Caltech 1 ­ 3 February 2017
The Galactic Renaissance:
A Symposium in Honor of Judy Cohen
Impact of non­LTE
on determinations of atmospheric parameters and chemical abundances of VMP stars
Lyudmila Mashonkina
(Institute of Astronomy, RAS, Russia)
P. Jablonka, P. North (EPFL, Switzerland)
Y. Pakhomov, T. Sitnova (INASAN, Russia)
What is meant by non­local thermodynamic equilibrium (NLTE)?
 Do not use Saha­Boltzmann equations,
 atomic level populations ni from balance between
population and de­population processes
(statistical equilibrium, SE),
 Maxwellian velocity distribution, with Te = TA = Ti ,
 solution of coupled SE and radiative transfer equations: ∑ n j ( R ji +C ji ) =ni ∑ ( Rij +C ij )
+
j ≠i
i= 1, .. . NL
j ≠i
Rij(Jν ) ­ radiative, Cij(T,N ) ­ collisional rates
Take care of
• completeness of
model atom,
• accuracy of
atomic data !
NLTE effects for different chemical species
Na, Mg, Ca:
abundance differences
between NLTE and LTE
for VMP giants.
NLTE effects have different
magnitude and sign
for different species
in given atmosphere.
NLTE is important for study
of element abundance pattern
of given star.
Mashonkina+2017 (in prep)
NLTE effects, Fe I
6000/4.0
4500/1.0
Lines of Fe I
[Fe/H]
-3 (●)
-2 (△ )
0 ( □)
Lines of Fe I
(Mashonkina et al. 2016, AstL., 42, 606)
Non­LTE abundance corrections
∆NLTE = log εNLTE - log εLTE
depend on stellar parameters.
NLTE is important for study
of stellar samples covering
broad Fe range.
Homogeneous set of atmospheric parameters and
non­LTE abundances was determined for two stellar samples
51 dwarfs (subgiants),
­2.6 ≤ [Fe/H] ≤ 0.2,
Hamilton/Shane spectra:
R ≈ 60 000, S/N > 100, 3700­9300 Å
Sitnova, Zhao+2015 (ApJ, 808, 148)
Zhao, Mashonkina+2016 (ApJ, 833, 255)
23 giants,
­4 < [Fe/H] < ­1.7,
selected from Cohen+2013 (ApJ, 778, 56),
Burris+2000 (ApJ, 544, 302),
our previous studies.
Spectra from archives:
R > 40 000, S/N > 100, Milky Way comparison sample
for study of VMP stars
in dSphs
Mashonkina, Jablonka+2017 (in prep)
Effective temperatures
✷ Dwarfs: IRFM method (Alonso+1996, Casagrande+2011)
✷ Giants
V­I, V­J, V­K colours (12 stars from Cohen+2013  CCT13),
V­J, V­H, V­K colours (11 remaining stars).
Surface gravities
Non­LTE analysis of Fe I/Fe II, using model atom from Mashonkina+2011
 Fe I: overionisation, weakened lines, positive ∆NLTE
 Fe II: ∆NLTE < 0.01 dex for [Fe/H] > -4
Non­LTE leads to higher log g:
≤ 0.1 dex, [Fe/H] > -1.5,
> 0.5 dex, [Fe/H] < -2.5
Spectroscopic method was tested with two samples.
(i) 20 dwarfs with accurate πHip (σπ < 10%, d < 100 pc)
Collisions with H I:
Steenbock & Holweger (1984)
formula, scaling factor SH = 0.5
Δlog g(Gaia ­ Sp) = ­0.01±0.10
●
Gaia DR1, □ Hipparcos
Dwarf sample:
Fe I ­ Fe II
abundance differences
with final Teff / log g.
⚫ non-LTE, ○ LTE
(ii) VMP giants in dSphs with known distance:
Sculptor (11 stars), [Fe/H] = (­4, ­2.2)
 Ursa Minor (10), (­3.1, ­2)
 Fornax (1), ­3.4
 Sextans (2), ­2.6, ­2.8
 Boötes I (8), (­3.8, ­1.5)
 UMa II (3), ( ­3, ­2.3)
 Leo IV (1), ­2.6

­ photometric Teff (from literature or computed in this study)
­ log gd using (log g, Teff, Mbol, М = 0.8 Msun) relation,
Sources of spectra and/or published EWobs:
Cohen&Huang 2010, Frebel+2010, 2016, Gilmore+2013, Jablonka+2015,
Kirby & Cohen (2012), Norris+2010, Simon+2010, 2015, Tafelmeyer+2010,
Ural+2015
8
VMP stars in dSphs: Fe I ­ Fe II in NLTE and LTE
(filled and open symbols)
OK !
[Fe/H] ≿ ­3.7: Fe I/Fe II is fulfilled in NLTE (SH = 0.5)
log gsp was derived for 22 halo giants with [Fe/H] ≥ ­3.45,
HE1357­0123 ([Fe/H] = ­3.9): log g(CCT13) + 0.2
9
Fe I/Fe II imbalance at [Fe/H] ≾ ­3.7 ?
4800/1.56/-4.0
 Fe I, Fe II
Fe I: do not use Eexc < 1.2 eV
due to possible 3D effects.
SH = 0.5
Scl07-50, Fe I - Fe II =
0.09 dex (LTE),
0.36 dex (NLTE)
Fe II 4923, 5018 Å can only be detected, EW > 30 mÅ.
Uncertainty in log gf ? -1.26 (MB09), -1.39 (RU+0.11), -1.32 (VALD)
✓
✓
Downward revision of Teff would help, in part:
ΔTeff = -200 K ⇒ 4600/1.49/-4 ⇒ Fe I - Fe II = 0.16±0.24
Can we derive Teff /log g of the UMP stars reliably?
- Photometry: the calibration at [Fe/H] < -4 ?
- Spectroscopy: poor in Fe I (Eexc > 2 eV) and Fe II lines.
10
Checking derived atmospheric parameters with Ti I/Ti II
following NLTE method by Sitnova+2016, SH = 1
MW
filled / open
symbols
show
NLTE / LTE
OK ! [Fe/H] ≿ ­3.2: Ti I/Ti II supports Teff/log g

[Fe/H] ≾ ­3.2: too high Teff ?
poor in Ti I lines ?
rough treatment of Ti+H collisions ?
11
Advanced treatment of Ti + H I collisions:
Ti(n) + H(n=1) ↔ Ti+ + H‾
Ti(n) + H(n=1) ↔ Ti(n') + H(n=1)
applying Quantum Fitting Method (QFM, Ezzeddine+, 1612.09302)
Ti I
Ti II
NLTE abundance corrections for Ti I and Ti II:
based on QFM (filled symbols) and Drawin (1969, open symbols)
QFM does not remove Ti I/Ti II imbalance
12
Checking atmospheric parameters with YY evolutionary
tracks
М = 0.75 Msun
Thick disc () and halo () TO and subgiant stars
sit well on their evolutionary
tracks [Fe/H] = -2.75, …, -0.75
VMP giants in dSphs and MW
 Teff > 4600 K: OK !?
М = 0.8 Msun
 Cooler stars sit on [Fe/H] = ­2 track.
⊙ log gGaia
outlier: HD 8724 (4560/1.29/-1.76)
13
Non­LTE abundances
Dwarf sample
17 non-LTE species:
Li I, C I, O I, Na I, Mg I,
Al I, Si I, K I, Ca I, Sc II,
Ti II, Fe I-II, Cu I, Sr II,
Zr II, Ba II, Eu II
(Zhao, Mashonkina+2016)
VMP giant sample
MW and dSphs
9 non-LTE species:
Na I, Mg I, Al I, Si I, Ca I,
Ti I-II, Fe I-II, Sr II, Ba II
(Mashonkina, Jablonka+2017,
in prep)
Model atmospheres: MARCS (Gustafsson+2008)
14
α/Fe trends in Milky Way
[Mg/Fe]
[Si/Fe]
•dwarfs
•giants
[Ca/Fe]
[Ti/Fe]
Similar MP plateau for [Mg/Fe], [Si/Fe], [Ca/Fe], [Ti/Fe]
dwarfs, [Fe/H] < ­0.9: 0.29±0.07 0.32±0.07 0.33±0.07 0.30±0.05
VMP giants: 0.36±0.13 scatter 0.36±0.11 0.28±0.10
15
α/Fe trends in dSphs, NLTE versus LTE
Scl, UMi, Sex, Fnx dSphs and UMa II UFD follow the MW trends.
11-1-4276
 NLTE makes Ca following
Mg, Ti in given galaxy.
ET0381
UMi NLTE LTE
[Mg/Fe] 0.30 0.28
[Ca/Fe] 0.24 0.08
[Ti/Fe] 0.32 0.31
41
 NLTE confirms outliers:
- Scl ET0381 (Fe-rich)
- Scl 11-1-4276 (Mg,Ca-poor)
16
α/Fe trends in dSphs
Boötes I: decline in [α/Fe]
(support Gilmore+2013)
11-1-4276
Boo-41: NLTE finds
equal and high abundances
from Ti I and Ti II.
ET0381
Leo IV­S1: solar α/Fe
41
17
Neutron­capture elements Sr, Ba in Milky Way
[Sr/Fe]
[Ba/Fe]
MW
•dwarfs large spread in [Sr/Fe] and [Ba/Fe] at [Fe/H] < ­2.5
•giants
In line with
Francois+2007 (LTE),
many later papers
18
Sr and Ba in dSphs
•Scl, •Umi, Fnx, ▪Sex, ▴BooI, ✭LeoIV •MW
[Ba/Fe]
[Sr/Fe]
Ba/Fe ratios of dSphs are close
to the Ba/Fe floor of MW halo.
MW stars can be even more
depleted in Sr than the dSphs.
MW halo: two channels of
n­capture element production.
⊙ [Sr/Fe] ≾ ­0.4, subsolar Sr/Ba,
r­process.
• [Sr/Fe] ≿ ­0.4,
upward trend of Sr/Ba. Source ?
dSphs: it takes a certain mass
to get Sr rich compared to Ba
19
Summary
NLTE is important for study of ­ element abundance patterns of given star,
­ elemental ratios of stellar sample covering broad Fe range.

Precise and accurate surface gravities can be derived
from Fe I/Fe II in the [Fe/H] ≿ ­3.7 domain,
using 1D­NLTE approach.

LTE: underestimates log g up to 0.5 dex and [Fe/H] up to 0.3 dex.
UMP stars: methods of Teff/log g determination need
to be further tested. 
Summary
α-elements
­ MW, [Fe/H] < -0.9: [Mg/Fe] ≅ [Si/Fe] ≅ [Ca/Fe] ≅ [Ti/Fe] ≅ 0.3.
­ Classical dSphs follow α/Fe trend of MW in VMP regime.
­ Boötes I: decline in α/Fe (Gilmore+2013) is confirmed.


n­capture elements
­ MW halo, two production channels:
r­process  [Sr/Fe] ≾ ­0.4, subsolar Sr/Ba,
?  [Sr/Fe] ≿ ­0.4, upward trend of Sr/Ba. ­ dSphs, it takes a certain mass to follow MW.
Scl, UMi: similar two channels,
Boötes I: r­process.
For practical use
Fe I, Ti I-Ti II, Ca I lines in
4000 K ≤ Teff ≤ 5000 K, 0.5 ≤ log g ≤ 2.5, -4 ≤ [Fe/H] ≤ 0 domain
∆NLTE online for given Teff/log g/[Fe/H]:
(Mashonkina et al. 2016, AstL., 42, 606)
http://spectrum.inasan.ru/nLTE/
Complementary slides
4670/1.13/-3.7
SH = 0.5
slope: -0.024 dex/eV
Absolute abundances
from lines of Fe I ()
and Fe II ()
Fe I: do not use
Eexc < 1.2 eV, EW > 120 mÅ.
Fe I - Fe II = 0.01 dex (LTE), 0.30 dex (non-LTE)
- Fe II 4923, 5018 Å can only be detected, EW > 53 mÅ.
Uncertainty in log gf ? -1.26 (MB09), -1.39 (RU+0.11), -1.32 (VALD)
- Too high Teff ? ΔTeff = -170 K ⇒ 4500/1.07/-3.7
4500/1.07/-3.7
SH = 0.5
Fe I - Fe II = 0.07±0.16
slope: 0.002 dex/eV
24
Impact of non­LTE on Mg I and Ca I
­ UV overionisation ⇒ ∆NLTE > 0 for weak lines, ­ photon loss in strong lines at τ0 < 1 ⇒ ∆NLTE can be < 0
for Mg Ib, Mg I 5528, Ca I 4226 !
25
Ca abundances of [Fe/H] < ­3.5 stars
 Only Ca I 4226 Å was measured in the most MP Scl stars
07-50 (-3.9), 11_1_4296 (-3.7), 6_6_402 (-3.66).
 Ca I 4226 Å: lower abundance compared with Ca I subordinate lines.
Scl031-11 (4670/1.13/-3.6): ∆ = -0.65 dex, LTE, 3 subordinate lines,
-0.87 dex, NLTE
4226 A
• Strong overionisation in deep layers
⇒ 4226 Å wings are weakened.
• Photon loss in line itself at log τ < -1
⇒ 4226 Å core is strengthened.
Net: ∆(4226 Å) < ∆(subordinate lines)
Departure coefficients of Ca I levels.
Resonance transition is 4s-4p.
For Scl 07-50, 11_1_4296
use Ca II 3933 Å.
NLTE ∆(4226 - 3933) = -0.59 dex
and -0.27
26dex.
Impact of non­LTE on Sr II and Ba II
Lines of Sr II, Ba II can be strengthened or weakened
depending on log g, Teff , element abundance Abundance differences
between NLTE and
LTE for VMP giants
27