2 Data selection and analysis

2.1 RXTE

For our investigation of GRS 1915+105 we used public RXTE data from November 1996 to September 2000 provided by HEASARC. We selected Proportional Counter Array (PCA) and High Energy X-Ray Timing Experiment (HEXTE) data of 139 $\chi$-state observations from 89 different days. The selection of the datasets was based on the $\chi$-states defined by Belloni et al. (2000) and on PCA light curves provided by E. H. Morgan. Data with more than 30 min away from the South Atlantic Anomaly and without X1908+075 in the HEXTE background were selected. The lack of obvious variability in the light curve and its spectral hardness allows long continuous exposure times and therefore high signal-to-noise ratios. Typically, the spectra during this state show blackbody emission arising from an accretion disk and a dominating power law hard energy tail.

We reduced the RXTE data using the standard reduction script REX included in the HEAsoft5.04 package. We restricted the analysis to Standard 2 binned data of all layers of PCU0 of the PCA from 3-25 keV and HEXTE cluster 0 from 20-190 keV only. We used PCARSP 7.10 to produce a particular response matrix for each PCU0 dataset. In order to account for part of the uncertainties in the PCU0 instrument we added a systematic error of 1% as recommended (Remillard).

The X-ray spectral fitting was done using XSPEC 11.0 (Arnaud 1996). A consistent model should fit all RXTE $\chi$-state data. Therefore we tested several X-ray radiation models and finally selected a model consisting of (i) photoelectric absorption (WABS; Balucinska-Church & McGammon 1992), (ii) a spectrum from an accretion disk consisting of multiple blackbody components (DISKBB) and (iii) a power law spectrum reflected from an ionized relativistic accretion disk (REFSCH; Fabian et al. 1989; Magdziarz & Zdziarski 1995). Several attempts in the past to fit GRS 1915+105 spectra have shown complicated residuals in the soft X-ray band suggesting the existence of emission and/or absorption features near 6.4 keV, the energy of the Fe K$\alpha$ line (Kotani et al. 2000). Because of the low spectral resolution of $\sim$1 keV of the PCA at this energy an additional line fit gives no meaningful results. Therefore we ignored all energy bins from 4.5 to 8.5 keV while fitting our model. We fixed the hydrogen column density at $N_{\rm H}=5\times10^{22}$ cm^-2 as determined by Greiner et al. (1994) with ROSAT. It has been shown that the disk blackbody + power law assumption strongly overestimates the flux at lower energies compared to the thermal and non-thermal comptonization models used by Vilhu et al. (2001) and Zdziarski et al. (2001), respectively. This explains the smaller $N_{\rm H}$( $2{-}3\times10^{22}$ cm^-2) values they found for $\chi$-state observations. However, because we ignored the energy bins from 4.5 to 8.5 keV the amount of data bins needed to adjust the hydrogen column density was to small to let $N_{\rm H}$ be a free parameter.

We fixed 9 of the combined 16 model parameters (Table 1), leaving free the accretion disk temperature, $T_{\rm bb}$, and relative normalization, $K_{\rm bb}\propto R_{\rm in}^2$, the power law photon index, $\Gamma$, with $N\propto E^{-\Gamma}$ and relative normalization, $K_{\rm po}$ (photons/keV/cm²/s at 1 keV), the reflection index, R, the ionization parameter, $\xi$, and a factor to account for the relative normalization between PCU0 and HEXTE cluster 0. If not stated otherwise, we plot 1$\sigma$ errors for each parameter of interest.

 

 

Table 1: Fixed parameters of the spectral model.

fix. parameter	value
N$_{\rm H}$	$5\times 10^{22}$ cm-2
cutoff energy	no cutoff
redshift	0
Z>2 element abundances	1
iron abundance to abundance above	1
inclination angle	70$^{\circ}$
power law index for reflection emissivity	-2
inner disk radius	6 GM/c2
outer disk radius	1000 GM/c2

2.2 GBI and RT

The appearance of steady radio emission in $\chi$-states suggested us to search for correlations between RXTE data and radio data at 15 GHz (Ryle Telescope = RT) and 2.25 GHz (Green Bank Interferometer = GBI). GRS 1915+105 shows variability at all frequencies on time scales of seconds to hours. For a useful statement in $\chi$-states a suitably selection of corresponding datasets is therefore required.

According to the general interpretation that the radio emission is synchrotron emission from ejected plasma in sporadic or continuous jets (Fender et al. 1995) the radio flux should reach a maximum 15 min after the actual ejection (Mirabel et al. 1997) and therefore after a possible determining X-ray event.

Figure 1 shows the 15 GHz radio flux, $F_{\rm R}$, from the RT for JD 2450898-2450913 together with two RXTE observations. The RT observed GRS 1915+105 several times a day with five minute exposures.

Because the radio exposure is much shorter than the X-ray exposure ($\sim$hour) it is important to select radio data simultaneous with the RXTE observations. 37 of the 139 analyzed RXTE observations have simultaneous RT data and 9 have simultaneous GBI data.

Still, the selection of simultaneous radio observations is non-trivial. Occasionally the radio emission varies also during a single $\chi$-state X-ray observation, whereas no variability is seen in X-ray count rate and hardness ratio (5.2-60 keV/2-5.2 keV) (Fig. 2). But the variation of $F_{\rm R}$ during a $\chi$-state observation is negligible compared to the uncertainties of the individual radio measurements and to the variation between different RXTE observations. Therefore, the radio fluxes were averaged for each RXTE observation.

Figure 1: Variation of the 15 GHz flux (crosses) between JD 2450898 and JD 2450913. The dotted lines mark the time of RXTE observations from 29.03.1998 ( ${\rm JD} \sim 2450902$) and 06.04.1998 ( ${\rm JD} \sim 2450910$). GRS 1915+105 showed low radio emission until JD 2450900.5. The source was dominated by strong, variable radio emission following JD 2450905. No statement can be made about the time in between, because of the lack of radio observations during the RXTE observation from 29.03.1998. Also for the RXTE observation from 06.04.1998 the radio flux is uncertain because of the strong variation before and after the X-ray exposure.

Figure 2: 15 GHz light curve ( RT, upper pannel), 2-60 keV PCA light curve (middle pannel) and hardness ratio $\frac{5.2-60~{\rm keV}}{2-5.2~{\rm keV}}$ (lower pannel) of the $\chi$-state from 15.09.1998 (JD 2451071.9). While the radio light curve shows structured variability, none is seen in X-ray intensity and spectrum. The large black dot (marked for clarity with the dotted lines) in the upper panel represents the averaged radio flux over the RXTE observation used for further analysis.