Core-collapse, superluminous, and gamma-ray burst supernova host galaxy populations at low redshift: the importance of dwarf and starbursting galaxies
Summary (8 min read)
Introduction
- The authors reanalyse low-redshift SLSN and LGRB hosts from the literature (out to z < 0.3) in a homogeneous way and compare against the CCSN host sample.
- More exotically, some models postulate that LGRBs and/or SLSNe may arise as the result of runaway collisions in young and dense star clusters (van den Heuvel & Portegies Zwart 2013).
- Thus, it is still unclear to what extent the environmental properties of SLSNe and LGRBs (low-mass, lowmetallicity, and high sSFR) reflect their specific physical influences (progenitor and explosion mechanism).
- Thirdly, the sample must be able to securely distinguish CCSNe from Ia SNe for all transients, ideally via spectroscopy.
- The authors investigate star formation within the CCSN host galaxy sample and compare to a sample of SLSN and LGRBs.
2.1 Core collapse supernovae
- The authors drew their CCSN sample from ASAS-SN, since it is shallow (mV ,limit ∼17 mag) but is all-sky, so the SNe it finds are bright and generally very nearby.
- The authors also included any SNe that were not discovered by ASAS-SN, but were ‘recovered’ in their data and therefore do not have an ASAS-SN name designation.
- There were some ambiguous classifications that the authors removed from the sample.
- Also.
- A mosaic showing their ASAS-SN CCSN host galaxy sample is provided in Figs 1 and 2.
2.2 Superluminous supernovae
- The authors collated their initial SLSN sample based on archival SLSNe in the literature.
- These SLSN candidates and their properties are summarized in Table 2.
- Thumbnail images of each host are shown in the bottom row of Fig. 1; the physical scale is the same as for the CCSN hosts, with a yellow scale bar of 2 arcsec.
- The authors restricted their analysis to SLSNe with a redshift of z < 0.3 for two main reasons.
- First, including distant SLSNe could have caused incompleteness in the sample due to the increased difficulty 2We did not include SN1995av and SN1997cy since their classifications are unclear: SN1997cy could be a SN Ia or IIn and SN1995av may have been associated with a LGRB.the authors.
2.3 LGRBs
- The authors LGRB sample consists of all z < 0.3 LGRBs discovered prior to the end of 2017 with an associated optical counterpart: a supernova, an optical afterglow, or both.
- This sample was comprised of 17 LGRBs; 12 of which had confirmed SN associations and 5 without any reported SN (see Table 3).
- These appear to have genuinely different progenitors (such as compact binary mergers) and/or explosion mechanisms from ordinary SN-associated long-duration GRBs, a possibility that makes scrutiny of their host properties particularly relevant.
3.1 CCSN host multiwavelength data
- The galaxies in their CCSN sample are nearby (z < 0.08), so most were detectable in all-sky multiwavelength surveys.
- Therefore, their primary image and source catalogues were public surveys.
- The NSA is a unified catalogue of galaxies out to z ∼ 0.05, optimized for nearby extended objects since the flux measurements are derived from reprocessed SDSS images with a better background subtraction (Blanton et al. 2011).
- Possible SLSNe-I from Quimby et al. (2018) are indicated by a∗; host analysis is done, but not included the SLSN statistical analysis due to uncertainty about the nature of the classification.
- In the cases of galaxies with small angular size, the aperture was usually adequate, but in the case of high-mass, extended galaxies the aperture often missed a substantial fraction of the low surface-brightness flux in the outskirts of the galaxy, thus the authors redid the 2MASS photometry for these sources.
3.2 Procedure for CCSN hosts
- The authors performed aperture photometry using the PYTHON programme PHOTUTILS.
- 3We used an elliptical aperture and a curve-of-growth technique.the authors.
- The authors derived the uncertainties on these photometric measurements by using the galaxy aperture to determine the brightness of the background sky.
- For image calibration, the authors used catalogues of stars (PS1 Object Catalogue, 2MASS Point Source Catalogue, and the SDSS Imaging Catalogue) to calculate the zero-point for each image.
3.3 Galaxies requiring special attention
- Some host galaxies in their sample required extra care when performing photometry and when fitting SED models.
- These galaxies were either diffuse, low surface-brightness galaxies, galaxies which showed signs of interaction with nearby galaxies, galaxies contaminated with foreground stars (or other objects), or galaxies where 3https://github.com/astropy/photutils/tree/v0.3 4SN 2003ma pierces through the Large Magellanic Cloud.
3.3.1 Interacting galaxies
- A significant number of host galaxies (in the both CCSN and extreme-SN samples) showed evidence of physical companions, some of which appeared to be in the process of interacting or merging.
- This system would barely be detectable as two individual galaxies if it was discovered at a similar redshift (z∼ 0.2) to the SLSN or LGRB sample; therefore, the authors quoted two different measurements for photometry: one of the entire system and one of the single galaxy from which the SN originated.
- The authors used the photometry of the system for the SED fit.
- There was a small, red object to east of the host galaxy (see panel 5 in Fig. 1).
- For this reason, the authors were careful not to include this object in the photometry aperture.
3.3.2 Unclear host galaxy
- This SN was originally reported to TNS as being hosted by the elliptical galaxy NGC 2444, which is interacting with NGC 2445.
- SN 2017ati was originally reported to TNS as a hostless supernova.
- When the authors looked at a larger image of the field, the SN was located between two galaxies.
- This placed the supernova ∼10 kpc (36 arcsec) away from the galaxy nucleus.
- In their analysis the authors assigned the SN to the nearest galaxy since this would be how they would treat this SN if it were at a typical SLSN redshift.
3.3.3 Foreground star contamination
- These hosts were large and extended objects low surface-brightness hosts.
- This host galaxy has a small background galaxy and a few foreground stars covering the host.
- The authors removed the flux from these stars in the images.
- Therefore in each case, the aperture was chosen carefully so that the stellar flux was not included in the flux measurements.
3.3.4 Active galactic nuclei
- The authors checked if any of the host galaxies in their sample had an observable AGN present.
- While visual inspection of the host galaxy suggested that the AGN is unlikely to contribute significantly to the optical flux measured in SDSS/PS1, it could contribute more significantly to the IR flux, which could in turn affect the SED derived parameters including ages of the stellar populations, SFRs and also dust attenuation in the host galaxy.
- Hence, for 14de the authors excluded NIR photometry for the SED fit.
- The authors also checked the ALLWISE colours (W1–W2 and W3–W2) of the host galaxies as another diagnostic to test whether an AGN was present in the host galaxies (see fig. 12 of Wright et al. 2010).
- The authors also obtained a spectrum of ASAS-SN 14ma in Taggart et al. (in preparation) from the WHT and they found a line ratio of log ([N II]6583 /H α) = −0.83, indicating that AGN contribution was minimal.
3.4 Literature photometry
- Photometry of the SLSN and LGRB hosts was gathered primarily from the published literature.
- For clarity, all sources are listed in Table 4.
- Photometry is not corrected for Galactic foreground extinction.
- All photometry is available online in a machine-readable form.
- If the uncertainties were not given in the photometry from the literature, it was assumed that they were negligible and the authors therefore assign an uncertainty of 0.01 mag when performing the SED modelling.
3.5 New LGRB and SLSN host photometry
- The authors supplemented the SLSN and LGRB photometry from the literature with new photometry from a variety of sources, detailed below.
- Most of the LGRB hosts in their sample were observed using the Infrared Array Camera (IRAC; Fazio et al. 2004) on the Spitzer Space Telescope (Werner et al. 2004) as part of the extended Swift/Spitzer Host Galaxy Legacy Survey (SHOALS; Perley et al. 2016a).
- Data from some archival programmes were also reanalysed using a consistent methodology.
- The companion spiral is approximately 6 magnitudes brighter and offset by 6.5 arcsec; subtraction of its halo also left some residuals in the sky background.
- As a result, in both these cases the uncertainty on the host flux is relatively large.
3.5.2 Keck / MOSFIRE
- LGRB 130702A was observed in imaging mode using the MultiObject Spectrograph for Infrared Exploration (MOSFIRE; McLean et al. 2010, 2012) at Keck Observatory on the night of 2014 Jun 16 in the J and Ks filters.
- The authors reduced these data using a custom pipeline.
- The resolution of these images (and of archival optical data) are sufficient that there are no issues with background contamination from the nearby galaxies.
- Aperture photometry was performed in a standard fashion using nearby 2MASS standards.
3.5.3 Palomar / WIRC
- The authors reduced these data using their custom pipeline, which included cleaning of noise signatures associated with the replacement-detector.
- Aperture photometry was performed in a standard fashion using nearby 2MASS standards.
3.5.4 Palomar / P60
- LGRB 150818A was observed extensively with the CCD imager on the Palomar 60-inch robotic telescope (Cenko et al. 2006) as part of a campaign to follow-up the supernova associated with this event (Sanchez-Ramirez et al. in preparation).
- A series of late-time reference images in griz filters were taken on 2016 February 14 for the purposes of galaxy subtraction against the earlier supernova imaging; the authors employed these here to measure the host flux in these bands.
3.6 CCSN distances
- The authors did not have their own spectroscopy for each CCSN host galaxy.
- The fractional distance errors from peculiar velocities could have has implications for the analysis of their hosts.
- This model accounted for peculiar velocities due to the Virgo Cluster, the Great Attractor and the Shapley Supercluster and was typically a 6–8 per cent correction.
- The authors estimated the uncertainty based on data from the Bright Transient Survey (Fremling et al. 2020).
- Therefore the authors adopted this uncertainty estimate in the distance.
4.1 Spectral energy distribution fitting
- To quantify the stellar parameters of the host galaxies, including stellar mass and SFR, the authors modelled the spectral energy distribution (SED) of each host galaxy using UV through NIR photometry.
- If the reduced chi-squared 1 (before the Monte Carlo sampling) and the SED photometry was well-sampled in the UV, optical and IR, the authors applied the additional uncertainty to the photometry.
- Distribution of the physical properties plotted against redshift for each host galaxy sample.
- Each upper panel is a Gaussian kernel density estimation of each physical property.
- Watson et al. (2011) studied the mid-infrared spectrum and did not find any evidence for PAH emission in the host of LGRB 031203.
4.2 Redshift evolution correction
- The overall SFR density of the Universe, and of individual galaxies, rises rapidly with increasing redshift (e.g. Lilly et al. 1996), making it likely that the rare, luminous SNe that are typically found at higher redshifts than common, less luminous SNe will tend to be found in galaxies with higher SFRs simply on account of the effects of cosmic evolution.
- To make a direct comparison between their samples and to avoid systematic errors introduced by cosmic evolution, the authors corrected for redshift evolution in SFR by empirically re-scaling all SFRs to z = 0.
- SFRcorrected = SFRmeasured SFRMS(M,0) SFRMS(M,z) (1) We parametrized the main sequence as a power law, as in equation (2).the authors.
- The SFR and sSFR parameters have not been corrected, unless specifically indicated in the text and figure caption.
- The authors provide the derived physical parameters from SED fits without applying this SFR correction in Tables B1–B3.
4.3 Sequence-offset parameter
- As an alternative to applying a redshift evolution correction to the SFR to deal with cosmic evolution, the authors defined a metric of star formation intensity, the ‘sequence-offset’ parameter ( S).
- Photometry are not corrected for Galactic foreground extinction.
- All photometry is available online in a machine-readable form.
5 R ESULTS
- LGRBs and the ASAS-SN CCSN.the authors.
- Basic statistical properties of each sample are summarized in Table 6.
- Uncertainties (1σ ) are calculated using a simple bootstrap.
5.1 Basic properties of CCSN hosts and comparisons to nearby star-forming galaxies
- A key goal of their study is to produce a uniform and unbiased (by galaxy mass) sample of CCSN hosts, providing a galaxy-luminosityindependent tracer of the sites of star formation in the local Universe.
- Most LVL galaxies are observed to populate the main sequence of star-forming galaxies, where mass and SFR are strongly correlated in a fairly narrow band of sSFR between 10−9 and 10−10 yr−1.
- Similarly, small but statistically significant differences are also seen in other parameters (SFR, sSFR, and sequence offset).
- Using their sample, the authors measure the fraction of CCSNe in dwarf galaxies and the fraction in ‘starburst’ galaxies.
- The authors use the Bayesian beta distribution quantile technique to derive the 1σ uncertainties, following methods outlined in Cameron (2011).
5.2 Basic properties of exotic SN hosts
- In Figs 4(b)–(d), the authors also plot the mass and SFRs of the ‘exotic’ SN samples in comparison to local galaxies.
- These populations are clearly quite different from ordinary CCSNe.
- SN-less LGRBs have sSFR of −9.6(0.4), which is more consistent with the CCSN population.
- This effect can be seen more clearly in Fig. 5, which shows specific star formation versus stellar mass.
- Perley et al. (2016c) and Schulze et al. (2018) also noted that many SLSN-I host galaxies in PTF and SUSHIES samples are undergoing intense star formation.
5.3 Relative rates of SN sub-types
- While the authors can qualitatively observe that the distributions of certain samples in Figs 3–5 seem similar or dissimilar, this is not a statistical statement.
- The authors employ several different methods to quantify the significance and model the nature of these apparent differences below.
5.3.1 Cumulative distribution tests
- In Fig. 6, the authors show the cumulative distributions of mass, SFR, sSFR, and sequence offset for each of their galaxy samples.
- The step sizes of local galaxies in LVL are weighted by star formation to create a population consistent with one that traces star formation.
- The CCSNe and LVL samples have remarkably similar sSFR and S distributions, while the rarer SN sub-types seem to show different distributions in most properties.
- These differences can be tested formally using Anderson–Darling tests.
- The authors compute the Anderson–Darling (AD) statistic, and associated p-value, for each pair of samples and for each parameter of interest: stellar mass, SFR, sSFR, and the sequence offset parameter ( S).
5.3.2 Relative rate formalism for univariate comparisons
- While the Anderson–Darling tests above confirm that differences exist between some distributions, they do not tell us anything about the degree or quantitative nature of the differences between any two distributions.
- The authors empirically re-scale all SFRs to z = 0 for all host galaxy samples (CCSNe, SLSNe-I, SLSNe-II and LGRBs) using the procedure in 4.2.
- Note that because windows within 1 dex overlap, values of R within 1 dex of each other are not fully independent.
- To calculate the confidence intervals on the relative rate the authors draw a new CCSN sample and a new SLSN sample from the original samples (with replacement) for 1000 bootstrap iterations.
5.3.3 Relative rate formalism for bivariate comparisons
- Their relative-rate formalism above can be extended to ascertain whether a difference in distributions associated with a control parameter (e.g. stellar mass) can completely explain an observed difference in distributions for another parameter (e.g. SFR).
- To test this, the authors reweight the comparison sample (sample ‘B’).
- The weights for each galaxy in the comparison sample are interpolated from the relative rate for the control parameter.
- The authors use the same confidence intervals derived from the univariate bootstrap procedure and rescale them using the same factor to the weighted relative rate.
5.4 SLSNe-I versus CCSNe
- The relative rate, R, of SLSNe-I versus CCSNe is plotted in the lefthand panels of Fig. 7 as purple dashed lines with the 2σ confidence intervals in a lighter colour against sSFR, sequence offset, redshift corrected sSFR scaled to z ∼ 0 and stellar mass.
- The rate is also enhanced for galaxies with a sequence offset parameter S > 5, which corresponds to galaxies with SFR > 5 times that predicted of galaxies on the main sequence with the same stellar mass and redshift.
- The right-hand panels of Fig. 7 show the original relative rates as a purple dashed line.
- This may hint that the rate of SLSNeI production is increased as a result of high sSFR and low stellar mass.
- A larger sample size should help to solidify this claim.
5.5 LGRBs versus CCSNe
- Using the same method as described above, the authors also calculate the relative rate R of LGRBs versus CCSNe in Fig.
- Given the rather limited low-z LGRB sample, the results are generally less constraining than for SLSNe, and the authors cannot conclusively (for any 1- dex bin) state that R = 1 for LGRBs versus SNe, given this analysis.
- Formally, the relative rate of LGRBs is enhanced in galaxies with sSFRs exceeding 10−9 yr−1 (after correcting for redshift evolution) by a factor of ∼3; it is enhanced in galaxies with sequence offsets >2 by a factor of approximately 2, and it is enhanced in low-mass dwarfs <108 M by a factor of approximately 2.5.
- As with SLSNe, these effects are degenerate and given the small sample sizes, the authors cannot yet determine which parameter (if any) is the primary cause of the differences.
5.6 SLSNe-I versus LGRBs
- The authors can also compare the LGRB and SLSN-I host populations directly against each other.
- In their work, the authors find that SLSNe-I and LGRBs are statistically consistent with being drawn from the same galaxy populations in terms of all measured parameters (see Table 7), similar to the findings of Japelj et al. (2018).
- The authors note that due to their selection of nearby events, their sample size for LGRBs is smaller than in these studies.
- In terms of sSFR, the authors do not find any statistical differences (pAD = 0.11).
- Both LGRBs and SLSNe-I have a higher median logarithmic sSFR than CCSNe –9.6(0.1).
5.7 SN-less LGRBs versus LGRB-SNe
- To address whether the sub-population of ‘SN-less’ LGRBs may represent a distinct class from the remainder of LGRBs, the authors compare the host properties of the five events above to the remainder of the sample (Table 6).
- Only a few per cent (2+2−1) of CCSN hosts are undergoing starbursts with rapid star formation sSFR > 108 yr−1, all of which are dwarf galaxies with stellar masses <109 M . (iv) LGRB SN and SLSN-I host populations exhibit similar host galaxy properties.
- The authors also acknowledge useful feedback from R. Lunnan, M. Modjaz, S. Schulze, S. Vergani, J. Japelj, A. Gal-Yam, and useful conversations with A. Wetzel and D. Bersier.
- Some of the data presented herein were obtained at the W. M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration.
- The authors recognize and acknowledge the very significant cultural role and reverence that the summit of Maunakea has always had within the indigenous Hawaiian community.
Did you find this useful? Give us your feedback
Citations
659 citations
148 citations
105 citations
62 citations
References
11,309 citations
10,384 citations
10,039 citations
9,835 citations
7,182 citations
Related Papers (5)
Frequently Asked Questions (12)
Q2. What is the primary motivation for this exercise?
While their primary motivation for this exercise will be to compare this sample to ‘exotic’ supernova types (SLSNe and LGRBs) in order to constrain their progenitors, their CCSN sample is also useful for studying the nature of star formation at low-redshift: few galaxy surveys are complete beyond the dwarfgalaxy 109 M limit, with those that are typically confined to small volumes.
Q3. How many times did the authors run the SED fit on each photometry?
The authors sampled from the distribution 1000 times and then ran the SED fit on each set of ‘noisy’ photometry and used the 16-to-84th percentile of each parameter as an estimate of its uncertainty.
Q4. How many supernovae were removed from the sample?
In addition, the authors imposed a minimum distance cut out to 10 Mpc, meaning that one supernova (AT 2014ge) was removed from their sample.
Q5. What was the contribution of emission lines to the modelled spectra?
The contribution of emission lines to the modelled spectra was based on the Kennicutt (1998) relations between SFR and UV luminosity.
Q6. What was the contribution of H and [O II] lines to the photometry?
The contribution of Hα and [O II] lines to the photometry was included for galaxies with dust free colour bluer than (NUV–r)ABS ≤ 4 and the intensity of the emission lines was scaled according to the intrinsic UV luminosity of the galaxy.
Q7. What law was used to determine the emission of dust in the galaxy?
Dust attenuation in the galaxy was applied to the SED models using the Calzetti et al. (2000) extinction law for starburst galaxies.
Q8. How did the authors remove the galaxy on the west of the image?
The authors used the programme GALFIT (Peng et al. 2002) to model and subtract any contaminating objects from the image and then used the procedure outlined in Section 3.2 to perform aperture photometry on the galaxy.
Q9. What did the authors do to check for a clear nuclear point source?
Since the authors did not have spectra for every galaxy in their sample, the authors also inspected the images of each host (see Fig. 1) to check for a clear nuclear point source.
Q10. Why did the authors limit their sample to a declination greater than 30?
The authors limited their sample to a declination greater than −30◦ because uniform, public, deep optical survey data is not available across the entire Southern hemisphere.
Q11. What was the effect of the Monte Carlo sampling on the photometry?
If the reduced chi-squared 1 (before the Monte Carlo sampling) and the SED photometry was well-sampled in the UV, optical and IR, the authors applied the additional uncertainty to the photometry.
Q12. How do the authors rescale the relative rate of SLSNe-I?
The authors use the same confidence intervals derived from the univariate bootstrap procedure and rescale them using the same factor to the weighted relative rate.