Histogram of Relative Orientation (HRO)

• Histogram of Relative Orientation (HRO) is a statistical method that quantifies the alignment between interstellar density gradients and magnetic field directions using histogram analysis.
• It employs Gaussian derivative kernels and image-differentiation techniques on column-density and polarization maps to study orientation patterns across various scales.
• HRO analyses reveal density-dependent transitions from parallel to perpendicular alignments, providing key insights into magnetic regulation and star formation processes.

Histogram of Relative Orientation (HRO) is a statistical method for quantifying how density or column-density structures are oriented with respect to the magnetic field in the interstellar medium. Introduced by Soler et al. for simulations and mock observations, and later applied to nearby molecular clouds with Planck, the method assigns a relative angle at each voxel or pixel between a density-gradient vector and a magnetic-field-related direction, and then studies the histogram of those angles, often as a function of density or column density (Soler et al., 2013, Collaboration et al., 2015).

1. Origin, scope, and object of measurement

The conceptual core of HRO is the comparison between two local directions: the orientation of matter structure and the orientation of the magnetic field. In the original formulation, the structure is represented by the density gradient, because the gradient is perpendicular to iso-density or iso-column-density contours; comparing the magnetic field to the gradient is therefore equivalent to comparing the field to the contour geometry itself (Soler et al., 2013). In observational implementations, the field direction is usually inferred from dust polarization, whereas the structure is traced by a column-density map or an intensity map, and the analysis is performed pixel by pixel rather than by fitting explicit filaments (Collaboration et al., 2015).

HRO is fundamentally a relative-orientation diagnostic rather than a magnetic-field reconstruction method. It quantifies whether structures are preferentially parallel, perpendicular, or randomly oriented with respect to a known plane-of-sky magnetic field. In that sense, it is distinct from methods that infer field direction directly from gradients. The broader review literature places HRO alongside DCF-based methods and the polarization-intensity gradient method as one of the principal tools for assessing the dynamical role of magnetic fields in star formation, with the specific strength that it captures how preferential alignment changes with column density (Liu et al., 2022).

2. Mathematical formulation and angle conventions

Two closely related angle conventions coexist in the literature. In the original three-dimensional formulation, the angle $\phi$ is defined between the magnetic field $\mathbf{B}$ and the density gradient $\nabla n$,

$$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$

and the HRO is built in $\cos\phi$, because random 3D vectors are uniformly distributed in $\cos\phi$ rather than in $\phi$ itself (Soler et al., 2013). In this convention, $\cos\phi \sim 0$ means the field is perpendicular to the gradient and therefore parallel to the iso-density contours, whereas $|\cos\phi|\sim 1$ means the field is parallel to the gradient and therefore perpendicular to the contours.

In the Planck two-dimensional formulation, the observable polarization pseudo-vector $\hat{\mathbf{E}}$ is used directly. The polarization angle is obtained from Stokes $\mathbf{B}$0 and $\mathbf{B}$1 via

$\mathbf{B}$2

and the projected magnetic-field orientation is perpendicular to that angle. The relative angle is then written as

$\mathbf{B}$3

with $\mathbf{B}$4 (Collaboration et al., 2015). Under this convention, $\mathbf{B}$5 corresponds to $\mathbf{B}$6 parallel to the gradient and therefore $\mathbf{B}$7 parallel to the iso-column-density contours, while $\mathbf{B}$8 corresponds to $\mathbf{B}$9 perpendicular to those contours.

Later implementations often rewrite the same geometry directly with the magnetic field and an intensity gradient, for example

$\nabla n$0

This is algebraically equivalent up to the usual $\nabla n$1 rotation between the polarization pseudo-vector and the magnetic field, so the critical point is always to state whether the angle is defined relative to the gradient or relative to the contour tangent (Hu et al., 2019).

The gradient itself is computed by standard image-differentiation procedures. The original HRO work emphasized Gaussian-derivative kernels so that smoothing and differentiation are performed simultaneously and multiple physical scales can be studied by varying the kernel size (Soler et al., 2013). Planck Collaboration Int. XXXV used convolution with Gaussian derivative kernels,

$\nabla n$2

and defined the orientation of iso-$\nabla n$3 contours from the gradient components (Collaboration et al., 2015). Other studies employ finite differences or local polynomial fits, but the invariant ingredient is the use of a local gradient field as the normal to the structure.

3. Observational workflow and computational practice

Operational HRO analyses require two maps on a common grid and at a matched resolution: a scalar structure tracer and a polarization-derived field tracer. In nearby-cloud studies, the structure tracer is often a Herschel or Planck column-density map, or a dust optical-depth proxy such as $\nabla n$4; the field tracer is the plane-of-sky orientation derived from Stokes $\nabla n$5 and $\nabla n$6 (Collaboration et al., 2015). In dense-core studies with JCMT POL-2, the same logic is used after smoothing the polarization map to the Herschel resolution and reprojecting the column-density map onto the polarization grid (Perry et al., 2024).

A standard workflow is: compute the gradient field, compute the relative angle at each selected pixel, bin the selected pixels by column density, build an HRO for each bin, and then compress each histogram into a scalar statistic. Planck Collaboration Int. XXXV smoothed the maps to an effective angular resolution of $\nabla n$7 FWHM, used $\nabla n$8, and divided the selected pixels into 25 equal-population bins in $\nabla n$9 for most clouds, with each HRO built from 12 angle bins of width $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$0 (Collaboration et al., 2015). Dense-core analyses use analogous binning but with higher-resolution maps and source-specific masks (Perry et al., 2024).

Pixel selection is not incidental. Planck Collaboration Int. XXXV combined a gradient mask with a polarization mask, requiring cloud-dominated gradients and reliable polarization, including cuts equivalent to $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$1 and $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$2 (Collaboration et al., 2015). In the Ophiuchus core analysis, the retained pixels satisfied $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$3, $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$4, and $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$5 (Perry et al., 2024). These cuts matter because HRO shape is sensitive to low-SNR polarization angles, poorly defined gradients, and line-of-sight confusion.

The same procedural structure extends beyond dust column density. Molecular-line studies compute HROs between dust-polarization angles and gradients of integrated line intensity. Synthetic-observation work likewise computes HROs for optically thin and optically thick tracers separately, which makes the method sensitive not only to density regime but also to radiative-transfer depth effects (Mazzei et al., 2023).

4. Shape parameters, companion statistics, and regime diagnostics

The simplest HRO product is the histogram itself, but most studies reduce that histogram to one or more scalar diagnostics. In the original three-dimensional work, Soler et al. defined a shape parameter

$$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$6

where $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$7 is the area under the central part of the HRO and $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$8 is the area in the extremes. For the 3D HRO in $$\phi \equiv \arctan\left( \frac{\mathbf{B} \times \mathbf{\nabla}n}{\mathbf{B} \cdot \mathbf{\nabla}n} \right),$$9, $\cos\phi$0 corresponds to $\cos\phi$1 and $\cos\phi$2 to the two edge intervals $\cos\phi$3 and $\cos\phi$4 (Soler et al., 2013). Positive $\cos\phi$5 indicates a histogram peaked near $\cos\phi$6, negative $\cos\phi$7 a histogram peaked near $\cos\phi$8.

Planck Collaboration Int. XXXV used a normalized variant,

$\cos\phi$9

with $\cos\phi$0 integrated over $\cos\phi$1 and $\cos\phi$2 over the two edge ranges $\cos\phi$3 and $\cos\phi$4 (Collaboration et al., 2015). The notation is the same but the normalization differs, so comparisons across papers require attention to the exact definition.

The Projected Rayleigh Statistic (PRS) is the most common companion diagnostic. In one widely used form,

$\cos\phi$5

with sign convention chosen so that positive values indicate one preferred orientation and negative values the orthogonal one (Micelotta et al., 2018). The review literature emphasizes that PRS uses all angles rather than only the central and extreme histogram bins, and therefore often yields smaller error bars than the histogram shape parameter (Liu et al., 2022). Gradient-based studies also compare HROs with the alignment measure

$\cos\phi$6

which carries similar alignment information in a different normalization (Hu et al., 2019).

Across simulations, HRO-derived statistics are strongly regime dependent. Soler et al. found that the relative orientation changes from parallel to perpendicular in regions above a critical density $\cos\phi$7 only in the highest magnetization case, and that the change decreases in lower magnetization cases (Soler et al., 2013). A later non-gravitating turbulent study sharpened this statement: only simulations with an initially strong magnetic field, $\cos\phi$8, show a change in the preferential orientation from mostly parallel at low densities to mostly perpendicular at higher densities; compressive turbulence alone is not capable of inducing the transition observed toward nearby molecular clouds (Körtgen et al., 2020). In self-gravitating simulations, the inclusion of gravity increases the number of dense structures perpendicular to the magnetic field, reflected as lower PRS values for denser regions, and observed-cloud comparisons favor sub-Alfvénic models (Barreto-Mota et al., 2021).

Tracer dependence is equally important. In synthetic molecular-line observations of a trans-Alfvénic cloud, a transition to perpendicular alignment appears above $\cos\phi$9 and is recovered in most optically thin tracers, whereas for $\phi$0CO the transition does not occur because the $\phi$1 surface lies largely in front of the cloud midplane, so the line mainly probes lower-density gas (Mazzei et al., 2023).

5. Empirical results across clouds, cores, envelopes, and the Galactic Center

The canonical observational result was established on nearby Gould Belt clouds. Using Planck 353 GHz polarization and column-density maps, Planck Collaboration Int. XXXV found that in most clouds the relative orientation changes progressively with increasing $\phi$2, from preferentially parallel or having no preferred orientation to preferentially perpendicular, with the switch occurring at $\phi$3 (Collaboration et al., 2015). The same approximate transition column density appears in L1688 when Planck and HAWC+ HRO analyses are combined, with the HAWC+ study showing consistent perpendicular relative alignment down to scales of $\phi$4 pc (Lee et al., 2021).

Subsequent studies showed that the phenomenology can be richer in denser regions. In Serpens Main, HRO analysis found magnetic fields parallel to filaments in less dense filamentary structures where $\phi$5, perpendicular to filaments in dense filamentary structures with star formation activity, perpendicular to density gradients again at $\phi$6 as a signature of core formation, and parallel to density gradients once more at $\phi$7, interpreted as magnetic fields being dragged in by infalling material (Kwon et al., 2022). The review literature summarizes this broader picture as a robust parallel-to-perpendicular change with increasing column density, with a possible transition back from perpendicular to random alignment at still higher column density (Liu et al., 2022).

At dense-core scales, geometry becomes a central part of the interpretation. The Ophiuchus core analysis with JCMT POL-2 and Herschel maps showed that high-aspect-ratio ellipsoidal cores produce strong HRO signals, while low-aspect-ratio cores tend toward $\phi$8 even for simple linear fields. In the corresponding toy models, linear fields oriented by less than $\phi$9 from the core minor axis yield constant HROs with $\cos\phi \sim 0$0, whereas hourglass fields produce a minimum in $\cos\phi \sim 0$1 at intermediate densities; the observed Ophiuchus cores showed no signature of hourglass fields on those scales (Perry et al., 2024). At protostellar-envelope scales of order $\cos\phi \sim 0$2 au, the BOPS survey found weak evidence for systematic evolution with column density alone and concluded that magnetization level also plays a crucial role: weakly magnetized envelopes exhibit predominantly parallel or random alignment, while strongly magnetized ones show perpendicular configurations even at moderate densities (Cai et al., 5 Feb 2026).

In more extreme environments, HROs can encode different physics. In the Central Molecular Zone, FIREPLACE used HRO and PRS to show that CMZ molecular clouds span a range of orientations from parallel to perpendicular, and argued that these orientations depend on the prevalence of gravitational shear in the Galactic Center. The same analysis found a preferred perpendicular relative orientation between the far-infrared magnetic field and prominent non-thermal filaments, supporting the interpretation of a more pervasive vertical field in the Galactic Center and yielding a dynamical upper limit $\cos\phi \sim 0$3 mG for that vertical component (Paré et al., 2024). In the western $\cos\phi \sim 0$4 Car GMC, HRO analysis found a transition from mostly parallel to mostly perpendicular alignment at $\cos\phi \sim 0$5 for the full sample and a similar threshold of $\cos\phi \sim 0$6 in a subset of clumps, while other clumps showed flat or reversed trends attributed to external environmental forces from a nearby H II region (Barnes et al., 24 Apr 2025).

6. Relation to other techniques, limitations, and interpretive cautions

HRO is often used together with PRS, DCF, and alternative structure-extraction methods, but it is not itself a field-tracing technique. It requires polarization to supply the magnetic-field orientation; in that sense it differs fundamentally from the Intensity Gradients Technique and the Velocity Gradient Technique, which attempt to infer the field from the gradients themselves. The same comparison work emphasizes that HRO is a statistical diagnostic of density-field geometry, not an autonomous magnetic-field tracer (Hu et al., 2019).

Methodological choices can materially alter the result. Comparisons between the Rolling Hough Transform and the gradient technique showed that when both methods are applied to the same pixels, HRO-based trends are consistent, but when each method uses its own native pixel selection the differences can be substantial and may even reverse the inferred trend with column density (Micelotta et al., 2018). Core-scale work likewise shows that a weak HRO signal may be geometry limited rather than physically random: low-aspect-ratio cores naturally produce $\cos\phi \sim 0$7 even for ordered linear fields (Perry et al., 2024).

Several caveats recur across the literature. Projection biases 2D relative-orientation distributions toward appearing more parallel, although Planck Collaboration Int. XXXV argued that this bias does not erase a true perpendicular trend at high column density (Collaboration et al., 2015). Grain-alignment efficiency, line-of-sight averaging, angular resolution, and masking all affect the measured histograms. Gradient weighting is another issue: critique from the intensity-gradient literature showed that weighting a histogram of $\cos\phi \sim 0$8 by $\cos\phi \sim 0$9 can distort the histogram shape and even flip the sign of the inferred shape parameter, motivating the use of unweighted histograms in $|\cos\phi|\sim 1$0-space rather than weighted histograms in $|\cos\phi|\sim 1$1-space for that particular application (Hu et al., 2019).

The most general interpretive caution is that HRO does not encode a single physical mechanism. In nearby molecular clouds, the observed parallel-to-perpendicular transition is usually interpreted with simulations as evidence for trans-to-sub-Alfvénic dynamics and an increasing role for gravity (Liu et al., 2022). In the CMZ, however, the same relative-orientation statistics are interpreted in terms of gravitational shear rather than the Galactic-disk paradigm (Paré et al., 2024). In massive clumps near strong feedback sources, flat or reversed HRO trends can reflect external compression rather than self-gravitating magnetic regulation (Barnes et al., 24 Apr 2025). This suggests that HRO is most informative when embedded in a broader analysis that includes kinematics, field-strength estimates, source geometry, and environmental context.