimates. pendent. Thisresultsinapositiondependentdecreasein Canthe X-rayobservationsprovideanyevidence for a the fraction of detected events and when doing imaging substructureinAbell1689,andwhataretheimplications spectroscopy for extended sources, we need to account onthemassinferredfromtheX-raydata? Inanattempt forthis effect. By generatinganAncillary Response File to answer this, we analyze XMM-Newton EPIC pn and (ARF)foreachsourcespectrumregion,usingXMMSAS EPIC Mos data to measure the mass profile of A1689 5.4.1 command arfgen we calculate an averageeffective andto investigatethe spatialstructureofthe cluster. 2 § area for each region considered by us (see below). contains the details of the observations and data reduc- tion with the XMM-Newton as well as with summary of Background subtraction theROSAT,AscaandChandradata; 3coversthemeth- § odsofspectralfittingoftheXMMdataandpresentsthe In order to correctly account for particle-induced and analysis of cluster asymmetry; 4 derives the mass and CosmicX-rayBackground,itwouldbeoptimaltoextract § the slope of the mass distribution in the core; and 5 a background spectrum from the same detector region § presentstheinferences aboutthe structureofthe cluster collected at the same time as the source spectrum. This inferred from the spatial analysis. The paper concludes isofcourseimpossible,andthebackgroundcanbetaken with the summary in 6. from a source-freeregionof the detector (other than the § target, but near it), or can be estimated using blank 2. ABELL1689X-RAYOBSERVATIONS sky data. We adopt the former method, noting that the 2.1. XMM-Newton observation backgroundisnotentirelyconstantoverthefieldofview; however,itcanbeassumedtobeapproximatelyconstant Data preparation with the exception of the fluorescent Cu line at 8 keV Abell 1689 was observed with XMM-Newton for 40 in pn which is the strongest contaminant emission line. ks on December 24th 2001 during revolution 374. For Using this assumption, we can effectively subtract the imaging spectroscopy we use data from the European particlebackgroundbyextractingaspectruminasource PhotonImagingCamera(EPIC)detectorsMos1,Mos2 free region in the same exposure. For pn data in the and pn. Both Mos cameras were operating in the Full 7.8 8.2 keV range is excluded due to the strong spatial − Frame mode whereas pn was using the Extended Full dependence of the Cu internal emission. Frame mode. The Extended Full Frame mode for pn UsinganARFgeneratedforasourceregiononaback- is appropriate for studying diffuse sources since it has ground subtracted spectrum will not take into account lowertimeresolutionandsoitislesssensitivetocontam- that the spectrum used as background was extracted ination from photons being detected during readout of from a different region where the vignetting was higher, the CCDs. These events (so called Out-Of-Time events) due to the larger off-axis distance of the background re- show up as streaksacrossthe X-rayimage andare espe- gion. This will have the effect that the CXB compo- ciallybothersomewhenthegoalofanobservationisspa- nent in our source spectrum will be under-subtracted tially resolved spectroscopy of diffuse sources. All cam- and the net spectrum will contain some remnant of the eras used the Thin filter during the observation. CXB. The particle induced background however is not EPIC background is comprised mainly of three com- vignetted and therefore should leave no remnant. ponents. The external particle background consists pri- To estimate the CXB component in our exposure we marily of soft protons (E < 100 keV) being funneled takethe eventsoutsideofthe field-of-viewtobethe par- p through the mirrors and causing a time variable flaring ticle background. Another spectrum is then extracted signal in the detector. The internal particle background fromalargesourcefreearea,locatedawayfromtheclus- is mainly due to high energy particles interacting with ter. The particlebackgroundis normalizedandremoved the detector material and causing a roughly flat spec- fromthisspectrum. Theresultingspectrumisfittedtoa trumwithflourescentemission-linescharacteristictothe brokenpowerlawmodeltodeterminetheCXBflux. The detector material. This component varies over the de- incorrect vignetting correction of the CXB is found to tector surface. The third source of background is the causeanatmost2%over-estimationofthefluxwhichis cosmic X-ray background (CXB) which is roughly con- thecaseintheoutermostannulusinouranalysis(seebe- stant in time but varies over the sky. low). However since the vignetting is energy-dependent, Foralldatareductionweusethe softwareandcalibra- the incorrect vignetting may cause a small ( 0.3 keV) ∼ tion data implemented in XMM Science Analysis Soft- shift in temperature for the outermostregion. The over- ware(SAS)5.4.1. Toexcludetheeventscontaminatedby all uncertainty in the background is estimated to be at protonflares,weproduce lightcurvesinthe 10 15keV most a few percent. − band where the true X-ray signal is low. We screen the We compareour backgroundsubtractionmethod with datausingaconstraintonthetotalcountrateoflessthan themethodofusingXMMstandardblankskydatacom- 1.5 ct s−1 for Mos and 1.1 ct s−1 for pn in this band, piled by Lumb (2002); we find that both methods give leaving an effective exposure time of 37 ks for Mos and similar fit parameters. However, given the difficulty of 29 ks for pn. This screening corresponds to a limit on normalizingthe CXB, the differenthigh-energyleftovers 4 Andersson & Madejski from proton-flare subtraction and the different inter- is certainly not background related. It is possible that nal particle background which occurs when using back- the pn low-energy discrepancy can be due to incorrect ground data from a different exposure, we chose to use treatment of charge collection at lower energies (S. in-field(ratherthanblank-sky)background. Wefindthis Snowden, priv. comm.). To resolve the discrepancy is more robustin keeping the overallshape of the source between Mos and pn we decided to compare with Asca spectrum uncontaminated. In all our analysis we use a GIS/SIS, ROSAT PSPC and Chandra ACIS-I data. background region of a circular annulus with inner and outer radii of 6′ and 8′ respectively. In fact, recent work 2.2. ROSAT, Asca and Chandra observations by Lumb et al. (2003) suggests that the in-field back- Besides the discrepancy regarding the softest X-ray ground method is probably more accurate. The clus- band for the Abell 1689 data between the Mos and pn ters in the above paper, however, have smaller apparent detectors aboard XMM-Newton, the spectral fits to the angular size which makes the background method more Chandra data presented by Xue & Wu (2002) imply a reliable. higherabsorbingcolumn,6.7 1.5 1020cm−2 thanthe Galactic value of 1.8 1020±cm−2×(Dickey & Lockman × 1990). Sincesuchexcessabsorptionisnotcommonlyde- Spectral fitting tected in X-ray data for clusters, this requires further Inanalyzingthevignettingcorrected,backgroundsub- investigation. To determine if there is indeed any addi- tracted radial count-rate profile, we fit it to a conven- tionalcomponentofabsorptionbeyondthatattributable tional beta model, S(r) (1+r/rc)−3β+0.5, where S(r) to the Galactic column – and assess the reliability of ∝ is the source surface brightness at radius r. The fit the softest energy band of the pn vs. Mos data – we gives rc = 91.2 0.7 h−1 kpc and β = 0.72 0.01 with used the most sensitive soft (< 1 keV) X-ray data for χ2/dof.=46us±ingdataoutto700h−1kpc. ±Clearlythis thisclusterobtainedpriortotheXMM-Newtonobserva- modelisnotaverygoodfit;weshowithereonlyforcom- tions, collected by the ROSAT PSPC. The ROSAT ob- pleteness and comparison with previous work, and note servation conducted during July 18-24, 1992 (available that it is not used in the subsequent analysis. The bad from HEASARC) yielded 13.5 ks of good data. We ex- fitaboveresultsfromthefactthattheclusteremissionis tracted the ROSAT PSPC counts from a region 3′ in morepeakedinthecorethanthebestfitbetamodel. To radius, centered on the nominal center of X-ray emis- obtaina generalidea ofthe properties ofthe cluster and sion. For background,we selecteda source-freeregionof comparethis withprevious results,we extractedspectra the same PSPC image. Using these data over the nomi- for the central 3′ (356 h−1 kpc) region, centered on the nalenergyrange0.15- 2.1keVwith the standardPSPC X-ray centroid at 13h11m29s.4 01◦20′28′′. This radius response matrix applicable to the observation epoch, we − corresponds to 0.32 r or 3.9 r . Background spectra performed a spectral fit to a simple, single-temperature 200 c were extracted from source-free regions from the same MEKAL model with soft X-ray absorption due to gas exposure. For XMM pn we use single and double pixel with Solar abundances, using the XSPEC package as events only whereas for XMM Mos we also use triple above. In the fit we use metal abundances of 0.27 So- and quadruple pixel events. lar obtained in the XMM Mos fit above. While the For spectral fitting we use the XSPEC (Arnaud limited bandpass of ROSAT PSPC precludes an accu- 1996) software package. We fit the data in ratedeterminationofthe temperature(the bestfitvalue the 0.3 10.0 keV range using the MEKAL is kT = 4.3+1.2 keV, 90% confidence regions quoted), − −0.8 (Mewe, Gronenschild, & van den Oord 1985; the PSPCdata providea goodmeasureofthe absorbing Mewe, Lemen, & van den Oord 1986; Kaastra 1992; column: the best value is 1.9 0.3 1020 cm−2, cer- Liedahl, Osterheld, & Goldstein 1995) model for the tainly consistent with the Gala±ctic va×lue. We note that optically thin plasma and galactic absorption. With the thedifferencebetweenthetemperatureinferredfromthe absorption fixed at the Galactic value, NH = 1.8 1020 PSPCfitandthatobtainedfromtheXMM-Newtondata cm−2 (Dickey & Lockman 1990), and assum×ing a asaboveisaresultofthelimitedbandpassofthePSPC, redshift z = 0.183 we arrive at a temperature of located much below the peak of the energy distribution 9.35 0.17 keV and a metal abundance of 0.27 0.03 of the cluster photons. The measurement of the absorb- Solar±for both Mos cameras with χ2/dof. = 95±1/824. ing column, however, clearly indicates that the column From the pn camera we get 8.25 0.15 keV and inferred from the Chandra observation by Xue & Wu 0.23 0.03 Solar with χ2/dof. = 142±4/863. The best (2002) is not correct, and might be due to instrumen- ± fit models with residuals can be seen in Fig. 2 for tal effects. We conclude that the absorbing column is Mos (Left) and pn (Right). The pn temperature is consistent with the Galactic value. in disagreement with Mos data and the reason for To obtain further constraints on the absorbing col- this effect can be seen from the residuals below 1 keV umn, we also used the Asca GIS and SIS data together for pn (Fig. 2 (Right)). Fitting the pn data above 1 with the PSPC data for an independent constraint on keV we get 9.33 0.20 keV and 0.24 0.03 Solar with the continuum radiation in the fitting process. We per- χ2/dof. = 892/±722, in better agree±ment with Mos. formed standard extraction of data from all Asca de- The temperature from Mos is in agreement with that tectors, also from a region from a region 3′ in radius found by Asca, 9.02+−00..4300 keV (Mushotzky & Scharf for the source, and a source-free region of the same im- 1997). This observation does not suffer from pile-up, age for background. We performed a spectral fit si- nor is the low energy spectrum sensitive to background multaneously to data from the PSPC and four Asca subtraction. The background uncertainties below 1 keV detectors. To account for possible flux calibration dif- for this pn spectrum are less than 1 % whereas the pn ferences, we let the normalization among all the dif- soft excess is sometimes higher than 10 %. The excess ferent detectors run free. We used the energy range XMM Newton observation of Abell 1689 5 Fig. 2.—BestfitMEKALmodelsfortheXMMMos(Left)andpn(Right)datafromthecentral3′ regionofAbell1689. Theabsorbing column was fixed at the galactic value (Dickey& Lockman 1990). Fig. 3.—The ratiooftheXMMMos(Left)andpn(Right)spectra ofAbell1689 tothebestfitmodel determinedfromtheROSATand Asca analysis. of 0.7 9.0 keV and 0.5 8.5 keV for the Asca GIS 9.43+0.16 keVandmetallicityof0.26 0.02Solar. While − − −0.15 ± andSIS camerasrespectively. Since Asca detectors (and this fitted value of absorption is formally inconsistent in particular, the SISs) often return spectral fits with with the Galactic and ROSAT-inferred values, this is a excess intrinsic absorption (Iwasawa,Fabian, & Nandra relatively small difference, which might be due to the 1999) we also let the absorbing column be fitted inde- slightly imperfect calibration of the Mos detectors or pendently for the GIS, SIS and PSPC detectors. Tem- the assumption of isothermality made by us for this fit peratureandmetalabundancesweretiedtogetherforall (since T and n are correlatedinthe fitting procedure). H datasets in the fit. The optical redshift z = 0.183 was We notethatusing the ROSAT valueforabsorptionwill used. ThejointfitofROSATandAscadatagivesusthe give us a somewhat lower measure on the temperature best fit temperature of 9.1 0.5 keV and abundances of (see below). ± 0.25 0.06 Solar, in agreement with the values quoted Finally, we reduced the Chandra data for Abell 1689 ± by Mushotzky & Scharf (1997), and the absorption (for using the mostrecentreleaseofdata reduction software; the PSPC data) 1.75 0.08 1020 cm−2, in agreement specifically,duetothedegradationoftheChandraACIS- ± × withtheGalacticHI21cmdata. Weshallusethisvalue I low-energy response correction for the charge transfer for the absorption in the subsequent analysis. inefficiency (CTI) is necessary, and we applied this to This best fit ROSAT/Asca model was compared with theChandradata. AsofChandradataanalysissoftware thedatafromthesameregionintheXMM-Newtoncam- package CIAO ver. 2.3 this correction can be applied eras giving an unfitted reduced χ2 of 1.39 for pn and in the standard event processing. It is also necessary to 1.06forbothMoscamerascombined. Theratioofthese account for the ACIS excess low-energy absorption due spectra to the ROSAT/Asca model can be seen in Fig. to hydrocarbon contamination. We used the acisabs 3. From this result we conclude that Mos low energy code for the correction to the auxiliary response func- response is more consistent with previous data and sub- tion. The event grades used in the ACIS analysis were sequently,wechoosetoignoreallpndatabelow1.1keV. GRADE=0,2,3,4 and 6. Re-fitting the XMM data from the above region using TheclusterwasobservedwiththeChandraACIS-Ide- 0.3 10.0 keV for Mos data and 1.1 10.0 keV for tector array at two separate occasions for 10 ks each on − − pn leaving the absorbing column as a free parameter 2000-04-15 and 2001-01-07. Spectra were extracted for yields n = 1.08 0.16 1020 cm−2, a temperature of the central 3′ region centered on the X-ray centroid at H ± × 6 Andersson & Madejski Fig. 4.—The ratio of the Chandra ACIS-I spectra of Abell 1689 to the best fit model determined from the ROSAT and Asca analysis, before (Left)and after (Right)the acisabs correction. Data used for the spectrum include eventswith ACIS grades 0, 2, 3, 4, and 6. 13h11m29s.4 01◦20′28′′ followingtheCIAO2.3Science gion. Pointsources with intensities greaterthan 3σ over − Threads for extended sources. Fig. 4 shows the ratio the average are excluded. The outer radii of the annuli of the combined Chandra data to the best fit model de- are as follows: 15′′, 25′′, 35′′, 47.5′′, 60′′, 75′′, 95′′, 125′′, termined from ROSAT and Asca above. The left spec- 165′′, 230′′ and 350′′. trumisbeforeandtherightaftertheacisabscorrection. Average cluster properties were determined using all Fitting the corrected spectrum to an absorbed MEKAL annular spectra simultaneously out to 693 h−1 kpc model in the energy range 0.3 - 7.0 keV yields absorp- (0.61 r , 7.6 r ) using the same spectral fitting pro- 200 c tionof1.7 0.6 1020cm−2,atemperatureof12.4 1.1 cedures as in Section 2.1.0. We use the optical redshift ± × ± keV and abundances of 0.34 0.10 Solar using the op- of z = 0.183 (Teague et al. 1990) and the line of sight tical redshift z=0.183. The di±screpant absorption found absorptionof N =1.75 1020 cm−2 as measured from H × by Xue & Wu (2002) (using data 0.7 - 9.0 keV) is ap- ROSAT data, and also consistent with the Galactic col- parently corrected for by acisabs and the value is in umn density (Section 2.2). The mean temperature of agreement galactic absorption. However there is still a the cluster is found to be kT = 9.00+0.13 keV and the −0.12 largediscrepancyintemperature,whichalsocanbeseen mean abundance 0.25 0.02 Solar. We note that leav- from the high energy ends of Fig. 4. Part of this effect ing the redshift as a fr±ee parameter gives a best fit red- can possibly be attributed to the high energy particle shift z = 0.173 0.003 (90% confidence range) which backgroundbut more likely to uncorrectedinstrumental is considerably le±ss than measured via optical observa- effects. Repeating all above steps for single pixel events tions. Considering Mos and pn data separately gives only(GRADE=0)inordertoachievehigherspectralac- z =0.171 0.002 and z =0.178 0.003 respectively. curacygivesusabestfittemperatureof7.2 0.4keV.We Inorder±totakeintoaccountth±ethree-dimensionalna- ± cannotaccountforthedifferencesbetweenthetwoChan- ture of the cluster we consider the spectrum from each dra data sets (using GRADE=0 vs. GRADE=0,2,3,4 of the 11 annuli to be a superposition of spectra from a and 6 events) nor between the results of the Chandra number of concentric spherical shells intersected by the and XMM spectral fits. We note here that the photon same annulus. The sphericalshells havethe same spher- statistics resulting from the XMM observation is supe- icalradiiastheprojectedradiioftheannuli. Weassume rior to that in the Chandra data, and since our analysis that each shell has a constant temperature, gas density does not require the superior angular resolution of the andabundance. Thevolumeforeachannulus/shellinter- Chandramirror,welimittheanalysisbelowtotheXMM sectionis calculatedto determine howlargea fractionof data. emissionfromeachshellshouldbeattributedtoeachan- nulus. Assigninga spectralmodelto eachsphericalshell 3. SPECTRALANALYSIS we can simultaneously fit the properties of all spherical 3.1. Temperature and metallicity distribution for a shells. spherically symmetric model In practice we use the method of Arabadjis, Bautz, & Garmire (2002) where we have To obtain a radial profile of cluster gas temperature, a matrix of 11 11 MEKAL models with absorption in abundance, and density, we first make the assumption × XSPEC. In the process of fitting the data in XSPEC, that the cluster is spherical and that above properties eachdatagroup(consistingofthreedatafiles: pn,Mos1 are only functions of radius. For this, we divide the im- age of the cluster into 11 concentric annuli out to 5′50′′ and Mos2 spectra for each annulus) is fitted using (693 h−1 kpc) centered on the X-ray centroid. For each the same set of models. For the central annulus (the one with zero inner radius so it’s actually a circle), annulus, we extract spectra from all EPIC cameras,and which will intersect all 11 spherical shells, we will need we set the inner and outer radii of each region by re- to apply 11 MEKAL models to represent these. This quiring that each annulus contains at least 9000 counts per each Mos camera and 13000 for pn. This allows us meansthatwe havetoapply11MEKALmodelsto each datagroup (annulus) with absorption where each model to derive a reliable estimate of temperature in each re- XMM Newton observation of Abell 1689 7 Fig. 5.— The radial metallicity distribution of Abell 1689 as Fig. 6.— The radial temperature distribution of Abell 1689 as derived from the spectral deprojection fitting. derived from the spectral deprojection fitting. been studied in detail by e.g. Pratt & Arnaud (2002) represents the properties of one spherical shell. The whofindthatcorrectingforthe PSFredistributiongives normalization of each shell model is set to be the ratio a profile that is consistent with an uncorrected profile. of the volume that shell occupies in the cylinder that is Abell 1689 has a temperature profile without large tem- the annulus/shell intersection to the volume it occupies perature variations and no large central flux concentra- in the central annulus. Abundance and temperature tion. Weconcludethattheeffectoffluxredistributionin are tied together for the models representing the same our case will be small and we do not attempt to correct spherical shell. Of course not all annuli intersect every for this. The PSF is also weakly energy dependent, and shell, and for those shells not intersected by the annulus to quantify its possible effect on the observed tempera- to which they are attributed, the model normalization ture profile, we calculate the energy dependent flux loss will be zero. The matrix of MEKAL models is thus from the central annulus and the effect on the central triangular and can be fitted directly to the spectra we temperature. The difference in flux loss between various have extracted from annular regions in the data. This energy bands (ranging from 1.5 to 7.5 keV) for the on- approach allows for all data to be fitted simultaneously, axis PSF is 3%. We find that for a cluster with an ∼ and we do not have the problem with error propagation assumed temperature of 9 keV, this could give an error which occurs when starting to fit the outermost annulus of the central temperature by at the most 0.5 keV. We and propagating inward subtracting contributions from note that this is a maximum difference since in practice, each previous shell. thefluxgainedfromouterannulicouldsomewhatreduce The metallicity profile (Fig. 5) from the deprojection this effect by working in the opposite manner. showssignsofincreasingabundancetowardthecenterof The luminosity of the cluster in the 0.5 10.0 keV − the cluster. Intemperature(Fig. 6)wefind anapparent band, calculated from the best fit model above (with decrease for large radii (kT <8 keV), an effect that has z =0.183)givesLX (EdS) =1.02 1045 h−2 ergs s−1 for × been seen in analysis of other clusters with XMM (see an EdS (Ω = 1.0, Ω = 0.0) Universe or L = M Λ X (CDM) e.g. Pratt & Arnaud (2002)). Gas dynamic simulations 1.21 1045 h−2 erg s−1 for a CDM (Ω = 0.3, Ω = M Λ of the formation of galaxy clusters also show a decline 0.7) ×Universe. This corresponds to a bolometric lu- of temperature at large radii (Evrard & Metzler 1996). minosity of L = 1.63 1045 h−2 erg s−1 or bol (EdS) We do not find a significant cooling in the cluster core L = 1.94 1045 h−2 e×rg s−1. All above val- bol (CDM) with the highest temperature (kT 9.5 keV) near the × ∼ ues should include 10% as the absolute calibration er- core radius: in fact, we will show in section 3.2 that the ror of XMM. From Chandra analysis, Xue & Wu (2002) temperature profile is not symmetric around the cluster find L =1.66 0.64 1045 h−2 ergs s−1 whereas center. Wenotethatforcompleteness,wealsoperformed bol(EdS) ± × Mushotzky & Scharf (1997) find L = 1.77 the above analysis with the best-fit value of the redshift bol(EdS) × inferred from the X-ray data alone, and while the ex- 1045 h−2 ergs s−1 from Asca data, both in agreement act values of temperature and elemental abundances are with our results. This does not provide any new in- slightly different, about 0.2 keV lower for temperature, formation regarding the location of Abell 1689 in the the general trends in the radial runs of the parameters Luminosity-Temperature(Mushotzky & Scharf1997)re- are the same. lation and it is still in a close agreement with the trend The limited point spread function (PSF) of the XMM suggested by other clusters. mirrors is a potential problem especially for the annuli 3.2. Asymmetry analysis locatedclosetothecentersincethosearenotmuchlarger thanthe PSFFWHM of 6′′. Someofthe flux incident With the good quality XMM data, it is possible to ∼ on the central (circular) region will be distributed over verify the result of Xue & Wu (2002) that there is no the outer annuli and vice versa. This flux redistribu- discrepancy between the optical and X-ray centers of tion will have the effect of smoothing out the measured the cluster. We determine the center of X-ray emis- temperature profile since all annuli will have some flux sionforAbell1689usingXIMAGE commandcentroid. that originally belong in other annuli. This effect has We also include a measurement of the lensing center 8 Andersson & Madejski 17:00 18:00 19:00 -1:20:00 21:00 22:00 23:00 1 arcmin = 119 h-1 kpc (EdS) 24:00 13:11:40 30 20 Fig. 7.—Spatial distributionof spectral fit temperatures (6-10 keV)with superimposed logarithmic X-ray contours forA1689. 6keVis marked as black, 10 keV is white, and intermediate temperatures are various shades of gray. oftheICMradialtemperaturedistributionaswellasthe Table 1. Comparison of the position of the cluster center of Abelloffset of optical and X-ray redshifts prompted us to an- 1689frommembergalaxies,gravitationallensing andX-raydata alyze the spatial structure of the cluster. In Fig. 7 we show the spatial temperature distribution. Spectra were Method Center Ref. extractedinrectangularregionsandfittedusingthesame R.A.(J2000.) Dec. (J2000.) method as in Section 2.1.0. The temperature in the fig- X-ray(ROSAT) 13h11m29s.1 −01◦20′40′′ 1 ure scales linearly from6 keV(black) to 10 keV (white). X-ray(Chandra) 13h11m29s.45 −01◦20′28′′.06 2 The errors on the temperature of the inner 16 regions Lensing 13h11m29s.6 −01◦20′29′′ 1 are 0.5keV,whiletheerrorsontheouter8regionsare Optical 13h11m29s.44 −01◦20′29′′.4 3 ∼ X-ray(XMM) 13h11m29s.4 −01◦20′28′′ 4 ∼1.0 keV. We clearly see an asymmetry in the temper- ature around the cluster center with an overall increase toward the northeast. References. — (1) Allen (1998) ; (2) Xue&Wu (2002); (3) To check the consistency of these results and to in- Ducetal.(2002);(4)Thisstudy. crease our accuracy, we re-group the data in larger spa- tial regions and perform a fit using the same above pro- cedure. We first fit the data keeping the redshift frozen fromDuc et al.(2002) andthe X-raymeasurementfrom at the optical value. Temperatures derived from this fit ROSAT by Allen (1998) (Table 1). We find that all val- areshownin Fig. 8 (Left). Thereafter, we leavethe red- uesagreewithin 3′′ exceptthe ROSATestimate,the off- shift as a free parameter and re-fit the data. The fitted setofwhichweattributetouncertaintiesinROSATHRI redshifts for these regions are shown in Fig. 8 (Right). astrometry. This apparently perfect agreement among All errors in Fig. 8 are 90% confidence limits. In the X-ray, lensing, and optical centers leads us to conclude temperaturemap,weseeacleardiscrepancybetweenthe that the ICM density peak and the central dominant northernandsouthernpartoftheclusterwithahintofa galaxy is probably located at the bottom of the dark temperaturegradientinthesouthwest–northeastdirec- matterpotentialwell. Still,theapparentnon-uniformity XMM Newton observation of Abell 1689 9 9.30+-0.34 keV 0.172+-0.003 8.99+-0.30 keV 0.185+-0.006 8.29+-0.37 keV 0.167+-0.008 8.08+-0.31 keV 0.170+-0.009 1 arcmin = 119 h-1 kpc (EdS) 1 arcmin = 119 h-1 kpc (EdS) Fig. 8.— The spatial distribution of temperature (left) and redshift (Right) in the central region of Abell 1689 with 90% confidence intervals. tion. The redshift map reveals a high redshift structure sentially the same peak position of 8.051 0.001 keV. ± to the east at z =0.185 0.006 separated from the rest This corresponds to a possible artificial redshift offset of ± of the cluster at z 0.17. Analyzing the Mos and pn maximum ∆z =0.0005. To explain the difference in the ∼ dataseparatelyforthishighredshiftregiongivesabroad offsetmeasuredbyusasaninstrumentaleffect,wewould minimuminχ2 atz =0.187 0.008forMoswhereaspn havetohaveanoffset(sayattheFeKlineat6.7keV)of ± shows several minima in the z =0.165 0.200 range. 62eV,andnotonthe orderof1eV,asinferredfromthe − This region may indicate a subcluster falling inward Cu K instrumental line. Hence we conclude that there awayfromtheobserver. Thisisfurthersupportedbythe is no gain shift that could alter our redshift measure- optical data, indicating that there are also high-redshift ments between pn CCDs 4 and 7. According to XMM giant elliptical galaxies in this region (Fig. 1): it is in- calibration documentation (Kirsch 2003) the magnitude teresting to speculate if this is actually the remains of a of calibration errors for pn & Mos should be no larger cluster core? Especially intriguing is the apparent coin- than 10 eV. cidence between the smaller subcluster as suggested by 4. MASSPROFILE strong lensing (Miralda-Escud´e & Babul 1995) and our high redshift gas region approximately 1′ northeast of 4.1. Mass calculation the main cluster. Another possibility is that the red- Ifweassumethattheclusterissphericalwithasmooth shift variation is due to large bulk motions of the intra- static gravitational potential and that the intra-cluster cluster gas. It has been shown in gas dynamic simula- medium is a pressure-supported plasma, we can employ tions that clusters with apparently relaxed X-ray pro- thehydrostaticequilibriumequation. ThecircularX-ray filescanhavecomplexgas-velocityfieldsandbefarfrom isophotes (Fig. 1) generally indicate that a cluster is in relaxed (Evrard& Metzler 1996). This kind of motion dynamicalequilibrium. The hydrostaticequationcanbe could give rise to non-thermal emission from shocks etc. written as (Sarazin 1988) : In our analysis we cannot distinguish between the two kT (r) r d ln T (r) d ln ρ (r) possibilities of bulk motion and subclustering. M(r)= g g + g (1) Thismeasurementofnon-uniformredshiftdistribution − Gmpµ (cid:18) d ln r d ln r (cid:19) ofthe X-rayemitting gasis important,andtoverifyifit where M(r) is the enclosed total gravitating mass en- couldbeduetoinstrumentaleffects,weinvestigatedthis closed within a sphere of a radius r, T (r) and ρ (r) g g in more detail. The data in the regions above are from are temperature and density of the ICM at r, µ is the different CCD chips in the pn data whereas for Mos all mean particle weight and m is the proton mass. Using p data are from the same chip. Since the pn camera pro- thetemperatureandnormalizationsfromthespectralde- videsabouthalfofthedata,wewanttoverifythatthere projectionfittingwecalculatethetotalgravitatingmass. are no gain shifts between the CCDs, which, if present, Errors are treated by error propagation. could easily cause such an effect. Most of the cluster Themassdistributionisfittedtoasingularisothermal emission is on the CCD chips 4 and 7. To test for any sphere (SIS) possible gain offset, we extracted the data from each of 2σ2r the chips individually to verify the position of the inter- M(r)= r (2) G nalfluorescentCuKαline,withthedominantcomponent where σ is the 1-dimensional velocity dispersion, which at 8.0478 keV. The spectra in the range 7.8 8.2 keV r − is used here for a comparison with previous mass esti- are fitted to a Gaussian profile: both datasets yield es- mates. 10 Andersson & Madejski The predicted density profile from CDM hierarchal clustering according to Navarro et al. (1997) for dark matter halos is ρ(r) δ c = (3) ρ (z) (r/r )(1+r/r )2 crit s s where ρ (z) is the critical energy density at halo red- crit shift z and δ is characteristic density defined by c 200 c3 δ = (4) c 3 [ln(1+c) c/(1+c)] − wherec=r /r istheconcentrationofthehalodefined 200 s as the ratio of the virial radius r to r , which in turn 200 s isacharacteristicradiusintheNFWmodel. Thecritical Fig. 9.—Spherical mass profile of Abell1689 (pluses) with best density at redshift z for a flat (Ω0 =1) Universe is fit NFW model (solid line) and singular isothermal sphere model (dot-dashed). The singularisothermal sphere isclearly notavery ρ (z)= 3H02[Ω (1+z)3+Ω ] (5) good fit to the data (see also Table 2). For comparison, the mass crit M Λ of the intra-clustergas is included (asterisks). 8πG where H is the Hubble constant, Ω and Ω are the 0 M Λ current contributions of matter and vacuum energy re- spectively to the energy density of the Universe. inthe Allen et al. (2003) sampleandsignificantlylower More recent numerical studies suggest a steeper core thanthemean. Wefindthatinourestimatefgas hasnot slope and a sharper turn-over from small to large radii convergedtoaconstantandthismighthelpexplainpart (Moore et al 1999). Both models can be generalized as of the discrepancy. However,it may be the case that for many clusters f does not converge until well beyond ρ gas ρ(r)= 0 (6) r2500. (r/rs)γ[1+(r/rs)α](β−γ)/α Comparingthemassandtemperatureatr2500 ofAbell 1689 to the M-T relation derived for a set of relaxed (Zhao1996),whereγandβcharacterizethedensityslope clusters (Allen & Fabian 2001) shows a low mass for the at small and large radii respectively whereas α deter- temperature of Abell 1689. The M-T relation predicts minesthesharpnessintheturn-over. Moststudiesagree on β = 3 but the value of γ is still being debated. The H(z)/H0 M2500 = 4.5 1014M⊙( 10%) for a 9 keV × × ± clusterwhereH(z)istheHubbleconstantattheredshift NFW and Moore profiles fit into the parameter space ofthecluster. ForAbell1689wefindH(z)/H M = (α,β,γ) as (1,3,1) and (1.5,3,1.5) respectively. 0 2500 We choose to fit our data to the NFW model (Eq. 3) 2.4 1014M⊙( 15%),significantlylower. The×aboveval- × ± ues were derived assuming a flat ΛCDM cosmologywith which, when integrated over r, yields Ω = 0.3, Ω = 0.7 and h = 0.7. The unusually low M Λ M(r)=M [ln(1+r/r )+(1+r/r )−1 1] (7) massmaybe due tothe factthatthe massofAbell 1689 0 s s × − seems to increase steadily beyond r . However, we where M = 4πρ δ r3. We find that the data give the 2500 0 c c s note that a lower mass (than would be predicted by the best fit for the NFW model with c=7.2+1.6 and r = −2.4 200 M-Trelation)is notanuncommonfeaturefor unrelaxed 1.13 0.21h−1 MpcwhereastheSISfitgivesσr =918 clusters (cf. Smith et al. (2003)). 27 km± s−1. The total mass data and models are show±n For completeness, we note that the calculated total in Fig. 9 together with the mass of the X-ray emitting massincludestheICMandgalaxymasscontributionsas gas Mgas. Model parameters are summarized in Table well as the dark matter. The proper way of fitting the 2. For a cosmology with ΩM = 0.3 and ΩΛ = 0.7, the NFW model would be to subtract these contributions best fit NFW model changes to c = 7.7+−12..76 and r200 = prior to performing the fit. The NFW model for the 1.31 0.25 h−1 Mpc. actual dark matter halo is not used in this paper; we ± Another cosmologically important quantity for clus- only calculate the total mass profile. ters is the fraction of the total mass that is in the X- ray emitting gas, f = M /M . Allen et al. 4.2. Core slope gas gas TOT (2003) analyze data from the Chandra Observatory for The mass data were fitted to a simple power law 10 dynamically relaxed clusters between z = 0.09 and (M(< r) rγ) in the ranges 20 90 h−1 kpc and z = 0.46 and measure an average redshift independent 200 500 ∝h−1 kpc. We find the b−est fit of the slope f =0.108 0.014 at the r radius (where the total − gas 2500 of the matter profile to to be 1.73 0.34and 0.72 0.32 ± massdensityis2500timesthecriticaldensityatthered- ± ± for small and large radii, respectively. This corresponds shiftofthecluster). Thecosmologywherethisisvalidis tototalmassdensityslopes(ρ rα)ofα= 1.27 0.34 a flat ΛCDM cosmology with ΩM = 0.291+−00..004306 assum- and α = 2.28 0.32, in go∝od agreement−with±what ingΩ h2 =0.0205 0.0018andh=0.72 0.08. Theer- is expected−from±numerical simulations of CDM hierar- b ± ± roronf aboveisthermsdispersionoftheAllen et al. chal clustering. We note here that Bautz & Arabadjis gas (2003) sample which is comparable to the individual er- (2002) have measured the density profile of Abell 1689 rorsonf . With ourbest fit NFW modelin the above for r < 100 h−1 kpc using Chandra data and found gas cosmology we find f = 0.072 0.008 for Abell 1689 α 1.3. gas ± ∼− at the r radius. This is lower than all the clusters We do not observe a flattening of the core density 2500