Mon.Not.R.Astron.Soc.000,1–??(2013) Printed2February2015 (MNLATEXstylefilev2.2) Kinematics and Host-Galaxy Properties Suggest a Nuclear Origin for Calcium-Rich Supernova Progenitors 5 Ryan J. Foley1,2⋆ 1 0 1Astronomy Department, University of Illinois at Urbana-Champaign, 1002 W. Green Street, Urbana, IL 61801, USA 2 2Department of Physics, University of Illinois at Urbana-Champaign, 1110 W. Green Street, Urbana, IL 61801, USA n a J Accepted .Received;inoriginalform 9 2 ABSTRACT ] Calcium-rich supernovae (Ca-rich SNe) are peculiar low-luminosity SNe Ib with rela- E tively strongCa spectrallines at∼2months after peak brightness.This classalsohas H anextendedprojectedoffsetdistribution,withseveralmembersoftheclassoffsetfrom . their host galaxies by 30 – 150 kpc. There is no indication of any stellar population h atthe SNpositions.Using asample of13Ca-richSNe, wepresentkinematic evidence p that the progenitors of Ca-rich SNe originate near the centers of their host galaxies - o andare kickedto the locations of the SN explosions.Specifically, SNe with smallpro- r jected offsets have large line-of-sight velocity shifts as determined by nebular lines, t s while those with large projected offsets have no significant velocity shifts. Therefore, a the velocity shifts must not be primarily the result of the SN explosion. There is an [ excess of SNe with blueshifted velocity shifts within two isophotal radii (5/6 SNe), 1 indicating that the SNe are moving away from their host galaxies and redshifted SNe v on the far sides of their galaxies are selectively missed in SN surveys. Additionally, 7 nearly every Ca-rich SN is hosted by a galaxy with indications of a recent merger 0 and/or is in a dense environment. At least 6–7 host galaxies also host an AGN, a 6 relativelyhighfraction,againlinkingthenuclearregionofthesegalaxiestoexplosions 7 occurringtensofkpcaway.Weproposeaprogenitormodelwhichfitsallcurrentdata: 0 The progenitor system for a Ca-rich SN is a double white dwarf (WD) system where . 1 at least one WD has a significant He abundance. This system, through an interac- 0 tion with a super-massive black hole (SMBH) is ejected from its host galaxy and the 5 binary is hardened, significantly reducing the merger time. After 10 – 100 Myr (on 1 average),the system explodes with a large physical offset. The rate for such events is : v significantly enhanced for galaxies which have undergone recent mergers, potentially i making Ca-rich SNe new probes of both the galaxy merger rate and (binary) SMBH X population. r a Key words: supernovae—general, supernovae—individual (SN 2000ds, SN 2001co, SN 2003H, SN 2003dg, SN 2003dr, SN 2005E, SN 2005cz, SN 2007ke, SN 2010et, SN 2012hn, PTF09dav, PTF11bij, PTF11kmb), galaxies—individual (2MASS J22465295+2138221, CGCG 170-011, IC 3956, NGC 1032, NGC 1129, NGC 2207, NGC 2272, NGC 2768, NGC 4589, NGC 5559, NGC 5714, NGC 7265, UGC 6934) 1 INTRODUCTION classes mentioned above. It turns out that there were about a dozen new classes of “exotic” or “peculiar” tran- For the past century, systematic supernova (SN) searches sients lurking in the shadows. These classes include lu- have discovered thousands of SNe. Nearly all of these SNe minous SNe IIn (e.g., Smith et al. 2007), Type I super- fall into 3 classes: Type Ia, Type II, and Type Ib/c. How- luminousSNe(Quimbyet al.2011),kilonovae(Berger et al. ever, over the last decade with the implementation of large 2013; Tanvir et al. 2013), SNe Iax (Foley et al. 2013), and SN searches, we have begun to discover many astrophys- SN 2006bt-like SNe(Foley et al. 2010). ical transients that do not fall into the well-delineated One class, the SN 2005E-like SNe (Perets et al. 2010, hereafter, P10),also knownas “Calcium-rich SNe”(orsim- ⋆ E-mail:[email protected] ply “Ca-rich SNe”; Filippenko et al. 2003; Kasliwal et al. (cid:13)c 2013RAS 2 Foley 2012, hereafter, K12), are particularly interesting. These in other works (P10; Kawabata et al. 2010; Sullivan et al. SNe are technically of Type Ib, having distinct He lines 2011; K12; Valenti et al. 2014). We include PTF09dav in in their spectra near maximum brightness. However, they this analysis, although at maximum light, it had much are less luminous and faster fading than normal SNe Ib lower ejecta velocities than other members (Sullivan et al. (P10; K12) and are often found in early-type galaxies 2011)anduniquelydisplayshydrogenemissionatlate-times (P10; Lyman et al. 2013). Additionally, several members of (K12). the class, including SN 2005E have large projected offsets To this sample, we add PTF11kmb. PTF11kmb was (>10kpcanduptoatleast150kpc)fromtheirhostgalaxies discoveredon24.24August2011(alltimesareUT)byPTF (P10;K12;Yuan et al.2013;Lyman et al.2014;thiswork). (Gal-Yam et al. 2011).A Keck/LRISspectrum obtained28 Relative to other SNe Ib, Ca-rich SNe also have a dis- August 2011 was used to classify PTF11kmb as a SN Ib tinctspectroscopicevolution.Afterabout2months,theSNe (Gal-Yam et al. 2011). We use this spectrum as well as a show strong forbidden lines, indicating that the ejecta are Keck/LRISspectrum from 26.37 November2011 toclassify at that point mostly optically thin. This is a much faster PTF11kmb as a Ca-rich SN (AppendixA). transition than for most SNe Ib. This “nebular” spectrum Basic host-galaxy information was obtained through is also distinct, showing extremely strong [CaII] λλ7291, theNASA/IPACExtragalacticDatabase(NED)andispre- 7324 emission relative to that of [OI] λλ6300, 6363, giving sented in Table A1. Four SNe have somewhat ambiguous the name to the class (Filippenko et al. 2003). Modeling of hosts, and we discuss each below. theejecta also indicatesthat themass-fraction ofCa in the SN 2003H occurred between two merging galaxies, ejecta is also extremely high (∼1/3; P10). NGC 2207 and IC 2163. Their recession velocities differ by P10 examined the rate of Ca-rich SNe in the Lick Ob- 24±25kms−1,andthisdifferencedoesnotaffectourresults. servatorySupernovaSearch(LOSS;Filippenko et al.2001), SN2003Hisoffsetby8.73and5.75kpcfromNGC2207and finding that they occur at 7±5% the rate of SNe Ia. K12 IC 2163, respectively, and the difference in offset does not foundarate>2.3%thatofSNeIaforthePalomarTransient affect any results. Factory (PTF) sample. Therefore, Ca-rich SNe are some- what common eventsand cannot comefrom extremely rare SN2007keoccurredinaclusterenvironment(AWM7) progenitor scenarios. nearthebrightestmember,NGC1129.Itisalsonearagroup Since most Ca-rich SNe have early-type hosts (P10; member,MCG+07-07-003. Thedifferenceintheirrecession Lyman et al. 2013), a massive star origin is unlikely. More- velocities is 227±23 km s−1, with NGC 1129 having the over, there is no indication of star formation at the posi- larger redshift. This difference does not affect any results, tionoftheSNeindeeppre-orpost-explosionimaging(P10; butisnoteworthy.TheSNwasoffset by16.71 and8.20 kpc Perets et al. 2011; K12; Lyman et al. 2014). from NGC 1129 and MCG+07-07-003, respectively. Again, Because of the large offset distribution, it has been thisdifferencedoesnotaffectanyresults.Giventherelative suggested that Ca-rich SNe occur in dwarf galaxies (K12; sizes of these galaxies and large offset from MCG+07-07- Yuanet al. 2013) or globular clusters (Yuanet al. 2013); 003, NGC 1129 is a reasonable choice for the host galaxy. however, thedeep limits on any stellar light at theposition Nonetheless, given the cluster environment, it is possible of the SNe rule out such possibilities (Lyman et al. 2014). that SN 2007ke originated from a faint cluster member or These observations require that the progenitors of Ca-rich was formed in the intracluster medium. SNe be born elsewhere and travel a significant distance to SN 2010et occurred 37.6 – 69.2 kpc from three galax- where they explode (Lyman et al. 2014). With this deter- ies at redshifts consistent with the SN redshift. The closest mination, Lyman et al. (2014) suggested that Ca-rich SN galaxy, CGCG 170-011 is also relatively large. Using the progenitors wereneutronstar(NS)–whitedwarf(WD)bi- NED reported major and minor 2MASS Ks isophotal axes nary systems which are kicked by the SN that created the (withareferencevalueofKs=20.0 magarcsec−2)andpo- NSandthenundergoamergeraftertravelingfarfromtheir sition angle, SN 2010et was offset 5.5 isophotal radii from birth site. CGCG 170-011. SN 2010et is 69.2 kpc and 16.5 isophotal In this manuscript, we present a kinematic and host- radiifromtheothernearby,largegalaxy,CGCG170-010.A galaxystudyofCa-richSNe.Byexaminingthevelocityshift thirdgalaxy,SDSSJ171650.20+313234.4, is46.9kpcandis distribution,theprojectedoffsetdistribution,andtheangle too small and faint for a 2MASS isophotal measurement. offset distribution, as well as the host-galaxy properties of Based on offset, the most likely host is CGCG 170-011, the sample, we constrain the progenitor systems and origin and we assume this for the rest of the analysis. K12 de- ofCa-richSNe.InSections2and3,wedefineourSNsample termined that no source exists at theposition of SN 2010et and discuss their host-galaxy properties. In Sections 4 and with MR < −12.1 mag, but this does not completely rule 5, we measure the projected offset and velocity shift distri- out a possible dwarf galaxy as thehost of SN2012et. butionsforoursample.InSection6,weanalyzethesedata. Finally, PTF11kmb is far from any galaxy. PTF11kmb WepresentabasicprogenitormodelinSection7.Wediscuss hasaredshiftofroughlyz =0.017asdeterminedbySNfea- our results and summarize ourconclusions in Section 8. tures (Gal-Yam et al. 2011, and confirmed in our analysis). TheclosestgalaxylistedinNEDis4.2arcminawayandhas no redshift information. However, if it is at the distance of PTF11kmb, it would be offset by ∼80 kpc. The three clos- 2 SUPERNOVA SAMPLE est galaxies with redshifts are 2MASX J22224094+3613514 Our sample consists of all known (and published) Ca-rich (offset by 4.5 arcmin, 83.0 kpc, and 36.6 isophotal radii; SNe. We begin with the Lyman et al. (2014) sample of Ca- cz =4500±25 km s−1), UGC 12007 (offset by 6.5 arcmin, rich SNe. This sample includes 12 SNe originally reported 127.3kpc,and15.0isophotalradii;cz=4829±33 kms−1), (cid:13)c 2013RAS,MNRAS000,1–?? Progenitors of Ca-Rich SNe 3 Figure1.DSSimagesofCa-richSNhostgalaxies,witheachcenteredinindividualpanels.Eachpanelis8′×8′ exceptforPTF11kmb, whichis16′×16′.AllpanelshavethesamescaleandarealignedwithNorthupandEasttotheleft.EachSNpositionismarkedwith abluecircle.Anellipseattwicetheisophotal radiusisshowninred. and NGC 7265 (offset by 7.3 arcmin, 150.0 kpc, and 10.2 maximum brightness (e.g., K12). At these phases, and for isophotal radii; cz=5083±26 km s−1). all spectra examined here, Ca-rich SNe have strong [CaII] WhileNGC7265isveryfarphysicallyfromPTF11kmb, λλ7291, 7324 emission, especially relative to that of [OI] itistheclosest intermsofisophotalradiiandhasthemost λλ6300, 6363. similar redshift to that derived from the SN. We therefore adoptNGC7265asthehostofPTF11kmbforthisanalysis. All spectra (except for that of PTF11kmb) were previ- 3 HOST-GALAXY PROPERTIES ously published (Kawabata et al. 2010; P10; Sullivan et al. 3.1 Evidence for Recent Host-galaxy Mergers 2011; K12; Valentiet al. 2014). We obtained some of these data through WISERep (Yaron & Gal-Yam 2012). In Figure 1, we present Digitized Sky Survey(DSS) images Almost all SNe in our sample have only one late-time of the host galaxies of the Ca-rich SN sample. The host spectrum. For thefew SNe with multiple late-time spectra, galaxies of Ca-rich SNe are clearly atypical as previously weprimarilyanalyzethelatest high-qualityspectrumavail- noted(P10;Lyman et al.2013);thesampleishighlyskewed able. The spectra typically had phases of ∼2 months after to early-typegalaxies. (cid:13)c 2013RAS,MNRAS000,1–?? 4 Foley However,previousstudieshadnotnotedthestrongev- idence for recent host-galaxy mergers. In fact, nearly every Ca-rich SN host galaxy shows some indication of a recent 1.0 merger and/or is in a very dense environment where the likelihood of recent galactic mergers is much larger than in ) AGN thefield. β 0.5 H St actinOgf (tNhGeC1322h0o7s)t agnadlaxioens,e oinse ais dcislteuarrblyedinstperi-- III]/ arfor 10et ral (2MASS J22465295+2138221; Sullivan et al. 2011). O 0.0 mi NGC 1129 has a disturbed morphology indicative of a re- g ([ ng03dg 11bij cent merger (Peletier et al. 1990). There are also four S0 lo −0.5 03dr 01co galaxies (IC 3956, NGC 1032, NGC 2272, and NGC 7265), which are often thought to be the result of recent mergers Comp (e.g.,Moore et al.1999)orwhichmayhavebeen“harassed” −1.0 (e.g.,Moore et al.1998).Fromthesetracersalone,7/13Ca- rich host galaxies havesome indication of a recent merger. −1.0 −0.5 0.0 0.5 Additionally, at least 8/13 host galaxies are in a group log ([N II]/Hα) orclusterenvironment.Thisisanexceedinglyhighfraction, evenforelliptical galaxies.Moreover,7/13hostgalaxiesare Figure 2. Baldwin-Phillips-Terlevich (BPT) diagram eitherthebrightestgroupgalaxy(BGG)orbrightestcluster (Baldwinetal. 1981) for the Ca-rich SN host galaxies with galaxy (BCG). As these galaxies tend to sit at or near the SDSS spectra. The curves delineating the star-forming region center of the group/cluster potential, they are more likely (below the dashed line), the AGN region (above the solid line) to havehad recent mergers than typicalcluster members. andthe“composite”region(between thedashedandsolidlines) The relative rate of Ca-rich SNe to SNe Ia is roughly aretakenfromKewleyetal.(2001)andKauffmannetal.(2003) 7%(P10).Usingtherelativefractionof“recentmerger”host asimplementedinKewleyetal.(2006). galaxies for the two groups, the Ca-rich SNe would have a relativerateofonly2±2%thatofSNeIaingalaxieswithno indicationsofrecentmergersand11±8%thatoftheSNIa incompletesurveyofCa-richhostgalaxies.Althoughtheto- rate in galaxies with some indication of a recent merger. tal AGN fraction may be as high as ∼43% when thelowest In total, 11/13 Ca-rich SN host galaxies either have luminosity galaxies and AGN are included (e.g., Ho 2008), some evidence for a ongoing/recent merger or are in dense thisfractioncanbelowerforspecificgalaxysamples.Forin- environments.UsingtheLOSSSNIasample(Leaman et al. stance, an SDSS spectroscopic sample of >120,000 galaxies 2011), we find that 52% of SN Ia host galaxies have simi- foundthat18%ofallgalaxieshostedanAGNasdetermined larhostproperties.Usingbinomialstatistics,thereisonlya from line diagnostics (Kauffmann et al. 2003). For the host 1.6%chancethatSNeIaandCa-richSNecomefromsimilar galaxieswithSDSSspectra,3/5galaxiesalsohostanAGN. host populations in terms of their “recent merger” proper- Using binomial statistics for this subsample with the SDSS ties.SinceCa-richSNandSNIahostgalaxieshaveindistin- occurrencerate,thereisa4.7%chancethattheCa-richhost guishable morphology and star-formation ratedistributions galaxies are typical. If we expand this to thelarger, incom- (P10; Lyman et al. 2013), the merger/environment is likely pletesample,wefinda<0.4–2.0%chance(dependingonif6 asignificantfactorintherelativerates.Thegalaxiesinrich or7galaxieshostanAGN).However,wenotethatifweuse environments or that have had recent mergers have an en- a 43% occurrence rate, these differences are not significant hanced rate of Ca-rich SNerelative to that of SNeIa. for any subsample. For such a rate, >9/13 galaxies must host an AGN for a significant difference. Thereislikelyaconnectionbetweenthenuclearregions 3.2 A High Occurrence of AGN ofthesegalaxiesandtheSNeoffsetupto150kpcinprojec- In addition to the different indications of merger activ- tion. Sincethe typical dutycycle of an AGN is ∼107 years, ity, several host galaxies also host an active galactic nu- theinformationfromthenucleustoSNlocationwouldhave cleus (AGN).Previously identifiedAGNincludeNGC 2207 to travelat a very high speed of &1000 kms−1. (Kaufman et al. 2012), NGC 2768 (V´eron-Cetty& V´eron 2006), NGC 4589 (Hoet al. 1997; Nagar et al. 2005), and IC 3956 (e.g., Liu et al. 2011). Additionally, NGC 1032 has 4 GALACTIC OFFSETS a radio-to-far infrared flux ratio consistent with an AGN (Drakeet al.2003).WealsoexaminedtheSloanDigitalSky Ca-richSNehavebeenfoundveryfarfromtheirhostgalax- Survey (SDSS) emission line quantities for the host galax- ies(e.g.,P10;K12).However,thisisnotexclusivelythecase. iesobservedbySDSS(e.g.,Tremonti et al.2004).Forthese Forinstance,SN2003dgwasoffset4.1′′ fromitshostgalaxy five galaxies, we identify CGCG 170-011 as a likely AGN, corresponding to a projected distance of only 1.7 kpc. confirmthatIC3956hasa“composite”spectrumindicative UsingNEDvaluesfortheangular sizescale andoffsets of AGN activity, and identify NGC 5559 as having a likely betweentheSNeandhost-galaxy,wedetermineaprojected composite spectrum (Figure 2). offset for each SN. The typical uncertainty in the offsets is At least 6–7 (depending on the classification of ∼0.1′′,correspondingtoprojectedoffsetuncertaintiesofonly NGC 1032) of the 13 Ca-rich host galaxies also host an 0.01 – 0.07 kpc. All SN properties, including those relative AGN. This is a large fraction considering that this is an to theirhost galaxies, are presented in Table A2. (cid:13)c 2013RAS,MNRAS000,1–?? Progenitors of Ca-Rich SNe 5 InFigure1,wepresentDigitizedSkySurvey(DSS)im- 1.0 ages of the host galaxies of the Ca-rich SN sample. In each image, we display an ellipse corresponding to two isophotal radii for each galaxy and theSN position. 0.8 TodeterminetheisophotalradialoffsetforeachSN,we use the 2MASS Ks isophotal elliptical parameters (with a reference valueof Ks=20.0 magarcsec−2)for each galaxy. a e For each SN position, we determine a multiplicative factor r 0.6 A suchthatanellipsewiththesameaxisratio,positionangle, al and center will intersect with the SN position. This multi- n o plicativefactoristhenumberofisophotalradiiatwhichthe cti 0.4 SN is offset. ra F There are many sources for isophotal parameters, but 2thMesAeSpSapraromveidteerssamlaarygehaavnedshoommeowgheanteoluarsgseouerrcroe.rsHfoowrepvaerr-, 0.2 0.4 kpc/" ticulargalaxies.Themostsuspectvalueisthepositionangle 0. forNGC2207.The2MASSpositionangleis70◦,whileother 1 k p sources have values of 97◦ and 141◦. It appears that these c/" 0.0 differencesarecausedbythedifferentpositionanglesofthe 0 50 100 150 200 250 300 bulge/bar and the spiral arms. Nonetheless, NGC 2207 has Projected Offset (kpc) anaxisratioof0.68, sotheexactanglehasonlyamarginal affectonthemeasuredisophotalradialoffsetforSN2003H. Some galaxies are highly inclined, which is important Figure 3. Fraction of the region within a certain projected off- set from a particular galaxy observed by KAIT. The dotted, for understanding the SN offsets. For instance, SN 2003dr long-dashed, short-dashed, and solid lines correspond to scales is offset only 2.7 kpc (in projection) from its host galaxy of0.1,0.2,0.3,and0.4kpc/′′,respectively.Thearrowsrepresent, nucleus, but it is offset nearly perpendicular to its nearly fromsmallertolargeroffsets, the projectedoffsets ofSN 2005E, edge-on host galaxy (minor axis relative to major axis of PTF11bij,andPTF11kmb,respectively. 0.16). Thus SN 2003dr is offset by 2.9 isophotal radii from thenucleus of its host galaxy. discovered Ca-rich SNe, the median offset is 37.6 kpc, and 4.1 Galaxy Targeted SN Search Efficiency thereforeKAITshouldhavebeenabletodiscoversuchSNe in nearly all of their targeted galaxies. SN 2005E, which SomeCa-rich SNehaveextremely large angularoffsets. For was discovered by LOSS (Graham et al. 2005), was offset ′ ′ instance, SN 2005E and PTF11kmb are offset 2.3 and 7.3 by 24.3 kpc. In fact, all Ca-rich SNe, with the exception from their host galaxies, respectively. Galaxy-targeted SN of PTF11kmb, are offset less than 3.9′, meaning that they searches, such as LOSS, could potentially miss several Ca- would land within the KAITFOV. rich SNe because they are offset beyond the field-of-view Although KAIT will not be able to discover SNe with (FOV) of the camera. This is particularly worrisome since extremelylarge offsets, especially for theclosest galaxies, it PTF,whichrunsanuntargetedsearchandhasalargeFOV shouldbeabletosamplethebulkofthepopulation discov- camera, has only discovered (and announced) Ca-rich SNe ered by PTF. Although there are likely systematic effects that have a projected distance of at least 34 kpc and up to related to humans identifying a transient far offset from a 150 kpc. host galaxy as a true SN, there is no technical reason for a Tobetterunderstandtheefficiency oftheLOSSsearch lack of Ca-rich SNewith large offsets in theLOSS sample. at finding Ca-rich SNe, we perform a simple calculation. Since LOSS is not clearly missing Ca-rich SNe with TheKAITcamerahasa7.8′×7.8′ FOV.Byassumingthat largeoffsets,it isreasonable toassumethatPTFismissing a targeted galaxy is centered in the middle of the FOV, we someCa-rich SNewith small offsets (ornot reportingthese can determine the fraction of projected offsets observed for objects as Ca-rich SNe). This may be caused by the diffi- a galaxy at a given distance. cultyof detectingfaint sourceson topofsmall galaxies, the Figure 3 displays the results from this calculation. For difficulty of obtaining high-quality spectra of faint sources ′′ very nearby galaxies, where the scale is 0.1 kpc/ , KAIT on top of relatively bright galaxies, or a preference to ob- observes 10, 50, and 100% of the area within 83, 37, and tain multi-epoch, high-quality spectra of SNe Ib with large 23kpc,respectively.SincemanyCa-richSNehavepeakab- offsets. solute magnitudes of M ≈ −16 (P10; K12), KAIT will be PTF not detectingornot classifying someCa-rich SNe magnitude limited for Ca-rich SNe beyond ∼80 Mpc. At projected on top of galaxies may also explain the “low” this distance, the scale is ∼0.4 kpc/′′. For this scale, KAIT rateforCa-richSNerelativetotheKAITratemeasurement observes 10, 50, and 100% of the area within 333, 149, and (P10; K12). 93 kpc, respectively. A typical Ca-rich host galaxy within the KAIT sample may have ∼0.3 kpc/′′, where KAIT ob- serves 10, 50, and 100% of the area within 250, 112, and 5 VELOCITY SHIFTS 70 kpc, respectively. The farthest offset in the Ca-rich SN sample is Forbidden lines, which are produced by optically thin ma- PTF11kmbwithanoffsetof150kpc.OftheremainingPTF- terial, are extremely useful for understanding the velocity (cid:13)c 2013RAS,MNRAS000,1–?? 6 Foley distribution of SN ejecta along our line of sight. In partic- ular, forbidden emission lines provide the opportunity to Weighted measure the bulk velocity offset of the SN ejecta along our line of sight. Peak Gaussian For Ca-rich SNe, there are two optical forbidden line 10 complexes which may be studied, [OI] λλ6300, 6363 and [CaII] λλ7291, 7324. As a defining property of this class, the SNe have much stronger [CaII] lines than [OI]. As a result,themajority ofoursamplehasweakorundetectable nt 8 [OI] lines. However, the [CaII] lines are sufficiently strong a t for proper kinematic measurements in all spectra used for s n this analysis. o C 6 + 5.1 Velocity Measurement Method λ f For a single emission line, there are three typical measure- e ments used to determine the bulk velocity offset: the peak v of emission, the center of a Gaussian fit to the emission ati 4 l profile, and the emission-weighted velocity. If a line profile e R is Gaussian, the same velocity would be measured from all three methods. However, if there is significant skewness to theprofile or multiple peaks, themethods could differ. 2 For [CaII] λλ7291, 7324 in SN spectra, the doublet is not typically resolved, and this is the case for Ca-rich SNe as well. However, the lines are separated enough such that a single Gaussian is generally not a good description of the 0 data. Moreover, the line profiles are generally moderately skewed, making even the sum of two Gaussian profiles (off- −5 0 5 set to match the wavelength offset of the doublet) a poor Velocity (103 km s−1) description of thedata. Asanexample,Figure4showsthe[CaII]λλ7291,7324 doublet for SN 2005E. For all line profiles, we subtract the Figure4.Continuum-subtracted[CaII]λλ7291,7324profilefor underlying continuum by linearly interpolating across the SN 2005E (black curve). The spectrum is also reflected across zerovelocity(dotted greycurve)todemonstratetheskewnessin line profile. The shape of the profile is complicated and far the tails of the feature. A best-fit double Gaussian profile with from Gaussian. The profile peaks blueward of the nominal rest wavelengths corresponding to those of the [CaII] doublet is centralwavelengthwithalargedropinemissionaroundzero shownasthereddashedcurve.Thevelocityshiftscorresponding velocity. Although this has been an indication of dust for- to the peak of the smoothed line profile, the emission-weighted mation in SN ejecta (e.g., Smith et al. 2008), the long tail center oftheprofile,andtheshiftofthedouble-Gaussianprofile of redward emission (extending perhaps 3000 km s−1 fur- aremarkedwithsmallverticalblue,gold,andredlines. ther to the red than the blue) suggests that the line profile cannot beexplained simply bydust reddening.Instead,the profileislikelyindicativeofacomplexvelocitystructurefor be influenced by small density perturbations in the ejecta, theSN ejecta. andisstronglyaffectedbynoiseinthespectra.Wetherefore ClearlyasingleGaussianisnotanappropriatedescrip- donot use thismethod of our analysis. tion of the [CaII] profile for SN 2005E. We have also at- The final possibility is the emission-weighted velocity. tempted to fit the profile with two Gaussians where their Relative to this velocity, half of the emission is blueshifted rest-framewavelengthoffsetscorrespondtothedifferencein (andhalfredshifted).Thismethodusesallinformationfrom wavelengths for the doublets, and their widths and heights the line profile, is not affected by the two emission lines, areconstrainedtobethesame.Theseconstraintsshouldre- is not significantly affected by noise, and is not affected sult in a perfect match to the data if the emitting material by skewness. Moreover, this velocity has a specific physical were described by a simple velocity distribution. However, meaning.Iftheejectaarecompletelyopticallythinandeach SN2005Ehasamorecomplexvelocitydistribution.Because Ca atom in the ejecta has an equal probability of emitting of the poor Gaussian fits for many spectra in our sample, aphotonatthesewavelengths,thisvelocitywill correspond we do not use this method for ouranalysis. to the systemic line-of-sight velocity. As this is clearly the Asecondpossibilityistousethepeakofemission.Since bestvelocity measurement fordetermininganyline-of-sight most spectracontain noiseatthelevelthatwouldaffect re- velocity shifts, we use this method for our analysis. sults(orpossiblysmallscalestructure),wesmooththespec- trabeforedeterminingthepeak.ForSN2005E,thevelocity of the peak of emission is offset from the best-fit Gaussian 5.2 Systematic Uncertainties velocity by1100 kms−1.Thepeak oftheemission isapar- ticularlypoortraceroftheentireSNejectasinceitlacksany While the method chosen to measure of the velocity shift informationabouttheshapeofthevelocitydistribution,can is relatively robust, we investigate possible systematic un- (cid:13)c 2013RAS,MNRAS000,1–?? Progenitors of Ca-Rich SNe 7 certainties, which dominate the uncertainty of the velocity Additionally, we examined the consequences of phase measurement. onthevelocityshift.Unfortunately,onlyoneCa-richSNhas To determine the uncertainty, we performed a Monte sufficientdatatoexamineanyevolution.SN2010et hastwo Carlo simulation for each spectrum. For each spectrum, we nebular spectra of reasonable quality that are separated by smoothedthe[CaII]lineprofileandmeasuredtheemission- morethanaweek.Atepochsofroughly62and87daysafter weighted velocity. Then using noise properties matched to peakbrightness,wemeasureavelocityshiftof−330±60and thedata, we simulated 1000 spectra for each spectrum. We −430±60 kms−1, respectively. These measurements differ measured the emission-weighted velocity of each simulated by only 1.1 σ, but may represent a slight velocity gradient spectrum,randomlychangingtheregionsusedtodetermine of −4±3 kms−1day−1. the continuum. There was no bias in these measurements, SNe Ia have been shown that their velocity shifts get with most shifts from the noise-free spectrum being one or redder with time (e.g., Silverman et al. 2013), eventually twopixels.Thisdemonstratestherobustnessofthismethod plateauingataparticularvelocityastheejectabecomeopti- to noisy spectra and with different choices for the contin- callythin.WhilethisisapossibilityforCa-richSNe,wenote uum. We use the standard deviation of the shifts from the thatSN2012hn, whichhasthelargest velocityoffset and is noise-free measurement as the uncertainty in the emission- blueshifted also has one of the latest spectra (+150 days; weighted velocity. Valentiet al. 2014). Thereisanadditionalpotentialsystematicuncertainty Currently, there are not enough data to determine the related tothecontinuum subtraction.Ourmethodassumes velocity evolution of Ca-rich SNe with phase. Nonetheless, alinearunderlyingcontinuum.However,ifthecontinuumis we note that the phase of a SN’s spectrum and its galactic dominated by agalaxy spectrum, alinear continuum is not offsetareuncorrelated (correlation coefficient ofr=0.006), necessarily appropriate. An example of likely host-galaxy suggesting any potential correlation between offset and ve- contaminationforaSNspectrumisshowninFigure5,where locity isnot caused bya correlation between phase andoff- we show the spectrum of SN2003dg. set. We examine the effect of subtracting a galaxy contin- uumratherthanalinearcontinuum.Forthistask,wechoose anelliptical templatespectrum,whichappearstobeanex- cellent match to the continuum for SN 2003dg, and a red- 5.3 Measured Velocity Shifts denedSctemplatespectrumsinceitshostgalaxyisclassified as Scd. Using the elliptical and Sc template spectra result In Figure 6, we present the continuum-subtracted [CaII] invelocityshiftsthataresystematically 240and80 kms−1 λλ7291, 7324 profiles for thefull Ca-rich SN sample. There bluerthanthelinearcontinuumresultinginavelocity shift is a large diversity to the line profiles with roughly equal of −1360 and −1200 kms−1, respectively, compared toour numbers having symmetric, skewed blue, and skewed red nominal value of −1120 kms−1. profiles. We have examined all spectra, and only two other For each spectrum in Figure 6, we mark the emission- spectra potentially have host-galaxy contamination. For weighted velocity.We present these data in Table A1. SN2001co,thecontinuumfluxislowenoughthatthechoice For four SNe (2005E, 2010et, PTF09dav, and of continuum does not result in any difference in velocity PTF11bij),K12measured[CaII]velocityshifts.Theirmea- shift. For SN 2003dr, both the elliptical and Sc templates surements deviate from ours by −80 kms−1 on average, result in a systematic blueshift of 80 kms−1. with a median difference of −69 kms−1. The largest dif- Whileitispossiblethatourchoiceofcontinuumisbias- ference is for PTF09dav, where we measure a shift of ingthemeasuredvelocityshiftsforSNe2003dgand2003dr, −110±60 kms−1, while K12 measure 250 kms−1. While we choose to use the results from having a linear contin- K12 do not describe their method for measuring the veloc- uum.Thischoiceisconservativegivenourresults(theother ity shift or list any uncertainties, this represents at most a choices strengthen our findings). It also provides an easier 5.9-σ difference (using only theuncertaintyassociated with path to reproducing themethodology with futuredatasets. our measurement). All other differences are <200 kms−1 Following the method of P10 (and references therein), indicating a general agreement in thesemeasurements. we also attempted to subtract a “photospheric” SN contin- For 9 Ca-rich SNe, we could also measure an [OI] ve- uumfromeachspectrumassumingthattheunderlyingcon- locity shift, albeit with much larger uncertainties. Perform- tinuumiscausedbyphotosphericSNemission.Wematched ing a Bayesian Monte-Carlo linear regression on the [CaII] ourspectratothoseofotherSNeIbwithsimilarspectralfea- and [OI] velocity shifts (Kelly 2007), we measure a slope turesawayfromthenebularlines.Wethenusedthesespec- and offset of 0.9±0.9 and 0.2±0.5, respectively. There- tra as the underlying continuum. Doing so did not change fore, velocity measurements from the [OI] feature are con- themeasured velocityin anycase.Forinstance,performing sistentwiththoseof[CaII], providingadditionalconfidence thesamecontinuumsubtractionusedbyP10forSN2005E, thatthe[CaII]velocityshiftsmeasureanybulkoffsettothe themeasured velocity offset shifted by <1pixel. ejecta. This is not surprising giventherelative strength of the Five of 13 Ca-rich SNe have absolute velocity shifts of nebular lines to the continuum. In the case of SN 2005E, <300kms−1 relativetotheirhostgalaxyrestframe,which which has one of the strongest continua, the continuum is are consistent with coming from galactic motion. However, only∼15%thefluxofthecontinuum-subtractedpeak[CaII] 8SNehavelargershifts,upto1700kms−1 relativetotheir flux. Therefore, even 20% changes in the continuum level host galaxy rest frame. Suchlarge velocity shiftsmustorig- result in only a ∼3% change in the flux at any given wave- inate from either an asymmetric SN explosion or extreme length. line-of-sight motion for theprogenitor. (cid:13)c 2013RAS,MNRAS000,1–?? 8 Foley Sc 14 SN 2003dg E Linear 10 12 8 10 λ f e 6 8 v i t a l Re 6 4 4 2 2 0 0 4000 5000 6000 7000 8000 −5 0 5 Rest Wavelength (Å) Velocity (103 km s−1) Figure5.Left:OpticalspectrumofSN2003dg(black).Overplottedaretemplateelliptical(red)andreddened(E(B−V)=0.35mag)Sc (blue)spectra.Thegalaxyspectranear[CaII]λλ7291,7324isnotnecessarilylinear.Right:Correspondingcontinuum-subtracted[CaII] λλ7291, 7324 profile for SN 2003dg. The black, red, and blue spectra correspond to a linear continuum, elliptical galaxy, or reddened Sc galaxy subtracted from the SN spectrum. The resultingvelocity shifts areplotted above the profile. For this example, the choice of continuumresultsinasystematicdifferenceinthevelocityshiftof−240and−80 kms−1fortheellipticalandSctemplates,respectively. 6 ANALYSIS SNe with small and large projected offsets are drawn from parent populations with different velocity shifts. This dis- 6.1 The Velocity-Offset Correlation tinction is relatively insensitive to the exact distance used For the Ca-rich SN sample, there is a large range of ve- to separate the subsamples. Any chosen separation ranging locity shifts (Section 5). However, this range changes with from 7.1–16.7 kpcproducessignificantly differentsubsam- projected offset. SNe with small physical projected offsets ples. have a much larger range of velocity shifts than those with Wealsonotethatwhilewemaybesystematicallymiss- largeprojected offsets.Inparticular,withinaprojected off- ingCa-rich SNenearthecentersof their host galaxies with set of 8 kpc, 4/7 Ca-rich SNe have absolute velocity shifts redshiftedvelocityshifts(Section4.1),theinclusionofthese of >500 km s−1, while only 1/6 Ca-rich SN beyond 8 kpc SNe, if they have a similarly extreme velocity distribution has such a large absolute velocity shift (Figure 7). as those with blueshifted velocity shifts, would only make We also display these data as cumulative distribution thetails of thedistributions morediscrepant and difference functions(CDFs)forCa-richSNewithinandoutsideapro- between the subsamples more significant. jectedoffsetof8kpc(Figure8)whichalsoshowsthediffer- This result is incredibly indicative. First, if the veloc- ent line-of-sight velocity distributions for thesesubsamples. ity shifts were primarily caused by the explosion or a bi- To examine the significance of this difference, we employ nary orbital velocity, there should be no correlation with theAnderson-Darlingstatisticaltest.TheAnderson-Darling projected offset. Therefore, the velocity shifts are almost test is similar to the Komogorov-Smirnov test except it is certainly the result of extreme line-of-sight motion for the more sensitive to differences in the tails of distributions. progenitor. To account for the SNe with extreme velocity Such a test is particularly useful for this analysis since we offsets, some progenitor systems must be given a velocity would naively expect any velocity shift distribution to be kick of &1500 kms−1. centeredatroughlyzerovelocityanddifferencestobeinthe Thereisanadditionalcluewhichindicatesthattheve- tailsofthedistributions.Ifthedistributionsarenormal,the locity shifts are not caused bytheexplosion. Velocity shifts Anderson-Darling test reduces to the Komogorov-Smirnov areseeninSNeIa(e.g.,Maeda et al.2010)andthoseshifts test. arethoughttobeabulkoffsetofthecoreoftheejectarela- The Anderson-Darling statistic for these two subsam- tivetotheouterlayers.However,SNIaexplosionsappearto plesresultsinap-valueof0.031,indicatingthattheCa-rich be constrained to have relatively small velocity shifts com- (cid:13)c 2013RAS,MNRAS000,1–?? Progenitors of Ca-Rich SNe 9 2 2 ) ) 1 1 − − s s m m k 1 k 1 3 3 0 0 1 1 ( ( y y cit cit o o el 0 el 0 V V ht ht g g Si Si − − of −1 of −1 − − e e n n Li Li −2 −2 0 50 100 150 0 5 10 15 20 Projected Offset (kpc) Projected Offset (Isophotal Radii) Figure 7. Line-of-sight velocity shifts as a function of projected offset (left) and isophotal radial offset (right) for the Ca-rich sample. Velocityuncertaintiesaretypicallysmallerthanthedatapoints.Only4/13SNehavearedshift,whileonly1/6SNewithintwoisophotal radii(and8kpc)havearedshiftand3/7outsidetwoisophotalradii(and8kpc)haveredshifts. pared to the full-width at half-maximum (FWHM) of their Ca-richSNprogenitorsystemsoriginateinthediskorhalo, emission lines. Using a sample of 22 SNe Ia (Blondin et al. offset from the galaxy’s nucleus, and (in some cases) are 2012), we find that the most extreme velocity shifts are kickedto large offsets. <20% that of their FWHM. For the Ca-rich SN sample, 3/13havemoreextremevelocityshifts(relativetoFWHM) 6.2 An Excess of Blueshifts than any SN Ia in the Blondin et al. (2012) sample. Figure 9 shows the CDFs of velocity shifts relative to ThevelocityoffsetdistributionfortheCa-richSNsampleis FWHMsfortheCa-richandSNIasamples. TheAnderson- skewedtotheblue.Nineofthe13SNeareblueshifted,while Darlingstatisticforthesetwosubsamplesresultsinap-value the remaining 4 are redshifted. This is particularly striking of 0.040, indicating that these samples are drawn from dif- when examining the 6 SNe within two isophotal radii of ferent parent populations. This result implies that either theirhost galaxies (Figure7).Only asingle SNin thissub- Ca-rich SNe come from significantly more asymmetric ex- sample is redshifted. However, 3/7 of the Ca-rich SNe with plosions than SNe Ia or there is an additional line-of-sight larger offsets are redshifted. The likelihood of having such velocitycomponentforsomeCa-richSNe.Thelatteriscon- skewed distributions are only 17.5% and 18.8% for the full sistent with a correlation between velocity shift and pro- sample and small-offset subsample, respectively. This cal- jected offset. culation accounts for the possibility that the distribution Thevelocity-offsetcorrelationfurtherindicatesthatthe could be skewed in the opposite direction as well (if only line-of-sight motion for the progenitor must be correlated consideringblueshifts,thelikelihoodishalfofthatreported with the projected offset. The most obvious explanation is above). that most progenitor systems originate near the centers of Although this is not a statistically significant result, it their host galaxies, are given a large kick, and when they is noteworthy because of a physically motivated reason for ultimately explode, have the velocity kick imprinted in the suchadifference.Iftheprogenitorsystemswerekickedfrom kinematicsoftheejecta.Foraprogenitorthattravelsalong the centers of their galaxies with velocities larger than the ourlineofsight,therewillbenoprojectedoffsetfortheSN, escape velocity of thegalaxy, then thevelocity shift will be but its line-of-sight velocity offset will belarge. Meanwhile, blueshiftedforSNeonthenearsideofthegalaxy.Insucha for a progenitor that travels perpendicular to our line of scenario, redshifted SNewill beonthefarsideof theirhost sight,theSNwillhavealargeprojectedoffsetbutavelocity galaxies,andthereforewillbemorelikelytosufferfromdust offset close to zero (although the SN explosion or orbital extinction.Thus,Ca-richSNewithredshiftedvelocityshifts motion of the progenitor system, as well as measurement willbehardertodetect,resultinginunder-representationin errors may result in some velocity offset). thefull sample. Therefore, Ca-rich SNeprogenitorsystemsdonottypi- ThisisfurthersupportformostCa-richSNprogenitors callyoriginatefromglobularclusters,in-fallingdwarfgalax- originating from near the centers of their host galaxies. If ies, the intracluster medium, or other large-offset compo- their progenitors came from bound orbits (such as from a nentsof agalaxy. globularclusterpopulation),thereshouldbeanequalnum- From the velocity-offset data alone, it is possible that berofblueshiftedandredshiftedSNeonthenearsideofthe (cid:13)c 2013RAS,MNRAS000,1–?? 10 Foley 1.0 03dg 0.8 2 kpc 12 −1.12 n 05cz o 2 kpc cti 0.6 a −1.03 r F e v 03dr ati 10 03.5 k6pc mul 0.4 dproj < 8 kpc u C 00ds 4 kpc d > 8 kpc −0.19 0.2 proj 12hn 7 kpc 8 −1.73 0.0 t n 01co −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 ta 7 kpc Line−of−Sight Velocity (103 km s−1) s n −0.47 o Figure 8. Line-of-sightvelocity shiftCDFsfor the Ca-richSNe C withaprojectedoffset<8kpc(blackline)and>8kpc(blueline). + TheSNewithsmallprojectedoffsetshaveawiderdistributionof fλ 6 line-of-sightvelocityshifts(p=0.031). e 03H v 9 kpc i t −0.18 a l e 07ke R 17 kpc 1.0 0.47 4 05E 24 kpc 0.8 −0.14 n o 11bij cti 34 kpc ra 0.6 F 0.58 e v 2 1308et kpc ulati 0.4 −0.43 m u C 09dav 0.2 42 kpc −0.11 Ca−rich SNe SNe Ia 11kmb 0.0 150 kpc 0 −0.6 −0.4 −0.2 0.0 0.2 0.26 Velocity Shift / FWHM −6 −4 −2 0 2 4 6 Figure9.CDFsofvelocityshiftsrelativetonebularlineFWHMs Velocity (103 km s−1) fortheCa-rich(black)andSNIa(red)samples.SomeCa-richSNe have much larger velocity shifts relative to the widths of their linescomparedtotheSNIasample.Thisindicatesanadditional Figure6.Continuum-subtracted[CaII]λλ7291,7324profilesof velocitycomponent.TheAnderson-Darlingstatisticforthesetwo the Ca-rich SN sample on a velocity scale relative to the host- subsamplesresultsinap-valueof0.040. galaxyframe.Thespectraareorderedbyphysical projectedoff- set.Acircleindicatestheemission-weightedvelocityofeachfea- ture. The uncertainty for each velocity measure is plotted, how- ever, most uncertainties are smaller than the plotted symbols. The offset and velocity (in 103 km s−1) are listed next to each spectrum. galaxy,andthereforenonoticeabledifferenceintheobserved population. Furthermore, for progenitors in-falling towards the host galaxy for the first time, we would expect more redshifted SNe,the opposite of what is observed. (cid:13)c 2013RAS,MNRAS000,1–??