1. Massive star feedback – from the first stars to the present Jorick Vink (Keele University)

2. Outline Why predict Mass-loss rates? (as a function of Z) Monte Carlo Method Results OB, B[e], LBV & WR winds Cosmological implications? Look into the Future

3. Why predict Mdot ? Energy & Momentum input into ISM

4. Massive star feedback NGC 3603

5. Why predict Mdot ? Energy & Momentum input into ISM

6. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution

7. Evolution of a Massive Star O B[e]

8. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution –Explosions: SN, GRBs

9. Progenitor for Collapsar model Rapidly rotating Hydrogen-free star (Wolf-Rayet star) But…… Woosley (1993)

10. Progenitor for Collapsar model Rapidly rotating Hydrogen-free star (Wolf-Rayet star) But…… Stars have winds… Woosley (1993)

11. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution –Explosions: SN, GRBs –Final product: Neutron star, Black hole

12. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution –Explosions: SN, GRBs –Final product: Neutron star, Black hole –X-ray populations in galaxies

13. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution

14. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution Stellar Spectra

15. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution Stellar Spectra –Analyses of starbursts

16. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution Stellar Spectra –Analyses of starbursts –Ionizing Fluxes

17. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution Stellar Spectra

18. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution Stellar Spectra Stellar “Cosmology”

19. From Scientific American

20. The First Stars Credit: V. Bromm

21. The Final products of Pop III stars (Heger et al. 2003)

22. From Scientific American

23. Why predict Mdot ? Energy & Momentum input into ISM Stellar Evolution Stellar spectra “Stellar cosmology”

24. Observations of the first stars

25. Goal: quantifying mass loss a function of Z (and z) What do we know at solar Z ?

26. Radiation-driven wind by Lines dM/dt = f (Z, L, M, Teff) STAR Fe Lucy & Solomon (1970) Castor, Abbott & Klein (1975) = CAK Wind

27. Radiation-driven wind by Lines dM/dt = f (Z, L, M, Teff) Abbott & Lucy (1985)

28. Momentum problem in O star winds A systematic discrepancy

29. Monte Carlo approach

30. Approach: Assume a velocity law Compute model atmosphere, ionization stratification, level populations Monte Carlo to compute radiative force

31. Mass loss parameter study

32. Monte Carlo Mass loss comparison No systematic discrepancy anymore ! (Vink et al. 2000)

33. Lamers et al. (1995) Crowther et al. (2006)

34. Monte Carlo Mass-loss rates  dM/dt increases by factor 3-5 (Vink et al. 1999)

35. The bi-stability Jump HOT Fe IV low dM/dt high Vinf Low density COOL Fe III high dM/dt low Vinf High density

36. Stars should pass the bistable limit During evolution from O  B LBVs on timescales of years

37. LBVs in the HRD Smith, Vink & de Koter (2004)

38. The mass loss of LBVs Stahl et al. (2001) Vink & de Koter (2002) Data Models

39. Stars should pass the bistable limit During evolution from O  B LBVs on timescales of years Implications for circumstellar medium (CSM) Mass-loss rate up ~ 2 wind velocity down ~ 2 CSM density variations ~ 4

40. SN-CSM interaction  radio Weiler et al. (2002)

41. Mass Loss Results from Radio SNe OB star? WR?

42. SN 2001ig & 2003bg Soderberg et al. (2006) 2003bg 2001ig Ryder et al. (2004)

43. Progenitors AGB star Binary WR system WR star LBV

44. Progenitors AGB star Binary WR system WR star LBV Kotak & Vink (2006)

45. Assumptions in line-force models Stationary One fluid Spherical

46. Polarimetry – from disks

47. Depolarisation

48. Asphericity in LBV: HR CAR (Davies, Oudmaijer & Vink 2005) SURVEY: asphericity found in 50%

49. Variable polarization in AG CAR (Davies, Oudmaijer & Vink 2005)  RANDOM: CLUMPS!!

50. Assumptions in line-force models Stationary One fluid Spherical Homogeneous, no clumps

51. Success of Monte Carlo at solar Z O-star Mass loss rates Prediction of the bi-stability jump Mass loss behaviour of LBVs like AG Car  Monte Carlo mass-loss used in stellar models in Galaxy

52. O star mass-loss Z-dependence (Vink et al. 2001)

53. O star mass-loss Z-dependence Kudritzki (2002) --- Vink et al. (2001)

54. O star mass-loss Z-dependence

55. Which metals are important? At lower Z : Fe  CNO solar Z low Z Fe CNO H,He Vink et al. (2001)

56. WR stars produce Carbon ! Geneva models (Maeder & Meynet 1987)

57. WR stars produce Carbon ! Geneva models (Maeder & Meynet 1987)

58. Which element drives WR winds? -C  WR mass loss not Z(Fe)-dependent -Fe  WR mass loss depends on Z host

59. Z-dependence of WR winds Vink & de Koter (2005, A&A 442, 587) WC WN

60. Corollary of lower WR mass loss:  less angular momentum loss  favouring the collapse of WR stars to produce GRBs  Long-duration GRBs favoured at low Z

61. Conclusions Successful MC Models at solar Z O star winds are Z-dependent (Fe) WR winds are Z-dependent (Fe)  GRBs Low-Z WC models: flattening of this dependence Below log(Z/Zsun) = -3  “Plateau”  Mass loss may play a role in early Universe

62. Future Work Solving momentum equation Wind Clumping Compute Mdot close to Eddington limit

63. Mass loss & Eddington Limit Vink (2006) - astro-ph/0511048 ~ Gamma^5

64. Future Work Solving momentum equation Wind Clumping Compute Mdot close to Eddington limit Compute Mdot at subsolar and Z = 0

65. From Scientific American

67. Non-consistent velocity law Beta = 1 WC8

68. Wind momenta at low Z Vink et al. (2001) Mokiem et al. (2007) Models (Vink) Data (Mokiem)

69. Two O-star approaches 1. CAK-type  Line force approximated  v(r) predicted CAK, Pauldrach (1986), Kudritzki (2002) 2. Monte Carlo  V(r) adopted  Line force computed – for all radii  multiple scatterings included Abbott & Lucy (1985) Vink, de Koter & Lamers (2000,2001)

70. Advantages of our method Non-LTE Unified treatment (no core-halo) Monte Carlo line force at all radii Multiple scatterings  O stars at solar Z & low Z LBV variability & WR as a function of Z

71. The bi-stability Jump HOT Fe IV low dM/dt high V(inf) Low density COOL Fe III dM/dt = 5 dM/dt HOT V(inf) = ½ vinf HOT High density = 10 HOT

72. The reason for the bi-stability jump Temperature drops  Fe recombines from Fe IV to Fe III  Line force increases  dM/dt up  density up  V(inf) drops  “Runaway”

73. Quantifying the effect of the velocity law

74. Can we use our approach for WR stars? Potential problems: –Are these winds radiatively driven? –Is Beta = 1 a good velocity law? –Do we miss any relevant opacities? –What about wind clumping?

75. B Supergiants Wind-Momenta Vink, de Koter & Lamers (2000)

76. New Developments: Hot Iron Bump Fe X --- Fe XVI Graefener & Hamann (2005) can “drive” a WC5 star self-consistently  Use Monte Carlo approach for a differential study of Mass loss versus Z

77. The bi-stability jump at B1 Lamers et al. (1995) Pauldrach & Puls (1990)