3 RXTE results

3.1 Long term behavior of the source in 1999

To follow the flux history of V4641 Sgr in 1999 we used publicly available data of RXTE/PCA scans over the Galactic Center region, that were performed almost bi-weekly during the whole year. The statistical significance of the data and the accuracy of the background subtraction allow us to detect any source down to the level of approximately 1-2 mCrab (if the Galactic diffuse emission do not contribute much to the detected X-ray flux at the position of the source). The method of Galactic Center map construction and the extraction of the source flux is described in Revnivtsev & Sunyaev (2002).

For the first time the V4641 Sgr was statistically significantly detected in PCA scan on Feb. 18, 1999 (Markwardt et al. 1999) and since then it was seen in almost every scan. We present the obtained light curve of V4641 Sgr in Fig. 1. The results of our analysis (61 data points) are in agreement with the 7 data points reported in Markwardt et al. (1999) and Markwardt et al. (1999).

After almost 6 months of moderate X-ray activity, during the first half of Sep. 1999 V4641 Sgr demonstrated several X-ray outbursts. The first, the weak one, was detected by BeppoSAX and RXTE/ASM on Sep. 10.1, 1999. The source X-ray flux reached $\sim$300 mCrab (in't Zand et al. 2000). On Sep. 14-15, 1999 three more powerful flares were detected. The segment of the light curve of V4641 Sgr around Sep. 14-15, 1999 is presented in Fig. 1 (lower panel). It is seen that the source rose up to 4, 12 and 2 Crabs on Sep. 14.9, Sep. 15.7 and Sep. 15.9 respectively. The observed large changes in X-ray flux (factor of 10 at least) occurred on the time scales of one-two hours. The X-ray luminosities of the source in the energy band 1-12 keV during these outbursts could be estimated to be $\sim1\times 10^{39}$ ergs/s, $\sim3.4\times 10^{39}$ and $\sim7\times 10^{38}$ ergs/s respectively (with adopted distance to V4641 Sgr $\sim9.5$kpc, 2001).

The last X-ray flare finished on $\sim$Sep. 15.95, 1999 by the rapid drop in X-ray flux from the level of $\sim$100 mCrabs by a factor of 10 and after this the source returned to the quiescent state - $\la$1 mCrab. Note, that at such low flux levels some contamination from the Galactic diffuse emission is possible. Our estimates of the contribution of the Galactic diffuse emission at the position of V4641 Sgr showed that it is small, but not negligible. As a very conservative estimation of the source flux the detected 1 mCrab flux at the position of V4641 Sgr should be treated as the upper limit.

3.2 Orbital modualtion of the X-ray flux

In spite of the fact that the significance of the source detection in the RXTE/ASM data is lower than in the RXTE/PCA data (during one day of observations), ASM points have advantage from the point of view of their quasi-uniform coverage over the period of Feb.-Sep., 1999. This helps us to search for the source period in X-ray data. Periodicities of V4641 Sgr were sought by means of the Lomb-Scargle periodograms (Lomb 1976; Scargle 1982; Press et al. 1992). In Fig. 2 we present the Lomb-Scarge periodogram obtained for V4641 Sgr lightcurve taken in the period of steady state activity $\sim$Mar. 13- Aug. 20, 1999 ($\sim$TJD 11250-11410)

Figure 2: The Lomb-Scargle periodogram of the lightcurve of V4641 Sgr , during period of moderate source activity, Mar. 13-Aug. 20, 1999. The value of the optical period is marked by an arrow.

It is seen that almost at the position of detected optical period P=2.8173 days (Orosz et al. 2001) a peak at the X-ray Lomb-Scarge periodogram is present. This peak corresponds to the period $P_{\rm x}=2.84\pm0.03$ days. The uncertainty of the period value was estimated by a Monte-Carlo bootstrap method, assuming the gaussian distribution of values of ASM lightcurve points. The false-alarm probability of this detection (taking into account the number of trial periods) is slightly less than 10^-3, if we assume the exponential distribution of Lomb-Scargle power values. It is not very high significance to rely on X-ray data alone. However, the marginally detected X-ray periodicity has the value of the period that coincides within 1-$\sigma$ errors with the firmly detected optical one P=2.8173 days. This strongly supports the detection of the binary period in X-rays.

Figure 3: The X-ray lightcurve of V4641 Sgr in the energy band 1.5-3.0 keV folded with the optical photometric period P=2.8173 days. The reference time T0 taken from the paper of Orosz et al. (2001). The phase 0 corresponds to time when the normal star is located between the black hole and observer.

We have folded the lightcurve of V4641 Sgr with the measured optical period of the system in order to search for the orbital dependence of the X-ray flux. The obtained orbital profile of the X-ray flux (1.5-3.0 keV, lowest ASM energy channel) is presented in Fig. 3. In order to make the comparison of optical and X-ray folded lightcurves easier we have used the same reference time T₀ as Orosz et al. (2001) in their Fig. 3. It is seen that the folded X-ray lightcurve demonstrates peak, when the black hole is located between the observer and the optical star, and the minimum - when the star is located between the observer and the black hole. We also detected strong dependence of the amplitude of orbital modulations of X-ray flux on the photon energies: for the lowest energies, 1.5-3.0 keV, the amplitude of the sinusoidal variations is $36\pm8$%, in the energy band 3-5 keV - $25\pm5$%, and in the energy band 5-12 keV the modulations is undetectable with an upper limit <15%(2$\sigma$). It should be noted, that the observed energy dependence of the X-ray orbital modulations and the position of the X-ray minimum on this modulation suggest that the detected X-ray variations could be caused by an absorption in the line of sight near the optical star. This suggests that the inclination of the system is close to 65-70$^{\circ}$, in agreement with the optical data (Orosz et al. 2001).

3.3 The spectral evolution of the source

3.3.1 Period of "quiescent'' activity in Feb.-Sep., 1999

The scan observations give us an important opportunity to follow the spectral shape of the source during a year. Unfortunately the acceptable "on-source'' exposure of a scan observation is only of the order of 10-20 s. Therefore the statistics in the obtained spectra is quite poor. During the period Feb.-Sep. 1999 the source was detected with quite soft spectrum - which could be roughly described by the model of bremsstrahlung emission with temperatures $kT\sim$2-3 keV or by multicolor disk model (Shakura & Sunyaev 1973) with the inner disk temperature $kT\sim1.5$ keV. The typical spectrum of V4641 Sgr at that time is presented in Fig. 4 (upper panel).

The spectrum of V4641 Sgr obtained by BeppoSAX observatory on Mar. 13, 1999 (see in't Zand et al. 2000) has the statistically significant emission line at the energy $\sim$7 keV with equivalent width $\sim$270 eV. Therefore we also searched for the emission line in the PCA scan data. Unfortunately, due to poor statistics, the emission line could not be detected in a single PCA scan observation with an upper limit on its equivalent width EW< 0.7-1.0 keV. However, a fit to spectrum of Sgr, averaged over the period Feb. 18-Sep. 02, 1999 gives an emission line at the energy $E_{\rm line}=6.60\pm0.08$ keV with the equivalent width $EW=360\pm90$ eV.

Figure 4: The spectrumn of V4641 Sgr averaged over period Mar.-beginning of Sep. 1999. The lower spectrum represents the set of spectra obtained during the period Sep. 6-Sep. 12, 1999.

The X-ray spectrum of the source strongly changed in the beginning of Sep. 1999 after the dip in the light curve (see Fig. 1). The photon spectral index hardens - to $\alpha\sim$1 - and the emission line became stronger - $EW\sim$ 1-2 keV (with typical uncertainty $\sigma_{EW}\sim$200-300 eV). The spectrum averaged over Sep. 6-12, 1999 is presented in Fig. 4 (lower panel). The spectrum resembles the one obtained by BeppoSAX on Sep. 10, 1999 (in't Zand et al. 2000). Note, that approximately simultaneously with the dramatic changes in the X-ray spectral properties of the source the increase of the source optical activity was detected (Kato et al. 1999).

3.3.2 Period of flaring activity (Sep. 14.8-15.9, 1999)

During the outburst activity on Sep. 14-15, 1999, the source demonstrated at least three powerful X-ray flares (Fig. 1), two of which were detected by the ASM instrument. ASM data indicated that during the flares the spectral hardness of the source was generally anticorrelated with the X-ray flux.

Between the two ASM flares, on $\sim$Sep. 14.9, the source flux dropped by a factor of 20 at least and became undetectable by the ASM. Fortunately, at this time, on Sep. 15.1, 1999 a PCA scan observation was performed, allowing us to investigate the spectrum of V4641 Sgr between the two flares. During this observation a remarkable spectrum was obtained (Fig. 5). The spectrum is dominated by the emission line at $E_{\rm line}=6.63\pm0.08$ keV with enormous equivalent width of $EW=2.4\pm0.3$ keV. Remarkably, this spectrum is very similar to an X-ray spectrum of SS433 (e.g. Margon 1984).

Figure 5: The spectrum of V4641 Sgr on Sep 15.1, 1999, between the two bright X-ray flares (see Fig. 1).

3.3.3 The pointed RXTE observation (Sep. 15.89-15.95, 1999)

The third of the detected bright flares occurred during the decaying part of the outburst and was missed by the ASM because of $\sim$1.5 hour gap between the ASM points. Coincidentally, it occurred during the pointed observation of RXTE. As it was previously mentioned by Markwardt et al. (1999) the spectrum of the source at the peak of the X-ray light curve during this RXTE observation was quite hard and resembled the typical spectra of the black holes in the low/hard spectral state with the 3-50 keV photon index $\alpha <$2.0, the cutoff at the energies 100-200 keV, and pronounced fluorescent Fe line at 6.4 keV.

Figure 6: The spectral evolution of V4641 Sgr during the first 1500 s of the RXTE pointed observation. The solid lines in the lower panel represented the intervals used for the accumulation of the broad band spectra of V4641 Sgr (see Fig. 7).

However, time resolved spectral analysis of the RXTE/PCA data revealed significant quantitative and qualitative evolution of the source spectrum during $\approx$1500 s of the pointed RXTE observation. In Fig. 6 we present the light curve of the source with 16 s time resolution, softness ratio (3-5 keV to 15-20 keV), the centroid energy of the Gaussian line and its equivalent width as a function of time. The position of the line and its equivalent width were determined using a simple power law + Gaussian line approximation of the spectrum in the 3-12 keV energy band. In the uppermost panel we also show behavior of the integrated fractional rms (in percents) of the source flux variations (3-20 keV energy band, 0.5-10 Hz frequency range). In order to illustrate the spectral evolution of V4641 Sgr we show in Fig. 7 the broad band spectra accumulated during four intervals, marked in Fig. 6 (lower panel) by the horizontal lines.

As is apparent from Figs. 6 and 7 the RXTE observation can be divided into two parts with qualitatively different spectral properties with the boundary at $t\sim 1100$ s (in the units of Fig. 6) corresponding to the final drop of the X-ray flux and change of the iron line energy from $\approx$6.4 keV to $\approx$6.6-6.8 keV.

During the first part the source had a strong emission line centered at $\sim$6.4 keV with the equivalent width varying between 200 and 900 eV, sufficiently hard spectrum extending to the hard X-ray energies and significant aperiodic variability with fractional rms $\sim$15-40%. In order to qualitatively illustrate the character of the spectral evolution, we show in Fig. 8 three spectra, accumulated during individual 16 s intervals, corresponding to significantly different values of the line equivalent width and fractional rms.

The Figs. 6 and 8 demonstrate that the spectral evolution can be qualitatively understood as a result of absorption/reprocession in the extended medium with varying absorption column density. Indeed, assuming that the primary spectrum does not change significantly, decrease of the absorption column density would lead to the apparent softening of the outgoing spectrum and decrease of the equivalent width of the fluorescent iron line. If the absorbing/reprocessing medium has a significant spatial extend with the light crossing time of $\sim$10-50 s the variations of the primary emission would be smeared out in the reprocessed emission, the effect depending on the fraction of the scattered/reprocessed emission in the outgoing radiation. Thus, decrease of the absorption column density would lead to increase of the apparent fractional rms. It should be noted, that absorption by the neutral medium with solar element abundances does not adequately explain the observed spectra - certain ionization of the absorbing gas is required by the data. The maximal $N_{\rm H}L$value should be of the order of $\sim{\rm few}~10^{23}~{\rm cm}^{-2}$.

We have not found any strong soft component in the spectrum during this part of the observation. However, this issue is rather complicated taking into account the absence of the spectral data at energies lower than 3 keV.

Figure 7: The spectra of V4641 Sgr, accumulated over time intervals, shown in Fig. 6. The change in the hardness and in the strength of the fluorescent line is clearly seen.

Figure 8: Three spectra of V4641 Sgr obtained at different times. Open circles represent the spectrum obtained at $\sim$120th s (in time units used if Fig. 6), filled circles - at $\sim$220th s, and crosses - at $\sim$420th s, scaled down by factor of 2.5 to match the other two at energies higher than $\sim$20 keV. All spectra were accumulated over 16 s time intervals.

After the final drop of the X-ray flux at $t\sim 1100$ s a significant softening of the spectrum occurred and the line shifted from $\approx$6.4 keV to $\approx$6.6-6.8 keV indicating a significant change of the emission regime.

At the maximum of X-ray lightcurve, when the absorption was presumably weak the source had ordinary hard spectrum, typical for black holes in the low spectral state (Figs. 7 and 9). Remarkably, V4641 Sgr have demonstrated this type of spectrum while it's luminosity exceeded by $\sim$10 times that of Cyg X-1 in the hard and, probably by $\sim$3-5 times in the soft spectral states. Let us mention that masses of black holes in both systems are comparable. Therefore the difference in luminosity might give an information about the accretion rate.

Figure 9: Comparison of the spectra of V4641 Sgr , accumulated during the peak of the observed light curve (Fig. 6) and Cyg X-1 in the hard state. The distance to Cyg X-1 was assumed to equal to 2.5 kpc.

3.4 Short-term variability

The only data suitable to study short term variability of V4641 Sgr during the period of flaring activity in Sep. 1999 are that of the pointed RXTE observation discussed in the previous subsection. The brief description of the source variability during this observation can be found in Wijnands & van der Klis (2000). During the first $\sim$1100 s of the observation the V4641 Sgr was found to be strongly variable (rms amplitude $\sim$50%) with a power law power spectrum in the 10^-2-5 Hz frequency range. At the frequency of $\sim$5 Hz the power spectrum changed from $P\sim f^{-1}$ to $P\sim f^{-2}$. It is interesting to note that in spite of similarity of the spectral properties of V4641 Sgr with the spectral properties of Cyg X-1 in the hard state the power spectrum of V4641 Sgr flux variability is more similar to that of Cyg X-1 in the soft state.

We discuss below change of the aperiodic variability properties with time, photon energy dependence of the fractional rms and time delay of the reflected emission.

As was shown in the previous subsection, during the RXTE observation the source demonstrated a strong spectral evolution accompanied with significant change of the fractional rms (Fig. 6). Similar to the spectral properties, the variability level changed significantly after the rapid drop of the X-ray flux at $t\sim$ 1100th s (Fig. 6). The fractional rms droped from $\sim$30-40% to the level, undetectable with PCA, with the $2\sigma$upper limit of $\sim$1-2% in the $5\times10^{-3}$-10 Hz frequency band.

A notable feature of the energy dependence of the fractional rms is the decrease near the energy of the Fe K$_{\alpha}$ line. Note that some indications on such behavior could be noticed in Fig. 3 of Wijnands & van der Klis (2000). However the authors concentrated on the properties averaged the entire outburst. In Fig. 10 we present the dependence of rms amplitude of the source variability calculated for two different periods - intervals <300th s in the time units of Fig. 6 (high EW of the line) and $\sim400$-700th s (low EW of the line). One can see that there exist a definite dip approximately at the position of the Fe 6.4 keV fluorescent line. Moreover, it is seen that the stronger the line in the source's energy spectrum, the stronger the dip at the rms-energy dependence. The presented rms-energy dependence clearly demonstrate that the fluorescent Fe line is less variable than the continuum. The simplest interpretation of this fact could be the smearing of the reprocessed emission (in particular the flux in the Fe fluorescent line) because of the finite light crossing time of the reprocessing medium (see e.g. discussion in Revnivtsev et al. 1999, or Gilfanov et al. 2000).

Figure 10: The dependence of the rms amplitude of X-ray variability of V4641 Sgr on the photon energy during the episode of low equivalent width of the Fe line during two time intervals - <300th s (lower points) and $\sim$350-700th s (upper points) in the units of Fig. 6. The dashed line are the splines for the clarity. Dips at the place of Fe line mean that the line flux is less variable than the continuum flux.

Figure 11: The crosscorrelation between the continuum X-ray flux from V4641 Sgr and the flux in the Fe fluorescent line, calculated over period $\sim$350-700 s in the time units of Fig. 6. The error bars were estimated using Monte Carlo bootstrap method. The positive values of delay correspond to the Fe line flux being delayed with respect to the continuum.

The reprocessing in medium of large light crossing time can also result in the time delay between the direct continuum emission and the photons of the reprocessed spectrum, in particular - the Fe line photons. In order to check this hypothesis we have crosscorrelate the continuum flux of V4641 Sgr and the flux in the fluorescent Fe line. Flux in the Fe line was taken from the spectral approximation used for Fig. 6. In order to avoid possible contamination by long term trend in the parameters we used only data during the period with relatively stable value of equivalent width of the fluorescent line - from $\sim$350th to $\sim$700th s (in the time units of Fig. 6). The obtained crosscorrelation is presented in Fig. 11. The time interval of the acceptable data is rather short and we could not calculate the uncertainties on the crosscorrelation function directly from the data. Therefore we estimated the error bars using Monte Carlo bootstrap method. From Fig. 11 one can see that the crosscorelation function is strongly asymmetric with respect to the zero delay, implying that the flux in the Fe fluorescent line is delayed with respect to the continuum flux. The approximate time of the delay is $\tau\sim50$ s.