Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://star.arm.ac.uk/preprints/2008/518.pdf Äàòà èçìåíåíèÿ: Mon Feb 11 12:54:09 2008 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 02:49:12 2012 Êîäèðîâêà:

The Coherent Radio Emission from the RS CVn Binary HR 1099 O. B. Slee
A B C

A, C

, W. LawsonA , G. RamsayB

Australia Telescop e National Facility, CSIRO, P.O. Box 76 Epping, NSW 2121, Australia Armagh Observatory, College Hill, Armagh, BT61 9DG, Northern Ireland, UK Email: bruce.slee@csiro.au

Accepted for Publication Feb 2008.
Abstract: The Australia Telescope was used in March­April 2005 to observe the 1.384 and 2.368 GHz emissions from the RS CVn binary HR 1099 in two sessions, each of 9 h duration and 11 days apart. Two intervals of highly polarised emission, each lasting 2-3 h, were recorded. During this coherent emission we employed a recently installed facility to sample the data at 78 ms intervals to measure the fine temporal structure and, in addition, all the data were used to search for fine spectral structure. We present the following observational results; (1) 100% left hand circularly polarised emission was seen at both 1.384 and 2.368 GHz during separate epochs; (2) the intervals of highly polarised emission lasted for 2­3 h on each occasion; (3) three 22 min integrations made at 78 ms time resolution showed that the modulation index of the Stokes V parameter increased monotonically as the integration time was decreased and was still increasing at our resolution limit; (4) the extremely fine temporal structure strongly indicates that the highly polarised emission is due to an electron-cyclotron maser operating in the corona of one of the binary components; (5) the first episode of what we believe is ECME (electro-cyclotron maser emission) at 1.384 GHz contained a regular frequency structure of bursts with FWHM 48 MHz which drifted across the spectrum at 0.7 MHz min-1 . Our second episode of ECME at 2.368 GHz contained wider-band frequency structure, which did not permit us to estimate an accurate bandwidth or direction of drift; (6) the two ECME events reported in this paper agree with six others reported in the literature in occurring in the binary orbital phase range 0.5 ­ 0.7; (7) in one event of 8 h duration, two independent maser sources were operating simultaneously at 1.384 and 2.368 GHz. We discuss two kinds of maser sources that may be responsible for driving the observed events that we believe are powered by ECME. One is based on the widely reported `loss-cone anisotropy', the second on an auroral analogue, which is driven by an unstable `horseshoe' distribution of fast-electron velocities with respect to the magnetic field direction. Generally, we favour the latter, because of its higher growth rate and the possibility of the escape of radiation which has been emitted at the fundamental electron cyclotron frequency. If the auroral analogue is operating, the magnetic field in the source cavity is 500 G at 1.384 GHz and 850 G at 2.368 GHz; the source brightness temperatures are of the order TB 1015 K. We suggest that the ECME source may be an aurora-like phenomenon due to the transfer of plasma from the K2 subgiant to the G5 dwarf in a strong stellar wind, an idea that is based on VLBA maps showing the establishment of an 8.4 GHz source near the G5 dwarf at times of enhanced radio activity in HR 1099. Keywords: Stars: activity -- stars: individual: HR 1099 -- stars: binary -- radio continuum: stars -- radiation mechanisms

1

Intro duction

It has b een known for some years that the very active RS CVn binary HR 1099 occasionally emits highly p olarised, narrow-band emission at L (1.4 GHz) and S bands (2.4 GHz), (e.g. White and Franciosini 1995, Jones et al. 1995). Similar episodes have b een recorded in the emissions of some dMe flare stars, a notable example b eing that from Proxima Centauri (Slee, Willes and Robinson 2003). Recently, p eriodic bursts of coherent emission have b een detected from an ultra-cool dwarf star by Hallinan et al. (2007). In addition, attempts have b een made over the past few years to detect coherent radio emission from double degenerate binary systems (Wu et al. 2002) and in degenerate star-planet systems (Willes & Wu 2004). In these systems, the authors suggest that unip olar induction (UI) could b e a driving mechanism, comparable to that which drives the Jupiter-Io decametric emission. Evidence for such emission has b een detected in the candidate double degenerate system RX J0806+15 (Ramsay et al. 2007).

1


2

Publications of the Astronomical Society of Australia

Recent observations by Osten and Bastian (2006, 2007) of the highly p olarised emission from the dwarf flare star AD Leo, using time resolutions of 10 and 1 ms resp ectively, showed a rich variety of frequency and temp oral structure, which ranged from diffuse bands to narrow-band, fast-drift striae. On one occasion the authors conclude that the mechanism is coherent plasma emission, while in another flare at a later date they favour the electroncyclotron maser (ECME). By analogy with the short highly p olarised solar bursts containing structure as short as 40 milli-seconds (Dulk 1985), the stellar emissions have b een attributed to coherent sources in stellar coronae, but up to now synthesis instruments have not p ossessed the millisec time resolution to prop erly compare stellar and solar coherent emissions. In the solar case, the shortest spike-like bursts are often attributed to electron cyclotron maser emission (ECME), while the coherent emissions resp onsible for the much longer bursts of Typ es I - V are due to plasma emission (c.f. Dulk 1995). It is clear that in order to distinguish b etween these two typ es of coherent emission from stars, one needs an improvement in the time resolution of synthesis telescop es of two orders of magnitude plus the ability to track the bursts of emission in frequency over at least 100 MHz. During 2004 one of us (WW) modified the correlator for the compact array (ATCA) of the Australia Telescop e to deliver samples at the rate of 12.8 s-1 , which is 128 times faster than that used for earlier work on stellar emissions with the ATCA. The on-line display software (VIS) was also modified to display the flux density in the Stokes V parameter (circularly p olarised comp onent of Stokes I); this prompts the observer to switch to the high sampling rate of 12.8 s-1 only at times when highly circularly p olarised emission of sufficient intensity is seen on the display. This observing system was utilized during two long observing sessions with the ATCA on March 28 and April 08, 2005, and resulted in our confirming the detection of strong coherent emission on b oth occasions.

2

Observations

We observed HR 1099 with the ATCA in a 6 km configuration for 9 h on b oth 28/3/05 and 8/4/05, recording simultaneously at 1.384 and 2.368 GHz with the usual integration time of 10 s. The flux density calibrator was B1934-638 and the phase calibrator (interleaved with the target integrations) was J0336-019. After the calibration, the full data-set was mapp ed to check that the radio image of HR 1099 was not contaminated by side lob es of surrounding field sources. Since HR 1099 is situated near the celestial equator, the E­W configuration of the ATCA resulted in very p oor angular resolution in declination, so that we needed to b e sure that the side lob es of surrounding field sources were minimized. The left hand panel of Figure 1 shows a cleaned map within a square area of 26 â 26 around HR 1099; the restoring b eam with FWHM 569 â 3.8 and ma jor axis in PA = 0 deg is depicted in the extreme lower left. It is clear the u-v data forming this map are likely to b e contaminated by the numerous field sources visible in this map. We therefore modelled the field using the clean comp onents from Figure 1 (left hand panel) and subtracted the modelled field to produce the much b etter map of Figure 1 (right hand panel), which has a dynamic range of 182 (maximum / minimum contour levels); the measured rms level in clear areas around HR 1099 is 70 µJy/b eam. It is esp ecially pleasing that the maximum side lob es from HR 1099 itself are only 0.6% of its p eak flux density. This modelling procedure was used at 1.384 GHz and 2.368 GHz on the data for b oth March 28 and April 08; the resulting modified uv data sets were used in the following analysis. The averaged total flux density (Stokes I) of each 25 min integration was then measured using the MIRIAD task UVFIT and plotted against the mid-UT of each integration. A similar set of measurements was made and plotted for the circularly p olarized comp onent Stokes V. So far we have b een referring only to the standard 10 s integrations. Our monitoring of the on-line display had enabled us to recognize the presence of strong, highly circularly p olarised emission. During the one or two 25 min integrations when this emission was strongest, we switched to the alternative high-resolution system, the results b eing available in a separate data file. These data were also separately calibrated in MIRIAD, the calibrators having also b een sampled at the rate of 12.8 s-1 . To analyse this data, we devised sp ecial software that enabled us to apply the task UVFIT to six sets of integrations automatically, with sampling intervals that varied b etween 0.078 and 40 s. This resulted in six sets of flux densities together with their rms fitting residuals. In the present exp eriment we needed to examine the sp ectral distribution of the continuum across the bandwidth of the receiver at two frequencies centred on 1.384 and 2.368 GHz. These frequencies were recorded simultaneously, with the four Stokes parameters available at each frequency. The IF output consisted of 13 indep endent frequency channels, each of 8 MHz width and spaced 8 MHz apart, p ermitting us to display the sp ectrum over a total width of 104 MHz, resulting in total fractional bandwidths at the lower and higher frequencies of 0.075 and 0.044 resp ectively. The fractional bandwidths of the thirteen 8 MHz IF channels were 0.00578 and 0.00036 at the two frequencies. We describ e the observations separately for each of the recording dates b elow.


www.publish.csiro.au/journals/pasa

3

Figure 1: Left): A cleaned total intensity map at 1.384 GHz of the field centred on HR 1099 on March 28 2005, constructed from nineteen 25 min integrations with the data sampled at 10 s intervals. The highest contour level is 33.62 mJy/beam and the lowest is 0.93 mJy/beam. The restoring beam of FWHM = 569â3.8 arcsec is shown in the extreme lower left corner. Right): The field surrounding HR 1099 after modelling the field in the left hand panel (with HR 1099 masked) and subtracting the uv components of the field sources. The highest contour is 33.73 mJy/beam and the lowest is 0.21 mJy/beam. The rms over the clear area surrounding the central unresolved plot of HR 1099 is 70 µJy/beam. The restoring beam of FWHM = 569â3.8 arcsec is shown in the extreme lower left corner.

2.1

Observations of March 28

Figure 2 shows the mean flux densities of the 25 min integrations, using the standard 10 s sampling interval. First, the Stokes I measurements in panel (a) at 1.384 and 2.368 GHz display markedly different variability, with the lower frequency showing increases that are not reproduced at 2.368 GHz. The higher frequency shows a steady intensity of 35­37 mJy, while the 1.384 GHz emission displays an increased level over the first 4h and a much larger increase in the last 2 h. A glance at panel (b) suffices to show that the p olarised intensity (Stokes V) at 2.368 GHz is close to zero while the same marked 1.384 GHz p eaks that were seen in panel (a) are clearly reproduced in panel (b). These p eaks are reproduced in the fractional circular p olarisation at 1.384 GHz shown in panel (c). Figure 2 clearly demonstrates that we are observing relatively narrow-band highly p olarised emission at 1.384 GHz, but the full extent of the fractional p olarisation will not b e evident until the full band width of 104 MHz, utilized in the results of Figure 2, is sub divided into its thirteen 8 MHz-wide frequency channels. The intensity of the first p olarised section at 1.384 GHz, shown in panel (b) of Figure 2, was not high enough to actuate the high-resolution mode of op eration; this was brought into op eration for one of the 25 min integrations near the strong p eak at 08:08 UT.

2.2

Analysis of high time-resolution data

Our analysis of the high time-resolution data was intended to elucidate the temp oral structure of the highly p olarised emission. We did this by allocating the data to six bins containing the samples integrated over 40, 10, 2.5, 0.625, 0.156 and 0.078 s and then finding the average Stokes V intensity in each bin. Call the mean flux 2 density associated with each bin, < St >, say, and its variance t , where t signifies the integration time of samples assigned to that bin. In addition, the ith individual flux density in that bin has its residual mean square value, 2 2 i , which is dep endent on the system noise and will also make a contribution to t . Therefore, to compute the 2 variance due to the star's intensity variability b etween samples in the bin, t , we have:
2 t 2 2 = t - < l >

(1)

in which the l have b een averaged over all the samples in the bin. The rms value of the stellar variability is thus;
t


2 = t - <

2 t

>

1/2

(2)

Next, to assign a meaningful measure of this variability, we define a modulation index: M = t / < S t > (3)


4

Publications of the Astronomical Society of Australia

45 Stokes I / mJy 43 41 39 37 35 1.0

(a)
Stokes V / mJy

-20 -15 -10 -5 0 3.0 5.0 7.0 -0.5 (c) V / I at 1.384 GHz -0.4 -0.3 -0.2 -0.1 0.0 1.0 3.0 5.0 7.0 9.0 1.0

(b)

3.0

5.0

7.0

9.0

9.0

Mid-UT / h March 28 2005

Figure 2: The Stokes I and V flux densities from the 25 min integrations on March 28 2005 are plotted against the mid-UT of each integration. Panel (a) shows the I flux densities at 1.384 GHz (filled circles joined by full lines) and the I flux densities at 2.368 GHz (open squares joined by dotted lines). Panel (b) shows the V flux densities at 1.384 GHz (filled circles joined by full lines) and at 2.368 GHz (open squares joined by dotted lines). Panel (c) chows the corresponding fractional values of circular polarisation (V/I) with their error bars. The error bars attached to the 1.384 GHz flux densities in panels (a) and (b) are the rms residuals from the task UVFIT; the error bars at 2.368 GHz are not shown but are similar in amplitude.

2.3

Modulation index for March 28 2005

The modulation indices for the coherent emission of 28 March are plotted in Figure 3a, which shows a steadily increasing value for M as the integration time falls to 0.078 s. It seems probable that M will approach 100% at still lower values of integration time, suggesting that temp oral structure as low as the 2­3 ms seen at times by Osten and Bastian (2007) may b e present. However, the presence of such fine temp oral structure does not necessarily discriminate b etween the alternative mechanisms of plasma emission and ECME. Additional attributes of the radiation such as its fractional instantaneous bandwidth and its frequency drift rate are necessary in making this distinction.

2.4

Observations of April 08 2005

Figure 4 shows the total and p olarised flux densities for the 9-h observation of April 08 in panels (a) and (b) resp ectively, while panel (c) plots the p olarised fraction V/I. It is evident that a significant event occurred at 2.368 GHz in the last three hours of the observation, but on this date there was also a slowly decreasing level of left-handed circular p olarisation at 1.384 GHz (negative Stokes V). In this resp ect, the emissions differ significantly from those on March 28, when the stonger event occurred at 1.384 GHz and a negligible level of Stokes V was seen at 2.368 GHz. The intensity levels of Stokes V at 1.384 GHz were never high enough to actuate the high-time resolution sampling, but we were able to record two high-resolution sections at 2.368 GHz centred on 07:00 UT and 08:27 UT. The Stokes I and V flux densities were similar to those attained on March 28. The two high-resolution sections were binned in the same manner as that describ ed in Section 2.2 and their modulation indices (M) were computed.

2.5

Modulation indices for April 08 2005

Figure 3b shows that the modulation indices steadily increase with decreasing integration time and are still increasing at our resolution limit of 78 ms. A comparison of Figures 3a and 3b indicates that the modulation indices of the 1.384 and 2.368 GHz coherent emissions show similar b ehaviour. Unfortunately, we do not have the capability to increase our time and frequency resolution for continuum observations with the existing ATNF correlator; such observations would b e required to b etter differentiate b etween p ossible coherent emission mechanisms.


www.publish.csiro.au/journals/pasa

5

70

50
Modulation Index / %

(a)

March 28 1.384 GHz Modulation Index / %

60 50 40 30 20 10 0 10-1

(b)

April 08

2.368 GHz

40 30 20 10 10 -1 10 0 10 1
08:32 UT

08:27 UT 07:00 UT

100 Integration time / s

101

Integration time / s

Figure 3: Plots of modulation index (de the high-time resolution sampling at 78 centred on 08:32 UT on March 28 2005. 2.368 GHz coherent emission centred on

fined by equation (3) in Section 2.2) versus integration time for ms. Panel (a) shows the relationship for the 1.384 GHz sample Panel (b) shows the relationships derived from two intervals of 07:00 and 08:27 UT on April 08 2005.

70

-30

(a)
60 Stokes V / mJy Stokes I / mJy -20

(b)

50

-10

40 1.0 3.0 5.0 -0.4 V / I at 2.368 GHz -0.3 -0.2 -0.1 0.0 1.0 3.0 5.0 7.0 9.0

0 1.0 3.0 5.0 7.0 9.0

(c)

7.0

9.0

Mid-UT / h April 08 2005

Figure 4: The Stokes I and V flux densities from the 25 min integrations on April 08 2005 are plotted against the mid-UT of each integration. The description of the three panels is identical to that of the caption to Figure 2.


6

Publications of the Astronomical Society of Australia

3

The sp ectral structure in coherent emission

The IF output of the ATCA consisted of 13 indep endent frequency channels, each of 8 MHz width and spaced 8 MHz apart, p ermitting us to display the radio sp ectrum over a total width of 104 MHz. For this task we refer to the 25 min integrations of the Stokes V data in Figures 2b and 4b, confining our attention to the several integrations that comprise the most highly p olarised sections in these two figures. In this sp ectral analysis we can utilize b oth low and high time resolution integrations. The task UVSPEC in the MIRIAD software was used to compute the averaged Stokes V and I flux densities and their errors in each of the 8 MHz-wide channels for the selected integrations. These flux densities were then plotted against the mid-frequency of each channel.

3.1

The spectra on March 28

Figure 5 shows the sp ectral structure and its dynamic b ehaviour at 1.384 GHz over the seven 25 min integrations that form the highly p olarised event in Figure 2b. The high time resolution data is centred on the panel lab elled 08:28 UT; here we show b oth the Stokes I and V sp ectra. The Stokes I sp ectrum closely mimics the Stokes V in the remainder of the panels. The increase in Stokes V is within a few p ercent of that in Stokes I in the seven panels, with an average value of (V /I ) = 0.99 (weighted by variance-1 ). Thus, as far as we can determine, the emission is 100% circularly p olarised in the left-handed sense. The details from successive frames clearly suggest that bursts of highly p olarised emission drift relatively slowly through the total 104 MHz bandwidth at 0.7 MHz min-1 . If one fixes attention on the burst in the top left panel, we note that by 07:16 UT it has drifted along the frame to lower frequencies. By 07:44 UT, this burst has practically drifted b elow 1.34 GHz and another burst of 45 MHz to FWHM has app eared near the high frequency end of the band. At 08:28 UT this new burst is situated near the centre of the band and by 08:58 is approaching the low frequency end. By 09:26 UT it has reached the low frequency end of the band. The instantaneous bandwidth of the coherent emission is one of the critical parameters in any theoretical mechanism for its creation. The FWHM of the most completely delineated burst at 08:28 UT is 48 MHz, yielding a value of ( / )=0.035. This, however, is an upp er limit b ecause in the 22 min integration, from which the panel is constructed, the centre frequency of the emission should have drifted 15 MHz. Therefore, the true instantaneous FWHM is likely to b e 30 MHz, giving ( / ) 0.02. We note that Stokes V never falls completely to zero in any of the panels, b ecause there are always remnants of the preceding and following bursts occupying one or other of the ends of the 104 MHz bandwidth of the receiver. If the channel bandwidth could b e reduced to say 4 MHz, one might see that zero Stokes V is reached b etween bursts, although that would dep end on how regularly they are generated.

3.2

The spectra on April 08

Figure 6 shows the sp ectral structure at 2.368 GHz over the nine 25 min integrations that contribute to the highly p olarised event in Figure 4b. Data with high time resolution were recorded in the panels lab elled 07:05 UT and 08:33 UT; in the latter we show b oth the Stokes I and Stokes V data on the same intensity scale spacings. The Stokes I and V sp ectra are very similar in shap e and amplitude on all these scans. The larger error bars in the panel lab elled 08:18 UT are due to this integration having b een cut short in order to b egin the high time resolution observations that contribute to the next panel. The increase in Stokes V and I are very similar with an average value of (V /I ) = 0.99 (weighted by variance-1 ). Again, as for the observations of March 28, the emission is p olarised at close to 100% in the left-handed sense. The successive panels of Figure 6 show that although sp ectral structure is clearly present in the 2.368 GHz p olarised emission, it is difficult to assign a direction of drift. Unlike the drifting narrow-band bursts visible at 1.384 GHz in Figure 5, no individual burst can b e completely isolated. This is no doubt due to the much wider bandwidth occupied by the 2.386 GHz p olarized bursts. The FWHM bandwidth of these bursts is probably closer to twice our 104 MHz bandwidth, and one would need to b e able to trace the intensity structure over at least this frequency interval to decide on its direction of drift. One notes that the p olarised emission does not fall to near zero in any of these panels, indicating the presence of pronounced frequency overlapping of adjacent bursts.

3.3

Orbital phase dependence

VLBA observations of HR 1099 by Ransom et al. (2002) provide reasonable evidence that during times of high radio activity the magnetospheres of the binary comp onents had a high degree of interaction such that two compact 8.4 GHz sources were seen, the stronger source centred on the K2 sub-giant and the weaker on the G5 dwarf. If our highly p olarised coherent emission were to also originate near the G5 dwarf at times of enhanced activity, one may exp ect that the phase dep endencies of the two phenomena would b e similar. We have searched the literature to determine the ep ochs of other events of long-lasting coherent emission from HR 1099 (greater than one hour in duration and Stokes V 100%). Table 1 gives the result of phase binning the eight occurrences of such emission. Although the statistics are rather p oor, this table shows that there has b een a definite tendency for the coherent events to favour the orbital phase range 0.33­0.83 with a minor p eak in the


www.publish.csiro.au/journals/pasa

7

50 06:48 UT LH circular (mJy) LH circular (mJy) 40 30 20 10 0 50 40 30 20 10 0 50 40 30 20 10 0 1.34 1.36 1.38 1.40 1.42 1.44 Frequency / GHz 50 09:26 UT LH circular (mJy) 40 30 20 10 0 1.34 1.34 1.36 1.38 1.40 1.42 1.44 07:44 UT LH circular (mJy) Frequency / GHz LH circular (mJy) 1.34 1.36 1.38 1.40 1.42 1.44 07:16 UT LH circular (mJy) LH circular (mJy)

50 08:08 UT 40 30 20 10 0 50 40 30 20 10 0 50 08:58 UT 40 30 20 10 0 1.34 1.36 1.38 1.40 1.42 1.44 Frequency / GHz 1.34 1.36 08:28 UT 1.38 1.40 1.42 1.44 80 70

V

60 I mJy 50

I

40 30

1.36 1.38 1.40 Frequency / GHz

1.42

1.44

Figure 5: The spectral structure in the highly polarised 1.384 GHz data of March 28 across the 104 MHz bandwidth of the compact array. The panels are arranged in increasing values of mid-UT of the 25 min integrations making up the data. Six of the 7 panels show the structure in Stokes V only, but the panel at 08:28 UT is derived from the high-time resolution sampling and shows both the Stokes I and V spectral structure. The frequency channel at 1.41 GHz is missing due to an interference problem. The error in the flux measurements are similar to that shown in Figure 6.


8

Publications of the Astronomical Society of Australia

32 28 LH circular (mJy) 24 20 16 12 8 32 LH circular ( mJy) 28 24 20 16 12 8 32 28 LH circular (mJy) 24 20 16 12 8 32 28 LH circular (mJy) 24 20 16 12 8 2.32 2.34 2.36 2.38 2.40 2.42 Frequency / GHz 32 28 LH circular (mJy) 24 20 16 12 8 2.32 2.34 2.36 2.32 2.34 2.36 2.38 2.40 2.42 07:31 UT Frequency / GHz 07:05 UT LH circular (mJy) 2.32 2.34 2.36 2.38 2.40 2.42 Frequency / GHz06:32 UT LH circular (mJy) 06:05 UT LH circular (mJy) 05:21 UT LH circular (mJy)

32 28 24 20 16 12 8 32 28 24 20 16 12 8 2.32 32 28 24 20 16 12 8 32 28 24 20 16 12 8 2.32 2.34 2.36 2.38 2.40 2.42 Frequency / GHz 09:00 UT 2.32 2.34 2.36 2.38 2.40 2.42 82 Frequency / GHz 08:33 UT 78 2.34 2.36 2.38 2.40 2.42 Frequency / GHz 08:18 UT 2.32 2.34 2.36 2.38 2.40 2.42 Frequency / GHz 07:59 UT 07:31 UT

V

I mJy
70 66 62 58

74

I

2.38

2.40

2.42

Frequency / GHz

Figure 6: The spectral structure in the highly polarised 2.368 GHz data of April bandwidth of the compact array. The panels are arranged in increasing values of integrations making up the data. Eight of the nine panels show the structure in panel at 08:33 UT is derived from high-time resolution sampling and shows the sp Stokes I and V.

0 8 a cr o s s mid-UT Stokes V ectral var

the 104 MHz of the 25 min only, but the iation in both


www.publish.csiro.au/journals/pasa

9 Phase 0.0­0.17 0.17­0.3 0.34­0.5 0.50­0.6 0.67­0.8 0.83­1.0 No. of Events 0 1 2 3 2 0 Ref (1 (1 (1 (2 ) ,2) ,2,3) ,3)

3 0 7 3 0

Table 1: The phase dependance of the ECME events. We have used the ephemeris of Fekel (1983) to determine the orbital phase of the events. References: (1) Osten et al (2002), (2) Jones et al (1993), (3) this paper.

phase range 0.50­0.67. Ransom et al. (2002) used the same orbital ephemeris to derive the phase range over which they observed the 8.4 GHz source near the G5 dwarf to move from 0.72­0.76. The secondary source was already at full strength at the start of their observations, so the phase range could b e credibly extrap olated to earlier phases to coincide with the p eak in the coherent emission in Table 1. This coincidence in phase range will b e discussed in Section 5.

4

Summary of observational results
1. Highly circularly p olarised emission ( 100% left-hand) was seen in two observing sessions separated by 11 days. 2. The intervals of strong, highly p olarised emission lasted from 2­3 hrs (Figures 2 & 4). 3. In the first observing ep och, the highly p olarised emission was seen only at 1.384 GHz (Figure 2). 4. During the second ep och observation, a strong, highly p olarized event was observed at 2.368 GHz, with weaker, highly p olarised 1.384 GHz emission partially overlapping the stronger 2.368 GHz emission (Figure 4). 5. Three 22 min integrations were made at high time resolution (0.078 s), enabling us to show (Figure 3) that the modulation index of the Stokes V intensity increased as the integration time was reduced and was still increasing at our resolution limit. 6. During the first ep och event, coherent bursts with FWHM width of 48 MHz The second ep och of coherent emission but its significantly wider FWHM did n frequency drift. emission at 1.384 GHz contained a regular frequency structure of that drifted across the sp ectrum at 0.7 MHz min-1 (Figure 5) . at 2.368 GHz also contained definite sp ectral structure (Figure 6), ot p ermit an accurate estimate of its bandwidth nor its direction of

The following observational conclusions can b e drawn from our data:

7. The two long-lasting coherent events rep orted in this pap er conform with six others rep orted in the literature, occurring preferentially in the orbital phase range of 0.50­0.67.

5

Discussion

It is well known that coherent radio emission is emitted from a variety of celestial ob jects ranging from the Sun, Earth, Jupiter and Saturn to flare stars and some close binaries. In the case of the Sun (Melrose & Dulk 1982), in the flare star AD Leo (Lang et al. 1983), and in the flare star YZ CMi (Lang & Wilson 1988), high time resolution measurements have shown temp oral structure of tens of millisec, which indicates from a light-travel time argument that the linear size of the emitting region should b e only a few times 3 â 108 cm. First, we check the p eak brightness temp eratures reached by these highly p olarised bursts, using a convenient formula due to Osten and Bastian (2007): TB = 6 â 1014 S t
ms mJ y

(Dpc / GH z t

ms

)

2

(4)

Where SmJ y is the p eak p olarised flux density, Dpc is the stellar distance, (29 p c) GHz is the frequency and is the burst duration, here taken as the time resolution of the sampling (78 ms). In the first ep och observation at 1.384 GHz, the maximum p olarised flux of 48 mJy (see Figure 5) results in TB > 2.1 â 1015 K. In the second ep och observation at 2.368 GHz, the maximum p olarised flux density of > 32 mJy (see Figure 6) yields TB > 4.7 â 1014 K. If the temp oral structure in these bursts is as low as 5­10 ms, the brightness temp eratures would b e two orders of magnitude higher. It is clear that this completely p olarised


10

Publications of the Astronomical Society of Australia

emission can not b e due to an incoherent mechanism such as that producing thermal emission or gyro-synchrotron radiation. Next, it is imp ortant to compare the total time intervals of coherent emission on a M4 V flare star (AD Leo) and the close binary HR 1099 of sp ectral typ e K3 IV + G5 V. Our Figures 2 and 3 show that the events on HR 1099 last for 2­4 hrs as compared to 1 min for the flares rep orted by Osten and Bastian (2006, 2007). One questions whether this more than two orders of magnitude difference can b e compatible with the same mechanisms op erating in sources with similar linear sizes, coronal electron densities and magnetic fields in two very dissimilar stellar systems. Perhaps the analysis of Osten and Bastian (2006) in their Figure 6 of a moderate intensity, 60 s flare on AD Leo in the frequency range 1162­1568 MHz is the most suitable to compare with the results for HR 1099 in our Figure 5. Osten and Bastian have plotted their results with a degraded time resolution of 200 ms and a degraded frequency resolution of 7.8 MHz. Our data are plotted with a time resolution of 78 ms and frequency resolution of 8.0 MHz. The imp ortant conclusions to come from this comparison are the frequency drift rates of the bursts and their instantaneous bandwidths. Osten and Bastian find a drift rate of 52 MHz s-1 , while for HR 1099 we measure a frequency drift rate of 0.012 MHz s-1 , i.e. more than 3 orders of magnitude slower, although the direction of drift is from high to low frequencies in b oth exp eriments. A second significant difference b etween the two sets of data is the instantaneous bandwidth of the bursts; in AD Leo, / > 0.29, while in HR 1099, / > 0.036 a difference of a factor of 8. The third outstanding difference b etween the ab ove sets of data is the difference in burst occurrence rates. In their Arecib o data of June 2003, Osten and Bastian (2006) detected only two short (60 s) bursts in 16 hr of data, spread over four days. In our HR 1099 data of March 28 2005, we see two long-lasting intervals of coherent emission, with the more intense episode consisting of two distinct bursts slowly traversing our 104 MHz band width. Our bursts may contain much finer structure that is smoothed out by our 78 ms time resolution, but its absence in our data makes a comparison with the higher resolution data of Osten and Bastian (2007) a less-fruitful exercise. Perhaps an even more relevant comparison may b e made b etween our data and the slowly varying decimetric coherent emission from the dMe flare star YZ CMi (Lang & Wilson 1988). The emission lasted for 5h and consisted of a numb er of discrete bursts, each of 10 min duration and 100% p olarised in the left-handed sense. Four of these bursts were shown to contain narrow-band structure with a fractional bandwidth of / =0.02, but their integration time of 10 s did not p ermit the authors to investigate their temp oral structure. The authors place an upp er limit of < 0.05 MHz s-1 on the frequency drift of these bursts, a value consistent with our first ep och value of 0.012 MHz s-1 . The alternative mechanisms of ECME and plasma emission are discussed, but no definite preference was given. Comparing the prop erties of the coherent emission which we have detected in HR 1099 with that from other stellar systems, we favour ECME as a more likely mechanism for our HR 1099 emission b ecause: (1) its 100% p olarisation is much higher than that achieved in most solar burst of Typ es I­V; (2) its duration is much longer than all solar bursts except Typ e IV; (3) its frequency-drift rate is orders of magnitude lower than solar bursts and the short flares from AD Leo rep orted by Osten and Bastian (2006, 2007); (4) its temp oral structure is much finer than that found in solar bursts, except p erhaps that in Typ e 1 solar noise storms and in the solar decimetric `spike' emission. Regarding the location of the coherent source in the close binary, we must consider at least two p ossibilities: (1) in the corona of either of the comp onents; (2) a coherent source resulting from the interaction b etween the magnetospheres of the comp onents. We shall consider b oth options b elow.

5.1

The single-star source of coherent emission

It is tempting to interpret our highly p olarised and highly time-structured microwave emission in terms of solar analogues, which have b een studied most extensively at metric and decimetric wavelengths and are describ ed comprehensively in the b ook `Solar Radiophysics' edited by McLean and Labrum (1985). Whether one can apply the models develop ed for solar bursts of Typ es I­V and their associated continuum emissions to a binary consisting of stars of differing sp ectral typ es is op en to serious question. The solar Typ e I noise storm has the long duration, high p olarisation and fine intensity structure that most resembles our emissions from HR 1099, but this is confined to metre waves ­ we do not, therefore consider this a likely cause for the events we have detected in HR 1099. The so-called solar `spike bursts', emitted in the frequency range 0.3­3 GHz, p ossess similar fine time and frequency structure to that contained in the coherent emission from HR1099 and are highly p olarised, suggesting that such a source in the corona of either of the binary comp onents is a distinct p ossibility. The most frequently discussed driver of solar spike bursts is an anisotropy in the pitch angle distribution ab out magnetic field lines, commonly called `the loss-cone', producing an electron-cyclotron maser. First, it must b e noted that one necessary condition for an electron cyclotron maser to op erate is that the source-region plasma has a relatively low plasma density and/or a relatively high magnetic field strength, such 1/2 that the ratio of the electron plasma frequency, p = 8.98 â 103 ne MHz to the electron-cyclotron frequency, nc = 2.80 H MHz, is small; here, H is the magnetic field strength in gauss and ne the electron density cm-3 . For p /c << 1, the highest maser growth rate is in the x-mode at the fundamental ( p ), but it is highly likely


www.publish.csiro.au/journals/pasa

11

that fundamental maser emission would b e reabsorb ed as it propagates into harmonic absorption bands in regions with lower magnetic field strengths. Consequently, the strongest maser emission likely to reach the observer is x-mode emission at the second harmonic (Melrose & Dulk 1982). However, without knowledge of the predominant magnetic field p olarity of the active region producing the radiation, the p olarisation mode (x-mode or o-mode) cannot b e ascertained from the observed handedness. Second-harmonic emission at 1.4 GHz corresp onds to a source region field of 250 G, and the condition p /c < 1 then requires source region plasma densities < 6 â 109 cm-3 . This plasma density requirement can b e met almost anywhere in the stellar corona ­ even at the low coronal height of 1.003 R the Baumbach-Allen model of the solar corona (Allen 1973) yields a plasma frequency of only 183 MHz, well b elow either of our observing frequencies of 1.384 and 2.368 GHz.. Plasma densities will, of course, b e enhanced in the converging magnetic field lines of coronal loops near their footprints. At this p oint, we may consider whether the source models develop ed to account for the observed auroral kilometric radiation (AKR) could b e of assistance in interpreting the coherent sources in HR 1099. The instructive reviews by Ergun et al. (2000), and Treumann (2006) show that AKR electron-cyclotron emission was measured by satellites that actually traversed the auroral source cavities, allowing detailed theoretical models to b e checked by measurements of AKR frequency, p olarisation mode, plasma frequencies, magnetic and electric field strengths and the temp oral and frequency structure of AKR. These authors find that, although a loss cone is generated, it is insufficient to amplify the AKR waves to levels that could account for the high-brightness, fine structured sources within the cavity. They prop ose an alternative driver, consisting of an unstable `horse-shoe' or `shell' distribution of fast electron velocities with resp ect to converging field lines in the presence of a parallel electric field and a generally low plasma density, referring to this source region as the `auroral cavity'. In this case, the direction of the earth's field is known with resp ect to the satellite's path through the region and so the p olarisation mode can b e identified. The horse-shoe distribution is apparently a much more efficient way of increasing the growth rate of AKR, which is preferentially emitted p erp endicular to the magnetic field lines in the right-hand circularly p olarized x-mode. Just as imp ortantly, the fundamental of the electron-cyclotron frequency can escap e the cavity, yielding a much stronger observed intensity of AKR than would have b een observed with the loss-cone model. Other characteristics of the AKR that mimic the coherent emission from HR 1099 include its fine temp oral and frequency structure. Bursts of AKR with fractional bandwidth as low as 0.01 and temp oral structure of 100 ms can drift through the sp ectrum at various rates, but a good example is shown in Treumann's Figure 6 . This illustrates a consistent drift from high to lower frequencies of 7­8 kHz s-1 , comparing well with the drift rate of the coherent event in our Figure 5 of 12 kHz s-1 from high to lower frequencies. However, we should b e cautious in applying the auroral model to stellar coronae, where we are dealing with magnetic fields and ambient plasma densities that are orders of magnitude higher than auroral values. Nevertheless, the aforementioned similarities suggest that this model should b e explored in detail for stellar coherent emission. It is feasible that source regions having aurora-like general structure could b e found in the converging magnetic field lines of coronal loops, but detailed modelling using realistic field strengths and plasma densities is required to validate this prop osal.

5.2

The two-star coherent source

If the observed coherent radiation from HR 1099 is due to the interaction b etween the magnetospheres of its comp onents it would seem that the more active K2 subgiant could p ossess a strong stellar wind capable of transferring a relatively dense plasma with imb edded magnetic field to the corona of the G5 dwarf. Ransom et al's (2002) VLBA maps show that a relatively weaker 8.4 GHz gyro-synchrotron source was detected near the p osition of the G5 star only when the system was in a state of high activity. A mapping of four equal sub divisions of the data resulted in this source moving its p osition in a manner consistent with the orbital motion of the G5 dwarf ab out the K2 subgiant. We have noted in section 3.3 that the orbital phases at which the published coherent events were detected app ear to favour an orbital phase range falling within that deduced from the p ositions of the secondary 8.4 GHz source. The fact that no 8.4 GHz coherent emission was detected from either star during this observation by Ransom et al. does not rule out its presence at considerably lower frequencies. All occurrences of coherent radiation from HR 1099 have b een detected at frequencies b elow 5 GHz, usually at ab out 1.4­2.4 GHz, where most lower frequency observations have b een made. Gyro-synchrotron emission of comparable strength often coexists with the coherent emission as our Figures 2 and 4 demonstrate, but, although this does not mean that the two radiation typ es necessarily emanate from the same source region in the corona, they could have a common exciting agent. Considering the geometry and phase dep endence of the coherent events as compared with that of Ransom et al's secondary source, we prop ose that an auroral analogue may op erate, in which comparatively dense plasma is transferred from the stellar wind of the K2 subgiant into the p oloidal field of its companion. The dwarf 's magnetic p oles would likely b e visible, due to the low inclination (38 degrees) of the orbital plane to the sky. Field strengths near these p oles would b e considerably higher than for the Sun, due to the much more rapid rotation rate. However, we stress again that a thorough modelling of the situation needs to b e done b efore the auroral analogue can b e established.


12

Publications of the Astronomical Society of Australia

5.3

Slow changes in the intensity and spectral distribution

Figures 2 and 4 demonstrate that b oth intensity and sp ectral parameters vary over an interval of nine hours and from day to day. In March 2005 (Figure 2) the coherent event was confined to 1.4 GHz and varied in intensity by a factor of ten b etween the two p eaks separated by at least 7 h. Our p oor angular resolution does not p ermit us to decide whether the two p eaks are due to separate coronal sources or whether the plasma density and/or magnetic field strength varies within the one maser source. A VLBI network with an angular resolution of one milliarcsec at 1.4 GHz will eventually b e able to decide b etween these alternative interpretations. Figure 4 (for April 2005) illustrates a phenomenon that has not b een rep orted hitherto. Here we observe that coherent radiation is present simultaneously at widely separated frequencies. At 1.4 GHz its intensity drops continuously to almost zero over the 8 h observation, while at 2.4 GHz its intensity rises continuously to a p eak near the end of the observation; there is little, if any, correlation b etween the intensities at the two frequencies, suggesting that two separate coherent sources are op erating indep endently and with differing plasma densities and/or magnetic field strengths. In order to resolve this dilemma, one is faced with the task of simultaneously op erating a VLBA network at two widely separated frequencies to achieve an angular resolution of one milliarcsec at the lower frequency. Such angular resolutions could b e achieved at these decimetric wavelengths by a satellite op erating with a network of earthb ound radio telescop es.

Acknowledgments
We thank Dr. Mark Wieringa for his modifications to the VIS software that enabled the on-line display of b oth the Stokes V and I intensities. Dr. Vincent McIntyre created software that was essential to the analysis of the high-time resolution data. Prof. E. Budding's useful comments are appreciated, as were those of Dr. J. Caswell. The insightful comments of the referee resulted in considerable improvements to the content of the pap er. The Australia Telescop e Compact Array is part of the Australia Telescop e, which is funded by the Commonwealth of Australia for op eration as a National facility managed by CSIRO.

6

References

Allen, C..W. 1973, Astrophysical Quantities, 3rd edition, p176, The Athlone Press Dulk, G.A., 1985, Ann. Rev. Astron. Astrophys., 23, 169 Ergun, R.E., et al. 2000, ApJ, 538, 456 Hallinan, G., et al. 2007, ApJ, 663, L25 Jones, K.L., Brown, A., Stewart, R.T., Slee, O.B. 1996, MNRAS, 283, 1331 Fekel, F.C., 1983, ApJ, 268, 274 Lang, K.R, Wilson, R.F. 1988, ApJ, 326, 300 Lang, K.R. et al. 1983, ApJ (Letters), 272, L15 Osten, R.A. et al. 2004, ApJS, 153, 317 Osten, R.A., Bastian, T.S. 2006, ApJ, 1016 Osten, R.A., Bastian, T.S.. 2007, astro-phys, arXiv:0710.5881 Melrose, D.B., Dulk, G.A. 1982, Ap.J, 259, 844 McLean, D.J., Labrum, N.R. eds, 1985, Solar Radiophysics, CUP Ramsay, G. et al. 2007, MNRAS, 382, 461 Ransom, R.R. et al. 2002, ApJ, 572, 487 Slee, O.B., Willes, A.J., Robinson, R.D. 2003, PASA, 20, 257 Treumann, R.A. 2006, Astron Astrophys. Rev., 13, 229 White, S.M., Franciosini, E. 1995, ApJ, 444, 342 Willes, A.J., Wu, K. 2004, MNRAS, 348, 285 Wu, K., Cropp er, M., Ramsay, G., Sekiguchi, K. 2002, MNRAS, 331, 221