Strong Excess Faraday Rotation on the Inside of the Sagittarius Spiral Arm

Magnetic fields play an important role in the physics of the interstellar medium on a wide range of scales (e.g., Beck 2015). Plasma ejected by stellar winds and supernova explosions expands into the surrounding magnetized interstellar medium to form magnetized super bubbles (e.g., Ferrière et al. 1991; Tomisaka 1998; Stil et al. 2009), driving a complex feedback cycle between star formation and the magnetic field. The strength and structure of the magnetic field are important for understanding its origin and effect on the interstellar medium (Klessen & Glover 2016). On a Galactic scale the magnetic field is best observed through Faraday rotation of radio waves, by which the polarization angle θ of a linearly polarized source changes with wavelength, λ, according to Δθ = ϕλ². The Faraday depth ϕ is defined as

$$\begin{eqnarray}&&\phi =\displaystyle \frac{{e}^{3}}{2\pi {m}_{e}^{2}{c}^{4}}\int {n}_{e}{B}_{\parallel }{dl}=0.81\int \left(\displaystyle \frac{{n}_{e}}{{\mathrm{cm}}^{-3}}\right)\left(\displaystyle \frac{{B}_{\parallel }}{\mu {\rm{G}}}\right)\left(\displaystyle \frac{{dl}}{\mathrm{pc}}\right),\end{eqnarray} \tag{ 1 }$$

with ϕ in rad m⁻², c the speed of light, n_e the density of free electrons (mass m_e, charge −e), ${B}_{\parallel }$ the component of the magnetic field along the line of sight, and l the distance along the line of sight (Klein & Fletcher 2015). The integral is evaluated from the source to the observer such that positive ϕ corresponds to ${B}_{\parallel }$ toward the observer. Rotation measure (RM) is the slope of the relation between Δθ and λ². In its simplest form, RM = ϕ, but superposition of waves that experience different amounts of Faraday rotation can make RM a function of wavelength and different from ϕ (e.g., Sokoloff et al. 1998).

Most Galactic Faraday rotation is believed to originate in the warm ionized medium (WIM; Heiles & Haverkorn 2012). Observable Faraday rotation can arise from a plasma whose free–free continuum and spectral line emission are undetectable (Uyaniker et al. 2003), yet inversion of the integral in Equation (1) requires assumptions about the geometry, electron density, and the magnetic field configuration in the object. In case of the Galactic magnetic field, this inversion benefits from a unique combination of measurements along diverging lines of sight, and information about n_e from pulsar dispersion measure DM,

$$\begin{eqnarray}&&\mathrm{DM}=\int {n}_{e}{dl},\end{eqnarray} \tag{ 2 }$$

with DM measured in pc cm⁻³, and n_e and l in the same units as in Equation (1). Pulsar DMs indicate a higher plasma density in the spiral arms, but the sampling is not yet dense enough to map structures on smaller scales consistently (Taylor & Cordes 1993; Cordes & Lazio 2002).

Large sections of the Milky Way disk have been surveyed for polarization of extragalactic sources and pulsars to map the Galactic magnetic field (e.g., Broten et al. 1988; Clegg et al. 1992; Brown & Taylor 2001; Brown et al. 2003, 2007; Taylor et al. 2009; Han et al. 2018; Schitzeler et al. 2019), and converted into complete maps of Galactic Faraday rotation by Oppermann et al. (2012, 2015) and Hutschenreuter & Ensslin (2019). The inner first Galactic quadrant is covered relatively sparsely by the survey of Van Eck et al. (2011) that targeted polarized sources selected from the NRAO VLA Sky Survey (NVSS; Condon et al. 1998). Jansson & Farrar (2012) combined Faraday rotation through the disk and the halo with polarization of diffuse emission into a coherent model of the large-scale Galactic magnetic field. This large-scale field has a strength of a few μG and a direction that varies with distance from the Galactic center. The number of reversals and the precise geometry of the Galactic magnetic field is still a matter of debate (e.g., Brown et al. 2007; Han et al. 2018).

The contribution of spiral arms to the Faraday depth of the Galaxy affects the geometry and field strength in the remainder of the disk. Brown & Taylor (2001) discussed deviations in RM in the direction of H ii regions along the local Orion-Cygnus arm. Vallée et al. (1988) reported excess Faraday rotation of −75 rad m⁻² in the region of the Scutum arm tangent. In this Letter we present the first results of Faraday rotation of compact polarized sources from the The Hi/OH/Recombination line (THOR) survey in the region of the Sagittarius (Sgr) arm tangent.

The THOR survey (Beuther et al. 2016) covers the inner Galaxy in the longitude range 145 < ℓ < 674 and latitude −125 < b < 125 with the Karl G. Jansky Very Large Array (VLA) in C configuration in L band (1–2 GHz). The survey includes the 21 cm line of atomic hydrogen, OH lines, several radio recombination lines, and the continuum in 512 channels from 1 to 2 GHz. The 21 cm line and total intensity continuum were combined with archival data from the VLA Galactic Plane Survey (VGPS; Stil et al. 2006) and the Effelsberg continuum survey by Reich et al. (1990) at 1.4 GHz only. For the other spectral lines and continuum polarization, only the C-configuration data exist, sampling the continuum at 1.5 GHz on angular scales from ∼5' to ∼15''.

Radio frequency interference (RFI) flagging and standard flux and phase calibration of the continuum data were described by Beuther et al. (2016). Polarization calibration and imaging was done in CASA following standard procedures. The single phase calibrator observed during an observing session was used for polarization calibration, with 3C 286 used for polarization angle calibration. The visibilities were averaged into 8 MHz channels before imaging of Stokes I, Q, and U to reduce data volume and improve the signal to noise ratio per channel for cleaning. The noise is approximately 0.4 mJy beam⁻¹ per 8 MHz channel. Bihr et al. (2016) and Wang et al. (2018) compiled a list of compact continuum sources in the THOR survey. Sources with peak brightness more than 10 mJy beam⁻¹ were selected for polarization analysis. The input catalog is essentially complete at 10 mJy, but detection of the polarized signal depends on the local noise level, which can be raised by nearby bright diffuse emission. Our sample includes additional polarized components of resolved sources, but excludes occasional entries that were considered a part of bright diffuse sources after visual inspection. No other selection was made, but we follow the common implicit assumption that compact sources detected in polarization are extragalactic. Pulsars are rarely detectable in cm continuum imaging surveys such as the THOR continuum catalog (e.g., Dai et al. 2016).

Analysis of the THOR polarization image cubes begins with Faraday Rotation Measure Synthesis (RM synthesis; Brentjens & De Bruyn 2005). The complex polarization ${ \mathcal P }$ as a function of wavelength λ is expressed in terms of the normalized Stokes parameters q = Q/I and u = U/I as ${ \mathcal P }=q+{iu}$. The dimensionless Faraday depth spectrum $\tilde{{ \mathcal F }}(\phi )$ is obtained by the Fourier transform

$$\begin{eqnarray}&&\tilde{{ \mathcal F }}(\phi )={\int }_{-\infty }^{\infty }{ \mathcal P }(\xi )W(\xi )\exp \left[-2\pi i\phi \xi \right]d\xi ,\end{eqnarray} \tag{ 3 }$$

where ξ = λ² for ξ > 0 and the weight function W(ξ) = 0 where no measurements exist, including ξ < 0. Multiple values of ϕ can arise from blending different rates of Faraday rotation, by integrating over frequency, solid angle, or different emission regions along the line of sight. This is referred to as Faraday complexity. It gives rise to fractional polarization changing with wavelength, and a nonlinear relation between polarization angle and λ². The rotation measure transfer function (RMTF) serves as the point-spread function in Faraday depth. It is the Fourier transform of the function W(ξ). RFI flagging reduces Faraday depth resolution to 101 rad m⁻² in the median and raises the side lobes of the RMTF, as shown in Figure 1.

Zoom In Zoom Out Reset image size

Significant Faraday rotation within a single frequency channel leads to depolarization. If the width of a channel in λ² is expressed as δλ², the maximum observable Faraday depth is (Brentjens & De Bruyn 2005),

$$\begin{eqnarray}&&| {\phi }_{\max }| =\displaystyle \frac{\sqrt{3}}{\delta {\lambda }^{2}}.\end{eqnarray} \tag{ 4 }$$

For 8 MHz channels at 1.5 GHz this amounts to $| {\phi }_{\max }| =4.0\,\times {10}^{3}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$, and at 1.8 GHz to $| {\phi }_{\max }| =7.1\times {10}^{3}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$, so we have sensitivity to somewhat higher Faraday depth from the highest observed frequencies, with reduced sensitivity because of the smaller effective bandwidth (see Pratley & Johnston-Hollitt 2019).

In this first exploration of the polarization survey, we present Faraday rotation of compact polarized extragalactic sources in the longitude range 39° < ℓ < 52°, and Faraday depth in the range −6000 rad m⁻² < ϕ < 6000 rad m⁻². In view of the high RMs encountered for 47° < ℓ < 49°, this region was also analyzed at full spectral resolution (2 MHz channels between 1.6 and 1.9 GHz) up to $| {\phi }_{\max }| =2.5\times {10}^{4}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$. No detections were found outside the initial search range, but three more high RM sources were added to the sample. Sources detected in polarization were analyzed with the RM Clean algorithm (Heald 2009) and QU fitting (Law et al. 2011), using the RMtools¹³ package of C. Purcell. Conceptually, RM synthesis resembles imaging of the visibilities in radio interferometry followed by a Clean deconvolution, while QU fitting resembles fitting a source in the visibility plane.

We present RMs for 127 compact polarized sources detected in the longitude range 39° < ℓ < 52°. Figure 1 shows Faraday rotation of the source G48.561−0.364 with RM = 2396 ± 8 rad m⁻². The Stokes Q and U spectra are well fitted by a model that depolarizes gradually to zero on the long-wavelength side of the band. RM Synthesis shows a distinct peak and RM Clean components spread in a range that reflects the gradual depolarization at longer wavelengths. Complex Faraday rotation and occasional high RMs in extragalactic sources are seen across the sky and believed intrinsic to the source (e.g., O'Sullivan et al. 2017). We adopt the location of the peak of $| \tilde{{ \mathcal F }}|$ as the RM, and identify it with the Faraday depth of the Galaxy. The value of the peak is the polarization degree of the source expressed as a percentage of Stokes I. This is consistent with previous surveys that had a smaller bandwidth. The RMs derived from RM synthesis are in general closely consistent with those derived from QU fitting. Figure 2 shows Faraday depth spectra before RM Clean for the 16 sources with the highest RM. Table 1 lists all detections that have RM > 1000 rad m⁻². The complete sample is available in the online material.

Zoom In Zoom Out Reset image size

Figure 3 shows RM as a function of Galactic longitude. We see a spike in Faraday rotation near ℓ ≈ 48°, in the longitude range of the Sgr arm tangent (Georgelin & Georgelin 1976; Beuermann et al. 1985; Beuther et al. 2012; Vallée 2014). In the range 47° < ℓ ≤ 49°, the mean RM is $\langle \mathrm{RM}\rangle =1.63\,\times {10}^{3}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$ with standard deviation σ_RM = 1.0 × 10³ rad m⁻² (23 sources). For ℓ > 49° we find only one RM > 800 rad m⁻² and a mean $\langle \mathrm{RM}\rangle =4.19\times {10}^{2}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$ and σ_RM = 2.4 × 10² rad m⁻² (29 sources), while we find many times RM > 800 rad m⁻² for ℓ ≤ 47° with a mean $\langle \mathrm{RM}\rangle =6.45\,\times {10}^{2}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$ and σ_RM = 3.9 × 10² rad m⁻² (75 sources). Subtracting the mean $\langle \mathrm{RM}\rangle$ for 49° < ℓ < 52° as an estimate of the Faraday depth of the remainder of the Galaxy (see models in Van Eck et al. 2011), the Sgr arm tangent contributes $\langle \mathrm{RM}\rangle =(1.2\pm 0.2)\times {10}^{3}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$, with standard deviation σ_RM ≈ 1.0 × 10³ rad m⁻². Several degrees from the tangent, the Sgr arm still adds ∼200 rad m⁻² in the mean and significant scatter to the total Galactic Faraday depth. The ratio $\langle \mathrm{RM}\rangle /{\sigma }_{\mathrm{RM}}\approx 1.7$ for each of the three regions. This may indicate anisotropy in the random component of the magnetic field (e.g., Brown & Taylor 2001; Beck 2015).

Zoom In Zoom Out Reset image size

We find good agreement for the sources measured by Van Eck et al. (2011) that lie within the THOR survey (large red dots in Figure 3). Many of our RMs are significantly higher than the range −129 rad m⁻² < RM < 831 rad m⁻² in Van Eck et al. (2011). The highest RMs in THOR are well above the threshold for bandwidth depolarization in the NVSS (Condon et al. 1998) from which their sample was selected. The green curve in Figure 3 shows the prediction of Galactic Faraday depth from Oppermann et al. (2015), and the cyan curve shows the prediction from Hutschenreuter & Ensslin (2019) that includes free–free emission as a prior for Faraday depth amplitude. The THOR RMs indicate that the Galactic Faraday depth is significantly higher than previously realized. This part of the Galactic plane is outside the survey of Schitzeler et al. (2019), but these authors also applied a threshold $| \mathrm{RM}| \lt 1000\,\mathrm{rad}\,{{\rm{m}}}^{-2}$ in their sample selection.

Figure 3 (right) shows the distribution of pulsar DMs in our (ℓ, b) range. There appears to be a discontinuity in DM between the highest DM pulsars and the remainder of the sample, up to 300 pc cm⁻³ around ℓ ≈ 48°, although the number of pulsars is small. If the two are related, the mean $\langle \mathrm{RM}\rangle \approx {10}^{3}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$ and the ΔDM ≈ 300 pc cm⁻³ suggest a mean magnetic field ${B}_{\parallel }=4\pm 1\ \mu {\rm{G}}$. The statistical error assumes 20% uncertainty in both $\langle \mathrm{RM}\rangle$ and ΔDM. A higher density of low-latitude pulsars is required for further investigation. Han et al. (2018) analyzed pulsars with $| b| \lt 8^\circ$ and 45° < ℓ < 60°. These pulsars show a more continuous distribution of DM, and a mean field ${B}_{\parallel }=1.4\pm 1.0\ \mu {\rm{G}}$.

Our highest RM values exceed those published for sources behind Galactic H ii regions. Vallée & Bignell (1983), Purcell et al. (2015), and Ma et al. (2019) reported $| \mathrm{RM}|$ up to 633 rad m⁻² behind the Gum nebula. Harvey-Smith et al. (2011) found $| \mathrm{RM}| \lesssim 300$ rad m⁻² behind a sample of large-diameter, high-latitude, H ii regions. Savage et al. (2013) and Costa et al. (2016) found $| \mathrm{RM}|$ up to 1383 rad m⁻² behind the Rosette Nebula. The massive star formation region W4 (Gray et al. 1999; Costa & Spangler 2018) and the Cygnus X region (Brown et al. 2003) have $| \mathrm{RM}|$ up to 1500 rad m⁻². Some of these studies are directly or indirectly subject to the effects of bandwidth depolarization in the NVSS, which is known to list fewer polarized sources behind H ii regions (Stil & Taylor 2007). Still, observations of Galactic H ii regions to date show $| \mathrm{RM}| \lesssim 1500\,\mathrm{rad}\,{{\rm{m}}}^{-2}$, with the high $| \mathrm{RM}|$ confined to regions outlined by the thermal radio emission. Very strong Faraday rotation has been detected from pulsars near the Galactic center, with $| \mathrm{RM}|$ up to 6.7 × 10⁴ rad m⁻² (Eatough et al. 2013; Schnitzeler et al. 2016), and $| \mathrm{RM}|$ up to 5800 rad m⁻² from the nonthermal filaments (Lang et al. 1999; Paré et al. 2019). In the Galactic disk, two pulsars, PSR J1841−0500 (Camilo et al. 2012) and PSR J1839−0643 (Han et al. 2018) have RM ≈ −3000 rad m⁻², and the magnetar PSR J1550−5418 has RM = −1860 rad m⁻².

In order to test whether the high RMs result from H ii regions along the line of sight, their positions were cross matched with the catalog of Galactic H ii regions of Anderson et al. (2014). This catalog is based on the mid-infrared WISE all sky survey, which provides a more sensitive census of H ii regions than the available radio continuum surveys. The 22 μm emission of stochastically heated small dust grains is spatially correlated with the ionized gas in an H ii region (Anderson et al. 2011). The catalog is complete in the first quadrant for H ii regions around a single star of spectral type O9.5 or earlier (W. Armentrout et al. 2019, in preparation). Table 1 lists the H ii region nearest to the line of sight selected by two criteria from the catalog of Anderson et al. (2014). The first listed H ii region is the nearest considering the angular distance expressed in units of the H ii region radius (d₁/R), and the second H ii region is the nearest selected by angular distance only (d₂). For each case, the angular distance and the separation in terms of the H ii region radius (R) is listed. We excluded the H ii region G49.048−0.886, because it is not in the most recent version (2.2) of the online catalog.¹⁴

Most high RMs are found away from detectable H ii regions. Figure 4 shows RMs of extragalactic sources (+), pulsars (×), and pulsar DMs (white squares) on a color-composite image of the Galactic plane that combines WISE 4.6 μm, 12 μm, and 22 μm, Midcourse Space Experiment 8 μm, and THOR radio continuum images. H ii regions appear as pink or bright yellow nebulae, while supernova remnants appear as blue nebulae. The highest RM is found on a line of sight that passes within ∼50 pc from the star formation region W51, which is the 9th most luminous source of free–free emission in the Galaxy (Rahman & Murray 2010) at a distance of 5.4 kpc (Sato et al. 2010). The median $| \mathrm{RM}|$ for 79 sources whose line of sight passes more than 2 times the radius R from the nearest H ii region is 700 rad m⁻². This includes most of the high RM sources listed in Table 1. The median $| \mathrm{RM}|$ for 14 sources whose line of sight passes within the radius of an H ii region (d₁ < R) is 439 rad m⁻². The difference is hardly significant considering the sample size and the large scatter in RM. The RM spike is about a degree (≈100 pc) from W51, on the inside of the spiral arm. This is upstream when considering the motion of gas and stars through the spiral arm.

Zoom In Zoom Out Reset image size

The lack of correlation of high RMs with the H ii region catalog of Anderson et al. (2014) indicates the high RMs arise in the more diffuse WIM. The excess RMs are all positive, indicating a mean magnetic field component toward the observer in the region in which they arise. Several of our RMs are significantly higher than RMs from star formation regions with strong thermal radio continuum emission. Emission from a Faraday rotating plasma may be unobservable if the plasma is spread out over a long distance along the line of sight, but a large RM from a region with small line of sight depth must result in a high emission measure unless the magnetic field strength is elevated. We consider two options: associate the RM spike at ℓ ≈ 48° with a confined region, e.g., the wall of a super bubble or an extended ionized halo around W51, with line of sight dimension L ≈ 100 pc, or with the diffuse WIM in the spiral arm, with line of sight dimension L ≈ 1 kpc. Within L we assume a filling factor of unity. A smaller filling factor would be equivalent to reducing L. As an order of magnitude estimate, we write $\langle \mathrm{RM}\rangle \approx 0.81{n}_{e}{B}_{\parallel }L\approx {10}^{3}\,\mathrm{rad}\,{{\rm{m}}}^{-2}$ for 47° < ℓ < 49°. We focus the discussion on $\langle \mathrm{RM}\rangle$, which includes several sources far from W51, noting that the extreme RMs near ℓ = 485, b = −06 may be affected by W51 (compare Figure 4 with the radio recombination line map of W51 presented by Liu et al. 2019).

The case L = 100 pc then implies a mean ${n}_{e}{B}_{\parallel }\,\approx 10\,\mu {\rm{G}}\,{\mathrm{cm}}^{-3}$. For a mean line of sight magnetic field ${B}_{\parallel }\,=5\,\mu {\rm{G}}$, this implies n_e ≈ 2 cm⁻³, and emission measure 400 pc cm⁻⁶. The free–free emission from such a region would have a brightness temperature of 0.6 K at 1.4 GHz, for plasma temperature T_e = 10⁴ K. The implied emission measure for the highest RM ≳ 3000 rad m⁻² is an order of magnitude higher. We see structures on scales of 05 with a contrast ∼1 K in the THOR continuum image, but no excess that traces the higher RMs. On the other hand, free–free emission from super bubble walls is clearly detected around W4 (Costa & Spangler 2018) and W47 (Beuther et al. 2016). The high RM sources of Costa & Spangler (2018), including O10, are enclosed by the contour T_B = 5.7 K, while the Galactic background is in the range 4.9–5.1 K. Also, Gray et al. (1999) found that depolarization of diffuse emission is closely correlated with total intensity. A stronger mean magnetic field ${B}_{\parallel }\gtrsim 15\,\mu {\rm{G}}$ (${B}_{\parallel }\gtrsim 45\,\mu {\rm{G}}$ for RM ≈ 3 × 10³ rad m⁻²) would reduce the implied emission measure beyond detection in the radio continuum image. Although the mean RM ≈ 10³ rad m⁻² at ℓ ≈ 48° is comparable to model predictions for super bubbles (Stil et al. 2009; Costa et al. 2016; Costa & Spangler 2018), the highest RMs are significantly larger.

If, however, the RM spike arises in the WIM of the Sgr arm with L ≈ 1 kpc, the product ${n}_{e}{B}_{\parallel }\approx 1\,\mu {\rm{G}}\,{\mathrm{cm}}^{-3}$. For mean ${B}_{\parallel }=5\,\mu {\rm{G}}$, the implied electron density is n_e = 0.2 cm⁻³. This is an order of magnitude higher than the volume-average density of the WIM in the midplane (Gaensler et al. 2008), but less than n_e ≈ 0.9 cm⁻³ reported by Langer et al. (2018) for compressed WIM in the Scutum spiral arm. Compression of the WIM in the spiral arm will also increase the magnetic field strength, so it is fair to say that n_e ≲ 0.2 cm⁻³. The emission measure of this plasma ≲40 pc cm⁻⁶ would be undetectable in the THOR continuum images. Association of the RM spike with compressed WIM in the Sgr arm also explains the difference in $\langle \mathrm{RM}\rangle$ and σ_RM between ℓ > 49°, and ℓ < 47° where the line of sight intersects the spiral arm. The line of sight component of the large-scale magnetic field may also be larger as a result of the compression of the plasma.

The Sgr arm passes inside the solar circle and is known in the southern sky as the Carina arm (e.g., Georgelin & Georgelin 1976). Association of the RM spike with the spiral arm would be confirmed if a similar structure, with the sign of $\langle \mathrm{RM}\rangle$ inverted, was found near the tangent of the Carina spiral arm around ℓ ≈ 285° (Vallée 2014). This area was previously covered by the Southern Galactic Plane Survey (SGPS; Haverkorn et al. 2006; Brown et al. 2007). The average SGPS RM at these longitudes is positive but close to zero (Brown et al. 2007), with a large range, −547 rad m⁻² < RM < 862 rad m⁻². It is possible that a narrow region (Δℓ ≲ 2°) with excessive RM has eluded detection. The SGPS was observed with twelve 8 MHz channels between 1336 and 1432 MHz (Haverkorn et al. 2006), frequencies where our high-RM sources depolarize in 8 MHz channels. The SGPS was also sensitive to diffuse Galactic emission that may have interfered with the detection of faint compact sources. Han et al. (2018) show pulsars for $| b| \lt 8^\circ$ near the Carina arm tangent with −800 rad m⁻² < RM < 900 rad m⁻². A denser rotation measure grid is required to confirm or reject the presence of a negative RM spike near the Carina arm tangent.

We find that $\langle \mathrm{RM}\rangle$ is 2.2 × 10² rad m⁻² (54%) higher and σ_RM is 1.5 × 10² rad m⁻² (63%) higher for ℓ ≲ 47° than for ℓ ≳ 49°. Such a sudden, substantial increase is not seen in current models of the magnetic field (e.g., Van Eck et al. 2011), but it is far too extended to ascribe to a single H ii region. If the Sgr arm raises the mean and the dispersion of RM for ℓ ≲ 47°, we expect its effect to peak at the arm tangent (47° ≲ ℓ ≲ 49°), where the line of sight through the arm is longest, and presumably aligned with the magnetic field. Association of the RM spike with an extended halo or bubble around W51 leaves unexplained why the RM is enhanced only on one side of W51, why we detect no excess thermal radio continuum from the direction of high RMs, and why $\langle \mathrm{RM}\rangle$ and σ_RM are higher for ℓ ≲ 47°. For these reasons, we favor association of the observed RM structure with the Sgr arm. This interpretation is consistent with the observation of polarized emission from the inside of spiral arms in galaxies with a strong spiral shock (Beck 2015, for a review), although these structures may be dominated by compressed turbulent field with little Faraday rotation (Fletcher et al. 2011).

This implies that Faraday rotation from spiral arms may be much stronger than previously thought, depending on the angle between the line of sight and the spiral arm. In particular, departures of the actual spiral arm from a perfect logarithmic spiral on kpc scales may impose systematic RM structure on scales of several degrees in Galactic longitude. The impact of the new RM data on the currently favored model for the Galactic magnetic field by Jansson & Farrar (2012) cannot be determined without detailed modeling. The high Faraday depth of the arm suggests it needs to be known accurately to model the magnetic field in the remainder of the disk.

These results emphasize the importance of sensitivity to high $| \mathrm{RM}|$. The initial polarization data product from the VLA Sky Survey (VLASS)¹⁵ is envisioned to be coarse cubes with 128 MHz channels. These cubes will have ϕ_max = 2160 rad m⁻² at mid-band (3 GHz). Another consequence of high $| \mathrm{RM}|$ associated with spiral arms is the potential effect on Faraday rotation of fast radio bursts (FRBs; Lorimer 2018). An FRB was recently identified 4 kpc from the center of a luminous early type or lenticular spiral galaxy by Bannister et al. (2019). If these bursts are associated with neutron stars in young supernova remnants (e.g., Piro & Gaensler 2018), they may well be embedded in spiral arms. The results presented here suggest that some FRBs could have RM in the thousands from their host galaxy without the need to invoke an extreme environment.

We present first results from the THOR polarization survey of Faraday rotation of compact radio sources. The THOR data reveal a strong spike in RM up to 4219 rad m⁻² at the Sgr arm tangent, with several sources exceeding 2 × 10³ rad m⁻². On the low-longitude side, where the line of sight intersects the Sgr arm, the mean $\langle \mathrm{RM}\rangle$ is higher by 2.2 × 10² rad m⁻² (54%), and the standard deviation σ_RM is 1.5 × 10² rad m⁻² (63%) larger than at higher longitude, where the line of sight does not intersect the Sgr arm. The combined pattern supports association of the RM spike with the large-scale structure of the spiral arm.

The strong Faraday rotation arises along lines of sight that do not intersect H ii regions detected by WISE at 22 μm. If the plasma is confined to any structure that has line of sight depth L ≈ 100 pc, e.g., the magnetized wall of a super bubble, its free–free emission should have been detected in the THOR continuum image unless the mean line of sight magnetic field ${B}_{\parallel }\gtrsim 15\,\mu {\rm{G}}$. We hypothesize that the RM spike arises from compressed magnetized WIM in the spiral arm, upstream from the major star-forming regions. This hypothesis can be tested in the future because it implies a similar structure, with the sign of RM inverted, at the Carina arm tangent (ℓ ≈ 285°) that may have eluded detection in existing surveys.

The excess Faraday depth of the Sgr arm tangent is several times the total Faraday depth of the remainder of the Milky Way disk. The arm acts as a very significant and structured Faraday screen inside the Galaxy. By inference, spiral arms in other galaxies may have similarly large Faraday depth. Some FRBs may have RM in the thousands without the need to invoke an extreme environment if these bursts are associated with neutron stars in spiral arms.

J.M.S. acknowledges the support of the Natural Sciences and Engineering Research Council of Canada (NSERC), 2019-04848. H.B., Y.W., and J.S. acknowledge support from the European Research Council under the European Community's Horizon 2020 framework program (2014–2020) via the ERC Consolidator Grant From Cloud to Star Formation (CSF)' (project number 648505). H.B. and J.S. also acknowledge support from the Deutsche Forschungsgemeinschaft in the Collaborative Research Center (SFB 881) "The Milky Way System" (subproject B1). F.B. acknowledges funding from the European Union's Horizon 2020 research and innovation programme (grant agreement No. 726384). R.J.S. acknowledges an STFC Ernest Rutherford fellowship (grant ST/N00485X/1) and HPC from the Durham DiRAC supercomputing facility. S.C.O.G. and R.S.K. acknowledge support from the Deutsche Forschungsgemeinschaft via SFB 881 (subprojects B1, B2, and B8) and from the Heidelberg cluster of excellence EXC 2181 STRUCTURES: A unifying approach to emergent phenomena in the physical world, mathematics, and complex data? funded by the German Excellence Strategy. This research was conducted in part at the Jet Propulsion Laboratory, which is operated by the California Institute of Technology under contract with the National Aeronautics and Space Administration (NASA). The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. The authors acknowledge the use of the RMtools package written by Cormac Purcell. J.M.S. thanks Y. K. Ma for his comments on the manuscript. The authors thank the anonymous referee for thoughtful comments on the manuscript.