1. Martin Asplund, Paul Barklem, Andrey Belyaev, Maria Bergemann,
3D and NLTE analysis for large stellar surveys
Karin Lind
Uppsala University, Sweden
Martin Asplund, Paul Barklem, Andrey Belyaev, Maria Bergemann,
Remo Collet, Zazralt Magic, Anna Marino, Jorge Meléndez, Yeisson Osorio

2. Outline Introduction 1D LTE/NLTE Worst-case scenarios Recent progress
Calibration techniques
Practical implementation
Applications
3D LTE/NLTE
Observational tests
Mg : 1D/<3D>/LTE/NLTE
Ca : 1D/<3D>/3D/LTE/NLTE

3. Motivation Galactic archaeology by chemical tagging of FGK stars
Statistics : Soon > 106 stars
Precision (S/N, wavelength range) :
σ[X/H] < 0.1dex, σTeff<150K, σlog(g)<0.3dex
Accuracy (assumptions: 1D, LTE, atomic data) :
σ [X/H]< 0.5 dex, σTeff<400K, σlog(g)< 1 dex

4. Methods Model atmosphere Detailed rad. Transfer
1D/<3D>/3D LTE D/3D LTE/NLTE
R. Collet

5. NLTE line formation

6. (1D) N-
Is it really necessary?
Is it safe?

7. Worst-case scenario I NaD lines in metal-poor horisontal branch stars
Lind et al. 2011, Marino et al. 2011
B-I

8. Worst-case scenario II
OI 777nm triplet at very low metallicities
LTE trend
Fabbian et al. 2009

9. Input data for NLTE analysis
Energy levels + oscillator strengths + photo-ionization cross sections
Red boxes : have sufficient(?) data
Blue boxes : missing e.g. QM photo-ionisation, but NLTE still attempted

10. Input data for NLTE analysis
Blue boxes : QM hydrogen collisions exist or will exist

11. Input data for NLTE analysis
Most important free parameter in NLTE
modelling of Fe is FeI+HI collisional cross-section
Black – LTE
Blue – NLTE with no hydrogen collisions
Solar neighborhood MDF Halo MDF [X/Fe] vs [Fe/H]

12. Calibration techniques: ionisation balance
Korn et al. 2003
FeI/FeII ionisation equilibrium calibrated using Hipparcos gravities
 S(H)=3

13. Calibration techniques: excitation balance
Bergemann & Gehren 2008
“Thus, NLTE can solve the discrepancy between the abundances derived from the MnI resonance triplet at 403 nm and excited lines, which is found in analyses of metal-poor subdwarfs and subgiants”
 S(H)=0.05

14. Calibration techniques: CLV
Allende Prieto et al. (2004)
Solar centre-to-limb variation of OI lines

15. Practical implementation I
“Curves-of-growth” from UV-NIR:
3200 FeI lines
107 FeII lines
Teff=6500K
log(g)=4.0
ξ=2km/s
ΔNLTE
Lind et al. (2012)

16. Practical implementation II
Pre-computed departure coefficients  NLTE synthesis
T. Nordlander

17. FeI NLTE grid
Lind et al. (2012)

18. Application : metal-poor stars
LTE
NLTE
+PHOT
Ruchti et al. (2012)

19. Application : metal-poor stars
LTE
NLTE+PHOT
Serenelli et al. (2013)

20. 3D (LTE/NLTE)
Is it really necessary?
Is it safe?

21. Stagger grid
Magic et al. 2014

22. Abundance patterns 3D N-LTE Keller et al. (2014) Dashed –200 Msun PISN
Solid – 60Msun fallback 3D
N-LTE

23. Worst-case scenario III
Li isotopic abundances 3D
N-LTE
Lind et al. 2013
Asplund et al. 2006

24. Observational tests: the Sun
Pereira et al. 2013
“We confronted the models with observational diagnostics of the [solar] temperature profile: continuum centre-to-limb variations (CLVs), absolute continuum fluxes, and the wings of hydrogen lines. We also tested the 3D models for the intensity distribution of the granulation and spectral line shapes. ”
“We conclude that the 3D hydrodynamical model is superior to any of the tested 1D models.”

25. Observational tests: low [Fe/H]
Klevas et al. 2013
FeI line assymmetries
in the metal-poor
giant HD122563

26. 1.5/3D + NLTE LiI : Asplund et al. 2003, Sbordone et al. 2010
OI, FeI : Shchukina et al. 2005
OI : Pereira et al. 2010, Prakapavičius et al. 2013
LiI, NaI, CaI : Lind et al. 2013

27. Ways forward Model LTE/NLTE Time Performance 1D LTE Seconds NLTE
Minutes (seconds using interpolation) 3D
Hours
Days
The ultimate goal, reference point
<3D>

28. Mg b in a VMP SG 1D LTE 1D NLTE <3D> LTE <3D> NLTE
“No” free parameters!
HD140283
Teff=5780K
log(g)=3.7
[Fe/H]=-2.4
1D LTE
1D NLTE
<3D> LTE
<3D> NLTE
Yeisson Osorio

29. Ca in a VMP dwarf LTE 1D 3D <3D> NLTE HD19445 Teff=6000K
log(g)=4.5
[Fe/H]=-2.0

30. Ca in a VMP dwarf LTE 1D 3D <3D> NLTE HD19445 Teff=6000K
log(g)=4.5
[Fe/H]=-2.0

31. Ca in a EMP TO Start ? Goal G64-12 Teff=6430K log(g)=4.0 [Fe/H]=-3.0
Bullets: Optical CaI lines
Squares: NIR CaII triplet

32. Ca in a EMP TO Start ? Goal Bullets: Optical CaI lines
Squares: NIR CaII triplet

33. Ca in a EMP TO Start ? Goal Bullets: Optical CaI lines
Squares: NIR CaII triplet

34. Ca in a EMP TO Start ? Goal Bullets: Optical CaI lines
Squares: NIR CaII triplet

35. Ca in a EMP TO Start ? Goal Bullets: Optical CaI lines
Squares: NIR CaII triplet

36. Ways forward A : NLTE-sensitive, B : not NLTE-sensitive Model LTE/NLTE
Time
Performance 1D
LTE
Seconds
Varied
NLTE
Minutes (seconds using interpolation)
Improves for A
No change for B 3D
Hours
May worsen for A
Improves for B
Days
The ultimate goal, reference point
<3D>