ASTRONOMY & ASTROPHYSICS APRIL I 1997, PAGE 51 SUPPLEMENT SERIES Astron. Astrophys. Suppl. Ser. 122, 51-77 (1997) A calibration of Geneva photometry for B to G stars in terms of T eff , log g and [M/H ] M. K¨ unzli 1 , P. North 1 , R.L. Kurucz 2 , and B. Nicolet 3 1 Institut d’Astronomie de l’Universit´ e de Lausanne, CH-1290 Chavannes-des-Bois, Switzerland 2 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, Massachusetts 02138, U.S.A. 3 Observatoire de Gen` eve, CH-1290 Sauverny, Switzerland Received April 16; accepted June 26, 1996 Abstract. We have used recent Kurucz models and nu- merous standard stars to improve the calibration of the Geneva photometric system proposed a few years ago. A new photometric diagram for the classification of interme- diate stars (8500 ≤ T eff ≤ 11000 K) is proposed and fills a gap that the previous calibration had left open. Evidence is given for a clear inadequacy of the new Kurucz models in the region of the parameter space where convection be- gins to take over radiation in the star’s atmosphere. This problem makes the determination of the surface gravity difficult, but leaves that of the other parameters appar- ently unaffected. The determination of metallicity is con- siderably improved, thanks to the homogeneous spectro- scopic data published recently by Edvardsson et al. (1993). Instead of showing the traditional diagrams, we chose to publish the diagrams of the physical parameters with the inverted grids inside, i.e. the lines of constant photometric parameters Key words: stars: general; atmospheres; fundamental parameters 1. Introduction The determination of the fundamental atmospheric pa- rameters of stars remains an everlasting preoccupation of most stellar astronomers. The imminent release of the Hipparcos results certainly makes this problem especially acute. On the other hand, new and hopefully more realistic atmosphere models have become available in the last few years (Kurucz 1991, 1993, 1994), as well as a large amount of precise, homogeneous [Fe/H] values (Edvardsson et al. 1993). We found it worthwhile, therefore, to revise the al- ready published calibrations (North & Nicolet 1990, here- after NN90; Kobi & North 1990, hereafter KN90) and Send offprint requests to: M. K¨ unzli to take this opportunity to fill the gap that existed for stars with effective temperatures between 8500 and about 10500 K. The efficiency of the Geneva system is similar to that of the widely-used uvbyβ one, so it would be a pity not to exploit it fully. The only drawback of the Geneva system, compared to the uvbyβ one, is its sensitivity to interstellar reddening for A and cooler stars. In the following, we shall present successively the three temperature domains which are treated separately (as in the uvbyβ system), discussing together the reference (or standard) stars and the theoretical grids. The possible ap- plications and limits of this calibration are discussed in the conclusion. 2. The B stars 2.1. Methods and results Following Cramer & Maeder (1979), we use here the reddening-free parameters X and Y which have, as these authors showed, the optimum efficiency for determining the effective temperature and the surface gravity respec- tively. Although the definition of the X and Y parameters is given in several papers (Cramer & Maeder 1979; NN90), we recall it here for convenience: X =0.3788 + 1.3764 U - 1.2162 B1 - 0.8498 B2 (1) -0.1554 V 1+0.8450 G Y = -0.8288 + 0.3235 U - 2.3228 B1+2.3363 B2 (2) +0.7495 V 1 - 1.0865 G where U , B1, B2, V 1 and G stand for the Geneva colour indices [U -B], [B1-B], [B2-B] etc. Let us recall that the Z parameter allows the separation of the Bp stars (mostly of the Si and SiCr types) from the normal B stars (Cramer & Maeder 1980); it will not be used here, however. The synthetic colours U , B1 etc. have been computed by one of us (BN) using recent Kurucz models with scaled
27
Embed
A calibration of Geneva photometry for B to G stars in ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ASTRONOMY & ASTROPHYSICS APRIL I 1997, PAGE 51
SUPPLEMENT SERIES
Astron. Astrophys. Suppl. Ser. 122, 51-77 (1997)
A calibration of Geneva photometry for B to G stars interms of Teff, log g and [M/H ]
M. Kunzli1, P. North1, R.L. Kurucz2, and B. Nicolet3
1 Institut d’Astronomie de l’Universite de Lausanne, CH-1290 Chavannes-des-Bois, Switzerland2 Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, Massachusetts 02138, U.S.A.3 Observatoire de Geneve, CH-1290 Sauverny, Switzerland
Received April 16; accepted June 26, 1996
Abstract. We have used recent Kurucz models and nu-merous standard stars to improve the calibration of theGeneva photometric system proposed a few years ago. Anew photometric diagram for the classification of interme-diate stars (8500 ≤ Teff ≤ 11000 K) is proposed and fills agap that the previous calibration had left open. Evidenceis given for a clear inadequacy of the new Kurucz modelsin the region of the parameter space where convection be-gins to take over radiation in the star’s atmosphere. Thisproblem makes the determination of the surface gravitydifficult, but leaves that of the other parameters appar-ently unaffected. The determination of metallicity is con-siderably improved, thanks to the homogeneous spectro-scopic data published recently by Edvardsson et al. (1993).Instead of showing the traditional diagrams, we chose topublish the diagrams of the physical parameters with theinverted grids inside, i.e. the lines of constant photometricparameters
The determination of the fundamental atmospheric pa-rameters of stars remains an everlasting preoccupationof most stellar astronomers. The imminent release of theHipparcos results certainly makes this problem especiallyacute. On the other hand, new and hopefully more realisticatmosphere models have become available in the last fewyears (Kurucz 1991, 1993, 1994), as well as a large amountof precise, homogeneous [Fe/H] values (Edvardsson et al.1993). We found it worthwhile, therefore, to revise the al-ready published calibrations (North & Nicolet 1990, here-after NN90; Kobi & North 1990, hereafter KN90) and
Send offprint requests to: M. Kunzli
to take this opportunity to fill the gap that existed forstars with effective temperatures between 8500 and about10500 K. The efficiency of the Geneva system is similar tothat of the widely-used uvbyβ one, so it would be a pitynot to exploit it fully. The only drawback of the Genevasystem, compared to the uvbyβ one, is its sensitivity tointerstellar reddening for A and cooler stars.
In the following, we shall present successively the threetemperature domains which are treated separately (as inthe uvbyβ system), discussing together the reference (orstandard) stars and the theoretical grids. The possible ap-plications and limits of this calibration are discussed in theconclusion.
2. The B stars
2.1. Methods and results
Following Cramer & Maeder (1979), we use here thereddening-free parameters X and Y which have, as theseauthors showed, the optimum efficiency for determiningthe effective temperature and the surface gravity respec-tively. Although the definition of the X and Y parametersis given in several papers (Cramer & Maeder 1979; NN90),we recall it here for convenience:
X = 0.3788 + 1.3764 U − 1.2162 B1 − 0.8498 B2 (1)
−0.1554 V 1 + 0.8450 G
Y = −0.8288 + 0.3235 U − 2.3228 B1 + 2.3363 B2 (2)
+0.7495 V 1− 1.0865 G
where U , B1, B2, V 1 and G stand for the Geneva colourindices [U−B], [B1−B], [B2−B] etc. Let us recall that theZ parameter allows the separation of the Bp stars (mostlyof the Si and SiCr types) from the normal B stars (Cramer& Maeder 1980); it will not be used here, however.
The synthetic colours U , B1 etc. have been computedby one of us (BN) using recent Kurucz models with scaled
52 M. Kunzli et al.: Calibration of Geneva photometry
solar metallicities and a constant microturbulent velocityξt = 2 km s−1 (Kurucz 1993). The passbands used werethose determined by Rufener & Nicolet (1988). The Xand Y parameters computed in this way are very similarto those obtained for older models, because the additionalline opacity of the new models affects essentially the ul-traviolet rather than the visible part of the energy distri-bution. As before, these synthetic parameters do not re-produce exactly the observations and should be corrected.However, we adopted for this particular point another phi-losophy than that generally adopted to date (Lester et al.1986; Moon & Dworetsky 1985 etc.). Instead of compar-ing the observed colour indices with those interpolatedin the “direct” grids of synthetic colours from the knownfundamental parameters, we preferred to compare the fun-damental physical parameters with those interpolated inthe inverted grid from the observed colours. Briefly, theinversion of a grid implies an iterative, two-dimensionalspline interpolation in the “direct” grid (where colours aregiven for regularly spaced values of physical parameterslike Teff and log g) and results in an “inverted” grid givingthe physical parameters for regularly spaced values of thephotometric parameters. In other words, we first invert thegrid of the synthetic X and Y parameters once and for all,following the method described by NN90; then, we obtainfor the standard stars interpolated physical parameters,which can be compared with the fundamental ones. Foreffective temperature, we use the quantity θeff = 5040/Teff
rather than Teff itself because θeff varies linearly with theX parameter and the range of Teff is large. In this way,the rms scatter around the mean trend is roughly constant,while it would vary strongly if we used Teff directly; thisis much safer from the point of view of the least-squaresfit, and is equivalent to give a lower weight to the higheffective temperatures. We obtain
and we plot ∆θeff vs. θeff(fundamental) in Fig. 1. Thetrend can be fitted by a straight horizontal line in thepresent case, because the slope indicated by the least-squares method is smaller than its uncertainty. The in-terpolated reciprocal effective temperature will then becorrected using the formula:
θeff = θi − 0.008 (4)
where θi stands for the interpolated value of θeff . Theadvantage of this method over the previous one is thatthe grids need to be inverted only once, while differentcorrections can be tried thereafter, for example as newfundamental data are published. The fundamental starsare those used by NN90, supplemented by new data fromAdelman (1988) and Adelman et al. (1993). The Adelmaneffective temperatures cannot be considered as purely fun-damental because they are partly based on a comparisonbetween the observed energy distribution and a theoretical
one. However, the Balmer lines were also used to estimatethese temperatures, which appear a posteriori quite con-sistent with the purely fundamental ones of Code et al.(1976). In any case, these temperatures are evidently in-dependent from any possible systematic error in the pass-bands of the Geneva system.
Fig. 1. Difference between interpolated and fundamental θeff
values vs. fundamental θeff for the hot stars. The fitted hori-zontal line is shown; see Table 1 for the key to the symbols
The fundamental Teff values are listed in Table 5, to-gether with the interpolated and corrected values. Theuncertainties of the fundamental values are quoted fromtheir authors, while those of the interpolated values areestimated from the photometric errors (for a photomet-ric weight P = 1), as described in NN90. The θeff and Teff
values obtained from the observed colours by interpolationin the corrected grids are compared with their respectivefundamental values in Figs. 2a and 2b.
One clearly sees in Fig. 2b that for Teff , the scatter in-creases strongly towards small values of the X parameter,i.e. towards high temperatures, where the sensitivity of thephotometry to temperature is known to strongly decrease.On average, the rms scatter of the difference amounts to751 K. For X > 0.4 (Teff ≤ 21000 K), the scatter re-duces to 386 K, while it increases to 1388 K for X < 0.4.This scatter is mostly attributable to errors in the funda-mental data. Their contributions amount to about 96% ofthe total scatter. Photometric errors induce only a smalldispersion. There is a small systematic zero-point shiftof −73 K, essentially due to the hot stars, which wereweighted differently by using θeff instead of Teff to definethe correction. A shift of about −180 K was present withthe previous calibration of NN90.
M. Kunzli et al.: Calibration of Geneva photometry 53
a
b
Fig. 2. a) Difference between photometric and fundamentalθeff values vs. the X parameter. The continuous line is themean while the broken lines define the average rms scatter; seeTable 1 for the key to the symbols. b) Same as a), but for Teff .The horizontal line is arbitrarily set to zero. Notice the largeincrease of the scatter towards small values of X, i.e. towardsthe hotter stars
The difference between the interpolated and funda-mental log g values follows the trend shown in Fig. 3, andthe interpolated values have to be corrected according tothe equation
log g = log gi + 0.46− 4.83 10−5 T ∗eff (5)
where log gi is the interpolated surface gravity while T ∗eff isthe interpolated and corrected Teff . The fundamental val-ues are listed in Tables 2 and 3, as well as the interpolatedand corrected ones with their standard deviations. Table 2lists the eclipsing binaries, for which the most fundamentalvalues of log g can be determined, and the non-eclipsingbut well-known stars Sirius and Vega. In Table 3 we listthe members of the Orion OB1 association, whose surfacegravity is inferred from the models of internal structureof Schaller et al. (1992) for isochrones with log t = 6.8
Fig. 3. Difference between interpolated and fundamental log gvalues vs. the photometric Teff for the hot stars. The regressionline is the adopted correction
Fig. 4. Difference between photometric and fundamental logg values vs. the X parameter. The continuous line is the meanwhile the broken lines define the average rms scatter
(subgroup c) and 5.7 (subgroup d). Compared with thevalues given by NN90, the fundamental log g values givenhere are about 0.06 dex smaller. This is due to the newopacities used by Schaller et al. (1992). Figure 4 comparesthe photometric and fundamental values of log g; we seethat a very good accuracy can be achieved, of the order of0.10 dex, provided the star is not too hot. On the wholerange of B stars, the rms scatter of the residuals is only
54 M. Kunzli et al.: Calibration of Geneva photometry
σ = 0.09 dex and it is chiefly due to errors in photometricdata.
Finally, the inverted and corrected grid for solar metal-licity is shown in Fig. 5, in the form of a diagram log gvs. Teff containing the lines of constant X and constantY parameters. Although such a diagram is unusual, it al-lows graphical interpolation with the same efficiency asthe usual photometric diagrams where lines of constantphysical parameters are shown.
2.2. Accuracy of the numerical interpolation
The reliability of the inverted grids and of the bicubicspline interpolation used to determine the physical pa-rameters has been tested in the following way. Knowingthe synthetic colours of each atmosphere model, we deter-mined the corresponding physical parameters Teff and logg by interpolation in the inverted, but uncorrected grid.Then, we could verify that the interpolated Teff and logg values correspond to those defining the model to within1 K for Teff and 0.01 dex for log g. This holds true notonly for this particular grid, but also for all other gridspresented below; the metallicity [M/H], for cool stars, isalso interpolated with an accuracy better than 0.01 dex.
3. The intermediate stars
With the previous calibrations, the stars whose effectivetemperature lies between 8500 and about 10500 K couldnot be properly dealt with. This is why we have definednew photometric parameters we call pT and pG, whichare sensitive to effective temperature and surface gravityrespectively. These parameters have been defined in a sim-ilar way as the a0 and r ones of the uvbyβ photometry, andhave about the same sensitivity to the corresponding phys-ical parameters. However, pT and pG are not reddening-free, contrary to a0 and r, because the Geneva systemlacks an equivalent to the β index for the intermediateand cool stars. Therefore, the interstellar reddening mustbe either negligible or known and corrected to make thecalibration meaningful. These parameters are defined by:
pT = B2− V 1 + 0.1 X + 0.0635 (Y − 0.2 d)− 0.006 (6)
and their values for solar-composition Kurucz models (notcorrected by standard stars) are represented in Fig. 6. This
figure can be compared with Fig. 4 of Moon & Dworetsky(1985). If the colour excess E(B2 − V 1) is known, thecorrection of the pT and pG parameters has to be donethrough the relations
pT0 = pT − E(B2 − V 1) (10)
pG0 = pG− 1.1 E(B2 − V 1). (11)
The standard stars used to correct the grid for Teff arelisted in Table 4 while those used to correct log g are listedin Table 5-8. The standard stars for Teff are taken from es-sentially the same sources as for the B stars. Table 5 liststhe eclipsing binaries already used by Moon & Dworetsky(1985) while Tables 6-8 give the members of the Orion as-sociation, of the Pleiades and of IC 2391 respectively. Thecolour excess of the stars in Orion was determined fromthe intrinsic colours of Cramer (1982), and the pT and pGparameters were corrected for it. For the Pleiades, the pTand pG parameters have been corrected for a mean colourexcess E(B2−V 1) = 0.052 (Nicolet 1981) and for IC 2391,they have been corrected for E(B2− V 1) = 0.005 (North& Cramer 1981). For the Pleiades and IC 2391, the funda-mental log g values were deduced from the estimated effec-tive temperatures using the isochrones at log t = 8.0 and7.7 respectively. The differences between the fundamentalTeff and log g and their values interpolated in the (uncor-rected) inverted grids from the observed Geneva colours ofthe standard stars are shown in Figs. 7 and 8 respectively.For effective temperatures, we obtain:
Teff = 778 + 0.9009× Ti (12)
and for log g:
log g = log gi + 3.38− 5.975 10−4 T i + 2.46 10−8 T 2i .(13)
The inverted and corrected grids themselves are rep-resented for the three metallicities [M/H] = −1, 0,+1 inFigs. 9-11. We show these three diagrams, just to makeclear the effect of the metallicity on the Teff and log g es-timates. As in the case of the X and Y parameters, werepresent the physical parameters log g vs. Teff with thelines of constant pT and pG parameters, rather than thereverse. Note that for these stars, the metallicity is as-sumed to be known a priori so that the most relevant gridcan be used; the colours are not sensitive enough to metal-licity to give a significant estimate of it, except perhapsfor a few hot Am stars.
The comparison between the fundamental and photo-metrically determined Teff and log g is shown in Figs. 12and 13 respectively. The residual rms scatter is σ(Teff ) =197 K for the effective temperature, and σ(log g) = 0.135dex (c.g.s.) for the logarithmic surface gravity respectively.As for the hot stars, σ(Teff) is mostly due to errors inthe fondamental data and in the same proportion. Thisscatter represents the uncertainty in the determination ofthese physical parameters, but we insist that interstellarreddening must be negligible or corrected for.
M. Kunzli et al.: Calibration of Geneva photometry 55
0.27
5
0.25
0
0.20
0.15 0.1
1.95
1.90
1.80
1.70
1.60
1.50
1.40
1.30
-0.2
25
1.20
1.10
1.00
0.90
0.80
0.70
0.60
-0.05
0.50
0.40
0.30 0.20
0.00
0.15 0.10 0.
05
0
0.05
Fig. 5. Inverted and corrected grid with solar metallicity for the hot stars. Roughly horizontal lines are those of constant valuesof the Y photometric parameter, while the vertical ones are lines of constant X parameter. The iso-X lines are separated by aninterval of 0.05 magnitudes, while the iso-Y lines are separated by 0.025. For clarity, not all lines are labelled
5
4.5
4
3.5
3
2.5
1200
0 1150
0 1100
0 1050
0 1000
0
9750 95
00 9250
9000
8750
8500
8250
8000
Fig. 6. Direct, usual grid showing the (uncorrected) iso-Teff and iso-log g lines in the pG vs. pT diagram. These lines are nicelyorthogonal, apart from a small region at high gravity and small Teff . This diagram is quite similar to the r∗ vs. a0 one of theuvbyβ photometry
56 M. Kunzli et al.: Calibration of Geneva photometry
Fig. 7. Difference between interpolated and fundamental Teff
values vs. fundamental Teff for the intermediate stars. The re-gression line is shown. See Table 4 for the key to the symbols
Fig. 8. Difference between interpolated and fundamental log gvalues vs. photometric Teff for the intermediate stars. The lineis the fitted least-squares parabola
4. The cool stars
This category includes the late A, the F and the early Gstars. We did not explore the cooler stars although Kuruczmodels now exist for them, because their calibration isa delicate matter and was already explored by Grenon(1978, 1982) and Grenon & Golay (1979). The philosophyof the present calibration is roughly the same as in KN90:it uses the d vs. B2−V 1 and them2 vs. B2−V 1 diagrams,
which are sensitive to effective temperature, surface grav-ity and metallicity.
4.1. Effective temperature and surface gravity
The B2 − V 1 index is mainly sensitive to Teff , while thereddening-free parameter d is sensitive to both Teff andlog g. The new Kurucz models seem more realistic thanthe previous ones in the sense that the synthetic B2−V 1indices computed from them are closer to the observedvalues.
In the d vs. B2 − V 1 diagram, however, there wasa conspicuous change in slope of the iso-log g lines forTeff ≈ 8000 K at high gravity and Teff ≈ 6500 K atlog g ≈ 2.5. This change in slope seems to be linked withthe onset of convection in the superficial layers of the star’senvelope, and was especially conspicuous in the c1 vs.b−y diagram of Stromgren’s photometry (see e.g. Kurucz1991). Since then, a conceptual error has been detected inthe ATLAS9 code by Dr. F. Castelli (1996). The changesin slope occur when the convective flux is zero at the lastdepth in the model because the convection zone is wholelycontained in the atmosphere. The program computed con-vection differently depending on whether the last valuewas zero or not. The program has now been changed tobe consistent and the change in slope has been greatly re-duced or removed. Kurucz has recomputed the convectivemodels and fluxes and will distribute them on CD-ROMs(Kurucz 1996a, b). We have computed new Stromgrenand Geneva colours (the new Geneva colours were kindlycomputed for us by Dr. David Bersier). Figure 14 shows acomparison between the new grid (full lines) and the “old”one (i.e. before the change to ATLAS9, dotted lines) in thed vs. B2−V 1 diagram for solar metallicity: there is a sig-nificant difference, the new grid looking much smoother.The grids represented in this figure have not been cor-rected to match any standard star. Although the differ-ences involve essentially the cool stars, the “intermediate”grids are also concerned in the vicinity of their cool edge,i.e. for Teff = 8000 up to 8750 K. Therefore the interme-diate grids have also been recomputed.
Now, although the problem of the slope discontinu-ity of the iso-log g lines has been solved in the models,yet another problem remains in the observed main se-quence of the Hyades. The same grid for the new mod-els is represented in Fig. 15 together with the observedmain sequence of the Hyades cluster: one clearly sees asudden change of slope in the observed sequence aroundB2 − V 1 = 0.16 or Teff = 7000 K, which is not repro-duced by the models. This feature is not seen only in theHyades, since it is present also in the Praesepe cluster.It had not been noticed by KN90, because these authorsused only the Pleiades cluster, where the scatter is larger.Though the internal structure models foresee a very smalland gradual decrease of log g with increasing mass anddecreasing Teff on the isochrone, they are quite unable
M. Kunzli et al.: Calibration of Geneva photometry 57
0.260.24 0.22 0.20 0.18 0.16
0.140.12
0.100.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.1
4
-0.1
2
-0.1
0
-0.0
8
-0.0
6
-0.0
4
-0.0
2
0.00
0.02
0.04
0.06
0.080.
10
0.12
Fig. 9. Inverted and corrected grid log g vs. Teff for the intermediate stars having [M/H] = −1.0. The iso-pT and iso-pG linesare the vertical and horizontal lines respectively
0.20 0.18 0.16 0.14 0.12
0.100.08
0.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.1
4
-0.1
2
-0.1
0
-0.0
8
-0.0
6
-0.0
4
-0.0
2
0.000.02
0.040.
06
0.08
0.10
0.12
0.14
Fig. 10. Same as Fig. 9, but for [M/H] = 0.0, i.e. solar metallicity
58 M. Kunzli et al.: Calibration of Geneva photometry
0.22
0.200.18 0.16
0.14
0.120.10
0.080.06
0.04
0.02
0.00
-0.02
-0.04
-0.06
-0.08
-0.10
-0.1
8
-0.1
6
-0.1
4
-0.1
2
-0.1
0
-0.12-0.14
-0.16-0.0
4
-0.0
2
0.00
0.02
0.04
0.08
0.10
Fig. 11. Same as Fig. 9, but for [M/H] = +1.0. Notice that the metallicity effects are non-negligible at the low-Teff end of thegrid
to account for the sharp decrease around 7000 K the ob-served sequence would imply if the atmosphere modelswere entirely realistic. Therefore, something is missing inthe atmosphere models.
The standard stars with known Teff are listed inTable 9 with their sources. They are taken essentiallyfrom Blackwell & Lynas-Gray (1994), who relied on theinfrared flux method. We did not use temperature esti-mates based on spectrophotometry, since they generallyuse a fit to the same Kurucz models and are therefore notfundamental nor independent from our photometric ap-proach. The difference between the interpolated Teff andthe fundamental one is shown in Fig. 16. The interpola-tion was done taking luminosity effects into account, us-ing log g estimates essentially from the previous versionof our calibration, i.e. from KN90. The conspicuous gapseen between Teff = 7400 and 7900 K is that of Bohm-Vitense (1970), corresponding to the onset of convection(see also Mendoza 1956; Bohm-Vitense & Canterna 1974;Jasniewicz 1984). Since this gap represents a kind of phys-ical discontinuity, we have fitted two different functions toeither side of it: a horizontal straight line for the hot side(and the gap itself), and a regression line for the cool side.The correction in Teff takes the form:
Ti ≤ 6905 : ∆Teff = −640 + 0.12 Ti (14)
Ti > 6905 : ∆Teff = 193. (15)
Figure 17 compares the fundamental and photometric val-ues of Teff . The rms scatter of the differences amounts to54 K. The scatter induced by photometric errors is negli-gible in comparaison with errors in the fondamental data.The contribution are respectively 7% and 93%.
The standard stars with known log g are listed inTables 10 and 11, and belong to the Hyades and IC 2391open clusters respectively. Their log g values have beendetermined from the models of Schaller et al. (1992) as-suming ages log t = 8.8 for the Hyades and log t = 7.7 forIC 2391. The needed effective temperature was estimatedfrom the previous calibration of KN90; the interstellar red-dening was considered negligible for the Hyades, while amarginally significant colour excess E(B2 − V 1) = 0.005(North & Cramer 1981) was assumed for IC 2391. InTables 10 and 11, the last two columns give the log g val-ues and their uncertainty respectively, obtained from thegrids corrected by Eqs. (14-15) above, not from the uncor-rected grids. This holds for all tables where standard starsare listed, for all three physical quantities Teff , log g and[M/H].
The difference ∆ log g between the interpolated andfundamental values is represented in Fig. 18a for theHyades and IC 2391 clusters, as a function of T ∗eff , the cor-rected effective temperature. A dip about 0.3 dex deep isclearly visible at Teff ≈ 7000 K, which reflects the changeof slope of the observed sequence mentioned above.
M. Kunzli et al.: Calibration of Geneva photometry 59
Fig. 12. Difference between the photometric and fundamentaleffective temperatures for the intermediate stars (assumed tohave solar metallicities). The solid line is the mean and thebroken lines define the rms scatter
Since the members of the Hyades cluster draw an iso-metallicity line in the ∆ log g vs. T ∗eff diagram (except forthe Am stars of course), it is not possible to explore thepossible effects of metallicity with these stars alone. Usingother old clusters is of little help because their range inmetallicity is small. Therefore, we used the numerous fieldF and G stars studied spectroscopically by Edvardsson etal. (1993, hereafter EAGLNT93), which span about 1.2dex in [Fe/H] and have empirical surface gravities derivedfrom Stromgren photometry. Even if their log g valuesare not fundamental stricto sensu, they are much betterthan nothing as shown in Figs. 18b, c and d. In Fig. 18b,where only the stars of EAGLNT93 with [Fe/H] ≥ 0.1are shown together with those of the Hyades, one seesthat the field stars of EAGLNT93 are very well super-posed on the Hyades, which tends to validate a posteriorithe surface gravities given by EAGLNT93. In Figs. 18cand d are represented the EAGLNT93 stars with −0.1 ≤[Fe/H] ≤ +0.1, and −0.5 ≤ [Fe/H] ≤ −0.3: clearly, thecorrection ∆ log g is very much dependent on metallicityfor these cool stars. Fortunately, the metallicity depen-dence becomes vanishingly small at high effective temper-atures (i.e. Teff > 6600 K), so that the correction of log gcan be safely defined by the cluster stars in that range ofTeff .
The correction is not very simple, and we had to fitseveral different functions of the type:
log g = log gi + a+ b T ∗i + c T ∗2i (16)
where T ∗i stands for the interpolated and corrected effec-tive temperature. The coefficients a, b and c are listed in
Fig. 13. Difference between the photometric and fundamentallog g values for the intermediate stars (assumed to have solarmetallicities). The solid line is the mean and the broken linesdefine the rms scatter
Table 12. For Teff > 6916 K, we fitted a straight line ratherthan a parabola, so that the c coefficient is identical withzero there. Strictly speaking, this correction is valid onlyin the case of unevolved stars. The surface gravity of F gi-ants will not be estimated correctly with this calibration.We made many different attempts to extend the validityof our calibration to giants, using e.g. very old clusters orwell-classified field F giants, but none proved satisfactory.The present calibration in terms of log g must then be con-sidered as being limited to unevolved, or only very slightlyevolved stars. As a rule of thumb, we may say that it isvalid for log g ≥ 4.0. The comparison between the inter-polated and fundamental values is shown in Fig. 19. Therms scatter of the differences is σ = 0.15 dex. But thescatter is larger for the cool stars than for the hotter ones,because the iso-log g lines come closer together as Teff di-minishes. The stronger dependence on metallicity effectsfor cool stars probably also contributes to the larger σ.
4.2. Metallicity [M/H]
As in KN90, the metallicity is determined using the m2
vs. B2−V 1 diagram. Fortunately, the recent Kurucz mod-els are much more realistic in this diagram than were thepreceding ones. In Fig. 10 of KN90, one sees how the gridof iso-Teff and of iso-[M/H] lines folds again as [M/H]increases from +0.5 to +1.0; at the same time, Teff variesstrongly as [M/H] increases (at given B2 − V 1). Suchcomplicated behaviour induced severe problems for thegrid inversion as well as in the final 2-d interpolation ofthe physical quantities, even though KN90 dropped themost perturbing points. Furthermore, it was impossible
60 M. Kunzli et al.: Calibration of Geneva photometry
5
4.5
4
3
3.5
2.5
800082
50850087
509000
7750
7500
7250
7000
6750
6500
6250
6000
5750
5500
Fig. 14. Comparison between the models before (dotted lines) and after the correction (solid lines) proposed by Castelli (1996).Notice the much smoother iso-log g lines. These grids are the original ones, i.e. they have not been corrected to fit any standardstar
5
4.5
4
3.5
800082
50850087
509000
7750
7500
7250
7000
6750
6500
6250
6000
5750
5500
Fig. 15. Comparison between the models after the correction, and the observed main sequence of the Hyades cluster. Noticehow the observed sequence crosses the line log g = 4.0 near Teff = 7000 K. The grid is not corrected to fit any standard star
M. Kunzli et al.: Calibration of Geneva photometry 61
Fig. 16. Difference between interpolated and fundamental Teff
values vs. fundamental Teff for the cool stars. Notice the gapof Bohm-Vitense between 7400 and 7900 K. The correction isfitted by a horizontal line at high Teff and across the gap, andby a regression line at low Teff . See Table 9 for the key tosymbols
Fig. 17. Difference between photometric and fundamental Teff
values vs. B2−V 1 for the cool stars. The solid line is the meanand the broken lines define the rms scatter. See Table 9 for thekey to symbols
a
b
c
d
Fig. 18. a) Difference between interpolated and fundamentallog g values vs. photometric Teff for the Hyades. Notice theconspicuous dip at Teff = 7000 K. b) Same as a), but for fieldstars of EAGLNT93 having [M/H] > 0.1 (triangles), for thehotter Hyades stars (full dots) and for members of the youngcluster IC 2391 (crosses). c) Same as b), but for field stars ofEAGLNT93 having −0.1 < [M/H] < +0.1. d) Same as b), butfor field stars of EAGLNT93 having −0.5 < [M/H] < −0.3.Notice the very strong metallicity effect
to obtain straightforward estimates of the physical pa-rameters of Am stars. On the contrary, the new modelsshow no more folding, i.e. the iso-Teff curves now have amonotonous behaviour and they are, furthermore, remark-ably straight.
The standard stars used have been taken essentiallyfrom the huge work of EAGLNT93, which contains ho-mogeneous, high-resolution spectroscopic results for 189F and G stars. 157 of these stars have been measuredin the Geneva photometric system, which allows an ex-cellent calibration of our diagram in terms of [Fe/H] (weconsider here [Fe/H] to be equivalent to [M/H], although
62 M. Kunzli et al.: Calibration of Geneva photometry
Fig. 19. Difference between the photometric and fundamentalvalues of log g vs. B2 − V 1 for the cool stars. The solid lineis the mean and the broken lines define the average rms scat-ter. Notice the larger scatter at large B2 − V 1. In the key tosymbols, “Ed” stands for “Edvardsson et al. (1993)”
it is not quite true for very metal-deficient stars). Thesestandard stars are listed in Table 13. Some other, hotterobjects taken from other sources (Perrin et al. 1977; Cayrelde Strobel et al. 1992 and especially Burkhart & Coupry1989) are listed in Table 14. To correct the grid, we hadto use a more complicated method than that of KN90. Wefirst defined a preliminary correction ∆[M/H]0 as a func-tion of T ∗eff (the interpolated and corrected Teff), which isthe only step KN90 had done. A plot of ∆[M/H]0 vs T ∗eff
does not show any trend, only a zero point-shift
∆[M/H]0 = 0.148 (17)
where the subscript 0 refers to this zeroth-order correction,while T ∗eff is the temperature obtained by interpolation inthe grids corrected by Eq. (14). This correction is shown inFig. 21. In a second step, we plotted the residual ∆[M/H]against the spectroscopic values [Fe/H]: there is a cleartrend, as shown in Fig. 20. This trend varies slightly witheffective temperature, according to the relation:
where [M/H]0 is the interpolated metallicity, correctedby Eq. (17) above; this relation was found using a least-squares fit. Therefore the difference ∆[M/H] is not somuch a function of B2 − V 1 or temperature, but chieflyof metallicity [M/H] itself. The latter dependence hadbeen overlooked by KN90, essentially because of the muchsmaller number of standard stars they used. This correc-tion allows us to obtain a photometric metallicity of the
Fig. 20. Difference between the photometric (i.e. interpolatedand corrected by Eq. (16)) and fundamental values of [M/H]vs. photometric [M/H], for stars with 5600 ≤ Teff ≤ 6500 K.Notice the remaining trend
Hyades cluster which agrees well with the accepted spec-troscopic value: we obtain [M/H] = +0.08, while Cayrelet al. (1985) give [Fe/H] = +0.12 ± 0.03 for the samestars.
It is now possible to treat Am and other metal-lic stars like any other stars: for the typical Am star63 Tau = HR 1376, we obtain Teff = 7247 ± 59 K,log g = 4.09 ± 0.08 and [M/H] = +0.55 ± 0.06.The effective temperature we find is in excellent agree-ment with the value Teff1 = 7250 − 7400 K esti-mated by Smalley (1993a), who uses the infrared fluxmethod and corrects for the presence of a cool com-panion. The surface gravity agrees perfectly with thatobtained by Smalley (1993b) from the uvbyβ colours(log g = 4.13). Our metallicity is a bit smaller than,but still agrees well with that of Smalley (1993b) whofound [M/H] = 0.651 ± 0.095. Notice that he adoptedTeff1 = 7570 K, which seems too high, and lowering Teff1
would imply a lower [M/H] as well. On the other hand,Burkhart & Coupry (1989) found [Fe/H]= +0.4 on thebasis of high-resolution spectroscopy, in very good agree-ment with our estimate.
The difference between our photometric [M/H] andthe fundamental [Fe/H] values is shown in Fig. 22. The av-erage rms scatter of the differences amounts to σ = 0.097,which is similar to, though slightly larger than the scat-ter found by EAGLNT93 (their Eqs. (13-15)) for the dif-ferences between their [Fe/H] values and the [M/H] val-ues derived from uvbyβ photometry. When only the starsstudied by EAGLNT93 are considered, our rms scatterdrops to a value quite similar to theirs. Therefore the
M. Kunzli et al.: Calibration of Geneva photometry 63
Fig. 21. Difference between the interpolated and fundamental[M/H] vs. photometric Teff . The regression line correspondingto Eq. (16) is shown. The key to the symbols is explained inTable 14; “Ed” stands for “Edvardsson et al. (1993)”
capability of the Geneva system to estimate metallicitiesis excellent; in this regard, this system is quite competitivewith the uvbyβ one.
The inverted and corrected grids are shown in Figs. 23and 24. Figure 23 is a log g vs. Teff diagram, showing thesolar-metallicity grid with lines of iso-B2 − V 1 and iso-d parameters. Figure 24 is a [M/H] vs. Teff diagram forlog g = 4.0, showing the iso-B2− V 1 and iso-m2 lines.
5. Conclusion
We have presented an updated and complete calibrationscheme of the Geneva photometric system in terms of ef-fective temperature, surface gravity and metallicity forall B to mid-G stars of the main sequence or just aboveit. This calibration can be applied to giant B, A and Fstars but not to red giants, and it will not give a reli-able estimate of the surface gravity of giant F stars . Itcannot be used either for B to G supergiants, since it isbased on LTE atmosphere models and no supergiant hasbeen included in our set of standard stars. Reddened Bstars can be dealt with (provided the reddening is nottoo large, i.e. not greater than about E[B − V ] ≈ 0.6and the reddening law is standard). For cool stars(Teff <∼ 7000 K), metallicities can be safely estimated inthe range −2.0 <∼ [M/H] <∼ +0.3 dex, and the upper limitof this range extends to about +0.6 in the case of Amstars, which are hotter (7000− 8000 K).
The new features of this calibration are:
– A new, simpler way of fitting the theoretical grids tothe standard stars
Fig. 22. Difference between the photometric and fundamental[M/H] vs. B2 − V 1 for the cool stars. The symbols are thesame as in Fig. 18. The solid line represents the mean whilethe broken lines define the rms scatter
– The possibility to estimate the physical parameters ofintermediate stars
– The use of new, generally more realistic Kurucz models– A more reliable estimate of the metallicity of cool stars,
thanks to the large number of metallicity standardsand to the smoother behaviour of the new Kuruczmodels in the m2 vs. B2− V 1 diagram
– The possibility to estimate the physical parameters ofmetallic-line stars
The weakness of our calibration lies in the inadequacyof the atmosphere models around the transition be-tween radiative and convective atmospheres, i.e. nearTeff ∼ 7000 K. We have tested different ways of treat-ing the convection, but found no simple way to re-produce the change of slope of the cluster sequences.We also checked whether a change of the micro-turbulent velocity Vturb could account for that; thisis a reasonable assumption, since it is known em-pirically that Vturb increases with Teff up to about8000 K (EAGLNT93, Coupry & Burkhart 1992). But, theincrease of Vturb from cool (5500 K) to hot (7000−7500 K)stars being observationally both small (about 2 km s−1)and smooth, it cannot account for the observed sequenceof the Hyades in the d/B2 − V 1 diagram, since thissequence has a slope which increases abruptly around7000 K. One would have to make both Vturb increasee.g. from 2 to 4 km s−1 and overshooting disappear atTeff = 7000 K, to reproduce the observed sequence, butthis would appear extremely ad hoc. Therefore, the cure
64 M. Kunzli et al.: Calibration of Geneva photometry
0.5
0.55
0.6
0.65
0.7
0.8
0.9
0.47
5
0.42
5
0.37
5
0.32
5
0.27
5 0.22
5
0.17
5
0.12
5
0.07
5 0.02
5
-0.0
25
-0.0
75
-0.125
1 1.051.1
1.15
1.21.25
1.31.35
1.41.45
1.51.55
1.6
1.651.7
1.75
1.8
1.85
1.9
1.95
2
Fig. 23. log g vs. Teff diagram for the cool stars with solar metallicity, showing the inverted and corrected grids with iso-d andiso-B2− V 1 lines. B2 − V 1 varies here between −0.125 and +0.475 and d between 0.5 and 2.0. Notice the tightening of thelines around Teff ≈ 7000 K, which reflects the change in slope of the observed Hyades’ sequence
Fig. 24. [M/H] vs. Teff diagram for the cool stars with log g = 4.0, showing the inverted and corrected grids with iso-m2 andiso-B2− V 1 lines. B2− V 1 varies between −0.05 to +0.50 by steps of 0.05 magnitudes. m2 varies between −0.18 to −0.58 bysteps of 0.04 magnitudes
M. Kunzli et al.: Calibration of Geneva photometry 65
seems far from straightforward. As a result, the photomet-rically estimated surface gravities of cool stars are not soreliable as one would expect from the accuracy and homo-geneity of the Geneva data. The most reliable values arethose obtained for unevolved, solar-metallicity stars.
The calibration we have just presented is not completein the sense that it does not give explicitly the absolutemagnitude, bolometric correction, mass, colour excess anddistance of the star. In particular, we have dropped the de-termination of the mass which was offered by NN90 andKN90; the reason is that the Barcelona group (Prof. F.Figueras and co-workers) has devised a code which in-terpolates the mass and age of a star from its effectivetemperature and gravity, in evolutionary tracks from var-ious authors including Schaller et al. (1992). We had noreason to duplicate their work. The other physical param-eters can be found using calibrations published by otherauthors. For the intrinsic colours (hence interstellar red-dening) of O, B and early A stars (hereafter “hot stars”),see Cramer (1993), who updated an earlier work (Cramer1982). Intrinsic colours of B2 to M0 stars have also beenestimated by Hauck (1993). The bolometric correction ofthe hot stars can be obtained from a formula given inAppendix by Cramer (1984a); a formula giving Teff as afunction of X is also given in that paper, but it shouldbe considered as superseded by our work. A calibration ofthe X and Y parameters in terms of Crawford’s β index isworth mentioning too (Cramer 1984b): it allows to detectHβ emission when both Geneva and β photometric dataare available. The absolute magnitude of hot stars can beobtained from a recent work by Cramer (1994), which su-persedes an earlier calibration (Cramer & Maeder 1979).The absolute magnitude of A and F stars (excluding su-pergiants) can be obtained from the calibration of Hauck(1973). The intrinsic colours of A and F supergiants havebeen estimated by Meynet & Hauck (1985). Finally, theGeneva system has been calibrated for G, K and M-typestars essentially by Grenon (1978, 1982) and Grenon &Golay (1979), as mentioned above in Sect. 4.
A fortran code has been written, which applies our cali-bration to stars measured in the Geneva system. This codeis available by anonymous ftp at the Centre de Donneesde Strasbourg (CDS), following the instructions given inA&A 280, E1-E2 (1993). This code uses several ascii filescontaining the inverted grids, which are, of course, alsoavailable. The (uncorrected) Geneva colours of the Kuruczmodels are also available at the CDS.
Acknowledgements. This work was supported in part by theFonds National de la Recherche Scientifique. We thank Mr.David Bersier (Geneva Observatory) for having computed thecolours of the corrected models of cool stars.
A(93): Adelman et al. (1993). A(88): Adelman (1988). C: Code et al. (1976).L: Lanz (1987). Le: Leggett et al. (1986).
1. log g: calibration of North & Nicolet (1990).2. log g: Cayrel de Strobel et al. (1992).3. log g: taken from North & Nicolet (1990), Table 1.4. Weighted mean of the results of the sources mentioned in column 8.5. Lanz's temperatures have been increased by 1.15%.6. Observed Geneva colours corrected for the contribution of the visual companion.
M. Kunzli et al.: Calibration of Geneva photometry 67
Table 2. Eclipsing binaries and other stars used as surface gravity standards (hot stars)
HD HR/DM X Y Teff log g Sources Remarks log g ±σfrom X,Y
1. Teff estimated from the X parameter and Cramer’s (1984b) calibration.2. Individual log g are 3.47 and 3.18 for the primary and the secondary componentsrespectively. The adopted log g value is a mean, weighted by the respective visualluminosities.3. Observed Geneva colours corrected for the contribution of the visual companion.
Table 3. Stars of the Orion association used as surface gravity standard (hot stars)
HD Brun X Y Teff log g Sub-group log g ±σnumber from X,Y
36629 25 0.404 -0.015 20245 4.22 c 4.25 0.1636655 50 1.241 0.011 11770 4.30 c 4.34 0.0836842 246 0.854 -0.006 14757 4.29 c 4.34 0.1236883 330 0.854 -0.012 14757 4.29 c 4.39 0.1136899 342 1.546 -0.059 9450 4.31 c 4.28 0.04
414 1.444 0.002 10550 4.31 c 4.27 0.0536918 417 0.825 -0.003 14994 4.29 c 4.30 0.1236939 437 1.196 0.013 11940 4.31 c 4.33 0.0936936 440 0.735 0.002 15781 4.28 c 4.21 0.1336939 442 1.099 0.018 12300 4.30 c 4.27 0.1036958 480 0.609 -0.005 17093 4.27 c 4.21 0.14
508 1.161 0.026 11900 4.30 c 4.24 0.1036983 520 1.360 0.014 11080 4.31 c 4.28 0.0736998 529 1.326 0.028 11290 4.31 c 4.23 0.0837000 552 0.562 -0.012 17689 4.26 c 4.26 0.1436999 581 0.950 -0.013 14015 4.29 c 4.44 0.1137025 621 0.652 -0.020 16609 4.27 c 4.37 0.1337017 632 0.432 -0.028 19702 4.24 c 4.39 0.1637059 736 1.251 0.003 11580 4.31 c 4.38 0.0837062 760 0.616 -0.014 17011 4.29 d 4.30 0.1437058 761 0.443 0.001 19376 4.23 c 4.09 0.1637060 776 1.487 0.017 10350 4.31 c 4.17 0.0537114 920 1.369 -0.012 10970 4.31 d 4.39 0.0637129 940 0.506 -0.001 18255 4.25 c 4.13 0.1537150 980 0.418 -0.006 19968 4.27 d 4.16 0.0637174 992 1.402 -0.015 10580 4.31 d 4.37 0.0537334 1109 0.485 -0.008 18789 4.25 c 4.20 0.15
68 M. Kunzli et al.: Calibration of Geneva photometry
Table 4. Standard stars of effective temperature (intermediate stars)
HD HR pT◦ pG◦ Teff ±σ log g AV Sources Remarks Teff ±σfrom pT, pG
B(94): Blackwell et al. (1994). B(91): Blackwell et al. (1991). Le: Leggett et al. (1986).L: Lanz (1987). A(88): Adelman (1988). A(93): Adelman et al. (1993).C: Code et al. (1976).
1. log g: calibration of North & Nicolet (1990).2. log g: calibration of Kobi & North (1990).3. log g: taken from North & Nicolet (1990), Table 1.4. log g: taken from Kobi & North (1990), Table 15. log g: Cayrel de Strobel et al. (1992).6. Weighted mean of the results of several sources mentioned in Col. 9.7. Lanz’s temperatures have been increased by 1.15%.
Table 5. Eclipsing binaries used as surface gravity standards for intermediate stars
HD HR pT◦ pG◦ Teff log g Remarks log g ±σfrom pT, pG
B(94): Blackwell et al. (1994). B(91): Blackwell et al. (1991). Le: Leggett et al. (1986).MD: Moon et Dworetsky (1985). C: Code et al. (1976).
1. [M/H]: calibration of Kobi & North (1990).2. log g: calibration of Kobi & North (1990).3. [M/H]: Cayrel de Strobel et al. (1992).4. log g: Cayrel de Strobel et al. (1992).3. log g: taken from Kobi & North (1990), Table 1.
M. Kunzli et al.: Calibration of Geneva photometry 71
Table 10. Stars of the Hyades used as surface gravity standards (cool stars)
Nbr. HD/DM (B2− V 1)◦ d m2 Teff log g log g ±σfrom grids