Mon.Not.R.Astron.Soc.000,1–9(2008) Printed9January2009 (MNLATEXstylefilev2.2) SDSS J080449.49+161624.8: A peculiar AM CVn star from a colour-selected sample of candidates G.H.A. Roelofs,1(cid:63) P.J. Groot,2 D. Steeghs,3 A. Rau,4 E. de Groot,2 T.R. Marsh,3 G. Nelemans,2 J. Liebert,5 P. Woudt6 1Harvard–Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA 9 2Department of Astrophysics, Radboud University Nijmegen, PO Box 9010, 6500 GL Nijmegen, The Netherlands 0 3Department of Physics, University of Warwick, Coventry CV4 7AL, UK 0 4Astronomy Department, California Institute of Technology, Pasadena, CA 91125, USA 2 5Steward Observatory, University of Arizona, Tucson, AZ 85721, USA n 6Department of Astronomy, University of Cape Town, Rondebosch 7700, South Africa a J 9 Accepted...Received9January2009 ] h ABSTRACT p We describe a spectroscopic survey designed to uncover an estimated ∼40 AM CVn - o stars hiding in the photometric database of the Sloan Digital Sky Survey (SDSS). We r haveconstructedarelativelysmallsampleofabout1,500candidatesbasedonacolour t s selection, which should contain the majority of all AM CVn binaries while remaining a small enough that spectroscopic identification of the full sample is feasible. [ We present the first new AM CVn star discovered using this strategy, 2 SDSSJ080449.49+161624.8,theultracompactbinarynatureofwhichisdemonstrated v using high-time-resolution spectroscopy obtained at the Magellan telescopes at Las 4 Campanas Observatory, Chile. A kinematic ‘S-wave’ feature is observed on a period 7 P = 44.5±0.1min, which we propose is the orbital period, although the present orb 9 data cannot yet exclude its nearest daily aliases. 3 The new AM CVn star shows a peculiar spectrum of broad, single-peaked helium . 1 emissionlineswithunusuallystrongseriesofionisedhelium,reminiscentofthe(inter- 1 mediate) polars among the hydrogen-rich Cataclysmic Variables. We speculate that 8 SDSSJ0804+1616 may be the first magnetic AM CVn star. The accreted material 0 appears to be enriched in nitrogen, to N/O(cid:38)10 and N/C>10 by number, indicating : v CNO-cycle hydrogen burning, but no helium burning, in the prior evolution of the i donor star. X Key words: stars: individual: SDSSJ080449.49+161624.8 – binaries: close – white r a dwarfs – novae, cataclysmic variables – accretion, accretion discs 1 INTRODUCTION detailedobservationalstudiesofknownsystems(e.g.Roelofs etal.2006b,2007a)andmoredetailedphysicalmodels(e.g. The AM CVn stars are a small, but growing class of ultra- Deloye et al. 2005, 2007; Bildsten et al. 2006, 2007; Yun- compactbinariesthatconsistofawhitedwarfaccretoranda gelson 2008), but in large part also due to an increase in (semi-)degenerate helium-transferring donor star. They are the number of objects known (e.g. Woudt & Warner 2003; characterized by their short orbital periods (below the or- Woudtetal.2005;Roelofsetal.2005;Andersonetal.2005, bitalperiodminimumforhydrogen-richdonors)andthelack 2008). The Sloan Digital Sky Survey (SDSS; York et al. of hydrogen in their spectra. There are currently 22 mem- 2000) has proven to be an important resource for the dis- bers, including the new system reported here and the two covery of new AM CVn stars, with six new systems so far. ultrashort-periodcandidatesHMCnc(RXJ0806+15,Israel More importantly, since the SDSS sample is relatively well et al. 2002) and V407 Vul (RXJ1914+24, Cropper et al. defined,thosesixsystemsprovideinformationabouttheun- 1998). A recent overview is given by Nelemans (2005). derlyingGalacticpopulationthatisdifficulttoobtainfrom Our understanding of the AM CVn stars has been in- theratherrandomindividualdiscoveriesfromthepre-SDSS creasing rapidly in the last few years. This is due to more era (Roelofs, Nelemans & Groot 2007b). We have shown in Roelofs et al. (2007b) that the pho- (cid:63) E-mail:[email protected] tometric database of the SDSS, Phase I (SDSS-I, or Data (cid:13)c 2008RAS 2 Roelofs et al. Release5)shouldcontainatotalofabout50AMCVnstars down to a magnitude of g = 21 (but with substantial un- certainty of a factor 2–3 due to, in part, the small sample of six systems from the spectroscopic database). We have started an extensive observing program using an array of telescopes to uncover this ‘hidden’ population of AM CVn stars. The significance of this sample will not only be to double or triple the total known population, but more im- portantlytoincreasethenumberofsystemsthathavebeen foundin arelativelywell-understoodwaybyan even larger factor,whichwillallowforamuchmoredetailedcomparison of their population with theoretical predictions. The outline of this paper is as follows. In Section 2 we Figure1.ColoursoftheknownAMCVnstarsintheSDSS-DR6 discussthetotalsampleofAMCVncandidatesweselected (stars),togetherwithoursampleofcandidates(dots).Thedashed from the SDSS, Data Release 6 (SDSS-DR6; Adelman- line (BB) shows the blackbody cooling track, and the solid lines (DA,DB)showmodelsequencesforhydrogenandheliumatmo- McCarthyetal.2008).InSection3wediscussourstrategy spherewhitedwarfs,respectively.SDSSJ0804+1616isshownas for spectroscopic identification of this sample, and in Sec- the open star symbol. Arrows indicate reddening vectors for an tion 4 we present the first new AM CVn star found in this extinctionA(g)=0.2. way. We discuss the newly discovered system in Section 5 and conclude in Section 6. 2 SAMPLE SELECTION As shown in Roelofs et al. (2007b), the known AM CVn stars span a range in (u−g,g−r) colour space that lies substantiallyandslightlyabovetheDAandDBwhitedwarf coolingsequences,respectively.TheDAsaremuchredderin u−gduetotheBalmerjumpintheirspectra,whichisabsent in the hydrogen-deficient AM CVns and DBs. See Figure 1 inwhichwehaveplottedthe(u−g,g−r)and(g−r,r−i) colourdiagramsoftheknownAMCVnstars,thecoloursof Figure 2. Distributions of apparent g-band magnitude and theAMCVncandidatesselectedhere,andmodelcoloursof Galacticg-bandextinctionA(g)inoursampleshowninFig.1. the cooling sequences of DA and DB white dwarfs of mass logg = 8.0. The latter were calculated from model spectra bulk of the AM CVns (as well as hot, single white dwarfs) kindly provided by D. Koester. down to an apparent magnitude g = 20.5 in the SDSS will All object colours are the dereddened colours from the be several 100pc, i.e. several dust scale heights, above the SDSS-DR6, which were corrected for the full Galactic red- GalacticPlane.Italsoimpliesthat,forthefaintestsystems, dening according to Schlegel et al. (1998). The assumption we are still covering the AM CVns out to about one scale isthat,atthehighgalacticlatitudesspannedbytheSDSS, height of the Galactic thin disk population (∼300pc) if we almostalloftheAMCVnswillbefurtherfromtheGalactic include objects as faint as g =20.5. At fainter magnitudes, Plane than the scale height of dust in the Galaxy. Models we anticipate that spectroscopic follow-up will become ex- of the AM CVn population (Nelemans et al. 2001, 2004a) ceedingly difficult. showthatthesystemsareexpectedtobeseveralGyroldon Guided by the colours of the known AM CVn stars, average,measuredfromthezero-agemainsequence:similar and attempting to reduce contamination by other sources to old thin-disk main-sequence stars that are characterized asmuchaspossible(seebelow),wehavequeriedtheSDSS- by a Galactic scale height ∼300 pc. Galactic dust, on the DR6photometricdatabasewiththefollowingrelativelysim- otherhand,isknowntohaveascaleheightofabout100pc ple set of criteria (colours are dereddened): (Savage & Mathis 1979). Angular momentum losses due to the emission of gravitational waves will cause an AM CVn 1. Object is a point source binarytoevolvetoanorbitalperiodlongerthan30minutes 2. Object is not saturated in any filter in only ∼100Myr, counted from orbital period minimum. 3. g<20.5 Combined with reasonable assumptions about the star for- 4. u−g<min[0.14,1.35(g−r)+0.32]−σ u−g mation history and the distribution of the delay times be- 5. −0.42+σ <g−r<0.02−σ g−r g−r tween star formation and the birth of AM CVn stars, this 6. −0.33+σ <r−i<0.03−σ r−i r−i implies that ∼98% of the local AM CVn population should where have P >30min. In this ‘long’ orbital period regime the orb q AM CVns are expected (Bildsten et al. 2006) and observed σu−g = σu2+σg2+σE2(u−g) (1) (Thorstensen 2003; Roelofs et al. 2005, 2006a, 2007a) to be and similar for the other colours; we have assumed an un- dominated in the optical by an accreting white dwarf of certainty in the reddening of typical temperature 10,000–20,000K, with a corresponding absolute magnitude of M ≈10−13. This implies that the σ =0.5E(u−g) (2) g E(u−g) (cid:13)c 2008RAS,MNRAS000,1–9 SDSSJ0804+1616: a peculiar new AM CVn star 3 and the uncertainties σ are the photometric errors on blended’inthephotometricdatabase;howeverwechosenot u,g,r,i u,g,r,ifromtheSDSS.Wetaketheuncertaintyinthered- to rely on those flags since a comparable number of seem- dening to be a fraction of the reddening estimate, as sug- ingly fine single point sources were similarly flagged. Our gested by Schlegel et al. (1998), except that we assume a final visual check weeded out a relatively small number of larger fraction to further reduce the influence of sources in 65 sources from the sample. (relatively rare) fields with high reddening scattering into Assuming a completeness of 90%, our final sample of our colour selection. Schlegel et al. (1998) derive an aver- 1,523 candidates should contain ∼40 new AM CVn stars, age scatter σ = 0.16E, but this is for objects that are based on the six AM CVn stars found in the SDSS and the E reddened by the full Galactic value; our sample includes completenessofSDSSspectroscopyderivedinRoelofsetal. Galactic sources at lower-than-full reddening which leads (2007b). Spectroscopic identification of the sample should toalargerscatter.Whilethefractionalerrorshouldthusbe thus result in 1 AM CVn star per ∼40 objects, although inthe0.16<σ /E <1range,theexactchoiceisarbitrary. the number will depend on the exact characteristics of the E The resulting sample is plotted in Figure 1, and the population. Determining these characteristics is the aim of distributionsofg-bandmagnitudeandextinctionareshown the study. in Figure 2. On the blue and sparsely populated side in (u−g,g−r) the colour selection includes the 10.3-minute orbital period AM CVn star ES Ceti (Warner & Woudt 3 SPECTROSCOPIC IDENTIFICATION 2002), the top-left star in both panels of Fig. 1, which is characterized by strong Heii emission lines and a very blue The known long-period systems are all characterized by continuum. The bulk of the population, at orbital periods strong Hei emission lines in the optical, with Hei 5876 >30min, are expected to be cooler than ES Cet, since they equivalent widths ranging from 30˚A in V406 Hya and areexpectedtohaveamuchlowerpresent-dayaccretionrate SDSSJ1240−0159 (P = 34 and 37 min, respectively; orb and have had a cooling time of >100Myr since the short- Roelofsetal.2006a)to90˚AinV396Hya(P =65.1min; orb period, high accretion rate phase (see Bildsten et al. 2006). Ruiz et al. 2001). If we assume that the currently unknown The known long-period systems are indeed observed to be population at P >30min has similar spectral character- orb redder in u−g, g−r and r−i. istics (see Roelofs et al. 2007b for a more detailed discus- Ontheredsideinu−g,carehasbeentakentominimize sion), spectra with a resolution as low as R = 300 and a thecontributionfromtheDAandDBwhitedwarfsequences signal-to-noise ratio (S/N) as low as 7 per pixel will suffice by cutting off the sample parallel to the blackbody cooling to positively identify all AM CVns among the candidates. track, as shown in Figure 1. The number of objects in the Figure 3 shows the spectrum of the new AM CVn star sample obviously increases rapidly with a redder cut-off on SDSSJ0804+1616(discussedbelow)degradedtoresolutions thisside.Whiletheexactnumberofcandidatesinthesam- ofR=600andR=300,andaS/Nof7perpixel.Itfurther- pleisarbitrary,thecut-offwaschosensuchthatthenumber more shows the same spectra again but with the equivalent ofobjectsremainedmanageablewhilestillencompassingan widths of all spectral lines reduced by a factor 2, to make estimated ∼90% of the total population, based on the 9 themmorecomparabletotheAMCVnstarswiththeweak- known emission-line AM CVn stars in the SDSS-DR6 pho- est lines of the known long-period systems. Each resolution tometric database (plotted in Figure 1). It is important to elementissampledbyexactlytwopixels,givingaS/Nof10 notethatthecompletenessofspectroscopicfollow-upinthe per resolution element. SDSS, i.e. the fraction of photometric objects that have an Byconsideringthefullopticalspectrum,AMCVnstars SDSS spectrum, increases towards the more densely pop- like the ones we know today will clearly stand out even in ulated DA white dwarf locus in (u−g,g −r) that is cut thelowest-qualityspectrashown.Evensystemswithweaker off from our sample, as shown in Roelofs et al. (2007b). lines will still show a clear hint of helium emission, that Thesampleof9knownAMCVnsintheSDSSphotometric canbeconfirmedfromasecond,higher-qualityspectrum.If database,ofwhichsixwerediscoveredintheSDSSspectro- one only requires spectra of such low resolution and S/N, scopicdatabase,isthusactuallyslightlybiasedtowardsthe it becomes feasible to observe the entire sample of 1,523 red cut-off in u−g, making it unlikely that a large fraction objectswithrelativelysmalltelescopes,despitethefactthat ofAMCVnsfalloutsidethiscut-off.Itshouldbepossibleto the sample goes deep. moreaccuratelyjudgeandcorrectfortheincompletenessof The question is, of course, whether there exist systems oursampleaposteriori,oncethesampleofspectroscopically that,forsomereason,havemuchweakeremissionlinesand identified AM CVns has been enlarged. maygoundetectedusingthisobservingstrategy.Thesecan- Weremovedknown,previouslyidentifiedobjects(based notberuledoutgiventhattheknownsystemshavelargely onasimbad1 query)fromthelistofsourcesselectedbythe been selected on their emission lines. However, we consider criteria above, which mostly affected objects at the bright itunlikelythattheserepresentalargefractionofthepopu- end of the sample, and we checked the remaining objects lation. There is a clear trend of increasing equivalent width by eye, removing obvious ‘mistakes’ of the SDSS photo- of the emission lines with orbital period in the known AM metric data reduction pipeline. These were mostly com- CVns(Fig.10inGrootetal.2008),presumablyasaresult pact star forming regions flagged as point sources in exten- of the accreting white dwarf cooling towards longer orbital sive starburst galaxies, as well as stars close to other very periods, leading to a drop in the continuum flux against brightstars.Manyofthesewereflagged‘blended’&‘notde- which the emission lines from the accretion disc are mea- sured (a possible additional effect is the expected increase inthephysicalsizeofthedisctowardslongerperiods).This 1 http://simbad.u-strasbg.fr/ suggests that systems with weaker emission lines will pre- (cid:13)c 2008RAS,MNRAS000,1–9 4 Roelofs et al. Figure 3. Simulated low-resolution, low signal-to-noise ratio spectra of the new AM CVn star, SDSSJ0804+1616, to illustrate the requirements for spectroscopic identification of AM CVn stars. The spectral range is typical for a low-resolution set-up. Left and right panelsshowspectralresolutionsR=600andR=300,respectively,whereeachresolutionelementissampledbytwopixels.Allspectra haveasignal-to-noiseratioof7perpixel,or10perresolutionelement.Inthelowerpanelsthestrengthsofthelineshavebeenreduced byafactor2,toaHei5876equivalentwidthof30˚A. dominantly have short orbital periods, P (cid:46) 30min. As orb Date UT Instrument Exposures mentioned before, such short-period systems should repre- (exp.time) sent only a small fraction of the total population. As in the case of the colour selection of our sample, the incom- 2008/03/19 00:55–03:03 IMACSf/4 46(150s) pletenessofthespectroscopicidentificationofAMCVnscan 2008/04/17 23:25–01:02 MagE 29(180s) be judged more accurately a posteriori, when all the actual 2008/04/18 23:19–01:05 MagE 32(180s) spectra have been taken, and when it is clearer what the 2008/04/19 23:21–01:16 MagE 35(180s) distribution in emission line strength is. Spectra have so far been obtained of 15% of the sam- Table 1. Summary of our Magellan observations of SDSSJ0804+1616. Weather in all cases was good, with an ple, mostly at the bright end, in the Spring of 2008 using averageseeingof0.6(cid:48)(cid:48). severaltelescopes,predominantlythe1.5-mTillinghasttele- scope at Mt. Hopkins, Arizona equipped with the FAST spectrograph, and the 2.5-m Isaac Newton Telescope on La PalmawiththeIntermediateDispersionSpectrograph.One In order to look for spectral line variations that might oftheobjectsobservedtodate,SDSSJ080449.49+161624.8 indicate orbital motion in the candidate AM CVn star, we (hereafterSDSSJ0804+1616),showedtheanticipatedspec- obtained phase-resolved spectroscopy on the Magellan tele- trum of helium emission lines but no hydrogen, and will be scopesatLasCampanasObservatory,Chile,aslistedinTa- discussed in detail in the next section. ble1.WeinitiallyusedtheIMACSf/4camerawiththe600 lines/mmgratingtoobtainatotalof46consecutivespectra of150sexposuretimeeachonMarch19,2008.Thespectral 4 THE FIRST CONFIRMED AM CVN STAR resolution was 3.0˚A. Wavelength calibration was obtained from helium–neon–argon comparison lamp exposures, with 4.1 Observations and data reduction 0.2˚A root-mean-square residuals. SDSSJ0804+1616,ag=18.2objectinoursample(seealso Following the tentative detection of spectroscopic vari- Fig. 1), was found to be a strong helium-emission-line ob- ability on a ∼40-min timescale, we obtained a longer series ject based on a spectrum obtained 25 February 2008 with of96spectraof180seachon17,18and19April2008using the2.5-mIsaacNewtonTelescope,LaPalma,andtheInter- the new Magellan Echellette (MagE) spectrograph, cover- mediate Dispersion Spectrograph. Unusually strong lines of ing the entire optical range from the atmospheric cut-off Heiiwerenoted,althoughcontaminationwiththehydrogen in the far-blue to approximately 9500˚A in the far-red, at Balmer series could not be ruled out for some of them. a resolution R ∼ 5000 or about 1.2˚A. Detector read-out (cid:13)c 2008RAS,MNRAS000,1–9 SDSSJ0804+1616: a peculiar new AM CVn star 5 time between exposures was 23 seconds in ‘fast’ mode, un- binned. Thorium–argon arc spectra were taken before and after the science exposures on each night, and dispersion solutions were obtained from third-order polynomial fits to the reference lines in the extracted echelle orders, leaving 0.05˚Aroot-mean-squareresiduals.Thedispersionsolutions for the science exposures were interpolated from the com- parisonarcsolutions;arcdriftwasobservedtobenegligible over the time-span of our observations. Series of short and longincandescentlampexposures,providingmedian-filtered flat field images for the red and blue sides of the CCD re- spectively,werecombinedintoasinglenormalisedflatfield. The echelle orders were flux calibrated using comparison spectra of the spectrophotometric standard star HD60753. At wavelengths > 5000˚A, we further corrected for telluric absorption lines using comparison spectra of the DA white dwarf standard LTT3218. Correction for atmospheric ex- Figure5.ProfileoftheHei7065lineinSDSSJ0804+1616com- tinction and wavelength-dependent slit-losses remained im- pared to the more typical profile in GP Com (GP Com data perfectsincethestandardstarhadtobeobservedatalower fromSteeghsetal.inpreparation).Thelineprofilesarewellre- airmass than SDSSJ0804+1616; this could affect the over- solved,withresolutionelementscorrespondingtoapproximately all slope of the spectrum. The corrected echelle orders were twobins.TherightpanelshowstheaverageoftheHei4387,4471, 4921,5875,6678and7065lines,andaGaussianfitindicatedby mergedwhileweightingpixelsinoverlappingordersbynum- thedottedline. ber of counts, and finally the wavelength scale was trans- formed to the heliocentric rest-frame. SDSSJ0804+1616 (and have been observed in spectra of several other AM CVns: Groot et al. 2001; Roelofs et al. 4.2 Results 2005,2006a),alleviatingtheproblemoftheoriginofallthe nitrogen for SDSSJ0804+1616 (but not for GP Com). 4.2.1 Average spectrum The strong emission-line complexes of neutral nitrogen Figure 4 shows the average spectrum of SDSSJ0804+1616. seen in GP Com are absent, or at the very least least much It exhibits a plethora of emission lines, mainly of helium. weaker, in SDSSJ0804+1616 despite the clear presence of EmissionlinesclosetothehydrogenBalmerseriesarefound nitrogen in its ionised form (see Table 2). tobeHeii,basedontheircentralwavelengthsandstrengths The emission lines in SDSSJ0804+1616 have a full- relativetotheHeiilinesat5411and4541˚Athatarenotco- width at half-maximum of about 1,500 km/s and are ap- incidentwithhydrogen.Theemissionlinesatthefar-redend proximately Gaussian in shape, except for a slight excess ofthespectrumdonotcoincidewiththehydrogenPaschen of flux at low velocities. See Figure 5 for a close-up of the series,anditthusappearsthereisnoevidenceforthepres- Hei7065lineincomparisonwiththesamelineinGPCom. ence of hydrogen. Thelatterhasthemoretypicaltriple-peakedprofileseenin A series of lines observed at 4630, 5679, 5941, 6170, AM CVn stars: a double-peaked profile from the accretion 6242, 6482 and possibly 6610˚A are identified with Nii. disc,similartotheprofilesseeninnon-magnetic,hydrogen- These, together with the observed Heii lines, indicate tem- richCVs,plusaverynarrowspikeuniquetotheAMCVns, peratures of ∼20,000–27,000K in at least part of the ac- which is thought to arise from (very close to) the surface cretion flow. However, several strong carbon and oxygen of the accreting white dwarf and to be related to the ac- linesthatwouldbeexpectedatthesetemperaturesaremiss- cretionprocess(seeMarsh1999;Morales-Ruedaetal.2003, ing entirely from the spectrum, most notably Cii at 4267, and see Ruiz et al. 2001; Roelofs et al. 2006a, 2007c for 6785,7119and7236˚A,andeitherOi7774orOii4070-4077, emission-lineprofilesinanumberofotherAMCVns).None 4591˚A.Highlynon-solarabundanceratiosofN/O(cid:38)10and of the emission-line profiles of SDSSJ0804+1616 show any N/C>10bynumberareneededtosufficientlysuppressthe evidence for a classical double-peaked component from a carbon and oxygen lines relative to nitrogen regardless of disc, indicating a different geometry of the accretion flow. theexacttemperaturestructure,assumingasimplespectral Table2liststhewidthsandequivalentwidthsofmostof modelofagasmixtureinlocalthermodynamicequilibrium thelinesobservedinthespectrumofSDSSJ0804+1616,and (LTE): the same model that was used to derive a similar compares them to those observed in the 46-minute binary (yet stronger) limit of N/C>100 for the 46-min binary GP GP Com. Com (Marsh 1991, as updated by Nelemans et al. 2004b). Thisstronglynon-solarabundancepatternhasbeenseenas 4.2.2 Spectral line variability evidence for CNO-processing in GP Com, with the added complication that the apparent absence of metal lines such Variability in the spectral lines, typically in the form of a as Si in that system suggests a metal-poor initial compo- sinusoidal ‘S-wave’ in spectroscopic time-series, is observed sition, which would require that the observed N have been in most of the AM CVn stars and all of the long-period, dredged up and transferred during the prior AGB phase of emission-line systems (Nather et al. 1981; Ruiz et al. 2001; whatisnowtheaccretingwhitedwarf.TheSiii6347&6371 Roelofs et al. 2005, 2006a, 2007c). Following the procedure linesthatwere‘missing’inGPComareclearlyobservedin of Nather et al. (1981) we compute a Lomb–Scargle peri- (cid:13)c 2008RAS,MNRAS000,1–9 6 Roelofs et al. Figure 4. Average spectrum of SDSSJ0804+1616. Thin and thick solid lines indicate neutral and ionised helium lines, respectively, whiledottedlinesindicatemetallinesaslabelled.SeeFigure5foradetailedviewoftheshapesoftheemissionlines. Figure 6. Lomb–Scargle periodogram of the flux ratios of the redandbluewingsoftheemissionlinesusedforFigure7. odogram of the ratios of fluxes in the red and blue wings of the emission lines of SDSSJ0804+1616. See Figure 6. A Figure 7. Trailed spectrum of SDSSJ0804+1616, composite of strong variability peak is seen centered on 32.4 cycles per theHeiandSiiilinesbetween5000–7400˚A.Asinusoidal‘S-wave’ day,or44.5minutes.Theshortobservingbaselineandequal variation is seen on a period of P = 44.5±0.1min. Higher- orb spacing of the individual blocks of spectra create strong velocity emission, up to 1500 km/s, is seen to vary on the same aliases at 1 cycle/day intervals, but regardless of the spec- period.Thegrey-scaleindicatestherelativefluxdensities. tral lines or velocity limits chosen for the red-wing/blue- wingperiodogram,the44.5-minuteperiodcomesoutasthe strongest peak. Figure 7. It shows a clear S-wave with a velocity ampli- Since the red-wing/blue-wing method does not opti- tudeof160±10km/s,aswellasasymmetrichigher-velocity mally utilize the variability information contained in the emission extending to about 1,500km/s on the red side. spectroscopic time-series, we also do a ‘double-blind’ (but Based on the sinusoidal S-wave feature we propose that subjective) test to judge by eye at which folding period, P =P =44.5±0.1min. fold orb in the range 20–120 minutes, the strongest variability pat- A similar picture is seen at the ±1 cycle/day aliases of tern occurs. This yields P = 44.5±0.1min, a perfect this period, as expected, while for the ±2 cycles/day (and fold match to the period obtained from the Lomb–Scargle peri- moredistant)aliasesthepatternfades.Novariabilityisseen odogram. The corresponding trailed spectrum is shown in on periods other than aliases of the main signal. We con- (cid:13)c 2008RAS,MNRAS000,1–9 SDSSJ0804+1616: a peculiar new AM CVn star 7 5 DISCUSSION Line FWHM(kms−1) EW(−˚A) The first new AM CVn star from our survey of candi- SDSSJ0804 GP Com datesfromtheSDSSposesaninterestingopticalspectrum: Heii4340 2300±300 3.3±0.4 − devoid of hydrogen, but rich in nitrogen, with strikingly Hei4387 1550± 50 22.6±1.0 12.8±0.3 strong lines of ionised nitrogen and helium. The high ni- Hei4471 1500±100 55.0±1.0 39.9±0.3 trogenabundanceputsconstraintsonthepriorevolutionof Heii4541 2400±150 8.0±0.8 − the donor star, in particular the amount of helium burn- Nii4630 2200±150 10.0±1.0 1.0±0.3 ing that may have taken place. Detailed helium star mod- Heii4686 1500±150 78.0±2.0 24.5±0.3 els for the donor stars in AM CVns predict typical sur- +Hei4713 face abundance ratios N/C=0.001−1 at orbital periods Heii4860 1900±100 8.0±0.8 − Hei4921 1460± 50 48.0±1.0 29.2±0.3 >30min(Yungelson2008),whileforwhitedwarfdonorstars Hei5015 1600±150 62.0±1.0 47.8±0.3 one expects N/C≈0.1 (if the CNO cycle never operated +Hei5047 during the donor’s hydrogen-burning phase) or N/C(cid:29)1 Heii5411 1900±75 12.8±0.5 − (if the CNO cycle has converted most of the primordial Nii5680 1930±50 15.0±1.0 3.3±0.3 C and O into N). This suggests that the donor star in Hei5876 1250±50 60.0±1.0 77.7±0.3 SDSSJ0804+1616 has not had a phase of helium burning, Nii5942 ··· 9.0±1.0 − implying that the helium-star model for the formation of Nii6170 ··· 2.2±0.2 − AM CVn stars (e.g. Iben & Tutukov 1991) does not ap- Nii6242 ··· 2.3±0.2 − ply to SDSSJ0804+1616. The lower limits of the observed Nii6484 ··· 4.5±0.4 − N/C and N/O abundances could be marginally consistent Heii6559 1700±50 20.0±1.0 − Hei6678 1240±50 50.0±2.0 60.6±0.3 with the lowest-possible amount of helium burning in the Hei7065 1240±50 34.8±0.5 54.3±0.3 modelsofYungelson(2008),correspondingtoaheliumstar Heii7177 ··· 2.2±0.4 − that is already almost filling its Roche lobe upon reaching Hei7281 1280±50 20.0±1.0 27.2±0.4 the zero-age helium main sequence, but this requires fine- Ni7440 − 14.1±0.4 tuning. The ‘evolved-CV’ model for the formation of AM Oi7774 − 5.0±0.4 CVn stars (Podsiadlowski et al. 2003) predicts essentially Ni8200 ··· †19.0±2.0 51.1±0.5 thesameCNOabundancesastheheliumwhitedwarfmodel, Ni8700 ··· †15.0±2.0 104.6±0.6 and is thus compatible as far as the observed CNO abun- Caii3933 1400±100 8.0±0.8 − dances are concerned. Siii6347 ··· 7.0±0.7 − While the LTE emission line model used to derive the +Siii6371 relativeCNOabundancesisrathersimple,itseemsunlikely that a more detailed treatment including non-LTE effects Table 2. Equivalent width (EW) and Gaussian full-width at couldsucceedinsuppressingalltheexpectedlinesofCand half-maximum (FWHM) for the most prominent spectral lines O while reproducing the observed lines of N, without an inSDSSJ0804+1616.Estimatederrorsaremainlyduetouncer- actualnitrogenoverabundance.Itwouldhoweverbeworth- tainty in the continuum level. The last column shows GP Com while to try to apply such models (e.g. Nagel et al. 2004). data from Marsh (1991). Non-detections, marked ‘−’, imply an AnalternativehypothesiscouldbethatsedimentationofC EWof0.0±0.3.Valuesmarked‘···’couldnotbemeasuredreli- ably.†Uncertainidentification;EWofNicomponentlikelylower and O in the outer layers of the donor star precludes these (andpossiblyzero). elements from showing up in the spectrum, or that we are seeing the previously sedimented N in what is now just the inner few 0.01M of the original donor. This mechanism (cid:12) is unlikely to work, however, since none of these elements should sink into a degenerate helium donor star, given the identicalmass-to-chargeratiosoftheC,N,OandHenuclei involved. siderthe±1cycle/dayaliasespossible,andthemoredistant A large nitrogen abundance is far from unique among aliases unlikely, to represent the actual orbital period. The the AM CVn stars, rather the opposite: anomalously large timecoverageofthepresentdataisinsufficienttotrackthe N/C line ratios in the far-UV spectrum of V396 Hya were S-wavefromonenighttothenextandbreakthealiases,but noted by Ga¨nsicke et al. (2003), while the X-ray spectra the data suffice to firmly establish the ultracompact nature of several others were best modelled with CNO-cycled or ofthebinary,therebyconfirmingitsmembershipoftheAM evenhighernitrogenabundances(Ramsayetal.2005,2006). CVn family. Combined with the results of Yungelson (2008), this would WiththeS-waveperiodinhand,werevisitourIMACS suggest that most of the known AM CVns do not har- dataobtainedonemonthearlier.NoS-waveisseeninanyof bour formerly helium-burning donor stars, even though the the lines or combination of lines, nor for any alias of the S- large observed luminosities of the short-period AM CVns waveperiod.AlthoughtheIMACSdatahaveaslightlylower are much more readily explained if one does assume such resolution, lower signal-to-noise ratio, and shorter baseline, donors (Roelofs et al. 2006b, 2007a). we would have expected to see a sign of the S-wave in the Further information about SDSSJ0804+1616 may be corresponding trailed spectrum if it were as strong as in obtainedfromtheprofilesofitsemissionlines.Theirunique theMagEspectra.Weconcludethatthepatternoforbital- single-peakedness is not simply an inclination effect, since period variations must change on timescales of one month. thelinesareverybroadcomparedtothoseinotheremission- (cid:13)c 2008RAS,MNRAS000,1–9 8 Roelofs et al. lineAMCVnsthatstillclearlyshowadouble-peakedaccre- forScientificResearch.BasedondatatakenattheMagellan tion disc profile (e.g., Roelofs et al. 2005). The most likely TelescopesatLasCampanasObservatory,Chile,andatthe explanation,alsoinviewofthelargeNii/NiandHeii/Hei Isaac Newton Telescope in the Observatorio del Roque de line ratios, is that the system is magnetic: broad, single- los Muchachos of the Instituto de Astrof´ısica de Canarias, peaked,almostGaussianemissionlineswithstrongseriesof Spain.ThisstudymakesuseoftheSloanDigitalSkySurvey; ionised helium are characteristic of magnetic CVs and the see http://www.sdss.org/collaboration/credits.html spectrum of SDSSJ0804+1616 is a perfect match, save for for the full acknowledgment. Special thanks to P. Berlind the absence of hydrogen (e.g. Schwartz et al. 1998). Opti- and the 1.5-m telescope crew at F.L. Whipple Observatory cal polarimetry should be pursued to confirm the magnetic (a facility of the Smithsonian Institution) for their valuable nature of the system, and near-infrared spectroscopy might contribution, and to L. Yungelson for helpful discussion. revealcyclotronemissionfeaturesasfurtherevidenceforthe Figure 1 makes use of white dwarf model spectra kindly presence of strong magnetic fields. provided by D. Koester. A final, intriguing question is whether we should not have found more than one AM CVn star amongst the can- didatesobservedsofar.Thehit-ratedoesappeartobebelow REFERENCES expectations (Section 2), but may at this point simply suf- fer from low-number statistics and the fact that we have so Adelman-McCarthy J., et al., 2008, ApJS, 175, 297 faronlytargetedthebrightendofthesample,whichmeans Anderson S. F., et al., 2005, AJ, 130, 2230 thatwehaveprobablyobservedadifferentpopulationofob- Anderson S. F., et al., 2008, AJ, 135, 2108 jects than the average of the entire sample. Once the entire BildstenL.,TownsleyD.,DeloyeC.J.,NelemansG.,2006, sample has been completed, a full analysis as described in ApJ, 640, 466 Roelofs et al. (2007b) will have to be done, which will give Bildsten L., Shen K. J., Weinberg N. N., Nelemans G., a revised and more accurate space density for the Galactic 2007, ApJ, 662, L95 AM CVn population. Cropper M., Harrop-Allin M. K., Mason K. O., Mittaz J. P. D., Potter S. B., Ramsay G., 1998, MNRAS, 293, L57 Deloye C. J., Bildsten L., Nelemans G., 2005, ApJ, 624, 934 6 CONCLUSION Deloye C. J., Taam R. E., Winisdoerffer C., Chabrier G., We have described a spectroscopic survey aimed at identi- 2007, MNRAS, 381, 525 fying all of the ∼40 expected emission-line AM CVn stars Ga¨nsicke B.T., et al., 2003, ApJ, 594, 443 in an estimated 90% complete sample of 1,523 candidates, Groot P. J., Nelemans G., Steeghs D., Marsh T.R., 2001, drawn from a colour selection of the Sloan Digital Sky Sur- ApJ, 558, L123 vey,DataRelease6,downtog=20.5.Theaimistogreatly Groot P. J., et al., 2008, MNRAS, submitted increasethesizeandcompletenessofthesampleofAMCVn Iben I. Jr., Tutukov A. V., 1991, ApJ, 370, 615 starsfromtheSDSS,allowingforamuchmoreaccurateas- Israel G. L, et al., 2002, A&A, 386, L13 sessment of the total Galactic population. Marsh T. R., 1999, MNRAS, 304, 443 We have presented time-resolved spectroscopy con- Marsh T. R., Horne K., Rosen S., 1991, ApJ, 366, 535 firming the first new AM CVn star from this survey, Morales-RuedaL.,MarshT.R.,SteeghsD.,Unda-Sanzana SDSSJ080449.49+161624.8,andmeasuredanorbitalperiod E., Wood J. H., North R.C., 2003, A&A, 405, 249 ofPorb =44.5±0.1min,orpossiblyoneofitsnearestdaily Nagel T., Dreizler S., Rauch T., Werner K., 2004, A&A, aliases.Thesingle-peakedemissionlinesaswellasthestrong 428, 109 lines of ionised helium, both unique among the long-period NatherR.E.,RobinsonE.L.,StoverR.J.,1981,ApJ,224, AM CVns known to date, are very much reminiscent of 269 themagneticsystemsamongthehydrogen-richCataclysmic NelemansG.,PortegiesZwartS.F.,VerbuntF.,Yungelson Variables, and we suggest that SDSSJ0804+1616 may be L. R., 2001, A&A, 368, 939 the first ultracompact version of such a system. Nelemans G., Yungelson L. R., Portegies Zwart S. F., The strong lines of nitrogen but non-detections of ex- 2004a, MNRAS, 349, 181 pected lines of carbon and oxygen suggest that the donor Nelemans G., Jonker P. G., Marsh T. R., van der Klis M., star in SDSSJ0804+1616 contains highly non-solar abun- 2004b, MNRAS, 348, 7L danceratiosoftheseelements.Wehavederived,fromasim- NelemansG.,2005,inASPConf.Ser.Vol.330,TheAstro- ple LTE model, lower limits of N/C>10 and N/O(cid:38)10 by physicsofCataclysmicVariablesandRelatedObjects.As- number. This indicates that the donor star has burned hy- tron. Soc. Pac., San Francisco, p. 27 (astro-ph/0409676) drogen via the CNO cycle, while it is very unlikely that it Podsiadlowski Ph., Han Z., Rappaport S., 2003, MNRAS, has had a subsequent helium-burning phase, effectively rul- 340, 1214 ingoutthe‘helium-star’formationscenarioforthissystem. RamsayG.,HakalaP.,MarshT.,NelemansG.,SteeghsD., Cropper M., 2005, A&A, 440, 675 Ramsay G., Groot P. J., Marsh T., Nelemans G., Steeghs D., Hakala P., 2006, A&A, 457, 623 ACKNOWLEDGMENTS Roelofs G. H. A., Groot P. J., Marsh T. R., Steeghs D., GHARissupportedbyNWORubicongrant680.50.0610to Barros S. C. C, Nelemans G., 2005, MNRAS, 361, 487 G.H.A. Roelofs. DS acknowledges an STFC Advanced Fel- Roelofs G. H. A., Groot P. J., Marsh T. R., Steeghs D., lowship. GN is supported by the Netherlands Organization Nelemans G., 2006a, MNRAS, 365, 1109 (cid:13)c 2008RAS,MNRAS000,1–9 SDSSJ0804+1616: a peculiar new AM CVn star 9 Roelofs G. H. A., Groot P. J., Marsh T. R., Steeghs D., Nelemans G., 2006b, MNRAS, 371, 1231 Roelofs G. H. A., Groot P. J., Benedict F., McArthur B., MarshT.R.,SteeghsD.,Morales-RuedaL.,NelemansG., 2007a, ApJ, 666, 1174 Roelofs G. H. A., Nelemans G., Groot P. J., 2007b, MN- RAS, 382, 685 Roelofs G. H. A., Groot P. J., Steeghs D., Marsh T. R., Nelemans G., 2007c, MNRAS, 382, 1643 Ruiz M. T., Rojo P. M., Garay G., Maza J., 2001, ApJ, 552, 679 Savage B. D., Mathis J. S., 1979, ARA&A, 17, 73 SchlegelD.J.,FinkbeinerD.P.,DavisM.,1998,ApJ,500, 525 Schwartz R., et al., 1998, A&A, 338, 465 Thorstensen J. R., 2003, AJ, 126, 3017 Warner B., Woudt P. A., 2002, PASP, 114, 129 Woudt P. A., Warner B., 2003, MNRAS, 345, 1266 Woudt P. A., Warner B., Rykoff E., 2005, IAU Circular, 8531, 3 York D. G., et al., 2000, AJ, 120, 1579 Yungelson L. R., 2008, Astron. Lett., 34, 620 (cid:13)c 2008RAS,MNRAS000,1–9