1997pasp 109 883R

Publications of the Astronomical Society of the Pacific
109: 883-906, 1997 August
Galactic Globular-Cluster Metallicity Scale from the Ca II

Triplet. I. Catalog
Glen A. Rutledge, James E. Hesser,1 and Peter B. Stetson

National Research Council of Canada, Herzberg Institute of Astrophysics, Dominion Astrophysical Observatory,
5071 West Saanich Road, RR5, Victoria BC V8X 4M6, Canada
Electronic mail; [email protected]
Mario Mateo1
Department of Astronomy, University of Michigan, 821 Dennison Building, Ann Arbor, Michigan 48109-1090
Electronic mail: [email protected]
Luc Simard
Department of Physics and Astronomy, University of Victoria, P.O. Box 3055, Victoria, British Columbia V8W 3P6, Canada
Electronic mail: simard beluga.phys.uvic.ca
Michael Bolte
UCO/Lick Observatory, University of California, Santa Cruz, California 95064
Eileen D. Frtet.
Maria Mitchell Observatory, 3 Vestal Street, Nantucket, Massachusetts 02554
Yannick Copin
Ecole Normale Suprieure de Lyon, 46, allee d'Italie, 69364 Lyon cedex 07, France
Received 1997 January 23; accepted 1997 May 22
ABSTRACT. We have obtained 2640 CCD spectra with resolution ~ 4 in the region 7250-9000 for
976 stars lying near the red-giant branches in color-magnitude diagrams of 52 Galactic globular clusters.
Radial velocities of ~ 16 km s_1 accuracy per star determined from the spectra are combined with other
criteria to assess quantitative membership probabilities. Measurements of the equivalent widths of the
infrared calcium triplet lines yield a relative metal-abundance ranking with a precision that compares
favorably to other techniques. Regressions between our system and those of others are derived. Our
reduction procedures are discussed in detail, and the resultant catalog of derived velocities and equivalent
widths is presented. The metal abundances derived from these data will be the subject of a future paper.
1. INTRODUCTION ters, and especially for the crucial clusters near the Galactic
center, we cannot hope to test models of Milky Way for-
The absolute and relative ages of globular clusters in the
Galaxy and in the nearest Local Group galaxies provide mation in a compelling fashion.
unique constraints on cosmology and early epochs of galaxy In 1989 we carried out a photometric and spectroscopic
formation. However, the ages of globular clusters cannot be program at Las Campanas Observatory that was aimed at
determined, even in a differential sense, without knowledge developing a highly precise relative ranking of globular-
to high precision of their chemical composition (or metal- cluster abundances from measurements of the Ca triplet
licity). An error of 0.3 dex in the overall heavy element lines in the near-infrared spectra of 12-15 probable red
abundance of a cluster usually denoted by [Fe/H] giant members of each of 52 clusters. The early work of
corresponds to an error of about 3 Gyr in the age derived Armandroff and Zinn (1988, hereafter AZ88) used the Ca
from fitting an otherwise absolutely correct isochrone to triplet lines formed in the integrated light of Galactic globu-
main-sequence photometry of perfect accuracy. Even for lar clusters, from which it appeared that an internal preci-
some bright, nearby clusters, recent careful abundance mea- sion of 0.15 dex per star was possible provided chromo-
surements differ by more than this amount, which reflects spherically active stars were avoided. Moreover, by working
the challenges of detailed analyses from stellar spectra. in the infrared, sensitivity would be reduced to the high and
Without reliable metallicity determinations for many clus- variable reddening towards many clusters of great interest
for the evaluation of formation scenarios for the Galaxy. At
1
Visiting Astronomer, Las Campanas Observatory. the time the project began, it was a relatively unexplored
883 1997. Astronomical Society of the Pacific
Astronomical Society of the Pacific Provided by the NASA Astrophysics Data System
884 RUTLEDGE ET AL.
empirical approach which had been applied primarily to The adopted data for the clusters we observed are pre-
integrated or composite light, and our goal was to acquire sented in Table 1, where the columns are, respectively: (1)
data of such quality on individual giants that we could the running cluster identification number used in Fig. 4;
assess thoroughly and independently the optimum proce- (2),(3),(4) the NGC, other cluster, and IAU names; (5),(6)
dures and relative merits of this technique. galactic longitude and latitude in degrees; (7) the visual
Since we undertook this project, several others have ex- magnitude of the horizontal branch level, (8),(9) the
ploited with great effect the Can triplet technique applied radial velocity and associated uncertainty, vH and ();
to individual giants for estimating abundances of globular (10) the mean reddening for the cluster; (11) the central
clusters, with particular emphasis upon distant and/or sparse velocity dispersion from Pryor and Meylan (1993, hereafter
objects (see, e.g., Armandroff and Da Costa 1991 (hereafter PM93); (12),(13) the metallicity and associated uncertainty
AD91); Olszewski et al. 1991; Armandroff et al. 1992 of the cluster taken from Zinn and West (1984, hereafter
(hereafter ADZ92); Da Costa et al. 1992 (hereafter ZW84); (14) the standard deviation adopted for the V pho-
DAN92); Suntzeff et al. 1992 (hereafter S92); 1993 (here- tometry, which is used in Sec. 6 during the robust line
after S93); Da Costa and Armandroff 1995 (hereafter fitting technique to determine the reduced equivalent width,
DA95); Geisler et al. 1995 (hereafter G95); and Suntzeff W, of the cluster (this value is estimated from the scatter
and Kraft 1996 (hereafter SK96)). While these programs in the CMDs published by the authors from which we
have provided numerous results of widespread interest, the adopted the photometry (see the Appendix), and represents
original motivation of our program remains. In this paper a combination of both the photometric errors and differen-
we describe how we optimized our reduction of the spectral tial reddening within the cluster). The data from columns
data (Sec. 3) to provide radial velocities (Sec. 4) and 5-10 were taken from a 1994 version of the Harris (1996)
equivalent widths (Sec. 5), compare our prescriptions and electronic MWGC catalog (hereafter referred to as the
results with those of other workers (Sec. 5.5), and present a MWGC catalog), and references can be found in the Ap-
catalog of the individual stellar results (Sec. 7). Following pendix.
the AD91 prescription, the cluster reduced equivalent
widths, W, are calculated (Sec. 6). A companion paper
discusses the calibration of our cluster W values to [Fe/H] 3. EXTRACTIONS AND CALIBRATIONS
values, and the astrophysical implications of our results.
Since there was not an overscan region on our detector,
2. OBSERVATIONS the bias level of each frame was estimated from the mean
level of the bias frames. This was a satisfactory approach
Spectra were obtained at the Las Campanas Observato- for run one, where the bias level remained constant
ry's 2.5-m Dupont Telescope equipped with the modular at ~ 550 ADU. However, due to a CCD electronics prob-
spectrograph and the Canon 85 mm //1.2 camera. A GG lem, the bias level in run two varied between 540 and 650
495 filter was used to block the second and higher spectral ADU on timescales of a few hours.
orders. The TI No.2 detector (800X 800 thinned CCD; read- The illumination response along the slit resulted in
out noise = 11 e~ pix-1; gain = 1.35 e~ per ADU; a 21% reduction in transmission from one end of the slit
scale = 0"85 pix-1) was used with an 831 1 mm-1 (8000 to the other. This effect was independent of slit rotation,
blaze) grating, which produced a dispersion of 2.19 and was found in both the object and fiat-field frames. By
pix-1 and spectral coverage from 7250-9000 . The normalizing the flat-field frames along the dispersion axis
8 ' .'25 slit provided an instrumental spectral resolution only (with the IRAF2 task response), we could use them to
of ~ 4 . remove the illumination response. For run two, where the
Observations were obtained on two 1989 runs: (1) April bias level of each frame was uncertain, the illumination
13-20 and (2) July 13-21. Of the 52 clusters observed, 23 response may not have been removed correctly by this pro-
were observed during the first run only, 26 were observed cedure. For these data we chose sky windows on both sides
during the second run only, and three were observed during of, and immediately adjacent to, the stellar spectrum being
both runs to check the consistency of our results. In each extracted so that a low-order fit between the windows
cluster, spectra were obtained for 10-20 stars selected from would account satisfactorily for any residual errors in the
published color-magnitude diagrams (CMDs) to lie on the illumination response and bias level. Spectra in crowded
red giant branch (RGB) and, if proper-motion data were fields were not extracted from run two data when windows
available, to be likely proper-motion members. Probable appropriate for accurate sky subtraction could not be iden-
asymptotic branch (AGB) stars were avoided, as were hori- tified adjacent to a spectrum.
zontal branch (HB) stars, and known variable stars near the We removed cosmic rays with the IRAF task cosmic-
RGB tip. Slit positions were chosen to contain at least two rays, while any remaining cosmic rays seen in a visual
stars per spectrograph rotation. inspection of the two-dimensional images we removed us-
Each star was observed two or three times consecutively, ing IRAF's imedit. We rectified residual distortion
with an Fe-Ar arc taken before and after each sequence for
the wavelength calibration. Occasionally the same star was 2
observed on different nights, or with a different slit orien- IRAF is distributed by the National Optical Astronomy Observatories,
which is operated by the Association of Universities for Research in
tation, to check for systematic effects in our results. Expo- Astronomy, Inc., under cooperative agreement with the National Science
sure times for a single frame ranged from 2 to 17 min. Foundation.
Ca TRIPLET I. CATALOG 885
Table 1
Adopted Cluster Data
^HB Vh o-(vH) E(B-V) [Fe/H] o-([Fe/H]) (/)
No. NGC Name IAU /11 (mag) (km s-1) (km s-1) (mag) (km s ) (dex) (dex) (mag)
1 104 47 Tue C0021 723 305.9 -44.9 14.06 -18.7 0.2 0.04 11.5 -0.71 0.08 0.10
2 288 C0050-268 152.3 -89.4 15.38 -46.4 0.4 0.03 2.9 -1.40 0.12 0.10
3 362 COIOO711 301.5 -46.2 15.43 223.5 0.5 0.05 6.4 -1.27 0.07 0.10
4 1261 C0310554 270.5 -52.1 16.70 53.3 9.1 0.01 -1.31 0.09 0.10
5 2298 C0647359 245.6 -16.0 16.11 100.5 9.1 0.13 -1.85 0.11 0.15
6 2808 C0911-646 282.2 -11.3 16.19 99.7 2.9 0.23 13.4 -1.37 0.09 0.15
7 3201 C1015461 277.2 8.6 14.80 494.6 0.2 0.21 5.2 -1.61 0.12 0.15
8 4372 C1223 724 301.0 -9.9 15.30 60.8 6.5 0.45 -2.08 0.15 0.20
9 4590 M68 C1236264 299.6 36.1 15.68 -95.1 0.6 0.04 2.5 -2.09 0.11 0.15
10 4833 C1256706 303.6 -8.0 15.45 203.1 5.6 0.33 -1.86 0.09 0.20
11 5286 C1343 511 311.6 10.6 16.20 53.6 4.1 0.25 8.0 -1.79 0.11 0.20
12 5897 C1514208 342.9 30.3 16.35 23.0 8.0 0.08 -1.68 0.11 0.15
13 5904 M5 C1516+022 3.9 46.8 15.06 51.8 0.5 0.03 5.7 -1.40 0.06 0.10
14 5927 C1524505 326.6 4.9 16.60 -100.5 5.3 0.47 -0.30 0.09 0.20
15 5986 C1542376 337.0 13.3 16.50 92.3 5.9 0.27 -1.67 0.10 0.20
16 6093 M80 C1614228 352.7 19.5 15.86 7.3 4.1 0.18 12.4 -1.68 0.12 0.20
17 6101 C1620720 317.7 -15.8 16.60 191.1 13.7 0.04 -1.81 0.15 0.10
18 6121 M4 C1620264 351.0 16.0 13.45 70.0 0.5 0.36 4.2 -1.33 0.10 0.20
19 6144 Cl624259 351.9 15.7 16.60 142.7 7.7 0.32 -1.75 0.15 0.20
20 6171 M107 C1629- 129 3.4 23.0 15.70 -33.8 0.3 0.33 4.1 -0.99 0.06 0.20
21 6218 M12 C1644 018 15.7 26.3 14.90 -43.5 0.6 0.17 4.5 -1.61 0.12 0.20
22 6235 C1650220 358.9 13.5 16.70 86.9 3.9 0.36 -1.40 0.15 0.20
23 6254 MIO C1654 040 15.1 23.1 14.65 75.4 1.0 0.28 6.6 -1.60 0.08 0.20
24 6266 M62 C1658300 353.6 7.3 16.30 -68.0 3.2 0.47 14.3 -1.28 0.15 0.20
25 6273 M19 C1659262 356.9 9.4 16.40 129.4 6.9 0.37 -1.68 0.15 0.20
26 6304 C1711 294 355.8 5.4 16.25 - 105.0 8.7 0.52 -0.59 0.23 0.20
27 6352 C1721 484 341.4 -7.2 15.28 -114.6 6.3 0.21 -0.51 0.08 0.20
28 6366 C1725-050 18.4 16.0 15.65 - 122.6 0.5 0.69 1.3 -0.99 0.25 0.15
29 6362 C1726670 325.6 -17.6 15.34 -13.0 0.6 0.09 2.8 -1.08 0.09 0.20
30 6397 C1736536 338.2 -12.0 12.87 18.8 0.1 0.18 4.5 -1.91 0.14 0.10
31 6496 C1755-442 348.0 -10.0 16.47 -98.4 7.7 0.13 -0.48 0.15 0.10
32 6522 C1800300 1.0 -3.9 16.85 -8.7 5.6 0.50 6.7 -1.44 0.15 0.20
33 6535 C1801 003 27.2 10.4 15.73 -215.1 0.5 0.32 2.4 -1.75 0.15 0.15
34 6528 C1801 300 1.1 -4.2 17.10 162.2 5.2 0.62 0.12 0.21 0.15
35 6544 C1804250 5.8 -2.2 14.90 -16.6 6.5 0.74 -1.56 0.15 0.25
36 6541 C1804437 349.3 -11.2 15.30 -153.8 3.1 0.12 8.2 -1.83 0.15 0.15
37 6553 C1806259 5.2 -3.0 16.60 -26.9 4.1 0.84 -0.29 0.11 0.20
38 6624 C1820 303 2.8 -7.9 16.11 54.3 0.6 0.27 5.4 -0.35 0.15 0.15
39 6626 C1821 249 7.8 -5.6 15.70 15.2 1.2 0.41 8.6 -1.44 0.15 0.15
40 6638 C1827255 7.9 -7.2 16.50 9.7 6.8 0.40 -1.15 0.15 0.20
41 6637 M69 C1828 323 1.7 -10.3 16.00 39.6 4.7 0.17 -0.59 0.19 0.15
42 6681 M70 C1840323 2.9 -12.5 15.60 218.7 1.2 0.07 5.1 -1.51 0.14 0.15
43 6712 C1850-087 25.4 -4.3 16.25 -107.7 0.6 0.46 4.3 -1.01 0.14 0.15
44 6717 Pal 9 C1852227 12.9 -10.9 16.60 1.8 7.1 0.21 -1.32 0.15 0.25
45 6723 C1856-367 0.1 -17.3 15.50 -81.9 6.3 0.06 -1.09 0.14 0.15
46 6752 C1906600 336.5 -25.6 13.70 -27.4 2.7 0.04 4.5 -1.54 0.09 0.15
47 6809 M55 C1936310 8.8 -23.3 14.40 174.9 0.4 0.07 4.9 -1.82 0.15 0.15
48 6981 M72 C2050- 127 35.2 -32.7 16.90 -288.8 8.2 0.05 -1.54 0.09 0.15
49 7089 M2 C2130010 53.4 -35.8 16.05 -6.7 2.1 0.05 8.2 -1.62 0.07 0.15
50 7099 M30 C2137234 27.2 -46.8 15.10 -184.3 1.0 0.04 5.6 -2.13 0.13 0.15
51 7492 C2305-159 53.4 -63.5 17.63 -188.5 8.5 0.00 -1.82 0.30 0.15
52 Pal 12 C2143-214 30.5 -47.7 17.13 28.5 1.5 0.02 -1.14 0.20 0.10
in the images (manifested by curved night sky lines near the ing the reductions; no stars were extracted which fell on
edges of the frames) with the IRAF tasks fitcoords and these columns.
transform. Spectral extractions were made with IRAF's apextract
Several bad columns and pixels were noted before the tasks. Two sky windows, with a minimum of 15 pixels
observations were taken, and stars were placed on the slit to each, were chosen on either side of the star, and a linear fit
avoid them. Several other unreliable sections of the CCD between the median values in the two windows was used to
were mapped and avoided during spectral extractions. define the sky level at the position of the stellar spectrum.
Charge skimmed columns, in which the percentage of elec- The windows were chosen to be as close to the star as
trons skimmed varied with time, were also discovered dur- possible, while still adhering to the run two constraints
886 RUTLEDGE ET AL.
mentioned earlier. To facilitate the placement of the sky

windows, a maximum of one bad column or charge
skimmed column was permitted to lie in a window and be
dealt with by the medianing process. The arc spectra were
extracted with the identical parameters used for the stellar
extractions. The two arc exposures associated with a given
star were averaged and the resultant digital spectra were
logically connected to the appropriate extracted stellar spec-
trum.
We calculated the dispersion solution for each spectrum
with a FORTRAN program (similar to IRAF's World Coor-
dinate System, which was unavailable at the time of the
reductions) that did not alter the pixel binning or the pixel
values, but rather wrote the coefficients of the dispersion
solution to the headers of the individual spectra. For each
arc, the program found every line above a threshold value
and separated from all other lines by at least two pixels, and
fit each with a Moffat function of exponent four to establish
an accurate pixel center. A Legendre polynomial with five
terms was fit to give the wavelength dispersion solution.
This was done consecutively for all the spectra of a given
run, and the mean and of the residual (XcalcXlab) for
each line was calculated. If the mean residual was greater
than 0.5 , then the line was not used for the final disper-
sion solution. The remainder of the fines were weighted Fig. 1Comparison of spectra with different S/N values (estimated as
such that explained in the text) illustrate the degradation as the S/N values drop
below ~ 15.
w = 1.0 if ^ 0.02 ,
w = 0.02/ if > 0.02 .
suggested that our data could be used for independent ve-
We applied these new weights in the final dispersion solu- locity determinations. The procedures adopted are described
tion calculation for each spectrum. In the end, 16 fines in below.
the wavelength interval 7272-8668 A were used.
To estimate the S/N of each spectrum, we used two
wavelength windows: 8580-8620 and 8700-8800 A.
These windows were chosen such that none of the 4.1 Cross Correlations
globular-cluster Ca triplet lines would be velocity shifted
into them. In each of these regions a robust line fitting We used a FORTRAN program to determine velocities by
technique was used to fit a straight line to the pixel values, cross correlation against a template spectrum. NGC 6809
where the absolute deviation was minimized in the fit rather star 2441 (-4-41), S/N = 180, served as our template.
than the square of the deviation. Let be the pixel value All spectra were continuum normalized and rebinned to a
divided by its fitted value, and let the mean and average log scale. The correlation function, C(AX), between the
deviation of all in a window be denoted as Nmem and program spectra (P) and the template spectrum (77) was
Ndey (after clipping by 3 X Ndty). The S/N for each window calculated to be the sum of ()( ) between
was then estimated to be Nmem/Ndev, and the S/N for the template :=: 8350 and 8750 (which corresponds to
spectrum in the relevant Ca triplet region was taken to be . rest ~ 8345 -8745 ) for a large range of values.
the average S/N calculated for the two windows. Figure 1 These limits were chosen to avoid telluric H2O features
shows a plot of the Can triplet region for four of our between ~ 8100-8300 A, and ^ 8800 . The maximum
program spectra having S/N values ranging from 12 to 125. value of C(AX) and the adjoining 3 values were
The distribution of S/N values for all of our program spec- then fit by a parabola whose center provided the initial
tra can be found in Fig. 2. The ~ 1% of the spectra with velocity estimate. The program spectra were then Doppler
S/N ^ 15 were not analyzed further. shifted by the initial velocity estimate and the cross corre-
lation repeated to get an additional velocity shift which
ensured that the same wavelength region in each spectrum
was being used in the correlation; this process was repeated
4. RADIAL VELOCITIES
until the velocity shift was stable to well within our errors.
Radial velocities aid in identifying bona fide cluster The final velocity was used to shift the spectra to the tem-
members, particularly for those clusters projected against plate velocity, so that the band windows used in Sec. 5
populous star fields. While for membership assessment, to calculate the equivalent widths (EWs) were aligned
only relative radial velocities are required, experimentation properly.
Ca TRIPLET L CATALOG 887
300.0 60.0
50.0
200.0 40.0
30.0

I
100.0
20.0 -
10.0 -
50.0 100.0 150.0

S/N 50.0 100.0 150.0
<S/N>
Fig. 2A histogram of the distribution of the S/N for the 2640 spectra
analyzed. Fig. 3The standard deviation of the velocity measurements for stars
observed consecutively at least twice with the same exposure times are
4.2 Velocity Errors compared to the mean S/N values for each set of multiple observations of
the same star.
Given a stable Cassegrain spectrograph insensitive to
rotation angle and changing gravity vector, the uncertainties position, so comparisons below are made between the me-
in our velocities will be dominated by slit centering errors. dian of each group taken under the different conditions. The
In order to achieve maximum throughput as well as reliable absolute value of the difference between two observations,
relative velocities, considerable care was spent in the each with standard error , has an expectation value of
slit rotation process to ensure that the prime pair of \2; this formula was used below to estimate the standard
stars was well centered on the 1.25^ wide slit. This error where appropriate.
width corresponded to 1.47 pixels on the image plane, or Star 3204 in NGC 3201 (see Table 9) was observed
3.22 ~ 97 kms-1 at 8500 A. As shown below, our consecutively 19 times during run one at positions ranging
velocities per star appear to be accurate to ~ 16 km s-1, over the entire length of the slit. The standard deviation of
after centering and other uncertainties are considered. the velocity was 14.2 kms-1, and no significant trend of
derived velocity was found as a function of position along
4.2 Internal Errors the slit. There were 223 and 37 stars observed on more than
one night of runs one and two, respectively; there were no
Our observational procedure ensured a large sample of significant differences found from night to night, and the
stars for each run that were observed at least twice con- standard errors derived from the mean absolute value of the
secutively with the same exposure time. From these, the star differences were 16 14 (s.d.) kms-1, and 12 12
standard deviation of the velocity measurement could be (s.d.) kms-1, respectively, where (s.d.) indicates the stan-
determined and compared to the mean S/N of the spectrum.
dard deviation of the sample, as distinct from the standard
The results are found in Fig. 3. The median standard de- deviation of the mean, (s.d.m.). Fourteen and seven stars
viation for run one, 7.7 km s" ^ and for run two, 8.0 kms-1, were observed in runs one and two, respectively, with dif-
are indicative of the internal precision and correspond to ferent spectrograph rotation angles. The standard errors de-
measuring shifts between the program spectra and the tem- rived from the mean absolute differences were, respectively,
plate spectrum at the ^-0.1 pixel level, which is typical for
149 (s.d.) kms-1 and 22 10 (s.d.) kms-1; in neither
cross correlation techniques. case was a trend observed as a function of rotation angle.
From 15 stars observed in common between the two runs,
4.2.2 External Errors the mean difference between the velocities (run two-run
Several tests enable us to characterize the external accu- one) was -3.510.1 (s.d.) kms-1.
racy of our data; these include comparison of observations An independent estimate of our uncertainties taking into
on different nights within a run, observations on different account the internal velocity dispersion of the clusters was
runs, and observations with different spectrograph rotations. made by comparing our velocities with results from Sec.
We generally have three consecutive spectra for every slit 4.4. Our calculated dispersion, 0&1, for each cluster for
888 RUTLEDGE ET AL.
which we had ten or more stars was compared to the inter- The mean error of unit weight (m.e.l) was calculated as
nal velocity dispersion, ^, as given by Pryor and Meylan follows:
(1993). The mean excess in dispersion for 25 such clusters 1/2
is ae = (^caic-^)0'5 = 16.05.7 (s.d.) km s-1. m.e.l =
^ ( V i V template) . .
2.3,
In summary, while the internal precision of an individual
^(,)
stellar velocity appears to be ~ 8 kms-1, the more rel-
evant external uncertainties (arising from centering errors, where the sum is over the 16 clusters used and = 15.
flexure, etc.) are 16 kms-1. The velocity of each star The uncertainty in our template velocity zero point is then
relative to its cluster velocity given in the MWGC catalog is taken to be (i;template) = m-e.l (1/2( i;))-05
presented in Table 9. = 2.2 kms-1. If we accept the cluster velocity errors
given in the MWGC catalog, then the m.e.l value above
4.3 Template Velocity Zero Point and Cluster Velocities indicates that const 2.3, in the definition of (), will
give a realistic external error estimate for v. It is unclear
As noted earlier, star 2441 (-4-41) of NGC 6809 (M55) why m.e.l is 2.3 rather than 1; either the normal error
was chosen as our template for the cross correlations. Since estimate, crobs / a/TV, is not appropriate for our error distri-
we did not observe radial-velocity standard stars, its ex post bution, or the errors quoted in the MWGC catalog are
facto choice was based upon it being a relatively high S/N underestimated (e.g., perhaps the quoted errors are more
observation of a globular-cluster giant from a cluster with a correct estimates of internal errors, rather than external
well determined radial velocity and velocity dispersion. To errors).
produce velocities on the standard system we need to assign In summary, we assign a velocity of 172 2.2 km s-1 to
a velocity to 2441. We could have chosen to use the star 2441, and our external uncertainties for the cluster
MWGC catalog value (174.90.4 kms"1) for the cluster, velocities are, (;) ~ 2.3 X aohs/yl, where is the num-
and ignore the possibility that this star might have a detect- ber of stars entering the mean.
able offset. The latter possibility seems ruled out by the
unpublished measurements of Pryor and collaborators who 4.4 Velocity Results
used the radial-velocity scanner on the Canada-France- Our cluster results are presented in Table 2, where the
Hawaii Telescope, where the velocity of 2441 was found to columns are respectively, (1), (2), (3) as in Table 1; (4), (5),
be 177.60.5 kms-1 relative to the cluster mean velocity (6) our mean cluster velocity, v, external error, cr(v),
(for a 20 star dataset), 176.60.9 kms-1, and a cluster and the number of stars used to estimate the mean
velocity dispersion of 3.8 kms-1. We chose, however, to velocity, N; (7), (8) the difference l? = v vH, and
set the template velocity zero point by minimizing the dif- ( vH) = {2() +2(H))0'5, where the // values
ference between our cluster velocity estimates and those can be found in Table 1; (9), (10), (11) the reduced equiva-
given in the MWGC catalog, as we now describe. lent width, W of the cluster, its associated uncertainty,
Our initial cluster velocity estimate was the median ve- a(W'), and the mean error of unit weight, m.e.l, in the fit
locity after five iterations of 3 clipping. Our final cluster of the cluster as described in Sec. 6. The difference between
velocity estimate also accounted for the central internal ve- each star's velocity and the cluster velocity given in the
locity dispersion, ^, of each cluster. The latter values, MWGC catalog is plotted in Fig. 4, and the individual
listed in Table 1, are from Pryor and Meylan (1993). If the stellar velocities are tabulated in Table 9 in Sec. 7, below.
cluster was not listed by Pryor and Meylan, then a typical While the goal of this project was not to determine
value of 5 kms-1 was used. Using our estimate of the accurate cluster velocities, it is reassuring that our cluster
external error in the measurement of the velocity of a single velocity estimates generally agree well with the estimates
observation of a single star, cre = 16 km s-1 (see Sec. 4.2) given in the MWGC catalog, with only five clusters having
for our measurement error, the observed dispersion for each differences greater than 3: NGCs 2298, 5897, 6101, 6553,
cluster should be = (cr^t+^)0 5. The mean of all and 6981. For NGCs 2298, 5897, and 6101, G95 recently
stars within 3aobs of our initial velocity estimate form our reported velocities of 150.41.3 kms-1, 102.91.0 km
final cluster velocity estimate, v. If these stars are drawn s-1, and 364.3 1.9 km s-1, respectively. In all three cases
from a normal distribution with = ^, then the vari- the differences between our velocities and theirs are less
ance in can be estimated as 2{) ~ const2Xc^bs/A^, than 1.
where is the number of stars entering the mean, and const Finally, NGCs 6235, 6528, and 6681 presented particular
is a constant that is determined below. challenges when identifying stars to define the initial veloc-
To obtain the template velocity zero point, we compared ity estimate. For NGC 6235, there was no obvious grouping
our velocity determinations of 16 clusters for which the of velocity measures in our sample; therefore, the catalog
MWGC catalog quotes velocity errors < 1 km s-1, and for velocity was used as the first estimate of this cluster's
which our estimate was based upon >10 stars. Let H and velocity and only five stars satisfied the iterative clipping
be the catalog and our values, respectively. For each procedures to determine our velocity, which differs from the
cluster, the difference. ; = v vH, and the variance, catalog value by 0.1. NGC 6528, with (l,b) = (1.1,
2( ;) = 2{) + {H)2, were calculated (where the 4.2), is projected onto the dense star fields of the Ga-
value of const in 2(;) was set to one in this analysis). The lactic center. However its high radial velocity made it rea-
velocity of the template, template ^ was taken to be the sonable to reject all the stars with velocities near 0 kms-1,
weighted mean [w, = 1/2( ^)] of viy 172 kms-1. which are most likely bulge stars, when determining
Ca TRIPLET 1. CATALOG 889
Table 2
Cluster Velocity and Reduced EW Results
o-(v) () W (')
No. NGC Name (km s"1) (km s_1) (km s-1) (km s-1) () () m.e.l
1 104 47 Tue -26.5 8.1 31 -7.8 8.1 4.51 0.04 1.49
2 288 -65.7 10.3 13 -19.3 10.3 3.65 0.06 1.32
3 362 225.5 11.9 11 2.0 11.9 3.72 0.07 1.16
4 1261 73.2 12.1 10 19.9 15.1 3.77 0.09 1.86
5 2298 144.4 9.0 18 43.9 12.8 2.24 0.05 1.30
6 2808 80.1 9.9 23 -19.6 10.3 3.75 0.08 2.13
7 3201 499.6 7.0 30 5.0 7.0 3.41 0.03 1.14
8 4372 70.4 7.1 29 9.6 9.6 1.94 0.05 1.33
9 4590 M68 -84.9 8.5 19 10.2 8.5 1.58 0.06 1.51
10 4833 194.1 7.5 26 -9.0 9.4 2.27 0.05 1.41
11 5286 65.7 11.8 12 12.1 12.5 3.03 0.08 1.45
12 5897 97.6 9.9 15 74.6 12.7 2.24 0.07 1.56
13 5904 M5 58.6 5.9 43 6.8 5.9 3.75 0.06 1.93
14 5927 -123.8 8.8 19 -23.3 10.3 4.79 0.05 0.99
15 5986 86.7 11.0 12 -5.6 12.5 3.16 0.09 1.36
16 6093 M80 11.8 10.6 19 4.5 11.4 2.86 0.06 1.20
17 6101 363.6 10.6 13 172.5 17.3 1.95 0.11 2.47
18 6121 M4 70.1 6.6 33 0.1 6.6 3.83 0.05 1.52
19 6144 174.3 13.5 8 31.6 15.6 2.20 0.05 0.80
20 6171 M107 -31.1 8.7 19 2.7 8.7 3.99 0.05 1.01
21 6218 M12 -37.3 9.5 16 6.2 9.5 3.85 0.10 1.73
22 6235 88.7 17.1 5 1.8 17.6 3.54 0.11 1.52
23 6254 MIO 83.1 9.9 16 7.7 9.9 3.42 0.07 1.70
24 6266 M62 -62.1 9.3 28 5.9 9.8 3.95 0.07 1.64
25 6273 M19 138.0 11.5 11 8.6 13.4 2.69 0.10 1.99
26 6304 -107.8 9.0 18 -2.8 12.5 4.84 0.05 1.02
27 6352 -122.8 8.0 23 -8.2 10.2 4.73 0.07 1.55
28 6366 -113.7 9.5 15 8.9 9.5 4.70 0.05 1.09
29 6362 -16.0 9.6 15 -3.0 9.6 3.93 0.07 1.44
30 6397 17.8 8.7 19 -1.0 8.7 2.21 0.06 1.99
31 6496 -129.3 19.1 4 -30.9 20.6 4.70 0.08 1.10
32 6522 -18.3 9.3 18 -9.6 10.9 3.47 0.09 1.72
33 6535 -204.8 14.0 7 10.3 14.0 2.74 0.27 3.33
34 6528 212.2 13.5 8 50.0 14.5 5.41 0.14 2.57
35 6544 -33.0 11.0 12 -16.4 12.8 3.53 0.09 1.42
36 6541 -163.5 12.4 11 -9.7 12.8 2.72 0.05 1.02
37 6553 8.4 8.4 21 35.3 9.3 5.13 0.09 1.55
38 6624 38.6 8.4 21 -15.7 8.4 4.66 0.05 1.09
39 6626 42.3 10.4 16 27.1 10.4 4.05 0.08 1.51
40 6638 22.4 11.0 12 12.7 13.0 4.31 0.10 1.62
41 6637 M69 40.4 8.2 22 0.8 9.4 4.48 0.07 1.59
42 6681 M70 216.8 17.1 5 -1.9 17.2 3.14 0.05 0.66
43 6712 -105.5 10.9 12 2.2 10.9 4.11 0.06 1.49
44 6717 Pal 9 29.0 8.8 19 27.2 11.3 3.78 0.11 1.50
45 6723 -100.3 9.9 15 -18.4 11.7 4.07 0.08 1.73
46 6752 -21.9 10.1 14 5.5 10.5 3.42 0.04 1.03
47 6809 M55 158.9 11.5 11 -16.0 11.5 2.69 0.05 1.30
48 6981 M72 -359.6 9.6 16 -70.8 12.6 3.53 0.10 2.13
49 7089 M2 4.1 12.4 11 10.8 12.6 3.28 0.10 1.76
50 7099 M30 -177.2 15.8 6 7.1 15.8 1.72 0.12 2.08
51 7492 -214.2 11.5 11 -25.7 14.3 2.98 0.16 2.47
52 Pal 12 13.7 15.6 6 -14.8 15.7 4.57 0.15 2.34
Notes to Table 2
The velocity results for NGCs 6235, 6528, and 6681 are affected by heavy field star contamination (see Sec.
4.4).
890 RUTLEDGE ET AL.
400.0 density of cluster stars and field stars for each cluster at the
radii which was observed. To first order, this number can be
estimated as the total number of cluster stars in our sample
over the total number of field stars, as defined above (i.e.,
200.0 Nc/f = 818/158). The probability that a star is a cluster
member on the basis of its velocity relative to the cluster
mean is,
Nc,f PcG ( ^ star ' ^ ' O'obs)
0.0 Py, =
elf Pci^staif ^ ^obs) -^0(^ star'^'Afield)
This approach to assigning individual star membership
probabilities offers at least two advantages compared to
-200.0 simple clipping. First, it allows for the greater ease of
distinguishing between field and cluster stars when the clus-
ter has a high velocity. Second, a star whose velocity de-
viates from the cluster mean in the direction of ; = 0 km
-400.0 s"1 is more likely to be a field star than one that deviates in
the opposite direction. These two effects are reected in our
probability scheme since the field star population is cen-
tered on = 0 kms-1, and the cluster population is cen-
-600.0 tered on v. Membership probabilities thus calculated are
0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0 listed for each star in Table 9.
Cluster Number (see Table 1)
Fig. AThe difference between the median velocity calculated for each 5. Ca TRIPLET EQUIVALENT WIDTHS
star and the cluster velocity given in the Harris (1996) MWGC catalog is
plotted for each of the 52 clusters observed. The numbers on the abscissa Measurement of the EW of an absorption feature is usu-
are linked with NGC numbers in Table 1. ally done by defining continuum bandpasses on each side of
the feature, and linearly interpolating the average or median
the initial estimate; eight stars survived the clipping to enter intensities in each of the bandpasses to define the con-
our final velocity, which differs from the catalog value by tinuum at the feature wavelength. The EW is then the in-
3.4. However, membership remains something of a con- tegral over the feature bandpass of the difference between
cern for any cluster with such contamination problems. the continuum and the feature (see Sec. 5.2). The feature is
Similarly, NGC 6681 is close to the bulge, but has a high defined either directly by the spectral intensities or by some
radial velocity. Since there was a group of stars close to the analytical function fitted to them (see Sec. 5.1). For Can
velocity quoted in the catalog, and a scattering of stars with triplet work, it is also necessary to combine the three lines
velocities closer to 0 kms-1, the median velocity of the in some manner to get a net Call index, which we will
five stars close to the catalog velocity was used as a first denote as SCa (see Sec. 5.3). Table 3 summarizes the
estimate of this cluster's velocity, which produced a final approaches used by previous practitioners of Can triplet
cluster velocity estimate that differs from the catalog value work, where the columns are, respectively: (1) the refer-
by .. Due to these problems, the cluster velocities we enced paper; (2) the method used to combine the three Ca 11
obtain, and our assignment of membership probabilities for triplet lines into a single Ca index for the star, Ca; (3)
stars from these three clusters, are less reliable than those the method used to define the line feature in the spectrum;
for the other 49 clusters. (4) the reference for bandpass limits used to define the
continuum and feature regions. Table 4 defines the band-
4.5 Cluster Membership Probabilities passes that were adopted by other authors and by
ourselves,3 where the columns are, respectively: (1) the pa-
For each cluster, a Gaussian probability distribution per where the bandpasses are defined; (2) the Can triplet
function (PDF) was defined, Poi^stai^ ,), which repre- line name defined by its rest wavelength; (3) the line center
sents the probability that a star with velocity vstai was adopted by other authors and derived by ourselves; (4) the
drawn from the assumed cluster Gaussian velocity distribu- limits of integration for the line; (5), (6) the limits used to
tion with mean and = [Cobs-^ ^(^)]5 as defined in define the continuum on the blue and red side of the line.
Sec. 4.3. A field star population was defined by selecting all Since there were many multiply observed stars in our
stars which differed from their cluster velocity by more than sample, a series of tests were performed to determine the
2. Of the 976 stars in our sample, 158 fit this criterion. optimum method of calculating SCa. In these tests, the
These presumed field stars form a symmetrical distribution mean, < EW > , and standard deviation, cr(EW), of the
about i; = 0 km s~1, and are satisfactorily fit by a Gaus-
sian with 1 = 75 kms-1. Therefore, the field PDF 3
was defined as PG^starA^eid) To define accurately the Note that the instrumental resolution (usually defined by the FWHM of
the arc lines) was ~ 3 for the earlier studies, with values ranging from
probability of a star being a velocity member of a given 2.5 (Olszewski et al. 1991) to 4.8 (Armandroff and Zinn 1988),
cluster, it is also necessary to know the relative number compared to our 3.4-4.8 (see Sec. 5.2).
Table 3
Comparison of SCa Calculation Techniques
Paper XCa Line fitting Bandpassesa
AZ88 98+ 8542+ 8662 Numerical AZ88
AD91 ^-8542+^8662 Gaussian AD91
Olszewski et al. (1991) -98 + ^-8542 + ^8662 Gaussian AZ88
ADZ92 ^-8542 + ^-8662 Gaussian AD91
DAN92 ^8542 + ^8662 Gaussian AD91
S92b ^ 8542 + 8662 Gaussian AZ88
S93 9 + . Gaussian AZ88
G95 ^-8542 + ^-8662 Numerical AZ88
DA95 ^-8542 + ^-8662 Gaussian AD91
SK96 ^-8542 + ^-8662 Gaussian AZ88
This paper (TP) 0.5 8498 + 8542 + 0.6 8662 Moffat TP
Notes to Table 3
a
b
See Table 4 for the different bandpass limits.
In the original paper, the sum of all three lines was used, but in the Appendix of S93, the S92 XCa values
were expressed as the sum of the two strongest lines; it is the latter values that we are referring to in this
paper.
EW was calculated for each line in every star that was ferences from the observed profile,6 except for slightly un-
observed at least twice consecutively with the same expo- derestimating the wings. On the other hand, the Gaussian fit
sure time.4 A diagnostic, a, defined as the median cr(EW) underestimates the depth of the line by 5%, overesti-
divided by the median < EW > , provides essentially an mates the full width at half maximum (FWHM) by ~ 10%,
estimate of the inverse S/N for the EW. A technique was and grossly underestimates the depth of the wings. We have
then developed to minimize a. Table 5 presents the a therefore adopted the Moffat fitting technique for the re-
values for the various tests described below, where the mainder of our analysis. To quantify the differences be-
columns are, respectively, (1) the technique used to calcu- tween the line fitting techniques, we performed a spectrum-
late the equivalent widths; (2) the run that the test was by-spectrum comparison for 1866 spectra to relate the Ca
performed on; (3), (4), (5) the values for each of the three index (see Sec. 5.3) calculated with the Numerical method,
Can triplet lines. SCa(A0, the Gaussian method, 2Ca(G), and the Moffat
method, 2Ca(M). A least-squares fit results in
5.1 Line Fitting Technique :SCa(A0 = 1.005( +0.003). 2Ca(M)
Three different techniques were tested for measuring -0.047( 0.013) rms
equivalent widths: direct numerical integration, fitting the
= 0.12
line with a Gaussian function, and fitting the line with a
Moffat function of exponent 2.5. Initially the AZ88 line XCa(G) = 0.995( 0.003). XCa(M)
bandpasses were used, and for the continuum bandpasses
the AZ88 definition was used for 8498 the AD91 -0.084( 0.005) rms
definition was used for 8542 and 8662 (see Table 4), which = 0.04,
we will refer to as the AZ88/AD91 continuum bandpasses.5
Between both runs, there was a total of 600, 764, and 750 which indicates that there are slight zero-point shifts be-
stars used for 8493, 8542, and 8662, respectively. The tween the techniques, but the slopes are consistent with
first six rows of Table 5 show that compared to the Nu- unity. The larger rms value for the Numerical technique is
merical method, the Gaussian method shows an improve- consistent with its larger a value (Table 5).
ment in of 10% on average, and compared to the
Gaussian method, the Moffat method shows an improve-
5.2 Line and Continuum Bandpass Windows
ment of ~ 4% on average. In addition, when plotting the
deviations from the fitted profile for all our spectra, it is Next, we optimized the continuum bandpasses for
clear that the Moffat fit shows no obvious systematic dif- the line bandpasses of Armandroff. Our continuum band-
passes, listed in Table 4, were chosen to be as large as
4 possible, while still avoiding the telluric H2O features be-
In Sees. 5.1 and 5.2, the sample was also restricted to stars with success-
ful EW calculations for all three line fitting methods to avoid bias arising tween ~ 8100-8300 and ^ 8800 . As seen in rows
from one method having a higher success rate than another. For the 7 and 8 of Table 5, our new continuum windows reduced a
remaining tests, we restricted the samples to stars which had successful
EW calculations from the Moffat line fitting technique for all three Ca by ~ 9% for the 8498 line, while the 8542 and \ 8662 lines
triplet lines, which was necessary for our XCa to be calculated.
5 6
AD91 significantly modified the continuum bandpasses defined by AZ88 The exponent for the Moffat function was chosen to be 2.5 since, from a
for 8542 and 8662 to optimize their equivalent width calculations. visual inspection, it was found to fit our observed line profile the best.
892 RUTLEDGE ET AL.
Table 4
Ca Triplet Line and Continuum Bandpasses
Feature Line center Line bandpass Blue continuum Red continuum
Paper name () () bandpass () bandpass ()
AZ88 ^8498 8498 8490-8506 8474-8489 8521-8531
^ 8542 8542 8532-8552 8521-8531 8555-8595
8662 8653-8671 8626-8650 8695-8725
AD91 8542 8532-8552 8474-8489 8559-8595
v 8662 8653-8671 8626-8647 8695-8754
8662
TP 8498 8498.1 8490-8506 8346-8489 8563-8642
^8542 8542.3 8532-8552 8346-8489 8563-8642
8662 8662.4 8653-8671 8563-8642 8697-8754
were insignificantly affected. We quantified the differences strongest for 8498, which ranges from 0.082 (10 ) to
between the two sets of continuum windows by making a 0.063 (3.5 ) for run one, and 0.091 ( 10 ) to 0.074
star-by-star comparison for 764 stars. This allowed us to (3.5 ) for run two. For bandpasses ^ 4.5 , the
relate the Ca index calculated with the Armandroff con- change in a becomes insignificant. Thus, for each line, the
tinuum windows, 2Ca(A), and with our larger continuum largest bandpass for which there was no significant increase
windows, 2Ca(L). A least-squares fit results in, in a from the 3.5 band was analyzed further. The
values chosen for Xg^g, 8542, and 8662, 4 , 4.5 ,
SCa(A) = 1.071( 0.004) 2Ca(L)
and 4.5 , respectively, are about half the size of the
-0.19( 0.02) rms original Armandroff line bandpasses. The a values for these
line bandpasses represent a 16.5% reduction, on average,
= 0.13 , from the Armandroff line bandpasses.
which indicates that a small scaling factor and zero-point Although smaller line bandpasses significantly decrease
offset exists between the two methods. All stars that lie the a values, they also dramatically increase sensitivity to
significantly off this relation were previously noted to have changes in the instrumental profile, which in turn, induce
strong TiO absorption, and were not further included in our systematic errors in our calculated equivalent widths. Un-
analysis. fortunately, from an analysis of the arc lines associated with
Finally, we adopted the new continuum bandpasses and each spectrum, the FWHM was essentially constant for run
tested a variety of line bandpasses that ranged from 3.5 one at 3.4 , while for run two, the FWHM was 4.9
to 10 about each line center. Our line fitting technique for night one, ~ 4.4 for nights two to six, and 3.8
simultaneously fit the amplitude, function-width parameter, for the remainder of the nights. The changes in run two
and central wavelength, Xc, for each line of each spectrum. are attributed, ex post facto, to changes in the spectrograph
The median kc for each line from all spectra are listed in focus. To quantify the effects of these changes in the
Table 4, which are almost identical to the laboratory wave- FWHM of the instrumental profile on our choice of line and
lengths defined for these lines. The feature bandpasses were continuum bandpasses, a group of 532 pairs of spectra were
defined around these median \c values. In general, we collected, such that (i) the spectra were of the same star,
found, as one might expect, that a decreases monotonically and (ii) the FWHM of the arc lines associated with the two
from the larger to the smaller bandpasses. The effect is spectra differed by more than 0.1 (the maximum differ-
ence was 0.8 ). Let EW be the difference between 2Ca
Table 5 calculated from the spectra with the larger arc FWHM and
Ca Triplet or Diagnostic for Different EW Calculation Techniques the smaller arc FWHM, and let |AFWHM| be the absolute
Techmque Run ^-8498 -8542 ^-8662 value of the difference between the FWHM of each of the
arc lines. When we use the AZ88 Une bandpasses and our
Numerical 1 0.118 0.042 0.067 continuum bandpasses (see the TP bandpasses in Table 4),
2 0.117 0.053 0.068
a least-squares fit results in,
Gaussian 1 0.088 0.042 0.064
2 0.101 0.051 0.073 kEW = 0.008( 0.10) I AFWHM|-0.04( 0.03),
so no significant trend is present. However, when we make
Moffat 1 0.088 0.041 0.059
2 0.101 0.049 0.069 the same comparison with the narrow line bandpasses and
our continuum windows, we obtain,
Moffat 1 0.080 0.040 0.059
(new cont.) 2 0.092 0.047 0.070 kEW = - 0.57( 0.07) I AFWHM|+ 0.05( 0.02),
Moffat 1 0.065 0.034 0.050 which clearly shows that iht EW of a line is underestimated
(new cont. + new line) 2 0.071 0.036 0.057 as the FWHM of the instrumental profile increases, as one
would expect. Had our instrumental resolution remained
constant throughout the two runs, it is clear from the final a ter giants, they determined that inclusion of the 8493 line
diagnostic in Table 5 that the narrower line bandpasses added more noise than signal to 2Ca, and thus excluded it
derived herein would have represented improvements over from the sum. Since we have a large sample of multiply
the AZ88 values. With the data available, however, it is observed stars, we revisited this issue.
evident that the larger line bandpasses are preferable, and Our goal is to determine a weighted mean for SCa,
thus they were adopted for the remainder of our work. Our XCa = W W{\8498) + ^W( .8542) + ^3^^(^8662^
final bandpasses (TP) are presented in Table 4. where the weights minimize a for XCa. For each line in
each star in our sample we calculated the mean, variance
5.3 Combining the Can Triplet Lines: SCa and covariance, (vl,v2,v3), and
(cl2, cl3, c23), respectively. The median of these values
In their original work, AZ88 measured Ca indices from was calculated for a sample and, respectively, denoted as
integrated cluster light and they defined 2Ca to be the sum ml0,m20,m30,i;I0,u20,i;30,cl20,cl30, and c230. A
of the three triplet lines. When AD91 used individual clus- value equivalent to a defined earlier was then calculated as
I
{w \ Q + w '2 V2Q + w
2 1
Q + 2w iw 2C\2Q + 2w iw + 2w 2W
w ! w 10 + W2 w 20 + 30
Even though the covariances were less than the variances star in Table 9 is the error in the weighted mean,
by at least a factor of ten, they were retained. We proceeded [ElAj2(XCa)]"05, where the sum is over the number of
as follows. We set the value of W2 to 1, and we minimize spectra per star.
a with a downhill simplex method (Press et al. 1992) by A number of tests for systematic effects in SCa deter-
letting Wi and W3 vary. For observing run one, the minima minations were possible, which also allowed us to estimate
found were wl = 0.55 and w3 = 0.62, while for run two the external errors in XCa. From 15 stars observed in com-
they were Wj = 0.48, and W3 = 0.68. The average, mon between runs one and two, the mean difference in
rounded to the nearest tenth, for each weight was taken to 2Ca was found to be 2Ca(run two) 2Ca(run one)
give final weights of W! = 0.5, W2 1.0, and w3 = = 0.007 0.15 (s.d.) . NGC 3201 star 3204 was observed
0.6. These weights yield a final a value for SCa of 0.032 consecutively 19 times at different slit positions. The stan-
for run one, and 0.034 for run two. If, on the other hand, dard deviation in SCa was 0.17 for a mean S/N of 62,
both W3 and W2 were 1.0 and W! was 0.0, the respective a which is ~ 1.4X a(XCa); no trend in XCa was seen as a
values would be 0.035 for run one, and 0.037 for run two. function of position on the slit. There were 223 and 37 stars
It thus appears that in our data, inclusion of the 8498 observed on different nights of run one and run two, re-
feature increases the S/N of XCa, and we therefore include spectively. No significant offsets were found between any
it with the weights just defined. of the nights, and the standard errors derived from the
overall absolute value of the differences between nights
was 0.12+0.12 [ S/N = 61 + 23; ~ l.OXCT(XCa)] and
0.090.07 [S/N = 71+15; ~ 0.9(2^)] for runs
5.4 EW Errors one and two, respectively. Fourteen stars from run one and
seven stars from run two were observed with different spec-
In order to combine observations with varying S/N in the trograph rotation angles. The standard errors derived from
most effective way, and to measure the uncertainty in SCa the mean absolute value of the difference in 2Ca
for single observations, the standard deviation in the XCa, were 0.160.09 (s.d.) and 0.170.10 (s.d) , respec-
<T(SCa), was computed for 594 stars that were consecutively. For their respective mean S/N of 53 and 43,
tively observed with identical configurations and exposure these correspond to error excesses of ~ l.lX(T(2Ca)
times, and compared to the mean signal-to-noise, and ~ 1.0X(XCa).
< S/N > , for each star. These values are plotted in Fig. 5. In summary, our external EW errors appear to be con-
The solid line overplotted on these data represents our es- sistent with our internal errors.
timate of the single observation internal as a function of
the S/N of the spectra, and was calculated as follows. A
5.5 Transformations
running median over 20 stars for both a(SCa) and
< S/N > was calculated, and a quadratic fit was per- We have derived transformations between our XCa sys-
formed which is taken to be the error estimate for tem, TP, and those of others. Observations of individual
< S/N > less than 100, and for < S/N > greater than stars in common with various sources are plotted in Fig. 6.
100, a constant (Ca) = 0.067 was assumed. We as- The error bars for the abscissa represent our external error
sume our errors are approximately Gaussian, and weight the estimates, while the error bars for the ordinate represent the
observations as l/2 when calculating the mean SCa for a errors quoted by the other authors. The dotted lines on the
star that was observed multiple times. The listed for each graph represent a one to one correlation, while the solid line
894 RUTLEDGE ET AL.
1.5
o Run 1
Run 2
LO
50.0 100.0 150.0

<SIN>
Fig. 5As in Fig. 3, but for the standard deviation of the XCa measure-
ments for stars observed consecutively at least twice with the same expo- 6.0 2.0 4.0
sure times.
Ia(TP)
represents our least-squares fit to the data, which allows for Fig. 6A comparison of the XCa values in globular-cluster stars from
errors in both directions (Stetson 1989). The regression co- various sources. The regression coefficients are found in Table 6.
efficients for the fit 2Ca(other) = m- 2Ca(TP) + ?, are
listed in Table 6, where the columns are respectively, (1) NGC 104 L5622, since this star has weak TiO bands
the data source; (2), (3) the slope of the fit and its corre- present, and SK96 suggests that the line strengths may be
sponding uncertainty; (4), (5) the intercept of the fit and its varying in this star.
corresponding uncertainty; (6) the mean error of unit weight If we assume that our estimates for a(SCa) reflect our
for the fit; (7), (8) the approximate minimum and maximum true external errors, then the m.e.l values for S92, S93,
SCa(TP) for which the regression is valid; and (9) the G95, and SK96, which are greater than one, suggest that
number of stars in common, N. The m.e.l value was cal- they have underestimated their errors. For G95, the m.e.l
culated as, m.e.l = (2/2)0'5~0'5, where the sum is value reduces from 1.75 to 1.18 if the two most deviant
over all the stars used, e is the deviation along the ordinate points are removed. DA95 compared their SCa values to
of a star from the fitted relation, is calculated for each those of other authors as well, but did not publish regres-
star as (mV2[ 2Ca(TP)]+(72[ SCa(other)])05, and is sions. Their results are qualitatively in agreement with ours,
the number of stars used minus two. The uncertainties listed which suggests that the DA95 and S93 XCa values are on
for the coefficients are the formal uncertainties from the fit the same system, the AD91 XCa values show a positive
multiplied by the m.e.l value for the fit. The fit to the SK96 slope with respect to the DA95/S93 system (i.e., the differ-
data was done excluding the star with the largest SCa, ence between AD91 2Ca and DA95/S93 XCa becomes
Table 6
Transformation Regressions
b Sb m.e.l mm max
Paper dm () () () ()
AD 91 1.26 0.08 -0.89 0.06 0.92 3.9 6.0
ADZ92 1.24 0.10 -0.73 0.05 0.51 4.2 5.4 5
DAN92 1.09 0.07 -0.38 0.07 0.92 2.6 5.2 6
592 1.04 0.04 -0.11 0.05 1.45 2.9 5.7 7
593 1.10 0.02 -0.33 0.02 1.35 2.1 5.5 37
G95 1.03 0.10 0.35 0.07 1.75 2.2 4.6 21
DA95 1.23 0.10 -0.88 0.05 0.49 4.0 5.5 5
SK96 1.06 0.02 -0.09 0.03 0.94 2.4 5.7 14
Table 7 technique, allowing for errors in both directions, is

Published A(SCa)/A(V) Slopes jij = mXi j+bj, where yij is the SCa value of star i in
m m cluster y ; is the VHB V value of star i in cluster j; m
Paper ( mag-1 ) ( mag-1 is the slope, A(2Ca)/A(VHB V), which is assumed to
be constant for all clusters; and bj is the intercept or re-
AD91 0.62 0.01 duced equivalent width, W', of cluster j. The errors, ,
ADZ92 0.66 0.07
DAN92 0.72 0.04 assigned to the Xi j values were assumed to be constant for
S93 0.64 0.03 a cluster, and are listed in Table 1 as cr(V). They represent
G95 0.61 0.04 our estimate of the errors in V from a combination of
DA95 0.61 0.03 photometric errors, and errors due to differential reddening
TP 0.64 0.02
within the cluster. The errors, , assigned to the
values are our estimates of the external errors in the XCa of
the star, and are listed in Table 9 as (Ca).
larger as SCa increases), and the ADZ92 SCa show a The weight of each star in the fit is w = fpfrV 2,
small excess of ~ 0.25 over the DA95/S93 SCa for the where the final w ,fp , and fr are listed for each star in
small range of SCa in common. DA95 found a small posi- Table 9. Thevalue for a star is a constant throughout the
tive slope between the DAN92 XCa and the DA95/S93 iterative fitting, and is calculated as,
system, whereas we find that all three systems are essen-
tially the same. fp = if ^ is not available,
We attribute the differences in slope and zero point be-
tween studies to a combination of continuum window defi- fp = Pv if > 0.2,
nitions (see Sec. 5.2), line fitting techniques (see Sec. 5.1), fp = 0 if ^ 0.2,
SCa definitions (see Sec. 5.3), and slight changes in the
instrumental resolution and throughput properties of the where the and Pv values for each star are the probabili-
spectrograph and detector used (see Sec. 5.2). ties of membership based on proper motions and velocities
(see Sec. 4.5), respectively (listed in Table 9). Since stars
6. CLUSTER REDUCED EQUIVALENT WIDTHS: were selected to be close to the RGB of the cluster, and are
W found as close to the cluster center as the published pho-
tometry, finder charts, and stellar density permitted, (both of
As demonstrated by AD91, the change in SCa as a which increase their probability of being cluster members),
function of Vhb V for stars within a cluster has a constant the proper-motion probabilities were used conservatively to
slope, A(SCa)/A(yHB - V), for all clusters. This result assess membership. The attached to each star is calcu-
was corroborated by subsequent studies for stars with lated as = + y)0'5, where m0 is the estimate of
Vhb " 0, where the mean slopes from the various the slope from the previous iteration. The fr value assigns
papers are presented in Table 7; the columns are, respec- lower weight to stars which lie significantly off the locus
tively, (1) paper reference, (2) mean slope from the paper, defined by the other stars in the cluster, and is calculated as,
and (3) the uncertainty in the slope. In most cases, a stan-
dard slope of 0.62 mag"1 was adopted, which allows for 1
the calculation of a reduced equivalent width, W, for a f "
1 + [|6|/( m.e.l)]'0
cluster, where the SCa of each star in the cluster is cor-
rected to the level of the horizontal branch by subtracting where is the star's SCa deviation from its cluster fit in
0.62(VhbV), the mean of all cluster members is the previous iteration; is calculated as described above;
taken. Since our method of calculating Ca differs from the and are constants discussed below; and m.e.l is the mean
previous studies, we do not expect, a priori, that our slope error of unit weight for the cluster which is calculated as
will be the same as theirs. Since SK96, and S92 showed m.e.l = (62/2)05^~05, where the sum is over the num-
that the slope reduces to ~ 0.35 for RGB stars with ber of stars in the cluster with fp ^ 0.75, and is one less
Vhb- V ^ 0, we have calculated W for each cluster ex- than this. The constants y and were set to 3 and 4,
cluding these stars, which represents only ~ 4% of our respectively, which assigns equal values of fr to all stars
total sample. We have also excluded stars which appear to within ~ 2 m.e. 1 of the cluster fit, while fr drops rapidly
be AGB stars, HB stars, have an uncertain ID, or lie sig- to 1/2 when the star is ~ 3m.e.l from the cluster fit. The
nificantly off the locus defined by the other stars in the iterations were continued until there was no significant
SCa, Vhb V plot Furthermore, the weight of a star in the change in m.e.l for any cluster. For each cluster, the final
fitting procedure is dependent on its radial velocity and intercept, W, its uncertainty, (\'), and the m.e.l value
proper motion, as described below. are listed in Table 2, where a(W') is the formal uncer-
The technique we chose to calculate W for each cluster tainty in the intercept from the fitting technique multiplied
uses an iterative, robust algorithm to fit one slope and 52 by m.e.l for the cluster. An m.e.l value greater than one
intercepts simultaneously to all our data. The intercepts indicates that the scatter about the best fit line is larger than
represent the value of XCa at the level of the horizontal what would be expected from the adopted errors in XCa
branch for each cluster, and thus are equivalent to W. The and V alone. Since we are generally confident in our error
analytic function fit to our data by an iterative least-squares estimates for XCa (see Sec. 5.4), the excess scatter in
896 RUTLEDGE ET AL.
most clusters is likely due to underestimating the errors in tal branch; (4) the dereddened BV color of the star de-
V, including noncluster members in our sample, including termined with (3-V) from the MWGC catalog; (5) the
non-RGB stars in our sample, differential reddening within weighted mean XCa of the star, calculated according to
a cluster, or, as a remote possibility, a Ca spread in the Sec. 5; (6) the external error of the weighted mean SCa; (7)
cluster. The inhomogeneous nature of the photometry used the probability of cluster membership from proper-motion
to establish the ordinate in Fig. 7 seems likely to be the data; (8) the probability of cluster membership from our
dominant source of scatter. calculated velocities (see Sec. 4.5); (9), (10), (11) the/p,
The slope that we obtain from this technique is fr values and final weight, w, respectively, used in our
0.64 0.02 mag-1, with m.e.l = 1.6, where the uncer- fitting technique to calculate the reduced equivalent width,
tainty quoted is the formal uncertainty from the fitting tech- W, of the cluster (see Sec. 6 for technique, and Table 2 for
nique multiplied by m.e.l. The m.e.l value here is calcu- Wf values); the weight has been scaled in each cluster so
lated as it was for individual clusters, except the sum is that the highest weight star in the cluster has w = 1 ; (12)
over the number of stars in all clusters with fp ^ 0.75, and the number of spectra analyzed for the star; (13) the nights
is this number minus 53 (i.e., the number of clusters used that the star was observed, where the first number indicates
plus one). This slope is consistent with the slopes that the observing run, and the other numbers indicate the spe-
previous authors have found, even though our transforma- cific night of observation. Therefore, 1-1 represents 1989
tion results (Sec. 5.5) suggest that our slope should be April 13, 1-2 represents 1989 April 14,..., and 2-1 repre-
slightly shallower. This is most likely due to the fact that sents 1989 July 13. If a star was observed on more that one
other authors have not allowed for errors in V when calcu- night for a given run, then it will have more than one
lating the slope, which causes the fitting technique to un- number following the dash. Finally, column (14) indicates
derestimate the true slope (Stetson 1989). This method of when notes are found at the bottom of the table.
calculating the mean slope is more effective than simply The values of vH, and (B V) for each cluster
taking the mean of slopes fit independently, since clusters can be found in Table 1. The references for these values, as
with stars having a small range in VHB V, and therefore well as the references for the Table 9 star names, photom-
possessing little slope information, simply add noise to the etry, and the proper motions (in columns 1,3,4, and 7,
latter method, whereas they do not affect the former. The respectively) can be found in the Appendix. Notes for a
final fit for each cluster is plotted in Fig. 7. given cluster, or stars in the cluster, can also be found in the
We performed two experiments to investigate whether Appendix. In Fig. 7, the Vhb- ^ magnitudes are plotted
the slope, A(SCa)/A(yHB - V), is a function of [Fe/H]. In
against the 2Ca values in the left panel. The light line
the first, every cluster was fit by our technique indepen-
represents the robust line fit (see Sec. 6) to the cluster stars,
dently, and the resulant slopes of each cluster were plotted
while the bold lines represent the line fits to three well
against their respective [Fe/H] from ZW84, which are listed
studied clusters spanning a large range in [Fe/H] and each
in Table 1. The mean slope was 0.620.02 mag-1, and
having low reddening. The CMD of the same stars is plot-
no significant trend with [Fe/H] was observed, although
ted in the right panel to give an indication of possible
there was some indication that the slope may become
steeper for the more metal-rich clusters. To examine further non-RGB stars, and to allow assessment of the photometry,
any possible [Fe/H] dependence of the slope, the clusters where a large scatter is likely indicative of sizable photo-
were split into four metallicity bins, with each subset ana- metric errors, differential reddening, inclusion of undetected
lyzed following the precepts described above for the full nonmembers, or some combination thereof. The RGBs of
sample. The results are presented in Table 8, where the the same three fiducial clusters used in the left panels are
columns are, respectively, (1) the [Fe/H] range as defined plotted in the right panels with bold lines. The relative
by ZW84 (see Table 1), (2) the number of clusters in that placement, with respect to the fiducial lines, of the cluster
[Fe/H] range, (3), (4) the slope, A(XCa)/A(VHB V), and line fit in the left panels, and the RGB stars in the right
uncertainty in the slope, derived for the clusters in the bin, panels give an indication of the accuracy of the (B V)
and (5) the m.e.l value for the fit. As seen in Table 8, none values listed in the MWGC catalog. The details for these
of the slopes for the metallicity bins differs significantly figures are given in the figure caption. The interpretation of
from the 0.64 0.04 A mag-1 found for the whole sample, the data compiled in this paper will be presented in subse-
nor is there compelling evidence that more metal-rich clus- quent publications.
ters differ in this regard. Accordingly, we will continue to We thank Las Campanas night assistants Angel Guerra
assume that the slope is constant for all metallicities. With
and Fernando Peralta for their cheerful, effective support,
better photometry and membership information, this point
and W. E. Kunkel for setting up the spectrograph and in-
would be well worth revisiting in the future.
structing J.E.H. in its use for run one. The large quantity of
photographic work required to prepare the extensive, large-
7. THE CATALOG
scale finding charts essential to the observations was ably
The results for individual stars are given in Table 9, performed by Dave Duncan at DAO. We thank Gianni
where the columns are, respectively: (1) the star name; (2) Marconi, Hugh Harris, Don Temdrup, and Sergio Ortolani
the difference between our calculated velocity of the star for sending us ASCII copies of their photometry, in some
(see Sec. 4), and the cluster velocity given in the MWGC cases before publication. Thanks also to Bill Harris for
catalog; (3) the V magnitude of the star above the horizon- compiling the MWGC catalog, for insightful discussions.
Ca u TRIPLET I. CATALOG 897
2.0 3.0 4.0 5.0 6.0 2.0 3.0 4.0 5.0 6.0 0.5
ljCa() UCa(A)
2.0 3.0 4.0 5.0 6.0 0.5 2.0 3.0 4.0 5.0 6.0
(b) () (d) ECa ()
Fig. 7For each cluster, the V magnitude above the horizontal branch (F^ values found in Table 1) is plotted against the XCa in the left panel, and the
color-magnitude diagram is plotted in the right panel, where the 5 - F values have been dereddened by the cluster E(B-V) given in Table 1. The circles
represent stars that have fp values greater than 0.75, while the triangles represent the remainder. The-1stars that were not used to compute the reduced
equivalent 2width, W, are labeled with a (X) symbol. The slope in the left panel is 0.640.02 mag . The error bars in the left panel are calculated
as = {m a2{V) + 2(^))05, where a(V) is assumed to be constant for a cluster and is listed in Table 1, and aiXCa) is listed for each star in Table
9 (plotted if > 0.1 ). The light line in the left panel represents the robust line fit for the stars in the cluster (see Sec. 6), while the bold lines represent
the robust line fit for the three clusters (from left to right) NGC 4590 (M 68), NGC 5904 (M 5), and NGC 104 (47 Tue), all of which have well determined
[Fe/H] and W values and low E(B-V) values. In the right panel, the RGB fiducials for each of these clusters is plotted (in the same order from left to
right); they were taken, respectively, from McClure et al. (1987) (bright end from Harris 1975); Sandquist et al. (1996); and Hesser et al. (1987).
898 RUTLEDGE ET AL.
2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5 2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5
(f) () (B-V)0 (h) ZCa () (B-V)0
Fig. 7(Continued)
and for a helpful referee's report. G.A.R. would like to MWGC catalog. Since this is a dynamic catalog, the refer-
thank the NRC for funding during this project. ence sources for the data we adopted are listed below for
each cluster. The cluster coordinates were taken from Djor-
govski and Meylan (1993). The sources for the E(B-V)
measurements were from Reed et al. (1988); Webbink
APPENDIX: THE PROGRAM CLUSTERS
(1985); and Zinn (1985). The catalog documentation states,
The reference cluster data (see Table 1) were taken from "In addition to the three major sources listed above, mea-
the 1994 June version of the Harris (1996) electronic surements of E(B ~V) from the individual color-magnitude
NGC 6496
NGC 6522
NGC 6535
2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5 2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5
Ia<A) (B~V)0 UJaU) (B-V)0
NGC 6528
NGC 6544
NGC 6541
2.0 3.0 4.0 5.0 6.0 1.0 2.0 3.0 4.0 5.0 6.0 1.0
XCa() (B-y)o Ia() (B-V)0
Fig. 7(Continued)
studies (sources given for V^) were employed whenever more recent sources are also available for smaller lists of
they appeared to be well calibrated. The final adopted red- objects; in many cases these are based on large samples of
denings are the straight averages of the given sources (up to stars from CORAVEL or multi-object echelle spectra with
four per cluster)." The cluster radial velocities were mainly very high precision ( 1 km s"1 or less) and almost totally
taken from Armandroff and Zinn (1988), Hesser et al. supersede any previous data. The adopted r for each clus-
(1986); Webbink (1981); and Zinn and West (1984). Again, ter is the average of the available measurements, each one
the catalog documentation states, "However, numerous weighted inversely as the published uncertainty." These
900 RUTLEDGE ET AL.
I I I I
"/ /
6553*
1 {I 1 1 1 1I 1 1 1 I 1 1 1 1I 1 1 1 I1 1 1 1 I I I I I 1
X 6624 '
'1 11 ''1 1 1 1I 1 1 1 I1 1 1 1 1I 1 1 1 I1 1 1 r1 I1 1 1 1 1 1 1r 1 1 1 1I 1
6626~
2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5

ZCa () (B-V)0
1 . I I I 1 ! < . . I 1 1
/ ^
6638~
1' 1 ''1 1 1 1 1! 1 1 1I 1 1 1 1 1 1 1 1 1 !1 1 1I 1 1 1 1 I1 1 1 1 1 I1 1 1
1
- 1Yl NGC 6637 ~

'1 I ''1 1 1 I 1 1 1 I I 1 1 1 I 1 1 1
1
I ! ' I M 69I I
/ /*
' NGC
M 70 6657
.1...1.,,1...1,..1...
2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5 2.0 3.0 4.0 5.0 6.0 0.5
(n) ZCa () (B-V)0 (P) UCa(A)
Fig. 7(Continued)
more recent sources are listed below for individual clusters charged coupled device, (2) PG for photometry done with a
in the form H : author, author .... photographic plate, and (3) RR Lyrae if the determination is
More generally, for each cluster we list individually the specifically referred to them.
relevant data sources, as follows. Proper-motion data (): values listed in Table 9 are
Horizontal-branch magnitudes (Vhb): these were mea- cited in the notes here in the form : author.
sured from the sources, which are cited in the form : Stellar identifications and photometric zero points:
type of data author, type of data author,..., where the type sources are cited in the form ID: author (letter), author (let-
of data is one of: (1) CCD for photometry done with a ter), ..., where the letter is used to preface the star ID in
Ca U TRIPLET 1. CATALOG 901
3.0 3.0
2.0 2.0
1.0 1.0
0.0 NGC 7089 0.0 .^/1
///
111111111I1111111I111 1I 1 1 1 1 I1 1 1 1 M2
! '
1 1 1 1 I1 1 t! 1 I1 i 1 1 1 1 1 1 1 1 1 1I 1 1 1 I1 1 I 1 1 1 1 I 1 1 1 1 I 1 1
3.0 3.0
2.0 2.0
1.0 1.0
0.0 // 0.0
X/ / / NGC 7099
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 I1 I ,1 I1 /,1 .I' I1 1, 1, ,1 MI. 30I1 1, 1. 1 I1 1 1 1 1 1 1 1 1I 1 1 1 1I 1 1I 1 i 1 1 ti 1I 1 1 1 1 I1 1 1 1 1 I1 1 1
3.0 3.0
2.0 2.0
1.0 1.0
0.0 0.0
2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5 2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5
(q) Ja () (B-V)0 (r) Ia () (B-V)0
Fig. 7(Continued)
Table 9 and for discussion of any photometric zero-point with L to within 0.02 mag, so no corrections were
adjustments. Because a 0.1 V magnitude zero-point offset applied.
between photometry used for the HB and RGB stars would NGC 288 - vH: Peterson et al. (1986); Pry or et al.
lead to a systematic shift in the calculated [Fe/H] (1991); Pryor and Meylan (1993); V^: CCD - Bergbusch
of ~ 0.025 dex for metal-poor clusters, and more for (1993); ID: Alcaino and Liller (1980c)[A]; Olszewski et al.
metal-rich clusters, zero-point offsets larger than 0.1
(1984)[0]. Using ten stars, the V photometry of A was
mag between photometric systems were always applied to
brighter than O's by -0.08 0.09 mag, so the straight
the RGB stars to bring them onto the Vm system. These
mean was taken for these stars for both V and B V. No
adjustments are listed below as Vm VletteT = ...; in
stars from A or O were in common with the Bergbusch
more complicated scenarios, they are explicitly given. When
star-by-star comparisons were performed, the refers to study, but Bolte (1992, see Fig. 6) finds good agreement
the standard deviation of the sample. We also indicate how between his photometry and O. Bergbusch found that his
the photometry from different sources was combined to V photometry was ~ 0.064 mag fainter than Bolte, so no
give the V values listed in Table 9. If there are no com- corrections were applied.
ments for a cluster, then the photometry was taken un- NGC 362 - vH : Fischer et al. (1993); : PG - Harris
changed from the single source listed. (1982); : Tucholke (1992b); ID: Harris (1982)[H].
NGC 104 = 47 Tue - H: Armandroff and Da Costa NGC 1261 - Vhb : CCD - Ferraro et al. (1993); ID:
(1986); Meylan et al. (1991); Meylan and Mayor (1986); Ferraro et al. (1993)[F], Alcaino (1979b)[A]. All photom-
VW CCD - Hesser et al. (1987); : Tucholke (1992a); etry was taken from F.
ID: Lee (1977b)[L]; Hesser et al. (see their Fig. 11, and NGC 2298 - Vhb CCD - Janes and Heasley (1988); ID:
Appendix ) found that their V photometry agrees
Alcaino and Liller (1986a)[A]. No stars from A were in
common with the Janes and Heasely study, but Vhb in both
Table 8 studies indicates that there is not a significant zero-point
Metallicity (XCa)/A(yHB-V) slopes
difference.
Metallicity NGC 2808 - Vhb : CCD - Ferraro et al. (1990); ID:
range ( mag ^ ( mag- ') m.e.l
Harris (1975)[H]; Harris (1978). The photometry was taken
-0.8 < [Fe/H] 9 0.73 0.10 1.4 from Harris (1975), and recalibrated according to Harris
-1.2 < [Fe/H] = -0.8 7 0.54 0.11 1.4 (1978). We could not reproduce Ferraro
-1.8 < [Fe/H] = -1.2 26 0.62 0.06 1.6 et al. 's comparison to the Harris (1978) data, so from our
[Fe/H] = -1.8 10 0.67 0.09 1.6
all 52 0.64 0.04 1.6 comparison of 14 stars, the V photometry of Ferraro
et al. was brighter than Harris (1978) by 0.060.04 mag.
902 RUTLEDGE ET AL.
Table 9
Individual star Data and Resultsa
Star Vhb-V (5-y)o SCa a(SCa) Run-
(km s_1) (mag) (mag) () () P fr Night Notes
NGC 104 47 Tue
L5622 17 2.23 1.66 6.08 0.08 97 90 2-3 0,2,11
L3730 -11 2.21 1.59 6.08 0.05 96 95 0.95 0.98 0.67 2-258
L8704 -1 2.20 1.57 5.98 0.03 98 95 0.95 1.00 1.00 2-3678
L4741 -8 2.12 1.57 6.13 0.04 98 95 0.95 0.76 0.69 2-2356
L4729 13 2.08 1.48 5.65 0.04 98 92 0.92 0.89 0.68 2-2356
L4745 -33 1.90 1.38 5.61 0.06 97 92 0.92 0.99 0.50 2-2
L5623 -15 1.84 1.43 5.58 0.10 97 95 0.95 1.00 0.25 2-3
L8705 32 1.80 1.36 5.60 0.05 96 76 0.76 1.00 0.54 2-3678
L5739 -19 1.66 1.33 5.67 0.06 98 95 0.95 1.00 0.57 2-3
L4628 -1 1.53 1.27 5.39 0.05 98 95 0.95 0.99 0.62 2-2
L6732 8 1.44 1.26 5.40 0.05 95 93 0.93 1.00 0.60 2-6
L6728 5 1.28 1.23 5.61 0.08 94 0.94 0.88 0.33 2-6
L6717 18 1.19 LU 5.14 0.09 98 90 0.90 0.99 0.26 2-6
L4737 -73 1.16 1.22 5.63 0.30 98 26 0.26 0.99 0.01 2-2
Notes to Table 9
a
NoteThis Table will be included in Vol. 9 of the AAS CD-ROM Series. The first few lines of the table are presented here to show its form and content.
(0) Star not used to compute reduced EW for cluster, (1) see cluster notes in the Appendix, (2) weak TiO present, (3) strong TiO present, (4) possible AGB
star, (5) possible HB star, (6) non-cluster member based on position in CMD, (7) crowded field, (8) uncertain ID, (9) star has weak XCa for V^V
position, (10) star has strong XCa for Vhb- ^ position, (11) SK96 suggest that line strength is variable.
and the B V photometry was redder by 0.09 0.06. Fer- 3) found that their V photometry was 0.1 mag fainter
raro et al. photometry was used for stars when available, than S, so Vhb- Vs = 0-1
and otherwise Harris (1978) data was used with no correc- NGC 5904 = M5 - vH: Olszewski et al. (1986);
tions. Peterson et al. (1986); Rastorguev and Samus (1991); Vhb
NGC 3201 - vH: Cot et al. (1995); CCD - CCD photometric RR Lyrae - Storm et al. (1991); :
Brewer et al. (1993); ID: Lee (1977c)[L]. Star 2405 was Cudworth (1979); ID: Buonanno et al. (1981)[B]. From the
listed in both the PE and PG data of L; we assumed that the level of Vhb in the CMD of B, there does not appear to be
PG value was correct. Brewer et al. find systematic differ- a systematic V offset from Storm et al.
ences with the photometry of L, but this effect was iS 0.1
NGC 5927 - Vhb: CCD - Sarajedini and Norris (1994);
mag in V, for the central part of the cluster, and less severe
CCD - Friel and Geisler (1991); ID: Menzies (1974b)[M].
in the outskirts. Since our stars were selected far from the
center, no corrections were applied. Sarajedini and Norris found that their V photometry was on
average 0.2 mag fainter than M, so VHB- VM = 0.2.
NGC 4372 - VfjB : CCD - Alcaino et al. (1991); ID:
Alcaino (1974a)[A]. Alcaino et al. find no systematic dif- NGC 5986 - Vhb: CCD - Bond et al. (1994); ID: Harris
ferences with the A photometry. et al. (1976)[]. From the level of Vhb in CMD of H,
there does not appear to be a systematic V offset from Bond
NGC 4590 = M68 - ;^ : Pry or and Mey lan (1993); et al.
V^hb : CCD - McClure et al. (1987); ID: Harris (1975)[H];
Alcaino (1977a)[A]. Using twelve stars, the V photometry NGC 6093 = M80 - Vhb: PG - Harris and Racine
of H was brighter than A by 0.03 0.01, so no adjust- (1974); ID: PG - Harris and Racine (1974)[H].
ments were made. McClure et al. (see their Fig. 3) over- NGC 6101 - Vhb: CCD - Sarajedini and Da Costa
plotted the CMD of H on their data; no significant differ- (1991); ID: Alcaino (1974b)[A]; Marconi (personal commu-
ence in Vm was found. We obtained two spectra of HI 184 nication) [M] M found that the V photometry of Sarajedini
with low S/N ~ 20. Although we determine this star to be and Da Costa was on average 0.08 mag brighter than M, so
a 94% probable velocity member, S93, who obtained higher ^HB- = -0.08.
precision velocity data, showed that this star is not a veloc-
NGC 6121 = M4 - vH: Clementini et al. (1994);
ity member, and thus it was not used in our W analysis.
Peterson and Latham (1986); Peterson et al. (1986); Ras-
NGC 4833 - Vuq: PG - Menzies (1972); ID: Menzies torguev and Samus (1991); Vhb: PG - Cudworth and Rees
(1972)[M]. (1990); : Cudworth and Rees (1990); ): Lee
NGC 5286 - Vhb : PG - Harris et al. (1976); ID: PG - (1977a)[L]. Photometry was taken from Cudworth and
Harris et al. (1976)[H]. Rees.
NGC 5897 - Vhb : CCD - Ferraro et al. (1992); ID: NGC 6144 - Vhb: PG - Alcaino (1980); ID: Alcaino
Sandage and Katem (1968)[S]. Ferraro et al. (see their Fig. (1980)[A].
NGC 6171 = M107 - Da Costa and Seitzer Sarajedini and Norris was ~ 0.05 mag brighter than R, so
(1989); Piatek et al. (1994); Pryor et al. (1987); V^ : PG- Vrb- -0.05. Note that the star A68 = R111/112
Cudworth et al. (1992); : Cudworth et al. (1992); ID: was listed as one star in A's photometry, but was resolved
Sandage and Katern (1964)[S]. Photometry taken from Cud- into two stars in the photometry of R. Stars Rill and R112
worth et al. have identical magnitudes (V = 15.19) and almost iden-
NGC 6218 = M12 - vH : Harris et al. (1983); Pryor et tical colors (B-V = 1.40 and 1.39 for Rill and R112,
al. (1987); Rastorguev and Samus (1991); VI4B: PG - Ra- respectively). We assumed it was one star with V =
cine (1971); ID: Racine (1971) (R - personal communica- 15.19, since we did not resolve these two stars. This will
tion). not affect our results since the surface gravity and tempera-
NGC 6235 - Vm : FG - Liller (1980a); ID: Liller ture of both stars should be very similar.
(1980a)[L]. NGC 6522 - Vhb : CCD - Temdrup and Walker (1994,
NGC 6254 = MIO - vH: Rastorguev and Samus personal communication; photometry without star names);
(1991); Vhb : CCD - Hurley et al. (1989); ID: Harris et al. ID: Arp (1965)[A]. Photometry was taken from Temdrup
(1976)[H]. Hurley et al. found that their V magnitudes were and Walker, except for star A15, which did not have a
0.18 mag fainter than H, so VH = 0.18. B V mag listed, so the value given in A was used. The
star A116 was resolved into three fainter stars by Temdrup
NGC 6266 = M62 - V^: CCD BV - Caloi et al. and Walker, and was not used in our analysis. The photom-
(1987); ID: Alcaino (1978)[A]. Using 14 stars the V pho- etry for this star is from A.
tometry of Caloi et al. was fainter than A by 0.140.04
mag, so Vhb- = 0.14. NGC 6535 - vH: Pryor and Meylan (1993); Vhb CCD
- Sarajedini (1994); ID: Liller (1980b)[L]; Sarajedini
NGC 6273 = M19 - V^: PG - Harris et al. (1976);
(1994)[S]. Photometry was taken from Sarajedini.
ID: Harris et al. (1976)[H].
NGC 6528 - Vhb : CCD - Ortolani et al. (1992); ID: van
NGC 6304 - Vhb : CCD - Davidge et al. (1992); ID:
den Bergh and Younger (1979)[VY]; Ortolani et al.
Hesser and Hartwick (1976)[H]. Our estimate of the
(1992)[0 (personal communication)]. Photometry was taken
level in was 16.15, whereas the estimate from Davidge et
from O, except -42, which was taken from VY. Using
al. was 16.25, so Vm~ VH = 0.1.
four stars fainter than V = 16.5, the V photometry of O
NGC 6352 - Vhb: CCD - Sarajedini and Noms (1994); was fainter than VY's by 0.05 0.04, and the 5 V
ED: Sarajedini and Norris (1994)[S]; Alcaino (1971)[A]; photometry was redder by 0.24 0.03, so a correction of
Hartwick and Hesser (1972)[H]. Using 17 stars, the V pho- (0,+0.24) was applied to (V, B-V) for the VY photom-
tometry of A was fainter than H by only 0.03 0.09, and etry of -42.
the BV photometry of A was redder than H by only
0.0060.1. Using ten stars, the V photometry of S was NGC 6544 - Vhb: PG - Alcaino (1983); ID: Alcaino
fainter than H by 0.25 0.09, and the 5 V photometry of (1983)[A].
S was bluer than H by 0.21 0.28. Photometry was NGC 6541 - Vhb: PG - Alcaino (1979a); ID: Alcaino
taken from S when available. Otherwise, the mean of H and (1979a) [A].
A was used or, for the cases where H photometry was not NGC 6553 - Vhb: CCD - Ortolani et al. (1990);
available, A was used. When photometry was taken from H ): Hartwick (1975)[H]; Ortolani et al. (1990)[O
and A, 0.25 was added to V, and 0.2 was added to B V. (personal communication)]. Photometry was taken from O.
NGC 6366 - vH: Oa. Costa and Seitzer (1989); V^: Star -3 = 0140 was not used in this analysis, since it
CCD - Harris (1993); ID: Pike (1976)[P]. From the level of is part of the RGB turnover as shown in Fig. 3(b) of O. Star
VHB in the CMD of P, there does not appear to be a -59 was not used due to strong TiO bands (the photom-
systematic V offset from Harris. etry for this star was taken from H).
NGC 6362 - vH: Pryor and Meylan (1993); : CCD NGC 6624 - vH'. Pryor et al. (1989); Pryor and Meylan
- Alcaino and Liller (1986b); ID: Alcaino (1972)[A]. Using (1993); Vhb: CCD - Sarajedini and Norris (1994); ID:
twelve stars, the V photometry of Alcaino and Liller was Liller and Camey (1978)[L]; Richtler (1995)[R]. Photom-
fainter than A by only 0.08 0.1 mag, so no correction etry was taken from R for all stars except LIV 150 and LI
was applied. 102, for which L's photometry was used. Sarajedini and
NGC 6397 - Vhb : digitized PG - Alcaino et al. (1987); Norris, as well as R, found systematic differences with L
ID: Alcaino (1977b)[A]; Cannon (1974)[C]; Alcaino et al. which were correlated with V. Using Figs. 2 and 3 of R, a
(1987)[AB]. Photometry taken from AB, which is a recali- (V, B-V) correction of (+0.2,-0.2) was applied to LIV
bration of the digitized PG photometry of Alcaino and 150, and (+0.1,-0.1) was applied to LI 102. From the
Liller (1980a). level of Vhb in the CMD of R, there do not appear to be
NGC 6496 - Vm : CCD - Sarajedini and Norris (1994); systematic differences with the V photometry of Sarajedini
ID: Armandroff (1988)[A]; Richtler (1995)[R]. Photometry and Norris.
was taken from R. Due to the high quality photometry in NGC 6626 - H'. Pryor et al. (1989); Pryor and Meylan
both studies, it is evident that the in the CMD of (1993); Vhb: PG - Rees and Cudworth (1991); ID: Alcaino
904 RUTLEDGE ET AL,
(1981)[A - written A(ring no.)-(star no.)]. Photometry was NGC 6981 - Vhb * PG - Dickens (1972a); ID: Dickens
taken from Rees and Cudworth, except A1-80, and A2-125, (1972a)[D].
for which A's photometry was used. Using eleven stars, the NGC 7089 - vH: Armandroff and Da Costa (1986);
V photometry of Rees and Cudworth was fainter than A by Vhb: PG - Harris (1975); : Cudworth and Rauscher
0.1 0.08 mag, and the 5-V photometry was bluer by (1987); ID: Harris (1975)[H]. Using nine stars, the V pho-
0.03 0.02 mag, so the (V, B V) corrections applied to tometry of was fainter than Cudworth and Rauscher by
the A stars was (+0.1,0). 0.004 0.07 mag, so the V magnitudes were taken to be the
NGC 6638 - Vhb : PG - Alcaino and Liller (1983); straight mean of and Cudworth and Rauscher, while the
spectroscopy of C-type RR Lyrae - Smith and Stryker B V magnitudes were simply taken from H.
(1986); ID: Alcaino and Liller (1983)[A]. NGC 7099 - : Pryor and Meylan (1993); Vhb : CCD
NGC 6637 = M69 - VHB: CCD - Sarajedini and - Bolte (1987); ID: Dickens (1972b)[D - PG magnitudes,
DP - PE magnitudes]; Alcaino and Liller (1980b)[A]. Pho-
Norris (1994); ID: Hartwick and Sandage (1968)[H- note
tometry was taken from D, or DP. No stars were in com-
that "n" implies the star is from the inner circle); Richtler
mon between Bolte and D, but Buonanno et al. (1988) find
(1995)[R]; Sarajedini and Norris (1994)[S]. Photometry was no significant difference between their photometry and D,
taken from either R or S as indicated by the star names. and since = 15.1 for both the photometry of Buon-
Stars that only had photometry from H were not used due to anno et al. and Bolte, no correction was applied to D. DP 17
their photometric uncertainties (see Fig. 13 of S, and Fig. 4 has an uncertain ID.
of R). From the level of in the CMD of R, the V
photometry of Sarajedini and Norris was 0.1 mag fainter NGC 7492 - Vhb: CCD - Cot et al. (1991); ID: Buo-
than R, so 0.1 was added to the V photometry of R. nanno et al. (1987)[B]; Cuffey (1961)[C]. Photometry was
taken from B. The star CR has an uncertain ID.
NGC 6681 = M70 - vH: Pryor et al. (1989); Vhb"
Pal 12 - vH: Armandroff and Da Costa (1991); Vhb:
CCD - Mittermeier et al. (1994); ID: Hams (1975)[H].
CCD - Stetson et al. (1989); ID: Stetson et al. (1989)[S];
From the level of yHB in the CMD of H, there does not
Harris and Cantema (1980) [H]. Photometry was taken from
appear to be a systematic V offset from Mittermeier et al. S, except H4122, which was taken from H.
NGC 6712 - vH: Grindlay et al. (1987); ^ PG "
Cudworth (1988); : Cudworth (1988); ID: Sandage and
Smith (1966)[S]. Photometry of Cudworth was used for all
stars except SB 67 and SA34, for which the photometry of S
was used. Using eight stars, the V photometry of Cudworth REFERENCES
was fainter than S by 0.120.04 mag, and the B V pho-
Alcaino, G. 1971, A&A, 11, 7
tometry was bluer by 0.02 0.06 mag, so the ( V, B V)
Alcaino, G. 1972, A&A, 16, 220
corrections applied to the S stars was (+0.12,0). Alcaino, G. 1974a, A&AS, 13, 345
NGC 6717 = Pal 9 - VUB : CCD - Jensen et al. (1994); Alcaino, G. 1974b, A&AS, 18, 9
ID: Goranskii (1979)[G]. From the level of V^ in the Alcaino, G. 1977a, A&AS, 29, 9
CMD of G, the V photometry of Jensen et al. was ~ 0.7 Alcaino, G. 1977b, A&AS, 29, 397
Alcaino, G. 1978, A&AS, 32, 379
mag fainter than G, so Vhb = 0.7. All the G pho- Alcaino, G. 1979a, A&AS, 35, 233
tometry was measured by eye except stars G16, G23, G15, Alcaino, G. 1979b, A&AS, 38, 61
and G24, which were measured with an iris-diaphragm pho- Alcaino, G. 1980, A&AS, 39, 315
tometer. Star G35 has an uncertain ID. Alcaino, G. 1981, A&AS, 44, 191
NGC 6723 - 7hb: CCD - Fullton and Carney (1993); Alcaino, G 1983, A&AS, 52, 105
Alcaino, G., and Liller, W. 1980a, AJ, 85, 680
ID: Menzies (1974a)[M]. From the level of V^q in the
Alcaino, G., and Liller, W. 1980b, AJ, 85, 1330
CMD of M, there does not appear to be a systematic V Alcaino, G, and Liller, W. 1980c, AJ, 85, 1592
offset from Fullton and Carney. The star MII-7 was not Alcaino, G., and Liller, W. 1983, AJ, 88, 1166
used in our analysis since it appears to be a blue HB star. Alcaino, G., and Liller, W. 1986a, A&A, 161, 61
NGC 6752 - Vhb: PG - Buonanno et al. (1986); ID: Alcaino, G, and Liller, W. 1986b, AJ, 91, 303
Alcaino, G., Buonanno, R., Caloi, V., Castellani, V., Corsi, C. E.,
Alcaino (1972)[A]; Cannon and Stobie (1973)[C]; Buon-
lannicolo, G, and Liller, W. 1987, AJ, 94, 917
anno et al. (1986)[B]. Using twelve stars, the V photometry Alcaino, G., Liller, W., Alvarado, F., and Wenderoth, . 1991, AJ,
of C was fainter than by 0.03 0.05, so the V magnitudes 102, 159
were taken to be the straight mean of C and B, while the Armandroff, T. E. 1988, AJ, 96, 588
BV magnitudes were simply taken from C. The star A9 Armandroff, T. E., and Zinn, R. 1988, AJ, 96, 92 (AZ88)
= B2403 was not used as it is most likely a variable, or Armandroff, T. E., and Da Costa, G. S. 1986, AJ, 92, 111
was contaminated in B's study, who obtained a V magni- Armandroff, . E, and Da Costa, G. S. 1991, AJ, 101, 1329
tude 0.35 mag fainter than A and C. (AD91)
Armandroff, T. E., Da Costa, G. S., and Zinn, R. 1992, AJ, 104,
NGC 6809 - vH: Pryor et al. (1991); Pryor and Meylan 164 (ADZ92)
(1993); Fhb: PG - Lee (1977d); ID: Lee (1977d)[L]. Arp, H. 1965, ApJ, 141, 43
Bergbusch, P. . 1993, AJ, 106, 1024 Harris, W. E., and Racine, R. 1974, AJ, 79, 472
Boite, M. 1987, ApJ, 319, 760 Harris, W. E., and Cantema, R. 1980, ApJ, 239, 815
Boite, M. 1992, ApJS, 82, 145 Harris, W. E., Racine, R., and de Roux, J. 1976, ApJS, 31, 13
Bond, H. E., Hams, W.E., Hams, G.L.H., Mittermeier, R., and Hartwick, F. D. ., and Hesser, J. E. 1972, ApJ, 175, 77
Palmer, . 1994, in preparation Hartwick, F. D. A. 1975, PASP, 87, 77
Brewer, J. P., Fahlman, G. G., Richer, . B., Searle, L., and Hartwick, F. D. ., and Sandage, A. 1968, ApJ, 153, 715
Thompson, L 1993, AJ, 105, 2158 Hesser, J. E., and Hartwick, F. D. A. 1976, ApJ, 203, 113
Buonanno, R., Corsi, C. E., and Fusi Pecci, F. 1981, MNRAS, Hesser, J. E., Shawl, S. J., and Meyer, J. E. 1986, PASP, 98, 403
196, 435 Hesser, J. E., Harris, W. E., VandenBerg, D. ., Allwright, J. W.
Buonanno, R., Caloi, V., Castellani, V., Corsi, C., Fusi Pecci, F., B., Shott, P., and Stetson, P. B. 1987, PASP, 99, 739
and Gratton, R. 1986, A&AS, 66, 79 Hurley, D. J. C, Richer, . B., and Fahlman, G. G. 1989, AJ, 98,
Buonanno, R., Corsi, C. E., Ferraro, L, and Fusi Pecci, F. 1987, 2124
A&AS, 67, 327 Janes, . ., and Heasley, J. N. 1988, AJ, 95, 762
Buonanno, R., Caloi, V., Castellani, V., Corsi, C. E., and Ferraro, Jensen, N., Harris, W. E., and Bond, . E. 1994, in preparation
L 1988, A&AS, 74, 353 Lee, S. -W. 1977a, A&AS, 27, 376
Caloi, V., Castellani, V., and Piccolo, F. 1987, A&AS, 67, 181 Lee, S. -W. 1977b, A&AS, 27, 381
Cannon, R. D., and Stobie, R S. 1973, MNRAS, 162, 227 Lee, S. -W. 1977c, A&AS, 28, 408
Cannon, R. D. 1974, MNRAS, 167, 551 Lee, S. -W. 1977d, A&AS, 29, 1
Clementini, G., Merighi, R., Pasquini, L., Cacciari, C, and Liller, M. H. 1980a, AJ, 85, 673
Gouiffes, C. 1994, MNRAS, 267, 83 Liller, M. H. 1980b, AJ, 85, 1480
Cot, P., Richer, H. B., and Fahlman, G. G. 1991, AJ, 102, 1358 Liller, M. H., and Carney, B. W. 1978, ApJ, 224, 383
Cot, P., Welch, D. L., Fischer, P., and Gebhardt, . 1995, ApJ, McClure, R. D., VandenBerg, D. ., Bell, R. ., Hesser, J. ., and
454, 788 Stetson, P. B. 1987, AJ, 93, 1144
Cudworth, . M. 1979, AJ, 84, 1866 Menzies, J. 1972, MNRAS, 156, 207
Cudworth, . M. 1988, AJ, 96, 105 Menzies, J. 1974a, MNRAS, 168, 177
Cudworth, . M., and Rauscher, B. J. 1987, AJ, 93, 856 Menzies, J. 1974b, MNRAS, 169, 79
Cudworth, . M., and Rees, R 1990, AJ, 99, 1491 Meylan, G., and Mayor, M. 1986, A&A, 166, 122
Cudworth, K. M., Smetanka, J. J., and Majewski, S. R. 1992, AJ, Meylan, G., Dubath, P., and Mayor, M. 1991, ApJ, 383, 587
103, 1252 Mittermeier, R., Harris, G. L. H., Bond, . E., Harris, W. E., and
Cuffey, J. 1961, MNRAS, 122, 363 Palmer, K. 1994, in preparation
Da Costa, G. S., and Seitzer, P. 1989, AJ, 97, 405 Olszewski, E. W., Cantema, R., and Harris, W. E. 1984, ApJ, 281,
Da Costa, G. S., and Armandroff, T. E. 1995, AJ, 109, 2533 158
(DA95) Olszewski, E. W., Peterson, R. C., and Aaronson, M. 1986, ApJ,
Da Costa, G.S., Armandroff, T. E., and Norris, J. E. 1992, AJ, 302, L4
104, 154 (DAN92) Olszewski, E. W., Schommer, R. ., Suntzeff, . B., and Harris,
Davidge, T. J., Harris, W. E., Bridges, T. J., and Hanes, D. A. H. 1991, AJ, 101, 515
1992, ApJS, 81, 251 Ortolani, S., Barbuy, B., and Bica, E. 1990, A&A, 236, 362
Dickens, R. J. 1972a, MNRAS, 157, 281 Ortolani, S., Bica, E., and Barbuy, B. 1992, A&AS, 92, 441
Dickens, R. J. 1972b, MNRAS, 157, 299 Peterson, R. C, and Latham, D. W. 1986, ApJ, 305, 645
Djorgovski S., and Meylan G. 1993, in Structure and Dynamics of Peterson, R. C., Olszewski, E. W., and Aaronson, M. 1986, ApJ,
Globular Clusters, ASP Conf. Series 50, ed. G. Meylan and S. 307, 139
Djorgovski (ASP, San Francisco), p. 325 Piatek, S., Pryor, C, McClure, R. E., Fletcher, J. M., and Hesser,
Ferraro, F. R., Clementini, G., Fusi Pecci, F., Buonanno, R., and J. E. 1994, AJ, 107, 1397
Alcaino, G. 1990, A&AS, 84, 59 Pike, C. D. 1976, MNRAS, 177, 257
Ferraro, F. R., Fusi Pecci, F., and Buonanno, R. 1992, MNRAS, Press, W. H., Teukolsky, S. ., Vetterling, W. T., and Flannery,
256, 376 B.P. 1992, Numerical Recipes in Fortran, 2nd ed. ( Cambridge,
Ferraro, F. R., Clementini, G., Fusi Pecci, F., Vitiello, E., and Press Syndicate of the University of Cambridge)
Buonanno, R 1993, MNRAS, 264, 273 Pryor, C, McClure, R. D., Fletcher, J. M., and Hesser, J. E. 1987,
Fischer, P., Welch, D. L., Mateo, M., and Cote, P. 1993, AJ, 106, in Globular Cluster Systems in Galaxies, IAU Symp. 126, ed.
1508 J. Grindlay and A. G. D. Philip (Dordrecht, Reidel), p. 661
Friel, E. D., and Geisler, D. 1991, AJ, 101, 1338 Pryor, C, McClure, R. D., Fletcher, J. M., and Hesser, J. E. 1989,
Fullton, L K., and Carney, B. W. 1993, BAAS, 25, 1407 AJ, 98, 596
Geisler, D., Piatti, A. E., Clari, J., and Minniti, D. 1995, AJ, 109, Pryor, C, McClure, R. D., Fletcher, J. M., and Hesser, J. E. 1991,
605 (G95) AJ, 102, 1026
Goranskii, Y. P. 1979, SvA, 23, 284 Pryor, C, and Meylan, G. 1993, in Structure and Dynamics of
Grindlay, J., Bailyn, C, Mathieu, R., and Latham, D. 1987, in Globular Clusters, ASP Conf. Series 50, ed. G. Meylan and S.
Globular Cluster Systems in Galaxies, IAU Symp. 126, ed. J. Djorgovski (ASP, San Francisco), p. 357 (PM93)
Grindlay and A. G. D. Philip (Dordrecht, Reidel), p. 659 Racine, R 1971, AJ, 76, 331
Harris, H. C. 1993, AJ, 106, 604 Rastorguev, A. S., and Samus, . N. 1991, SvAL, 17, 388
Harris, H. C., Nemec, J. M., and Hesser, J. E. 1983, PASP, 95, Reed, B. C., Hesser, J. E., and Shawl, S. J. 1988, PASP, 100, 545
256 Rees, R. F., and Cudworth, . M. 1991, AJ, 102, 152
Harris, W. E. 1975, ApJS, 29, 397 Richtler, T. 1995, A&AS, 109, 1
Harris, W. E. 1978, PASP, 90, 45 Sandage, ., and Katem, B. 1964, ApJ, 139, 1080
Harris, W. E. 1982, ApJS, 50, 573 Sandage, ., and Katem, B. 1968, ApJ, 153, 569
Harris, W. E. 1996, AJ, 112, 1487 Sandage, ., and Smith, L. L. 1966, ApJ, 144, 886
906 RUTLEDGE ET AL.
Sandquist, . L., Boite, M., Stetson, P. ., and Hesser, J. . 1996, Suntzeff, . B., Schommer, R. ., Olszewski, E. W., and Walker,
ApJ, 470, 910 A. R. 1992, AJ, 104, 1743 (S92)
Sarajedini, A. 1994, PASP, 106, 404 Suntzeff, . B., Mateo, M., Temdmp, D. M., Olszewski, E. W.,
Sarajedini, ., and Da Costa, G. S. 1991, AJ, 102, 628 Geisler, D., and Weller, W. 1993, ApJ, 418, 208 (S93)
Sarajedini, ., and Norris, J. E. 1994, ApJS, 93, 161 Suntzeff, N. B., and Kraft, R. P. 1996, AJ, 111, 1913 (SK96)
Smith, . ., and Stryker, L. L. 1986, PASP, 98, 453 Temdrup, D., and Walker, A. 1994, AJ, 107, 1786
Stetson, P. . 1989, Image and Data Processing, ed. B. Barbuy, E. Tucholke, H. J. 1992a, A&AS, 93, 293
Janot-Pacheco, . M. Magallaes, and S.M. Viegas, published Tucholke, H. J. 1992b, A&AS, 93, 311
by Departamento Astronoma, Instituto Astronmico e Geof- van den Bergh, S., and Younger, F. 1979, AJ, 84, 1305
sico, Universidade de Sao Paulo, C.P. 9638, Sao Paulo 01065, Webbink, R. F. 1981, ApJS, 45, 259
Brazil Webbink, R. F. 1985, in Dynamics of Star Clusters, IAU Symp.
Stetson, P. ., Vandenberg, D. ., Boite, M., Hesser, J. E., and 113, ed. J. Goodman and P. Hut (Dordrecht, Reidel), p. 541
Smith, G. H. 1989, AJ, 97, 1360 Zinn, R. 1985, ApJ, 293, 424
Storm, J., Carney, . W., and Beck, J. 1991, PASP, 103, 1264 Zinn, R., and West, M. J. 1984, ApJS, 55, 45 (ZW84)

1997pasp 109 883R

Uploaded by

Copyright:

Available Formats

1997pasp 109 883R

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

1997pasp 109 883R

Uploaded by

Copyright:

Available Formats

Publications of the Astronomical Society of the Pacific

109: 883-906, 1997 August

Galactic Globular-Cluster Metallicity Scale from the Ca II

Glen A. Rutledge, James E. Hesser,1 and Peter B. Stetson

883 1997. Astronomical Society of the Pacific

mentioned earlier. To facilitate the placement of the sky

50.0 100.0 150.0

50.0 100.0 150.0

Table 7 technique, allowing for errors in both directions, is

2.0 3.0 4.0 5.0 6.0 0.5 1.0 1.5

- 1Yl NGC 6637 ~

You might also like