TheAstrophysicalJournal,782:3(7pp),2014February10 doi:10.1088/0004-637X/782/1/3 (cid:2)C2014.TheAmericanAstronomicalSociety.Allrightsreserved.PrintedintheU.S.A. NuSTARANDSWIFTOBSERVATIONSOFTHEFASTROTATINGMAGNETIZEDWHITEDWARFAEAQUARII TakaoKitaguchi1,HongjunAn2,AndreiM.Beloborodov3,EricV.Gotthelf3,TakayukiHayashi4, VictoriaM.Kaspi2,VikramR.Rana5,StevenE.Boggs6,FinnE.Christensen7,WilliamW.Craig6,8, CharlesJ.Hailey3,FionaA.Harrison5,DanielStern9,andWillW.Zhang10 1RIKENNishinaCenter,2-1Hirosawa,Wako,Saitama351-0198,Japan 2DepartmentofPhysics,McGillUniversity,RutherfordPhysicsBuilding,3600UniversityStreet,Montreal,QuebecH3A2T8,Canada 3PhysicsDepartmentandColumbiaAstrophysicsLaboratory,ColumbiaUniversity,NewYork,NY10027,USA 4TheInstituteofSpaceandAstronauticalScience/JAXA,3-1-1Yoshinodai,Chuo-ku,Sagamihara252-5210,Japan 5CahillCenterforAstronomyandAstrophysics,CaliforniaInstituteofTechnology,Pasadena,CA91125,USA 6SpaceSciencesLaboratory,UniversityofCalifornia,Berkeley,CA94720,USA 7DTUSpace-NationalSpaceInstitute,TechnicalUniversityofDenmark,Elektrovej327,DK-2800Lyngby,Denmark 8LawrenceLivermoreNationalLaboratory,Livermore,CA94550,USA 9JetPropulsionLaboratory,CaliforniaInstituteofTechnology,Pasadena,CA91109,USA 10NASAGoddardSpaceFlightCenter,Greenbelt,MD20771,USA Received2013October31;accepted2013December17;published2014January17 ABSTRACT AE Aquarii is a cataclysmic variable with the fastest known rotating magnetized white dwarf (P = 33.08 s). spin Compared to many intermediate polars, AE Aquarii shows a soft X-ray spectrum with a very low luminosity (L ∼ 1031 erg s−1). We have analyzed overlapping observations of this system with the NuSTAR and the Swift X X-ray observatories in 2012 September. We find the 0.5–30 keV spectra to be well fitted by either an optically thinthermalplasmamodelwiththreetemperaturesof0.75+0.18,2.29+0.96,and9.33+6.07 keV,oranopticallythin −0.45 −0.82 −2.18 thermal plasma model with two temperatures of 1.00+0.34 and 4.64+1.58 keV plus a power-law component with −0.23 −0.84 photonindexof2.50+0.17.Thepulseprofileinthe3–20keVbandisbroadandapproximatelysinusoidal,witha −0.23 pulsedfractionof16.6%±2.3%.Wedonotfindanyevidenceforapreviouslyreportedsharpfeatureinthepulse profile. Key words: accretion, accretion disks – novae, cataclysmic variables – stars: individual (AE Aquarii) – whitedwarfs–X-rays:stars Online-onlymaterial:colorfigures oftypicalIPs.ItsX-rayspectrumhasbeenmodeledasemission 1.INTRODUCTION fromanopticallythinthermalplasmawithseveraltemperature AE Aquarii (hereafter AE Aqr) is a cataclysmic variable components, similar to those seen from other IPs. However, binary system classified as a member of the DQ Herculis or the highest temperature found in such models was 4.6 keV intermediate polar (IP) class (Patterson 1994). It is a non- (Itohetal.2006),whichissignificantlylowerthantheaverage eclipsingclosebinarysystematadistanceof102+42 pc(Fried- kT ≈22keVfoundinthe22IPsdetectedbyINTEGRAL(Landi −23 jung1997),consistingofamagneticwhitedwarf(primary)anda etal.2009).Forthesereasons,themechanismandlocationof K4–5Vstar(secondary)withanorbitalperiod,P =9.88hr. theX-rayemissionarestilluncertain(e.g.,Choietal.1999;Itoh orbit The 33.08 s period makes AE Aqr the fastest-spinning mag- etal.2006;Mauche2009). netic white dwarf. The pulsations were originally discovered Another intriguing feature reported by a Suzaku observa- in the optical (Patterson 1979), then confirmed in soft X-rays tion in 2005 is that AE Aqr may emit non-thermal hard (Pattersonetal.1980)andtheultraviolet(Eracleousetal.1994). X-rays with a very narrow pulse profile at the spin period IntheDQHerculisclass,thewhitedwarfisgenerallythought (Terada et al. 2008), suggesting that AE Aqr may acceler- to possess a magnetic field (B ∼ 105–7 G) strong enough to ate charged particles in a fashion similar to rotation-powered channel the accretion flow from the secondary to the poles of pulsars (e.g., Kaspi et al. 2006 for a review). However, the thewhitedwarf.Accordingly,hardX-raysareproducedbythe Suzaku observation in 2006 did not reproduce the earlier re- shock-heatedgas,whichreachestemperaturesofafewtensof sult,leavingthedetectionofnon-thermalX-raysfromAEAqr keVnearthesurface.Thewhitedwarfphotosphere,heatedby uncertain. thehardX-rays,emitsultravioletlight.Theseemissionsexhibit In this paper, we present broad-band X-ray observations of spin modulation caused by the varying aspect of the accret- AE Aqr obtained with NuSTAR and Swift. Section 2 details ing poles as the white dwarf rotates (see Patterson 1994 for a the observations, data reduction, and background modeling. review). These more sensitive observations in the hard X-ray band can AEAqrstandsoutasanunusualmemberoftheDQHerculis helpmeasurethemaximumtemperatureofthethermalplasma class. It displays strong flares of broad-band emission, from in AE Aqr and test the presence of any beamed non-thermal radio to X-rays. In addition, possible TeV γ-ray flares have component.WedescribethespectralmodelinginSection3and been reported (Meintjes et al. 1994), although they have not the timing analysis in Section 4. The results and a possible yetbeenconfirmedwithmorerecentTeVCerenkovtelescopes interpretationarediscussedinSection5.Throughoutthispaper, (Mauche et al. 2012). The persistent X-ray luminosity of AE allerrorsaregivenatthe90%confidencelevelunlessotherwise Aqr(∼1031ergs−1)istwoordersofmagnitudelowerthanthat stated. 1 TheAstrophysicalJournal,782:3(7pp),2014February10 Kitaguchietal. 10-2 10-3 1 -V e k 1 -Counts s10-4 -1-1 keVs10-3 s 10-4 1.4 nt u o o C ati 1 R 10-5 0.6 3 4 5 7 10 20 30 40 50 70 Energy (keV) 3 4 5 7 10 20 30 40 Figure1.Comparisonof(red)thebackgroundmodelproducedbynuskybgd Energy (keV) with (black) the actual blank-sky spectrum in the AE Aqr observation. The bottompanelshowsthespectralratiooftheactualdatatothemodel. Figure2.X-rayspectraofAEAqrobservedwiththetwoNuSTARtelescopes, (Acolorversionofthisfigureisavailableintheonlinejournal.) (black)FPMAand(red)FPMB.Thecrosspointsandsolidhistogramsarethe background-subtractedspectraandbackgroundmodels,respectively. (Acolorversionofthisfigureisavailableintheonlinejournal.) 2.OBSERVATIONSANDDATAREDUCTIONS regions of the same observation, and then generates a back- 2.1.NuSTAR groundspectrumwithinanyregionswiththebest-fitparameters 2.1.1.ObservationLogandDataScreening (Wiketal.inpreparationforamoredetailed). Three blank-sky spectra for each telescope were extracted WeobservedAEAqrwithNuSTAR(Harrisonetal.2013),the from annular regions of radius 120(cid:5)(cid:5)–270(cid:5)(cid:5), 270(cid:5)(cid:5)–370(cid:5)(cid:5), and first focusing hard X-ray (3–79 keV) telescopes in orbit, from 370(cid:5)(cid:5)–740(cid:5)(cid:5)centeredonAEAqrtomodelthebackgroundspectra 2012September4,19:20UTtoSeptember7,18:50UT.NuSTAR of the two NuSTAR telescopes. The blank-sky spectra were was operated in its default mode throughout the observation. well fitted by the model with χ2/dof = 1217.6/1124. The The acquired data with the observation ID of 30001120 were background count spectrum in the source region was modeled processedandscreenedinthestandardwayusingtheNuSTAR withanexposure20timeslongerthantheactualonetoreduce pipelinesoftware,NuSTARDASversion1.1.1,withtheNuSTAR statistical errors. In order to verify the background model, the calibration database (CALDB) version 20130509. The data backgroundspectruminablank-skyregion3.(cid:5)5fromthesource werefilteredforintervalsofhighbackground,includingEarth in a northeasterly direction was simulated and compared to occultationsandSouthAtlanticAnomaly(SAA)passages.This the actual spectrum extracted from the same region. The ratio resulted in a total of 125 ks of dead-time corrected exposure oftheactualspectrumtothemodelinFigure1wasfittedbya time. constantfactorof0.98±0.04withχ2/dof =15.7/27,showing All photon arrival times were corrected to the solar sys- the background model is consistent with the actual spectrum tem barycenter using the JPL DE200 ephemeris and the withinstatisticalerrors. Chandra -derived coordinates (20:40:09.185, −00:52:15.08; Figure 2 shows the AE Aqr spectra obtained with the two J2000)whichhavesub-arcseconduncertainties.Thesourcepho- NuSTAR hard X-ray telescopes (FPMA and FPMB) and the tonswereextractedfromacircularregionofradius1.(cid:5)0centered background models. The highly ionized iron line can be seen onAEAqr.Thesourcespectraweregroupedwithaminimumof around 6.7 keV. The total count rate in the 3–30 keV energy 50countsperbin.Thebackgroundspectraweregeneratedwith bandafterthebackgroundsubtractionanddead-timecorrection the background modeling tool, nuskybgd (D. R. Wik et al. in is(9.2±0.1)×10−2 countss−1. preparation). A detailed description of the background model- ingisgiveninSection2.1.2.Thetelescoperesponsefiles,ARF andRMFwerealsogeneratedbyNuSTARDAS. 2.2.Swift 2.1.2.BackgroundSpectralModeling Swift(Gehrelsetal.2004)observedAEAqrsimultaneously Since AE Aqr is a faint source in hard X-rays, the NuSTAR with NuSTAR from 2012 September 6, 06:07 to 07:53 UT. background spectrum must be carefully subtracted from the The Swift X-ray telescope (XRT; Burrows et al. 2005) was sourcespectrum.Blank-skyobservationsshowthebackground operated in the photon counting mode during the observation. rate varies in an energy dependent way with detector position TheXRTdata(observationID00030295037)wereprocessedin by a factor of ∼2 (Harrison et al. 2013). In addition, the thestandardmannerwithxrtpipeline(ver.0.12.6)inHEASoft internal background rate varies in time due to changes in the (ver. 6.13). The total exposure time after the data screening cosmicradiationintensityassociatedwiththegeomagneticcut- is 1.54 ks. The source photons were extracted from a circular offrigidity,aswellastheelapsedtimesinceSAApassage. regionofradius1.(cid:5)2centeredonAEAqr.Thesourcespectrum In order to perform accurate background subtraction, the wasrebinnedtohaveatleast20countsperbin.Thebackground background spectrum was empirically modeled using the spectrumwasextractedfromanannularregionofradius1.(cid:5)7–7.(cid:5)0 NuSTAR background fitting and modeling tool, nuskybgd centered on the source and was scaled by an area ratio of the (revision52).nuskybgdfitsblank-skyspectrainuser-selected sourceregiontothebackgroundregion. 2 TheAstrophysicalJournal,782:3(7pp),2014February10 Kitaguchietal. Table1 Best-fitParametersoftheThree-temperatureModel 0.1 Parameter Unit Value V1− 0.01 e ZNkTH1 (10((2Zk0e(cid:6)cVma))−2) 00..<77658−−++20000...2..11147835 k s stnuo1− 1100−−34 F1b (10−12ergs−1cm−2) 1.97−+00..9845 C 10−5 kFT22b (10−12e(rkgeVs−)1cm−2) 32..5239−−++0000....95866524 10−46 (a) APEC x 3 kT3 (keV) 9.33−+62..0178 2 F3b (10−12ergs−1cm−2) 2.26−+00..9996 χ 0 CFPMB/CFPMAc 1.00±0.03 −2 CXRT/CFPMAd 0.94−+00..1292 −4 χ2/dof 247.1/201 0.5 1 2 5 10 20 Energy (keV) Notes. aSolarabundancefromWilmsetal.(2000),inwhichthesolarironabundance relativetohydrogenis2.69×10−5.NotethatthesolarabundancefromAnders& 0.1 Grevesse(1989),inwhichthesolarironabundanceis4.68×10−5,isemployed V−1 0.01 e inotherAEAqrarticles(e.g.,Itohetal.2006;Teradaetal.2008;Oruru& k Meintjes2012). s −1 10−3 bUnabsorbedfluxin0.5–10keV. nts 10−4 cCross-normalizationfactorbetweenNuSTAR/FPMBandNuSTAR/FPMA. u o dCross-normalizationfactorbetweenSwift/XRTandNuSTAR/FPMA. C 10−5 (b) APEC x 2 + Power law 10−6 4 3.SPECTRALANALYSIS 2 3.1.SpectralFittingwithMulti-temperatureModels χ 0 −2 The X-ray spectra of AE Aqr observed with ASCA (Choi −4 et al. 1999), XMM-Newton (Itoh et al. 2006), and Suzaku 0.5 1 2 5 10 20 (Teradaetal.2008)havebeenmodeledusinganopticallythin Energy (keV) thermalplasmaemissionmodelwithafewdifferenttemperature Figure3.X-rayspectraof(green)Swift/XRT,(black)NuSTAR/FPMA,and components in the same way as for other IPs. Therefore in (red) NuSTAR/FPMB overlaid with (a) the best-fit three-temperature APEC the joint fitting of Swift and NuSTAR spectra, we adopted an modeland(b)two-temperaturemodelwithapower-lowemission.Themodel emission model from a collisionally ionized diffuse plasma, componentsaredrawnwithdottedlines.Thebottompanelsshowfitresiduals intermsofσ witherrorbarsofsizeone. APEC(Smithetal.2001).EachAPECmodelwasconstrainedto (Acolorversionofthisfigureisavailableintheonlinejournal.) have common metal abundances relative to solar from Wilms etal.(2000),butwasallowedtohavedifferenttemperaturesand normalizations. We use the tbabs model (Wilms et al. 2000) Although other IP spectra commonly contain a strong neu- to account for interstellar and self-absorption. In addition, the tral (or low-ionized) iron line around 6.4 keV with a compa- cross-normalizationfactorsofXRT/FPMAandFPMB/FPMA rable equivalent width to highly ionized iron emission around wereallowedtovary. 6.7keV(e.g.,Ezuka&Ishida1999;Yuasaetal.2010),wedo InordertodeterminethenumberofAPECcomponentswith notdetectthisfeatureclearlyintheAEAqrspectrum.Infact, differenttemperatures,weaddednewAPECcomponentsoneby XMM-Newtonspectraconstrainedtheupperlimittothe6.4keV one until the fit was not significantly improved, as determined equivalentwidthtobe<61eV(Itohetal.2006).Wetriedtoadd by the F-test. Improvement of the fit is significant with the a narrow Gaussian emission model with a center energy fixed chanceprobabilityof(cid:7)1%untilthenumberofcomponentsis at6.4keVtothethree-temperatureAPECmodel,andsimultane- increasedtothree,whereaslittleimprovementwasfoundbythe ouslyfitthespectraofSwiftandNuSTARwithit.Theintensity addition of a fourth component with the chance probability of oftheneutralironlineis1.33+1.26 ×10−14ergs−1cm−2,which −1.33 8.4%derivedfromtheχ2 reductionfrom247.1(dof=201)to correspondstoanequivalentwidthof37.9+41.7 eV.Ontheother 241.0(dof=199).The0.5–30keVspectraofSwiftandNuSTAR hand,theintensityofthehighlyionizediro−n3l7i.n9earound6.7keV with the best-fit three-temperature APEC models are shown in was determined to be 1.11+0.11 ×10−13 erg s−1 cm−2, or an −0.09 Figure 3(a), the parameters of which are listed in Table 1. equivalentwidthof495+365 eV,withthelinecentralenergyof w9.i3t+h6.p1rekveiVo,usisrecsounlstsidfeorarbrleyfehriegnhceer. tThhaenhpirgehveiostustleymopbesraetruvreed, 6.732+−00..006134keV.Therefo−r5e3,theneutralironlineismuchweaker −2.2 thanthehighlyionizedironline. forthissource(4.6keVfromItohetal.2006),andapproaches the average temperature of 22 keV for any other IPs (Landi 3.2.SpectralFittingwithNon-thermalEmission et al. 2009). The metal abundance, 0.76+−00..1173Z(cid:6), is consistent with that determined by XMM-Newton and Suzaku. The total In order to search for possible non-thermal hard X-rays as luminosityin0.5–10keVis9.3+1.0×1030ergs−1foranassumed suggested by Terada et al. (2008), we simultaneously fitted −2.2 distanceof100pc. the spectra of Swift and NuSTAR with a multi-temperature 3 TheAstrophysicalJournal,782:3(7pp),2014February10 Kitaguchietal. Table2 Best-fitParametersoftheTwo-temperatureModelwithPower-lawEmission (a) 3-10 keV Parameter Unit Value 100 NH (1020cm−2) <82.2 Z (Z(cid:6)) 1.14−+00..6323 kT1 (keV) 1.00−+00..3243 2Z1 F1a (10−12ergs−1cm−2) 2.01−+11..2053 50 kT2 (keV) 4.64−+10..5884 F2a (10−12ergs−1cm−2) 1.86−+00..9861 Γ 2.50−+00..1273 FPLb (10−12ergs−1cm−2) 4.07−+11..6560 0 CFPMB/CFPMA 1.00±0.03 (b) 10-20 keV CXRT/CFPMA 0.81−+00..2129 10 χ2/dof 244.5/201 Notes. aUnabsorbedfluxofeachthermalemissioncomponentin0.5–10keV. 21 bUnabsorbedfluxofthepower-lawmodelin0.5–10keV. Z 5 plus power-law emission model. In the same way as the determination of the number of APEC components described above,wefirstfitthespectrausingasingleAPECwithapower- 0 lawmodel,andthenaddednewAPECmodelsonebyoneuntil 33.04 33.06 33.08 33.1 the fit was not significantly improved as determined by the Period (s) F-test. We find two temperatures with power-law emission is Figure4.Z2periodogramsacquiredwithNuSTARintheenergybandsof(a) thestatisticallyfavoredmodel. 1 3–10and(b)10–20keV. The Swift and NuSTAR spectra with the best-fit model are shown in Figure 3(b), the parameters of which are listed in Table3 Table2.Comparedwiththethree-temperatureAPECmodel,the PulseProfileParametersaforAEAqr fit with the two-APEC model with the power-law emission is EnergyRange PeakPhaseb PulseFraction χ2/dof slightly but not significantly preferred. We cannot distinguish (keV) (%) whether the hard X-ray component detected with NuSTAR is thermalornon-thermalemission. 3–20 0.30±0.02 16.6±2.3 17.2/16 3–6 0.31±0.04 14.8±3.1 11.7/10 6–10 0.31±0.03 19.9±3.7 8.4/10 4.TIMINGANALYSIS 10–20 0.22±0.10 14.2±8.0 2.5/4 4.1.SpinPeriodDetermination Notes. aBest-fitparametersforafittotheNuSTARpulseprofilespresented We examined the hard X-ray pulse profile for evidence of in Figure 5 using a sinusoid model with a constant offset, C: the narrow pulsation reported by Suzaku (Terada et al. 2008). C(1.0+fPcos(2π(φ−φ0))), where φ0 is the peak phase and fP Previous measurements have shown that soft X-rays below isthepulsefraction. 10 keV are sinusoidally modulated with a single peak in the bPhase0correspondstothatofFigure5. spinphaseofthewhitedwarf(e.g.,Mauche2006).Inthiscase, theZ12-statistic,orRayleightest(Buccherietal.1983)ismore 4.2.PulseProfile sensitive to determine the spin period than the epoch folding techniquewiththeχ2test(Leahyetal.1983). Figure5presentsthepulseprofilesinseveralenergybands, Figure 4(a) shows the Z2 periodogram using NuSTAR folded on the best determined period, with the background 1 3–10 keV data with barycenter-corrected time. We found the subtracted using the model of Section 2.1.2. The profiles are peakat33.0769±0.0004swithZ2 =124.1.Withintheuncer- wellrepresentedbysinusoidalfunctions(seeTable3).Wefind 1 tainty,theperiodisconsistentwithpreviouslymeasuredvalues no evidence for significant variation in the phase or relative intheopticalband(Patterson1979;deJageretal.1994),andat modulationamplitude(orpulsefraction)withenergy.Themean X-rayenergies(Choietal.1999;Choi&Dotani2006;Mauche pulse fraction, 16.6%±2.3%, is consistent with that found in 2006;Teradaetal.2008;Oruru&Meintjes2012). thequiescentphasefromXMM-Newtonobservations(Itohetal. Wealsosearchedforasignalinthe10–20keVenergyband. 2006;Choi&Dotani2006). The hard X-rays above 20 keV were ignored to improve the Wesearchedforanarrowpeakinthepulseprofileasreported signal-to-noise ratio since the source rate from AE Aqr falls byTeradaetal.(2008)basedontheSuzakuobservationof2005. below that of the background in that energy range (Figure 2). Weinvestigatedthepulseprofileusingdifferentnumbersofbins Evidence for a pulsation was found also in this band at the percyclerangingfrom5to50.Furthermore,weexaminedthe expected pulse period, 33.0765±0.0016 s, with Z2 = 11.1, profile using 29 bins per cycle as per the Suzaku analysis and 1 corresponding to a null hypothesis probability of 0.38%, or reproduced 1000 pulse profiles, each having a different start 2.9σ detectionsignificance(Figure4(b)). timeatphasezero.Thenwelookedforanysignificantoutliers 4 TheAstrophysicalJournal,782:3(7pp),2014February10 Kitaguchietal. fluxin2005(Teradaetal.2008).Thederivedpower-lawindex, 2.5±0.2,isalsoinconsistentwiththeSuzakuvalue,1.1±0.6, 0.1 and is steeper than those found for rotation-powered pulsars, which range from 0.6 to 2.1 (Gotthelf 2003; Kaspi et al. 1 2006).HerewenotethatthepreviouslyreportedSuzakufitting -s s 0.09 errorsdidnotincludethesystematicuncertaintyoftheSuzaku/ nt HXD-PIN background model because the authors described u o C 0.08 that the hard X-ray flux was consistent with the narrow pulse flux in the spin profile. However, the systematic uncertainty, 3% (Fukazawa et al. 2009), is comparable to the background- 0.07 subtractedcountrate,andthereforeshouldnotbeignored(see (a) 3-20 keV Figures6and7ofTeradaetal.2008forcomparisonoftherates). Itispossiblethatthediscrepancyofthenon-thermalemission parameters is due to the neglected systematic error in Terada etal.(2008). Furthermore,wefindnoevidenceofthenarrowpulseprofile 1 0.05 -s withapulsefractionofnearly100%andadutycycleof∼0.1 s nt reported by Terada et al. (2008). In contrast with the Suzaku u o result, we marginally detect the sinusoidal modulation in the C 10–20keVenergybandatasignificancelevelof2.9σ.Inorder 0.04 todeterminewhetherornotNuSTARissensitivetothenarrow pulse profile suggested by Suzaku, we estimated its count rate (b) 3-6 keV considering the NuSTAR telescope response and assuming the Suzakuresult.Theexpected10–20keVcountratewithinaduty 0.04 cycle of 0.1 is 0.086±0.003 count s−1, 10 times higher than thetime-averagedratemeasuredwithNuSTAR,indicatingthat NuSTARwouldhavebeencapableofsignificantlydetectingthe -1s 0.035 narrowpulsation. nts ApossibleexplanationofthedifferencebetweentheSuzaku u and NuSTAR results is that AE Aqr may have varied and Co 0.03 its hard X-ray flux decreased. However, the soft X-ray flux determined with NuSTAR in 2012 is consistent with previous observationsmadewithSuzakuin2005(seeTable1),withinthe 0.025 cross-normalizationfactorofNuSTARtoSuzaku/XISof∼15% (c) 6-10 keV (basedonsimultaneousobservationsofcalibrationtargets).Itis difficulttoexplainamechanismonlyinhardX-raysthatwould 0.012 makethefluxvarywithoutcorrelationofthermalsoftX-rays,a partofwhichalsomodulateswiththespinperiod. -1s 5.2.PossibleInterpretationoftheObservedEmission s 0.01 unt Two energy sources could, inprinciple, power the observed Co X-rayluminosityLX ∼1031ergs−1:liberationofgravitational energyofaccretingmatterandtherotationalenergyofthewhite 0.008 dwarf. The white dwarf was reported to spin down with a rate (d) 10-20 keV P˙ =5.64×10−14ss−1(deJageretal.1994),whichcorresponds 0 0.5 1 1.5 2 tospindownpowerLsd ∼1034ergs−1 (cid:8)LX.Thissuggeststhe Phase possibilitythattheX-rayemissionofthewhitedwarfisfedby Figure 5. AE Aqr pulse profiles folded on the best period of 33.0769 s in rotation(Ikhsanov&Beskrovnaya2012).Therotation-induced (a) 3–20, (b) 3–6, (c) 6–10, and (d) 10–20 keV energy bands. The dashed electric field could accelerate particles to high Lorentz factors curvesshowthebest-fitsinusoidalmodels.Theepochofphase0corresponds ifthemagnetosphereisplasma-starvedandtheelectricfieldis to56174.5inMJD. notscreened.Synchrotronemissionwassuggestedasapossible radiative mechanism in such a model (Terada et al. 2008). It from the sinusoidal model with a 95% confidence level. We is, however, unclear how the observed luminosity, spectrum, concludethatthereisnoevidenceforanadditionalsharppulse and pulse profile observed by NuSTAR would be produced by componentinourdata. relativisticparticles.Theefficiencyofsynchrotronemissionis expected to be small, since the particles are accelerated along 5.DISCUSSION themagneticfieldlines.Emissionfromacceleratedparticlesis 5.1.ComparisonofNon-thermalHardX-RayEmission alsoexpectedtobestronglybeamedalongthefieldlines,which withPreviousSuzakuResults isinconsistentwiththeobservedbroadpulseprofileinFigure5. Emission powered by accretion is a plausible mechanism. The 10–30 keV flux determined with NuSTAR is 5.0+−11..34 × The standard model of accreting IPs involves an accretion 10−13ergs−1cm−2,abouthalfofthebest-fitSuzaku/HXD-PIN column heated by the accretion shock and cooled by thermal 5 TheAstrophysicalJournal,782:3(7pp),2014February10 Kitaguchietal. bremsstrahlung(Aizu1973).ItexplainswelltheX-rayspectra effect (Wynn et al. 1997), whereby most of the mass lost by ofmanyIPs(seeCropperetal.1999forGingadata,Suleimanov thesecondaryisbeingflungoutofthebinarybythemagnetic et al. 2005 for RXTE, Landi et al. 2009 for Swift/XRT and field of the rapidly rotating white dwarf. Norton et al. (2004, INTEGRAL/IBIS, Brunschweiger et al. 2009 for Swift/BAT, 2008) performed numerical simulations of accretion flows in and Yuasa et al. 2010 for Suzaku). The main parameters of magnetic cataclysmic variables with wide ranges of magnetic the model are the free-fall velocity v = (2GM /R )1/2 moment,μ = 1032–36 Gcm3,andP /P = 10−3–1.They and the accretion rate per unit area, aff= M˙/4πRW2Df,WwDhere demonstrated that IPs with a very smspainll raotrbioit of P /P < WD spin orbit f is the fraction of the white dwarf surface occupied by the 10−2 like AE Aqr (P /P = 9.3×10−4) show magnetic spin orbit accretion column. The standard model assumes that the shock propellereffects.Therefore,thetallaccretioncolumncouldbe iTshreadshiaotcivkeaaltnidtudpeinhne(cid:2)d tvotbtrh,ewwhheriteevdw=avrfffs/u4rf∼ac1e,08hc(cid:7)msR−W1Dis. wfoormuleddailnsoIPbsewexitphecPtespdin/toPohrbaivte<lo1w0−X2-,rtahyoulugmhisnuocshitiseys,sttehmuss thepostshockvelocityoftheaccretinggasandtbr ∼0.3/a sis making them more challenging to identify. AE Aqr is the thebremsstrahlungcoolingtimescale(withaexpressedinunits only currently known example. It is also possible that nearly ofgs−1cm−2).ThenetaccretionrateM˙ isdeterminedfromthe synchronoussystemssimilartoEXHydraecouldpossessatall X-TrahyislusmceinnoasriiotywLoXul=dGimMplWyDMM˙˙/R∼W2D.× 1014 g s−1 for AE accocnrseitsitoenntcwoliuthmtnh.eTphicetulorewoafcrcinregt-iloinkeraatcecrientiosuncdhisscyusstesemdsbiys Aqr, and the condition a (cid:8) 10−2 g s−1 cm−2 is satisfied if Nortonetal.(2008). f (cid:7)10−3.Suchasmallfmaybepossibleif,e.g.,theaccretion flow in AE Aqr is concentrated in a thin wall of the accretion 6.SUMMARY column (as Basko & Sunyaev 1976 suggested for accreting neutronstars).However,ontheassumptionofMWD ∼0.7M(cid:6) dwWaref AhaEveAqarnaolybzseerdveXd-wraiythdNatuaSToAfRthfeorm1a2g5nektsizaenddwSwhiitfet determined with the optical measurements (e.g., Watson et al. for 1.5 ks. The 0.5–30 keV spectra are well characterized 2006;Echevarr´ıaetal.2008),ourcalculationsofthespectrum by either an optically thin thermal plasma model with three predictedbythestandardaccretionmodeldonotprovideagood temperatureswiththehighesttemperatureof9.33+6.07 keV,or fitfortheNuSTARdata—theobservedspectrumwiththehighest −2.18 temperatureof9.3+6.1 keVissofterthanthepredictedpostshock anopticallythinthermalplasmamodelwithtwotemperatures −2.2 plusapower-lawcomponentwithphotonindexof2.50+0.17.The temperatureof29keV. −0.23 3–20 keV pulse profile shows a sinusoidal modulation, with a Twomodificationsmightmaketheaccretionmodelconsistent pulsed fraction of 16.6%±2.3%. We find no evidence for the withNuSTARobservations.(1)TheshockaltitudehinAEAqr previouslyreportedsharpfeatureinthepulseprofile. couldbesignificant,perhapsevencomparabletothewhitedwarf TheobservedsoftX-rayspectrumwiththebroadpulseprofile radius.Suchtallaccretioncolumnswithlowaccretionrateswere is hard to explain as a result of rotation-powered emission. recently studied by Hayashi & Ishida (2014). The shallower Instead accretion-powered emission is more likely, although gravitationalpotentialatthehighaltituderesultsinareduction the observed spectrum is softer than predicted by the standard ofthepostshocktemperaturebelow20keVtypicalofotherIPs. accretion column model. We suggest two modifications of the ThetallaccretioncolumninAEAqrwaspreviouslyproposed standard model to explain the AE Aqr spectrum: the shock asthesourceofthelargehotspotsonthewhitedwarfsurface altitude could be high, comparable to the white dwarf radius, inferredfromobservationsbytheHubbletelescope(Eracleous andcyclotronemissionwithB >106Gcouldadditionallycool et al. 1994) (2) The X-ray spectrum emitted by the accretion downtheaccretionplasma.Detailedcalculationsofsuchmodels column could be affected by additional radiative losses due to willhopefullyreproducethespectrumandpulseprofileofAE cyclotronemissionintheinfraredband.Cyclotronemissionis importantwhenB > 106 G,andmaycarryawayasignificant Aqrwiththewhitedwarfmassconsistentwiththatdetermined withtheopticalmeasurements. partoftheaccretioncolumnenergy,especiallyatlargealtitudes wheretheplasmahasthehighesttemperature.Thiseffectcould This work was supported under NASA Contract No. explaintherelativelysoftextendedspectrumofAEAqrabove NNG08FD60C,andmadeuseofdatafromtheNuSTARmission, 3 keV. Detailed calculations of such models will be presented aprojectledbytheCaliforniaInstituteofTechnology,managed elsewhere. by the Jet Propulsion Laboratory, and funded by the National In addition to the low X-ray luminosity and soft spectrum, AeronauticsandSpaceAdministration.WethanktheNuSTAR AE Aqr has another special feature—weak absorption of soft X-rays,whichcorrespondstoN <8×1021cm−2,lowerthan Operations, Software and Calibration teams and the Swift Op- H theN >1022cm−2observedinmanyIPs(e.g.,Ezuka&Ishida erations team for support with the execution and analysis of H theseobservations.ThisresearchhasmadeuseoftheNuSTAR 1999;Yuasaetal.2010).Thisfeaturecouldbeexplainedbythe Data Analysis Software (NuSTARDAS) jointly developed by lowerdensityoftheaccretioncolumn,andneedstobefurther theASIScienceDataCenter(ASDC,Italy)andtheCalifornia investigatedwithdetailedmodels.AlsonotethatAEAqrdoes InstituteofTechnology(Caltech,USA)andtheXRTDataAnal- not show evidence for a neutral (or weakly ionized) iron line. ysisSoftware(XRTDAS)developedundertheresponsibilityof AmongotherIPspreviouslystudiedinthehardX-rayband,EX HydraehasthelowestluminosityLX ∼ 1032 ergs−1 andalso ASDC. T.K. was supported by Japan Society for the Promo- the lowest column with N (cid:2) 1022 cm−2 (Suleimanov et al. tion of Science (JSPS) Grant-in-Aid for Young Scientists (B) H (No.24740185). 2005).Theshapeofits3–30keVspectrumisnotfarfromthat Facilities:NuSTAR,Swift ofAEAqr,althoughEXHydraeisanearlysynchronoussystem with an orbital period of 98.3 min and a long spin period of REFERENCES 67.0minutes(Belleetal.2005),unlikeAEAqr. The much smaller accretion rate of AE Aqr compared to Aizu,K.1973,PThPh,49,1184 many other IPs is thought to be due to the magnetic propeller Anders,E.,&Grevesse,N.1989,GeCoA,53,197 6 TheAstrophysicalJournal,782:3(7pp),2014February10 Kitaguchietal. Basko,M.M.,&Sunyaev,R.A.1976,MNRAS,175,395 Kaspi,V.M.,Roberts,M.S.E.,&Harding,A.K.2006,inCompactStellar Belle,K.E.,Howell,S.B.,Mukai,K.,etal.2005,AJ,129,1985 X-RaySources,ed.W.Lewin&M.vanderKlis(Cambridge:Cambridge Brunschweiger, J., Greiner, J., Ajello, M., & Osborne, J. 2009, A&A, Univ.Press),279 496,121 Landi,R.,Bassani,L.,Dean,A.J.,etal.2009,MNRAS,392,630 Buccheri,R.,Bennett,K.,Bignami,G.F.,etal.1983,A&A,128,245 Leahy,D.A.,Elsner,R.F.,&Weisskopf,M.C.1983,ApJ,272,256 Burrows,D.N.,Hill,J.E.,Nousek,J.A.,etal.2005,SSRv,120,165 Mauche,C.W.2006,MNRAS,369,1983 Choi,C.-S.,&Dotani,T.2006,ApJ,646,1149 Mauche,C.W.2009,ApJ,706,130 Choi,C.-S.,Dotani,T.,&Agrawal,P.C.1999,ApJ,525,399 Mauche,C.W.,Abada-Simon,M.,Desmurs,J.-F.,etal.2012,MmSAI,83,651 Cropper, M., Wu, K., Ramsay, G., & Kocabiyik, A. 1999, MNRAS, Meintjes,P.J.,deJager,O.C.,Raubenheimer,B.C.,etal.1994,ApJ,434,292 306,684 Norton,A.J.,Butters,O.W.,Parker,T.L.,&Wynn,G.A.2008,ApJ,672,524 de Jager, O. C., Meintjes, P. J., O’Donoghue, D., & Robinson, E. L. 1994, Norton,A.J.,Wynn,G.A.,&Somerscales,R.V.2004,ApJ,614,349 MNRAS,267,577 Oruru,B.,&Meintjes,P.J.2012,MNRAS,421,1557 Echevarr´ıa, J., Smith, R. C., Costero, R., Zharikov, S., & Michel, R. Patterson,J.1979,ApJ,234,978 2008,MNRAS,387,1563 Patterson,J.1994,PASP,106,209 Eracleous,M.,Horne,K.,Robinson,E.L.,etal.1994,ApJ,433,313 Patterson, J., Branch, D., Chincarini, G., & Robinson, E. L. 1980, ApJL, Ezuka,H.,&Ishida,M.1999,ApJS,120,277 240,L133 Friedjung,M.1997,NewA,2,319 Smith,R.K.,Brickhouse,N.S.,Liedahl,D.A.,&Raymond,J.C.2001,ApJL, Fukazawa,Y.,Mizuno,T.,Watanabe,S.,etal.2009,PASJ,61,17 556,L91 Gehrels,N.,Chincarini,G.,Giommi,P.,etal.2004,ApJ,611,1005 Suleimanov,V.,Revnivtsev,M.,&Ritter,H.2005,A&A,435,191 Gotthelf,E.V.2003,ApJ,591,361 Terada,Y.,Hayashi,T.,Ishida,M.,etal.2008,PASJ,60,387 Harrison,F.A.,Craig,W.W.,Christensen,F.E.,etal.2013,ApJ,770,103 Watson,C.A.,Dhillon,V.S.,&Shahbaz,T.2006,MNRAS,368,637 Hayashi,T.,&Ishida,M.2014,MNRAS,inpress Wilms,J.,Allen,A.,&McCray,R.2000,ApJ,542,914 Ikhsanov,N.R.,&Beskrovnaya,N.G.2012,ARep,56,595 Wynn,G.A.,King,A.R.,&Horne,K.1997,MNRAS,286,436 Itoh,K.,Okada,S.,Ishida,M.,&Kunieda,H.2006,ApJ,639,397 Yuasa,T.,Nakazawa,K.,Makishima,K.,etal.2010,A&A,520,A25 7