THE EVOLUTION OF RED SUPERGIANTS TO SUPERNOVA
Emma Beasor NASA Hubble Fellow NOIRLab
Ben Davies (LJMU), Nathan Smith (U. Arizona), Nate Bastian (LJMU), Bob Gehrz (U. Minnesota), Don Figer (Rochester), Jacco van Loon (Keele) SOFIA Tele-talk, July 15th 2020 Outline
Introduction to RSGs Evolution of RSGs to SN Mass-loss Age determinations Summary
!2 RED SUPERGIANTS
Evolved massive stars (8-25 M⊙) Direct progenitors to Type II SN - powerful test of stellar evolutionary theory
!3 EVOLUTION OF A 15M⊙ STAR
H-burning shell
H-burning core He-burning core Envelope Envelope
Core contracts… star swells up
!4 EVOLUTION OF A 15M⊙ STAR
H-burning shell
H-burning core Envelope He-burning core
Envelope
Core contracts, star swells up.
Lots of convection in the!5 envelope EVOLUTION TO RSGPHASE log(L/L 4.5 5.5
⊙ 50,000 Main sequence Main ) 30M 20M ⊙ 10M ⊙ ⊙ 20,000 credit: BenDavies !6 Evolutionary tracks…. Evolutionary 4,000 log(T/K)
Red Supergiants THE PATH TO SUPERNOVA
He-core gets exhausted, fuses into carbon core, which fuses into oxygen, which fuses into neon… Core gets heavier and heavier No more nuclear reactions Star collapses onto the core…
Fe
Core-collapse supernovae
!7 THE PATH TO SUPERNOVA
What kind of SN depends on the appearance of the progenitor at core collapse Strong winds peel away H-rich envelope envelope RSGs live ~106 yrs, mass-loss timescale (M/Mdot) is about the same Whole envelope can be peeled off through lifetime
!8 TWO OPTIONS…
‘Wolf-Rayet’ (hot RSG/YSG progenitor progenitor) H-rich envelope intact Stripped/H-poor SN (type Unstripped SN (Type II) Ibc)
!9 We know RSGs explode as II-P SNe SN —> check archival images —> identify progenitor
Pre-explosion photometry + some assumptions allow us to find the terminal luminosity of the progenitor and infer a mass
!10 Red supergiant problem…
observed… predicted…
Maximum IIP progenitor mass ~16Msun?
Smartt et al. 2009, updated in Smartt 2015 !11 EVOLUTION FROM MS TO RSG
G. Meynet & A. Maeder: Stellar evolution with rotation. V 111
Meynet & Maeder
60M⊙ - skip RSG phase, straight
40M⊙ - short RSG phase,
25M⊙ - RSG phase, dies Fig. 8. Evolutionary tracks for non–rotating (dotted lines) and rotating (continuous lines) models with solar metallicity. The rotat- ing models have an initial velocity vini of 300 km s−1. For purpose of clarity, only the first part of the tracks for the most massive stars (M ≥ 40 M⊙) is shown. Portions of the evolution during the W–R phase for the rotating massive stars are indicated by short– dashed lines. The long–dashed track for the 60 M⊙ model corresponds to a very fast ro- −1 tating star (vini ∼ 400 km s ), which fol- lows a nearly homogeneous evolution. Only the beginning of its evolution is shown. !12 University of Sheffield - Oct
as red supergiants at Teff below 4000 K, while the non–rotating 5.3. Masses and mass–luminosity relations models spend a significant part of the He–burning phase in the When rotation increases, the actual masses at the end of both blue part of the HR diagram: for the non–rotating 15 and 20 the MS and the He–burning phases become smaller (cf. Tables 1 M⊙ models, respectively 25 and 20% of the total He–burning and 2). Typically the quantity of mass lost by stellar winds dur- lifetime is spent at log T ≥ 4.0. The behaviour of the rotat- eff ing the MS is enhanced by 60–100% in rotating models with ing models results mainly from the enhancement of the mass v = 200 and 300 km s−1 respectively. For stars which do not loss rates. This effect prevents the formation of a big interme- ini go through a Wolf–Rayet phase, the increase is due mainly to diate convective zone and therefore favours a rapid evolution the direct effect of rotation on the mass loss rates (in the present toward the RSG phase (Stothers & Chin 1979; Maeder 1981). models through the formula proposed by Friend & Abbott 1986) Let us note that the dispersion of the initial rotational velocities and to the higher luminosities reached by the tracks computed produces a certain distribution of the above behaviours. with rotation. The fact that rotation increases the lifetimes also Very interestingly, for the 12 M⊙ model a blue loop appears contributes to produce smaller final masses. For the most mas- when rotation is included. This results from the higher lumi- sive stars (M ≥ 60 M⊙), the present rotating models enter the nosity of the rotating model. The higher luminosity implies that Wolf–Rayet phase already during the H–burning phase (see also the outer envelope is more extended, and is thus characterized Maeder 1987; Fliegner & Langer 1995; Meynet 1999, 2000b). by lower temperatures and higher opacities at a given mass co- This reduces significantly the mass at the end of the H–burning ordinate. As a consequence, in the rotating model during the phase. first dredge–up, the outer convective zone proceeds much more As indicated in Sect. 5.1, the initial distribution of the ro- deeply in mass than in the non–rotating star. Typically in the tational velocities implies a dispersion of the luminosities at non–rotating model the minimum mass coordinate reached by the end of the MS. This effect introduces a significant scatter the outer convective zone is 6.6 M⊙ while in the rotating model in the mass–luminosity relation (Langer 1992; Meynet 1998), it is 2.6 M⊙. This prevents temporarily the extension in mass of in the sense that fast rotators are overluminous with respect to the He–core and enables the apparition of a blue loop. Indeed their actual masses. This is especially true in the high mass star the lower the mass of the He–core is, the lower its gravitational range in which the luminosity versus mass relation flattens. This potential. According to Lauterborn et al. (1971, see also the dis- may explain some of the discrepancies between the evolutionary cussion in Maeder & Meynet 1989), a blue loop appears when masses and the direct mass estimates in some binaries (Penny the gravitational potential of the core Φ is inferior to a crit- c et al. 1999). ical potential Φ depending only on the actual mass of the crit Let us end this section by saying a few words about the star which is about the same for the rotating and non–rotating mass discrepancy problem (see e.g. Herrero et al. 2000). For model. This explains the appearance of a blue loop in the 12 some stars, the evolutionary masses (i.e. determined from the M⊙ rotating model. For the 9 M⊙ model, the minimum mass theoretical evolutionary tracks) are greater that the spectroscop- coordinate reached by the outer convective zone is not much ically determined masses. Interestingly, according to Herrero et affected by rotation and the models with and without rotation al. (2000), only the low gravity objects present (if any) a mass present very similar blue loops. discrepancy. Even if most of the problem has collapsed and was …MODEL DEPENDENT
what changed..? abundances, opacities, overshooting… mass-loss rate implementation
!13 MASS-LOSS RATE IMPLEMENTATION
log(L/L⊙) in addition, in Ekstrom+ 2012: if L > 5LEdd, Ṁ = Ṁx3 (kicks in at ~20M⊙) Ṁ : Main sequence Ṁ : Vink / de Jager / Nugis & Lamers / van Loon Kudritzki
5.5
30M⊙ Supergiants 20M⊙ 4.5 10M⊙ Red Yellow Supergiants Yellow
50,000 20,000 4,000 log(T/K) credit: Ben Davies !14 EMPIRICAL MASS-LOSS RELATIONS
Mass-loss rates are not calculated from first principles
!15 Mauron & Josselin 2010 EMPIRICAL MASS-LOSS RELATIONS
Mass-loss rates are
not calculated from Reid+ 1990 first principles dusty RSGs van Loon+ Lots of internal 2005 scatter ± x10 Groenewegen+ 2009
de Jager Bonanos+
!16 ̣ ̣ ̣ No longer used for OBA radio…) excesses, absline analysis, methodologies (mid-IR Highly heterogeneous metallicities…) (masses, sample Highly heterogeneous
1988A&AS...72..259D !17 Needs abitofanupdate… REAPPRAISAL OF RSG MASS-LOSS
By targeting RSGs in clusters, we can assume all RSGs
are the same Z and Late stage same Mini RSGs We can use L as a proxy for
evolution evolution
Early stage RSGs NGC 2100
!18 SOFIA - new data (PI N Smith) WISE - archival
MSX - archival
SPITZER - archival
!19 HOW DO WE MEASURE MASS-LOSS RATES?
Dust layer
Dust layer absorbs and re-emits photons Mass-loss can be RSG measured by modeling mid-IR excess
!20 HOW DO WE MEASURE MASS-LOSS RATES?
The shape of this bump tells us about the dust composition, for RSGs it’s silicate rich dust Light absorbed by dust here Flux
Re-emitted here
Wavelength !21 HOW DO WE MEASURE MASS-LOSS RATES?
We model this emission using a radiative transfer code (DUSTY), and derive a mass-lossflux rate for each RSG in the cluster
wavelength !22 NGC2100
Dust shell mass increases with Tight correlation… evolution Fixed initial mass and Z
Beasor & Davies 2016
!23 HOW DO MASS-LOSS RATES CHANGE WITH INITIAL MASS?
χ Per NGC 2100 RSGC1
Repeat for clusters of different ages (and hence RSGs of different initial masses..) NGC 7419 NGC 2004 !24 A MASS-DEPENDENT MASS-LOSS RATE PRESCRIPTION
Same gradient to within the errors…
Beasor et al. (2020) !25 A MASS-DEPENDENT MASS-LOSS RATE PRESCRIPTION
increasing initial mass
Beasor et al. 2020 !26 COMPARISON TO OTHER PRESCRIPTIONS
Our prescription de Jager Offset = 1.4
RMS = 0.45 RMS = 0.52 Residuals [ log(Mdot/Msun) ]
Lower scatter Scatter slightly higher No offset Avg offset = 0.13 BUT much worse for higher luminosity stars…
Beasor et al. 2020 !27 COMPARISON TO EVOLUTIONARY MODELS
20 M star Solid line - ⊙ current implementation Dashed line - our ~1M ⊙ prescription
~10M⊙
Quiescent mass- loss cannot remove the H envelope….
Beasor et al. 2020 !28 What does it all mean…
Observed H-poor SN fraction ~ 1/3 Back of the envelope IMF calculation… H-poor % stars 8-30M⊙ ~ 85% % stars >30M⊙ ~ 15%
Smith et al. 2011
!29 What does it all mean…
If mass-loss rates were higher… could explain this discrepancy
H-poor % stars 8-16M⊙ ~ 60% % stars >16⊙ ~ 40%
Smith et al. 2011
!30 What does it all mean…
But, mass-loss rates are lower Single star evolution cannot explain the H-poor observed SN rate Strong evidence for most H-poor SN being the products of binary interaction
Smith et al. 2011
!31 CONCLUSIONS: Part 1
There is no observationally motivated reason to increase mass-loss by factors of 3 or more in stellar evolution models
RSGs that evolve as single stars do not shed their envelope via quiescent mass loss
Single stars between 20-30M⊙ do not lose enough mass to evolve blueward
The relative number of stripped/unstripped SN events predicted by single star models is way off
Something else (binaries??) is removing the envelope
!32 HOW WELL DO WE KNOW CLUSTER AGES?
χ Per NGC 2100 RSGC1
NGC 7419 NGC 2004
!33 USING THE CMD
Bastian et al. 2018 For old and intermediate age clusters (>50Myr), many observational effects can’t be explain by SSP… e.g. blue stragglers
!34 CMD of NGC 7419
RSGs MSTO
!35 METHOD 1: brightest TO star
!36 METHOD 2: luminosity function of the TO
Slightly more robust?
!37 METHOD 3: lowest luminosity RSG
G. Meynet & A. Maeder: Stellar evolution with rotation. V 111
Constant luminosity Fig. 8. Evolutionary tracks for non–rotating (dotted lines) and rotating (continuous lines) models with solar metallicity. The rotat- ing models have an initial velocity vini of 300 km s−1. For purpose of clarity, only the first part of the tracks for the most massive stars (M ≥ 40 M⊙) is shown. Portions of the evolution during the W–R phase for the rotating massive stars are indicated by short– dashed lines. The long–dashed track for the Meynet & Maeder 2000 60 M⊙ model corresponds to a very fast ro- −1 tating star (vini ∼ 400 km s ), which fol- lows a nearly homogeneous evolution. Only !38 the beginning of its evolution is shown.
as red supergiants at Teff below 4000 K, while the non–rotating 5.3. Masses and mass–luminosity relations models spend a significant part of the He–burning phase in the When rotation increases, the actual masses at the end of both blue part of the HR diagram: for the non–rotating 15 and 20 the MS and the He–burning phases become smaller (cf. Tables 1 M⊙ models, respectively 25 and 20% of the total He–burning and 2). Typically the quantity of mass lost by stellar winds dur- lifetime is spent at log T ≥ 4.0. The behaviour of the rotat- eff ing the MS is enhanced by 60–100% in rotating models with ing models results mainly from the enhancement of the mass v = 200 and 300 km s−1 respectively. For stars which do not loss rates. This effect prevents the formation of a big interme- ini go through a Wolf–Rayet phase, the increase is due mainly to diate convective zone and therefore favours a rapid evolution the direct effect of rotation on the mass loss rates (in the present toward the RSG phase (Stothers & Chin 1979; Maeder 1981). models through the formula proposed by Friend & Abbott 1986) Let us note that the dispersion of the initial rotational velocities and to the higher luminosities reached by the tracks computed produces a certain distribution of the above behaviours. with rotation. The fact that rotation increases the lifetimes also Very interestingly, for the 12 M⊙ model a blue loop appears contributes to produce smaller final masses. For the most mas- when rotation is included. This results from the higher lumi- sive stars (M ≥ 60 M⊙), the present rotating models enter the nosity of the rotating model. The higher luminosity implies that Wolf–Rayet phase already during the H–burning phase (see also the outer envelope is more extended, and is thus characterized Maeder 1987; Fliegner & Langer 1995; Meynet 1999, 2000b). by lower temperatures and higher opacities at a given mass co- This reduces significantly the mass at the end of the H–burning ordinate. As a consequence, in the rotating model during the phase. first dredge–up, the outer convective zone proceeds much more As indicated in Sect. 5.1, the initial distribution of the ro- deeply in mass than in the non–rotating star. Typically in the tational velocities implies a dispersion of the luminosities at non–rotating model the minimum mass coordinate reached by the end of the MS. This effect introduces a significant scatter the outer convective zone is 6.6 M⊙ while in the rotating model in the mass–luminosity relation (Langer 1992; Meynet 1998), it is 2.6 M⊙. This prevents temporarily the extension in mass of in the sense that fast rotators are overluminous with respect to the He–core and enables the apparition of a blue loop. Indeed their actual masses. This is especially true in the high mass star the lower the mass of the He–core is, the lower its gravitational range in which the luminosity versus mass relation flattens. This potential. According to Lauterborn et al. (1971, see also the dis- may explain some of the discrepancies between the evolutionary cussion in Maeder & Meynet 1989), a blue loop appears when masses and the direct mass estimates in some binaries (Penny the gravitational potential of the core Φ is inferior to a crit- c et al. 1999). ical potential Φ depending only on the actual mass of the crit Let us end this section by saying a few words about the star which is about the same for the rotating and non–rotating mass discrepancy problem (see e.g. Herrero et al. 2000). For model. This explains the appearance of a blue loop in the 12 some stars, the evolutionary masses (i.e. determined from the M⊙ rotating model. For the 9 M⊙ model, the minimum mass theoretical evolutionary tracks) are greater that the spectroscop- coordinate reached by the outer convective zone is not much ically determined masses. Interestingly, according to Herrero et affected by rotation and the models with and without rotation al. (2000), only the low gravity objects present (if any) a mass present very similar blue loops. discrepancy. Even if most of the problem has collapsed and was Results…
Big disagreement in ages between the methods
This is seen for all clusters in our sample
!39 What’s going on..?
Rotation can’t explain the offset… v/vcrit =0.95 is really really fast
Mergers???
!40 Testing with synthetic clusters… - single stars
Single stars only. TO method underestimates the age by quite a lot
!41 Testing with synthetic clusters… - binary fraction of 50%
Binary fraction of 50%.
Even worse for TO… RSGs do better.
!42 CASE STUDY: Westerlund 1
Beasor et al. (submitted)
Supposedly a very young Galactic cluster (~4Myr), and massive (105 Msun). !43 CASE STUDY: Westerlund 1
Beasor et al. (submitted)
First time we’ve been able to attempt a bolometric luminosity for these RSGs… !44 CASE STUDY: Westerlund 1
4.5 Myrs
5.0 Myrs
5.6 Myrs
6.3 Myrs
7.1 Myrs W237 W26 7.9 Myrs
8.9 Myrs
10.0 Myrs W20
11.2 Myrs
W75
Beasor et al. (submitted) !45 CONCLUSIONS: Part 2
Using the cluster turn-off to estimate age will cause ages to be underestimated
Using red supergiants allows a binary independent age to be determined
There could be lots of mergers/mass transfer systems in young clusters
Westerlund 1 probably isn’t as young as people first thought
!46