1: \documentclass[12pt,preprint]{aastex}
2: \usepackage{graphicx}
3: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: %\usepackage{txfonts}
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: %
7: \shorttitle{Multiple Stellar Populations in GCs}
8: \shortauthors{Milone et al.}
9:
10: \newcommand{\hst}{{\it HST\/}}
11:
12: \begin{document}
13: \title{The ACS Survey of Galactic Globular Clusters. III. The Double
14: Subgiant Branch of NGC 1851\thanks{Based on observations with the
15: NASA/ESA {\it Hubble Space Telescope}
16: under the programs GO-10775 and GO-5696,
17: obtained at the Space Telescope
18: Science Institute, which is operated by AURA, Inc., under NASA
19: contract NAS 5-26555.}}
20:
21: \author{ A.\ P.\ Milone\altaffilmark{1},
22: L.\ R.\ Bedin\altaffilmark{2},
23: G.\ Piotto\altaffilmark{1},
24: J.\ Anderson\altaffilmark{3},
25: I.\ R.\ King\altaffilmark{4},
26: A.\ Sarajedini\altaffilmark{5},
27: A.\ Dotter\altaffilmark{6},
28: B.\ Chaboyer\altaffilmark{6},
29: A.\ Mar\'\i n-Franch\altaffilmark{7},
30: S.\ Majewski\altaffilmark{8},
31: A.\ Aparicio\altaffilmark{7},
32: M.\ Hempel\altaffilmark{5},
33: N.\ E.\ Q.\ Paust\altaffilmark{2},
34: I.\ N.\ Reid\altaffilmark{2},
35: A. Rosenberg\altaffilmark{7},
36: M.\ Siegel\altaffilmark{9}}
37: \altaffiltext{1} {Dipartimento di Astronomia, Universit\`a di
38: Padova, Vicolo dell'Osservatorio 3, Padova, I-35122, Italy}
39: %email{antonino.milone-giampaolo.piotto@unipd.it }
40: \altaffiltext{2} {Space Telescope Science Institute, 3700 San Martin Drive,
41: Baltimore, MD 21218, USA}
42: \altaffiltext{3}{Department of Physics and Astronomy, Mail Stop 108,
43: Rice University, 6100 Main Street, Houston, TX 77005, USA}
44: % \email{jay@eeyore.rice.edu}
45: \altaffiltext{4}{Department of Astronomy, University of Washington,
46: Box 351580, Seattle, WA 98195-1580, USA}
47: % \email{king@astro.washington.edu}
48: \altaffiltext{5}{Department of Astronomy, University of Florida,
49: 211 Bryant Space Science Center, Gainesville, FL 32611, USA}
50: \altaffiltext{6}{Department of Physics and Astronomy, Dartmouth College,
51: 6127 Wilder Laboratory, Hanover, NH 03755, USA}
52: \altaffiltext{7}{Instituto de Astrof\`\i sica de Canarias, E-38200 La
53: Laguna, Canary Islands, Spain}
54: \altaffiltext{8}{Dept. of Astronomy, University of Virginia,
55: P.O. Box 400325, Charlottesville, VA 22904-4325}
56: \altaffiltext{9}{University of Texas, McDonald Observatory, 1
57: University Station, C1402, Austin TX, 78712}
58: % \and
59: % }
60: % }
61: %\date{Received Xxxxx xx, xxxx; accepted Xxxx xx, xxxx}
62: %__________________________________________________________________
63: %
64:
65: \begin{abstract}
66: %
67: Photometry with {\sl HST\/}'s ACS reveals that the subgiant branch (SGB)
68: of the globular cluster NGC 1851 splits into two well-defined
69: branches. If the split is due only to an age effect, the two SGBs
70: would imply two star formation episodes separated by $\sim$ 1 Gyr. We
71: discuss other anomalies in NGC 1851 which could be interpreted in
72: terms of a double stellar population. Finally, we compare the case of
73: NGC 1851 with the other two globulars known to host multiple
74: stellar populations, and show that all three clusters differ in
75: several important respects.
76: \end{abstract}
77: %
78: \keywords{globular clusters: individual (NGC 1851) ---
79: Hertzsprung-Russell diagram }
80:
81:
82: %________________________________________________________________
83: %
84:
85:
86: %%%
87: %
88: \section{Introduction}
89: %
90: %%%
91:
92: For many decades, globular clusters (GC) have been considered the
93: simplest possible stellar populations, made up of stars located at the
94: same distance, formed at the same epoch and from the same material.
95: Although anomalies had been noted from time to time in the abundances
96: of a number of individual elements
97: (see discussion in Gratton, Sneden, \& Carretta 2004),
98: the description of populations in terms of a
99: helium abundance and an overall heavy-element abundance seemed firm.
100: The ``second-parameter problem'' has continued to be a pain for many
101: decades, and unusual HB morphologies were turning up more often, but the
102: implications for the origin and evolution of GC stars remained
103: ambiguous.
104: Similarly, theoreticians had predicted that self-enrichment
105: (Cotrell \& Da Costa 1981, Ventura et al.\ 2001, Ventura \& D'Antona 2005, Maeder \& Meynet 2005)
106: or mergers might generate multiple populations in clusters, but such
107: predictions remained controversial
108: (see, e.g., Fenner et al.\ 2004, Bekki \& Norris 2006),
109: and had little observational basis.
110: Most important for our discussion here, the idea of one clear sequence
111: in each part of the HR diagram of each cluster stood firm.
112:
113: Now, however, the paradigm of GC hosting simple stellar populations
114: has been seriously challenged by the discovery of multiple
115: evolutionary sequences
116: in $\omega$~Centauri (Bedin et al.\ 2004)
117: and NGC 2808 (Piotto et al.\ 2007, P07). In both clusters the
118: evidence that stars must have formed in distinct bursts is the
119: presence of multiple main sequences (MSs). However, the two objects
120: differ in at least two important aspects: The stars of $\omega$ Cen
121: have a large spread in metal content, whereas in NGC 2808 only oxygen
122: and sodium are observed to vary much. Omega Cen shows at least four
123: distinct subgiant branches (SGBs), implying a range of ages, while the narrow
124: turnoff in NGC 2808 implies that there is little or no difference in
125: age among its populations. One more cluster, M54, shows a complex
126: color-magnitude diagram (see, e.g., Layden \& Sarajedini 2000),
127: including a bifurcated SGB (see paper IV of this series,
128: Siegel et al.\ 2007). This cluster has been shown,
129: however, in too many papers to cite here, to be a part of the
130: Sagittarius dwarf galaxy that is in process of merging into the Milky
131: Way, and very possibly the actual nucleus of that galaxy. (Actually,
132: it is still matter of debate which parts of the color-magnitude
133: diagram of M54 represent the cluster population and which ones are due to the
134: Sagittarius stars.) Even though $\omega$ Cen could very well
135: represent a similar situation, we feel that M54 is very different from
136: the clusters that we discuss here, and we will therefore not include
137: it in the discussions of this paper.
138:
139: The puzzling observational facts for $\omega$~Cen and NGC 2808
140: call for a more careful analysis of the MS, turnoff (TO), and subgiant
141: branch of other GCs. In this respect, the color-magnitude
142: diagram (CMD) database from the \hst\ Treasury program GO-10755 (see
143: Sarajedini et al.\ 2007) provides a unique opportunity to search for
144: anomalies in the different evolutionary sequences of other Galactic
145: GCs.
146: Even though the narrow F606W $-$ F814W color baseline
147: %
148: is far from ideal for identifying multiple main sequences, our
149: observing strategy was devised to have a very high signal/noise ratio
150: at the level of the TO, and therefore the CMDs are perfectly suitable
151: for identifying multiple TO/SGBs, such as those as found in $\omega$~Cen.
152: (New \hst\ observations specifically devoted to the identification of
153: multiple MSs are already planned for other massive GCs, in
154: GO-10922 and GO-11233.)
155:
156: Indeed, a first look at the entire Treasury database showed at least
157: one other GC whose CMD clearly indicates the presence of multiple
158: stellar populations: NGC 1851. Although NGC 1851 is a massive GC with
159: a prototypical bimodal horizontal branch (HB), not much attention has
160: been devoted to the study of its stellar population. The most complete
161: photometric investigation of this cluster is by Saviane et al.\
162: (1998), and the most extended spectroscopic analysis was done over 25
163: years ago, by Hesser et al.\ (1982). Saviane et al.\ (1998) confirmed
164: the bimodal nature of the HB of NGC 1851, but did not find any other
165: anomaly in its CMD. Hesser et al.\ (1982) found that three out of
166: eight bright red giant branch (RGB) stars have extremely strong CN
167: bands.
168:
169: The present paper is based on new photometry of NGC 1851, from
170: \hst/ACS imaging and from archival WFPC2 images. Section 2 describes
171: the data sets, and \S\ 3 presents the color-magnitude diagram from the
172: ACS data, with a split in the subgiant branch clearly visible. In \S\
173: 4 we use the WFPC2 data, both for proper-motion elimination of field
174: stars and to construct a CMD with a broader wavelength baseline which
175: we use to set a severe upper limit on any spread in the main sequence.
176: Section 5 discusses the properties of the SGB and of the horizontal
177: branch, and shows that the two parts of each have the same spatial
178: distribution. Section 6 considers the ages of the two populations,
179: and \S\ 7 is a discussion and summary.
180:
181: %__________________________________________________________________
182: %
183: \section{Observations and Measurements}
184: %
185: The main database used in this paper comes from {\sl HST} ACS/WFC
186: images in the F606W and F814W bands, taken for GO-10775 (P.I.\ Sarajedini).
187: In addition, we use {\sl HST} WFPC2 archive data from GO-05696 (P.I.\
188: Bohlin) to extend the wavelength range to the blue, as well as to
189: obtain proper motions. The data sets are summarized in Table 1.
190:
191: %\begin{center}
192: \begin{table}
193: %\centering
194: \label{t1}
195: \begin{tabular}{cccc}
196: \hline\hline DATE & EXPOSURES & FILT & PROGRAM\\
197: \hline
198: %-------------------------------------------------------------------------------------
199: \multicolumn{4}{c} {WFPC2}\\
200: \hline
201: April 10, 1996 & 4$\times$900s & F336W & 5696 \\
202: \hline
203: \multicolumn{4}{c} {ACS/WFC}\\
204: \hline
205: July 1, 2006 & 20s$+$5$\times$350s & F606W & 10775 \\
206: July 1, 2006 & 20s$+$5$\times$350s & F814W & 10775 \\
207: %----------------------------------------------------------------------
208: \hline
209: \end{tabular}
210: \caption{Description of the data sets used in this work.}
211: \end{table}
212: %\end{center}
213:
214: The photometric and astrometric measurements were made for both WFPC2
215: and ACS using the algorithms described by Anderson \& King (2000, 2006).
216: Each image was reduced independently, with the observations averaged to
217: produce a single flux for each star in each filter.
218: %
219: We put the WFC photometry into the Vega-mag system following the
220: procedures in Bedin et al.\ (2005) and adopting the encircled energy
221: and zero points as given by Sirianni et al.\ (2005). WFPC2 data were
222: photometrically calibrated following Holtzman et al.\ (1995). The
223: proper motions were derived as in Bedin et al.\ (2006).
224:
225: Breathing can change the focus of the telescope, which can result in
226: small spatial variations of the PSF relative to the library PSF, and
227: consequently in small spatial variations of the photometric zero
228: point. In order to deal with any residual PSF variation, we used a
229: method similar to the one described in Sarajedini et al.\ (2007) to
230: perform a spatial fit to the color residuals relative to the
231: main-sequence ridge line, and remove them. In this case,
232: however, the corrections were made only to colors, rather than along a
233: reddening line, because, as noted in Section 3, the reddening of NGC
234: 1851 is so small that what we are correcting is surely differences
235: between the spatial variations of the PSF in the two filters, rather
236: than differential reddening. The spatially dependent correction was
237: at the level of $\sim$0.01 magnitude, and merely sharpens the
238: sequences a little.
239:
240:
241: \section{The ACS/WFC Color-Magnitude Diagram}
242: %
243: %%%
244:
245: %__________________________________________________________________
246: %
247: \begin{figure}
248: \epsscale{.80}
249: \plotone {f1.ps}
250: \caption{{\it Left}: Original CMD. {\it Right}: CMD corrected for spatial
251: variation of the photometric zero point of color.
252: }
253: \label{f1}
254: \end{figure}
255: %__________________________________________________________________
256:
257: In Figure~\ref{f1} we show the color-magnitude diagram for the
258: ACS/WFC photometry. The left panel shows the original CMD. The main
259: result is already evident:\ the subgiant branch (SGB) is split into two
260: distinct branches. The right panel shows the CMD after the correction
261: for the spatial variation of the color zero point. The SGB split
262: is still there, and there are clearly two sequences down to the
263: cluster TO. No separation is evident in the main sequence. It must be
264: immediately noted that because of the narrowness of MS below the TO,
265: the SGB broadness can be due neither to any residual spatial variation
266: of the zero point nor to any differential reddening, which is expected
267: to be negligible because of the low reddening of NGC 1851
268: [$E(B-V)=0.02$, Harris 1996].
269:
270: The two SGBs remain separated by $\sim 0.1$\ mag in the magnitude
271: interval $18.0<m_{\rm F606W}<19.0$, over the color interval $0.5<m_{\rm
272: F606W}-m_{\rm F814W}<0.6$. Hereafter we will refer to the two SGBs
273: as bSGB and fSGB, where b stands for brighter and f stands for fainter.
274:
275: Omega Cen and NGC 2808 also show evidence of multiple populations,
276: but we note that the split in the SGB of NGC 1851 is quite different
277: from what we found in the former two cases. In NGC 2808, the main
278: sequence splits into three distinct sequences a couple of magnitudes below
279: the turnoff, but the SGB shows no evidence of splitting. In Omega Cen,
280: both the SGB and the MS show evidence of splitting. Here, it is only
281: the SGB that is split. We will see that the main sequence of NGC1851
282: appears consistent with a single population.
283:
284:
285: %%%
286: %
287: \section{The WFPC2 Data}
288:
289:
290: %%%
291: %
292: \subsection{Proper motions for field decontamination}
293: %
294: %%%
295:
296: We used the archival WFPC2 images as a first epoch to determine proper
297: motions. WFPC2 has a field of view that is only $\sim$50\% of the ACS
298: field, and the crowding prevented us from measuring good WFPC2
299: positions for stars within 500 ACS/WFC pixels of the cluster center.
300: This somewhat reduces the size of our sample.
301:
302: In the left panel of Fig.~\ref{f2} we show the CMD, using the F336W
303: magnitudes from the WFPC2 images and the F814W magnitudes from the ACS
304: images. The second column of panels shows the proper-motion diagrams
305: of the stars for four different magnitude intervals. It should be
306: noted that we measure proper motions relative to a reference frame
307: made up of only cluster members, so that the zero point of our motions
308: is the mean motion of the cluster. In the third column the magnitudes
309: of the proper motion vectors are plotted against stellar magnitude.
310: The errors clearly increase toward fainter magnitudes. The line was
311: drawn in order to isolate the stars that have member-like motions, and
312: was derived as follows. First, we note that in our study of SGB stars
313: it is more important to have a pure cluster sample than a complete
314: one, so we make a conservative choice. Taking each interval of 1
315: magnitude in F336W, we begin by estimating the sigma of the member-like
316: motions. To do this we find the size of proper motion that includes
317: 70\% of the stars; for a bivariate Gaussian this radius should lie at
318: $1.552\sigma$, where $\sigma$ refers to each dimension. We then plot
319: a point at $2.448\sigma$, which is the size that should include 95\%
320: of the cluster stars. We use spline interpolation to draw a line
321: through these points, and choose all the stars to the left of the
322: line. By giving up 5\% of the cluster stars we effectively exclude
323: nearly all of the field stars. The rightmost panel shows the CMD of
324: our chosen cluster stars.
325:
326: In Fig.~\ref{f3} we summarize the results. The top left panel shows a
327: zoomed CMD around the SGB for the stars in our proper-motion-selected
328: sample. Stars marked in red are are bSGB stars (119 objects), and the
329: ones in blue are fSGB stars (88 objects). In the bottom right panel is
330: the vector point diagram for the same stars. The two SGBs seem to have
331: the same proper motion distribution. To show this more clearly, in the
332: adjacent panels we plot normalized histograms of the separate components
333: of the proper motions, for the stars of the three groups separately.
334: Gaussian fits are also shown.
335:
336:
337: %__________________________________________________________________
338: \begin{figure}
339: \epsscale{.80}
340: \plotone{f2.ps}
341: \caption{{\it Left}: F336W vs.\ F336W $-$ F814W CMD for all stars.
342: {\it Second column}: Proper-motion
343: diagram of the stars in the left panel, in intervals
344: of 1.5 magnitudes. {\it Third column}: The
345: total proper motion relative to the cluster mean proper
346: motion. {\it Right}: The
347: proper-motion-selected CMD. }
348: \label{f2}
349: \end{figure}
350: %__________________________________________________________________
351: %
352:
353: \begin{figure}
354: \epsscale{.80}
355: \plotone{f3.ps}
356: \caption{{\it Top left}: The same CMD as in Fig.~\ref{f2}, zoomed
357: around the SGB, bSGB stars in red, fSGB stars in blue.
358: {\it Bottom right:} proper motions, with same color
359: coding. The other two panels show histograms of the
360: individual components of proper motion.}
361: \label{f3}
362: \end{figure}
363: %__________________________________________________________________
364: %
365:
366:
367: A side-benefit of proper-motion cleaning is that it also removes from
368: the CMD stars
369: %
370: that did not have well-measured positions. These stars tend to have
371: poor photometry as well, so removing them naturally improves the quality
372: of the CMD.
373:
374:
375:
376: %%%
377: %
378: \subsection{Wide color baseline}
379: \label{ss_widecolor}
380: %
381: %%%
382:
383: The second benefit of the WFPC2 data is that it provides a much better
384: color baseline than the ACS data alone; our study of NGC 2808 (P07)
385: showed that a wider color baseline yields a more significant splitting
386: of the main sequence. The CMD in Figure~\ref{f2} shows F336W vs.\ the
387: F336W $-$ F814W color. The SGB split is even more clear in this figure
388: than in the $V-I$ CMD of Figure~\ref{f1} from the ACS data alone --- but
389: what is more important is that we now have a better opportunity to study
390: any possible color spread in the main sequence. Any limits on the
391: intrinsic width of the main sequence will translate directly into upper
392: limits on spreads in the helium and heavy-element abundances, since age
393: differences have little effect below the turnoff. (We note, however,
394: the possibility that simultaneous changes in He and heavy elements could
395: have effects that offset each other.)
396:
397: Fig.~\ref{f4} shows the color distribution of the MS stars. The left
398: panel repeats a part of the proper-motion-selected CMD that we showed
399: in Fig.~\ref{f2}. In order to derive the main sequence ridge line
400: (MSRL) we divided the CMD into intervals of 0.2 mag in the F336W band
401: and computed the median color in each interval. By fitting these
402: median points with a spline, we obtained a raw fiducial line. Then for
403: each star we calculated the difference in color with respect to the
404: fiducial, and took as the $\sigma$ the location of the 68th percentile
405: of the absolute values of the color differences. All stars
406: %
407: with a color distance from the MSRL greater than $4\sigma$ were
408: rejected, and the median points and the $\sigma$ were redetermined. The
409: middle panel of Fig.~\ref{f4} shows the same
410: CMD as the left panel, but after subtracting from each star the MSRL
411: color appropriate for its F336W magnitude. Finally, the right panel
412: shows the histograms of the color distributions in five different
413: magnitude intervals.
414:
415: The color distribution is fairly well reproduced by a Gaussian, and
416: there is no evidence of a split of the MS, such as observed in
417: $\omega$~Cen or NGC 2808. Since most of the error in color will come
418: from the WFPC2 photometry, which will at best have an error of 0.02
419: mag (Anderson \& King 2000), a 0.01 mag color error in the WFC
420: (Anderson \& King 2006) will give us an expected color error of 0.025
421: mags from measuring error alone. As Fig.~\ref{f4} shows that the MS
422: broadening can be represented with a Gaussian with $\sigma\sim0.05$,
423: we obtain an upper limit for the intrinsic color dispersion of the MS
424: of NGC 1851 of $\sigma_{MS}=0.04$.
425:
426:
427: %__________________________________________________________________
428: \begin{figure}
429: \epsscale{.80}
430: \plotone{f4.ps}
431: \caption{{\it Left}: The same CMD as shown in the right-hand panel
432: of Fig.~\ref{f2}, with our MSRL overplotted; {\it Middle}: The CMD,
433: rectified by subtraction of the MSRL; {\it Right}: Color
434: distribution of the rectified CMD. The $\sigma$ in the inset are
435: those of the best-fitting Gaussians.}
436: \label{f4}
437: \end{figure}
438: %__________________________________________________________________
439: %
440:
441:
442: %%%
443: %
444: \section{Two Stellar populations in NGC 1851}
445: %
446: %%%
447: \label{SGB}
448:
449: In this section we will derive the basic properties of the two SGBs,
450: and consider
451: %
452: whether other features in the CMD may indicate the presence
453: of two stellar populations in NGC 1851.
454:
455:
456: \subsection{SGB population ratios}
457:
458: \begin{figure}
459: \epsscale{.80}
460: \plotone{f5.ps}
461: \caption{
462: {\it Top}: F606W vs.\ F606W $-$ F814W diagram zoomed around the SGB. The
463: red continuous lines show the adopted fiducial line for the bSGB; the
464: other lines are explained in the text. {\it Middle}: The same diagram,
465: after subtraction of the bSGB fiducial line. The lines are explained in
466: the text. {\it Bottom}: Distribution in magnitude of the bSGB stars
467: (red) and fSGB (blue). The numbers indicate the fraction of stars in
468: each SGB, as the area of the best-fitting Gaussians, shown in the plot.
469: (See the text for more details.)
470: %
471: }
472: \label{f5}
473: \end{figure}
474:
475: Figure \ref{f5} shows how we fitted Gaussians to the magnitude
476: distributions of the bSGB and fSGB stars to estimate their number ratio.
477: The top panel of the figure is a blow-up of the part of the SGB where
478: the split occurs. The red line is a fiducial line through the bSGB. We
479: drew it by marking the middle magnitude level of bSGB at four equally
480: spaced points and drawing a line through them by means of a spline fit.
481: In the middle panel the magnitude of each star has had subtracted from it
482: the magnitude of this fiducial line at the color of the star. For the
483: analysis that was to follow, we divided the color range into four
484: sections, as shown.
485:
486: In each color section we estimated the fraction of stars in each of the
487: SGBs, as follows. Our aim was to fit the magnitude distributions in the
488: two branches by a pair of overlapping Gaussians, but we felt the need to
489: eliminate outliers, especially in the bluest color section. This we did
490: by making a preliminary least-squares fit of the Gaussians, using all
491: of the stars. We then repeated the solution, but omitting the stars
492: that lay more than $2\sigma$ above the midpoint of the Gaussian for bSGB
493: or more than $2\sigma$ below the midpoint of the Gaussian for fSGB
494: (using in each case the sigma of the corresponding Gaussian).
495: In the middle panel of Fig.~\ref{f5} the horizontal lines show the
496: midpoint of the final Gaussians. The red dashed line runs $2\sigma_b$
497: on the bright side of the bSGB, and the blue dashed line runs
498: $2\sigma_f$ on the faint side of the fSGB (where the $\sigma$'s are
499: those of the best-fitting Gaussians in each color section, fitted to the
500: distributions of the bSGB and fSGB stars, respectively). The stars
501: rejected consist of field stars, objects with poor photometry, or
502: binaries (brighter than the bSGB). We have also marked, in the top
503: panel of Fig. \ref{f5}, the values of the Gaussian centers of bSGB with
504: large dots and the spline through them with continuous lines. The
505: crosses are the Gaussian centers of fSGB lowered by $2\sigma_f$, and the
506: dashed lines are a spline fitted through the crosses.
507:
508: The bottom section of Fig.~\ref{f5} shows the histogram of the magnitude
509: distribution in each color section of the middle panel, and the
510: best-fitting Gaussians. The numbers reported in the figures give the
511: percentage of total area under each of the two Gaussians, i.e., the
512: percentage of bSGB (red) and fSGB (blue) in each interval.
513:
514: When we added up, for each SGB, the areas under the four Gaussians
515: fitted to the separate color sections, we found that 55\% of the stars
516: belong to the bSGB, and 45\% to the fSGB. Although the expected sigma
517: of samplings from a binomial distribution of more than 1000 stars, with
518: a probability of 0.55, is less than 2\%, we estimate that our observed
519: 55/45 split really has an uncertainty that is more like 5\%. This
520: estimate is based on the differing results of several different ways of
521: fitting the Gaussians (e.g., as simple a thing as changing the bin-width
522: of the histograms).
523:
524: As for the intrinsic width of each SGB, the dispersions in magnitude of
525: the two Gaussians in the color sections where the two SGBs are well
526: separated are of the order of 0.02 mag, consistent with the
527: uncertainties in our corrections for spatial variation of the PSF.
528: (Paradoxically, WFC colors are more accurate than WFC magnitudes,
529: because these corrections are what limits the accuracy of high-S/N
530: photometry, and in the case of colors the corrections are only for the
531: difference in behavior of the PSF through two different filters.)
532:
533:
534: \subsection{Spatial distribution of the SGB stars}
535:
536: Figure \ref{f6} shows the spatial distributions of the two SGB
537: components.
538: We select the two SGB subsamples as shown in the
539: bottom-right panel, and plot the spatial distributions in the
540: bottom-left panel. Stars from both groups have similar spatial
541: distributions. This is also confirmed by the
542: cumulative radial distributions shown in the top panel. The
543: Kolmogorov-Smirnov statistic shows that in random samplings from the
544: same distribution a difference this large would occur 11\% of the
545: time, which is very reasonable for the hypothesis that the two SGBs
546: have the same distribution.
547:
548:
549: %__________________________________________________________________
550: \begin{figure}
551: \epsscale{.80}
552: \plotone{f6.ps}
553: \caption{{\it Lower right}: Selection of the bSGB stars (red dots)
554: and fSGB stars (blue dots). {\it Lower Left}: Spatial distribution
555: of the bSGB and fSGB stars. {\it Top}: Cumulative radial
556: distributions of the bSGB and fSGB stars.}
557: \label{f6}
558: \end{figure}
559: %__________________________________________________________________
560: %
561:
562:
563:
564: %%%
565: %
566: \subsection{The horizontal branch stars}
567: %
568: %%%
569:
570: %
571: % rHB 242 (63%) younger
572: % bHB 143 (37%) older
573: %
574: Since the WFC data set included short exposures, we can also examine
575: the horizontal branch (HB) population. Figure~\ref{f7} shows the
576: upper part of the F606W vs.\ F606W $-$ F814W CMD. The NGC 1851 HB is
577: clearly bimodal. Interestingly enough, the red HB contains $63\pm7$\%
578: of the total red + blue HB stars (242 stars), and the blue part
579: contains $37\pm9$\% of the red + blue HB stars (143 stars). Note that
580: some stars populate the RR Lyrae gap. It is difficult to extract RR
581: Lyrae variables from the small number of exposures that went into
582: Fig.~\ref{f7}, with each star caught at random phases in each color
583: band. In any case, the contribution of the RR Lyraes to the total
584: population of the HB stars is $<$10\% (see Saviane et al.\ 1998 and
585: Walker 1992). As in the case of the bSGB and fSGB, stars from the red
586: and blue HB groups have similar spatial distributions (see top right
587: panel of Fig.~\ref{f7}). This is also confirmed by the cumulative
588: radial distributions shown in the bottom panel. The Kolmogorov-Smirnov
589: statistic shows that in random samplings from the same distribution a
590: difference this large would occur 17\% of the time, so here too there
591: is no significant indication of a difference in their radial
592: distribution.
593:
594: Neither for the SGB stars nor for the HB stars is the distribution of
595: the two samples significantly different. It would obviously be
596: desirable to strengthen the statistics by combining the two sets of
597: counts, but unfortunately this is not possible, because
598: the incompletenesses in star counts at the HB and at the SGB level are
599: different.
600:
601: %__________________________________________________________________
602: \begin{figure}
603: \epsscale{.80}
604: \plotone{f7.ps}
605: \caption{As in Fig.~\ref{f6}, but for the HB.}
606: \label{f7}
607: \end{figure}
608: %__________________________________________________________________
609: %
610:
611: On the basis of the relative numbers of red to blue HB stars and of
612: bSGB to fSGB populations,
613: it is very tempting to connect the bSGB (which includes 55\% of the
614: SGB stars) with the red HB, and the fSGB (45\% of the SGB stars) with
615: the blue HB stars.
616:
617:
618: \section{The ages of the two populations}
619:
620: The dispersion in the MS in F336W $-$ F814W is less than 0.04 mag.
621: Such a narrow main sequence implies that either the two SGB
622: populations have the same metallicities and the same helium abundances
623: or some fortuitous combination of abundances that results in no MS
624: split. Comparison with the theoretical models of Dotter et al.\
625: (2007) provides an upper limit to a possible dispersion in [Fe/H] of
626: 0.1 dex or in helium abundance of $\Delta Y= 0.026$, if each of the
627: two is assumed to act alone. We are not suggesting that such spreads
628: exist; we merely state upper limits to what might be present.
629:
630: We note that if we just increase the He abundance, there will be no
631: appreciable effect on the SGB luminosity. Helium alone cannot explain
632: the observed SGB split. On the other hand, a population 0.2 dex more
633: metal rich would have an SGB that is fainter by $\sim0.12$ magnitude,
634: and therefore it would be able to explain the SGB split. However, such a
635: high difference in [Fe/H] is ruled out by the narrowness of the MS. We
636: have also looked at the color distribution of the red giant branch (RGB)
637: and found no evidence of a split, and found an RGB color dispersion
638: consistent
639: %
640: with the photometric error, confirming that any dispersion in [Fe/H]
641: alone must be much smaller than 0.2 dex.
642:
643: Note, however, that increasing [Fe/H] makes the MS redder, while
644: increasing helium makes the MS bluer. Thus if, as a mere working
645: hypothesis, one increases [Fe/H] by 0.2 dex (enough to explain the SGB
646: split), while changing $Y$ from 0.247 to 0.30, then the two effects
647: cancel each other out, and the MS would have the same color. The same
648: would happen on the RGB:\ an [Fe/H] content 0.2 dex higher would move
649: the RGB to the red, but the RGB would be moved to the original color in
650: the presence of a He content $Y \sim 0.30$. In this way we would be able
651: to reconcile the SGB split with the low dispersions in the MS and RGB.
652: (Clearly, this kind of conspiracy cannot be ruled out till we have
653: high-resolution spectra to measure the metal content of the NGC 1851
654: stars.) However, there is additional observational evidence which seems
655: to rule out such an unlikely coincidence. A change in [Fe/H] of +0.2 dex
656: would imply an HB 0.03--0.04 magnitude fainter, while the suggested $Y$
657: enhancement would end with a bluer HB that is $\sim0.2$ magnitude
658: brighter. The resulting sloping of the horizontal branch is not
659: observed in Fig.~\ref{f7}.
660:
661: The only remaining possibility of explaining the SGB split is therefore
662: to assume
663: %
664: that in NGC 1851 there are two stellar populations with similar
665: metallicity and He content, but with different ages.
666: Figure~\ref{f5} shows that the two SGBs are separated vertically by
667: $\sim0.1$ magnitude in F606W. Comparison with the theoretical models of
668: Dotter et al.\ (2007) shows that this separation corresponds to an age
669: difference of $\sim 1$ Gyr, with the fSGB
670: older than the bSGB.
671:
672:
673: %__________________________________________________________________
674: %
675: \section{Discussion}
676:
677: In this paper we have shown that the SGB of the Galactic GC NGC 1851 is
678: split into two distinct branches.
679: If the split is interpreted in terms of age, its width implies that the
680: two populations were formed at epochs separated by $\sim1$ Gyr.
681: %
682: As discussed in the previous section, the age difference seems a
683: possible explanation for the peculiar SGB.
684:
685: The SGB split is not the only observational evidence of the presence of
686: more than one population in NGC 1851, though it is probably the clearest
687: one. We have already noted that NGC 1851 has a bimodal HB, and that
688: there might be a connection between the two SGBs and the two HB
689: sections. The relative frequency of stars in the different branches
690: implies that the progeny of the bSGB are in the red part of the HB, while
691: the progeny of the fSGB are on the blue side of the instability strip.
692: This is what one would expect qualitatively from stellar evolution
693: models, though an age difference of only 1 Gyr is not enough to move
694: stars from the red to the blue side of the HB. Our models indicate
695: that to move stars from the red to the blue side of the RR Lyrae gap would
696: require an age difference of $\sim$ 2--3 Gyr.
697: %
698: Some additional parameter, or combination of parameters, must be at work
699: in order to explain the morphology of the HB of NGC 1851.
700:
701: There is another relevant observational fact that must be mentioned.
702: Hesser et al.\ (1982) found that three out of eight of their bright RGB stars
703: show ``extraordinarily strong'' CN bands. These stars also show
704: enhanced Sr II and Ba II lines, and lie systematically on the red side
705: of the RGB. In other words, $\sim$40\% of their sample of bright RGB
706: stars contains CNO-processed material. This material could come from
707: the interior of these stars through mixing processes, but after what
708: we have learned from $\omega$~Cen and NGC 2808 we cannot exclude
709: the possibility that this processed material comes from a
710: first generation of stars that polluted the gas from which the
711: CN-strong stars have formed. This hypothesis would be further
712: supported if Sr and Ba are also confirmed to be enhanced (as the
713: spectra of Hesser et al.\ [1982] seem to imply), as these elements
714: cannot be produced in the low-mass stars presently on the RGB of NGC
715: 1851.
716:
717: Unfortunately, the results by Hesser et al.\ (1982) are based on a
718: very limited sample of stars; a more extended spectroscopic
719: investigation is clearly needed. Meanwhile, it is rather instructive
720: to look at the CMD published by Grundahl et al.\ (1999), based on
721: Str\"omgren photometry. Among the 15 CMDs in their Fig.\ 1, NGC 1851
722: shows by far the broadest RGB, with some hint of a bimodality. This
723: bimodality is not visible in our narrow-color-baseline F606W vs.\
724: F606W $-$ 814W CMD, but the capability of Str\"omgren photometry to
725: distinguish stellar populations with different metal content (in
726: particular CN content) is well known, and the CMD of Grundahl et al.\
727: (1999) tends to confirm the results of Hesser et al.\ (1982). We have
728: investigated the possibility that an enhancement of C or N, or of both
729: elements, can be the cause of the observed SGB split. Hesser et al.\
730: (1982) found that model spectra with [C/A] = +0.2 and [N/A] = +0.5
731: (where they define [A/H] as the logarithmic relative abundance of all
732: heavy elements in the theoretical models), or else model spectra with
733: with [C/A] = 0 and [N/A] = +1.0 could fit the observed spectra equally
734: well. A detailed analysis of the effect of C and N overabundances is
735: beyond the scope of the present paper. Furthermore, before such an
736: analysis is justified we will need better abundance constraints, which
737: would require fitting a larger sample of stellar spectra with modern
738: atmospheric models.
739: %
740:
741: Here we note that one possible consequence for a second generation of
742: stars is indeed an increased CNO abundance from mass lost after the
743: third dredge-up in intermediate-mass AGB stars from an earlier
744: generation (see discussion in Ventura \& D'Antona 2005 and Karakas et
745: al.\ 2006). The spectroscopic results of Hesser et al.\ (1982)
746: suggest enhanced levels of C and N but do not include an analysis of
747: the oxygen abundance, due to a lack of measurable features in the
748: spectra. (This further underlines the need for a new study.) It is
749: however possible that the C + N + O amount is enhanced to some extent,
750: though at the moment we cannot quantify this enhancement. The CNO
751: enhancement is an important observational input, as the level of
752: enhancement would allow us to better identify the possible polluters
753: from the first generation of stars.
754:
755: The presence of CNO enhancement would affect the age difference between
756: the two populations. A CNO enhancement of +0.3 dex would
757: reduce the age difference between the two SGBs to 300--500 Myr.
758: Under this hypothesis, we estimate a gap in the SGB of $\sim0.07$ magnitude
759: due to the composition alone, and an additional 0.03--0.05 mag
760: displacement due to the age difference, fully accounting for the observed
761: magnitude difference between the two SGBs.
762:
763: Larger CNO enhancements would further reduce the age gap between the
764: two generations of stars, down to a few hundred Myr, as expected in
765: the intermediate-mass-AGB-star pollution scenario (Ventura et al.\
766: 2002). We note that if the CNO enhancement in second-generation stars
767: of NGC 1851 is confirmed, NGC 1851 might differ from other globular
768: clusters, in which the sum of CNO elements for normal and
769: self-enriched stars seems to be constant (Cohen \& Melendez 2005).
770: New spectroscopy of stars in NGC 1851, along with the photometry
771: presented in this paper and detailed modeling, can significantly
772: improve our understanding of this intriguing situation.
773:
774: NGC 1851 is the third globular cluster for which we have direct evidence
775: of multiple stellar generations. All three clusters seem to differ in
776: several important respects, however:
777:
778: \begin{itemize}
779:
780: \item
781: In $\omega$~ Cen, the multiple populations manifest themselves both in
782: a main-sequence split (interpreted as a bimodal He abundance; see
783: Bedin et al.\ 2004, Norris 2004, and Piotto et al.\ 2005) and in a
784: multiplicity of SGBs (interpreted in terms of large age variations,
785: $\gg 1$ Gyr; see Villanova et al.\ 2007, and references therein),
786: which implies at least four different stellar groups within the same
787: cluster.
788:
789: \item
790: In NGC 2808, the multiple generation of stars is inferred from the
791: presence of three MSs (interpreted in this case too in terms of three
792: groups of stars characterized by different helium contents; see
793: discussion in Piotto et al.\ 2007), and further confirmed by the
794: presence of three groups of stars with different oxygen abundances.
795: It is also consistent with the presence of a multiple HB
796: (D'Antona \& Caloi 2004). However, in NGC 2808 the TO-SGB
797: regions are so narrow that any difference in age between the three
798: stellar groups must be significantly smaller than 1 Gyr.
799:
800: \item
801: In the case of NGC 1851, we have evidence of two stellar groups from
802: the SGB split, which apparently implies two stellar generations,
803: formed with a time separation of $\sim 1$ Gyr. This hypothesis is
804: further confirmed by the presence of a group of RGB stars with strong
805: CN bands (distinct from the majority of CN-normal RGB stars), and
806: enhanced Sr and Ba, and possibly the presence of a bimodal HB. In the
807: case of NGC 1851 there is no evidence of MS splitting, which implies
808: that any difference in helium or other heavier element content between
809: the two populations should be small.
810:
811: \end{itemize}
812:
813: Apparently, not only are GCs not single-stellar-population objects,
814: containing stars formed in a single star burst, but the evidence
815: emerging from the new exciting \hst\ observations presented in this
816: paper --- as well as from the \hst\ observations of $\omega$ Cen and NGC
817: 2808 --- is that the star-formation history of a globular cluster
818: can vary strongly from cluster to cluster. We are still far from
819: understanding what governs the different star-formation histories, and
820: this is an exciting and challenging question for future investigations.
821: At the moment, we can only note that the three clusters in which multiple
822: generations of stars have been clearly identified ($\omega$ Cen, NGC
823: 2808, and NGC 1851), and the two other clusters strongly suspected to
824: contain more than one stellar generation (NGC 6388 and NGC 6441; see
825: Caloi \& D'Antona 2007, and Busso et al.\ 2007) are among the ten most
826: massive clusters in our Galaxy. This evidence suggests that cluster mass
827: might have a relevant role in the star-formation history of GCs.
828:
829: We should note finally that the cases of $\omega$ Centauri and M54,
830: both probably associated with mergers of other galaxies into the Milky
831: Way, suggest the possibility that all globular clusters that have
832: complexities in their CMDs are likewise merger remnants. This is an
833: attractive speculation, but we think that it does not yet have enough
834: observational support.
835:
836:
837:
838:
839:
840:
841: \begin{acknowledgements}
842: The authors wish to thank the referee, Francesca D'Antona, for useful
843: discussions. The USA authors acknowledge the support for Program
844: number GO-10775 provided by NASA through a grant from the Space
845: Telescope Science Institute, which is operated by the Association of
846: Universities for Research in Astronomy, Incorporated, under NASA
847: contract NAS5-26555.
848:
849:
850:
851: \end{acknowledgements}
852:
853: %%%
854: \begin{thebibliography}{}
855: %%%
856: \bibitem[Anderson(2000)]{A00} Anderson, J., \& King, I.\ R. 2000, \pasp,
857: 112, 1360
858: \bibitem[Anderson(2006)]{A06} Anderson, J., \& King, I.\ R. 2006, ACS
859: Instrument Science Report 2006-01 (Baltimore:\ STScI)
860: \bibitem[Bedin(2004)]{B04} Bedin, L. R., Piotto, G., Anderson, J.,
861: Cassisi, S., King, I. R., Momany, Y., \& Carraro, G. 2004, ApJ, 605, L125
862: \bibitem[Bedin(2005)]{B05} Bedin, L. R., Cassisi, S., Castelli, F.,
863: Piotto, G., Anderson, J., Salaris, M., \& Pietrinferni, A. 2005, MNRAS,
864: 357, 1038
865: \bibitem[Bedin(2006)]{B06} Bedin, L.\ R., Piotto, G., Carraro, G., King,
866: I.\ R., \& Anderson, J. 2006, A\&A, 460, L27
867: \bibitem[Bekki(2006)]{bk06} Bekki, K., \& Norris, J.\ E. 2006, ApJ, 637, L109
868: \bibitem[Busso(2007)]{BU07} Busso et al. (2007), A\&A, in press
869: \bibitem[Caloi(2007)]{C07} Caloi, V., \& D'Antona, F. 2007, A\&A, 463,
870: 949
871: \bibitem[Cohen(2005)]{Co05} Cohen, J.\ G., \& Mel\'endez, J. 2005,
872: AJ, 129, 303
873: \bibitem[Cotrell(1981)]{C81} Cotrell, P.\ L., \& Da Costa, G.\ S. 1981,
874: ApJ, 245, L79
875: \bibitem[Dantona(2004)]{d04} D'Antona, F., \& Caloi, V. 2004, ApJ,
876: 611, 871
877: \bibitem[Dotter(2007)]{D07} Dotter, A., Chaboyer, B., Jevremovic, D.,
878: Baron, E., Ferguson, J.\ W., Sarajedini, A., \& Anderson, J. 2007, AJ,
879: 134,376
880: \bibitem[Fenner(2004)]{F04} Fenner, Y., Campbell, S., Karakas, A.\ I.,
881: Lattanzio, J.\ C., \& Gibson, B.\ K. 2004, MNRAS, 353, 789
882: \bibitem[Gratton(2005)]{G05} Gratton, R., Sneden, C., \& Carretta, E. 2004,
883: ARA\&A, 42, 385
884: \bibitem[Grundahl(1999)]{G99} Grundahl, F., Catelan, M., Landsman, W.\
885: B.,
886: Stetson, P.\ B., \& Andersen, M.\ I. 1999, ApJ, 524, 242
887: \bibitem[Harris(1996)]{H96} Harris, W. E. 1996, AJ, 112, 1487 (Feb.\
888: 2003 revision)
889: %
890: \bibitem[Hesser(1982)]{H82} Hesser, J.~E., Bell, R.\ A., Harris, G.\ L.\
891: H., \& Cannon, R.\ D. 1982, AJ, 87, 1470
892: \bibitem[Holtzman(1995)]{Ho95} Holtzman, J.\ A., Burrows, C.\ J.,
893: Casertano, S., Hester, J.\ J., Trauger, J.\ T., Watson, A.\ M., \&
894: Worthey, G. 1995, \pasp, 107, 1065
895: \bibitem[Karakas(2006)]{K06} Karakas, A.\ I., Fenner, Y., Sills, A.,
896: Campbell, S.\ W., \& Lattanzio, J.\ C. 2006, ApJ, 652, 1240
897: \bibitem[Layden(2000)]{L00} Layden, A.\ C., \& Sarajedini, A. 2000,
898: ApJ, 119, 1760
899: \bibitem[Maeder(2006)]{M06} Maeder, A., \& Meynet, G. 2006, A\&A, 448, L37
900: \bibitem[Norris(1996)]{N96} Norris, J.\ E., Freeman, K.\ C., \&
901: Mighell, K.\ J. 1996, ApJ, 462, 241
902: \bibitem[Norris(2004)]{N04} Norris, J.\ E. 2004, ApJ, 612, 25
903: \bibitem[Piotto(2005)]{P05} Piotto, G., et al. 2005, ApJ, 621, 777 (P05)
904: \bibitem[Piotto(2007)]{P07} Piotto, G., et al. 2007, ApJ, 661, L53 (P07)
905: \bibitem[Sarajedini(2007)]{S07} Sarajedini, A., et al. 2007, AJ, 133, 290
906: \bibitem[Saviane(1998)]{S98} Saviane, I., Piotto, G., Fagotto, F.,
907: Zaggia, S., Capaccioli, M., \& Aparicio, A. 1998, A\&A, 333, 479
908: \bibitem[Sirianni(2005)]{Si05} Sirianni, M., et al. 2005, \pasp.
909: 117, 1049
910: \bibitem[Sollima(2005)]{S05} Sollima, A., Pancino, E., Ferraro, F.\
911: R., Bellazzini, M., Straniero, O., \& Pasquini, L. 2005, ApJ,
912: 634, 332
913: \bibitem[Ventura(2001)]{v01} Ventura, P., D'Antona, F., Mazzitelli, I.,
914: \& Gratton, R. 2001, ApJ, 550, L65
915: \bibitem[Ventura(2002)]{v02} Ventura, P., D'Antona, F., \& Mazzitelli, I. 2002,
916: A\&A, 393, 215
917: \bibitem[Ventura(2005)]{vda05} Ventura. P., \& D'Antona, F. 2005, ApJ, 635, L149
918: \bibitem[Villanova(2007)]{V07} Villanova, S., et al. 2007, ApJ,
919: in press, astro-ph 0703208 (V07)
920: \bibitem[Walker(1992)]{W92} Walker, A. \ R. 1992, \pasp,
921: 104, 1063
922: %%%
923: \end{thebibliography}
924: %__________________________________________________________________
925: %%%%
926:
927:
928: \end{document}
929:
930: