Non-Periodic Phenomena in Variable Stars
IAU Colloquium, Budapest, 1968
SPECTRAL VARIATIONS OF P CYGNI
MART DE GROOT
Sonnenborgh Observatory, University of Utrecht, The Netherlands
ABSTRACT
From a careful study of 35 high-dispersion spectrograms of P Cygni it is
concluded that the spectroscopic data do not confirm the conclusion
of Magalashvili and Kharadze that P Cygni is a W UMa type system. It is
found that many of the absorption lines are double, the hydrogen
absorption lines even triple. This is attributed to line formation in
different shells. In the outer shell variations with a period of 114 days
lead to observed radial velocity variations between -180 and -240 km/sec.
A preliminary conclusion about the velocity field in the atmosphere of P Cygni
is drawn.
P Cygni, in the Henry Draper Catalogue classified as B1p, has been
known as a variable star from the year 1600 when it was discovered as a third
magnitude nova by the Dutch chartmaker, geographer and mathematician
Willem Janszoon Blaeu. The early history of the light variation of this star
is nearly unequaled and rather puzzling. However, since 1880 P Cygni has
been of nearly constant brightness. Some observers have reported irregular light
variations with an amplitude up to 0.2 magnitude (e.g. Nikonov 1936, 1937).
About one year ago Magalashvili and Kharadze reported some interesting two and
three color observations of P Cygni. From their observations made during
the period 1951-1960 they concluded that P Cygni is a W UMa system with a period
of 0.500656 days and with amplitudes of 0.10m and 0.08m for the primary and
secondary minimum respectively. (Magalashvili and Kharadze 1967a, b.)
When these results were first reported in the Information Bulletin
on Variable Stars (no. 210) P Cygni was put on a constant observation program
for 5 nights by Alexander and Wallerstein (1967) who reported that their
observations did not reveal any variations of the brightness of P Cygni and
thus did not confirm the observations made by Magalashvili and Kharadze.
In this paper some facts pertaining to the character of these light
variations are presented from a different point of view. We have been working
upon a collection of high-dispersion spectrograms of P Cygni, covering the
period 1942-1964. From the study of some 35 spectrograms the following
facts have been established.
l. On most of the spectrograms the lines of hydrogen, many lines of He I
and the strongest lines of Fe III show besides the nearly undisplaced emission
line two shortward displaced absorption components. In the case of the hydrogen
lines with Balmer number n >= 9 there often are even three components with
velocities of about -95, -125 and -210 km/sec (cf. Figure 1).
2. The radial velocity of the most shortward displaced component of the
hydrogen lines is not constant but shows variations which after a closer
inspection have a period of 114 days. Other lines do not show this periodicity.
Fig. 1. Profile of H10 lambda 3797 showing three absorption components.
3. There are variations in the relative intensities of different absorption
components. These variations seem to be rather irregular.
With this information let us consider again the conclusion of Magalashvili
and Kharadze about the binary nature of P Cygni. Should the fact that the
spectral lines often are double be regarded as a proof that P Cygni is a binary
The two absorption components which appear at the positions of the hydrogen
and helium are of comparable strength. This means that a companion star
should not be more than one magnitude fainter than the main star. If this
statement were true, then also other spectral lines of the companion should
be visible in the spectrum of P Cygni, this providing more double absorption
lines. This is not the case. Only the hydrogen and some of the helium lines
are double. One might think of a late B type companion with few strong spectral
lines except those of hydrogen. But then the Si II spectrum and the line of
Mg II at lambda 4481 should be more prominent than the lines actually
observed in the spectrum of P Cygni.
Furthermore, the mean velocity of approach, as derived from the two
absorption components of the hydrogen lines equals about -170 km/sec.
If the duplicity of the lines were a proof of the binary nature of P Cygni this
figure would mean either that the system as a whole has a velocity of -170
km/sec with respect to the sun, or that the W UMa binary is surrounded by
a large expanding atmosphere. The first suggestion is not acceptable because
it leaves unexplained the fact that all the emission lines lie at an average
displacement of about -15 km/sec. Also a velocity of -170 km/sec is impossible
to combine with the membership of P Cygni of the galactic cluster NGC 6871.
The second suggestion is difficult to maintain because the two components
always fall in the same limited radial-velocity intervals between -180 and
-240 km/sec and between -120 and -160 km/sec respectively, but their relative
intensities change. If these were two lines from the spectra of different
stars their radial velocities should pass through all values between say -120
and -240 km/sec.
In order to find out if the intensity ratio between the two absorption
components or their radial velocities show any correlation with the phase
of the light variations given by Magalashvili and Kharadze (1967a) the phases
of all the plates of this study were determined and in Figure 2 are shown plotted
against the intensity ratio of the two absorption components at -210 and
at -125 km/sec. The intensity ratios for the Balmer lines, H9, H10, H11
and H12 were used, since these lines are essentially free from blends and nearly
always show the two components concerned. The same phases are also shown
plotted against the radial velocities of the components of H gamma and H9 at
about -210 km/sec and of H9 at about -125 km/sec (see Figure 3).
In both figures there is much scatter. In Figure 2 this is caused by the
roughness of the visual intensity estimates that were made on the spectrograms
while measuring them for their radial velocity. In Figure 3 much of the scatter
is introduced by unresolved double or triple absorptions. No convincing evidence
appears of a change either in the intensity ratio or the radial velocity
in a period of 0.500656 day.
One must conclude that the result of Magalashvili and Kharadze, that
P Cygni is a W UMa system, though very interesting from the points of view
of stellar evolution and of explaining nova outbursts, is not supported by the
spectroscopic information.
As is indicated above the radial velocity of the most shortward displaced
component of the hydrogen lines shows variations with a 114 day period. This
results is more fully illustrated in Figure 4 which shows the radial velocities
of H beta, H gamma, H delta, H9, H10 and H11 against their phases in the 114 day
period. It is found that all these lines show very much the same variations
with corresponding phases and amplitudes. In evaluating Figure 4 one should
keep in mind that many of the points in the lower part of the diagram at small
and at large phases are from dates on which the H-lines did not show all three
components. These points then are either the results of blends between the third
and second component, or they are only the second components the third being
absent. In both cases these points give lower limits to the radial velocity
of the third component.
Not only are the phases and amplitudes of these variations about the same,
but also the mean value around which the radial velocity varies is strikingly
similar for the various lines studied. If one assumes a unique relation between
radial velocity and level in the stellar atmosphere, which in fact is a unique
relation between radial velocity with respect to the star and the distance
from the stellar surface, Figure 4 could be explained in either of two ways:
1. At some high level in the atmospheres of P Cygni there is a layer
which shows periodic velocity fluctuations. The velocity of that particular
part of the atmosphere varies with a 114 day period between -180 and
-240 km/sec.
2. The velocity field in the stellar atmosphere is fixed. The variations
are introduced by variations in the opacity of the atmosphere. Sometimes we
can only see as deep as the layer with a velocity of -240 km/sec and half
a period later we see a deeper layer with a velocity of -180 km/sec.
Fig. 2. Intensity ratio of second and third components of hydrogen lines
against phase of Magalashvili and Kharadze.
Fig. 3. Radial velocity of some hydrogen absorption lines against phase
of Magalashvili and Kharadze; open circles: second component of
H gamma; dots: third component of H9; crosses: second component of H9.
Fig. 4. Radial velocity of the most shortward displaced absorption components
of various hydrogen lines against phase in the 114-day period;
a: H beta lambda 4861;
b: H gamma lambda 4340;
c: H delta lambda 4101;
d: H9 lambda 3797;
e: H10 lambda 3797;
f: H11 lambda 3770.
Before trying to decide which of these explanations should be chosen
it is investigated whether similar variations are found in the behaviour of
other spectral lines.
This has been done for the second absorption component of H delta, H9,
H10 and H11; the results are shown in Figure 5. It is clear that the general
pattern of Figure 4 is not retained. The variations are more at random. This
means that second absorptions are formed in a layer where no radial-velocity
fluctuations or opacity variations of the stellar atmosphere occur.
Fig. 5. Radial velocity of the second absorption component of various
hydrogen lines against phase in the 114-day period;
a: H delta lambda 4101;
b: H9 lambda 3835;
c: H10 lambda 3797;
d: H11 lambda 3770.
The same results are obtained for the radial velocities of the helium lines.
From different series the best measured lines were selected and their radial
velocities plotted against the phase in the 114 day period in Figure 6.
The lines at lambda lambda 3964, 4471, 4387 and 4120 are used for this purpose.
There are no indications of variations in that part of the atmosphere where
these helium lines are formed. The second components of the lines at lambda 4387
and at lambda 4120 have radial velocities of about -180 km/sec and this value
is well below the value found in the case of the varying velocity of the third
components of the hydrogen lines. For the two other lines, lambda lambda 3964
and 4471, the second components have radial velocities of nearly -200 km/sec.
This value is about equal to the velocity minima of the third components
of the hydrogen lines. That no variations are found in the case of lambda 4471
may be due to the small number of measured second components. For lambda 3964
the mean velocity of the second component is -193 km/sec whereas the third
hydrogen absorption component with smallest radial velocity, H11, still gives
-208 km/sec. The conclusion is that even the radial velocity of lambda 3964 is
not subject to variations because this line is formed just below the layer of
the atmosphere in which the variations occur.
Fig. 6. Radial velocity of all components of some He I lines;
a: lambda 3964;
b: lambda 4471;
c: lambda 4387;
d: lambda 4120.
First components are indicated by open circles, second
components by dots, and unresolved pairs by triangles.
The influence of the emission lines upon the measured radial velocities has
been investigated also. The tendency is that a strong line fills in a larger
part of the adjacent absorption and thus will cause a larger absorption velocity
to be measured. By studying the radial velocities and the line profiles
simultaneously it is possible to separate this "emission-line effect" from the
influence upon the radial velocity of the velocity gradient of the atmosphere.
It appears that the corrections to be applied in correcting for the emission-line
effect are always smaller than 15 km/sec. The effect of the stratification
of the atmosphere which can be determined from a study of the radial velocities
of lines from ions with different ionization potentials is much larger than
the emission-line effect.
Fig. 7. Absorption minus emission radial velocity against Ionization
Potential. The size of the dots is a measure for their weight.
The results obtained by previous investigators (Struve 1935, Kharadze 1936)
about the dependence of the radial velocity upon the ionization potential
are confirmed. This dependence is found best to show up if the ionization
potential is plotted against the velocity difference absorption minus emission
instead of plotting it against the absorption velocity only (cf. Figure 7).
If we now combine all these results into one general picture of the
atmosphere of P Cygni we find the following: Material from the stellar surface
is driven away from the star. While moving outward it is accelerated up to
a maximum velocity of about 240 km/sec. Beyond that point the velocity
stays constant or may even decrease a little. The matter in the extended
atmosphere is concentrated into three spherical shells of gas each giving rise
to one of three absorption components. These shells are stationary; the
particles move outward through the shells with high velocity. In the velocity
range -80 to -140 km/sec the velocity increases not very much with the
distance to the star but at higher levels (where the velocity is between -180
and -240 km/sec) the velocity changes more rapidly (see Figure 8). If now
a varying opacity according to our previous second assumption permits one
to see deeper into the atmosphere the result is that at high velocities one
really sees into a layer with smaller velocity, while in the deeper layers one
sees about the same velocity. This explains why the velocity of the third
component is varying while the first and second components only scatter
about their mean value.
Fig. 8. Tentative picture of the outward velocity in the extended
atmosphere of P Cygni against the distance to the stellar surface.
The next step is to find out if and how it is possible to fulfil the equation
of continuity in this case. Furthermore, it is possible from spectroscopic
criteria about the relation between emission intensity and dilution factor to
give a more accurate height scale to Figure 8. From the study of absorption
equivalent widths it is then possible to deduce values for the densities of the
different shells which will complete the present provisional picture. This work
is hoped to be completed in the next few months.
I am indebted to the Mount Wilson and Palomar Observatories, to the
Dominion Astrophysical Observatory, to the Lick Observatory and to the
Haute Provence Observatory for the spectrograms which form the underlying
material for this investigation.
The stimulating remarks and comments of Prof. Anne B. Underhill I gratefully
acknowledge.
REFERENCES
Alexander, Th. and Wallerstein, G., 1967, Pub. astron. Soc. Pacific 79, 500.
Kharadze, E. K., 1936, Z. Astrophys. 11, 304.
Magalashvili, N. L. and Kharadze, E. K., 1967a, Inf. Bull. Var. Stars No. 210. (IBVS N°.210)
Magalashvili, N. L. and Kharadze, E. K., 1967b, Observatory, 87, 295.
Nikonov, V. B., 1936, Abastumansk. astrofiz. Obs. Gora Kanobili Bull. 1, 35.
Nikonov, V. B., 1937, Abastumansk. astrofiz. Obs. Gora Kanobili Bull. 2, 23.
Struve, O., 1935, Astrophys. J. 81, 66.
DISCUSSION
Fernie: I think one can dismiss the W UMa hypothesis for this star more simply.
From its position on the H-R diagram, P Cyg must have a radius of
about 100 R_sun. For a binary companion to have a period of 0.5 day,
the primary would have to have a mass of the order of 10^5 M_sun, and the
companion an orbital velocity of ~ 10^4 km/sec.
An alternative explanation might be that P Cyg is a beta CMa star. However,
this too would require large and rapid variations in radial velocity, which
your observations do not show. Also, current photometry at Toronto is
in agreement with Wallerstein's finding that there are no significant
short-period light variations in P Cyg at present.
De Groot: From the observations made by the Russian observers it seems
to me that there certainly are brightness fluctuations on P Cygni. As
you said, another possibility could be that P Cygni is a beta CMa star. It
might well be that an even more detailed spectrophotometric study
would reveal the radial-velocity variations you mentioned. Although
P Cygni lies somewhat out of the general region of occurrence of the beta
CMa stars we should have an open mind for new findings. And why not
find a hot, overluminous beta CMa star some day?
Fernie: That is true, but the beta CMa variables are confined to the low-luminosity
classes near the main-sequence. P Cygni is too bright to be a beta CMa
variable. However, I agree you that one should always have an open
mind for new findigs.
Hutchings: The preliminary velocity field presented by Dr. De Groot is in
general agreement with similar results I have obtained for this star and
three others of similar type.
Note added on May 20, 1969: For a more thorough discussion of the
light variations of P Cygni see Luud's contribution at this colloquium and M.
de Groot, 1969, Bull. Astr. Inst. Netherlands 20, 225-273.