4. Comparison with observations
4.1. ROSAT PSPC X-ray data
Once blobs have formed and detached from the Local Bubble boundary, the remaining holes (or depressions, depending on the ratio of blob to wall extension) in the hydrogen distribution can cause corresponding changes in the absorption of soft X-rays from distant emission regions. This should be accompanied by spectral changes as softer X-rays are more heavily attenuated than harder X-rays.
The ideal data base for such an analysis of absorption and emission variation of soft X-rays is the ROSAT XRT/PSPC All-Sky Survey: sky maps of intensities and their uncertainties of the soft X-ray background (SXRB, keV) have been produced in 7 energy bands with spatial resolution (see Snowden et al. 1995, 1997, and references therein). These maps have been carefully cleaned from non-cosmic contaminations (e.g., particle events, scattered solar X-rays) and point sources have been masked. Also telescope vignetting and deadtime corrections have been applied.
The ROSAT survey maps of the Loop I region at low energies, primarily the R1 band (PSPC channels 8-19, keV) indicate the presence of a coherent neutral gas sheet filling the interaction ring as found by EA95. Judging from the known optical depth in the R1 band the mean column density of this gas sheet is cm-2, because it is blocking most of the R1 band radiation coming from behind.
This can be seen in Fig. 5 which shows ROSAT survey maps of the Loop I region in the ROSAT R1 (left, keV) and R2 energy bands (right, keV). The green line is a small circle fit to the radio continuum (RC) Loop I (Berkhuijsen 1972). The maps are centered on the centre of RC Loop I , galactic north is up. The dashed white lines mark the contours of the HI ring formed by the interaction of Loop I with the LB as described in Egger & Aschenbach (1995). The intensities of both images are scaled in a way that the regions outside (north) of the interaction zone appear to be equally bright. Hence, the intensity ranges are counts s- 1 arcmin-2 and counts s- 1 arcmin-2 for R1 and R2, respectively. One can see that the interaction ring is casting equally deep shadows onto Loop I on both of the low energy images due to its high absorbing column density of cm-2 which corresponds to for both bands. Also the view through the ring into the interior of the Loop I superbubble is inhibited by intervening neutral gas. However, the absorption is stronger in R1 than in R2, which roughly corresponds to optical depth unity in R1 as will be shown below. This is clear evidence for the existence of a coherent neutral gas sheet or "wall" separating the LB from Loop I.
Quantification of the absorbing column density using the R2/R1 band ratio is not trivial since we also have to take into account a certain amount of foreground emission in both bands coming from the LB. To make a rough estimate we compare the mean intensities of three separate regions, each about in size, one of them inside and two outside the interaction zone. We will use the labels A, B and C for the upper left, the upper right and the lower box (drawn in the ROSAT images in Fig. 5, respectively. The three regions have been chosen such that the spectra of the background sources are roughly equal: K. This precondition, however, is not very critical as will be shown below. Assuming now that the LB foreground emission is constant we obtain the following equations:
where , , , , , and are the total median intensities within the three regions in the two energy bands in units of counts s- 1 arcmin-2. and is the LB foreground emission, , , , , , and are the absorbed background intensities for bands R1 and R2, respectively. In order to cancel the foreground we subtract:
Hence, we obtain the hardness ratios and . Fig. 10a in Snowden et al. (1997) shows that for a R2/R1 ratio of we get an absorbing column density of cm-2 assuming a thermal background emission of . This result is almost insensitive to variations in background temperature between . cm-2 is therefore a conservative value. If we assume that the line of sight through the LB is somewhat larger at higher galactic latitudes (fields A and B) than it is in the galactic plane (field C), as is indicated by some current models, then the R2/R1 contrast is even higher for the absorbed background sources. Another way to get a rough estimate for the wall column density is to use the model of Snowden et al. (1998) for the LB quantities for and . For position C we would then obtain and . With and the hardness ratio is . Using Fig. 10a of Snowden et al. (1997) this corresponds to cm-2. These numbers are obtained by hardness ratio analysis only, i.e. by analysis of only two bands, but are nevertheless sufficient to show X-ray evidence for the existence of a coherent (large-scale) neutral gas sheet between the Local Bubble and Loop I. In the next section we will show a more detailed spectral analysis. We want to note that in the first estimate statistical errors are negligible () but systematic errors may be stronger due to the separation of regions A and B from C and the assumed spectral constancy of the local and distant emission regions. In the second approach errors are difficult to obtain from Snowden et al. (1998) as this study aimed at a large-scale description of the local and distant 1/4 keV X-ray sky by smoothing filters and does not provide uncertainties for individual sky regions.
On smaller scales, however, there are several directions where the H I wall appears to be disrupted, which is indicated by X-ray enhancements and spectral variations in the R1 and R2 bands: The R1 band count rate is only slightly increased, the R2 band rate (channels 20-41), however, shows significantly higher values (see Fig. 6) as compared to surrounding regions. This can be interpreted in terms of still considerable absorption for the soft R1 band X-rays, and beginning transparency for R2 band emission. In this region of enhanced X-ray intensity also a local minimum of the galactic neutral hydrogen column density is found. This supports the idea of blobs that have detached from the boundary and have left a remaining boundary with decreased density.
4.2. Spectral fits
Now we want to quantitatively examine the spectral variation visible in the R1 and R2 band. Since the ROSAT PSPC (as any proportional counter) has only a moderate energy resolution, the number of free spectral model parameters is limited to a maximum of 5. Our input model consists of three emission components (similar to Egger 1995). One is the extragalactic background, which we assume to follow a power-law spectrum with full galactic absorption derived from 21 cm maps. All parameters for this component are kept fixed during the fits but have been determined separately at high galactic latitudes. We have also included a local unabsorbed thermal component representing the Local (Hot) Bubble as Raymond-Smith plasma (in collisional ionization equilibrium), where the temperature was fixed to the canonical value of keV in agreement with other authors (e.g., Snowden et al. 1990, 1998; Egger 1995). The emission measure is a free parameter since we allow for local intensity variations. The third component is due to distant thermal emission from the Loop I superbubble with free temperature , emission measure , and foreground absorption (which has an upper limit of less than 1.5 times the galactic value to account for uncertainties in the determination of the galactic ). The values for the galactic neutral hydrogen column densities have been obtained by scaling the IRAS m map to data by Dickey & Lockman (1990), since the latter have only a spatial resolution of . There is no gain in using a more complicated model because having 4 free parameters and 7 data points is close to the feasible limit of the fit procedure.
The fits were performed for two regions inside the interaction ring (as defined by EA95), a bright area (A1) at and a darker one (A2) at (see Fig. 6). These regions were chosen to lie away from sky map portions with excessive background contamination to avoid effects of over- or undercorrections of these components (e.g., long-term X-ray enhancements) by the background cleaning processes. Cross-checking with the ROSAT Bright Source Catalog 4 (Voges et al. 1999) we convinced ourselves that the contribution by point sources was negligible. Moreover, we tried to avoid a bias due to the galactic bulge emission. Finally, the regions should not be separated too much to minimize variations of the distant emission component. Our selected areas A1 and A2 are otherwise typical regions and the restrictions have only been imposed in order to isolate the effects of absorption and spectral variation. The size of each region is pixel of , i.e. square degrees. This corresponds to a diameter of the blobs of pc for a distance estimate of the wall of pc, which is in good agreement with the estimated sizes derived earlier.
The enhanced region A1 shows a deficiency of hydrogen by cm-2 ( cm-2, cm-2) compared to the neighbouring region without enhancement (A2). For both areas we derived a distant temperature of keV. In Fig. 7 we illustrate the significance of our results by contour plots for both regions (A1 dashed, A2 solid) for parameters (x-axis) and (y-axis). Contour levels are , , and (from inside to outside). There are no significant changes in the properties of the distant thermal emission component (temperature, emission measure).
The column density contrast is of the order of the column density of the HI shell around Loop I ( cm-2, see e.g. Egger 1995) which is consistent with the view that most of the shell in the direction to the enhancement has been disrupted into blobs. Such blobs are not a rare event as will be shown in a separate paper (Freyberg et al. 2000), in which we report on the search for H I blobs in the data of the Leiden/Dwingeloo 21 cm survey. This suggests that a hydromagnetic instability is working on a larger scale.
4.3. Interstellar absorption lines
The existence of a wall is not only supported by X-ray spectral analysis, but also by independent interstellar absorption line measurements. EA95 and Egger (1998) have shown that there is a distinct rise in the H I column densities of stars in the direction towards the ring between the Local Bubble and Loop I. At a distance of about 70 pc, jumps from cm-2 to cm-2. We have reanalyzed the absorption data compiled by Fruscione et al. (1994) by using recent HIPPARCOS distances. The result is shown in Fig. 8, corresponding to Fig. 3 of Egger (1998). The left panel summarizes for which sample stars HIPPARCOS data are available (filled symbols), and whether these are "close" (triangles) or "distant" (squares) with respect to a distance of pc. The right panel still shows a distinct rise in , at a somewhat larger distance as some stars are now slightly further away. However, the main conclusion, namely the existence of an interaction ring, is still clearly visible.
In principle all the stars with cm-2 could be located at large distances even far behind the Loop I. However, we have checked that more than half of the 40 stars in the direction of the ring are located within 140 pc from the Sun. There are 6 stars between 89 and 140 pc with cm-2, and one star at 130 pc with cm-2 (however, this star is located in the direction of the H I enhancements of the ring). We further emphasize that all stars at pc have cm-2, i.e. at the same level as before. Therefore we conclude that the HI enhancement cannot be due to neutral hydrogen beyond Loop I and the jump in should be interpreted as a dense absorbing ring. As Loop I is a superbubble with ongoing star formation in which gas is continuously heated it is a valid assumption that the pressure is higher than in the adjacent Local Bubble where no evidence for an active stellar cluster is found. Therefore the interaction shell (wall) is pushed towards the Local Bubble with respect to the ring. Thus the distance of pc to the ring may serve as an upper limit for the distance to the wall, while parts near the centre direction of Loop I may be closer.
The compilation by Fruscione et al. (1994) contains column densities derived by various methods. We have analyzed a subsample of stars with directions towards the ring and towards the wall inside the ring with (our X-ray analysis was restricted to this range as well). In the left panel of Fig. 9 we have shown the absorption-distance-relation derived from Na I measurements, whereas in the right panel this relation derived from all other methods (like Ly , Mg II , EUV,...) is displayed. The Na I absorption column densities tend to be systematically below the ones obtained by other methods. This may be possibly due to ionization and/or abundances effects which influence the conversion factor of Na I to H I . The conversion factor by itself is a source of error in contrast to methods evaluating hydrogen directly (e.g., Ly ), and the correlation between and shows considerable scatter (one order of magnitude) (cf. Ferlet et al. 1985). Diamond et al. (1995) have used Na I absorption line studies to map the distribution of interstellar matter. Only two stars of their sample are in our field of interest (numbers 46 and 47 of their Table 3) and both are not in conflict with our model of a wall inside an interaction ring.
In the recent literature the Na I method is extensively described e.g. by Welsh et al. (1998). Unfortunately in the wall region, for a distance less or equal 100 pc, there are not many background stars with measured Na I column densities available. Therefore Welsh et al. conclude that it is difficult to derive the detailed morphology of the LISM using Na I absorption measurements.
© European Southern Observatory (ESO) 2000
Online publication: January 29, 2001