3 The spectrum of HD108

Figure 1: Composite spectrum of HD108 from blue to IR: the $\lambda \lambda$3875-3942 domain comes from the 1986 data, 3942-4300 from the 1994 data, 4300-4930 from the 1999 data, 6500-6710 from the 1997 data, 7000-7100 from the 1990 data and 8430-11000 from the 1989 data.

The spectrum of HD108 is quite peculiar, showing a great number of emission lines. The blue part of the spectrum is dominated by the presence of strong N III $\lambda \lambda$4634-41 and He II $\lambda$4686 emission lines (see Fig. 1). As for many Of stars (Underhill et al. 1989), these lines stand on top of a broad emission bump between 4600 and 4700 Å. In addition, Si III $\lambda \lambda$4552, 4568, 4574 and various O II emission lines, as well as the unidentified Of emissions situated near $\lambda \lambda$4485, 4504 are also present. All Balmer lines except H$\alpha$ appear as P Cygni profiles during most of our observing runs, but their morphology changes with time (see Sect. 6). On the contrary, some lines always appear in absorption, for example Si IV $\lambda \lambda$4088, 4116 and He II $\lambda \lambda$4200, 4542. Some interstellar features can also be seen in Fig. 1: besides the Ca II H and K lines, there are diffuse interstellar bands (DIB's) around 4428, 4726 and 4763 Å, as well as a CH line at 4300 Å.

The H$\alpha$ and He I $\lambda \lambda$6678, 7065 emissions dominate the red part of the spectrum. In our 1989 near-IR spectrum, the Paschen lines of Hydrogen appear in emission, as well as He I $\lambda$10830 which is very intense. He II $\lambda$10124 is also present, but in absorption.

 

 

Table 2: Mean heliocentric radial velocities of He II $\lambda \lambda$4200, 4542. We have also included the RVs of N III $\lambda \lambda$4634, 4641 and He II $\lambda$4686 for completeness. The date of the observation is given in the format HJD-2440000.

Run	Date	RV4200/4542	RV4634/4641	RV4686
		kms-1	kms-1	kms-1
Jul. 1986	6627.582	-66.4
	6628.548	-47.0
Jul. 1987	7007.536	-68.5
	7007.544	-79.2
	7007.549	-72.9
	7007.554	-74.2
	7008.502	-84.4
	7008.517	-80.4
	7008.524	-84.5
	7009.526	-82.9
	7009.538	-78.4
	7010.549	-74.8
	7010.558	-73.4
	7011.554	-72.8
	7011.561	-75.8
	7011.568	-77.4
	7011.572	-74.9
	7012.566	-77.2
	7012.573	-71.9
	7013.540	-70.7
	7013.545	-64.7
	7013.549	-70.2
Aug. 1987	7016.601	-69.2	-75.4	-72.2
Aug. 1991	8490.478	-93.5	-93.3
	8490.489	-86.4	-85.1
	8490.505	-85.2	-84.0
	8492.739	-94.2	-89.9
	8496.590	-84.6	-80.4
	8497.581	-77.0	-72.6
Oct. 1993	9264.447	-61.4
	9264.452	-67.3
Aug. 1994	9576.553	-44.2
	9577.542	-64.0
	9579.522	-64.8
	9580.481	-58.0
	9581.535	-58.8
	9582.511	-63.8
	9583.533	-59.3
Aug. 1996	10316.650	-58.9	-63.4	-56.9
	10316.658	-58.3	-63.0	-57.0

Run	Date	RV4200/4542	RV4634/4641	RV4686
		kms-1	kms-1	kms-1
Aug. 1996	10318.644	-56.8	-66.8	-53.9
Jul. 1997	10638.574	-60.8	-63.0	-41.4
	10639.557	-61.7	-61.5	-54.8
	10642.573	-59.1	-60.3	-53.4
Sep. 1998	11066.635	-64.1	-77.0	-65.6
	11070.629	-65.0	-76.8	-64.9
	11072.638	-62.8	-76.6	-61.7
Nov. 1998	11132.377	-73.1
	11132.421	-66.5
	11133.330	-69.4
	11133.380	-68.2
	11133.432	-69.7
	11133.589	-71.6
	11134.378	-67.9
	11134.448	-66.9
	11135.339	-68.4
	11135.394	-69.8
	11135.458	-66.4
	11136.373	-67.0
	11136.439	-69.5
	11137.401	-67.6
	11137.467	-67.8
Jul. 1999	11374.597	-70.1	-75.7	-68.5
	11376.532	-75.7	-78.2	-69.6
	11378.530	-71.2	-75.8	-65.1
Aug. 1999	11396.600	-68.8	-74.4	-66.4
	11403.635	-73.4	-79.2	-68.9
	11406.634	-69.7	-76.0	-67.1
	11407.639	-71.6	-77.3	-69.4
Sep. 2000	11810.586	-69.4	-79.6	-66.9
	11810.595	-70.6	-79.6	-67.5
	11811.567	-68.6	-78.5	-67.4
	11811.575	-68.6	-78.9	-67.2
	11812.594	-71.6	-80.0	-70.4
	11813.603	-67.8	-76.5	-67.8
	11814.595	-69.7	-79.4	-66.2
	11815.596	-69.8	-79.8	-69.4
	11818.531	-66.3	-77.9	-64.4
	11819.560	-68.3	-79.2	-66.7
	11821.561	-72.0	-80.7	-69.3

3.1 Spectral type

HD108 has been classified successively as O6fqk (Plaskett & Pearce 1931), O8fp (Morgan et al. 1955), O7If (Conti & Alschuler 1971), O6:f?pe (Walborn 1972) and finally O7IIIfpe (Underhill 1994).

Applying Conti & Alschuler's (1971) criterion to our data, the spectral type varies between O4 (Aug. 1987) and O7.5 (Sep. 2000). A similar type of variation has already been reported by Beals (1950), who found for HD108 an O6 type in 1938 and O7 in 1945. A closer inspection of this situation indicates that the He II $\lambda$4542 line profile is quite constant while the spectral type variations are due to the varying shape of the He I $\lambda$4471 line which sometimes appears with a P Cygni profile (see Sects. 5 and 6), revealing that this line is not completely of photospheric origin. This casts doubt on the possibility of applying the Conti & Alschuler criterion to this star.

For the luminosity class, we use the criterion based on the value of $\log(W'') = \log{\frac{EW(\rm {{Si~{\sc{IV}}\ }} \lambda\,4088)}{EW(\rm {{He~{\sc{I}}\ }} \lambda\,4143)}}$. The equivalent widths (EWs) of these two absorption lines show no strong variations, and result in a supergiant classification, regardless of the year of observation. Even though it is possible that He I $\lambda$4143 could also be contaminated by a variable emission component, this effect should be small with regard to the stability of our result. Using polarimetric data, Fox & Hines (1998) also favored a supergiant classification. The presence of a strong He I $\lambda$10830 emission is also consistent with a supergiant classification, although this emission is also observed in Oe-type objects (Vreux & Andrillat 1979).

The presence of N III $\lambda \lambda$4634-41 and He II $\lambda$4686 in emission justifies an "f'' tag. Moreover, a "p'' tag can be added because of the presence of many emission lines, more numerous than those commonly found in Of spectra. Finally, the emission in H I Balmer lines results in the addition of an "e'' tag. Choosing the latest spectra, i.e. the least affected by possible emission in He  I, we can derive a probable O7.5Ifpe type for HD108. If the He I $\lambda$4471 absorption continues to strengthen over the next years (see Sects. 5 and 6), it is possible that we will finally recover an O8 type as found by Morgan et al. (1955).