OMC-1 dust polarisation in ALMA Band 7: Diagnosing grain alignment mechanisms in the vicinity of Orion Source I

We present ALMA Band 7 polarisation observations of the centre of the OMC-1 region of the Orion molecular cloud. We find that the polarisation pattern observed in the region is significantly altered by the radiation field of the $>10^{4}$ L$_{\odot}$ high-mass protostar Orion Source I. In the optically thick disc of Source I, polarisation is likely to arise from dust self-scattering, while in material to the south of Source I - previously identified as a region of 'anomalous' polarisation emission - we observe a polarisation geometry concentric around Source I. We demonstrate that in this region the extreme luminosity of Source I can shorten the radiative precession timescale to be less than the Larmor timescale for moderately large grains ($>0.005-0.1\,\mu$m), causing these grains to be aligned by Radiative Alignment Torques (RATs) to precess around the radiation anisotropy vector (k-RATs) rather than the magnetic field direction (B-RATs). This is the first time that k-RAT alignment has been observed outside of a protostellar disc or AGB star envelope. Elsewhere in OMC-1, we find that grains remain aligned perpendicular to the magnetic field direction, and that the magnetic field geometry is consistent with that inferred in the region from lower-resolution single-dish observations. The persistence of this geometry over orders of magnitude in size-scale and density suggests that the magnetic field is dynamically important and plays a significant role in mediating star formation in OMC-1.


INTRODUCTION
The role of magnetic fields in star formation, and particularly in high-mass star formation, remains poorly constrained. Until recently, there was a lack of observational evidence for the magnetic field morphology in the high-density interstellar medium (ISM). Dust emission polarimetry is a long-standing means of inferring ISM magnetic field properties (Davis & Greenstein 1951); however, observations have in the past been strongly surface-brightness-limited. The polarimetric capabilities of the new generation of submillimetre telescopes, including the Atacama Large Millimeter/submillimeter Array (ALMA), have made magnetic fields in dense, star-forming gas newly accessible (e.g. Cortes et al. 2016;Kwon et al. 2019). However the density regimes now observable have brought with them complications in interpretation of polarisation observations, as the mechanisms by which dust grains can gain a preferential alignment proliferate at high densities (Davis & Greenstein 1951;Gold 1952;Lazarian & Hoang 2007a,b;Kataoka et al. 2015;Hoang et al. 2018;Kataoka et al. 2019).
Most observations of submillimetre dust polarisation at very high densities have been in protostellar discs (e.g. Hull & Zhang 2019, and refs. therein). However, sites of high-mass star formation are another important high-density ISM environment. In this work, we present ALMA Band 7 (345 GHz; 880 µm) observations of the OMC-1 region, at the centre of the Orion Molecular Cloud, a nearby site of high-mass star formation (e.g. Bally 2008).
The OMC-1 region, at the centre of the well-studied 'integral filament' in the Orion A molecular cloud, is located at a distance of 388 ± 5 pc (Kounkel et al. 2017), and consists of two dense clumpsthe northern Becklin-Neugebauer-Kleinmann-Low (BN/KL) clump (Becklin & Neugebauer 1967;Kleinmann & Low 1967), and the southern Orion S clump (Batria et al. 1983;Haschick & Baan 1989). In this paper we focus on the centre of the BN/KL clump, an active site of star formation which hosts an extremely powerful wide-angle explosive molecular outflow, with multiple ejecta known as the 'bullets of Orion' (Kwan & Scoville 1976;Allen & Burton 1993). The young stars BN, Source I, and x, located in the core of the BN/KL clump, have proper motions consistent with their having been colocated ∼ 500 years ago, leading to the suggestion that the BN/KL outflow is the result of a dynamical interaction between these sources (Gómez et al. 2005). The dynamic age of the BN/KL outflow is also ∼ 500 yr (Zapata et al. 2009), and the kinetic energy released by the interaction is comparable to the energy in the outflow (Kwan & Scoville 1976;Gómez et al. 2005), supporting this interpretation. An alternative explanation for the BN/KL outflow is a protostellar merger (Bally & Zinnecker 2005). Debate over what combination of interaction, decay and merger produced the BN/KL outflow continues (e.g. Luhman et al. 2017;Farias & Tan 2018); however, the approximate age, high energy and impulsive nature of the outflow are well-established.
Source I drives a separate, slower, bipolar outflow along an axis perpendicular to its direction of motion (Plambeck et al. 2009). BN and Source I appear to be recoiling from a common centre Luhman et al. 2017); nonetheless, the outflow from Source I is symmetric about an axis approximately perpendicular to the direction of motion, despite the significant ram pressure on the source as it ploughs through its surroundings at ∼ 12 km s −1 . This discrepancy could be ascribed to the outflow being shaped by a strong magnetic field. Hirota et al. (2020) recently observed a highly uniform polarisation structure in SiO emission associated with the outflow, suggesting a field strength of ∼ 30 mG, strong enough to prevent distortion of the outflow by ram pressure.
Complete depolarisation is observed on the position of BN/KL in single-dish observations (Schleuning 1998;Houde et al. 2004;Pattle et al. 2017). This depolarisation, over a single telescope beam, results from an approximately elliptical polarisation pattern in the dense centre of OMC-1, as observed using BIMA (Rao et al. 1998) and the SMA (Tang et al. 2010). In regions where dust grains are aligned with their major axes perpendicular to the magnetic field direction, polarisation vectors can be rotated by 90 degrees to trace the plane-ofsky magnetic field (Davis & Greenstein 1951;Andersson et al. 2015). Thus Tang et al. (2010), observing the 870µm dust continuum with the SMA, inferred that the magnetic field in the region is radial, and centred on the outflow. From this they suggested two hypotheses: (1) a toroidal field in a magnetised, differentially rotating 'pseudo-disk' in the centre of OMC-1, or (2) the magnetic field is being dragged into a radially-symmetric morphology by the explosive outflow.
However, in recent years, a number of different grain alignment mechanisms have been suggested to explain the polarisation properties of dust emission in different environments in the very highdensity ISM. As well as the traditional interpretation of polarised dust emission as tracing the plane-of-sky magnetic field direction, an effect usually ascribed to Radiative Alignment Torques (B-RATs; Lazarian & Hoang 2007a;Andersson et al. 2015), alternative mechanisms include supersonic mechanical grain alignment (the Gold effect; Gold 1952), Mechanical Alignment Torques (MATs; Lazarian & Hoang 2007b;Hoang & Lazarian 2016), a variation on Radiative Alignment Torques in which grains precess around the radiation anisotropy vector rather than the magnetic field direction (k-RATs; Lazarian & Hoang 2007a;Tazaki et al. 2017), and dust self-scattering (Kataoka et al. 2015). In this work we will discuss the polarisation pattern of the BN/KL region in the context of these various mechanisms by which polarised emission can arise.
The structure of this paper is as follows: in Section 2 we describe our observations. In Section 3 we interpret the polarisation distributions seen across the region. Our conclusions are summarised in Section 4.

DATA
We observed three overlapping fields in OMC-1 in ALMA Band 7 (345 GHz, 870µm) polarised light. These observations were taken in ALMA Cycle 6 on 2019 April 9. We observed one track on each source, for a total of 1.5 hours of observing time, in array configuration C43-3. The data have project code 2018.1.01162.S. The three fields were centred on the Orion Hot Core, with R.A. The data were calibrated and imaged using CASA version 5.4.0 by the Observatory. The phase calibrator was J0529-0519 and the polarisation calibrator was J0522-3627. The tclean parameters used for imaging were Briggs weighting, with a robust parameter of 0.5, and a cell size of 0.060 . The output maps have a restoring beam  Contour levels are 0.1, 0.3, 0.5 and 0.7 Jy/beam. Note that the Stokes I map is more dynamic-range-limited than is the polarised intensity map. Centre: Polarised intensity overlaid with contours of CARMA 86 GHz SiO v = 0 emission averaged over the velocity range −10 to +20 km s −1 , tracing the Source I outflow (Plambeck et al. 2009). Contour levels are 0.05, 0.1, 0.2, 0.5, 1.0, and 2.0 Jy/CARMA beam. Right: An illustration of the division of regions in OMC-1 used in this work. The centre of the BN/KL explosion (Rodríguez et al. 2017) is marked with a black star and the centre of the Wright & Plambeck (2017) ring feature is marked with a black cross. The mean magnetic field direction observed on large scales (116 c ir c E of N) is shown as a black vector in the inset box, and its associated polarisation direction (26 • E of N) is shown as a grey vector. In each panel, the synthesised beam is shown in the lower left-hand corner. size of 0.54 × 0.43 , oriented −74 • E of N, and a maximum recovered size scale of 4.7 . Integrated Stokes Q, U and I maps were produced; note that we do not consider circular polarisation (Stokes V) in this work. We linearly mosaicked these maps using the Miriad task linmos. The mosaicked Q, U and I maps are shown in Figure 1. Our RMS noise in the mosaicked Stokes Q and U maps is 0.62 mJy/beam. For the purpose of the analysis in Section 3 below, we regridded the Stokes Q, U and I maps to a 0.25 pixel grid, approximately Nyquist-sampled on the major axis of the beam.
Polarised intensity is calculated as and is shown in Figure 2. Polarised intensity is thresholded at 6 mJy/beam. Polarisation angle is given by Polarisation angles are shown in Figure 3. We note that while po-larisation angle segments are referred to as vectors for convenience, they are not true vectors due to the ±180 • ambiguity in polarisation direction. We do not calculate polarisation fractions in this work, as the Stokes I map is much more dynamic-range-limited than are the Stokes Q and U maps, as shown in Figure 2, and so any calculation of polarisation fraction is likely to produce artificially large values, particularly in low-surface-brightness regions.

Polarised line emission in OMC-1
Emission from OMC-1 includes contributions from a plethora of spectral lines (e.g. Pagani et al. 2017). Fortunately for our purposes, most of these lines will contribute little polarised signal, and so our Stokes Q and U maps are dominated by continuum emission. The most significant spectral line in our spectral window is the 12 CO J = 3 → 2 line at 345.8 GHz. The Gaussian component of this line has a FWHM linewidth of 8 km s −1 around the 9 km s −1 velocity of the region, while its high-velocity line wings, associated with the  (Furuya & Shinnaga 2009;Zapata et al. 2009;Coudé et al. 2016). The Gaussian line component is not well-resolved in our observations, and is likely to be significantly self-absorbed (Furuya & Shinnaga 2009), while the wings occupy ∼ 8% of the total bandwidth. CO emission in outflows can be polarised up to a maximum of ∼ 3% by the Goldreich-Kylafis effect (GK effect; Goldreich & Kylafis 1982;Ching et al. 2016), which has previously been observed in OMC-1 (Houde et al. 2013). Polarised CO line emission contributes a small amount to our observed Stokes Q and U emission but, as discussed in Section 3.2.3, this effect does not systematically change our observed polarisation angles. We also observe marginally-resolved line polarisation features at 345.6 GHz and 345.9 GHz. We tentatively associate these features with the methanol (CH 3 OH) v t = 0 → 2 transition, noting that polarisation has previously been observed at lower frequencies in methanol masers (e.g. Surcis et al. 2015). Despite these effects, our integrated Stokes Q and U maps are dominated by the dust continuum emission, and the polarisation geometry which we observe is consistent with that shown at slightly lower resolution in ALMA Bands 3 and 6 by Hull et al. (2020) (note that these authors show polarisation vectors rotated by 90 degrees).

RESULTS AND DISCUSSION
Throughout most of the ISM, polarised dust emission can be reliably assumed to arise from dust grains aligned with their major axis perpendicular to the magnetic field direction the leading theory for explaining which is Radiative Alignment Torques (B-RATs; Lazarian & Hoang 2007a). However, at extremely high densities, such as those found in the centre of OMC-1, this assumption starts to break down. In extreme environments such as these, possible cause of polarised dust emission include: (1) Alignment by B-RATs, the standard mechanism throughout the ISM.
(2) Dust self-scattering, an effect seen in protoplanetary discs, in which polarisation arises from Rayleigh scattering from large dust grains (Kataoka et al. 2015).
(3) Gold alignment: mechanical alignment of dust grains in a supersonic gas flow, with grain major axes parallel to the flow direction (Gold 1952 We consider these alternatives when interpreting the polarised emission from dust in OMC-1, introducing each mechanism in detail as it arises. When considering possible causes of polarised emission in OMC-1, we divided the regions into the following sub-regions: (1) Source I, (2) the Anomalous Region, also referred to as the Fork, (3) the Ridge, (4) MF4/MF5, (5) the Compact Ridge, also referred to as MF1. These sub-regions are labelled on Figure 2. The mean and median polarisation angles, as well as the number of independent measurements, in each region are listed in Table 1. We discuss each region in turn below.
Throughout the following discussion we principally compare the polarisation geometry in each region to three models: polarisation arising from grains aligned (1) with their major axes perpendicular to the large-scale 116 • E of N magnetic field (polarisation angle 26 • E of N) observed on larger scales in the region, (2) such that their major axes trace concentric circles around the centre of the BN/KL explosion, and (3) such that their major axes trace concentric circles around Source I. In case (1) we expect to find some variation from the large-scale mean field direction, which is measured at resolutions 10 (Houde et al. 2004;Ward-Thompson et al. 2017), but use the mean direction as a simple model for purposes of comparison. We also consider the dust self-scattering model for Source I, as discussed below.

Source I
Source I is a well-studied highly luminous high-mass protostar with a collimated SiO outflow (e.g. Hirota et al. 2015), as shown in Figure 2. Its mass has been a matter of discussion, with its velocity, approximately half that of the 10 M B star BN, from which Source I appears to be recoiling, suggesting a mass ∼ 20 M , while high angular resolution observations of the rotation curves of H 2 O and salt lines imply a central mass of 15M (Ginsburg et al. 2018). However, rotation curves of emission lines from the base of the bipolar outflow suggest a mass in the range ∼ 5 − 8 M (Kim et al. 2008;Matthews et al. 2010;Plambeck & Wright 2016;Hirota et al. 2017;Kim et al. 2019). Moreover, the high velocity of the source with respect to the surrounding dense medium is belied by the symmetry of its outflow, which would be expected to be significantly bowed by ram pressure . These apparent contradictions have recently been reconciled by polarisation observations of the SiO emission associated with the outflow, suggesting a magnetic field sufficiently strong to shape the bipolar outflow and to cause sub-Keplerian gas dynamics at the base of the outflow, leading to the mass underestimate .
The polarisation geometry of Source I is shown in Figures 4 and 5. Source I is not well-resolved in our observations, with only two independent measurements across the disc. The mean polarisation angle of Source I is 53 • .5 ± 8 • .6, and the median is 52 • .0 ± 5 • .4, inconsistent with the polarisation having arisen from a 116 • magnetic field, but consistent with being parallel to the minor axis of Source I (53 • ; Ginsburg et al. 2018). The polarisation geometry is also broadly consistent with that expected for a polarisation pattern concentric around BN/KL. The three model geometries which we consider for Source I are shown in Figure 4. Kataoka et al. (2015) introduced the dust self-scattering mechanism for producing polarised emission in protostellar discs, wherein polarisation arising at a given wavelength arises from Rayleigh scattering from dust grains with sizes comparable to that wavelength. This mechanism can produce a polarisation pattern concentric around the protostellar position, or aligned with the disc minor axis, consis-  to the the large-scale magnetic field direction (hypothesised alignment mechanism: B-RATs). Centre right: polarisation vectors concentric around the BN/KL explosion centre (hypothesised alignment mechanism: v-MATs). Far right: polarisation vectors perpendicular to the major axis of the Source I disc (parallel to minor axis; polarisation hypothesised to arise from dust self-scattering). All maps are shown on 0.25 (approximately Nyquist-sampled) pixels. The synthesised beam size is shown in the lower left-hand corner of each plot. tent with what we see in Source I. The conditions for polarisation arising from dust self-scattering to produce uniform polarisation aligned with the disc minor axis, as seen in Source I, are given by Sadavoy et al. (2019) as an inclined (i>60 • ) disc with optically thick dust emission. Wright et al. (2020) find a spectral index ∼2 along the disc midplane, consistent with optically thick dust emission. This spectral index increases to ∼3 at the disc edges, suggesting that dust emission is optically thin in the periphery of the disc. Wright et al. (2020), observing at 340 GHz, set a lower limit to the disc inclination of 79 • ±1 • (with major and minor axes 99 × 19 au) at a disc brightness temperature contour of 400 K, and 74 • ± 1 • (239 × 45 au) at a disc brightness temperature contour of 25K, but note that the observed geometry suggests that the inclination is closer to 90 • . Similarly, Matthews et al. (2010) measure an inclination ∼ 85 • from SiO masers close to the disc. Source I thus meets the conditions for our observed polarisa- tion geometry to arise from dust self-scattering. While we note that the polarisation geometry is also consistent with being concentric around BN/KL, dust self-scattering is an established mechanism for producing polarisation in protostellar discs, the necessary conditions for which are matched in Source I, and so we consider it to be the probable source of the dust polarisation which we observe. This implies the existence of significant dust growth and coagulation within the Source I disc. While grain growth is expected in protostellar discs (e.g. Kwon et al. 2009), for dust self-scattering to be observed, there must be a significant population of spherical dust grains with size ∼ λ/2π (Kataoka et al. 2015), or non-spherical grains with sizes λ/2π (Kirchschlager & Bertrang 2020). At 870µm, this implies the existence of a population of dust grains with sizes ∼ 140 µm or larger in the Source I disc.
We note that Hirota et al. (2020) found an upper limit continuum polarisation fraction of 1% in the Source I disc at 96 GHz (3.1 mm). These observations, with a synthesised beam size of 0.05 , did resolve the Source I disc, and so if scattering were important, some polarisation signal would be expected. If scattering is significant at 345 GHz but negligible at 96 GHz, this puts strong constraints on the grain size distribution in the Source I disc, with a significant population of grains with sizes ∼ 140 µm, but a cut-off in grain size at < 500 µm. Higher-resolution polarisation observations of Source I at 345 GHz are required to confirm the dust self-scattering hypothesis, and to better constrain the grain size distribution in the disc.

Anomalous region/Fork
There is a significant 'fork' visible to the south of the Hot Core in the polarised intensity and polarisation angle maps, as shown in Figures 2 and 3. Vectors on the western side of the fork have a typical polarisation angle significantly different both to that across the Ridge and to that in the eastern arm. Rao et al. (1998), observing OMC-1 in polarised light with BIMA, identified this as an 'anomalous region', with polarisation vectors significantly different from elsewhere in OMC-1, and inconsistent with being perpendicular to the large-scale field direction. The vectors which we see are consistent with their observations. Rao et al. (1998) suggested that grains in this region are mechanically aligned by the Gold effect, driven by the Source I outflow. However, our higher-resolution observations show that the polarisation pattern in the Anomalous Region/Fork is not consistent with being parallel to the Source I outflow. Wright & Plambeck (2017) suggest that much of the dust emission in the Fork originates from the walls of the cavity formed by the bipolar outflow from Source I, based on the spatial coincidence of dust emission and SiO emission tracing the outflow. The Source I outflow is shown in Figure 2 In the following discussion we consider both the case in which emission arises from the outflow cavity wall, and that in which it arises from the ambient medium of OMC-1. Figures 6 and 7 show that the polarisation vectors in the Fork are inconsistent with the 26 • polarisation direction associated with the large-scale magnetic field, but broadly consistent with being concentric around either the BN/KL explosion centre, the centre of a 'ring feature' identified by Wright & Plambeck (2017) (discussed below), or Source I. Figure 6 shows each of these polarisation geometries, along with the absolute residual angles between the models and the observed polarisation geometry. The BN/KL-concentric model is consistent with the observations in the eastern arm of the Fork, but systematically different by ∼ 20 • in the western arm. The ringfeature-concentric model is consistent with the observations in the south of the region, but not consistent in the north; we discuss this further below. The Source I-concentric model is broadly consistent with the observed polarisation geometry across the region. We consider six hypotheses to explain the observed polarisation pattern: (1) grains aligned by B-RATs with respect to a distorted magnetic field; (2) polarisation arising from the GK effect in CO emission associated with the Source I outflow; (3) polarisation arising from scattering of emission from Source I; (4) polarisation arising from supersonic mechanical alignment (Gold alignment) induced by either the BN/KL explosion or the Source I outflow; (5) grains aligned by subsonic v-MATs, induced by (a) the shock associated with the passage of ejecta from the BN/KL explosion through the region; (b) shocks associated with the Source I outflow; (6) grains aligned by k-RATs, perpendicular to the radiation gradient associated with Source I. We consider these hypotheses in turn below, and then as a check on our analysis confirm that the gas damping timescale is sufficiently long in the region to allow a preferential dust precession axis to exist. The following discussion is summarised in Table 2.

Distorted magnetic field
As shown in Figures 6 and 7, the polarisation pattern in the Fork is inconsistent with that predicted based on the large-scale mean field direction. Tang et al. (2010) hypothesised that the polarisation geometry of OMC-1, observed at 870µm using the SMA, arose from grains aligned perpendicular to a magnetic field which had been significantly distorted from its initial configuration. They proposed two hypotheses for how the field had been distorted: (1) that the observed polarisation indicated that the density structure of the centre of OMC-1 forms a rotating 'pseudo-disk' around the centre of the BN/KL explosion, with a toroidal magnetic field. We consider this hypothesis to have been disfavoured by ALMA studies better determining the line-of-sight distances and velocities of the various OMC-1 clumps (Pagani et al. 2017).
(2) that the BN/KL explosive outflows have realigned the magnetic field to be radial around the explosion centre (i.e. polarisation vectors are concentric around the BN/KL centre). Our higher-resolution observations suggest that while vectors elsewhere in OMC-1 are not consistent with being concentric around the BN/KL explosion centre, they are marginally consistent with being so in the Fork. However, the polarisation ge-  Table 2. A summary of potential mechanisms for producing polarised dust emission in the Anomalous Region/Fork. 'AM' refers to polarisation in the Fork arising from the ambient medium of OMC-1; 'OCW' to polarisation arising from the cavity walls of the Source I outflow.  (2017) ring feature (hypothesised alignment mechanism: v-MATs). Far right: polarisation vectors concentric around Source I (hypothesised alignment mechanism: k-RATs). All maps are shown on 0.25 (approximately Nyquist-sampled) pixels. The synthesised beam size is shown in the lower left-hand corner of each plot. Table 3. A summary of the timescales which we estimate in the Anomalous Region/Fork. Note that a −5 = (a/10 −5 cm); a −5 = 1 indicates grain size a = 0.1 µm.
ometry which we observe is more consistent with being concentric around Source I than around the BN/KL explosion centre, suggesting that if the grains remain aligned with the magnetic field in the region, the field is likely to be radial around Source I. This could potentially indicate a highly poloidal field in the Source I outflow.

Larmor timescale
We will consider alternative explanations for the observed polarisation geometry in the Fork, which do not require such wholesale reorganisation of the magnetic field. For dust grains to be aligned with respect to the magnetic field direction, rather than some other axis, the timescale for precession around the magnetic field direction (the Larmor timescale, τ Lar ) must be shorter than all other precession timescales (e.g. Hoang & Lazarian 2016). Therefore we estimate τ Lar in the Fork, for comparison with other timescales.
We note that if the dust grains in the Fork were to have superparamagnetic inclusions (e.g. Lazarian & Hoang 2019, and refs. therein), χ could be made considerably larger, correspondingly decreasing τ Lar .

Goldreich-Kylafis Effect
We consider the possibility that the polarisation we observe in this region is produced by polarised 12 CO emission. Polarised CO emission arises from the GK effect, in which the splitting of rotational transitions J into magnetic sublevels M causes ∆M = ±1 transitions to emit radiation linearly polarised perpendicular to the magnetic field, and ∆M = 0 transitions to emit radiation parallel to the magnetic field (Goldreich & Kylafis 1982). If these two polarisations have different optical depths, the net polarisation direction will be either perpendicular or parallel to the plane-of-sky magnetic field direction. The effect is most pronounced in regions of strong radiation anisotropy with optical depths ∼ 1, and is usually associated with protostellar outflows (Ching et al. 2016). However, we note that 12 CO emission in OMC-1 is strongly self-absorbed (Furuya & Shinnaga 2009). While we do observe polarisation arising from the CO line emission in the region, its contribution to our integrated Stokes Q and U images is minimal. The bright features in Q and U in the Fork visible in Figure 1 persist across our observed frequency range, and are not associated with the 12 CO line emission around 345.8 GHz. We measured the mean Stokes Q and U emission in both the eastern and western arms of the Fork, both over the full bandwidth of the observations and by excluding frequencies in the ranges 345.  GHz. The former range excludes the one of the narrow features which we tentatively ascribe to polarised CH 3 OH emission; the latter excludes both the wide 12 CO J = 3 → 2 emission feature, the other narrow feature, and the frequency range between them. Thus, the clipped frequency ranges contain only contribution from the polarised dust continuum.
In the western arm, the mean channel value of Stokes Q over the full range is Q = −6.0 ± 1.6 mJy/beam, while in the clipped range, Q = −6.6 ± 0.4 mJy/beam. The mean Stokes U values over the full range is U = −3.6 ± 1.0 mJy/beam, while in the clipped range, U = −3.7 ± 0.7 mJy/beam. The difference in mean polarisation angle is less than 1 • , with the mean polarisation angle over the full range being 105.6 • and the mean angle over the clipped range being 104.8 • .
In the eastern arm, the mean channel value of Stokes Q over the full range is Q = −4.6 ± 1.0 mJy/beam, while in the clipped range, Q = −4.9 ± 0.5 mJy/beam. The mean Stokes U values over the full range is U = +3.3 ± 2.3 mJy/beam, while in the clipped range, U = +3.8 ± 0.9 mJy/beam. The difference in mean polarisation angle is again less than 1 • , with the mean polarisation angle over the full range being 72.3 • and the mean angle over the clipped range being 71.1 • .
We thus conclude that while line polarisation may have the effect of slightly increasing the scatter on the polarisation angles which we observe in the Fork, our integrated maps are dominated by dust continuum polarisation. Line polarisation does not systematically affect the distribution of polarisation angles which we measure, and the GK Effect is not responsible for polarised emission in the Fork.

Dust self-scattering
The polarisation pattern in the Fork is consistent with being concentric around Source I. This could imply that polarisation arises from scattering of light from Source I (Kataoka et al. 2015). We note however that as is the case in the Source I disc, this would require grain sizes ∼ 140µm. If the emission from the Fork arises from the ambient medium of OMC-1, this level of grain growth appears impossible. If the emission arises from the Source I outflow cavity, and if such large grains exist in the Source I disc as is suggested by the Source I polarisation pattern, it could be hypothesised that they might be entrained into the outflow. However, transport of such large dust grains, as well as their avoiding destruction in the outflow in sufficient number to produce the observed polarisation pattern, does not seem likely (Giacalone et al. 2019). We thus discount this hypothesis, while noting that we cannot definitively rule it out.

Supersonic mechanical (Gold) alignment
For completeness, we note the possibility of supersonic mechanical alignment (Gold alignment;Gold 1952). If the polarised emission in the Fork arises from the ambient medium of OMC-1, and is associated with BN/KL shocks, Gold alignment would produce a radial polarisation pattern around the centre of the BN/KL explosion. If the emission in the Fork arises from the Source I outflow cavity walls, Gold alignment would produce polarisation parallel to the Source I outflow, or radial around Source I. All of these geometries are inconsistent with the observed polarisation pattern shown in Figure 3, and so we do not consider Gold alignment further. Lazarian & Hoang (2007b) proposed the Mechanical Alignment Torques (MATs) mechanism, in which grains drifting relative to gas are aligned by mechanical torques to have their long axes perpendicular to the precession axis of the grain. This precession axis is typically the magnetic field direction (B-MATs, Lazarian & Hoang 2007b;Hoang et al. 2018), but can in some environments be the velocity vector of the gas/dust drift (v-MATs, Lazarian & Hoang 2007b;Hoang & Lazarian 2016). The v-MAT alignment mechanism can occur when the velocity difference between the gas and dust is subsonic, and when the mechanical alignment timescale (the precession time around the gas flow), τ mech , is shorter than the Larmor precession timescale, τ Lar (cf. Hoang et al. 2018, Sec. 6.4). This mechanism further requires the dust grains to have significant helicity, which is acquired through coagulation. (Brauer et al. 2008;Ormel et al. 2009;Hirashita 2012). Hoang et al. (2018) discuss environments where gas/dust drift is likely to be induced, concluding that such drift may be triggered by cloud-cloud collisions, radiation pressure, ambipolar diffusion, or gravitational sedimentation. Gas/dust drift occurs across shock fronts (McKee et al. 1987), and so can plausibly expected to be occurring in OMC-1, either in the aftermath of the BN/KL explosion, or in shocks within the Source I outflow. Shocks in OMC-1 are thought to be continuous (C-shocks; e.g. Colgan et al. 2007), supporting the hypothesis that the magnetic field in the region is dynamically important (Draine 1980).

Mechanical Alignment Torques
If the emission in the Fork arises from the ambient medium of OMC-1, rather than from the Source I outflow cavity walls, we might expect grains to be mechanically aligned by shocks associated with the BN/KL explosion ejecta. While the observed polarisation geometry is, in the western arm of the Fork, inconsistent with being concentric around the BN/KL explosion centre, Wright & Plambeck (2017) identified a ring of emission near SMA 1 in HCN 354.5 GHz and H 3 CN 354.7 GHz emission, which they interpreted as evidence for passage of debris from the BN/KL explosion. The ring has ∼2 km s −1 expansion velocity and a dynamical age of ∼700 yr, consistent with the approximate age of the BN/KL explosion, if its expansion has been somewhat decelerated. As well as considering grain alignment concentric around Source I, we consider grain alignment concentric around the centre of the ring, in case mechanical alignment were induced by the shock associated with these particular ejecta. In the south of the Anomalous Region/Fork (i.e. at larger radii), the polarisation pattern is more consistent with being concentric around the position of the ring than it is with being concentric around the BN/KL explosion centre, but the model fails at positions near the ring centre.
Alternatively, if the polarised emission in the Fork arises from the Source I outflow cavity walls, grains cannot be aligned by shocks Figure 8. A comparison of the radiative precession (τ r a d, p ), Larmor (τ L ar ), gas damping (τ g a s ) and mechanical alignment (τ me c h ) timescales which we estimate in the Anomalous Region/Fork, as a function of grain size a. Note that the shortest timescale determines the precession axis; thus, small grains are aligned by B-RATs to precess around the magnetic field direction, and large grains by k-RATs to precess around the radiation anisotropy vector. The minimum value of τ me c h which we show assumes ω ≈ ω t h , i.e. that grains are rotating at the thermal angular velocity. The dark grey shaded region marks the grain size distribution in the diffuse ISM (Draine & Li 2007); the light grey shaded area shows the maximum extent of grain growth likely in OMC-1 (a ma x < 500 µm; see Section 3.1), although we note that grains larger than a few microns are unlikely to be found outside of the Source I disc.
associated with the BN/KL explosion, as Source I and its associated outflow is moving behind these shock fronts (e.g. Hirota et al. 2020). In this case, v-MAT alignment could instead be induced by shocks associated with the expansion of the bipolar outflow into its surroundings. Polarisation vectors might then be expected to be radial around Source I, perpendicular to the surface of the outflow, or less ordered, depending on the nature of the outflow shocks.

Mechanical alignment timescale
The timescale for alignment by v-MATs is given by (Lazarian & Hoang, ApJ subm.), where c s is gas sound speed, ∆v is gas/dust velocity difference, ω is the grain angular velocity, ω th is the thermal angular velocity, and Θ is the angle between the grain axis of major inertia and the direction of radiation. We takeŝ ∼ 1 and sin 2Θ ∼ 0.5. The velocity difference between gas and dust in C-type shocks is not well-characterised, potentially taking any value between zero and the shock velocity, depending on environment (Wardle 1998;Guillet et al. 2007). However, the condition for v-MATs is ∆v < c s , and so equation (5) becomes τ mech 72 ω ω th year.
Comparison of equations 4 and 6 suggests that τ mech τ Lar .
The requirement for τ mech < τ Lar to hold is the physically implausible condition ω/ω th 1, i.e. grains would have to be rotating subthermally. The timescale for v-MAT alignment thus remains too long for this mechanism to be likely to be the main cause of grain alignment in the Fork.

Radiative Alignment Torques
Under the Radiative Alignment Torques (RATs) paradigm of grain alignment, grains are efficiently aligned when they can be spun up to suprathermal rotation by an anisotropic radiation field (Dolginov & Mitrofanov 1976;Lazarian & Hoang 2007a). As with MATs, the grains will align with their long axes perpendicular to their precession axis. In the large majority of ISM environments, the precession axis can be presumed to be the magnetic field direction (B-RATs Lazarian & Hoang 2007a). However, in the presence of a sufficiently strong and anisotropic radiation field, the precession axis can instead be the radiation anisotropy vector, and so grains will be aligned with their major axes concentric around the source driving the radiation field (k-RATs; Tazaki et al. 2017). The condition for k-RATs to dominate over B-RATS is that the radiative precession timescale must be shorter than the Larmor timescale, i.e. τ r ad, p < τ Lar (Lazarian & Hoang 2007a;Tazaki et al. 2017). Alignment by k-RATs has not previously been observed outside of protostellar discs, with the exception of a potential detection in the envelope of an evolved star (Andersson et al. 2018).
The brightest source in OMC-1 is Source I, with a luminosity > 10 4 L (Menten & Reid 1995). As shown in Figure 6, the polarisation pattern in the Fork is quite consistent with being concentric around Source I, potentially suggesting that the dust grains in the region are aligned by k-RATs, driven by the radiation field of Source I. In the following section we estimate the radiative precession timescale τ r ad, p in the Anomalous Region/Fork. We do not include other sources of radiation in OMC-1 in this analysis, as it is the strongly anisotropic radiation field of Source I which we hypothesise is driving k-RAT alignnment in the Fork.

Radiative precession timescale
The radiative precession timescale is given by Tazaki whereT d = T d /15 K and T d is dust temperature; u r ad is the energy density of the radiation field in the region under consideration; u is the energy density of the standard interstellar radiation field (ISRF), given by Tazaki et al. (2017) as u = 8.64 × 10 −13 erg cm −3 ;λ is the mean wavelength of the incident radiation spectrum, γ is radiation field anisotropy, and |Q Γ | is the RAT efficiency. This formulation of τ r ad, p assumes grains to be rotating at the thermal angular velocity, i.e. ω ≈ ω th (Lazarian & Hoang 2007a). We again takeρ ∼ 1 andŝ ∼ 1. We expect 0.1 < γ ≤ 1, as γ ∼ 0.1 in the diffuse ISM (Draine & Weingartner 1996), and γ 1 in the immediate vicinity of a protostar (Tazaki et al. 2017). As we are specifically considering the radiation field from Source I, which we expect to be strongly anisotropic, we take γ ∼ 1 (note that this implies that radiation from Source I is effectively unobscured in the Fork). We further take |Q Γ | ≤ 0.4 (Lazarian & Hoang 2007a;Tazaki et al. 2017), and so γ|Q Γ |/0.01 40.
The luminosity of Source I is not well-characterised, but is thought to be > 10 4 L (Menten & Reid 1995). The plane-of-sky separation between the Fork and Source I is ∼ 3.5 , which at a distance of 388 pc corresponds to ∼ 2 × 10 16 cm (∼ 1400 au). We thus estimate the radiation energy density in the vicinity of the Fork to be where L is the luminosity of Source I and R is the separation between Source I and the Fork.
The effective brightness temperature of Source I is ∼ 1500 K (Reid et al. 2007), and so from Wien's Law, we infer a peak emission wavelength of ∼ 1.9 µm. We thus takeλ/1.2µm ∼ 1.6. Dust temperature T d can be estimated for silicates using the relation T d ≈ 16.4 u r ad u 1 6 K (9) (Draine 2011). Using our value of u r ad from equation (8), we estimate T d ≈ 135 K in the Fork, and soT 1 2 d ≈ 3. Combining these estimates, equation (7) becomes τ r ad, p 1.7 × 10 −5 a 1 2 −5 year. (10) Comparing this to the Larmor timescale in the Fork, as given in equation (4), we find the condition for k-RATs to dominate over B-RATs, τ r ad, p < τ Lar , is equivalent to or equivalently, This suggests that in the vicinity of Source I, τ r ad, p < τ Lar will hold for larger paramagnetic dust grains, and so we can plausibly expect to see a polarisation pattern arising from k-RATs. While highly uncertain, the minimum values of a for which τ r ad, p < τ Lar that we find are plausible grain sizes in a dense molecular cloud (e.g. Draine & Li 2007). The maximum grain size in the diffuse ISM is 0.25 − 0.3 µm (Mathis et al. 1977;Draine & Li 2007), indicating that while k-RATs could potentially dominate over B-RATs in the vicinity of Source I even in relatively pristine ISM material, the grain growth which is likely to have occurred in such a dense environment (Ysard et al. 2013) makes τ r ad, p < τ Lar more likely to hold. Moreover, if grains are indeed aligned by k-RATs downstream of the shocks associated with BN/KL ejecta and/or the expansion of the Source I outflow, it suggests that these shocks have not destroyed all of the larger dust grains in the cloud.
Note that in the preceding analysis, we have implicitly assumed that the polarised emission arises in a location where there is minimal obscuration of Source I. Such obscuration would introduce absorption, re-emission and scattering of radiation, reducing the anisotropy γ in the radiation field and so increasing τ r ad, p . This might support the interpretation of the emission in the Fork as arising from the Source I outflow cavity walls, as the radiation field of Source I will be significantly less obscured on the cavity walls than it will be in the ambient material of OMC-1.

Gas damping timescale
A further requirement for grains to precess around any given axis is that the precession timescale around that axis is shorter than the gas damping timescale τ gas , the characteristic timescale of grain randomisation by gas collisions (Lazarian & Hoang 2007a). The highly ordered polarisation geometry of the Fork -and across OMC-1 -strongly suggests that the dust grains are not randomised. However, as a check on our previous analysis, we estimate the gas damping timescale in the Fork. Hoang & Lazarian (2016) give τ gas as τ gas = 6.6 × 10 4ρŝ− 2 3 a −5 Γ −1 300 K 1 2 cm −3 T 1 2 gas n year, where Γ is a factor of order unity characterising grain geometry and T gas is gas temperature. We take Γ ∼ 1, continue to takeŝ ∼ 1 andρ ∼ 1 and n H ∼ 10 6 cm −3 , and assume T gas ∼ T dust ≈ 135 K. Equation (13) thus becomes τ gas ∼ 1.7 a −5 year.
Comparison of equation (14) with equations (4) and (10) shows that there is no value of a −5 at which τ gas is the shortest timescale. τ gas < τ r ad, p holds only for unphysically small grains, with a −5 < 10 −10 , at which size τ Lar τ gas would hold if such grains existed.
Conversely, τ gas < τ Lar only for unphysically large grains, with a −5 > 10 3 − 10 5 , at which size τ r ad, p τ gas . These timescales are summarised in Table 3 and illustrated in Figure 8. For the values of τ Lar and τ r ad, p which we find in the Fork, τ gas would need to be smaller by at least five orders of magnitude for a regime to exist in which it is the shortest timescale. Thus grains in the Fork cannot have their alignments randomised by gas collisions faster than they can be induced to precess around either the magnetic field direction or the radiation anisotropy gradient by RATs.

Discussion of grain alignment in the Anomalous Region/Fork
This analysis, summarised in Tables 2 and 3, and illustrated in Figure 8, suggests that moderately large grains in the vicinity of extremely luminous sources such as Source I can be aligned by k-RATs rather than by B-RATs. This is the first time that this effect has been seen outside of a protostellar disc or AGB star envelope. We emphasise that our estimates of both τ r ad, p and τ Lar are highly uncertain. However, if our hypothesis is incorrect, and the grains in the Fork remain aligned with respect to the magnetic field, then a wholesale reorganisation of the field to be approximately radial around Source I must have taken place, apparently exclusively in this region.

Main Ridge
The main Ridge of OMC-1 (hereafter 'the Ridge') is an active site of ongoing star formation, an elongated structure which contains a number of dense cores (e.g. Hirota et al. 2015). Most famous amongst these is the Hot Core (Ho et al. 1979), a dense but apparently externally-heated and starless structure (Zapata et al. 2011) separated from Source I by ∼ 1 .
The polarisation pattern in the Ridge is strongly peaked on the 26 • E of N polarisation direction perpendicular to the large-scale magnetic field, as shown in Figure 10, with deviations in the Hot Core, and on a position NE of Source I and disconnected from the main body of the Ridge. The polarisation pattern in the Ridge is not consistent with being either concentric or radial around either the centre of the BN/KL explosion or around Source I, suggesting that the grains remain aligned by B-RATs. The polarisation vectors in the Ridge, rotated by 90 • to trace the magnetic field direction, are shown in Figure 11. We exclude from this figure the vectors tentatively associated with the Source I outflow, as discussed in Section 3.3.1, below.
We detect little polarised emission in the Ridge south of the Hot Core; particularly, we do not see polarised emission associated with the source SMA 1 (Beuther et al. 2005), although the Anomalous Region/Fork borders on this source. We similarly detect little polarisation on the north-western side of the Ridge. A possible explanation for this is a lack of a dominant polarisation mechanism in these regions.
Where polarised emission is detected in the Ridge, its direction is consistent with that predicted if the large-scale magnetic field direction persists to the highest-density and smallest-scale structures in OMC-1. If the magnetic field direction is indeed consistent over orders of magnitude in size scale, it suggests that the field remains dynamically important at the highest densities. On larger scales in molecular clouds, magnetic fields are consistently found to be perpendicular to filamentary structures where (a) the filament is gravitationally unstable and (b) the magnetic field is, on scales larger than the filament, dynamically important (Soler et al. 2013;Planck Collaboration et al. 2016). Although the Ridge is not a filament in the usual sense, it does meet these conditions, further suggesting that the magnetic field plays a significant role in mediating star formation in OMC-1.
The polarisation pattern in the Hot Core is broadly similar to that the rest of the ridge, with some deviation on the south-western side of the core. This deviation, although not well-resolved, is somewhat suggestive of the pinched field predicted for strongly magnetised dense cores. A dynamically important magnetic field is broadly expected to support a prestellar core against, and to impose a preferred direction on, gravitational collapse (Mouschovias 1976), producing the classical 'hourglass' magnetic field indicative of ambipolar-diffusionmediated gravitational collapse (e.g. Fiedler & Mouschovias 1993). However, it is not clear why such an hourglass morphology would only be apparent on one side of the core. Higher-resolution polarisation observations are required in order to understand the role of magnetic fields in the evolution of the Hot Core.
We note for completeness that in the south-west side of the Hot Core, the polarisation vectors are also consistent with being elliptical around Source I, with e 0.9, as would be expected for k-RAT alignment concentric around Source I in material displaced along the line of sight with respect to Source I. This is the only region in the Ridge where polarisation vectors are consistent with k-RAT alignment. However, as argued below, it appears unlikely that k-RATs Figure 11. Magnetic field vectors in the Ridge, obtained by rotating the polarisation vectors by 90 • , on the assumption that grains are aligned by B-RATs. We exclude the vectors tentatively associated with emission from the Source I outflow.
can dominate over B-RATs in the Ridge, and there is no clear reason for the Hot Core to be the exception to this.

Source I outflow?
Polarisation is detected at a position north-east of Source I, and disconnected from the Ridge. This region, labelled as 'Source I outflow?' in Figure 2, has polarisation vectors approximately perpendicular to those in both the Ridge and Source I, and thus are inconsistent both with the 26 • E of N polarisation direction perpendicular to the large-scale field direction and with being concentric around the BN/KL explosion, as can be seen in Figure 9. These vectors have orientations qualitatively similar to the SiO polarisation vectors detected by Hirota et al. (2020) in the north-eastern lobe of the Source I outflow, perhaps suggesting that this polarised emission arises from dust in the outflow cavity walls, as is hypothesised for the Fork (Wright & Plambeck 2017), or entrained by the outflow. We note, however, that the size scale of the SiO measurements is quite different to our observations  observed ∼ 1 around Source I), and so assigning this emission to the Source I outflow is speculative.
If the dust grains in the outflow are aligned by B-RATs, they could be tracing a helical magnetic field structure ). However, the vector orientations are also qualitatively similar to being concentric around Source I, as shown in Figure 9. This might suggest that grains in this region could instead be aligned by k-RATs, as we hypothesise in the Anomalous Region/Fork. It seems plausible that the extreme conditions apparently giving rise to k-RATs in the Fork to the south-east of Source I might also be expected to arise in the north-western outflow cone; however, we do not have sufficient evidence to conclusively determine the grain alignment mechanism in this region.

Why are grains in the Ridge not aligned by k-RATs?
With the exception of the handful of vectors tentatively associated with the Source I outflow, the polarisation pattern in the Ridge is inconsistent with being induced by k-RATs driven by Source I. Given that the Ridge is at a similar or smaller distance to Source I than is the Anomalous Region/Fork, this raises the question of how its grains  The requirement for k-RATs to dominate over B-RATs is τ r ad, p < τ Lar . The material of the Ridge has a significantly higher volume density (7.3 × 10 8 cm −3 ; Favre et al. 2011) than its surroundings and so is at higher A V . For much of the Ridge, the emission from Source I will also be obscured by its disc. The effect of this obscuration will be to decrease u r ad and make the radiation field less anisotropic (decreasing γ), thereby increasing τ r ad, p , although the mean wavelengthλ will increase, potentially mitigating this effect. We note also that the polarised emission which we see in the Ridge mostly arises from its eastern side, away from Source I. Moreover, if the emission in the Fork traces the Source I outflow cavity walls, then the hypothesised k-RAT alignment in the Fork will made more efficient by the lack of obscuration of Source I in the outflow cavity. This effect is less likely to apply in the Ridge, although the northern part of the Ridge could be impacted on by the Source I outflow, depending on the relative orientations of the Ridge and the outflow.
We also expect τ Lar to decrease in high-density material, as magnetic field strength B is expected to scale with density such that B ∝ n 0.5 or B ∝ n 0.66 (e.g. Crutcher 2012). We thus expect B to increase by a factor ∼ 10 − 20 in the Ridge over its value in the more diffuse surrounding material, correspondingly decreasing τ Lar .
While these two effects are difficult to quantify, the polarisation geometry which we observe on the eastern side of the Ridge suggests that between them they are sufficient to result in τ r ad, p > τ Lar , allowing B-RATs to dominate over k-RATs, despite the proximity of Source I.
We note that the increase in density will also decrease τ gas as shown in equation (13), although this will be mitigated by a decrease in T gas as the radiation field of Source I is increasingly obscured (cf. equation 9). Nonetheless, as discussed in Section 3.2.10, a density increase of two orders of magnitude would not be sufficient to make τ gas < τ Lar hold for physically plausible grain sizes, particularly if τ Lar is itself shortened in the Ridge.

MF4/MF5
Polarisation vectors in MF4 and MF5 (Favre et al. 2011; collectively known as the Northwest Clump) are similar both to the pattern pre-   (1) the centre of the BN/KL explosion (light blue, solid outline), (2) Source I (red, dashed outline). The polarisation angle associated with the mean 116-degree magnetic field direction is marked. Angles are measured on 0.25 (approximately Nyquist-sampled) pixels dicted for alignment perpendicular to the large-scale field direction, and to that for being concentric around the BN/KL outflow centre or Source I. The two clumps are at a similar distance to the BN/KL explosion centre as is the Ridge, and have complex substructure, with each consisting of three distinct velocity components (Pagani et al. 2017). We detect only 6 independent beams over MF4/MF5. As shown in Figures 12 and 13, the polarisation pattern in MF4/MF5 is more consistent with that expected for concentric polarisation around BN/KL than with the mean field direction or with being concentric around Source I, which could suggest that the grains are aligned by v-MATs. Unlike in the Fork, we do not have specific evidence for the recent passage of shocked ejecta which could have cause the grains to become mechanically aligned, but its proximity to BN/KL suggests that this mechanism could be plausible. However, the same argument that τ mech τ Lar applies in MF4/MF5 as in the Fork, again disfavouring v-MATs as the source of grain alignment in the region. Alternatively, the grains could be aligned by B-RATs to be perpendicular to an ordered field whose direction deviates slightly from the average large-scale field. An argument in favour of this interpretation, as opposed to k-RAT alignment driven by Source I, is that MF4 and MF5 represent significant density peaks (Pagani et al. 2017). We argued that grains in the centre of the Ridge are sheltered from the effects of the BN/KL explosion and/or the radiation field of Source I, and a similar argument can be made for MF4/MF5. However, MF4 and MF5 are somewhat less dense than the Ridge (n 2 = 2.2 × 10 8 cm −3 , compared to n 2 = 7.3 × 10 8 cm −3 in the Ridge; Favre et al. 2011) and so any such effect might be slightly less pronounced. We also expect k-RAT alignment efficiency to drop precipitously with distance from Source I, due both to energy density decreasing with R 2 and to increasing obscuration of Source I causing a decrease in anisotropy γ. We further find that the observed polarisation in MF4/MF5 is not consistent with any elliptical polarisation pattern around Source I with e < 1.
While it is difficult to conclusively determine the grain alignment mechanism in MF4/MF5, the likelihood that τ mech τ Lar disfavours v-MATs, and the disagreement between the observed polarisation pattern and that predicted to arise from polarisation concentric around Source I disfavours k-RATs, leaving B-RATs as the most probable source of grain alignment in the region.

Compact Ridge/MF1
The Compact Ridge (also known as MF1) is a  yet been influenced by the BN/KL explosion, and so place it at least 10 000 AU, and likely ∼ 20 000 AU, distant from the explosion centre along the line of sight. It is thus unlikely to be physically associated with the other dense clumps which we observe. The polarisation pattern in MF1 is inconsistent with being concentric around either BN/KL or Source I, and broadly similar to the polarisation pattern expected for grains aligned perpendicular to the large-scale 116-degree field, as shown in Figures 14 and 15. This result is in keeping with the hypothesis that the region is at a significant distance from the other clumps considered here. We can with some reliability in MF1 expect grains to remain aligned by B-RATs, and so we show the polarisation vectors, rotated by 90 • to trace the magnetic field direction, in Figure 16. There is significant ordered variation in the magnetic field direction across MF1, and the implied mean and median magnetic field direction values (147 • and 149 • , respectively) are similar to, but do not match, the average large-scale field direction.

SUMMARY
We have presented ALMA Band 7 polarisation observations of the centre of the OMC-1 region of the Orion Molecular Cloud.
We divided OMC-1 into five regions: Source I (a massive outflowdriving protostar), the Anomalous Region/Fork, the Main Ridge, MF4/MF5, and the Compact Ridge/MF1. Our key findings are as follows: (i) In Source I, we found a polarisation geometry parallel to the minor axis of the Source I disc, consistent with polarisation arising from dust self-scattering. The Source I disc is optically thick and viewed almost edge-on, supporting this interpretation.
(ii) In the Anomalous Region/Fork, a region in which emission may arise from the Source I outflow cavity walls, we found a polarisation geometry consistent with being concentric around Source I, and marginally consistent with being concentric around the centre of the BN/KL explosion or the centre of a ring of emission likely formed by recent passage of ejecta from the BN/KL explosion. We compared the mechanical alignment timescale τ mech to the Larmor timescale τ Lar in the Anomalous Region/Fork, finding τ mech τ Lar , indicating that grains are unlikely to be aligned by subsonic mechanical alignment torques (v-MATs) induced by the passage of shocks associated with the BN/KL explosion or associated with the Source I outflow. We compared the radiative precession timescale τ r ad, p for emission from Source I to the Larmor timescale in the Anomalous Region/Fork, finding τ r ad, p < τ Lar for moderately large grains (> 0.005 − 0.1 µm), indicating that grains in this region are likely to be aligned by radiative torques to precess around the radiation anisotropy gradient (k-RATs), i.e. to be perpendicular to the gradient of intensity from Source I. This is the first time that this effect has been observed outside protostellar discs or AGB envelopes, and favours the interpretation of emission in the region as arising from the Source I outflow cavity walls, as Source I must remain relatively unobscured for k-RATs to dominate in this manner.
(iii) In the Main Ridge, we found a polarisation geometry inconsistent with k-RAT or v-MAT alignment, and a polarisation geometry consistent with that of the large-scale magnetic field in the region, and so determined that grains are aligned perpendicular to the magnetic field (B-RAT alignment). The highly uniform magnetic field geometry and consistency with the large-scale magnetic field suggest that the magnetic field in the Ridge is dynamically important. We identified an area of polarised emission north-east of Source I possibly arising from the Source I outflow. Grains in this region could trace a helical magnetic field in the outflow or be aligned by k-RATs.
(iv) In MF4/MF5, we found B-RAT, k-RAT and v-RAT models all to produce predictions similar to the observed polarisation geometry. As τ mech τ Lar is likely to continue to hold in this region and the observed polarisation geometry is slightly but systematically different to that predicted for k-RATs driven by Source I, we concluded that grains are most likely aligned by B-RATs, tracing a magnetic field deviating slightly from the large-scale magnetic field direction.
(v) In the Compact Ridge/MF1, likely located sufficiently far from the BN/KL explosion and Source I to remain uninfluenced by their effects, we expect grains to remain aligned by B-RATs. We here found a polarisation geometry similar to, but showing ordered deviation from, being perpendicular to the large-scale magnetic field direction, again suggesting a dynamically important field.
Our observation of grains which are likely to be aligned by k-RATs rather than by B-RATs in the vicinity of Source I demonstrates the care which must be taken in the interpretation of polarisation observations in extreme environments in the interstellar medium. However, our results elsewhere suggest that the magnetic field in the centre of OMC-1 remains largely uniform over orders of magnitude in size-scale and density, and so is dynamically important and plays a significant role in mediating star formation in the region.

DATA AVAILABILITY
The data used in this paper are available in the ALMA Science Archive, under project code 2018.1.01162.S.