Astron. Astrophys. 350, 447-456 (1999)

5. Results and discussion

The data reduction process resulted in high-quality images of the Stokes Q and U emission, over much of the first Galactic quadrant. In addition to being shown herein (Figs. 3-5), these data are all available on-line; see the end of the paper for details.

It should be noted, however, that the observing technique (using scan lengths of approximately ), as well as the reduction procedures applied to the data, attenuate structure on scale-sizes of the order of and larger. Hence, it must be stressed that Stokes Q and U emission components on these (and larger) angular scales will not be correctly represented in the survey data.

Note that the intensity calibration of the high-latitude data () is accurate to approximately 5%.

5.1. Distribution of polarised intensities

The distribution of polarised intensities in the first and fourth Galactic quadrants is shown in Fig. 1. This (and Fig. 2) contain data from the Effelsberg survey for Galactic longitudes greater than , and the Parkes survey (Duncan et al. 1997) for longitudes less than . Before plotting, the Effelsberg data were convolved to the beam size of the Parkes survey (), and regridded to match the pixel size and spacing of the Parkes images ().

Fig. 1. Mean polarised intensity values over the Galactic Plane, binned into intervals of longitude. Data for are from the Effelsberg 2.695 GHz survey, and data with are from the Parkes 2.4 GHz survey. The dashed line shows the mean values of points with ; the solid line for points with . The small dashes seen across the figure indicate the standard deviation of each point (these are placed at both ), and hence provide a measure of the uniformity of the polarised intensities in each bin.

Fig. 2. The distribution of polarisation position-angles along the Galactic Plane, binned into intervals of longitude, showing the orientation of the electric vector of the received radiation. Data for are from the Effelsberg 2.695 GHz survey, and data with are from the Parkes 2.4 GHz survey. The dashed line shows the mean values of points with ; the solid line for points with . A vector of angle is oriented perpendicular to the Galactic Plane, and rotates anti-clockwise as the angle increases. The small dashes scattered across the figure indicate the standard deviation of each bin (these are placed at both ).

Fig. 3. a  A grey-scale image detailing the polarised intensities over the longitude range . The rms noise is approximately 2 mJy beam area-1, with a resolution of . The grey-scale wedge is labelled in units of mJy beam area-1. b  Polarisation position-angles over the same region, showing the orientation of the electric vector of the received radiation. The length of each vector is proportional to the intensity of the polarised emission at each point, with a maximum polarised intensity of 880 mJy beam area-1. The vector lengths are clipped at 100 mJy beam area-1. Data are blanked wherever the intensity falls below 2. A vector is plotted every 12´.

Fig. 4. a A grey-scale image detailing the polarised intensities over the longitude range . The rms noise is approximately 2 mJy beam area-1, with a resolution of . The grey-scale wedge is labelled in units of mJy beam area-1. b  Polarisation position-angles over the same region, showing the orientation of the electric vector of the received radiation. The length of each vector is proportional to the intensity of the polarised emission at each point, with a maximum polarised intensity of 680 mJy beam area-1. The vector lengths are clipped at 100 mJy beam area-1. Data are blanked wherever the intensity falls below 2. A vector is plotted every 12´.

Fig. 5. a A grey-scale image detailing the polarised intensities over the longitude range . The rms noise is approximately 2 mJy beam area-1, with a resolution of . The grey-scale wedge is labelled in units of mJy beam area-1. b  Polarisation position-angles over the same region, showing the orientation of the electric vector of the received radiation. The length of each vector is proportional to the intensity of the polarised emission at each point, with a maximum polarised intensity of 200 mJy beam area-1. The vector lengths are clipped at 75 mJy beam area-1. Data are blanked wherever the intensity falls below 2. A vector is plotted every 12´.

Examination of Fig. 1 reveals a general symmetry in the polarised intensity distribution in the first and fourth Galactic quadrants. First, bright emission is seen at both low and high latitudes, around the Galactic Centre (GC). This is predominantly due to the presence of a bright "plume" of polarised emission in this region, detected by the 2.4 GHz Parkes survey (Duncan et al. 1998). The Effelsberg data confirm that this plume extends no further than approximately into the first Galactic quadrant.

Second, bright, large-scale polarisation is seen in both the northern and southern sections of the Plane, at distances from the GC of between and . It should be noted, however, that the characteristics of these two regions appear quite different. In the south, the low and high latitude data are comparable, but with slightly higher values closer to the Plane. By contrast, the northern region is somewhat fainter overall, and is present in the high latitude curve alone. Hence, the polarisation enhancements are less dependent on latitude (i.e. more diffuse) in the south, but show a strong concentration to in the north. The polarised intensity distribution associated with this northern enhancement will be further discussed in Sect. 5.4.

Third, about from the GC, both the low and high latitude curves tend towards minima of approximately 30 mJy beam area^-1 (with the exception of a low latitude "bump" near longitude). The standard deviations are also smallest in these regions. Although not shown, the southern data continue this trend to a longitude of (the edge of the survey), prompting the idea that this represents a quasi-uniform "background" level of polarised emission (Duncan et al. 1997). The addition of the Effelsberg data, which exhibits similar minima, supports this hypothesis.

5.2. Distribution of polarisation position angles

The distribution of polarisation position-angles in the first and fourth Galactic quadrants is shown in Fig. 2. As in Fig. 1, the data are averaged into intervals of Galactic longitude.

Considering Fig. 2, it can be seen that the data on either side of the GC exhibit very different characteristics. Whilst southern longitudes contain several areas of large-scale structure in the position-angles, such regions are essentially absent from northern longitudes. Furthermore, the distribution of standard deviations is much more uniform over positive longitudes. This suggests that, on scale-sizes of degrees, the polarisation position-angles from the Effelsberg survey () are more random and uncorrelated than those from the Parkes survey ().

5.3. Differences between the Effelsberg and Parkes surveys

The data presented in Figs. 1 and 2 (and particularly that in Fig. 2) indicate significant differences between the Effelsberg and Parkes surveys. In particular, the Effelsberg survey appears less sensitive to the largest scale-sizes of structure than does its southern counterpart (scale-sizes of the order of ).

To confirm this interpretation, large-scale structure was approximately subtracted from the Stokes Q and U images for longitudes of . The subtraction was performed using the method of "unsharp masking", as used by Sofue & Reich (1979). Structure on angular scales and larger was subtracted from the Parkes images. Polarised-intensity and polarisation position-angle maps were then generated from the Q and U components.

The filtered Parkes data displayed characteristics much more like those of the Effelsberg survey. Specifically, large-scale structures in the position-angle data were removed, and the standard deviations appeared much more uniform. Interestingly, almost all of the large-scale polarised-intensity structure remained unchanged. This is also consistent with the observations, because the "minimum" polarised intensity (approximately 30 mJy beam area^-1), as seen in Fig. 1, is similar in both the Effelsberg and Parkes data.

The differences between the Effelsberg and Parkes surveys, discussed above, result primarily from the differences in observing procedure. As noted in Sect. 3, the Effelsberg observations used scan lengths of typically . In contrast, the Parkes survey employed scan lengths of between and . It is because of these longer scans that the data reduction procedure is able to retain more of the large angular-scale emission and structure.

5.4. Bright polarisation near longitude

In Fig. 1, bright, polarised emission can be seen at , between longitudes of approximately and . The polarised-intensity images of this region of the Plane appear in Figs. 3a and 4a, and the polarisation position-angle maps are shown in Figs. 3b and 4b. Note that the position-angle map shows the electric vector of the received radiation.

The bright emission seen in this region at latitudes reaches a maximum intensity of approximately 75 mJy beam area^-1 near , . Most of this high-latitude emission appears as a series of large patches. As with polarisation structure seen in other regions of the Plane, the vector orientations are distinctly "cellular", over angular scales of the order of . Within these cells, the vectors are generally aligned well with each other (indicating that differential Faraday rotation is small over these areas), with regions of depolarisation at cell boundaries. In some areas, the vectors are seen to lie approximately perpendicular to filamentary structures (e.g. the bright filament near ), indicating both that the rotation measures (RMs) produce negligible differential rotation across these features (less than 30 rad m^-2), and that coherent magnetic fields exist within the regions.

Fig. 6 shows the distribution of polarisation angles for latitude ranges of both and . As in Figs. 1 and 2, the data are binned into longitude intervals of . The peaks in these graphs are fairly regularly spaced, with separations of between and . Fig. 6 clearly shows the presence of coherent, quasi-periodic structure over several tens of degrees of longitude. It is interesting to speculate as to the origins of this structure. Whilst data from the Parkes southern survey also shows evidence of large-scale features in the polarisation position-angle data, no similar quasi-periodic structure has been found.

Fig. 6. The distributions of mean polarisation angles over the latitude ranges of (upper graph) and (lower graph) are shown. As in Fig. 2, the angles have been binned into intervals of longitude, and show the orientation of the electric vector of the received radiation. A vector of angle is oriented perpendicular to the Galactic Plane, and rotates anti-clockwise as the angle increases. The small dashes scattered across the figure indicate the standard deviation of each bin, and are placed at both . The peaks in the curves show separations of between and .

As mentioned in Sect. 5.1, the large and brightly polarised regions are almost entirely restricted to longitude ranges of between and from the GC, over both positive and negative longitudes (see Fig. 1). This implies that a relationship between Galactic structure and the presence of the brightly polarised regions may exist. Indeed, longitudes of to are coincident with the Sagittarius spiral arm, and to with the Carina arm (Beuermann et al. 1985; Vallée 1997). This highlights the possibility that we are detecting polarised radio emission from these spiral arms. Between longitudes of and , the line of sight through the Sagittarius arm is quite long, starting at approximately 2 kpc and continuing to approximately 8 kpc (Beuermann et al. 1985, assuming a distance from the Sun to the GC of 8.5 kpc).

5.4.1. Anticorrelation with HI data

The possibility of detecting polarised emission from distant spiral arms, and the quasi-periodic nature of the polarisation angles seen in Fig. 6 prompted the examination of HI data (Hartmann & Burton 1997) over this section of the Plane, as this could provide distance information.

Fig. 7 shows both the HI emission and polarised intensity over a section of the Plane. Data below show very intense HI emission (greatly exceeding the dynamic range of the figure), and no polarised emission. The lack of polarisation here is not unexpected, because the increase in thermal electron density and the more turbulent field near will produce strong depolarisation. Indeed, this trend can be seen throughout much of the data presented in Figs. 3-5.

Fig. 7. A grey scale image of the HI gas over a section of the Plane examined in Fig. 6, averaged over a velocity interval of +30 to +39 km s-1. The contours are those of polarised intensity, obtained from Figs. 3 and 4. Before plotting, the distribution of polarised emission was appropriately smoothed to a resolution of , so as to better match the resolution of the HI data. Note the anticorrelation between regions of bright polarised emission, and those of enhanced HI . Data below latitudes of have been blanked. The HI intensities towards negative latitudes have been multiplied by 2. Contour levels are: 1.7, 2.3, 3 and 4 Jy beam area-1. The grey scale wedge is labelled in K.

However, examining the distribution of polarised intensities in Fig. 7 for , a clear tendency for regions of polarised emission to be anticorrelated with regions of bright HI emission is also seen. The velocity ranges over which this is evident are from +25 to +40 km s^-1 between longitudes of approximately , and +35 to +50 km s^-1 between longitudes of . This is consistent with the HI emission originating at kinematic distances of between 1.8 and 2.5 kpc (Brand & Blitz 1993). We again note that the Sagittarius spiral arm is located at distances of between approximately 2 kpc and 8 kpc over these longitudes.

The anticorrelation can arise through two principal means. First, if the polarised radio emission is produced at the same distances as the HI regions (1.8-2.5 kpc), then the polarised emission must be suppressed in regions of dense HI by internal depolarisation effects (Sokoloff et al. 1998). Second, the majority of the polarised emission may be produced at distances greater than those of the HI regions (i.e. between distances of 2.5 and 8 kpc), and then depolarised on its passage through these regions of denser HI by external Faraday dispersion (Sokoloff et al. 1998).

5.4.2. A Faraday "screen" in the Sagittarius spiral arm

If the polarisation detected over this region of the Plane were due to an increase in the intensity of the synchrotron emission, we would expect increases in the polarised emission to correlate with increases in the total-power flux. Specifically, the polarised patches should have counterparts in total-power of at least 1.4 times the intensity. However, examination of total-power structure on the same angular scales shows no correlation with polarised intensity. This is also true of the emission on smaller angular scales (as shown in Figs. 3 and 4). Thus, we suggest that the detected anticorrelation is the result of the Faraday depolarisation of more distant polarised emission by a "Faraday screen", which is correlated with the HI discussed before, rather than the result of increases in the synchrotron emissivity. Note that the existence of such screens has been suggested by several other authors, working with interferometric telescope data. Specifically, Wieringa et al. (1993) deduced the presence of a local ( pc) Faraday-rotating screen from their work at 325 MHz. Similarly, Gray et al. (1999) find evidence for a Faraday screen in a region of the Perseus spiral arm.

From the anticorrelation with HI over the velocity ranges noted above, it is probable that this depolarising "screen" is located on the near side of the Sagittarius spiral arm, at distances of between 1.8 and 2.5 kpc.

The polarised emission upon which the depolarising screen acts cannot originate from more than a few kpc behind the screen. This is because more distant emission is depolarised by differential Faraday rotation within the spiral arm (Sokoloff et al. 1998; Burn 1966).

5.4.3. Sources of the anticorrelation with HI

It is of interest to consider whether HI -associated HII may be the cause of the general anticorrelation seen in Fig. 7. That is, are the areas coincident with bright HI emission depolarised because of a greater density of thermal electrons associated with the brighter regions of HI .

Observations by Reynolds et al. (1995) of a number of HI complexes towards medium and high Galactic latitudes detected thermal particles with column densities of the order of  cm^-2 or less. Much of the HI seen in Fig. 7 appears as patches of the order of to in size, corresponding to 70-100 pc at the kinematic distance of the HI regions (approximately 2 kpc). This leads to a density of thermal electrons of up to  cm^-3 (assuming a filling-factor of unity). Using a (uniform) magnetic field strength of G (as adopted by Heiles 1996for the Solar vicinity), this results in an RM of  rad m^-2 or less, which will produce rotations of the polarisation angles of or so at 2.7 GHz. It is unlikely that the average values of n exceed 0.15 cm^-3 by a large factor, because of the total-power emission which would result. For example, increasing n by a factor of 10 would elevate the emission measures to  cm^-6 pc (and to even higher values for filling-factors less than unity). Such emission measures would be detectable in the total-power survey data, and show up as a better anticorrelation between the total-power and polarised intensity images.

If we assume, then, that  cm^-3 is an average value across the telescope beam, and that variations in n within the beam area rise to maxima of several times this value, we expect variations in the rotation of the polarisation angles within the beam of the order of . Although an order-of-magnitude estimate, this value is not sufficiently small for us to exclude HI -associated thermal electrons as a source of significant depolarisation, and hence as a possible mechanism for the production of the anticorrelation seen in Fig. 7.

Another mechanism for producing the anticorrelation between polarised emission and HI is "tangling" of magnetic fields. In this case, the magnetic field within the denser regions of HI is tangled by the turbulent motions of the gas, resulting in no coherent Faraday rotation within the beam and leading to depolarisation. Outside the denser regions of HI the magnetic field is more uniform, resulting in less beam depolarisation. We note that a similar "tangled field" idea has been suggested to account for the low degrees of polarisation observed coincident with the arms of external galaxies (Beck et al. 1996, 1998). Here, this mechanism would require the magnetic field to be tangled on scales , which corresponds to  pc at a distance of 2 kpc.

Of course, both mechanisms may play a significant role in depolarising the emission.

5.4.4. The quasi-periodic structure

Examination of the data shows the polarisation angles of the brightly polarised regions (which give rise to the quasi-periodic patterns in Fig. 6) to deviate typically up to from one extremum to the other. At a distance of 2 kpc, the distance between peaks of approximately longitude corresponds to  pc. If we consider the rotation to occur over a similar distance (i.e. that the "depth" of the structures is approximately 250 pc also), then a maximum value of

is required to produce this rotation at a frequency of 2.7 GHz, where and are the mean thermal electron density and line-of-sight component of the magnetic field, respectively, within the Faraday rotating region.

The strength of the uniform component of the magnetic field adopted by Heiles (1996) for the Solar vicinity of 2.2 µG and the electron density model presented by Taylor & Cordes (1993) lead to a typical value of G cm^-3 for z equal to 150 pc. As we are looking predominantly along the Sagittarius arm, the uniform component of the field is predominantly oriented along the line of sight. Thus, the required amount of rotation of the polarisation vectors can be supplied by the magneto-ionic medium of the Sagittarius spiral arm; no special field or electron density enhancements are required.

Although no enhancements in field strength or thermal electron density appear to be needed to produce the required Faraday rotation, a particular field geometry is still required. Fig. 6 shows vectors to be oriented at both positive and negative position angles. If this behaviour is produced through the action of Faraday RMs of opposite sign, then alternations of the direction of the line-of-sight component of the magnetic field are needed. Based on the calculations above, the required strength of this alternating field component is G.

We again note that, at distances of approximately 2 kpc, these polarised features are several hundred pc in size. Reaching z heights of several hundred pc above and below the Sagittarius spiral arm, it is possible that these magnetic field structures are produced by a mechanism similar to the Parker instability (e.g. Parker 1966; Giz & Shu 1993).

5.4.5. Local Faraday rotation?

Whilst the bright, polarised emission seen over the longitude range of is probably produced at distances of between 2.5 and 8 kpc, we have not established the distance at which the quasi-periodic angle structure (Fig. 6) is imposed. Magnetic fields exhibiting similar "wave-like" patterns are known on much smaller linear scales (e.g. the Taurus dark cloud complex, as investigated by Goodman et al. 1990). It is therefore possible that this structure in the polarisation angles is produced through the Faraday effects of a relatively local magnetic field feature.

However, if the variations in the polarisation angles were imposed at relatively local distances ( pc, say), then the magnetic field structures would be of correspondingly smaller linear size. This implies a much smaller path length over which the polarisation angles must be rotated, requiring far larger values for the field strength and/or thermal electron density over the region of rotation. For example, at a distance of 300 pc, the path length for Faraday rotation shrinks by an order of magnitude. Hence, an order of magnitude increase in over the value in Eq. 1 would be required to produce the observed rotation. It is difficult to see how such increases could come about.

Furthermore, we note from Fig. 6 that the polarisation angles at positive Galactic latitudes are not correlated with those at negative latitudes. If a local magnetic field feature of modest linear size was responsible for the structure in Fig. 6, it is likely that the angles would show some correlation.

© European Southern Observatory (ESO) 1999

Online publication: October 4, 1999
helpdesk.link@springer.de