C. A. Karl1 - R. Napiwotzki1 - U. Heber1 - S. Dreizler2,3 - D. Koester4 - I. N. Reid5
1 - Dr.-Remeis-Sternwarte, Astronomisches Institut der Universität
Erlangen-Nürnberg, Sternwartstr. 7, 96049 Bamberg, Germany
2 -
Universitäts-Sternwarte Göttingen, Geismarlandstr. 11, 37083 Göttingen,
Germany
3 -
Institut für Astronomie und Astrophysik, Sand 1, 72076 Tübingen,
Germany
4 -
Institut für Theoretische Physik und Astrophysik, Universität Kiel,
24098 Kiel, Germany
5 -
Space Telescope Science Institute, Baltimore, USA
Received 9 June 2004 / Accepted 12 November 2004
Abstract
The sharp H
NLTE line cores of hydrogen-rich (DA) white dwarfs
allow their projected rotational velocities to be determined.
High resolution optical spectra of 22 stars obtained with the Keck I telescope
are matched by synthetic spectra computed from a grid of NLTE model
atmospheres. In this paper, the third in a series on white
dwarf rotation, we concentrate preferentially on DA white dwarfs
with convective atmospheres, i.e. with
< 14 000 K. Previous
analyses found DA white dwarfs hotter than 14 000 K to be very slow
rotators and rarely show any spectroscopically detectable rotation.
For 19 of our programme stars we were able to derive projected rotational
velocities or upper limits.
No rotation could be detected for seven
stars in our sample.
However twelve stars show significant line broadening.
In the case of the ZZ Ceti star G 117-B15A, the observed H
line
profile cannot be matched by a rotationally broadened profile as
its line core is too narrow.
Combining our results with those from two similar studies,
we have obtained information on the rotation or other line broadening
mechanisms
(such as caused by magnetic fields) of 56 DA white dwarfs.
The fraction of rotating DA white dwarfs whose line profiles can be matched
for a vanishing projected rotation velocity is high for hot white dwarfs
with radiative atmospheres (25 out of 28).
Amongst the cool white dwarfs with presumedly convective
atmospheres, only for a few stars (8 out of 22) has no additional
line broadening
to be invoked to explain their observed H
line profiles.
We conjecture that the physics of H
line formation in convective
DA white dwarf atmospheres is not yet sufficiently well understood and
additional observational and theoretical efforts have to be made.
Key words: stars: white dwarfs - stars: variables: general - stars: rotation - stars: convection - stars: atmospheres
White dwarfs are the compact remnants of low and intermediate mass stars
(
). Due to the conservation of angular momentum white dwarfs
should be very fast rotators, if a significant fraction of the angular
momentum of the progenitor stars were preserved. Although the spectral lines
of white dwarfs are generally very broad, the existence of sharp NLTE cores
of the hydrogen H
line in DA white dwarfs allows one to
determine (projected) rotational velocities
for these stars.
Several groups (Greenstein et al. 1977; Pilachowski & Milkey 1984, 1987; Koester & Herrero 1988; Heber et al. 1997; Koester et al. 1998) applied this method to measure rotational velocities of 53 white dwarfs. A compilation of the existing sample is provided by Koester et al. (1998).
Table 1: Atmospheric parameters, masses, radii and projected rotational velocities of the programme stars. Line profiles that cannot be matched at all by rotational profiles are marked as peculiar (see text).
For most white dwarfs, only upper limits on the rotational
velocities of the order of 20-40 km s-1 are available. Thus the general
conclusion is that white dwarfs are slow rotators.
This result is supported by astereoseismological
investigations of pulsating white dwarfs. Rotation periods of
the very hot white dwarfs PG 1159-035
(Winget et al. 1991) and
PG 0122+200 (O'Brien
et al. 1996), the DB star GD 358
(Winget et al. 1994)
and ZZ Ceti stars (e.g. Kleinman et al. 1998)
have been determined from the rotational splitting of
oscillation modes. The determined periods of the stars range from 1.17 to
1.6 days, corresponding to rotational velocities of 1 km s-1.
Polarization measurements of magnetic white dwarfs also indicate that
most of these objects are very slow rotators (see Table 1 in Wickramasinghe
& Ferrario 2000). However, magnetic white dwarfs may not be
representative of the entire class of white dwarfs because
the evolution of angular momentum could have been influenced by the strong
magnetic fields.
These low rotational velocities indicate that the angular momentum of the
stellar cores is efficiently transported to the stellar surface and carried
away by the mass lost during the course of stellar evolution.
However, the emerging
picture is challenged by the puzzling results presented in the recent
investigation of Koester et al. (1998, henceforth Paper II).
The authors analyzed the
H
NLTE cores of 25 DA white dwarfs spanning the temperature range
7000 K to 30 000 K. Most white dwarf spectra were compatible with zero
rotational velocity. However, all three ZZ Ceti variables in that sample
showed significant extra broadening of the observed line cores.
Rotational broadening profiles, however, did not match the observations
well.
Therefore Koester et al. (1998) argue that the observed broadening is not due to rotation and discussed velocity fields caused by pulsation and convection but find them to be far too small to explain the observed spectra. Recently it became possible to measure radial velocity variations of ZZ Ceti stars (Van Kerkwijk et al. 2000; Kotak et al. 2002b; Thompson et al. 2003), which confirmed that the amplitudes indeed are as low as predicted. Even if the broadening were caused by rotation it would be difficult to understand how white dwarfs could spin-up when entering the instability strip during their evolution - and spin-down again after leaving it.
It is obvious that more data are needed to improve the statistics near the ZZ Ceti strip and to pin down the temperature range in which this phenomenon occurs. This will also help to identify the underlying mechanism. In this article we present the results of the analysis of the NLTE cores of 22 DA white dwarfs covering the temperature range from 6800 K to 20 000 K. This is an extension of our previous investigation (Heber et al. 1997, henceforth Paper I) to lower temperatures.
The analysis is based on the spectra obtained by Reid (1996)
with the HIRES spectrograph of the Keck I telescope. Reid (1996) used
these spectra to measure the gravitational redshift of white dwarfs in
common proper motion pairs and open clusters.
In our previous analysis (Paper I) we
modeled the line profiles of 13 hotter stars (
K)
of this sample and determined upper limits on
.
No significant
rotation was detected in any case. The analysis of the remaining 22 useful
spectra of this sample is presented in this paper.
This paper is organized as follows.
In Sect. 2 we summarize the available atmospheric parameters and derive
stellar masses. In Sect. 3 the spectral analysis of H
profiles is
described and the resulting projected rotational velocities are presented
in Sect. 4.
The latter are discussed in
the concluding Sect. 5.
The atmospheric parameters
and
need to be known in
order
to compute synthetic H
line profiles.
Atmospheric parameters (
,
)
determined spectroscopically were preferred when available.
Values for 12 stars were taken from Bergeron et al. (1995a) and
one from Claver et al. (1998).
These authors used a mass-radius relation and the gravity to determine
radii and masses. We adopted the masses as given in the cited papers.
For five of those (mostly cooler) stars, atmospheric parameters have also been obtained from photometry and trigonometric parallaxes (Bergeron et al. 2001). These values were used for intercomparison (see below).
For all other programme stars we had to rely on photometric temperature estimates. Effective temperatures of two stars were derived from multicolour photometry by Weidemann & Koester (1984). For the remaining stars we have either determined the effective temperature from photometry listed in the catalog of McCook & Sion (1998) using the colour calibrations of Bergeron et al. (1995b) or we adopted the temperatures derived by Reid (1996).
Gravitational redshifts were combined with the mass-radius-relation to derive masses and radii (Reid 1996). We then calculated the surface gravities from the latter.
The error budget was calculated as described in Napiwotzki et al. (1999) for the surface gravities of all stars and for most of the temperatures. Errors for photometric temperatures were estimated from the errors of the colors.
The temperatures range from 6800 to 20 000 K and
from 7.70 to 8.50.
Only four stars (WD 0220+222, WD 0810+234, WD 0836+201
and WD 1211+392) have effective temperatures exceeding 14 000 K.
Their atmospheres are therefore radiative.
The masses lie between 0.43 and 0.92
.
The parameters of the observed stars used for the subsequent analysis are summarized in Table 1.
The low gravity derived by Bergeron et al. (2001) is not covered by our model grid. Therefore some extrapolation was inevitable. Despite the large differences in atmospheric parameters the two resulting limits on the projected rotation velocity are quite similar (see Table 1).
G 117-B15A = WD 0921+354 is a known ZZ Ceti which has been studied in great detail (see Kepler et al. 2000, and references therein).
LP 375-51 = WD 1147+255 and G 216-B14B = WD 2258+406 lie near the red edge of the ZZ Ceti instability strip but are non-variable (Giovannini et al. 1998). The atmospheric parameters derived from spectroscopy (Bergeron et al. 1995a) are consistent with those derived from photometry and trigonometric parallax (Bergeron et al. 2001).
LP 135-154 = WD 1510+566 and G 172-4 = WD 0030+444:
and
were derived by Weidemann & Koester (1984)
from analysis of multichannel photometry.
Wolf 672 A = WD 1716+020 lies near the blue edge of the ZZ Ceti instability strip but is non-variable (Giovannini et al. 1998). The effective temperature derived from spectroscopy (Bergeron et al. 1995a) is consistent with that derived from photometry and trigonometric parallax (Bergeron et al. 2001), whereas the spectroscopic gravity is somewhat higher than derived from the trigonometric parallax.
G 154-B5B = WD 1743-132 The projected rotation velocity has also been determined in Paper II.
G 156 - 64 = WD 2253-081 The atmospheric parameters derived from spectroscopy (Bergeron et al. 1995a) are consistent with those derived from photometry and trigonometric parallax (Bergeron et al. 2001).
LP 591-117 = WD 0251-008
We used the photometric data found in the Mc Cook & Sion catalogue
(1999) and the color calibrations from Bergeron et al. (1995) to derive the
effective temperature. However, the temperatures calculated from
the Broadband B-V and U-B (15 200 K and 160 000 K resp.) differs
from the temperature derived using Strömgren b-y (11 900 K).
The mean temperature of 14 400 K cannot reproduce the H
line profile.
Reid (1995) estimated a lower effective temperature of
= 9000 K
while Silvestri et al. (2001) found 6453 K.
In order to clarify these large differences we fitted the H
line wings
in our Keck spectrum (neglecting the line core) using the model grid described in
Sect. 3.1.
From the best fit we obtained 7700 K (see Fig. 1, which is
close the mean value of Reid and Silvestri et al.
To derive
we used this temperature for the further analysis of the H
line core and adopted
= 8.05 as computed from Reid (1995).
![]() |
Figure 1:
H![]() ![]() |
Open with DEXTER |
In Paper I we used self consistent,
line blanketed NLTE model atmospheres which, however, neglected convection.
Therefore we had to restrict our analysis to white dwarfs hotter than
14 000 K. The atmospheres are radiative for
16 000 K.
Below this
and depending on the choice of gravity the atmospheres become convective,
however the convection is inefficient and has almost no influence on the
atmospheric structure as long as
12 500 K.
In Paper II LTE models were used that account
for convection (mixing length theory using ML2/ = 0.6, see Paper II
for details) to determine the temperature - pressure
stratification of the atmosphere.
With this stratification kept fixed, the line formation was computed in
NLTE, solving the rate equations for the occupation number at each depth.
The dominant continuum absorption mechanism in cool DA is
absorption which
is also not included in the NLTE calculations.
Therefore in Paper II H- opacity was included
as a background absorption and calculated in the LTE approximation.
The model grid used for our analysis is the same as used in Paper II. It
covers the parameter range
= 7000 K to 25 000 K in steps of
1000 K and
= 7.5, 8.0 and 8.5.
The technique for the analysis of the H
line profiles is described
in Paper I. Briefly,
the synthetic
line profiles are interpolated from a
grid of model spectra (convolved with the instrument profile, a Gaussian
with 0.14 Å FWHM).
Only the innermost part (about
20 Å) of the line profile was used for comparison with
observations. Wavelength scales were adjusted by cross correlating the observed
and the synthetic spectra.
The rotational profile (in the notation of Gray 1992)
was used for the line broadening of the synthetic spectra and a limb
darkening coefficient of
following Pilachowski et al.
(1984) was adopted.
Since no continuum point can be found close to the line cores, the
observation and the synthetic profiles were adjusted at some preselected
wavelength points outside of the spectral range used in the fitting process.
For each star a grid of synthetic spectra was produced by varying
in steps of 1 km s-1. Four additional sets of synthetic spectra
per star were obtained by varying the effective temperature
and surface gravity within their error limits.
These sets were matched to the observed spectra by calculating the
deviations using a modified version of Bergeron's and Saffer's
program (Saffer et al. 1994).
Subtracting the minimum
we obtain a relation of
with
for each combination of
and
(see Figs. 2 and 3).
and
confidence level are read from these plots at
and
,
respectively.
Previous analyses have used either a
(e.g.
Paper II) or a
criterion (Paper I).
Therefore we give both values for our programme stars in Table 1.
For the purpose of comparison we have recalculated the 1
limits
for the stars of Paper I and list them in Table 2.
Projected rotational velocities for 19 of our 22 programme stars have been
determined.
For three stars (WD 0628-020, WD 0913+442 and WD 2047+809)
no reasonable fit could be obtained (cf. Sect. 4.1).
Resultes are listed in Table 1 were errors at the 1
and 3
level are given.
For seven stars of our sample the best fit was achieved for vanishing rotation
(
within the 1
error limits).
The quality of the fits are demonstrated for four non-rotating stars
in Fig. 2; spectra of different noise levels are displayed.
![]() |
Figure 2:
H![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 3:
H![]() ![]() ![]() ![]() |
Open with DEXTER |
Another twelve stars showed extra broadening.
As can be seen from
Fig. 3, line profiles broadened by rotational profiles
with
between 20 km s-1 and 72 km s-1 match the
observations well.
Strong magnetic fileds can broaden lines
due to Zeeman splitting as well.
Koester et al. (1998) discovered
magnetically broadened line profiles
for some of their objects.
However,
no hints of Zeeman triplets can be
found among our spectra, not even
"flat bottom'' line profiles
(cf. Figs. 7 to 9 in Paper II).
Moreover, from the most extreme
line broadening of 0.35 Å (WD 1710+683)
we deduce an upper limit for the magnetic
fields of 17.5 kG only.
Thus we exclude any significant broadening
due to magnetic fields.
Table 2:
Upper limits (
and
)
for the projected
rotational
velocities of the stars from Paper I.
limits have not been
published previously.
WD 0921+354:
the fit for this ZZ Ceti variable, however,
is peculiar. The formal best fit is obtained for
= 36 km s-1 (upper panel of Fig. 4).
However, inspection of the line profile
fit
reveals that the observed line core is
narrower than predicted by the synthetic line profile.
This is demonstrated in
Fig. 5, where we compare the line profile fit for WD 0921+354
to those of the non-rotating star WD 1716+020 and the rotating star WD 1706+333.
We have to conclude that the line broadening of the ZZ Ceti star
WD 0921+354 cannot be due
to rotation, but must be caused by an unknown physical mechanism.
The same conclusion was reached in Paper II for three ZZ Ceti stars.
WD 0518+333 is a special case among our program stars.
Two parameter sets had to be used.
There is no evidence for rotation at the 3
level, whereas if we
apply the 1
criterion and use the parameters derived by Bergeron
et al. (2001) the star would be marginally rotating.
WD 1240+754 and WD 2258+406: rotation of these stars is
detected at a 1
level (
= 30 km s-1 and
20 km s-1 resp.) but not at the 3
level for which only rather high
upper limits of 62 km s-1 and 42 km s-1 can be quoted.
![]() |
Figure 4:
H![]() ![]() ![]() ![]() |
Open with DEXTER |
WD 0628-020, WD 0913+442 and WD 2047+809 show broadened lines, too. However, in these cases the broadening is very different from a rotational profile as can be seen from Fig. 6 in the case of WD 0628-020. While a sharp line core can be reproduced by a synthetic line profile and vanishing rotation, the near line wings are far too broad. The formal best fit gives a very poor fit to the observation.
To improve the fit quality we changed the
stellar parameters within larger error limits than indicated in
Table 1.
We found that
hardly influences the shape of the line profile,
while a decrease of
by
1000 K will reasonably
improve the fit of the core and the near line wings.
In this region the shape of the observed line profile is
now very similar to the synthetic profile, whereas the outer line wings
are not well matched, as can be seen from Fig. 7.
![]() |
Figure 5:
H![]() |
Open with DEXTER |
Since the innermost line cores of these "peculiar'' objects show neither the typical Zeeman splitting nor "flat bottom'' profiles, we exclude magnetic fields as the reason for this puzzling effect. An anomalous limb darkening can also be excluded since the rotational profile is insensitive to the variation of this coefficient (see Pilachowski et al. 1984). Thus we have to conclude that their line profiles cannot be reproduced by any combination of the stellar parameters. However, we cannot rule out that these mismatches are artefacts caused by inadequate data reduction.
Obviously these three white dwarfs need to be reobserved before any firm conclusion can be drawn. Therefore we shall exclude these stars from further discussion.
![]() |
Figure 6:
H![]() ![]() ![]() ![]() ![]() |
Open with DEXTER |
![]() |
Figure 7:
H![]() ![]() |
Open with DEXTER |
Table 3 summarizes the results from the present study (19 stars) as well as those of Papers I (13 stars) and II (25 stars). However, WD 2329-291, analyzed in Paper II, was shown recently by Lisker et al. (2004) to be a sdB star. Therefore we excluded it from further discussion of the remaining 56 white dwarf stars.
We divide the complete sample into a hot one
with radiative
atmospheres and a cool one
with presumedly
convective atmospheres and divide the sample with a borderline
at
K.
There are, however, six stars with
close to
the borderline, i.e. in the transition zone from radiative to convective
atmospheres. Therefore we consider a group of hot white dwarfs consisting of
all programme stars with
K (WD 1919+145) or higher and a
group of cool white dwarfs with
K (WD 0921+354) or less.
For all but three of our 28 stars with
K
there is no hint of rotation,
with upper limits ranging from 1 km s-1 to 24 km s-1.
However, the line broadening is more likely to be due to magnetic
fields in WD 1531-022 (see Paper II). Therefore
only two rotating object (WD 0836+201 and WD 2039-682)
are found in the hotter temperature range,
which are amongst the fastest rotating DA stars
known (80 km s-1 and 66 km s-1, respectively).
For lower temperatures we find no hint of extra broadening only in 8 out of 22 stars with upper limits to the projected rotational velocites ranging from 7 km s-1 to 35 km s-1.
For the remaing stars, extra broadening has to be accounted for. Again the broadening is probably caused by magnetic fields in two cases (WD 1953-011 and WD 2105-820, see Paper II).
ZZ Ceti stars For the four ZZ Ceti variables the line profiles are not well matched by rotational profiles (see Paper II and Fig. 4). Unlike the three ZZ Ceti stars studied in Paper II, WD 0921+354 displays a very narrow line core.
Non-variables There are
the remaining eight stars whose H line profiles can reasonably well be
matched by rotationally broadened synthetic line profiles with projected
rotational velocities ranging from 20 km s-1 to 72 km s-1.
Table 3:
Summary of projected rotational velocities (1
detection limit)
and stellar parameters of all stars from this paper and Papers I and II.
No hint of extra broadening can be found for four of the six stars near the
borderline
K with upper limits to the projected rotational
velocity of 7 km s-1to 35 km s-1. Two stars can be matched by assuming rather
slow rotation,
= 23 km s-1and 28 km s-1.
In the latter case an even lower projected
rotational velocity of 13 km s-1was derived in Paper II,
but no error could be estimated.
Comparing the results for the cool DAs with supposedly convective atmospheres
to that for the hot DAs with supposedly radiative atmospheres, we find a
striking difference. While there is no indication of rotation or any other
extra broadening mechanism in the vast majority of the hot stars, we have
to call upon extra broadening for the majority of the cool stars, as
only in three
cases are magnetic fields the likely cause. For the four ZZ Ceti stars
rotation can be excluded as the broadening mechanism because that would be
inconsistent with the analysis of their pulsation patterns. The H
lines of the other stars can be reasonably well matched by rotational profiles.
Three stars are excluded from
this discussion because their H
line profiles are very peculiar.
These objects are amongst the coolest stars in our sample.
![]() |
Figure 8:
![]() |
Open with DEXTER |
We analyzed the H
line profiles of 22 hydrogen-rich white
dwarfs (DA).
Synthetic spectra were calculated from LTE model atmospheres by
treating the statistical equilibrium of hydrogen in detail in NLTE in order
to model the sharp H
line core reliably. The
synthetic spectra were broadened by rotational profiles and a
minimization method was used to determine projected
rotational velocities for 19 of our 22 programme stars from high resolution
spectra obtained with the HIRES spectrograph at the Keck I telescope.
We combined our measurements with data from previous analyses (Heber et al.
1997; Koester et al. 1998) which were based on similar
spectra and model atmospheres to derive a homogenous set of
i
measurements of 56 DA white dwarfs.
This sample contains four ZZ Ceti variables and four probably magnetic white
dwarfs.
In the latter cases the widening of the line core is caused
by Zeemann splitting.
As already noted in Paper II the results for hot stars with supposedly radiative atmospheres differ significantly from those for the cooler stars with supposedly convective atmospheres. Our sample contains 28 hot and 22 cool DA stars. Excluding the magnetic white dwarf from the hot sample, it was possible to match the observed profiles with the synthetic spectra well for the remaining 27 stars. Only in two cases was significant rotation detected. For all other hot stars the spectral analysis did not reveal any rotation.
For more than half of the cooler DA stars, however, the H profile
did show evidence of additional broadening.
The spectra of ten cool stars were consistently matched by line profiles
with projected rotation velocities between 23 km s-1 and 72 km s-1.
Close inspection of the line
profiles revealed significant deviations from the rotational profile for
the four ZZ Ceti stars.
All four ZZ ceti stars in our sample show evidence of peculiar line
broadening. In principle, the rotation velocity of a ZZ Ceti star can be
determined independently by measuring the
rotation-induced splitting of the pulsation modes. As noted already in
Paper II the
low rotation periods (1 day) derived from asteroseismology
are at variance with the spectroscopic results. Hence the line broadening
cannot be caused by rotation and other broadening mechanisms have been
invoked. Line-of-sight variations have indeed be discovered
in the ZZ Ceti stars G 29-38 (up to 5 km s-1, Van Kerkwijk et al. 2000)
and HS 0507+0434B (of 2.6 km s-1, Kotak et al. 2002b) and were attributed
to horizontal surface motions associated with g mode pulsation.
The measured amplitude, however, is too small to account for the line
broadening observed in the four ZZ Ceti stars in our sample (36 km s-1 to
45 km s-1). Recently, Thompson et al. (2003) reobserved G 29-38
at high spectral resolution and found radial velocity changes as
large as 16.5 km s-1, but concluded
that neither rotation nor pulsation can explain the truncated shape
observed for the ZZ Ceti star's H
line core.
Therefore the physical mechanism still remains unclear.
In view of the ZZ Ceti results, it may be premature to conclude that
the extra broadening that we see in the non-variable cool DAs is due
to rotation.
The average rotation velocities in a sample of hot white
dwarfs should be the same as in a sample of cool
white dwarfs.
Moreover, one would expect similar broadening
for all stars within a given parameter range.
It is very puzzling, for example, that WD 0807+484 does not show
any sign of extra broadening, while WD 1710+683 - having
almost the same
and
- does.
Thus we conjecture that the physics of H
line formation in
convective
white dwarf atmospheres is not yet sufficiently well understood.
Additional parameters may have to be taken into account.
For example, the helium abundance in a convectively mixed
atmosphere might be related to the mass of the hydrogen
envelope and thus could be different in two stars having
otherwise identical parameters.
New high resolution observations of excellent S/N for a large sample of both
hot and cool
white dwarfs are required to gain insight into the rotational properties of
white dwarf stars and possible other broadening
mechanisms.
Recently the ESO Supernova Ia Progenitor Survey (SPY) provided high resolution
spectra of more than 1000 white dwarfs (Napiwotzki et al. 2003).
Many of these are of sufficiently high
S/N to allow precise measurements of the H
lines of
hundreds of white dwarfs.
Acknowledgements
C.K. gratefully acknowledges financial support by the Deutsche Forschungsgemeinschaft (grant Na 365-x/2-1). The authors thank the referee for constructive comments and suggestions.