1. Constraints on the Velocity Distribution in the NGC 3783 Warm Absorber T. Kallman, Laboratory for X-ray Astrophysics, NASA/GSFC ●

2. Kaspi et al 2002  900 ksec Chandra HETG observation of NGC3783  >100 absorption features  blueshifted, v~800 km/s  broadened,  vturb~300 km/s  emission in some components  fit to 2 photoionization model components  Piecewise continuum  Not a full global fit

3. Kaspi et al 2002  900 ksec Chandra HETG observation of NGC3783  >100 absorption features  blueshifted, v~800 km/s  broadened,  vturb~300 km/s  emission in some components  fit to 2 photoionization model components  Piecewise continuum  Not a full global fit

4. Krongold et al 2004  >100 absorption features  blueshifted, v~800 km/s  broadened,  vturb~300 km/s  emission in some components  fit to 2 photoionization model components  Fe M shell UTA fitted using Gaussian approximation  Full global model

5. Krongold et al 2004  >100 absorption features  blueshifted, v~800 km/s  broadened,  vturb~300 km/s  emission in some components  fit to 2 photoionization model components  Fe M shell UTA fitted using Gaussian approximation  Full global model

6. Krongold et al 2004  fit to 2 photoionization model components  Ionization parameter and temperature are consistent with coexistence in the same physical region  Disfavored existence of intermediate ionization gas due to shape of Fe M shell UTA  But used simplified atomic model for UTA Curve of thermal equilibrium for photoionized gas  Flux/pressure

7. Chelouche and Netzer 2005 Combined model for dynamics and spectrum Assumes ballistic trajectories Favors clumped wind

8. Blustin et al. (2004)  Fitted the XMM RGS spectrum using global model  Also find evidence for two components  Omit Ca  Include line-by-line treatment of M shell UTA, but still miss some  Claim evidence for higher ionization parameter material  require large overabundance of iron

9.  Work so far on fitting warm absorber spectra has concentrated on the assumption of a small number of discrete components  This places important constraints on the flow dynamics, if it is true  There is no obvious a priori reason why outflows should favor a small number or range of physical conditions  In this talk I will test models in which the ionization distribution is continuous rather than discrete, and discuss something about what it means  Previous tests of this have invoked simplified models for the Fe M shell UTA which may affect the result

10. Proga and Kallman 2004 Disk winds have smooth density distributions on the scales which can be Calculated…

11.  Work so far on fitting warm absorber spectra has concentrated on the assumption of a small number of discrete components  This places important constraints on the flow dynamics, if it is true  It is not obvious why outflows should favor a small number or range of physical conditions  I will examine the robustness of this result and discuss some implications Outline

12.  Lines are not symmetric  blue wing is more extended in lines from ions which occur at smaller ionization parameter (eg. O VIII L  vs. S XVI)  Suggests that higher velocity material is less ionized (Ramirez et al. 2005) I will not discuss details about Line Profile Shapes, widths and blueshifts

13. 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7  2 ~64905/8192 As a start, fit to a continuum plus Gaussian absorption lines. Choose a continuum consisting of a double power law plus cold absorption (  1 =1.5,  2 =3, log(N H )=22.5; This is a cheat because it already takes into account some of the absorbtion, but does not strongly affect the results)

14. 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7  2 ~64602/8192 Absorption lines are placed randomly and strength and width adjusted to improve the fit.

15. 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7  2 ~59716/8192 Absorption lines are placed randomly and strength and width adjusted to improve the fit.

16. 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7  2 ~51773/8192 Absorption lines are placed randomly and strength and width adjusted to improve the fit.

17. 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7  2 ~43787/8192 Absorption lines are placed randomly and strength and width adjusted to improve the fit.

18. 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7  2 ~28526/8192 Absorption lines are placed randomly and strength and width adjusted to improve the fit.

19. 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7  2 ~16070/8192 Absorption lines are placed randomly and strength and width adjusted to improve the fit.

20.  2 ~9915/8192 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7 Absorption lines are placed randomly and strength and width adjusted to improve the fit.

21.  requires ~950 lines, Ids for ~100   2 ~9915/8192  300 km/s<v/c<2000  Represents best achievable  2 (by me)  Allows line Ids  Shows distribution of line widths, offsets Results of notch model:

22.  requires ~950 lines, Ids for ~100   2 ~9915/8192  300 km/s<v/c<2000  Represents best achievable  2 (by me)  Allows line Ids  Shows distribution of line widths, offsets Ionization parameter of maximum ion abundance vs. line wavelength for identified lines

23. Favored region

24. Photoionization Models  Full global model  Pure absorption (maybe misses something important?)  Single velocity component  Xstar version 2.1l  Inner M shell 2-3 UTAs (FAC; Gu); >400 lines explicitly calculated  Chianti v. 5 data for iron L  Iron K shell data from R matrix (Bautista, Palmeri, Mendoza et al)  Available from xstar website, as are ready-made tables  Not in current release version, 2.1kn4  Xspec ‘analytic model’ warmabs  Not fully self consistent: assumes uniform ionization absorber, but this is small error for low columns.  Xspec11 only at present, not quite ready for prime time

25. Comparison of model properties X* release (2.1kn4)X* beta (2.1l)warmabsothers Xspec interfacetables analyticvarious Atomic dataKB01K04, chianti5 various ‘real’ slabyynvarious Energy resolved??yvarious Self-consistent SEDyyn? nlteyyy? radiative transfer nnn? dynamicsnnnn

26. 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18 19 wavelength (A) single Component Fit, log  =2.4  2 ~27811/8192, voff=700 km/s vturb=300 km/s Missing lines near 16-17A

27. 2 Component Fit, log  =2.4, 0.1  2 ~18516/8192, voff=700 km/s vturb=300 km/s 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7 Better, Some discrepancies, Missing lines

30. Fitting absorption only is fraught, due to influence of scattering/reemission

31. Si VII-XI K lines

32. Al XIII Al XII

33. Fe XX-XXII

36. Fe XXII

37. Fe XXI

45. Comparison with previous work NetzerKrongoldBlustinMe Log(U)-0.6,-1.20.760.45 Log  1 ) 3.7,3.12.252.42.2 Log(N 1 )22.2 22.4521.4 Log(U)-2.4-0.78-1.55 Log  2 ) 0.690.720.30.2 Log(N 2 )21.921.620.7320.4

47. 2 component fit:

48. What if the distribution is continuous instead?

49.  2 ~20071/8192, voff=700 km/s vturb=300 km/s 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7 Continuous distribution, 0.1<log  2.4 Over-predicts the absorption Due to intermediate iron ions

50. Continuous distribution, 0.1<log  2.4

51. What happens if we iteratively search for the best fit, starting with a continuous distribution?

55. The best fit turns out to be the same 2 component fit…

56. What about extended power law distribution?

57.  2 ~21826/8192, v off =700 km/s v turb =300 km/s Power law distribution, extended to -0.8<log  <3.2 There can be some gas at Higher ionization parameter. Fit is slightly worse 3 4 4 5 5 6 6 7 7 8 8 9 9 10 10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18 19 wavelength (A)

58. A summary of  2 Notch11945 single27811 2 component18516 continuous20071 extended22871 best17445

59. What does this mean? Why do we preferentially see only a few components in  ? First, consider the most likely conditions… Radius (timescales scale the same with radius) v vir =  2GM/R) 1/2 ~300 km/s => R~10 17 v 300 -2 M 6 cm density  =L/nR 2 ~10 2 erg cm/s => n~ 10 8 L 44 R 17 -2  2 -1 cm -3 Physical thickness N=n  R ~10 22 cm -2 =>  R~10 14 n 8 -1 N 22 cm So  R/R <<1. Absorber is thin compared with radius

60. Now consider the timescales… Cooling timescale t cool =kT/(n ) ~10 4 n 8 -1 T 5 s Flow timescale t flow =  R /v ~10 6.5  R 14 v 300 s So it looks like thermal equilibrium should be OK. Sound crossing time t exp =  R /c s ~10 7.5  R 14 s Pressure equilibrium is questionable

61. If the gas is in photoionization equilibrium: temperature vs. 

62. Equilibrium temperature vs.  Ionization parameters don’t quite line when calculated this way…

63. Krongold et al 2004 Curve of thermal equilibrium for photoionized gas  Flux/pressure

64. Krongold et al. 2002 SEDs

65. Krongold et al. 2002 SEDs me Finding stable temperatures at observed ,T requires some fine tuning of the SED

66. Another way of think about this: cooling curve is multi- valued, and appears to correspond with derived ionization parameters If so, then the absence of higher  material is associated with rapid increase in cooling For  >3. Implies some non-photoionization heating. Mass loss estimates based on observed  and v would be valid

67. What if the gas is not in ionization equilibrium….  If the heating timescale is short compared with the flow time,then what we observe is gas in a transient state on its way to being fully ionized  The net heating rate increases steeply for  >3, so the high ionization state may represent conditions where the gas accumulates, but is not in equilibrium  This would require N 22 < 10 -2.5 T 5 v 300 -1  Tiny filaments?  What we see is superposition of many,  All in a transient state on their way to being highly ionized  Would occur if the dynamics do not provide enough density for equilibrium at T<<T IC Mass loss rate is then determined by the heating time, not the flow time

68. Contours of constant heating (black) and cooling (blue) timescale vs  and T for density 10 8 ~10 4 s shorter longer

69. Which is a better model: this

70. Or this?

71. Summary  In spite of a priori expectations, continuous distribution of ionization conditions is an inferior fit to (some) high quality X-ray spectra  Implies some special meaning for inferred ionization parameters:  equilibrium: favored locus in ( ,T). Requires fine tuning, but does not appear to work in detail, since inferred ,T is not stable  Observed ,T could be stable if some other source of heating is acting  non-equilibrium: high  is where the gas spends the most time before heating runs away toward Compton equilibrium. Requires very thin clouds

72. ● Extra slides

73. What does this mean?  Why would warm absorber gas favor certain ionization parameters/densities?  Thermal properties? But what is the mechanism?  Time dependence? If so, t cool > t flow  t cool ~n 2, t flow ~d/v  requires d 18 T 5 <10 -9 n 10 v 7. How?  equilibrium?  2 phase idea?  Is this the assymptotic state for the outflow?  Source variability?

74. Krongold et al 2005 ● Claim detection of variability in UTA 10  ● No variability in high ionization component ● Claim that this implies clumping of low ionization component, since in smooth flow flux change would produce change in location of absn, not strength or ionizaation. ● But this only applies if gradient is small. This result would be better phrased as implying that gradient is large compared with change in location of ionization surfaces.

75. 2 component fit: What about a continuous distribution?

77. What about extended power law distribution?

78. Result: We end up with the same 2 components

79. Now let’s see what happens when we iteratively fit for the ionization parameter distribution. Trial:

81.  2 ~20071/8192, voff=700 km/s vturb=300 km/s 15 16 16 17 17 18 18 19 wavelength (A) 11 12 12 13 13 14 14 15 7 8 8 9 9 10 10 11 3 4 4 5 5 6 6 7 Continuous distribution, 0.1<log  2.4 Over-predicts the absorption Due to intermediate iron ions

82. Another way of think about this: cooling curve is multi- valued, and appears to correspond with derived ionization parameters

83. Log(  )=1 1.9 2.8

84. Log(  )=1 1.9 2,8 3.7 4.6 Red=cooling rate Black=heating rate Another way of think about this: cooling curve is multi-valued, and appears to correspond with derived ionization parameters