2 Observations

Table 1 gives the journal of observations. Several images in U, B, V, R and I Bessel filters were taken together with three different H$\alpha$ interference filters ( $\Delta\lambda = 65$ Å; obs. A-E). Standard procedures were used for bias subtraction and flat-field correction. Figure 1 shows the B filter image of the field of RXJ0806.3+1527 together with the ROSAT HRI positional uncertainties. Down to the limiting R magnitude of $\sim$25.3 only one object was found consistent with the HRI error circles. This was at RA=08$^{\rm h}$06$^{\rm m}$22 $.\!\!^{\rm s}$9 and Dec=+15$^{\circ}$27$^\prime$31 $.\!\!^{\prime\prime}$0 (uncertainty of $\sim$0 $.\!\!^{\prime\prime}$5; equinox 2000), fully consistent with the previously identified blue object (I99). The U, B, V, R and I magnitudes were determined to be 19.6, 20.7, 21.1, 21.0 and 20.9, respectively (magnitude uncertainties of $\sim$0.1) on March 1999.

Time-resolved photometry (TRP) in the B band was first obtained with the ESO 3.6m telescope equipped with EFOSC2 (obs. F) over an interval of 1800s. Pulsations at the 321s X-ray period were detected, with a pulsed fraction (semiamplitude of modulation divided by the mean source magnitude) of 15 $\pm$ 4% (90% confidence level) confirming the correctness of the optical counterpart identification.

Based on this result, we observed again the source in the B, V and R bands with the 3.6m Telescopio Nazionale Galileo (TNG) equipped with DoLoReS (obs. G). TRP at a time resolution of 20s was obtained in each of the three filters ($\sim$2hr per filter) for a total duration of 7hr. Differential light curves were accumulated by subtracting the source magnitudes with the mean magnitude of 10 reference stars within the field of view. Figure 2 (upper panel) shows the whole light curve of RXJ0806.3+1527 normalised, as an example, to one of the reference stars. A best period of 321.5 $\pm$ 0.3s was obtained by fitting the phases of the modulation obtained over 6 different intervals of $\sim$4000s exposure each. This value is fully consistent with that detected in X-rays (321.25 $\pm$ 0.25s), to within $\sim$1 part in 10³. The shape of the B, V and R modulations could be well fit by the sum of two sinusoids (fundamental plus 2nd harmonic; see Fig. 2). Pulsed fractions of 13.9 $\pm$ 0.5%, 14.2 $\pm$ 0.6% and 13.2 $\pm$ 0.6% (90% confidence level) were determined for the B, V and R band, respectively. In order to search for additional flux modulations up to periods of hours, we merged the B, V and R light curves, by normalising their average flux to the average B band flux and calculated a power spectrum (see Fig. 2); no significant signal was detected (99% confidence level upper limit of 1.5%) in the 1min-5hr period range, other than that at 321s. Aperiodic flickering is apparent in the optical light curve.

Figure 2: TNG DoLoReS B (from start to 2.1 hr), V (2.1-4.5 hr) and R (4.5 hr to the end) merged light curve for the optical counterpart of RXJ0806.3+1527 (upper panel). Power spectrum density with superimposed the 99% confidence level threshold for signals (lower left panel). Merged B, V and R light curve folded to the best period of 321.5s, with superimposed the best fit (lower right panel).

Several ESO Very Large Telescope (VLT) medium resolution (6 Å) spectra were obtained with 30min exposures in the blue band with FORS1+2 (see Rupprecht & Böhnhardt 2000 for instrument description; obs. B and H). Each spectrum was analysed independently. The summed and normalised spectrum obtained during obs. H is shown in Fig. 3 (bias subtracted, flat-field corrected and calibrated in wavelength). No significant absorption features were detected, while several faint (equivalent width, EW, of $\sim$-1 $\div$ -6 Å) and broad (full width half maximum, FWHM, of $\sim$20-30 Å) emission lines are apparent. The lines at 5411 Å, 4541 Å, 4199 Å, 4025 Å and 3923 Å are unambiguously identified with HeII Pickering lines. The lines at 6560 Å, 4859 Å, 4338 Å, 4100 Å and 3968 Å correspond to the even terms of this series. NIII/CIII emission lines around 4640 Å and 5270 Å are also detected testifying that recombination processes are occurring in the system. Intensity changes (up to a factor of $\sim$2 in EW) and shape of the emission lines were detected from night to night and even within the same night.

Figure 3: VLT FORS1 medium (6 Å; 3800-6000 Å) and low (30 Å; above 6000 Å) resolution spectra obtained for the optical counterpart of RXJ0806.3+1527. Numerous faint emission lines of HeI and HeII (blended with H) are labeled.

VLT low resolution (30 Å) spectra were also obtained to sample the red part of the spectrum (obs. D and I; exposure times of 15-30min). The low and medium resolution spectra collected in 2001 were flux-calibrated using the spectrum of the V = 13.56 sdB spectrophotometric standard GD108. The calibrated flux is dominated by a steep blue continuum with no intrinsic absorption lines (see Fig. 4). The shape of the optical continuum is thermal and implies a temperature of $T_{\rm bb} > 4 \times 10^4$K together with negligible reddening of $E(B-V) \leq$ 0.01. Its normalisation is such that its blackbody emitting radius is $R_{\rm bb} \sim 570 (d/100$pc $)(T_{\rm bb}/4\times 10^4$K)^-1/2km, with d the source distance.

Spectral information at higher photon energies was obtained by analysing the Extreme Ultraviolet Explorer (EUVE) survey and the RASS data. The EUVE upper limits were extrapolated from data taken from on-line a strophysical databases. The relevant archival RASS data were retrieved and photon arrival times extracted within a 2$^\prime$ radius region centered on the peak of the emission. Photons were also extracted from a nearby region, far from other detected sources, so as to have a similar background level. The source ROSAT PSPC Pulse Hight Analyser (PHA) rates were grouped so as to contain a minimum of 10 photons per energy bin (after background subtraction), resulting in 4 statistically independent energy bins. A blackbody model gave a good fit (by using a maximum likelihood and C statistics) with a characteristic temperature of $T_{\rm bb}\sim(6\pm^{13}_{4})\times10^{5}$K (90% confidence level) and an absorption column of $1.9 \times 10^{20}$ cm^-2 (however the 90% uncertainty includes all values <6 $\times$ 10²⁰ cm^-2) and a 0.1-2.0keV unabsorbed flux of $\sim$ $4\times10^{-10}$ergcm^-2s^-1. We also inferred an unabsorbed flux of $\sim$ $5\times10^{-9}$ergcm^-2s^-1 at the peak of the emission of the 321s modulation. By combining the X-ray data point and EUVE upper limits with the optical continuum from the VLT observations, we obtained a best fit blackbody spectrum for a temperature of $T\sim 2.6 \times$ 10⁵K. This value, however should be treated with caution as it depends crucially on the assumption that the spectrum of RXJ0806.3+1527 from the optical to the X-rays is due to a single blackbody emission component. We note that the vastly different amplitude of the 321s modulation in the optical and the X-rays argues against this possibility.

Figure 4: VLT FORS1 flux-calibrated medium and low resolution spectra with superimposed, from the left to the right, the  U, B, V, R and I photometric data.

Finally, we used the 1951 and 1991 Palomar Observatory Sky Survey plates, digitized to produce the Guide Star Catalogue-II, in conjunction with the ESO frames to put an upper limit of 0.02 $^{\prime\prime}$yr^-1 on the proper motion of this object. Assuming that we are seeing only the reflex peculiar motion of the sun, this would put the object at a distance of $\geq$100pc.