You are on page 1of 21

Astronomy & Astrophysics manuscript no.

main
January 2, 2017

c
ESO
2017

The role of inelastic hydrogen collisions in NLTE iron abundance


determination:
Introducing the quantum fitting method (QFM)
R. Ezzeddine1,2,3 , B. Plez1 , T. Merle4 , M. Gebran5 , and F. Thvenin6
1
2
3

arXiv:1612.09302v1 [astro-ph.SR] 29 Dec 2016

4
5
6

Laboratoire Univers et Particules de Montpellier, Universit de Montpellier, CNRS, UMR 5299, Montpellier, France
Joint Institute for Nuclear Astrophysics, Center for the Evolution of the Elements, East Lansing, MI 48824, USA
Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
Institut dAstronomie et dAstrophysique, Universit Libre de Bruxelles, CP 226, Boulevard du Triomphe, 1050 Brussels, Belgium
Department of Physics and Astronomy, Notre Dame University-Louaize, PO Box 72, Zouk Mikal, Lebanon
Universit Cte dAzur, Observatoire de la Cte dAzur, CNRS, UMR7293, CS 34229, F-06304 Nice Cedex 4

Received / Accepted
ABSTRACT
Context. Determination of high precision abundances has and will always be an important goal of all spectroscopic studies. Iron

abundance, especially, plays a vital role as it is a pre-requisite to any chemical abundance analysis.
Aims. We investigate the role of hydrogen collisions in NLTE spectral line synthesis, by introducing a new general recipe to determine
inelastic charge transfer and bound-bound hydrogen collisional rates, based on fitting the quantum rates of several elements (Na i, Mg i,
Al i and Si i). We call this recipe the Quantum Fitting Method (QFM).
Methods. A new complete iron model atom was developed for the purpose of testing this method. It was first tested using a silicon
model atom and then applied on iron.
Results. Fe i and Fe ii line-by-line abundances were calculated for a sample of 24 benchmark stars of well-determined nonspectroscopic fundamental atmospheric parameters. Our results demonstrate that the QFM can be applicable in NLTE calculations
when detailed quantum rates are not available. Our NLTE iron calculations show that inelastic hydrogen charge transfer collisional
processes dominate over excitation, especially for metal-poor stars. Fe i and Fe ii ionization equilibrium was secured for most stars,
however attaining excitation equilibrium conditions predict lower effective temperatures than non-spectroscopic values for the metalpoor stars. Our results are also in good agreement with recent 1D and 3D NLTE analysis of metal-poor stars using unpublished
quantum data.
Conclusions. Our proposed quantum fitting method for estimating hydrogen collision rates is shown to be useful in NLTE calculations,
and in determining high accuracy iron abundances. Most importantly, it can furthermore be extended to other elements where quantum
calculations do not exist, and replace the unreliable classical widely used Drawin approximation.
Key words. atomic processes - line: formation - stars: abundances - stars: atmospheres - stars: late-type

1. Introduction
In the era of large-scale surveys tracing the history of stellar populations (e.g., RAVE, Steinmetz et al. 2006; APOGEE, Allende
Prieto et al. 2008; Gaia-ESO, Gilmore et al. 2012), there is an
important need for high accuracy detailed chemical composition
determinations.
Iron is a particularly important case as its numerous lines are
prominent in many stellar spectra, and the iron ionization equilibrium is often used to spectroscopically determine the surface
gravity of cool stars. Iron is a usual proxy for the metallicity, and
it is used as a reference in galactic chemical evolution relative to
which other elemental abundances are compared.
Most stellar spectroscopic analyses adopt the assumption of
Local Thermodynamic Equilibrium (LTE). This assumption is
however not always valid, especially when collisions are not
frequent enough to ensure equilibrium excitation and ionization
populations. Collisions, in cool stellar atmospheres, are mainly
due to electrons and hydrogen atoms. Electrons have larger thermal velocities than hydrogen atoms by a factor of order 43, and
unless they are less numerous than hydrogen atoms, they are

believed to be responsible for most collisions (Lambert 1993).


In low-metallicity stars, electron densities are lower and NLTE
effects are expected to grow stronger, as well as in giants and
supergiants rarefied atmospheres (Thvenin & Idiart 1999; Asplund 2005; Lind et al. 2011). In F, G, and K-type stars, iron
is mostly ionized, and small departures from ionization equilibrium may lead to large variations of Fe i lines, and hence of the
derived Fe i abundances, while mostly not affecting the Fe ii lines
(Bergemann et al. 2012; Mashonkina et al. 2016).
NLTE modeling requires a good knowledge of a bulk of
atomic data for each element under consideration. Large experimental and theoretical efforts have been devoted to measure and
calculate, e.g., energy levels, g f -values, broadening parameters
(e.g. Anstee & OMara 1995; Castelli & Kurucz 2010; Peterson & Kurucz 2015; Barklem et al. 2015). Hydrogen collision
rates remain, however, a major source of uncertainty in NLTE
calculations for cool stars (Asplund 2005; Barklem et al. 2010).
Quantum mechanical collision rates have been computed only
for a small number of atoms (Li, Belyaev & Barklem 2003; Na,
Barklem et al. 2010; Mg, Belyaev et al. 2012; Al, Belyaev 2013;
Article number, page 1 of 21

A&A proofs: manuscript no. main

Si, Belyaev et al. 2014 and Ca, Barklem 2016; Belyaev et al.
2016) over the past decade.
No quantum calculations have been published yet for iron.
In the lack of these calculations, the Drawin (Drawin 1968,
1969a,b; Lambert 1993) approximation is customarily used to
estimate the rates. This approximation is originally derived from
the classical Thomson (1912) e + atom ionization rate equation.
The Drawin approximation has been applied to allowed boundbound (bb) and ionization bound-free (b f ) transitions, where it
has been shown to overestimate the collisional rates by orders of
magnitude (Barklem et al. 2010). A fudge-factor, SH , has been
applied in most studies to try to correct for the inadequacies of
the Drawin approximation. SH -factors between 0.1 and 3 have
been proposed for Fe i (e.g. Gehren et al. 2001; Korn et al. 2003;
Mashonkina et al. 2011; Bergemann et al. 2012), showing the inadequacy of this description further discussed in Barklem et al.
(2011).
Recently, charge transfer rates, corresponding to the ion-pair
production and mutual-neutralization processes: A + H
A+
+ H , have been shown to be larger than the bb-rates (Barklem
et al. 2010). They thus play an important, and in some cases
even dominating, role in the collisional processes (Lind et al.
2011; Osorio et al. 2015; Guitou et al. 2015). They must thus be
included in NLTE calculations.
In Ezzeddine et al. (2016), we used the Drawin approximation to estimate charge transfer rates, in the absence of any other
approximation. We defined two scaling-factors SH for bb and
b f , and SH (CT) for charge transfer rates, and tried to calibrate
them using benchmark stars spectra ( Cen A, HD 140283 and
the Sun). Our method used a 2 -minimization of differences between observed and calculated iron lines equivalent widths for
different sets of SH and SH (CT). We could not find a combination
of SH and SH (CT) that would work for the three stars. In previous
studies, not including charge transfer rates, the SH value was also
found to be star dependent (Mashonkina et al. 2011; Bergemann
et al. 2012).
This inability of the Drawin approximation to properly reproduce the behavior and magnitude of hydrogen collision rates
calls for another approach, as long as no quantum calculation
are made available. This is the motivation for our work, which
introduces a semi-empirical method to estimate the hydrogen
collision rates for both bb and charge transfer processes. It is
based on the observation that the few well determined quantum
rates all behave similarly, and can be described with simple
fitting functions. We test the method on Si and then use it to
estimate H-collisional rates for iron. Finally we use our best
estimate of the collisional rates to determine iron abundances for
a sample of 24 benchmark stars. We hereafter call this method
the "Quantum Fitting Method" (QFM).
This article is divided as follows: In Sec. 2, we introduce the
Quantum Fitting Method and its application to estimate the hydrogen collision rates for both charge transfer and de-excitation
rates. The method is first tested on Si in Sec. 3, and then applied
to Fe in Sec. 4, with a specially developed iron model atom, and a
set of benchmark stars with well determined parameters. Fe i and
Fe ii abundance determinations for the benchmark stars using the
QFM, and comparison to previous work are detailed in Sec. 5.
Finally, conclusions are presented in Sec. 6.

2. The Quantum Fitting Method


Our semi-empirical Quantum Fitting Method to estimate the
hydrogen collisional rates for both charge transfer and (deArticle number, page 2 of 21

)excitation processes is motivated by the similar behavior of the


quantum-mechanical hydrogen collision rates for Na (Barklem
et al. 2010), Mg (Belyaev et al. 2012), Al (Belyaev 2013) and
Si (Belyaev et al. 2014) as a function of transition energies E.
The QFM is based on fitting these rates, as shown below.
2.1. Charge Transfer (CT) rates

We first plot in Fig. 1 the mutual-neutralization (downward


charge transfer) rate coefficients QCT = hvi1 of Na, Mg, Al and
Si from Barklem et al. (2010), Belyaev et al. (2012), Belyaev
(2013) and Belyaev et al. (2014) respectively. We only include
the rate at T = 6000 K, as the rates at other temperatures show
a similar behavior. We discuss the temperature dependence
further below. The rates for all four elements have a peak of
loghvi 7 cm3 s1 arising at 1/E between 0.5 and 1 eV1 .
Here E = Eup Elow is the energy difference between the
upper and the lower level of the transition. As the upper state
consists of H + Fe+ , its energy is the iron ionization potential
decreased by 0.754 eV, the H electron affinity.
We chose an inverse energy scale representation (see Fig. 1)
to allow simple linear fits of most of the data. We ignore rates
that lie 4 orders of magnitude below the peak. Such rates will
not contribute significantly. We are thus able to describe the CT
rate coefficients QCT of all four elements as a function of 1/E
with the following recipe:

log QCT

a1  E
=

a2 E

1
E0
1
E0




+ log Qmax

for

1
E

<

1
E0

+ log Qmax

for

1
E

>

1
E0

(1)

where a1 and a2 are respectively the positive and negative


slopes of the linear fits, and Qmax is the maximum rate at E1 0 .
The temperature dependence of the CT fitting parameters a1 ,
1
a2 , E
and log Qmax were also tested, showing a roughly linear dependence of a1 on T , while log Qmax depends linearly on
log T . a2 and E1 0 do not vary significantly with temperature.
The temperature dependencies for the case of Si, as an example,
are shown in the upper panels of Fig. 2. It follows that a1 and
log Qmax can be generally written as a function of temperature T
as:

 

a1 = a11 10T 4 + a10

Qmax = ( T 4 ) Q10000

(2)

10

where Q10000 = 10Q0 is a constant equal to Qmax at 10000 K.


a11 , a10 , and Q0 , as well as a2 and 1/E0 obtained for the four
elements are shown in Tab. 1. a11 and a10 vary within less than
20% from a11 ' 2.00 and 40% from a10 ' 5.00 respectively for
all four elements, except for Al where a11 = 7.46. This can be
attributed to Al having the least number of transitions between
the four elements in the fit. Similar values of 1/E0 ' 0.75 eV1 ,
' 0.34 0.03 and Q0 ' 7.5 0.4 are obtained for all
elements, except again for Al where = 0.56 and Mg where
1/E0 = 0.63 eV1 .
Overall, for T [3000, 7000], the temperature range of interest in cool star atmospheres, a1 varies within < 20% while
Qmax varies within 1 order of magnitude.
1

hvi is the Maxwellian-averaged hydrogen collision cross-section.

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

Fig. 1. Charge transfer rate coefficients at T = 6000K (in logarithmic scale) for Na, Mg, Al and Si (filled triangles) from the quantum calculations
of Barklem et al. (2010), Belyaev et al. (2012), Belyaev (2013) and Belyaev et al. (2014) respectively and their corresponding fits (dashed lines).
1
The rate coefficients hvi as a function of the inverse transition energies E
for all four atoms show the same pattern of behavior with energy
1
1
3 1
peaking around loghvi 7 cm s for E between 0.5 and 1 eV .

El
Na
Mg
Al
Si

a11
1.50
2.12
7.46
1.20

a10
7.92
7.36
3.74
4.90

a2
2.60
3.42
4.90
3.52

1
E0

0.72
0.63
0.74
0.75

0.37
0.35
0.56
0.34

Q0
7.17
7.34
7.43
7.47

b31
0.04
0.01
0.88
0.08

b30
0.08
0.05
3.67
0.23

b21
0.48
0.09
5.89
0.70

b20
0.09
0.21
25.04
2.05

b11
1.69
0.54
8.43
1.36

b10
1.02
0.67
33.95
3.51

b0
9.84
9.05
10.87
9.45

Table 1. Charge transfer and bb hydrogen collision rates fitting coefficients obtained for Na, Mg, Al and Si.

2.2. De-excitation (bb) rates

bb fitting parameters as:

The de-excitation (downward bb) rates for Na, Mg, Al and Si


show larger scatter than charge transfer rates as more transitions
are included, and a decreasing behavior with increasing transition energies, E, with large variations in magnitudes between
the maximum and minimum rates (up to 10 orders of magnitude).

b3

b2

b1

b0

= b31 log T + b30


= b21 log T + b20
= b11 log T + b10
= b0

(4)

The rates for all four elements can be best fit as a function of
E using a third degree polynomial:
where b30 , b31 , b20 , b21 , b10 , b11 and b0 for the four elements
are displayed in Tab. 1.
log Qbb = b3 E 3 + b2 E 2 + b1 E + b0

(3)

In addition, b3 , b2 and b1 are linearly dependent on log T ,


while b0 does not vary much with temperature as is demonstrated
for Si in the lower panels of Fig. 2. This allows us to write the

The polynomial fitting coefficients for the bb rates vary more


severely from one atom to another than for the CT coefficients,
which can be attributed to the large scatter in rates. Al fitting
coefficients again show larger differences from the other three
atoms. b0 , which is the scaling coefficient of log Qbb , varies between ' 9.0 and ' 11.0 for the four elements.
Article number, page 3 of 21

A&A proofs: manuscript no. main

Fig. 2. Temperature - rates fitting coefficients dependence of a1 , a2 ,


b3 and b0 (lower panels) respectively for the Si model atom.

1
E0

3. Testing on Silicon
The similar behavior of the quantum rates for charge transfer
and bb transitions observed for the four atoms triggered us to
further test the QFM with the prospect of applying it to other
atoms, such as Fe.
We therefore tested the QFM on Si where comparison with
quantum hydrogen collision rates calculated by Belyaev et al.
(2014) is possible. We tested (i) whether the QFM is able to reproduce the NLTE calculations made using the quantum data,
and (ii) the effects of varying the fitting parameters from their
nominal values. For that purpose we developed a Si i / Si ii model
atom.

3.1. Silicon Model atom


3.1.1. Energy levels

The Si i energy levels and the Si ii ground level (up to 8.15 eV)
were extracted from the NIST2 database (Martin & Zalubas
1983). The total number of Si i fine structure levels used in our
model is 296. This number is comparable to previous studies and
is equal to that used by Shchukina et al. (2012) in their NLTE,
3D-model atmosphere study of the silicon abundance in the Solar photosphere.

and log Qmax for charge transfer (upper panels), and de-excitation b1 , b2 ,

3.1.2. Radiative transitions

Line data for 8674 bb radiative transitions between the energy


levels were extracted from the VALD33 database (Piskunov
et al. 1995; Ryabchikova et al. 1999; Kupka et al. 1999, 2000;
Ryabchikova et al. 2011, 2015). The oscillator strengths, radiative damping coefficients and wavelengths are extracted from
the Kurucz database4 (Kurucz & Bell 1995). The Van der-Waals
hydrogen collisional broadening coefficients were calculated
using the ABO theory (Anstee & OMara 1995; Barklem &
OMara 1997; Barklem et al. 1998) for available transitions, and
otherwise using the Unsld approximation (Unsold 1955), enhanced by a fudge factor of 1.3 as recommended by Gustafsson
et al. (2008).
All the levels in the Si model atom were also coupled to
the Si II ground level via b f radiative (photoionization) transitions. The b f photoionization tables were calculated by Nahar (1993) using the close-coupling approximation and the Rmatrix method. They include photoionization cross-sections for
264 Si i LS-coupling terms of multiplicities 1 and 3, up to n = 10
and l = 8. The photoionization cross-sections for the mean levels were distributed onto the fine structure levels using their
corresponding statistical weights. Due to the large number of
points in the tables and the sharp resonance peaks in the data,
the cross-sections were smoothed and re-sampled. For terms of
multiplicity 5 and for JJ-coupling terms for which no photoionization table exists, Kramers hydrogenic approximation (Travis
& Matsushima 1968) was used to calculate the threshold crosssections.
3

https://www.nist.gov/pml/atomic-spectra-database/

Article number, page 4 of 21

http://vald.astro.uu.se/
http://kurucz.harvard.edu/

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

Fig. 3. De-excitation rate coefficients at T = 6000 K (in logarithmic scale) for Na, Mg, Al and Si (empty circles) from the quantum calculations of
Barklem et al. (2010), Belyaev et al. (2012), Belyaev (2013) and Belyaev et al. (2014) respectively and their corresponding fits (dashed red lines).
The logarithmic of the rate coefficients hvi as a function of transition energies E for all four atoms show a similar pattern of decreasing behavior
with energy which can be fit by a third degree polynomial.

3.1.3. Collisional transitions

Star
log (Si)
log (Si)

All levels in the Si i / Si ii model are coupled via inelastic


collisions with electrons and neutral hydrogen atoms.
For allowed electron collisional bb transitions, the Seaton
(1962a) impact parameter approximation
was used to calculate

the effective collisional strengths e . For forbidden transitions,


Seaton (1962b) approximation is used with a collisional strength
= 1, as recommended for neutral atoms by Allen (1973). All
the levels were coupled to the Si ii ground level through electron
collisional ionization using Bely & van Regemorter (1970)
dipole impact approximation.
For hydrogen collisions, different recipes were used for our
tests, as described in Sec. 3.3 below.
3.2. Model atmospheres

Three reference stars were chosen to perform the tests on


the QFM, namely the Sun, the red giant Arcturus, and the
metal-poor sub-giant star HD 140283.
The adopted input stellar atmospheric parameters (effective
temperature, T eff ; surface gravity log g; metallicity [Fe/H]; and
microturbulent velocity, t ) are from Heiter et al. (2015a) and
Jofr et al. (2014) and are shown in Tab. 4. 1D, LTE MARCS
model atmospheres (Gustafsson et al. 2008) were interpolated to the corresponding atmospheric parameters using the

Sun
7.549
0.016

HD 140283
5.303
0.094

Arcturus
7.287
0.051

Table 2. Adopted values for Si abundance for the three stars used in our
calculations and their respective uncertainties. References are found in
the text.

interpolation routine interpol_marcs.f written by Thomas


Masseron5 . The NLTE radiative transfer code MULTI2.3 (Carlsson 1986, 1992) was used to compute the level populations
and the lines equivalent widths. Background line opacity files
excluding Si lines were included in the MULTI2.3 calculations,
to account for line blanketing effects. The opacity files were
calculated for each stars metallicity and microturbulent velocity
using the MARCS opacity sampling routines.
The Solar Si abundance value adopted in the calculations is
taken from the NLTE, 3D calculations of Shchukina et al. (2012),
while those for the other two stars are from Jofr et al. (2015)
where line-by-line abundance analysis was performed for each
star. The abundance values and their corresponding uncertainties
are shown in Tab. 2.

http://marcs.astro.uu.se/software.php
Article number, page 5 of 21

A&A proofs: manuscript no. main

3.3. Method
3.3.1. Test 1: Is the QFM able to reproduce the results of the
quantum data?

The reference Si model atom (hereafter referred to as the QM


model) was built using the quantum rates from Belyaev et al.
(2014) for the H+Si collisional rates for the de-excitation and
charge transfer processes. Belyaev et al. (2014) computed
rates for mean Si i levels only. The collisional rates of the
mean levels were distributed on the fine structure levels in our
Si i /Si ii model atom according to their statistical weights.
In order to test the effects of using the QFM instead of the
QM data from Belyaev et al. (2014) on the NLTE calculations,
hydrogen collision rates from Eqns. 1, 2, 3 and 4 were included,
using the Si i CT and bb fitting parameters from Tab. 1. This
model atom is hereafter referred to as the QFM model.
Results are shown in Fig. 4, where the relative differences
between the equivalent widths computed in NLTE using
the QFM, EWQFM (NLTE), and those using the QM model,
EWQM (NLTE), are displayed for the three test stars (left panels):
the Sun, Arcturus and HD 140283. All the Si i lines in the
Si i / Si ii model atom lying between 2000 25000
were included and color-coded as a function of the line reduced
equivalent widths, (EWQM (NLTE)/). The relative differences
between the QM and the QFM for all the lines (left panels)
are small for the Sun (< 5%), with larger scatter found for
Arcturus and HD 140283, but with most lines lying within 20%
from EWQM (NLTE). The scatter increases toward the saturated
lines (green circles color-coded in Fig. 4), which correspond
to 5 < EWQM (NLTE)/ < 4. These lines correspond to
transitions at the highest E & 5 eV (see Fig. 3) for the lowest
bb rates, which have been ignored in the QFM fit since they are
4 magnitudes lower than the maximum rate. On the other hand,
the scatter remains small for stronger lines (yellow circles),
corresponding to EWQM (NLTE)/ > 4, except for Arcturus
which has the lowest T eff = 4286 K among the three stars.
The relative differences between the NLTE and LTE EW for
the QFM versus the QM model are also shown (right panels of
Fig. 4). To quantify this difference, we define the two quantities
X and Y, representing the relative differences between the NLTE
and LTE EW for both the QM and QFM models, such that:

Xi =

 EW i (NLTE) EW i (LTE) QM

,
EW i (NLTE)
 EW i (NLTE) EW i (LTE) QFM
Yi =
EW i (NLTE)

(5)

The dispersion between Y and X is more severe toward the weak lines (blue circles), corresponding to
EWQM (NLTE)/ < 6. This is strongly seen for the cases
of Arcturus and HD 140283 where the scatter is generally larger
than the Sun. This shows that the QFM can overestimate the
QM rates for the weak lines, i.e. shifting the EWQFM (NLTE)
closer to LTE than the QM model (Figs. 1 and 3 show this
overestimation for the CT and bb rates respectively).
Article number, page 6 of 21

The mean difference between Y and X for all the lines, N,


can be expressed as:

hY Xi =

N
1 X i
(Y X i )
N i=1

(6)

We plot (Y X) for all the lines between 2000


25000 in Fig. 4 (small right black panels). From that we compute the standard deviation for (Y X), denoted as (Y X) for
the 3 stars. We obtain hY Xi = 0.012 and (Y X) = 0.017
for the Sun, hY Xi = 0.022 and (Y X) = 0.033 for Arcturus
and hY Xi = 0.021 and (Y X) = 0.086 for HD 140283.
This shows that while the scatter between the QM and QFM
can grow up to 10%, more pronounced for Arcturus and
HD 140283 than the Sun, the systematic differences between
both methods only extends to < 4% for all three stars.

3.3.2. Test 2: What is the effect of varying the fitting


parameters?

To test the effects of varying the different fitting parameters


from their reference values in Tab. 1 (hereafter denoted as the
reference fit), modified Si model atoms were used where the
Si i fitting parameters for charge transfer and de-excitation were
varied around their nominal values.
Each parameter was respectively decreased and increased
from its reference value as shown in Tab. 3. a11 and are kept to
their reference values of 1.2 and 0.35 respectively. Changing
one parameter at a time, NLTE Si i equivalent widths were
computed for each model, hereafter referred to as EWmod , and
were compared to EWref . This was done for all three test stars.
To quantify the differences between EWmod and EWref , the
mod
relative differences, Yrel
, were computed from all the lines, N,
which lie within [2000, 25000] as:

mod
Yrel
=

N
i
i 
EWref
1 X  EWmod
i
N i=1
EWref

(7)

mod
mod
The standard deviations of Yrel
, Yrel
, are listed in Tab. 3
for all the models.
Varying the CT fitting parameters a1 , a2 , and E1 0 at
T = 6000 K within 50% of their reference values (see Fig. 5),
result in a small percentage of variation in the computed EWmod
mod
(Yrel
< 0.03, i.e. < 3%). Also, varying Q0 within 1 order
of magnitude from its reference value, shows slightly larger
scatter (Yrelmod < 0.07) than other parameters. This shows that
Q0 plays an expected important role in shifting the computed
EWmod , as it scales and shifts the rates by orders of magnitudes
from their nominal values.

Varying the bb fitting parameters b31 , b30 , b21 , b20 , b11


and b10 within 50% from their reference values and varying
b0 within 1.0 orders of magnitude respectively, results in a
relative difference up to 0.2 (i.e. 20%) from the QFM.
These results show that a variation in the bb fitting parameters from its reference value, as presented in Fig. 5 for CT and
bb, propagate onto a relative difference of < 10% and a scatter

YX

Arcturus

5000

10000 15000 20000 25000 30000


HD140283

5000

0.4
0.2
0.0
0.2
0.4

0.2
0.15
0.1
0.05
0.0
-0.05
-0.1
-0.15
-0.2
-0.25

10000 15000 20000 25000 30000

4.5
0.1

0.0

0.1

0.2

0.3

5.0
0

0.4
0.4
0.2
0.0
0.2
0.4

0.3
0.2
1
0
0.0
-0.1
-0.2

log(EWQM (NLTE)/)

0.3
0.2
0.1
0.0
0.1
0.2
0.30

10000 15000 20000 25000 30000

4.0

5000 10000150002000025000

YX

0.1
0.0
0.1
0.2
0.30

5000

0.3
0.05
0.0
0.2 -0.05
-0.1
-0.15
0.1
0
0.0
0.1
0.20.2

5000 10000150002000025000

0.2

5.5
0.0

0.2

6.0

0.4

YX

0.05
0.00
0.05
0.10
0.150

Sun
Y = [ EW QFM (NLTE) - EW(LTE) ] / EW QFM (NLTE)

( EW QFM (NLTE) - EW QM (NLTE) ) / EW QM (NLTE)

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

6.5
0

5000 10000150002000025000

7.0
0.4
0.2
0.0
0.2
0.4
X = [ EW QM (NLTE) - EW(LTE) ] / EW QM (NLTE)

()

Fig. 4. NLTE equivalent widths calculated using the QM models versus those calculated using the QFM model for Si lines lying between 2000
25000 (left panels). Agreement within < 5% for the Sun and < 20% for Arcturus and HD 140283 is found between both models. Relative
differences between NLTE and LTE are also shown in the right panels for the QFM (denoted Y) versus the QM (denoted X) models for the Sun,
Arcturus and HD 140283 respectively. Lines are color-codes as a function of their reduced line strengths EWQM (NLTE)/. (Y X) as a function
of are shown in the small right black panels.

fitting coefficient

reference value

mod
Yrel
(102 )

model value
Sun

Q0

4.9
-3.5
0.75
-7.5

2.0
-2.0
0.5
-6.5

0.80
0.19
1.02
4.41

b31
b30
b21
b20
b11
b10
b0

-0.08
0.23
0.69
-2.05
-1.35
3.5
-9.75

-0.001
0.1
0.2
-1.0
-0.5
2.0
-8.0

3.57
4.94
4.01
6.42
1.88
5.73
9.88

a10
a2
1
E0

Arcturus HD 140283
CT coefficients
1.03
1.65
0.42
0.98
1.44
1.78
5.76
6.56
bb coefficients
6.25
8.72
5.27
9.38
6.87
9.87
7.52
8.73
6.64
8.62
8.74
9.50
10.87
12.54

mod
Yrel
(102 )

model value
Sun

Arcturus

HD 140283

7.0
-6.0
1.00
-8.5

0.11
0.22
1.29
4.79

1.14
0.86
1.55
5.88

2.26
1.23
1.75
6.51

-0.5
0.5
0.9
-3.0
-2.0
5.0
-11.0

4.26
3.57
7.92
2.71
7.59
3.87
9.52

5.27
7.66
8.02
6.77
8.57
7.48
11.53

9.85
8.72
8.72
9.70
9.95
8.64
13.05

Table 3. Si+H rates fitting parameters variations from the reference values for which NLTE calculations were performed.

of Y < 5% in the computed NLTE equivalent widths as compared to the reference QFM values. This means that the acceptable range of variation around the optimal values can be quite
broad.

4. Application to Iron
The similarity in the behavior of the mutual neutralization and
de-excitation quantum rates as a function of transition energies

for the different chemical species, and the comparable values obtained for the fitting parameters, especially for charge transfer
rates, motivated us to apply the QFM recipe to iron, instead of
the Drawin approximation to estimate the collision rates. The
aim in this section, is therefore to calibrate the iron fitting coefficients of the QFM using a well defined set of benchmark
stars. In what follows, we will introduce a newly developed
Fe i /Fe ii /Fe iii model atom (hereafter denoted as our iron model
atom) which was used for that purpose.
Article number, page 7 of 21

log bb (cm3 . s1 )

log

CT

(cm3 . s1 )

A&A proofs: manuscript no. main

4
6
8
10
12
14
16
18
20
220.0
7
8
9
10
11
12
13
14
15
16 0

QFM fit
Q max [ 9. 0, 6. 0]
1/E 0 [0. 50, 0. 75]
a 1 [2. 0, 6. 0]
a 2 [ 6. 0, 2. 0]
QM rates, T = 6000K

0.5

1.0

1.5

1
E

2.0

2.5

3.0

3.5

b0 = 8. 0
b0 = 11. 0
QFM fit
QM rates, T = 6000K

E (eV)

Fig. 5. Si i +H charge transfer (upper panel) and de-excitation (lower panel) QM rates and their reference QFM fits (black dotted lines). Models
with fitting coefficients varied from their nominal values within the ranges specified on the plots are also shown. a11 , a2 (dark and light shaded
areas respectively) and E1 0 (red horizontal line) are varied by 50% from their nominal values. Qmax (vertical blue line) and b0 (dotted blue and
green lines in lower panel) are varied by 1.0 orders of magnitude from their reference values respectively. The variations lead to < 10% relative
differences with respect to the reference QFM fit.

4.1. Iron model atom


4.1.1. Energy levels

Our iron model atom was built up using all the Fe i energy
levels extracted from the NIST database (846 levels) from the
experimental analysis of Nave et al. (1994), up to 7.815 eV.
The model also includes the predicted high lying Fe i levels
(66 levels) from Peterson & Kurucz (2015), up to 8.392 eV.
The coupling between these high-lying levels and the low-lying
levels correspond to UV transitions, which are important for the
determination of abundances of stars from their UV spectra (Peterson 2011, 2013). Transitions between these high-lying levels
correspond to IR transitions which are used for iron abundances
determination in luminous red giants in dust-obscured regions
like the bulge, bar, and disk of the Milky Way (Majewski 2010).
The model atom was completed with all Fe ii levels from
Nave & Johansson (2013), with 1027 levels up to 19.83 eV,
corresponding to 189 spectroscopic terms. Only the ground
level of Fe iii was included.
To reduce the amount of computing time, as well as memory requirements, the number of levels were reduced by using
a super-level algorithm, in which sets of actual levels are combined into super-levels. The energy levels were combined into
super-levels as follows:
The fine structure levels of the ground and first excited states
of Fe i were included, up to 2 eV.
Article number, page 8 of 21

All other fine structure levels were combined into mean levels according to their statistical weights. The excitation potential of each mean level I (E I ) is a weighted mean of the
excitation potential of the corresponding fine structure levels i (Ei ), computed for a typical temperature T = 5000 K:
PN
Ei gi eEi /kT
E I = Pi=1
N
Ei /kT
i=1 gi e

(8)

where gi is the statistical weight of each level i, and the denominator is the mean level partition function at the temperature T .
After sorting the mean levels in increasing energy order, all
Fe i mean levels lying within an energy interval of 250 cm1
were combined into super-levels. The excitation energy of
each super level was computed in the same way as the mean
levels energy.
For Fe ii , mean levels lying within an increasing energy difference interval starting from 250 cm1 at the Fe ii ground
level up to 2500 cm1 toward the Fe iii ground level were
combined.
Our iron model super-atom contains 365 Fe i and 58 Fe ii levels,
as well as the continuum Fe iii level.

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

4.1.2. Radiative Transitions

We used the VALD3 database to extract all the Fe i and Fe ii radiative bb transitions accounting to 79594 Fe i lines and 113964
Fe ii lines. The UV and IR lines (1568 lines) corresponding to
transitions from and to the predicted high lying levels (Peterson
& Kurucz 2015) were added, increasing the total number of
Fe i lines to 81162.
Individual transitions belonging to levels combined into
superlevels were combined into super-transitions. Oscillator
strengths were averaged using the same method as for the energy,
with a super-level partition function at T = 5000 K. Line broadening parameters (radiative or collisional), were averaged using the levels weights gi .
Our final iron model includes 24182 Fe i transitions and
1256 Fe ii transitions combined from the individual lines.
In addition, all levels were coupled to the next ion ground
level via b f photoionization transitions. For Fe i levels, we used
the photoionization cross-section tables calculated by Bautista
(1997). For Fe ii , we used the computation by Nahar & Pradhan
(1994) (ab-initio R-matrix close-coupling approach) from the
TOPBASE project. The tables correspond to calculations for 52
LS Fe i terms (for n 10 and l 9) of spin multiplicities 1,
3, 5 & 7, and for 83 Fe ii LS terms of spin multiplicities 2, 4,
6 & 8, coupled through photoionization to the ground Fe ii and
Fe iii levels respectively. For levels with no quantum calculations
and for JJ coupling terms, Kramers hydrogenic approximation
was used (Travis & Matsushima 1968).
TOPBASE photoionization energies for each level in the
model were shifted to match the threshold ionization energies in
NIST, and cross-sections were smoothed and resampled to 200
frequency points.
In the super-atom model, the bound-free rates were averaged
using the super-level partition functions.
4.1.3. Collisional transitions

We included inelastic collisional rates with both electrons and


neutral hydrogen atoms.
electron collisions:
For Fe i , quantum electron-impact excitation rates have not
yet been calculated for most transitions. Effective excitation
collisional strengths calculations by Pelan & Berrington (1997)
for the fine-structure transitions involving the ground level and
the first excited level using the R-matrix method were included.
For all other transitions, the excitation cross-sections were estimated using Seaton (1962a) impact parameter approximation
for allowed transitions and that of Seaton (1962b) for forbidden
transitions. The cross-sections were then averaged over a
Maxwellian velocity distribution to obtain the rate coefficients
he vi.
For Fe ii , the situation is better, and we could include the
rate coefficients calculated by Zhang & Pradhan (1995) and
Bautista & Pradhan (1996) for 10011 excitation transitions
among 142 fine structure levels (R-matrix method). The impact
parameter approximation was used for all other transitions. For
b f transitions, Bely & van Regemorter (1970) approximation
was used to calculate the collisional cross-sections for both

Fe i and Fe ii .
hydrogen collisions:
The inclusion of the inelastic hydrogen collision rates in the iron
model using the QFM is explained in Sec. 4.2 below.
4.2. Method

We implement hydrogen collisional rates using Eqns. 1- 3 to


calculate the Fe i + H charge transfer and excitation rates respectively. For Fe ii + H bb collisional rates, Lamberts (1993)
derivation of the Drawin approximation was used since no rates
have been calculated for ionized species.
4.2.1. Benchmark stars

The calibration of the QFM hydrogen collision rates is obtained


by minimizing the difference between calculated and measured
equivalent widths in a sample of benchmark stars. This requires
to reduce as much as possible the number of other unknowns,
especially the stellar atmospheric parameters used in the spectral
synthesis computations.
We therefore use the FGK Gaia Benchmark stars proposed
by Heiter et al. (2015a). These stars are reference objects
for the calibration of all the observed Gaia stars. They are
therefore carefully studied from the ground. They have been
the subject of numerous spectroscopic investigations (Jofr
et al. 2014; Blanco-Cuaresma et al. 2014), which allowed for
the comparison of their parameters determined via different
methods. They have been chosen such that their atmospheric
parameter values cover a wide range of T eff , log g, and [Fe/H]
representing different stellar types and populations. In addition,
their atmospheric parameters were determined as independently
from spectroscopic methods as possible.
In Heiter et al. (2015a) and Jofr et al. (2014), the T eff
of the stars were determined using limb-darkened angular
diameters LD and bolometric fluxes Fbol 6 with uncertainties
less than 3% reported by the authors. Uncertainties of 3% in
LD and 5% in Fbol translate into 1.5% and 1% uncertainties in T eff respectively. The masses were determined from
asteroseismic scaling relations when available for stars with
solar-like oscillations, from binaries relations (for stars in binary
systems), or evolutionary tracks for others. Surface gravities
were calculated using the fundamental gravity relation7 , where
the stellar radii were calculated from the angular diameters LD
and the distances which were determined from their parallaxes
. Uncertainties on log g were found to be less than 0.1 dex for
dwarfs and hotter giants, and up to 0.3 dex for the cooler giants.
The input metallicities and microturbulent velocities of the stars
are adopted from Jofr et al. (2014).
We use 24 stars of the suggested FGK Gaia benchmark stars
from Heiter et al. (2015a), which we considered had reliable
atmospheric parameters. They are listed in Table 4 with their
fundamental atmospheric parameters, and uncertainties.

Using the relation T eff =

g=

Fbol

1/4

( LD
)1/2 .
2

GM
R2

Article number, page 9 of 21

A&A proofs: manuscript no. main

Q0 > 8.5 in the case of the Sun.

4.2.2. Observational data


8

The spectra of the stars were collected using the UVES spectrograph at the VLT9 by the Gaia-ESO collaboration. The sources
of the UVES spectra are from the Advanced Data Products collection of the ESO Science Archive Facility, which were reduced by the standard UVES pipeline version 3.2 (Ballester et al.
2000), and from the UVES Paranal Observatory Project UVESPOP library which were reduced with tools written especially for
that library (Bagnulo et al. 2003).

For the three stars, a2 does not play an important role with
almost identical 2 values obtained for all models, with slightly
smaller values for a2 > 6.0 for Arcturus and HD 140283. It
can be clearly seen that Qmax plays the most important role in
the 2 variation. Interestingly, we find a consistent set of best
fitting parameters for Fe i charge transfer for all 3 stars where:
a10 = 6.0 , a2 = 6.0 ,

1
E0

= 0.75 and Q0 = 8.5 (which

corresponds to log Qmax = 8.42 at T = 6000 K).


4.2.3. Linelist and measured equivalent widths

The Fe i and Fe ii linelists and their log g f values chosen in the


analysis of the NLTE calculations in this work is a subset of the
Gaia-ESO survey golden linelist version 4 (v4) (Heiter et al.
2015b).
The observed equivalent widths were measured from the
UVES spectra using the automatic code DAOSPEC (Stetson &
Pancino 2008) by the Gaia-ESO working group "Bologna" in
Jofr et al. (2014). The lines are listed in Tab. 9.
4.2.4. QFM iron coefficients determination

Charge transfer
The best fitting parameters for Fe i charge transfer rates are
first determined through a comparison of the calculated EWmod
with the measured EWobs using a 2 minimization. For each
of the 3 comparison stars (the Sun, Arcturus and HD 140283),
a grid of NLTE models is calculated by varying the fitting
coefficients around their starting values, chosen equal to the Si
QFM parameters. The fitting coefficients a10 , a2 , E0 and Q0
variation intervals are listed in Tab. 5. a11 = 1.2 and = 0.35
were kept constant at their reference Si values. A total of 243
models were calculated for each star.

De-excitation
To find the best fitting coefficients for the de-excitation rates, the
fitting parameters from Eqn. 3 and 4 were varied for the values
shown in Tab. 5 in steps of param . The charge transfer rates were
kept constant at the best fitting values obtained above. Hence a
total of 5103 models was computed for each star. Similarly, the
calculated EWmod were compared to the measured EWobs using
a 2 minimization.
The fitting coefficient b0 is found to play the most important
role in varying the calculated EWcalc similar to that of log Qmax
for charge transfer rates. It can be seen in Fig. 7 that for values
smaller than b0 = 10 and greater than b0 = 8, EWcalc are
driven away from their corresponding EWobs with 2 increasing
to large values. A clear minimum is obtained for the three stars
at 8.5. The best fit is at b31 = 0.1 , b30 = 0.2 , b21 = 0.7 ,
b20 = 2.0 , b11 = 1.5 , b10 = 3.5 and b0 = 8.5 .
300

180

Sun

90

Arcturus

160

250

140

200

HD140283

80
70

a10
a2
1
E0

Q0

b31
b30
b21
b20
b11
b10
b0

step

100

param
param
param
param

= 2.0
= 2.0
= 0.25
= 1.0

param
param
param
param
param
param
param

= 0.1
= 0.1
= 0.2
= 1.0
= 0.5
= 1.0
= 1.0

Table 5. Fe i +H CT and bb fitting rate coefficients variation range of


values and the steps param used.

The results for the 3 stars at a11 = 2.0 are shown in Fig. 6.
For Arcturus and HD 140283, a clear minimum is obtained at
Q0 = 8.5. The position of this minimum is independent of the
other parameters values, within their range of variation. This
minimum is compatible with an almost flat 2 minimum for
8
9

http://www.eso.org/sci/facilities/paranal/instruments/uves.html
http://www.eso.org/public/teles-instr/vlt/

Article number, page 10 of 21

150

variation interval
CT coefficients
[2.0, 6.0]
[6.0, 2.0]
[0.5, 1.0]
[-12.5,-4.5]
bb coefficients
[0.3, 0.1]
[0.1, 0.3]
[0.5, 0.9]
[3.0, 1.0]
[2.0, 1.0]
[2.5, 4.5]
[12.0, 6.0]

parameters

120
100

50

80

50

40

60

013 12 11 10 9 8 7 6 5

b0

60

4013 12 11 10 9 8 7 6 5

b0

3013 12 11 10 9 8 7 6 5

b0

Fig. 7. 2 values obtained for different values of b0 for QFM deexcitation rates.

4.3. Testing the role of inelastic hydrogen collisions

It is interesting to assess what role does the hydrogen inelastic


collisions play in NLTE calculations, especially charge transfer
processes that have seldom been included. We therefore compare computed LTE and NLTE equivalent widths for 3 cases: (i)
including all H collisions, (ii) removing H charge transfer collisions and (iii) removing all H collisions. We perform these
calculations for three benchmark stars (the Sun, Procyon and
HD 140283). We inspect the relative equivalent width departure
from LTE in each case, for Fe i and Fe ii lines separately. This is
shown in Fig. 8. Obviously, without H-collisions, the equivalent
widths are driven far from LTE, as electron rates alone cannot
ensure LTE. Again, we see that Fe ii lines are weakly affected,

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method
Star name (ID)
Metal-poor
HD 122563
HD 140283
HD 84937
FGK dwarfs
Eri
 For
Cet
18 Sco
Sun
HD 22879
Cen A
Ara
Hyi
Vir
Boo
Procyon
HD 49933
 Eri
FGK Giants
Arcturus
HD 220009
Leo
HD 107328
Gem
 Vir
Hya

T eff [K]

T eff [K]

log g

log g

[Fe/H] [dex ]

[Fe/H] [dex]

t [JOF14] [Kms1 ]

t [This work] [Kms1 ]

4587
5522
6356

60
105
97

1.61
3.58
4.06

0.07
0.11
0.04

2.64
2.36
2.03

0.22
0.10
0.08

1.3
1.3
1.5

1.8
1.0
1.5

4954
5123
5414
5810
5777
5868
5792
5902
5873
6083
6099
6554
6635
5076

30
78
21
80
1
89
16
66
45
41
28
84
91
30

3.76
3.52
4.49
4.44
4.44
4.27
4.31
4.30
3.98
4.10
3.79
4.00
4.20
4.61

0.02
0.08
0.02
0.03
0.0002
0.04
0.01
0.03
0.02
0.02
0.02
0.02
0.03
0.03

+0.06
0.60
0.49
+0.03
+0.03
0.86
+0.26
+0.35
0.04
+0.24
+0.32
+0.01
0.41
0.09

0.05
0.10
0.03
0.03
0.05
0.05
0.08
0.13
0.06
0.07
0.08
0.08
0.08
0.06

1.2
1.2
1.1
1.2
1.2
1.2
1.2
1.2
1.3
1.4
1.4
1.8
1.9
1.1

1.0
0.9
0.7
0.9
0.8
1.0
1.2
0.9
1.3
1.3
1.5
1.8
1.0
0.7

4286
4217
4474
4496
4858
4983
5044

35
60
60
59
60
61
40

1.64
1.43
2.51
2.09
2.90
2.77
2.87

0.09
0.12
0.11
0.13
0.08
0.02
0.02

0.52
0.74
+0.25
0.33
+0.13
+0.15
+0.16

0.08
0.13
0.15
0.16
0.16
0.16
0.20

1.3
1.3
1.1
1.2
1.1
1.1
1.1

1.8
1.7
1.8
1.9
2.0
1.5
1.5

Table 4. The stars adopted in our study: fundamental parameters including the adopted T eff and their uncertainties T eff , and log g and their
uncertainties log g from Heiter et al. (2015a), [Fe/H] , [Fe/H] and microturbulent velocities t from Jofr et al. (2014) and those determined in
this work (see Sect. 5 below).

as they are the dominant Fe ionization stage in cool stars. The


effects on Fe ii equivalent widths amount to less than 10% in the
case of the metal-poor star HD 140283 (except some rare cases
that can reach down to 40%). Including bb hydrogen collisions
drives the lines slightly closer to LTE (represented by the dotted
horizontal lines in Fig. 8) in the case of the Sun and Procyon,
as electron collisions contribute at a similar level. The shift towards LTE is more notable for HD 140283, where free electrons
from metals are more sparse in its atmosphere. The inclusion of
charge transfer collisions, on the other hand, has a strong effect
in all three star models. Neglecting these collisions would lead
to largely overestimated NLTE effects, specially for Fe i lines.
Including these collisions in NLTE calculations, would thus be
necessary in order to reproduce the observations. We stress that
it is particularly important to include charge transfer collisions
with hydrogen in NLTE calculations for iron, as already shown
for Li and Na by Lind et al. (2011) and Barklem et al. (2003)
respectively. In the case of Mg, however, Osorio et al. (2015)
showed that both H-collisions, including charge transfer, and
electron collisions had different impacts on different types of
lines in the Mg atom. It is thus possible that once quantum-based
computations of electron collision rates with iron are available,
our understanding of the effects of different types of collisions
on Fe would yet improve.

5. Iron abundance determination


We included the best fit coefficients for charge transfer and
de-excitation rates from Sec. 4.2.4 in our iron model atom, in
order to determine NLTE Fe i and Fe ii abundances for the Gaia
benchmark stars chosen in Sec. 4.2.1. The equivalent widths are
calculated using MARCS model atmospheres interpolated to the
atmospheric stellar parameters of each star from Tab. 4. The line
list is that tested on the Sun (see Sec. 4.2.3), where blended lines
were removed for each star. All abundances are normalized to

the Solar Fe abundance (Fe) = 7.45 from Asplund et al. (2000).


Line profiles were calculated using a Voigt profile function with
a maximum of 80 frequency points.
For each star, we compute equivalent widths for a range of
iron abundances and microturbulences, centered around [Fe/H]
and t from Heiter et al. (2015a) (see Tab. 6), and keeping T eff
and log g fixed to their values. The microturbulence velocities
and iron abundances are then determined simultaneously using a 2 -minimization algorithm on all the EW of the lines.
Once the microturbulence velocities are set, line-by-line Fe i and
Fe ii abundances are derived and then averaged to obtain the final iron abundances and their standard deviation errors, quoted
in Tab. 6. This process is done for both the LTE and NLTE cases.
The derived microturbulence velocities are shown in Tab. 4.
The NLTE effect (or NLTE correction) defined as the difference between NLTE and LTE average abundances of all lines,
[Fe/H] = [Fe/H]NLTE - [Fe/H]LTE , was calculated for each star
as well. The number of Fe i and Fe ii lines used in the abundance
analyses of the stars are included in Tab. 6. The results are categorized into the different stellar types proposed by Heiter et al.
(2015a).
5.1. Results and Discussion

We plot the NLTE Fe i and Fe ii abundances for the different


types of stars as a function of excitation potential and reduced
equivalent widths (log(EW/)) in Fig. 9 for the metal poor stars,
Fig. 13 for the FGK dwarfs and Fig. 14 for the FGK giants.
5.1.1. Metal-poor stars

Fig. 9 and Tab. 6 show that larger iron abundances are obtained
from Fe ii lines than from Fe i for the three metal-poor stars
in our sample. However, a non-negligible abundance trend
Article number, page 11 of 21

A&A proofs: manuscript no. main

a 2 =-2.0

Sun
140
1
E0 =0.5
120
100
80
60
13 12 11 10 9 8 7 6 5 4
140
1
E0 =0.75
120
100
80
60
13 12 11 10 9 8 7 6 5 4
140
1
E0 =1.0
120
100
80
60
13 12 11 10 9 8 7 6 5 4

Q0

a 2 =-4.0

a 2 =-6.0

Arcturus
HD140283
80
500
1 =0.5
1
450
75
E0
E0 =0.5
400
70
350
65
300
250
60
200
55
150
5013 12 11 10 9 8 7 6 5 4 10013 12 11 10 9 8 7 6 5 4
80
500
1 =0.75
1
450
75
E0
E0 =0.75
400
70
350
65
300
250
60
200
55
150
5013 12 11 10 9 8 7 6 5 4 10013 12 11 10 9 8 7 6 5 4
80
500
1
1
450
75
E0 =1.0
E0 =1.0
400
70
350
65
300
250
60
200
55
150
5013 12 11 10 9 8 7 6 5 4 10013 12 11 10 9 8 7 6 5 4
Q0

Q0

Fig. 6. 2 obtained as a function of the different CT fitting coefficient a2 , E1 0 and Q0 respectively at a10 = 2.0. Vertical panels display results
obtained for different E1 0 values, while different symbols display different a2 values. Panels of each vertical column represent the results for one
star respectively. All stars display a clear minimum 2 at Q0 8.5 independent of the other parameters.

Fe i lines display positive abundance corrections to LTE


up to +0.18 dex for HD 84937, +0.11 dex for HD 122563
and +0.13 dex for HD 140283, while Fe ii abundances have
negligible negative corrections of 0.00 dex for HD 122563 and
HD 84937, 0.01 dex for HD 140283. The large Fe i abundance
NLTE corrections are due to the over-ionization of Fe i lines
by the UV radiation of non-local origin (Asplund et al. 2000;
Bergemann et al. 2012; Mashonkina et al. 2011), which decreases the Fe i level populations with respect to equilibrium
values, i.e. weakening Fe i lines relative to LTE. The majority
ion is Fe ii in this temperature range, as demonstrated in Fig. 10
for HD 122563. This explains why Fe ii lines show negligible
NLTE effects, as their number densities are much less affected
than Fe i densities. The slight Fe ii negative corrections are due
to UV pumping from Fe i levels which enhance the Fe ii level
populations relative to LTE (Mashonkina et al. 2016).
NLTE effects are enhanced in the low metallicity and low
gravity regimes (e.g. HD 122563) due to the decrease of the
number of collisional perturbers in the line formation region
of the atmosphere. Departure coefficients of selected Fe i and
Fe ii levels as a function of optical depths at 5000 are plotArticle number, page 12 of 21

100
1.4
80

1.2
1.0

N NLTE /N LTE

Iron ionization fractions (%)

with EP exists for the three stars as well, most pronounced for
HD 122563 (/EP = 0.09 dex eV1 ). Note also the trend
with reduced equivalent width for HD 84937. These trends may
be due to errors in stellar parameters (most probably T eff , see
below).

60

40

0.8
0.6
0.4

20

Fe I
Fe II
Fe III

0.2
0
4

log 5000

0.0 4

log 5000

Fig. 10. NLTE iron ionization fractions (left panel) and total departure coefficients (NNLTE /NLTE ) (right panel) as a function of optical
depth 5000 for Fe i (full black lines) and Fe ii (dashed blue lines) and
Fe iii (dotted red lines) species in HD 122563.

ted in Fig. 11. Comparison with the Sun shows that departures
from LTE are larger in HD 122563.
As noted above, an abundance trend with line excitation exists for all three stars, which could be due to a T eff error. Uncertainties from Heiter et al. (2015a) range from 60 K to 105 K
for these stars, see Tab. 4. We therefore run additional models

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

0.05

0.05

Sun

HD140283

Procyon
0.0

0.00

0.00
0.2

0.05

(EWNLTE EWLTE)/EWLTE

0.05

0.10

0.4

0.10

0.15

0.6

0.15

0.20

0.30

0.8

0.20

0.25

all H collisions
no charge transfer collisions
no H collisions

1.0

0.25

0.35
0.30
4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5
5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0

log (EWLTE/)

log (EWLTE/)

FeI
FeII
6

log (EWLTE/)

Fig. 8. Impact of inelastic hydrogen collisions on calculated equivalent widths. Three cases are shown for each test star: all H-collisions included,
(ii) no charge transfer H-collisions, and (iii) no H-collisions. The relative equivalent width difference between LTE and NLTE is displayed for both
Fe i and Fe ii lines for each case.

5.1.2. FGK Dwarfs

The NLTE Fe ii abundances are very close to the LTE abundances, mostly slightly lower (by at most -0.03 dex). Fe i abun-

1.2

departure coefficient (b)

with variations of T eff and recompute the 2 . The best fit T eff is
deemed as the value for which the slope for the Fe i abundance
trend as a function of lower level excitation potential is minimized. We get an excellent fit for HD 122563 with T eff =4400 K,
[Fe i /H] = 2.63 0.04 and [Fe ii /H] = 2.57 0.03. This
value is 187 K lower than the fundamental one from Heiter et al.
(2015a), which is based on a direct angular diameter measurement (1.2% uncertainty) but a calibrated bolometric flux (4.7%
uncertainty). We also determine the best fit T eff =4450 K in LTE
which is found to be 50 K higher than the NLTE value. Lineby-line Fe i and Fe ii abundances at the best fit LTE and NLTE
T eff are shown in Fig. 12. A better agreement between Fe i and
Fe ii abundances is obtained in NLTE than LTE at these temperatures.
We also show that decreasing T eff by 22 K for HD 140283
and 106 K for HD 84937 can remove the abundance trends,
which also produces a much better agreement between Fe i and
Fe ii abundances, except for HD 140283. This indicates that
while our determined T eff agrees with the fundamental one
(within 22 K), a higher log g value is required to produce
ionization equilibrium. This result is shown in Sect. 5.2 and
Tab. 8 below, where increasing log g to 3.65 dex produces a
better agreement. We recommend a log g value of 3.65 0.05 for
HD 140283, in agreement with Creevey et al. (2015). The iron
abundances obtained with these new T eff are shown in Tab. 7.

Sun

1.0
0.8

HD122563

1.0
0.8

0.6

a5D
z7D o
x5D o
y3F o
v3D o
a6D
a4G
z6F o

0.4
0.2
0.0 5

1.2

1 0

log5000

0.6

a5D
z7D o
x5D o
y3F o
v3D o
a6D
a4G
z6F o

0.4
0.2
2 0.0 5

1 0

log5000

Fig. 11. Departure coefficients for the Sun and HD 122563 as a function
of optical depth 5000 for selected Fe i (black lines) and Fe ii (red lines)
energy levels.

dances are systematically higher in NLTE, by up to +0.08 dex


(Procyon, HD 49933). The abundance dispersion is similar in
LTE and NLTE, for both Fe i and Fe ii . The total NLTE correction ranges between +0.02 and +0.07 dex, negligible as compared to those obtained for the metal-poor stars. However, the
consistency between Fe i and Fe ii abundances is not better in
NLTE than in LTE.
Inspection of Fig. 13 shows that for a number of these stars,
there is a trend of abundance with excitation. This increases the
dispersion, and may shift the Fe i and Fe ii abundance from one
another. For most stars Fe i and Fe ii abundances are consistent
within the error bars. The extreme cases,  For (+0.18 dex Fe i Article number, page 13 of 21

A&A proofs: manuscript no. main

NLTE
Star name (ID)
Metal-poor
HD 122563
HD 140283
HD 84937
FGK dwarfs
Eri
 For
Cet
18 Sco
Sun
HD 22879
Cen A
Ara
Hyi
Vir
Boo
Procyon
HD 49933
 Eri
FGK giants
Arcturus
HD 220009
Leo
HD 107328
Gem
 Vir
Hya

LTE

NLTE correction

[Fe i /H]

[Fe ii /H]

[Fe/H]

[Fe i /H]

[Fe ii /H]

[Fe/H]

[Fe/H]

N Fe i

N Fe ii

2.66 0.11
2.43 0.05
2.01 0.09

2.46 0.06
2.38 0.03
2.03 0.10

2.65 0.11
2.43 0.06
2.11 0.09

2.77 0.11
2.56 0.05
2.19 0.07

2.46 0.04
2.39 0.07
2.03 0.04

2.76 0.13
2.57 0.09
2.21 0.18

0.11
0.13
0.10

90
47
39

4
3
2

0.07 0.11
0.56 0.10
0.50 0.09
0.05 0.10
0.01 0.07
0.86 0.08
0.31 0.09
0.39 0.13
0.04 0.07
0.16 0.08
0.37 0.03
0.00 0.07
0.49 0.07
0.10 0.12

0.09 0.12
0.69 0.05
0.52 0.05
0.02 0.08
0.05 0.06
0.91 0.06
0.20 0.07
0.26 0.08
0.08 0.03
0.19 0.07
0.34 0.03
0.03 0.07
0.48 0.07
0.02 0.08

0.06 0.12
0.57 0.11
0.50 0.09
0.04 0.11
0.006 0.06
0.86 0.08
0.31 0.09
0.39 0.13
0.04 0.07
0.16 0.08
0.37 0.03
0.004 0.07
0.49 0.07
0.09 0.12

0.03 0.12
0.62 0.09
0.53 0.09
0.006 0.10
0.04 0.07
0.92 0.08
0.27 0.09
0.35 0.11
0.09 0.07
0.11 0.09
0.33 0.15
0.07 0.07
0.57 0.06
0.11 0.13

0.09 0.12
0.68 0.05
0.50 0.05
0.01 0.09
0.07 0.04
0.90 0.06
0.21 0.07
0.26 0.09
0.07 0.03
0.19 0.08
0.34 0.22
0.03 0.07
0.46 0.08
0.04 0.08

0.04 0.12
0.63 0.09
0.53 0.09
0.007 0.10
0.04 0.08
0.92 0.08
0.26 0.09
0.35 0.11
0.09 0.07
0.11 0.09
0.33 0.16
0.06 0.08
0.56 0.07
0.11 0.13

0.02
0.06
0.03
0.05
0.05
0.06
0.04
0.05
0.05
0.05
0.04
0.06
0.07
0.02

207
209
196
228
117
145
216
218
209
231
188
140
87
215

20
18
13
26
14
19
24
26
25
29
18
24
7
13

0.55 0.10
0.71 0.09
0.26 0.16
0.26 0.10
0.12 0.10
0.09 0.09
0.07 0.09

0.47 0.12
0.85 0.08
0.19 0.23
0.36 0.11
0.05 0.16
0.06 0.11
0.06 0.08

0.55 0.10
0.74 0.09
0.28 0.17
0.28 0.10
0.12 0.15
0.06 0.09
0.08 0.09

0.60 0.10
0.77 0.09
0.23 0.17
0.33 0.09
0.06 0.15
0.02 0.09
0.01 0.08

0.44 0.13
0.83 0.08
0.24 0.23
0.34 0.12
0.09 0.17
0.10 0.11
0.06 0.09

0.60 0.10
0.78 0.09
0.26 0.17
0.34 0.09
0.06 0.15
0.03 0.10
0.02 0.08

0.05
0.04
0.02
0.06
0.06
0.06
0.06

217
213
223
203
237
221
207

20
17
24
23
16
26
24

Table 6. NLTE iron abundance calculations using our final iron model and the non-spectroscopic stellar parameters from Heiter et al. (2015a)
for our sample of benchmark stars. The columns correspond to the the NLTE and LTE Fe i , Fe ii and Fe average abundances and their standard
deviations. The NLTE corrections [Fe/H] and the number of Fe i and Fe ii lines used in the analyses are also shown.

5.1.3. FGK Giants

HD122563
2.2

[Fe/H]

2.4
2.6
2.8
3.0
3.2 1

Teff = 4450 K
0

2.2

LTE
1

= 0. 04

[Fe/H] = 2. 63

2.4

[Fe/H]

= 0. 10

[Fe/H] = 2. 78

FeI
FeII

2.6
2.8
3.0

Teff = 4400 K

3.2
1

non LTE
1

EP(eV)

Fig. 12. LTE (upper panel) vs. NLTE (lower) Fe i and Fe ii abundances as a function of excitation potential EP for the metal poor
star HD 122563 using the best fit "free" T eff = 4400 K in NLTE and
T eff = 4450 K in LTE.

Fe ii difference) and Ara (+0.15 dex), both show a significant


trend of abundance with line excitation.
To remove those trends, we minimize the Fe i abundance
trends as a function of EP for  For, 18 Sco, Cent A and
HD 49933 by decreasing T eff by 123 K, 60 K, 42 K and 135 K
respectively. This values are within < 2 from the fundamental
values from Heiter et al. (2015a) for  For and HD 49933, and
within < 1 for 18 Sco and Cent A. This allows obtaining a
much better agreement in Fe i and Fe ii abundances as compared
to using the values from Heiter et al. (2015a) (see Tab. 7 and
Fig. 15)
Article number, page 14 of 21

In these stars, the NLTE Fe ii abundances are also slightly lower


than their LTE counterparts, with differences of up to -0.07 dex
(for HD 220009, the coolest, lowest gravity star of the sample).
The Fe i NLTE abundances are sytematically higher than the LTE
ones, like in the dwarfs, the differences ranging from +0.03 to
+0.08 dex. NLTE corrections range from +0.02 to +0.07 dex.
Again, the star with the largest NLTE Fe i , Fe ii discrepancy
(HD 107328, with 0.13 dex Fe i - Fe ii difference), shows a rather
large abundance trend with line excitation. Heiter et al. (2015a)
explains that the fundamental log g for HD 107328 was found to
be 30% smaller than the spectroscopic one. We find that decreasing T eff by 146 K, decreases the difference down to 0.03 dex.
The error bar on T eff quoted by Heiter et al. (2015a) is 59 K.
5.2. Comparison to the literature NLTE iron abundances

We compare our NLTE Fe i and Fe ii abundances with the


results of 1D, NLTE calculations by Mashonkina et al. (2011)
(hereafter MAS11), Bergemann et al. (2012) (hereafter BER12)
and Amarsi et al. (2016) (hereafter AMR16) for 5 stars we
have in common: the Sun, Procyon, HD 84937, HD 140283
and HD 122563. These studies used iron model atoms with a
large number of levels and transitions, comparable to what we
used. For hydrogen collisions, MAS11and BER12 used the
Drawin approximation with a scaling factor of SH = 1, and did
not include any charge transfer collisional rates. AMR16 used
unpublished quantum calculations. Comparisons to their work
provide an important verification of our QFM.
For the three metal-poor stars in common, our iron abundances agree with AMR16 values within the error bars (values

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

5.0
4.9
4.8
4.7
4.6 (
4.5
1

FeI
FeII

/ EP) = 0. 09
0

HD122563

(Fe)

metal poor stars


( / EW) = 0. 02
2

7.5
6

7.5
6

7.0

6.5

6.0

5.5

5.0

4.5

6.5

6.0

5.5

5.0

4.5

6.0

5.5

5.0

4.5

4.9
4.8
4.7

(Fe)

HD140283

5.0

( / EP) = 0. 02
1

5.9
5.8
5.7
5.6
5.5
5.4
5.3 (
5.2
1

( / EW) = 0. 007
7.0

HD84937

(Fe)

5.1

/ EP) = 0. 05
0

( / EW) = 0. 24
2

EP(eV)

7.5
6

7.0

6.5

log(EW/)

Fig. 9. NLTE Fe i (black circles) and Fe ii (red triangles) abundances as a function of excitation potential EP (left panels) and reduced equivalent
widths log(EW/) (right panels) for the metal-poor stars in our sample using the QFM and our final Fe model atom.

Star
HD 122563
HD 140283
HD 84937
 For
18 Sco
Cen A
HD 49933
HD 107328

fund.
T eff
4587
5522
6356
5123
5810
5792
6635
4496

free
T eff
4400
5500
6250
5000
5750
5750
6500
4350

fund. free
T eff
187
22
106
123
60
42
135
146

[Fe i /H]
2.63 0.03
2.45 0.05
2.10 0.04
0.57 0.06
0.15 0.10
0.39 0.10
0.49 0.07
0.20 0.08

[Fe ii /H]
2.57 0.03
2.38 0.02
2.11 0.02
0.50 0.05
0.08 0.09
0.35 0.07
0.39 0.03
0.17 0.09

[Fe/H]
2.62 0.04
2.42 0.06
2.11 0.04
0.57 0.07
0.15 0.10
0.39 0.09
0.49 0.08
0.19 0.08

Table 7. Differences between fundamental and "free" effective temperatures obtained by forcing no trend in Fe i abundance as a function of EP in
NLTE for stars which exhibited such trends. Fe i , Fe ii NLTE iron abundances and their uncertainties obtained with these "free" T eff values are
also presented.

shown in Tab. 8). Our model atoms differ, as do the codes used.
We did not try to match their line list. For HD 140283, different
T eff ( 70 K difference) and log g ( 0.07 dex difference) were
used in the studies. We therefore also derive the Fe i and
Fe ii abundances for this star using a model atmosphere with the
same parameters as AMR16, which gave closer values to their
result.
Our abundance determination for HD 84937 is 0.17 dex
and 0.10 dex higher for Fe i and 0.05 dex and 0.09 dex higher
for Fe ii than those of MAS11 and BER12 respectively. For
HD 140283, comparison to BER12 (no calculations for this
star by MAS11) shows a 0.05 dex lower abundance for Fe i ,
while the Fe ii values are 0.06 dex higher. A notable difference
is the input stellar parameters where a 250 K higher T eff , and
0.12 dex higher logg values were used by BER12. We, therefore,
perform an additional NLTE computation for this star using

the same parameters as BER12 (i.e. T eff = 5777 K and log g =


3.70 dex). Similar to HD 84937, we find 0.05 dex and 0.11 dex
higher abundances than BER12 for Fe i and Fe ii respectively.
For HD 122563, our Fe i abundance is 0.09 dex and 0.08 dex
lower than those obtained by MAS11 and BER12 respectively
upon using fundamental parameters which are 13 K and 78 K
in T eff lower than MAS11 and BER12 respectively. For accurate comparison, we also compute abundances with same
parameters as BER12 (T eff = 4665 K and log g = 1.64 dex),
and similarly obtain 0.06 dex higher Fe i abundance than BER12.
The agreement of our determinations with the analysis of
AMR16 indicates that our method performs better than the
Drawin approximation in estimating H-collisional rates. It
can presumably be extended to other elements for which no
H-collisional data exist.
Article number, page 15 of 21

A&A proofs: manuscript no. main


Object
Sun

Procyon

HD 84937

HD 140283

HD 122563

MAS11
BER12
This work
MAS11
BER12
This work
MAS11
BER12
AMR16
This work
BER12
AMR16
This work(a)
This work(b)
This work(c)
MAS11
BER12
AMR16
This work(a)
This work(c)

T eff
5777
5777
5777
6510
6543
6554
6350
6408
6356
6356
5777
5591
5522
5591
5777
4600
4665
4587
4587
4665

log g
4.44
4.44
4.46
3.96
3.94
4.00
4.09
4.13
4.06
4.06
3.70
3.65
3.58
3.65
3.70
1.607
1.64
1.61
1.61
1.64

Fe I
7.540.09
7.440.05
7.460.07
7.320.07
7.370.04
7.450.07
5.320.07
5.390.07
5.450.07
5.490.09
5.070.09
4.980.09
5.020.05
4.940.09
5.120.09
4.980.10
4.870.12
4.700.07
4.790.11
4.930.11

Fe II
7.560.05
7.440.04
7.400.06
7.420.05
7.400.04
7.480.07
5.370.04
5.330.04
5.400.04
5.420.10
5.010.04
5.060.04
5.070.03
5.070.10
5.120.08
4.890.07
4.950.05
4.970.04
4.990.06
4.960.06

Table 8. This works NLTE Fe i and Fe ii abundances compared to


those of Mashonkina et al. (2011) (MAS11), Bergemann et al. (2012)
(BER12) and Amarsi et al. (2016) (AMR16) for the Sun, Procyon,
HD 84937, HD 140283 and HD 122563. Notes: (a) Using same fundamental parameters as in Heiter et al. (2015a). (b) Using same parameters
as in Amarsi et al. (2016). (c) Using same parameters as in Bergemann
et al. (2012).

For Procyon, a 0.1 dex abundance difference is found


between Fe i and Fe ii by MAS11 (7.32 and 7.42 dex for Fe i and
Fe ii respectively), which agrees well with the determinations
by BER12 within uncertainties (though a smaller difference
of 0.07 dex was found by BER12). Larger values, though,
are determined in this work (7.49 and 7.42 dex for Fe i and
Fe ii respectively). It has to be noted that while the input T eff in
the three studies agree within uncertainties, the log g used in
this work is slightly larger by 0.04 dex. For the Sun, our values
agree within 0.04 dex with BER12, however significantly higher
values were reported by MAS11 (up to 0.16 dex higher in
Fe ii ).
The large discrepancies between the results, computed obtained in this study and Amarsi et al. (2016) and those from
MAS11 and BER12, computed at the same stellar parameters,
can be explained by the different inputs for the hydrogen collisions adopted in our and these studies, which is reflected
more severely in the metal-poor star. The Drawin approximation
(adopted by MAS11 and BER12) overestimates the collisional
rates as compared to the QFM Barklem et al. (2010), i.e. shifting
the abundances to lower values and closer to LTE than what is
expected by the QFM. These results demonstrate that previous
studies, adopting the Drawin approximation, have also underestimated the NLTE corrections.

6. Conclusions
The work in this paper shed the light on the role of hydrogen
collisions in NLTE iron modeling in cool stellar atmospheres,
which is usually modeled with the semi-classical Drawin
approximation.
We introduce a new method to calculate hydrogen collisional
rates, which we call the Quantum Fitting Method (QFM), which
is based on fitting the existing quantum calculations of a few
Article number, page 16 of 21

elements, namely, Na, Mg, Al and Si. We show that a general


fitting recipe for charge transfer rates can be used for all four
elements. A different recipe is adopted for de-excitation rates as
well.
The fitting recipes are first tested on Si, where NLTE calculations using the QFM are shown to reproduce the equivalent
width calculations using the real quantum data within a scatter
of < 10% for the Sun and < 20% for Arcturus and HD 140283.
These dispersions, however, reflect onto NLTE EW corrections
of less than < 4% for all the three stars. Our calculations also
demonstrate that a variation in the fitting parameters within
50% from their reference values, only alters the results of the
calculations within < 10%, and thus the range of acceptable
fitting parameters determined for each atom is rather broad.
The QFM was then extended to neutral iron, for which no
quantum data have been published yet. The general form of the
excitation and charge transfer collisional rate fits obtained for
other elements were conserved and used. The fitting parameters
for each of charge transfer and excitation rates were varied at
once, thus creating a multidimensional grid of parameters for
which NLTE calculations were computed. A best fit minimum
was found at the same set of parameters for three stars of
different stellar types. This emphasizes that the QFM is quite
robust.
The best fit model was used to determine 1D, NLTE
and LTE Fe i and Fe ii abundances for a sample of 24 Gaia
benchmark stars. Non-negligible NLTE effects ( 0.13 dex)
were determined for the metal-poor stars while more negligible
corrections were obtained for solar-type stars, as expected and
in agreement with previous studies. Our results also agree very
well with the most recent iron abundance calculations using
quantum rates by Amarsi et al. (2016) showing that our method
is successful in reproducing the results of the actual quantum
data for H collisions and can be used to replace them in the case
of the lack of the latter. Our QFM based iron abundances are
mostly larger than the previous calculations using the Drawin
approximation, showing larger NLTE effects than previous
studies (Mashonkina et al. 2011; Bergemann et al. 2012; Lind
et al. 2012). This shows, once again, that the classical Drawin
approximation not only over estimates the H collisional rates,
but also the derived abundances from Fe i lines.
We also show that inelastic charge exchange collisions with
H atoms play a very important role in NLTE iron abundance
computations as it can dominate over excitation processes,
especially in metal-poor stars, and the neglect to include it can
lead to wrong abundance determination.
Our results also predict that lower T eff are required to remove any trend in the determined abundances as a function of
excitation potential for the metal-poor stars than the fundamental
photometric values (such as the InfraRed Flux Method, IRFM).
This decrease also reduces significantly the difference between
Fe i and Fe ii , as well as decreasing the abundances down to
0.10 dex. These predictions agree with previous studies such as
Creevey et al. (2012, 2015) who upon fixing log g from nonspectroscopic values (e.g. asteroseismology, parallaxes) found
that lower T eff in NLTE are required than LTE to remove any
abundance trend with line excitation.
A main interesting contribution of the QFM is its generality and the ability to utilize it in NLTE calculations for other

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

atoms prone to NLTE effects in cool stars, e.g. O i, S i, K i, Sc i,


Ti i, Co i, Sr i, ... (See e.g. review by Bergemann & Nordlander
(2014)) where no quantum rates have been calculated or published yet. We highly recommend the adoption of this method in
NLTE computations as opposed to using the Drawin approximation and the global scaling factor SH .
Acknowledgements. We thank Paul Barklem for the valuable discussions and insights on inelastic hydrogen collisions throughout the past years. We thank and
acknowledge the GES-CoRoT collaboration for providing us with the UVES
spectra for the benchmark stars, which have been useful in this work. R.E. acknowledges support from a JINA-CEE fellowship (Joint Institute for Nuclear
Astrophysics - Center for the Evolution of the Elements), funded in part by the
National Science Foundation under Grant No. PHY-1430152 (JINA-CEE). T.M.
is supported by a grant from the Fondation ULB. This research is partially financed by the CEDRE PHC program, and the CNRS Programme National de
Physique Stellaire. This work has made use of the VALD database, operated at
Uppsala University, the Institute of Astronomy RAS in Moscow, and the University of Vienna.

References
Allen, C. W. 1973, Astrophysical quantities
Allende Prieto, C., Majewski, S. R., Schiavon, R., et al. 2008, Astronomische
Nachrichten, 329, 1018
Amarsi, A. M., Lind, K., Asplund, M., Barklem, P. S., & Collet, R. 2016, MNRAS
Anstee, S. D. & OMara, B. J. 1995, MNRAS, 276, 859
Asplund, M. 2005, ARA&A, 43, 481
Asplund, M., Nordlund, ., Trampedach, R., & Stein, R. F. 2000, A&A, 359,
743
Bagnulo, S., Jehin, E., Ledoux, C., et al. 2003, The Messenger, 114, 10
Ballester, P., Modigliani, A., Boitquin, O., et al. 2000, The Messenger, 101, 31
Bard, A., Kock, A., & Kock, M. 1991, Astron. and Astrophys., 248, 315, (BKK)
Bard, A. & Kock, M. 1994, Astron. and Astrophys., 282, 1014, (BK)
Barklem, P. S. 2016, Phys. Rev. A, 93, 042705
Barklem, P. S., Anstee, S. D., & OMara, B. J. 1998, PASA, 15, 336
Barklem, P. S., Anstee, S. D., & OMara, B. J. 2015, abo-cross: Hydrogen broadening cross-section calculator, Astrophysics Source Code Library
Barklem, P. S., Belyaev, A. K., & Asplund, M. 2003, A&A, 409, L1
Barklem, P. S., Belyaev, A. K., Dickinson, A. S., & Gada, F. X. 2010, A&A,
519, A20
Barklem, P. S., Belyaev, A. K., Guitou, M., et al. 2011, A&A, 530, A94
Barklem, P. S. & OMara, B. J. 1997, MNRAS, 290, 102
Bautista, M. A. 1997, A&AS, 122, 167
Bautista, M. A. & Pradhan, A. K. 1996, A&AS, 115, 551
Bely, O. & van Regemorter, H. 1970, ARA&A, 8, 329
Belyaev, A. K. 2013, A&A, 560, A60
Belyaev, A. K. & Barklem, P. S. 2003, Phys. Rev. A, 68, 062703
Belyaev, A. K., Barklem, P. S., Spielfiedel, A., et al. 2012, Phys. Rev. A, 85,
032704
Belyaev, A. K., Yakovleva, S. A., & Barklem, P. S. 2014, A&A, 572, A103
Belyaev, A. K., Yakovleva, S. A., Guitou, M., et al. 2016, A&A, 587, A114
Bergemann, M., Lind, K., Collet, R., Magic, Z., & Asplund, M. 2012, MNRAS,
427, 27
Bergemann, M. & Nordlander, T. 2014, NLTE Radiative Transfer in Cool Stars,
ed. E. Niemczura, B. Smalley, & W. Pych, 169185
Blanco-Cuaresma, S., Soubiran, C., Jofr, P., & Heiter, U. 2014, A&A, 566, A98
Carlsson, M. 1986, Uppsala Astronomical Observatory Reports, 33
Carlsson, M. 1992, in Astronomical Society of the Pacific Conference Series,
Vol. 26, Cool Stars, Stellar Systems, and the Sun, ed. M. S. Giampapa & J. A.
Bookbinder, 499
Castelli, F. & Kurucz, R. L. 2010, A&A, 520, A57
Creevey, O. L., Thvenin, F., Berio, P., et al. 2015, A&A, 575, A26
Creevey, O. L., Thvenin, F., Boyajian, T. S., et al. 2012, A&A, 545, A17
Drawin, H.-W. 1968, Zeitschrift fur Physik, 211, 404
Drawin, H. W. 1969a, Zeitschrift fur Physik, 225, 470
Drawin, H. W. 1969b, Zeitschrift fur Physik, 225, 483
Ezzeddine, R., Merle, T., & Plez, B. 2016, Astronomische Nachrichten, 337, 850
Fuhr, J. R., Martin, G. A., & Wiese, W. L. 1988, Journal of Physical and Chemical Reference Data, Volume 17, Suppl. 4. New York: American Institute of
Physics (AIP) and American Chemical Society, 1988, 17, (FMW)
Gehren, T., Korn, A. J., & Shi, J. 2001, A&A, 380, 645
Gilmore, G., Randich, S., Asplund, M., et al. 2012, The Messenger, 147, 25
Guitou, M., Spielfiedel, A., Rodionov, D. S., et al. 2015, Chemical Physics, 462,
94
Gustafsson, B., Edvardsson, B., Eriksson, K., et al. 2008, A&A, 486, 951

Heiter, U., Jofr, P., Gustafsson, B., et al. 2015a, A&A, 582, A49
Heiter, U., Lind, K., Asplund, M., et al. 2015b, Phys. Scr, 90, 054010
Jofr, P., Heiter, U., Soubiran, C., et al. 2015, A&A, 582, A81
Jofr, P., Heiter, U., Soubiran, C., et al. 2014, A&A, 564, A133
Korn, A. J., Shi, J., & Gehren, T. 2003, A&A, 407, 691
Kupka, F., Piskunov, N., Ryabchikova, T. A., Stempels, H. C., & Weiss, W. W.
1999, A&AS, 138, 119
Kupka, F. G., Ryabchikova, T. A., Piskunov, N. E., Stempels, H. C., & Weiss,
W. W. 2000, Baltic Astronomy, 9, 590
Kurucz, R. & Bell, B. 1995, Atomic Line Data (R.L. Kurucz and B. Bell) Kurucz CD-ROM No. 23. Cambridge, Mass.: Smithsonian Astrophysical Observatory, 1995., 23
Kurucz, R. L. 2007, Robert L. Kurucz on-line database of observed and predicted
atomic transitions
Kurucz, R. L. 2013, Robert L. Kurucz on-line database of observed and predicted
atomic transitions
Lambert, D. L. 1993, Physica Scripta Volume T, 47, 186
Lind, K., Asplund, M., Barklem, P. S., & Belyaev, A. K. 2011, A&A, 528, A103
Lind, K., Bergemann, M., & Asplund, M. 2012, MNRAS, 427, 50
Majewski, S. R. 2010, in IAU Symposium, Vol. 262, IAU Symposium, ed. G. R.
Bruzual & S. Charlot, 99110
Martin, W. C. & Zalubas, R. 1983, J. Phys. Chem. Ref. Data, 12, 323
Mashonkina, L., Gehren, T., Shi, J.-R., Korn, A. J., & Grupp, F. 2011, A&A,
528, A87
Mashonkina, L. I., Sitnova, T. N., & Pakhomov, Y. V. 2016, Astronomy Letters,
42, 606
Nahar, S. N. 1993, Phys. Scr, 48, 297
Nahar, S. N. & Pradhan, A. K. 1994, Journal of Physics B Atomic Molecular
Physics, 27, 429
Nave, G. & Johansson, S. 2013, ApJS, 204, 1
Nave, G., Johansson, S., Learner, R. C. M., Thorne, A. P., & Brault, J. W. 1994,
ApJS, 94, 221
Osorio, Y., Barklem, P. S., Lind, K., et al. 2015, A&A, 579, A53
Pelan, J. & Berrington, K. A. 1997, A&AS, 122, 177
Peterson, R. C. 2011, ApJ, 742, 21
Peterson, R. C. 2013, ApJ, 768, L13
Peterson, R. C. & Kurucz, R. L. 2015, ApJS, 216, 1
Piskunov, N. E., Kupka, F., Ryabchikova, T. A., Weiss, W. W., & Jeffery, C. S.
1995, A&AS, 112, 525
Ryabchikova, T. A., Pakhomov, Y. V., & Piskunov, N. E. 2011, Kazan Izdatel
Kazanskogo Universiteta, 153, 61
Ryabchikova, T. A., Piskunov, N. E., & Kurucz, R. L. 2015, Physics Scripta, 90,
article id. 054005
Ryabchikova, T. A., Piskunov, N. E., Stempels, H. C., Kupka, F., & Weiss, W. W.
1999, Physica Scripta Volume T, 83, 162
Seaton, M. J. 1962a, Proceedings of the Physical Society, 79, 1105
Seaton, M. J. 1962b, in Atomic and Molecular Processes, ed. D. R. Bates, 375
Shchukina, N., Sukhorukov, A., & Trujillo Bueno, J. 2012, ApJ, 755, 176
Steinmetz, M., Zwitter, T., Siebert, A., et al. 2006, AJ, 132, 1645
Stetson, P. B. & Pancino, E. 2008, PASP, 120, 1332
Thvenin, F. & Idiart, T. P. 1999, ApJ, 521, 753
Thomson, J. J. 1912, Philosophical Magazine, 23, 499
Travis, L. D. & Matsushima, S. 1968, ApJ, 154, 689
Unsold, A. 1955, Physik der Sternatmospharen, MIT besonderer Berucksichtigung der Sonne.
Zhang, H. L. & Pradhan, A. K. 1995, A&A, 293, 953

Article number, page 17 of 21

A&A proofs: manuscript no. main

7.85
7.80
7.75
7.70
7.65 (
7.60
1
7.65
7.60
7.55
7.50
7.45
7.40
7.35 (
7.30
1
7.3
7.2
7.1
7.0
6.9
6.8 (
6.7
1
7.7
7.6
7.5
7.4
7.3
7.2 (
7.1
1

Eri
4

4.0
6

/ EP) = 0. 05

4.0
6

/ EP) = 0. 01

4.0
6

/ EP) = 0. 03

4.0
6

3.5
6

EP(eV)

4.0
6

/ EP) = 0. 05

4.0
6

/ EP) = 0. 07

4.0
6

/ EP) = 0. 02

2.5
6

/ EP) = 0. 01

3.5
6

3.0
6

3.5
6

/ EP) = 0. 03

3.5
6

/ EP) = 0. 01

4.0
6

2.5

2.0

1.5

1.0

0.5

3.0

2.5

2.0

1.5

1.0

0.5

2.5

2.0

1.5

1.0

0.5

18 Sco

3.5

Cet

3.0

3.5

3.0

( / EW) = 0. 02

0.5

( / EW) = 0. 06

3.5

( / EW) = 0. 06

1.0

( / EW) = 0. 08

/ EP) = 0. 04
0

1.5

3.0

2.5

( / EW) = 0. 005
3.5

3.0

Sun

/ EP) = 0. 0005

2.0

2.0

log(EW/)

1.5

1.0

0.5

HD22879

2.5

2.5

2.0

1.5

1.0

0.5

2.5

2.0

1.5

1.0

0.5

2.5

2.0

1.5

1.0

0.5

Cen A

3.0

( / EW) = 0. 08
3.5

3.0

Ara

3.5

( / EW) = 0. 13
3.5

3.0

Hyi

( / EW) = 0. 05

For

( / EW) = 0. 02
2.0

1.5

1.0

0.5

Vir

7.7
7.6
7.5
7.4
7.3 (
7.2
1
8.0
7.8
7.6
7.4
7.2 (
7.0
1

( / EW) = 0. 02
EP(eV)

3.0

2.5

2.0

log(EW/)

1.5

1.0

0.5

Boo

6.8
6.6
6.4
6.2 (
6.0
1
8.2
8.0
7.8
7.6
7.4 (
7.2
1
8.4
8.2
8.0
7.8
7.6 (
7.4
1

/ EP) = 0. 01

/ EP) = 0. 0007
0

/ EP) = 0. 0005
0

( / EW) = 0. 04
2.5

( / EW) = 0. 03
3.0

2.0

1.5

1.0

Procyon

7.7
7.6
7.5
7.4
7.3 (
7.2
1

FeI
FeII

2.5

2.0

1.5

1.0

2.5

2.0

1.5

1.0

HD49933

7.8
7.6
7.4
7.2 (
7.0
1
7.4
7.2
7.0
6.8
6.6 (
1
7.2
7.0
6.8
6.6
6.4 (
1
7.9
7.8
7.7
7.6
7.5
7.4
7.3 (
7.2
1

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

FGK Dwarfs

( / EW) = 0. 04
3.0

Eri

(Fe)

( / EW) = 0. 03
EP(eV)

3.5

3.0

2.5

2.0

log(EW/)

1.5

1.0

0.5

Fig. 13. NLTE Fe i and Fe ii abundances as a function of excitation potential EP (left panels) and measured equivalent width EW (right panels) for
the FGK dwarf stars in our stars list using our final iron model atom.
Article number, page 18 of 21

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

( / EW) = 0. 05

4.0
6

4.0
6

4.0
6

4.0
6

3.0
6

3.5
6

3.5
6

/ EP) = 0. 01
0

/ EP) = 0. 06
0

3.0

2.5

2.0

1.5

1.0

0.5

2.5

2.0

1.5

1.0

0.5

2.5

2.0

1.5

1.0

0.5

2.5

2.0

1.5

1.0

0.5

( / EW) = 0. 05

/ EP) = 0. 01
0

3.5

3.5

3.0

Leo

HD220009

/ EP) = 0. 003

Arcturus

FeI
FeII

( / EW) = 0. 08
3.5

3.0

HD107328

(Fe)

(Fe)

(Fe)

(Fe)

FGK Giants
7.3
7.2
7.1
7.0
6.9
6.8 (
6.7
1
7.0
6.9
6.8
6.7
6.6
6.5 (
6.4
1
8.1
8.0
7.9
7.8
7.7
7.6
7.5
7.4 (
7.3
1
7.6
7.4
7.2
7.0
6.8 (
1

( / EW) = 0. 09
3.5

3.0

8.0

Gem

(Fe)

7.8
7.6
7.4
7.2

( / EP) = 0. 006
1

( / EW) = 0. 20
2.5

2.0

1.5

1.0

7.8

7.5
7.4
7.3

(Fe)

Vir

7.6

( / EP) = 0. 01
1

7.9
7.8
7.7
7.6
7.5
7.4
7.3 (
7.2
1

( / EW) = 0. 03
3.0

2.5

2.0

1.5

1.0

2.5

2.0

1.5

1.0

Hya

(Fe)

7.7

/ EP) = 0. 004
0

( / EW) = 0. 02
EP(eV)

3.0

log(EW/)

Fig. 14. NLTE Fe i and Fe ii abundances as a function of excitation potential EP (left panels) and measured equivalent width EW (right panels) for
the FGK giant stars in our stars list using our final iron model atom.

Article number, page 19 of 21

A&A proofs: manuscript no. main

Free Teff

( / EW) = 0. 02
2

6.5
6

7.5
6

7.5
6

6.5
6

6.2
6

7.0
6

/ EP) = 0. 003

6.5
6

/ EP) = 0. 008

6.2
6

6.0

5.5

5.0

4.5

HD140283

/ EP) = 0. 008

HD122563

FeI
FeII

/ EP) = 0. 008
1

7.0

6.5

6.0

5.5

5.0

4.5

6.5

6.0

5.5

5.0

4.5

( / EW) = 0. 01
7.0

For

( / EW) = 0. 02

HD84937

/ EP) = 0. 005

7.4
7.3
7.2
7.1
7.0 (
6.9
1

/ EP) = 0. 008
1

6.0

5.5

5.0

4.5

4.0

( / EW) = 0. 03
6.0

5.8

5.6

5.4

5.2

5.0

4.8

4.6

4.4

18 Sco

( / EW) = 0. 05

Cent A

/ EP) = 0. 0003
0

( / EW) = 0. 002

6.0

5.5

5.0

4.5

4.0

( / EW) = 0. 002
6.0

5.5

5.0

4.5

4.0

( / EW) = 0. 01
EP(eV)

Fig. 15. Fe i and Fe ii abundances obtained from best fit "free" T eff in Tab. 7

Article number, page 20 of 21

6.5

HD49933

8.1
8.0
7.9
7.8
7.7
7.6 (
7.5
1
8.0
7.9
7.8
7.7
7.6
7.5
7.4
7.3 (
7.2
1
7.2
7.1
7.0
6.9
6.8
6.7 (
6.6
1

/ EP) = 0. 006

HD107328

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

(Fe)

4.7
4.6
4.5
4.4
4.3 (
4.2
1
5.2
5.1
5.0
4.9
4.8
4.7
4.6 (
4.5
1
5.40
5.35
5.30
5.25
5.20
5.15 (
5.10
1
7.1
7.0
6.9
6.8
6.7 (
6.6
1

6.0

5.8

5.6

5.4

5.2

log(EW/)

5.0

4.8

4.6

4.4

R. Ezzeddine et al.: NLTE iron abundance determination in cool stars: Introducing the (QFM) method

element

Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I

(air)
[]

EP
[eV]

logg f

EW
[m]

EW
[m]

ref

element

4787.827
4794.354
4799.406
4808.148
4809.938
4869.463
4875.877
4877.604
4892.859
4905.133
4962.572
4986.223
4999.112
5023.498
5029.618
5031.914
5054.643
5088.153
5104.438
5129.631
5180.056
5197.936
5228.376
5243.776
5247.050
5253.021
5267.270
5285.127
5293.959
5294.547
5295.312
5320.036
5322.041
5326.143
5361.625
5379.574
5401.267
5409.133
5417.033
5441.339
5452.088
5472.708
5473.163
5491.832
5522.446
5539.280
5543.936
5546.506
5549.949
5560.211
5567.391
5577.025
5584.765
5618.632
5619.595
5636.696
5651.469
5652.317
5679.023
5680.240
5691.497
5717.833
5731.762
5741.848
5760.344
5848.126
5849.683
5855.076
5858.778
5861.108
5873.212
5881.280
5909.972
5927.789
5987.065
6034.035
6093.643

2.998
2.424
3.640
3.252
3.573
3.547
3.332
2.998
4.218
3.929
4.178
4.218
4.186
4.313
3.415
4.371
3.640
4.154
4.283
3.943
4.473
4.301
4.220
4.256
0.087
2.279
4.371
4.435
4.143
3.640
4.415
3.642
2.279
3.573
4.415
3.695
4.320
4.371
4.415
4.313
3.640
4.209
4.191
4.186
4.209
3.642
4.218
4.371
3.695
4.435
2.609
5.033
3.573
4.209
4.387
3.640
4.473
4.260
4.652
4.186
4.301
4.284
4.256
4.256
3.642
4.608
3.695
4.608
4.220
4.283
4.256
4.608
3.211
4.652
4.796
4.313
4.608

-2.604
-3.950
-2.130
-2.690
-2.620
-2.420
-1.900
-3.050
-1.290
-1.730
-1.182
-1.290
-1.640
-1.670
-1.950
-1.570
-1.921
-1.680
-1.590
-1.670
-1.160
-1.540
-1.190
-1.050
-4.949
-3.840
-1.596
-1.660
-1.770
-2.760
-1.590
-2.440
-2.802
-2.071
-1.330
-1.514
-1.820
-1.200
-1.580
-1.630
-2.802
-1.495
-2.040
-2.188
-1.450
-2.560
-1.040
-1.210
-2.810
-1.090
-2.568
-1.543
-2.220
-1.255
-1.600
-2.510
-1.900
-1.850
-0.820
-2.480
-1.450
-0.990
-1.200
-1.672
-2.390
-1.056
-2.890
-1.478
-2.160
-2.304
-2.040
-1.740
-2.587
-0.990
-0.429
-2.312
-1.400

36.86
10.90
31.24
23.68
16.90
20.21
54.27
17.33
52.12
32.50
53.31
42.27
30.80
23.52
47.44
23.21
34.21
34.92
30.10
44.62
45.24
29.79
56.16
58.89
65.21
17.80
27.84
24.84
25.80
11.25
25.12
17.46
59.47
36.93
41.62
57.98
21.27
50.06
30.99
28.56
12.75
40.69
16.23
10.01
40.36
17.00
59.97
47.09
7.39
48.96
62.74
9.27
36.89
47.68
30.84
17.9
15.66
23.53
55.18
8.59
37.01
57.6
55.38
30.62
21.69
38.96
6.61
20.85
12.05
6.98
16.78
14.77
30.49
40.96
73.56
7.44
29.58

0.25
0.48
0.49
0.24
0.43
0.02
1.21
0.20
0.22
0.20
0.22
1.88
2.60
4.06
0.68
0.21
0.32
1.16
2.54
8.05
0.04
1.13
1.45
0.77
0.81
0.25
0.67
0.15
0.32
0.28
0.32
2.46
0.32
0.76
1.12
0.33
0.19
0.57
0.34
0.25
0.29
1.44
1.04
0.27
0.30
0.07
0.25
0.94
1.59
0.40
3.26
0.19
1.35
0.19
0.54
1.68
0.37
0.29
0.36
0.80
0.88
2.82
1.80
0.54
1.56
0.66
0.45
0.35
0.12
0.16
0.42
0.72
0.92
0.49
2.03
0.31
1.33

K07
K07
K07
K07
FMW
K07
K07
K07
FMW
K07
K07
K07
K07
K07
FMW
K07
K07
K07
FMW
FMW
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
FMW
FMW
K07
K07
FMW
K07
K07
K07
FMW
FMW
FMW
K07
K07
BK
K07
FMW
K07
K07
K07
K07
BKK
K07
K07
K07
FMW
K07
K07
K07
FMW
FMW
K07
K07
K07
K07
K07
K07
K07
K07
FMW
K07
K07
K07
K07
K07
K07
K07
K07

Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe I
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II
Fe II

(air)
[]

EP
[eV]

logg f

EW
[m]

EW
[m]

6096.664
6127.906
6151.617
6157.728
6165.360
6170.506
6173.334
6180.203
6187.989
6200.312
6226.734
6229.226
6270.224
6271.278
6293.924
6297.792
6311.500
6380.743
6436.406
6469.192
6481.870
6569.214
6574.227
6581.209
6591.313
6593.870
6597.559
6608.025
6609.110
6625.021
6627.544
6703.566
6710.318
6716.236
6739.521
6745.956
6750.151
6752.707
6786.858
6810.262
4993.350
5100.654
5132.661
5256.931
5264.802
5284.103
5325.552
5414.069
6149.245
6238.385
6239.942
6369.459
6456.379
6516.076

3.984
4.143
2.176
4.076
4.143
4.796
2.223
2.728
3.943
2.609
3.884
2.845
2.858
3.332
4.835
2.223
2.832
4.186
4.186
4.835
2.279
4.733
0.990
1.485
4.593
2.433
4.796
2.279
2.559
1.011
4.549
2.759
1.485
4.580
1.557
4.076
2.424
4.638
4.191
4.607
2.807
2.807
2.807
2.891
3.230
2.891
3.221
3.221
3.889
3.889
3.889
2.891
3.903
2.891

-1.830
-1.399
-3.295
-1.160
-1.473
-0.440
-2.880
-2.591
-1.620
-2.433
-2.120
-2.805
-2.470
-2.703
-1.717
-2.737
-3.141
-1.375
-2.580
-0.730
-2.981
-0.380
-5.004
-4.679
-2.081
-2.420
-0.970
-3.930
-2.691
-5.336
-1.590
-3.060
-4.764
-1.836
-4.794
-2.500
-2.618
-1.204
-1.970
-0.986
-3.684
-4.197
-4.094
-4.182
-3.130
-3.195
-3.160
-3.580
-2.841
-2.600
-3.573
-4.110
-2.185
-3.310

36.54
46.87
48.14
60.59
44.51
75.93
66.38
53.63
47.2
70.01
27.7
36.12
52.28
22.87
11.79
72.6
26.57
50.03
8.74
55.84
61.08
79.59
26.41
17.51
7.61
80.04
40.31
15.97
63.49
14.05
26.56
35.58
16.67
14.96
11.78
8.57
71.84
34.53
24.59
46.55
35.3
18.4
19.2
15.9
43.8
59.5
41.2
23.9
33.9
43.1
11.4
17.0
59.2
56.3

0.96
1.05
0.78
1.65
0.32
6.79
1.29
4.97
0.98
1.76
0.52
2.39
2.08
0.60
0.17
0.86
4.32
0.55
0.29
5.13
2.91
4.17
0.17
0.27
0.13
0.22
0.19
0.23
5.94
1.34
0.48
0.76
0.69
1.23
0.34
1.52
1.07
1.04
6.36
1.04
1.15
0.48
0.59
0.42
0.44
1.62
0.54
0.41
0.51
0.82
0.29
0.31
0.38
1.19

ref

K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K07
K13
K13
K13
K13
K13
K13
K13
K13
K13
K13
K13
K13
K13
K13

Table 9. Solar linelist used to determine iron abundances abundances for GBS. References respectively correspond to: K07 : (Kurucz 2007), FMW
: (Fuhr et al. 1988), BK: (Bard & Kock 1994), BKK: (Bard et al. 1991), K13: (Kurucz 2013).

Article number, page 21 of 21

You might also like