Spectroscopic Abundances in the Open Cluster, NGC 6819

0.4 MB·English

by  Donald B. Lee-Brown

Accepted for publication in the Astronomical Journal 01/16/15
PreprinttypesetusingLATEXstyleemulateapjv.5/2/11
SPECTROSCOPIC ABUNDANCES IN
THE OPEN CLUSTER, NGC 6819
Donald B. Lee-Brown and Barbara J. Anthony-Twarog1
DepartmentofPhysicsandAstronomy,UniversityofKansas
Lawrence,KS66045-7582,USA
[email protected],[email protected]
Constantine P. Deliyannis1
DepartmentofAstronomy,IndianaUniversity
5
Bloomington,IN47405-7105,USA
1
0
2 Evan Rich2
DepartmentofPhysicsandAstronomy,UniversityofKansas
n
Lawrence,KS66045-7582,USA
a
and
J
0 Bruce A. Twarog
2 DepartmentofPhysicsandAstronomy,UniversityofKansas
Lawrence,KS66045-7582,USA

] Accepted for publication in the Astronomical Journal 01/16/15
R
S ABSTRACT
. High-dispersion spectra of 333 stars in the open cluster NGC 6819, obtained using the HYDRA
h
spectrograph on the WIYN 3.5m telescope, have been analyzed to determine the abundances of
p
iron and other metals from lines in the 400 ˚A region surrounding the Li 6708 ˚A line. Our spectra,
-
o with signal-to-noise per pixel (SNR) ranging from 60 to 300, span the luminosity range from the
r tip of the red giant branch to a point two magnitudes below the top of the cluster turnoff. We
t
s derive radial and rotational velocities for all stars, as well as [Fe/H] based on 17 iron lines, [Ca/H],
a [Si/H], and [Ni/H] in the 247 most probable, single members of the cluster. Input T estimates
eff
[
for model atmosphere analysis are provided by (B V) colors merged from several sources, with

1 individual reddening corrections applied to each star−relative to a cluster mean of E(B V) = 0.16.
−
v Extensive use is made of ROBOSPECT, an automatic equivalent width measurement program; its
3 effectiveness on large spectroscopic samples is discussed. From the sample of likely single members,
7 [Fe/H] = 0.03 0.06, where the error describes the median absolute deviation about the sample
− ±
9 median value, leading to internal precision for the cluster below 0.01 dex. The final uncertainty
4 in the cluster abundance is therefore dominated by external systematics due to the temperature
0 scale, surface gravity, and microturbulent velocity, leading to [Fe/H] = 0.02 0.02 for a sub-sample
. restricted to main sequence and turnoff stars. This result is consistent−with o±ur recent intermediate-
1
bandphotometricdeterminationofaslightlysubsolarabundanceforthiscluster. [Ca/Fe],[Si/Fe],and
0
5 [Ni/Fe]aredeterminedtobesolarwithintheuncertainties. NGC6819hasanabundancedistribution
1 typical of solar metallicity thin disk stars in the solar neighborhood.
: Subject headings: open clusters and associations : individual (NGC 6819), stars : abundances
v
i
X
1. INTRODUCTION clusterinourprogramtomaptheevolutionofLiamong
r
a NGC 6819 is an old (2.3 Gyr) open cluster whose fun- stars of varying mass as they evolve from the main se-
quencetothetipofthegiantbranchandbeyond. Delin-
damental properties have garnered increasing attention
eationofsuchamaprequiresreliableestimatesofstellar
inrecentyears. AsdetailedinAnthony-Twarog,Deliyan-
temperature, luminosity, and metallicity, especially iron,
nis, & Twarog (2014), (hereinafter Paper I), its age
and, indirectly, the cluster reddening, distance modulus
places it in a sparsely populated range of open cluster
and age.
ages, making it an invaluable testbed for stellar models
Data on individual cluster stars and the surrounding
ofstarsjustabovetheSun’smass. Itwaschosenasakey
field have expanded due to the cluster’s location in the
Kepler field, bringing the added potential for asteroseis-
WIYNOpenClusterStudyLXV
1Visiting Astronomer, Kitt Peak National Observatory, Na- mic insight into the structure of individual stars in a
tionalOpticalAstronomyObservatory,whichisoperatedbythe variety of evolutionary states (Gilliland et al. 2010).
Association of Universities for Research in Astronomy (AURA) Reliable radial velocities are available for a large sample
undercooperativeagreementwiththeNationalScienceFounda-
of stars extending well below the cluster turnoff (Hole et
tion.
2Current affiliation: Department of Physics & Astronomy, al. 2009)(hereinafterH09)andimprovedproper-motion
UniversityofOklahoma,Norman,OK73019,USA memberships by Platais et al. (2013) (hereinafter PL)
2 Lee-Brown, Anthony-Twarog, Deliyannis, Rich, Twarog
havesupplantedtheolderdataofSanders (1972). Com- certaintiesonourfinalresults,andplacesthedatawithin
prehensive broad-band (Rosvick & VandenBerg 1998; thecontextofbothstellarandgalacticevolution. Sec. 6
Kaliraietal. 2001;Yangetal. 2013)andintermediate- summarizes our conclusions.
band (Paper I) photometric surveys allow precise deter-
mination of each star’s position in the color-magnitude 2. SPECTROSCOPIC DATA
diagram (CMD), while definitively demonstrating (PL, Spectroscopic data were obtained for target stars in
PaperI)thattheclustersuffersfromvariableforeground NGC 6819 using the WIYN 3.5-meter telescope 3 and
reddening with a range of ∆E(B V) = 0.06 mag, HYDRA multi-object spectrograph over 13 nights from
−
not surprising given the cluster’s galactic coordinates September and October 2010, June 2011 and February
(l,b=74◦,+8◦) and a distance of 2.4 kpc from the Sun. 2013. The positions of the target stars within the clus-
Until recently, the weakest link in the discussion of the ter (V,B V) CMD are shown in Figure 1. Six con-
propertiesoftheclusteranditsstarshasbeenthecluster figurations−were designed to position fibers on a total
metallicity. AsdetailedinPaperI,metallicityestimation of 333 stars. The brightest configurations include stars
from cluster members has been plagued by small num- near V 11, while the fainter configurations reach to
ber statistics often coupled with incorrect assumptions V 16.∼5. Individual exposures ranged from 10 to 90
of uniform and/or anomalously high reddening, leading min∼utes, with accumulated totals of 2.5 to 4.5 hours for
toaconsensusviewthatNGC6819wasmetal-rich,with stars in the brightest configurations, 10 to 13 hours for
[Fe/H] +0.1. Precise intermediate-band photometry the configuration with stars of intermediate brightness,
∼
on the extended Stromgren system (Paper I) strongly and over 14 hours total for the faintest configurations.
contradicts this claim, producing [Fe/H]= 0.06 0.04 Ourspectracoverawavelengthrange 400˚Awidecen-
from 278 single, unevolved F stars. − ± tered on 6650 ˚A, with dispersion of 0.2∼˚A per pixel. Ex-
AlthoughafullerdiscussionoftheLithiumabundances aminationofThorium-Argonlampspectraindicatesthat
for our large sample in NGC 6819 is in preparation the line resolution comprises 2.5 pixels, yielding a spec-
(Deliyannis, Twarog, Lee-Brown & Anthony-Twarog tral resolution over 13,000.
2015), one preliminary result - the identification of a Li- Thedatawereprocessedusingstandardreductionrou-
rich giant star - was reported in Anthony-Twarog et al. tines in IRAF4. These routines included, in order of ap-
(2013). We employ a combination of spectral synthesis plication, bias subtraction, division by the averaged flat
andcurve-of-growthanalysistoconvertmeasuredequiv- field, dispersion correction through interpolation of the
alentwidthsfortheLilineat6707.8˚AtoLithiumabun- comparisonspectra,throughputcorrectionforindividual
dances, A(Li). Our curve-of-growth analysis method de- fibers using daytime sky exposures in the same config-
pends on subtracting an estimated contribution by an uration, and continuum normalization. After flat field
ironlineat6707.4˚Afromtheequivalentwidth,forwhich division and before the dispersion correction, the long-
accuratetemperatureandironabundanceinformationis exposure program images were cleaned of cosmic rays
required. Given the criticality of a precise estimate of using “L. A. Cosmic”5 (van Dokkum 2001). Real-time
the stellar Fe abundance for derivation of Li from spec- skysubtractionwasaccomplishedbyusingthedozensof
troscopy,especiallyamongthegiants,itwasdecidedthat fibers not assigned to stars and exposed to the sky for
all measurable lines within the wavelength range of our each integration. Composite spectra for each of the six
HYDRA high dispersion spectra of the Li 6708 ˚A re- configurationswereconstructedbyadditivecombination.
gion would be used to estimate the cluster metallicity The signal-to-noise ratio per pixel (SNR) may be esti-
for Fe and other elements. In addition to testing the mated two ways: first, by direct inspection of the spec-
photometricabundance,followingthepatternlaidoutin trawithinIRAF’sSPLOTutility,usingmeanvaluesand
our earlier investigations of the open clusters NGC 3680 r.m.s. scatterfromarelativelyline-freeregionor,second,
(Anthony-Twarogetal. 2009)andNGC6253(Anthony- by construction from output files of the ROBOSPECT
Twarogetal. 2010),suchdatacouldrevealtheeffectsof software suite. We found the ROBOSPECT values to
stellarevolutiononsurfaceabundancesthroughcompar- be entirely consistent with hand-measured SNR values
isonoftheredgiantstotheturnoffstars,whilepinpoint- and quote the ROBOSPECT-derived values in Table 1,
ing the place of the cluster within the global context of computed from the relatively line-free 6680-6694 ˚A re-
galactic chemical evolution. gion. SNR estimates reflect the statistics characterizing
The layout of the paper is as follows. Sec. 2 details the summed composite spectra.
the spectroscopic observations and reduction of images
touniformlywavelength-calibratedandcontinuum-fitted 3. STELLAR PROPERTIES
spectra. Sec. 3 explains the collation of radial-velocity, 3.1. Radial Velocities and Proper Motions
photometric,proper-motion,andreddeningdatausedto
An initial spectroscopic sample of probable cluster
define the crucial temperatures, surface gravities, and
members was taken from the valuable radial-velocity
microturbulent velocities for single-star cluster mem-
bersthat,togetherwithappropriatemodelatmospheres, 3 TheWIYNObservatoryisajointfacilityoftheUniversityof
translatethespectroscopicmeasurestoabundances. Sec. Wisconsin-Madison, Indiana University, Yale University, and the
4 supplies a detailed examination of ROBOSPECT, an NationalOpticalAstronomyObservatory.
automatedline-measurementprogramusedforthisstudy 4 IRAF is distributed by the National Optical Astronomy Ob-
servatory,whichisoperatedbytheAssociationofUniversitiesfor
which makes internally consistent and rapid abundance
Research in Astronomy, Inc., under cooperative agreement with
estimationforhundredsofstarsfeasible. Sec. 5presents theNationalScienceFoundation.
theabundanceresultsforindividualstarsandthecluster 5 http://www.astro.yale.edu/dokkum/lacosmic/, an IRAF
as a whole, delineating the impact of the parametric un- script developed by P. van Dokkum (van Dokkum 2001); spec-
troscopicversion.
Spectroscopic Abundances in NGC 6819 3
survey of NGC 6819 by H09. All stars brighter than 3.2. Effective Temperature, Surface Gravity, and
V 16.75 with radial-velocity membership probabili- Microturbulent Velocity
∼
ties greater than 50% were identified as candidates for
In keeping with our approach to spectroscopic abun-
the present study. Stars classed as double-lined spec-
dance determination for previous clusters in this pro-
troscopic binaries were eliminated; single-lined systems
gram, our default scheme for determining model atmo-
wereretainedsincetheexistenceofthecompanionwould
sphere input temperatures is based upon photometric
have minimal impact on line measurement. Stars were
color, specifically B V. Four sources of broad-band
not eliminated based upon their position in the CMD −
BV dataexist. Asafirststep,wemergedtheCCDpho-
to avoid biasing the sample against stars undergoing
tometry presented by H09 as PHOT98 and PHOT03.
anomalous evolution.
From923starsbrighterthanV =16.7,excluding5stars
Individual stellar radial velocities were derived from
with absolute residuals greater than 0.10 mag, the mean
each summed composite spectrum utilizing the Fourier-
offsets, in the sense (PHOT98 - PHOT03), in V and
transform,cross-correlationfacilityFXCORinIRAF.In
B V are 0.005 0.022 and +0.003 0.022, respec-
this utility, program stars are compared to stellar tem- − − ± ±
tively. While the dispersion in both sets of residuals was
plates of similar effective temperature (T ) over the
eff satisfactory, further analysis revealed that a large frac-
wavelength range from 6575 ˚A to 6790 ˚A, as well as a tion of the scatter in V is the result of a nonlinear ra-
narrower region in the vicinity of Hα alone. Typical un- dialgradientamongtheresiduals, reaching∆V = 0.05
certainties in the individual radial velocities were esti- mag for stars near the core of the cluster but +0.0−1 for
matedat1.15km/sec. TheFXCORutilityalsoprovides stars in the outer regions of the frame. Milliman et al.
measures of the line widths within each spectrum, from (2014) have independently discovered the same effect
whichrotationalvelocitiesmaybeinferred. Observations in a comparison between the VI photometry of Yang
of a pair of radial-velocity standards were obtained dur- et al. (2013) and PHOT03, implying that the primary
ingeachrunandprocessedusingthesameprocedureap- source of the trend in the current comparison must lie
plicable to the cluster. Comparison of the velocity zero- withthePHOT03database. Additionalevidenceinsup-
pointsusingstandardvaluesfromtheGeneralCatalogof port of this conclusion comes from a direct comparison
Radial Velocities (Wilson 1953) allowed transformation between the V magnitudes from Yang et al. (2013)
of the cluster data to the standard system, within the and the intermediate-band data from Paper I where the
uncertainties of the measurements. residuals do not exhibit a radial dependence. More im-
Asafirstcheckonourspectroscopicdata,wecancom- portant from our perspective is the absence of a radial
pare our measured radial velocities with those of H09 to trend among the residuals in B V between the two
identify discrepant stars or long-term variables. Elimi- samples, PHOT98 and PHOT03. −Application of a small
nating 29 stars classed as spectroscopic binary members colorterm, (B V) =0.986(B V) +0.006, reduces
03 98
ofNGC6819,theremaining304singlestarshaveamean thescatterinth−eresidualsto 0.0−21mag. ThePHOT98
radial velocity of 2.65 1.36 (s.d.) km/sec. The same B V data were converted to±the PHOT03 system and
samplefromH09hasam±eanvelocityof2.38 0.99(s.d.) ave−raged for stars common to both samples.
±
km/sec. The dispersion in the residuals between these Next, the composite H09 data were transformed to
samplesis1.06km/sec;thepredicteddispersionfromthe ouradoptedstandardofRosvick&VandenBerg (1998).
quoted errors for each star is 1.19 km/sec. The residu- From470starswithV brighterthan16.5commontothe
als for only three stars (WOCS 1007, 2016 and 56018) two datasets, removing 10 stars with residuals in B V
fall more than three sigma from the mean; one of these, greater than 0.05 mag, the mean residual, in the se−nse
WOCS 2016, has unusually broad lines (presumably due (RV-H09), is +0.006 0.014 mag. A weak color de-
to high rotational velocity) which leads to a larger than pendence among the r±esiduals was found, (B V) =
RV
average uncertainty in the final radial-velocity estimate. 1.007(B V) , and applied prior to the me−rger of the
H09
By contrast, the sample of 29 spectroscopic binaries has two data−bases, eliminating the small zero-point offset.
ameanclustervelocityof3.45km/secandadispersionof Finally, a transformation was derived between the
9.8km/sec;thesamestarsfromH09exhibitadispersion B V indices of Kalirai et al. (2001) and the previ-
of9.78km/sec. Weconcludethatallsinglestarmembers ous−ly merged B V data on the Rosvick & VandenBerg
as classified by H09 are confirmed as such by our data. (1998) system;−comparisons of the V mags of Rosvick
Table 1 contains the derived mean radial and estimated & VandenBerg (1998) and Kalirai et al. (2001) have
rotational velocities for each star in our survey. already been discussed in Paper I.
At the start of our program, the only NGC 6819 Eliminating 11 stars with absolute residuals greater
proper-motion study available was that of Sanders than 0.1 mag and applying a small color term, (B
(1972), which proved inadequate for reliable identifica- V) =1.025(B V) 0.018,846starsbrightertha−n
RV KA
tion of probable members, particularly at the fainter V =16.5exhibit−ameanr−esidualinB V of0.000 0.018
limit of interest. Fortunately, the comprehensive sur- mag. − ±
veybyPLcoverstheappropriaterangeinbothareaand Allstarswithabsoluteresidualsgreaterthan0.05mag
depth, generating probabilities for all stars in our sam- wereindividuallychecked. IfB V estimateswereavail-
ple except one. Of 332 stars selected via radial velocity, able from more than two sour−ces and one source was
59 have proper-motion membership probabilities below clearly the origin of the discrepancy, that value was
50%, 43 of which are in single digits. The PL values for dropped. If only two sources of B V existed, an in-
individual stars may be found in Table 1. The different dependent check on the predicted B− V was attempted
stellar categories are identified in the CMD, Fig. 1. usingthepublishedV I dataofYan−getal. (2013). In
−
caseswherenoresolutionofthediscrepancywaspossible,
4 Lee-Brown, Anthony-Twarog, Deliyannis, Rich, Twarog
the averaged value from all sources was retained. Aver- 4. SPECTROSCOPIC PROCESSING WITH
aged B V indices from the four transformed primary ROBOSPECT
−
data sets are given in Table 1 along with the number of
With the expansion in spectroscopic samples from
sources and each star’s estimated individual reddening
a few dozen stars in previous cluster work (Anthony-
correction.
Twarog et al. 2009, 2010) to a few hundred in this
As noted earlier, PL identified and mapped variable
and future analyses, manual measurement of equivalent
reddening across the field of NGC 6819, a result con-
widths (EW) for individual lines is now prohibitive. To
firmed in Paper I. We used the map of individual red-
overcome this obstacle, the automated line-measuring
dening values derived by PL to estimate by spatial in-
program, ROBOSPECT (Waters & Hollek 2013) (here-
terpolation the degree of reddening affecting each of the
inafter WH), has been utilized. For full details, the
stars in our spectroscopic sample. Individual reddening
reader is referred to WH; a brief outline of the pro-
estimates, with a range of 0.033 about our adopted
gramoperationandourprocedureisprovidedhere. RO-
±
mean value of E(B V) = 0.16, were applied to each
BOSPECT measures EW by first determining the con-
−
star’s (B V) color. A temperature for each star based
tinuum level and constructing a noise profile across the
−
on its dereddened (B V) color was then derived using
wavelength range of interest using an iterative process.
−
two primary color-temperature calibrations.
With the initial continuum and noise levels established,
For dwarfs, we continue to use a calibration (Deliyan-
spectral lines are tagged at locations specified by a user-
nis,Steinhauer&Jeffries 2002)consistentwithprevious
supplied line list, and other potential lines are automat-
spectroscopic studies by this group, namely:
ically identified based on significant deviations of the lo-
T = 8575 5222.7(B V) +1380.92(B V)2 +
eff − − 0 − 0 calspectrumfromthecontinuum. Fortheautomaticline
701.7(B V) [[Fe/H] 0.15] K.
0 identification, the significance threshold is user-specified
− −
For giants, the T -color-[Fe/H] calibration of Ram´ırez
eff and based on the current noise solution. After lines
&Mel´endez (2005)wasused. Wenotethatfor13stars,
are identified though these two processes, they are sub-
it was possible to obtain temperatures from both color-
tracted from the spectrum and the continuum process is
temperature calibrations. Temperatures derived from
repeated and the noise level refined. Through multiple
the Deliyannis, Steinhauer & Jeffries (2002) calibration
iterations, the best-fit continuum and individual line so-
are 41 6 K higher for these mostly subgiant stars.
lutionsarereached. Forthisstudy,eachspectrumwasin-
±
Surface gravity estimates (log g) were obtained by di-
dividually corrected in ROBOSPECT for radial velocity
rect comparison of V magnitudes and B V colors for
and run through 25 iterations of continuum fitting and
oursampleof333starstoisochronesfrom−theY2 compi-
line estimation using a gaussian line profile with three-
lation (Demarque et al. 2004), constructed for a scaled
sigma automatic line identification and no least-squares
solar composition with [Fe/H] = -0.06 and an age of 2.3
line deblending. All other parameters for the program
Gyr, essentially the same as the comparison presented
were set to default values.
in Paper I. The isochrone’s predicted magnitudes and
An issue of obvious concern with any automated pro-
colors were adjusted to match the cluster’s reddening,
cedureisthepotentialforunreliableEWduetolowSNR
E(B V)=0.16 and apparent distance modulus, 12.40.
spectra, line blending, or inaccurate radial-velocity cor-
−
Forstarsthatappeartobebluestragglers, surfacegrav-
rection. With interactive measurement, such issues can
ities were estimated by comparing photometric informa-
be flagged and corrected or eliminated but, with RO-
tion to a grid of younger isochrones of similar composi-
BOSPECT, flawed EW measures propagate to followup
tion.
steps in the abundance determination process, poten-
We are fortunate to have independent confirmation of
tially distorting the final results. It is expected that the
theloggvaluesforthegiantbranchstars,thankstotheir
uncertainty in EW will increase with decreasing SNR,
status as Kepler objects of interest. We compared log g
potentially reaching a level where the subsequent uncer-
values from our isochrone match to log g values inferred
taintyinastar’scalculatedabundancerendersituseless.
using asteroseismology from Kepler data, as presented
Similarly, we can expect the EW uncertainty to increase
by Basu et al. (2011) for 21 giants. The seismological
as individual lines suffer increased blending from neigh-
gravitiesareinremarkableagreementwiththeisochrone-
boringlines. Finally,asignificanterrorinradialvelocity
inferred gravities, differing by an insignificant amount,
will cause ROBOSPECT to misidentify the wavelength
0.02 0.04 where the listed error is the standard devi-
of line centers. To ensure that our abundances are mini-
− ±
ation.
mallyaffectedbytheseissues,avarietyoftestswerecon-
Inputestimatesforthemicroturbulentvelocityparam-
ducted to determine if our line list and ROBOSPECT
eter were constructed using various prescriptions. For
input parameters lead to robust results over the range
dwarfs within appropriate limits of T and log g, the
eff in SNR and radial-velocity uncertainties spanned by our
formula of Edvardsson et al. (1993) was used; a simi-
spectra. SinceROBOSPECTwillbeusedinfutureanal-
lar formulation by Ram´ırez, Allende Prieto, & Lambert
yses with similar input parameters, the level of detail
(2013) extends to slightly lower temperatures and grav-
supplied is somewhat greater than usual.
ities and was used for some stars. For giants, a gravity-
Thelinelistusedduringthisstudywasconstructedby
dependent formula, V = 2.0 0.2 log g, was used. For
t visually identifying relatively isolated, medium-strength
−
somesubgiantsandcandidatebluestragglers,nosuitable
(EW 10 to 200 m˚A) lines in the solar spectrum, using
formula for Vt was found other than the purely gravity- the so∼lar atlas of Wallace, Hinkle, & Livingston (1998)
dependent expression employed by the SDSS collabora-
as a guide. Atomic information for each line was re-
tion in their DR10 data release and discussion of the
trieved from the VALD database (Kupka et al. 1999).
abundance analysis pipeline for APOGEE spectra (Ahn
Log gf values given by VALD were then modified to
et al. 2014).
Spectroscopic Abundances in NGC 6819 5
force-fit abundances from ROBOSPECT analysis of so- ing) is similar in solar spectra and in the spectra of pro-
lar spectrum EW to abundances in the 2010 version of gram stars, and is not a function of SNR to any de-
theabundanceanalysissoftware,MOOG(Sneden 1973). tectable extent, so we anticipate no effect on our abun-
We used day-time sky spectra, taken as part of the cali- dance estimates based on differential analysis with re-
brationdatasets,assolarspectraforthisnormalization. spect to the sun.
MOOG analysis was conducted using the abfind driver Using our measured radial velocities and constructed
and a solar model (Kurucz 1995) with the following at- line list, EWs were measured for all 333 stars in our
mospheric parameters: (T ,log g,v ,[Fe/H]) = (5770 observing program. We discarded negative EW (RO-
eff t
K, 4.40, 1.14 km/s, 0.00). BOSPECT’s designation of emission lines) due to non-
Our final line list contains 22 lines of interest (17 Fe, 3 convergent fitting solutions, artifacts of cosmic ray re-
Ni, 1Ca, 1Si), presentedalongwiththerelevantatomic moval,largenoisespikes,ornonexistentlinesinthemea-
parametersinTable2. Tominimizetheimpactofblend- suredspectrum. Approximately4%ofthemeasuredlines
ing,allselectedlinesareatleast0.5˚Afromanylinewith of interest returned negative values.
EWgreaterthan5m˚Ainthesolarspectrum. Tofurther
minimize measurement distortions in spectra with lower 5. ABUNDANCE DETERMINATIONS
SNR, the line list input for ROBOSPECT included any
Modelatmosphereswereconstructedforeachprogram
significant feature within 2 ˚A of a line of interest. EW
starusingthegridofKurucz (1995)andtheinputT ,
forthesefeaturesweredisregardedduringtheabundance eff
log g and microturbulent velocity values listed in Table
analysis. Theextraneousentrieswereaddedafterobserv-
1. As a reminder, each star’s T estimate is based
ing unusual amounts of scatter among the EW for lines eff
on its B V color, individually dereddened based on
with close neighbors, particularly at lower SNR. −
its spatial position within the field of the cluster. For
To evaluate the performance of our line list and to
any star that is not actually a cluster member, this spa-
quantify the effects of a spectrum’s SNR on EW values,
tiallyinterpolatedreddeningcorrectionmaybespurious;
we tested ROBOSPECT on 100 solar spectra, 25 each
the gravity estimate based on an isochrone matched to
with mean SNR of 160, 130, 95 and 70. These represent
the cluster CMD almost certainly would be as well. For
typical SNR for our program stars (see Fig. 2). We note
suspected binaries, the color might reflect the combined
that WH include an evaluation of ROBOSPECT’s per-
lightofthetwostarsandaffecttheassignedtemperature.
formance over a wide range of line strengths and SNR;
WhilethismightnotbealargeissueforSB1, afewcan-
our test results are consistent with theirs. The mean
didate SB2 were highlighted by examination of spectra:
standarddeviationsinEWforourfourtestsampleswere
WOCS7009,26007,35025,9004and60021. Asaconser-
4.2, 4.7, 7.5, and 11.1 m˚A, in order of decreasing mean
vative approach, we restricted our analysis to stars with
SNR. The uncertainty this scatter introduces into our
membership probablity 50% (PL) and no evidence of
abundanceresultsissmallrelativetotheuncertaintydue ≥
binarity. Wealsoexcludedfromourfinalanalysisseveral
toexternalsystematics. AtallSNRstested,weobserved
stars with (B V) greater than 1.35, for which severe
0
nosignificantcorrelationbetweenthemeanEWofaline −
line-blending makes reliable measurement of equivalent
and its associated uncertainty. Taken together, these re-
widths problematic. This left us with a sample of 247
sults indicate that our line list is robust over the range
single, member stars.
in mean EW and SNR of interest.
Each star’s measured equivalent widths and model
The effect of potential errors in the determined radial serve as input to the abfind routine of MOOG to pro-
velocity on EW measurement was evaluated by artifi-
duce individual [A/H] estimates for each measured line
cially increasing and decreasing our best estimate of ra-
for each star. A large volume of spectroscopic informa-
dialvelocitybyaflatamountinthesetof100solarspec-
tion is available for interpretation from over 7000 mea-
tra. For a shift of 2.5 km/s, corresponding to a radial-
sured lines, with up to 17 [Fe/H] values for each star.
±
velocityuncertaintygreaterthanthatof99%ofourpro-
Before constructing median [Fe/H] values for each star
gram stars, the mean deviation in EW between the true
and for the sample as a whole, the following filters were
and adjusted spectra was 0.7 m˚A. When the shift was applied to the individual abundance estimates: abun-
−
increased to 5.0 km/s, corresponding to a radial-velocity dances based on equivalent width measurements smaller
uncertainty greater than that of all program stars (in- than three times the expected error in equivalent width
cluding the rapid rotator WOCS 2016), the mean devi- for each star based on the star’s SNR, were not con-
ation in EW increased to 3.5 m˚A. In both cases, RO- sidered, nor were equivalent widths for very strong lines
−
BOSPECTcorrectlyidentifiedessentiallythesamenum- (EW 200m˚A).Wenotethattheconsequencesofusing
berofentriesfromthelinelistasithadwiththecorrect 200 or≥250 m˚A as the upper limit for EW are negligible.
radial velocity. We conclude that the effect of radial-
Estimated abundances from individual lines deviant by
velocity error on our abundance results is negligible.
more than 1 dex from solar were not included in the de-
Finally,wecomparedROBOSPECTresultswithman-
termination of [Fe/H] for each star.
ual EW measurements for 18 random program stars us-
Figures 2 and 3 show the spread of stellar [Fe/H] esti-
ing a 10-line subset of our line list. Manually obtained
mates for each star as a function of the individual spec-
EWs were measured using IRAF’s SPLOT tool. RO-
trum SNR and unreddened B V color for each star.
BOSPECT’s results largely agree with the manual val- −
A few outliers with SNR below 50 are obvious, as are a
ues,withROBOSPECT’sEWonaveragebeing7.7 4.2
few of the brightest (and coolest) stars with higher than
m˚A (s.d.) lower than the manual results. We note±that typical abundance. Figure 3 suggests that abundances
this consistent underestimate of EW by ROBOSPECT forthecoolerstarsappeartobeslightlybelowtherange
(or, over-estimate using SPLOT and interactive measur- for the stars nearer the turnoff. We further divided our
6 Lee-Brown, Anthony-Twarog, Deliyannis, Rich, Twarog
sample of 247 likely members into stars bluer than and Table 3 includes the median [Fe/H], along with the
redder than (B V) = 0.60 to separate out the cooler number of iron lines included in the estimate, as well
0
−
and more evolved stars and re-examined the effect of fil- as estimates of [Ca/H], [Si/H] and [Ni/H] for 333 stars.
ters that precede the construction of each star’s median Stars not part of the set of 247 single, probable mem-
[Fe/H] abundance. Do some of our 17 iron lines produce bersstarsareflaggedwithanoteindicatingnon-member
abundances in dwarfs that are unacceptably noisy? of binary status. The statistic that accompanies each
By examining the effect of imposing different criteria [Fe/H] estimate is the MAD. We note that the 29 stars
for excluding small EW measures, we were able to high- not included in these estimates of cluster [Fe/H] due
light several iron lines, which for dwarfs, lead to biased to evidence of binarity, yield a similar median [Fe/H]
[Fe/H] values in the sense that the abundance values for value if the suspected double-lined spectroscopic bina-
these lines in dwarf stars are higher than the ensemble ries are excluded. For member stars which are mostly
median value by several tenths. The reason for this is SB1, [Fe/H]= 0.05 0.06, consistent with the more
− ±
clear; if the median EW for a particular absorption line conservative sample that excludes binaries.
is small, imposing a 3-sigma cut will preferentially pre- Conversions of photometric parameters to T and
eff
serve stars in the sample with larger EW, and conse- V are fundamentally discontinuous for dwarf and giant
t
quentlylarger[Fe/H]values. Weexcludedlinesindwarfs stars; to assess the effects of parameter choices on the
forwhichimposing3-sigmacutschangedthemedianEW derived abundances, each star was reanalyzed with at-
forthelinebymorethan25%incomparisonwithamore mospheric parameters incremented and decremented by
generous1-sigmacut. Asareminder,theselinesaremea- thefollowingamounts: 100KforT ,0.25forlogg and
eff
sured in the solar spectrum at considerably higher SNR 0.25 km/sec for V . A summary of the extensive results
t
so we may still place considerable confidence in the log is shown in Figure 5 where once again median values
gf values determined from those lines. and variances indicated by median absolute deviations
As this still leaves information from several thousand are compiled and shown for stars in different color bins.
separateironlinemeasurements, wewereabletofurther This figure illustrates some well-known dependences of
explore the consequences of these filters. A diagnostic abundancedeterminationsonstellarparameters,namely
diagramispresentedinFigure4,demonstratingthelack thatmainsequencestarabundancesareverysensitiveto
of a trend in the median [Fe/H] for each of our 17 iron T estimates,whileevolvedstarsdemonstrateahigher
eff
lines with wavelength, where the median value is con- sensitivity to surface gravity and consequently to micro-
structed from the full sample of 247 probable-member, turbulent velocity estimation schemes.
single stars for lines denoted by black symbols; the six Although thediscrepancy between the giant andnear-
lines denoted by red symbols represent measures for gi- turnoff samples is small, it is of value to explore modest
ant stars only. The upper panel of this figure shows the parameter changes that could eliminate it. Reference
MAD (median absolute deviation) statistic for each line, to Figure 5 suggests that errors in log g have relatively
a robust estimator of variance, constructed by compar- little impact; we remind the reader that the log g val-
ingthe[Fe/H]valueforeachstartothemedianvaluefor ues employed are consistent with values suggested by
that particular absorption line. asteroseismic analyses of NGC 6819 giants. The rela-
For the subset of 247 single member stars bluer than tion used for microturbulent velocities for the evolved
(B V) <1.35,themedian[Fe/H]valueis 0.03 0.06, stars is more tenuous; a fairly modest decrement of 0.2
0
− − ±
with abundances for the other elements that are entirely km/sec would raise the giant abundances to match the
consistent with solar: 0.01 0.10, 0.01 0.06 and [Fe/H] representative of the near-turnoff stars. Finally,
− ± − ±
0.00 0.06for[Ca/H],[Si/H]and[Ni/H],respectively. Of increasedabundancesforbothregionsoftheCMDwould
±
these247stars, 200aremainsequenceandturnoffstars, accompany higher temperatures, though a simple incre-
withamedianironabundanceof 0.02 0.05, whilethe ment to the mean cluster reddening will have differing
− ±
remaining47coolandevolvedstarsshowalower[Fe/H], effects for the two classes of stars, both because of the
0.09 0.05. We note that this apparent discrepancy slightcolor-dependenceoftheeffectontheabundanceas
− ±
between giant and dwarf [Fe/H] values is nearly identi- shown in Figure 5 and the larger increment implied for
cal to that found in NGC 3680 (Anthony-Twarog et al. warmer stars resulting from the same higher reddening.
2009). To define the final estimate of [Fe/H] for the cluster,
For all of these quantities, the indicated error is the we default to the dwarfs, where the sensitivity to the
MAD statistic; for a large sample size drawn from a changes in the three key input parameters (T , V , log
eff t
normally distributed population, the normalized MAD g)iscollectivelyminimized,thecolorrangeissmall,and
statistic, MADN = 1.48 MAD, approximates well the theparametricdifferencesbetweentheprogramstarsand
samplestandarddeviation,permittinganestimateofthe the solar reference spectra are significantly smaller than
traditional standard error of the mean through division for the evolved stars. For log g and V we adopt 0.1 dex
t
by (N 1)1/2. For the sample of 247 single, probable- and 0.2 km/sec as generous estimates for the uncertain-
membe−r stars, the MAD statistic implies a value for ties. From Fig. 5, these translate into 0.003 and 0.008
MADN = 0.09; as this statistic is a good proxy for a dex,respectively. Asusual,themoresensitiveparameter
standard deviation, a standard error of the mean of 0.01 is the temperature, tied to the B V color transforma-
−
is implied. These values can be confirmed by computing tion,whichhastwoprimarycomponents,theuncertainty
averages, standard deviations and s.e.m. values in linear inthemeanclusterreddening,E(B V),andtheuncer-
−
space; for these stars, the average [Fe/H], computed in tainty in the photometric zero point. From Paper I, the
linear space and then converted to logarithmic values, is photometric internal and external errors combined lead
0.02+0.07 (standard deviations). to an uncertainty of 0.007 in E(B V), while the pho-
− −0.13 tometric merger of the four broad−-band surveys above
Spectroscopic Abundances in NGC 6819 7
implies an uncertainty in the B V zero-point of 0.004 sample, we can minimize the impact of these parame-
−
mag. Combined, the (B V) scale is reliable to 0.008 ters by looking only at the 200 unevolved stars since the
0
−
mag. On our temperature scale, this translates to ap- red giant abundances have a higher degree of sensitiv-
proximately 32 K or, from Fig. 5, 0.012 dex in [Fe/H]. ity to microturbulent velocity. From these stars alone,
Combining all three parameters, one arrives at a final [Fe/H] = 0.02 0.02, where the uncertainty includes
− ±
external uncertainty of 0.015 dex, much larger than the both internal and external errors. The lower metallicity
internalprecisionofthespectroscopicaverage. From200 compared to the commonly adopted value of [Fe/H] =
turnoff stars, the mean [Fe/H] for NGC 6819 is deter- +0.09 (Bragaglia et al. 2001) continues a trend first in-
mined to be 0.02 0.02. dicatedbytheintermediate-bandphotometryofPaperI,
− ±
Among the prior determinations of NGC 6819’s heavy whereprecisionphotometrywasusedtoderiveareliable
element abundance, the result published by Bragaglia et cluster reddening and confirm the existence of variable
al. (2001) has anchored a string of moderately super- reddening across the face of the cluster (PL), leading to
solar values. Their study of three clump giants uti- subsolar [Fe/H] from either m or hk indices.
1
lized the high dispersion (R 40,000) spectrograph on Within the context of Galactic evolution, NGC 6819
∼
the Galileo Italian National Telescope (TNG). Though exhibits no features which distinguish it from a typical
the sample size was small, their analysis incorporated thin-diskpopulationformedwithinthelast5Gyrswithin
hundreds of lines, including nearly 100 iron lines, six 1 kpc of the Sun’s galactocentric distance. Its [Fe/H]
of which were Fe II lines to permit an accurate ver- placesitatthemeanforopenclustersatthesolargalac-
ification of surface gravity. The reddening value of tocentric radius, a distribution which exhibits a scatter
E(B V) = 0.14 they adopted, however, was some- atonlythe 0.1dexlevel(Twarog,Ashman,&Anthony-
− ±
what smaller than the mean value derived in Paper I Twarog 1997). ItsabundanceratiosrelativetoFeareall
andutilizedhere,E(B V)=0.16. Werepeatedparam- consistent with solar, a pattern which holds true within
−
eter determinations for the three giants studied by Bra- the scatter for Ni, Ca, and Si for stars with [Fe/H] 0.0,
∼
gaglia et al. (2001), including independent determina- whethertheyareclassedasmembersofthethinorthick
tions of spatially-dependent reddening and T values, disk population (see Hinkel et al. (2014) and references
eff
loggvaluesfromisochronecomparisonsandmicroturbu- therein). NGC 6819 forms a rich and ideal link for tests
lentvelocityvaluesbasedonthesurfacegravities. Using of stellar evolution theory midway in age between NGC
the estimates of ∆[Fe/H] for each of the principal at- 5822(Carraroetal. 2011)andM67(Sandquist 2004),a
mospheric parameters cited by Bragaglia et al. (2001), characteristicwhichwillprovevaluableintheanalysisof
their equivalent width analysis would yield an average the evolution of Li for intermediate mass stars evolving
abundance 0.07 dex lower with our atmospheric param- off the main sequence and up the giant branch.
eters, lowering their [Fe/H] for the cluster from +0.09 to
+0.02 0.03, essentially solar.
± The authors express their thanks to Imants Platais
for supplying additional data to compute spatially-
6. DISCUSSION AND CONCLUSIONS
dependentreddeningestimatesinthecluster. Drs. Ryan
High dispersion spectra of 333 potential members of Maderak and Jeff Cummings assisted as visiting as-
the old open cluster, NGC 6819, have been processed tronomers at KPNO for runs in 2010, supplying sup-
and analyzed. From the wavelength region near Li 6708 port and advice for the spectroscopic data collection.
˚A, abundances have been obtained for Fe, Ni, Ca, and NSF support for this project was provided to ER while
Si for 247 highly probable, single-star members ranging he was an undergraduate at the University of Kansas
from the tip of the giant branch to well below the top throughNSFgrantAST-0850564viatheCSUUREREU
of the CMD turnoff. Using up to 17 Fe lines per star, program at San Diego State University, supervised by
the cluster exhibits a slightly subsolar [Fe/H], a conclu- Eric Sandquist. NSF support for this project to BJAT,
sionwhichremainsunchangedifthedataaresortedinto BAT and DLB through NSF grant AST-1211621, and
evolved stars (47) or turnoff/unevolved stars (200). Due to CPD through NSF grant AST-1211699 is gratefully
to the cumulative impact of a few thousand Fe lines, the acknowledged. Extensive use was made of the WEBDA
standard error of the mean for [Fe/H] tied to internal 6 database operated at the Department of Theoretical
errors is below 0.01 dex, leaving the systematics of the PhysicsandAstrophysicsoftheMasarykUniversity, the
temperature, log g, and microturbulent velocity scales TOPCATsuite7 theViennaVALD,oftheMOOGsuite
as the overwhelming source of uncertainty in the final of spectroscopic analysis software, and of ROBOSPECT
cluster metallicity. Once again, thanks to the size of the software.
REFERENCES
Ahn,C.P.,Alexandroff,R.,AllendePrieto,C.,etal.2014,ApJS, Anthony-Twarog,B.J.,Deliyannis,C.P.,Rich,E.&Twarog,B.
211,17 A.2013,ApJ,767,19
Anthony-Twarog,B.J.,Deliyannis,C.P.,Twarog,B.A., Basu,S.,Grundahl,F.,Stello,D.,et.al.2011,ApJ,729,L10
Croxall,K.V.,&Cummings,J.2009,AJ,138,1171 Bragaglia,A.,Carretta,E.,Gratton,R.G.,etal.2001,AJ,121,
Anthony-Twarog,B.J.,Deliyannis,C.P.,Twarog,B.A., 327
Cummings,J.D.,&Maderak,R.M.2010,AJ,139,2034 Carraro,G.,Anthony-Twarog,B.J.,Costa,E.,Jones,B.J.&
Anthony-Twarog,B.J.,Deliyannis,C.P.&Twarog,B.A.2014, Twarog,B.A.2011,AJ,142,127
AJ,148,151(PaperI) Deliyannis,C.P.,Twarog,B.A.,Lee-Brown,D.B.&
Anthony-Twarog,B.J.2015,inpreparation.
Deliyannis,C.P.,Steinhauer,A.,&Jeffries,R.D.2002,ApJ,
577,L39
6 http://www.univie.ac.at/webda Demarque,P.,Woo,J.-H.,Kim,Y.-C.,&Yi,S.K.2004,ApJS,
7 http://www.star.bristol.ac.uk/mbt/topcat/ 155,667(Y2)
8 Lee-Brown, Anthony-Twarog, Deliyannis, Rich, Twarog
Edvardsson,B.,Andersen,J.,Gustafsson,B.etal.1993,A&A, Ram´ırez,I.,AllendePrieto,C.&Lambert,D.L.2013,ApJ,764,
275,101 78
Gilliland,R.L.,Brown,T.M.,Christensen-Dalsgaard,J.,etal. Rosvick,J.,&VandenBerg,D.A.1998,AJ,115,1516
2010,PASP,122,131 Sanders,W.L.1972,A&A,19,155
Hinkel,N.R.,Timmes,F.X.,Young,P.A.,Pagano,M.D.,& Sandquist,E.L.2004,MNRAS,347101
Turnbull,M.C.2014,AJ,148,54 Sneden,C.1973,ApJ,184,839
Hole,K.T.,Geller,A.M.,Mathieu,R.D.,etal.2009,AJ,138, Twarog,B.A.,Ashman,K.,&Anthony-Twarog,B.J.1997,AJ,
159(H09) 114,2556
Kalirai,J.S.,Richer,H.B.,Fahlman,G.G.,etal.2001,AJ,122, vanDokkum,P.G.2001,PASP,113,1420
266 Wallace,L,Hinkle,K.&Livingston,W.1998,An atlas of the
Kupka,F.,Piskunov,N.,Ryabchikova,T.A.,Stempels,H.C.,& spectrum of the solar photosphere from 13,500 to 28,000 cm 1
Weiss,W.W.1999,A&A,138,119 (3570 to 7405 ˚A),NSOTechnicalReport#98-001. −
Kurucz,R.L.1995,inIAUSymp.149,TheStellarPopulationsof
Galaxies,ed.B.Barbuy&A.Renzini(Dordrecht:Kluwer),225 Waters,C.Z.,&Hollek,J.K.2013,PASP,125,1164(WH)
Milliman,K.E.,Mathieu,R.D.,Geller,A.M.,etal.2014,AJ, Wilson,R.E.,1953,CarnegieInst.WashingtonD.C.Publ.601
148,38 Yang,S.-C.,Sarajedini,A.,Deliyannis,C.P.,etal.2013,ApJ,
Platais,I.,Gosnell,N.M.,Meibom,S.,etal.2013,AJ,146,43 762,3
(PL)
Ram´ırez,I.&Mel´endez,J.2005,ApJ,626,465
Spectroscopic Abundances in NGC 6819 9
Fig. 1.— Color-magnitude diagram of spectroscopic sample. Blue open circles denote single, probable member stars. The few large
magentaasterisksmarkstarswithverycoolcolors,whileblackcrossesmarkprobablenonmembersandredfilledsquaresdenoteprobable
memberspectroscopicbinaries.
10 Lee-Brown, Anthony-Twarog, Deliyannis, Rich, Twarog
Fig. 2.—For247single,probablememberstarsinNGC6819,themedianironabundancerelativetosolarisshownasafunctionofthe
spectrumSNR.