Review of Working Group Results and Plenary Session Discussion

This section is an editorial review of the results in Tables 1 through 7, as well as the discussion of those results during the plenary session (the transcript of the plenary session debate is in the following section of these proceedings).

1. From twenty parameters to four

Dick White, the plenary session moderator, opened the discussion by pointing out a principal component analysis (PCA) recently carried out by Andrea Dobson and Bob Donahue (Dobson, A. K., & Donahue, R., 1998, How Many Parameters Are Needed to Describe a Star?: Principal Component Analysis as an Astrophysical Tool, to appear in The Tenth Cambridge Workshop on Cool Stars, Stellar Systems, and the Sun, R. A. Donahue and J. A. Bookbinder, eds.), in which the authors find that four parameters, as opposed to the canonical three, are required to decribe a star.

In preparing the plenary session presentations, White distilled the information in Table 1 into the following.

Table 1A. Data of Table 1 as shown in plenary session
Parameter Rankings Parameter Rankings
T_{eff} 3 ± 2, 4 ± 1 * V magnitude 0 ± 0, 4 ± 1
Cousins photometry 4 ± 1, 2 ± 1 * v \sin i 2 ± 0, 0 ± 0
B-V 2 ± 1, 4 ± 0 * Inclination 0 ± 0, 0 ± 0
Strömgren photometry 5 ± 0, 3 ± 2 * Microturbulence 1 ± 1, 0 ± 0
R'HK 4 ± 1, 3 ± 2 * B field 1 ± 0, 0 ± 0
[Fe/H] 4 ± 0, 4 ± 0 * Energy distribution 3 ± 2, 2 ± 0
MK type 5 ± 0, 5 ± 0 * Photometric variability 3 ± 2, 1 ± 0
Johnson other than B-V 3 ± 0, 3 ± 1 * Spectroscopic var. 1 ± 0, 0 ± 0
Mbol 5 ± 1, 5 ± 0 * Location 2 ± 1, 1 ± 0
log g 3 ± 2, 0 ± 0 ~~~ Space velocity 2 ± 1, 2 ± 0

Any parameters given a combined ranking >= 6 by the working groups received an asterisk, as shown above. The choice of 6 as a cutoff score was arbitrary, and the numbers themselves were produced by ten independent groups. The rankings were debated for some 20 minutes during the plenary session, but in the end, Radick pointed out that the asterisked quantities did in fact reduce to four parameters: five temperature or temperature equivalent, two luminosity or equivalent, one activity parameter, and composition. Thus, this part of the debate finished full-circle, with a confirmation, at least in this extemporaneous setting, of Dobson & Donahue's four-parameter result. Garrison pointed out that the redundancy available along the four axes of this parameter space was invaluable for consistency checks.

2. From four parameters to two

Only two parameters emerged in Table 1 with a ranking of "5" from both groups that discussed them, where "5" meant that parameter has been essential and highly beneficial in solar analog surveys over the past few decades. These were the star's MK spectral type, and its bolometric luminosity. Perhaps the MK result is unsurprising, since the Sun defines G2 V, but clearly there has been significant insistence that a posited solar twin closely resemble the Sun, in the spectroscopic sense defined by the MK system.

As a supplement to a near-solar MK type, the star's absolute energy output, Mbol, emerged as a prime requirement for selecting a solar twin. However, the absolute energy distribution was rated lower, even though it was cited as perhaps the most important observational quantity one could obtain. Concerns were raised about (1) systematic errors in the calibrations and (2) the lack of a large number of stars for which good energy distributions are available. White and Hall demurred, pointing out that very large compilations of stellar energy distributions, which agree well with published distributions for standard stars, have been made by the Russian groups and are largely unknown in the West. However, Lockwood also noted that published energy distributions for the Sun itself are in poor agreement, so we might expect some difficulty in identifying a solar twin in that sense.

Other parameters on the list were rated lower because of a recurring concern, launched by Buzasi during the panel discussion, that even if a quantity is of fundamental importance to understanding the character of a star, it is useless to us if it cannot be measured precisely or yields ambiguous results. Spirited debate revolved around [Fe/H], which has been widely employed in solar analog searches but for which different workers have obtained different results for the same objects.

3. On the mandate and pitfalls of huge databases

Several photometric parameters were given in the list of parameters, and this produced (1) a specific call for a shift in photometric systems and (2) a more general note of caution regarding the inertia a well-established and huge database of observations can develop.

Discussion of the various photometric parameters given in the list produced strong agreement regarding the use of Strömgren photometry in selecting solar analogs. The Strömgren system was generally agreed to be somewhat more desirable than the Cousins system, and far superior to Johnson. The venerable B-V was simultaneously lambasted as much too metallicity-sensitive, and praised for the vast number of stars for which it has been determined. But Skiff drew attention to the work of Erik Olsen, who has performed Strömgren photometry of tens of thousands of solar-like stars. Other systems were mentioned (e.g., Tycho and Geneva), but there was little support for the observing effort required to develop the databases on those systems to the level that has been achieved with Johnson and Strömgren.

Inevitably the question arose as to why Olsen's work, for example, was not more familiar to the attendees, and why, if B-V was good only as a rough guide for finding solar-like star, a shift to b-y as the standard had not taken place years ago. The circular answer, of course, was that we use the huge B-V database because that is the huge database we have had. Two serious reservations exist about following this path.

First, the scope and undeniable value of such a database often creates a mandate for its use, and can overly constrains investigators' perspectives. The more a database such as B-V grows, the harder it becomes to shift to another standard of measurement, even if that standard is a superior quantity. <0> Second, and perhaps more insidious, is what happens once a quantity has been used and measured by many different investigators, using different instruments, referenced to different standard stars. The data from these many references can easily be placed in a blender and mixed into a large database, but, as Garrison pointed out in detail during the panel discussion, the resulting potion may well be lethal. Extreme caution in the selection of standard stars and the use of a standardized database is essential; or, if various references are combined, scrupulous attention to the discussions of calibrations and errors that the authors must provide if their paper is to be considered useful.

4. Neglected parameters

Having dispensed with essential and useful parameters, the workshop turned to those that have been so far ignored; the results are summarized in Table 2.

The most popular parameter in Table 2 was companionship. Some confusion surfaced as to whether the working groups had meant companionship in the sense of binarity and the circumstellar environment, or the presence of planets. Both were accepted as interesting parameters to apply to solar twins. Guinan suggested that components of wide, non-interacting binaries (e.g., the alpha Cen system) could be included in solar analog lists.

A clear demand for the incorporation of space-borne satellite observations of solar-like stars emerged in discussions of Question 1B (cf. Table 2), asking which quantities have been neglected in the solar analog search. Selection of solar analogs has so far been mostly with optical and near-UV data, yet as Rottman discussed in his talk, the relative variability of the Sun becomes increasingly pronounced at shorter wavelengths, exceeding a factor of 10 at coronal wavelengths. A significant problem here is the lack of time-series UV and X-ray data for numerous stars, the large IUE database notwithstanding. Extension of the "other-than-optical" idea into the radio regime also appeared twice on the working groups' lists.

There was general acknowledgment that the neglect of the parameters above has been technology-limited. At the time Hardorp initiated the solar analog search, IUE had just been launched, and high-precision planetary detections (e.g., Marcy and Mayor) were still in the future. However, the consensus was that current and planned solar and stellar space-borne observatories could and should used to study the upper atmospheres of solar twin candidates in detail.

5. Four parameters again

Given that the current state of affairs is not entirely satisfactory, where should we go from here? This was the focus of Question 2 (results in Tables 3, 4, and 5), which asked working groups to identify parameters that must be used in any future solar twin searches, and to rank them if the group considered it useful or desirable.

The essence of Tables 3 through 5 is (1) we have been more or less on the right track, and (2) we are best served by a conservative approach to selecting our solar twins.

Table 3 shows the 1-to-5 rankings assigned by different groups (group numbers are given in parentheses). No parameters were supplied a priori; groups were free to choose whatever four they thought were most important. Examination of the first five entries in Table 3 yields the same result as that in Table 1: there is luminosity MV, abundance (metallicity), activity (assuming "activity" and "Prot" to be related), and temperature. Duplicate parameters appear in the rest of the entries. In agreement with the attendees' opinions in Table 1, Strömgren was the only photometric system mentioned in this part of the discussion. The general irreverence of the attendees in debating these weighty matters appear in the final entry of Table 3.

A significant number of the working groups (4 out of 10) explicitly said that a ranking of the desirable parameters was not possible, supporting the general feeling that (a) a solar twin must embody as full a description of the star as possible, which appears to require four parameters, and (b) different researchers will have different requirements of their solar twins (see, e.g., Wade's panel presentation, and the unanimous response to Question 2B).

Those groups that did rank the parameters took a conservative approach. Four out of seven chose "colors" or "MK Type" as most important parameter; "colors" were agreed in the plenary session discussion to mean well-defined, homogeneous databases of, e.g., Strömgren colors. Two of the remaining three groups ranked Strömgren colors second. A strong sentiment therefore emerges for whittling down the stellar sample with well-defined, well-behaved guides before proceeding to specific discriminants such as abundances, rotation rate, and effective temperature. The bolometric luminosity also reappears here as one of the parameters one must know.

6. And the winner is...

Fifteen frequently-cited solar analogs were given to the working groups for evaluation and ranking, to settle the issue (for at least the next six months) of which star is our current best solar twin. Table 7 speaks for itself; the following points stand out.

(1) The star 18 Sco (HD 146233 = HR 6060) is the clear winner, being placed first on every group's list.

(2) Four stars appeared consistently at the top of most lists. These top four have been separated by a line in Table 7. They are 18 Sco, 16 Cyg A & B (HD 186408 and 186427) and 51 Peg (HD 217014).

(3) A second, less well-defined group of four appears between rankings 5 and 8. These include HD 44594 (a former candidate for best solar twin), alpha Cen A, and HD 76151. A few interlopers appear in this group as well.

(4) Most of the remaining stars on the lst given to the working groups appear between rankings 9 and 15. These include the Hyades stars (HD 1835 and van Beuren 64) and others which are known to be young and highly active.

(5) Two stars, alpha Cen A and HD 10307, were not on the initial list, but were recommended by at least one group as targets deserving of further study.

Cayrel de Strobel suggested that the terms solar-like star, solar analog, and solar twin be applied to stars that increasingly closely resemble the Sun, with the last term only applied to those that are nearly indistinguishable from the Sun in terms of any parameter one observes. Table 7 likewise falls into three well-defined parts. Only the first four stars were considered reasonable solar twins by the workshop, with 18 Sco superior to the other three.

Below ranking 4, we move from the realm of solar twins to solar analogs. Workers using the results of this workshop can probably use stars in the lower part of Table 7 for applications where solar-analogous stars are required, but with a greater degree of caution to the literature regarding that star than might be needed for the twins. The star HD 44594, for example, is slightly evolved, while HD 76151 has larkedly greater chromospheric activity than does the Sun. For some applications, this might be irrelevant, but the purists at the workshop labeled these star as decidedly unsolar in at least one important way.

The stars below ranking 8 were placed there wth almost no dissension. The question asked working groups to rank "twins" rather than analogs, and there was no question that these stars cannot be considered twins. Many of the participants ranked these stars as solar analogs, though there are a few (the very active HD 1835, and kappa Cet (G5V), for example), that might well be demoted to the realm of solar-like stars.

7. Conclusion

The solar analog search began in 1978, when Johannes Hardorp published The Sun Among the Stars. I. A Search for Solar Spectral Analogs (A&A, 1978, 63, 383). A vitriolic debate concerning the veracity and pitfalls of Hardorp's method developed, and The Sun Among The Stars. III. (A&A, 1980, 91, 221) is a bitter tract in which the science is overshadowed by Hardorp's acerbic responses to his critics. In 1985, Bob Garrison wrote that Hardorp "set out, with good intentions, to fnd a solar analog...[but] he has muddied the waters considerably." (in IAU Symposium 111, The Calibration of Fundamental Stellar Quantities (Dordrecht: Reidel), ed D. S. Hayes et al., p. 25). In 1998, during the panel discussion at our workshop, Garrison told us that Hardorp visited him in Toronto six months before his death and said, "Bob, you were right."

These were the four most telling words of the workshop, for they remind us that we tread on very delicate ground. Without Hardorp's pioneering papers, and the tremendous effort he devoted to inventing this line of inquiry for the rest of us, our workshop probably would not have happened. But "You were right" must have been an unhappy conclusion to reach at the end of one's life.

Throughout the workshop, and these proceedings, we see calls for extreme caution in working through this very complex subject. We have the exhaustive effort de Mello and da Silva have expended on one single star, Garrison's comments regarding calibration of standards and indiscriminate culling of the literature, Soderblom's remarks about the curious state of some of the HIPPARCOS parallaxes, Farnham & Schleicher's dramatically divergent reflectance spectrum of Ganymede, and Burlov-Vasiljev's plots of the substantial discrepancies in different attempts to measure the energy distribution of the Sun itself. Several times during the workshop, we heard the forlorn comment that picking a solar twin was nearly impossible because the Sun doesn't even always look like the Sun.

During the discussion after her talk, Giusa Cayrel stated that "We are not done." The Solar Analogs Workshop certainly did not solve the problem, but we selected a current best solar twin, and we defined the problems we must solve more clearly. We hope the next two decades of solar analog research will benefit.