An archival search for type Ia supernova siblings

By searching the Open Supernova Catalog, an extra-galactic transient host galaxy database, and literature analyses, I present the largest sample of type Ia supernova (SN Ia) siblings to date. The sample comprises 158 galaxies, consisting of 327 confirmed SNe Ia - over 10 times larger than existing sibling SN Ia samples. SN siblings share host galaxies, and thus share global environmental properties and associated systematic uncertainties. This makes them valuable for both cosmological and astrophysical analyses; for example, sibling SNe Ia allow for comparison of environmental properties within the same galaxy, progenitor comparisons, rates analyses, and multiple calibrations of the Hubble-Lema\^{\i}tre constant. This large sample will provide a variety of new avenues of research, and be of great interest to the wider SN Ia community. To give an example use of this sample, I define a cosmology sub-sample of 44 siblings; and use it to compare light-curve properties between sibling pairs. I find no evidence for correlations in stretch ($x_1$) and colour ($c$) between pairs of siblings. Moreover, by comparing to a comparable set of random pairs of SNe Ia through boot-strapping, I find that siblings are no more similar in $x_1$ and $c$ than any random pair of SNe Ia. Given that siblings share the same hosts, differences in $x_1$ and $c$ between siblings cannot be due to global galaxy properties. This raises important questions regarding environmental systematics for SN Ia standardisation in cosmology, and motivates future analyses of sibling SNe Ia.


INTRODUCTION
As more supernovae are discovered in wide-field time domain surveys, the likelihood of finding multiple type Ia supernovae (SNe Ia) that have occurred in the same galaxy is increasing.This likelihood will only further increase with the dawn of the next generation of transient surveys, with the Rubin Observatory Legacy Survey of Space and Time (LSST) predicted to find ∼ 800 galaxies hosting multiple SNe Ia (Scolnic et al. 2020).These SNe are termed 'siblings' (Brown 2014), and they provide a novel data set to study the effects of galaxy environmental properties on the standardisation of SNe Ia in cosmology, and to provide further insight into potential progenitor scenarios.
SNe Ia are important cosmological probes as standardisable candles due to their low intrinsic peak absolute magnitude dispersion.This dispersion can be further reduced using standard corrections based on correlations between SN Ia brightness and light-curve stretch ' 1 ' (the 'brighter-slower' relation;Rust 1974;Pskovskii 1977;Phillips 1993) and brightness and optical colour '' (the 'brighter-bluer' relation; Riess et al. 1996;Tripp 1998).
Despite these corrections, there is a remaining brightness dispersion currently unaccounted for.A potential cause of this remaining dispersion is considered to be due to host galaxy effects because of clear correlations between brightness (or 'Hubble residual' -the difference between observed and model brightness) and host galaxy properties.The most well-known of these is the 'mass step' in ★ E-mail: lisa.kelsey@port.ac.uk which SNe Ia associated with high-mass host galaxies standardise to brighter luminosities than those in lower-mass hosts (Sullivan et al. 2010;Kelly et al. 2010;Lampeitl et al. 2010;Gupta et al. 2011;Johansson et al. 2013;Childress et al. 2013;Uddin et al. 2017Uddin et al. , 2020;;Smith et al. 2020;Ponder et al. 2021;Popovic et al. 2021).Alongside this stellar mass correlation, other environmental properties such as star formation rate and galaxy colour also show this clear correlation (e.g., Lampeitl et al. 2010;Sullivan et al. 2010;D'Andrea et al. 2011;Childress et al. 2013;Pan et al. 2014;Wolf et al. 2016;Uddin et al. 2017;Kim et al. 2019;Kelsey et al. 2021Kelsey et al. , 2023)).In modern cosmological analyses, a host galaxy correction term is included in the cosmological fit, typically applying two different corrections for high-and low-mass hosts.
As sets of SN Ia siblings share the same host galaxies, they share global environmental properties, along with systematic uncertainties from redshift, peculiar velocities and gravitational lensing.Therefore, any remaining differences in brightness,  1 , , or Hubble residual between sets of siblings cannot be due to global host galaxy properties, potentially uncovering a concern with the current formalism in large surveys where siblings are corrected for in the same way.Instead, these differences may be due to sub-galactic differences in environments, dust distributions, progenitor scenarios, explosion mechanisms, or as-yet unknown physics.Thus, the use of siblings could determine if sub-galactic differences are the dominant factor in environmental standardisation, as suggested by local environment studies such as Rigault et al. (2013Rigault et al. ( , 2015Rigault et al. ( , 2020) ) and Kelsey et al. (2021Kelsey et al. ( , 2023)).
By comparing eight sibling pairs from the Dark Energy Survey (DES), Scolnic et al. (2020) found no better agreement between SN properties (with the tentative exception of light-curve stretch,  1 ) in the same galaxy as from any random pair of galaxies (as was also found by Scolnic et al. 2022, for a similar analysis using the Pantheon+ sample), and concluded that at least half of the intrinsic scatter in Hubble residuals is not from global host galaxy properties.Scolnic et al. (2020) also found that the sibling pair in the DES sample with the most differing distance modulii have SNe located on opposite sides of their host galaxy.Similarly, by finding a difference of 2 mag in rest-frame B-band peak magnitude between another set of siblings, with one SN being very red (high  value) whilst having a similar  1 value to its sibling, Biswas et al. (2022) found evidence of sibling SNe experiencing differing levels of dust attenuation, as was also suggested by Elias-Rosa et al. (2008) for SN2002cv and SN2002bo in NGC 3190.Gall et al. (2018) found that distances for siblings SN2007on and SN2011iv in NGC 1404 differed by up to 14%, despite having similar decline rates.By studying SN1980N and SN1981D in NGC 1316, Hamuy et al. (1991) found that distances for the two siblings differed by ∼ 0.1mag.When this was expanded by Stritzinger et al. (2010) to include the newer siblings in NGC 1316 (SN2006dd and SN2006mr), this distance dispersion was reduced to ∼ 4 − 8%.Hoogendam et al. (2022) found that the fast-declining SN siblings SN1997cn and SN2015bo (which also exhibited near-identical spectra) were consistent in distance at the 0.06 level, and with distances derived using surface-brightness fluctuations, perhaps indicative that they exploded under similar circumstances.Additionally, DerKacy et al. ( 2023) measured distance modulii to NGC 5018 using the siblings SN2002dj and SN2021fxy and found them to be consistent at 1.2.Furthermore, Burns et al. (2020) found that a sample of 12 literature siblings gave an average consistency in distance of 3%.Ward et al. (2023) also found distances for a trio of SNe Ia (SN1997bq, SN2008fv, and SN2021hpr) in NGC 3147 were consistent, with a low standard deviation of ≤ 0.01 mag.This range of findings across the literature gives clear motivation for the further analysis of larger samples of sibling SNe Ia.
Complementary to their potential to improve cosmological standardisation, SN Ia siblings are particularly useful for cosmology by providing multiple independent distance measurements to the same host galaxy if they also share their host with a Cepheid variable star.This gives a new, innovative calibration method for the Hubble-Lemaître constant ( 0 , the rate of the expansion of the Universe).By calibrating the zero-point of Hubble diagrams from large surveys using individual sibling SN Ia independently, different values of  0 are measured, providing uncertainties on the overall measurement (Gallego-Cano et al. 2022).With a large sample of siblings, this can be compared for each set, providing a new independent measurement of  0 which may help to understand the remaining Hubble Tension.
Alongside their use for cosmology, SN Ia siblings are valuable tools for investigating potential progenitor stellar populations by comparing differences in star formation rate, dust extinction and metallicity.Although not the focus of this work, by compiling samples of galaxies that contain many different types of SNe (i.e., not just SNe Ia) as has been done for the Zwicky Transient Facility (ZTF) (Graham et al. 2022) or for a historical data set (Anderson & Soto 2013), understanding can be gained into rates of SNe (Graur et al. 2017a,b), and preferences for different classes of SN to explode in different types of galaxies.
However, despite the quality and wealth of data for these siblings found in recent large surveys such as DES and ZTF, they suffer from the same critical flaw.The rate of detection of siblings in a single survey is limited by survey length.As the rate of SNe Ia is higher in more massive galaxies (Mannucci et al. 2005;Sullivan et al. 2006;Graur et al. 2017b;Wiseman et al. 2021), it is more likely that the siblings found in such single surveys are associated with high-mass galaxies (log(/M ⊙ ) ≥ 10), e.g., as found by Scolnic et al. (2020) for the siblings found in the 5-year long DES survey.This means that for cosmology, the current sample of sibling SNe Ia would predominately lie on the same side of the mass step.Additionally, more massive, passive hosts typically host faster (lower  1 ) SNe Ia, (Riess et al. 1995;Hamuy et al. 2000;Howell 2001;Sullivan et al. 2010;Graur et al. 2017b;Wiseman et al. 2021).Due to the known correlation between  1 and age (Howell et al. 2009;Neill et al. 2009;Johansson et al. 2013;Childress et al. 2014;Wiseman et al. 2021), this survey bias means that the single-survey siblings are more likely to occur from older progenitors.By combining data from multiple surveys and archival data, more combinations of siblings may be found, where each sibling may have been found by a different survey.As the elapsed time between siblings will be longer in this case, siblings can be found from galaxies with lower SN Ia rates, and hence lower masses.By having more variety in duration between siblings, there may be more diversity in environmental properties.
In this work, I search the Open Supernova Catalog (Guillochon et al. 2017) for SNe Ia that share a host galaxy, and cross-match using the transient host galaxy database produced by Qin et al. (2022, hereafter Q22).By combining this archival sample with additional sibling SNe Ia published in the literature, I present the largest sample of sibling SNe Ia to date, which is over 10 times larger than prior sibling samples.This sample is non-exhaustive, and some siblings may have been missed if not released publicly.The sample of sibling SNe Ia is constantly increasing, so I encourage the community to use this sample, and continue to add to it as more siblings are observed.
The layout of this analysis is as follows.In Section 2 I discuss the creation of the sample and briefly consider some curiosities within it.In Section 3, I identify a 'cosmology' sample based on the suitability of the siblings for cosmological analyses and perform a cosmological analysis of the light-curve properties, before summarising my conclusions in Section 4.

ARCHIVAL SIBLING IDENTIFICATION
I begin by discussing the creation of the sample.

The Open Supernova Catalog
First, I use the Open Supernova Catalog (OSC; Guillochon et al. 2017) to identify a sibling SNe Ia sample.The OSC was created to be a community-driven project to collect SN metadata spanning the electromagnetic spectrum, and a range of epochs of observations.This was intended to be a complete collection of public data of all known SNe, updated daily with new observations, open access and available to all in searchable catalogs on the internet.Unfortunately, in April 2022, the project ceased updating after an irrecoverable database corruption,1 and there are no plans to resurrect the project.Despite this, the data that the project collated is freely available on the project github page and can be accessed through an API,2 so provides a near-complete record of SNe up to April 2022.
By querying the OSC API, I obtained a large sample of transients that had been classified as SNe Ia.To be a Ia, I simply require that the object has the OSC tag claimedtype: Ia, but this requirement is not exclusive, meaning that the objects may also have sub-classifications, for example claimedtype: Ia Pec or claimedtype: Ia-91bg.I grouped these objects by host galaxy name to find cases where a galaxy hosted more than one SN Ia (siblings).As some surveys use different naming conventions for their galaxies, I matched on all galaxy aliases to ensure that no siblings were missed.Additionally, some SN surveys use different naming conventions for their SNe; for example, a SN from the All Sky Automated Survey for SuperNovae (ASAS-SN, Kochanek et al. 2017) will have a different supernova designation; so they may appear on the OSC twice with different names.To combat this, I cross-checked all the aliases of the identified 'siblings' on the Transient Name Server (TNS) and removed those that were duplicates. 3n checking the individual SNe on the TNS, I also verified the positions (ra and dec) and host galaxy association of each object, ensuring that they matched other public data, so that the sibling classification was robust.As an example, SN2006hj and SN2006mz are both reported on the OSC as being associated with a range of different galaxies, but have SDSS J211021.14+005557.5 in common, suggesting they may be siblings.Their ra values are very similar, but have dec values over a degree different, meaning it is very unlikely they are from the same host and this SDSS designation may have been incorrectly assigned.I therefore remove this pair from the sample.
A few objects are identified as being associated with galaxy clusters on the OSC, where the individual host galaxy within the cluster has not been determined or is disputed.As I require definitive host galaxy identification for sibling analysis, these may not be true siblings and so such objects are not included in the sample.
Although some are considered as the likely result of a SN Ia explosion, I do not include in this sample objects that were listed on the OSC as SN remnants (SNRs).Whilst I find a number of galaxies that contain multiple SNRs (such as the Milky Way and the LMC), it is difficult to confirm without-a-doubt which are the result of a SN Ia explosion (although see van den Bergh 1988;Krause et al. 2008a,b, for some examples of the classification of SNRs).Therefore I do not consider them SN Ia siblings for the purposes of this sample.
Using the OSC alone, the resultant sample comprises 113 galaxies, which contain 236 confirmed SN Ia siblings.

Qin et al. (2022) Extragalactic Transient Host Galaxy Database
Close to when the OSC ceased updating, Q22 published a database of extragalactic transients and their associated host galaxy properties. 4his database cross-identifies transients with known host galaxies, and updates those for whom the host was uncertain, or unknown, with newly identified host candidates.I used this database to cross-check and to supplement my findings from the OSC.In a similar way to the OSC search, I identified transients identified as SNe Ia in this database and grouped these by host galaxy name.For all bar one sibling set, the associated host galaxy did not change, but I found a few instances where OSC had not identified a set of siblings where this database did.After a visual inspection, I agree that the hosts identified in Q22 are the likely hosts and add them to the sample.There were a few additional cases where the OSC identified a set of siblings where one SN was detected after the creation of the Q22 database, so does not appear in that sample as a sibling.
Ignoring the cases where one SN in a sibling set was found after the creation of the Q22 database, I find three cases where siblings from the OSC are not siblings in the Q22 database.These are as follows: (i) If using the OSC data alone, SN2000dk and SN2015ar are labelled as originating from the same galaxy.However, after cross-checking on the TNS and in the Q22 database, this is likely not the case.In both other cases, SN2000dk is associated with NGC 382 and SN2015ar with NGC 383.NGC 382 and NGC 383 are interacting galaxies so it can be unclear which galaxy individual SNe are associated with if they are far from any one galaxy.In this case, as it is unclear if they are truly siblings or not, these objects are removed from the sample.(ii) In the OSC sample, SN2010af is located in PGC 36268 with SN2004fy.However, in the Q22 database, it is labelled as also being associated with NGC 3172 along with SN2017gla.This is the only such duplicate in the sample.These three SNe are all relatively close together, but as there is considerable uncertainty over the host association it is unclear if they are siblings, so they are removed from the sample.(iii) On the OSC and TNS, SN2007fq and SN2000dd are both associated with the galaxy MCG -04-48-19, where SN2007fq is close to the galaxy centre and SN2000dd is far from the galaxy light.However, in the Q22 database, which benefits from combining recent galaxy survey data, these are associated to two different, but close, galaxies.SN2007fq is reported as in ESO 528-G019 (MCG -04-48-19) and SN2000dd is located directly on a faint extended source, given the designation WISEA J203500.06-230556.9.This host may have been too faint in the original imaging of SN2000dd, hence the previous host association.As there is now doubt over the host association, this pair is removed from the sample.
The sample from Q22 adds an additional sibling, SN2020uoo, to the existing SN2017ets-SN2020kyx pair in CGCG 137-064.In all other cases, this sample confirms existing SN Ia sibling sets from the OSC sample, or adds independent sets of SN Ia siblings to the sample.
Combining the OSC sample from Section 2.1 with the 296 SNe Ia in 143 galaxies found in the Q22 sample gives a total of 148 galaxies containing 307 SN Ia siblings.

Literature Siblings
Not all surveys reported their transients to the OSC, or released them as part of a public data release that were queried by the OSC and Q22, and instead have individual sibling sample papers.In the literature, some samples of SN Ia siblings have been collated for the purposes of cosmology or as an additional product of a large cosmological survey.Therefore, they will not have undergone the same selection cuts as outlined in this analysis.However, to boost the sample size, I add any literature SN Ia siblings that are not already present in the sample.The siblings referred to in the large Brown (2014), Burns et al. (2020) and Pantheon+ (Scolnic et al. 2022) samples are found in the OSC and in the Q22 database, so the objects listed here are only those which do not: From DES, Scolnic et al. (2020), there are eight sets of siblings, consisting of 16 SNe Ia: From ZTF, Graham et al. (2022), some objects are already present in the sample due to having been already added to the OSC, the Q22 database, or both.Presented here are the two sets of siblings, consisting of four SNe Ia that were not already in the sample: and SN2019uej (ZTF19acnwelq)  • SN2019bbd (ZTF19aaksrgj) and SN2020hzk (ZTF20aavpwxl)

Combined Sample
Combining the SN Ia siblings found in the OSC with those in the Q22 database, and adding those literature examples that do not appear in either, I have a final sample of 158 galaxies, containing 327 confirmed SN Ia siblings -well over 10 times larger than prior sibling samples.I present this full sample in the appendix at the end of this paper in Table B1 of Appendix B.

Curiosities
I have identified a number of particularly unusual or notable SN Ia siblings within the sample.Here I give a short overview of each.
Siblings originally labelled as one object Typically, sibling SNe are well-separated within their host galaxies.However, there exist some recorded cases where surveys identified siblings as a single object.
Within DES, objects found within a 1 ′′ radius are automatically assigned to one candidate object.However, the high redshift coverage of DES means that entire galaxies at a higher redshift may be only around 1 ′′ in diameter.The light-curve of one event (DES16C3nd) has been manually identified as consisting of two clear SN signatures which were separated in time by 200 days (Scolnic et al. 2020).The light curves of these objects were classified separately and were both found likely to be SNe Ia.These objects were thus given the designations DES16C3nd 0 and DES16C3nd 1 .As both are SNe Ia, they are present in the sample.
In a similar situation to DES16C3nd within the DES survey, ZTF identified a pair of sibling SNe Ia which exploded within 200 days of each other (Biswas et al. 2022).These were found at nearly identical positions, with a separation of 0.6 ′′ corresponding to a projected distance of 0.6kpc.The second transient, SN2020aewj, was classified spectroscopically as a normal SN Ia, whilst the first transient, SN2019lcj, had to be classified photometrically.As the best-fit template reported in Biswas et al. (2022) was to a normal SN Ia, this set of siblings is present in the sample.
Alongside these, Soraisam et al. (2021) found a similar such object in the real-time alert stream of ZTF, ZTF20aamibse (AT2020caa).This object displayed a characteristic SN light curve in 2020, with a re-brightening in 2021.The spectrum of the measured peak in 2021 has been classified as a SN Ia, however there are no spectra for the 2020 event.The authors consider that this event is likely evidence of SN siblings, but cannot conclusively determine if both of the events are SN Ia in nature.For this reason, AT2020caa is not included in the sample.
There may be other such objects lurking in many SN surveys, and I encourage the community to carefully investigate their samples, as these peculiar objects may be unrecognised SN siblings.

Four sibling SNe Ia
The vast majority of the sample consists of galaxies hosting a pair of siblings.Whilst there are a few galaxies that contain three SNe Ia, to date only one galaxy, NGC 1316 (Fornax A), has been found to contain four SNe Ia.These are: SN1980N, SN1981D, SN2006dd and SN2006mr, and have been extensively studied by Stritzinger et al. (2010).
Longest time between siblings Considering the difference in time between sets of sibling pairs, and all pairwise combinations of those consisting of three or four SNe Ia, the average time between siblings in the data is 11 years, considerably shorter than that predicted for a Milky Way-like galaxy (approximately 1 every 200 years).The longest duration between sibling SNe Ia in the sample is the 81 years between between siblings in NGC 4636: SN1939A (discovered by Fritz Zwicky, one of the early pioneers of SN research, Zwicky 1939, for an overview see Graur 2022) and SN2020ue (discovered by prolific amateur SN hunter Koichi Itagaki, Itagaki 2020).I present a Sloan Digital Sky Survey (SDSS) gri composite image of the locations of these in Fig. 1 to give an example sky image displaying the locations of siblings with respect to their host galaxy.As can be seen, SN1939A is located close to the nucleus of the bright NGC 4636, whilst SN2020ue is found towards the outer reach of the galaxy.

COSMOLOGY WITH SN IA SIBLINGS
As an example use-case for this large sibling sample, I perform a brief analysis of their light-curve properties and discuss how they could be used in cosmology.

Photometry
First, I obtain photometry for each object in the sample.I require that all the data be publicly available.Most objects have photometry recorded on the OSC, which can be accessed by querying the OSC API.I use this photometry as a starting point, and supplement with high-quality public ZTF photometry on the ALeRCE ZTF Explorer 5 .Some SNe do not have publicly available photometry, whilst others do not have enough photometry to constrain a suitable model fit (in some cases, only one epoch of 'detection' photometry is available).These objects are rejected from the cosmology sample before lightcurve fitting.
For the DES siblings, the photometry is not publicly available.However, these have been previously fit with SALT2 in Scolnic et al. (2020), so I use the light-curve parameters recorded there.Additionally, the Pantheon+ siblings have already been fit in the same way by Scolnic et al. (2022), so I can use those light-curve parameters as-well and take the averages of the multiple measurements for each Pantheon+ SN.

Light curve fitting
In order to perform cosmology with the SN Ia siblings, the photometry needs to be of high quality so that the light curves can be accurately fit by models.I fit the light curves of the sample of sibling SNe Ia using the SALT2 fitter (Guy et al. 2010) implemented in SNCosmo (Barbary et al. 2022), with the redshifts defined as the host galaxy redshift for the shared host galaxies, consistent with the methods of Scolnic et al. (2020Scolnic et al. ( , 2022)).These fits return the overall amplitude ( 0 ), the light-curve stretch ( 1 ) and the light-curve colour ().As each sibling SN has been observed by different instruments, in different filters and at different times, I give SALT2 all available public data for each object, and it evaluates the appropriate wavelength range for fitting based on the redshift given for each SN, removing data outside of this range.To account for the fact that some SN have very late-time observations listed on the OSC, I apply a loose epoch cut of < 100 days post maximum brightness (as reported on the OSC) to constrain the fit based on the requirements of SALT2.I also apply fixed Milky Way dust extinction at the SN location using the Schlegel et al. (1998) dust map.
As the fits are of varying quality for each SN given the range of photometric data on the OSC, I apply a loose quality cut where fits are rejected based on their  2 and degrees of freedom from the SALT2 fit.To retain as many siblings as possible at this stage, I simply remove those where  2 = inf and dof = 0, indicating that the fitting has failed.
Ignoring those siblings that appear in the DES or Pantheon+ samples, I have 103 SNe Ia which pass these loose SALT2 requirements.I can then combine with the DES and Pantheon+ samples, and from the 327 confirmed SN Ia siblings, 131 are successfully fit with SALT2, or have published SALT2 properties from DES (Scolnic et al. 2020) or Pantheon+ (Scolnic et al. 2022).To these objects, I apply standard Joint Light-Curve Analysis (JLA)-like light-curve quality cuts (Betoule et al. 2014), i.e.: This results in a total of 91 SN Ia siblings, with selection cuts summarised in Table 1.In some cases, one SN will pass the quality requirements, whilst its sibling does not.In order to compare properties between siblings, I require that at least two SNe from a host galaxy pass the quality cuts, resulting in 21 sibling host galaxies, containing 44 SNe Ia (some host galaxies contain more than two SNe Ia which pass cuts).I visually inspect all SNe that pass these cuts, and confirm that all have successful fits.By considering all pairwise combinations, I have 25 pairs of siblings which pass the strict cosmology cuts.These are presented in Table 2.

Light-curve properties
I first compare the differences in light-curve properties  1 ,  and peak B-band apparent magnitude (m B ) for all pairs of SN siblings in the cosmology sample.These comparisons are presented in Fig. 2, with Pearson  values to interpret potential correlations in the data.
As there are uncertainties in both the x and y axes, I use a Monte Carlo technique to account for the uncertainties in the correlation.By taking the x and y uncertainties around each data point as a sampling area, I obtain a random location within the defined area for each SN pair and calculate the  value.I then apply this technique 10,000 times, taking the mean and standard deviation of the calculated  values as the defined  value and uncertainty on  for the data.This is repeated for  1 ,  and m B .
As shown in Fig. 2, the only quantity that is correlated between siblings is the apparent magnitude m B , with  = 0.995 ± 1.762 × 10 −4 indicating a very strong positive correlation, and a -value = (3.853± 1.856) × 10 −25 , meaning that the null hypothesis can be rejected and the correlation is highly significant.This is to be expected as m B is primarily dependent on the distance to the SN (or the redshift), so should be correlated for SNe at the same redshift.
For  1 and , it can be shown that there is considerable scatter, and the -value and -values for each are comparable.For  1 , the -value is measured at  = 0.422 ± 0.063.Whilst this may be considered to be a weak positive correlation, the -value of 0.046 ± 0.037 suggests that this is not significant and the x and y data (i.e., the  1 values between sibling pairs) may be unrelated, assuming a typical -value threshold of 0.01.This is also similar for , with  = 0.419 ± 0.050 and  = 0.044 ± 0.029; meaning that the  values between sibling pairs may be unrelated.A larger sample of cosmological siblings is needed to further investigate this.
Given that I have cosmological siblings that span a large redshift range, it is prudent to investigate potential trends with redshift.Thus, in Fig. 3, I present the differences in  1 and  between sibling pairs (Δ 1 and Δ) as a function of redshift.No clear trends are evident, but I note that the overall scatter in Δ 1 and Δ seems to reduce towards higher redshifts.This however is likely due to small number statistics, as I only have five pairs of SNe with  > 0.1.This may also be due to Malmquist bias, as at high redshift we are more likely to observe the brighter, high-stretch SNe.
I also investigate potential trends with the angular separation between sibling pairs, and with the difference in the date of peak magnitude between sibling pairs, finding no evidence of any trend for either, as shown in Fig. 4.
The scatter in Δ 1 , Δ and Δm B provides an interesting way to compare siblings to a sample of random pairs of SNe, as in Scolnic et al. (2022).I find the standard deviations () of the Δ 1 , Δ and Δm B values and present them in Table 3.
To compare the results obtained for SN Ia siblings to a standard sample of SNe Ia (i.e., non-siblings), I perform a simple boot- MNRAS 000, 1-14 ( 2023) strapping analysis.I take the sample of SNe Ia from Pantheon+ (Scolnic et al. 2022), make a cut on redshift to match the siblings in my cosmology sample, and apply the same JLA-like light-curve quality cuts from Section 3.2.This gives 1418 SNe Ia to sample from.I select 25 pairs of random SNe from this sample, and repeat the Monte Carlo analysis to obtain Pearson  values and to calculate scatter in light-curve properties.I repeat this 10,000 times and obtain the means and standard deviations of the data.I study the effect of performing this boot-strapping analysis on smaller redshift sub-samples in Appendix A.
For the  values comparing light-curve properties between random pairs of SNe, I find that there are consistently no correlations for the different properties.In detail, I find for  1 the -value is measured at  = 0.002 ± 0.057 with a -value of 0.503 ± 0.137; for ,  = 0.004 ± 0.073 with a -value of 0.497 ± 0.167; and for m B ,  = 0.001 ± 0.003 with a -value of 0.501 ± 0.008.Whilst the lack of any correlation suggests that one could argue that sibling SNe are more correlated than any random pair of SNe, the associated -values are consistent with a lack of correlation across both data and the boot-strap sample.Therefore I can only confidently say that the m B values for SN siblings are more similar than any other random pair of SNe, which is to be expected.
Comparing between data and boot-strapped standard deviations () in the differences in light-curve properties between sibling pairs can provide more answers.Presented in Table 3 are the different  values for the sibling data and for any random pair of SNe.As can be seen, the boot-strapped  values are higher than that for the data, with the most significant difference (at 6 significance) being for the difference in scatter in Δm B , which will be due to the siblings being co-located in the same galaxy.For Δ 1 and Δ, the differences in scatter are < 2 different.Therefore, I can say that  1 and  are no more similar between SNe Ia associated with the same host galaxy than for any random pair of SNe Ia.This finding raises important questions over the validity of global host galaxy corrections in the standardisation of SNe Ia for use in cosmology.
This finding for  1 disagrees with Scolnic et al. (2020Scolnic et al. ( , 2022)), which both find tentative evidence that  1 is more correlated between siblings than random pairs of SNe.This difference may be due to different light-curve quality cuts in our samples, or their smaller  Assuming a simple reference cosmology of  = 0.148 and  = 3.112 based on the values for Pantheon+ in Scolnic et al. (2022), the standard deviation on the difference in distance modulus between siblings in this analysis is 0.185 mag.This is less than the 0.32 mag value found by Scolnic et al. (2022) for the Pantheon+ siblings alone, but is still more than the ∼ 0.15 mag dispersion on the Hubble diagram (e.g.Brout et al. 2022), agreeing with the conclusions that siblings are no more similar than any random pair of SNe Ia.This is in contrast to studies such as Burns et al. (2020), who find that the relative distances between siblings agreed within 3%.However, these were all spectroscopically confirmed SNe Ia with well-sampled light-curves, consistent photometry and data for sets of siblings from the same telescopes.Whereas the sample compiled by my analysis comprises a mixture of those features.In particular, Burns et al. (2020) note that their largest distance discrepancies (5 − 10%) arise when siblings are observed with different telescopes.This may be due to systematics in the photometric systems, but may also be due to differences in the SNe themselves.As postulated in Section 1, to have been observed by the same survey, it is more likely that the SNe Ia will have come from a more massive galaxy, hosting faster SNe Ia, from older progenitors.Contrasting with Burns et al. (2020), in this analysis the siblings with the most differing distance modulii (0.725 mag) are SN2019bbd and SN2020hzk, both observed by ZTF.Despite having similar  1 values, the differences in  between the siblings is high, suggesting perhaps differing levels of dust attenua-tion or differing progenitor properties, as also found by Elias-Rosa et al. (2008) and Biswas et al. (2022) for other sets of siblings.Future analysis is needed to understand how much of this difference across sets of siblings is an environmental effect, and how much is a systematic from combining different telescope data.Regardless of the cause, understanding these differences between SNe Ia from the same galaxy is vital as we enter the new era of precision cosmology.

DISCUSSION AND CONCLUSIONS
In this analysis I have identified the largest sample of SN Ia siblings to date by searching the Open Supernova Catalog, the Q22 Extragalactic Transient Host Galaxy Database, and the literature.This results in a combined sample of 158 galaxies, containing 327 SN Ia siblings, over an order of magnitude larger than prior samples.
By creating a sub-sample based on photometry quality and applying JLA-like light-curve quality cuts, I identified a cosmology sample of 25 pairs of siblings, which I used to compare light-curve properties between sets of siblings.I found that the only property that was correlated between siblings was the peak apparent magnitude m B , which is expected due to sets of sibling SNe being at the same redshift.There was no correlation in either  1 or  for SNe associated with the same host galaxy, suggesting that these properties are not consistent for SNe Ia associated with the same galaxies.
Using a sample of non-sibling SNe Ia from the Pantheon+ sample, I compare the siblings to random pairs of SNe Ia, finding that  1 and  values for siblings are no more similar than for any random pair of SNe Ia.In the current cosmology formalism, environmental effects are corrected for by considering the mass of the galaxy that each SN is associated with, meaning that sibling SNe Ia are subject to the same corrections.My finding that their  1 and  values are no more similar than any random pair of SNe raises important questions about the validity of this method, as this global host galaxy correction does not account for sub-galactic environmental differences.The local environments of sibling SNe Ia should be investigated and compared to light-curve properties to understand if using local corrections may better account for environmental property standardisation in cosmology as suggested by Kelsey et al. (2021Kelsey et al. ( , 2023)).It is clear that SN siblings are a vital tool to understand the remaining dispersion in SN Ia luminosities.
As we enter the new era of large surveys, the number of SN Ia siblings will dramatically increase.Scolnic et al. (2020), for example, predict around 800 pairs in LSST, alongside the future matching of newly discovered SNe with archival SNe sharing the same host galaxy.This will allow for further analysis of differences between sibling pairs, and for the detailed investigation of their environments.small redshift bins.Given siblings are at the same redshift, whilst the Pantheon+ sample spans a wide redshift range, this test will allow further understanding of the similarities or differences between siblings and random pairs of SNe Ia.
Given the wide redshift range of the siblings, I create three subsamples as follows: a) 0 <  ≤ 0.01, containing 11 pairs, b) 0.01 <  ≤ 0.1, containing 9 pairs, and c) 0.1 <  ≤ 0.648, containing the 5 DES pairs.Breaking down the sample into further sub-samples with a smaller redshift difference creates a large number of bins containing only two pairs of siblings, making statistical comparison difficult.More cosmology-quality siblings will need to be found to improve the statistical power of this type of analysis.
For each defined subsample, I cut down the Pantheon+ sample to match the subsample redshift range, then perform the boot-strapping analysis as outlined in Section 3.3 for each case.Comparisons between light-curve parameters in the different subsamples are presented in Fig. A1, with the -and -values for the siblings displayed in the top left corners.The Pearson -values and corresponding values for the sibling data and the boot-strapped data for each bin are presented in Table A1, with the  values in Table A2, comparable to Table 3 in the main analysis.
As can be seen in Fig. A1, and highlighted by Table A1, the sibling data are more strongly correlated than the boot-strap data for all properties in all redshift bins.This might suggest that the siblings are more similar than random pairs in small redshift bins, however, as in the main analysis, the -values for all properties except m B are consistent with a lack of correlation in both data and the bootstrap sample.This means that only the m B values for SN siblings are more similar than any other random pair of SNe, even when only considering subsamples of small redshift bins.I do note however that the  1 -value for the lowest redshift bin may be considered weak evidence that stretch is more correlated for siblings than for random pairs of SNe Ia at low redshift due to being close to the -value threshold when accounting for the uncertainty.
As in the main analysis, I compare between data and boot-strapped standard deviations () in the differences in light-curve properties between sibling pairs.These are presented in Table A2.For both Δ 1 and Δ, the boot-strapped and data  values are consistently less than 2 in significance, so it can therefore be concluded that  1 and  values are no more similar for SN siblings than for any random pair of SNe in the same redshift bin.For m B , the data and bootstrapped  values are more consistent in redshift bins than they were when considering the entire sample, with only the lowest redshift bin having > 3 significance indicating siblings being more similar than any random pairs.This bin also contains the siblings with the most similarities in  1 when compared to random pairs at a close to 2 difference.These results agree with my findings and conclusions from the main analysis, that sibling SNe Ia are no more similar than any random pair of SNe Ia.

Figure 1 .
Figure 1.SDSS gri composite colour image of NGC 4636, the host galaxy in the sample hosting the siblings that are most differing in explosion date, with SN locations indicated by green circles.The SN closest to the galactic core is SN1939A, whilst SN2020ue is towards the edge of the galaxy.Imaging data obtained from SDSS Data Release 17, Abdurro'uf et al. (2022).

Figure 2 .Figure 3 .
Figure 2. Comparison of light-curve parameters  1 (top panel),  (middle panel) and m  (lower panel) for all pairs of SN siblings in the cosmology sample.The blue dashed line represents the  =  line where sibling lightcurve parameters are equal.Presented in the top left corner for each plot are the Pearson  values for the data, the associated -values, and uncertainties on both.

Figure 4 .
Figure 4. Differences in light-curve parameters  1 (top panel) and  (lower panel), as a function of (a) angular separation between SNe for all pairs of SN siblings in the cosmology sample, and (b) the difference in the date of peak magnitude for all pairs of SN siblings in the cosmology sample.

Table 1 .
Summary of selection requirements

Table 2 .
Cosmology sample of SN Ia Siblings.

Table 3 .
Standard deviation  of the differences in light-curve properties between SN 1 and SN 2 for the data and for a sample obtained through bootstrapping.

Table A1 .
Pearson -values and corresponding -values for boot-strapping analysis.

Table A2 .
Standard deviation  of the differences in light-curve properties between SN 1 and SN 2 for the data and for a sample obtained through boot-strapping for a range of different bins divided by redshift.

Table B1 :
Full SN Ia Sibling Sample