Since Dr. Harvey has already mentioned the main results of my observations, I want to restrict myself to two topics:

1. i) a brief description of the improvements of my method as compared to that used by Stenflo (1973),

2. ii) some puzzeling aspects of my results which might be discussed in this meeting.

The main differences between the two methods of simultaneous Zeeman polarization measurements used by Stenflo (1973) and by myself are based on suggestions by Schröter (1973):

1. a) use of three instead of only two differently split lines;

2. b) use of less temperature sensitive lines;

3. c) use of model atmospheres instead of a Milne-Eddington model for the calculation of the ‘calibration curves’;

4. d) observation of selected Ca+K features in order to avoid the statistical method with the scatter-plot diagram.

Historically, attempts to model the temperature structure of faculae have generally suffered from a rather basic contradiction. Models which were based on center to limb measurements of the continuum contrast of faculae disagree with models that are based on measurements of line profiles in faculae. The “continuum” models predict line weakenings which are of larger amplitude than what is observed, and the “line profile” models predict a continuum contrast that is less than what is observed. Chapman (1976) discusses this problem in some detail. It is the purpose of this paper to show that there is a fundamental reason for this historical contradiction between line profile measurements and continuum contrast measurements: The line profile and the continuum contrast of a given facular are both a function (the two functions are different) of the size of that facula. The first indication of this fact was given by Frazier (1971). Figure 1 shows the contrast of faculae in the core of the line Fe Iλ 525. 0 nm, and in the continuum, as a function of the observed magnetic flux. One can see immediately that the contrast in each channel depends on φ in a much different manner. Therefore, one can conclude that the shape of the entire line profile will vary as a function of φ. On the basis of Figure 1, we must expect that this variation of the line profile will be continuous from infinitesimally small faculae up through very large faculae, and indeed, all the way up to pores.

Photospheric faculae are now believed to be closely associated with the small-scale solar magnetic field. In order to obtain reliable observations of solar magnetic fields, one needs to have a good description of faculae, which is presently lacking. The problem in obtaining good observational data is severe because faculae are usually not spatially resolved, particularly so in the case of spectroscopic observations. Proper use of spectroscopic observations also requires some knowledge of solar velocity fields and atomic physics in the case of a non LTE analysis. Many of these problems can be avoided by making use of the wings of the Ca II K-line. The wing of this line is unaffected by magnetic and velocity effects. The formation of the line has become increasingly well understood and most of the wing (with the exception of the inner 1 - 2 Å) is formed in LTE. The line is so strong that its formation spans the whole depth of the photosphere.

In an attempt to interpret the observed properties of small scale magnetic fields at the solar surface, a set of models has been calculated based on the assumption of a magnetostatic equilibrium. The basic assumptions made are:

1. i. The observed magnetic elements are magnetostatic flux tubes.

2. ii. The efficiency of convective heat transport inside the tube is reduced with respect to that in the normal convection zone; the horizontal convective heat transport in the tube is suppressed completely by the magnetic field.

3. iii. Close to the tube, horizontal convective heat transport is reduced due to the proximity of the magnetic field.

Harvey has reported in this discussion observations of mean downflows in magnetic elements that increase in speed with depth. It is therefore pertinent to ask what role such flows may play in the production and maintainance of the flux concentrations.

Equilibrium Models

The first thing to be said is that such flows cannot directly influence the horizontal momentum balance, i.e. their dynamic pressure is negligible.

A steady internal flow such as that proposed by Ribes and Unno (1975) for chromospheric concentration would require supersonic flows with very large curvature at photospheric levels. There is no evidence for transverse flows of such magnitude.

Recent interest in the subject of this discussion gained momentum by an observation which did not concern magnetic fields directly. The filigree which Dr. Richard Dunn on Sacramento Peak found 2 Angstrom off the center of Hαis a bright and crisp structure in the photosphere with a width o: 1/5 arcsec. It was described in proper detail by Dunn and Zirker (1973). Even in the printed pictures in their paper one clearly sees one step beyond the solar granulation. The filigree is certainly related to the small scale structure of the photospheric magnetic field, but it is not yet clear whether the flux elements are exactly cospatial and have the same small dimensions. Simon and Zirker (1974) concluded froi spectra that the field structure is wider than the filigree. On the other hand Harvey (1976) in his excellent review of the observations has also presented the arguments of several authors who conclude that the sizes of the flux elements are as small as those of the filigree. This discrepancy certainly needs further study before such even more delicate questions as the spatial extent of the downdraft inside and around the flux elements can be reliably answered from observations. The theoretical interpretation of the downdraft depends on this answer as different sources of the mass flow are involved: the overlying atmosphere and the convergent mass flow of the surrounding convection. The latter stays partly outside the flux element, partly diffuses into it with an efficiency that might be enhanced by convection of still smaller scale.

The idea of organizing a joint discussion on the impact of ultraviolet observations on spectral classification arose during the IAU symposium n»72 held at Lausanne, in the honour of W.W. Morgan. It became evident at that time that the wealth of spectroscopic and spectrometric data already obtained from various orbiting telescopes should shed some new light on the problem of spectral classification. I whish to thank here the members of the Organizing Committee who planned this Joint Discussion: W.P. Bidelman, A.D. Code, C. Jaschek, and K. Nandy. I also thank the authors of invited reports and contributed papers, as well as those who participated in the discussions.

A spectral classification system is a morphological device for arranging in order a finite series of observations of individual objects, One has thus three elements: an “order”, i.e. an abstract discontinuous scheme, a series of objects and a certain type of observation (spectrogram) that is performed on all objects of the sample. The “order” is in principle arbitrary; one could think for instance of an order based on richness of lines, or on the presence of hydrogen lines, etc. However such an idealization is clearly simplistic, because it would imply that the classifier is completely unaware of theory -but he is not, if for no other reason than simply that he would never get his PhD. Thus one should add the constraint that the scheme one adopts be physically sound.

The methods of spectral classification from the low dispersion ultraviolet spectra obtained with the S2/68 experiment in the TD1 satellite have been described. The bright stars, the spectra of which are photometrically accurate, can be divided into natural groups according to the spectral appearance of the features. These features vary in strength with spectral type and luminosity, and enable separation between main sequence and luminous stars. The limits for these stars are at BO to at AO. For fainter stars the spectral data have been combined to obtain narrow band magnitudes at several wavelengths. These photometric bands have an effective width of 100 A. An ultraviolet photometric system which enables determinations of spectral type and luminosity of early type stars is described and the results for about 3000 stars is presented. The photometric system considered here consists of the ultraviolet colour indices (m2740−V,) (m2190−V) and (m1490−V).

The experiment S2/68 has supplied a large number of ear type spectra in the wavelength region 1350–2730 A. The resolution is 37 A, which is about equivalent to a reciprocal dispersion of 1850 A/mm. We had at our disposal about one thousand spectra of stars brighter than 6m5.

The availability of such material has lead us to establish a spectral classification system based only upon features visible in the spectrum. A second step was to compare this system to the MK classifications.

It turned out to be easy to establish a temperature sequence based upon convenient intensity ratios of line features and somewhat more difficult to establish luminosity criteria. Among the B type stars it is possible to distinguish main sequence objects, supergiants and intermediate objects.

Ultraviolet low resolution spectra of most of the B stars observed with the S2/68 experiment onboard the TD1 satellite in the spectral range 1350 A - 2500 A are computer-investigated in order to define a grid of standard behaviors (J.M. Vreux and J.P. Swings, 1976 Astron. Astrophys., in press). The behavior of the relative intensities of the Balmer and Paschen continua is studied as a function of the spectral type and the visible photometric index Q: both relations are shown to be luminosity dependent. A comparison to theoretical predictions and to previous studies is also presented. A measure of the slope of the pseudo-continuum drawn between ~ λ 1660 A and ~ λ 2550 A is studied in the same way: the effect of luminosity is discussed in connection with recently proposed temperature scales for super-giants. The behavior of the depths of the absorption features at λλ 1550 A, 1940 A, 2000 A, 2055 A, 2105 A, 2340 A and 2395 A with respect to the spectral type, the Q index and the luminosity class is also briefly presented. Special emphasis is given to the absorption features at λ 1550 A in connection to the discussion of the Felll lines in the ultraviolet spectra of early B stars by J.P. Swings, M. Klutz, J.M. Vreux and E. Peytremann (1976, Astron. Astrophys. Suppl. 25, 193). Preliminary results of a study of high resolution Copernicus spectra in the A 1550 A region are also presented: it is shown that the role of Felll remains predominant for stars of spectral type as early as about BIII - BIIII (J.P. Swings and J.M. Vreux, to be published).

In this contribution we will discuss the extension of visual photoelectrical photometry into the ultraviolet and its potential impact on stellar classification. As is the case in the visual, classification by photoelectrical photometry in the ultraviolet has some important advantages over classification by inspection of individual spectra. Studies that require global characteristics of stellar properties -such as studies of stellar distribution, extinction properties of interstellar dust, galactic structure - often require large numbers of observations to derive statistical properties. In photoelectric photometry such a large number of observations can usually be obtained in a reasonable amount of time and relatively conveniently processed without subjective criteria being applied to the data. Comparatively, the classical way to obtain stellar classification is a complex process that requires many steps.

During the Skylab I, II and III missions, ultraviolet spectra were obtained in 188 fields with a. 15 cm aperture objective-prism spectrograph. The instrument has been described by Henize, Wray, Parsons, Benedict, Bruhweiler, Rybski and O’Callaghan (1975; hereafter referred to as Paper I).

The spectra cover the wavelength region from 1300 to 5000 Å and have a resolution of 2 Å at 1400 Å and 12 Å at 2000 Å. Absorption and/or emission lines of C II λ 1335, Si IV λ 1394, 1403 and C IV λ 1549 are visible in more than one hundred stars. The lines of Si IV and C IV are found to be particularly sensitive to stellar temperature and luminosity. Since these lines are visible in spectra of moderate to low resolution it is clear that they should be of special interest in any UV classification system for faint stars. This paper investigates the correlation of the intensities of the C IV and Si IV lines with MK spectral type, and presents a preliminary classification scheme for 04 to B2 stars based on these lines.

Ultraviolet stellar fluxes have been presented for 140 bright stars in the spectral region from 1200 A to 3600 Å in a graphical and tabular form (A. D. Code and M. R. Meade, 1976). The spectra represent a subset of 0A0-2 spectrometer data on file at the National Space Science Data Center. The monochromatic flux is given in units of ergs cm−2 sec−1 Å−1 with a spectral resolution of about 22 Angstroms in the region from 3600 Å to 1850 Å and approximately 12 Angstroms from 1850 Å to 1160 Å.

The monochromatic flux, Fλ, is based upon the absolute calibration described by Bless, Code and Fairchild (1976).

The empirical temperature scale of Code, Davis, Bless, and Hanbury Brown has been recalibrated in terms of an ultraviolet color derived from OAO-2 photometry, and mean temperatures have been computed for MK spectral types from O9 to A0. The dereddened color (1910-V)o was chosen as a temperature indicator because it is not strongly affected by lines or continuum edges. This parameter was computed for 16 stars with spectral types earlier than A2, luminosity classes III-V, and normal spectra, whose temperatures had been measured by Code et al. Plotting 8eff against (1910-V)o, we obtained the relation.θeff = 0.111 (1910-V)o + 0. 565. Next, the quantity (1910-V) was computed for over 150 stars in the same spectral-type range which had been observed by OAO-2, and a mean value was obtained for each spectral type. (Class III stars were excluded for types B5 and later because they were found to be systematically redder than classes IV and V - possibly an evolutionary effect.) Finally, the mean values of (1910-V)o were converted to effective temperatures using the formula given above. The resulting temperature scale is in good agreement with the scales of Morton and Adams, Schild, Peterson and Oke, and Code et al. for types B3 and later, but between types B0.5 and B2 the new scale and the Code scale are significantly hotter than the others. A particular application of the new calibration is the determination of small temperature changes occurring in early-type variable stars, since in this case the large and uncertain reddening correction drops out. For the β Cephei variables δ Cet and γ Peg, we obtained ΔT = 550°K and 185°K, respectively. This work was supported by the National Aeronautics and Space Administration under grants NSG 5004 and NSG 5069.

Observations shortward of the hydrogen Lyman limit provide sensitive determinations of stellar temperatures and interstellar absorption. Such data are of particular value in studies of hot white dwarfs, for which a large fraction of the emission occurs in the extreme ultraviolet band (100-1000 Å). Observations of HZ 43 and Feige 24 have been obtained with the Apollo-Soyuz extreme ultraviolet telescope; both stars are copious EUV emitters, with 4 × 10−9 and 3 × 10−9 erg/cm2 sec in the 170-620 A band respectively. The EUV data combined with optical spectrophotometry, allow their temperatures to be estimated as 80, 000 and 60, 000 K respectively. The corresponding interstellar neutral hydrogen column densities are ~ 4 × 1018 cm−2.