Publications of the Astronomical Society of the Pacific 106: 376-381, 1994 April Spectroscopy of Long-Period Variables in 3 Alessandra Giovagnoli Dipartimento di Astronomia, via Zamboni 33, 40100 Bologna, Italy Electronic mail: alessandrag@alma02.bo.it Jeremy Mould 1 Palomar Observatory, Caltech 105-24, Pasadena, California 91125 Electronic mail: jrm@deimos.caltech.edu Received 1993 September 7; accepted 1994 January 4 ABSTRACT. Spectrophotometry of 36 long-period variable stars in 3 is presented. Most of them show strong TiO bands. These spectra together with published photometry indicate that the red variables exceeding 5X10 4 Lq are M supergiants, evolved from young massive stars which are prominent in this galaxy. A small number of lower-luminosity stars are present in this sample. These are asymptotic-giant-branch stars. We identify two of them as carbon stars. There are no strong S stars. 1. INTRODUCTION Much remains to be learned about the late stages of stellar evolution. Although the core hydrogen-burning phase is the clock which regulates the evolution of a stellar population (e.g., Renzini and Buzzoni 1986), later evolutionary phases in which stars become pulsationally unstable offer invaluable information on stellar structure, much of which remains to be properly assimilated. Many of the important events in the life of a star, like supernova explosion and ejection of a planetary nebula, occur during the red giant or supergiant phase. Since the long-period variables (LPVs) are in the phase that immediately precedes the above events, their study is very important for the comprehension of those events. Significant progress has been made in recent years in obtaining a more consistent understanding of Cepheid variable stars (Chiosi et al. 1992). The study of LPVs remains in its infancy, however. A classification scheme for red variable stars is well established, but physical understanding lags behind. Study of core helium-burning supergiants and asymptotic-giant-branch (AGB) stars in the Magellanic Clouds (Wood et al. 1983; Reid and Mould 1985; Mould and Reid 1987) has shed some light on these matters. From these works it is clear that LPVs are useful indicators of star-formation history, of carbon star formation, of dredge-up events, and of s-process element production. Investigation of the LPVs in other Local Group galaxies is essential to extend these efforts. Kinman et al. ( 1987) and Mould et al. (1990) started the survey of 3, a spiral galaxy like the Milky Way. We present here the results of a spectroscopic study. Current address: Mount Stromlo & Siding Spring Observatories, Australian National University, Private Bag, Weston Creek P.O., A.C.T. 2611, Australia. 2. OBSERVATIONS AND CALIBRATION Our LPV sample is drawn from Table II of Kinman et al. (1987), who provide accurate positions and finding charts for these stars. Spectra were obtained with the 4SHOOTER spectrograph (Gunn et al. 1987) at the Hale telescope of Palomar Observatory over the period 1990 September 10-12. We employed the 300 line mm -1 grism to obtain 15 Â resolution spectra with typical exposure times of 10-20 min. The seeing was approximately 1 arcsec during the run and conditions were photometric. All these spectra were wavelength and flux calibrated using the package FIGARO on the Caltech VAX cluster. We subtracted the sky from the spectra and calibrated in wavelength using He and Ne arc lines. Flux calibration was relative to observations of the standard star HD 19445 (Oke and Gunn 1983). As part of this reduction, the data were rebinned between 4600-8200 Á. Examples of final reduced spectra are shown in Fig. 1, where the flux is in micro-janskys and the wavelengths are in Angstroms. 3. SPECTRAL CLASSIFICATION For spectral classification, we used a series of narrowband photometric indices shown in Table 1. A continuum is defined for each index, using two sidebands, and the flux in this continuum is compared with the observed flux in the central passband (Mould 1976). In Table 2 are presented the data for the individual 3 stars. The strength of the Na D line (A = 5896 Á) and the CaH molecular band (A = 6830 Â) are used as a dwarfgiant discriminator in K and M stars. The Na D index is plotted against the continuum slope between 7050 and 7530 À (30 Ä bandpasses) in Fig. 2(a). The slope is the flux ratio in these bandpasses expressed in magnitudes; it is in no sense a true continuum measure: both these bandpasses are blanketed by molecular bands. Most of the stars have Na D strength <0.25 mag with a few strong sodium line stars. Given that strong sodium and strong CaH [Fig. 376 1994. Astronomical Society of the Pacific
LONG-PERIOD VARIABLES IN 3 377 Q14906 5000 5500 6000 6500 7000 7500 8000 Wavelength Angstroms Q29631 Wavelength Angstroms Fig. 1 Four examples of spectra of LPVs in 3. Wavelength Angstroms 2(b)] does not occur together in these stars, and that we have selected periodic variable stars, it is not surprising that none of the sample seems to be a dwarf. As an indicator of carbon stars, we used C 2 A5635, and its strength is shown in Fig. 3. The rise in the C 2 index in M stars with decreasing temperature indicates that this index is TiO blanketed, and so only significant departures from the trend are carbon star candidates. In our sample in Table 1 Photometric Indices Index Sideband 1 Band Sideband 2 Species DA5635 DA5896 DA6180 DA6270 DA6475 DA6708 DA6830 DA7100 DA7812 5380-5400 5845-5865 6230-6255 6680-6697 7460-7540 5610-5630 5880-5905 6165-6200 6250-6275 6475-6510 6702-6712 6790-6870 7060-7140 7792-7832 5645-5665 5940-5960 6580-6620 6310-6330 6580-6620 6717-6731 7460-7540 7460-7540 8096-8136 C2 Na TiO l3 CN ZrO Li CaH TiO TiO fact only Q17068 and P23152 (Mould et al. 1990) show evidence of carbomoxygen > 1 on the surface. In Fig. 4 we show the strength of the ZrO at 6472 Â which can be used to identify S stars. At the present resolution, this is a difficult feature to recognize or to rule out. Strong S stars are seen on the upper AGB in the Magellanic Clouds (Wood et al. 1983) and weaker ones in the LMC by Lundgren ( 1988). Nevertheless, we can draw two conclusions from the present sample with confidence: strong S stars are not present in the sample, and we see no difference, on average, between the ZrO band strengths of stars brighter than the AGB tip in 3 and those fainter than the tip. The majority of the stars occupy the zone at lower values, and there is therefore generally no evidence of the s process in stars in our sample. The lithium index at 6707 A also yielded negative results (cf. Smith and Lambert 1989). TiO bands, >1 = 6180 Á and 7100 Â (Figs. 5 and 6, respectively), were used to classify the stars in terms of spectral type. From the Reid and Mould (1985) LMC sample, we derived two relations between the TiO band at >17100 and the spectral type quoted in their Table 1. Two
378 GIOVAGNOLI AND MOULD Star 7050- (J-K) Na TiO 7530 D(5896) D(6180) Q24798 Q21740 Q17068 Q18180 Q12254 Q6508 Q6512 P6836 Q29250 P23324 P32202 P28986 PI1172 P3482 Q29631 Q21765 Q21736 Q29211 P23182 Q31267 P24355 P16071 P29926 P29476 P16339 P21091 P32117 P19844 Q14906 P20917 P6283 P18694 P25017 Q31100 Q18569 P22330 1.175 1.894 1.053 2.449 2.079 2.237 1.264 2.427 1.030-0.516 1.326 1.101 1.976 1.227 1.303 1.211 1.266 1.565 1.609 0.254 1.889 1.676 1.520 1.910 1.272 2.104 2.213 2.145 1.388 1.261 1.062 1.520 1.870 1.356 2.789 0.9 1.14 1.57 1.08 1.2 1.01 0.9 1.10 0.88 0.89 1.10 1.01 1.39 0.95 1.17 1.14 1.09 1.31 1.09 0.95 1.16 1.15 1.04 1.13 1.13 1.18 1.63 1.05 1.04 1.12 0.88 1.26 1.08 1.05 1.22 1.39 0.171 0.190 0.320 0.187 0.126 0.275 0.084-0.152-0.130-0.015 0.120 0.028 0.245 0.058-0.078 0.427 0.182 0.098 0.113 0.408 0.163 0.570 0.133 0.155-0.315 0.981-0.249 0.099 0.169 0.296 0.031 0.084 0.152 0.435 0.322 0.474 0.185 0.800 0.483 0.506 0.312 0.880 0.437-0.052 0.362 0.306 0.444 0.297 0.184 0.249 0.367 0.613 0.398 0.685 0.654 0.477 0.559 0.577 0.364 0.607 0.155 0.652 0.555 0.303 0.399 0.520 0.416 0.365 1.235 Table 2 3 Sample ZrO C2 TiO TiO Sp D(6475) D(5635) D(7100) D(7812) 0.002 0.037 0.060 0.168-0.055-0.076 0.051 0.139 0.051 0.016 0.066 0.058-0.038 0.048 0.064-0.138 0.034 0.045 0.050-0.010-0.010 0.000-0.025-0.036-0.010-0.182-0.176-0.047 0.000 0.070 0.062-0.075 0.031 0.055 0.053 0.050 0.007 0.513 0.586 0.328 0.070 0.060-0.326-0.246-0.015-0.047 0.055-0.244 0.045-0.005 0.168-0.006-0.014 0.126-0.059 0.229 0.106 0.095 0.167 0.047 0.579 0.419 0.575 0.217-0.191-0.161 0.408-0.300 0.288 0.333 0.576 0.441 0.926 0.592 0.756 0.315 0.669 0.355-0.030 0.383 0.323 0.586 0.314 0.132 0.478 0.366 0.522 0.400 0.500 0.823 0.545 0.670 0.872 0.483 0.791 0.850 0.856 0.522 0.410 0.326 0.648 0.496 0.348 0.938 0.023 0.153-0.397 0.361 0.255 0.182-0.152 0.506-0.163 0.216 0.043-0.081 0.309-0.043-0.148 0.327 0.027 0.108 0.190 0.390 0.299 0.290 0.315 0.151 0.411 0.423 0.435 0.171-0.083 0.317 0.045 0.085 0.479 C A M0? Remarks trace ZrO strong CaH Mis-id? poor sp strong CaH relations are necessary because for the late-type M stars, the band strength dispersion of the sample becomes large. The equation y= 1.3687 + 8.188* is used for TiO band strengths <0.6 and y=3.095 +1.638* for greater values. In these relations, y represents the M spectral subtype (0-9) and x the band strength. In addition, we considered another TiO band at A = 7810 (Fig. 7) as an indicator at later subtypes (Wing 1967). The quoted spectral types are shown in Table 2. 4. DISCUSSION We have derived the absolute magnitude distribution over spectral type for the stars of our sample. In Figs. 8 and 9, our distribution and the Magellanic Cloud one, derived from Wood et al. (1983), are shown. The magnitude is calculated assuming m Af=24.4, and m bol is from Mould et al. (1990). In 3, the carbon stars lie at M ho i > 6 and the M stars are at M hox < 6. In general these characteristics agree with the behavior of the Magellanic Clouds: the presence of M supergiants shows that the prominent population in 3 is of massive stars. The two distributions are not necessarily different because the 3 sample is very incomplete due to the greater distance and the magnitude cut-off of the Kinman et al. (1987) survey at r~20 mag. Stars brighter than M hol = 7.1 mag must be supergiants, because for larger luminosities the core mass of an AGB star exceeds the Chandrasekhar mass (Wood et al. 1983). This picture has been modified slightly by recent calculations in which envelope burning is important (Sackman and Boothroyd 1992). Overluminous AGB stars are short lived, however. In Fig. 8 we see only M stars in the brightest two bins. This is consistent with the expectation that they are supergiants. As is clear from Fig. 9, in the Magellanic Clouds the luminosity range ( 4, 6) is dominated by C(40%) and S(15%) stars. In 3 we see just two carbon stars. It is not surprising that, in our small
LONG-PERIOD VARIABLES IN 3 379 Q cd 55 CD CT) CO id Q Fig. 3 C 2 band strength vs. continuum flux ratio. Second, the TiO band strength (on which spectral types are based) will increase with increasing metallicity. This effect has been demonstrated in globular clusters by Mould and McElroy (1978). We can obtain an indication of the magnitude of this effect from Fig. 10 of Frogel and Whitford (1987). In their comparison of field giant stars and Baade s window giants, they show that the {J K) values of to M5 stars are averagely 0.1 mag bluer at constant spectral type in the more metal rich bulge of the Galaxy. This effect, which is the same as we see in the comparison of {J K, type) for 3 and the LMC, must be comprised Fig. 2 (a) Na D line strength vs. continuum flux ratio, (b) CaH line strength vs. continuum flux ratio. sample, we have seen no S stars. A deeper LPV survey of 3 is required. In Figs. 10 and 11, the plots of spectral type against ( J K) for 3 and LMC are shown. At a given spectral type the LMC stars tend to be redder than the 3 ones, in fact there are no stars with {J K) < 1.04 in the LMC sample whereas the 3 stars are as blue as {J K) =0.88. Three factors contribute to this behavior. First, the metal abundance in Population I in 3 is, on average, a factor of two higher than in the LMC, judging by the oxygen abundance (Pagel and Edmunds 1981 ). The Hayashi track on which red supergiants reside in 3 will therefore lie at cooler temperatures than that of the LMC. The corresponding increase in {J K) with metallicity from this effect is sso.07 mag, judging by the giant branches of globular clusters (Frogel et al. 1983). (7050-7530) Fig. 4 ZrO band strength vs. continuum flux ratio. Stars fainter than the AGB tip are plotted with open symbols.
380 GIOVAGNOLI AND MOULD Fig. 5 TiO band strength vs. continuum flux ratio. of an increase in ( J K) of 0.07 mag and an increase in spectral type of approximately three subtypes. The third factor influencing this comparison is reddening. The foreground reddening for the LMC and 3 is similar and only 0.03 in (J K). Differences in circumstellar reddening can also in principle contribute to the comparison of 3 and LMC colors. The sequence of bluer colors at given spectral type from the LMC, through 3, to the Galaxy argues that the metallicity effect on band strength is the dominant effect here. This explanation is plausible provided that the reddening of LPVs is no larger in 3 than in the LMC. That is a different conclusion from that of Humphreys et al. (1984) from JHK photometry of non variable M su- Fig. 7 TiO band strength vs. continuum flux ratio. pergiants, but is consistent with JHK photometry of these LPVs (Mould et al. 1990). 5. CONCLUSIONS The majority of the stars in the 3 LPV sample are M supergiant stars. Subtype classification shows that there are only a few stars of really late type. Judging by their J K colors, the effective temperatures of these supergiants at a given subtype are closer to those of the Galaxy (warmer) than those of the Magellanic Clouds. This can be understood as a metallicity effect. A few stars at lower luminosity are asymptotic-giant-branch stars. Spectral classification of these AGB stars show that there are two carbon stars in the sample. There are no strong S stars in the o P o' 00 co Q Mbol Fig. 8 The luminosity function of 3 LPVs from Table 2. Spectral classes are indicated.
LONG-PERIOD VARIABLES IN 3 381 Mbol Fig. 9 The luminosity function of LMC LPVs from Wood et al. ( 1983). Spectral classes are indicated. sample, either at the AGB tip or at fainter magnitudes. In the present limited sample there is no significant difference between the luminosity distribution of M, S, and C stars of 3 LPVs and that of Magellanic Cloud LPVs. AG would like to thank the SURF program at Caltech for a fellowship and the Osservatorio Astronómico di Bologna for travel assistance. JRM would like to thank the (J-K) Fig. 10 J K color as a function of spectral type for 3 LPVs. The line indicates the relation for Galactic supergiants (Johnson 1966) after transformation to the CIT system. Fig. 11 Same as Fig. 10 for LMC supergiants. staff of Palomar Observatory for their help. We are grateful to Shaun Hughes for help with flux calibration of these data and for reading and commenting on the manuscript. REFERENCES Chiosi, C, Bertelli, G., Bressan, A., Wood, P., and Mateo, M. 1992, ApJ, 387, 320 Frogel, J. A., and Whitford, A. E. 1987, ApJ, 320, 199 Frogel, J. A., Cohen, J. G., and Persson, S. E. 1983, ApJ, 275, 773 Gunn, J., Carr, M., Danielson, G. E., Lorenz, E., Lucinio, R., Nenow, V., Smith, J., Westphal, J., Schneider, D., and Zimmerman, B. 1987, Opt. Eng., 26, 779 Humphreys, R., Jones, T., and Sitko, M. 1984, AJ, 89, 1155 Johnson, H. 1966, ARAA, 4, 193 Kinman, T. D., Mould, J., and Wood, P. R. 1987, AJ, 93, 833 Lundgren, K. 1988, A&A, 200, 85 Mould, J. 1976, ApJ, 207, 535 Mould, J., and McElroy, D. 1978, ApJ, 223, 824 Mould, J., and Reid, N., 1987, ApJ, 321, 156 Mould, J., Graham, J. R., Matthews, K., Neugebauer, G., and Elias, J. 1990, ApJ, 349, 503 Oke, J. B., and Gunn, J. E. 1983, ApJ, 266, 713 Pagel, B., and Edmunds, M. 1981, ARAA, 19, 77 Reid, N., and Mould, J. 1985, ApJ, 299, 236 Renzini, A., and Buzzoni, A. 1986, in Spectral Evolution of Galaxies, ed. C. Chiosi and A. Renzini (Dordrecht, Kluwer), p. 195 Sackman, I. J., and Boothroyd, A. 1992, ApJ, 392, L71 Smith, V., and Lambert, D. 1989, ApJ, 345, L75 Wing, R. F. 1967, Ph.D. thesis, Univ. of California, Berkeley Wood, P. R., Bessell, M. S., and Fox, M. W. 1983, ApJ, 272, 99