short-period line profile variations in the be star μ centauri

13
Mon. Not. R. Astron. Soc. 324, 1041–1053 (2001) Short-period line profile variations in the Be star m Centauri L. A. Balona, 1w D. J. James, 2 W. A. Lawson 3 and R. R. Shobbrook 4 1 South African Astronomical Observatory, PO Box 9, Observatory 7935, Cape Town, South Africa 2 School of Physics and Astronomy, University of St Andrews, St Andrews, Fife 3 School of Physics, University College UNSW, Australian Defence Force Academy, Canberra, ACT 2600, Australia 4 Research School of Astronomy and Astrophysics, Australian National University, Canberra, ACT 0200, Australia Accepted 2001 January 30. Received 2001 January 29; in original form 2000 July 24 ABSTRACT We present new intensive photometric observations of the Be star m Cen for several seasons which support a period close to 1 d. We also present high-resolution spectroscopic data consisting of 302 spectra in 1999 and 864 spectra in 2000, all obtained within a two-week observing run in each season. We use stacked grey-scale plots of spectra, from which the mean line profile has been removed, to examine the profile variations. We find that most nights show one residual absorption feature, moving from blue to red, visible in all helium and metal lines and also clearly visible in Ha and other lines formed in the circumstellar environment. We therefore conclude that this feature is of circumstellar origin. In addition, a residual absorption feature moving from red to blue is sometimes seen at irregular intervals. We find that the residual absorption feature frequently strays outside the projected rotational velocity limit and that on occasions it remains well within this limit. The radial velocity data reproduce only two of the six frequencies previously found in the star. We point out that this by no means implies that the star is a multiperiodic, non-radial pulsator. Photometric data obtained over several seasons indicate a period very close to 1 d and not the 0.5-d period found from the radial velocities. We describe an outburst which occurred during the run and which resulted in increased Ha emission two nights later. It is clear that outbursts in Be stars are localized events and that the gas released by outbursts is probably responsible for localized increased absorption, resulting in periodic light and line profile variations. Key words: line: profiles – stars: early-type – stars: emission-line, Be – stars: individual: m Cen. 1 INTRODUCTION Many Be stars are variable with periods of about one day – about the same as the expected periods of rotation in these stars. Indeed, Balona (1990, 1995) found a correlation between the period and the projected rotational velocity, from which it can be concluded on statistical grounds that the observed period cannot differ from the period of rotation by more than a few per cent. For those stars that have been observed both photometrically and spectro- scopically, the period of the light variation is the same as the period of the line profile variation. Nevertheless, multiperiodic behaviour of the radial velocities has been reported from time to time, suggesting that non-radial pulsation (NRP), rather than rotational modulation, may be responsible for the variations. Some observers have also reported multiperiodicity in the light curves of Be stars. The criterion for accepting that a star is multiperiodic is to show that the same periods are present in two or more independent data sets. These analyses are nearly always based on only a few nights of data and cannot be said to demonstrate multiperiodicity. Longer and therefore more reliable data sets, such as those of Cuypers, Balona & Marang (1989), have consistently failed to confirm multiperiodicity. However, two periods have been found in the light curve of 28 CMa (S ˇ tefl, Aerts & Balona 1999). About half of the periodic Be stars have a double-wave light curve (two unequal maxima or minima). A feature in a single-wave light curve can deepen and become a double-wave light curve and vice versa (Balona, Sterken & Manfroid 1991). While NRP has been the favoured explanation for a long time, recent observations of h Cen (Balona 1999), z Tau (Balona & Kaye 1999) and other stars suggest that the periodic variability may be due to gas trapped into corotation close to the star, perhaps by a weak magnetic field. It is difficult to distinguish between the two hypotheses, as both can lead to similar line profile and light variations. However, in h Cen the repetitive pattern is too complex to be understood as NRP. In z Tau it is clear that the periodic variations do not arise in the photosphere but are associated with the circumstellar material. q 2001 RAS w E-mail: [email protected]

Upload: l-a-balona

Post on 06-Jul-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Short-period line profile variations in the Be star μ Centauri

Mon. Not. R. Astron. Soc. 324, 1041±1053 (2001)

Short-period line profile variations in the Be star m Centauri

L. A. Balona,1w D. J. James,2 W. A. Lawson3 and R. R. Shobbrook4

1South African Astronomical Observatory, PO Box 9, Observatory 7935, Cape Town, South Africa2School of Physics and Astronomy, University of St Andrews, St Andrews, Fife3School of Physics, University College UNSW, Australian Defence Force Academy, Canberra, ACT 2600, Australia4Research School of Astronomy and Astrophysics, Australian National University, Canberra, ACT 0200, Australia

Accepted 2001 January 30. Received 2001 January 29; in original form 2000 July 24

A B S T R A C T

We present new intensive photometric observations of the Be star m Cen for several seasons

which support a period close to 1 d. We also present high-resolution spectroscopic data

consisting of 302 spectra in 1999 and 864 spectra in 2000, all obtained within a two-week

observing run in each season. We use stacked grey-scale plots of spectra, from which the

mean line profile has been removed, to examine the profile variations. We find that most

nights show one residual absorption feature, moving from blue to red, visible in all helium

and metal lines and also clearly visible in Ha and other lines formed in the circumstellar

environment. We therefore conclude that this feature is of circumstellar origin. In addition, a

residual absorption feature moving from red to blue is sometimes seen at irregular intervals.

We find that the residual absorption feature frequently strays outside the projected rotational

velocity limit and that on occasions it remains well within this limit. The radial velocity data

reproduce only two of the six frequencies previously found in the star. We point out that this

by no means implies that the star is a multiperiodic, non-radial pulsator. Photometric data

obtained over several seasons indicate a period very close to 1 d and not the 0.5-d period

found from the radial velocities. We describe an outburst which occurred during the run and

which resulted in increased Ha emission two nights later. It is clear that outbursts in Be stars

are localized events and that the gas released by outbursts is probably responsible for

localized increased absorption, resulting in periodic light and line profile variations.

Key words: line: profiles ± stars: early-type ± stars: emission-line, Be ± stars: individual:

m Cen.

1 I N T R O D U C T I O N

Many Be stars are variable with periods of about one day ± about

the same as the expected periods of rotation in these stars. Indeed,

Balona (1990, 1995) found a correlation between the period and

the projected rotational velocity, from which it can be concluded

on statistical grounds that the observed period cannot differ from

the period of rotation by more than a few per cent. For those stars

that have been observed both photometrically and spectro-

scopically, the period of the light variation is the same as the

period of the line profile variation. Nevertheless, multiperiodic

behaviour of the radial velocities has been reported from time to

time, suggesting that non-radial pulsation (NRP), rather than

rotational modulation, may be responsible for the variations.

Some observers have also reported multiperiodicity in the light

curves of Be stars. The criterion for accepting that a star is

multiperiodic is to show that the same periods are present in two

or more independent data sets. These analyses are nearly always

based on only a few nights of data and cannot be said to

demonstrate multiperiodicity. Longer and therefore more reliable

data sets, such as those of Cuypers, Balona & Marang (1989),

have consistently failed to confirm multiperiodicity. However, two

periods have been found in the light curve of 28 CMa (SÏtefl, Aerts

& Balona 1999). About half of the periodic Be stars have a

double-wave light curve (two unequal maxima or minima). A

feature in a single-wave light curve can deepen and become

a double-wave light curve and vice versa (Balona, Sterken &

Manfroid 1991).

While NRP has been the favoured explanation for a long time,

recent observations of h Cen (Balona 1999), z Tau (Balona &

Kaye 1999) and other stars suggest that the periodic variability

may be due to gas trapped into corotation close to the star, perhaps

by a weak magnetic field. It is difficult to distinguish between the

two hypotheses, as both can lead to similar line profile and light

variations. However, in h Cen the repetitive pattern is too complex

to be understood as NRP. In z Tau it is clear that the periodic

variations do not arise in the photosphere but are associated with

the circumstellar material.

q 2001 RAS

w E-mail: [email protected]

Page 2: Short-period line profile variations in the Be star μ Centauri

The strongest case presented for NRP in these stars is to be

found in m Centauri (HR 5193, HD 120321; B2IV±Ve). Rivinius

et al. (1998a) found that the radial velocity (as measured by the

position of the absorption cores) is multiperiodic. There are four

periods, all within 2 per cent of P � 0:505 d: The other two

periods are near 0.28 d. Moreover, it has always been supposed

that sudden changes in the emission-line strengths of Be stars, i.e.

outbursts, occur at random times. Rivinius et al. (1998b) found

that the times of maximum amplitude, calculated from the beating

of the periods, are correlated with the times of outburst. They

suggest NRP may play an important role in generating sudden

mass-loss episodes (outbursts) in m Cen.

While the detailed analysis of the line profile variations in

m Cen by Rivinius et al. (1998a) appears to leave little room for

doubt regarding multiperiodicity, it seems to be unique among the

Be stars. One of the characteristics of the Be stars is that they vary

randomly on all time-scales. The periodic variations in the light

curves are characterized by large changes in shape and amplitude.

The presence of several stable periods in m Cen over many years

is therefore surprising. A period of almost exactly 0.5 d also

poses some important questions. The data were obtained

exclusively from one site (ESO in Chile). It is evident that any

period close to half a day will mean that almost the same phase is

sampled at the same time each night. The analysis of Rivinius et al.

(1998a) is based on observations where, for the most part, only

one spectrum per night was obtained. Furthermore, there is no

evidence from Cuypers et al. (1989) for a 0.5-d period in the

multi-site photometry of the star, instead there is a periodicity

close to 1 d, or possibly a double-wave period of nearly 2 d. If the

NRP interpretation is correct, the amplitude at 0.5 d must be too

small to be detected in the light curve and a separate explanation is

required for the photometric 1-d period.

On the basis of these results, and bearing in mind the

importance of the star in the NRP/rotational modulation debate,

we decided to observe the star ourselves. We obtained over 300

echelle spectra of m Cen during a continuous two-week observing

run on the SAAO 1.9-m reflector in 1999 March. In addition,

StroÈmgren photometry was obtained over the same period using

the 0.5-m reflector at SAAO. We again observed the star in 2000

May, this time obtaining over 850 echelle spectra over a two-week

period at SAAO. These observations span almost a full 0.5-d

period on each night, so it is unlikely that aliasing will be a serious

problem if the period is indeed close to 0.5 d.

In this paper we present and discuss these new data and include

an analysis of previously unpublished photometry from the 1989,

1990 and 1991 seasons. Due to the intensive time coverage of our

spectroscopic observations, we are able to show that the behaviour

of the moving absorption features is more complex than

previously supposed.

2 O B S E RVAT I O N S

Observations at SAAO were obtained using the GIRAFFE echelle

fibre-fed spectrograph attached to the Cassegrain focus of the 1.9-m

telescope. The GIRAFFE spectrograph has a resolving power of

about 32 000. The 1024 � 1024 TEK CCD chip gives a resolving

power of 0:06±0:09 �A per pixel. A Th±Ar arc lamp was used for

wavelength calibration with arc spectra taken at regular intervals

to calibrate possible drifts. Flat-fielding was accomplished by

illuminating the camera with uniform light using a tungsten

filament lamp and a diffusing screen. The blaze correction was

determined by measuring the response across each order when the

fibre was illuminated by a tungsten lamp.

Observations were made during the period 1999 March 23±

April 6 and again during 2000 May 9±21. For the 1999 season, the

wavelength range was set to be 3960±5880 �A spread over 46

orders. Exposure times were normally 15 min for a S/N ratio of

about 100±200 per pixel in the middle of the wavelength range. A

total of 302 spectra of m Cen was obtained (see Table 1). For the

2000 season, the wavelength coverage was 4220±6710 �A spread

over 49 orders. The fibre was aligned more accurately during this

time, which resulted in all the spectra having a S/N ratio in excess

of 200 per pixel, and frequently approaching or exceeding 300 per

pixel, with exposure times between 5 and 7 min. A total of 864

spectra was obtained (Table 1).

In order to study the line profile variations, the continuum needs

to be determined. Placing the continuum by hand is not an easy

process. We preferred to rectify the spectra with the aid of a

synthetic spectrum calculated using the spectrum code (Gray &

Corbally 1994). To do this, we calculated a synthetic spectrum

covering the full echelle wavelength range using a Kurucz solar-

abundance model atmosphere, appropriate to a B2IV star and

broadened with v sin i � 130 km s21: To rectify the spectrum, each

echelle order was divided by the synthetic spectrum and the result,

Table 1. Observing log for the GIRAFFE echellespectra obtained at SAAO for the 1999 and 2000seasons. The Julian day, with respect to JD245 1000, is shown for the first and last spectrumof m Cen on the given night. N is the number ofspectra of m Cen obtained on that night.

1999 2000Start End N Start End N

261.38 261.46 7 674.22 674.67 80262.35 262.66 26 675.20 675.66 95263.35 263.68 27 676.19 676.31 16264.34 264.66 29 677.19 677.65 71265.34 265.65 26 678.19 678.65 76267.60 267.62 2 679.23 679.65 76268.51 268.57 6 680.20 680.36 28269.29 269.67 31 681.21 681.65 73270.33 270.63 25 682.24 682.64 73271.32 271.59 24 683.19 683.65 68272.30 272.67 33 684.19 684.65 76273.31 273.67 32 685.19 685.64 72274.28 274.67 34 686.24 686.62 60

Table 2. Log of photometry of m Cen obtained atSAAO. Except for the 1999 season where the uvbyobservations were obtained, all these data wereobtained through the StroÈmgren b filter only. Thetable shows the start and end of the run relative to JD244 0000; N is the number of observations. The lasttwo columns contain results of a period analysis. Theperiod and amplitude correspond to that of the highestpeak found in the periodogram.

Season Start End N Period Amp

1987 6959 6983 147 1.05 0.0231988 7236 7275 187 1.04 0.0071989 7621 7649 128 0.98 0.0121990A 7929 7957 70 1.03 0.0171990B 7985 8000 77 0.97 0.0071991 8335 8348 79 0.69 0.0071999 11262 11275 112 0.79 0.006

1042 L. A. Balona et al.

q 2001 RAS, MNRAS 324, 1041±1053

Page 3: Short-period line profile variations in the Be star μ Centauri

after some manipulation, fitted with a third-order polynomial.

Direct comparison of the rectified spectra with the synthetic

spectrum showed that the above procedure is quick and effective.

Photometry was obtained using the Modular Photometer

attached to the 0.5-m reflector of SAAO, Sutherland, during

1989, 1990, 1991 and simultaneously with the echelle data of

1999. For the 1999 season, we obtained photometry through all

four StroÈmgren uvby filters. For the other seasons, the star was

observed only through the StroÈmgren b filter. In all cases HR 5063

and HR 5439 were used as local comparisons. All three stars were

observed in sequence through a light neutral density filter. HR

5063 turned out to be slightly variable, so the SAAO results were

reduced with HR 5439, as the only comparison star. A log of the

photometric observations obtained at SAAO is shown in Table 2.

For completeness, we include the data obtained during 1987 and

1988 and published in Cuypers et al. (1989).

3 P H Y S I C A L PA R A M E T E R S

The Hipparcos parallax of m Cen is 6:19 ^ 0:71 mas; from which

we obtain MV � 22:64 ^ 0:25 assuming a mean unreddened

apparent magnitude of V � 3:4; the mean of the values in Cuypers

et al. (1989). The star is classified variously as B2Ve, B3Ve or

B2IV±Ve. Using the calibration of Popper (1980) and assuming

an uncertainty of half a spectral type, we have log Teff � 4:32 ^

0:05; BC � 22:15 ^ 0:21: The radius is then estimated to be

R=R( � 6:2 ^ 1:7: Rapid rotation will increase the equatorial

radius relative to the polar radius, perhaps by 25 per cent or

thereabout.

The calibration of Balona (1995) gives R=R( � 4:8; 6.6 and 8.4

for B2 classes V, IV and III respectively. For a B2 main-sequence

star, Porter (1996) finds a polar radius of 5.4 R(. The value

obtained above is in good agreement with these other estimates.

Assuming a typical mass of �10 ^ 1�M( for a B2 star, we find

log g � 3:9 ^ 0:3; and a critical rotational velocity of 550 ^

120 km s21: The shortest rotation period is 0:56 ^ 0:10 d: The

calibration of Porter (1996) gives a critical rotational velocity of

about 480 km s21.

The projected rotational velocity of m Cen is 130 ^ 9 km s21

(Brown & Verschueren 1997). We determined v sin i from various

helium and metal lines by using the intrinsic line profile calculated

from the spectrum program and a non-linear least squares

optimizing algorithm. Results, shown in Table 3, confirm the

value obtained by Brown & Verschueren (1997). From all lines in

both seasons we find v sin i � 127 ^ 3 km s21: From the helium

lines only we get v sin i � 123 ^ 3 km s21; or v sin i � 129 ^

1 km s21 if lines showing signs of emission (4471, 4922 and 5016)

are excluded. From the metal lines we find v sin i � 137 ^

3 km s21: The star is slowly rotating for a Be star and can be

classified as a `pole-on' Be star. Apart from v sin i, the radial

velocity of the line was also left as a free parameter. Table 3 lists

the radial velocities from the least-squares solution. A strange

result is that the metal lines have considerably higher radial

velocities than the helium lines.

4 T H E P E R I O D ± P H OT O M E T RY

Cuypers et al. (1989) obtained a large number of photometric

observations from Chile and South Africa, specifically to detect

coherent periodicities. The star was particularly active and they

were unable to find any clear period except possibly for a

frequency f � 0:48 cycle d21 (double-wave curve) or f � 0:96

cycle d21. The same frequency seems to be present in the 1988

data, from which they concluded that the best representation is

that of a double-wave light curve with P � 2:1017 d: They were

unable to detect any periodicity near 0.5 d.

In Fig. 1 we show the phase dispersion minimization period-

ograms for all the seasons. Except for the 1991 and 1999 seasons,

the power is located predominantly at 1 cycle d21. Table 2 shows

the period and amplitude that correspond to the highest peaks

in the periodograms. Very similar results are obtained by a

Fourier-type periodogram analysis. The only season in which a

frequency of 2 cycle d21 may have been present is 1990, but at

a weaker level than either the 1.0 or 0.5 cycle d21 peaks. These

data clearly demonstrate that there is no significant variation near

a period of 0.5 d. The Hipparcos and Tycho data sampling is too

poor to extract a reliable period.

It may be argued that the 1-d period suggested by the

photometry is, in fact, an alias of a much longer period.

Examination of the light variations shows, however, that the star

varies significantly during the course of a night, indicating that the

period is indeed of the order of 1 d. Except for the 1987 season and

the last three nights of the 1991 season, the star did not show any

large light changes characteristic of an outburst. The periodogram

for the 1991 season excludes these nights. We conclude that a

period close to 1 d is present in m Cen and that a 0.5-d periodicity

is absent in the photometry.

5 T H E P E R I O D ± R A D I A L V E L O C I T I E S

By the term `radial velocity' we mean the position of minimum

intensity of an absorption line. Clearly, the interpretation of this

measurement is model dependent. For NRP of a low degree, the

radial velocity can be used as a trace of the physical pulsation

periods, but this breaks down for modes of high degree which

Table 3. Projected rotational velocity, v sin i (km s21),obtained by non-linear least squares for various lines(laboratory wavelengths given). The standard devia-tion of the fit, s , is in continuum units. For the heliumlines, the range of fit was 2300 to 1300 km s21,while for the other lines it was 2200 to 1200 km s21.The last column gives the radial velocity of the line(km s21) as obtained from the fit.

Line Season v sin i s RV

He i 4026.359 1999 130 0.0053 5.5He i 4143.761 1999 133 0.0022 2.6He i 4387.929 1999 128 0.0036 7.4

2000 126 0.0023 6.8He i 4471.682 1999 119 0.0048 7.6

2000 110 0.0062 6.7He i 4713.376 1999 130 0.0014 4.0

2000 126 0.0025 6.2He i 4921.931 1999 124 0.0039 3.7

2000 115 0.0042 6.3He i 5015.678 1999 113 0.0055 21.0

2000 105 0.0078 3.5He i 5047.738 1999 130 0.0051 0.0

2000 130 0.0038 7.0C ii 4267.167 1999 152 0.0052 13.9

2000 133 0.0026 14.1Si iii 4552.622 1999 131 0.0019 20.0

2000 135 0.0017 21.3Si iii 4567.841 1999 136 0.0017 27.7

2000 135 0.0018 30.0

Short-period line profile variations in m Centauri 1043

q 2001 RAS, MNRAS 324, 1041±1053

Page 4: Short-period line profile variations in the Be star μ Centauri

gives rise to a series of moving bumps across the line profile. The

radial velocity only samples a small region near the centre of

the line profile. Discontinuities in the radial velocity occur as the

series of bumps moves across the centre of the profile. The

pulsational velocity itself is continuous and is related to the radial

velocity in a way which cannot easily be quantified. The period

obtained in this way will be affected by harmonics and, possibly,

linear combinations of the harmonics. The radial velocity of a

spotted star is affected in a similar way. In this case, the radial

velocity is not necessarily related to the period of the star (the

rotation period) but depends on the number and geometrical

distribution of the spots. By taking the radial velocity periods at

face value, one is imposing certain constraints in the interpretation

of the physical cause of the variations which may not be valid.

Turning now to our 1999 and 2000 echelle data, we calculated

the radial velocities of 10 helium lines, and the Si iii lines at 4553

and 4567 AÊ . Phase dispersion minimization periodograms of these

data are shown in Fig. 2. The peaks at 1.0 and 2.0 cycle d21 are the

most prominent features. In this case the Nyquist frequency is

high (about 50 cycle d21), so the problem discussed above does

not apply and it is certainly possible that the peak at 2.0 cycle d21

may be a real pulsational frequency. There is, however, little to

distinguish this frequency from the equally plausible 1.0 cycle d21.

We now proceed to determine what periodic Fourier com-

ponents are required to describe the radial velocity variations. We

do this by consecutive pre-whitening of the periodograms, always

taking the highest peak as the most probable periodic component.

We stop when we have reached a false alarm probability (Scargle

1982) of 3 per cent. Results are shown in Table 4. Clearly, a large

number of Fourier components are required to fit the observations

adequately. This is the type of analysis used by Rivinius

et al. (1998a), and we expect to recover their values. They

obtained the following frequencies (cycle d21): 1.988 368,

1.970 370, 2.022 151, 1.936 641, 3.55 360, 3.582 470 from the

FLASH/HEROS 1992±1997 data. The only frequencies that we

find in common are 1.988 368 and 3.553 60 cycle d21. However,

we need to ask the question of whether these frequencies refer to

real pulsation frequencies or whether they are merely a mathe-

matical description of the radial velocity, unrelated to the physical

period(s) in the star.

Figure 1. Phase dispersion minimization (PDM) periodograms of

StroÈmgren b obtained during several seasons. The frequency is in

cycle d21 and the amplitude in arbitrary units.

Figure 2. PDM periodograms of the radial velocities of several helium

lines and two Si iii lines from the SAAO GIRAFFE echelle spectra. The

1999 and 2000 data have been combined except for He i 6678 for which

only the 2000 data are available. The frequency is in cycle d21 and the

amplitude in arbitrary units.

1044 L. A. Balona et al.

q 2001 RAS, MNRAS 324, 1041±1053

Page 5: Short-period line profile variations in the Be star μ Centauri

Instead of measuring the position of maximum line absorption,

we can also use the centroid of the line profile as a measure of the

radial velocity. In a pulsating star, the centroid radial velocity can

be directly related to the pulsational velocity (Dziembowski

1977). Table 5 shows all the significant frequencies in three

helium lines. The centroid for each line includes most of the wings

except for He i 6678. For this line, the range was truncated to

prevent the inclusion of line emission in the wings. There is no

obvious common periodicity in all the three lines, though two

of them do show quite strong frequency components near

2 cycle d21. We suspect that the reason why the centroid is a

poor indicator of the periodicity in the star is because all the lines

are affected in some way by the circumstellar medium. Quasi-

random variations are introduced which mask the basic 0.5-d

periodicity. In a pulsating star the variations can be decomposed

into a superposition of spherical harmonics, allowing the

pulsational frequencies to be recovered in the centroid. This is

clearly not the case in m Cen and emphasizes the need for caution

in interpreting the radial velocities.

In Fig. 2 a peak near 3.6 cycle d21 �P � 0:28 d� is visible in the

helium lines, but absent in the Si iii lines. This is also found, but

not discussed, by Rivinius et al. (1998a). In terms of the

pulsational model, one would have to suppose that the absence

of this frequency component in the Si iii lines is a level effect. In

other words, there is a rapid increase or decrease of pulsational

velocity with height. The problem with this interpretation is that

there is really not much of a difference between the effective level

of formation of Si iii and the helium lines in a normal atmosphere.

The total absence of this frequency component in the silicon lines

is therefore quite puzzling. Curiously, this frequency is absent

from the centroid variation of the helium lines (Table 5). In our

opinion, this reflects the fact that the radial velocity variations do

not represent a meaningful physical property of the star.

The equivalent width variations of several lines were examined

for possible periodicities. All lines showed a long-term variation

in both seasons, but a periodogram analysis revealed no obvious

consistent periodicity. None of them showed any power near a

period of 0.5 d.

6 L I N E P R O F I L E VA R I AT I O N S

As we have mentioned, the physical interpretation of radial

velocity depends on the model being used. The radial velocity

essentially samples the position of the line core and gives no

information regarding most of the line profile. To avoid this

problem, it is essential that the structure of the line profile

variations be fully explored. The most convenient way of doing

this is to examine time sequences of the line profiles from which

the underlying, constant, profile has been removed. We used the

mean profile for a particular line during a season for this purpose.

It should be borne in mind that the mean line profile is not

necessarily the same as the line profile of a normal, non-variable,

non-emission star with the same physical parameters. In Fig. 3 we

show the mean profiles of the helium lines. The profiles for the

two seasons are in good agreement except for He i 5876, where

emission is present in the wings in 2000, though absent (or

weaker) in 1999. He i 5016 shows emission wings in 2000 but not

in 1999. Fig. 4 shows the mean profiles of some metal lines and

the hydrogen lines. The metal line profiles are in good agreement

in both seasons.

After dividing each spectrum by the mean spectrum for the

corresponding season, we obtain what we will call `difference

spectra'. In Fig. 5 we show grey-scale representations of time

sequences for the difference spectra of He i 4922 for four

consecutive nights in 1999 and 2000.

One important point to note is that the moving residual

Table 4. Analysis of the periodic components of the radialvelocities of He i 4922. The first column contains the data set,the number of observations and the standard deviation of themultiperiodic fit. The second column gives the frequency, incycle d21, and its standard error (in the last one or two digits).The third column is the fitted velocity amplitude (km s21) andits standard error. The fourth column is the phase and itsstandard error, in units of the period, with respect to JD245 1260.000.

Data set Frequency Amplitude Phase

All 1.98451 (^7) 6.8^ 0.4 0.458^ 0.0091166 3.55569 (^10) 7.0^ 0.3 0.226^ 0.0088.17 1.95188 (^13) 6.1^ 0.4 0.322^ 0.010

5.51429 (^14) 3.9^ 0.3 20.132^ 0.0140.12783 (^18) 3.7^ 0.3 0.032^ 0.0167.08989 (^19) 3.7^ 0.3 0.029^ 0.0150.30966 (^22) 3.5^ 0.3 20.170^ 0.0165.58564 (^24) 2.9^ 0.3 20.277^ 0.0193.63190 (^28) 2.6^ 0.3 20.292^ 0.021

1999 3.553 (^5) 5.7^ 0.6 0.254^ 0.015302 7.097 (^6) 4.8^ 0.5 20.061^ 0.0186.77 3.027 (^7) 4.4^ 0.5 20.315^ 0.020

5.490 (^7) 4.2^ 0.5 0.054^ 0.0201.242 (^9) 3.9^ 0.5 0.158^ 0.022

2000 1.977 (^2) 11.8^ 0.4 20.496^ 0.006864 4.544 (^4) 7.1^ 0.4 20.088^ 0.0098.29 2.570 (^7) 4.2^ 0.4 20.276^ 0.015

0.132 (^7) 3.8^ 0.4 20.499^ 0.0194.630 (^8) 4.3^ 0.4 0.193^ 0.0156.087 (^7) 4.5^ 0.4 20.259^ 0.0140.287 (^8) 5.6^ 0.4 20.411^ 0.0122.272 (^9) 3.8^ 0.4 20.457^ 0.017

Table 5. The same as Table 4, but for thecentroid of the line.

Frequency Amplitude Phase

He i 6678: 2000

1.968(^2) 9.0^ 0.2 0.396^ 0.0030.092(^3) 5.3^ 0.2 0.115^ 0.0051.573(^5) 4.4^ 0.2 20.001^ 0.0061.782(^5) 3.5^ 0.2 0.136^ 0.0070.338(^8) 3.8^ 0.2 0.101^ 0.0070.438(^8) 3.3^ 0.2 20.010^ 0.007

He i 4922: 1999 1 2000

1.2340(^1) 3.2^ 0.1 0.409^ 0.0061.3920(^1) 1.9^ 0.1 0.251^ 0.0091.1816(^1) 1.8^ 0.1 20.012^ 0.0101.7023(^1) 2.0^ 0.1 20.356^ 0.0091.2844(^1) 2.3^ 0.1 20.380^ 0.0081.7813(^1) 1.9^ 0.1 20.424^ 0.0090.8730(^2) 1.4^ 0.1 20.442^ 0.013

He i 4388: 1999 1 2000

0.3707(^1) 1.4^ 0.1 0.392^ 0.0091.7695(^1) 1.8^ 0.1 20.443^ 0.0061.9830(^1) 1.0^ 0.1 0.085^ 0.0120.1569(^1) 1.2^ 0.1 0.393^ 0.0110.7043(^1) 1.4^ 0.1 0.213^ 0.0080.0012(^2) 8.9^ 0.1 20.247^ 0.002

Short-period line profile variations in m Centauri 1045

q 2001 RAS, MNRAS 324, 1041±1053

Page 6: Short-period line profile variations in the Be star μ Centauri

absorption features are prominent in all the helium lines, including

He i 6678, a line which is clearly affected by the circumstellar

emission (Fig. 3). They are also visible in the C ii 4267, Mg ii

4481, Si iii 4553 and Si iii 4565 lines. The variations in the helium

lines and the metal lines are almost the same. Of even greater

significance, however, is that moving residual absorption features

are clearly distinguished in the hydrogen lines, though with more

difficulty. Fig. 6 shows the stacked difference spectra of the Hbline for four nights in the 2000 season. Similar figures are

obtained for Ha .

One of the remarkable features of the line profile variations is

the distinctive cross-over effect (particularly obvious on JD

245 1265). This occurs when a residual absorption feature moving

from blue to red encounters another residual absorption feature

moving from red to blue. In 1999 the cross-over appears on JD

245 1263.52, 245 1265.48 and 245 1273.65. In 2000 we have

cross-overs on JD 245 1675.57, 245 1679.48 and 245 1682.61. The

residual absorption feature moving from red to blue was not

reported by Rivinius et al. (1998a), presumably due to their poor

data sampling and the fact that it occurs at sporadic intervals. It is

also difficult to understand the range in the visibility of the cross-

over features. Fig. 5 shows that the crossover on JD 245 1265.48 is

prominent, while at other times it is more difficult to distinguish.

Fig. 6 shows that the cross-over effect is visible in the Hb line.

In Fig. 7 we show the locus of the ridge line velocity of the

residual absorption and emission features as a function of time for

He i 4922. By `ridge line' velocity we mean the position of

maximum residual absorption or emission intensity in the

difference profile at any given time. It is essentially a trace of

the residual absorption or emission features shown in the grey-

scale representations. These velocities were calculated by finding

the minimum (maximum) of a parabola fitted to the bottom (top)

of the residual absorption (emission) feature. A similar figure is

obtained for all helium and metal lines. It is apparent that there are

rather gross changes from night to night which deserve further

study. In Fig. 8 we show a plot of the residual absorption ridge line

velocity as a function of time. In this figure we have marked the

maximum limb velocities, assuming v sin i � 130 km s21 offset

for the mean radial velocity of this line as given in Table 3

(15 km s21). It is evident that there are frequent and significant

excursions outside these boundaries, but there are also a number

of nights where the residual absorption feature stays well within

the v sin i limit. We refer, in particular, to nights JD 245 1681, JD

245 1683 and JD 245 1686.

Another anomaly that is apparent from the grey-scale difference

spectra and from Fig. 7 is that there is quite often a static

absorption or emission feature at the red or blue wing which can

persist for several hours. This is much like the effect seen in hCen (Balona 1999), where alternate absorption and emission

features persist in the line wings. In m Cen, the travelling

absorption feature seems to linger on the limb of the star for many

hours.

7 T H E P E R I O D F R O M L I N E P R O F I L E

VA R I AT I O N S

Given that the difference spectra exhibit quite complex moving

structures, a simple scalar quantity such as radial velocity will not

be suitable for determining the period because it samples a very

small part of the profile. The photometry, however, shows that

only one period can be extracted from the data, the remaining

variation appearing to be random. This is quite typical of Be stars.

Sporadic line profile variations are also quite normal in the Be

stars and are discussed by Rivinius et al. (1998a) (see also Peters

1998). This, of course, means that extra care has to be taken before

Figure 3. Mean rectified profiles of the helium lines for the 1999 season

(thin line) and the 2000 season (thick line). The abscissa is in units of km s21

as measured from the laboratory wavelength of the line. The red side of

He i 5876 and the blue side of He i 5048 are affected by telluric lines.

Figure 4. The same as Fig. 3 but for the C ii 4267 line (blue emission is an

instrumental effect), the Mg ii 4481, Si iii 4553, 4567 and hydrogen lines.

1046 L. A. Balona et al.

q 2001 RAS, MNRAS 324, 1041±1053

Page 7: Short-period line profile variations in the Be star μ Centauri

concluding that the line profile variations are due to multiple

periods in a pulsating star.

Balona et al. (2000) assumed that since the cross-over effect

was visible approximately every alternate night (at least for the

first few nights), the frequency must be close to 1.5 cycle d21, and

suggested that P � 0:66 d � f � 1:52 cycle d21�: They noted,

however, that departures from strict periodicity were present

during the run (the 1999 data set described here). The additional

information we now have from the 2000 season shows quite

clearly that we cannot expect a strict periodic or multiperiodic

solution. The data can be described by one period, but, in addition,

there are random variations which cannot be reproduced by a

finite multiperiodic solution. The situation is quite similar to the

photometric behaviour, which is well described by only one

period, but with amplitudes and shapes that are highly variable.

It is clear from the period analysis described above that if one

can extract a single period from these data, the corresponding

frequency must be close to 0.5, 1.5, 1.0 or 2.0 cycle d21. If the

period is close to one day or half a day, then the line profile

variations should repeat from night to night except for a small

phase shift. This is true for the Rivinius et al. (1998a)

multiperiodic solution, because all the frequencies are within a

narrow range and cannot be resolved in a time span of two weeks.

To select the best period, we constructed a large number of

grey-scale difference spectra of the helium lines phased with

different periods and selected, by visual inspection, those periods

-200. 0. 200.0.7

0.6

0.5

0.4

0.3

0.2

-200. 0. 200. -200. 0. 200. -200. 0. 200.

-200. 0. 200.0.7

0.6

0.5

0.4

0.3

0.2

-200. 0. 200. -200. 0. 200. -200. 0. 200.

Figure 5. Grey-scale plots of stacked difference spectra of He i 4922 for four consecutive nights in the 1999 season (top) and the 2000 season(bottom). Time

runs from top to bottom, with the fractional part of the Julian date shown. The Julian dates for the 1999 season are, from left to right, 245 1262±245 1265 and

for the 2000 season 245 1682±245 1685: The black bar in the JD 245 1684 panel is a result of a few grossly underexposed spectra and should be ignored. The

abscissa is in units of km s21 as measured from the laboratory wavelength of the line. Dark areas represent residual absorption, light areas residual emission.

A typical residual absorption depth is 1±3 per cent of the continuum level.

Short-period line profile variations in m Centauri 1047

q 2001 RAS, MNRAS 324, 1041±1053

Page 8: Short-period line profile variations in the Be star μ Centauri

for which the diagrams showed the best coherency. We

concentrated on frequencies in the range of 1.0, 1.5 and

2.0 cycle d21 only. In Fig. 9 we have phased the 1999 and 2000

data for He i 4922 on the best periods: f � 1:52 cycle d21 �P �0:66 d�; f � 0:98 cycle d21 �P � 1:02 d� and f � 1:977 cycle d21

�P � 0:51 d�: Fig. 10 shows grey-scale plots of the Ha , Hb , He i

6678 and Si iii 4553 lines from the 2000 season phased with P �1:02 d; which is just twice the 0.51-d period found by Rivinius

et al. (1998a). Of course, we must also consider the 0.51-d period

as one of the possibilities. The new data show that the most likely

period is either P � 1:02 or 0:51 d; though the 0.66-d period is not

entirely excluded.

One of the most powerful clues to the correct period is to be

found in the photometric variations. It is clear that a period close

to 1 d is present in several seasons and must have a physical cause.

The fact that it is twice the shortest possible spectroscopic period

must surely be significant. One can either suppose that the true

period is 0.51-d, in which case a different explanation is required

for the 1-d photometric period, or assume that we are dealing here

with a fairly typical case among Be stars ± the double-wave

phenomenon. A substantial fraction of periodic Be stars have light

curves consisting of alternate deep and shallow minima (or

maxima), forming a double-wave light curve. Stars have been

known to change from a double-wave to a single-wave light curve

and vice versa (Balona et al. 1991). The double-wave phenom-

enon has never been investigated spectroscopically. In this

interpretation, the spectroscopic period of 0.51 d is half the true

period of 1.02 d.

8 S H A R P A B S O R P T I O N S P I K E S A N D A N

O U T B U R S T

In addition to the features migrating across most of the full width

of the profiles, Rivinius et al. (1998a) report conspicuous narrow

absorption spikes appearing close to the wings from time to time.

Their occurrence is strongly enhanced during precursor and early

phases of a line emission outburst. These spikes can occur in both

the blue and red wings, but are never seen in both wings

simultaneously. Rivinius et al. (1998a) suggest that both blue and

red spikes are part of the same pattern, separated by about 0.25 d.

A related phenomenon is the simultaneous appearance and

disappearance of a very extended absorption in the other wing,

taking the shape of an extended linear ramp. Rivinius et al.

(1998a) mention that during outbursts, when spikes are observed

more regularly, a cyclic behaviour becomes apparent.

With our intensive coverage, the nature of these absorption

spikes is clarified. They occur whenever the normal absorption

feature is at the blue or red wing. As already mentioned, the

travelling absorption feature is often prolonged at the wings,

giving rise to a long-lasting absorption spike in one of the wings.

This does not necessarily mean that there is a true absorption in

the wings. It is possible that the mean helium line profile is

slightly narrower in the wings than the true photospheric helium

line profile, a possibility that is supported by the smaller v sin i

obtained from the helium lines relative to the metal lines.

Nevertheless, the important point is that there is a variation in the

wings of the line profile which could be a variation from net

absorption to net emission or perhaps just a change in the

emission-line strength at the limbs. It is not possible at this stage

to distinguish between these possibilities. In Fig. 11 we show a

time sequence of He i 4922 line profiles for JD 245 1685, for

which the grey-scale representation is one of those shown in

Fig. 5. As can be seen in Fig. 11, a sharp absorption spike

appears, in this case on the red wing, at the times when the

absorption feature is at its maximum positive velocity. The

prolongation of the absorption on the red or blue wings can be

seen in Fig. 7 on many nights.

Another phenomenon which happens at sporadic intervals is

what is called an outburst. This is the sudden increase of Haemission (see, for example, Hanuschik et al. 1993) or emission in

the line wings of, for example, He i 6678 as described by Rivinius

et al. (1998a) and Peters (1998). Hanuschik et al. (1993) describe

-200. 0. 200.0.7

0.6

0.5

0.4

0.3

0.2

-200. 0. 200. -200. 0. 200. -200. 0. 200.

Figure 6. Grey-scale plots of stacked difference spectra of Hb for four nights in the 2000 season. From left to right, 245 1675, 245 1682, 245 1683, 245 1686.

Note that the central two panels refer to the same nights as the first two bottom panels in Fig. 5. The typical depths of the residual absorption features are

about 1±3 per cent in both Hb and He i 4922. The abscissa is in units of km s21 as measured from the laboratory wavelength of the line.

1048 L. A. Balona et al.

q 2001 RAS, MNRAS 324, 1041±1053

Page 9: Short-period line profile variations in the Be star μ Centauri

four major Ha outbursts of m Cen occurring within an interval of

200 d in 1987. Rivinius et al. (1998c) have characterized an

outburst by four primary phases: (i) a period of relative

quiescence, (ii) a sudden �5±13 d� drop in the strength of all

circumstellar emission lines (the `precursor' phase), (iii) a

subsequent rapid �2±15 d� increase in the emission strength (the

outburst itself) and (iv) a transition phase evolving into relative

quiescence.

It is interesting to note that although He i 6678 in our data

frequently shows emission in the red and/or blue wings, the

difference profiles show exactly the same behaviour as any other

helium or metal line. On JD 245 1677, we noted a sudden increase

of emission in the blue wing of this line (Fig. 12) followed by a

more gradual decrease. Strangely, on the night of the outburst, the

moving absorption feature is difficult to distinguish in grey-scale

plots of the other helium lines, but is quite clear in He i 6678 and

the Ha and Hb lines, indicating its circumstellar origin. The

central depth of the moving absorption feature shows a sudden

increase from its mean value of about 2 per cent continuum units

at the beginning of the night to a maximum of about 5 per cent on

JD 245 1677.3, the same date as maximum red wing emission, just

two hours later. It decreases again to the mean value by the end of

the night. The moving absorption feature in Hb shows the same

behaviour, reaching a maximum of 5 per cent at about the same

time. However, there are large variations in the depth and the trend

is less clear. The Ha line is contaminated with telluric absorption

features which makes such an analysis rather difficult. Apart from

He i 6678 and Hb , the moving absorption feature in other lines

does not appear to undergo any significant change at the time of

the outburst.

In Fig. 13 we show the strength of the Ha emission relative to the

continuum (E/C ratio) for the red and blue wings and the central

absorption dip as a function of time. The figure looks rather similar

to the schematic diagram of an outburst phase shown by Rivinius

Figure 7. Loci of the ridge lines of the absorption (filled circles) and

emission (lines) features for all nights of the 1999 (top) and 2000 (bottom)

seasons as measured in the He i 4922 line. Panels are labelled with respect

to JD 245 1000. The fraction of the HJD and the velocity in km s21 are

shown. The small filled circles for JD 245 1271 and 245 1677 are to

indicate the uncertain nature of the loci on these two nights.

Figure 8. Loci of the ridge lines of the absorption features (filled circles)

plotted as a function of time for the 1999 (top) and 2000 (bottom) seasons

for He i 4922. The horizontal lines are the extreme limb velocities

assuming v sin i � 130 km s21:

Short-period line profile variations in m Centauri 1049

q 2001 RAS, MNRAS 324, 1041±1053

Page 10: Short-period line profile variations in the Be star μ Centauri

et al. (1998c, their fig. 3) and confirms the characteristics that they

describe. Maximum Ha emission occurs at about JD 245 1682, five

days after the outburst seen in He i 6678. It seems that the original

outburst occurred close to the photosphere, affecting only the He i

6678 line, and that the material ejected from this outburst only

reached the circumstellar disc after a few days.

The Hb emission-line strength varies from E=C � 0:87 to

0.91 during the same period and shows similar behaviour.

The spacing between the two emission peaks at Ha remains

constant at 102:2 ^ 0:06 km s21: For Hb the corresponding

spacing is 147:5 ^ 0:2 km s21: As can be seen in Fig. 13, there

is considerable variation on a short time-scale in the intensities of

all three features, particularly in the intensity of the emission

strength of the red peak. This variation has a period of

approximately 0.5 d. The ratio of the violet and blue peaks (V/R

ratio) varies with the 0.5-d period as well. In Ha , the range of V/R

variation is 0:7±1:0 and in Hb it is 0:9±1:1:There appears to have been no outburst during the 1999 season.

In this season the height of the central reversal in Hb remained at

1.00 for the whole run. This is much the same as in the 2000

season. The V/R variation is clearly periodic with P � 1:067 d:Unlike in the 2000 season, the periodogram peak at this period is

considerably higher than for the period 0.486 d. This supports the

photometric period and our deduction that the true period is nearly

1 d. The spacing between the two emission peaks at Hb remains

constant at 113:7 ^ 0:14 km s21; much lower than in 2000.

We were unable to detect the presence of any Fe ii emission or

absorption lines. The Si ii lines are present in absorption.

According to Rivinius et al. (1998c), Fe ii lines are good indicators

of a major outburst. The absence of these lines in our spectra

indicates that the outburst described here is a relatively minor one.

Rivinius et al. (1998b) found that the times of maximum

amplitude, calculated from the beating of the periods in their

multiperiodic solution, are correlated with the times of the

outburst. The actual outburst dates predicted by their multi-

periodic NRP model are JD 245 1669 and JD 245 1689. The

-200. 0. 200.

.0.2

.4.6

.81.

01.

21.

4P

hase

-200. 0. 200.

-200. 0. 200.

.0.2

.4.6

.81.

01.

21.

4P

hase

-200. 0. 200.

-200. 0. 200.

.0.2

.4.6

.81.

01.

21.

4P

hase

-200. 0. 200.

Figure 9. Grey-scale plots of stacked difference spectra of He i 4922 for

the 1999 (left panels) and 2000 (right panels) data phased with frequencies

of f � 1:52 cycle d21 (top panels), f � 0:98 cycle d21 (middle panels) and

f � 1:977 cycle d21 (bottom panels). The abscissa is in units of km s21 as

measured from the laboratory wavelength of the line.

-200. 0. 200.

.0.2

.4.6

.81.

01.

21.

4P

hase

-200. 0. 200.

-200. 0. 200.

.0.2

.4.6

.81.

01.

21.

4P

hase

-200. 0. 200.

Figure 10. Grey-scale plots of stacked difference spectra from the 2000

season phased with f � 0:98 cycle d21 �P � 1:02 d�: Left to right: top

panels ± He i 6678, Si iii 4553; bottom panels ± Ha , Hb . The abscissa is

in units of km s21 as measured from the laboratory wavelength of the line.

1050 L. A. Balona et al.

q 2001 RAS, MNRAS 324, 1041±1053

Page 11: Short-period line profile variations in the Be star μ Centauri

outburst we observed occurred on JD 245 1677, almost exactly

halfway between the two predictions.

9 D I S C U S S I O N

NRP was originally suspected as the cause of the periodic

behaviour in Be stars, because of the strong qualitative

resemblance of the line profile variations to those predicted by

NRP models and to the general perception that all B-type stars

pulsate. There have only been two quantitative studies in which

modelling of line profiles is attempted, both for 28 CMa. Balona,

Aerts & SÏtefl (1999) were unable to fit the profile variations with

the generally accepted idea that the modes in the Be stars are

prograde NRP waves. However, Maintz et al. (2000) were able to

fit the line profiles with a mode which is retrograde in the

corotating frame, but prograde in the observer's frame.

The line profile variations in m Cen are rather different from

what might be expected from NRP. For example, in NRP the

stacked difference spectra create a typical `barber pole' diagram in

which an absorption feature travels from a velocity of 2v sin i to

1v sin i. A moving feature due to NRP must always travel the full

profile between these limits, but, as Fig. 8 shows, this is not the

case in m Cen. It is not physically possible to have NRP that

produces a double-wave radial velocity curve while at the same

time generating a single-wave light curve. The photometric period

of 1 d in m Cen cannot therefore be explained by NRP, even if

NRP is assumed for the line profile variations. It is also difficult to

understand how the random occurrence of a red-to-blue moving

absorption subfeature can be explained by NRP.

Other mechanisms are required to explain the many sporadic

features seen in the line profiles, even if NRP is assumed. In order

to understand rapid changes that occur in the light curves and the

complex, variable, moving absorption features in the line profiles

of Be stars, it is evident that considerable surface activity or highly

variable obscuration is required. At the same time, there is no

doubt that a characteristic period can be obtained from the

photometric and spectroscopic data. This period is close, or equal,

to the period of rotation. As we have seen, NRP has great

difficulty in accounting for the behaviour of the moving

absorption features in m Cen. A starspot likewise does not have

the desired properties and, in any case, cannot reproduce the large

radial velocity to light amplitude ratios (Balona 1995). We do

know, without doubt, that the Be stars are losing mass and that this

mass loss originates in the photosphere. If the mass loss is

occurring at one or two localized points on the photosphere, we

automatically have a mechanism which can generate the line

Figure 11. He i 4922 line profiles during the night of JD 245 1685 (also

shown in grey-scale representation in Fig. 5). Each line profile is a mean of

five observed profiles. Notice the increase in absorption in the red wing at

a fixed velocity of about 70 km s21 (indicated by the dashed line).

Figure 12. He i 6678 line profiles during the night of JD 245 1677 showing

an outburst resulting in increased blue wing emission. Later, the emission

in the red wing also increases.

Short-period line profile variations in m Centauri 1051

q 2001 RAS, MNRAS 324, 1041±1053

Page 12: Short-period line profile variations in the Be star μ Centauri

profile and light variability without postulating NRP or any other

mechanism. The gas emerging from the localized area will cause

an obscuration of the photosphere and give rise to the variations

that we observe. In order to produce a distinct period, however, the

gas clouds need to be confined to their place of origin, perhaps by

a magnetic field, so that they are forced into corotation with the

star. In other words, we suggest that the periodic line profile and

light variations are a direct consequence of localized mass loss.

The most convincing evidence for the corotating gas clouds is

given by h Cen (Balona 1999). In this star NRP is ruled out owing

to the complex nature of the moving absorption features and the

presence of alternate emission and absorption at the stellar limb.

The He i 6678 line profiles in z Tau are strongly affected by

circumstellar material, yet the moving absorption features are

clearly seen and must be formed above the photosphere (Balona &

Kaye 1999). The line profile variations in the pole-on Be star 28

CMa are described very well by a cloud at an intermediate latitude

(Balona et al. 1999) though, as mentioned above, a peculiar mode

of pulsation can also be made to fit the profile variations (Maintz

et al. 2000). In e Cap (Balona & Lawson 2001), all the helium

lines are affected by emission, yet the difference profiles show the

same periodic pattern, strongly indicating that the periodicity is

due to circumstellar material and not in the photosphere.

Magnetically controlled corotating circumstellar clouds have

also been proposed to explain the far-UV variations in g Cas

(Smith, Robinson & Hatzes 1998; Smith & Robinson 1999).

In m Cen the moving absorption subfeature is seen in lines

which are clearly affected by the circumstellar material: in He i

6678 which shows strong wing emission and in Ha and Hb . As in

other Be stars mentioned above, there is little doubt that the

periodicity arises in the circumstellar material and not in the

photosphere. Two diametrically situated clouds could, we feel,

explain the gross properties of the line profile variations in m Cen,

but it is clear from the wealth of new information described here

that complex mechanisms are at work.

If we assume that the true rotation period is 1.02 d, and we

accept a stellar radius of R=R( < 7 ^ 1:7 and v sin i �130 km s21; then the equatorial velocity v � 350 ^ 85 km s21

and the angle of inclination is i � 228 ^ 68: It is evident that the

clouds must be situated close to the equator in order to produce the

phase diagram shown in Fig. 9. The prolongation of the absorption

at the limb probably implies that the clouds are quite large (in 28

CMa, the line profile variations can be modelled by a cloud with a

radius of about 408: Balona et al. 1999). We do not know whether

true variable residual absorption is involved or whether there is

simply a variation in the emission at the limbs. In either case, it is

difficult to account for the phenomenon without invoking a cloud

of gas just above the photosphere.

The 1-d period of the light curve can be understood if the light

distribution from the star has a dipole character. For example, if

we assume that the clouds are situated above the north and south

magnetic poles and that the surface temperature is a maximum at

one of the poles and a minimum at the other, we will get a

photometric period of one day and a spectroscopic period of

half a day. Another possibility might be that the light from the

star is redistributed by the scattering of a disc inclined to the

equator.

The major difficulty with this model is to understand the

excursions outside and within v sin i. Absorption cannot occur

outside the limb. If true variable residual absorption is involved,

we need to assume that the feature is Doppler shifted by a mass

motion of the gas cloud. A mass motion in the direction of rotation

will lead to an excursion outside v sin i, while a motion in the

opposite direction will confine the moving feature to within this

limit. From Fig. 8, we estimate that a mass velocity of about

20 km s21 will be sufficient. The occasional red-to-blue moving

absorption feature, which appears to be formed at a somewhat

higher altitude above the photosphere, may be detached clumps at

a higher latitude seen beyond the pole.

It is unfortunate that such flexibility needs to be introduced in a

model, but it is difficult to see how this can be avoided in any

other hypothesis. We know that the Be stars are highly complex

and this seems to be another manifestation of the complexity. It is

beyond the scope of this work to attempt a model of these

complex variations. At this early stage we do not, in any case,

possess enough information to attempt this task. What is required

is further intensive observations of other Be stars in order to

improve our understanding of these enigmatic stars.

We were fortunate to observe quite a large outburst, leading to

the rapid development of an emission wing in He i 6678 followed

by a substantial increase in the Ha emission after five days. At the

time of the outburst, the normal moving absorption feature is

difficult to distinguish, as if the whole star is obscured by veiling.

In fact, the feature is more easily visible in Ha and He i 6678 than

in the other helium and metal lines on this night. The outburst

seems to have been located near the approaching limb. A similar

outburst was seen and studied on the visible hemisphere of h Cen

(Balona 1999), where its localized nature was evident.

Figure 13. The height relative to the continuum (E/C ratio), of the red

emission peak (bottom panel), the central absorption (middle panel) and

the blue emission peak (top panel) in Ha as a function of time. The rise in

emission coincides with the outburst shown in Fig. 12.

1052 L. A. Balona et al.

q 2001 RAS, MNRAS 324, 1041±1053

Page 13: Short-period line profile variations in the Be star μ Centauri

1 0 C O N C L U S I O N S

We do not confirm the multiple periods previously found by

Rivinius et al. (1998a) in the radial velocities. The apparent

multiperiodicity appears to be a result of poor sampling

combined with a period very close to a submultiple of one

day. Evidence from the light variations leads us to believe that

the true period of m Cen is 1.02 d. We believe that the

periodicities from the radial velocities do not correspond to any

physically meaningful process, but are a result of complex line

profile variations arising mostly from the circumstellar material.

As in many other Be stars, we find that periodic variations are

present in the lines which are clearly affected by the

circumstellar material. We suggest that the periodic light and

line profile variations in Be stars arise as a result of localized

corotating clouds formed by outburst events.

Our observations of m Cen have revealed several previously

unknown characteristics. Perhaps the most important finding is

that the line profile variations are far more complex than

previously supposed. The complexity is only apparent with

intense data sampling. One surprise is the occasional, apparently

random, occurrence of a travelling residual absorption feature

moving from red to blue instead of blue to red.

It is clear that, for further progress, intensive spectroscopic

studies involving a large number of spectra obtained within a

relatively limited time span of a week or two are essential. We

need to determine characteristics common to all periodic Be stars

so that an idea of the physical conditions of the gas clouds can be

determined.

AC K N OW L E D G M E N T S

We thank Drs Rivinius, Baade, SÏ tefl, Stahl, Wolf and Kaufer (the

HEROS group) for kindly sending us the radial velocity data used

in their paper, Rivinius et al. (1998a).

R E F E R E N C E S

Balona L. A., 1990, MNRAS, 245, 92

Balona L. A., 1995, MNRAS, 277, 1547

Balona L. A., 1999, MNRAS, 306, 407

Balona L. A., Kaye A. B., 1999, ApJ, 521, 407

Balona L. A., Lawson W. A., 2001, MNRAS, 321, 131

Balona L. A., Sterken C., Manfroid J., 1991, MNRAS, 252, 93

Balona L. A., Aerts C., SÏ tefl S., 1999, MNRAS, 305, 519

Balona L. A., James D., Lawson W. A., Shobbrook R. R., 2000, in Smith

M. A., Huib H., Fabregat J., eds, ASP Conf. Ser. 214, Proc. IAU

Colloq. 175, The Be Phenomenon in Early-Type Stars. Astron. Soc.

Pac., San Francisco, p. 220

Brown A. G. A., Verschueren W., 1997, A&A, 319, 811

Cuypers J., Balona L. A., Marang F., 1989, A&AS, 81, 151

Dziembowski W. A., 1977, Acta Astron., 27, 203

Gray R. O., Corbally C. J., 1994, AJ, 107, 742

Hanuschik R. W., Dachs J., Baudzus M., Thimm G., 1993, A&A, 274, 356

Maintz M., Rivinius Th., Tubbesing S., Wolf B., 2000, in Smith M. A.,

Huib H., Fabregat J., eds, ASP Conf Ser. 214, Proc. IAU Colloq. 175,

The Be Phenomenon in Early-Type Stars. Astron. Soc. Pac., San

Francisco, p. 244

Peters G. J., 1998, ApJ, 502, L59

Popper D. M., 1980, ARA&A, 18, 115

Porter J. M., 1996, MNRAS, 280, L31

Rivinius Th., Baade D., SÏ tefl S., Stahl O., Wolf B., Kaufer A., 1998a,

A&A, 336, 177

Rivinius Th., Baade D., SÏtefl S., Stahl O., Wolf B., Kaufer A., 1998b, in

Bradley P. A., Guzik J. A., eds, ASP Conf. Ser. 135, A Half-Century of

Stellar Pulsation Interpretations. Astron. Soc. Pac., San Francisco, p. 343

Rivinius Th., Baade D., SÏ tefl S., Stahl O., Wolf B., Kaufer A., 1998c,

A&A, 333, 125

Scargle J. D., 1982, ApJ, 263, 835

Smith M. A., Robinson R. D., 1999, ApJ, 517, 866

Smith M. A., Robinson R. D., Hatzes A. P., 1998, ApJ, 507, 945

SÏtefl S., Aerts C., Balona L. A., 1999, MNRAS, 305, 505

This paper has been typeset from a TEX/LATEX file prepared by the author.

Short-period line profile variations in m Centauri 1053

q 2001 RAS, MNRAS 324, 1041±1053