2 The X-ray properties of 1ES 1927+654

Figure 1: ROSAT All-Sky Survey light curve of 1ES 1927+654 obtained between 1990 July 11 (Julian date 2448084.1) and 1990 July 16 (left panel) and 1990 December 11 and 1990 December 21 (right panel). Persistent, rapid and giant X-ray variability is detected. The maximum amplitude variation is a factor of 15, at a significance level greater than 6 $\sigma$.

The X-ray properties of 1ES 1927+654 discussed below were obtained with the ROSAT and Chandra satellites. The ROSAT data cover the survey observations carried out between July 1990 and July 1991 as well as a ROSAT pointed observation obtained in December 1998.

For the ROSAT All-Sky Survey observations the source counts were obtained using a circular source cell with a radius of 11.1 arcmin. The number of source plus background counts within this cell was $4255 \pm 65$. The background was determined from a source-free cell with a radius of 11.5 arcmin, located in the scan direction through the centroid position of the source and centered at $\rm RA(2000) = 19^h22^m26.1^s$, $\rm Dec(2000) = 65^{\circ}44'17''$. The number of background counts normalized to the source cell size was $338 \pm 18$. The net counts are therefore $3917 \pm 62$, resulting in a mean count rate of $\rm 1.22 \pm 0.02\ counts\ s^{-1}$. For each path of the source through the ROSAT PSPC detector the corresponding background was subtracted.

Figure 2: The two upper panels give the ROSAT All-Sky Survey light curves in the soft (channels 11-41) and hard (channels 51-201) ROSAT energy bands referred to as S and H, obtained between between 1990 July 11 and 1990 July 16 and 1990 December 11 and 1990 December 2. The lower panel gives the resulting hardness ratio $\rm HR1 = (H{-}S)/(H+S)$.

1ES 1927+654 was again observed during the final observation period of the ROSAT satellite in 1998 December 8 between 18:17:21 UT and 18:41:01 UT with the PSPC detector. The total exposure time was 1338 s. This observation was affected by some anomalies (cf. Sect. 2.1 of Boller et al. 2000) and we have performed a careful data quality check to ensure that our results were not affected.

In addition, 1ES 1927+654 was observed with Chandra during the guaranteed time programme. The Chandra observation was performed using the low energy transmission grating spectrograph (LETGS, see Brinkman et al. 2000), which comprises the low energy transmission grating (LETG) and the high resolution camera microchannel plate detector for spectroscopy (HRC-S). 1ES 1927+654 was observed from March 20, 2001 at 13:20 UT until March 21, 2001 at 06:58 UT. The total exposure obtained from all good time intervals, corrected for deadtime is 63 110 s. During the whole observation no times with high background were detected. The raw data was reprocessed and the spectrum extracted using CIAO version 2.2.1 and CALDB release 2.15 following the corresponding science threads available at the Chandra X-ray Center.

The overlapping higher spectral orders were taken into account by using the LETGS grating response matrices (version July, 2002) which include 1st to 6th order. Separate files are provided for the negative and positive orders. The fits with the same parameters were performed simultaneously with the extracted negative and positive orders.

The Chandra X-ray centroid position of 1ES 1927+654 is $\rm RA(2000) = 19^h 27^m 19.61^s$, $\rm Dec(2000) = 65^{\circ}33' 54.86''$, which is in excellent agreement with the optical position (cf. Fig. 9). The positional accuracy is about one arcsec. The unprecedented positional accuracy of Chandra confirms the identification of the strongly variable X-ray source with the distant galaxy. All optically bright objects in Fig. 9 (labeled as 2-6) have been spectroscopically identified as late G or early K stars. This rules out any significant contribution of them to the X-ray flux of 1ES 1927+654. The positional accuracy of ROSAT is less sensitive than Chandra, however, the centroid position of 1ES 1927+654 in the ROSAT All-Sky Survey of $\rm RA(2000) = 19^h27^m19.2^s \pm 1.3^s$, $\rm Dec(2000) = 65^{\circ}33'58'' \pm 20''$ is consistent with the Chandra position.

2.1 The strong, persistent and rapid X-ray variability of 1ES 1927+654

In the following we constrain the timing properties obtained from the ROSAT All-Sky Survey, ROSAT pointed and Chandra observations. The plotted error bars in the light curves correspond to 1 $\rm\sigma$ in the Poisson regime (cf. Gehrels 1986). As a conservative approach we have calculated the total errors of the counts using the relation: $\rm 1.0 + \sqrt{(counts\ +\ 0.75)}$.

Figure 3: ROSAT PSPC pointed light curve obtained in December 1998. A strong X-ray flare lasting for about 100 s is detected near the end of the observations. The corresponding isotropic energy is $\rm 10^{46}\ erg$ (see Sect. 2.1).

1ES 1927+654 was observed during the ROSAT "mini-survey'' for about 5 days between 1990 July 11 and 16 with a total exposure time of 254 s. During the normal survey scan operations, 1ES 1927+654 was observed for about 11 days between 1990 December 11 and 21. Figure 1 shows the ROSAT PSPC survey light curve of 1ES 1927+654. The left panel refers to the `mini-survey' observations in July 1990 and the right panel gives the count rate variations during the survey scan observations in December 1990. During the mini-survey the paths of 1ES 1927+654 through the PSPC detector lasted only between 5.6 and 7.8 s and the source passed the PSPC detector 49 times. During the December observations the source passed the PSPC detector 143 times and the exposure times of the individual scans range between 5.6 and 26.1 s.

Summing up all paths of 1ES 1927+654 through the PSPC detector during all survey observations, results in a total exposure time of 3200 s. Most interestingly are the unusually large amplitude and persistent variability, making this object the second radio-quiet AGN showing this type of behavior, the first being IRAS 13224-3809 (Boller et al. 1997). As the exposure time per path during the "mini-survey'' is only about 1/3 of the December 1990 observations, we have rebinned the light curve shown in the left panel of Fig. 1 by summing up three paths of the source through the PSPC detector, resulting in comparable exposure times per data point between the two surveys. The mean separation between the data points is 4.8 hours in the left panel and 1.6 hours in the right panel. Averaging the four data points with count rates below 0.3 $\rm counts\ s^{-1}$, gives a mean number of counts of 3.76 $\rm\pm$ 3.12, collected within 22.93 s. The four data points above 2.4 $\rm counts\ s^{-1}$ result in a mean number of counts of 60.18 $\rm\pm$ 8.80, obtained within a 23.74 s exposure interval. The resulting count rates are $0.164 \pm 0.136$ and $2.534 \pm 0.370$ $\rm counts\ s^{-1}$, for the low and high states, respectively. The maximum amplitude variability is of a factor of 15, at a significance level greater than 6 $\rm\sigma$.

Figure 4: ROSAT PSPC pointed light curves in the soft (channels 11-41) and hard (channels 51-201) ROSAT energy bands obtained on 1998 December 8. The lowest panel gives the resulting HR1 hardness ratio.

Figure 5: The Chandra LETG light curve in the 0.3-7 keV band also shows several strong flaring events with doubling time scales less than 400 s.

In the following we constrain the spectral variability of 1ES 1927+654 based on the examination of appropriate hardness ratios. In Fig. 2 we show the light curve of 1ES 1927+654 during the 1990-1991 All-Sky Survey observations in the soft (channels 11-41; 0.1-0.4 keV) and hard (channels 51-201; 0.5-2.0 keV) ROSAT energy bands, as well as the resulting hardness ratio light curve. A constant model fit to the data points in the top panel ( $\chi^2 = 55.2$ for 10 d.o.f.) and the middle panel of Fig. 2 ( $\chi^2 = 180$ for 70 d.o.f.) can be rejected with >99.9 per cent confidence. Significant spectral variability is detected during the ROSAT All-Sky Survey observations of 1ES 1927+654. The variations in the soft and hard ROSAT energy bands are not correlated with the variations of the hardness ratio, making it difficult to draw further conclusions of the underlying physical emission mechanism causing the strong X-ray variability. This is confirmed by performing simple power law fits to different count rate intervals (e.g. from 0-1.5 and 1.5-3  $\rm counts\ s^{-1}$) and finding no significant differences in the photon index. Simple relations as found in other variable AGN, e.g. the source becomes steeper when the flux increases, do not seem to occur in 1ES 1927+654.

In Fig. 3 we present the ROSAT PSPC light curve on 1ES 1927+654. Using a bin size of 10 s, a strong X-ray flare becomes apparent between 1110 and 1280 s after the beginning of the observations. The count rate fluctuations significantly exceed that of the variations caused by the ROSAT wobble. We have searched for associated spectral variability during the flaring event. In Fig. 4 we demonstrate that the strong flux variability is not correlated with significant spectral changes, however, the hardness ratio light curve is still significantly variable (a constant model fit can be rejected with >99.9 per cent confidence).

The Chandra light curve (Fig. 5) indicates rapid count rate variations with doubling time scales down to about 400 s. The Chandra timing properties further support the interpretation that the variable X-ray emission arises within a few Schwarzschild radii from the central black hole.

2.2 Spectral properties

Figure 6: Power-law fit to the ROSAT All-Sky Survey observations as well as to the pointed observations with their corresponding residua. The neutral absorbing column density is significantly above the the Galactic column density of $N_{\rm H,Gal} \rm = (7.2 \pm 0.4) \times 10^{20}$ cm-2 (Dickey & Lockman 1990; Elvis et al. 1994).

Figure 7: Upper panel: contour plot of $\chi ^2$ as a function of the absorbing column $N_{\rm H,fit}$ and the photon index $\Gamma$ for five confidence levels of 68.3, 95.4, 99.7, 99.99 and 99.99999 per cent for the ROSAT Survey observations (upper panel) and the ROSAT PSPC pointed observations (lower panel). The absorption is larger than the Galactic column (dashed vertical line) at the 3 $\sigma$ level and at the 5 $\sigma$ level during the pointed observations.

Figure 8: Power-law fit to the Chandra LETG observations of 1ES 1927+654. The spectrum was fitted by leaving the neutral oxygen abundance free. The oxygen abundance is found to be consistent with solar values. If we fix the Galactic absorption to $\rm 7.2 \times 10^{20}$ cm-2, we find an intrinsic absorbing column density of $\rm (7.3 \pm 0.3)\times 10^{20}$ cm-2.

As described in the previous section, significant spectral variations are present, however, the count rate changes do not seem to be correlated with the variations in the hardness ratio. The spectral fitting results given below, represent the time-averaged spectra from each dataset in the 0.1-2.4 keV energy band for the ROSAT observations and in the 0.3-7.0 keV band for the Chandra LETG observations, respectively.

A simple power-law fit, where the absorption column density and the photon index are allowed to be free parameters, provides an acceptable fit to the ROSAT All-Sky Survey PSPC data ( $\chi^2 = 66$ for 82 d.o.f.; cf. Fig. 6). The soft X-ray absorption of $N_{\rm H,fit} \rm = (9.1 \pm 1.4) \times 10^{20}~ cm^{-2}$ is larger than the Galatic column towards 1ES 1927+654 of $N\rm _{H,gal} = (7.2 \pm 0.4) \times 10^{20}~ cm^{-2}$ (Dickey & Lockman 1990; Elvis et al. 1994) at the 3 $\sigma$ level. The photon index is $2.6 \pm 0.3$. These results are robust to changes in the number of data points included in the fit. Using the spectral parameters for 1ES 1927+654, as displayed in Fig. 6, results in a mean 0.1-2.4 keV flux of $f \rm = 6.6 \times 10^{-11}$ erg cm^-2 s^-1, corresponding to an isotropic luminosity of $L = \rm 4.6 \times 10^{43}$ erg s^-1.

A simple power-law fit to the PSPC pointed observation is shown in Fig. 6 (lower panel). The absorbing column during the pointed observation is in excess of the Galactic column above the 5 $\sigma$ limit (Fig. 7). No significant changes in the photon index are detected, however, the mean count rate during the pointed observation is 2.03 $\rm counts\ s^{-1}$, a factor of 2 larger than the ROSAT All-Sky Survey observations in 1990. The isotropic 0.1-2.4 keV energy emitted in the strong X-ray flaring events (cf. Fig. 3) is $E \rm = 1.7 \times 10^{46}\ erg$.

The spectral energy distribution of 1ES 1927+654 as obtained with Chandra can be fitted with a weak black body (accounting for the soft X-ray excess emission) with kT = 0.01 keV and a power law with a photon index of $\rm (2.5 \pm 0.2)$. However, the black body temperature remains unconstrained. A simple power-law model also provides an acceptable fit. If we fix the Galactic absorption to $\rm 7.2 \times 10^{20}$ cm^-2, we find an intrinsic absorbing column density of $\rm (7.3 \pm 3.0)\times 10^{20}$ cm^-2, which is about a factor of 2 greater than that found during the ROSAT observation (however, the difference is less than 2 $\rm\sigma$). As the Chandra LETG observation gives the highest intrinsic soft X-ray absorption, we use that value for comparison with the extinction values derived from the optical observations. The photon index in the 0.3-7.0 keV band is $\rm 2.7 \pm 0.2$. The isotropic luminosity in the 0.3 to 7.0 keV band is $\rm 2.1 \times 10^{43}\ erg\ s^{-1}$.

The strongest absorption edge from neutral oxygen at 0.537 keV is clearly visible in the LETG spectrum. To constrain the oxygen abundance relative to hydrogen, we have fitted the spectrum by leaving the oxygen abundance free. The spectral fit is still acceptable but does not significantly improve the spectral fitting results. The oxygen abundance is found to be consistent with solar values. Other element abundances can not be constrained with the present statistics of the Chandra LETG observation.

If we assume that (i) the empirical relation between interstellar X-ray absorption and optical extinction of $A_V = 4.5 \times 10^{-22} N_H$ (Gorenstein 1975) applies to 1ES 1927+654; (ii) that the dust grains are optically thin to X-rays, i.e. that the amount of X-ray absorption by the dust grains is not much different from that by an equivalent mass of material in gaseous form; (iii) that the X-ray and optical radiation travel through the same matter; (iv) that the obscuration is not a strong function of time and (v) a Galactic gas to dust ratio, then the neutral hydrogen column densities derived from the X-ray spectra can be converted into X-ray A_V values. The maximum intrinsic X-ray A_V derived from the ROSAT and Chandra observations is 0.33 (following Gorenstein 1975) or 0.58 (following Predehl & Schmitt 1995).