5 Discussion

  
5.1 The radio-infrared correlation

The integrated $\lambda$6 cm radio flux density $S_{6~\rm cm}$(Tables 5 and 6) is correlated with the integrated $\lambda60~ \mu$m far-infrared flux density $S_{60~\mu \rm m}$ (Tables 1 and 2) as shown in Fig. 2. The correlation coefficient is $0.97\pm0.02$. NGC 1559 lies well above the fitted line, i.e., its radio emission is "too high'' compared with its far-infrared emission (cf. Sect. 5.2). This is possibly also true for NGC 1097. NGC 986 and NGC 7552 are "too radio-faint'', possibly due to the incomplete uv coverage of our observations.

Figure 2: The radio-far-infrared correlation for the sample of barred galaxies: integrated radio continuum flux density at $\lambda6~{\rm cm}$ versus the integrated far-infrared flux density at $\lambda60~ \mu$m. NGC names are indicated for bright galaxies.

The radio-far-infrared correlation has been studied in detail for large samples of barred and non-barred galaxies. The average flux density ratio $S_{6~\rm cm}/S_{60~\mu \rm m}$ (where $S_{6~\rm cm}$ is measured in mJy and $S_{60~\mu \rm m}$ in Jy) is $3.0\pm0.3$ for the RC2 galaxies in the sample of de Jong et al. (1985) and $2.3\pm0.1$ for the 134 galaxies observed by Unger et al. (1989, scaled to $\lambda$6 cm). The slope of the fitted line in our Fig. 2 is $2.6\pm0.3$, in agreement with these results. The radio continuum luminosity of spiral galaxies is also closely related to the far-infrared luminosity (Condon 1992; Niklas 1997).

Unger et al. found no significant difference in the radio/far-infrared ratio either between Hubble classes or between barred and non-barred galaxies. The average value of $S_{6~\rm cm}/S_{60~\mu \rm m}$ in our sample also indicates no general excess of radio emission from barred galaxies.

A close correlation between radio continuum emission and dust emission in the far-infrared has been found within many galaxies (Bicay & Helou 1990; Hoernes et al. 1998), and between radio continuum and the mid-infrared emission ( $\lambda15~\mu$m) in the spiral galaxy NGC 6946 at all spatial scales (Frick et al. 2001). ISOCAM images at $\lambda7~\mu$m and $\lambda15~\mu$m are available for several galaxies in our sample: NGC 1097, 1365, 1433, 1672, 4535 and 7552 (Roussel et al. 2001a). The similarity to our radio maps is striking and shows that the relationship holds not only for the integrated flux densities and luminosities, but also for spatial scales down to our resolution.

As synchrotron emission dominates at radio wavelengths longer than about $\lambda$3 cm, the radio-infrared correlation cannot be explained solely by thermal processes. Various interpretations are dicussed by Hoernes et al. (1998). As suggested by Niklas & Beck (1997), the radio-[far-]infrared correlation for bright galaxies holds if the magnetic field is connected to the star formation rate where the gas clouds may serve as the physical link. For radio-weak galaxies, however, the far-infrared emission is dominated by cold dust heated by the general radiation field which is not related with recent star formation (Hoernes et al. 1998). Most of the galaxies in our sample are bright enough to ensure that their far-infrared emission is indeed a measure of star formation intensity.

In normal spiral galaxies, cool gas and magnetic fields are compressed in various shocks, followed by an increase in star formation. However, large-scale shock fronts in a galaxy do not always enhance star formation. For example, the non-barred spiral galaxy NGC 2276 interacts with some external (intracluster) gas, so that a large-scale shock front forms on the leading side producing a ridge of strong total and regular magnetic field without significant effect on star formation (Hummel & Beck 1995). As a consequence, this galaxy deviates from the radio-far-infrared correlation.

Barred galaxies also host large-scale shock fronts, identified with dust lanes. However, shock fronts in bars are non-standard shocks in that they have enhanced velocity shear across them, similar to bow shocks. If the shear rate $\partial V_i/\partial x_j$exceeds the inverse time for star formation, the gas density enhancement in the shock does not trigger star formation. If the magnetic field is compressed in the shock, the ratio of radio/far-infrared flux densities would be higher than normal. However, for our sample this ratio and the average total field strengths (Sect. 5.3) are similar to those of non-barred galaxies. This indicates that large-scale field compression in the bar is generally small and that the magnetic field is not frozen into the flow in the regions of strong compression and shear (the dust lanes). Nevertheless, the average surface brightness in radio continuum and far-infared increases with increasing bar length (see Sect. 5.2).

5.2 Radio emission and bar strength

Several quantitative measures of bar strength have been suggested. Most of them are based on purely geometric parameters such as the bar axial ratio, where a smaller value of b/a means a stronger bar (Martin 1995; Aguerri 1999; Chapelon et al. 1999; Abraham & Merrifield 2000). Smallest values of b/a in our sample are found in NGC 1300, 1433, 1493, 1559, 3059, 3359 and 7552 (see Tables 1 and 2), but only NGC 1559 and NGC 7552 have a high surface brightness in radio continuum (see Tables 5 and 6) and far-infrared. NGC 1365 has the highest radio flux density $S_{6~\rm cm}^*$in our sample, although the aspect ratio of its bar is relatively small (b/a=0.51). However, it hosts the longest bar ($\simeq$29 kpc) in the sample. NGC 1559, 1672 and 7552 are fainter mainly because they are smaller. Their (distance-independent) radio surface bightness values $I_{6~\rm cm}$(measured in $\mu$Jy/beam area) are similar to or even larger than that of NGC 1365 (see Col. 11 in Tables 5 and 6), and the same is true for the typical far-infrared surface brightness, which is a measure of star formation rate per surface area. Figure 3 confirms that the radio surface brightness $I_{6~\rm cm}$is uncorrelated with the aspect ratio b/a (correlation coefficient of $0.58\pm0.15$).

Figure 3: Variation of the radio surface brightness $I_{6~\rm cm}$at $\lambda$6 cm with the deprojected aspect ratio b/a of the bar.

A physically motivated measure of the bar strength has been introduced by Buta & Block (2001) and Block et al. (2001) based, following Combes & Sanders (1981), on the maximum amplitude of the tangential gravitational force relative to the mean axisymmetric radial force. Thus defined, the strength parameter $Q_{\rm b}$ is sensitive not only to the bar ellipticity, but also to its size and mass, and is related to the quadrupole moment of the bar potential (P. Englmaier, priv. comm.). A reliable estimation of $Q_{\rm b}$ involves careful analysis of near-infrared galactic images (Quillen et al. 1994). Buta & Block (2001) and Block et al. (2001) determine $Q_{\rm b}$ for a selection of galaxies, but only six of them belong to our sample. Therefore, we consider a simpler (and admittedly incomplete) measure of the bar strength described in what follows.

Figure 4: Variation of the radio surface brightness $I_{6~\rm cm}$at $\lambda$6 cm with the relative bar length 2a/d25(logarithmic scales). Deviating points are marked by their NGC name.

The quadrupole moment, with respect to the major axis, of a homogeneous triaxial ellipsoid with semi-axes $a,\ b$ and c is given by $\frac{1}{5}M(2a^2-b^2-c^2),$where M is the mass of the ellipsoid (Landau & Lifshitz 1976). Assuming that the vertical scale height of the bar is much smaller than its size ( $c \ll a, b$) (e.g., Buta & Block 2001), the quadrupole moment normalized to M₂₅R₂₅⁴ (with M₂₅ the mass within the radius R₂₅) is given by

$$\Lambda = \left(\frac{a}{R_{25}}\right)^4 \frac{b}{a} \left(... ...rac12}\frac{b^2}{a^2}\right) \frac{\sigma_{\rm b}}{\sigma}\;,$$ (1)

where $\sigma_{\rm b}$ and $\sigma$ are the average mass surface densities of the bar and within R₂₅, respectively, and we have omitted numerical factors of order unity. The relative bar length a/R₂₅=2a/d₂₅, rather than the bar axial ratio, is the dominant factor in $\Lambda$. Moreover, 70% of galaxies in our sample have b/a>0.4, and $\Lambda$ varies just by 30% for 0.4<b/a<1. Therefore, b/a is a poor measure of bar strength, especially for our sample, as it does not discriminate well enough between galaxies with $b/a\ga0.5$.

Since $\Lambda$ depends strongly on the relative bar length 2a/d₂₅, we can reasonably expect that radio emission is correlated with this parameter. This expectation is confirmed by the high correlation between the radio surface brightness $I_{6~\rm cm}$at $\lambda6~{\rm cm}$ and the relative bar length (correlation coefficient of $0.86\pm0.06$, with NGC 1559 excluded - see below), confirmed by Student's t test. Figure 4 shows this correlation in logarithmic scales. Although the scatter is stronger than that in Fig. 2, the correlation is not weaker than other correlations discussed for barred galaxies in the current literature (cf. Chapelon et al. 1999). We conclude that a stronger bar results in an overall enhancement of the total radio emission in the bar region despite a relatively weak compression of the regular magnetic field near the dust lanes, as discussed in Sect. 5.1. As noted by Block et al. (2001), longer bars can produce more extensive deviations form axial symmetry in the gas velocity because the relative tangential force is stronger when the end of the bar is farther from the (axisymmetric) bulge; this may be the physical reason for the correlation shown in Fig. 4.

Figure 5: Total intensity contours and the observed B-vectors of polarized intensity (E-vectors turned by 90$^\circ$, uncorrected for Faraday rotation) of NGC 1097, overlayed onto an optical image kindly provided by H. Arp. The contour intervals are at 1, 2, 3, 4, 6, 8, 12, 16, 32, 64, 128, 256 $\times$the basic contour level which is 700, 500, 500 and 400 $\mu$Jy/beam area at $\lambda$22, 18, 6.2 and 3.5 cm, respectively. A vector of 1 $^{\prime \prime }$ length corresponds to a polarized intensity of 20 $\mu$Jy/beam area. The half-power width of the synthesized beam is shown in the corner of each panel.

With the most strongly deviating galaxy excluded (NGC 1559), the data shown in Fig. 4 can be fitted with a power law

$$I_{6~{\rm cm}}\propto (2a/d_{25})^{1.5\pm0.4}.$$

A similar dependence is valid for the far-infrared surface brightness $I_{60~\mu\rm m}$ (exponent $1.5\pm0.5$).

Figure 6: Total intensity contours and the observed B-vectors of polarized emission of NGC 1300, overlayed onto an optical image from the Digitized Sky Surveys. The contour intervals are 1, 2, 3, 4, 6, 8, 12, 16, 32, 64, 128, 256 $\times$ the basic contour level, which is 150, 100, 50 and 40 $\mu$Jy/beam area at $\lambda$22, 18, 6.2 and 3.5 cm, respectively. A vector of 1 $^{\prime \prime }$ length corresponds to a polarized intensity of 10 $\mu$Jy/beam area.

We have been unable to include the dependence on the surface mass densities into our measure of the bar strength, and this plausibly contributes into the scatter of the data points around the fit. It is difficult to say whether or not $\sigma_{\rm b}/\sigma$ is correlated with 2a/d₂₅. If $\sigma_{\rm b}/\sigma$ is independent of 2a/d₂₅, the above fit implies an approximate scaling

$$I_{6~{\rm cm}}\propto \Lambda^{0.4\pm0.1}.$$

There are a few deviations from the above correlation (see Fig. 4). NGC 1559 has the largest ratio of radio to far-infrared flux densities and the highest radio surface brightness, and so deviates strongly from the radio-infrared correlation as well (see Fig. 2). NGC 1559 is not a member of any group or cluster of galaxies and has no nearby companion (Zaritsky et al. 1997). High-resolution radio and optical observations are required.

Figure 7: Total intensity contours and the observed B-vectors of polarized emission of NGC 1365, overlayed onto an optical image taken at the ESO 3.6 m telescope by Lindblad (1999). The basic contour levels are 700, 500, 200 and 100 $\mu$Jy/beam area in decreasing wavelength order, the contour intervals are as in Fig. 5. A vector of 1 $^{\prime \prime }$ length corresponds to a polarized intensity of 20 $\mu$Jy/beam area.

On the other hand, NGC 1300, 1433 and 3992 are radio-weak in spite of their relatively long bars (Fig. 4). Their far-infrared flux density and thus their star formation rate is low. Apart from an usually small value of $\sigma_{\rm b}/\sigma$for these galaxies, other reasons for these deviations are concievable. Martinet & Friedli (1997) argue that some galaxies with strong bars have settled into a quiescent state after an episode of vigorous star formation which has transformed most of the gas into stars. Alternatively, Tubbs (1982) and Reynaud & Downes (1998) found indications for suppression of star formation in fast flows of the gas along the bar. The field strength should be low in the first case, because there is not enough gas to hold the field or the dynamo is not able to maintain a strong magnetic field. In the second case the field should be strong, but the galaxy does not host enough cosmic-ray electrons to generate strong synchrotron radiation. This can be verified by comparing regular magnetic field strengths deduced from Faraday rotation and polarized intensity from further radio observations with higher sensitivity.

Figure 11: Total intensity contours and the observed B-vectors of polarized emission of NGC 3992, overlayed onto an optical image from the Digitized Sky Surveys. The contours and the vector scale are as in Fig. 6.

Figure 12: Total intensity contours and the observed B-vectors of polarized emission of NGC 4535, overlayed onto an optical image from the Digitized Sky Surveys. The contours are as in Fig. 6. A vector of 1 $^{\prime \prime }$ length corresponds to a polarized intensity of 20 $\mu$Jy/beam area.

Measurements of the star formation efficiency SFE may also help: In the first case, the content of molecular gas should be low, with a SFE similar to that in spiral galaxies, while in the second case the SFE should be exceptionally small. Existing data seem to favour a higher SFE in barred galaxies compared to non-barred ones (Young 1993), but the infrared luminosity is dominated by the central region where star formation is triggered by gas inflow (see Roussel et al. 2001b). The SFE in the bar itself (and its possible suppression by a fast gas flow) should be subject to future investigations.

Figure 14: Total intensity contours and the observed B-vectors of polarized emission of NGC 7479, overlayed onto an optical image from the Digitized Sky Surveys. The contours are as in Fig. 6. A vector of 1 $^{\prime \prime }$ length corresponds to a polarized intensity of 20 $\mu$Jy/beam area.

5.3 Magnetic field strength

The estimates of the total magnetic field strength in our Galaxy, derived from $\gamma$-ray data and the local cosmic-ray energy density (Strong et al. 2000), agree well with equipartition values from radio continuum data (Berkhuijsen, in Beck 2001), so that the equipartition assumption is a useful estimate, at least on scales of more than a few kpc.

From the integrated flux density $S_{6~\rm cm}$ at $\lambda$6 cm and the solid angle of the integration area, the surface brightness $I_{6~\rm cm}$and the corresponding equipartition strength of the total magnetic field $B_{\rm tot}$ were computed (Tables 5 and 6), assuming for all galaxies a thermal contribution to the surface brightness at $\lambda6$ cm of 20% and a spectral index $\alpha\rm _n$ of the nonthermal emission of 0.85, which is the mean value for spiral galaxies of type Sb and later (Niklas et al. 1997).

Spectral indices $\alpha$ between $\lambda 22$ cm and $\lambda6$ cm ( $S_{\nu}\propto \nu^{-\alpha}$) can be computed from our VLA data (given in Table 5). The values lie in the range 0.71 and 0.97 which is in the range typical of normal spiral galaxies (Niklas 1995). Nonthermal spectral indices $\alpha\rm _n$ cannot be determined with data at only two frequencies.

We adopted the standard cosmic-ray proton-to-electron ratio K of 100, a pathlength through the disc of 1 kpc/$\vert\cos i\vert$, and assumed that the regular field is in the galaxy's plane and the random field is statistically isotropic. Uncertainties in any of these parameters of $\le$50% lead to an error of $\le$13% in $B_{\rm tot}$. We estimate the total error in $B_{\rm tot}$ to be about 30%. The relative errors between galaxies are smaller because some of the input parameters (e.g. the proton-to-electron ratio) are not expected to vary strongly from one galaxy to another.

With the above assumptions, $B_{\rm tot}$ is related to the average synchrotron volume emissivity $\epsilon$and surface brightness I (neglecting a term weakly varying with inclination i) by

$$B_{\rm tot}\propto\epsilon^{1/(\alpha_{\rm n} +3)}\quad (\mbox{where }\epsilon \propto I \vert\cos i\vert).$$

Note that the equipartition field strengths are about 10% larger than the field strengths derived from the standard minimum-energy formula (which should be used with caution, see Beck 2000).

The average total magnetic field strength $B_{\rm tot}$, according to Tables 5 and 6 (representing the average synchrotron emissivity), is a function of neither Hubble type (SBb-SBc) nor luminosity class (I-III), which has also been found for a much larger sample of barred and non-barred spiral galaxies (Hummel 1981). The average total field strength $B_{\rm tot}$ is $10\pm3~\mu{\rm G}$ for our sample, similar to the average minimum-energy field strength of ${\simeq}8~\mu{\rm G}$for the large galaxy sample (Hummel et al. 1988) and to the mean equipartition value of $11\pm4~\mu{\rm G}$ of the sample of 146 late-type galaxies calculated by Fitt & Alexander (1993), corrected to K=100. Niklas (1995) derived a mean equipartition value of $9\pm3~\mu{\rm G}$for his sample of 74 spiral galaxies. Hummel (1981) also found no significant emissivity difference between barred and non-barred galaxies.

The following galaxies have the strongest total magnetic field in our sample, as evidenced by their high radio surface brightness: NGC 1365, 1559, 1672 and 7552 (see Tables 5 and 6). This indicates that the total field strength is highest for galaxies with the (relatively) longest bars, with the exception of NGC 1559 that has a short bar (see Fig. 4).

The last column in Tables 5 and 6 gives the average equipartition strength $B_{\rm reg}$ of the resolved regular magnetic field, derived from the polarized surface brightness averaged over the galaxy. $B_{\rm reg}$, in contrast to $B_{\rm tot}$, depends on the linear resolution within a galaxy and thus on its physical size, its distance and its inclination. However, $p_{\lambda }$ and $B_{\rm reg}$ in Tables 5 and 6 do not correlate with distance of the galaxy. As a test, we smoothed the $\lambda$6 cm map of NGC 1097 by enlarging the beam size from 30 $^{\prime \prime }$ to 60 $^{\prime \prime }$ and to 90 $^{\prime \prime }$  which corresponds to increasing the galactic distance by factors 2 and 3. The degree of polarization decreased from 8.5% to 7% and 6%, respectively, and the strength of the resolved regular magnetic field decreased from 4.3 to 4.0 and $3.6~\mu{\rm G}$, respectively, remaining above the sample average. Hence the values of $p_{\lambda }$ and $B_{\rm reg}$in Tables 5 and 6 seem to depend only weakly on distance to the galaxies, implying that our observations generally resolve most of the structure in the regular magnetic field, at least for large galaxies and at distances of up to about 40 Mpc.

Average polarized surface brightness (and thus $B_{\rm reg}$) values are similar for the galaxies of our sample. The exceptions are NGC 1097 and NGC 1559 with $B_{\rm reg}\simeq 4~\mu{\rm G}$, above the average of $2.5\pm0.8~\mu{\rm G}$. NGC 1097 probably drives a strong dynamo where field amplification is supported by shear in the velocity field (Moss et al. 2001, Paper II). The degree of polarization at $\lambda$6 cm, signature of the degree of uniformity of the resolved field, is also high in NGC 1097 (see Fig. 25). NGC 1559, 1672 and 7552 are similar candidates for a strong dynamo, but the present radio observations (Fig. 26) have insufficient linear resolution at the relatively large distances of these galaxies to reveal the true strength of the regular fields and their detailed structure.

NGC 1300, NGC 3992 and NGC 4535 have the highest degrees of polarization but only low total surface brightness. They host weak but ordered magnetic fields with spiral patterns, similar to those in non-barred galaxies (Beck 2000).

5.4 Bars and global magnetic field structure

A classification system of barred galaxies was introduced by Martinet & Friedli (1997), based on the axis ratio b/a (see Tables 1 and 2) and on the star formation rate (SFR) measured by the far-infrared luminosity. Galaxies with large b/a are generally weak in star formation (class I), but some have a high SFR (class II). Galaxies with small b/a have a large spread in SFR: from high (class III) to weak (class IV). Galaxies of class IV in Martinet & Friedli have strong bars, but low SFR (see Sect. 5.2).

Figure 17: Total intensity contours of NGC 1433, overlayed onto an optical image from the Digitized Sky Surveys. The contours are 1, 2, 3, 4, 6, 8 $\times 100~\mu$Jy/beam area.

Here we propose that there are basic differences among barred galaxies concerning their magnetic field structure and strength which may reflect physical properties of barred galaxies like the gas flow, the shock strength in the bar and the presence of a circumnuclear ring.

Firstly, barred galaxies can have low radio luminosity because they are small (NGC 1313, 1493 and 5068), or because their gas content and star formation activity is small in spite of their large bars (NGC 1300 and 1433). Little or no polarization is detected in these galaxies. In galaxies with small bars the radio continuum morphology is formed as a result of star formation in the spiral arms, as in NGC 2336, 3359, 3953, 3992, 4535, 5643, and also M 83 observed previously by Beck (2000). The bar is of little importance for the overall radio properties of these galaxies. The average degree of radio polarization (i.e., the degree of field regularity) seems to be controlled by the spiral structure rather than the bar, being low in flocculent spirals and high when massive spiral arms are present. Regular fields are often enhanced in interarm regions between optical spiral arms, e.g. in NGC 3359, NGC 4535 and M 83, similar to non-barred galaxies.

Secondly, galaxies with long bars and strong star formation have a high radio luminosity and a strong total magnetic field ( $B_{\rm tot}\ge10~\mu{\rm G}$) (NGC 1097, 1365, 1672, 2442 and 7552, and also NGC 3627 observed previously by Soida et al. 2001). NGC 1097, 1365, 1672 and 7552 have a high polarization surface brightness and a strong regular field which is enhanced upstream of the shock fronts in the bar. The magnetic field lines upstream of the dust lanes are oriented at large angles with respect to the bar and turn smoothly towards the dust lanes along the major axis of the bar. This is accompanied by large-scale field enhancements associated with, e.g., strong shear in the velocity field and/or strong dynamo action rather than enhanced gas density. Gas inflow along the bar may lead to circumnuclear rings which have been detected already in NGC 1097 (Hummel et al. 1987; Gerin et al. 1988), NGC 2442 (Mihos & Bothun 1997), NGC 7552 (Forbes et al. 1994a, 1994b) and possibly in NGC 1365 (Sandqvist et al. 1995), and should be searched for in the other radio-bright galaxies.

Figure 19: Total intensity contours and the observed B-vectors of polarized emission of NGC 1559, overlayed onto an optical image from the Digitized Sky Surveys. The basic contour levels are 300 and 100 $\mu$Jy/beam, the contour intervals and the vector scale are as in Fig. 6.

NGC 7479 is anomalous in the radio range as it possesses a nuclear "jet'' (Laine & Gottesman 1998). Indications of a weaker nuclear jet have been found in NGC 1365 by Sandqvist et al. (1995).

Figure 20: Total intensity contours and the observed B-vectors of polarized emission of NGC 1672, overlayed onto an optical image from the Digitized Sky Surveys. The basic contour levels are 100 and 40 $\mu$Jy/beam, the contour intervals and the vector scale are as in Fig. 6.

Figure 21: Total intensity contours and the observed B-vectors of polarized emission of NGC 2442, overlayed onto an optical image from the Digitized Sky Surveys. The contours and the vector scale are as in Fig. 19.

For NGC 986, 1559 and 3059 the resolution and sensitivity of the present observations are insufficient to reveal their detailed field structure.

Figure 23: Total intensity contours and the observed B-vectors of polarized emission of NGC 5643, overlayed onto an optical image from the Digitized Sky Surveys. The basic contour level is 100 $\mu$Jy/beam area. The contour intervals and the vector scale are as in Fig. 6.

Figure 24: Total intensity contours and the observed B-vectors of polarized emission of NGC 7552, overlayed onto an optical image from the Digitized Sky Surveys. The basic contour level is 100 $\mu$Jy/beam area. The contour intervals and the vector scale are as in Fig. 6.

Figure 25: Polarized intensity contours and the observed B-vectors of NGC 1097, NGC 1365, NGC 4535 and NGC 7479. The contour intervals are 1, 2, 3, 4, 6, 8, 12, 16 $\times$ the basic contour level, which is 50, 30, 30 and 30 $\mu$Jy/beam in the order of increasing NGC number. A vector of 1 $^{\prime \prime }$ length corresponds to a polarized intensity of 20 $\mu$Jy/beam area.

Figure 26: Polarized intensity contours and the observed B-vectors of NGC 1559, NGC 1672, NGC 2442 and NGC 7552. The contour intervals are 1, 2, 3, 4, 6, 8, 12 $\times$ the basic contour level, which is 50 $\mu$Jy/beam. A vector of 1 $^{\prime \prime }$ length corresponds to a polarized intensity of 10 $\mu$Jy/beam area.

NGC 7552 is a special case. Its radio surface brightness is high (i.e., the total magnetic field is strong, see Table 6), but still too low to be consistent with its far-infrared flux density (see Fig. 2). NGC 7552 hosts a starburst ring and may drive a "galactic superwind'' (Forbes et al. 1994a). As a member of a galaxy group, it may be subject to tidal interactions. It seems possible that the magnetic field is still not strong enough to hold the large number of cosmic-ray electrons produced due to the high star formation activity. However, major distortions of our radio map by instrumental effects cannot be excluded. Further radio observations are required.