THE EXTENT OF MAGNETIC FIELDS AROUND GALAXIES OUT TO z ∼ 1^*

The origin and evolution of magnetic fields in galaxies over cosmic time is still observationally largely unconstrained. For normal galaxies at significant look-back times, statistical studies of the Faraday rotation effect on luminous polarized background sources provide the most effective way to probe magnetic fields. In Bernet et al. (2008, 2010), we presented evidence that quasars with strong intervening Mg ii absorption lines in their optical spectra (with equivalent width EW₀ > 0.3 Å) have a significantly broader distribution of rotation measure (RM) than those without. Bernet et al. (2010) showed that this was unlikely to be caused to any indirect correlation with the quasar redshift since the effect was not present for sightlines with weaker Mg ii absorption. Since strong Mg ii absorption is known to be generally associated with the halos of normal galaxies, the simplest interpretation was that ∼10 μG large scale magnetic fields exist in or around galaxies out to z ∼ 1.3 (Bernet et al. 2008).

In this Letter, we study the radio properties of background quasars at different impact parameters from the Mg ii host galaxies responsible for the enhanced RM. A similar approach has been used to study individual nearby galaxies, e.g., M31 (Han et al. 1998), the LMC (Gaensler et al. 2005) and the SMC (Mao et al. 2008), which have numerous polarized background sources available. For distant galaxies, this is not viable and a statistical approach is necessary.

Our study reveals that strong magnetic fields are present around galaxies out to large impact parameters of the order of 50 kpc. Applying recent results from the study of Mg ii systems (Bordoloi et al. 2011, 2012), this suggests that the ubiquitous winds in high redshift galaxies (Weiner et al. 2009; Bordoloi et al. 2011; Rubin et al. 2010) are highly magnetized. Furthermore, this finding provides support for the idea that magnetized outflows help remove small-scale helicity from galactic disks (Shukurov et al. 2006), preventing the quenching of the α–Ω dynamo mechanism (Vainshtein & Cattaneo 1992) thought to generate the large-scale magnetic fields in galactic disks (Brandenburg & Subramanian 2005).

Where required, we assume a concordance cosmology with h = 0.71, Ω_M = 0.27, and Ω_Λ = 0.73.

We obtained images of the fields of 28 radio quasars with strong Mg ii absorption lines selected from the sample of Bernet et al. (2008).

Images were taken in three bands with the EFOSC2 instrument at the New Technology Telescope in P82 from 2008 October 30–2008 November 2 and in P85 from 2010 March 15–2010 March 18. The quasar fields with absorbers in the redshift range 0.4–0.8 were observed with g, r, i filters and those with absorbers in the range 0.8–1.4 with the r, i, z filters. These filters were chosen in order to straddle the 4000 Å break at the redshifts of the absorber. The total exposure times in each filter varied between 600–6000 s, depending on the redshift of the absorbers. We aimed to detect galaxies down to 0.1 L^⋆3 at the absorber redshift. To facilitate the subtraction of the point-spread function of the quasar, the camera was used in the 1 × 1 binning mode, giving a pixel scale of 0.12 arcsec pixel⁻¹.

In order to identify the Mg ii host galaxies, we place the galaxies in the quasar field on a color–color diagram and compare them with the theoretical loci of galaxies of different types computed from spectral energy distributions (SEDs) from Coleman et al. (1980) as a function of redshift. The photometry of the galaxies was done using SExtractor (Bertin & Arnouts 1996). The flux was measured in each filter using circular apertures with a diameter of typically 30 pixels, corresponding to 3.6 arcsec. This was reduced when necessary to avoid contamination by neighboring objects.

To identify the host galaxy we proceed as follows.

• 1.  
  Measure the (r − i) and (g − r) or (i − z) colors of all the galaxies within 120 kpc of the quasar at the redshift of the absorber. In cases where there was no detection of a galaxy in one of the bands, upper limits were calculated for that band.
• 2.  
  All galaxies that have colors in the color–color diagram consistent with the locus of SED-types at the redshift of the Mg ii system (see Figure 1) are considered to be candidate host galaxies. When, in regions of color–color space, there is a degeneracy between SED-type and redshift, a morphological classification was done in order to separate these two quantities.
• 3.  
  For objects very close to the quasar, within ∼2'', it was not possible to perform accurate photometry of the galaxies. We assume that all three objects within ∼2'' are the host (see Chen et al. 2010).
• 4.  
  In cases where there were two candidate host galaxies, both with consistent colors, the one closer to the quasar was selected when the impact parameter differed by more than a factor two. For cases where they differed by less than a factor two, the impact parameters were averaged (PKS2204−54, 4C+19.34, 4C+13.46, PKS0506−61, PKS0038−020).

Zoom In Zoom Out Reset image size

The properties of the identified host galaxies are given in Table 1. While there is inevitably a certain arbitrariness to the identification of the host galaxies, we emphasize that this process was carried out blind with respect to the RM values of the quasars in order to preserve the statistical integrity of the analysis. For the quasar fields 4C+06.41 (Lanzetta et al. 1995) and 4C+19.44 (Kacprzak et al. 2008), spectroscopic redshifts of galaxies are available and those agree with our choice of host galaxies.

Table 1. Properties of Candidate Host Galaxies

Quasar	zquasar	R.A.	Decl.	zMg ii	W0(2796)	ΔR.A.	ΔDecl.	Ang. Sep.	D	mg	mr	mi	mz	Type	Symbol
		(J2000)	(J2000)		(Å)	(arcsec)	(arcsec)	(arcsec)	(kpc)
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)	(16)
4C-02.55	1.043	12:32:00.0	−02:24:05	0.39524	2.03	0.6	−1.9	2.0	10.6	...	...	...	...	ucl.	$\Box$
4C-02.55	...	...	...	0.75689	0.30	1.6	−0.1	1.6	11.8	...	...	...	...	ucl.	$\Box$
MRC0122−003	1.07	01:25:28.8	−00:05:56	0.39943	0.47	−16.8	24.8	30.0	159.9a	...	...	...	...	ucl.	x
MRC0122−003	...	...	...	0.3971	...	−8.4	−12.1	14.7	76.1b	22.21	20.88	20.52	...	Sbc	x
PKS2326−477	1.299	23:29:17.7	−47:30:19	0.43195	0.38	3.7	−6.8	7.7	43.1	...	21.49	20.92	20.64	Ell	•
PKS2326−477	...	...	...	1.26074	0.66	1.6	5.4	5.6	47.3	...	24.83	24.14	23.80	Sbc	$\Box$
4C+06.41	1.270	10:41:17.1	+06:10:17	0.44151	0.69	9.4	2.2	9.7	55.0	22.96	21.14	20.49	...	Ell	$\Box$
4C+19.44	0.720	13:57:04.4	+19:19:07	0.45653	0.85	1.6	7.8	7.9	45.8	23.03	21.67	21.17	...	Sbc	•
PKS1244−255	0.633	12:46:46.8	−25:47:49	0.49286	0.68	−3.9	2.3	4.5	27.1	24.47	22.92	22.25		Sbc	•
OC-65	0.733	01:41:25.8	−09:28:44	0.50046	0.53	−0.4	−1.4	1.5	9.1	...	...	...	...	ucl.	•
PKS0130−17	1.022	01:32:43.5	−16:54:49	0.50817	0.59	−3.9	−4.7	6.0	37.1	23.50	22.44	22.14	...	Scd	•
4C+19.34	0.828	10:24:44.8	+19:12:20	0.52766	1.00	−5.6	4.0	6.9	43.2	24.83	23.33	22.84	...	Sbc	○
PKS1615+029	1.339	16:17:49.9	02:46:43	0.52827	0.31	5.7	−2.3	6.2	38.8	24.31	23.01	22.50	...	Sbc	•
4C+01.24	1.024	09:09:10.1	+01:21:36	0.53587	0.44	−6.8	0.9	7.0	44.1	23.06	21.71	21.26	...	Sbc	•
PKSB1419−272	0.985	14:22:49.2	−27:27:56	0.55821	0.44	11.2	8.0	13.7	88.6	23.36	21.85	20.91	...	Ell	•
OX+57	1.932	21:36:38.6	+00:41:54	0.62855	0.60	5.6	−0.3	5.6	38.1	...	...	...	...	ucl.	•
OX-192	0.672	21:58:06.3	−15:01:09	0.63205	1.40	2.8	1.6	3.2	21.8	24.33	22.68	21.39	...	Ell	•
PKS0420−01	0.915	04:23:15.8	−01:20:33	0.63291	0.77	5.7	−13.3	13.5	92.5	24.09	22.45	21.68	...	Ell	•
3C208	1.112	08:53:08.6	+13:52:55	0.65262	0.62	0.3	−6.6	6.6	45.5	...	21.90	20.70	20.34	Ell	$\Box$
3C208	...	...	...	0.93537	0.40	−3.5	6.7	7.6	60.4	...	23.07	22.40	22.09	Scd	$\Box$
PKS0038−020	1.178	00:40:57.6	−01:46:32	0.68271	0.35	−10.8	−3.8	11.5	78.2	23.82	22.88	22.26	...	Scd	○
PKS2204−54	1.206	22:07:43.7	−53:46:34	0.6877	0.73	−9.7	−5.3	11.1	78.7	23.35	22.40	22.10	...	Scd	○
PKS0839+18	1.270	08:42:05.1	+18:35:41	0.71118	0.56	−5.4	5.4	7.6	54.7	22.70	21.71	21.14	...	Scd	•
4C+13.46	1.139	12:13:32.1	13:07:21	0.77189	1.29	−0.6	1.8	1.9	14.1	...	...	...	...	ucl.	○
4C+6.69	0.99	21:48:05.4	+06:57:39	0.79086	0.55	0.3	−5.8	5.8	43.4	23.82	22.74	22.15	...	Scd	•
PKS0402−362	1.417	04:03:53.7	−36:05:02	0.79688	1.80	1.8	2.5	3.1	23.2	23.28	22.61	21.83	...	Scd	•
PKS2223−05	1.404	22:25:47.2	−04:57:01	0.84652	0.60	−2.1	6.7	7.0	53.9	...	23.47	22.26	21.49	Ell	•
PKS0506−61	1.093	05:06:43.9	−61:09:41	0.92269	0.49	1.8	4.0	4.4	34.6	...	23.42	22.64	22.33	Scd	○
PKS0112−017	1.365	01:15:17.1	−01:27:05	1.18965	0.90	2.0	−2.6	3.3	27.4	...	...	...	...	ucl.	•
PKS0332−403	1.445	03:34:13.7	−40:08:25	1.20898	0.79	3.4	10.3	10.9	91.0	...	23.66	22.66	21.74	Ell	x
PKS0332−403	1.445	03:34:13.7	−40:08:25	0.8		−3.4	3.3	4.8	36.1c	...	23.31	21.93	21.43	Ell	x
PKS1143−245	1.940	11:46:08.1	−24:47:33	1.24514	0.30	10.0	0.3	10.0	84.0	...	23.70	23.19	22.53	Sbc	$\Box$
PKS1143−245	...	...	...	1.52066	0.46	2.2	−1.7	2.8	23.9	...	...	...	...	ucl.	$\Box$
OQ135	1.612	14:23:30.1	+11:59:51	1.36063	0.51	−12.9	−3.1	13.3	112.8	...	23.62	22.87	22.12	Sbc	•
Quasar fields with alternative candidate galaxies
PKS0130−17	1.022	01:32:43.5	−16:54:49	0.50817	0.59	11.8	6.0	13.2	81.0	22.61	21.76	21.42	...	Sbc	...
4C+19.34	0.828	10:24:44.8	+19:12:20	0.52766	1.00	2.2	−3.4	4.0	25.0	...	...	...	...	ucl.	...
PKS1615+029	1.339	16:17:49.9	02:46:43	0.52827	0.31	3.9	4.8	6.2	38.8	22.66	21.19	20.43	...	Ell	...
4C+01.24	1.024	09:09:10.1	+01:21:36	0.53587	0.44	11.7	10.3	15.5	97.8	24.16	22.53	21.80	...	Ell	...
PKS0038−020	1.178	00:40:57.6	−01:46:32	0.68271	0.35	13.2	8.3	15.6	110.3	23.43	22.64	22.06	...	Scd	...
PKS2204−54	1.206	22:07:43.7	−53:46:34	0.6877	0.73	3.5	−7.4	8.2	58.1	24.12	24.04	23.56	...	Irr	...
PKS0839+18	1.270	08:42:05.1	+18:35:41	0.71118	0.56	−3.3	7.5	8.1	58.2	24.03	22.64	21.91	...	Sbc	...
4C+13.46	1.139	12:13:32.1	13:07:21	0.77189	1.29	3.1	0.4	3.1	23.0	24.13	22.30	21.14	...	Ell	...
PKS0506−61	1.093	05:06:43.9	−61:09:41	0.92269	0.49	−7.6	−1.4	7.7	60.6	...	23.08	22.39	22.07	Scd	...
PKS1143−245	1.940	11:46:08.1	−24:47:33	1.52066	0.46	−9.1	−3.6	9.7	82.9	...	23.61	23.42	22.79	...	Scd

Notes. Columns: (1) name of the source; (2) redshift of the source; (3) and (4) quasar's coordinates; (5) redshift of the Mg ii system; (6) equivalent width W₀(2796); (7) and (8) coordinates of galaxies with respect to quasar; (9) angular separation; (10) impact parameter D; (11)–(14) apparent magnitudes in g, r, i, z; (15) type of galaxy; (16) symbol used in Figure 2. ^aFrom Chen et al. (2001). ^bWeak Mg ii absorber at z = 0.3971. ^cGalaxy at z ∼ 0.8.

Download table as:  ASCIITypeset image

The RMs in this work are selected from the sample of Kronberg et al. (2008) at Galactic latitudes |b| > 30°. At least three polarization angles, typically measured at wavelengths around 6 cm, were used for the RM determination (for more details, see Bernet et al. 2012).

3.1. |RM| versus Impact Parameters

In Figure 2, we plot the observed |RM|, uncorrected for Galactic foreground, for the 28 sightlines containing strong Mg ii absorption systems against the impact parameters D to the identified host galaxies. The filled circles show sightlines with a single Mg ii system and a uniquely identified intervening galaxy. The open circles correspond to the case when two candidate host galaxies are plotted at an average impact parameter (see (4) above). The crossed symbols correspond to the case in which a foreground galaxy lies closer to the line of sight (LOS) than the Mg ii host galaxy and which is therefore likely to have contaminated the RM value. The open squares indicate quasars sightlines with two strong Mg ii systems. Full details about each sightline are provided in Table 1.

Zoom In Zoom Out Reset image size

Figure 2 shows that the |RM| increases significantly toward sightlines with smaller impact parameters. All |RM| values >50 rad m⁻² are at D < 50 kpc. At a significantly larger impact parameter, >60 kpc, the RM is mostly given by the Milky Way contribution and observational error (horizontal dash line), which amount to ∼20 rad m⁻² (Bernet et al. 2008). This is similar to the 68% percentile spread of the |RM| for quasars without Mg ii absorption, i.e., to $|\rm {RM}|_{68} \sim 25 \;\rm {rad}\; \rm {m}^{-2}$ (Bernet et al. 2010). Note that the Galactic foreground as in, e.g., Oppermann et al. (2012), does not introduce any trend with impact parameter, while it contributes to our sources a median $|\rm {RM}| 12.6\; \rm {rad}\; \rm {m}^{-2}$.

A Pearson test for correlation for the circles in Figure 2 shows that |RM| and D are anticorrelated with ρ = −0.43, corresponding to a chance probability of p = 4.8%. A two-sided Kolmogorov–Smirnov (KS) test for the same data reveals a chance probability that the |RM| distributions below and above 40 kpc are drawn from the same distribution of p = 2.2%.

This analysis corroborates the above result that the RM contribution is negligible for sightlines beyond 60 kpc. We can therefore try to repeat the KS test including the two sightlines with closer foreground galaxies without Mg ii absorption (× symbols), and also the four sightlines with two Mg ii systems (open squares), assuming that only galaxies with D < 50 kpc contribute RM. This lowers the above chance probability to p = 1.5% and 1.9% if we split the sample at 40 and 50 kpc, respectively.

Using the sightlines with one Mg ii absorber at D < 50 kpc, we determine an observed dispersion of RM $\sigma _{{\rm obs}} \sim 65\; \rm {rad}\: \rm {m}^{-2}$. The contribution from the intrinsic RM, the Milky Way RM, and the observational error is found from sightlines without absorbers to be ∼25 rad m⁻². Subtracting in quadrature and multiplying by (1 + z_Mg ii)², we obtain a rest frame RM dispersion for the Mg ii absorbers $\sigma _{{\rm Mg\,\scriptsize{II}}} \sim 150\; \rm {rad}\: \rm {m}^{-2}$.

3.2. Inhomogeneity of the RM Screens

In Bernet et al. (2012), we studied the effect of depolarization due to inhomogeneous Faraday screens in intervening galaxies at redshift z, with RM dispersion σ_RM and covering factor f_c. We predicted a wavelength λ dependent effect, $p(\lambda ^{2})/p_{0}=f_{c}\exp (-2\sigma _{{\rm RM}}^{2}(1+z)^{-4}\lambda ^{4}) +(1-f_{c}),$ where p₀ is the intrinsic polarization. Since, according to Figure 2, the observed RM dispersion increases for a smaller impact parameter, we expect the degree of polarization to follow a similar pattern.

In order to test the depolarization potentially suffered by our sources, we use the ratio p₂₁/p_1.5, where p₂₁ and p_1.5 are the degrees of polarization at 21 and 1.5 cm, from Taylor et al. (2009) and Condon et al. (1998), and from Jackson et al. (2010) and Murphy et al. (2010), respectively. Since, in general, the short and long wavelength emission originate from different components of the radio source (i.e., the compact core and the radio lobes), the above ratio is not a "measure" of depolarization. However, due to the wavelength dependence of the depolarization effect, the ratio p₂₁/p_1.5 will statistically be lower the stronger the depolarizing effect along the LOS.

In Figure 3, we plot the ratio p₂₁/p_1.5 versus the impact parameters D to the galaxies. The solid and open circles indicate if the quasar redshift is >1.0 or <1.0, respectively. Due to the low number of quasar fields for which both p₂₁/p_1.5 and D are available, this plot also includes two sightlines with multiple absorbers, shown as solid squares. For one of them (PKS1143−245), one galaxy is at 21 kpc and the other at 84 kpc, so it is likely that only the closer one is contributing. For the other (4C-02.55), both galaxies are close, at 10.6 kpc and 11.8 kpc from the sightline, respectively, so both are likely contributing to the observed RM.

Zoom In Zoom Out Reset image size

A clear trend is visible in Figure 3 in the left panel, whereby the lower the impact parameter, the lower the value of p₂₁/p_1.5. Kendall's τ test shows that, for the overall sample, p₂₁/p_1.5 and D are correlated with τ = 0.30 and a chance probability of 11.6%. Knowing that quasars with z_QSO < 1.0 are intrinsically more depolarized than those above this redshift (Bernet et al. 2012), we split the sample according to this redshift. This effectively separates out two objects with low p₂₁/p_1.5 values at high impact parameters. Repeating the test for the z_QSO > 1.0 sample shows that p₂₁/p_1.5 and D are strongly correlated with τ = 0.56 and a chance probability of only 1.7%. For comparison, the p₂₁/p_1.5 for sightlines from the whole Bernet et al. (2008) sample that do not have Mg ii absorption systems and are at z_QSO > 1.0 are plotted in the right panel in Figure 3.

As in Bernet et al. (2012), a Faraday screen can be modeled as a collection of cells of size l_c, the magnetic field coherence length. The (rest frame) RM dispersion of the cells on a Faraday screen of depth L is then written as:

$$\begin{equation} \sigma _{{\rm RM}}=\sigma _{c}\sqrt{L/l_{c}}\propto Bn_{e}l_{c}\sqrt{L/l_{c}}, \end{equation} \tag{ 1 }$$

where σ_c∝Bn_el_c is the RM of a single cell. We can now relate σ_RM to the (rest-frame) RM dispersion characterizing sightlines through a galactic Faraday screen, $\sigma _{{\rm Mg\,\scriptsize{II}}}$, by

$$\begin{equation} \sigma _{{\rm RM}}=\sigma _{{\rm Mg\,\scriptsize{II}}}\sqrt{N/f_{c}}, \end{equation} \tag{ 2 }$$

where N is the number of surface cells covering the inhomogeneous screen and f_c its covering factor. Using the value of σ_Mg ii estimated at the end of last section, we see that the decrease in p₂₁/p_1.5 toward lower impact parameters can be explained by the observed increase in |RM| alone (Figure 2) for N ∼ a few. In general, however, the coherence scale of the RM, i.e., N, and f_c might also change as a function of D.

The simultaneous increase toward small impact parameters of the RM dispersion, on the one hand, and depolarizing effects as probed by p₂₁/p_1.5, on the other, is consistent with and corroborates previous results (Bernet et al. 2008), suggesting the presence of significant magnetic fields in and around galaxies at distances of several tens of kiloparsecs, acting as inhomogeneous Faraday screens (Bernet et al. 2012).

In principle, the observed RM can be attributed to magnetic fields either in the disk or in the halo of the intervening galaxies. The RM associated with the disk magnetic field can be probed using a simple axisymmetric spiral magnetic field model and a distribution of free electrons, n_e, both with radial scale length and vertical exponential scale height of 30 kpc and 1.8 kpc, respectively (Gómez et al. 2001; Gaensler et al. 2008), and with normalizations, B₀ = 10 μG, n_e0 ∼ 0.03 cm⁻³, at the galactic center. A simple Monte-Carlo model for the RM from a sample of galaxies at random orientation along the LOS then shows that a pure disk magnetic field is not able to account for the observed |RM| versus D reported in Figure 2, unless the normalizations or the vertical and radial scales are made unrealistically large. This implies that, while the disk magnetic field may contribute to the observed RM, other components are most likely present.

It is now established that galactic winds appear in galaxies with star formation rates (SFRs) at low (Bouché et al. 2012), intermediate (Bordoloi et al. 2011; Weiner et al. 2009; Rubin et al. 2010), and high redshifts (Pettini et al. 2002). Bordoloi et al. (2011) find that Mg ii absorption in foreground edge-on galaxies at 0.5 < z < 0.9 shows a strong azimuthal dependence within 50 kpc, indicating the presence of bipolar outflows around the disk rotation axis. Bordoloi et al. (2012) show that most Mg ii quasar absorption systems also lie within 45° of the minor axis and at D < 40 kpc.

Since our Mg ii systems show statistically the same properties, the sightlines in our sample will pass through regions above the poles (this will be soon tested with Hubble Space Telescope imaging of most of the quasar fields in our sample). Further magnetized outflows, although at significantly smaller distances of a few kiloparsec, have also been detected in local galaxies (Haverkorn & Heesen 2012). Therefore, it is likely that the high RM values in Figure 2 are associated with magnetized outflows.

Can we obtain meaningful column densities of free electron N_e in the outflows? To calculate N_e, we use a simple wind model based on the work of Bouché et al. (2012). We assume that the outflow rate is proportional to the SFR, $\dot{M}_{{\rm out}}=\eta \, \rm {SFR}$. In addition, for a biconical geometry and constant outflow velocity v_out, $\dot{M}_{{\rm out}} \simeq ({\pi }/{2})\mu N_{g}D v_{{\rm out}} \Theta _{{\rm max}},$ with N_g the gas column density measured at impact parameter D from a sightline transverse to the outflow, Θ_max the opening angle and μ the mean atomic weight (Bouché et al. 2012). For a fully ionized gas, we can then estimate the free electron column density as N_e ≈ 9 × 10¹⁹ cm⁻²(η/0.5)(SFR/10 M_☉ yr⁻¹) × (v_out/200 km s⁻¹)⁻¹(D/30 kpc)⁻¹(Θ_max/30°)⁻¹, which is essentially the same as the obtained estimate in Bernet et al. (2008) for typical Mg ii systems, based on H i measurements and an estimate of the ionization correction. This shows that the column density at large D can be substantial in outflows of normal galaxies. Using Equations (1) and (2), we then infer a magnetic field strength

$$\begin{eqnarray} B_{\parallel } & =& 54\ {\rm \mu G} \left (\frac{\sigma _{{\rm Mg\,\scriptsize{II}}}}{150\ {\rm rad\: m}^{-2}} \right) \left (\frac{s}{10\ {\rm kpc}} \right) \left (\frac{L}{10\ {\rm kpc}} \right)^{0.5} \nonumber \\ && \times \left (\frac{f_{c}}{0.5} \right)^{-0.5} \left (\frac{N_{e}}{9 \times 10^{19}\ {{\rm cm}}^{-2}} \right)^{-1} \left (\frac{l_{c}}{3\ {\rm kpc}} \right)^{-1.5}, \qquad \end{eqnarray} \tag{ 3 }$$

where we have set the number of independent RM cells to $N=s^{2}/l_{c}^{2}$, with s the projected linear size of the source at z_Mg ii and have assumed that the scale of the RM fluctuation, l_c, is the same along the sightline and in the plane of the sky. This value of B is higher than the ∼10 μG obtained in Bernet et al. (2008) because here we consider the RM screens to be inhomogeneous.

This simple analysis shows that magnetized winds can account for the observed |RM| at large impact parameter reported in Figure 2. This requires a magnetic field strength B of several tens of μG, which is considerably larger than the few μG fields observed in large scale outflows in nearby galaxies, e.g., NGC 5775 (Tüllmann et al. 2000) and NGC 4666 (Dahlem et al. 1997).

However, fields of such strength are required in high redshift galaxies by the lack of evolution in the FIR–radio correlation (Condon 1992; Ivison et al. 2010; Sargent et al. 2010). In fact, the radio-emitting electrons would otherwise mostly radiate their energy through inverse Compton scattering on the cosmic microwave background and/or starlight, both significantly higher at z ∼ 1 due to cosmological expansion and a ∼10 × higher SFR, respectively. Since this is not observed, the total magnetic field must have been larger by ∝max [SFR^1/2, (1 + z)²] ∼ 4. If this was the case for both large and small scale fields, then values of the order of a few tens of μG are found for z ∼ 1 galaxies, if our reference low-z field is the Galactic one at several μG.

The existence of magnetized outflows in normal galaxies has important implications for α–Ω models of galactic dynamos. Galactic dynamos are thought to be responsible for the origin of large scale magnetic fields in spiral galaxies. However, their efficiency can be severely limited by conservation of magnetic helicity once the small scale magnetic field reaches equipartition with the small scale kinetic energy of interstellar gas (Vainshtein & Cattaneo 1992; Brandenburg & Subramanian 2005). However, galactic winds can transport magnetic helicity away from the plane of the galaxy and restore the efficiency of α–Ω dynamos, as proposed by Shukurov et al. (2006). Assuming again the coexistence of large and small scale fields, our observations support the occurrence of this process and could represent a first observational link between galactic dynamos and magnetized winds at intermediate redshifts.

Understanding the nature of magnetic fields in and around intermediate redshift galaxies deserves further work.

We thank an anonymous referee for valuable comments. This research has been supported by the Swiss National Science Foundation and made use of observational facilities of the European Southern Observatory (ESO).