The observed mass loss vs. metallicity relation

1
The observed mass loss vs. metallicity relation

• Alex de Koter

University of Amsterdam
2
The observed mass loss vs. metallicity relation

• Alex de Koter
• Rohied Mokiem, Chris Evans, S. Smartt, J. Puls,
  A. Herrero, F. Najarro, M.R. Villamariz

University of Amsterdam
3
The problem of determining dM/dt(Z)

• Large set of OB stars of sufficient baseline in Z
• GAL, LMC, SMC
• sampling of spectral sub-types and luminosity
  classes
• Line blanketed, unified model atmospheres
• dM/dt
• Z
• Homogeneous analysis
• automated (well defined) method
• realistic errors
• Physics? among others
• clumping
• velocity structure

4
Z diagnostics

• Ideally
• stars (not nebulae)
• that reflect the composition of O stars for which
  dM/dt is determined (so, e.g. B V or O stars
  themselves)
• that do not show products of nucleosynthesis (no
  fast rotators, no Supergiants)
• with sufficiently strong abundant optical
  lines (no fast rotators, high S/N)
• of elements driving the wind (Fe dominant)

5
Z diagnostics

• Unfortunately
• B V (and O themselves) have only few and weak
  optical Fe lines (still Rolleston et al. 2003)
• But
• Supergiants only show products of CNO - not too
  relevant for driving (e.g. Venn 1999)
• one may resort to modeling UV spectrum, plenty Fe
  lines (Bouret et al. 2003)
• Findings are
• AFG I O(UV) ?Fe 0.2 Fe? B V ? Fe 0.3 Fe?
  for SMC

6
dM/dt diagnostics

• Ha, other optical lines (e.g. He II 4686)
• difficult for low dM/dt (lt few times 10-8 M?/yr)
• UV resonance lines
• most sensitive dM/dt diagnostic
• trace ions, sensitive to X-rays, shocks, atomic
  data
• Radio flux
• limited to Galactic stars

7
Detailed line-blanketed unified studies

• FASTWIND (Puls et al. 2005)
• dM/dt from Ha
• H, He, Si explicit
• fast individual /- errors
• CMFGEN (Hillier Miller 1998)
• dM/dt from Ha and UV resonance lines
• allows treatment of clumping
• H, He, CNO, Si,S, Fe explicit
• slow typical errors

8
Detailed line-blanketed unified studies
Code Study GAL LMC SMC
fastwind Puls et al. 1996 (unblocked) 24 6 8
Repolust et al. 2004 24
Massey et al. 2004 9 11
Massey et al. 2005 15 5
Mokiem et al. 2005 (GA), Herrero et al. 2002 12
Mokiem et al. 2005 (GA) 31
Mokiem et al. in progress 20

cmfgen Crowther et al. 2002 3 1
Hillier et al. 2003 2
Bouret et al. 2003 6
Evens et al. 2004 2 6
Martins et al. in press 11
9
Detailed line-blanketed unified studies
Code Study GAL LMC SMC
fastwind Puls et al. 1996 (unblocked) 24 6 8
Repolust et al. 2004 24
Massey et al. 2004 9 11
Massey et al. 2005 15 5
Mokiem et al. 2005 (GA), Herrero et al. 2002 12
Mokiem et al. 2005 (GA) 31
Mokiem et al. in progress 20

cmfgen Crowther et al. 2002 3 1
Hillier et al. 2003 2
Bouret et al. 2003 6
Evens et al. 2004 2 6
Martins et al. in press 11
10
Analyzed sample of O stars
11
Analyzed sample of O stars
12
Homogeneous analysis

• FASTWIND CMFGEN compare well
• good agreement, except for ionising flux lt 400 A
  and He I singlets (Puls et al. 2005)
• CMFGEN TLUSTY compare well (Bouret et al. 2003)
• would benefit from automated analysis

13
Automated method (Mokiem et al. 2005)

• Genetic Algorithm based optimization
• FASTWIND stellar atmosphere model
• PIKAIA GA optimization (Charbonneau 1995)
• Fitting of optical H,He lines and V magnitude
• L, Teff, g, YHe, vsini, vturb, dM/dt, ?
• Homogeneous analysis of large samples

14
Optimization technique
Teff g YHe vturb dM/dt ? v?
15
O5 If star example
HeI 4471 HeII 4541 HeII 4686

• GA method gives fit of at least the quality
  obtained by
• experts in modelling
• Well determined errors
• Helps to resolve issues related to mass
  discrepancy
• One fit per day

16
Clumping in O stars winds

• Evidence for clumping in O stars
• Eversberg et al. 1998, (stochastic substructure
  He II)
• Crowther et al. 2002, Hillier et al. 2003 (PV)
• Bouret et al. 2003 (O V 1371)

17
Clumping in O stars winds

• dM/dt / vf invariant
• clumping leads to ? dM/dt
• f 0.01 0.2
• Clumping physics not well known (Owocki et al.)
• radial behaviour f ?
• potentially different observed clumping in V
  and I stars
• interclumped medium? (assumed void)
• dimension of clumps? (assumed to be small)
• time dependence?

18
Modified Wind Momentum
log (dM/dt v8 vR) x log (L/L?) log Do x
1/(a-d) 0 (all thin) 1 (all thick)
19
Modified Wind Momentum
log (dM/dt v8 vR) x log (L/L?) log Do x
1/(a-d) 0 (all thin) 1 (all thick)

• Theory no rotation, clumping, spherical outflow
  -predicts a unique relation, for fixed metal
  content
• (Vink et al. 2000, Pauldrach et al. 2003, Puls et
  al. 2003)

20
Galactic O stars
21
Galactic O stars
22
Galactic O stars
? Repolust et al. 2004
23
LMC O stars
24
LMC O stars
25
SMC O stars
26
SMC O stars
27
Steep turn in MWM at log(L/L?) lt 5.25
Bouret et al. 2003
28
Steep turn in MWM at log(L/L?) lt 5.25
Bouret et al. 2003
not even corrected for clumping!
29
Why this break-down at low L??

• NOT bi-stability jump, as this causes a jump up
  in dM/dt
• 2 early B I stars that do not appear to obey the
  turn-down may be at cool side, having higher
  dM/dt due to different set op driving lines

30
Why this break-down at low L??

• Perhaps derived rates are wrong
• ionization predictions depend on X-rays (shocks),
  treatment of clumping, atomic data
• discrepancy Ha and UV dM/dt determination
• uncertainties in v8
• Perhaps predictions are wrong
• for low density winds Sobolev approx. for line
  force may break down (Owocki Puls 1999)
• ion decoupling (? Krticka Kubat)
• Perhaps low dM/dt stars have a different nature
• Vz stars

31
Steep turn in MWM at log(L/L?) lt 5.25
Martins et al. 2005
32
Steep turn in MWM at log(L/L?) lt 5.25
33
SMC O stars
34
GAL, LMC, SMC O stars
35
GAL, LMC, SMC O stars
36
Conclusions Remarks

• Number of O stars analyzed rapidly increasing
• multi-object spectroscopy (such as FLAMES
  program)
• automated fitting
• fast codes
• Model atmosphere codes in good agreement
• Not clear whether dM/dt diagnostics in agreement
  in low-brightness O stars
• UV vs Ha problem

37
Conclusions Remarks

• dM/dt(Z) in terms of MWM(Z)
• appears to be a kink in MWM slope at
  log(L/L?) 5.25, not predicted by current
  theory, leading to much lower dM/dt
• uncorrected supergiants yield 0.4 dex
  higher dM/dt than Vink et al. predictions
• bright OB stars dM/dt Z-m, m 0.7-0.9
  (corrected for v8(Z) following Leitherer et al.
  1992), in agreement with theory

38
Conclusions Remarks

• If clumping important
• dM/dt of O stars may all (?) be overestimated by
  factors 3

39
End
40
End
41
Large sample analysis

• Fundamental parameters of massive stars
• 100 analyzed until now
• Multi-object spectrographs
• VLT-Flames Survey
• 100 hours VLT time
• Galactic, SMC and LMC fields
• will double total number of analyzed stars
• calibration of fundamental parameters
• requires an automated method

42
How can we study massive stars?
43
How can we study massive stars?
44
Optimization technique
Teff g YHe vturb dM/dt ? v?
45
The automated method

• FASTWIND stellar atmosphere model
• Puls et al. 2005, AA, 435, 669
• PIKAIA genetic algorithm optimization
• Charbonneau1995, ApJS, 101, 309
• Fitting of optical hydrogen and helium lines
• Teff, g, YHe, vsini, vturb, dM/dt, ?

46
The observed mass loss vs. metallicity relation

• Alex de Koter

University of Amsterdam
47
O5 If star example
HeI 4471 HeII 4541 HeII 4686
48
First results

• Galactic sample analysis
• 12 early type stars
• previously analyzed by eye
• Spectroscopic mass
• g, R ? Mspec
• Evolutionary mass
• Teff, L, evolutionary track ? Mevol

49
Vz Stars
50
In conclusion

• The method works
• 12 object galactic sample fitted
• good agreement with earlier analyses
• improved determination of fundamental parameters
• Currently applied to a sample of 100 stars
• VLT Flames survey