HAWK-I infrared supernova search in starburst galaxies
M. Miluzio, E. Cappellaro, M.T. Botticella, G. Cresci, L. Greggio, F. Mannucci, S. Benetti, F. Bufano, N. Elias-Rosa, A. Pastorello, M. Turatto, L. Zampieri
aa r X i v : . [ a s t r o - ph . C O ] M a r Astronomy&Astrophysicsmanuscript no. hawki˙article˙aa˙def c (cid:13)
ESO 2018February 4, 2018
HAWK-I infrared supernova search in starburst galaxies ⋆ M. Miluzio , E. Cappellaro , M.T. Botticella , G. Cresci , L. Greggio , F. Mannucci , S. Benetti , Bufano, F. ,Elias-Rosa, N. , A. Pastorello , M. Turatto , Zampieri, L. Department of Astronomy, Padova University, Vicolo dell’Osservatorio 3, I-35122, Padova, Italye-mail: [email protected] INAF, Osservatorio Astronomico di Padova, vicolo dell’Osservatorio 5, Padova, 35122 Italy INAF, Osservatorio Astronomico di Capodimonte, Salita Moiariello 16, Napoli, 80131 Italy INAF, Osservatorio Astrofisico di Arcetri, Largo Enrico Fermi 5, Firenze, 50125 Italy Departamento de Ciencias Fisicas, Universidad Andr´es Bello, Av. Rep´ublica 252, Santiago, Chile Institut de Ci´encies de l’Espai (IEEC-CSIC), Facultat de Ci´encies, Campus UAB, Bellaterra, 08193 SpainReceived: ????; Revised: ??????; Accepted: ?????
ABSTRACT
Context.
The use of SN rates to probe explosion scenarios and to trace the cosmic star formation history received a boost from anumber of synoptic surveys. There has been a recent claim of a mismatch by a factor of two between star formation and core collapseSN rates, and di ff erent explanations have been proposed for this discrepancy. Aims.
We attempted an independent test of the relation between star formation and supernova rates in the extreme environment ofstarburst galaxies, where both star formation and extinction are extremely high.
Methods.
To this aim we conducted an infrared supernova search in a sample of local starburts galaxies. The rational to search inthe infrared is to reduce the bias due to extinction, which is one of the putative reasons for the observed discrepancy between starformation and supernova rates. To evaluate the outcome of the search we developed a MonteCarlo simulation tool that is used topredict the number and properties of the expected supernovae based on the search characteristics and the current understanding ofstarburst galaxies and supernovae.
Results.
During the search we discovered 6 supernovae (4 with spectroscopic classification) which is in excellent agreement with theprediction of the MonteCarlo simulation tool that is, on average, 5 . ± . Conclusions.
The number of supernovae detected in starburst galaxies is consistent with that predicted from their high star formationrate when we recognize that a major fraction ( ∼ ffi ciency and high extinction. Key words.
Stars: supernovae: general - Galaxies: starburst - Galaxies: star formation - Infrared: galaxies - Infrared: stars
1. Introduction
The rate of supernovae (SNe) is a key quantity in astrophysicsthat provides a crucial test for stellar evolution theory and aninput for the modeling of galaxy evolution with direct impacton the chemical enrichment and the feedback mechanism. Core-collapse SNe (SN CC), because of their short-lived progenitors,trace the current star formation rate (SFR). Conversely, for anadopted SFR, measurements of the SN CC rates give informa-tion on the mass range of their progenitors as well as the slopeof the initial mass function at the high mass end. SN Ia, resultingfrom the thermonuclear explosion of a white dwarf in a binarysystem, show a wide range of delay times from star formation toexplosion. Therefore, the SN Ia rate reflects the long-term starformation history of the parent stellar system. Recently, it hasbeen claimed that a significant fraction of SN Ia have a short de-lay time, possibly as short as 10 years (Mannucci et al. 2006).Like for CC SN, the rate of such prompt SN Ia events is expectedto be proportional to the current SFR.In one of the early attempts to compare the SN and SF rates,Cappellaro et al. (1999) found that the SN CC rate in galaxieswith di ff erent U − V color matches the predicted SFR when ⋆ ESO proposal: 083.D-0259, 085.D-0335, 085.D-0348, 087.D-0494, 087.D-0922. GTC proposal: GTC50-11B adopting a mass range 10 M ⊙ < M <
40 M ⊙ for the SN CCprogenitors.In the last decade there was a enormous improvement in themeasurement of the cosmic SFR with the careful combinationof many di ff erent probes (eg. Hopkins & Beacom 2006). A mostrelevant feature is that the SFR reaches a maximum at a redshift z ∼ ff ort was also devoted to the measurementof the cosmic SN rate: although much of the focus wasfor type SN Ia, a few estimates of the SN CC rates werealso published both for the local Universe (Li et al. 2011a)and at high redshifts (Dahlen et al. 2004; Cappellaro et al.2005; Botticella et al. 2008; Bazin et al. 2009; Graur et al. 2011;Melinder et al. 2012; Dahlen et al. 2012). While the new localSN CC rate confirms previous results, with a much better statis-tics and lower systematic errors, the evolution with redshift wasfound to track very well the SFR evolution, considering the largeuncertainties in the extinction corrections. Again, to best matchthe observed SN and SF rates it was argued that the lower limitfor SN CC progenitor had to be ∼
10 M ⊙ (Botticella et al. 2008;Blanc & Greggio 2008).At about the same time, following a di ff erent line of research,the analysis of archival images allowed the identification of the
1. Miluzio et al.: HAWK-I SN search precursors for a number of nearby SN CC. From the often veryscanty but precious photometry, and using stellar evolution mod-els, one can estimate the SN precursor mass. The uncertain-ties are in general quite large, as confirmed from the discrep-ancy in the mass estimates from di ff erent groups, but this anal-ysis suggests a lower limit for SN CC progenitors of 8 ± ⊙ (Smartt 2009). If this value is adopted, the observed SN rateswould result a factor two smaller than those expected from theobserved SFR. This was identified by some authors as a ”SNrate problem” (e.g. Horiuchi et al. 2011. While one should re-mind that the uncertainties on SFR rate calibrations are still large(Botticella et al. 2012; Kennicutt & Evans 2012), it also true thatthere is a number of possible biases in the SN rate estimates. Thetwo most severe are the possible underestimate of a large popu-lation of faint SN CC and / or the underestimate of the correctionfor extinction (Horiuchi et al. 2011; Mattila et al. 2012).In particular, Mannucci et al. (2007); Cresci et al. (2007)and, more recently, Mattila et al. (2012) argued that a significantfraction of SN CC remains hidden in the nuclear region of star-burst galaxies, with a loss of up to ∼ / LIRGs). This e ff ect is expected to be more important athigh redshift because of the larger fraction of starburst galaxies.Indeed, when a correction for this hidden SN fraction is includedin the rate calculation the discrepancy between SN and SFrates at high redshifts seems to disappear (Melinder et al. 2012;Dahlen et al. 2012; the ”missing fraction” correction adopted inthese works was from Mattila et al. (2012). It is currently un-clear if this e ff ect is large enough to explain also the discrepancyobserved in the local Universe with somewhat conflicting evi-dences from the statistics of SNe in the Local Group galaxies(Botticella et al. 2012; Mattila et al. 2012) and large sample SNsearches (Li et al. 2011a).Entering in this debate, we planned for an infrared SN searchin a sample of local starburst galaxies (SBs). The idea was toverify the link between SN and SF rates in an environment wherestar formation is very high, 1-2 order of magnitude higher thanin normal star-forming galaxies. By observing in the K-band wewere aiming to reduce the bias due to extinction (A K ∼ . V ).The idea is not new. A first attempt of a dedicated SN searchin SBs was performed in the optical band by Richmond et al.(1998). During the search only a handful of events were de-tected leading the authors to conclude that the rate of (unob-scured) SNe in SBs is the same as in quiescent galaxies. A sim-ilar conclusion was reached by Navasardyan et al. (2001), againbased on optical data. As for infrared SN search, after a few un-successful attempts (Grossan et al. 1999; Bregman et al. 2000),the first results of a systematic search in SBs were reported byMaiolino et al. (2002) and Mannucci et al. (2003). They foundthat the observed SN rate in SBs was indeed one order of mag-nitude higher then expected for the galaxy blue luminosities butstill 3-10 times lower than would be expected from the far in-frared (FIR) luminosity. Among the possible explanation for theremaining discrepancy, they suggested extreme extinction in thegalaxy nuclear regions (A V > ffi cient spatial resolution to probethe very nuclear regions. The reliability of the use of NIR searchfor obscured SNe in the nuclear and circumnuclear regions of ac-tive starburst galaxies was also investigated by Mattila & Meikle(2001) taking into account in particular the problem of extinc-tion. They conclude that with a modest investment of observa-tional time it may be possible to discover a number of nuclearSNe. A negative search for transients in NICMOS images re-trieved from the Hubble Space Telescope archive suggests that the same biases likely a ff ect also space-based, high spatial reso-lution observations (Cresci et al. 2007).The same approach was used by Mattila et al. (2007b) butwith ground based, adaptive optics (AO) assisted observations.The application of this technique led to the discovery of a hand-ful of SNe (Kankare et al. 2008, 2012) but not yet to an estimateof the SN CC rate.Until now, about a dozen SNe have been discovered by IRSN searches, not all with spectroscopic confirmation. The num-ber is higher if we include also events first detected in the opticaland re-discovered by the IR searches. Therefore the statistics isstill very low and many of the original questions are still unan-swered. This gave us the motivations to make a new attempt ex-ploiting the opportunity o ff ered by HAWK-I, the infrared cameramounted at the ESO VLT telescope.The paper is divided in two parts: the first part describe theobserving program, namely the galaxy sample and the searchstrategy in Sect. 2.1, the data reduction in Sect. 2.3, the SNdiscoveries and classification in Sect. 2.4 while in Sect. 2.5 wedetail the procedure to estimate the search detection e ffi ciency.The second part is devoted to the description of a simulationtool which is used to predict, based on our current knowledge ofSBs properties and on the specific features of our SN search, thenumber of expected SN detections (Sect. 3). Finally, we comparethe number and properties of the expected and observed events(Sect. 4) and draw our conclusions (Sect. 5).Throughout this paper we assume the following cosmolog-ical parameters: H =
72 km s − Mpc − , Ω Λ = .
73 and Ω M = .
2. The SN search program
Starbursts are galaxies with very high star formation rate, of theorder of 10-100 M ⊙ yr − compared to the few M ⊙ yr − of nor-mal star forming galaxies in the local universe. Given that in atypical galaxy the very high SFR will rapidly consume the gasreservoir, it is thought that the starburst is a temporary phase inthe galaxy evolution. The fact that many SBs are in close pairs orhave disturbed morphologies point to the interaction as a dom-inant, although possibly not unique, reason of the phenomena(Gallagher 1993). The ultra-violet radiation from young, mas-sive stars heats the surrounding dust and is re-emitted in the farinfrared. Indeed the most luminous SBs in the local Universeare LIRGs , with 11 < log( L IR / L ⊙ ) <
12, and ULIRGs, withlog( L IR / L ⊙ ) >
12 (Sanders & Mirabel 1996).For our project we selected from the IRAS Revised BrightGalaxy Sample (Sanders et al. 2003) a sample of SBs with to-tal infrared (TIR) luminosity log( L TIR / L ⊙ ) >
11 and redshift z < .
07. With the additional requirement that the targets are ac-cessible from Paranal in the April to September observing sea-son (to fit in one of the ESO allocation period) we retrieved asample of 30 SBs.The list of SBs is reported in Tab. 1. Along with the galaxyname and equatorial coordinates (cols. 1-3) we report the helio-centric redshift (col. 4), log L TIR and log L B (cols. 5 and 6; cf.Sect. 3.1.1), the Hubble type (col. 7), the SFR and the expectedSN rates (cols 8, 9) derived from L TIR as described in Sect. 3.1.1.Galaxy data have been retrieved from NED . In the last column The NASA / IPAC Extragalactic Database (NED) is operated by theJet Propulsion Laboratory, California Institute of Technology, undercontract with the National Aeronautics and Space Administration.2. Miluzio et al.: HAWK-I SN search
10 11 12 logL/L ⊙ L TIR L B Fig. 1: Distribution of the B and FIR luminosities for the SBgalaxies of our sample.we listed (in boldface) the designation of the SNe discovered inour search which are the basis for our analysis. For completenesswe also list (in italics) the SNe discovered by other SN searchesoutside our monitoring period. The distribution of L B and L TIR are compared in Fig. 1 showing that, as typical for SBs, L TIR ison average a factor ten higher than L B , whereas for normal starforming galaxies L TIR ∼ L B . We notice that almost all galax-ies are LIRGs and only two are ULIRGS. Most galaxies of thesample are isolated ( ∼ − / interacting galaxies or contain double nuclei, signature of arecent merger. Several galaxies of the sample are asymmetrical,disturbed, or show warps, bars and tidal tails. To search SNe in the selected SB sample we used the HAWK-Iinstrument installed at the ESO VLT telescope at Cerro Paranal(Chile). HAWK-I is a NIR (0 . − . µ m ) wide-field imagerwith a mosaic of four Hawaii-2RG detectors. The total field ofview is 7 . ′ × . ′ with a scale of 0 . ′′ / pix. Even in poorseeing conditions ( > . / N ∼
10 for a K =
20 magnitude star with a 15 min exposure.The infrared light curves of SNe evolve relatively slowly,remaining within one / two magnitudes from maximum fortwo / three months (Mattila & Meikle 2001) and therefore an IRSN search does not require frequent monitoring. We planned foran average of three visits per galaxy per semester, for a total of80-100 visits. The monitoring campaign was scheduled in ser-vice mode and we did not set tight constraints for the sky con-ditions. This and the relatively short duration of the observingblocks made the program well suited as filler. We notice that wehad no influence on the actual scheduling of the observationswhich followed the rules of the ESO service mode scheduler.Eventually, the fraction of useful observing time was 100%of the allocated time in the first season, and 70% in the second and third semesters. The log of the observations is reported inTab. 2 where for each galaxy we list the epoch of observations(MJD), the seeing (FWHM in arcsec), and the minimum andmaximum magnitude limit for SN detection across the image(cf. Sec. 2.5). In total, we obtained 210 K-band exposures (ex-posure time 15min), with an average of about 3 visits per galaxyper semester. Because of the time loss, three galaxies were notmonitored in the last two seasons.It turned out that the average image quality was quite good:for ∼
90 % of the exposures the seeing was less than 1 . ′′ , withan average FWHM across the whole program of 0 . ′′ . For data reduction and mining of the HAWK-I mosaic images wedeveloped a custom pipeline that integrates di ff erent, publiclyavailable, recipes and tools in a Python environment.The pipeline consists of four sections:1. pre-reduction, astrometric calibration and production ofthe stacked mosaic image. For these steps we use theESO HAWK-I pipeline recipes in
EsoRex , the ESO RecipeExecution Tool ;2. subtraction of images taken at di ff erent epochs using ISIS (Alard 2000) for the PSF matching;3. search for transient candidates in the di ff erence image using Sextractor (Bertin & Arnouts 1996). The candidates wereranked based on their
Sextractor measured parameters andsubmitted to the operator for visual inspection and valida-tion;4. estimate of the detection e ffi ciency through artificial star ex-periments performed for each of the search images (detailsin Sect.2.5).The raw images were retrieved from the ESO archive as soonas they became available, and immediately reduced to allow foractivation of follow-up spectroscopy of transient candidates.For the pre-reduction, we followed the reduction cascade de-scribed in the HAWK-I pipeline manual including dark subtrac-tion, flat field and illumination corrections, background subtrac-tion, distortion correction, astrometric o ff set refinement, com-bination of the di ff erent exposures and stitch of the 4 detec-tors in a single mosaic image. Actually, it turned out that theESO pipeline recipes for background subtraction and o ff set re-finement do not provide satisfactory results for our images. Themain reason is the extended size of our sources and the conse-quent large dithering we had adopted. To address this issue weimplemented custom recipes for the two afore mentioned reduc-tion steps.The most critical step of the data reduction is the image sub-traction, in particular in the proximity of the nuclear regionsof the galaxies. First of all we need to choose a proper refer-ence image, usually the image with the best seeing obtained atleast three month before (or in some case after) the image to besearched. We also need to choose the proper parameters for theimage di ff erence procedure (see Melinder et al. 2012 for an ex-tensive discussion). An additional problems arises because in thedistributed version of ISIS , the program automatically selects thereference sources for the computation of the convolution kernel.Owing to the small number of sources in our extragalactic fields, http: // / sci / software / cpl / esorex.html ftp: // ftp.eso.org / pub / dfs / pipelines / hawki / hawki-pipeline-manual-1.8.pdf 3. Miluzio et al.: HAWK-I SN search Table 1: The SB galaxy sample. The last column report our 6 SNe (in bold) with other SNe discovered previously in the galaxysample. galaxy R.A. Dec. redshift log L TIR log L B Hubble SFR SN rate SNedesignation J2000.0 [L ⊙ ] [L ⊙ ] type [M ⊙ yr − ] [SN yr − ]CGCG011-076 11 21 13.3 -02 59 08 0.025 11.28 10.35 2.9 32.1 0.38CGCG043-099 13 01 49.9 +
04 20 01 0.037 11.59 10.51 3.4 65.4 0.77ESO148-IG002 23 15 46.6 -59 03 14 0.045 11.94 10.82 7.9 148.8 1.75ESO239-IG002 22 49 39.6 -48 51 01 0.043 11.75 10.88 -0.1 95.7 1.13ESO244-G012 01 18 08.6 -44 27 40 0.023 11.32 10.20 5.3 35.7 0.42ESO264-G036 10 43 07.0 -46 12 43 0.023 11.24 10.70 1.5 29.1 0.34ESO286-IG019 20 58 27.4 -42 38 57 0.043 11.95 11.13 10.0 151.6 1.78ESO440-IG058 12 06 53.0 -31 57 08 0.023 11.33 10.17 99.0 36.3 0.43ESO507-G070 13 02 51.3 -23 55 10 0.021 11.44 10.67 6.8 46.6 0.55IC1623A / B 01 07 46.3 -17 30 32 0.020 11.63 10.42 6.0 72.7 0.86
PSN 2011
IC2545 10 06 04.2 -33 53 04 0.034 11.66 10.58 -0.1 77.7 0.92IC2810 11 25 47.3 +
14 40 23 0.034 11.60 10.61 1.5 68.0 0.80IC4687 / PSN 2010
IRAS12224-0624 12 25 02.8 -06 40 44 0.026 11.30 9.83 2.9 34.0 0.40IRAS14378-3651 14 40 57.8 -37 04 25 0.068 12.13 10.40 5.1 233.0 2.74IRAS16399-0937 16 42 39.2 -09 43 11 0.027 11.55 10.39 10.0 60.3 0.71IRAS17207-0014 17 23 21.4 -00 17 00 0.043 12.42 10.32 -50. 447.5 5.27IRAS18090 + +
01 31 40 0.029 11.63 10.54 2.0 71.7 0.84MCG-02-01-051 / MCG-03-04-014 01 10 08.5 -16 51 14 0.035 11.59 10.53 -5.0 66.1 0.78NGC0034 00 11 06.6 -12 06 27 0.020 11.43 10.34 -1.0 45.7 0.54NGC0232 00 42 46.5 -23 33 31 0.020 11.51 10.71 1.1 55.7 0.66
NGC3110 10 04 02.7 -06 28 35 0.017 11.29 10.94 3.3 33.0 0.39NGC5010 13 12 25.4 -15 47 45 0.021 10.84 9.79 -1.0 11.8 0.14NGC5331 13 52 16.6 +
02 06 08 0.033 11.60 10.92 3.0 67.1 0.79NGC6240 16 52 58.6 +
02 24 03 0.024 11.81 10.89 -0.3 108.4 1.28 , NGC6926 20 33 04.8 -02 01 39 0.020 11.25 11.38 5.6 30.4 0.36NGC7130 21 48 19.6 -34 57 05 0.016 11.34 10.75 1.2 37.1 0.44
NGC7592 23 18 22.2 -04 24 56 0.024 11.36 10.51 -1.0 38.4 0.45NGC7674 23 27 56.9 +
08 46 46 0.029 11.37 10.92 1.1 40.0 0.47 , , the reference source list in general includes the bright galaxy nu-cleus which, being very bright, has a significant weight in thedetermination of the kernel. This may cause some problems be-cause if at one epoch a SN occurs very close to the galaxy nu-cleus it can be included in the convolution kernel and e ff ectivelycancelled in the di ff erence image. We therefore modified the ISIS selection procedure to allow for exclusion of specific sources, inparticular the galaxy nuclei, from the reference list.Despite the e ff orts in many cases the di ff erence image showssignificant spurious residuals in correspondence to the galaxynuclear regions. The problem is most severe in case of im-ages with poor seeing ( FWHM > ′′ ) and / or reduced trans-parency. This is illustrated in Fig. 2 where we show two exam-ples of image di ff erence one for a search image with poor seeing( FWHM = . ′′ , left panel) and the other for a case with excel-lent seeing ( FWHM = . ′′ ). In both cases, the reference imagewas the same and had excellent seeing ( FWHM = . ′′ ).False detections due to residuals of the image subtraction werelargely removed by the requirement that the candidate had to bevisible at least in two consecutive epochs. During our monitoring campaign 6 transients were detected inat least two consecutive epochs separated by at least one month(finding charts are in Fig. 3). Four of them were spectroscopi-cally confirmed as SNe (three SN-CC and one SN Ia) and wewill argue in the following that also the other two transients, la-beled as probable SN (PSN), are likely SN CC (Tab. 3). SNe2010bt and 2010gp were discovered and announced before our Fig. 2:
Left panel : example of poor subtraction of images ofNGC 7130 with large seeing di ff erences ( FWHM = . ′′ , with FWHM = . ′′′ for the reference). Right panel : optimal subtrac-tion for two images with similar, good seeing (
FWHM = . ′′ ).In both panels the source in the lower left quadrant is SN 2010bt(cf. Fig. 3). The FOV in both panel is about 2’ × ff sets fromthe galaxy nucleus and projected linear distances from thegalaxy nucleus.For all transients K-band magnitudes were measured throughaperture photometry on the di ff erence images and calibratedwith respect to 2MASS stars in the field. Upper limits measuredon pre-discovery images were also estimated. For all transients,
4. Miluzio et al.: HAWK-I SN search
Table 2: The log of the observations with the epoch of observations (MJD), the seeing (FWHM in arcsec), and the minimum andmaximum magnitude limit for SN detection across the image (cf. Sec. 2.5).
CGCG011-076 ESO440-IG058 IRAS17207-0014 NGC0232
JD Seeing m limmax m limmin JD Seeing m limmax m limmin JD Seeing m limmax m limmin JD Seeing m limmax m limmin CGCG043-099
NGC3110
JD Seeing m limmax m limmin limmax m limmin ESO507-G070 limmax m limmin ESO148-IG002 limmax m limmin IRAS18090 + IC1623A
JD Seeing m limmax m limmin NGC5010 limmax m limmin limmax m limmin NGC5331ESO239-IG002 IC2545 limmax m lim min JD Seeing m limmax m limmin JD Seeing m limmax m limmin MCG-02-01-051 limmax m limmin IC2810
NGC6240 limmax m limmin limmax m limmin IC4687
ESO244-G012
JD Seeing m limmax m limmin MCG-03-04-014 limmax m limmin limmax m limmin NGC6926 limmax m limmin IRAS12224-0624 limmax m limmin ESO264-G036
NGC0034 limmax m limmin limmax m limmin IRAS14378-3651 limmax m limmin NGC7130 limmax m limmin IRAS16399-0937 limmax m limmin ESO286-IG019
NGC7674 limmax m limmin limmax m limmin NGC7592 limmax m limmin NGC0232 limmax m limmin Table 3: Information for the detected SNe.
SN2010bt SN2010gp SN2010hp SN2011ee PSN2010 PSN2011Host galaxy NGC7130 NGC6240 MCG-02-01-52 NGC7674 IC 4687 IC 1623ADistance modulus 34.08 35.07 35.24 35.39 34.21 34.54R.A. (SN) 21:48:20.22 16:52:57.39 0:18:50.01 23:27:57.34 18:13:40.213 01:07:46.229Dec. (SN) 34:57:16.5 2:23:16.4 -10:21:40.6 + ff set [ ′′ ] 9E 14S 22W 47S 2.6W 2.7N 9.3E 6.3S 2.6E 2.8N 4E 7S r [Kpc] 5.3 25.9 2.1 8.9 6.5 3.2SN type IIn Ia IIP Ic IIP Ic R max [mag] 15 . ± . . ± . . ± . . ± . . ± .
5K max [mag] 16 . ± . . ± . . ± . . ± . . ± . . ± . AB G [mag] 0.44 0.33 0.16 0.25 0.43 0.07 AB H [mag] 1 . ± . . ± . . ± . ∼ . − . ± . M R -19.2 -19.2 -17.9 -18.1 -17.9abs M K -18.1 -18.2 -18.0 -17.7 -16.3 / -17.3 -17.5MJD max R 55303 ± ± ± ± ±
20 55725 ± but PSN2010 in IC4687, we obtained some follow-up imagingin the optical or near-infrared domains. These observations werereduced using standard procedures in IRAF . When a referenceimage was not available, the SN magnitude was measured usingthe PSF fitting technique. Optical band magnitudes were cali-brated with respect to Landolt’s standard fields. Our photometryfor the six transients is reported in Tab.4. For the two transientswith no spectroscopic confirmation, the photometry will be usedto assess their nature.Spectroscopic observations were obtained for four candi-dates: epoch, spectral range and instruments are reported inTab.5. Data were reduced using standard procedure in
IRAF but for the X-Shooter spectra which were reduced using ver-sion 1.0.0 of the ESO X-shooter pipeline (Goldoni et al. 2006)with the calibration frames (biases, darks, arc lamps, and flatfields) taken during daytime. The extracted spectra, after wave-length and flux calibration, were compared with a library of tem-plate spectra using the
GELATO
SN spectra comparison tool(https: // gelato.tng.iac.es / , Harutyunyan et al. 2008). The best fittemplate SN, the SN type and phase are reported in Tab.5.Spectroscopic classification for PSN2010 in IC4687 was at-tempted, but the observed spectrum resulted too noisy for a safeclassification. The table includes the result of the spectroscopicobservations of SN 2010gp from Folatelli et al. (2010).We have used the available photometry and spectroscopy toput some constraints to the amount of extinction su ff ered by theSNe. Hereafter we will describe in some details the sparse infor-mation available for each transient. SN 2010bt was discovered on 2010 April 17.10 UT by Monard(2010). A spectrum taken on April 18.39 UT (Turatto et al.2010) shows strong resemblance to several type-IIn SNe, inparticular SN 1996L (Benetti et al. 1999) shortly after explo-sion (Fig. 4). A broad H α component is present indicating anexpansion velocity of about 3500 km s − (half width at zerointensity). SN 2010bt was independently re- discovered byus on May 25 and was observed in other 2 epochs. The ob-ject was not visible on a HAWK-I image taken in 2009 July26 (limit K = . A B = . ± . SN 2010gp was discovered on 2010 July 14.10 UT byMaza et al. (2010) with the 0.41-m PROMPT1 telescope lo-cated at Cerro Tololo. Folatelli et al. (2010) reported thespectroscopic classification as a type-Ia SN around maxi-mum light and with high expansion velocity of the ejecta.SN 2010gp was re- discovered independently by us on July21 and was observed in other 2 epochs. The object was notvisible on a HAWK-I image taken on 2010 May 26 (limit K = . A B = . SN 2010hp was discovered on a HAWK-I image taken on 2010July 21.3 UT (Miluzio & Cappellaro 2010). The object wasnot detected on 2009 Aug. 25 (limit K = . +
60 days, adopting a red-dening of about A B = . SN 2011ee was discovered on 2011 June 27.3 UT(Miluzio et al. 2011). The object was not detected on aK-band image taken on 2010 September 7 ( K > . PSN2010 in IC 4687 was discovered on 2010 May 21.3 UT inthe northern component of a galaxy triplet that include alsoIC 4686 and, 1 arcmin to the south of IC 4687, IC 4689.IC 4687 has a chaotic structure formed byf stars, gas anddust and a large curly tail. The transient was not detected ona K-band image taken on 2009 Aug. 8.1 ( K > .
6. Miluzio et al.: HAWK-I SN search (a) SN 2010bt (b) SN 2010gp (c) SN 2010hp (d) PSN2010 in IC 4687 (e) SN 2011ee (f) PSN 2011 in IC1623A
Fig. 3: K-band finding charts for the SNe of our list. The inserts show the transients as they appear in the di ff erence image.
7. Miluzio et al.: HAWK-I SN search
Table 4: Transient photometry. Estimated errors are given in parentheses.
MJD B V R I J H K Instr.
SN 2010bt
SN 2010gp > . > > > > > > > PSN 2010 > PSN2011 > > > = HAWK-I@VLT, E = EFOSC2@NTT, S = SOFI@NTT, L = RATCam@Liverpool, D = Dolores@TNG
Table 5: Log of spectroscopic observations.
SN MJD range(nm) res. (nm) Instrument best fit type Phase2010bt 55304.4 350-1000 1.4 EFOSC / NTT 2005gj IIn max2010hp 55454.2 350-1000 1.4 EFOSC / NTT 1999em Ia + / VLT S / N too low2011ee 55824.6 2.5 OSIRIS / GTC 2007gr Ic + / VLT 1994I Ic max2010gp 55401.0 354-886 Folatelli et al. (2010) 2002bo Ia just before max
We obtained an optical / infrared spectrum with X-Shooter atVLT on 2010 June 5. However, because of its very low S / N,we could not derive a convincing classification and there-fore we had to rely on the K-band photometry. Comparingthe K band absolute light curve of PSN2010 ( A B (host) = ff erent SN types we found agood match with SN 2005cs a prototype of under-luminoustype IIP SN (Pastorello et al. 2009), assuming that the detec-tion of PSN2010 was 2 months after the explosion. However,lacking color measurements, we could not constraint the ex- tinction and indeed, assuming a high extinction A B ∼ PSN2011 in IC 1623 was discovered on 2011 July 21.4 UT inthe western component of a galaxy pair. The object was notdetected on a K-band image taken on 2010 Sept. 5 ( K > .
8. Miluzio et al.: HAWK-I SN search ◦ A]0.00.20.40.60.81.0 f l u x (a) The spectrum of 2010bt, dereddened by A B = . ◦ A]0.00.20.40.60.81.01.2 f l u x (b) The spectrum of 2010hp, dereddened by A B = , ◦ A]0.00.20.40.60.81.01.21.4 f l u x +60d (c) The spectrum of 2011ee is compared to that of theSN Ic 2007gr at the maximum (top panel) and 60 daysafter the maximum. Fig. 4: Spectra of SN 2010bt, 2010hp, and 2011ee are shownalong with the best fitting templates. K a b s m a g IIP (2005cs)IIP (1999em)Ic (2007gr)Ia (2006X)
Fig. 5: K band absolute light curve of the PSN2010 (black dots),compared with those of template SNe. The empty circles showthe same data but assuming an extinction A B = A B = . ± . In order to derive the SN rate from the number of detected eventsit is crucial to obtain an accurate estimate of the magnitude de-tection limit for each of the search images and for di ff erent loca-tions in the images. As it has been shown in Fig. 2, the detectione ffi ciency is influenced by the sky conditions at the time of ob-servations (namely seeing and transparency) and by the transientposition inside the host galaxy.The magnitude limit for SN detection has been estimatedthrough artificial star experiments. The procedure we adoptedwas the following:1. fake SNe of di ff erent magnitudes are simulated with the PSFderived from isolated field stars;2. the image is segmented in a number of intensity contour lev-els. We took denser contours in the nuclear regions becausethe magnitude limit changes rapidly with background inten-sity;3. one fake SN of specific magnitude is randomly placed insidea chosen intensity contour;
9. Miluzio et al.: HAWK-I SN search K a b s m a g IIP (1999em)Ic (2007gr)Ia (2006X) R - I Fig. 6: K and R-I colours of PSN 2011, compared with those oftemplate SNe. For this plot we adopted A B = . ∆ R = . ff erence and transient detection pipeline;5. if the fake SN results in a detected transient, the experimentis repeated with a fainter artificial star until we have a nulldetection. The fainter magnitude for which the fake SN is de-tected defines the magnitude limit for the given backgroundintensity level;6. Steps 2 to 5 were repeated for each contour level three timesto enhance the statistical significance of the results. The av-erage value for each contour level has been adopted as themagnitude discovery limit for the given background inten-sity.To illustrate the results, a plot of the magnitude limit ver-sus background counts for four observations of the galaxyNGC 7130 is shown in Fig. 7. Each epoch is labelled with the im-age seeing, while the errorbar shows the range of limiting magni-tudes for the three experiments. The top x-axis shows the lineardistance in Kpc from the galaxy center.It can be seen that, as expected, the magnitude limit is lowerin the nuclear regions which, for a typical galaxy, correspond to1.5-2.0 kpc. Epochs with di ff erent seeing have similar magnitudelimits in the galaxy outskirts (typically K ∼
19 mag), while inthe nuclear region when seeing is poorer the magnitude limitis brighter (in the worst case even 5-6 mag brighter than in thegalaxy outskirts).
3. SN search simulation
To evaluate the significance of the detected events we elabo-rated a simulation tool that returns the number and propertiesof expected events based on specific features of our SN search, anumber of parameters describing our current knowledge of SBsand SN properties. The tool uses a MonteCarlo approach which
14 14.5 15 15.5 16 16.5 17 17.5 18 18.5 19 19.5 0 2000 4000 6000 8000 10000 12000 14000>5 2 1 0.3 0 M ag li m countsKpc 0.76"0.57"0.47"0.44" Fig. 7: Magnitude limit vs. background intensity for four di ff er-ent observations of NGC 7130. The points are labeled with therespective image seeing. Errorbars show the range of the magni-tude limits from the three di ff erent experiments.simulates the stochastic nature of SN explosions. By collectinga number of MonteCarlo experiments with the same input pa-rameters, we can test whether the observed events are within theexpected distribution. On the other hand by varying some of theinput parameters, we can test the influence of specific assump-tions. Our MonteCarlo (MC) simulation tool is built in a
Python en-vironment and makes use, for the di ff erent inputs, of standardvalues taken from the literature. For those that are more contro-versial, we will give references with some discussion. The basicingredients of the simulation are: – relevant data for the selected SBs, namely: redshift, galac-tic extinction, infrared fluxes from IRAS catalogues at 25,60 and 100 µ , B magnitude corrected for internal extinction ,Hubble morphological type. These data were retrieved fromNED; – information describing the SN properties for each of the SNtypes considered here: SNe Ia and core collapse events, in-cluding SNe IIP, IIL, IIn and Ib / c. K band template lightcurves were constructed starting from B template lightcurves (Cappellaro et al. 1997) and B − K k-corrections forthe given galaxy redshift (Botticella et al. 2008). We notethat the results of this procedure are in close agreementwith the K template light curves of Mattila & Meikle (2001).For the SN luminosity functions we adopted as referencethose of Li et al. (2011b), but we also tested for the possi-ble presence of a significant population of faint core collapsewhich may be suggested by the analysis of very nearby SNe(Horiuchi et al. 2011). From the LOSS project we adoptedalso the relative rates for the di ff erent SN types (Li et al.2011a); – details of the search campaign: log of observations, magni-tude detection limit for each observation as a function of thehost galaxy background intensity; – the number of SNe expected from a given star formationepisode. For core collapse SNe, this is determined only by The galaxy internal extinction was retrieved from HyperLeda(Paturel et al. 2003)10. Miluzio et al.: HAWK-I SN search the adopted mass range of the progenitors and IMF slope.In fact, for our purposes, we can neglect the very short timedelay from CC progenitor formation to explosion. For typeIa SNe we need to consider the realization factor, that is thefraction of events in the proper mass range which occurs insuitable close binary systems and the delay time distribution(Sect. 3.1.2).; – the depth and distribution of the extinction by dust inside theparent galaxies (cf. Sect. 3.1.3); – the star formation spatial distribution in the parent galaxies(cf. Sect. 3.1.4).Hereafter, we discuss our assumptions about the parametersof the simulation. The SFR in SBs can be estimated on the basis of the galaxytotal infrared luminosity (L
TIR ) under the assumption that dustre-radiates a major fraction of the UV luminosity, and after cali-bration with stellar synthesis models. In turn the TIR luminositycan be estimated from FIR flux measurements.Helou et al. (1988) provided a prescription for deriving theFIR emission from IRAS measurements:
FIR = . × − [2 . f ν (60 µ m ) + f ν (100 µ m )]where FIR is in W m − and f ν are in Jansky. FIR fluxesare converted into TIR fluxes by using the relation of Dale et al.(2001) log T IRFIR = a + a x + a x + a x + a x where x = log f ν (60 µ m ) f ν (100 µ m ) and [ a ( z = ≃ [0 . , − . , . , . , . L TIR = π D T IR
Finally, the relation between the SFR ( ψ ), and L TIR was de-rived by Kennicutt (1998) from SB galaxy spectral synthesismodel adopting 10-100 Myr continuous bursts and a SalpeterIMF as: ψ [M ⊙ yr − ] = L TIR . × [erg s − ] = L TIR . × [L ⊙ ] In general, the rate of SNe expected at a specific time, ˙ n S N ( t ),for a stellar population depends on the star formation history,the number of SNe per unit mass from one stellar genera-tion (labelled as SN productivity) and the distribution of delaytime from star formation to explosion for the specific SN type.Following the notation of Greggio (2005, 2010):˙ n S N ( t ) = Z t ψ ( t − τ ) k S N f S N ( τ ) d τ (1)where ψ ( t ) is the star formation rate, f S N is the distributionof the delay times τ and k S N is the supernova productivity. The equation shows that at a fixed epoch t since the beginning of starformation, the rate of SNe is obtained by adding the contributionof all past stellar generations, each of them weighted with theSFR at the appropriate time. Core Collapse SNe
For core collapse SNe the delay time from star formation to ex-plosion (2.5 Myr for 120 M ⊙ stars up to 40 Myr for 8 M ⊙ stars)is short compared with the typical SBs duration (200 −
400 Myr,McQuinn et al. 2009). Assuming that the SFR in the SB wasconstant during the past 40 Myr, the expected CC SN rate, ˙ n CC ,is proportional to the current SFR:˙ n CC = k CC × ψ The supernova productivity k CC is derived by integrating theIMF, φ ( m ), and assuming a CC progenitor mass ( M CC ) range: k CC = R M UCC M LCC φ ( m ) dm R M U M L m φ ( m ) dm where M LCC and M UCC are respectively the lower and uppermass limits for SN CC progenitors and M L , M U are the lower andupper stellar mass limit. To be consistent with the Kennicutt’sSFR calibration we adopted a Salpeter IMF, that is: φ ( m ) ∝ m − α with α = .
35 and 0 . ⊙ < M <
100 M ⊙ Assuming 8 < M CC <
50 M ⊙ for the CC progenitor massrange, k CC = .
007 M − . This number changes significantly ifwe adopt a di ff erent IMF, e.g. k CC = .
011 for a Kroupa IMF or k CC = .
039 for an extreme
Starburst
IMF (Dwek et al. 2011).We soon note however that, because the IMF enters also in theconversion from L TIR to ψ , the expected rate of SN events isalmost independent on the selected IMF (cf. Sect. 4.1) providedthe choice is consistent.More important is the assumption on the mass range forCC progenitors which is not well constrained. Actually, whilechanging the upper limit of the progenitor mass from 40 to100 M ⊙ makes a modest 10% increase in the CC SN produc-tivity, the lower mass limit is crucial, with k CC decreasing by30% if we adopted M LCC =
10 M ⊙ instead of the favored valueof 8 M ⊙ (Smartt 2009). SN Ia
Estimating the expected rate of SN Ia is complicated be-cause the delay time distribution f Ia , while still uncertain, cer-tainly ranges from short to very long time. In particular it hasbeen suggested that SN Ia can be divided into two classes,one with a short delay time whose rate scales with the cur-rent SFR (also called prompt ), and a second with a long de-lay time ( tardy ), whose rate scales with the average of theSFR along the entire galactic evolution (Scannapieco & Bildsten2005; Mannucci et al. 2006). While stellar evolution arguments(Greggio 2010, 2005; Greggio & Renzini 1983) and more recentdata (Maoz et al. 2012; Totani et al. 2008) suggest a continuousdistribution of the delay time instead of two distinct classes, theschematization is still a fair approximation that help in simplify-ing the problem of predicting the expected SN Ia rate in SBs.
11. Miluzio et al.: HAWK-I SN search
In general, for a galaxy of the local Universe, ∼
13 Gyr afterthe beginning of SFR, we can identify the contribution of thetwo components as follows (Greggio 2010):˙ n Ia (13) = k Ia × ψ C Z . f Ia ( τ ) d τ + ψ P Z . f Ia ( τ ) d τ ! (2)where ψ C and ψ P are the average SFR over, respectively, thelast 0.1 Gyr (current SFR) and from 0.1 to 13 Gyr ago (pastSFR). The SN productivity k Ia is the product of the number ofstars per unit mass in the adopted progenitor mass range (0.021for a Salpeter IMF and a mass range 3M ⊙ < M < ⊙ ) and therealization fraction, the actual fraction of systems which makea successful explosion ( ∼
5% according to the most recent es-timate) (Maoz & Mannucci 2012). We assume that SF historyin SBs can be described schematically with two components:a constant SFR during the galaxy evolution which created thegalaxy stellar mass, and an on-going episode of intense SFRwhich is the source of the strong TIR emission. Neglecting thecontribution of the ongoing SB to the galaxy stellar mass, wecan approximate ψ P ≃ M / × , and write Eq. 2 as follows:˙ n Ia (13) ≃ k Ia (cid:18) ψ C F pIa + M × F tIa (cid:19) where F pIa = < f pIa > × . F tIa = < f tIa > × ≃ − F pIa are the relative fraction of prompt and tardy events derived byintegrating the delay time distribution in the relevant time range.In our approximation ψ C can be derived from the observed L TIR and the galaxy mass from the K magnitude and B − K colors (cf.Mannucci et al. 2005).The relative contribution of the two SN Ia components hasbeen a very debated issue in the last few years, ranging from F pIa ∼
50% (Mannucci et al. 2006) to F pIa ∼
10% from standardstellar evolution scenarios (Greggio 2010). In our simulation weadopted as reference an intermediate value, F pIa ∼ Dust extinction in SBs is very high, especially in the nuclearregions. For instance, Shioya et al. (2001) found that fitting thespectral energy distribution of the nuclear region of Arp 220 re-quires a visual extinction A V >
30 mag. Actually, accordingto Engel et al. (2011), ”over most of the disk the near-infraredobscuration is moderate, but increases dramatically in the cen-tral tens of parsecs of each nucleus”. Similar high extinction, A V ∼
20, was found for the SB region of Zw 096 (Inami et al.2010).As a first order approximation, for our simulation we as-sumed that the extinction has the same distribution of the SF (seenext section) with a maximum value A V =
30 mag correspond-ing to the SFR peak and scaled linearly in the other regions.While this is a crude approximation, it turns out that the actualchoice of extinction correction has little impact for our simu-lation. In the nuclear, high extinction regions the SN detectionis limited by the reduced performance of the image subtractionalgorithm in these high surface brightness regions. At the sametime, our IR search is largely insensitive to variation in the (mod-erate) extinction of the outer galaxy regions.For the wavelength dependence of extinction we adopted theCalzetti’s law with R V = . ± . The spatial distribution of the SFR is a key ingredient of thesimulation. This is because we expect that SNe occur more fre-quently in the high SF regions where, on the other hand, our de-tection e ffi ciency is lower. In principle, the FIR emission whichis used to estimate the SFR would also be a good tracer of its spa-tial distribution. However, it turned out that the available MIRimaging for the galaxies of our sample (mainly obtained withthe Spitzer observatory) do not have enough spatial resolutionfor mapping the compact SB structures.Selected K-band images from our survey can have excellentresolution but, as is well-known, the near IR emission bettertraces the old star population, that is the galaxy mass distribu-tion more than the SFR distribution. Therefore for an estimateof the SFR concentration, we are forced to an indirect, statisticalapproach.Our starting point is the SB classification by Hattori et al.(2004), who derived a correlation between the global SBs prop-erties, such as FIR colors, and the compactness of the SF re-gions. These range from very compact ( ≤
100 pc) nuclear star-bursts with almost no star-forming activity in the outer regions(type 1), to extended starbursts with relatively faint nuclei (type4), with type 2 and 3 as intermediate cases. In addition, theyfound a trend for galaxies with more compact SF region showinga higher star formation e ffi ciency and hotter far-infrared color.They also found that the compactness of SF regions is weaklycorrelated with the galaxy morphology, with disturbed objectsshowing preferentially more concentrated SF. On the other hand,an appreciable fraction ( ∼ <
20 Kpc and galaxies that have anearby ( <
100 Kpc) companion at the same redshift were classi-fied as ”pairs” (P). The remaining objects were classified as ”sin-gle” (S). The classification of the SBs of our sample is listed inTab. 6 along with the galaxy FIR colors, log f / f , log f / f .We attributed to each galaxy a compactness class on the ba-sis of its correlation with the FIR colors as shown in Fig. 4 ofHattori et al. (2004) that for the object of our sample correspondsto our Fig. 8. As it can be seen, we also confirmed their claim ofa (weak) relation of FIR color and, as a consequence, compact-ness class with SB morphology.The next step is based on Soifer et al. (2000, 2001). For anumber of SBs galaxies they plotted the MIR and NIR emis-sion curve of growth finding that in general the MIR emissionis more concentrated, while only for few galaxies the MIR andNIR curves of growth show a similar trend. Actually we foundthat, to a first order approximation, the MIR emission profile of agiven galaxy can be matched by NIR profile powered to an expo-nent α which ranges between 1, when the two profiles are simi-lar, to 2, when the MIR emission is strongly concentrated. Whenwe classify the same galaxies with the compactness criteria ofHattori et al. (2004), we found that (as expected) the galaxieswith compact SF regions (type 1 −
2) are characterized by moreconcentrated MIR emission ( α = . − .
5, respectively), whilegalaxies with extended SF region (type 3-4) have similar MIR
12. Miluzio et al.: HAWK-I SN search
Table 6: Morphological classification, FIR colors and com-pactness classification for the SBs of our sample. FollowingHattori et al. (2004), type 1 have SF region <
500 pc, type 2 < > Galaxy Morp. log f / f log f / f Comp.class classCGCG011-076 S -0.20 -0.89 3CGCG043-099 S -0.19 -1.05 3ESO148-IG002 CP 0.01 -0.82 2ESO239-IG002 M -0.04 -0.79 2ESO244-G012 CP -0.10 -0.68 2ESO264-G036 S -0.34 -0.96 4ESO286-IG019 M 0.07 -0.81 2ESO440-IG058 P -0.23 -0.99 3ESO507-G070 S -0.08 -1.21 1IC1623A / B CP -0.14 -0.80 2IC2545 M 0.01 -0.88 2IC2810 P -0.22 -1.00 3IC4687 / + / -1.5 -1.0 -0.5log(f /f )−0.3−0.2−0.10.00.1 l og ( f / f ) SS CPM CPS MPS CPMP PS S CPM P PS MPP SCP CPSPCP P
Fig. 8: FIR color of the SBs of our sample. The di ff erent sym-bols identify the compactness class (Hattori et al. 2004) whilethe label show the morphological type. vs. NIR profiles ( α = . − .
0, respectively). As a reference,we notice that in the typical case of NGC 6240, assuming α = α = L K map, andadopting a power index α appropriate for the compactness classof the given SB galaxy (Tab. 7). Having defined all the ingredients of the simulation, we can nowdescribe how this proceeds. The simulation flowchart can besummarized as follows:1. for each galaxy of the sample, based on the estimated totalSFR and adopted progenitor scenarios, we compute the ex-pected number of SNe per year;2. a time interval is chosen so that 100 SNe are expected to ex-plode in the given galaxy in that period. The time intervalends with the last observations of the galaxy. Given the ex-pected SN rates it is for all galaxies much longer than theduration of our monitoring campaign. The reason to simu-late 100 events is to avoid having to deal with fractional SNnumbers for the di ff erent subtypes. We assign to each eventa random epoch of explosion chosen within the defined timeinterval;3. each SN is assigned to a random explosion site inside theparent galaxy according to the SFR spatial distribution;4. a peak magnitude is also assigned to each SN, with a ran-dom value derived from the adopted SN luminosity functionfor the specific subtype. We also associate to the event anextinction value randomly extracted from a gaussian distri-bution whose mean value depends on the position of the SN,and σ = /
4. Comparison between observed and expected SNdiscoveries
As we outlined above, from a large number of MonteCarlo simu-lation runs we obtain the distribution of the expected SN discov-eries. This is shown in Fig. 9, where each bin of the histogram isthe predicted probability of observing the specific number of SNdiscoveries whereas the dashed line marks the number of actualSNe discovered and the shaded area shows its 1- σ Poissonianuncertainty range.
13. Miluzio et al.: HAWK-I SN search
Table 7: Input parameters for the reference simulation L TIR / S FR calibration Kennicutt (1998)IMF SalpeterSFR distribution ∝ L α K for compactness class 1,2,3,4 α = . , . , . , . V ∝ S FR (max 30 mag)CC mass range 8 − ⊙ Ia components 30% prompty, 70% tardySN luminosity function LOSS (Li et al. 2011b)
Fig. 9: Histogram of the number of expected SNe from ourMonteCarlo experiments. The dashed vertical line indicates thenumber of observed events and the grey area its 1 σ -Poissonianuncertainty.We found that with the adopted simulation scenario and inputparameters we should have expected, on average, the discoveryof 5 . ± . σ ) the expectednumber is in the range 4-8 which is in excellent agreement withthe observed number of 6 events.The prediction of the simulation is that almost all SNe areCC (5.1 SN CC vs. 0.2 SN Ia), though in 10% of the experimentsat least one type Ia is found (that is what we have from the realSN search).The distributions of some of the expected and observed SNproperties are compared in Fig. 10. For the simulation, we showthe distribution across a large number of experiments (line-onlyhistogram) while the grey shaded histogram represents the actualobservations.The top panel in Fig. 10 shows the distribution of the appar-ent magnitudes at the discovery. The good agreement betweensimulations and observations is a crucial consistency check ofour estimates of the magnitude detection limit: if the discov-ered SNe were systematically fainter / brighter then expected, this (a) Expected (line-only) and observed (grey) Kmagnitude distribution at the discovery(b) Expected (line-only) and observed (grey) ex-tinction distribution. In light grey we indicate theallowed range for the extinction of PSN2010 (seetext).(c) Radial distribution of expected (line-only) anddetected (dark grey) SNe. For regular galaxies thesurface brightness decrease monotonically with ra-dial distance (the upper axis shows this correspon-dence for one of the galaxy of our sample). In lightgrey we show the distribution of injected artificialSNe (see text). Fig. 10: SN properties comparison
14. Miluzio et al.: HAWK-I SN search would indicate, respectively, an underestimate / overestimate ofthe search detection e ffi ciency.A comparison of the simulated vs observed extinction distri-bution is shown in the middle panel of Fig. 10. For the observeddistribution the case of PSN2010 for which extinction is am-biguous is shown in light grey. Again the simulation is in goodagreement with the observations. This argues in favor of the con-sistency of the input assumptions. The fact that, in our IR searchwe expect that most of detected SNe have low extinction ( ∼ A V < A V =
30 magfor the extinction towards the SN remnants of M82. Confirmingthe presence of high extinction (about A V ∼ A V ∼
16 mag for the SN2008cs, located at about 1.5 Kpc, relatively far from the galaxynucleus.Finally, in the bottom panel of Fig. 10 we compare the distri-bution of locations inside the host galaxy for the expected (line-only) vs observed (dark grey) SNe. The di ff erent location areidentified by the K band pixel counts: in general high countsoccurs in the nuclear regions while low counts are in the out-skirts (we use pixel counts instead of radial distances becausethe latter is di ffi cult to be defined for galaxies with irregular mor-phology or double nuclei. However a indicative correspondencefrom pixel count to radial distance is shown it the top axis of thefigure for a galaxy with regular morphology. There is a mild in-dication of a deficiency of observed events in regions with highpixel counts. Taken to face value this may suggest a minor over-estimate of the detection magnitude limit in the nuclear regions.Given the poor statistics we cannot derive definite conclusionsand therefore we will not elaborate further this issue.In the same figure we show also (in light grey) the distribu-tion of locations of the expected events for an ideal case wherethe magnitude detection limit in the nuclear regions is as deepas in the outskirts, and extinction is negligible. The experimentshows that the fraction of events that remains hidden to oursearch in the galaxy nuclear regions due to the combined e ff ectof reduced search e ffi ciency and high extinction is very high,being about 60% (cf. Mattila et al. 2012). One of the main source of uncertainty for the simulation is re-lated to the estimate of the magnitude detection limit, ma g lim . Forthe reference simulation, we adopted as ma g lim the mean valueout of three artificial star experiments conducted for a number ofselected positions inside the host galaxy (cf. 2.5). The dispersionof measurements, that is the uncertainty on ma g lim , is quite largewith a typical range of ∼ . ffi cultcases, it can be as large as 2 mag.To test the propagation of this uncertainty, we performedMonteCarlo experiments assuming alternatively the lower andhigher ma g lim out of the three experiments. We found that the predicted number of SNe is respectively 6 . ± . . ± . +
17% and −
11% with respect to the numbers from thereference simulation. The fact that the error is significant is thereason why we spent a significant e ff ort for a detailed estimateof the detection limit. SN Luminosity Function
In the reference simulation we use a gaussian distribution forthe SN luminosity function (SN-LF) with a mean value and dis-persion taken from Li et al. (2011b). However, Horiuchi et al.(2011), based on a small sample of very nearby SNe, claimedthat the faint end of the SN-LF is underestimated and SN CCfainter than mag ≃ −
16 could made up to 50% of the dis-tribution, to be compared with 20% of the sample of Li et al.(2011b). On the other hand, Mattila et al. (2012) argued thatHoriuchi et al. (2011) overestimated the fraction of intrisicallyfaint CCSNe since they neglect the host galaxy extinction fortheir SN absolute magnitudes. We performed a MonteCarlo ex-periment adopting the Horiuchi’s SN-LF and found that in thiscase the expected number of events would be low, only 3.3 onaverage. This is because most faint events are expected to fallbelow the search detection limit. The fact that the actual discov-eries are twice this number argues against a large fraction of faintSN-CC (cf. Botticella et al. 2012)
IMF and SN CC progenitor mass range
The IMF enters both in the estimate of the number of SN pro-genitors and in the calibration of TIR luminosity in terms of SFRrelation. However, the expected rate of CC SNe in our sample isvirtually independent of the IMF slope. Indeed, for a given totalmass of the parent stellar population, top heavy IMFs imply botha higher number of CC progenitors as well as a larger luminos-ity. Following Dwek et al. (2011) the number of CC progenitorsper unit mass is k CC = . , .
011 and 0.039 M ⊙− respectivelyfor a Salpeter, a Kroupa and a Starburst IMF , assuming that theprogenitors range from 8 to 50 M ⊙ . At the same time the totalluminosity of a SB forming stars with a SFR of 1 M ⊙ yr − over aperiod of 10 Myr (i.e. a 10 M ⊙ stellar population) is 4 . × ,7 . × and 2 . × L ⊙ again for a Salpeter, a Kroupa anda Starburst IMF, respectively. The M / L ratio of such SB is then0.0021, 0.0014 and 0.0004 (solar units) for the three IMFs, andthe expected number CC SNe originating from it is ≃ . L − ⊙ for all the three IMFs. Working out the numbers, it turnsout that the SN CC rates from a population with a given L TIR isalmost independent on the IMF, provided a consistent choice ismade.Crucial is instead the assumption of the SN CC progenitormass range, in particular the lower limit. Indeed if we adoptedan upper limit of 100 M ⊙ instead of the reference value of 50M ⊙ the expected number of SNe would be 5 . ± .
1, only ∼ ⊙ (instead of 8 M ⊙ ) results in an ex-pected number of SNe of 3.9 ± ∼
30% lower than theexpected rate obtained in the reference case.
Extinction
For the reference case we assumed that the extinction scales withthe SFR with a maximum value corresponding to the SFR peak A V =
30 mag. To test the uncertainty related to this assumptionwe made two di ff erent tests. In one experiment we maintained
15. Miluzio et al.: HAWK-I SN search the relation of A V with SFR but taking, alternatively, a peak ex-tinction value A V =
10 and A V =
100 mag. The experiment gaveas expected number of SNe 5 . ± . . ± .
3, respectively.In the second experiment we assume that the extinction is con-stant through the galaxy and is A V = . . ± . ff ect significantly the simulation or, conversely, that our ex-periment we cannot probe the extinction distribution. Star Formation Distribution
The spatial distribution of SFR is an important, and the mostuncertain, ingredient of the simulation. For instance, if we as-sume that the SFR is confined in the very inner regions, say inthe inner 3 −
500 pc, the resulting SNe will remain unaccessi-ble to our search. On the other hand, the fact that in some SBsthe SFR is extended has been confirmed by di ff erent studies (eg.McQuinn et al. 2012), not to mention that many of the SNe wehave discovered are at significant radial distances (cf. Tab. 3).As we described in Sect. 3.1.4 as proxy of the SFR distri-bution we use L α K where α range from 1 to 2 depending on thegalaxy compactness class (Tab. 6). To test for the uncertaintiesof this assumption we performed two simulations assuming thatfor all galaxy α is either 1 or 2. We obtained in the first casean expected rate of 8 . ± . . ± .
7. The latter occurs because when the SFR is more con-centrated, a large number of SNe remain hidden to our searchdue to the low search detection e ffi ciency in the nuclear regions.The conclusion is that the uncertainty in the adopted SF dis-tribution propagates with an error of ∼
50% on the expected SNnumber. We may consider that the actual good match of observa-tions with the reference simulation argues in favor of the adoptedprescription.
5. Summary and Conclusions
We have presented the analysis of an infrared SN search in asample of 30 nearby SB galaxies, conducted between 2009 and2011, with the goal to verify the link between star formationand SN rate. During our search we collected in total about 240observations discovering 6 SNe, 4 of them with spectroscopicconfirmation.How does this number compares with the expectation ?Answering this question requires a detailed characterizationof the SN search detection e ffi ciency, the galaxy properties (inparticular SF rate and spatial distribution) and the SN propertiesand progenitor scenarios. We included all these ingredients ina MonteCarlo simulation tool that, allowing for the stochasticnature of SN events, can be used to explore the distribution ofthe expected SN number and properties.First of all, we may remark that by itself the number of de-tected SNe is a proof of the high SFR in SBs. In fact if we com-pute the expected number of SNe in our survey based on theaverage SN rate per unit B luminosity or mass (Li et al. 2011a),we would predict the discovery of 0.5 events (or more precisely,50% of the simulation predict the discovery of one event andnone is expected in the other 50%). The observed number is oneorder of magnitude higher, which is consistent with the fact thatthe TIR emission of SBs is about ten times higher than for nor-mal SF galaxies with the same B luminosity. Indeed, it is well-known that the TIR luminosity is an excellent tracer for SFR, inparticular in SBs. When we adopt the SFR from L TIR as input for theMonteCarlo experiment, we find that the expected number ofSNe in our search is 5 . ± .
3, SNe in excellent agreement withobservations. In most cases we predict that only SN CC shouldbe discovered while in the actual search we did detect one type IaSN. Given that there is a sizable fraction of experiments (10%)when this is predicted to occur we do not elaborate further thisissue. Also, allowing for the low statistics, we find an excellentagreement between the predicted and observed SN properties,namely apparent magnitude at discovery, extinction and locationinside the host galaxies.We performed a number of tests to verify the dependenceof the simulation outcomes from the input parameters. For theSN search characterization we show that an accurate estimate ofthe magnitude limit for SN detection is crucial. This is why wespend a considerable e ff ort in artificial star experiments (possi-bly the single most expensive task of our project). For the galaxycharacterization the most uncertain input is the SF spatial dis-tribution. With some creativity, we devised a prescription thatseems to work, but it is certain that this is a place for improve-ments when new, high resolution SB maps will become avail-able. Instead, we found that our results are not sensitive to theuncertainty on the amount of extinction because where extinc-tion is very high (the dense SB regions) our search is limited bythe bright magnitude detection limit. SNe in these regions re-main hidden to our search almost independently on the amountof extinction. Based on our simulation we estimated that the frac-tion of hidden SNe is very significant, that is ∼
60% with an up-per limit of 75% if we account for the poissonian uncertaintiesin the number of detected events. Finally, for the SN progenitorscenarios the larger uncertainty is the lower limit of the progen-itor mass range. If we adopt a lower limit M LCC =
10 M ⊙ insteadof 8 M ⊙ as in the reference simulation, the expected number ofSNe would be 30% lower than observed.Our results appear in good agreement with those of previ-ous similar searches (Mannucci et al. 2003; Cresci et al. 2007;Mattila et al. 2007a, 2012, cf. Sect. 1). In broad terms, the over-all conclusion of all these studies can be expressed as follows:the number of (CC) SNe found in SBs galaxies is consistent withthat predicted from the high SFR (and the canonical mass rangefor the progenitors) when we recognize that a major fractionof the events remains hidden in the unaccessible SB regions.As stressed by Mattila et al. (2012), this has important conse-quences for the use of SN CC as probe of the cosmic SFR, be-cause the fraction of SBs is expected to increase with redshifts(cf. Melinder et al. 2012; Dahlen et al. 2012)While continuing to search for SNe in SBs, in optical andinfrared, can certainly help to improve the still low statistics, onemay argue at this point for a change of strategy.In this respect good example is the attempt to reveal someof the hidden SN CC through infrared SN searches which ex-ploits adaptive optics at large telescopes, eg. Gemini or VLT.The results are encouraging with the discovery of two SNe withvery high extinction, namely SN 2004ip with A V between 5and 40 mag (Mattila et al. 2007b) and SN 2008cs with A V ff ect (also in these cases no spectroscopic classificationwas obtained). The extinction towards SN 2011hi was revisedby Romero-Ca˜nizales et al. (2012) using Gemini-N data. They
16. Miluzio et al.: HAWK-I SN search demonstrate that this is most likely a SN IIP with A V of 5-7 mag.Because of the need to monitor one galaxy at the time and to ac-cess heavily subscribed large telescopes, this approach will notresult in large statistics though even a few events may be molstlyvaluable to explore the very obscured nuclear regions.On the other hand, a new opportunity that should be exploredis the piggy-back on wide field extragalactic surveys of the nextgeneration infrared facilities, in particular EUCLID. This wouldallow for the first time to perform IR SN searches on large sam-ple of galaxies exploring a range of SF activity and, by monitor-ing galaxies at di ff erent redshifts, probe the cosmic evolution. Acknowledgements.
We thank the referee, Seppo Mattila, for the careful readingand the very useful comments.We particular thank Anna Feltre (ESO), for her help inestimating the possiblecontribution by AGNs to the FIR luminosity of the galaxies and to Barbara LoFaro (Astronomy Department of Padova) for helpful discussions and sugges-tions.We acknowledge the support of the PRIN-INAF 2009 with the project”Supernovae Variety and Nucleosynthesis Yields”.E.C., L.G., S.B., A.P. and M.T. are partially supported by the PRIN-INAF 2011with the project ”Transient Universe: from ESO Large to PESSTO”.N.E.R. acknowledges financial support by the MICINN grant AYA08-1839 / ESP,AYA2011-24704 / ESP, and by the ESF EUROCORES Program EUROGENESIS(MINECO grants EUI2009–04170).F.B. acknowledges support from FONDECYT through Postdoctoral grant3120227 and from the Millennium Center for Supernova Science through grantP10-064-F (funded by ”Programa Bicentenario de Ciencia y Tecnologa deCONICYT” and ”Programa Iniciativa Cientiffica Milenio de MIDEPLAN”).
References